============== Page 1/1 ============== Michel MOULY Marie-Bernadette PAUTET The GSM System for Mobile Communications A comprehensive overview of the European Digital Cellular Systems This book is published by the authors. Correspondence, in particular for orders, but also for comments, should be mailed to: M. MOULY et Marie-B. PAUTET 49, rue Louise Bruneau F-91 120 PALAISEAU FRANCE Telephone : +33 1 69 31 03 18 Facsimile : + 3 3 1 69 31 03 38 Copyright © 1992, Michel MOULY and Marie-Bernadette PAUTET All rights reserved. No part of this book may be reproduced, translated, or utilised in any form or any means, electronic or mechanical, including photocopying, recording, or any information storage or retrieval system, without permission in writing from the authors. All drawings are original, and all corresponding rights reserved. The name and logo GSM (and "Global System for Mobile c°111.11191119#1194.0..M.W8iggred • ' 1 International Standard Book Number: 2-9507190-0-7 This book would not exist if a group of European people had not taken a conunon aim and worked hard together to reach it. This hook is dedicated to all these mothers and fathers of I I /Id \ \ M P FOREWORD • by Thomas Ma,,. Former Chairmn. ETSI TCIGSM At i t s meeting i n V i e n n a i n June, 1 9 8 2 . t h e C E P T Telecommunications Commission decided to set up a group to work out specifications for a pan-European cellular communication system for the 900 MHz band which had recently been allocated to land mobile use. The idea behind this decision was to create fur the first time a system that would end the traditional European fragmentation and incompatibility in the mobile field. It was clear to the CEPT that unless the opportunity y‘ as taken there and then. the 90() MI-li band would rapidly he allocated l'or different and incompatible sysiems in different countries. and that in view of the difficulty o f nt.. 11 ng another commonly available band. the opportunity would then he gone for decades to create a system with pan• European roaming. From day one the work was directed towards a second-generation cellular system. since many countries already had f i r s t -generation systems (TAGS. N I T. System C. etc.) in use or in the implementation stage. There was no decision or directive that the new system should he . . . w e there was uncertainty as to what transmission mode wouldd l,A ta ig best meet the requirements. However, there was agreement that the new system m u s t t a k e i n t o account recent developments i n t h e telecommunication field. such as CCITT Signalling System No. 7. ISDN. OSI and other powerful innovations. Gradually. it became clear that the most likely solution would be a fully digital one, provided the radio propagation problems could he overcome. since the digital mode would be far more adaptable to the needs of a modern system than the analogue one. The final proof of the feasibility of the digital system was given alter a very thorough series of measurements in various countries had been performed, above all at CNET in Paris, and the decision o r a digital solution could be taken. With that as a basis. the way forward to a (Nall \ new system. fit for the 1990s. was open. 8 T H E O ' GSM SYSTEM To prepare a description that goes into details Oil virtually every point of the system in such a way that the interested person feels tempted to read more. not less, is a very difficult task. requiring experts with a good ability to make complicated things easy to understand in plain language and above all to explain how the various parts of the system. described i n seemingly unrelated technical specifications, a r e interconnected. It is therefore extremely fortunate that the authors of this book, both o f whom are experts i n the field and have made very significant contributions to the GSM work. have on their own initiative produced a description of GS 1\I which in my view very well meets the needs described above. I wish t o congratulate the authors on their achievement and to express my thanks for the very gyeat effort they have spent. I am sure the hook will he of very great help to a large community, both in and outside Europe. in the years to conic. 41, The work on the system was made possible by a united European effort, both i n PTTs, research institutes and manufacturers, w h o participated from 1987 on. This kind of common effort towards a unified system i s different f r o m t h e earlier tradition i n European telecommunications, where national concerns have frequently led to a protectionist attitude, with resulting incompatibility and poor economy. Also the European Community was strongly in favour of the common system and made a significant contribution in securing the legal basis for setting aside the necessary frequencies in its member states. The result of the work by the GSM (first a CEPT working group. later on an ETSI group, and recently renamed SMG) is available through ETSI in the form of some 5200 pages of technical specifications for the first phase of the system. The work has attracted a great deal of interest even outside Europe, notably in Australia and south-east Asia. In fact. several countries have already taken f i r m decisions t o build GSM systems. For coming phases, more papers will follow in the years to come. Clearly, this huge amount o f specifications is totally unreadable t o anyone but a tiny group of specialists, just like a collection of law texts is incomprehensible to non-lawyers. This is inevitable. since an absolute requirement on a technical specification which is to be used by many different manufacturers, operators and regulatory bodies i n many countries, often as a basis for legally binding contracts, is that it must be complete. consistent and unambiguous (unfortunately, i t sometimes happens that this requirement is not fully met, despite all the efforts). Thus, the stringency requirement does not improve the legibility, and one may safely assume that very few people will ever master the whole set of specifications describing the system. At the same time. very many people will be involved in the implementation, procurement. operation and maintenance o f the system, which means that they arc in dire need o f another type o f description o f GSM, which provides them with the necessary overview and a lot o f details without resorting to the very formal description given in the specifications. This fact has been evident for a number of years, since the next lower level of description below the specification only consisted of various conference proceedings, useful in themselves, but in most cases very summary and usually without much co-ordination. Thoiiii, I hug 0 CON I F.N1S I I coNTENTs Foreword, by Thomas Haug 7 Contents 11 Preface 17 Chapter I - Setting the Scene 23 I.I. A l i t t l e Bit of 11 isloc I.I.I. B.G. (Before (1SNI) 1.1.2. The Cienesis of a Standard 1.1.3. Organisation of the ‘Vork 1.1.4. The GSM Mot' 1.1.5. Technical Choices I.I.O. The GSN1 Technical Specifications 1.2. Cellular Systems 1.2.1. General Aspects 1.2.2. Cellular Cmel:age 1.2.3. Radio Interface Nlanagement 1.2.4. Consequences of Nlobility 1.2.5. Roaming 1.3. G S M Functionalities 24 2-1 29 31 33 37 39 3') alt 43 44 dh 1.3.11 G S M : a M u h i N e n ice S y s t e m f o r the 1...er 1.3.2. GSM: a System for the Operahu Specifications Reference 7_2 75 Chapter 2 -1Architecture 79 2.0.1. The three Description Axes 1.0.2. Frontiers of the System: Where are the Borders of GSM? 2.0.3. Internal GSM Organisation 2.1. Sub-Systems 2.1.1. The Mobile Station (MS) 2.1.2. The Base Station Sub-System (BSS) 2.1.3. The Network and Switching Sub-System (NSS) 2.1.4. The Operation Sub-System (OSS) 84 87 St) St) 94 100 105 12 T H E I3 CONTENTS GSM SYSTEM it 2.2. Functional Planes 2.2.1. Layer Modelling 2.2.2. Transmission 2.2.3. Radio Resource Management (RR) 2.2.4. Mobility Management (MM) 2.2.5. Communication Management (CM) 2.2.6. Operation, Administration and Maintenance (DAM) 2.3. Interfaces and Protocols — An Overview 108 109 113 114 114 115 116 118 Specifications Reference 121 Chapter 3 - Transmission - 125 3.1. Modelling Principles 126 3.2. A n End-to-End View of Transmission 3.2.1. Speech 3.2.2. Non Speech Services 3.3. Transmission inside GSM 3.3.1. Architecture 3.3.2. Speech 3.3.3. Data Specifications Reference 128 128 132 149 149 154 166 184 Chapter 4 - The Radio Interface 187 4.1. The Needs 4.1.1. User Data Transmission 4.1.2. Signalling 4.1.3. Idle Mode 4.2. The Multiple Access Scheme 4.2.1. The Time Axis 4.2.2. The Frequency Axis 4.3. From Source data to Radio Waves 4.3.1. The Bursts 4.3.2. Interleaving and Channel Coding 4.3.3. Ciphering 4.3.4. Modulation Specifications Reference 188 189 190 191 195 197 217 227 231 238 248 249 258 Chapter 5 - Signalling Transfer 261 5.1. The Needs 5.1.1. Contiguous Entities 5.1.2. Relaying 5.1.3. Protocol Interworking 5.2. Linking 5.2.1. Structuring in Frames 5.2.2. Segmentation and Re-Assembly 5.2.3. Error Detection and Correction 5.2.4. Multiplexing 5.1.5. Flow Control LAPD and LAPDm Frames: a Summary 5.2.7. RLP Characteristics 5.3. Networking 5.3.1. Networking in the BSS 5.3.2. Networking in the NSS 5.3.3. Networking for Supplementary Services Management 5.3.4. Networking for Point-to-Point Short Messages Specifications Reference 26P Chapter 6 - Radio Resource Management 309 6.1. RR Functions 6,1.1. The Concept of RR-Session 6.1.2. Initialisation 6.1.3. Transmission Management H andover Preparation 6.1.5. Power Control and Timing Advance 6.1.6. Radio Channel Management 6.2. Architecture and Protocols 312 313 317 321 327 342 350 362 6.3. RR Procedures 6.3.1. Initial Procedures: Access and Initial Assignment 6.3.2. Paging Procedures 6.3.3. Procedures for Transmission Mode and Cipher Mode Management 6.3.4. I landover Execution 6.3.5. Call Re-Establishment 6.3.6. RR-Session Release 6.3.7. Load Management Procedures 6.3.8. SACCII Procedures 6.3.9. Frequency Redefinition 6.3.10.General Information Broadcasting 366 263 264 266 268 269 270 771 277 279 280 281 283 284 294 299 301 305 367 382 385 306 412 415 41$ 42(1 424 424 I flit I t T H E (iSNI SI'STEN1 Chapter 7 - Mobility and Security Management 433 7.1. Location Management 7.1.1. The Factors Determining the Service 7.1.2. Cell and PLMN Selection 7.1.3. Architecture 7.1.4. The Location Updating Procedures 7.2. Security Management 7.2.1. The Needs 7.2.2. The Functions 7.2.3. Architecture and protocols 7.2.4. The Signalling Mechanisms 7.3. Miscellaneous M M Functions 434 435 446 459 465 477 477 478 485 487 493 Specifications Reference 498 Chapter 8 - Communication Management 501 8.0.1. The Communication 8.0.2. Management Functions 8.1. Call Control 8.1.1. The Routing of Mobile Terminating Calls 8.1.2. Architecture 8.1.3. The Mobile Originating Call Establishment Procedure 8.1.4. The Mobile Terminating Call Establishment Procedure 8.1.5. The Interrogation Procedures 8. I .6. Call Release 8.1.7. I n -Call Functions 8.2. Supplementary services management 8.2.1. Architecture 8.2.2. Procedures 8.3. Short messages 8.3.1. Architecture 8.3.2. Mobile Originating Short Messages 8.3.3. Mobile Terminating Short Messages Specifications Reference 503 507 510 510 528 530 539 543 545 547 552 553 554 556 557 559 560 564 Chapter 9 - Set work Nlanagenient 567 9.1. Subscriber Management 9.1.1. Subscription Administration 9.1.2. Billing and Accounting 9.2. Maintenance 568 569 572 578 9.3. Mobile Station Management 9.4. System Engineering and Operation 585 9.4.1. Cellular Planning 9.4.2. Cell Conliguratitin 9.4.3. Network Engineering 9.4.4. Observations 9.4.5. Network Change Control 9.5. Architecture and Protocols 9.5.1. Management Network Architecture 9.5.2. Operation and Maintenance in the Traffic Handling Protocols 9.5.3. The H'I'S Mariatieinent nniti)col 9.5.4. The GSM Q3 Protocol Specifications Reference The List of the GSM Specifications Bibliography Index Message Index Index of Figures 591 593 613 622 627 628 633 633 637 638 640 646 649 667 673 693 697 tI PREFACE This book has been written with a set o f goals and a few categories of reader in mind. This is not a work aimed primarily at the layman, hut rather a t professionals. o r professionals-to-be i n t h e s p h e r e o f telecommunications. However, it should not be a hook accessible only to a handful o f specialists. W e have tried t o ‘vaymark a route leading a reader of scientific background and interest to the understanding of a full system. from its needs to its technical choices. The s p e c i f i c a t i o n o f t h e " G l o b a l S y s t e m f o r M o b i l e Communications"( G S E T 3 ' )l and of its sibling. DCSI800, was a long process. f u l l o f sometimes lengthy and often h o t debates. T h e major visible product o f these years o f work is a pack o f sonic 5000 pages. the GSM a n d D C S M O O Te c h n i c a l Specifications. I n d r a f t i n g t h e s e specifications, the experts have tried to minimise the risk o f ambiguous interpretation, rather than to ease understanding f o r the outside reader. The rationale behind most o f the choices, as well as the alternative ideas which have not been retained i n the filthl soilliinn. are nowhere t o he found in the specifications. Moreover. the purpose o f sonic features and the way to implement or use them efficiently are also to a large extent out of the scope o f the specifications. One o f the aims o f this hook i s t o (partially) f i l l this gap. I t offers a synthesis o f the G S M and DC'S I SOO specifications. which w i l l enable anyone w h i c h tlesires t o understand what is in the Specifications to get a good grasp of how the whole system is designed. I t w i l l i n n o w a y replace t h e official specifications f o r developers o r operators. but should increase the efficiency w i t h which they w i l l acquire knowledge o f the system. A s often as possible. topics have been presented from a different angle to the official specifications. describing the topics v a t o p -down approach. often f o r that purpose f i n i n g t o g e t h e r p i e c e s I - a n d i n W11.'21'011 s p e c i f i c ' a ' t i o n s . GSM a n d DCSA SOO are undoubtedly a m a j o r achievement i n modern cellular telephonv . ant. 1we hope this hook w i l l help to shoe i t . But the purpose i s not apologetics. neither is i t a d \ ertising. A critical view (nothing human is absolutely perfect) and personal ideas (indicated c u n u n r r ; i . i l I C d 6 > l i S N T 18 PI:II \ i t \I SYS'IlAl as such) have not been left aside. The reason for this choice is our hope that such constructive criticism will be helpful for future work, as GSM is certainly not the ultimate in its domain. GSM includes a number of firsts. but it will he followed by yet more elaborate systems. May the designers of new generations of systems find something of use in the subjective views they will find in this book. The system is also by itself an object of interest. The standardising bodies chose not to limit the scope of the Technical Specifications. as is often the case for radiocommunications, to the radio interface. GSM and DCS1800 are standardised as a complete system. and much o f the internal behaviour is documented. as well as the outline architecture of both the infrastructure and the mobile station. They encompass many different areas of the telecommunications field, and allow the individual presentation of each of those fields as a part of a whole. As such. GSM forms an interesting case study for students in the telecommunication field, so the hook is also aimed at this readership'-. As a consequence. a compromise has been sought between—on the one hand—using a layman's vocabulary and—on the other hand—drawing heavily from the technical jargon of the GSM committees. Good telecommunication basics are considered as the only prerequisite for reading this book. As a consequence o f this multiplicity o f goals, the technical difficulty o f the book varies greatly. We have spared neither basic explanations (certainly unnecessary for the specialists of each domain). nor some of the technical details which are the bread and butter of the life of us specialists, but maybe irrelevant for others. We have taken some effort to write the book so that it can be read by readers interested in only some of the topics, or who have neither time nor inclination to read the book from cover to cover. For these readers, we have compiled, at the end of the book, an index of the main terms and topics so that a reading of the entire book is not necessary to understand one subject. We have also tried t o organisi the text so that the difficulty o f the material increases as the reader progresses. The two first chapters are of general nature, whereas the others are targeted at some specific technical areas. Each technical chapter begins with a presentation o f the needs and principles, in plain terms, and becomes increasingly more detailed. The first portion of each of the technical chapters does not depend on the last portion of 'previous chapters, enabling two levels of reading for each of the technical domains. Where a topic has been elaborated in some finer An important part of the book is indeed based on courses taught by the authors in various engineering schools and universities as well as in the context of postgraduate education. detail within the main test. we have also made use of a smaller font. in order not to unduly interrupt the main flow f o r those who want the broader understanding. • Let us now have a general view of the contents. For starters. the first chapter is a general introduction. It sets the scene. including a brief history of the design of the system and hum the work is conducted within the standardisation arenas. The functions which the system was designed to fulfil are described in this general presentation. The rest of the book describes the system. follo i n g more or less a top-down approach. "I'lk• system is first described as a single object. as seen by the external world. Then its architecture is progressively re\ Med, with an increasing level of detail. Our main courses are the transmission chapters (chapters 3 to 51. which aim a t describing h o w information i s transported. and the signalling chapters (chapters 6 to 8). \\Inch describe how the constituents of the system cooperate to fulfil the functional requirements. The last chapter (chapter 9) deals with the system from the point o f view of the operator: how are networks built and managed? A number of annexes are included for reference, and as a digestif. One is a detailed list of the GSN-1 and DCS1800,rechnical Specifications. aimed at newcomers who may wonder at which end to start. Another annex is an index o f terms and acronyms. giving references to the book. A t the end o f each chapter of the main text. references t o t h e relevant parts o f the Technical Specifications are given. in order to facilitate the search for the official information. Several menus can be suggested. Those who are interested in a general view of GSM and DCS1800 should read the two first chapters (though not easy, the second one i s necessary t o acquire the basic vocabulary), and browse through the beginning o f each o f the other chapters. Those more interested in the specificities of cellular telephony as opposed to general telecommunications will be mostly interested in Chapter 4 (the radio interlace). Chapter 6 (dealing with radio resource management) and the beginning o f Chapter 7 (the management o f mobility). The entirety o f Chapters 5 to 8 will appeal mostly to those interested 11.telecommunication systems. Finally, the whole work. from a r t Hyat d in w e s h to cover to cover, is strongly recommendedp (or all'?) o f the Te c h n i l ' a l Specifications. Bon Jame( it! 20 T I PRI* \('N Ili GSM SYSTEM English Language Editors' Note Disclaimer This hook was written in that international stream of the English language which is being increasingly employed in Europe and other parts of the world as a medium o f communication b y people f o r wOom. generally, English i s n o t their mother tongue. Indeed, the G S M Specifications themselves were developed using a form of this language. The substance of this book reflects only the understanding of the GSM and DCS1800 Technical Specifications by the authors. I f it is true that both authors have taken part to a large extent in the design of the system, and that this book by and large tries to reflect the orthodox understanding of the Specifications, it can in no way be taken as the unique view of the committees who wrote the Specifications. The GSAI and DCSI800 Technical Specifications are the sole reference, and the authors disclaim any responsibility for any usage of the interpretations proposed in this book. In editing and correcting the text we have endeavoured to retain all the flavour. inventiveness and humour expressed by the authors whilst trying to remove elements which 'night obstruct the understanding by a native English speaker or lead to a misunderstanding of technical content. English is a rich, living language and we have allowed it to live a little! Acknowledgements We wish t o acknowledge a l l the friends and colleagues, i n particular f r o m FRANCE TELECOM a n d M AT R A COMMUNICATION, w h o helped us, with their expertise, to understand the variegated specialised domains that are to be travelled through when exploring the nooks and crannies of GSM. Among all those who provided assistance, special thanks should be given to the manuscript readers, whose constructive criticisms are the source of the possible quality of this piece of work, including Christian Casenave. Nicolas Demassieux, Jean-Louis Dornstetter, Philippe Duplessis, Philippe Dupuis, Michel Lambourg, and Alain Maloberti. As a particular mention, we wish to convey our deepest thanks to Paul Simmons and Tony Wiener, who accepted, in addition to a technical review of the book, the difficult task o f correcting the weird language resulting from French people attempting to write in "English". Remaining errors, technical or grammatical, are only to he blamed on the authors. Lastly, we wou)d like to thank our respective spouses for their cooperation, as well as Annc-Laure and Guillaume for their forbearance. Michel Mouly Marie-Bernadette Pautet 4P Paul Simmons Tony Wiener 1-III.(iSNIN '1 I M c . 1 SETTINGTHE SCENE 1.1. A Little B i t o f History 1.1.1. B.G. (Before GSM) 1.1.2. The Genesis of a Standard 1.1.3. Organisation of the Work 1.1.4. The GSM MoU 1.1.5. Technical Choices 1.1.6. The GSM Technical Specifications 1.2. Cellular Systems 1.2.1. General Aspects 1.2.2. Cellular Coverage 1.2.3. Radio Interface Management 1.2.4. Consequences of Mobility 1.2.5. Roaming 1.3. G S M Functionalities 1.3.1. GSM: a Multiservice System for the User 1.3.2. GSM: a System for the Operator Specifications Reference SETTINGTHE SCENE 24 24 28 29 32 33 37 Mass-market mobile telecommunications i s certainly one o f the major breakthroughs o f this end o f millennium. The possibility to make and receive calls through a small wireless handset. wherever you are. has an obvious appeal. The business opportunities are tremendous. since one can imagine that e v e n person (and n o t o n l y every home), including children. could be equipped provided the service is cheap enough. Many people agree that the sociological consequences b e important. much more than f o r videocommunications. Wireline telephony allows u s t o reach a place, i f someone is there to answer. Mobile telephony allows us to reach a particular person. wherever (almost) he o r she is. This w i l l greatly increase the accessibility t o people. and increase the feeling o f security. On the other hand, this increased accessibility can in many cases he a nuisance. and widespread social acceptability o f mobile telephony requires that users have a high degree o f control on the calls they receive (identification o f the calling party. forwardiqg o f calls t o a third party, message banks, etc.). 39 39 40 43 44 46 47 47 72 75 Mobile telecommunications is not a very recent technology, but it 0 THE CiSNISYSILNI s1.1 II\ ti I 111. st GSM was designed internationally, in stanu—disation committees. by the major European telecommunications operators and manufacturers. The understanding of the gain to be obtained by combining resources. and of the business opportunity offered by mass-market radiotelephony. resulted i n a substantial man-power and financial effort from the participants, thus making GSM a very dynamic project. This design-by-committee has resulted in a public set of detailed specifications encompassing all system areas. This book is based largely on this committee output and o n the authors' experience o f the discussions which were held during the elaboration o f the system specifications. This first chapter will be quite general, and aims at setting the stage, the main lines of the plot and some of the main actors. The rest of the casting will be presented in the next chapter. The present chapter is divided into three parts. The first part will set the historical background of GSM. Some text i s devoted t o the description o f h o w t h e standardisation committees worked; this will give some insight o f what happened behind the stage. The second part aims to provide the basic technical foundations of public cellular radio telecommunications. Most of the addressed points are developed from different angles throughout the other chapters, recurring as leitmotivs. The last part describes the services that GSM offers to the customers as well as to the system operators. This describes the objectives of the system. and the rest of the book explains how they are accomplished. 1.1. A LITTLE BIT OF HISTORY 1.1.1. B.G. (BEFORE GSM) Public radiocommunication requires sophisticated techniques, and therefore its evolution has always followed very closely the progress of electronics. The idea of instant communication regardless of distance is part of man's oldest dreams, and this dream became reality as soon as technique would allow it. The first implementations of radio waves for communication were realised i n t h e late nineteenth century f o r radiotelegraphy. Since then, i t has been a widely used technique for military communications. T h e f i r s t p u b l i c applications concerned broadcasting (sound, t h e n images): t h i s i s m u c h simpler than radiotelephony as the mobile terminal is only a receiver. The real boom in two-way public mobile radiocommunication systems took place right after the second world war, when the use of frequency modulation and of Figure 1.1 - the concept of cellular coverage Numerou. omnidirectional or .ectori.ed antenna •ite• allots%vide coverage. h phiting the geogriphical ;Area nil'' ak alapping cm erage area.. the lines shim the linuts of the in erlupping coverage of the hereusthe poi> gon. repre.cni the more usual non-overlapping representation. electronic techniques s u c h a s t h e vacuum v a l v e enabled t h e implementation of a real-scale telephony service for cars. The first true mobile telephone service was officiallv horn in St Louis (Missouri. Ii5A1 in 1946. Europe. recovering from war, followed a few y ears 4ater. The first mobile telephone networks were manually operated (that is to say that the intervention of an operator was required,to connect the call to the wireline network) and the terminals were heavy, bulky and expensive. The service area was restricted to the coverage of a single emission and reception site (single-cell systems). Little radio spectrum was available for this kind of services. since it was allocated with priority to military systems (this has not changed much!) and br:uadcasting tin particular television). As a consequence. the capacity of the 'early systems was small and saturation came quickly despite the high•cost, of terminals which deterred many a potential customer. Quality of service decreased rapidly with congestion. and the throughput sometimes fell drastically due to near-deadlock situations. Between 1950 and 1980. mobile radio systems evolved, to become automatic a n d t h e costs decreased d u e t o t h e introduction o f ,semiconductor technology. Capacity increased a little but remained too 26 Ii\l, I I I I TI II: (iSNI SYSTNNI i • small compared to potential demand: public radiotelephony remained a luxury product for a chosen few. Country Systems Freq. band Date of laurich Subscribers (thousands) During the 70's,' large-scale integration of electronic devices and the development o f microprocessors opened t h e d o o r t o t h e implementation of more complex systems. Because the coverage area of one antenna is mainly limited by the transmitting power o f mobile stations, systems were devised with several receiving stations for a single transmitting station. They allowed coverage of a larger area at the cost of additional infrastructure complexity. But the real breakthrough came with cellular systems, where both transmitting and receiving sites are numerous and whose individual coverage areas partially overlap (see figure 1.1). United Kingdom 'I'M'S 90(1 .1985 1200 450 1981 900 1986 Instead of trying to increase transmission power. cellular systems are based on the concept of frequency re-use: the same frequency is used by several sites which are far enough one from the other, resulting in a tremendous gain in system capacity. The counterpart is the increased complexity. both for the network and for mobile stations which must be able to select a station among several possibilities. and the infrastructure cost due to the number of different sites. The cellular concept was introduced by the Bell Labs. and was studied in various places in the world during the 70s. In the US, the first cellular system, the AMPS (Advanced Mobile Phone Service) became a reality in 1979 when the first pre-operational network was opened in Chicago, I l l i n o i s . I n t h e N o r t h -European countries, t h e telecommunication administrations together with some manufacturers devised the NMT (Nordic Mobile Telephone) system which aimed at a Scandinavian coverage. The system started operation i n Sweden i n September 1981, and shortly afterwards i n Norway, Denmark and Finland. Networks based on these two sets of specifications account for the great majority of mobile networks throughout the world in the early 90's. For example, the TACS, derived from AMPS, was put in service in the UK i n 1985. Most European countries have one o r more cellular networks today. Table 1.1 shows the major cellular networks in operation in Europe in early 1992. All these cellular systems a r e based o n analogue speech transmission with frequency modulation. They all use frequency bands either around 450 MHz or around 900 MHz. Their coverage is usually nation-wide and their capacity reaches several hundreds of thousands subscribers. The largest national system in Europe (composed o f two Scandinavia I 130(1 (Sweden. Norwa), Finland. Denmark) NMT France Radiocom 2000 NM! 450,900 450 Italy RTN15 TACS Germany • 1985 1989 300 90 450 9011 1985 1990 6(1 560 C-450 450 1985 600 Su iuerland NNII 900 1987 ISO The Netherlands NMI' 450 900 1985 1989 13(1 Austria NMT TAGS 4511 900 1984 1990 60 60 Spain NNIT TAUS 450 900 1982 199(1 60 6(1 Table 1.1 — Major cellulir Nystems in Europe in 1991 Variants of NN1T and TACS b ) far the most widespread • a n a l o g u e cellular system, in Europe. with Scandinavia leading the nay For N M I .ind UK T A ( ' S . countrywide coverage networks) is the British TACS with more than one million subscribers by 1990. The highest population penetration is held by the Scandinavian N M T. with more than 6% o f the Swedish and Norwegian population having a mobile equipment. However, these figures are much higher than the mean European values. For instance, the penetration factor in 1991 in France was about one tenth of the Swedish one. Mobile equipment evolved very rapidly during the late 80's. At the onset, only vehicle-mounted equipment could he built. In the mid-S0'. portable equipment appeared, with a weight of a few kilograms and an autonomy o f a few hours. Handheld equipment first appeared around. 1988; not yet small enough to lit in your pocket. but fitting nicely in your attaché-case. In 1990, the smallest terminals on the market were weighing less than 400 g. and fitted in a coat pocket. 28 T I IE GSM SYSTEM In parallel with this reduction in the size o f mobile equipment. prices have decreased tremendously. Expensive car phones, accessible to a happy few, have cleared the way for pocket terminals affordable by you and me. 1.1.2. THE GENESIS OF A STANDARD von- crrrrr rrrr rrrr rr r From t h e start o f t h e 80's, after N M T started operating successfully, i t became obvious t o several European countries that existing analogue systems had limitations. First, the potential demand for mobile services, even though systematically under-estimated in the early 80's, was larger than the expected capacity o f the existing analogue networks. Second, t h e different systems i n operation o f f e r n o compatibility for mobile users: a TACS terminal cannot access an NMT network, and neither can an N M T terminal access a TACS network. What's more, the design of a new cellular system requires such a large investment that no European country on an individual basis can afford this investment if the only return expected is on its own national market. All these circumstances pointed toward the design of a new system, done in common between several countries. The major prerequisite for a common radio system is a common radio bandwidth. This condition had already been met a few years before. in 1978, when it was decided to reserve a frequency band of twice 25 MHz at around 900 MHz for mobile communication in Europe. The need was clear and the major obstacle removed. It remained to organise the work. The world of telecommunication in Europe always was dominated by standardisation. The CEPT (Conference Europeenne des Postes et Telecommunications) is a standardisation arena which—in the early 80's—included the European Administrations o f Posts and Telecommunications of more than 20 countries. A l l these factors, both circumstantial and market-driven, led to the creation in 1982 of a new standardisation body Ovithin CEPT, whose task was to specify a unique radiocommunication system for Europe, at 900 MHz. The new-born "Groupe Special Mobile" (GSM) held its first meeting in December 1982 in Stockholm, under the chairmanship o f Thomas Haug, from the Swedish Administration. 31 persons from 11 countries were present at this first meeting. Slit'llNri II II S('ENli Date Achievement 1982 "Ciroupe Special Nlobile- is created within CEPT 1986 A Permanent Nucleus is set up 1987 Main radio transmission techniques are chosen. based on prototype evaluation 119861 iugy CISNI becomes an 1:1S1 technical committee 1990 The phase I GSN19001 specifications !drafted 1987-19901 arc fro/en DCS18(10 adaptation starts 1991 First s t e m s ate running (Telecom 91 exhibition. DC'S 15(8) specifications are frozen 1992 All major European CISN1900 operators begin commercial operations 'Fable 1.2 6 5 \1 project milestones The span or GSM work 1.10111 the yen start to commercial service extended over some 10 years. but the :tumid specification work did not start until 1987. In 1990. by request of the United Kingdom. the specification of a version of G5M adapted to the 1800 MHz frequency band was added to the scope o f the standardisation group. with a frequency allocation o f twice 75 MHz. This variant. referred to as DC'S1800 (Digital Cellular System 1800) is aimed at reaching higher capacities in urban areas for example for the type of mass-market approach known as PCN (Personal Communications Network). The elaboration of the GSM standard took almost a decade. Major milestones are shown in table 1.2. and the corresponding stages are described in more detail in the following. 1.1.3. ORGANISATION OF THE WORK The first two years of the (ISM were dedicated to discussions of the fundamental principles. The frequency of meetings and the number of participants increased steadily. At the beginning of 1984, three "working The term "GSM900- will be used to refer to the GSM standard at 9(10 M11/, when it i s necessary to differentiate i t From the 1)CSI800 standard. When such a distinction is not necessiny the term "GSM" will be used. encompassing both GSM900 and DCS18110, 30 ' n t h OSNI SYSTIIM S I T I I N G • 1 1 1 parties" were created: the meetings were split to allow a more in-depth technical work. The amount o f contributions and problems to solve increased steadily, and at the end o f 1985 it became obvious that the number of meetings was insufficient. It was then decided (for the first time i n CEPT) that the working parties would meet independently. reporting t o the G S M plenary meetings which still endorsed the decisions. The role of the different working panics had already been clearly established for some time: 1 , st'l 3 Technical Assembly • W P I (Working party 1), for the definition of services: 1 PT 12 • W P 2 (Working party 2 ) , 11w the specification o f radio transmission; this subject stayed dominant until 1987; • W P 3 (Working party 3), f o r all other issues, i.e., mainly network architecture, specification o f the signalling protocols and of the open interfaces between network entities. Later on, a fourth working party (WP4) \ \ as created t o deal specifically with the implementation of data services. During 1985, a detailed list of Recommendations to be output by the group, following the model of the technical Recommendations of the CCITT ( C o m i t e C o n s u l t a t i f International Telographique e t Telephonique), was discussed and settled. From 1986 onwards, work became centred a r o u n d a m a j o r objective: d r a f t i n g t h e s e recommendations. The list includes more than 100 recommendations, sorted in 12 series. A detailed action plan was generated to follow the progress of this huge task, and was updated at each meeting. In 1991, the list includes 130 recommendations, with a total of more than 5000 pages. These recommendations cover the full specification of the radio interface (i.e., the interface between the mobile stations and the infrastructure) as well as fairly detailed specifications of infrastructure architecture and of some of the interfaces and signalling protocols between network entities. In order to co-ordinate the work between the working parties and to manage the edition and the updating o f the recommendations, a "permanent nucleus" (PN) was created at the beginning o f 1986. I t consisted of a small team of full time members and was located in Paris. In 1988, the European Telecommunications Standard Institute (ETSI) was created and most o f the CEPT technical standardisation activities were transferred to this new body, including GSM. Contrary to CEPT. ETSI is not restricted to Administrations, but includes members from industry and user groups as well as operators. GSM had already anticipated this need by officially allowing industry representatives to participate directly to the working parties on an ad hoc basis since 1987. 1 GSM NA A \ \ S P S A I \ \ GSM 1 G S M 2 G S M 3 G S M 4 Figure 1..2 - Partial v i e t h e ICI-SI iqg,inh.ation ale ant to (ISM ETSI Technical committees such as OSSI. NA (Network Aspects) and SI'S (Signalling. Protocols and Switching; report to the Technical Ast.embl) ETSI Project 'leant 12 support: the uork Te c h n i c a l Commince (ISM. With the transfer to ETSI. most GSM Recommendations were due to become 1-ETSs (Interim European Telecommunications Standards). then ETSs. Becoming an ETS requires several stages o f approval. including public enquiries and voting. and this process takes several months. I n the meantime. GSM Recommendations are called "GSM_ Technical Specifications" (GSM TS). In this hook we will refer to the set of the GSM 'Technical Specifications as the SpecOhmions. When integrated in the ETSI structure. the Working Parties became Technical Sub-Committees STCs. or "Sub Technical Committees'' as the official ETSI terminology stands: GSM itself is a Technical Committee, 'V('. of ETSI) and are named GSN11, GSM2. GSM3 and GSM-). The Permanent Nucleus became the Project Team n"I 2 (PT 12) o f ETSI. Figure 1.2 shows the ETSI entities involved in the GSM project in 1990-1991. the date at which the first GSM Technical Specifications were published. 32 T H E At the end of 1991, activities concerning the post-GSM generation of mobile communications were added t o the scope o f the GSM Technical Committee, which was renamed S M G ("Special Mobile Group"), with Technical Sub-Committees SMG1 to 4 being the same as the previous GSM1 t o 4 , and SMG5 dedicated t o the post-GSM generation, the UMTS ("Universal Mobile Telecommunication System"). A sub-group first established under the responsibility of the Permanent Nucleus to draft specifications in the area of Operation and Maintenance became SMG6 in early 1992. The name of the group was changed from GSM to SMG to distinguish it from the 900 MHz system: the term GSM is still kept to refer to the standard and to the corresponding system. Moreover, the term GSM has been chosen as the commercial trademark for the European 900 MHz system, meaning in that context "Global System for Mobile communications", with the corresponding logo: GSITI The meaning of the dots in this logo, i f any, is unknown to the authors and is left to the reader's imagination. 1.1.4. THE GSM MoU In parallel with the drafting of technical specifications within the GSM committee, European public telecommunication operators (most of them GSM-operators-to-be) recognised the importance o f co-operation for commercial and operational aspects. and signed a Memorandum of Understanding i n Copenhagen, o n September 7 t h . 1987. T h i s memorandum, generally referred to as the GSM MoU, covers areas such as time-scales for the procurement and the deployment o f the system. compatibility o f numbering a n d routing plans. concerted service introduction, harmonisation o f t a r i ff principles and definition o f accounting procedures. The memorandum was later signed b y more operators, and amended in mid-1991 to accept members from non-CEPT countries and extend its scope to cover co-operation agreements with non-signatory bodies. The European operators which have signed the MoU in early 1992 are shown in figure 1.3. In addition to operators. regulatory bodies also signed t h e G S M MoU, such a s t h e D T I (Department o f Trade and Industry) in UK. DCS 1800 operators have their own association, which holds close relations with the GSM MoU. From 1990 onwards, GSM started to spread outside Europe: a number of other countries, such as The United Arab Emirates, Hong ;I NEEll G T111- M•I: • F. GSM SYSTEM OP • • • • • • • • •; c-.0.t • • • • • • • • • • 1 • • • • • • GSM MoU operator Figure 1.3 - European GSM ' .ignatories in early 1992 In n i o , t E u r o p e a n c o u n t r i e s . t w o Or m u t e o i l e r : 11 0 N h a v e b e e n l i c e n s e d f o r G S M andha\e`ighed the Memorandum or Understanding. Competition i. becoming the rule and more CotillIdes \k ill deregulate the area. Kong. N e w -Zealand and Australia. envisaged adopting t h e (iSNI standards. In 1992. Australian operators officially became the first nonEuropean signatories ()I' the GSM Moth 1.1.5. TECHNICAL CHOICES Some of the aims of the system \\ ere clear from the start: one of these aims was that the system should allow rice roaming o r the subscribers within Europe. Practically speaking, this means that a subscriber o f a given national network may access the service when travelling abroad. The same GSM mobile station must enable its user to call or he called anywhere within the international coverage area. 34 THE GSM SYSTFNI SFITTINC, II IN SChNI:. Services Network aspects —The system shall be designed such that mobile station, can he used in all participating countries. 35 —T h e identification plan shall he based on the relevant ( V I V I Recommendation. In addition to telephone traffic, the sysiem must allow maximum flexibility for other types of services. e.g. ISDN related services. —T h e numbering plan shall be based on the relevant C('11-1 Recommendation. The services and facilities offered in the PSTN/ISDN and other public networks should as far as possible be available in the mobile system. The system shall also offer additional facilities, taking into account the special nature or mobile communications. —T h e system design must permit different charging strictures and rates to he used in different networks. —F o r the interconnection of the mobile sw itching centres and location registers. an internationally standardised signalling system shall be used. —I t should be possible for mobile stations belonging to the system to be used on hoard ships, as an extension to the land mobile service. Aeronautical use of GSM mobile stations should be prohibited. No significant modification of the Ilse(' public networks must he required. —T h e GS\1 system shall enable implementation of common cos erage PLNINN. In addition to vehicle-mounted stations, the system shall he capable of providing for handheld stations and other categories of mobile stations. - Protection of signalling information and 'network control information must be provided l'or in the system. Quality of service and security Cost aspects - From the subscriber's point of view, the quality for voice telephony in the GSM system shall be at least as good as that achieved by the first generation 900 MHz analogue systems over the range ()I' practical operating conditions. The system parameters shall he chosen w 'oh a view to 111101 the cost Of the complete system. in particular the mobile units. —The system shall be capable of offering encryption of user information but any such facility should not have a significant influence on the costs of those parts of the system used by mobile subscribers who do nut require such facility. 'Fable 1.3 B a s i c requirements set out by (ISM {original test as written ht the committee in 19851 These sy stem ohjecti‘ es w ere :Treed and distrilYtited to the telecommunications indusny k ‘ itli the o f providing a clear franick‘ork for progiesmig the technical design. Radio frequency utilisation —The system concept to be chosen shall permit a high level of spectrum efficiency and state-of-the-art subscriber facilities at a reasonable cost, taking into account both urban and rural areas and also development of new services. —The system shall allow for operation in the entire frequency hand 890-915 MHz and 935-960 MHz. —The 900 MHz CEPT mobile communications system 11111tii co-exist with earlier systems in the same frequency band. It w a s clear as w e l l that t h e capacity M e r e ( ' b y the system s h o u l d be b e t t e r t h a n w i t h e x i s t i n g a n a l o g u e ilk:m.01.k. i n o r d e r t o l a k e m e r V without reaching saturation t o o q u i c k l y. As earls. a s 1 9 8 2 . t h e b a s i c r e q u i r e m e n t s f o r \ - 1 w e r e stated. They w e r e s l i g h t l y revised i n 1 9 8 5 , as reproduced v e r b a t i m i n table 1.3. and are s t i l l rather up-to-date. S i n c e M e n . portable and h a n d h e l d m o b i l e stations became the main objective. o v e r vehicle- m o u n t e d stations. 1.1.5.1. I S D N as Godfather At a b o u t t h i s t i m e , t h e specification o h t h e Integrated Service ira I N r t w n r k ( I SDN I w a s P h a s e l o c o m p l e t i o n a n d t h e . l l l i (iSNI SYS IltN1 NI. I I 1St; I III S t I S I telecommunication trends were such that GSM was necessarily designed as a system offering access t o a large variety o f telecommunication services: speech, data, voice and alphanumeric messaging. videotex. and so on. selected. Only the key features o f the transmission method were decided: they were summarised i n w h a t w a s called t h e "broad avenue". T h e narrowing would have to take place afterwards. within GSM2. The key. features were the following: From its origin i n 1984, and under the strong impulsion o f its chairman, the work in GSM3 was inspired by the principles of ISDN and its access protocols. T h e i r influence i s obvious w h e n reading t h e Specifications, b o t h i n t h e a r e a o f services d e f i n i t i o n a n d i n t h e description o f signalling protocols. The use o f a layered model such as the 0 5 1 (Open System Interconnection) model f o r t h e definition o f protocols is but an example of such an influence, and with regards to this aspect G S M differs f r o m the foregoing analogue radiocommunication systems. The kinship between G S M and I S D N allows some optimism when considering their integration, G S M networks being just another way to access the general telecommunication networks. • medium-sized b a n d ( 2 0 0 k l lz c a r r i e r separation). t o h e compared to narrow-hand systems (12.5 k H z or 25 k H z as i n existing analogue systems) o r wide-band systems (one o f the candidates proposed a 6 M l l z carrier spacing): 36 I 1.1.5.2. Born Digital General trends in the telecommunication arena made it unofficially clear f r o m t h e start t h a t t h e system w o u l d b e based o n d i g i t a l transmission, and that speech would be represented by a digital stream at a rate o f the order o f 16 kbit/s. The official decision. however, was not made until 1987. From 1984 t o 1986, G S M focused o n t h e means t o compare different technical possibilities for transmission (digital or analogue), in particular with regards t o their respective spectrum efficiency. I t was decided to compare several technical proposals on the basis of prototypes allowing actual radio transmission. In 1985, the French and West German P&T Administrations joined their efforts to order four studies leading to such prototypes. Comparative testing of eight prototypes, including these four plus four more Scandinavian prototypes was conducted in December 1986 a t t h e C N E T laboratories ( C e n t r e N a t i o n a l d ' E t u d e s d e s Telecommunications) near Paris, under the control o f the Permanent Nucleus. A l l these prototypes made use o f digital transmission, and most were proposed by industrial companies. 1.1.5.3. The "Broad Avenue" The results o f the comparison were reported at the beginning o f 1987. T h e discussions were difficult, because o f the prestige and the advance that would be conferred to a proponent were its solution to he chosen. To circumvent this problem, none o f the proposed solutions was • d i g i t a l speech transmission at a rate not exceeding 16 khit/s: • t i m e multiplexing o f order 8 . w i t h future evolution towards multiplexing of order 16 once a second-generation speech coder has been defined at a smaller rate; • s l o w frequency hopping capability. 1.1.6. THE GSM TECHNICAL. SPECIFICATIONS Originally. the w o r k was planned i n three phases: specification writing, validation, then f i e l d tests w i t h operational equipment. T h i s phasing was, however, modified l i t t l e b y little. A t the beginning, the specification o f all features of the GSM were scheduled to he ready at the same moment, i n time f o r validation and tests before an opening o f service i n 1991. There w a s o n e b i g exception. t h e h a l f -rate speech coding. Because i t was clear that speech encoding techniques evolve quickly, the plan was to have a first speech encoding algorithm at around 16 kbit/s (it uses 13 kbit/s). and a few years later. a more efficient scheme using half as many bits. About twice as many subscribers could then be accommodated within the same spectrum allocation, with a radio channel for carrying 13 kbit/s speech (a full rate channel) being replaced by t radio channels to carry speech encoded w i t h the future algorithm. The whole transmission system was designed from the start t o support the future h a l f -rate speech c o d i n g a s m u c h a s possible. T h e d a t e o f introduction of this speech coding, algorithm was not precisely planned. - T h e idea o f separating M u validation phase from the specification writing phase quickly disappear6d. A s time went,, it was more and more • obvious that it would not he adequate to write all the specifications first, i.e., i n a very short period o f time. T h e idea o f a phase 2 step as a functional enhancement o f phase I gradually crystallised. Around 1988. the idea matured. and it was agreed that the launch in 1991 would not be with the full palette of services. A t this date. services were split between two phases: t h e most important i t e m s l e f t f o r phase 2 w e r e m a n y supplementary. services. phase I being limited to the most common ones 38 T I I E 19 GSM SYSTEM (such as call forwarding and call barring), and alternate services (for instance speech alternating with data within the same communication). Later on. the capability of offering data services on half rate channels was transferred to phase 2, though all the required specifications existed. This last point was decided at the end of 1991, and was a side effect o f the ongoing work on half rate speech coding: it was considered better to open the possibility to change the specification of the half rate radio channel, in order to allow more flexibility in the transmission improvements. Such a change would be very difficult to implement i f half rate channels were already used for data services. During the w o r k concerning DCS1800, t h e need f o r n e w functionalities (such as national roaming) appeared. Some o f these functionalities are included in DCS1800 phase I . which is thus slightly richer than GSM900 phase I . T h e added functionalities w i l l b e incorporated in the phase 2 (a single set of the phase 2 GSM Technical Specifications will cover both DCS1800 and GSM900). In the middle o f 1988. first drafts o f the G S M Technical Specifications were available. From this date up to the middle of 1991. the Specifications evolved under a control procedure o f "change requests": each change had t o be justified i n terms o f objectives. complete (including impacts on all other applicable GSM Technical Specifications), and precise. Such a mechanism was necessary to trace all changes, i n particular at a period when manufacturers were already deeply involved i n the development process. The rules for accepted changes became more and more strict, so that at the end the only accepted changes were those necessary t o correct malfunctions o r dangerous ambiguities. Ideas for enhancement, not acceptable at a late date simply because not absolutely necessary, were not discarded, but generally forwarded to the phase 2 stage. Nowadays (at the date of writing, middle of 1992), a new set of GSM Technical Specifications is scheduled for approval soon. These phase 2 Technical Specifications will include not only the specifications of the new services (supplementary services and facsimile), but also numerous minor and major improvements. One o f the big issues that came to the fore in 1991 was upward compatibility. Mobile stations designed according to the phase I Specifications will exist for some time. and they will therefore coexist in the network with phase 2 mobile stations during several years. Similarly, infrastructure equipment throughout the whole system will not be updated to phase 2 specifications during one night. Upward compatibility is basically the problem of how to specify phase 2 mobile and infrastructure equipment so that this coexistence is possible. The notion o f an evolutive standard emerged quite late, and a number of small points in the phase I Sperification.c. not considered from this vantage at the time of decision. rendered upward compatibility difficult or even impossible in some areas: these parts \c etc then frozen out of future evolution. This problem limited quite severely which changes could he accepted for phase 2. and often made the technical solutions more complex. However, awareness o f the crossphase compatibility issue grew with time and a few general mechanisms have been introduced to reduce constraints in future phases. Phase 2 is not however the end of the stor. More enhancements are scheduled. Hall rate speech is not ascribed to any phase. and will then introduce another step. Its specifications are foreseen to he completed h‘ end 1993. To sum up. the original thinking was o f a one shot specification work. Nowadays. GSM is an evolving standard. In this domain also. GSM marks a boundary. The reasons for the change are the level o f complexity o f the SfieelfiCtIlioll.\ and the ever accelerating rate o f technical evolution. Future VLSI and microprocessor performances will allow more and more complex systems. The time between successive technical generations is now shorter than the life time of a system. the latter being constrained by financial considerations: the infrastructure cannot be replaced before ha ing been paid off. This trend has no reasons to slacken off. and the GSN1 story tells us how important it is to manage a standard as an evolutive object from the start. This book will deal mainly with the phase I Specificatiom o f GSM. since these are the only officially-approved standards at the date of writing. Now and then. some future schemes, or some evolutive trends. will he hinted at (in particular in relation with half rate channels, because full specifications exist for the management o f these channels and for their application to data services. even though these specifications mav still be changed). but not described in details. 1.2. CELLULAR SYSTEMS 1.2.1. GENERAL ASPECTS Even though the term o f network is often used when speaking about the GSM architecture (-PLMN". Public Land Mobile Network), it would be more proper to consider the system, as designed. as an access to existing telecommunication networks. aimed at users with GSM mobile stations. A GSM PLMN. as presented II> the Spergications, is indeed not 40 T H E GSM SYSTEM sETTINGillk 4I subscribers. For all other call configurations, the PLMN relies on fixed networks to route the calls. Most of the time the service provided to the user is the combination o f the access service provided b y a GSM network, and o f the service provided b y some fixed network. This approach has an important impact on the Specifications: for example. no switch hierarchy is defined, and transport of information between GSM machines is based on the usual 64 kbit/s units multiplexed onto 2 Mbit/s links, o r higher rate multiplexes. However, nothing i n the technical choices really limits GSM to be an access network. One can imagine in the future, i n some countries without extensive wireline-access infrastructure, that GSM, complemented with a suitable switch hierarchy, would be used as the basic telecommunication network. Considering GSM as another "local-loop" leads us to compare its characteristics with wireline telecommunication access, highlighting two major differences: • first, the wide-area mobility of subscribers leads them to change their point o f access t o the network; this poses a serious problem for the routing of calls toward subscribers; this is the realm of mobility management; • second, the link between the subscriber terminal and the fixed infrastructure is not permanent and is subject to fluctuating transmission requirements; this is the realm of radio resource management. The consequences o f both aspects will now be described, after some preliminary considerations on the cellular coverage. 1.2.2. CELLULAR COVERAGE The m a j o r problems w i t h r a d i o distribution arise f r o m electromagnetic wave propagation. With a decreasing weight and an increasing autonomy, the mobile terminals have a limited transmission range. Every telecommunication engineer will remember that the power of radio waves decreases with the inverse of the squared distance ((t-2): however, it must be remembered that this applies only in empty space. Between t w o stations close t o t h e ground, interference-creating reflections from the ground cannot be neglected, and it is very likely that obstacles intervene on the direct path between them. As a consequence. propagation at ground level in an urban environment is more difficult, and the received power varies typically with ct-4! Figure 1.4 — Cellular coverage representation Hexagons nicely pave the plane without overlapping and are commonly used for calculating theoretical frequency reuse in cellular systems. A second problem i s • spectrum scarcity: t h e number o f simultaneous radio communicatiolus supported by a single fixed station (a base station) is therefore limited. Cellular coverage allows ,a high traffic density •in a w e area despite both problems. at the expense o f infrastructure cost anti o f complexity. Because of the limited transmission range of the terminals, cellular systems are based o n a large number o f reception and transmission devices o n the infrastructure side (the base stations). scattered over the urea to •cover. and euch one covering a faith small geographical zone called a cell. The underlying image is the one of membrane composed o f epithelial cells. A minimum density o f filed stations allows low-pinker mobile stations to access s skint :111\VI here within a wide area: they are never ‘‘.- f a r from a base. station. Cells are often represented by hexagons, in order to model the s)slein by pax the plane with a single geometrical figure (see figure 1.4). Ilmcever. a typical coverage looks more like figure I . I . Spectrum scarcity is circumvented b> the reuse of radio resources. Frequencies used in a given cell are reused a few cells away, at a distance sufficient so that the unavoidable interference created by the close use of the same spectrum has fallen to an acceptable level. which depends in -12 ' r i f t : G S M SYSTEM reused in every ninth cell, a spectrum allocation of N frequencies allows N/9 carriers to be used simultaneously in any given cell. The total system throughput, often expressed in number of simultaneous calls per kni2 per MHz, c a n therefore b e increased b y reducing t h e c e l l s i z e . notwithstanding the limited spectrum available. In practice cell reduction has some other effects, and the design of a cellular system rarely copes for all cell sizes from 0 to infinity. In the case of GSM, the design was aimed at the beginning at medium-sized cells, of a diameter expressed in kilometres or tens of kilometres. Yet, the lower boundary is difficult to determine: cells of more than one kilometre radius should be no problem. whereas the system may not be fully suitable to cells with a radius below. say, 300 meters. One source of limitation is more economical than due to physical laws. The efficiency of the system decreases when cell size is reduced, and then the ratio between the expenditure and the traffic increases, a n d eventually reaches a p o i n t w h e r e economical considerations call for a halt. Another important point is the capacity of the system to move a communication from one cell to another rapidly. and GSM requires too long a time to prepare such a transfer to cope with fast moving users in very small cells. The cell size upper bound is more obvious: a first, non-absolute, limitation i n GSM i s a range o f 35 kilometres. Cells o f bigger size are possible, but require specially designed cell-site equipment and incur some loss in terms of maximum capacity. The number of sites to cover a given area with a given high traffic density, and hence the cost of the infrastructure, is determined directly by the reuse factor and the number of traffic channels that can be extracted from the available spectrum. These two factors are compounded in what is called the spectral efficiency of the system. Not all systems allow the same performance i n this domain: they depend i n particular on the robustness of the radio transmission scheme against interference, but also on the use of a number of technical tricks, such as reducing transmission during the silences of a speech communication. The spectral efficiency, together with the constraints on the cell size, determines also the possible compromises between the capacity and the cost of the infrastructure. All this explains the importance given to spectral efficiency. Many technical tricks t o improve spectral efficiency were conceived during the system design and have been introduced in GSM. They increase the complexity, but this is balanced by the economical advantages of a better efficiency. The major points are the following: • t h e control of the transmitted power on the radio path aims at minimising the average power broadcast by mobile stations as well as by base stations, whilst keeping transmission quality s r I \ ti I I S c I A I above a given threshold. This reduces the level of interference caused to other communications: • frequency hopping improves transmission quality at slow speeds through frequency diversity. and improves spectral efficiency through interferer diversity: • discontinuous transmission. w h e r e b y transmission i s suppressed when possible. allows a reduction in the interference level of other einnmunications. Depending on the type of user information transmitted. it is possible to derive the need for effective transmission. In the case of speech. the mechanism called VA D (Voice Activity Detection) allows transmission requirements to he reduced by an important factor (typically. reduced by half): • t h e mobile assisted handover. whereby the mobile station provides measurements concerning neighbouring cells. enables efficient handover decision algorithms aimed at minimising the interference generated b y t h e c a l l ( w h i l s t keeping t h e transmission quality above some threshold). We will come hack to this subject after the technical aspects of the standard have been described. i n Chapter 9 concerned w i t h t h e engineering of the system. 1.2.3. RADIO INTERFACE MANAGEMENT Since the number of available radio chapels is much smaller than the total number o f potential users ( a safe assumption!), channels enabling bi-directional communications are only assigned at need. This is a major difference with standard telephony. where each terminal i s continuously linked to a switch. be there a call in progress or not. Such perenniality of the link allows rather simple call set-up procedures: sonic device in the switch continuously monitors the line for changes bem the "on-hook- and "off-hook" status in order to detect outgoing calls. w*ereas the user terminal is always ready to detect a ringing tone carried by the subscriber line when an incoming call is received. In a mobile network, radio channels must he allocated and released dynamically, on a call basis. This function is additional to the usual fixed network call handling procedures. • Moreover. reaching the subscriber is not an easy problem. In GSM as well as in most other cellular systems. the user. when not engaged in a call, learns about an incoming call by listening to a specific channel. This 44 T H E SETUINC VI III. ` W I N E GSM SYSTEM channel carries messages called "paging messages": their role i s t o indicate that a given mobile subscriber is being called. Such a channel is broadcast in every cell, and the problem of the network is to determine in which cells a given subscriber should be called when needed. The setting u p o f any call, whether mobile originating o r terminating, requires specific means, by which the mobile station may access the system i n order t o get a channel. I n GSM, this access procedure is performed on a specific mobile t o base channel. This channel, together with the base to mobile broadcast channels transporting in particular the paging messages, is known as a common channel in GSM, since it carries information to and from many mobile stations at the same time. Conversely, channels allocated to a single mobile station for some period o f time are called dedicated channels. Based on this distinction, two "macro-states" of the mobile station may be defined: • i n idle mode, a mobile station listens to broadcast channels: it has no channel of its own. • i n dedicated mode, a bi-directional channel is allocated to the mobile station f o r its communication needs, allowing i t t o exchange point-to-point information with the infrastructure in both directions. The access procedure is the particular function which allows the mobile station to reach the dedicated mode from the idle mode. 45 send a paging message in all cells of the network when a call arrives. removing the need for the mobile station to advise the network o f its current location: this i s ubiquitous paging. The third method i s a compromise between the two first extreme methods. introducing the concept of location area. A location area, is a group of cells, each cell belonging to a single location area. The idehtity of the location area a cell belongs to is sent in the cell on a broadcast channel. thus enabling mobile stations to be informed of the location area they are int When a mobile station changes of cell, two cases may arise: • b o t h cells are in the same location area: the mobile station does not send any information to the network: • t h e cells belong to two different location areas: the mobile station informs the network o f its change o f location area (location updating). When an incoming call arrives, a paging message needs only be sent in those cells belonging to the location area where the mobile station has last performed location updating. This third method allows to halahcc the amount of paging messages (which increases when the location area includes more cells) with the amount o f location updatings (which increases when the location area includes less cells). GSM supports this method. 1.2.4.2. H an dover 1.2.4. CONSEQUENCES OF MOBILITY 1.2.4.1. Location M a n a g e m e n t The mobility of users in a cellular system is the source of major differences with fixed telephony, i n particular for incoming calls. A network can route a call towards a fixed user by simply knowing the network address (e.g., the telephone number) of this user, since the local switch to which the subscriber line is directly connected does not change. However, i n a cellular system, the cell i n which contact may h e established with the user changes when the user moves. I n order to receive incoming calls, a mobile user must first he located, i.e.. the system must determine in which cell he currently is. In practice, three different methods may be used t o gain this knowledge. In the first method, the mobile station indicates each change of cell to the network: this is systematic location updating at cell level. When a call arrives, a paging message needs to he broadcast only in the cell where the. rrinhan . . The preceding section (location management) deals w i t h the consequences o f mobility i n idle mode. I n dedicated mode, and i n particular when a call is in progress. the mobility of the user may also induce t h e need t o change t h e serving cell. i n particular when transmission quality drops below a given threshold. With a system based on large cells. the probability of such an event is small and the loss of a call in such conditions man be acceptable. I lawc%cr. the achin einem ot high capacities requires the reduction o f cell site. and maintaining the calls despite user mobility becomes an essential requirement to avoid a high degree of customer dissatisfaction. The process of automatically transferring a transaction in progress (a call in particular) from one cell to another to avoid the adverse effects of user movements is called handover. This process requires first some means o f detecting the need to change cell while in dedicated mode (handover, preparation). and second the means to switch a communication GSNI s y s n i m SLITING 111F S i t \ k from a channel in a given cell to another channel in another cell, ideally in a way not noticeable by the users, and at least keeping user disturbance to a minimum. unproved commercial interest. Similarly. the European 900 MHz hand may not he available in other parts of the world. and using a different bandwidth would preclude using the same mobile equipment for roaming. But in all such cases, it is possible to envisage another kind of roaming. based on using the subscriber-specific part of the mobile station only. in connection with a radio access part specific to each network. Such a combination w i l l enable users t o have a single subscription and he reached through the same directory number. whatever network they be roaming into. A way to achieve this aim is already included in GSM and will be expanded upon later on in thisfichapter. when the SIM (Subscriber Identity Module) is introduced. • 46 T I I E 1.2.5. ROAMING In telecommunication systems accessed through a fixed link, the choice of which network provides the service is done (when choice is possible!) at subscription, once and for all. When mobility is introduced. new horizons emerge. Because a mobile terminal is not on a leash. different networks can provide service to a given customer, depending on where he is. When different network operators co-operate, they can use this possibility to offer to their subscribers a coverage area much wider than any of them could do on its own. This is what is called roaming. and it i s one o f the major features o f the pan-European GSM, whose subscribers can enjoy European-wide coverage, whatever their national network of subscription. Roaming can b e provided o n l y i f some administrative and technical constraints are met, and these points w i l l be addressed i n various parts of this book. From the administrative point of view, issues like charging, subscription agreements, etc. must be solved between the different operators. The free circulation of mobile stations also requires regulatory bodies to agree on the mutual recognition of type approvals. From the technical point of view, some topics are a consequence of the administrative matters, such as the transfer of call charges or the transfer of subscription information between networks. Others are needed simply for roaming to be possible at all, such as the transfer of location data between networks, or the existence of a common access interface. This last point is probably the most important one. It requires a subscriber to have a single piece of equipment enabling him to access the different networks. To this avail, a common air interface has been specified, so that the user can access all the networks with the same mobile station. Other GSM-based (or DCS 1800-based) systems will be created. Roaming between these systems and the European GSM may not always be possible with the same mobile station. A possible limitation i s bandwidth. European countries have agreed to use a common part of the spectrum, at 900 MHz. Another band is already possible, at 1800 MHz. Roaming with the same mobile equipment i s not possible between GSM900 and DCS1800, except with dual band mobile station. of as yet 4 7 1.3. GSM FUNCTIONALITIES In this section we will describe the functions of GSM as seen 11\ the users. that is to say the services that are provided to users. abstracting the details of how it is clone. 1.3.1. GSM: A MULTISERVICE SYSTEM FOR THE USER GSM i s a multiservice system. allowing communications o f various types. depending on the nature ol• the transmitted information as perceived b y the end users. B y tradition, one distinguishes speech services from data services: in speech services. the information is voice. whereas the term "data services" groups everything else. such as text. images, facsimile. computer files. messages, and so on. GSM provides a large palette of the services offered to fixed telecommunication users, as will be described further on in this .ection. GSNI provides in addition a non-traditional set of services. the "short message services-. closer to the paging services (one-way radio messaging services) than to any service provided in fixed networks. These services will he described separately from the Other data services. More generally, the definition o f a telecommunication service includes more than just the nature o f transported information. Other characteristics o f the communication are also relevant. such as the 4IS T H E GSMSYSTEM S transmission configuration (point-to-point or point-to-multipoint, half or full-duplex) a n d t h e potential partners, b u t a l s o t h e possible "supplementary services", which refer to the possible control o f various aspects o f the communications by the user, and more administrative notions such as for example the charging aspects. Service provision depends on three independent factors: • t h e contents of the subscription held by the subscriber, in terms of services as well as i n terms o f geographical areas. The subscription packages offered by each operator will vary, as well a s t h e corresponding subscription rates, a n d each subscriber will perform a choice between these packages: • t h e capabilities of the network from which the user is getting service. A l l networks will not offer exactly the same range of services at a given date, and therefore the user might find some restrictions on the available services depending on the network in which he is currently roaming: • t h e capabilities of the equipment held by the user. For instance. it is obvious that faxes cannot be sent or received on a speechonly mobile station i f i t is not connected t o a proper fax machine. The basic services provided b y a telecommunication network. which consist in transmission media and in the means to set up calls, are distinguished from a number of supplementary features enabling users to better control the provision of these basic services, or to simplifying the daily use o f telecommunications. These features w i l l represent an important contribution to the user comfort, and include the possibility for the user to forward calls, to visualise the calling number, etc. Some of these features are performed locally by the terminal, and are not properly speaking provided by the network (though the limit is difficult to tell for the user). Other features are provided by the network, and are called the "supplementary services". Several reasons warrant t h e distinction between basic and supplementary services, including the fact that supplementary services apply generally t o several basic services (and hence a separate presentation is better than repetition): and also because in many existing networks these supplementary services must be asked and paid for in addition to the basic services. In the future it is quite likely that some of these supplementary features will be automatically part of a service package. This statement holds in particular for those services which improve network efficiency (such as the forwarding of calls which cannot be completed). G I I ING III. - 1 9 The description ofthe services provided by GSM follows. We will first address the speech ;rvices. the data sere ices and the short message services. T h e n t h e "supplementary services" w i l l h e considered. Afterwards, a small section will deal with the "local services", i.e.. the facilities ()tiered by the GSM terminals on their own. Then an important particularity o f the G S M mobile terminal w i l l b e described: t h e Subscriber Identity Module (SIM). Finally, the security teat ures of GSM. another complement of service for the users, will be addressed. 1.3.1.1. Speech Services The most important service provided by GSM is telephony. This service enables bi-directional speech calls to he placed between GSM users and any telephone subscriber reachable through the general telephony network. Fixed telephone subscribers world-wide can be reached as well as mobile network subscribers or subscribers of specific networks connected to a public telephone network. With the rise o f ISDN, telephony has been somewhat eclipsed by data services, which are often presented as the future o f telecommunications. However, speech remains and will, remain the most important service for mobile systems. The absence of a fixed wire and the arrival o f handheld terminals (less than 400 g. and fit to he carried in a pocket) make cellular telephony the foremost technique f o r inter-personal communications. the true telecommunication method extending t h e m o s t natural f o r m o f communications that human beings have used for maybe several hundred millennia. Following the GSM official terminology. emergency calling is a distinct service. derived from telephony. II allows the user of a 'nubile station to reach a nearby tinergenc) service (such as police or the lire brigade) through a simple and unified procedure. by dialling 112 (the number which has been agreed :is the standard emergency number throughout Europe,. Another service derived From lelePhonv is voice messaging. The Spec/Few/um do not identify this set ice as a separate one. but mon\ an operator will offer it as a basic feature. It enables a voice message to he stored for later retrieval by the mobile recipient. either because he was not reachable at the time o f the call_ or even because thst. L• calling park. chose to access directly the voice mailbox of the GSM subscriber.sul.iT 50 T I Ili C;SN1 SN'STENI S 1.3.1.2. Data Services As an heir of ISDN, GSM has been designed from the start to offer many data services. Basically, most services which are provided to fixed telephony users and to ISDN users (thus covering access to other more specialised networks such as Packet Switched Public Data Networks) have been included, as far as the limitations due to radio transmission allow. This corresponds to a very large number of different cases (the Specifications list some 35 services), t o cope with all the variants stemming from the history of telecommunications. Data services a r e distinguished m a i n l y b y t h e potential correspondents (users o f the telephone network. o f ISDN, o r o f specialised networks), by the nature of the end-to-end information flow (raw data, facsimile, videotex, teletex, ...), by the mode of transmission (packet or circuit, end-to-end digital or making use of an audio modem. synchronous or not, ...), by the nature of the terminal, and so on. For practical reasons, data services shall not be presented here with the full descriptive methodology used in the Specifications (i.e., as tables of attribute values), but shall be grouped according primarily to the type of potential correspondents, that is to say according t o the type o f network to which the other correspondent is a subscriber to. A summary of all data services as specified in the Specifications will then be given. I I I ' I \ r f l !IF St'k \ I. 5 I network, at the point of interworking with the PSTN. For the PSTN user wanting to communicate with a GSNI user. this approach effectively limits the choice o f modulations t o the ones available i n the GSNI network. However, this is not a ,strong restriction, since it' so desired by the operator. GSM can cope with the most widely used standards up to 9600 bit/s full-duplex. such as V.21. V.22. V.22bis, and so on, including V.32 and the modems specific to group 3 facsimile and videotex. • On the GSM user side. specific data terminals. adapted to mobilik (power supply. size). w i l l certainly sooner o r later he proposed b y manufacturers. Besides, the specifications cater for the connection of offthe-shelf terminal equipment (fax. personal computer, videotex terminal) to the mobile stations. This possibility enables the use o f standard equipment designed for PSTN usage. that i s t o say designed t o he connected to a telephone line (as most fax machines). or to a modem itself connected to the telephone line (as personal computers). Thanks to its multiservice nature. GSM will provide in these eases a service a hit better than the PSTN itself. beside the obvious advantages of mobility. The palette of supplementary services will become extensive. A point worth mentioning is the possibility for a subscriber to have several directory numbers. all corresponding to the same subscription. but to different services (for instance one number for speech. and another for • facsimile). Connections with ISDN Users Connections with PSTN Users The Public Switched Telephone Network (PSTN) is already used widely today for data transmission, with audio modems, not by vocation but simply because o f its wide availability. I t was then mandatory for GSM t o provide i t s users w i t h t h e same possibilities o f data communication as are available between PSTN users, in particular the most popular data services: group 3 fax and videotex. Unfortunately, this cannot be done as simply in GSM as in the PSTN (which in fact provides only "speech" paths, strictly speaking 3.1 kHz audio transmission paths), because o f the characteristics o f t h e radio transmission. Vo i c e transmission on the GSM radio interface is based on coding algorithms optimised for speech, making its representation unsuited t o standard modem signals. In GSM, the user does not need to provide a modem for such communications. The transmission between the mobile terminal and the network is fully digital, and an audio modem i s provided inside the Before going further, i t should he noted that the distinction between the ISDN and the PSTN can be quite fuzzy. In many countries. digital ISDN-like transmission and advanced signalling methods are already used on a wide scale within the telephony network, and the transition to ISDN is very smooth. Here we will use (improperly) the term PSTN to refer to the "Plain Old Telephone System". where analogue transmission or less modern signalling methods are used. If connections with PSTN users are a must because of the premint weight o f this network. connections with ISDN users are a must for a digital system because of the future expected importance o f ISDN. A n important point duringthe GSN1 design phase was t o ease as far as possible the interconnection with ISDN. so as to enable the integration of GSM in an ISDN system, with for instance local exchanges serving both GSM and ISDN users. 52 T H E Still. the 9600 bit/s upper bound of GSM data transmission puts a serious limitation o n the interworking with ISDN, which provides basically 64 kbit/s transmission (and higher rates i n the future with broadband ISDN). The weight of history being about the same for ISDN as for GSM, the former must also cope with the provision of data services with PSTN users. This opened a possibility o f interworking between ISDN and GSM, each network considering the other as it does PSTN. Interworking between GSM and ISDN can then be done using the standardised ISDN digital formats developed for PSTN adaptation and enable it to carry low rate digital streams over the 64 kbit/s links. This leads to a somewhat wider range of services than that possible with audio modems through the PSTN, and includes, e.g., asynchronous data transmission at rates over 2400 bit/s, for which no audio modems are specified internationally. The extreme form of this approach consists in using the ISDN as the PSTN (using the 3.1 kHz audio transmission mode and hence audio modems), and with this method the same services are available for communications between GSM users and ISDN users as with PSTN users. This possibility is opened in the Specifications, as an interim method. otherwise than between GSM users since no external network supports it. the transmission of raw data at I 2'kbit/s (this possibility is deemed to be marginal. since it will take some time before corresponding terminals are provided. and because data communications between GSM users will be a very marginal traffic). Connections with Packet Switched Public Data Network Users A Packet Switched Public Data Network (PSPDN). such as the French TRANSPAC network. is a general purpose data network using the packet transmission techniques. as opposed to circuit techniques as used for instance i n t h e P S T N . PSPDNs a r e u s e d primarily f o r communications w i t h o r between computers. They are also often supporting services like message handling systems, or remote data base interrogation. A PSPDN can typically he accessed in three different ways: • using a direct connection for the subscribers (X.25 access) of such a network: • through the PSTN o i r ISDN. via a PA D (Packet Assembler Disassembler, for access via a modem using an asynchronous modulation —X.2)S access) '• through the PSTN o r ISDN. via synchronous access (X.32 access. that is to sap X.25 plus some complements in particular 9 for the identification of the user): Connections between GSM Users GSM has not been designed as an independent network which provides primarily services between its users and, only as a side dish, with users of other networks. On the contrary it has been designed as an access network f o r the fixed telecommunication networks, and the services provided between GSM users have nothing specific. Indeed, in most cases an external network intervenes in a communication between GSM users. The goal is that the ISDN be this interworking network. At the beginning, i t will often be the PSTN instead, and the provided services will be those that can be supported by PSTN transmission. that is to say the same as are,provided between GSM users and PSTN users. When it is certain that the communication is digital end-to-end. using a 64 khit/s link, the service range is wider, the same as in the case of GSM to ISDN. This happens when the transit network is a modern PSTN, or ISDN, or when there is no transit network (for instance for communications local t o a GSM exchange). I n the last case the Specifications open the possibility for an original service, not provided 53 SETTING 1 HE SCENE GSM SYSTEM • a n d through the ISDN. using the capabilities o f ISDN t o transmit packet data. Asa result. several different methods are proposed to a GSM user which desires to communicate with a PSPDN subscriber. These different methods of interconnection are shown in figure 1.5. The differences arc mainly technical. but impact the user in the type or terminal he can use. the available data rate. in the way he has to enter the called number. in the need or not to subscribe to the PSPDN and consequently in how the calls are billed. The less specific method consists in acting as a PSTN subscriber. and to access a PAD through the PSTN. via asynchronous audio modems supported by the PSTN. This allows use of a simple data terminal on the GSM user side (case (a) in figure 1.5). For the GSM PLMN, this type of communication is perceived just as a plain data communication with c()Ille P S T N " C i l l I S I • r i l l t • C . : 11 3 1 111 tirlii'd l • C l I e l l T h i s r o m i i r e c 1 1 w G S M PSTN k (b) P A D 1/4. - , P A D • . < GSM (c) PSTN/ ' H A " ISDN f (d).' (e) ISDN (a), (b) PAD access (c), (d), (e) packet access PAD 1 P a c k e t assembler/dissassembler —1.<1--- p a c k e t interfaces (X.25 or X.75) non-packet interfaces Figure 1.5 — Ways of accessing a PSPDN from GSM The services offered to a GSM user for communication with a PSPDN user differ according to the capabilities of the GSM user's terminal, to the possible subscription he holds with the PSPDN, as well as to possibilities offered by network implementations. user to provide the PAD PSTN number himself, and also to be registered on the PSPDN, so that he is charged for the PSPDN connection (except if' reverse charging is used by the called PSPDN subscriber). The directory number of the desired correspondent is provided by the user afterward, once in contact with the PAD (double numbering). The communication can be set up only from a GSM user toward the PAD, not in the other direction. The approach is general, and enables users to access PSPDN from other countries (including the home country when roaming). The service is called "basic PAD access" in the GSM terminology. N e x — what is called the "dedicated PAD access" (case (b) in figure 1.5). This corresponds to a GSM service distinct from basic data communication through the PSTN. The subscriber must have a GSM subscription for the service, but is not required to have a registration with the PSPDN. The PSPDN part is billed to the GSM operator, and the full communication is charged by the GSM operator to the subscriber. The GSM user need not indicate any PA D number: he just indicates the directory number o f his correspondent and that he wants a "dedicated PAD access" communication, and the PLMN takes charge o f all the details (single numbering). As in the previous case, only mobile user originating calls are possible. Usually a given PLMN will provide this type of access only for a national PSPDN, and hence the service will not be available when roaming. The access rate can be up to 9600 bit/s. Then we find the possibility to access a PSPDN in packet mode (X.321 via the PSTN or the ISDN, used as raw data carriers (case (c) in figur(1.5, and the variant (d)). As with the "basic PAD access", the GSM user must also be registered with the PSPDN, and must provide a PSTN or ISDN number corresponding to the PSPDN access unit he wants to communicate with (double numbering). There also the PLMN (and the PSTN o r ISDN) does not see a specific communication, just a data communication toward some "subscriber", which is charged as such. The PSPDN part of the communication is directly billed to the user, in his quality as PSPDN subscriber. The transmission mode is synchronous, and depends on the access unit. GSM provides synchronous transmission at 2400, 4800 and 9600 bit/s. A specific terminal has to be provided by the user to support the X.32 protocol (manufacturers may provide in the future integrated mobile stations including the support of X.32). Calls can be set up from the mobile station as well as toward the mobile station, depending on PSPDN possibilities. As in the first case, an advantage of this approach i s the possibility t o easily access PSPDNs i n fiireign countries (including the home country when roaming). The final case is the direct interworking between the PLMN and the PSPDN (case (e) in figure 1.5). This is close to the previous case, but here the PLMN fulfils the functions o f interworking with the PSPDN. The advantage is that the user does not need to he registered with the PSPDN. The PSPDN part o f the connection is charged to the PLMN, which in turn puts it on the bill of the subscriber. The access to a PSPDN, according to one or several of' the means described above, is a service which can be offered to the users as such,. but is also the support o f some end-to-end services, using specific terminals, such as teletex and message handling. These services may he offered by PLMN operators separately from the simple PSPDN access. k 56 T H E GSM SYSTEM S E Connection with Circuit-Switched Public Data Networks Users Circuit-Switched Public Data Networks (CSPDN) are general purpose data networks, such as PSPDNS. and by and large aiming at the same kind o f services. They differ by using a circuit transmission approach, and as such are more fitted to the cases of intensive end-to-end traffic. CSPDNs are likely to disappear, replaced by ISDN whose basic data services are of the circuit type. A CSPDN is typically accessed directly from GSM, with a usernetwork interface according t o X.21. Another possibility consists i n transiting the ISDN, in which case the user-network interface is according to the ISDN standards. GSM can provide access to a CSPDN in the same way as ISDN can. However, in the GSM case, the transmission modes are limited to synchronous transmission at 2.4, 4.8 or 9.6 kbit/s. 1.3.1.3. Short Message Services The different data services listed in the previous section are not really adapted to the mobile environment. They are simply extensions to GSM subscribers of the services available to fixed subscribers. One of the problems i s that these data services normally require rather bulky terminals, compared to the size of a handheld. They are then suited to semi-fixed usage, such as temporary installations, or for use with vehiclemounted terminals (e.g., fax in the car). But none of the data services is totally fit for an integrated implementation in the handheld case (they require a complete computer keyboard and a comfortable-sized screen). Still some data services are of interest for persons on the move and not desiring to be encumbered with a bulky terminal. An example of such services is the paging service, where simple messages. a few tens o f octets long, can be received on a very small terminal. Since its designers recognised this need,'GSM was designed to support such a service, in order to spare GSM subscribers the trouble of carrying two terminals. one for speech, the other for paging. GSM enables the transmission of point-to-point short messages. and distinguishes between the "Mobile Terminating Short Message Service, Point t o Point" (SMS-MT/PP), f o r the reception o f short messages, and the "Mobile Originating Short Message Service, Point to Point" (SMS-MO/PP), enabling for instance a GSM user to send such a message to another GSM user. Another short message service is the "Cell I T I N G s c h s w. 5 7 of a general nature to be broadcast at regular intervals to all subscribers in a given geographical area. Let us examine i n more details what i s provided. by taking the example of a GSM user called Kevin. Point-to-Point Short Messages These services enable alphanumerical messages of several tens of characters to he sent from or to Kevin. In the mobile terminating case. the message is typically displayed on Kevin's (ISM terminal upon reception. The service provided b y GSM i s akin t o paging, but with many enhancements, which make use o f the other capabilities o f GSM. in particular the possibility to have a hi-directional dialogue between the mobile station a n d t h e network (paging services a r e based o n unidirectional links). For instance. the network is informed whether the message has been received or not by the mobile station: and messages can therefore be kept in the network in cases o f clylivery failure, and repeated once Kevin is known to be reachable. Another improvement is that the originator of the message can he advised or the outcome, which is not the case in standard paging systems. On the mobile station side. the last received messages can be stored in a nun-volatile memory. The way the message is sent from the original sender to the PLMN is operator dependent. This service is not an extension o f some fixed network service. a n d a s a result does n o t correspond t o a n \ standardisation prior to GSM. Generating methods as widespread as keying o n a dual-tone mulii-frequency phone allow many PSTN subscribers to access the service: however. the message contents are in that case restricted t o numerical characters. Other access possibilities include simple videotex terminals such as the French Minitel. Of course. any computer accessible through a packet data network would do. but access to the service would not be so popular! Delivery through a human operator could also be envisaged, but is not very cost effective for the PLMN operator. In the same field. voice recognition machines may he a futuristic solution to obtain a widespread access through the PSTN. CiSM subscribers are luckier than paging subscribers. since die Mobile Originating Point-to-Point short message service enables them to send short alphanumericalmessages to other panties. For the time being. these other parties are not precisely defined. They include certain!) the other GSM users (which will receive them according to the SMS-MT/PP service), but these other parties could also he subscribers to a paging system, an electronic mailbox or alternatively one could Foresee gate( av devices transforming a short message into. e.g.. a facsimile or any kind of format suitable for the equipment of the recipient. Interworking between SR T H E Broadcast Messages I The service called "Cell Broadcast" in GSM is another specific feature o f the system. I t consists in broadcasting digital information messages cyclically towards mobile stations in a given geographical area. A mobile station designed for this service can monitor continuously the broadcast messages and present them to the user, except when engaged in a bi-directional communication with the network. A classic example for cell broadcast usage is road traffic information. In the current specification state, these messages are neither addressed nor ciphered: any mobile station equipped for the service may receive and decode them. As a consequence, this service will usually not require a specific subscription. The Specifications leave open the way in which these messages are provided to the network; it is up to the network operators to define their own rules for the generation of such messages, which could be reserved for instance f o r public authorities o r allow wider usage such as advertising. Bearer service Partner networks 3,1 kHz ex-PLMN ISDN Data circuit duplex asynchronous PSTN, ISDN Offered rates Data Services: a Summary The Specifications define sen ices in the same way as ISDN does. They distinguish "bearer services" \\ hich correspond to the transportation of u s e r d a t a between t w o —term in:U-1110(1cm- interfaces, f r o m "teleservices" which are complete end-to-end services, including terminal capabilities. The list of hearer services appearing in the Specifications in giVn in table 1.4. Though identified as a separate bearer service, the basic "PAD accps circuit asynchronous" service requires in fact nothing specific in the GSM networks compared to the "data circuit duplex asynchronous" service. The same sort of remark also applies to several teleservices. as can be derived from table 1.5. For instance, the specificity of a teleservice such as videotex, compared t o the relevant bearer service, does not concern the GSM domain. Though videotex is identified as a separate teleservice in the Specifications (at least for phase 11. the GSM networks need not implement anything specific to oiler Weir customers access to videotex, o n top o f the basic support o f the "data circuit duplex asynchronous. 1200/75 bit/s— bearer service. Other examples are X.400 message handling systems and teletex. Conversely, some teleservices are supported by GSM i n the same way, though not mentioned by the Specifications. Examples are group 4 facsimile o r access t o voice messaging centres. 300, 1200; 1200/75, 2400, 4800, 9600 hit/s Teleservice Telepholw PAD access circuit asynchronous (basic or dedicated access) PSPDN " Data circuit duplex synchronous PSTN. ISDN CSI'DN 12(X), 2400, 4800, 9600 bills PSTN. ISDN one of the above. corresponding to the data part of the service Alternate speech/unrestricted digital 5L) SliftING TI Ili SCI'S I'. GSM SYSTEM Corresponding bearer set.% ice Emergenc> calls Shun message Nen ice NIT/P1' Short message Len ice Mt)/pit Short message cell broadcast . ' AdL ;weed MI IS access t X.40ot I )ai a al rcti i1duplex synchronous Videotex access (profiles 1. 2. or 3l RILE CircUll t i n p i C ‘ a ‘ ) M i l . 1 2 0 11 / ] 5 I ' l l ; , Data circuit duple‘ synchronous Speech followed by data PSTN. ISDN Data packet duplex synchronous PSPDN 2400. 4800. 9600 bills Teletex 12 O M unrestricted digital — 12 kbit/s Alternate spccehaacsimile group 3 Automatic facsimile group 3 Table 1.4 — Bearer services in the Spec if kasha's Table 1.5 Te l e s e n ices otlered in (1S\1 Bearer services offered in GSM arc similar to those offered in ISDN - 60 T H E SET 1 1 1 1 ; SCIAN GSM SYSTEM 61 1.3.1.4. Supplementary Services The "supplementary services" m o d i f y a n d e n r i c h t h e b a s i c services, mainly by allowing the user to choose how calls toward or from himself are treated by the network, or by providing him with information enabling an intelligent usage o f the services. These functions are not specific to GSM nor to cellular radio: most o f them are directly inherited from the fixed network, with minor modifications when needed to adapt to mobility. The aim o f GSM is to offer eventually the widest palette o f supplementary features. However, in phase I only a few of them are fully specified, and t h i s l i m i t e d set does n o t g i v e a f a i r image o f what supplementary services are. Many more are incorporated in phase 2, and their specifications are sufficiently stable at the date o f writing so that we will include their description i n this book (but not the details o f their implementation). In this section we w i l l list and describe these features as they are perceived b y t h e user. T h o u g h t h i s approach m a y b e considered superficial, many difficulties lie in this domain (including for the user!). in particular when several features interfere. Two aspects of supplementary services must be distinguished. The first one is how they modify or complement the calls. The second one is how the users can sea and ask f o r the various alternatives. A s for this second point, there i s a k i n d o f "profile" f o r each subscriber, which determines the network behaviour as far as his calls are concerned. The user has access to this profile and can modify i t (at least i f he has the subscription entitlement for that). This is done through commands issued from the terminals. A n o n -sophisticated approach consists i n entering abstruse key sequences (this is what is usually available in the PSTN o r with m a n y PA B X s ) . T h e availability o f a display a n d o f a m o r e comprehensive keyboard than on standard phone sets w i l l enable mobile manufacturers to provide more friendly user interfaces. such as a menudriven command structure. The supplementary services are often grouped f o r presentation according t o t h e i r c o m m o n a l i t y o f implementation, a s s e e n b y manufacturers. F o r a change. w e w i l l present them according t o the moment they impact a call, as seen b y the user. We w i l l then look at a call, and see h o w t h e treatment can h e adjusted b y t h e users (the originator of the call as well as the other party). This will leave aside the features impacting established communications, which w i l l be dealt with afterwards. You want your calls to be directed to your secretariat when you are busy? Just subscribe to CFB, and why not also CFNRc ? However. the interaction with BIC-roarriand even with BOIC-exHC will make the CF quiescent abroad regardless of your CUG... Figure 1.6 — The puzzle o f .upplementary services ConlinCreial llepartmenrs k i l l need io make shuttle packages for subscribers 0111 or all the supplementary services coniliinations and abbreviations in C.itiNI. ii' they want 10 as old situations such as the one shown here! Still within the preliminaries, it should be noted that each feature owns an official esoteric name. as well as an official abbreviation. Both will be introduced. though. as suggested in figure, 1.6. these abbreviations will hopefully he harmed b y the commercial blurts trying t o sell the service. So let us proceed )) ith a call between GSM:subscribers, and see all the steps \\ here some supplementary feature ma) intervene. a) F i r s t Cheek • To start a call. the GSM subscriber I say ) must he entitled N. He can ask the network to block all o r some o f the call attempts made under h i s subscription. ' I b i s s e e m i n g ' ) c u r i o u s request i s b e t t e r understood when one thinks about lendine one's terminal t o somebody else (a teenage daughter for instance). It could he useful in such cases to restrict the usage t o the reception o f calls. o r 10 low-cost calls. T h e activation and deactivation o f such a blocking can b e protected b y a password. allowing the control t o be kepi b y Bjorn himself. There are 62 T I I N s( GSM SYSTEM The prohibition from setting up any call is the "Barring of All outgoing Calls" (BAOC). Another classic service in this area is the barring of international calls (which have certainly the highest potential for heavy bills), but in a roaming environment, this can he misleading: whether a call is international or not depends on the user location. So in GSM there a r e t w o possibilities. T h e " B a r r i n g o f Outgoing International Calls" (BOIC) allows one to forbid call attempts toward a country other than the one where the user is currently located: and the "Barring o f Outgoing International Calls except those directed toward the home PLMN country" (BOIC-exHC) will only allow calls toward a party in the country of subscription. Another obstacle against starting a call is the fact that another is already in progress. Usually, the only solution is to end this call before starting a new one. But in some cases the user may wish to keep the existing communication, just briefly interrupting it. This is possible and is called putting the call "on hold". This corresponds to the "Call hold" (HOLD) facility. T h e other party m a y receive a n announcement indicating that the call has been put on hold, in order that he does not hang up; but there is no such announcement when the communication resumes. b) Treatment of a Call Toward a GSM User, the Second Stumbling Block Let us now consider how the call proceeds, when it happens to be directed toward a GSM subscriber (let us call her Nina) in order to cover the called party aspects. As a GSM subscriber, Nina can ask the network beforehand to react to received calls according to three possibilities: 1. The network tries to establish the call toward her. This is the normal case when no special facility is acting: The network simply rejects the call: this is what happens i f "Barring of All Incoming Calls" (BAIC) has been asked for by Nina; or if she is known to he reachable through a PLMN outside her country of subscription. and has asked for "Barring of Incoming Calls when roaming outside the home PLMN country" (BIC-roam); 3. The network redirects the call to a third party, say Hans, chosen in advance by Nina (and the call is then back to case I. with a new direction). This is what happens i f "Call Forwarding Unconditional" (CFU) has been asked for by Nina. b . ; To simplify the picture, the c a n be different for different services re.g.. one number for telephony and another for fax). c) Third Obstacle Assuming that the call still exists MICE the earlier obstacles, the network w i l l try t o set up Wiirn's call intended f o r Nina t o some subscriber, which will normally be Nina. but may be I Inns i f C a l has been applied. Let us consider the case \\ hen the call is still routed towards Nina. The network has then to try to contact Nina's mobile station. 13u1 this may happen w t to he possible. The network can in some cases he immediately awani that Nina cannot be reached. In GSM, the mobile station can (and should i f so decided by the operator) indicate to the network that it will be switched off. or that it enters an area where service cannot he provided t o the subscriber (this is the concept o f "IMSI detach. which is explored in Chapter 7). In other cases. an effective attempt to contact the mobile station is done, and kills. When the network knows of this situation. the call is either: I. Terminated, with a "sorry- message or a tone toward Bjorn (basic case): 1. Forwarded to a third party, let us say Hans again. chosen in advance by NMa (the call is hack to square I ). This is what happens i f "Call Forwarding o n mobile subscriber N o t Reachable" I CFN Rct has Seen asked for by Nina. This feature has no counterpart in a fixed PSTN. d) Fourth Impediment The next possible problem arises when the destination is found to he already busy. e.g.. Nina is already engaged in a communication. There again. Nina can lime programmed the lick\ ork in advance for (int- of several possibilities: I. The call is terminated. \\ ith a lot el) voice saying how sorry the nem ork operating company i s not t o have been able t o complete the call. or, mule often. with a bus\ tone. This is the basic case: 2. The call is forwarded to a third party. chosen in advance by Nina. This happens \ \ hen " C a l l Forwarding o n mobile subscriber Busy" (CHI) has been asked fur: 3. Nina is warned of the call, and can then choose what to do. This happens if "Call Waiting" (CW1 has been asked for. Nina can then either "reject" the call, which is then treated according to 1— o r 2—, as chosen i n advance; o r she stops (totally o r momentarily) her ongoing call to accept the new one. The user can be helped in this choice, as explained in the following item. e) Fifth, and Penultimate, Obstruction Supposing the call has cleared all the previous hurdles, the final destination (say Hans) is alerted. Then there are still two possibilities: Hans answers, and the call is completed. or he does not, either because he is absent, or because he does not want to. In the good old telephony network, there is usually no reason not to answer selectively, for lack of information. Moreover, how to reject the call is the source of a dilemma: either Hans waits until the caller runs out of patience (but the ringing can he quite nerve-racking), or he lifts the receiver and hangs up at once (hut this is not deemed very polite). GSM provides the means for better manners. First, Hans can have asked for the "Calling Line Identification Presentation" (CLIP), in which case he (or the machine) recognises who is calling, and can then take his decision as a consequence. Now, this is only true if the calling party (Bjorn still) has not asked for "Calling Line Identification Restriction" (CLIR), in which case he may have required that his number is nor presented. In some cases, involving authorities. this restriction can be lifted. Second, Hans has means to actively reject the call (mechanisms exist in the GSM protocols to do so. and it is a matter for the terminal to enable this feature). Whether on time out ( i f the called party does not react), or by active rejection, a "rejected.' call can be forwarded to another party. chosen in advance by Hans, if he has asked for the "Call Forwarding on No Reply" ICFNRy). I t i s worth mentioning that the number o f forwardings in a row is limited to prevent infinite looping of a call. For instance, if Hans's country limits this number to I, and if Hans has asked for CFNRy, the original call from Bjarn will be cleared since i t has already been forwarded from Nina to Hans. I ) Last Problem Call forwarding is line. but may not be what Bjorn was really asking for. He may wish at this stage to abort the call in case of its being forwarded. This would he more polite (and les expensive) if done before the final destination (Hans) hangs up. But. to achieve this. the knowledge of the forwarding is necessary. Such information can he provided to Bjiirn by telling :him the directory number. he finally gets. i.e., Hans's number. This happens i f Bjorn has asked l i w "Connected L i n e Identification Presentation" (CoLP). And. o f course, i f Hans did not forbid this presentation by asking for "Connected Line Identification Restriction" (CoLR). This steeplechase through a call allowed us to present most ()I' the features. but some escaped. They will now be described. Charging Both the connected-to and the calling party can ask to be informed in real time o f the progress.; of the call cost. This is the "Advice o f Charge" (AoC) facility. This feature in fact covers, different issues. It may be 5 simple indication. not guaranteed to the last decimal to he \Out will appear on the bill if the call was stopped at this moment: or it can be used for a real time charging tpayphone application. the payment being by coins. credit or debit card,. in which case the actual charge levied on the user is exactly what is indicated. Accuracy i s a real problem i n an international multi-operator environment. The roaming possibility means that the PLNIN in charge of the call (and issuing the toll ticket) is not necessarily the one which hills the subscriber. The call charge and the billed charge can he different. for instance i n different currencies. The relevant information on tariffs. excluinge rates. m u s t be exchanged between operators to ensure that the advice o f charge reflects accurately what the subscriber will later discover on his bill. At the time o f writing. discussions are still ongoing on the provision ()I' the advice of charge facility w h e na user is roaming out o f the network which holds his subscription. It is likek that this facility will not be offered initially to roaming subscribers. 66 G S M SYSTEM st. 1 I f s t i u n s t [N.1 can offer a number of functions .oca..y. without the help of a network. Examples inclide the dialling o f abbreviated numbers, the storage o f received short messages, the edition o f short messages, the automatic repeat of failed calls. the automatic answering of calls, and so on. In some cases, like i n the latter example. the same function can he fulfilled locally. i f the terminal implements it. or by the infrastructure. i f a voice messaging facility is provided. Multiparty "Multiparty" (MPTy) i s a facility enabling a user to merge several communications, so that everybody hears what everybody else says. This applies only to speech communications. The way to proceed is to start with an established communication, to put it on hold, to establish a second call, and to ask for the conference. The process can be repeated so that a given user can merge up to 5 communications in which he takes part. Because the others can do the same thing, there is potentially as many correspondents as wanted together. Each call keeps its identity. and can be separated temporarily, o r terminated independently from the others. Closed User Groups The notion o f "Closed User Group" (CUG) does not cover a simple atomic facility as the others do. I t refers to a complex set o f facilities centred around the concept of a group o f users who want to restrict their usage of the network to communications inside the group. The typical application i s f o r companies, providing terminals and subscriptions to their employees for professional usage. It is then possible to limit for a given subscriber the outgoing calls, or the incoming calls, or both, to calls inside the group. A password mechanism allows to check the belonging to the group. Things are even more complex because a given subscriber can be part of several groups. A detailed description of all the possibilities would take too long, and the reader is referred to the Specifications or to the CCITT Recommendations. Phasing of the Supplementary Services Out of all the features presented above, only the call barring and the call forwarding features are provided for in phase I (namely BACK'. BOIC, BAIC, CFU, CFB, CFNRy, and CFNRc. plus as a network option BOIC-exHC and BIC-Roam). The other features cited here w i l l he available in phase 2. 1.3.1.5. Local Features A GSM mobile station is quite a complex piece of machinery. and includes the capacity of a small computer. As an intelligent terminal, it • The standard places few constraints on the local features. Mobile station manufacturers may or may not include them in their products. Sonic are specified by the standard simply because they are provided in fact by the SIM (see next section). and not by the part built by the mobile station manufacturer. In fact the only real imposed constraints pertain to the automatic repetition o f call attempt. f o r which a number o f restrictions are put. to diminish the risk of overloading the networks with useless attempts. Another point worth noting is the existence of the '+' key. which is specified as a harmonised shortcut replacing the international prefix. whatever the convention of the network the user happens to get service from. For instance. w hen in Sweden. a GSNI user can call somebody in Italy by dialling +39 followed by the national number. instead of dialling 00939_ Another important advantage in so doing is that the stored "+39...- number w i l l he recognised correct I) b v a l l GSM PLNINs (including in Italy). and therelbre remains valid irrespective of roaming. 1.31.1.6. The Subscriber Identity Module A mobile station in any cellular network must he personalised. i.c.. associated with a given subscription. This is needed since the identity of the subscriber is not in a one-to-one correspondence with the physical medium used for access as in a wireline network. The usual approach is to store in a permanent niemor o f the machine the required infatuation. such as a subscription identifier. This is what is done in most analogue cellular networks tan exception is iite German C.' network). The approach in GSM is different. A GSM mobile station is split in two parts. one of which contains the hardware and software specific to the radio interface. and another ,which contains the subscriber-specific data: the Subscriber Identity .Module. or SIM. The SRI can be either a smart card. having the wellknown size o f credit cards, or afternatively it can be ''cut" to a much I lib CiSM SYS11:51 1/4,I1-1 I ;\ (I .1111. SYLNI. 69 order for calls toward their personal number to he routed to the rented terminal, and for the call charges to he put on the same bill as f o t o calls made through their handheld. 1ism • The SIM is also the custodian of much information involved in the local provision of services to the user. The SIM can he protected by a password. a PIN code (Personal Identity Number), similar t o the (typically 4-digit) PINs of credit cards. Unlike many credit card PINs; the GSM PIN may he chosen by the subscriber. The SIM may also contain a list o f abbreviated d i a l l i n g numbers, w i t h t h e corresponding alphanumerical index (for the name o f the correspondent for instance) and the type of call (speech. fax....). The SIM can also he used to stun; short messages. in particular those received when the user is not present. A inure technical application is the storage of a list of preference for the choice of a network when several are possible. Since the user will have to choose vi hich network he will get service from. tbr instance when he crosses an international boundary. the SRI stores information to make this choice automatically. taking into account the user's preferences. When real-time advice of charge becomes available on networks. the SRI will also be able to memorise this charging information, to keep the subscriber informed of his expenses. • This view must be somewhat qualified, because the insertion and removal of the SIM is not necessarily easy with all mobile stations. Since its small size does not make it easy to manipulate, it is not foreseen that plug-in SIMs will he easy to remove. and in some cases mobile manufacturers have even secured them in the handheld station by a screwed lid. But the possibility still remains for the user to change it. An interesting development for the user is the potentiality to read and modify part of the personal information stored in the SIM. This can of course he done using the keyboard o f a mobile station. but a more comfortable approach could also h e offered. using a card reader connected t o a personal computer. and relevant software t o enter abbreviated dialling numbers. to archive short messages on the computer. etc. Of course, this only holds for part of the data stored in the SIM, since most of the information is protected against alterations and in some cases even against reading. The scow of the SIN-1can even be extended beyond GSM, and the concept o f iimulti-application card is emerging. The compatibility o f the SIM specification with internationally-recognised ISO standards i n this domain makes the GSM application a good candidate for inclusion into a multi-application card. The concept of the SIM is vet in its infancy, and will undoubtedly become a basis fora liner interworking between a user and a terminal. The possibility to remove the SIM presents many advantages for the user beside its role as a key. For instance, i f his mobile station fails and must be taken to repairs, another one can be used for the interim period. It suffices to remove the SIM from one equipment and to put it in the other. Another example is the case of urban users, which have only a handheld, for reasons of economy. When needed, they can borrow a more powerful station to be used in the countryside, or rent a car equipped with a vehicle-mounted station. In all cases, they can use their own SIM, in Another portentous aspect o f the SIM is related to roaming. We have already seen how roaming can be achieved by using the same mobile equipment to get service from two different networks with a single subscription. We will call this kind of roaming "MS-roaming" (MS standing f o r mobile station), since there i s another possibility. The interface between the S I M and the rest o f the mobile station i s standardised in the Specifications, and this standard could provide a basis for roaming between PI_MNs having different Air Interfaces, which will Figure 1.7 —The two types of SIMs The plug-in SIM has been designed to enable smaller handhelds to be built. and will not be removed as often as the card-sited SIM. smaller format, called "plug-in SIM" (see figure I.7). This smaller format was introduced to put less constraints on the design of handhelds. The SIM is a kind of key. Once removed from the terminal, the latter cannot be used except for emergency calls (if the network permits), that is to say it cannot be used for any service which will impact the subscriber's bill. 70 SEITINti THI. m t . m , t T H E G S M SYSTEN1 PLMN A 7 1 which the user simply equipped with his SINI will be able to access an) telecommunication system. A PLMN B PLMN A 11 / 4 PLMN B y / sOi n . g l e SIM-ME interface SIM (a) SIM-roaming • 1.3.1.7. Security Functions A radio accessed network is inherently less secure than a fixed network. This comes from the possibility to listen to and to emit radio waves from anywhere. without tampering with operator's equipment. To correct a little this state of affair. several types of security functions have been introduced i n GSM i n order t o protect the network 4 a i n s t fraudulent access and t o ensure subscriber privacy. These functions include: • authentication o f t h e subscriber. t o prevent access o f unregistered users: (b) MS-roaming Figure 1.8—PLMN interlaces Inter-PLMN interfaces must be fully standardised to allow any kind of roaming: in addition, MS-roaming requires a standardised air interface, whereasSIM-roaming requires a standardised SIM-ME interface. GSMoffers the most flexible form of roaming. with full MS—and SIM-roaming. (the greyed rectangles indicate the areas where standardisation is required). be referred to as SIM-roaming. SIM-roaming does not offer the fully automatic network selection as MS-roaming does (except with dual- or multiple-mode mobile stations), but it allows inter-operability at a much larger scale between systems based on different radio techniques. Instead of carrying his mobile station, a user would only take his subscriber card With him and use a different mobile equipment to adapt it to the networks he wants to access (see figure 1.8). Moreover, SIM-roaming does not present any technical obstacle to the extension to any kind of telecommunication network, wire-accessed or radio-accessed, since the network aspects of the roaming issue do not depend on the access scheme used in each network. The SIM appears then as the technical vector for personat numbering, that is to say a means to provide each user with a single telecommunication number whatever the network the user happens to be connected to. This is an important topic at the date of writing, with the studies concerning UPT (Universal Personal. Telecommunications). Undoubtedly, the Specifications in this domain prefigure a future world-wide telecommunication system. i n • radio path ciphering. in particular ciphering o f all subscriber information to prevent third-party tapping: • subscriber identity protection. to prevent subscriber location disclosure. These facilities are not subscribed to. and are not under control Of the user. The reason for this is that it would he much more costly and complicated to manage a per-subscriber or per-call protection as far as ciphering and subscriber identity protection are concerned. A s far as authentication i s concerned. t h e issue i s n o t even relevant since authentication is of benefit to everyone apart from the subscriber being authenticated. All security functions involve the SIM. which is in fact the real subject o f authentication (a correct approach i s t o say that what i s authenticated is the SIM. and not the subscriber or the mobile station). A first consequence is that none of these functions are provided when the SIM is not inserted. Another consequence is that the physical presence of the SIM is absolutely necessary to;get most services. the only exception being the e m e r encc call. I t i s not possible f o r a mobile_ Jaaiion manufacturer to provide a mobile equipment which reads the SIM once. and then is able to provide services under the guise o f the subscriber when the SIM is removed Moreover. SIMs are so designed that it is very difficult to duplicate thet (except by their issuer. a network operator). Thus the combination of the security functions and of' the SINI provides a high protection of the users and of the 4tworks against fraudulent access. t. THE CISS,1 SISTENI SEATING TIIE SCENE 1.3.2. GSM: A SYSTEM FOR THE OPERATOR The user's v i e w o f a telecommunications network may h e restricted to the provision of means of .transmission. This is however not quite the whole picture. In the same way as a bus company consists of more than just motors and passenger seats, a telecommunication network is more than just switching offices and transmission links. Taking this image a bit further, a bus company can be described as including driving controls (steering wheel, gear lever, ...). drivers, technicians, breakdown tests and alarms, spare parts, a ticket sale system. traffic observation systems to adapt the routes or the schedule to passenger needs, and so on. All these components have their equivalent i n a telecommunications system. They correspond to the operator companies, possibly also service provider companies, and a set of machines and software. Their collective purpose is referred to as "operation and maintenance" (O&M). Operation and maintenance appears often as a hotchpotch. It can he sorted into three main areas: • subscriber management includes what is needed to manage subscriptions. The most obvious tasks is to log subscribers in and out o f the system. A GSM subscription can be rather complex, with the multiplicity of services and of supplemental.) features. All the corresponding parameters have to be accessible to the operator (or the service provider). A related aspect o f operation (maybe the foremost for the operator!) i s call charging. Call tickets have t o be built. collected at one point for each subscriber and hills have to he sent out. • network operation consists i n "driving" the network. I t includes features enabling the operator to observe the behaviour of the network such as system load, blocking rates. number of handovers between two given cells, and so on. This knowledge permits the operator to check the overall quality o f service perceived by the users, and to find out where bottlenecks are. Operation includes also the means t o modify the network configuration, i n order t o palliate problems detected b y • observation functions for instance, to prepare for a foreseen increase o f traffic, o r t o extend coverage. N e t w o r k modifications can be "soft" changes, done by signalling, e.g.. the modification of handover parameters to change the relative boundary between two cells, but also "hard" modifications. necessitating in-the-field interventions, e.g., to add transmission ,-An,nit‘, n r . • n o m . ,•;•,, , , , , , , , . ,,,,, , , , , r : 7 3 system, monitoring and parameter adjustment are computer assisted tasks which are typically centralised at a single site. • maintenance aims at detecting, locating and correcting faults and breakdowns. It bears some relationship with operation. The difference can be presented by analogy. On a car dash-board, some displays are aimed to: assist the driving, such as the speed dial, or the head-lights indicator. The driver may react on these displays by accelerating or decelerating, or by switching off the head-lights. This i s operation. Now, other displays aim a t indicating failures, or? impending failures, such a s water temperature or oil-levec The driver reacts to these displays by immediate elementary inaintenance actions, such as stopping the car for some time pr adding oil, or alternatively by taking the car to a mechanic. This is maintenance. The machines in a modern telecommunication system are able to detect some o f their own failures, o r even t o forecast impending failures through s e l f -testing, a n d t o react b y themselves. In a number o f cases, redundancy has been built into the machines so that a failed part can be deactivated and its role taken by another part. This transition can be automatic. Other palliative actions can be commanded remotely by the operator, after possibly a remote localisation o f the failed part. Maintenance includes also local actions t o replace failed machines. The issue of operation and Lintenance obviously goes Far beyond the scope of GSM or of cellular networks. Most of the work done in this field within the telecommunication community takes a broader view, and aims at all telecommunication systems. The focus o f the field i s the concept of TMN, the Telecommunication Management Network, which features w a y s t o interconnect a l l infrastructure machines t o a management network and ultimately t o centralised sites w h e ff work stations enable 4uman beings to control the whole system. GSM i s designed to f i t with the T M N concepts, but its specifications do not include the whole management network. However, a full series o f the Specifications (the 12 series) is devoted t o the application o f these concepts t o specific issues such as the GSM architecture and i t s specificity as a cellular network. On the other hand, sonic, operation and maintenance features are intermingled with the service provision to the user, and are tackled b y the technical specifications o f the GSM interfaces. In the field o f subscription management, a specific problem is the 74 T i l E G S m SYSTEM purposes) between networks when users are roaming. GSM provides technical means for this together with the means to enable in the other direction the transfer of subscriber information so that visited networks can deal with roamers. In the field of operation and maintenance proper, some effort has to be put into specifying the control interfaces for the machines close to the radio. The operation o f switches and o f terrestrial links is o f general application, and does not require specific study. On the other hand, radio cellular coverage introduces a number of new aspects to manage. Each cell has its own set o f attributes, such as the frequencies i t uses, the channels i t allocates, the list o f neighbour cells and other parameters needed to make optimal decisions f o r handovers, and so on. These attributes must be managed, not only locally, but taking into account interactions on a large scale. For instance, the frequency planning, i.e., the choice of the frequencies for each cell, must be decided globally so as to minimise inter-cell interferences. The automation o f these functions calls for sophisticated centralised operation equipment. I f the methods to—• perform these tasks are out of the scope of the Specifications, means to carry control information to and from the radio related machines are dealt with, enabling operators and manufacturers t o design ultimately an efficiently controlled radio coverage. In the structure of this book, we have also linked up operation and maintenance with some side topics which are not usually so. Operation as described above concerns exclusively the infrastructure equipment. But the mobile stations are an integral part of the system, and are substantial contributors to the quality of service as seen by the user. The operators' problem is that when something goes wrong the subscriber is usually not able to distinguish whether the fault lies with the mobile station or with the network, and generally the network is assumed to be the culprit. While the operators have full control of the SIM, as part of subscription management, they have only an indirect control on the main part of the mobile stations. Indirect methods provided by GSM include observations concerning the mobile stations which can be performed through the infrastructure, such as activity or call tracing, so as to detect and localise mobile stations creating problems. They also include the possibility to introduce databases so as to centralise information relative to machines, as opposed to subscribers. But the most important indirect control means is preventive and consists in testing the mobile stations to be put on the market before allowing their use. This is called "type approval", and is still at the date of writing a controversial issue in the standardisation work in SMG, in particular because it is considered one of the braking forces on the take-up of GSM on a large scale. SI..INC S C I.NI• 7 5 The problems linked with type approval are o f two different natures. One is technical, and comes from the complexities ()I' GSM us /compared with previous analogue cellular systems and even more with ,,kpireline devices such as telephone sets or modems. Testing GSM mobile ations is a rather complex task. The test specifications are included in he Specifications, but many people in the specification committees (lid not realise until late in the process the importance of the task. The test specifications were drafted b y a separate committee and underwent modifications quite a long time after the Main specifications were frozen. • Similarly, the development of the test devices was no easy task. Other problems of a non-technical nature concern administrative issues, and are related to roaming. One of the aim of GSM designers. as well as operators and international and national regulation bodies. was that type approval can be clone only once. and not once for each operator or country. One of the reasons behind this W... is simply of .a practical nature, but another is to prevent the introduction o f bias within the industrial competition. The definition of a suitable type approval method then must follow a very narrow line between different objectives. For instance, one issue al the heart of the discussion concerns the degree ()I' testing. On one side, sonic regulation bodies \\ ish as free a competition between mobile station manufacturers as possible. testing only that mobile stations do not jeopardise the functioning of other machines. On Hid other side, some operators wish to get some additional guarantee that mobile stations will provide a correct quality of service, testing their full conformance to the standard as well as some perkmnance aspects. 4 T h e Specifinnions do not resolve these issues, and it is not their • g o a l . Still. they provide, first through the interface specifications which clearly define the properties of the mobile stations relevant to the quality of service, and second through explicit mobile station test specifications. the technical basis on which the type approval specification can he built. SPECIFICATIONS REFERENCE The 02 series of the Specifications is devoted to the presentation of the services which may be offered to the user by a GSM PLMN. In particular. telecommunication services are defined in: TS GSM 02.02 o r bearer services: TS GSM 02.03 for teleservices: and TS GSM 02.04 fur supplementary services in general. together with more details in TS GSM 02.82 for call forwarding services and in TS GSM 02.88 for call barring services. Other TSs in the GSM 02.tix series will complete the set o f sunnlementary services for nhasr 76 T H E GSM SYSTEM Security aspects are described in TS GSM 02.09. A set of mobile station local features is given in TS GSM 02.07, together with the status of each feature (mandatory or optional). This set is by no means exhaustive. Similarly, a man-machine interface for the mobile station is degcribed in TS GSM 02.30. There again, requirements are kept. to a minimum to leave manufacturers some freedom of design. leaving the market to decide what is best for the user. 78 T H E ARO IITECTVHF GSM SYSTEM 7 9 2 ARCHITECTURE ARCHITECTURE 2.0.1. The three Description Axes 2.0.2. Frontiers of the System: Where are the Borders of GSM? 2.0.3. Internal GSM Organisation 2.1. Sub-Systems 2.1,1. The Mobile Station (MS) 2.1.2. The Base Station Sub-System (BSS) 2.1.3. The Network and Switching Sub-System (NSS) 2.1.4. The Operation Sub-System (OSS) 80 84 87 89 89 94 100 105 2.2. Functional Planes 108 2.2.1. Layer Modelling 109 2.2.2. Transmission 113 2.2.3. Radio Resource Management (RR) 114 2.2.4. Mobility Management (MM) 114 2.2.5. Communication Management (CM) 115 2.2.6. Operation, Administration and Maintenance (OAM) 116 2.3. Interfaces and Protocols — An Qverview Specifications Refprence GSM as a modern telecommunications system is a complex object. To the multi-service aspect it shares with ISDN. it adds all the difficulties coming f r o m cellular networks. A s such. i t s specification. i t s implementation and its operation are no simple tasks. Neither is its description. In the course of the specification of GSM, much effort was expended to sort out this complexity. In this hook, we have built upon. this foundation, so as to follow a structured approach for the presentation of the system. The purpose of the chapter is first to present this structure, both of the system and o f the book. G S M w i l l b e analysed i n terms o f subsystems, and the main machines and functional domains w i l l he identified. This is the first step or the system description. o f which the subsequent chapters can he considered as refinements. Another aim o f this chapter is to present the structuring method. which is of interest in itself. The question of how to deal with complex systems has been encountered i n many different domains such as 121 e l e c t r o n i c s , b i o l o g y. e c o l o g y, economics, a n d o f c o u r s e telecommunications. where the Open System Interconnection t r i l l model f o r data networks represents one step towards a structured approach of complex meshed networks. While the authors do not claim mastery o f this field, they think that the study o f the structure o f a concrete system like GSM and the general study o f systems can he 118 80 - 1 . 1 1 E A last, but important, aspect of this chapter is to introduce many terms of the GSM jargon. These terms and abbreviations are used heavily in the Specifications, in the GSM literature in general, and in the rest of this book. Though we try to limit the use of the technical jargon, it is unavoidable; when a concept or an object has to be referred to every third line, it is best to give it a name. physical grouping (machine) From one point o f view, a telecommunications system i s a collection o f electronic boards transferring analogue electrical signals through wires or electromagnetic waves. But there is more to a system than t h i s "reductionist" approach. equating t h e system w i t h electromagnetic field values or transistor states. The opposite viewpoint would consist in looking at the system as a black box, seen only through its interfaces with the external world. Though not satisfactory either, this approach is helpful in raising two fundamental questions: spatial distribution Figure 2.1 -No-dimen,ional view of a network • w h a t is part of the system (here, GSM) and what is not? P h ' i c a l g r o u p i n g . ( m a c h i n e s o r entities) arc represented bs v e r t i c a l b l o c k . . • w h a t does GSM interface with? whereas compenning functions are grouped iu hori/onial layers. each one corresponding to a different functional domain. These two questions will be answered in turn. But it would be somewhat frustrating t o stop there, only looking a t GSM from the outside. A middle way has been thought in this book between this black box approach and sheer reductionism. On this route, the system will be divided into pieces; each of these pieces will be described as a black box in itself, and the overall operation of the system will be presented as the interactions between the pieces. This approach results in a description of GSM as a set of interconnected co-operating sub-systems. But, as already expressed, GSM is more than a concatenation of sub-systems: some areas involve many pieces of equipment and cannot be described satisfactorily by looking at each sub-system independently. Therefore, we must additionally look at how GSM operates from two different viewpoints: a static one and a dynamic one. The static viewpoint enables us to identify and describe several Junctions which are fulfilled through t h e co-operation o f several machines. The term machine is used here to refer to an assembly o f interconnected system components, physically close t o each other, working together to perform identifiable tasks. The term function is often used in the technical literature (and in the Specifications) to refer to some abstract machine. The use here is closer to the basic meaning of the word. A function is something to fulfil, an activity. In a sound architecture, a o n , r t i i r n , " `• distributed . functional l e — plane (field of • co-operation) A increasing level of • abstraction 2.0.1. THE THREE DESCRIPTION AXES machine enrrecnntuic I n cnme l i o n n i n n \I AlC11111-.1.-11 GSM SYSTEM chni.1,1 I also he identified, which group similar functions in a system. These functional groupings will put together elementary functions, possibly from different machines. which fulfil by co-operation the same goal. The usual representation. such as used in the layering method of the Open System Interconnection model. consists in showing functional groupings over several machines a s "horizontal- layers. a machine being a "vertical" structure in this representation, as shown in figure 2.1. The dynamic view consists in looking at the events affecting the system. Events happen a t m a n y scales, t w i n i n icroseek-wds• f o r transmission aspects. i v years when one views the deployment o f a network. The descriptfn o f these events. their organisation and o f the way they trigger other events in turn is very important in understanding how the system performs its functions. The GSM architecture, which is the subject of this chapter, will Last be described in terms of machines. then through a functional layer view. The subsequent chapters (i.e.. the bulk of the book) will be devoted static f u n c t i o n a l view S N static e q u i p m e n t view N dynamic view Figure 2.2 — The three axes of the description GSM functions can be described along several axes, each one from a different and complementary viewpoint. to the study of functional planes in detail, going deeper into the role of each machine within each plane, with a substantial part of these chapters containing a description o f event sequences. Having thus covered the three complementary axes (as shown in figure 2.2), the description should enable the reader to get a full—and consistent—picture of GSM. Now, although this three-axes concept helps i n structuring the description o f the system, i t i s not sufficient t o tackle the overall complexity. The number of possible physical groupings (that is to say machines or abstract portions o f machines), or o f planes, or of event sequences, is still rather big. The second trick we will use is a recursive description, or top-down approach. Taking the horizontal axis as a first example, this approach consists of describing the system in a few (less than 5 when possible) sub-systems, analysing the interactions at this level whilst taking each subsystem as a black box, and then applying this method t i n each subsystem. Similarly. functional planes w i l l b e composed of sub-planes. themselves composed of sub-sub-planes, and so on. Likewise for the temporal sequences, which can be analysed from the large scale down to the small scale. This allows us to identify first general, compound events, o f substantial duration teat.. system deployment, a site installation, a communication), resolved into liner events (e.g., the call set-up, its yelease). then stepping up the resolution (e.g., procedures. like channel allocation, start o f ciphering). and so on down to elementary message exchanges. and finally to transmission and bit modulation. This is the -reductionist- part of the methodology. which must be balanced at each step by the analysis of the relationships betv eel) the sub-parts. in order not to lose the large-scale view. This will be helped considerably by the three-axes approach. each axis embodying relationships between entities along the other axes. In this book. this recursive splitting i s used as a descriptive approach. In the case o f telecommunications systems like GSM, this approach is also used. more or less s>stcmatically:. at the design stage. Another application. which w e w i l l see i n the .last Chapter. i s the information model used f o r the design (Li the network management software. When applied to description. but even more so when applied to design. the method requires a lot of care for the location of each split. The basic rule to reach a successful splitting is to group things in such a wav that there is more interworking within one group than between different groups. W h e n correctly applied. t h i s method leads t o a sound architectural m o d e l . w h i c h c a s e s g r e a t l s specification a n d implementation. Some effort has indeed been invested i n the GSM standardisation bodies to obtain a clear-cut architectural model. which will be used in this book. Thc architecture of GSM described in the SpeCiliratiolls will be referred to as the "canonical- architecture. This architecture does not describe t h e actual machines w h i c h w i l l b e found i n a G S M implementation. f i r s t because s o m e freedom h a s b e e n l e f t t o manufacturers in terms of physical grouping choicesi second because (he Specifications cover only a small p a t of the specifications or ad -Julia' machine. The canonical architecture can he seen as the description of a network model, serving as a template for an effective implementation. However, the interfaces. that i s t o say what happens between t w o canonical machines. are described and specified without abstraction. The latitude left t o manufacturers consists most often in the possibility to group canonical entities in a single machine (in which case canonical interfaces disappear). or to split an entity into several distinct. possibly distant. machines (thus creating "proprietary-manufacturer specified- interfaces). Throughout this honk. we will describe the system on the basis of the canonical architecture: additional in, .oration will be given on the implementation choices left to manufacturers. Physical groupings and their borders are two sides o f the same coin. In fact, borders between machines are extremely important in a system such as GSM, since the Specifications specify in fact mainly the behaviour of the system as seen on interfaces between machines and not the internal working of these machines (though this can be derived to a large extent from the external behaviour). It is then an important function of the architectural model to define the system interfaces, and the last section of this chapter will be devoted to this aspect. In this book, an interface represents the frontier between two machines which are i n contact v i a a transmission medium. I t should h e noted that the Specifications use in some places a wider meaning for this term, as for instance when referring t o the MSC t o H L R interface (where the "interface" may well include a full signalling support network). In the following, both the machines and the interfaces involved at each stage will be indicated, with a particular emphasis on the interfaces for which the exchange of information is specified in the Specifications. Interactions between functional layers can also be described in terms of interfaces. Because these interactions happen inside machines. they are not to be specified. However, a detailed and formal description of such interactions can often be found in the Specifications (as the notion of primitives between layers). In some cases this description takes up a substantial portion of the Specifications, and is justified by the quest for as little ambiguous specification as possible. It should be understood that in the Specifications the description of the interactions between layers is a model, which does not constrain implementation in any way, though it can be a useful guideline. In this book we will not refer to this formalism any more. It is now time to apply these fine principles, and we will start by the first splitting steps, first along the vertical axis (machines), then across the horizontal axis (functional domains). This analysis is then refined in the rest o f the book, where the dynamic view is the main subject of the signalling chapters. 2.0.2. FRONTIERS OF THE SYSTEM: WHERE ARE THE BORDERS OF GSM? The first task is to identify the extent of GSM itself as a whole, thereby also identifying its external interfaces, i.e., its interfaces with the rest of the world. OPERATOR .c) co 1/4 co 2.3 -- External interfaces of GSNI When looked at as a Had: box. GSNI shares borders ‘\ ith three major realms: users of the system obtaining service through their mobile station, other telecommunication nem (irks 1111'0Ligh f i l c h calls Mahn. and the operating company controlling the GSNI domain. . When l o o k & at from a distance. GSM i s part o f the Global Telecnunications Network. itself a part of the human organisation. As such. el5M is in direct contact with users (human beings or machines which 'are being provided with telecommunications services through GSM). with other telecommunication networks (e.g., the global telephony network) and with the personnel of the operating companies. These arc . indeed the three main external interfaces of GSM. as shown in figure 2.1 t Other interfaces with the external \\ orld exist. such as the contact of • achines with air, ground and power supplies ( w i l t we may tern) environmental aspects) as well as with other systems using the radio spectrum (electromagnetic compatibility—EMC—aspects). T h e s e pragmatic aspects, which are far from negligible for manufacturers and operators. i f not for users, are not directly related to the provision o f telecommunication services and will not be dealt with here. Let us now 86 T H E 0551 S Y S T E M look at the three main border lines, with respectively users, other telecommunication networks and operators. On the user side, the limit lies somewhere between the user himself, who can be•excluded from GSM, and the radio interface which represents the principal part o f the Specifications. But is everything between the user and the radio interface part o f GSM? The mobile stations are only partly specified by the Specifications: an example is that terminal equipment functional entities similar to those defined in ISDN are not defined specifically f o r GSM; another example is the manmachine interface of mobile stations, which is in no way specified in a binding manner in the Specifications, and could include functions which have nothing to do with GSM. The point of contact between GSM and the user lies therefore somewhere inside the mobile station. We will come back t o t h i s point when addressing t h e internal mobile station architecture. Now for the interface with other telecommunications networks. GSM is specified mainly as an access network, enabling the setting up of calls between G S M subscribers a n d subscribers t o o t h e r telecommunication networks. For practical reasons, machines belonging to GSM are most often kept separate from machines belonging to other networks: this comes from the division, which is now the general rule. between GSM operators and the fixed PSTN/1SDN operator(s). even in companies of marked administrative origin such as FRANCE TELECONI or DEUTSCHE BUNDES TELEKOM. Other choices exist: one could imagine telecommunication switches performing G S M functions as well a s managing PSTN/1SDN subscribers. T h i s i s n o t excluded b y t h e Specifications, but the canonical GSM architecture does not consider this possibility, and the interface between GSM and other telecommunication networks is clearly defined. This interface includes three aspects. The first and major one concerns the point where communications transit between GSM and another network: the GSM machine ( a switching exchange) at the corresponding contact point i s referred t o as a "Mobile services Switching Centre" (MSC). The second aspect of interfacing with fixed networks concerns the provision o f basic telecommunications transport between GSM machines. In many insumces, regulations do not allow GSM operators to operate the terrestrial links between their machines. often leaving them no other choice but leasing lines from a fixed network operator. This second point will be ignored, since in practice it has no effect on the functioning of GSM. A third aspect of interconnection with fixed networks, which is also within the realm of using external networks as support for GSM functions, concerns the routing of non-call related 1-int.t•nnn e l i f f n r n n t ( " I C k 4 ARoirriirrt -RE 87 management of location data for GSM users roaming abroad, requires an exchange of data with no direct relation d ith given calk. to be transmitted through an international signalling network. There again. GSM operators of many countries are foibidden to access the international signalling network directly, and most therefore interface with a fixed network operator in their country to transmit this signalling information between GSM entities (MSCs, location register's) of different networks. Although the point o f contact (at least for the transit of calls) is well defined, the interfaces between GSM and external networks are not specified in the Specifications. CCITT recommendations. as well as EIS! standards based on them. include such specifications. but usually national variants exist. Fixed network operators have their own peculiarities, and the machines in contact with their networks must be customised to meet the exact interface requirements specific to each network. In this book, as in the Specifications. the basis I b r description w i l l he the ( t i n ' Recommendations related to ISDN or telephony. Now conies the border between GSNI and operator personnel. As in the case of the user interface. the border is basically between machines and the human employees themselves. The set of machines intervening between t h i s b o u n d a r y a n d t h e m a c h i n e s h a n d l i n g t h e telecommunications traffic (as well as some parts of these machines) are globally referred to as the Operation and Maintenance Sub-system (OSS). It includes various entities such as workstations (or terminals) handling the man-machine interface w i t h operator personnel, and dedicated computers managing a number o f tasks required f o r operation and maintenance o f the system, as well as parts o f the software o f traffic handling machines themselves. Most of the OSS aspects are not specified by the SpeClj. Only a small part o f the interfaces between traffic handling entities defined i n the GSNI architecture i s related t o OSS functions, and interfaces between these entities and ()SS machines are only partially specified. However, the whole 0 5 5 area must he considered part o f GSNI—and-this would indeed be the opinion of many an operator --since what would a complex system such as GSNI he without Means to drive and maintain it? 2.0.3. INTERNAL. GSM ORGANISATION This quick study ()I' the GSM borders already hints at a first split of 'the internal GSM domain into sub--systems. The mobile station (NIS) and the OSS have already been identified as manifest sub-systems. The .....,;13;1101 INIri 1.0,13,•;• I• 1 dr bi 11.•" i • 1 1 ( 1 ‘ 0 role. my to ..... • / I a 11111 - 3 I i . 1 . 0 1 1[1' The interface between the BSS and the mobile station is the already introduced radio interface, whereas the interface between the BSS and the NSS has been named the A interface in the S'pec•i(ic•atians. The MS, BSS and NSS form the operational part o f the system, whereas the OSS provides means for the operator to control them. This model is shown in figure 2.4. OPERATOR On this scale, the interactions between the subsystems can he grouped i n t w o main categories. Thri bottom part o f the figure corresponds to a chain: 4 ,rri ir-dirrr , external networks (=> NSS *I) BSS ictI> M S G users Co cr- ‘<, co whose business as a whole is to provide transmission paths and means to establish them. This is the operational part o f the system. handling the telecommunications traffic. Above it we find the control part. composed o f the OSS and the operator. which interacts with the traffic handling part by observing and modif\ ing it so as to maintain or improve its functioning. 2.1. S U B -SYSTEMS Figure 2.4— GSM subsystem organisation Following logically the three borders of the GSM domain, CiSM can he defined as composed of subsystems which interact between themselves and with the outside world along the white border lines shown. provide and to manage transmission paths between mobile stations and either fixed networks or other mobile stations, and to provide the means for the users to set-up communications along these transmission paths. The canonical GSM architecture distinguishes two parts: the BSS (Base Station Sub-system) and the NSS (Network and Switching Subsystems). The BSS is in charge of providing and managing transmission paths between the mobile stations and NSS machines (namely the MSCs). including in particular the management o f the radio interface between mobile stations and the rest o f GSM. The NSS must manage the communications and connect mobile stations to the relevant networks or to other mobile stations. The NSS is not in direct contact with the mobile stations and neither is the BSS in direct contact with external networks. The term NSS is used by many operators and manufacturers, though not in the Specifications. This section will deal with each subsystem i n turn. The main purpose is to introduce a number of terms. not to exhaust the subject. The level of detail here is just sufficient to get a general idea of the functional splits. More complete architectural considerations will he found in each of the other chapters of the book, where ❑ more thorough description of the required functions will he found. 2.1.1. T H E M01111 .E ST VI ION (MS) The mobile station usual I \ represents the only equipment the user ever sees from the whole system. Examples taken from the first types of GSM mobile stations to be on the market are shown in figures 2.5 and 2.6. Mobile station types include not only vehicle-mounted and portable equipment. bur also handheld stations. which will probably make up most of the market. Figure 2.6—A GSM handheld nubile station th) courtes o f Aleatel RadituLlkplione With their ever-decreasing weight and volume, handheld mobile stations represent a ver) attractive product or the user. Figure 2.5—A GSM portable mobile station (by courtesy of Orbiteli The First GSM mobile stations on the market were portable mobile stations. weighing around 2 kilograms. and also capable of heing installed in a %chicle. But what clues a mobile station involve'? Beside generic radio and processing functions to access the network through the radio interface. a mobile station must otter either an interface to the human user (such as a microphone. loudspeaker, display and keyboard f o r the management o f speech calls), or an interface to some other terminal equipment (such :IN an interface towards a personal computer or a facsimile machine). or both of them. A n e f f o r t h a s been m a d e t o a l l o w o f f -the-shelf terminal equipment t o b e connected t o mobile stations ( f o r instance group 3 facsimile machines designed I c r connection t o the telephone nem k l . and specific terminal adaptation functions ha e been specified r i b i s purpose. However, all implementation choices are possible and left open to manufacturers. enabling f u l l y integrated compact mobile stations t o coexist with mobile stations featuring standard interfaces. The , : t i o n a l split betwien nubile termination. terminal adaptor and terminal equipment is very much related to the transmission needs of each service. and will he detailed in Chapter 3. Another. more significant. architectural aspect o f the mobile station relates t o the concept ))I' subscriber module. or SIM (Subscriber Identity Module. a slightly restrictive name, as inure than identity is involved). As • GSM' described in Chapter I . the SIM is basically a smart card (or a cut-out thereof). following ISO standards. containing all the subscriber-ielated information stored on the user's side of the radio interface. Its fanctionalities, besides this information storage capability, relate also to the confidentiality area. The rest o f the mobile station contains all the generic transmission and signalling means to access the network. The interface between the SIM and the rest of the equipment is fully specified in the Specifications, and is simply referred to as the -SIM-ME interface (ME stands for -mobile equipment''). The functions o f the SIM w i l l he studied i n detail i n Chapter 7. Figure 2.7 — Mobile station functional architecture The mobile station may be a standalone equipment for certain services or support the connection of external terminals. either directly or through relevant adaptation functions. This leads to the identification of three main functions, as shown in figure 2.7: • t h e terminal equipment, carrying out functions specific to the service, without any G S M -specific functions: e.g.. a f a x machine; • t h e mobile termination. carrying o u t . among others. a l l functions related to transmission on the radio interface: • possibly a terminal adaptor, which ausa s a gateway between the terminal and the mobile termination. A terminal adaptor is introduced w h e n t h e external interface o f t h e mobile termination follows t h e I S D N standard f o r a terminal installation, and the terminal equipment has a terminal-tomodem interface. • In this book as well as in the Specifications, the term Mobile Station (MS) shall generally include a Nlobile Equipment and a SINI. although a rare case exists where a mobile station could be operated without a SIM (i.e., reduced to the mobile equipment) for the handling of anonymous emergency calls when soliermitted by the network. The concept of a removable storage device for subscriber data has far-reaching consequences. In previous cellular systems, except for the German C-network which introduced the smart card concept at the 011ie when it was making its way in GSM committees. the personalisation of the mobile station required a non-trivial intervention, only possible for technical specialists and not for the operator's administrative clerks. This situation lead to sleveral dra h m o b i l e station could only he sold by specialist dealers. able not only to install the equipment in a vehicle. but also to act as an intermediary bet \\ een the user and the men ice provider to personalise the equipment. Should the mobile station fail (unfortunately not such a taw event I. it was difficult to provide the user with a replacement during the repair period. and almost impossible to allow the user to keep the same directory number during this time. The removable SIM simplifies these issues. and also brings other benefits. A potential user may of course buy a 'nubile equipment. but he may also lease or borrow it for any period of time. and change it as he wishes without a lot o f administration. A l l he needs is his own SIM. obtained through an operator or a service provider. independently of any kilt I I I I lit 1 t K V equipment choice. The last steps of the SIM pet sonalisation can he done easily through a small computer and a simple adapter. Mobile equipment will be for sale on a much larger-scale than ever before, since their acquisition will not require the intervention of an operator or a service provider. Car phones will still require installation in the vehicle, but portables o r handhelds w i l l encourage users t o b u y their mobile equipment from any store. More advantages can be envisaged. For instance, rented cars could be equipped with a mobile equipment usable with any SIM, whether userowned or also rented. The reverse situation may also bring benefit to subscribers, i f not to operators: a subscriber may change his serving operator without replacing his ME. But most o f all, as explained in Chapter 1, this personal chip secured in its plastic case and called SIM is the first brick in the building o f a personal communication system enabling wide-ranging mobility between different telecommunications networks. 4 - > " control flow OSS user data flow ti\ BSS NSS Figure 2.8 —'Thee external environment of the BSS 2.1.2. THE BASE STATION SUB-SYSTEM (BSS) Largely speaking, the Base Station Subsystem groups the infrastructure machines which are specific to the radio cellular aspects o f GSM. The BSS is in direct contact with mobile stations through the radio interface. As such, i t includes t h e machines i n charge o f transmission and reception on the radio path, and the management thereof. On the other side, the BSS is in contact with the switches o f the NSS. The role o f the BSS can be summarised as to connect the mobile station and the NSS, and hence the mobile station's user with other telecommunications users. The BSS has to be controlled and is thus also in contact with the OSS. The external interfaces of the BSS are summarised in figure 2.8. According to the canonical GSM architecture, the BSS i n c l u d e s _ . . , _ two types of machines: the BTS (Base Transceiver Station), in contact with the mobile stations through the radio interface, and the BSC (Base Station Controller), the latter being in contact with the switches of the NSS. The functional split is basically between a transmission equipment, the BTS, and a managing equipment, the BSC. I he BSS bridges the space between the 'nubile stations on one side (through the radio incr.:ice), and the switching functions on the other. It is controlled by the operator through the OSS. A BTS comprises radio transmission and reception devices. up to and including the antennas, and also all the signal processing specific to the radio interface. • BTSs can be considered as complex radio modems, and have little other function. A typical first-generation BTS consists of a few racks (2m high and 80 cm wide) containing a l l electronic (leviers necessary for the transmission functions, as shown in figure 2.9 for a GSM900 131'S and figure 2.10 for a DCSI800 BTS. The antennas are usually a few tens of meters away, on a mast, and the racks are connected to it through a feeder cable. A one-rack first-generation BTS is typically able to handle three to five radio carriers, carrying between 2 0 a n d 4 0 simallliirOtis communications. Reducing the BTS volume is important to keep down the cost of the cell sites. and progress can be expected in this area. An important component of the 1355. which is considered in the canonical GSM architecture as a part o f the WIS. is the TRAM. or Transcoder/Rate Adapter Unit. The i s the equipment in \\?111C11 the GSM-specific speech encoding and decoding is carried out. as well as the rate adaptation in case of data. Although the Specifications consider the Figure 2.10 - A DCSI8(l0 BTS thy courtesy of Nokia) The small rack shown is designed to be used outside, typically below the antenna mast. Figure 2.9 — A GSM BTS (b)' courtesy of Motorola) A one-rack BTS, suet; as the one shown, is typically able to handle up to 5 carriers. The picture shows the rack equipped for 3 carriers. TRAU as a sub-part of the BTS, it can be sited away from the BTS, and even more so since in many cases it is actually between the BSC and the MSC. Its remote position allows the advantage o f more compressed transmission between BTS and TRAU, and its impact will he discussed in detail in Chapter 3. The internal structure o f the BSS i s represented i n figure 2.11. On top o f the BTS, it shows the second "canonical" component of the BSS. the BSC. The BSC is i n charge o f all the radio interfa,ce management through the remote command of the BTS and the \IS. (BSC mainly the allocation and release of radio channels and the handawer overmanagement. The BSC is connected. on one side, to several BTSs and on the other side, lo the NSS (more exiled) to an MSC). I A BSC is in fact a small switch with a substantial computational capability. Its main roles are the managtoment t h e o f channelson the radio interface, and of the handovers. A typical RS(' consists o f one Or t\\0 racks, as shown in figure 2.12 and can !image up to sonic tells of ION.. depending on their traffic capacity. BSS Vt •13 Nef k t A interface a Abis interface radio interface Figure 2.1 I — BSS components and interfaces The Base Station Sub-system consists of BTSs, situated on the antenna sites, and of BSCs, each one in control of several BTSs. The concept of the interface between BSC and MSC, called the A interface, was introduced fairly early in the GSM standard elaboration process. Only later was i t decided to also standardise the interface between BTS and BSC, and this interface therefore bears the (not any more meaningful than A!) name of "Abis" interface. In the GSM vocabulary, a BSS means the set of one BSC and all the BTSs under its control, not to be confused with the BSS as the subsystem including all the BSCs and BTSs. Figure 2.12— A (1SNI I3SC (by courte.> N l a t r a Coninuniu mum, The BSC n here conNist. of t\‘‘) cabinets: a control cabinet !inkling the duplicated central 1..olitrol and ,.‘‘ itching functions. and a cabinet handling the interlace.. 2.1.3. THE NETWORKAND SWITCHING SUB-SYSTEM (NSS) The Network and Switching Sub-system, o r NSS, includes the main switching functions of GSM, as well as the data bases needed f o r subscriber data and mobility management. It is also sometimes called the switching s u b -system, w h i c h i s i n d e e d m o r e NSS appropriate since a GSM network includes the BSS as well as the NSS. The main role of the NSS is to manage the communications b e t w e e n t h e G S M u s e r s a n d o t h e r telecommunications network users. Within the NSS, t h e basic switching function i s performed by the MSC (Mobile services Switching Centre), whose main function i s t o co-ordinate the setting-up of calls to and from GSM users. The MSC has interfaces with the BSS on one side (through which MSC VLR it is,in contact with GSM users), and with the external networks on the other. The interface with external networks for communication with users outside GSM may require a Figure 2.14 — A GSM MSC thy courtesy of Matra I i.icsson Telecommunications An MSC includes several rows of cabinets, one of which is shown here. control How OSS gateway for adaptation (Interworking Functions. o r IWF), the role o f which may he more or less substantial depending on the type of user data user data flow and the network i t interlaces w i t h . T h e NSS also needs t o interlace 1 .4: BSS external networks to make use of their capability to transport user data or signalling between GSM entities. In particular. the NSS makes use of a signalling support network. at least partly external to GSM. follo‘‘ ing the • r'• NSS CCITT Signalling System 11-'7 protocols (and therefore usually referred to • las t h e S S 7 network): t h i s signalling n e t w o r k enables co-operative ?interworking between N S S machines ‘ ‘ ithin o n e o r several G S N I networks. T h e external interlaces 0 1 t h e N S S a r e represented schematically in figure 2.13. Figure 2.13—The external environment of the NSS The NSS, through its MSC, is in contact with the BSS and external networks. Like the BSS. it is controlled by the operator through the 055. NSS entities are also in contact with NSS entities of other GSM networks for the exchange of data through SS7 signalling networks. As a piece o f equipment, an M M . ' controls a f e w l i S C s and i s usually a rather b i g switching machine. w i t h a medium population penetration percentage, a typical MSC at the .date of writing is suitable Mr covering a regional capital and its surroundings, totalling say I million inhabitants. Such an MSC includes about h a l l a dozen racks. Figure 2.14 shows a GSM MSC. 1O2 T I Ako uncrt RI: IE GSM SYSTEM The interconnection of the MSC with certain networks requires the adaptation of the GSM transmission peculiarities to those of the partner network. These adaptations are the Interworking Functions (IWF). This term refers by extension to the functional entity in charge of them. It basically consists of a transmission and protocol adaptation equipment. It enables interconnection with networks such as PSPDNs (Packet-Switched Public Data Networks) o r CSPDNs (Circuit-Switched Public Data Networks), but it also exists when the partner network is simply the PSTN or the ISDN. lnterworking functions may be implemented together with the MSC function, o r they may be performed b y a separate equipment. In the second case, the interface between MSC and IWF is left open by the Specifications. Besides MSCs, the NSS includes data bases. Subscriber information r e l e v a n t t o t h e p r o v i s i o n o f telecommunications services i s h e l d o n t h e . Icy infrastructure side i n t h e H L R (Home Location Register), independently o f the actual location o f the subscriber. The HLR also includes some information related to the current location o f the subscriber. As a physical machine, an HLR is typically a standalone computer, without switching capabilities, and able to handle hundreds o f thousands o f subscribers. A functional subdivision o f the H L R identifies t h e Authentication Centre, o r AuC, the role o f which is limited to the management of security data for the authentication of subscribers. HH The second database function identified i n GSM i s the V L R (Visitors Location Register), linked to one or more MSCs, and in charge of temporarily storing subscription data for those subscribers currently situated in the service area o f the corresponding MSC(s), as well as holding data on their location at a more precise level than the HLR. In current practice, as will be explained in more detail in Chapter 7, a VLR function is always integrated with each MSC. 1 0 3 But the NSS contains more than MSCs, VLRs and HLRs. In order to set up a call towards a GSM user. this call is first routed to a gateway switch, referred to as GMSC. without any knowledge of the whereabouts of the subscriber. The gateway switches are in charge of fetching the location information and of routing the call towards the MSC through \\Mei) the subscriber can obtain service at this instant (the Visited MSC). To do this. they must first find the right HLR, knowing only the directory number of the GSM subscriber, and interrogate it. The gateway switch has an interface with external networks for which it provides gaiewaying as well as with the SS7 signalling network to interwork with other NSS entities. The term GMSC is somewhat misleading, because the GMSC function is not by technical necessity linked to an MSC. I t could he thought o f as an independent equipment, or as a function integrated in a digital telephony switch. However, charging cons erations explained in detail in Chapter q8 are such that gateway functioA will not for sonic time be set outside GSM networks, and economic considerations make it undesirable to have standalone machines for this function. This state of things results in the widespread implementation of the GMSC function in the same machines as the MSC function itself. GMSC Having seen the pieces. l e t u s look a t the glue. Depending upon national regulations, a GSM operator may or may not be allowed to operate the f u l l SS7 network between NSS machines. I f the GSM operator the full control o f this signalling network, then SS7 has Signalling Transfer Points (STPs) will probably be part of the NSS functions, and could be implemented either as stand-alone nodes or in the same machines as the MSCs, in order to optimise the cost o f the signalling transport between NSS entities (MSC/VLRs. GMSCs. HLRs....). Similarly. depending upon the terms of its license, a GSM operator may have the right to implement its own network f o r routing calls between GMSC and MSC. or even r o u t i n g outgoing calls as near as possible to the destination point he e using the fixed network. In this case, Transit Exchanges (TE. not9lo h e confused w i t h "Terminfl • I04 THEGSNI SYNTHS] I05 2.1.4. THE OPERATION SuB-SvstrEm (OSS) control flow As we described it in Chapter 1. the OSS has various tasks t o fulfil. A l l these tasks require interactions between some or all the infrastructure machines such as in the BSS or in the NSS and members of the operator company o r o f the •seryice providing companies. OSS Network operation and equipment maintenance concern 1 a l l machines (including by the way the OSS machines), whereas subscription management has impact on at least the HLR. tt•-• AuC userdata flow SS7 GMSC MSC VLR A - Figure 2.15—Internal structure of the NSS Beside switches such as the MSCsand GMSCs, linked between themselves through a fixed network structure which may or may not be part of the GSM network, theNSS includes databases (such as the HLR/AuC) interconhected through anSS7 signalling support network. This, picture shows the casewhen the VLR database is integrated with the MSC. Equipment" as used for the mobile station architecture) may be part of the GSM network as well, and there again may be implemented as standalone nodes or together with some MSCs. As a summary, figure 2.15 shows the main components of a GS\1 NSS and the interconnections between them. The canonical architecture described in the Specifications is not as specific on operation and maintenance aspects of the GSM as it is on the rest. A wide latitude is left to operators and manufacturers. One sound reason for this is that the issue is not specific to GSM. Operation and maintenance functions are necessarily implemented in existing networks. and a lot o f standardisation work on the issue is going on for general application to telecommunicationsnetworks. Let us look at some of the ways to conceive the OSS. Up to some point in the past, operation and maintenance actions were performed locally, by intervention on the site of each machine. To this avail, each equipment was provided with some man machine interface, such as through a local terminal. The co-ordination o f the actions on different machines was managed by human beings. Such an approach could still be adopted for GSM. at least at the very beginning. for experimental networks. For instance, subscription management could be done by manually entering the subscription data on an on-site terminal connected to the HLR. With the local-only approach, the 055 functions are simply spread in the 1355 and NSS machines. and the only " O W machines are the man-machine devices such as terminals. Now, with the evolution of the technology and of the complexity of telecommunications systems. the range o f possible actions on the system as well as the quantity o f inlOrmation t o he processed have increased tremendously. The local-only approach is then inefficient as soon as the number o f machines to operate becomes significant. Sonic centralisation is required. and this calls for separate pieces of equipment intervening between several o f the K a l e handling machines and the man4nachine devices. Moreover. these `pieces of equipment can perform some o f the co-ordination functions instead o f human operators. thus offering a better guarantee of consismicy hemeen the configuration of 106 T H E GSM SYSTEM the different machines. The ultimate centralised approach is the concept of T M N (Telecommunication Management Network), where a l l operation and maintenance machines compose a network which as a whole is linked to all the traffic handling machines. Once some centralisation i s applied, interfaces between traffic handling machines a n d O S S machines appear, requiring some specifications. I t is at this level that most of the substance related to operation and maintenance can be found in the Specifications. This part of the standard has been designed with the ideas of TMN in mind, in order to enable a smooth integration o f GSM networks with advanced operation and maintenance machines. We will now look in very general terms at the OSS architecture, and then focus, as the Specifications do, on the portion of OSS close to the traffic handling machines, especially to the BSS. In Chapter 1 , w e have already noted t h a t operation and maintenance covers functions in several rather independent domains. The architecture o f the OSS reflects these domains. For the needs o f this chapter, operation and maintenance functions are best split i n three domains, corresponding to three distinct sets of OSS machines. Network operation and maintenance proper calls for mediation between the operator personnel and all the machines. I f TMN principles are followed, the operation network i s linked o n one side t o the telecommunications machines (MSCs, BSCs, HLRs, and others, but not the BTSs which are accessed through the parent BSC). On the other side, the operation network is linked to workstations acting as man-machine interfaces. The Specifications identify the OSS machines directly in contact with BSS or NSS machines. These dedicated machines are called the OMCs, or Operation and Maintenance Centres, and are linked to a few telecommunications machines of the same category (and usually of the same manufacturer), e.g., an "OMC-R" (or OMC-Radio) would be in charge of a few BSCs and, through them, the corresponding BTSs. In a simple management architecture, the OMC is autonomous, and includes the man-machine interface f o r the control o f the traffic handling machines i t i s linked to. Such an O M C i s typically a standalone workstation. The OMCs can also become simply the gateways to an overall management network, acting as mediation devices, following the terminology of TMN. Subscription management corresponds t o tasks w h i c h a r c independent from the other operation and maintenance functions, and which may then be supported by machines separate from those involved in n e t w o r k noel-mit-In S i i h c r r i n t i o n n i n n a r r e m e n t b a r s t w n f n e e t c Akentrirrukt. 1 0 7 subscriber data management and call charging. The Specifications do not address at all the first aspect, and only lightly the oecond. Different architectural approaches will he adopted by the different operators. Subscriber data management involves only the MLR (including the AuC for the security related data) and dedicated OSS machines. f i r instance a network linking the HLR and the man-machine devices in the commercial agencies where subscribers are dealt with. This network. if it exists. can be autonomous. in contact with the rest of the system only at the HLR. The SIM is also affected at this stage. It has to he initialised consistently with the information held in the infrastructure. The SIM has then two phases in its life. In a first "administrative- phase. it is dealt with by the subscription management system. but is not active in the traffic handling part of the system. Then. once initialised. it is operational and can be inserted into a mobile equipment. The second facet of subscription management is call charging. In a mobile environment. the call tickets relative to a given GSM subscriber can be issued by many different machines, including all the MSCs the user may visit. GMSCs are also. as will be seen in Chapter S. a source of call tickets. Then machines are needed to centralise the billing data for each subscriber. Once again. the .Specifications do not address these machines. However. the H L R i s an obvious candidate t o h e the centralised equipment for charging information. i f only to act as the gateway with the subscription management system to which it is linked. Some steps have been taken to enable the 557 network to be used to carry charging information between the MSCs and GMSCs and the HLR. This solution could be a convenient one for the exchanges between networks of different countries, but others can he envisaged. Finally, the last domain o f operation and maintenance i s the management o f mobile stations. A part o f it is done within network operation, through the infrastructure machines. There is however one machine identified in the canonical GSM architecture and specific t o mobile station management. the FIR. or Equipment Identity Register. It is a database which stores data related to Mobile Nquipment. As explained earlier. a mobile station consists of a SIM and a Mobile Equipment t Mk). Subscriber data management concerns the SIM and is handled by the HLR and VLR. Because SIMs can he moved between MEN. overseeing subscribers does not imply overseeing mobile stations. Though the management of the mobile/equipment is not absolutely necessary for the operation o f telecommunications services, it holds some interest. e.g.. when searching for stolen mobile equipment or when monitoring mobile station misbehaviour. For this purpose. the EIR is in charge of storing the relevant ME-related data. It interfaces with other NSS entities and with ARCIIITECILKIt S % - 1 3 I V I .1 1 a I C I V I subscription management and charging network operation and dire,...- 7 maintenance NSS it SS. mobile equipment management U v according to their closeness in purpose. The question is what functions for what purpose? Of course, the division of the system into three subsystems already bears a strong functional flavour. This reflects the fact that the choice leading to grouping some functions in the same machines depends on the closeness o f those functions. However, a number o f functions are by essence distributed, and can be fulfilled only by the cooperation of distant machines. This is obvious in any telecommunications system. since the basic function. the management of communications. is distributed. As a consequencl. most functional entities described so far perform tasks in several spread functional areas. For instance, machines such as the MSC o r the BSS have necessarily some operation and maintenance functions; another example is the involvement of the MSC in BSS-related functions such as handover. 2.2.1. LAYER MODELLING Figure 2.16 - OSS organisation Three main areas are pan of OSS: - network operation and maintenance of telecommunications machines; - subscription management, charging and billing: and - mobile equipment management. 1.8 I the network operation system. The interface with NSS machines is, again, through the SS7 signalling support network. This is the reason why the EIR is often considered as part of the NSS, though its functional role sets it within the OSS. Figure 2.16 summarises the general organisation of the OSS. The whole subject will be addressed in more detail in Chapter 9. 2.2. FUNCTIONAL PLANES Up to this point, our description o f the GSM architecture has focused on the physical grouping o f functions, i.e., the underlying question was where are the functions implemented? The present section will take another angle, looking at the functions to fulfil, grouping them In the telecommunications domain. a powerful method to obtain a functional grouping is to use the Open System Interconnection model. Functions are grouped in functional planes. represented as stacked one upon the other. The lowest plane. devoted to the physical transmission of information between distant entities, relies on physical hardware media. whereas the highest one represents the view of external users. Each plane (or layer) provides services to the next layer up. these services being themselves enhancements o f the services provided by the next layer below. Machines or entities are represented vertically, the intersection between entity E and layer L corresponding to the functions fulfilled by E to contribute to the objectives of L. Besides this hierarchical organisation. based o n the notion o f service provided by one layer to another, there is an underlying temporal organisation.. In general. lower layers correspond to functions having a short time scale, whereas the upper layers will group long time scale fuwtions. . . _ . . t W i t h i n each layer. the entities co-operate to provide the required se, ice, through information exchanges. The rules of these exchanges are specified at reference points where the information flow crosses an interface between different entities. These rules are called the signalling protocols. The distinction between an interface and a protocol is important. An interface represents the point of contact between Iwo adjacent entities. and as such it can bear information flows pertaining to several different pairs of entities. i.e.. several protocols. For instance. the radio interface in GSM is the transit point for messages pertaining to several protocols: I I (1 ARC') IITB -It RE THE GSM SYSTEM A t MM + CM II user Operator S S I OAM CM MSC VLR RR MM • RR I \ transmission radio interface Figure 2.17—Protocols versus Interfaces The example of the radio interface shows the distinction between an interface (here, the point of contact between MS and BTS) and protocols between peer entities, some of which may be far away. Each of the GSM interfaces typically transports several protocol flows. RR refers to the protocol used for the management of the transmission over the radio interface, MM to the management of users' mobility, CM to the management of the communications, and SS to the setting of Supplementary Services parameters. Figure 2.18 " l h e functional planes of GSM The CiSM function. are modelled here in 5 functional plane. iTransmission. Radio Resource management. Mobility Managenwnt. Communication Management. and Operation, Administration and \ laimenaneet. which offer the user and the operator a large palette of services. For the purpose of protocol specification, the slicing of functions in planes must lead to fairly thin "slices", in order for earth one to he consistent and also t o escape f r o m tot4 big a complexity f o r the corresponding protocols. Let us riow look at these planes. identifying on the way their intersections with machines and the relevant protocols. between MS and BTS (for transmission), but also between MS and BSC (for the management of the transmission over the radio interface), MS and M S C ( f o r t h e management o f users' mobility, a n d o f the communications), o r even M S and HLR (for setting Supplementary Services parameters), as shown in figure 2.17. Five planes will be distinguished in this book, as shown in figure 2.18. At the bottom lies the basis of any telecomipunkations system, i.e.. the transmission plane. I t provides transmission means f o r t h e • communication needs o f the users, as well as for information transfer between the co-operating machines. Transmission is a domain for very This is an example of the analysis of an interface as a protocol stack, each element of the stack being related to the intersection o f a functional plane and of the interface. Still, signalling messages pertaining to a given protocol may be visible on several interfaces along their path, if the corresponding peer entities are not adjacent. The protocol then appears on several interfaces. Thus, the specification o f a protocol is_ • somewhat distinct f r o m t h e specification o f a n interface. T h e specification o f an interface can be reduced to the description o f its protocol stack. T h i s conceptual distinction i s i l l resolved i n the Specifications, where often what is called an "interface" specification is short t i m e s c a l e e v e n t s , f s .from m i c r o s e c o n d s t e . g . . h i t m o d u l a i i m i r. t o in f a r t a " n r n r n e n l " c n e r i f i r a t i n n seconds (for message transmission). The next plane u p i s concerned w i t h t h e management o f transmission resources. In telecommunications networks, these functions are usually grouped with the communication management functions, because fixed circuit management represents a small portion thereof, However, in the case of a cellular system such as GSM, the management of transmission resources on the radio path is a complex issue and i t warrants its own functiona Splane. This is called the radio resource management layer, o r RR y e r . The RR layer provides stable links 0- HLR ix T ' . MSC VLR GMSC, 41. CM GMSC MM HLR RR trans. MS B T S B S C MSC/ VLR Figure 2.19 —General protocol architecture of GSM This schematic view, which will be refined in the course of this book, shows how the GSM machines co-operate in each of the traffic handling functional domains. between the mobile stations and the MSCs, coping in particular with the movements of the users during the calls (handovers). The BSS performs most of the RR functions. From a temporal point of view, this plane and the two next ones deal with events on the scale of the call, that is to say from seconds to minutes. Next conies a small functional plane, which has not been grouped with communication management because of its strong GSM specificity. This Mobility Management layer, or MM layer, is in charge of managing subscriber data bases, and in particular the subscriber location data. An additional task o f the M M layer is the management o f confidentiality aspects, such as authentication. The SIM, HLR and AtiC are examples of machines mostly involved in MM activities. The M M layer adds to the transmission functions provided by the lower layers the means to track mobile users when not engaged i n communication, and the security related functions. The next inane is much less specific to (.ISNI. It makes use of the stable basis r o v i d e d b y t h e R R a n d M M layers t o provide telecommunications services t o t h e users. W e w i l l c a l l i t t h e Communication Management layer. or CM layer. It consists of several independent components, depending on the type of service. The NSS, mainly the MSC, obviously has a strong involvement in the CM layer. The RR. MM and CM layers provide the users with a high grade of service. But, to make the picture complete, another plane should be added, which makes use o f the transmission facility. This i s the Operation, Administration and Maintenance plane (0AM plane). which provides the means for operator actions. This layer is not strictly above the other ones in terms of services, since it does not enhance directly the service they provide for the user. It is characterised by a larger time scale compared to the other layers, ranging typically from hours or days to years. Its kinship with the OSS is obvious. As will be explained later, many interfaces in other sub-systems carry some OAM functionality. Let us now describe further the respective relationships o f each plane with the GSM traffic handling machines. the main principles o f which are summarised in figure 2.19 (‘‘ ith the exception o f the OANI plane). 2.2.2. TRANSMISSION Some of the GSM machines are concerned with transmission only. An obvious example is the transcoder and rate adaptor unit (TRAU), which is only concerned in adapting speech o r data representations. Another example is provided by a transit exchange. whose role is limited to the routing of signalling exchanges between distant NSS entities. But most other GSM machines also play a more or less complex role in transmission. The mobile station obviously does so. and so does the BSC, the MSC and the interworking 'function (1W1'). which may all be along the transmission path between t w o users. Conversely, some o f the machines have no relation to transmission, except for the minimum needs concerning signalling with the other machines. These include the data bases (HLR, VLR. EIR) and the OSS in general. As already mentioned, the transmission plane includes two more or less independent functions. The first one is to provide the means to carry user information (whether speech or data) on all segments along the path followed by a communication. The second one is io provide the means to carry signalling messages between entities. The transport of signalling is needed between adjacent machines (i.e.. MS to !ITS. w r s to BSC. BSC .1-1. 1 1 1 1 . 11 , 11 , 1 a l I IN to MSC), b u t also through networks such as the SS7 network used between NSS machines. Included i n t h i s plane a r e aspects indeed traditionally called transmission, such as modulation, coding, multiplexing, but also other aspects such as l o w level protocols t o format data, t o ensure proper sequencing, to correct errors through repetitions and to route information throughout networks. Three chapters in this book are dedicated to the transmission plane (Chapters 3, 4 and 5). Chapters 3 and 4 deal with the "traditional" transmission aspects, a l s o referred t o a s " l a y e r I " o r "physical layer" i n t h e Specifications, i n accordance w i t h t h e Open System Interconnection model. Chapter 4 is entirely devoted to the radio interface, because o f its prime importance i n G S M . Chapter 5 deals specifically with the additional functions provided b y " l i n k layers" or "network layers" (according t o ISO terminology) i n order to transport user data and signalling messages between communicating entities. 2.2.3. RADIO RESOURCE MANAGEMENT (RR) The role of the radio resource management layer is to establish and release stable connections between mobile stations and an M S C for the duration o f a c a l l f o r instance, a n d t o maintain t h e m despite user movements. I t m u s t c o p e w i t h a l i m i t e d r a d i o resource ( a n d t h e corresponding terrestrial resources) and share it dynamically between all needs. The functions o f the RR layer are mainly performed b y the M S and t h e BSC. I n addition, since t h e responsibility f o r the handover process l i e s entirely w i t h i n t h e R R l a y e r, p a r t o f t h e functions implemented in the MSC are within the RR domain, in particular the ones related to inter-MSC handovers. The detailed study o f the RR layer functions and protocols is to be found in Chapter 6, which is devoted to this subject. - 2.2.4. MOBILITY MANAGEMENT (MM) The machines concerned with mobility management are mainly the mobile station (and more precisely the SIM inside the mobile station), the HLR and the MSC/VLR. The management o f the security functions are done by the same machines, and more particularly by the AuC inside the HLR. The BSS is not concerned with the M M plane. The detailed description o f the M M protocols, including mobility management issues a n d security management issues, i s g i v e n i n Chapter 7. 1.0. 1 1 1 1 I I I 1 , 1 2.2.5. COMMUNICATION MANAGEMENT (CM) 1 The functions o f the communication management laver. o r CNI layer, cons' 't in setting up calls between users at their request, as well as of course ! aintztining these calls and releasing them. I t includes the means for t o user to have some control over the management o f ihe calls he originates o r receives. through the -Supplementary Services-. T h e variety o f the Communication Management functions makes it easier to describe as three suh-domains. 2.2.5.1. Call Control The MSC/VLRs. GNISCs, \ V I ' s and H i l t s , through basic call management functions. are able to manage most o f the circuit oriented services provided to GSM users, including speech and circuit data. This functional core represents a sub-part o f the C M laver, and is called CC (Call Control) in the Specilintfirms. An important aspect o f communication management, beside establishing. maintaining and releasing calls, is the routing function. i.e.. the choice o f transmission segments l i n k i n g distant users a n d t h e i r concatenation through switching entities. GSM mostly relies on external networks to perform this task, interfacing these networks through MSC's and G M S C s . T h e I W F m a y a l s o h a v e a switchi ng f u n c t i o n f o r communications t o a n d f r o m t h e networks i t interfaces w i t h . C a l l management requires access t o the substription data. i n order to check the profile o f the subscriber. and therefore the HLR also intervenes in the CM layer. 2.2.5.2. Supplementary Services Management A second aspect o f the C M layer concerns the management o f the so-called !supplementary services. A s explained through examples i n Chapter 1.*. users i n G S M have sonic control on the way their calls are handled by the network. This potentiality is descrihed as -supplement:1r\ services". each one o f them corresponding to sonic specific variation o f ----the w a y t h e basic service i s rendered to t h e user. T h e i m p a c t o f supplementary services on calls is mainly a CC function. I lowever, the management o f t h e supplementary s e r v i c e s t a t u s i t s e l f , i . e . . t h e I EU THE GSM SYSTEM modification o r checking o f their actual configuration, can be done through GSM independently of the calls. It is the object of a separate subpart of the CM plane: the SS management part. The entities involved in SS management are very few: the mobile station and HLR are the only entities involved. 2.2.5.3. Short Message Services The last aspect of the CM layer is related to the point-to-point short message services (SMS-PP). For the purpose of these services, GSM is in contact with a Short Message Service Centre (SM-SC). A service centre may be connected to several GSM networks. In each o f these, one or several functional entities are in charge of interfacing the SM-SC. They are basically gateway functions, with the general definition we gave previously. However, the Specifications have special terms f o r the gateway functions when applied to Short Messages. They define two types o f such entities: the SMS-GMSC for Mobile Terminating Short Messages (SMS-MT/PP) and the SMS-IWMSC for Mobile Originating Short Messages (SMS-MO/PP). The role of the SMS-GMSC is identical to the role of the GMSC for incoming speech or data calls. The role of the SMS-IWMSC is much less obvious and adds little value to the service except providing a fixed GSM point of interconnection for an SM-SC, rather than enforcing its connection to the SS7 network which would enable information transfer with any MSC. All CM layer functions will be addressed in detail in Chapter 8, the last of three chapters dedicated to signalling issues. 2.2.6. OPERATION, ADMINISTRATION AND MAINTENANCE (OAM) The OAM plane includes the functions which enable the operator to monitor and control the system. I n one direction, i t mediates the observation flow from machines to the operator. In the other direction, it enables the operator t o modify the configuration o f machines and functions. As a functional plane, it hovers over all the others, whilst not using the services provided b y the other planes except the basic transmission functions f o r t h e exchanges between t h e concerned machines. The kinship between the OAM plane and the OSS is obvious. The OSS is an integral part of the OAM plane. but all the machines in the 117 BSS and the NSS also contribute to the Operation and Maintenance functions. There are a variety o f small tasks incumbent o n these machines: they are often those of the smallest time scale and scope. For instance the raw information which forms the basis for the observation of the system's behaviour i s clearly issued inside the traffic handling machines themselves. The data are then transferred to OSS machines. Another example is the first level of automatic maintenance, such as self testing or immediate actions such as activation of redundant parts in case of failure, and this is also done by the machines. Some other examples can be found, such as call tracing. A more interesting point in the canonical GSM architecture is the operation of the BTSs. As defined in the Specifications, the BTSs are not in direct contact with OSS equipment. but are operated through their parent BSC. The BSC appears then as an OSS device. in charge o f centralising the operation of several other machines. In terms of interfaces and protocols. the organisation of the OAM plane has many different aspects, and most of them are outside the scope of the Specifications. Completely excluded are the independent network. if any, dedicated to the support of subscription management. and most of the network operation and maintenance network. TNIN-related work may * t h e basis for the interfaces and protocols in these domains. In many cases. the interfaces and even the protocols identified for traffic handling are used to support OANI related functions. We have already seen that the B R fulfils its functions by communicating with NSS machines through the SS7 network. The corresponding protocol functions are in fact specified within the MAP. In another area, many OAM functions are intermingled with traffic related functions. Examples are call tracing or the management of terrestrial links. This explains why protocols such a s the BSC-MSC protocol includes O A M related picedu res. Finally, some protocols dedicated to OAN1 are included within the Specifications. The first one appears on the Abis interface. and supports the functions needed for the operation of BTSs through the BSC. The others appear on the interface between the BSC and the ()SS proper. All O A M related functions w i l l h e addressed i n details i n Chapter 9. 118 T H E (ism SYSTEM 2.3. INTERFACES AND PROTOCOLS - AN OVERVIEW The details of the architecture of the protocols inside the different functional layers w i l l be found i n each o f the specialised chapters. However, because we use in this book a model and a terminology slightly different f r o m the one used i n t h e Specifications, some general explanations may be useful for the readers already used to the GSM literature. Unfortunately, this is difficult to present without hinting at or using notions or terms not yet presented, and can be skipped in a first reading. Basically, this section lists the main GSM protocols, with our terminology. The Specifications do not always give an abbreviation or a name to the protocols. This is the main reason f o r our introduction o f new terminology, in an area already riddled with acronyms and other jargon terms. However, there is a need for a short term used consistently to refer to each protocol. I n particular this name will be used systematically before a message name in the signalling chapters, so as to enable an easy distinction between messages. The reader unaware of the problem will understand better the need for the mention of the protocol when he reads 119 LRCIIITE(' I BTS CM :Inirriq_L RIL3-004 MM RIL3-MM RR anchor MSCVLIR relay MSC BSC HLR f o n . a f . r. , C r - MAP/D RSM l i S S M A P i M A P / E TCAP "... i ..... : ..: SCCP MTP S m D r C p P do- .•••b LAPDm S Crvi.;PF, L A P D I I 1 1 Figure 2.20—GSM signalling architecture GSNI machines (show it as 1/2ellical hnca and functional layers (shown as hod/cilia! liQ.ers ()I' bricks) demarcate protocols. a stack of which can be defined on each of the interfaces. that CIPHER M O D E C O M M A N D a n d ENCRYPTION C O M M A N D a r e t w o messages fulfilling basically the same purpose, but belonging to two different protocols. Figure 2.20 shows an overview of the signalling architecture in the machines of a GSM transmission chain, as far as the telecommunications functions are concerned (i.e., without showing explicitly the O A M functions). The horizontal axis corresponds to spatial distribution, starting with the mobile station on the leftmost part of the diagram and going through the various infrastructure machines on the way. The vertical axis corresponds t o the functional planes, starting from the bottom with transmission and going up through the different layers described in the previous section. The functions of these protocols will he described in the relevant chapters. Firstly, let us just briefly consider the stack of protocols on the radio interface, which identifies many of the important GSM protocols. At the very bottom, all transmission functions use protocols between MS and BTS. Then the RIL3-RR protocol enables MS and BSC to co-operate for the management of radio resources (RIL3 stands for Radio Interface Layer 3, following the name of TS GSM 04.08 where the RIL3-RR is specified together with two others). This protocol also appears on the Abis interface. Upper layer protocols (RIL3-MM and RIL3-CC) define the rules for signalling exchanges between the MS and NSS entities. These protocols also appear when looking at the Abis and A interfaces. Taking this last case as an example, the BTS and BSC are -transparentto these signalling exchanges, i.e., they just act as carriers for information whose semantics is irrelevant for them. Inside the NSS. each machine has in fact a single interface with the SS7 signalling support network. The corresponding stacks of protocols share the same lower layers: the protocols used for signalling transport on an SS7 network and referred to as MT? (Nlessage Transfer Part).:\hone MTP, the protocols may vary depending on the corresponding peer entities concerned. Call-related signalling between MSC% and external networks makes use of TUP. ISUP or national variants of these protocols. Non call-related signalling corresponds t o many different protocols. which are grouped together in the MAP (Mobile Application Part). In this book we w i l l distinguish the different protocols b y names such as MAP/B. MAP/C, and so on from B to I. For instance MAP/C is the protocol between a GMSC and an HLR. All these MAP/x protocols use the services provided b y t h e S S 7 p r o m o ' I C A P (Transaction Capabilities Application Part), itself using the service offered by the SS7 \ !WIWI EC I t RN THE GSM SYSTEM _ • - 1 MAP. F 2 I > EIR MAP 1 HLR MAP.I VLR +._ MAP _D 4 MARE 1 A MAP C M A R C + G MAP/C M S C ir VLR 1 . 1 4 1 / / Figure 2.21 — Stack of protocols on the SS7 interfaces All SS7 interfaces make use of MTP, but the MAP/x protocols require SCCP and TCAP on top, whereas call-related signalling for standard call management use ' U P or ISUP related protocols. protocol SCCP (Signalling Connection Control Part). The different MAP/x protocols are not stacked up, but used independently in parallel. The corresponding diagram is given in figure 2.21. In the Specifications, the presentation is a bit different. The MAP/x protocols are specified as a single "protocol". named the MAP, and what we call here "protocols" are termed "interfaces". These "interfaces" are referenced with letters (B to G, in no order of prominence). To keep a strong relationship with the Specifications. w e have used the letter referring to an interface to denote the corresponding protocol. No H and interfaces appear i n the M A P specification. Still, w e have defined MAP/H as the protocol for transfer o f short messages although the corresponding MAP "interface" in the Specifications is interface " E " between MSCs, which is also used for the management of handover. In _the same way, MAP/I was introduced here as the MS to I ILR protocol for supplementary services management, which o n l y appears i n t h e Speed/cations on the "D" interface between MSC and I II_R, in order to be closer to the conceptual approach o f this book. A l l the MAP/x protocols referred to in this book are summarised in figure 2.22. As can be seen from this figure. MAP includes a great variety of functions. The corresponding procedures will he described in the relevant functional chapters of this book, each individual MAP/x protocol being considered within the scope of the relevant CM. MM or RR planes. 4+ MSC E(B) 11(- - - - SMS-gateway MAP H 497 Figure 2.22 NEAP/e' 10 mApii protiteols The MAP interfaces defined in the Speeiji( tni,n,a hear letter, %%Inch hate been used as a basis to describe separate MAI'lx protocols for the sake of complian, e with the conceptual guideline, of this hook. This summary o f interfaces and protocols does not go into amdetail as far as the transmission layer is concerned. In particular. both user data and signalling make use of specific transmission techniques and "link layer- protocols which have not been described here. This will now be the purpose o f the next chapters o f this book. dedicated t o transmission. SPECIFICATIONS REFERFMTE In fact there are very f e w SperWeillioux devoted t o general explanations of the general structure and architecture of (iSM. Piec'es can be found here and there, but the main source of information (beside this hook!) is various articles i n specialised cora, a c e proceedings and magazines (see the bibliography). The only specification which appears as clearly devoted to the subject is TS GSM 03.02 (Network architecture). Its scope is almost entirely limited to the infrastructure aspects, principally of the NSS. This TS is basically descriptive, with few explanations. One o f its useful aspects is the reference to other TSs. The internal structure o f the mobile station in terms o f terminal equipment, terminal adaptor and mobile termination is presented in TS GSM 04.02. There is no TS introducing the distinction between the RR, MM and CM planes. Part o f the subject, unfortunately written in a very technical jargon, can be found in TS GSM 04.07, will deals with the modelling o f the interactions between layers o f the corresponding protocols. I I%.\ \ l V I V I : 1 1 1 ) N 3 TRANSMISSION TRANSMISSION 3.1. Modelling Principles 126 3.2. A n End-to-End View of Transmission 3.2.1. Speech 3.2.2. Non Speech Services 128 128 132 3.3. Transmission inside GSM 3.3.1. Architecture 3.3.2. Speech 3.3.3. Data 149 149 154 166 Specifications Reference 184 The first goal o f a telecommunication network k to pros isle a means of transmission het cell end users. What exactly is to he provided. and how. varies. according to the different kinds of information to he transported. and to the specific constraints of the dilThrent interlaces to he crossed. Both points are relevant to GSM. The multi-service nature of GSM requires that it interconnects with various kinds o f external networks. each with their own transmission requirements. A s f a r as internal interfaces are concerned. the radio interface is usually the local point in a cellular network. GSM is no exception, and the specifications of its radio interface include more original features than any other public radio interface yet developed. Though transmission on fixed links is more constrained b y existing standards. sonic n e w features have been introduced on the terrestrial links between GSM infrastructure machines. Despite this variety, the whole system must provide consistent endto-end transmission paths. taking into account different optimisation schemes on the successive segments along the way. This calls f o r translation functions between some of the transmission segments. and is a source of complexity. The purpose o f this and the next two chapters is t o give an overview of transmission, with a focus on the specihe aspects or GSM. A top-down approach w i l l he followed. This chapter will start with a presentation o f how end-to-end transmission paths are structured. The way in which user data is transformed along the way will he addressed: this includes the GSM digital speech: coding, and the various rate adaptation schemes for dat❑ services. Then transmission between the GSM infrastructure machines will be detailed.., was felt that the radio interface deserved a chapter of its own, which follows this one. Both chapters will focus on the circuit transmission mode, which covers almost all of the services offered by GSM. This leaves aside the packet mode transmission, which i s used i n GSM f o r the exchange o f control information between machines, and f o r the Short Message services. which will be described in Chapter 5, the third and last chapter dealing with the transmission functions. 3.1. MODELLING PRINCIPLES transmission path A upper layers .111PP•11111 lower layers end node When looked at as a whole. the transmission functions o f GSM give an impression of complexity. Because GSM is aimed at providing multiple services, there are a number of different types o f information flow to be transmitted, and a variety of networks to interface with. Three broad types of information can be identified: speech: various information formats such as text, images, facsimile. computer file, messages, and so on. grouped under the vague term o f data: and finally the internal signalling messages. Depending on the possible transit network, and on the network of the correspondent, the same kind of information may be transmitted in different ways. Moreover, the transmission methods also vary within the system, from one interface to another. In fact (luckily), this seemingly complex jungle can be structured in two directions. A first "horizontal" structure stems from the need for consistent end-to-end connections. A number o f them are defined, according, typically, to the type o f transmitted information and to the final network or equipment. A second "vertical" structure corresponds to a layering design approach. With such an approach, a given portion of the transmission path uses machines which need not have a full knowledge of the type of data to be transported. The basic principle is that the meaning of the information transported on a given physical connection can he peeled little by little like the different layers o f an onion, each layer representing a different level of knowledge. If we represent the layering along a vertical axis, with the raw information at the bottom, and the distilled product at the top, and if the I transmission path follows a horizontal axis, we obtain a representation such as the one in figure 3.1. Intermediate nodes need not be concerned with the knowledge o f upper layer semantics. A s a consequence, the specifications of the transfer mechanisms on the individual interfaces can intermediate node end node Figure 3.1 1.ayered approach Higher layers, dealing with the contents and presentation of the information. are typically dealt with at hoth ends of the transmission path. whereas lower layers may use different protocols on different segment, between intermediate nodes. be much simplified, by taking into account only the information attributes relevant for transporting it. If we apply this model, a first split along the vertical axis consists in separating two domains, which will he described in the two main sections of this chapter: • t h e upper and outermost end-to-end domain is concerned with the information as dealt with by the final users. This domain is by nature rather "horizontal", i.e., each type o f information warrants its own individual description (speech. different kinds of data, ...). When describing the end-to-end transmission path. special emphasis ,Af,i I he put on the interworking between GSM and other networks. At this stage. GSM will he considered as a "bladk box" obeying certain rules: • t h e lower and innermost transmission domain is concerned with the way in which the information. regardless of its upper-layer semantics, i s transported inside GSM. I n the corresponding section, the GSM "black b o x " w i l l h e opened. gradually unfolding its internal organisation and interfaces. I K . N . N ; ) NI T u l l I N 3.2. AN END-TO-END VIEW OF TRANSMISSION In this section we will address first the transmission o f speech between a GSM user and another user i n the GSM o r i n some telecommunications network accessible through the PSTN or the ISDN. Then the other types of transmission will be adressed, that is to say the transmission o f non-speech data between GSM users and users i n networks such as the PSTN, GSM, ISDN, packet or circuit data networks. 3.2.1. SPEECH Telephony is by far the most popular service offered by public networks, including fixed networks and mobile cellular networks. After a general presentation o f how voice is transported from mouth t o ear through these networks, we will attempt to describe the environment with which GSM must interwork. 3.2.1.1. General Overview Let us take the example of a speech call between a GSM user called Bernard and a subscriber of a fixed telephony network, called Fred. The function of the transmission plan is to transfer speech signals from Bernard's mouth to Fred's ear, as well as from Fred's mouth to Bernard's ear. The transmission path goes through GSM equipment from Bernard to some interworking point, and from there to Fred through the PSTN. Starting with Bernard, the first item encountered is the microphone of his mobile station. Inside this mobile station, the analog voice signal is transformed into a digital information stream at a rate of 13 kbit/s, which represents the speech signal. Other processes in turn change this digital bit stream into a high frequency analog signal transmitted over the air. After being detected by a base station antenna, this radio signal is processed to recover the digital signal representing speech, which is transported over coaxial cables towards a speech transcoder. From this incoming 13 kbit/s input, the speech transcoder derives another digital representation of the speech signal, at a rate o f 64 kbit/s (the standard used i n fixed network transmission). I t is routed through the mobile speech 7 i s trans.:oat., Sala/ o r m Accoustic plane Analog plane n a l inomminasid Digital 13 kbit/s plane Digital 64 kbit/s p l a n e 3. 2 Speech ropre.emaiinil, hcluecn Iv.0 end iNer, lake, place on siN eral irawanksion each one correspondinr. 10 a kW lereni repreNentation or the speech .11pial. services switching centre (\1S(') and various links and switches in the PSTN until i t reaches the local switch to which Fred's telephone is connected. There (or possibly ?elsewhere inside the PSTN), the analog speech signal i s reconstructed from the digital 64 kbit/s f l o c . and transported on Fred's subscriber line until it reaches his telephone. where the loudspeaker enables Fred to hear an acoustic signal which should be recognised as Bernard's voice. This elemental" description o f a telephony call involves several transmission modes for speech, and therefore several transmission planes. Starting with the GSM subscriber's end. these transmission planes unfold a s follows (see W i r e 3.2): acoustic transmission, analog transmission, digital transmission at 13 kbit/s (this transmission being performed in two different ways over the radio path and between the base station a n d t h e speech transcoder). a n d f i n a l l y another digital transmission mode in which speech is represented by a 64 kbit/s Closest to the end user is the acoustic transmission plane: digital transmission at 13 kbit/s and 64 kbit/s as well as analog transmission found further down the transmission path. The transmission means vary from one interface to another, even within the same transmission plane. 3.2.1.2. Transmission Segments and 1 sworking Functions In this description, the transmission mode at 13 kbit/s is clearly specific to GSM, and is therefore part of the GSM domain. The precise border between the GSM domain and the external world is however rather subjective, and is done by defining two reference points within the transmission path. The first point is between the user's mouth and the microphone: the handset is considered to be an integral part of the GSM domain. The second reference point is between the MSC and the switch of the external network (the PSTN in the example) to which the MSC is directly connected. This split considers the GSM as a distribution network. The transmission path is then separated into: • t h e local "loop" relating to Bernard which goes from him to his MSC. I t is a part o f GSM and is therefore specified in the Specifications; speech is i s m i t t e d digitally. coded in PCNI rPuke Code Modulation). at 64 kbit/s. But i n the case o f older PSTN equipment. speech i s transmitted b y analog means. T h e sole cxample described i n the Specifications is however the 64 kbit/s case. other possibilities being considered only as country-dependent interim solutions. The 6 4 k b i t / s d i g i t a l c o d i n g i s p a r t o f e v e r y b a s i c telecommunication curriculum. The analog signal is sampled at a rate of 8-kbit/s: this operation limits the bandwidth to 4 kHz. Each sample (which has an analog value) is given an integer value after the application of a logarithmic compression law known as A law. Each value is coded as an 8 bit symbol. The outcome is then a How at a rate of K kbytes/s. i.e.. 64 kbit/s. The transcoding between the analog signal and its digital A law representation includes an analog process (the pre-emphasis). sampling. a linear analog-to-digital conversion o f the samples giving a result on 13 bits, and finally a'coding process which transforms the 13 bit samples into an 8 bit code. All these processes are specified in more detail in the CCITT Recommendations. i n particular in Recommendations CCITT 0.711/0.714. The segments may be combined in different ways depending on the type of call. For example, a call between two GSM users that are both being served b y the same MSC w i l l only involve two distribution segments: these segments will be connected inside the MSC switching matrix. A call between two remote GSM users will include two GSM distribution segments and a long-distance segment between the two MSCs. In this case the two MSCs are connected as if the other was part of an external network, typically in the same way as with ISDN. The effect of the speech transmission methods used in the PSTN and even in the ISDN cannot he neglected by the GSM transmission. since they do not provide a true reconstruction of the original acoustic signal. In both cases. the high frequencies of the signal are filtered out. Fortunately, this does not raise a problem. The lower part of the spectrum is also distorted, and this has some undesirable consequences. In the • analog case, the 0-300 Hz hand is COMpleld) filieNd out. Even ‘N ith digital transmission. t h i s b a n d usually disappears i n t h e f i x e d correspondent's handset. T h e m a i n consequence i s that t h e t w o directions. from a GSM user to a PSTN user and vice versa. are not identical. In the first case, the signal is processed first 1-i‘ GSM machines. before any distortions introduced by the other network. The result is that transmission quality is. barring other differences. better in the mobile to fixed direction. The functions to be fulfilled in order to adapt between the outer world and the inner GSM transmission are very simple on the mobile station side, since the reference point is at the boundary of the acoustic transmission segment. O n t h e infrastructure side, t h e network interworking functions (or IWF, though the term i s used rarely for speech) depend on the mode used for representing speech in the transit network. With the future ISDN, and with modern PSTN equipment. Another network interworking problem for speech communication is the problem of echo. The final segment in the PSTN uses a two-wire cable, and there is necessarily a 2-wire/4-wire adaptation somewhere. which is a first source of echo. The other sources are local to the terminal installation. As will he seen. the GSM transmission introduces a large delay, amounting roughly t o half that which i s encountered with a communication involving a satellite link. Since it is considered that echo • t h e rest includes a long-haul segment, which goes from Bernard's MSC to Fred's local switch, and Fred's local loop, which goes from him to his local switch. Both segments belong to the general telephony network, and may include national networks of several countries in the case of an international call. I P. % . \ . , , 1 1 : 1 . 1 1 1 . is particularly disturbing to the user when delay is more than 25 ins, all echo control function must then he provided at the boundary between GSM and the PSTN to avoid the negative impact of a delayed echo. 3.2.2. NON SPEECH SERVICES Non-speech services, or data services, cover the exchange of a lot of different types o f information. Data transmission encompasses the exchange of text, of drawings, of computer files, of animated images, of messages, and so on. An important part of the information processing is done at the two extremities, in a machine most often outside the scope of the Specifications. We call such a machine "terminal equipment" or more simply "terminal", though in some cases it can be a complex installation. such as a videotex server or a message handling system. The main functions performed by a data terminal in the realm of end-to-end information are the following: • source coding, which transforms back and forth text, images. sound, etc. i n the international currency o f the world o f information which are the binary digits; • e n d -to-end protocol. dealing w i t h the organisation o f the communication, juggling with such concepts as pages, sessions, languages; • and, most important. the presentation of the information to the user, by display, sound generation, printing, and so on. In most cases it is possible to confine such processes to the ends of the transmission chain. This enables the reduction o f the number o f different cases which need to he taken into account by eliminating the need t o study the intermediate transmission devices. T h e relevant characteristics distinguishing the different cases are few, and include the bit rate, the acceptable transmission delay (fixed or variable) and the maximum acceptable degradation due to transmission errors. The concept of bearer capability is used to describe and to refer to what is provided by the intervening equipment, i.e., the transportation o f information between t w o user-network interfaces. A bearer capability i s then characterised mainly by the attributes listed above. Conversely, the concept o f teleservice 'corresponds t o the full chain, including end-to-end processing o f specific information. The detailed functions in the data terminals are outside the scope of this book, just as they are outside the scope o f the Specifications. This is natural. GSM being basically a distribution network. In most cases, only One of the two terminals at the ends of the transmission path is within reach of the Specifications. The terminal functions, and the corresponding end-toend protocols between the terminals. arc by necessity those that we used in the networks (ism interworks with. For instance, the way images are transported through CSM between two fax machines is the one used between PSTN users. because one o f the two machines may he in the PSTN. A terminal equipment can he for instance a facsimile device. a personal computer. a computer terminal or .t ideotex terminal. We will concentrate on the bearer capabilities used hem een the terminals. Looking closely at the structure of the transmission path. we immediately detect an important point: the boundary between GSM and the external network. USN! can be connected to a variety ()I' external networks. since we are not yet in the promised land of broadband ISDN where a single international long-distance 'lemur': supports all possible telecommunication services. Examples of neby orks include the good old PSTN (this ubiquitous telephone network \ \ Inch is still the principal carrier of data transmissions), Packet Switched Public Daly Nemorks (PSPDNs), such as TRANsuAt in France. or Circuit Switched Public Data Networks (CSPDNs). The existence of an external network divides the transmission path into two segments. The segment between the GSM user terminal and the boundary point is entirely \\ ithin GSNI. Rut the other segment. from the boundary point to the other terminal, is entirely outside the control o f GSM. and follows transmission rules that are specific to the external network. To reduce the number o f cases dealt w i t h b y transmission equipment within GSNI. despite the variety or interworking cases. tyy o generic functions are inserted on each side o f the GSM segment. as shown in figure 3.3. These functions enable GSM to deal with a small amount o f internal transmission nodes. and still accommodate the various imerworking needs. The adaptation function at the boundary between G S M a n d t h e external network i s called t h e /wove (1( interworking function, most often reduced to the last two terms, and often further reduced to "IWF". On the GSM user side, the functional part of the mobile station which performs the adaptation hem een a specific terminal equipment (TE) and the generic radio transmission part is called the Terminal Adaptation Function. or TAF. 11lEGSN1S1'STENI 134 End-to-end communication • other network G S M . TAF I N e r , IWF generic GSM \ r / transmission MSC VLR Figure 3.3 -- Data transmission planes Data transmission can he studied on three levels. the end-to-end transmission plane between terminal equipment. the TAF-IWF plane inside GSM, and the generic GSM transmission plane. TioNssusstox As already noted, most o f the end-to-end domain is out o f the scope o f the Specifications. But the adaptation functions (in TAF and 1WF) are of direct concern to GSM. For most of the data services, the tasks fulfilled by the adaptation functions can be inferred from the single knowledge of the bearer capability. There is however an exception of importance, which is the facsimile teleservice. In this case, the adaptation includes additional functions dedicated to facsimile, and will be further discussed later. So, with the sole exception of facsimile, the adaptation functions and the general configuration of the transmission paths depend mainly on 3 5 the bearer capability and on the external network. It appears that the key factor, at the origin of most of the differences. is the external network. As a consequence, we chose to present the different cases not service per service, but external network per external network, starting with the PSTN. 3.2.2.1. T h e PSTN Case Data (i.e., binary digit flows) can he transmitted over the speechoriented Public Switched Telephone Network (PSTN) via the use o f audio modems, which transform the hit stream in an analog signal which is constrained to the 300-34(X) Hz bandwidth carried by the PSTN. Most of the data services, including fax, videotex. are transported by the PSTN. A data connection between two PSTN users has the configuration shown in figure 3.4. case a. When considering a connection between a GSM user and a user it is clear the we must keep the half of the configuration concerning the PSTN user as i f he was in connection with another PSTN user. An audio modem must then appear somewhere between the GSM user and the PSTN-GSM interworking point. as shown in figure 3.4. case h. The user The TAF on one side and the IWF on the other side act as entry points into the GSM world. Their functions depend on the type of end-toend service. Conversely, GSM entities in-between TAF and IWF are not concerned with the end-to-end service, but solely with the bearer capabilities required to transport the corresponding data flow. Figure 3.3 shows how data transmission can be looked at on three different levels. the end-to-end level, t h e TA F - I W F level a n d t h e generic GSM transmission level. 1 user PSTN audio modem 3.1 kHz audio line audio modem I a PSTN user to PSTN user PSTN user GSM transmission audio modem 3.1 kHz audio line user audio modem e• M GSM user to PSTN user Figure 3.-1— Interconnection s, ith the PSTN GSN1 appears as an extended terminal-to-modem interface I III 1 3 1 I problem is then to design the GSM part of the transmission path. with the constraints imposed by the radio interface. For different reasons. chiefly because of the difficulty in designing an efficient and robust transmission scheme over radio for audio modem signals. the choice was to put the required modem on the infrastructure side, in fact in the interworking function. This means that, between the TA F and the RIVE the GSM network is only required to transport digitally provided data. Modem type Kale \ lode of transmission V.21 300 Hi'. as) nchronous V.22 1200 bills as) nchronous. synchronow, V.22bis 2400 Kith s)uchronous V.23 12011/75 bilis as) nehronotis Putting modems on the infrastructure side necessarily restricts the freedom of the user to use any kind of audio moderns (as in the PSTN). since they are constrained to use the types offered by the operator. However, in general, only standardised moderns are used in the PSTIN, and among all possible modulations, only a handful (mercifully!) have survived the standardisation process. Moreover, the available rates are in a limited series (defined by CCITT Rec. X.1), in which the typically used values are the bi-directional symmetric 300 bit/s. 600 bit/s, 1200 bit/s. 2400 bit/s, 4800 bit/s, and 9600 hit/s. and the asymmetric 1200/75 bit/s used in particular for the videotex service. V.26ter 2400 s)tieltrontius Between the two modems, the stream formatting falls into two categories, the asyhchronous format. and the synchronous format. The difference of names comes from the fact that in the asynchronous format the instant of transmission of the bits are not aligned on a regular time base, whereas they are in the synchronous case. In fact, the difference is more profound. as the asynchronous transmission is a small protocol by itself, grouping bits into characters, and providing flow control for example. The unit of transmission is a character, a grouping of 7 to 9 hits. A character is preceded and followed by special signals, the start and the stop "bits". The data rate represents the rate of transmission of bits within a character, but successive characters may be separated by a period of any length, as shown i n figure 3.5. I n the synchronous case, bits are start signal s t (m-1) bit o p signal x Figure 3.5 - An asynchronous data Clim The "start' and "stop- signals help the receiver to decode the flow of characters. where bit beginnings are not synchronised with a clock edge. The length x of the stop signal is not a multiple of the bit duration. hut must he greater than a minimum value. Typical values of n are 7. 8 or 9. V . 3 2 45(X). 9oln hit's s) nehronous Table 3.1 - Audio modem Dees supported IN) Ow Spo The most wide]) used audio modem types are supported I)) GSM. including lxith as) nehronous and 5) nelitonous modems at rates up to 9600 hit's. transmitted regularly and continuously. I n t h i s case, t h e u n i t o f transmission is the bit, and it is up to end-to-end conventions to group bits in one way or another. These aspects are important because the start/slop Format. used by the asynchronous services me still. despite their ancestral character, the most widely used form of data transmission. being used for instance the modern videotex. on the serial port o f personal computers. for the connection o f most computer terminals. or ill the interface with smart cards (including the SINI-NIE interface in (iSNIi. A finite and limited list o f modem bees, rates and transmission modes which can be supported for the diterworking between GSM and the PSTN has then b e e nestablished. It t o u r s all the rates previously listed. w i t h i n each case. t h e modes allowed h y t h e ' I 1 1 * Recommendations. The list is given in table 3.1. Not only do the C(T1T Recommendations specilw the modulation method used between Iwo modems over an audio line. hut also the interlace between the terminal and the modem tthe DIFI/DCF, interlace). All the eases belong to a wide family \\, Rich, in various haws. use the signals defined by ('CITY Rec. V.24. (hi the mobile station side. the interface between the TN and the transmission part follows the standard corresponding to the modem with which it Mien\ orks. A s a result. the user to network interlace is not at the same point ;is it is for a PSTN user. The interface is at the boundary of the Data Terminal F.quipment and the rest of the mobile station. and the Terminal Equipment is connected as it to a modem. T h e CiSNI transmission segment ;Ippears in this case as Heim? inserted b e t w e e n t h e a n d t h e m o d e m . T h e p r o b l e m i s M e n t o l i e as unobtrusive as possible. A few problems are encountered at Ibis and some steps are necessary for the adaptatioi..0 the GSM transmission. These functions are fulfilled in the TAF and the IWF. The classical interface between a modem and a computer includes several wires, on which several information flows are transmitted in parallel. Two of these wires carry the user data bit streams (one in each direction), and, in the case o f a synchronous transmission, two which carry the bit clocks. But the data exchanged between a modem and a terminal includes not only the data streams, but also additional signals, whose general purpose is to allow the terminal to control the modem. For example, the modem may work only in half-duplex, and a signal is then needed to set it to receive or transmit mode. Because in GSM the modem is remote from the terminal, these signals must be transported over the GSM transmission system. These signals account f o r the difference between the user bit rate (e.g., 9600 bit/s) and the carrying capability of GSM (e.g., 12000 bit/s). Because GSM is basically a serial means of transmission, a function for multiplexing and demultiplexing the basic rate and the added signals must be fulfilled in the TAF and in the IWF. Facsimile 1 Facsimile is considered as a Lev service for the marketing of (ISNI. but it is the source of many technical difficulties. Its specification ‘k as delayed, and even removed from phase I for some time before being reintroduced i n 1991. Some o f the difficulties come from the long transmission delay incurred by the transmission over the radio path. which appeared not to be compatible with the protocols used by group 3 facsimile machines. Another problem is the absence o f a standardised terminal interface other than the plain analog m u -wire connection for PSTN connection. The consequence o f this is the need for a specific adaptation function, both in the mobile station and in the IWF. The GS NI model distinguishes in fact a "generic.' Terminal Adaptation function (the one used for synchronous hearer capabilities) and an additional "tax adaptor" which is inserted between the facsimile machine and the generic TAF. The interface between the fax adaptor and the TAF is a terminal to• in, lb,dem digital interface, which supports if !Wed he automatic dialling and an vexing, and the various synchronous data rates used by fax group 3. The interface between the terminal proper and the lax adaptor is a 2-wire analog telephony interface. enabling the connection of existing facsimile A second problem comes from the asynchronous transmission mode. The transmission system of GSM is basically synchronous. This raises the problem of the adaptation between an asynchronous characteroriented flow and a synchronous one. Some function must be put to this avail in the TAF (on the mobile side) and the IWF (on the network side). Conversely a problem exists i n t h e case o f a synchronous transmission, because the clock between the modem and the terminal may, in some cases, be independent from the GSM clocking system. This- leads to a possible slight difference in frequency, a drift, which can add up to a non negligible value after some time. This issue is referred to as "Network Independent Clocking". The specification o f these functions was not something new, because the same problem o f interconnection with the PSTN exists in ISDN, and was already tackled within the scope of ISDN standardisation when GSM started to work on the issue. The solution adopted in GSM is derived from the ISDN specifications, with many parts used without modifications, both for simplicity and for ensuring an easy transition between GSM and ISDN. Moreover, most of the problems arise when interworking with other networks. This is why the details of how they are solved w i l l be explained later, so as t o allow a discussion o f the characteristics of some of these other networks and, in particular, ISDN. PSTN user u s e r 3.1 kHz audio a u d i o line a u d i o modem m o d e m ! a) PSI N user to PS 'f N L.S(" t GSfT.I user Fax adaptor audiol__ a u d i o modem m o d e m 1-% o l P S T N GSM transmission 3 . 1 kHz Ty audio a u d i o line modem a b) GSM user to PSTN use: F i g u r e 3 M - R e f e r e n c e t a l l I f i ! 2 0 1 8 111111 f i l l f a y Beside the "generic- terminal adaptation function of Ow GSM transmission chain. a lias adaptor is i o accommodate ‘tanildid PSTN t e r m i n a l s . 140 1 1 1 1 i ussur.. GSM SYSTEM machines. As a consequence. the fax adaptor includes the complex audio modern (V.21. V,27ter, and V.29) required to communicate with the terminal, a n d t h e overall transmission c h a i n h a s t h e complex configuration shown in figure 3.6. This is the functional point of view. Concrete implementations of this model will certainly include integrated products, where all functions will be grouped inside a single piece o f equipment. The market will decide on the opportunity o f such developments, which are certainly attractive from the technical point of view. For instance, integrating the facsimile terminal and the fax adaptor would remove two audio modems at once. But the fax adaptor is not reduced to a modem. I n addition, it manages a number o f higher level functions. I n a fixed facsimile communication, t h e e n d -to-end protocol specified i n C C I T T Recommendation T.30 enables the facsimile terminals to co-operate for the management of functions such as the choice of the modulation speed, the mutual identification, the page delimitation, the management of halfduplex, etc. This protocol is tampered with by the fax adaptor and by the IWF, in particular to avoid the problems raised by the long transmission delay. The specific functions required f o r the management o f the facsimile teleservice are confined t o the fax adaptor (on the mobile station side) and the IWF. They can be considered as an additional layer on top of the basic synchronous transmission capabilities of GSM, with the latter not being impacted by facsimile transmission. The specific management of facsimile would require quite some development if it was to be described in detail. It is certainly an interesting item, but it is quite specialised, and we chose not to deal with it any further within this book. The interested reader is referred to the TS GSM 03.45 and 03.46 for details, or the papers referenced in the bibliography. 3.2.2.2. The ISDN Case The interconnection with the Integrated Services Digital Network (ISDN) is a must for a modem digital telecommunication system. Speech raises no problem, but this is not the case for data services. The basic data service supported by the ISDN uses the capacity o f a bi-directional 64 kbit/s channel. By deliberate choice. GSM is not able to provide this service. Because it would use at least four times as much spectrum as a speech channel (and eight times in the future with the half rate speech coding scheme), a heavy price would have been charged to the user of this service. In these conditions, it is difficult to imagine it would have user PSTN I S D 1 4 1 user N 64 WI's • el line modemI 3 .audio 1 kHz '2.1:.!.'1% l • ' - audio I . RA I *cVirc.u.lilt0l modem ' . I a) PSTN user to !SON user rt user GSEIT I S D user N GSM 6 4 kbitts transmission c i r c u i t . T - - RA} + V . 1 1 0 RA b) GSM user to ISDN user I f Figure 3.7 — Interconnection N11111 ISDN The rate adaptation scheme ( R A I used h) I S D N to inter m i . ; with I'S-EN (carve eu has been implemented in GSNI. leading to the diagram o f case h. been much used. Moreover, it was clear From the start that the inclusion of a 64 kbit/s capability would have seriously impacted the design of the transmission system and increased the total system cost. The interconnection between GSM and ISDN fur data services is nevertheless possible. but in a devious way. To understand how, let us first note that compatibility with the good old ps'IN is a constraint. not only for GSM, but also for ISDN. The configuration o f a connection between an ISDN user and a PSTN user is shown in figure 3.7. case a. The end-to-end stream is rate -a.apted -adapted("RA- box in the figure) to I v transported o n 6 4 khit/s circuits. T h i s i s t h e subject t l ; ( t 'FIT Recommendation V. I 1 0 ( o r 1.4631. which specifies h o w an ISDN subscriber equipment must send data at 64 khit/s when in communication with a PSTN subscriber, which transmits and receives data through ji modem at, for instance. 4800 bit/s. The ISDN is able to carry the saute data services as those which are possible hetv een GSM and PSTN (and more). These services nay then he offered for interconnection between GSM and ISDN. The result is that both the terminals on the GSNI side and on the ISDN side act as if the other was in the analogue PSTN. This exemplifies the weight of history. Even though the services are similar i n both network.. the interconnection between Rte GSM and the ISDN. both using digital 142 c s x i 11:AN5MISSli SYSTENI transmission, i s radically different from the interconnection between either of them and the PSTN. Between GSM and ISDN there is no need for an audio modem. At the boundary point, the information is carried on a 64 kbit/s bit stream, with a portion of this rate corresponding to the hits exchanged between the end terminals, according to Recommendation V.110. The configuration of the connection, as presented in figure 3.7. case b, shows that each side is almost identical to the case where the other end is i n the PSTN (the difference being the audio modems). B y comparison with the PSTN to ISDN case (figure 3.7. case a). it can be seen that the GSM transmission replaces the modems and one of the rate adaptation functions. 3.2.2.3. The PSPDN Case GSM can offer the possibility to communicate between a GSM subscriber and a Packet Switched Public Data Network (PSPDN) subscriber, or more generally between a GSM subscriber and a subscriber that can be reached through a PSPDN. This possibility can he provided by different means, depending on the terminal on the mobile station side. and on the level of intervention from the GSM infrastructure. To explain the different cases, it is simpler to start from the ways in which fixed subscribers can access a PSPDN. There are three different means widely used to access the PSPDN, plus one of potential future utility when the user is an ISDN subscriber. The first is the direct access, where the subscriber is connected physically to the PSPDN. An example of an access interface, widely used, is the one specified by CCITT Recommendation X.25. The configuration (shown in figure 3.8. case a) usually includes a modem in the subscriber premises, but this modem is not an audio modern and it is part of the PSPDN. The terminal exchanges data with the network according to a high-level packet protocol (X.25 levels 2 and 3). The subscriber i s basically identified by the access line. The second method is a variation of the first. where the access is via the PSTN (see figure 3.8, case b). Audio modems must be used. The connection is established first through the PSTN, as a normal PSTN communication. When the PSTN subscriber initiates the call, he enters a PSTN number addressing the PSPDN entry port. Then a second number. referring to the end correspondent, will be provided from the terminal to the PSPDN (this i s the notion o f double numbering). T h e major user 1 PSPDN L- 4 3 user X.25 a! moo physical connectioil to UN: PSPDN user PSTN audio modem 6. 3.1 kHz audio line X.32 PSPDN user Iaudio • P H modem • b) packet access to the PSPDN throoph the PSTN Figure 3,5 — PSI'DN packet ;icces‘ A filed u‘er may access the PSPDN in a packet 'node either through a physical N.25 link to a packet handler I PH) of the PSPDN (case or through the PSTN via an X.32 procedure tea.e difference with the previous case is that it is not possible to identify the subscriber by the access line. This is why the access protocol is slightly modified i n this case t o convey the subscriber identity. X.25 thus modified is the protocol specified by CCITT X.32. However the terminal and the PSPDN still Use a packet protocol for communication. A third method makes use of a PAD function (a Packet Assembler Disassembler). It enables the user to have a simple terminal, which does not support a packet protocol. The major disadvantages of this solution are that the transmission data rate is limited, and that calls can he set up only at the initiative of the subscriber (incoming calls are not supported). The transmission uses an asynchronous protocol. with a characteroriented simple access protocol on top. \\ Inch is used Ibr numbering fur example (an example o f PAD access protocol is the one specified by CCITT Recommendations X.28). Access to a PAD can he direct. but the 144 ' i t I RA \s\ils!.,IIIN G S N I SYSTLNI user audio modem 3.1 kHz audio line GSM' user PSPDN PSTN P S T N GSM transmission y audio modem audici - PAD modem I -h user 3.1 kHz audio line audio modem 't • Li. X.25 X.28 GSM access to PSTN PSPDN Figure 3.9 - PAD access to a PSPDN audio modem A simple asynchronous terminal may still access a PSPDN through a PAD, which converts the asynchronous stream of user data into packets and Images the PSPDNpacket protocol. basic PAD access to PSPDN via the PSTN principle application i s f o r the access through t h e PSTN. T h e configuration is then as presented in figure 3.9. The last case i s t h e access through t h e I S D N . C C I T T Recommendation X.3 I specifies how an ISDN subscriber can access the PSPDN. It refers to two methods, one being in fact the X.32 access, using the ISDN as i f it was a PSTN (the configuration is then the same as in figure 3.8, case b). With the other method (shown in figure 3. IC), the user-network interface between ISDN and the subscriber includes X.25. The packet protocol is then used between the terminal and an ISDN machine, which in turn interworks with the PSPDN. This ISDN machine is in charge o f the identification problem. The subscriber only has to user a I S D N PSPDN u s e r 64 kbit/s circuit PH X.32 Figure 3. III - ISDN X.25 access ISDN offers the pos‘ibilit) for a subscriber to access an X.25 service directly. through an ISDN packet handler (PH). Figure 3.11 'Basic- packet (N.32) and I'.\ I) access in t 1S\I Basic interconnection modes \‘ ith the PST's. can be used fur X.32 or PAD access. \‘ idiom am spec di; t h e (IS n e m ork compared to the pure PSI:\ inter,..onn,•,:tion case. enter a single number. and does not has e to be registered in the PSPDN but only with the ISDN. The different schemes for access to a PSPDN from GiSNI are adaptations of the three last schemes (direct access is not suited for users that are by essence mobile users). The total number of combinations is in fact more than three. first because of the possibility for CiSN1 to intervene. principally t o suppress the need f o r double numbering. and second because of the possible interpositiorpol ly)N. An obvious possibility is to access a PSPDN through the PSTN. with GSM machines unaware that the call is ultimately for the PSPDN. Both X.32 and PAD access methods are possible. as shown in figure 3.1 I for the basic PAD access. The PSPDN access protocol is emplmeil between the terminal and sonit!. PSPDN machines. GSM being (ink a carrier. as is the PSTN. This requires double numbering. and i t also requires that the GSM suhscriher be a subscriber o f the PSPDN f o r charging reasons. This approach is of no concern to GSM.1/4shich does not prmide any additional sets ice than it does for t r a n s m k s i o n through the PSTN. Yet the method is cited in the Specilic‘itiems for the PAD cane. and is referred to " b a s i c PAD access''. 1 \ \ NSIISsIO \ user GSITEI P S P D N user modem m o d e m PA D 0.1 . a) dedicated PAD access to PSPDN from GSM user GSET:1 • X.32 modemF PSPDN user modem F pH b) dedicated direct access to PSPDN from GSM Figure 3.12—Dedicated GSM access to the PSPDN GSM can act as a direct bridge between the subscriber and a PSPDN, either via a PAD (case a) or via a packet handler (case b). Interworking through modems is shown. but direct digital interworking is also possible. Another possibility is access through the ISDN, as an ISDN user. along the lines shown in figure 3.10. This possibility is referred to in phase I (notion of "transfer mode packet"), but will be suppressed for phase 2, and it is likely it will never be used. The remaining possibilities are the more interesting, and mostly involve the GSM machines. They are the "dedicated PAD access", and the "dedicated packet access" (the second option does not appear explicitly in the phase I Specifications, where the packet access covers both what is described in the following paragraphs, and the access through the ISDN). In both cases, a single number is requested from the user, and no specific subscription with the PSPDN is required. The GSM interworks directly with the PSPDN, and the transmission does not necessarily make use o f audio modems, depending o n agreements between t h e G S M a n d t h e PSPDN operator. T h e transmission configuration is as shown in figure 3.12. The user-network interface is then between the terminal and the mobile station. The PSPDN access 1 . 1 . protocol (X._ or X.321 is still handled between the user terminal and a machine in the PSPDN. However, the GSNI machines (more precisely the IWF) are perfectly aware that it is a PSPDN access, and they do interfere with the protocol. mainly to add the required identification (which is specific to the PLMN and not t o the subscriber). The transmission requirements are then identical to the basic cases, with only the network interworking functions specific t o the PSPDN access. A s far as the PSPDN is concerned, its subscriber is the n t i ‘ l network operator. which is then in charge of dispatching (and reetworing) the PSI'I)N charges to the relevant subscribers. 3.2.2.4. The CSPDN Case Circuit Switched Public Data Networks are. a s their name indicates. telecommunication networks d o oted to data, and use circuit transmission. l i k e the ISDN o r the PSTN. as opposed t o packet transmission like in PSPDNs. The standardised user to network interface access for CSPDN follows CCITT X.21 or N.2 I his. GSM provides the specifications hai the support o f N.21 and X.2 Ibis terminals in the mobile station. as well as for interworking CSPDNs. This is done in a way which is Very similar to the previous cases, noting that for CSPDNs only synchronous access is provided. with rates equal t o 2400. 4800 o r 960(1 bit/s. ISDN specilkations already provi4e for these functions. and GSM simply inherited them. The rate adaptation specification in this case is CCITT N.30. which is very close to the synchronous case of V.I 10: the differences are almost entirely in the additional information signals. which are not the same in ('('I'll' N.2I and in c c i r r V.24, hut the structure of the I came is mitetly the same. The differences are only visible in the TA ! : and in the IWF. so the intervening machines are not involved. There are two ways identified in the specifications for accessing CSPDNs. Either GSM is directly interconnected. or a CSPDN is accessed through ISDN. There is no provision for access through the PSTN. In all cases the data flow at the interface between the IWF and the BSS follows the ISDN specifications. N o further adaptation i s needed a s the connection goes o n through t h e I S D N . I n t h e case o f direct interconnection, the M T performs the rate adaptation functions for the translation between the ISDN formal (N.30) and the format for CSPDN. user X.21 GS17.71° x.30 CSPDN X.71 3.3. TRANSMISSION INSIDE GSM user —IWF a) direct access to CSPDN from GSM user X.21 S ITI • I S D N CSPDN u s e r X.30 X . 3 0 p In the previous sections we have looked mainly at the aspects of transmission outside, or at the boundary. of GSM. It is time to tackle the innards of the system. The inner part o f the GSM transmission system extends from a point somewhere in the mobile station (inside the TAF for, data services, and where speech is an acoustic signal for the speech case) and the interworking point between G S M and external networks. Between these two points lie several machines and several interfaces. Our first task will be to present them. 3.3.1. ARCHITECTURE b) access to PSPDN via ISDN from GSM Figure 3.13 - CSPDN interconnection Access to nCSPDNcanbe provided directly or through the ISDN. which is defined b y CCITT X.71. In both cases the interworking is digital, without specific modems. Figure 3.13 illustrates the t w o interconnection configurations. 3.2.2.5. A Summary of the Different Cases Considering all the cases w e have seen, we have two main categories of data transmission schemes from the GSM point of view, depending on the presence or not of a modem in the IWF. However, the inner portion o f GSM transmission, up to and excluding the IWF, is basically the same in both cases. This will allow us to proceed with the study of the inner GSM domain in a unified approach. ignoring the nature of the external network and the presence or not of a modem inside the IWF. Let us start by looking a little bit more in detatt at the functions situated at the borders of GSM (the 1WF on one side, the TA F on the other) before describing the more internal parts ()I' GSM. The IWF is a set of functions fulfilling the adaptations necessary between GSM and external networks. As will he seen. it can he rather limited for speech toward the PSTN. and for basic data when interfacing with ISDN. But in other cases, such as facsimile. Interworking Functions can be quite extensive. The IWF as a function lick somewhere between the M S C a n d t h e external network i t interfaces w i t h . A f i r s t implementation approach is simply to put the IWF in the MSC. and this is the usual approach for simple cases such as spq0ch. For the complex cases, it can also be imagined to have special machines devoted only to the interworking functions, and linked to several M-Ses. This centralised approach i s sensible i f the traffic through the I W F is hut a small proportion of the overall traffic. This implementation is not precluded by the Specifications, but there is no standard specification of an MSC/IWF interface, and any such interlaces will he proprietary. Let us turn t o the mobile station side. The canonical GSM architecture identifies on one side the terminal (1Th. in direct contact with the user, and on the other the core fund ion:links of a mobile which are common t o a l l services. I n between l i e t h e Terminal Adaptation Functions. and ill addition For facsimile the Fa \ Adaptor. The piece of mobile station equipment which contains the functions common to all services is called in tlw Spec/put/him the Mobile Termination (MT). If we turn now to concrete implementations o f this functional model, we find a number of possibilities. di tiering by the grouping of the 150 T H E functions i n specific machines, and o n the interface specifications between these machines. The simplest case i s when everything is integrated, generic functions, terminal equipment a n d adaptation functions i f applicable. Such a machine i s called M TO (Mobile Termination type 0) in the Specifications. This is the approach retained for speech in all known implementations. Integrated mobile stations for other services will certainly appear sooner o r later, for instance for facsimile. The next simplest case, which will be our basis of work in the following for data applications, is when the TAF is totally integrated with the generic functions, and interfaces with the terminal through a classic modem t o terminal interface. This integrated device i s called MT2 (Mobile Termination type 2). Another identified possibility is when the external interface of the Mobile Termination is the ISDN " S " interface, to which o ff -the-shelf ISDN terminal equipment can be directly connected. In this case, the- —.machine is called MTI (Mobile Termination type 1). A terminal using a modem to terminal interface can still be connected to an MTI, provided an ISDN Terminal Adaptor (TA) is inserted. In this case, the Terminal Adaptation Functions (TAF) are spread between the M T I (where a synchronous adaptation is performed) and the TA (where for instance the synchronous/asynchronous adaptation i s performed). T h e different mobile station configurations are illustrated i n figure 3.14. I n the following, w e w i l l n o t distinguish t h e s e d i ff e r e n t physical implementations. We will refer to the Terminal Adaptation Functions in general, whatever their implementations: similarly, we will refer to the TE-TAF interface, which can be according to the case the interface between the TE and the MT, or the interface between the TE and the TA. Let us look now at what exists between the mobile station and the IWF. Along the transmission path, the canonical architecture of GSM distinguishes the BTS (Base Transceiver Station), the BSC (Base Station Controller) and the MSC. Between the mobile station and the BTS is a clear reference point, the radio interface. where the information crosses the space riding the 900 o r 1800 M H z electromagnetic waves. The BTS/BSC/MSC split is adequate for the study of the signalling aspects. But the MSC and the BSC have little role to play in the transmission chain. Historically, the BTS and the IWF were the main actors in the transmission scene, and only basic transmission functions were found between them. Then another piece o f equipment was introduced: the TRAU (Transcoder/Rate Adaptor Unit). which is definitely transmission equipment, and which was conceived to be distinct from the BSC or the MSC. The TRAU will take the starring role in this section. 15 1 TRAN.SN-11SS1ON GSM SYSTEM Mobile Termination (type 0) Terminal Equipment Mobile Termination (type 2) terminal to modern interlace Terminal Equipment Terminal Adaptor A Mobile Termination (type 1) V ISDN "S" interlace Figure 3.14 - Mobile stations configurations Mobile stations can he either tulle integrated. or include a separate terminal equipment connected to a Mobile Termination. through a Terminal A d a ptation junction \\Ind' can he either integrated or kepi as an independent piece of equipment, The rationale behind the existence of the TRAIL distinct limo the MSC and BSC. consists of several points. 'Ilk' implicit assumption during the elaboration of the concept of MSC was that it would he implemented more o r less as a modified ISDN switch. A s a consequence. the transmission at the level o f the MSC is very close to that of the ISDN specifications. In particular only: 64 khit/s circuits are switched. A s a consequence, the A interface must conform to the lower layers o f the ISDN specifications. Indeed. the 2 Nthit/s standard multiplex structure used on the A interface (and also on the Abis interface) is not specific to GSM, but follows the CCITT G.703 standard. Their basic usage is to carry 6 4 kbit/s circuits compliant w i t h t h e needs o f ISDN. T h e multiplexing is based on a 125 us cycle. each cycle transporting one octet 152 T I Ili GSM SYSTEM 64 kbit/s circuits, but, in addition, enables the transport of sub-multiple rates such as 32, 16 or even 8 kbit/s. This possibility is effectively of interest for GSM, which does not require connections of more than 16 kbit/s, and where the cost of internal terrestrial links (between BTS and BSC, and between BSC and MSC), usually leased by the operator, represents a substantial part o f the operational cost. A transmission method using only 16 kbit/s for user data (signalling is kept on 64 kbit/s links) was then devised, to allow this cost reduction which seems compelling despite some drawbacks. First, this introduces some extra delay for the transmission, and hence lowers the overall speech transmission quality. Second, it introduces a gateway function at the border between 16 and 64 kbit/s, which is really the purpose of the TRAU. The late introduction o f the TRAU, and the will to keep the switching capability of the MSC strictly similar to the one of an ISDN switch, is the source of its eccentric architectural location. The TRAU may be located in different places along the transmission chain, between the BTS, to which it belongs functionally, and (but not including) the MSC. One may then deduce that the only site possible when not on the BTS site i s the BSC site. This i s however not quite so: the implementations o f many manufacturers include a remote transcoder situated on the MSC site. The BSC, as a functional unit, is then "spread" over its own site and the MSC site, and includes the link between these two sites. Conversely, the BSC-MSC interface (or A interface) is situated on the MSC site, over a very short distance. Figure 3.15 shows the positions of the TRAU relatively to the other BSS machines. This somewhat artificial definition stems from historical reasons (i.e., reasons which were meaningful when the decision was taken, even though this meaning might have been lost through later evolution...). The definition was chosen in order to avoid the introduction of an option on the A interface, between transporting user data at 16 kbit/s or at 64 kbit/s, since operators have always been keen on limiting the number of options on the A interface to help multi-vendor inter-operability. Manufacturers also might have found an advantage in the decision of having a single rate of 64 kbit/s on the A interface, by avoiding the implementation of switching matrices at 16 kbit/s when the TRAU is put on the BTS side of the BSC matrix. As a consequence, the Specifications strictly speaking do not allow the placing of the transcoder inside the MSC. Every call between two GSM users must then undergo two transformations (from 16 kbit/s to 64 kbit/s and vice versa, entailing for speech two transcoding operations between the 13 kbit/s and the 64 kbit/s representations). even if the two users are connected to the same BTS. One may imagine that such a restriction could be removed in future phases of GSM... 153 !FR ANSM ISSION TRAU BSC BSC MSC VLR TRAU MSC VLR - - TRAU 1. BSC Abu; 'Mortace MSC VLR A HIWIlilcl! 16 kbit s transmission physical site 64 kbit $ transmission Figure .L15 P o s i t i o n , of t h e I \ the T R A t \‘1111.11 i , f u n : n o n a l b p a i l o f Lite f t I'S. can he installed in a remote location lop to the M s ( ' %itch to save link cosh between the 13.1's and Me T R A t thank, to the r e a , c in iran.inksion capacio i n t o l b to i l l M i n h . Because t h e T R A U i s t h e true intermediate equipment f o r transmission, and because of its architectural predi .cament. we k i l l not use the notion o f A arid Ahis iiiterfaces in this cection.,but instead the B S C BTS/IRAU and the -I I< At W I : tor 'FRAU/MSC) .nter..aces. I land the MSC in the ease 01 data) is simply ignored. as it has no special role as transmission equipment (hut some as a switching equipment!). In the following paragraphs we will descrihe the transmission o f speech. then data inside riSVI b u t we t r i l l exclude the details o f the radio interface, since they trill be studied in the next chapter. The relevant interfaces are in each case the following: • t h e radio interface: • t h e BTS-TRAU interface (which can he non-existent i t t r m c o n d o r c •In• c;tinto.1 •if Ile l54 T I I E • and the TRAU-IWF interface, or more generally the interface between the transcoder and the point of interconnect with other networks. points are the "elevators" between the different floors of figure 3.2. The following transcoding points are adentified inside the GSM domain: • Acoustic to Ant. g u e Electric transcoding, implemented in the microphone, an t h e reverse Analogue Electric t o Acoustic transcoding, implemented i n the loudspeaker: this type o f transcoding is not however specific to GSM: Across those interfaces, the transmission system has the task of carrying speech or binary flows for a variety of data services. This still represents quite a f e w different cases, and additional treatment is performed to obtain a higher uniformity. In the rest of the chapter we will tackle these aspects and the detailed transmission schemes inside the infrastructure, first for speech and then for data. The study of speech will take us to the very specific area of digital speech encoding, whereas for data we will discover the subtleties o f the numerous rate adaptation schemes. 3.3.2. SPEECH Digital speech transmission over a radio interface i n a mobile environment is quite a challenge. As already mentioned, a special digital speech coding algorithm is used in GSM, chosen for its low bit rate (13 kbit/s) and its resistance to high error rate conditions. The description of this algorithm will be the first topic of this section. This description is somewhere between the view of laymen (which the authors basically are in this field), and that of a specialist. Some emphasis will be given to some side features of voice transmission, such as vocal activity detection and discontinuous transmission, which are important for the spectral efficiency o f GSM. The rest o f the section i s devoted t o the rate adaptation which enables speech encoded with this algorithm t o be carried not only over the radio interface, for which i t was originally designed, but also over fixed digital links, between the BTS and the TRAU. Taking again figure 3.2 (page 129), the GSM transmission path for speech can be divided into the following segments: a) the mobile station; b) from the mobile station to the base station: the radio path; c) from the base station to the (remote) voice transcoder: d) from the voice transcoder to the MSC. The junction points separating the segments a) to d) described above correspond to places where a speech representation is changed to another one. These transcoding points are of major importance here, since the description o f a transmission scheme is intimately related to the description o f the corresponding transcoding functions. Transcndinp 155 FRANSNIISSIoN GSM SYSTEM • Analogue to 13. kbit/s Digital transcoding (and the reverse operation), implemented in the mobile station. • 1 3 kbit/s Digital t o 64 kbit/s Digital transcoding (and the reverse operation), implemented in the voice transcoder, either in the BTS or in the TRAU. This does not mean that the signal is transported exactly in the same way on all links between t w o transcoding points. The signal representation is adapted to the transmission medium i n intermediate processing points. The two main adaptations are: • adaptation o f t h e 1 3 O W % d i g i t a l representation f o r transmission on the radio path; • adaptation o f the same 1 3 kbit/s digital representation t o transmission on fixed links between the BTS and the voice transcoder in the TRAU. 3.3.2.1. Speech on the Radio Interface Let us dive into sonic more detail and examine the digital mode used t o represent the speech signal f o r transmission over the radio interface. The prime concern for the design of the speech transmission means on the radio path was spectrum efficiency. The goal was to use as low a data rate as possible while providing an acceptable level of quality. Since speech is considered as the prime service. these considerations have heavily influenced the whole design of the system. On the radio interface. two types of raw carrying capabilities are defined, the "full rate" channel (which deserve this title just because it was the first to be specified). and the "halt' rate" channel, which indeed makes use o f half as much radio resource as the previous one. The existence o f these two types collies from history: at the time of definition. .t was certain that it would he easy and quick to specify a digital speech coding at around 16 khit/s with the required quality. and it was foreseen dial it would be possible to do the same thing with half as many bits some years later. Because of the stress n n n , , , , , ; • i l l • • 0 1 1 1 1 1 1 1 1 r111.01.1 t h i b a h ' 610 130 i l l h possibility. It was therefore decided to define the system in two steps, starting with a less efficient coding scheme, but paving the way for a twofold increase in efficiency to be introduced as soon as possible. In the first phase of GSM, speech is then only defined for the full rate channel. The description o f the coding scheme for this type o f channel is presented in the following pages. A s o f the end o f 1991, commercial use of half rate speech is foreseen for 1994 or 1995. In Ia ) > 0 >0-- > O A A The aim of this section is to give the general principles of the GSM speech coding scheme. The bit-exact algorithms for coding and decoding are given in TS GSM 06.10; therefore a specialist shall find no better description than the one i n the Specifications. However, the nonspecialist will find here a general idea of the GSM speech representation. The easiest way to dive into this subject is to look first at the contents o f the transmitted signal, and its translation from 13 kbit/s to 64 kbit/s, which is performed on the decoder's side. Speech is cut into 20 ms slices: in fact, rather than saying that the transmission rate is 13 kbit/s, it would be more realistic to say that speech is transmitted using groups of 260 bits every 20 ms. Synchronisation (i.e., separation of the 260 bits blocks) relies on external means and not on information transmitted within the blocks. The radio interface makes out out In A Gtheral Principles >0V >0V L—-- - > 0 > O L The Speech Coding Algorithm The GSM speech coding scheme at 13 kbit/s is called RPE-LTP, which stands for Regular Pulse Excitation-Long Term Prediction. It aims a t producing a speech quality similar—when n o errors are added—to the fixed telephony network quality, with a much smaller rate in order to optimise the use of the radio spectrum. I I RANSNIISSli \ GSNI SYSTENI > O .1/1.O A -c • I — Ia sI shift register a multiplication by a scaling factor a Figure 3.17 —A linear filter and the corresponding inverse filler A linear filter combines the input signal with delayed representations of itsd 1 and IN usually represented bs a polynomial s l u m n \\ (Add correspond to: =a+ + + efi +/E5. and the inverse filter (shown below the littear filter o u l d correspond to 1/.1i:t. extensive use o f its complex synchronisation scheme: this is why the 13 kbit/s flow, structured i n frames ()I' 2(1 ins, does not include any information helping t h e receiving entity t o determine t h e frame boundaries. For each block. the output signal is reconstructed h) the receiver from an input signal (the excitation signal) which is filtered through a succession of filters (i.e., of linear transforms). as shown in figure 3.16. The Long Term Prediction—or LT V —filter is a very simple filter. which consists in adding the signal and its delayed image multiplied bx a factor b,.. the delay being A. samples. I he values of both \ , and a r e transmitted in the speech frame. once for every 5 nis slot. Figure 3.16 — Modelling of the 13 OUR speech signal —•or L P L -filter is the inverse of an The Linear Prediction 8111order linear filter. A linear filter o f n'h order performs a linear combination of the signal and of itself delayed by I . 2. .. n samples at 8 kHz. The coefficients of this filter vary from one block to another. and are transmitted in the speech frame. The stmt.:lure ()I' a linear filter and the inverse filter can be found in figure 3.17. Speech can be generated from an excitation signal S by applying three filters in succession: LW, LPC and a defined de-emphasis filter. The de-emphasis filter is pre-defined. and therefore requires no parameters to he transmitted. TRANSMISSION Bits per 5 ms Bits per 20 ms 15') The Speech Decoder The structure o f the speech decoder follows very closely the modelling described above. The major processing stages are as follows: LPC filter 8 parameters LTP filter N, (delay parameter) 7 28 br (gain parameter) 2 8 Sub-sampling phase 2 8 Maximum amplitude 6 24 • L T P filtering, involving the samples of the current 5 ins block and of the 3 previous ones: 13 samples 39 156 • L P C filtering, according to the transmitted coefficients; 260 • d e -emphasis filtering. Excitation signal 36 Total • Production o f the excitation signal by putting the 13 samples back to scale, and production o f an 8 kHz sampled signal by adding 27 null samples. according to the phase indication; Table 3.2 —Parameters transmitted in each speech frame 'to In each speech frame, 188 bits relate to the excitation signal. The other bits are the parameters of the LPC and LTP filters applicable during the 20 ins period. This is a simplified picture of the decoder; the interested reader shall find i n the Specifications the full description o f the quantizing methods, as well as the description of some additional processes which aim at smoothing the signal at the boundary between successive blocks. so as to avoid sharp transitions. The Speech Coder Last but not least, the excitation signal S itself is coded so that the set of all parameters—those of the above filters plus the description of the signal S—fits into 260 bits. S is sampled regularly ("RPE, Regular Pulse Excitation") at a rate of only 8/3 kHz. According to the signal processing theory, this allows to know accurately the information concerning only the lower 1.3 kHz bandwidth of S. The excitation signal at the input of the filters is reconstructed by inserting null value samples, so as to obtain a signal sampled at 8 kHz. From a spectral point of view, this results in a signal with spectral components above 1.3 k H z which are derived (second and third harmonics) from those below. The phase of the 8/3 kHz samples with regard to the 8 kHz samples can vary, and is transmitted once for every 5 ms slot. The coder is somewhat more complex than the decoder, since it must compute the different parameters so that the reconstructed signal will be as close as possible to the original one. like order of its processes is the reverse of the decoder's. Starting from the 64 kbit/s digital signal (i.e., speech samples at a rate of 8 kHz), the A law 8 hits samples are converted to 13 bit samples corresponding to a linear representation o f thh amplitude. This linear representation of the signal goes through a first filter (the pre-emphasis filter). 160 samples o f the resulting signal. representing,' a period o f 20 ins, are grouped together in a block; the next two stages (1..PC analysis and filtering) are performed on sucli blocks: The samples are coded using Adaptive Pulse Code Modulation (APCM). This coding i s called "adaptive" because the maximum amplitude and the ratio of each sample to this maximum value are coded separately. This differs from the usual 64 kbit/s coding, where each sample is directly coded using a fixed scale. Table 3.2 summarises all the parameters transmitted in each 260 bits frame (i.e., every 20 ms). I • t h e LPC filter is chosen in such a way as to minimise the energy of the signal obtained as the input signal filtered by the reverse LPC filter. Because the speech signal is inherently yeny redundant, a substantial gain tin term of energy reduction) may be achieved through such a choice of LP(' coefficients. When LPC filtering is applied by the decoder, the energy of the 1.13C filtered excitation signal will be amplified by the corresponding factor, whereas quantization errors _will h e amplified b y a smaller factor. The tolerance f o r quantization errors ()I' the excitation signal is then higher. thus allowing the use of less 160 T H E (ism sYsTFINt i I R A N S M I S S I t b bits to represent it than what would be needed for the original signal itself; Discontinuous Transmission, and Voice Activity Detection • t h e next steps are applied to the signal obtained after filtering by the reverse LPC filter, called the "short term residual signal"; An important side aspect of speech transmission is what is called in the GSM jargon the M X mode (Discontinuous Transmission). This corresponds to What has been sometimes called in the US literature the variable hit rate. It aims at increasing the system efficiency through a decrease of the interference level. by inhibiting the transmission of the radio signal when not required from an information point of view. This DTX mode is an optional alternative to the normal mode. Because the DTX mode slightly deteriorates the quality of transmission, in particular when used twice along a path, that i s t o say i n the case o f a communication between GSM users. the choice between the two modes can be done by the network on a call per call basis. The residual signal is then processed by blocks o f 40 samples representing 5 ms of speech, as follows:. • t h e coefficients of the long-term filter are calculated in such a way as t o 'minimise the residual power, with the following constraint: the delay R. must be in the range 40 to 120 samples (i.e., between 5 and 15 ms); this range has been chosen so that this delay corresponds usually to the fundamental frequency of speech, which is somewhere between 60 and 200 Hz depending on the speaker; • t h e short term residual signal is then filtered by the reverse LTP filter, and sub-sampled to a third of the original 8 kHz rate, with a phase chosen as the one corresponding t o the result o f maximum energy; the resulting samples are coded in APCM. Implementation Both the speech coding and decoding algorithms described above use many calculations. A s an indication, the coding and decoding of speech requires about 1.5 million elementary operations (addition or multiplication) per second. Digital Signal Processors—called DSPs—are hardware components specialising in this type o f calculation and are therefore commonly used in speech coder/decoder implementations. Speech coding raised a problem for the specification o f the type approval tests of the mobile stations. Subjective testing, which is used to compare coders, is very expensive, since it involves ratings by several independent human listeners. Moreover, in real operation, the coding and the corresponding decoding are done in machines from different makes (one is in the mobile station, the other on the infrastructure side); a complete test would then require to test all possible pairs of coder and decoder! S u c h testing w a s avoided b y specifying a complete implementation using integer arithmetic o n 1 6 and 3 2 bits. A s a consequence, the results produced by any implementation o f the coder and decoder are not allowed to differ—not even by one bit—from the specified one. In the DTX mode. the goal is to encode speech at 13 kbit/s when the user is effectively speaking. and otherwise at a bit rate around 500 bit/s. This low rate flow is sufficient to encode the background noise. which is regenerated for the listener (this is the notion of comfort noise) to avoid him thinking that the connection i s broken. The l o w rate encoding corresponds to a decreased effective radio transmission since. to be exact. the active speech flow is of one frame of 260 bits each 20 ms. and the inactive speech flow is o f one such frame each 480 ins. The choice of these values is related to other features of the radio interface. In order to implement such a mechanism, the source must be able to indicate when transmission is required or not. In the case of speech. the coder must detect whether or not there is some vocal activity. This function is called Voice Activity Detection, or 1';1/). A t the reception side, the listener's ear must not he disturbed he the sudden disappearance of noise and the decoder must therefore he able t o generate some "comfort noise" when no signal is received. One must he careful to distinguish hoween activity detection. which i s o n l y concerned i n deciding whether some part o f the information f l o w m a y a v o i d transmission. a n t i discontinuous transmission. A c t i v i t y detection i s i n n dependent (speech. various kinds o f data, w h e r e a s D I X i s hound t o transmission characteristics: it raises issues such as its impact on radio measurements. synchronisation recovery, etc. This section only deals with VA D and the generation o f comfort noise. which are specified respectively i n TSs GSM 06.31 and 06.32. The consequences o f DTX on radio resource management will be described in Chapter 6. 162 T i n , . GS \I SYSTEN1 T R VAD — Voice Activity Detection A s \liNsIWN 1 6 3 BTS. A basic principle guiding the specification is that no additional signalling flow exists between the BTS and the remote transcoder unit (TRAU): all the necessary information is carried in-band. Before lookinng at the contents of this information in detail, let us study the requirements to be fulfilled. The VA D algorithm is very closely linked to the speech coding algorithm. For each output frame, the coder provides an additional bit of information indicating whether the frame must be transmitted or need not be, depending whether the algorithm decides that it contains speech or background noise. The Requirements The decision is mainly based on a comparison between a threshold and a measure of the filtered signal energy. The role of the filter is to distinguish a speech signal from background noise. Both the threshold and the filter are continuously adjusted according to the characteristics of background noise which are evaluated during non-active periods. Synchronisation The 13 kbit/s speech encoded stream is structured in successive blocks of 260 bits every 20 ms. The stream itself does not contain any information allowing the receiver to determine the starting bit of a block: this must be provided separately. On the radio path. this is provided by the general synchronisation mechanism. On a 2 Mbit/s link, the longest cycle which can be guaranteed from one end to the other is the 125 ps cycle, which amounts to groups of 2 bits for a 16 khit/s link. A part of the remaining 3 kbit/s must then he used for the 20 ms synchronisation. Generation o f Comfort Noise Experience has shown that a listener is greatly disturbed when the background noise behind the speech suddenly stops. This would happen. regularly w i t h discontinuous transmission. A mean t o avoid the disturbance is to generate an artificial noise when no signal is received. The characteristics of this noise are regularly updated and transmitted to the receiving end by the speech coder situated at the other end. Time Alignment These characteristics are transported by specific frames called SID frames (Silence Descriptor frames). A SID frame is sent at the start of every inactivity period, and more are then sent regularly, at least twice per second, as long as inactivity lasts. In order for the receiver to be able to distinguish speech frames from SID frames, the latter uses certain combinations of values which cannot be found in an error free speech frame. Both the radio path and the 16 kbit/s interface use a 20 ins structure. Moreover, in the downlink direction, transmission on the radio path can start only when a while 20 ms block is received. There is then an optimum time relationship between the moment of the beginning of a block transmission on the radio path. and the end o f the reception of a block from the [ 6 khit/s link. The transmission suffers an additional delay, of up to 20 ms, if this relationship is not optimum. 3.3.2.2. Speech on the B T S - T R A U Interface For this reason, a small protocol. needing a part ol' the 16 kbitis stream, allows the BTS to control the phasing of the incoming 20 ins blocks generated by the TRAU, 4 Speech/Data and En/l11all Rate Discrimination On the infrastructure side, the functional entity where 13 kbit/s encoded speech and 64 kbit/s encoded speech are translated one into the other is the TRAU. When the TRAU is physically distant from the BTS. for instance on the MSC site, the 13 kbit/s stream must be carried between them, over standard digital links. This makes use of 16 kbit/s circuits. In fact, the 16 kbit/s bit stream contains more than the 13 kbit/s encoded speech bit stream. It also includes some auxiliary information to carry the 20 ms synchronisation (which cannot he derived from the 13 khit/s flow) and to allow the remote control of the transcoder by the In the future. when a half rate speech encoding scheme is specified. different capabilities will be required in the TRAt I. Beside. not only speech, but also data (as w i l l he seen i n the next section) can he transported on 16 khit/s links. .;I For all these reasons. in-hand information allows the TRAI. t o know what kind o f information i s received and then what type o f adaptation it must apply both for the uplink and downlink transmission. 164 ' 1 1 1 E 165 TRANSNIISSION GSM SYSTEM Reception Quality The receiver (demodulator and decoder) in the BTS is able to tell whether its correcting capabilities have been overwhelmed by the number of errors o r not. The information about whether a speech frame is regenerated correctly or not is very important for the transcoder, which will ignore a frame indicated as bad. Such an indication is therefore conveyed from the BTS to the TRAU for each frame. TRAU to BTS n t l BTS to MS I I 1 1 _J 1/4 4 8 0 ms 70- DTX Mode Aspects As already mentioned, two modes o f speech transmission are possible. In one mode, the speech flow is coded continuously at 13 kbit/s (one speech frame each 20 ms), irrespective o f whether the user is speaking or not. I n the second mode (DTX mode), the transmission alternates between speech-active phases, with a transmission o f one speech frame each 20 ms, and of speech-inactive phases, during which the transmission falls to a frame each 480 ms (only the comfort noise characteristics are transmitted). The mode can be different between the uplink and downlink part of the connection. The speech transcoder need not be told whether transmission in the uplink direction is in the DTX mode or not. A non-transmitted frame is seen as a bad frame, and the distinction between comfort noise frames and speech frames can be done on the basis of the frame contents. On the contrary, the speech transcoder must be told by the BTS which mode to apply for the downlink stream. In phase 1, this indication i s not supported b y the network signalling: this will be corrected for the phase 2, and one can assume that a number o f remote transcoder implementations w i l l anticipate this correction. Even in the D T X mode. the transcoder provides continuously correctly formatted frames, either encoding speech o r comfort noise. However, the BTS has t o decide whether o r not a frame must be transmitted to the mobile station. Obviously. all the frames encoding speech must be sent. In addition, from a speech transmission point of view, some of the frames encoding comfort noise must also be sent. The rule for this is that the last transmitted frame preceding a sequence of non transmission must be a comfort noise frame (see figure 3.18). The other frames (all comfort noise frames) may be transmitted or not, as the BTS Comfort noise frame I I Speech frame Figure 3.15 'Transmission of speech frames The TRAU continuously transmits frames. but the BTS nay stop transmitting some comfort noise frames according to some synchronisation rules. but must take care to send a comfort noise frame before entering this mode. wants. The distinction between speech frames and comfort noise frames can be done by looking at the frame contents. It would have been useful if the transcoder gave the information whether, from its point of view, the frame must be sent or may not be sent. This is not so. and the only auxiliary information w h i c h i s provided i s t h e redundant - S r ' information, which tells whether the frame is a speech frame or a comfort noise frame. It is the BTS which must take care that at least one comfort noise frame is transmitted before going to the I frame per 480 ms mode. Other Information For historical reasons. sonic other information, which can he regarded as useless or redundant. is added. We have already encountered one such case in the guise of the SF indication. Another example is the SID indicator (Silence Descriptor), used in the uplink direction to sort frames into three categories according to the number of " L s received among the 95 hits which are all set to -0- in a comfort noise traink.. Vet another piece o f redundant information is the Time Alignment Hag indication in the uplink. used to mark the one frame each 480 ms when in low rate. In fact. as far as can he seen. the helm lour of the transcoder is not affected by the value of this indication. TRANSMISSION THEGSM SYSTEM 166 Number of bits in uplink frames Number of bits in downlink frames Frame synchronisation 35 35 Discrimination betweenspeech and data, full rate and half rate 5 5 Time alignment 6 6 Badframe indication 1 DTX mode I (not in phase I) Other information (specified, mandatory but redundant) 3(BFI + TAF) I (SP) Speech block 260 260 Spare 5(6 in phase I) 9 Table 3.3—Contents of aspeech block for transmission at 16 kbit/s The block contains 316 bits, among which 260 represent speech. The 16 kbit/s Bit Stream Structure The 1 6 kbit/s b i t stream i s structured i n successive blocks occurring on average each 20 ms. A block contains 316 bits, leaving on average 4 bits between successive blocks to cope with the variations due to time alignment. The contents of the block is summarised in table 3.3. The details of the format, and the detailed protocol for starting and maintaining the time alignment, can be found in TS GSM 08.60. - 3.3.2.3. Speech on the TRAU-IWF Interface On a 64 kbit/s link, the standard G.711 speech transmission is used, with A law coding. 3.3.3. DATA We have seen in the first part of this chapter the general aspects of the connection between users when circuit data services are provided. The grand result was that between the mobile user terminal and some 1 6 7 point inside the IWF, the GSM segment was an over-stretched DTE (data terminal equipment, o r terminal) t o D C E (data communication equipment, usually modem) junction. This statement is true even when the IWF does not effectively include a modem. e.g.. when direct digital interworking i s possible. Seen like that the different cases f o r data transmission correspond to different cases o f "modems'•, with various data rates and other properties. We will devote the rest of this chapter to study how the synchronous GSM radio interlace has been made suitable to become part of the various types of terminal to modem connections. The design of the GSM data transmission was led by two main issues. The first one is simply how to transport a multi-wire interface such as a standard terminal t o modem interface, possibly using the start/stop transmission format. o v e r basically a single w i r e a n d synchronous medium. A similar issue exists i n ISDN. and technical answers are specified in the ISDN recommendations, such as V.110. The second problem is the radio interface itself. with its high raw error rate. Though an important design goal for a transmission system is to reduce the number o f different ways i n which the data has t o be transported, several connection types have been defined in GSM. They reflect two widely different approaches o f how to adapt a terminal to modern interface, and warrant a detour to explain the different types of connections which have been designed between the mobile station and the IWF. 3.3.3.1. The Connection Types In the case o f GSM the w i l l t o limit the number o f different transmission modes had to be pondered by transmission quality and delay considerations, and the result is a compromise. Transmission over mobile radio links is not by far as easy as over fixed lines. In particular. raw bit error rates of more than 10-3 are not uncommon on such links. For speech transmission, this does not cause a problem since the human ear is•very tolerant to noise, but it is unacceptable for most data services. For such services, error correction protocols do in fact exist, either end-to-end or on' some segments o f the transmission path. Ye t these protocols are !designed for the error conditions that are common over fixed lines. and these conditions are far more lenient than those found on the radio transmission. Besides. such error correction schemes are not even used for some data services. It was then necessary to provide error correction schemes inside the GSM transmission path. Information theory tells us that for a given raw bit rate and given error conditions, the characteristics of transmission when error correction I N is provided i s a compromise between the throughput o f data, the transmission delay and the remaining error rate. In the case of GSM, no single trade-off fits a l l the different types o f services. Out o f the possibilities, a short delay despite a relatively high error rate is better in some cases, whereas (for fax for instance) a long delay can be tolerated in order to achieve a better transmission quality. For these reasons, several types of connection are provided in GSM. A fundamental division exists among the different types. In one category, the error correction i s entirely done b y a forward error correction mechanism provided b y the radio interface transmission scheme. This case is referred to as "T" in the Specifications, as will be explained later. I n the other category. referred t o as " N T " i n the Specifications, an additional scheme i s used, where information i s repeated when it has not been correctly received by the other end. The T mode o f transmission i s derived f r o m t h e ISDN specifications (in particular from V.110), and the path between the TAF and the 1WF is seen as a synchronous circuit: the available throughput is constant, and the transmission delay is fixed. The information exchanged between the two entities include the user data, at some rate between 600 and 9600 bit/s, plus some auxiliary information, as in the case of ISDN. As we French say, qui peat le plies peat le moms, and it could have been imagined that a single transmission mode, able to transport the highest rate, could have been used. Simple rate adaptation schemes could have been designed for the transmission o f other rates. But because better protection can be provided by forward error correcting schemes when the rate to transmit is low, the preference went to the definition o f three different intermediate rates, despite the increased complexity. The lower user rates, up to and including 2400 bit/s, are grouped into a single category, as i f all were at 2400 bit/s. The two other intermediate rates correspond to user rates o f respectively 4800 and 9600 bit/s: User rates up to (and including) 4800 bit/s do not necessarily require the maximum channel capacity ("full-rate" channels) on the radio interface. b u t the highest rate o f 9600 bit/s does. T h e lower the intermediate rate, the better the achievable transmission performances, as shown in Table 3.4. The table also shows the different cases which will appear in the future, when "half-rate" channels are used. In the case of 9600 bit/s, or of 4800 bit/s on the half rate channel, the performance is not very satisfying, and this explains why a different approach (the NT mode) was specified in addition. In the N T approach, the transmission o n t h e G S M circuit connection is considered as a packet data flow (though the offered service, end-to-end, is most often a circuit service). Consequently, the available throughput varies with the quality o f basic transmission (the \ User rate Intermediate rate Channel p e Residual error rate 9600 bit/s 12 kbith, full rate (1.3 (.:; 4800 bit/s 6 kiln/s full rate halt' rate 0.01 c 0.3 ,; 5 2400 bit/s 3.6 OM% full rate half rate 0.001 1, 0.01 ,, Table 3. 4 - PerMrmanee versus rate in the T mode On a given radio channel. a lower data rate leases more room for an efficient forward error correction scheme, resulting in less residual errors. The figures for the residual error rates are esiracied from TS GSM 05.ofi. considering typical urban conditions with frequency hopping. higher the probability of error, the lower the throughput), as well as the transmission delay, but the resulting transmission quality in terms o f residual errors is much better than in the T approach. The basic rate is 12 kbit/s (6 kbit/s on the half-rate channel), whatever the user rate. and with the same forward error correction scheme as for the T connection. respectively for 9600 bilis and 4800 bills. Bits are considered grouped in successive frames o f 240 hits. The frames include redundancy bits to enable a receiver to detect remaining errors. and this is the basis for a repetition-when-needed correction scheme, the RLP protocol (Radio Link Protocol, the expanded name as often is misguiding, so we will use only the acronym). This protocol is operated between the TAF and the IWF. In rough terms, the user data stream (including auxiliary information) is sliced in blocks of 200 bits, sprinkled with some redundancy and sent to the other end within frames of 240 bits. Each frame is numbered. When received, the frame is tested for correctness thanks to the redundancy. II' found correct, the receiver acknowledges the reception (using the frame number). I f not, a negative indication (either explicit, or by default) is received by the sender which tries again. k i t is basically a link protocol. such as those used for the transport of signalling messages. More details on its functions and characteristics have therefore been grouped «Tether with the description of similar protocols. in Chapter 5. The insertion o f RI.P in the transmission chain raises a major problem. which is the data late. If the user rate of %DO hit/s is considered as an example. the addition of the auxiliary information leads (as in the T mode) to a rate of 12000 hit/s. The question is then ''here to find the overhead rate which i s needed f o r the error detection. the frame numbering, the acknowledgement indication. the repetitions. and so on. The trick was t o notice that i n many cases o f end-to-end data an I /U T H E TRANSNIISSIoN GSM SYSTEM important overhead exists already. For instance, the start/stop protocol is quite inefficient: t o each character o f 8 bits sent b y the user, the equivalent of at least 2 bits is added, in the form of the start and stop signals. This represents an overhead of at least 20%. Another case which is exploited is when a low level protocol is used, like LAPB in the case of an X.25 access or in the case of a fax transmission. The main functions of these protocols are framing, error correction by retransmission and flow control, all functions that are also fulfilled by the RLP protocol. In all cases, e.g., start/stop and LAPB, the idea is to replace them on the GSM segment, by some equivalent features adapted to the GSM transmission environment. Because o f this replacement, a connection using RLP is referred to as "Non Transparent", or NT in short. The plain circuit case is referred to as "Transparent", or T (once again, we will use henceforth only the acronym). An important consequence of this replacement concept is that the RLP approach can only be used in some specific service configurations (in fact only the cases cited above), for which the GSM machines know which low layer protocol is being used. To summarise, restricting o u r view strictly t o the TAF-IWF segment, abstracting the structure of the data flow performed by the TAF, the IWF and the external world, data transmission in GSM provides 4 different types of data connections in the first phase of the system. When half rate channels are introduced, they will be used to support the existing 2400 bit/s and the 4800 bit/s types of data connections, and one new type, the 4800 bit/s connection with RLP. In fact, the type of channel is not neutral. I t influences the transmission delay and the error statistics. Because these characteristics are important for some services, it is best to take the radio channel type into account when defining the types of connection. This means that the TAF and the IWF must know, or even control, this parameter. We then obtain 7 cases, summarised in Table 3.5. Seen from the outside world, the rate 9600 bit/s is offered with two compromises of quality and delay, rate 4800 bit/s will be offered with three compromises between quality, delay and radio spectrum utilisation. and rates up to 2400 bit/s with two compromises between quality, delay and radio spectrum utilisation. 3.3.3.2. T h e General Issue of Rate Adaptation Let us return to our muttons, and look at how the terminal to IWF connection is performed. Not very astonishingly, the T and NT cases are very different in this account. In the T mode, the GSM choice was to put the modem (if any) in the IWF, and to have the transmission specification , ()unlit, of service Name Helm Itwo-wa3. TA F-I WF I TCH/F9.6, T lim .330 ms TCH/F9.6. NT high > 330 ins TCH/F4.8 a l medium .130 ins TCH/F2.4 (T) medium 2110 ins TCH/114.8. I low NM ins I TCH/H4.8. NT high > NM ins TCH/H2.4 (Ti medium 6(11) HIS Table 3,S • The 7 CiSNI data connection types Channels such as the IC'11/1; rfull-rate- channel. 23 kbit/s raw hit rate) and in the future the TCII/I I ("hall-rate- channel) carr user data rates of op to 9600 hit/s. with different compromises lictween a n d quality. The quality estimation is only statistically indicative. since in chosen geographical spots even the 'ICI may lead to very good results, at the TRAU to IWF hued/ace (64 kbit/s) conform exactly with ISDN specifications, in particular 'with V.110 or X.63 (the two being rather close, we will focus on V. I 10 in the following). This fulfils all the needs. whether f o r the remote access t o a modem o r f o r direct ISDN interworking. An advantage of this choice is that no further adaptation is needed in the IWF in the cases of interworking with the ISDN without a modem, at least on the transmission level. For the NT cases, the transmission between the TAF and the IWF makes use of the RLP. which then has little involvement with the rate adaptation schemes o f ISDN. The connection is best described in two layers. The higher layer corresponds to the conversion o f the signals between the terminal and the IWF into something which can be carried by the RLP. This aspect involves the TA R and the IWF for the reverse conversion, and the RLP transfer in between. The second (and lo \\L.) layer includes the means to transport the RI.P frames between the TAF and the IWF. We will see that this lower layer uses means derived from the T mode (in particular this transport is basically a synchronous circuit). Some knowledge of V.110 is a key to the understanding o f the GSM transmission scheme. The detailed description will then start with a summary o f V.1 ID. seen as the basis f o r the interface specification between the TRAU and the IWF, followed by a description of how the I 'L.. LIJI.1 J 1 J I IIINI matter is tackled in ISDN. Then we will look in turn at the T and the NT mode, before looking at individual segments along the GSM TAF-IWF path. V. 11 0 Let us first look at how the different issues mentioned above are solved in ISDN, that is to say • t h e transport of the auxiliary information; terminal to modem signals modem in terminal signals mean sampling rate status of circuit HIM (data terminal ready) status of circuit 1117 (data so ready) each 1.25 ins or 2.5 ins status of circuit 105 (request For sending) state of circuit l09 (line signal detector) each 2.5 ins or 5 its status of circuit 106 t ready kw sending) each 2.5 ins or 5 Ins • t h e transport of an asynchronous flow over a synchronous link; Table 3.6 - V.1 10 transportation of model t control signals • t h e transport of a synchronous flow over a synchronous link using independent clocks. The auxiliary information transported in V.I ID 'ratites corresponds to the sampling of circuits between ISlli (terminal) and DCF. (modem). CCITT Recommendation V.110 specifies how ISDN supports a data terminal equipment (DTE) equipped with a V-series interface. It contains the specification o f an asynchronous/synchronous adaptation (the RAO function, f o r Rate Adaptation 0, though i t is not strictly speaking a rate adaptation like RAI o r RA2, which will be described later). It contains also the definition o f the auxiliary information to be carried, and the sampling rate, as well as the means to support network independent clocking. These specifications were used unmodified for the GSM application. The Modern Control Signals In V.110, the auxiliary information is limited to two additional signals in one direction (from the terminal), and three in the other. This additional information amounts to 8 bits every 5 o r 10 milliseconds (according to the bask rate), which are multiplexed together to form a single bit• stream. Table 3.6 lists the different signals, as well as their sampling rates. At first view, this is not too difficult. It seems enough to impose a short delay to the hits \\ hich arc not truly in line. This works line if the real (asynchronous) throughput is below the maximum (synchronous, capacity, so that the time lost in the delaying process does not add up. The maximum data rate may he theoretically the same before and after the adaptation, but in fact the frequency inaccuracy allowance is such that the incoming rate may he actually higher than the outgoing rate. The difference may be only a few percent or even a fraction of percent, but this is sufficient to cause trouble. The specification being there to answer all nasty problems, this case was foreseen, and it is allowed to skip no‘‘ and then a stop signal to compensate. The receiving side must then detect asynchronous flow 1 1 IT IT:T1tTi l I I M T 1 7 . 1 I I1 7 1 . ; 1. 117.1 .11-14 skip of one 1/2stop signal The RAO Function As mentioned previously, a n asynchronous data f l o w i s a succession o f characters, each typically 8 hits and preceded by a start "bit" and terminated by a stop "bit". On such flows, it is not required that the bit edges fall with a regular clock. Figure 3.5 (page 136) gives an example of such a flow. In ISDN as well as GSM. data transmission is only synchronous. and the role of the Rate Adaptation 0 is to transform the asynchronous flow into a synchronous one, as shown in figure 3.19. Like a sergeant major, the RAO must get the hits in line. synchronous flow [ I 111111 i i i f 11111111 i I h i l t L I F T 1 - 1 7 Fieure t.19 T h e ItAlt"atlaptantnr ltmction The role or RAO is to convert an as) nehninous rim, into a s)nchronons nue for transmission into systems .tell I I S M or GSNI. IRANS NI IS XII the deletion, and reinsert the missing stop element. The same problem of speed adaptation exists at the receiving end. I t i s solved there by shortening the duration of the stop signal. Network Independent Clocking I /? These mechanisms make•use of a set of 8 commands. such as "no change", "accelerate your clock toward the terminal by 20 ( I ' . "skip one bit". "add a bit of value I". and so on. The ISDN General Rate Adaptation Scheme The exact transmission rate through digital networks such as ISDN is imposed by a network clock. Now, the exact end-to-end rate may be different when one end is in the PSTN. A n audio modem used with a PSTN line may synchronise its transmission on its own clock. In this case, t h e • frequency tolerance i s 1 0 0 ppm, a v e r y h i g h value corresponding to one bit in excess or in default each second at 9600 bit/s. To cope with such cases, V.I 10 includes mechanisms which enable a rate adaptation unit to indicate frequency corrections to the other. These mechanisms also allow an indication of when a bit has to be skipped, or on the contrary to be added and in this case the bit value. asynchronous flow (50 to 19200 bit/s) RAO' RAO I synchronous flow When considering the end-to-end synchronous case. o r when looking at the portion o f the transmission path hem een t w o R A 0 functions in the case of the end-to-end asynchronous case. the data flok‘ consists i n a bi-directional synchronous f l o w a t the nominal rate. accompanied by some auxiliary information representing a rate of a few kbit/s. Rate adaptation proceeds in two steps. called respectively RA I and RA2. The whole process is shown in figure 3.20. The RA I function provides a bit Elm\ at the intermediate rate of 8 kbit/s or 16 kbit/s, according to the nominal rate to transport. RA I is not only a rate adaptation function. It also includes the multiplexing and demultiplexing between the auxiliar, information (modem control plus other signals) and the main flow, as slimy in figure 3.21. This is done in time multiplexing. that is to say that theimultiplexed hit flow is a regular alternation of bits of the main Ilow and of auxiliary information bits. This requires synchronisation between the multiplexer and the demultiplexer. which i s maintained b y transporting additional hits. The period o f recurrence for the multiplex structure is either 5 nm (for the 16 kbit/s intermediate rate) o r 10 ms (for the 8 kbit/s intermediate rate). and defines successive multiplex frames. By the way. these values of 5 and (600 to 19200 bit/s) RA1 I synchronous 8 or 16 kbit/s RA2 N RA2 Asynchronous S y n c h r o n o u s raw rate, • -• - - - e.g., 300 or 9600 bibs Y ' RAO ': Sarni:dog synchronous 64 kbit/s Intermediate rate (8 or 16 kbilis) RA1 • synch fill „ R A 2 fia Figure 3.20 — Rate adaptation in ISDN: the three steps Figure 3.21 R a t e adaptation in ISDN: R A I Rate adaptation in ISDN proceeds in three steps: —RAO when necessary for the asynchronous-synchronous conversion; —R A I u p w a n intermediate rate o f S or 16 kbit/s; —RA2 for "padding" of this intermediate r u e into 64 k h i t h channels. Pictured on the transmitter side, the R A I runt:lion multiplexes a synchronous user data f l o w w i t h sampled auxiliary inform:it ion and n e h r o n i s a i ion to obtain an intermediate rate 176 .rizANsmissii TIM GSM SYSTEM bit name information carried comment (V.110) SI. 53, S6. S8 (or SA) 54. 59 (or SB) status of circuit 108 (data terminal ready) or 107 (data set ready) depending on direction status of circuit 105 (request to send) or 109 (line signal detector) depending on direction X status of circuit 106 (ready to send) sent twice per frame El. E2, E3 real data rate This indicates the end-toend data rate network independent clocking (used in synchronous cases to control the remote clock when modems are not synchronised on the transport network) Codes a speed-up or slowing of the clock rate or bit skip or insertion E4, ES. E6 E7 40 ms synchronisation in the case of 600 bit/s only (X.30 compatibility); may also carry information relating to end-to-end flow control Table 3.7 — V. 110 auxiliary information 8 "status" bits (SI, S3. S4. S6. S8. S9 and two X) and 7 "E" bits (El to E7) carry auxiliary information in each V.I 10 frame 10 ms are one of the reasons for the choice of 20 ms as the frame period for speech in GSM. When seen at the transmitter side, RAI proceeds for the lower bit rates to a first rate adaptation. This is done for nominal rates below 4800 bit/s, by repeating each bit as many times as necessary to obtain 4800 bit/s. This raises no difficulty since all the lower standardised rates are sub-multiples of 4800 bit/s. Information is added in the digital stream to indicate the effective end-to-end rate. After this step. the data rate is either 4800 bit/s, or 9600 bit/s. A frame, with a period of 5 ins for a basic rate of 9600 bit/s, and of 10 ms for a rate of 4800 bit/s, then contains 48 user data bits. The auxiliary information hits are then added through time multiplexing, a s w e l l a s enough synchronisation b i t s t o obtain respectively 16 kbit/s or 8 khit/s. The auxiliary information accounts for 1 7 7 15 bits (see Table 3.7 for the list) and the synekonisation bits for 17 hits of each multiplexed frame. The RA2 function rate-adapts the intermediate rate to 64 kbit/s, by. simply adding 7 or 6 bits set to I in each octet. This can he seen as timemultiplexing betwcen the intermediate-rate flow and a constant flow consisting of all "I-s. In this case, no additional synchronisation hits are necessary. This comes from the fact that the basic ISDN channel is not strictly speaking a 64 kbit/s channel. but an 8 kilo-octet per second channel. This subtle distinction means that the grouping of hits by 8, in a way unambiguous for the two end parties. is a side product of the ISDN transmission scheme. This property. usually referred t o as " 8 k H z integrity'', allows to easily define sub-flows at rates which are an integral multiple of 8 kbit/s. 3.3.3.3. The GSM T Connections The transmission path bctwee t h e TAY' on the mobile user side and the IWF is functionally totally equivalent to what appears between the terminal to "modem- interface and the 64 kbit/s circuit in the case of an ISDN connection using V.110. So the RAO. RAI and RA2 functions will appear somewhere between the TA F (included) and the I W F (excluded). However the transmission over the radio interface must he introduced somewhere in the picture. Data transmission on the radio interface is not clone at 64 kbit/s. and V.110 obviously cannot be used in its pure form. A first idea could be to keep V.110 as it is with the exception of the RA2 function. which is very simple, and has clearly to do only with transportation over 64 khit/s circuits. Between R A I functions, the transmission i s done a t a n intermediate rate, lb khit/s or 8 kbit/s. which could have been fitted onto the transmission over the radio interface. Yet. the problem on the radio interface is to limit as much as possible the information to be transmitted. so that the maximum part o f the rov. throughput can he denoted In optimised redundancy. in order to maximise the transmission quality . When the V.110 bit stream at the intermediate rate is looked at. it becomes apparent that an important part o f the exchanged hits can he removed in GS%i. The first o f these are the synchronisation hits. Out of the 80 bits of a V.I 10 frame. 17 are used for synchronisation. (;NM radio transmission is based on a complex synchronisationscheme. and Were is no difficulty i n deriving the V I I I ) frame hottudaries from ilie GSM synchronisation (thanks i n particular t o the choice o r 21) ins as a fundamental GSM synchronisation period. \\ hich is a multiple of 5 and ll) ins). In fact. another important aspect 111 synehronisai loll collies from the forward error correcting scheme used over me radio interface. With such schemes, residual errors are grouped into bursts. corresponding to an ill-fated radio coding block. There are some advantages to map precisely the V.110 frames and the coding blocks. This is possible because the coding block recurrence has been chosen to be 20 ms. The rule is then simple: a radio coding block corresponds exactly to 2 or 4 V.110 frames. But 63 bits still remain per V.110 frame. Out of those, three are not transmitted over the radio interface, because they can be reconstructed by the receiver. These are bits El, E2 and E3, which indicate the true end-toend data rate. This does not correspond to new information between the mobile station and the infrastructure, since the rate i s transmitted separately by signalling means for setting-up purposes, and thus can be dispensed with. What remains consists finally of 60 bit frames, which can be seen as a simple subset of the original V. 110 frame. The resulting "intermediate rate" f o r GSM i s then 1 2 kbit/s (derived from the 16 kbit/s) or 6 kbit/s (derived from the 8 kbit/s). In fact, a third and lower rate has been introduced for user bit rates below 2400 bit/s, once again to optimise the redundancy. We have seen that in ISDN, rates below 4800 bit/s are rate-adapted to 4800 bit/s by simple bit repetition. In GSM, this simple repetition is done only up to 2400 bit/s. Because t h e same amount o f auxiliary information i s kept, the intermediate rate corresponding to user data rates of 2400 bit/s or less is then 3.6 kbit/s (2.4 + 1.2). The "V.110-like" frame in this case is not 60 bits long any more, but 36 bits long. The transformation from ISDN frames at 4800 bit/s to GSM frames is done simply by taking every other user bit. The reverse transformation consists in duplicating each user bit. Figure 3.22 shows how the radio interface is introduced in the rate adaptation chain, to be compared to the ISDN case of figure 3.21. The RAO function is performed on the mobile side, as well as the rate adaptation inspired by the ISDN RAI, called RAI'. This includes in the synchronous cases the network independent clocking control as defined in V.110. Asynchronous Sylk:hronous raw rate, eg300 or 9600 bi! s 1 ,- RAO + . data Imo 13 5 Of 1 2 K I M S M T san-pl nq I RAI' TAF Into rme(ha!, data rate 8 or :13 nb!!'• . . - synch . tnl RA2 kbil th. BTS + TRAU Figure 3.22 • Rate adaptation in GSNI Adaptation functions RAO (for asnehronons data only) and part of RAI called R:\ I ) are performed in the TAP (inside the mobile station). whereas the complement of R:\I and IZA2 ;He performed in the BT511.k.Al.. 3.3.3.4. The GSM NT Connections Since in any case the IWF has a lot to do for an NT connection. there is no reason why GSM N T connections need to strictly follow ISDN specifications as in the T case. The needs are basically to transport 'a flow o f 240 bit frames between the TA F and the ! W E using a maximum total rate of 12 kbit/s. The adaptation to ISDN is done, if need be, at the ends of the connection (TAF and IWF). On the infrastructure side the R A I '/RA1 function performs the translation between- the radio interface format and the ISDN format, and an RA2 function completes the ISDN adaptation. so that the data flow reaching the IWF is in a full ISDN format. However there are some advantages in using as close as possible transmission methods for different modes, and NT transmission has been designed to have a common core with the 9.6 OWN T connections. Indeed, the data rate effectively carried between the TAF and the IWF for 9600 bit/s T connections is also 12 kbit/s. as explained above. In addition. the distinction between user data and auxiliary information is irrelevaiii for the RAI' function or for the transmission over the radio interface. Lei us see how these common aspects are exploited. The difference between a V.110 frame and a radio rate adaptation frame is simple, and the translation between the two is easy. It is just a matter o f adding (respectively removing) the synchronisation bits. synchronising t h e V. 11 0 frames w i t h coding blocks: a n d adding (respectively removing) bits E l , E2, E3, whose contents are known thanks to signalling inside the GSM infrastructure. Let us first start with a difference requiring some explanation. A simple solution for the transmission of the auxiliary information on NT connections would have been to do the same as in the T cases. There would have been no major obstacle to this choice, and the result would have been elegant. A different choice was made. For 'I' connections. auxiliary information adds 12 bits for each period of 5 or 10 ms. This TRANSNIISSIO I I would have resulted in respectively 48 or 24 bits per RLP frame. This was felt to be a high overhead, of which at least a third is useless (bits E4 to E7), since network independent clocking signals are not used in NT connections. Moreover the rest o f the bits, which correspond to side signals to and from the terminal, rarely vary. A more complex approach has been chosen to reduce the load incurred by the auxiliary information in most cases. The key fact is that at most, three side signals from the terminal are sampled in each direction (see table 3.6). The idea was to transmit the values of these signals once per RLP frame, plus to indicate the transitions, if any, during the period corresponding to the user bits in the frame. The formatting is such that a minimum of 8 bits is consumed per frame, plus a further 8 bits to indicate a transition. So. if the signals are stable, which is the case when the connection is operational, only one octet per RLP frame is used for auxiliary information instead of 6. If the side signals are often changing. the auxiliary information may use much more than 6 octets per frame, but such cases cannot occur when effective transmission of user data takes place. only one RLP frame. Conversely, the E hits are set in the uplink direction according to the position o f the contents o f the V. I 10 frame in the corresponding radio block. Another important point for NT connections is the need for frame delimitation. As usual, frame delimitation can be easily obtained on the radio path as a side product o f the comprehensive synchronisation arrangement. Then, despite the difference of frame length (RLP frames are four times as long as V.110 frames for the 9.6 kbit/s T connection), it was possible to use exactly the same transmission scheme for the T and NT cases over the radio interface, and so it was done. (Well! This is somewhat a rewriting of history. In reality, the NT mode was contrived by starting from the T case, wondering how the latter could be used to obtain a better transmission quality!). In the asynchronous case, all the functionalities o f the start/stop protocol a r e fulfilled b y t h e N T m o d e functions. T h e frame synchronisation enables hits inside the frame t o he constructed into octets, thus removing the need for start and stop signals. Only the 7 or 8 bits of user information are transported (I fill bit is added in the case of 7-bit characters. so that in all cases there are 8 hits per character in the frame). At the receiver's end, the start and stop signals can he reinserted at their correct position. Note that the duration of the stop signal is lost. and as a consequence the relative timing of the characters is lost. but their order is kept. This just Takes asynchronous transmission even more asynchronous, and no application is known which suffers from this. Flow control is fulfilled i n the start/stop protocol either 1w using special characters, or by toggling modem control signals. The RI.I' provides its own flow control, in particular because it needs to regulate the flow when for instance too many repetitions are needed at one moment. which can happen in the case of had luck. So the start/stop flow control protocol can be relayed by the similar functionality of M.P. once the t p e o f Ilan Unfortunately the radio path is not the only segment in the way, and the RLP frame delimitation must also he transported between the BTS and the IWF. The V.110 frame delimitation is available, but is not sufficient, the RLP frames being larger. And there like a devil out of its box we find the three E bits: they are free for any use in the case of an NT connection. They are used to convey the frame delimitation. For T connections, the three E bits are dealt with in the RAI '/RA I translation. N T connections u s e a slightly modified R A I '/RA I translation, which manages the correspondence between the RLP frame delimitation over the radio interface and the one given by the V.I10 frame delimitation and the E hits. In the NT case, these bits are used to indicate the position of a V. I 10 frame in a group of four constituting an RLP block. I n t h e downlink direction. the R A I '/RA I uses this information to put the four frames of same RLP block in a same radio interface block. The size of the RLP block has been chosen according to the radio block size, such that errors affecting one coding block affect The next point to look at is the protocol conversion. It has already been mentioned that the R19 replaces the start/stop protocol or the packet protocol used by the terminal. The RLP between the TA F and the IWF provides the same functionality as the original protocol. but adapted to the GSM transmission. The conversion is done by relay functions in the TAF and the IWF. These functions depend o f course on the terminal protocols. There are three cases: the start/stop protocol. LAPB (X.25-21 and the protocol used for fax. The fax protocol is in essence the same as X.25-2 but with additional signalling so that transmission is basically identical and the three cases are effectively only two. The Spergicatimi.s. distinguish two relay protocols, namely the L2R-COP (Layer 2 Relay Character Oriented Protocol) and the L2R-BOP (Layer 2 relay B i t Oriented Protocol). control i n u s e i s h i m \ n. A l a s t i n t e r e s t i n g F u n c t i o n o f t h e s t a r t / s t o p protocol i s the 'break' signal. which is basically a violation o f the start/stop rules (a break signal is a stall signal longer than a character. i.e.. such that the stop signal arrives too late). The break signal i s used basically as a reset mechanism at the disposal o f the user. to he used when things are going strangely. A special method is provided in the relay protocol to convey an indication of the reception of a break signal. which is regenerated by the receiver. 4.) A last point for this presentation of the start/stop relay protocol concerns the efficiency of the method. If we take the worst case, which is a 9600 bit/s asynchronous connection with 7-bit characters and 1-bit long start and stop signals, we have a maximum throughput of 9600/9=1067 characters per second, that is 21 and one third characters per RLP frame period. A frame contains as a whole 240 bits, 40 of which being used for error detection, for frame numbering and for the acknowledgement and flow control protocol. A minimum of 8 is used for the modem control signals. There remains 192 bits, that i s t o say 2 4 octets (which corresponds to a data rate of 9.6 kbit/s exactly). A benefit of 2 2/3 octets is then obtained, allowing on average, one frame out of 8 to be a repeated frame. For user rates of 4800 bit/s or less, things are obviously much better. The relay method for protocols which deal with frames, such as HDLC, is rather different. Because the RLP replaces the protocol, the digital stream coming from the external world can be stripped from the overhead introduced by the replaced protocol. In the case of HDLC, this corresponds to the link layer header (2 octets per HDLC frame), the synchronisation overhead (a minimum of 1 octet) and the error detection overhead (2 octets). The remaining data consists o f chunks of variable length, which in general do not fit into the fixed length RLP frames. Frame delimitation is then needed, and this is done using a special status octet after the last octet of the content of an HDLC frame. The final gain is then 4 octets per HDLC frame. The relative gain depends on the HDLC frame length. The frame rate in RLP frames being exactly equal to the original rate, the breathing space obtained for repetition is never null, but is greater when the HDLC frames are smaller. 3.3.3.5. The BTS-TRAU Interface We have now presented the complete connection for data. Or so it seems. A last complication has to be studied, which is the transmission at 16 kbit/s. We have seen that in order for the operator to make economies on the cost o f internal links, a scheme has been contrived to transport speech on 16 kbit/s links. The same thing had to be done for data. At first view, it seems that a modified RA2 function would do. But this would not take into account the constraint that only in-band information can be used by the BTS to control the TRAU. The problem is then to distinguish speech from data with i n -band information, and there i s no room available in V.110 frames for this. TRAU 64 kbit,s MSC VLR RA2 synch _ eu other info - - BTSI • Intermediate bit rate (3.6, 6 or12 kbit/s) T R A U **Intermediate bit rate (8 or 16 kbit 5) Figure 3.23 —The split between BIS and TRAU kw rate adaptation The transmission of data between BTS and 'TRAU takes place at a rate of In kbit/s. but in-band control of the TRAU imposes all :W16011111 specific multiplexing stage. What has been chosen in order to solve this issue is to specify specific rate adaptation modes o n the BTS-TRAIT interface. rather — different from V.110, but compatible with the BTS-TRAU interface as specified for speech. This is schematically represented in figure 3.23. The bit stream is structured in 320 bit blocks. recurring each 20 ins. The synchronisation a n d data/speech discrimination h a s already been described for the speech case (see page 163). A notable difference is that the time alignment control. introduced for speech to prevent an additional transmission delay due to lack of synchronism with the radio path. is not used for data. To simplify the TRAU. the 63 bits resulting from the stripping of the 17 synchronisation bits from a V.110 frame are kept unmodified. Hence bits El. E2 and E3 are sent 2 or 4 times in the blocks ( whether the intermediate rate is S or 16 kbit/s1. Thanks to this approach. the TRAU does not need to be told whether the connection type is T or NT. nor has it to bother with RLP frame delimitation. I K . \ . 1!%11717,1t I N Role of the bits A umber of hits per frame (uplink = downlink) Frame synchronisation 35 Discrimination between speech and data, full rate and half rate 5 Intermediate rate on the TRAU/IWF interface 18 kbit/s or 16 kbit/s) I Data block 120 or 240 El, E2, and E3 bits (meaning as indicated in the previous sections) 2 times 3 or 4 times 3 Fill bits 1-I4 or 18 Spare hits 9 with the PSPDis case. involving communications in an X.25 mode or via a PAD respectively. TS GSM 09.04 deals with the C'SPDN case. The 07 series includes Specifications devoted t o the interface between the Mobile Termination and the Terminals. 'I'S GSM 07.01 is a general introduction t o t h e problem. I t presents t h e different configurations o f a mobile station. lists and describes the Functions needed in the TAF in all cases. and introduces the rest of the Series. TS GSM 07.02 and TS GSM 07.03 deal respectively with the asynchronous and synchronous cases. They contain in particular the description of the relay protocol used in the NT case. Table 3.8 —Contents of a data block on the BTS-TRAU interface TS GSM 03.45 and 03.46 deal with the facsimile services, when T and N T transmissions respectively are used. This i s where a l l the information concerning the fax adaptor can he found. TS GSM 03.43 deals specifically with the videotex service. It does not include important specifications, but is interesting also as presenting an example of service provision from a general standpoint. When the intermediate rate is 8 kbit/s. till bits replace 120 data bits and the missing El. E2 and E3 bits in the 320 bits block. TS GSM 03.10 describes the different types of connection from TAF to IWF. The bulk of it concerns the data services. This results in the block contents as summarised in table 3.8. The structure is the same in both directions (uplink or downlink). It is worth noting that, because the specification indicates that the fill bits (in the 8 kbit/s intermediate rate case) are set to "1" as in V.I10, the distinction between the 8 kbit/s and the 16 kbit/s cases is totally irrelevant for the transcoder, which can perform the same operations in both cases while still respecting the specification. TS G S M 03.50 presents t h e transmission path f o r speech communications. It 4ddresses in particular the echo problem. • T h e n we find the Specifications providing the detailed description of the interfaces specific to GSM. The whole 06 Series in devoted to speech transmission. Its focus point is TS GSM 06.10, which specifies the voice coding algorithm. The other Specifications in the Series deal with Discontinuous Transmission. Vocal Activity Detection and comfort noise. SPECIFICATIONS REFERENCE Rate adaptation for the data connections is specified in Ts u s m 04.21, for the radio interface, and in 'I'S GSM 08.20 for the A and Abis interface when 64 khit/s circuits are used. Transmission over 16 khit/s circuit is described in TS GSM 08.60, both for speech and for data. The topics addressed i n this chapter relate t o many different Specifications. We present them here in the order we think is the best to easily unfold the complexity, in particular for the data services. At a lower leiel the physical aspects of the transmission m a dui A and Abis interface - e dealt with respectively in 'I'S GSM 08.04 and TS GSM 08.54. Transmission over the radio path is the subject of the next chapter. The 09 series include a series o f Specifications devoted to the interworking between GSM and other networks. They are organised by type of network. TS GSM 09.07 concerns the interworking between GSM PLMNs and the PSTN or the ISDN. 'I'S GSM 09.06 and TS GSM 09.05 deal The RLP protocol (not including the relay part specific to the start/stop or LAP protocols) is specified in 'I'S GSM 04.22. 186 THE GSM SYSTEM I Hi 0 1 0 1 \ rtk • I I •"44 7 / . I . I -..... r.,,,, N T s e . : „ E R ' E • f i I f i c * E / 1 . ; / i . A c i 187 4 '41121". ir1j1 lo n - - 7 THE RADIO INTERFACE i . — ; . 4,1.iThe Needs 4 , t Y 1 , 1 J s c i P 4 4 T r a t i s m i s s i o n . s r P • 4 . 1 , 2 . ; s i g 4 3 , 11 4 1 4 4 4 w p i , 4.1 3 ' ,1 1 4 t", I W R I V • * I T : 1 1 , • - : 1 1 07..-;,±! Among all GSM herfaces, the radio interface is probably the most important one, at least i f a measure o f importance is the time spent for its specification. Several reasons explain w h y this interface between the mobile stations and the fixed infrastructure plays a leading part i n the system specification. • 412;.ThSM q ; oceis. Scheine- • • • , „Rli4't o t a 7;41j :;fi, 42r2••••Tiii#Svcy Axis a ' • ‘••• ,Itdtti-r11 ^ , n i i i l . 7 0 1 1 : ' V).- f t * * ;inii ' • ' 4,Farn g q ‘ l i c y 9 y e & . 4 4 •, First o f all, i t needs to be completely specified t o achieve a full compatibility between m o b i l e stations o f various manufacturers a n d networks o f different oberators. This is the key to one o f the main goals --of GSM, that is to say to obtain MS-roaming, for example allowing users to ,et service in different countries using a single mobile station. f f fiks. • •, ' .4,3,2,,IITItsrltityk.1tg and C k t n . p 4 c p a i n g l e nt, t ely.244.;:s.4:?.:misiftiatityth& ! : ;„; • Specifications' Reference • i;t• • . . 4 1 ! ' e.t). : '25 Second, t h e spectral efficiency o f a cellular system i s a k e y economic factor. and is determined entirely by the transmission over the radio interface. The economic importance o f spectral efficiency i s best understood when it is defined as the number o f cells needed to cover a given area and a given traffic with a given amount of radio spectrum. The better the efficiency the lower the number o f cells. Spectral . c . e.il .encv depends on the number o f simultaneous calls which can, be l i t i n the available spectrum, on the resistance o f the transmission 19 interference. which determines the frequency reuse factor, and o f a clever exploitation of a number of interference reduction techniques. such as emission power control, detection of silent periods w i l l + ) speech and optimised ha ndover decision methods. The radio interface o f GSM is an excellent example of the search for maximum spectral efficiency. 188 T H E GSM SYSTEM Two main aspects of the transmission over the GSM radio interface will be studied, the multiple access scheme, and the signal processing,._ from bits to radio waves. The multiple access scheme describes the way the radio spectrum i s shared between several simultaneous communications, occurring between different mobile stations, in different cells. In GSM, after epic discussions, the choice settled on a mixed Frequency Division and Time Division multiple access (FDMA and TDMA), spiced with frequency hopping. FDMA, which is used in particular to share the spectrum between neighbouring cells, is of a medium bandwidth type (200 kHz). The basic TDMA factor, i.e., the number o f calls obtained from a 200 kHz frequency band, is 8 for data at 9600 bit/s and for speech in a first phase, and will be 16 for speech in the near future. The TDMA scheme used in GSM is an interesting feature of the system, being based on complex cycles. The final ingredient of multiple access, frequency hopping, which consists in regularly changing the frequency of transmission for a given mobile station to base station connection, has been introduced for improving quality at low speeds as well as spectrum efficiency. Signal processing in GSM also presents interesting features. The digital modulation of the 900 MHz or 1800 MHz radio waves is quite classical, being Gaussian Minimum Shift Keying (GMSK). A less classical point is the emission and reception on a burst basis, with equalisation adapted burst per burst. GSM also makes use of various digital error correction codes with interleaving, with the intent to provide an excellent compromise between the spectrum usage of a channel and resistance to interference. This part of the presentation will also be the occasion to have a glance at ciphering. But before dealing in details with all these technical aspects, we will first consider the needs which were to be fulfilled, and which indeed explain the various design choices. The previous chapter has already identified the different kinds of user data flow that the system is able to transport. Other transmission needs, related to the exchanges between the mobile stations and the infrastructure for control purposes, have yet to be introduced. 4.1. THE NEEDS There are several reasons why data is transmitted to and from mobile stations. The purpose of this section is to identify them, but in doing so we will do a bit more. To make easily the link between tho •11 R A m o INTERFAA•E 180 needs and the transmission scheme, we will introduce at this stage the concept of channel. Different channels are used to transmit different simultaneous transmission flows. The methodology used hereafter is to present each identified need i n turn, introducing the corresponding channel types used in GSM to fulfil the need. 4.1.1. USER DATA TRANSMISSION The very first aim of a communication system is to transport user information. The first service to he offered to GSM users is speech. The radio interface must therefore accommodate h i -directional speech transmission. In order to limit the use of the radio spectrum, speech is represented by a binary signal at a rate of 13 kbit/s. through a coding scheme specific to the system and described in the previous chapter. Data services also require bi-directional transmission means. GSM only accommodates fairly small data rates: the raw data to be transmitted is a binary flow at a rate of 12, 6 or 3.6 kbit/s depending on the given service. These rates allow data services which are similar to those offered by a PSTN through modems using respectively 9.6. 4.81 and 2.4 khit/s rates. Most of the user services offered by GSM rely on one 01' these four transmission modes (speech and 3 data modes). One exception, however. consists in the transfer of short messages. which is implemented on the radio path by methods derived from those used to transfer signalling. Point-to-point short messages use means identical t o signalling transmission, and will therefore be described together with the means which support signalling, in Chapter 5. Cell Broadcast:short messages make use of a special channel, the CBCH (Cell Broadcast Channel), which will be dealt with in a next section. For all other services than short messages. a user engaged in a communication must have at his disposal'a portion of the radio interface uniquely devoted to the call for its duration. This corresponds to a special kind of channels: the TCHs (Traffic Channels). From the multiple access point of view, two kinds of TCHs are defined: • t h e T C H / F (where I ' stands f o r " f u l l rate") allows the transmission of 13 OMR speech or of data at I 2. 6 or 3.6 khit/s;, • t h e TCH/H (where 11 stands for "half rate") allows speech coded a t a rate anmnd 7 kbitis t o h e transported (the specification of this coding is currently under %%ayI. or data al a rate of 6 or 3.6 kbit/s. 190 T H E . GSM SYSTEM 4.1.2. SIGNALLING User data is not the only flow o f information t o be transported during a call. Signalling messages must also be conveyed; they allow the mobile station and the network to discuss the management o f several issues either related to the user (e.g., call in progress indications) or concerning technical aspects of the communication (e.g., preparation and execution o f handover). The establishment and release o f a call also require signalling exchanges. I n addition, w e w i l l see that some information exchanges between mobile stations and the infrastructure are needed even when no communication is in progress. 4.1.2.1. Signalling in Connection with a Call In order t o transport signalling data i n parallel w i t h the transmission o f a user data flow, GSM offers two possibilities. Each traffic channel comes with an associated low rate channel, used for the transport of signalling: the SACCH (Slow Associated Control CHannel). This bi-directional channel may carry about 2 messages per second in each direction, with a transmission delay of about half a second. It is used for n o n -urgent procedures, mainly t h e transmission o f the radio measurement data needed for the decisions concerning handover. The other needs for associated signalling, such as the messages to indicate the call establishment progress, or to authenticate the subscriber, or to command a handover, to give only a few examples, make use of the TCH itself. This usage shall be referred t o here as fast associated signalling, i n place o f the term FACCH (Fast Associated Control CHannel) used in the Specifications. Fast associated signalling does not indeed refer to a channel in our use of the term, but to a particular use of a TCH. The receiver is able to distinguish both uses by reading a binary information transmitted on the TCH and called the stealing flag (see page 237). During the initialisation and release phases, n o user data is transmitted and therefore signalling may use the channel without any conflict w i t h other types o f data. However, during the call, the transmission of fast associated signalling is done to the prejudice of user data: there is a loss o f user data, as i f transmission errors (of known location) had occurred. Hence the term -stealing". THE RADIO N i l : R I : A C E 191 4.1.2.2. Signalling outside a Call In some cases. there is a need to establish a connection between a mobile station and the network solely for signalling matters. be it at the user's demand (e.g.. call forwarding management, transmission of short messages) or for other management needs. such as location updating. A TCH/F or a TCH/F1 may he used f o r that purpose. This is however wasting a lot o f spectrum. since few messages are t o he exchanged (typically hull a dozen each way) and the channel would only be used for a very small portion of the time. In order t o increase system efficiency, a n additional type o f channel has been introduced. Its rate is very low and the only specified usage—so f a r —for this channel i s signalling (and short messages transmission). This channel shall be referred to here as TCH/S, where "8°' stands for "eighth". I f a TCH/H may be considered as half a TCH/F. then this small channel is one eighth of a TCH/F. This terminology diverges from the GSM one. where this channel is called SDCCH (Stand-alone Dedicated Control CHannel. n o t a very enlightening term). T h e characteristics o f such a channel are in fact very close to those o f a TCH/F or 14, the prevalent difference being its rate. The choice o f vocabulary in the Specifications stems from the fact that no user data mode has been specified for the TCH/8. However. this does not appear the most pertinent distinction. given the definition of a channel (see page 193)! Moreover, nothing would fundamentally preclude an evolution of the Specifications to use a TCH/8 as a oral channel, for instance for low rate data services like telemetry. A pragmatic reason for this change of terminology is to avoid in many places "TCH or SDCCH". As the other TCHs. a TCH/8 is hound to an SACCH. The notion of fast associated signalling also applies to a TCH/8, but for the moment it is a trivial matter, since it is the only usage of a TCH/8. 4.1.3. IDLE MODE The rarity of the radio spectrum does not allow each user of the — system to have his own TCH at all times. Traffic channels are therefore allocated to the users only when the need arises. This leads us to the basic distinction o f dedicated mode and idle mock. which is an essential concept in radiotelephony. 192 THE GSM SYS 1E51 Formally, a mobile station shall be considered in dedicated mode when a TCH is at its disposal. This corresponds to the phases when full bi-directional point-to-point transmission is possible between the mobile station and the infrastructure, for instance during a call established for the user, or to perform a location updating. TCHs and SACCHs are therefore referred to as dedicated channels in the Specifications. When a mobile station is active (powered on) without being in dedicated mode, it is deemed to be in idle mode. However, the mobile station is far from being idle... It must continuously stay in contact with a base station, listen to what this base transmits in order to intercept paging messages (to know i f its user is being called), and monitor the radio environment in order to evaluate its quality and choose the most suitable base station. In addition, there is one telecommunications service which is provided to the mobile stations in idle mode: the Cell Broadcast Short Message service. The transition between idle mode and dedicated mode requires some information exchanges between the mobile station and the base station (the "access" procedure). The mobile station indicates to the network that it needs a connection and the network indicates in return which dedicated channel it may use. All of these uses require specific transmission means which are grouped under the terminology "common channels". 4.1.3.1. Access Support In order to communicate with a base station, a mobile station must first become (and stay) synchronised with it. Two channels are broadcast by each base station to this avail: the FCCH (Frequency Correction CHannel) and the SCH (Synchronisation CHannel). Mobile stations in idle mode require a fair amount of information to act efficiently. Most often, a mobile station can receive, and potentially be received by, several cells, possibly ill different networks or even in different countries. I t has then t o choose one o f them, and some information is required for the choice, for instance the network to which each cell belongs. This information, as well as much other, is broadcast regularly in each cell, to be listened to by all the mobile stations in idle Mode. The channel for this purpose is the BCCH (Broadcast Control CHannel). I lilt I ) I ( ) IN I . \ ( ' I . 1 9 3 Let us now come to the access. We have seen that paging messages have to be broadcast to indicate to sonic mobile stations that a call toward its user is being set up. The access procedure itself includes a request from the mobile station and an answer frpin the base station. allocating a channel. Paging messages and messages indicating the allocated channel upon prime access are transmitted on the PAGCH (Paging and Access Grant CHannel). This terminology combines the t w o terms "PCH" (Paging CHannel) and "AGCH" (Access Grant CHannen used in the Specifications. Referring to two channels in that case is not consistent with the fact that the partition between PCH and AGCH varies in time. hence the concatenation in notation. All the common channels listed above (FCCH. SCH. BCCH. PAGCH) are "downlink" unidirectional channels. i.e.. they convey information from the networkito the mobile stations. The last type o f common channel, which allows the mobile stations to transmit their access requests to the network, is an "uplink" unidirectional channel. It is called the RACH (Random Access Cllannel). Its name indicates that mobile stations choose their emission time on this channel in a random manner. This results i n potential collisions between the emission o f several mobile stations, and will be studied in detail later. 4.1.3.2. T h e Cell Broadcast Short Messages Cell Broadcast shim ncssages require the means t o transmit around one KO octet mcssagr every t w o seconds from the network towards the mobile stations in idle mode. This corresponds to half the capacity o f a downlink TCH/K. I n each cell where this service i s supported, a special channel, a CBCII Well Broadcast Cl lannel i s used for broadcasting messages. A CBC11 is derived from a TOWS. Some special constraints exist for the design o f this channel, because of the requirement that i t can be listened t o i n parallel w i t h the BCCH information and the paging messages. 4.1.3.3. A Point of Terminology: What is a Channel? — As noted a few times in the previous sections. we introduced some new terms in this book, and we will avoid some other terms though the are of common use in the Sper/ficarions and in the GSM literature. This I Ilk 12A1)10 I \ I r1:1 W i t requires some justifications. The issue revolt' a r o u n d the notion of channel, which in our opinion is not used consistently in the GSM literature. CCITT officially defines a channel as "an identified portion of an interface". Though it can be debated at length, we interpret this definition in the way described below, w i t h which the terminology i n the Specifications does not always comply. First, a channel i s defined b y its characteristics i n terms o f transmission support, and not by its potential uses. Very often a channel is dedicated to only one usage, hence the mixture of concepts. Second, it should always be possible, before reception, to know the channel to which a given part of a data flow belongs, without having to look a t the contents o f this data f l o w. T h e knowledge o f the characteristics o f the channel (timing, frequency) should enable the recipient of the information to know whether the data may be for him or not. This method is used, e.g., i n the subscriber's loop o f a fixed telephony network (the line itself identifies the subscriber). Another means to identify the destination of data would be to include explicitly the destination's address in the data. This method is the one used, e.g., for postal dispatches, where the address o f the recipient is written on the letter or package, hence the term "addressee", In the latter case, we prefer not to speak of a channel for each address, but of a single channel shared between the different flows. One advantage o f defining a channel independently from the information carried b y the channel i s when describing the allocation of channels, or a configuration of channels. In such cases, what matters is the description of the part of the interface which can be used for a given usage. The GSM terminology mixes the concept of channel with its usage, and also mixes a channel (in the sense defined above) with a data flow identified by an internal address. In our view those approximations are detrimental t o the grasping o f the underlying concepts. Tw o major examples for the application of these principles have been encountered in the previous sections. First, the term PAGCH is used in this book for the type o f channel conveying both the paging f l o w and the channel allocation flow, instead of the GSM terminology PCH (for the "channel' carrying all the paging flow and some of the channel allocation flow) and AGCH ( f o r the complementary "channel" carrying o n l y channel allocations). Second, the term FACCH used in the Specifications will not be used here, since it refers to a specific use of a channel (the TCH) but I ` ) . , not to a pre-det...ed portion of the resource. and we will prefer to refer to fast associated signalling as a mode of use of the channel. The FACT!" ;gives an example of the conceptual difficulties: to which channel does the ,stealing flag belong? We understand t h a t diverging i n terminology f r o m t h e Specifications (or for that matter from the established usage in some field. as our using emission where most often transmission is used) will cause some difficulties. On the other hand. we hope that the new terms help the reading of this book, and that the balance is beneficial for the overall understanding of the GSM concepts. 4.2.' THE MULTIPLE ACCESS SCHEME Many techniques have been devised to create channels out o f a communication medium, in order to use them as separate links. Using the standard terminology, the radio interlace of GSM uses a combination of Frequency Division Multiple Access t FDMA) and Ti m e Division Multiple Access (TDMA). with a pinch o f Frequency Hopping f o r flavour. But the resulting mixture is something different than just the juxtaposition of the ingredients. A basic concept of the CiSM transmission on the radio path is that the unit of transmission is a series of about a hundred modulated hits. and is called a burst. Bursts have a finite duration, and occupy a [Mite part of the radio spectrum'. The a r e sent in time and frequency ss Wows. that we will call slots. Precisely. the central frequencies o f the slots are positioned every 200 kHz within the s) stem frequency hand I FDMA aspect). and they recur in time every 0.577 ins- - o r more exact's'. ever\ 15/26 ms (TDMA aspect). All slot time limits are simultaneous in a given cell. A time interval corresponding to simultaiwous slots will he referred to as a time slot. and its duration will be used as a lime unit. designated BP (Burst Period). In practice. there are specili ....eal.on and implementation tolerances. and some energy is sent outside the finite spectrum m. indtm defining the slot Tin. RAD101M-El(FACE THEGSM SYSTEM 196 1 9 7 in time. A channel has therefore a temporal definition giving, for each lime slot, the number of slots which are part of the channel. For the time being—and probably for quite a while...—a channel may only consist of 0 or I time/frequency slot per time slot.. ,The temporal definition o f a channel is cyclic, that is to say it repeats itself after sonic time. The simplest case of a cyclic definition is "I burst every n". As will he seen on the next pages, the GSM channels are rarely as. simple. frequency 200 kHz BP > 15/26 ms time slot Figure 4.1—A slot in the time and frequency domain Atransmission quantum in GSM (a burst) fits in a time and frequency window called a slot, which lasts about 577 ps and occupies a bandwidth of 200 kHz. A slot may therefore be pictured in a time/frequency diagram as a small rectangle 15/26 ms long and 200 kHz wide (see figure 4.1). Similarly, we will call a frequency slot (a radio frequency channel in the Specifications) a 200 kHz bandwidth as specified for GSM. The above paragraph is heavy with terms and definitions. Once again, there is some difference between the choice of terms here and the one of the Specifications, because the terms such as timeslot or burst are used with several different meanings in the Specifications. For instance, burst refers sometimes t o the unit time-frequency "rectangle", and sometimes to its contents. Likewise, timeslot is used to mean either what we call here "slot", or its time value, or also the cycle using one slot every eight slots in time. To use a given channel means to transmit bursts at specific instants in time, and at specific frequencies, that is to say in specific slots. Tu define a channel consists then to specify which slots can be used by, or are part of, the channel. Usually the slots of a channel are not contiguous In parallel to the time definition, the frequency definition o f a channel gives the frequency o f every slot belonging to a channel. I t rconsists basically of a function allocating a frequency to each time slot •where a channel has a slot, There exists fixed frequency channels, for which the frequency is the same for every slot, and frequency hopping channels, whose slots may use different frequencies. For bi-directional channels (e.g., a ICH), the two directions could have been defined in different ways. Howeverjor simplicity reasons, the channel definitions for the two directions are always related in a very elementary manner: a fixed frequency gap (the "duplex separation") o f 45 MHz (in the 900 MHz band) and of 75 MHz (in the 1.8 GHz band) and a time shift, which depends on the Channel type. separate two corresponding slots o f a given channel. In more academic terms, one direction is related to the other by a translation in time and in frequency. 4.2.1. THE TIME Axis The organisation o f a channel along the time axis can be quite complex. This organisation is always,cyclic. but the length of the cycle as well as the number o f slots in a cycle vary according to the type o f channel. The positioning of the cycles in time is achieved through system synchronisation. Each cell provides a reference clock, defining the time slots, but also a -dating- scheme lo which the cycles of all the channels are referred. In GSM. each time slot tand hence all the corresponding slots on the different frequencies) is given a number, which is known both by the base station and the mobile station, and which is part of the synchronisation information. The description o f a given channel (e.g., sent by the base •station to the mobile station) refers to this numbering scheme. The time slot numbering is cyclic, but of a very long cycle (3.5 198 T H E .1.11F RAD10 iNtFRFAuF GSM SYSTEM hours), which has been chosen as a multiple o f all cycles needed for multiple access. Within each period, any slot can be unambiguously referred to by its time slot number and its frequency slot number. 7 rmt' 0 8 BP cycle I 99 V time 1 4.2.1.1. Dedicated Channels 7 r f l ' T ' T T T 'T T T T T T TCH/F and its SACCH 7 7 . ; T, T I T. T 7 T 7 T T A TCH/F is always allocated together with its associated slow-rate channel (SACCH). The resulting group of channels bears unfortunately no specific name in the Specifications; and is therefore often confused with the TCH/F part only. Because we found the distinction useful, and to avoid long paraphrases, we choose to introduce a non GSM term. The group TCH/F plus SACCH shall be referred t o i n this book as a TACH/F. A TACH/F has a simple cyclic definition. It consists of one slot every 8 BP in each direction, i.e., one slot every 4.615 ms (or, better, 60/13 ms). A consequence of this definition is that it consists in the slots whose time slot number is 8 times an integer plus a value k between 0 and 7 specific to the channel. This value k is the phase modulo 8 of the numbers of the slots of the channel. It is called in the Specifications the Time slot Number (TN) of the channel. Let us point out that this meaning of Time slot Number is not the same as a few lines above, but TN is the GSM term and is the abbreviation of Time slot Number. In the following. we will use TN only in short form, to limit the confusion. 120 ms (26) • I t . . 7 . 7 T T T T T T 7 T 7 T S Z E T 2 2 5 2 : 7 0 ri4 7 " , r i l t : T. T I T, i 1 B P . BP ( 8 ) 1 5 / 2 6 ms Figure 4.2 — Choice of value of a burst period The exact value of a burst period is derived from a multiple of 20 ms 0 1 2 3 4 5 6 7 8 T 1 T T1sass 1 11 T 71.1471 111 T S 9 1 0 11 1 2 13 1 4 1 5 1 6 1 7 1 8 1 9 2 0 21 2 2 2 3 2 4 25 Figure 4.3 —Time organisation of a TAG1/1: A full rate TACH uses a cycle of 26 x S slots, Within each of these cycles. 24 slots are used for the TC'll and one slot for the corresponding SACCI I. In o r d e r t o s p r e a d t h e a r r i v a l o f S A M I m e s s a g e s at t i l e b a s e s t a t i o n . the cycles of two TAC'I Is using successive slots are separated hy V7 13I's (i.e., 12 x R. plus the difference of one sloth Eight different types o f TACH/F can therefore b e defined depending on their phase modulo 8 (their TN). Two TACH/F with the same phase consist of simultaneous slots. On the network side. 8 TACH/F with different TNs may be emitted by a single transmitter: at any given moment, a single TACH/F of the group is to be emitted. This is the core of the concept o f TDM (Time Division Multiplex). This is an important aspect of the system. leading to substantial savings in the base station. A TACH/F contains a TCH/F and an SACCH. The split between these two channels is also specified in the time domain. using a cycle of 26 TACH/F slots, i.e., 26 x 8 successive slots, or a period of 120 ms. The value 120 ms is an exact one: which was chosen as a multiple of 20 ins in order t o obtain some synchronism w i t h fixed networks. ISDN i n particular. We then can explain the reason of the strange -aloe of BR the BP is therefore exactly 120/26 x 8 ms. i.e... 15/26 ins (see figure 4.2). The TACH/F 26 slot cycit includes 24 slots on which TCH/F bursts are sent, I slot on which gn SACCH burst is sent, and ()tiepinl where no transmission takes place (see figpre • Coding follows cycles based on • the grouping o f 4 successive bursts, as will be explained on page 247. For the TCH/F, a cycle contains 6 times 4 bursts. However, for the SACCH. the full cycle. taking into account this grouping 4 by 4. lasts 4 x 26 x 8 = 104 x 8 BI's. i.e.. 480 ms. 200 ' r l i E TI II RADIO I • GSM Svsnim \ 1. The position in time of the full 104 x 8 cycle of a TACH/F is the same for all the TACH/F o f the same TN in a given cell. We will try to describe the relationship between the different TN in simple terms. The position in time of the TCH/F bursts (by opposition to the SACCH and unused slots) follows a 13 x 8 BPs cycle. The beginnings of this cycle are almost simultaneous for the TCH/F of different TNs, in the sense that the first slot of the 13 x 8 cycle of a TCH/F of TN I follows immediately the corresponding slot for a TCH/F of TN 0. Very exactly, they follow each other from 0 to 7. But the start of the SACCH cycle for the different TNs do not happen in the same 8 BP interval. Still, all the TACH/F have the same definition in the time domain, except for a translation in time. This translation invariance property is important for the design of a mobile station: it means that the scheduling of the treatment for all TACH/F is the same, the only impact o f the TN is that the scheduling starts at different moments. _ The reason for the shifting of the 104 x 8 BP cycles comes from.._ load considerations in the infrastructure. I f the SACCH had been almost simultaneous, the base station would have received the SACCH messages from all the mobile stations almost simultaneously every 480 ms, resulting in a very uneven load. In order to avoid that drawback, the cycle of a TA C H / F using T N n + I ( f o r n f r o m I t o 7 ) i s shifted (12 x 8) + 1 = 97 BPs from the one of a TACH/F using TN n. This shift can be seen in figure 4.3; since four bursts are necessary to build an SACCH message, the base station will process the SACCH messages corresponding to 8 TACH/F of TN 0..7 at 8 different moments evenly spread in time. Relationship between Uplink and Downlink As seen from the base station point of view, the organisation in the uplink direction is derived from the downlink one by a delay of 3 BPs. This delay o f 3 BPs i s a constant throughout GSM. I n fact, the convention is that the numbering of the uplink slots is derived from that of the downlink ones by a shift of 3 BPs: this choice allows the slots of one channel to bear the same TN in both directions. But the point o f view o f the mobile station i s affected by considerations about propagation delays which, even at the speed of light. are not negligible compared to the burst duration (the round-trip delay between an MS and a BTS 30 km apart is 200 ps). In a first step, we will ignore the problems due to propagation delays. and consider that a mobile station very close to the BTS sees the 8 BP cycle as shown in figure 4.4. 8 BPs = 60113 ms Transmission Reception Reception H T 23 56 7 Transmission T T 1-16F2 3 5 671 T Figure 4.4 - Cycle of 8 BPs seen by a 'nubile station close to the base Emission by a mobile station happens 3 BPs later than Reception. Typically, a mobile station will receive during one time slot. then shill in frequenc> by 45 Nll It to emit some time later (3 burst periods minus the correction in lime for propagation). then possibly shift again to monitor Other channels, and 'time to the adequate receive frequency to start the cycle all over again. Such a choice allows mobile stations t o avoid emitting and receiving simultaneously, thereby promoting easier implementations: the receiver in the mobile station need not he protected from the emitter of the same mobile station. When the mobile station is far front the I3TS, propagation delays cannot be neglected, and an exact 3 I3P shift canobt be maintained both at the MS and at the BTS. But it is imperative that the bursts received at the BTS fit correctly into the time slots. and as we will see this is not very roomy. Otherwise. the bursts from mobile stations using adjacent time slots could overlap. resulting in a poor transmission quality or even in a loss o f communication. The only solution i s that the mobile station advances its emission relatively to its reception by a time compensating the to and fro propagation delay. This value is called the liming advance. The exact shift between downlink and uplink as seen by the mobile station is then 3 BP minus the tinting advance. The timing advance value can be computed only by the BTS. and is then provided to the mobile station through signalling. This is why we have chosen to deal completek with this subject in Chapter 6. 202 T H E Hit: RA010 r • m i n . w i : GSM SYSTEM 2 0 3 TCH/H The half-rate TCH, or TCH/H, has led an eventful life. Until the beginning of 1991, a TCH/H was fully specified f o r data services, but not yet f o r speech. The following description w i l l refer to that specification. However, as explained in the previous chapter, "half-rate" speech coding studies (i.e., studies aiming at defining a speech coding scheme adapted to transmission on TCH/Hs), which were under way in early 1991, have shown that the time structure chosen for TCH/Hs was not necessarily optimum. It was therefore decided that the half-rate TCH was no longer part of the phase I Specifications, even for data services. • As for a TCH/F, a TCH/H is always allocated together with its SACCH, and this group of channels will be referred to as TACH/H. A TACH/H is defined in the time domain as consisting on average of one slot every 16 BPs, hence the "half-rate" compared to a TACH/F. The term "on average" is important here, because it is not true that all the TACH/Hs are defined as exactly each sixteenth slot, as the TACH/F is defined as exactly each eighth slot. It is true for half of the TACH/Hs, but not for the other, for whom the cycle is 13 x 16 BP long. The cycle is shown in figure 4.5. It should be noted however that in both cases, a TACH/H consists only of slots of the same TN. This state of affairs is unfortunate f o r t h e simplicity o f t h e description, a n d f o r t h e mobile station implementation (the two categories of TACH/H differ by more than a translation). It has no clear reason, and will quite likely be modified before half rate channel specifications are completed. As far as the time definition is concerned, there are 16 kinds of TACH/H, which are usually referenced by their TN (8 values) plus a sub channel number (sub-11k1, 0 or l). For a TACH/H of sub-TN 0, the time slots of the TCH/H are of even number modulo 16, 7 .1, i r P O i Jill i l i i Reception T r a n s m i s s i o n p Figure 4.6 - Cycle of 16 BPs seen by a nubile station S A mobile station using a TACH/H performs the same operations as fur a TACH/F (same timing between reception and emission). but does so only every other group of 8 bursts. whereas they are of odd number in the case ()I' sub- I N I . Now TACII/II of TN 0. 2. 4, h are well-behaved, in the sense they use exactly each sixteenth slot. whereas the others have the irregular cycle. The crafting of the TACH/H was done so that the two TACI o f the same TN but of different sub-TN have 110 simultaneous slot. and then may he grouped to form the equivalent of a TACH/F. Such TACH/H pairs and/or TACH/Fs may then be grouped in 8s for a single transmitter in a base station. .A transmitter may accommodate up to 16 TACH/Hs, or more generally a combination o f t sets, each set containing either a TACH/F or a pair of TACH/Hs. The split between a TCH/II and its S A C O ! i s done f o r the well behaving TACH/H along a cycle of 13 T.-WIWI slots. i.e.. 121) ills as for the TACII/F. The cycle contains 12 slots for the TCH/H bursts and one for the SAC('I I. A l l the slots o f the TACH/H arc then used for transmission. For the other category or TA(11/11. the cycle is less regular, us shown in figure 4.5 tor INs I or 7. 1 time PM - I a 8 BP cycle 2 x 8 BPs = 120/13 ms 4 l i i l a T I i i i '' \'' \ :TkalTt ,:j-r)7 Ilt'.1 iL ;T 0 1 2 3 4 5 6 7 8 9 10111213141516171819202122232425 T S T C H / H and associated SACCH for sub-TN 0 0 T C H / H and associated SACCH for sub-TN 1 Figure 4.5 - Time organisation of TACH/Hs A TACH/H, as defined provisionally, uses one slot every 16 in average. On "even" TNs, they use exactly one slot every 16. On "odd" TNs, the scheduling is not so regular, as can be seen on TN I or 7 (occurrences I I, 12, 13 for the channel of sub-TN I, and for occurrences 24. 25, 0 for the channel of sub-TN 0). As for the full-rate channel. a complete cycle hist'. 410 111% t I0-1 X S BPs) when coding is taken into account, and the start ()I' es des is defined in the same way. The reader may have noticed that the SACCHs hake the same time organisation as those associated with TCH/Fs. Hall of them (those with sub-TN (1) are emitted exacily_as they were associated with a T('11/1: of the same TN. the other half I those w ith soli-TN being emitted during the slot which would be free in the cycle of a I / 1 ' of the same TN. The timing offsets between TACT I/1 Is o r different T N s is similar to the one defined for TACH/Fs: a TACH/11 ()ITN n+ I is shifted 97 BPs compared to the TACI I/I I of TN n and o f different sub-TN. so that i t is shined twice 97 BPs compared to a TACH/H using TN n and the same sub-TN. The uplink and downlink directions are also related in the same manner as those o f full-rate channels: an uplink slot follows a downlink slot 3 BPs later (at the base station). As seen by the mobile station. the l b BPc!iele of a well-behaved TACT 1/11 can be pictured as shown in figure 4.6. ZU+ TCHI8 THE GSM SYSTEM The description of a TACH/8 (i.e.. a TCH/8 and its SACCH) si s o m e w h a t m o r e c o m p l e x than the o n e of the full. and e v e n of the half rate TACHs, because there exists many different kinds of TACH/8s from the point of view of the time organisation: • some may be grouped by 8 to form the equivalent of a TACH/F: they are called SDCCH/8 in the GSM terminology; • others may be grouped by 4 and combined with common channels to form all together the equivalent of a TACH/F: they are c a l l e d SDCC H/4 in the GSM terminology. All TACH/8 have many properties in common: they all follow a cycle of 102 × 8 BPs, where 8 slots are used for the TCH/8 bursts (a group of 4 slots separated by 8 BPs, then 51 × 8 BPs, then again a group of 4 slots separated by 8 BPs, ...) and 4 slots are used for the SACCH bursts (one group of 4 slots separated by 8 BPs). The attentive reader will have noticed that the length of the TACH/8 cycle (102) bears no simple relationship with the TACH/F cycle (26. and 4 times 26 is 104). The 8 8 origin of this choice lies n i the possibility to associate 4 TACH/8s with common channels; the latter follow a 51 × 8 BPs cycle, as will be explained in a few paragraphs. In order to keep some homogeneity between the different TACH/8s. this cycle has been used also for the TACH/8s crafted to be grouped by 8, though a cycle of 52 × 8 could have as easily been chosen. This difference between the length of the TACH/8 cycle and the one of the TACH/F or /H results in slightly different rates (2% difference) for the corresponding SACCHs. The TACH/8 vary in their phase relations between the TCH slots and the SACCH ones, as well as between the uplink and downlink directions. Figure 4.7 shows the time organisation for both categories of TACH/8. i the case of grouping by 4), but the notion of measurement and 2 n From the figures it is obvious that the TACH/8 cannot be derived one from each other by a simple translation in time. The result is that there are 12 different schedulings for mobile stations in connection on a TACH/8. In fact the figure shows 4 cases (2 in the case of grouping by 8. reporting period (dealt with in Chapter 6) results in 12 classes (the same notion does not have similar impacts on the TACH/F and the TACH/H) ' TACHI8 nO T Si grouping TACHnA S TA C H N O T S Figure 47. . . 'Timé I 101 99 83 57 59 55 The uplink eveles can be obtaine The picture show and I b o c k of During this evele. 2 block 24 tiSM SiSTLN1 4.2.1.2. C o m m o n Channels SCH slot follows each FCCH slot 8 BPs later. Each of those two channels uses 5 slots in each 51 x 8 BPs cycle, as shown in figure 4.8. General Organisation All common channels have bein defined with the intention of grouping them together in few combinations. Their time definitions are therefore all based or' the same cycle, i.e., 51 x 8 BPs. This cycle and the cycle o f traffic channels were deliberately chosen with different values (in fact, they were chosen not to have any common divider) in order to allow mobile stations in dedicated mode to listen to the synchronisation channel (SCH) and frequency correction channel (FCCH) o f surrounding base stations, both o f which carry the information needed for mobile stations to become and stay synchronised with a cell. With the numerical relationship between the cycle of the common channels and the cycle of the TACH/F or H, the bursts of the common channels file o ff one after the other in front o f the reception windows o f the mobile stations situated in surrounding cells, and in particular in front of the large window left open in a TACH/F cycle by the unused slot. These mobile stations are then able to receive—at least from time to time—a burst belonging to the FCCH or to the SCH. By so doing, they acquire the synchronisation information they need on surrounding cells, whatever t h e relation between t i m e bases o f neighbouring base stations (which may b e anything indeed). This functionality i s referred t o as "pre-synchronisation" and i s part o f handover preparation: the reader w i l l f i n d more o n that topic i n Chapter 6. FCCH and SCH Both the FCCH and the SCH have the same time structure: one 3 8 BP cycle 0 1 1 0 1 3 1 2 1 BCCH and PAGCH Apart from the SCH and FCCH. the other downlink common channels include the BCCH and the PAGCH, introduced on pages 192 and 193. The difference between these two channels lies more in their usage than in their transmission characteristics. It is indeed possible that, in later phases of the system. their respective size may be allowed to vary. e.g., in order for the BCCH to gain sonic capacity at the expense of the PAGCH capacity. Two kinds of PAGCH are defined, which have different capacities. hence which use a different number of slots per cycle. They do not hear a specific name in the Specifinniems: they will be called here PAGCH/F ("full") for the larger one and PAGCH/1' ("third") for the smaller one, by analogy with the dedicated channels terminology. A BCCH together with a PAGCH/F uses 40 slots per 51 x 8 BP cycle, all with the same TN. Figure 4.9 shows how these slots are spread. These 40 slots are built as 10 groups of 4. the four slots o f one group 7 111 SCH 4 3 A single set (FCCH + SCH) is broadcast in any given cell. In all cells, the slots of these channels have the same position within the 8 BP cycle, that is to say the same TN. This position is by thfinirion called TN O. Indeed no burst on the radio interface carries its TN, and the mobile station knows the TN of a slot only by reference to the FCCH and SCH. Every burst of the SCH indicates the remaining part of the time slot numbering (i.e., the remainder modulo 8). thereby enabling mobile stations to derive the numbering of all slots within the cell. FCCH time 0 1 207 M : A k A D M I N T I E R F A C E 1 0. 8 BP cycle HU 0 4 H U H U 11111111 11111111 11111111 1 2 6 1 11 B C C H time 1 1 1 PAGCH F 2 32 4 2 Figure 4.8 - time organisation of the FCCH and SCH Figure 4.9 -Title organisation of a BCCH and a PAGCH/I' Oncea mobile station has found anFCCH burst, it knows that the SCH slot canbe found KBPs later on the same frequency. TheBCCH uses 4 slots per 51 x RBP cycle. and 36 slots tire dedicated MlIwPaPilwalulArroc,nmeri,I 208 T H E GSM SYSTEM 8 BP cycle 209 I'llk RADIO rs If.kl. BCCH 8 BP cycle PA G C H / T time R A C H hl ,-,:r-,7-1-,--wriit tirin.-7,--,h-x,-,:-_:::::::::1-..trinii,-4-_,-_27:--•-ini-rrn-v, 11 " 1 " - - 7 ._.......,..,16....L....d..a-l k.i..-;:..'...:;.....—L.:.-...-L-,:....L.,.— • 2 6 1 2 , . 4 ‘ 4 IIIqlt f _11 11111111111111111111111 1 1 45 4 Figure 4.10—Time organisation of a BCCH and a PAGCH/T Figure 4.11 — Time organisation of a RAO I/11 A PAGCH/T uses 12 slots per 51 x 8 BP cycle, i.e., one third of a PAGCH/F. A RACH/H fits in the bursts left tree in the uplink h) 4 'FACI I/8s, i.e., it uses 27 bursts during each 51 x 8 BP cycle. being separated by 8 BPs and containing bursts from a single coding block. The BCCH uses the first quartet and the PAGCH uses the 9 other quartets. A BCCH together with a PAGCHIT uses 16 slots per 51 x 8 BP cycle, all with the same TN. Figure 4.10 shows how these slots are spread. These 16 slots are organised as 4 groups of 4. The BCCH uses the first quartet and the PAGCH uses the 3 other quartets (hence one third of the PAGCH/F). f In order to save spectrum. common channels are al‘vays arranged in groups; there arc 3 such possible combinations. The basic combination includes ( i n the downlink direction) a FCCH, a SCH, a BCCH and a PAGCH/F. all of the same TN. namely 0. The uplink direction contains a RACH/F. This arrangement is shown in figure 4.12. All these channels together use the same amount of resources as a TACH/F, which allows a single base station transmitter to manage such a combination plus 7 TACH/Fs (01TNs I to 71. RACH As for the PAGCH, two types of RACH exist: the RACH/F and the RACH/H. The RACH/F uses one slot every 8 BPs, so that its time organisation is similar to the one of a TACH/F in the uplink direction. On the other hand, a RACH/H only uses 23 slots in the 51 x 8 cycle, and its capacity is therefore slightly more than half the one of a RACH/F. The time organisation of a RACH/H is shown in figure 4.11 and allows the combination of a RACH/H with 4 TACH/8s. () ( 7 8 BP cycle Every cell broadcasts one single FCCH and one single SCH. As far as the BCCH, PAGCH and RACH are concerned, every cell supporting mobile access must have at least one of each (one may however imagine cells accessible only through handovers, in which case these channels would not be necessary). time ilavallirik m a t I I sum Waal ['AGO I I Ii. „ 1 1 . 6 Common Channels Combinations ft 1 0 .12 11 20 2 2 uplink I I RAC 111111111111111111111111111111111111111111111111111 Figure 4.12 — Ba.ie common chaotic] pattern A typical medium cell common channel pattern tines TN Hitt one carrier for [('('I I. SLI I. l i r ( ; V I I I ; and rt.\ LI 1.1' 21(J 1111:12 \ la( k i • THE GSM SYSTEM 2; 8 BP cycle FCCH B C C H downlink time 111 SCH P A G C H r T used for TACHreS . 0 2 1 . II . 6 1 0 111 2 2 1 uplink 20 R A C H 3031 u s e d 4 0 14 1 for TA C K 8 s 11 m u n n u m m u m i n 4 4 36 i f --4 5 \ 2 1 1 These extension sets are found on TN 2 (one extension ). INs 2 and 4 (two extensions), or TNs 2. 4 and 6 (three extensions). Why impose this constraint? The reasons are the tbllo w ing: • first. a l l common channels 0.* one cell must use die same downlink frequency (and therefore the same uplink frequency): this will be explained when exploring the frequency realm page /24; • second, cells o f very large radius may allow RACI I bursts to overflow into the next slot. as explained i n Chapter 6. I f consecutive TNs had been allowed f o r extension sets. they would not be compatible with such possibilities; • t h i r d and last. i t w a s desirable t o simplify t h e system complexity by minimising the number of different cases. Figure 4.13 — A common channel pattern for small capacity cells A typical common channel implementation of a small capacity cell combines 4 dedicated channels (TACH/8s) used for signalling with a set of common channels (FCCH, SCH. BCCH, PAGCH/T and RACH/H) The second combination has been introduced w i t h the small capacity cells in mind (which are usually not the smallest in geographical size...). When the capacity of a PAGCH/T and a RACH/F are not needed, the operator might be interested in combining: • i n the downlink direction, a PAGCH/T with the usual FCCH, SCH, BCCH plus 4 downlink TACH/8s: • i n the uplink direction, a RACH/H with 4 uplink TACH/8s. Such a combination is shown i n figure 4.13. A s for the basic combination, this one uses TN 0. Conversely, the third combination is used in very high capacity cells, in which a PAGCH/F and a RACH/F are not sufficient to cope with too high a traffic level. On top of a set of channels grouped as in the basic combination (i.e., w i t h PA G C H / F a n d RACH/F), a c e l l m a y accommodate u p t o 3 extension sets, each set following the third combination. An extension set contains the same channels as the basic combination, except the FCCH and SCH (they must be unique in the cell). A BCCH appears in each extension set, at least for two reasons: first, a part o f the information broadcast on the BCCH relates to the RACH of the same TN, and hence can differ from one TN to another; second, it is simpler for the mobile station to listen to bursts of a single TN only. CBCH A Cell Broadcast C7Hannel (CBCH) follows a cycle o f 8 x 51 x 8 BPs (lasting for about 2 stionds). where 4 times 4 slots are used. The allowed positions in the 51 x 8 BPs cycle. and the allowed TNs. are limited, so that there is no collision with the requirement to listen to other BCCH or PAGCH information. The CBCH can be seen as a sub-part of a TCH/8. Tw o cases are t o he distinguished. I f the common channel configuration is the small one with a PAGCH/T and a RACH/H, the CBCH can use the same T N (TN 0) and frequency as the common channels. It then uses slots that would otherwise belonu to one of the four TCH/8s which can use T N 0 and the beacon frequency. A second possibility, applicable whatever the common channels configuration. is for the CBCH to use TN 0 (but not on the beacon frequency). I . 2 or 3: the CBCH bursts must there again use a specific position in the 51 x 8 BP cycle, which would otherwise belong to a TCH/8. When a CBCH is used. the first block of the PAGCI I in the 51 x 8 cycle cannot be used for paging. All these rules ensure a minimum time between a CBCH burst and a burst belonuing to a block carrying a paging message. However. in this second case. it is allowed (hut not mandatory) to have a CBCH with a TN different than O. In this case. the mobile station in idle mode has to listen regularly to bursts of different TNs. a source of complexity for the scheduling of reception. This is the sole case where such a requirement exists within one cell. I .1 7 0 8 BP cycle 1time FCCH B C C H SCH P A G C H / F CBCH TN 7 TN 0 .1:1111111 1111111 TN1 0 8 1 2 Figure 4.14 - Time organisation of a Cell Broadcast CHannel (CBCH) ACBCH uses part of the capacity which would otherwise be allocated to a TACH/8. Theexample shows the CBCH when combined with TACH/8s on TN I. Inside the 8 x 51 x 8 BP cycle, the CBCH can be seen as a half downlink TCH/8, using four out of the eight 4-burst blocks. The example of a CBCH using TN I is given in figure 4.14. The four other blocks, i.e., the slots that would be used by the SACCH, and the uplink corresponding slots, are not used by the CBCH, and cannot be used for anything else. However, it is allowed to stop the transmission of the CBCH.in case of congestion, and then these resources can be used for a TACH/8 during such periods. 4.2.1.3. Channel Organisation in a Cell The above sections have mainly taken into account the mobile station point of view when describing the channels. A few ideas of how channels may be grouped in a cell have been introduced; let us now dig deeper into the management of time resources by a base station. An elementary transceiver can emit or receive continuously, but on a single frequency slot at any given instant: it may be able to change its frequency as often as once every burst period, but i t cannot emit or receive two bursts on different frequency slots during the same time slot. A base station usually contains several such elementary transceivers in order t o reach t h e desired capacity. T h i s concept o f elementary transceiver is used in the Specifications (mainly in the specification of the interface between BTS and BSC) under the term TRX (Transmitter Receiver). Channels Unused slots I TACH/F I out of 26 2 TACH/H of different sub-TNs none 8 TACH/8 of different sub-TNs 3 out of 51 1SCH + 1 FCCH + I BCCH + 1PAGCH/F + I RACH/F downlink: I out of 51 uplink: none 1BCCH + 1 PAGCH/F + I RACH/F downlink: 11 out of 51 uplink: none I BCCH + I PAGCHIT + 1RACH/H + 4 TACH/8 downlink: 3 out of 51 uplink: none Table 4.1—Possible combinations of channels °Ellie same IN T h e s a m e c a p a c i t y o f one s l o t e v e r y /t s l o t s m a y b e u s e d f o r v a r i o u s c o m b i n a t i o n s . s o m e e x a m p l e s o f kN, Iiich a r e g i v e n i n t h i s t a b l e . It is therefore desirable. in order to optimise implementation costs in a base station, to choose channels so that they form groups where at most one burst is emitted at any one time, and to fill the time slots within these groups as much as possible. In order to facilitate such groupings, the time organisation of the radio interface makes an extensive use of the 8 BP cycle. Every TRX is able to cope with 8 groups of channels. each group corresponding to a given TN. For instance, a group o f channels using the same TN may consist of one of the combinations listed in table 4. I (the CBCH is not included). A TRX may combine eight such groups, with only the constraints listed earlier on common channels, in particular the TN(s) to use. The three combinations using only TACHs may exist on any TN. In outer .to oomplete the picture. let us give a few examples of channel combinations in a cell. A small ( w a d i , e l l w i t h a single T R X w i l l typically he organised as follows: TN 0: FCCH, SCH. BCCH, PACCHrf. RAC H/1-21.4 TACH/S: TN I to 7: I TACH/F each. I III. K . W i t ) A medium capacity cell with 4 TRXs ma, aclude: one TN 0 group: FCCH, SCH, BCCH, PAGCH/F, RACH/F; twice 8 TACH/8; and 29 TACH/F. I burst. Inside, , 4 l 1 find some precise information on the limits between slots, as well as sufficient information to deduce the number of the time slot in the cycle of 8 x 26 x 51 x 2048 BPs. and hence its position in all useful cycles. From this moment onwards, the mobile station only needs to maintain its knowledge of slot boundaries and to add 1 at each BP! A large capacity cell with 12 TRXs may include: one TN 0 group: FCCH, SO-I, BCCH, PAGCH/F, RACH/F; one TN 2 group, one TN 4 group and one TN 6 group: BCCH, PAGCH/F, RACH/F; 5 times 8 TACH/8; and 87 TACH/F. 4.2.1.4. Synchronisation Acquisition, or "How did it all start?" Any reader who is not an expert in synchronisation might well at this stage ask the reasonable question: how does the mobile station manage to find the very first synchronisation with a cell, in order to read any of the channels defined in the previous paragraphs? What kind of bootstrap is there? A few explanations might clarify the issue. As already mentioned, the FCCH and the SCH are provided for helping the mobile station to acquire the synchronisation. More precisely, the successful reception of an SCH burst will give the mobile station all the information needed for synchronisation. T h e problem i s then t o find a n SCH burst. The specifications are such that an SCH burst always follows an FCCH burst 8 BPs later on the same frequency. Now an FCCH burst has a rather easily recognisable structure. A possibility is then to look for FCCH bursts, at all frequency slots and at all times. When such a burst is encountered, the mobile station is able to get some information out of it. First (as the name FCCH evokes), it is able to correct the frequency o f its internal time base i n order to ease the demodulation of other channel bursts—this is however not the main point here. What is more important for synchronisation acquisition is that the mobile station is able to have a rough idea of the boundaries between slots, and of the situation in time of the slots of TN 0 (since the FCCH uses by definition slots of TN W. Knowing how SCH slots are positioned relatively to FCCH slots, the mobile station may then proceed to find and demodulate an SCH 4.2.1.5. Frames This small section is not aimed at adding more explanations on the tinie organisation of the channels. The previous sections were to give all explanations, and i t is hoped they have reached their goal. Yet. the proposed description is quite different than what can be found in the Specifications, or in the main literature on GSM (which usually follows the same approach as the TSst. In the following paragraph we will try to bridge the gap. In the Specifications. the time description o f channels refer to "frames", a word which we did not use in this context. A frame is often presented as the succession o f n slots. In particular. a "TDMA framerepresents a succession of 8 consecutive slots. the accent being put on the grouping of slots rather than on the 8 BP cycle. The grouping vision is quite natural when dealing with the implementation o f a base station. which caters for many channels. But the cycle approach is much more natural as seen from the mobile station. which deals with few channels at the same time. This is why we preferred to put the stress on the concept of cycle (shown as an helix in the figures) rather than use the notion of "frame". However, since the numbering of slots in (ISN1 is very much based on frames, it is worth presentingthe GSM frame hierarchy now. In the Specifications, a TDMA frame most ()hen refers to a grouping of 8 time slots starting with one of TN 0. This allows to use the - T W A frame number", which is simply the quotient modulo 8 of the time slut numbers of the time slots in the frame. But because all the channels arc designed as having only slots of the same TN. the full time slot numbering is 110% er used in the Specifications. Instead, one finds the TDMA frame number (FN) plus the TN. There a r e o t h e r frames appearing i n t h e Specifications. corresponding to the major cycles. They are shown in figure 4.15. The "26 TDMA frame multiframe- is delined as a succession of 26 TDMA frames, and corresponds to the 26 x 8 BP or 120 ins cycle used in the definition of the TACH/F and the TACH/H. hyperframe = 2048 superframes 3h 28mn 53s 760ms Superframe = 26 x 51 multiframes (6.12 s) "26 multiframe" (120 ms) 0 1 2 24 25 "51 multiframe" (#235 ms) 0 1 2 3 48149 501 TDMA frame (4.615 ms) 01 1 1 1 1 1 17 Figure 4.15 — Hierarchy of frames The superframe is the smallest multiple of both the 51 T D M A frame cycle for common channels and the 26 TDMA frame cycle for dedicated channels. 204K superframes build an hyperframe, which serves as the basis for frame numbering. Similarly, the "51 T D M A frame multiframe" i s defined as a succession of 51 TDMA frames, and corresponds to the 51 x 8 BP cycle used in the definition of the TACH/8 and of the common channels. The "superframe" is a succession of 51 x 26 TDMA frames (6.12 seconds), and corresponds t o t h e smallest cycle f o r which the organisation of all channels is repeated. Note that this repetition abstracts some properties of the channel, for instance the SACCH coding period is not taken into account (the period would then be 4 superframes), neither is the internal structure of the PAGCH (which in some cases does not recur more rapidly than the hyperframe which will now be defined). The "hyperframe" i s t h e numbering p e r i o d . I t i s 2048 x 51 x 26 x 8 BP long, that is to say exactly 12533.760 seconds, or 3 hours, 28 minutes, 53 seconds, and 760 milliseconds. It is obviously a multiple of all previously cited cycles, and determines in fact all the cycles in the transmission on the radio path. It is in particular the smallest cycle for frequency hopping and for ciphering. 4.2.2. THE 2EQL f m l A x i s 4.2.2.1. T h e Available Frequencies GSM was first devised as a cellular system in a specific 900 MHz band, called "the primary band". This primary band includes two subbands o f 25 M H z each. 890-915 MHz and 935-960 MHz (see figure 4.16). This does not mean that the whole primary hand must be used for GSM in a given country. especially at the start of the system. Moreover. a given operator is rarely given more than a portion of this band, since most countries have several operators. However, every mobile station must he able to use the full band. in order not to impose constraints upon roaming users. In 1990, upon request of the United Kingdom. a second frequency band was specified for being used with the Specifications. This hand includes the two domains 1710-1785 MHz and 1805-I 880 M Ilz. i.e.. twice 75 MHz: three times as much as the primary 900 MHz band. --Mobile stations using this band are different from those using the primary band: using the same mobile equipment for roaming betiseen the two variants of the system. GSM900 and DCS1800. although not ruled out. is not envisaged in the near future. Another extension o f the primary band i s foreseen. I t should consist o f the band which is directly "below- the primary hand. For instance, an 8 MHz extension would raise the 900 MHz hands to 882915 MHz and 927-960 MHz. i.e., twice 33 MHz. 45 MHz I I I 890 9 0 0 9 1 0 9 2 0 9 3 0 9 4 0 9 5 0 25 MHz 960 f 25 MHz uplink (Mobile to Base) downlink (Base to Mobile) Figure 4.16 • GSM primary band The GSM primary hand includes Iwo 25 N111/ sob-hands around 91)0 MI I/. THE GSM SYSTEM 0 0 1 2 T H E R A ' ) ( c a r r i e r number) 10 219 I N T E R F A C E frequency try Sit 890.2 8 9 0 . 4 890.6 ( M H z ) t r IMO 200 kHz GSM band Carrier spacing is equal to 200 kHz. A guard band of 200 kHz between the edge of the band and the first carrier is needed at the bottom of each of the sub-bands; the carrier numbered 0 is often not used in practice. The central frequencies of the frequency slots are spread evenly every 200 kHz within these bands, starting 200 kHz away from the band borders (see figure 4.17). 124 different frequency slots are therefore defined in 25 MHz, and 374 in 75 MHz. The modulation spectrum is somewhat wider than 200 kHz, resulting in some level o f interference between bursts on simultaneous slots or on adjacent frequency slots. This is a nuisance mainly near the band borders, since this interference could disturb non-GSM applications in adjacent bands. The border frequencies are therefore usually avoided. The normal practice is not to use the frequency slots at the border (those numbered 0 and 124), except when a special agreement has been reached with the users of the adjacent band. As a consequence, the number of frequency slots which can be used in 25 MHz is usually limited to 122. 4.2.2.2. Frequency Hopping The radio interface o f GSM uses slow frequency hopping. Frequency Hopping consists in changing the frequency used by a channel at regular intervals. T h e origin o f this technique lies i n military transmission systems, where it was introduced to ensure secrecy and combat jamming. Publications in that domain distinguish Fast Frequency I M a S i Figure 4.17 —Carriers at the border of the GSM band V ter r e I M F time Figure 4.18—Slow Frequency I lopping in the tilne l equenc domain Frequency for a given channel nun change at each bunt. and remains constant during the iransmissi(14)1 a burst Hopping (FFH), where t h e frequency changes quicker than t h e modulation rate, from Sher Frequency !lopping (531:1-1). In GSM, the transmission frequency remains the same during the transmission o f a whole burst: GSM belongs therefore clearly to the slow hopping case. Figure 4.18 shows an example o f a time-frequency diagram l e r a frequency hopping channel. Slow frequency hopping was introduced in GSNI f o r two maw reasons. The first reason is frequency diversity. As shall he explained later, error-correcting codes are introduced i n the transmission chain. Such codes are based on redundancy: the data is made redundant in such a way that. even with a certain amount of errors. the original data ma.y_be reconstructed from what remains in the received flow. This redundancy is spread over several bursts. SF1.1 therefore ensures that this information is sent on several frequencies. and this improves transmission performance. To explain this, a digression concerning propagation is needed. Mobile radio transmission i s subject t o important short term amplitude variations when obstacles are involved: these variations are called Rayleigh fading. I n most cases, the emitting and receiving antennas are not within direct sight one with the other, and the received signal is the sum o f a number o f copies of one signal with different phases. For instance, if the path includes a reflection on an ohsincle. there 220 THE GSM SYSTEM R.‘1)10 IMERFAul. 221 has been evaluated to be around 6.5 dB. This advantage is o f prime importance for a system where a high proportion of handhelds is sought, since hand-held users are usually moving at a slow pace or not moving at all. 10 The second advantage o f frequency hopping i s interferer diversity, a property associated with Code Division Multiple Access (CDMA). In high traffic areas, such as large cities, the capacity o f a cellular system is limited by its own interferences caused by frequency reuse. The relative interference ratio (C//) may vary a lot between calls: C (the Carrier level) changes with the mobile station position relative to the base station, with the amount o f obstacles between them. etc.: / (the Interference level) changes depending on whether the frequency is being used by another call in some nearby cell, and it also varies according to the distance with the interfering source, its level, etc.. -10 -20 dB -30 -40 -50 -60 0 2 3 4 5 6 7 Figure 4.19 — Typical amplitude variations due to Rayleigh fading (the time unit is the time to move through one wavelength, e.g., 24 ms at 50 km/h for 900 MHz) are usually many other reflected paths, in particular when the reflector Is irregular and of a scale greater than the wavelength (buildings meet these criteria). The sum o f a l o t o f phase-shifted signals with a random distribution o f phases h a s a n envelope following t h e Rayleigh distribution. Figure 4.19 shows an example of the variation in time of the envelope of a Rayleigh affected signal. Since the aim of a system is usually to satisfy as many customers as possible, its maximum capacity is calculated based on a given (small!) Proportion o f calls subject to a noticeable decrease in quality due to interferences. Because of this concept of "worst case", the capacity of a system is better when, for a given mean C// value, the statistical spread --around this mean value is as small as possible. Let us consider a system where the interference level perceived b y a call is the mean o f the interference level caused by many other calls: then, the greater this number of interferers for a given total sum, the better the system. This is how interferer diversity operates. Now, the fading incurred by signals at different frequencies are not the same, and become more and more independent when the difference in frequency increases. With frequencies spaced sufficiently apart (say 1MHz), they can be considered completely independent. With frequency hopping, all the bursts containing the parts of one code word are then not damaged in the same way by Rayleigh fading. In a system such as the current analog ones, a call potentially receives interference from a small number of other calls (typically 2 to 6, depending on the reuse pattern). A t the other extreme. i n a CDMA system—very fashionable these days West o f the Atlantic—all calls interfere a little with all others. For the same mean interference value in the two systems, a call in the conventional system will either have a very good quality or be completely jammed, whereas a call in the CDMA rarely so had that system will always have sonic low l e vofeinterference. l I transmission would fail. Thus the interferer diversity can be used to increase system capacity. When the mobile station moves a t high speed, the difference between its positions during the reception of two successive bursts of the same channel (i.e., at least 4.615 ms) is sufficient to decorrelate Rayleigh fading variations on the signal. In this case Slow Frequency Hopping does no harm, but it does not help much either. However, when the mobile station is stationary or moves at slow speeds, SFH allows the transmission to reach the level of performance of high speeds. The gain The major drawback of CDMA systems is that their design usually leads to calls interfering in the same cell and in adjacent cells. What is good for reducing the spread around thetmean value causes the same mean value to decrease! GSM has been devised to avoid collisions inside one cell and between a certain number o f adjacent cells. I t allows however to spread the interference between many calls o f a potential interferer cell, instead of a single one as in conventional systems. 271 11 3 THE GSM SYSTEM An example should help to illustrate this principle. Let us consider a cell using 4 frequencies (cell A), and 2 potentially interfering neighbour cells (see figure 4.20). Each one of the latter two also uses 4 frequencies, two of them being part of the 4 frequencies used in the first cell. The figures listed in table 4.2 give an example o f the interference levels experienced by a mobile station in cell A when traffic is maximum. A call is deemed jammed when the sum of all interferences amounts to a C/I ratio o f less than 5.0 (7 dB). Without frequency hopping, the mobile station has an even chance to receive correctly the signal (if the allocated frequency is f1 o r f2), whereas with frequency hopping, the quality is correct in all cases. This example is not a deliberate choice of a rare case; such situations arise quite often and frequency hopping really creates a significant statistical gain. Mobile to base interference level on the frequency 11 11 — 13 14 0.10 (C/1=10 dB) (t.14 (C/1=8.5 dlt 1 0.25 (C/1.6 dB) 0.28 W/1=5.5 dB) Mobile to base average interference level R I 9 IC/1=7.2 (IB ) Base to mobile interference level on the frequency 0.2S 1(71=5.5 dB) 0.10 iC/1=10 dB) Base to mobile average interference level 0.19 t(71=7.2 tIBI Table 4.2—Interference e e l s lor 4 calls in each cell of figure 4.20 (without power control: lek els relative to %%anted signal level, • Without frequency hopping, perform:111LT can be compared with die first and third lines: with frequency hopping. it can be compared with the second and last lines. Interferer diversity helps to imprint: capacity for a given mean quality, 4.2.2.3. Hopping Sequences GSM allows a wide diversity—indeed almost an infinity- - o f different channels, when both time and frequency parameters are taken into account. A hopping sequence—i.e., the sequence o f couples (TN. frequency) allocated t o a channel-fmay use u p t o 6 4 different frequencies. O f course. t h e single frequency l i s t i s possible: i t corresponds to a fixed frequency channel. which appears here as a "degenerate" frequency hopping channel. Figure 4.20—Example of interfering cells with Slow Frequency Hopping Cells using the same frequencies. but with decorrelated hopping sequences, lead to interferer diversity. Hopping sequences are described for channels using one slot every 8 BPs. A hopping sequence is then a function o f the timeslot number modulo 8. For a channel using less than one slot every S BPS. the hopping frequencies are calculated by applying the same function. For example, if a TACH/F uses the following h y i n g sequence: I 2 3 4 I 2 3 ' 4 Then t h e corresponding TA C I I/1 I s h a l l use t h e following sequences: 4 sub-TN 0: sub-TN I: I 3 I 4 3 4 ... 224 T H E GSM SYSTEM Since the number o f bursts i n the uplink direction derives conventionally from the one in the downlink direction by a delay of 3 BPs, the hopping sequence (i.e., the function which associates a frequency to each TN modulo 8) in the uplink direction derives from the one in the downlink direction by simply adding 45 MHz. For a set o f n given frequencies, GSM allows 64 x n different hopping sequences to be built. They are described by two parameters, the MAIO (Mobile Allocation Index Offset) which may take as many values as the number of frequencies in the set, and the HSN (Hopping Sequence Number) which may take 64 different values. Two channels bearing the same HSN but different MAIOs never use the same frequency on the same burst. On the opposite, two channels using the same frequency list and the same TN, but bearing different HSNs, interfere for 1/n th of the bursts, as i f the sequences were chosen randomly. The sequences are indeed pseudo-random, except for the special case of HSN = 0, where the frequencies are used one after the other i n order. Pseudo-random sequences have been chosen because they have statistical properties similar to random sequences. Usually, channels in one cell bear the same HSN and different MAI0s: it is desirable to avoid interference between channels inside a cell. Adjacent cells are not interfering either, since they use disjointed frequency sets. In distant cells using the same frequency set, different HSNs should be used in order to gain from interferer diversity. I f this gain is sought, i t is best to avoid HSN = 0, which leads t o poor interferer diversity, even with non-identical frequency sets. 4.2.2.4. T h e Case of Common Channels There exists a restriction to the use of frequency hopping: common channels (FCCH, SCH, BCCH, PAGCH and RACH) must use a fixed --frequency. This constraint i s meant t o ease initial synchronisation acquisition (described on page 214): once the mobile station has found an FCCH burst, it will look for an SCH burst on the same frequency. Since this burst is too small to contain the description of a hopping sequence for the BCCH, the simplest-way is to put the BCCH on the same frequency as the SCH. If the PAGCH and the RACH were hopping channels, their hopping sequences could be broadcast o n the BCCH. This would however increase system complexity for little gain. The choice was that common channels on TN 0 never hop and all use the same frequency. 11A1)I() R P : \ l ' I • . 225 Similarly, extension sets of common channels are also forbidden from hopping and use the same frequency as the primary group, so that there is no need to transmit the description of their frequency organisation on the BCCH of TN 0. Another peculiarity related t o common channels i s that the frequency they use must he emitted continuously, even ir no information needs to be conveyed on some bursts. This is needed because mobile stations in neighbouring cells continuously perform measurements on this frequency, in order to determine the best cell they should listen to or to report measurements f o r handover preparation. When there i s n o information transfer request. a specific pattern is emitted (the fill frames). Because of these special roles of the frequency carrying the FCCH, we will refer it by a special name. the "beacon frequency" of the cell (in the Specifications it is in some places referred to as the BCCH frequency). 4.2.2.5. Channel Organisation in a Cell The description of the channels has been up to now focused on the description of one channel at a time, a point of view close to that of the mobile station. The point of view of a base station is somewhat different. We have seen in the time domain description that channels have been designed so as to use as well as possible a transmitter which is limited to one burst per burst period. This introduces the notion of TRX, which is the natural unit to measure the capacity of a base station (this unit is so natural it appears in the procedures between the 13'FS and the BSC). A TRX has then the capacity of 8 TACH/F. or 16 TACH/H. or 64 TACH/8 grouped by 8, or a lot o f other combinations. It is perfectly possible to build equipments with smaller capacities. but it seems they have no commercial interest, and the TRX is a concept used by all manufacturers. Another reason for this concept is the frequency allocation. A cell is usually allocated an integral number n o f frequencies. and the maximum capacity of such a cell corresponds to the capacity of n TRXs. This does not mean that there is a one to one relationship between frequencies and TRXs: this would he true only if frequency hopping was not used. But in most applications, a cell is equipped with exactly as many TRXs as allocated frequencies. In fact, it may happen that a cell is "7.g-hipped with less TRXs, for economy reasons. but this raises a specific problem that will be dealt with later. II: ('NMSip NTIN II11. I<.\1)10 frequencies A D 1 beacon frequency iV.rt C B 0 1 2 3 4 5 6 7 227 different TNs. For a given TN, the channels are grouped in the frequency domain in one or several sets including the channels which have at least one frequency i n common. I n fact, the G S M frequency hopping sequences are defined in such a way that the only rational approach is that all the channels in a group use the same set of frequencies. Hence the diagrams, which show such frequency/TN sets. The consequence o f these possibilities i s that • the bursts i n succession on the same frequency may belong to a rot o f different channels, and that there is very little logic in the succession. This is why the notion of TDMA is somewhat misleading with frequency hopping. Channels are not sharing a fre e n c y on a time division basis, the channels in a same frequency/TN u p are sharing several frequencies. TN A set of commonchannels + SCH + FCCH on beacon frequency B set of 5 TACH/F, hopping on 5 frequencies C set of 8 TACH/E + 5 TACH/F, hopping on 6 frequencies D additional set of common channels on beacon frequency andso on... Figure 4.2I—Frequency/TN groups TheGSM hopping sequencesare such that, for a given TN in a given cell, onemay define channel groups hopping on the sameset of frequencies. In the case of group C (to take an example). if all channels are not allocated, theBTS must still manage to emit on the beacon frequency in all slots of TN 1. Let us look at a typical cell with n frequencies and n TRXs. The channel organisation must include one common channel group with a FCCH and a SCH, consisting o f either (FCCH + SCH + BCCH + PAGCH/F + RACH/F), or a group with combined TACH/8 (FCCH + SCH + BCCH + PAGCH(T + RACH/H). Other common channel groups May be added in the first case. These choices are determined by load considerations. The common channels are non-hopping channels, and all use the same frequency, the beacon frequency. The rest of the resources are distributed among TACHs, with a ratio between TACH/F and TACH/8 which depends on load considerations. The constraints o n the hopping sequences are few. To show various possibilities, we will use the diagram in Figure 4.21. Such a diagram is based on the particularity that the channel configuration can be described independently for each TN, because no channel use bursts of Substantial gains in frequency and interferer diversity are obtained when at least 4 frequencies. and preferably more (say 8), are used in a hopping sequence. This causes a problem in the cases where for capacity —reasons a single TRX would he sufficient in a given cell. The operator may choose f o r the gain o f frequency hopping t o allocate more frequencies to that cell than the number of installed TRXs. A small problem then arises f r o m t h e necessity t o e m i t continuously on the beacon frequency. In cells o f small capacity, the operator may choose either to let the channels of TN other than h o p on only as many frequencies as there are TRXs (but the gain of frequency hopping is small), or on as many frequencies as available. In the latter case, an additional transmitter dedicated to the filling o f the common channels frequency is needed. Because the frequency/TN groups are not fully used by the installed TRXs. the continuous emission of the beacon frequency i s not guaranteed b y these T R X alone. The role o f the additional transmitter is to emit on the beacon frequency when it would have been used by one of the missing channels. 4.3. FROM SOURCE DATA TO RADIO WAVES Up t o this point, w e have addressed only h o w transmission resources are organised to be shared between users, not how they are used. In the previous chapter. we have seen that. if we restrict our view to the radio interface, all needs for user data transmission can he fulfilled 228 229 THIS RADIO INTERFACE THE GSM SYSTEM • with a few different transmission modes. Let us recall and name these modes, with a notation derived from the Specifications: • TCH/FS, T C H / F 9 . 6 , T C H / F 4 . 8 , T C H / F 2 . 4 a r e t h e transmission modes over a TCH/F, respectively for full rate speech, 12 kbit/s, 6 kbit/s and 3.6 kbit/s data. Speech • TCH/H4.8, TCH/H2.4 are the transmission modes which were defined over a TCH/H, respectively for 6 kbit/s and 3.6 kbit/s data (the speech mode on a TCH/H is not yet specified). There is a good chance that data modes stay as indicated here. To these modes must be added the different needs for signalling, such as fast associated signalling, and the transmission mode on the SACCH, BCCH and PAGCH. In these cases, the source data appear as an irregular f l o w o f blocks. Finally there i s the variety o f specific transmission needs coming from the RACH, the FCCH and the SCH. After the presentation o f the burst, and its divers species, the presentation will follow the fate of the information from data to radio waves. Radio emission involves several successive operations in order to convert source data into the final signal. Conversely, reception implies that a reverse series of operations be performed at the receiver's side until the original data is—approximately—regenerated. The sequence of such operations for speech is shown in figure 4.22. The process is identical for other user data and signalling. p Digitizing and I source coding J S Channel I coding J C d / We will see that, in detail, GSM transmission for the different modes presents at the same time some uniformity and some variety. The unifying concepts are the burst, which will be our first topic, and the modulation, which will be our last. On the other hand, the different modes differ in the details o f the error correction and error detection coding schemes. S e o e u c h r c h a n n e c o d i L e decoding J I e n l g X Interleaving De-interleaving Burst e formatting Burst formatting I _1— Deciphering Ciphering I _\--Modulation Demodulation The operations described i n this section are common t o all transmission modes. The following operations take place on the source side: • channel coding introduces redundancy into the data flow, increasing its rate by adding information calculated from the source data, i n order t o allow the detection o r even the correction of signal errors introduced during transmission. The result of the channel coding is a flow of code words; in the case of speech, for example, these code words are 456 bits long; Figure 4.22—Sequence of operations from speech to radio w a r s ... and hack to speech Alter having transformed speech into digital blocks, channel coding ;aids redundancy. The block. :ire inierlaat ed and spread into pieces. which are combined kith Ilags and a -midanthle" 141 build up the burst.. Ciphering is applied to these bursts and the resulting data is used to modulate the farriers. The reverse tranNionnations are performed on the revel er 4 • 230 T H E GSM SYSTEM TIIF. I n o 23 I • interleaving consists in mixing up the bits o f several code words, so that bits which are close to one another in the modulated signal are spread over several code words. Since the error probability of successive bits in the modulated stream is very much correlated, and since channel coding performance is better when errors are de-correlated, interleaving aims at decorrelating errors and their position i n code words. After interleaving, the f l o w o f information i s a succession o f blocks—one block for each channel burst; 4.3.1. THE BURSTS • ciphering modifies the contents of these blocks through a secret recipe known only by the mobile station and BTS. by the BTS i f the adjacent burst is not emitted. The Verifications allow dint do Inn enforce) the BTS to keep the anmlitude constant between two adjacent emitted burs's. • burst formatting adds some binary information to the ciphered blocks, in order to help synchronisation and equalisation of the received signal; the output o f this stage consists o f binary information blocks. Several kinds o f bursts are defined with regard t o their timeamplitude profile. e.g., "normal" bursts and access bursts. An example of the amplitude profile of a normal burst (used. e.g.. on tralTic channels) during its emission window is given in figure 4.23. together with the rinu mask defined in the Specifications and which gives the acceptable limits within which the amplitude profile must lie. Its constant amplitude part lasts 147 bit periods, i.e., 2.5 hits on each side ()I' the 142 informationcarrying bits. • modulation transforms the binary signal into an analog signal at the right frequency and at the right moment, according to the multiple access rules described starting on page 215; last but not least, this signal is radiated as radio waves. The receiver side performs the reverse operations as follows: • radio waves are captured by the antenna. The portion of the received signal which is of interest to the receiver is determined by t h e multiple access rules. T h i s portion undergoes demodulation with the help o f the additional information introduced during burst formatting. The result may consist in a succession o f binary information blocks. More sophisticated demodulators, however, are able t o deliver a n estimated probability o f correctness for each bit: this i s called "soft decision"; • deciphering modifies those bits b y reversing the ciphering recipe; since-this recipe is a bit-by-bit "exclusive or" with a ciphering sequence, it may be performed just as well when a soft decision process is applied; • d e -interleaving puts the bits o f the different bursts back in order to rebuild the code words; • last, channel decoding tries t o reconstruct t h e source information from the output o f the demodulator, using the added redundancy to detect or correct possible errors in the output from the demodulator. This operation is much more efficient when the demodulator indicates an a priori error likelihood of each bit. The burst is the transmission quantum of GSM. Its transmission takes place during a time window lasting (576 + 12/13) ps. i.e.. (156 + bit duration. Within this time interval, the amplitude ()I' emission rises from a starting value of 0 to its nominal value. The signal phase is then modulated to transmit a packet of bits. Alter that. the amplitude decreases until it reaches 0. This description is valid only for emission by the mobile Mallon. and for emission The packet of bits used to modulate the signal phase of a burst includes in general the useful part o f the information, plus a training sequence and three additional " 0 " bits at each end. Theoretically. the signal phase is obtained by applying the modulation method to an infinite series of bits, consisting here of the hit sequence of the burst preceded and followed by infinite series of "I". The three "0" bits added at the beginning and at the end of each burst avoid a loss of demodulation efficiency for the extreme information bits. An interesting question is why the choice was not to set them to " t " like the infinite sequence assumed outside the burst. The specification is such that the transition from " I " to the first "0" hit ()tithe burst, and from the last "0" bit of the burst to " I " fall exactly in the ramping portion of the burst amplitude profile. A property o f the modulation i s that i n absence of transition, the modulated signal is shifted toward the higher --Tfaltiencies, and the interference created by the ramping outside the frequency slot would then be greater than with a bit transition. The training sequence is a sequence of bits known by the receiver. There are several such training sequences defined in GSM. as will be seen in the next paragraphs. The signal resulting front the transmission of this training sequence allows the receiver to determine very precisely the 232 T H E RADID INTERFAck GSM SYSTEM 1,13 position of the useful -signal inside a reception window, and to have an idea of the distortion caused by transmission. These information are of prime importance to obtain good demodulation performances. Several burst formats are defined: • t h e access burst is only used in the uplink direction during initial phases when the propagation delay between the mobile station and the base station is not yet known. This is the case with the first access o f a mobile station on the RACH, or sometimes with the access o f a mobile station to a new cell upon handover. The access burst is a short burst; it is the only kind of burst used on the RACH; • t h e F and S bursts are used respectively on the FCCH and on the SCH. T h e y serve solely f o r initial synchronisation acquisition of a mobile station in a given cell. • t h e normal burst is a long burst used in all other cases. -6 147 bits -70 or 36dBm 1 0 1 0 4 - 1 48- . 14 -> 4 7056,13 410* 8 410* ( p s ) 1burst period (7500'13 us) 4.3.1.1. T h e Normal Burst Figure 4.23 — Amplitude profile of a "normal- burst A normal burst contains two packets o f 58 bits surrounding a training sequence of 26 bits (see table 4.3). Three "tail" bits (set to 0) are added on each side. The Specifications also include the guard time in the burst. The actual guard time is determined by the signal envelope (see figure 4.23). If we only consider the guard time as the period during which the signal is below -70 dB, its duration is about 30 ps. In the uplink direction, this guard time is barely enough to compensate for equipment inaccuracies and for multipath echoes i f they are spread on the maximum range allowed by demodulation. The propagation delay itself is compensated by the timing advance mechanism (see Chapter 6 for this topic). In the downlink direction, the guard time could have been chosen shorter, but it has been kept at the same value for symmetry reasons. Tail Information Training Sequence Information Tail 3 58 26 58 3 Table 4.3 — Contents of a no mai burst 116 information bits are spread on both sides of a midamble, or training sequence, of 26 bits. The time mask of a normal burst is specified \1/4 ith a constant amplitude during the "useful parr of the burst and power ramping at both ends. The power level during the guard time should inn exceed 10- (-70 dB of the "useful part" level. or l0-"'' W (-36 (Mink whichever the highest. The training sequence has been inserted in the middle of the burst in order to minimise its maximum distance with a useful hit. and is therefore sometimes called "tnidamble- h a s the same role as a preamble, but is in the middle of the burst). The only drawback of that position is the need for the receiver to memorise the first portion of the burst before being able to demodulate it. but this is a very mild constraint compared to the benefit gained from it. Eight different training segue 'es have been specified. Why not a single one? Let us consider the ca.. o f two similar interlering signals arriving a t the receiver at almost the same time. I f their training 'sequences are the same. there is no way to distinguish the contribution of each oalithem to the received signal. The situation is much clearer when the two training sequences differ. and are as little correlated as possible. Distinct training sequences will therefore be allocated to channels using the same frequencies in cells which are close enough to interfere with one another. 234 T H E llt I t \DIU IN 11.10 V 1 GSM SYSTEM 2 3 5 The 8 - training sequences have been chosen f o r their low correlation between one another as well as for the special shape of their autocorrelation function, which is meant to ease some demodulation techniques. Figure 4.24 shows the autocorrelation function o f one of these 8 training sequences, calculated between the central 16 bits and the whole 26 bit sequence. All 8 sequences share the central correlation peak surrounded by 5 "0" on each side. -30 87 bits )1, -70 or - -36 dBm 1 4—> 0 8 41 0> 4 1 7 6 / 1 3 4- I> - t > 410p 8 410> l > t (ps) . . access burst =87 bits _ normal burst = 147 bits burst period = 500'13 us Figure 4.25 —'time mask for an access Figure 4.24—Autocorrelation function of a GSM training sequence as a NRZ signal Every GSM training sequence shares this autocorrelation shape, calculated between the 16 central bits and the whole 26-bit sequence. 4.3.1.2. T h e Access Burst As already mentioned, the access burst is the only short burst defined in GSM. Its envelope during the time window is constrained by the time mask shown in figure 4.25. An access burst contains a 41-bit training sequence, 36 information bits and respectively 7 and 3 tail bits at the beginning and at the end (see table 4.4). The training sequence and the initial tail are longer than the ones o f normal bursts, in order to increase the demodulation success probability: the task is quite hard indeed, since the receiver starts from An access burst has the same ramping specifications as a normal hursi, but its useful duration is much shorter in order to account for propagation lime with distant mobile stations. scratch: it knows neither the reception level, neither the frequency error, nor the exact reception time! A single training sequence i s specified 11w access bursts: the interference probability i s indeed t o o stn:tll t o j u \ t i l \ t h e added complexity of multiple training sequences. It Tail Training sequence Information 7 41 36 Table 4.4 -- Contents of an access burst The training sequence is longer than in a normal burst.' and the useful information is sere short. Tail )36 237 'HIV RADIO INTIM:At:1i THE GSM SYSTEM Total delay Tail Information Training Sequence Information Tail 3 39 64 39 3 Table 4.5 - Contents of an S burst Base station Mobile station time Downlink delay A burst on the s a l bears less lamination than a normal burst. but has the same length 1142 bits) As the access burst is the first burst that a base station needs to demodulate in the uplink direction. the S burst is the first burst that a mobile station needs to (1em dulatein the down)1 4 direction. Therefore. its training sequence is longer than the one of a 1161:mai burst. The training sequence is unique, by necessity: the mobile statioi would otherwise not be able to know the sequence chosen by the •base station i f several sequences were defined. Uplink delay emission 4.3.1.4. T h e F Burst reception The F burst is a very peculiar burst. It is a long burst. whose sole use is to enable mobile stations to find and demodulate an S burst of the same cell. Figure 4.26 - Delay of an access burst The time offset of an access burst as seen by the base station is equal to twice the propagation delay. Since the propagation time between mobile station and base station is not known when such bursts are used, an access burst arrives at the base station with a time error of twice the propagation delay (see figure 4.26), compared to the reception window. The small duration o f the access burst is there to compensate for this effect: the mobile station has to be very far indeed for the burst not to fit in the reception window. Mobile stations may wander up to 35 km away from the base station before they miss the target. 4.3.1.3. T h e S Burst The S burst is only used in the downlink direction on the SCH. It is as long as the normal burst- 1 4 2 bits—but its contents are different (see table 4.5). • I t is the simplest of all bursts: all of its 14K bits are set to "01 With the modulation technique used. the resulting signal is a pure sine wave which frequency is 1625/24 kHz higher than the carrier central frequency (two successive bits with equal value lead to a phase shill o f Tc/2, as explained in the section dealing with modulation). 4.3.1.5. Preemption As explained on page 190, it is possible to insert signalling blocks in the TCHs, when those are normally used to transfer user data. In both cases normal bursts are used, but the channel coding differs. I t i s therefore advisable for the receiver to know before starting decoding whether the burst includes user data or signalling. This information is included through an addressing mechanism inside the burst. The receiver is therefore able to get the information alter demodulation. and thereby avoids the need to attempt decoding in two different ways. 238 T H E GSM SYSTEM The binary information indicating whether a coding block is user data or signalling is called the stealing flag. Two bits are used in the normal bursts for the purpose of carrying it, one bit for each half of the burst. They are the bits closest to the training sequence on each side. When a signalling block is transmitted through fast associated signalling on the TCH/F or TCH/H, it uses 8 burst halves, and the corresponding stealing flag bits are set to indicate signalling. On other channels than a TCH/F or a TCH/H, these bits are of no use. They are always set to "1" (meaning signalling). On channels other than a TCH/F and a TCH/1-1, the stealing flags could be considered as an extension of the training sequence, because they are known beforehand. Unfortunately, the good autocorrelation properties o f the training sequence are net extended. whatever the sequence. Curiously enough, half of the sequences bear the good properties when one more " 0 " is added on each side o f the central peak (Training Sequences 0, 1, 6 and 7). The other cases result in asymmetric correlation functions. T111:. li.\1110 INTERFACE 239 Interleaving consists in spreading the b bits of a code word into it bursts in order to change proximity relations between the bits. The larger the value of n, the better the transmission performance. On the other hand, the larger the value o f n, the longer the transmission delay. A compromise has to be sought. and depends on the channel usage. Several interleaving schemes are therefore specified in GSM. In order to avoid complex huerleaving schemes, it is best to have simple arithmetic relations between b, ii and the number of bits per burst (114 in the GSM case). For example, a code word of b=456 bits (i.e.. 4 x 114) could be spread into. e.g.: • 4 parts of 114 bits, each one filling up a whole burst: • 8 parts of 57 bits, each one filling up half a burst: or • 2 4 parts of 19 bits. each one using one sixth of a burst: or • 7 6 parts of 6 bits, each one using one 19th of a burst. 4.3.2. INTERLEAVING AND CHANNEL CODING The interleaving and coding schemes are all different for the different transmission modes. However, they are all from the same family, and we will present first the general aspects which apply to most of the cases. Then a splendid table w i l l summarise the features of different modes, and an example. speech. will be presented in more detail. 4.3.2.1. General Principles of Interleaving Interleaving is meant to decorrelate the relative positions of the bits respectively in the code words and in the modulated radio bursts. Why is that useful? Firstly, bit errors tend to occur consecutively rather than singly. This is true inside a burst, and conies both from error statistics on radio transmission, and from intersymbol interference introduced by the modulation. The burst structure itself is a reason for error grouping: because of the variations of reception and interference level in time (and in frequency i f frequency hopping is used), a burst can have a small number of errors where the next will have a lot. Secondly, it appears that it is more difficult to design efficient codes when several adjacent bits are in error: better performance can be achieved when errors are randomised. With these examples, a burst would include contributions from different code words, respectively I. 2, 6 and 19. The first two cases are used i n GSM and are fairly easy t o understand. A third case o f interleaving is also used (for data services), but it is much more complex than the three listed above. The interleaving for data at 9.6 kbit/s is derived from the last example above. and is referred to as "19 bursts interleaving". The careful reader will have deduced from the above examples that no regular interleaving scheme can be defined for spreading blocks of 456 bits onto 19 bursts: each piece would he 24 hits long. and therefore one burst has to f i t exactly 4.75 such pieces! What kind o f jigsaw puzzle i s that? I n fact. die interleaving is done on 22 bursts. A code word is cut into 22 pieces: 16 pieces of 24 hits. 2 pieces of 18 bits, 2 pieces of 12 hits and 2 pieces of 6 hits. A burst includes 4 pieces of 24 bits plus either one piece of 18 bits or two pieces of respectively 12 and 6 hits. :\ burst may therefore include contributions from either 5 or 6 different code w ord.. And that works! Just take a few minutes to be convinced by figure 4.27 overleaf. Good. Now why "19 bursts..? The easiest w ay to understand is to cut the -156 bits code word into 4 sub-hlocks of 114 hits each. These sub-blocks should then he Tread over 19 bursts in a regular manlier U. bits on each huNti. shilling ['loin I burst at each • sub-block. The current specification has been generated this scan rat some stage during the specification o f the '..Ctern. COLIC words were I 14 bit. Ringo. hence the -19 bursts" terminology. More o n interleaving shall b e said i n conjunction w i t h the description o f the coding schemes. scanning each transmission mode separately. 240 T H E GSM SYSTEM 'HIE RADIO INTH4.1,Ati 2 4 1 (r2) Figure 4.27 —"19 burst" Interleaving scheme for data services There are 4 different types of bursts o ith regard to the contribution the 5 contain in terms of data blocks. Each burst may contain 6. 12. I3 or generally 24 hits from a given block. The organisation repeats itself every 4 bursts (burst n+.5 would bear the suns organisation as burst n. but with contributions of blocks 13-4, 13-3. 13-2. 13—I. 13 and 13+I ). 4.3.2.2. General Principles of Channel Coding Channel coding aims at improving transmission quality when the signal encounters disturbances (significant noise when the reception level is low, interferences, multipath propagation. Doppler shill, and so On...). It results however in art increased number of bits. Coding consists i n adding t o the source data some redundant information calculated from this source information. Decoding makes use of this redundancy to detect the presence of errors or estimate the most probable emitted bits given the received ones. Errors arc detected when the transmitted redundancy is different from the one calculated with the received data. Depending on the transmission mode, radio path transmission uses different codes. Moreover, several codes are "concatenated" in certain cases, i.e., redundancy is added by applying a given code. then more redundancy is added by applying another code. The codes used in GSM are few: • b l o c k convolutional codes; these code's are only used f o r correction purposes. They achieve tremendous efficiency when they are combined with likelihood estimation such as that coming from the demodulator; • a F i r e code; this code i s dedicated t o the detection and correction of "burst)" errors. i.e.. errors which are grouped. It is----used i n concatenation after a block convolutional code. for which residual errors often conic out in groups; • simple parity codes used for error detection. After a brief description o f the concepts used in these codes, a sunbmary of all interleaving and coding characteristics for the different transmission modes is given (table 4.7, page 246). The example of full rate speech is then studied in detail. Once this case is understood, the reader should find it easy to understand the other cases described in the Specifications. 242 T I I E 4.3.2.3. Convolutional Codes and Block Codes Source block (12 hits) Addition of tail bits Strictly speaking, a convolutional code consists in transmitting the results o f convolutions (that is to say what is obtained by adding a sequence and shifted versions of itself) of the source sequence using c different convolution formulas. As a n example, f o r c=2, t w o result sequences would be transmitted. The first one could be obtained by applying the "exclusive or" operation (which i s t h e addition when speaking o f bits) t o respectively: • t h e source sequence, the source sequence delayed by one bit period, and the source sequence delayed by 3 bit periods; • t h e source sequence, the source sequence delayed by two.bit periods, and the source sequence delayed by 3 bit periods. The transmission b i t rate after coding i s twice the original information rate and is then redundant. Each information bit is "sent",6 times, each time as a part of a sum with two• other information bits. This multiplicity of sending will be used by the receiver to check the received flow, and possibly recover the original information f l o w even i f transmission errors occurred. The condensed form used to describe these recipes consists in using polynomials, here respectively D3+D+1 and D3+D2+1, where "+" stands for the exclusive-or and D for a delay of I bit. This method, as described, has two drawbacks: i t is adapted to infinite sequences, and it results in an efficiency ratio (ratio between the number o f useful bits in the source sequence and the number of bits actually transmitted) which is limited to fractions of the form 1/c. Two modifications of this basic scheme are used in GSM. First, the source sequence consists of finite blocks of p hits. They can be coded as described above, but with the addition of a few known bits at both ends of the block, called tail bits' . Second. an efficiency of the form ply was needed in one case, and is obtained by puncturing the code, i.e., by keeping only q bits among pn, through a pre-determined rule. For instance, i f we want to apply the code described above to a block of b=12 bits, we could add 3 bits positioned to "0" at the start and In fact, in the Specifications. the term tail hits is reserved for those added at the end; the first bits arc specified to he 0 by the specification of the initial state of the encoder. However, the difference is only superficial. and the presentation here keeps the symmetry. 243 fill'. RADIO INTERFACti GSM SYSTEM Delay=1 bit period Delay=2 bit periods Delay=3 bit periods 1st cony. sequence (15 bits) 2nd cony. sequence (15 bits) punctured 2nd sequence (8 hits) transmitted block (23 bits) 1001001 1010 1 0 0 0 1 0 0 1 0 0 11 0 1 0 1 0 0 . 0 0 0 01 0 01 0 0 I I 0 I 0 I 0 0 0 0 0 0 1 0 0 1 0 0 I I 0 I 0 10 0 0 0 0 0 1 0 0 1 0 0 11 0 1 0 1 0 0 0 II 0 0 I 0 001 0 0 I 0 0 I 10 I 0 0 1 0 1 I I 1(11 I I 1 1 0 0 1 1 1 1 I I 101011x10001 1001 101011 Table 4.6 - Example of a punctured convolutional code Every bit of the first convolutional sequence is kept. but only every other bit of the second convolutional sequence is kept. The convolutional sequences are obtained by applying the "exclusive or- operation to the sequence and its delayed self, after addition of fixed hits at both ends ()I' the original sequence. at the end of the sequence and therefore get bc=30 bits. In order to get a convolutional efficiency of. e.g.. 2/3. every other hit o f the second sequence could be omitted. The coded result uses 23 bits (see table 4.6). The actual efficiency coefficient i s slightly smaller than the convolutional efficiency, because of the tail hits. and amounts in that case to 12/23. The convolutional codes used in GSM are not much more complex than the one described above: the maximum degree o f the corresponding polynomials is 4. Now. convolutional codes of larger degreeshare usually much hetter performance. at the price not of bit rate but o f decoder complesit>. In GSM. hosek er, simple codes Bari been chosen because the performance gain is limited by the interleaving depth. The vinut of a code can be understood in broad terms as the length of the sequence of coded hits which must be analysed t o decode one information hit. Now . i t can I s h o w n . ttilh the assumption that bursts are mostlt all good or all had. that lisle gain is obtained if the span N greater than the interleaving depth. thc (ism o l t a i o n a l codes hen e h e r o chosen :Is simple as possible. w i t h a span just b i g enough fc.g.. 2 2 1111- the 9.6 kbit/s data transmission mode). 4.3.2.4. F i r e Code A Fire code is a conventional linear binary block code. that is to say (in simplified terms) it consists in transmitting in addition to the information bits a number of redundancy hits computed by exclusive-or formulas applied to the information hits. Moreover Fire codes derive from codes belonging to the "cyclic" code family (to he precise. the Fire 244 T H E Ti II. RAN( ) INTERFAcE GSM SYSTEM code used i n GSM i s a shortened cyclic code). I n this case, the redundancy computation formulas can be expressed as follows. The original sequence can b e used t o build a polynomial ( i n binary arithmetic), whose coefficients are the bits o f the sequence. The redundancy can be expressed as the coefficients o f another polynomial obtained as the remainder of the division of the polynomial representing the sequence by a pre-determined polynomial, characteristic of the code and called the generator polynomial. A Fire code has a generator polynomial designed to allow good detecting and/or correcting performance when errors happen in group4. The GSM Fire code uses the following generator polynomial: (X23 + 1) ( X " + X3 + 1) This polynomial being o f degree 40, the remainder has 40 coefficients, and then there are 40 redundancy bits. The properties of this code are such that error groups o f up to 11 bits can be detected and corrected. 4.3.2.5. P a r i t y codes These are linear block codes, derived, like the Fire code, from cyclic codes. Three o f those codes appear in the Specifications. For speech, a 3-bit redundancy code enables assessment of the correctness of the bits which are most sensitive t o errors i n the speech frame. Its detection capability is quite limited: all one-error patterns are detected, but there are patterns of two errors which are not detected. For the RACH and SCH bursts, codes of respectively 6 and 10 bits of redundancy are aimed a t detection, though they have adequate properties to be used for error correction (they have a minimum distance of 4, which means that they detect all patterns of 3 errors, and that they could be used to correct all single error cases). For those interested, w e g i v e here their generator polynomials, and the - factorisation thereof: RACH: X 6 + X S + X 3 + X 2 + X + 1 = ( X + 1 ( X 5 + X 2 + I ) SCH: X l " + X s + X ( ' + X S + X 4 +X2+1 = ( X 4 + X 3 + X 2 + X + 1 ) ( X 3 + X + 1 ) ( X 3 + X 2 + 1 ) 245 4.3.2.6. Decoding The Specifications only describe in detail the way in which the signal has to be transmitted. They do not impose explicit decoding methods. Still, many decoding algorithms exist, using different. more or less complex methods, or having different goals. The same code can he exploited at reception for error correction or for error detection. Usually some level o f detection o f the remaining errors is possible after error correction, but the performance of detection is much better i f no error correction is applied. The Specifications include however some constraints f o r the decoding process (and more generally on the reception chain): these constraints are provided as a minimum performance specification. For instance, they impose that the parity codes must he used only f o r detection. On the other hand, it is not so clear that using the Firq code for grouped-errors correction is needed to reach the required performance. For convolutional codes, the required performance is quite severe. There are many decoding algorithms which can be used in general. but here the choice seems limited. A first point is that it seems necessary that the demodulator provides for each bit not only the indication whether it is more likely a "1" or a "0" ("hard-decision"). but also its opinion about the probability that it was a "1" or a "0" ("soft-decision"). There is more information to be used in the second case by the decoder. A n obvious case, quite relevant in GSM, is the indication of an erased hit. that is to say a bit about which nothing can be provided by the demodulator (e.g.. because the whole burst was so interfered that i t was not possible to demodulate). In this case a hard-decision demodulator would provide garbage not indicated as such to the decoder, whereas a soft-decision demodulator would indicate the "garbageness" (the bit would he given as " I " or "0" fifty-fifty). It seems that some extra information beyond .harddecision is needed in GSM to reach the required performance. There are still many decoding algorithms f o r convolutional codes able t o use likelihood information provided by the demodulator. but it appears that only a maximum likelihood decoding (i.e.. finding the sequence o f highest likelihood) fits the requirements. 4.3.2.7. Interleaving and Coding: a Summary of Transmission Modes (X5+ I)( X2 +I) (X+1)(X+1) The latter code is a cyclic code. It is the best (35,25) code as given in the tables of Error Correcting Codes from Peterson (see bibliography). Table 4.7 summarises the interleaving and coding schemes used for the different transmission modes. The first column gives the channel and. where relevant, the transmission mode. T h e "input block" column indicates the size (in bits) o f the data block before channel coding. 246 T H E GSM SYSTEM TnE ItA1)10 INTERF.xcii 2 4 7 whereas the " o u t p u t b l o c k " c o l u m n indicates t h e r e s u l t i n g c o d e d block length (also in bits). I n the " c o d i n g " c o l u m n , codes are g i v e n i n the same order as they are a p p l i e d d u r i n g coding. D e c o d i n g proceeds o f course in the reverse order. 4.3.2.8. T h e TCH/FS Transmission Mode To a v o i d o v e r b u r d e n i n g t h i s b o o k a n d t h e reader, w e w i l l detail o n l y one case, speech. T h e table above gives the needed i n f o r m a t i o n f o r the other cases. every 20 ins). When errors occur, the received speech is disturbed in different ways depending on the role o f the bits in error. This is why it The i n p u t r a t e f o r t h e s p e e c h t r a n s m i s s i o n m o d e o n a f u l l r a t e traffic channel i s equal t o 13 kbit/s, b y b l o c k s o f 260 hits (i.e.. one b l o c k was decided to protect the 2 6 0 bits o f a source b l o c k in d i ff e r e n t ways: • 1 8 2 b i t s are protected b y a convolutional b l o c k c o d e w i t h a convolutional e ff i c i e n c y equal to 1/2; Channel and transmission mode Input Input rate block (kbit/s1 la TCH/FS lb 13 II Coding 50 Parity (3 bits) Convolutional 1/2 132 Convolutional 1/2 78 none TCH/F9.6 TCH/H4.8 12 6 Convolutional 1/2 240 punctured 1 bit out of I5 TCH/F4.8 6 Addition of 32.null bits Convolutional 1/3 120 Output p block Interleaving • t h e 78 other bits are not protected. On 8 half-bursts 456 Complex, on 22 unequal burst portions is X3 + X + 1. B i t s o f class I a are o f such i m p o r t a n c e that. i f any o n e o f them is w r o n g , the user w i l l hear a— p o t e n t i a l l y l o u d — n o i s e i n place o f a Complex, on 22 unequal burst • portions an extrapolation o f the preceding block. 1 1 1 The p o l y n o m i a l representing the detection code f o r category l a bits 456 72 Convolutional 1/6 456 On 8 half bursts TCH/H2.4 3.6 144 Convolutional 1/3 456 Complex, on 22 unequal burst portions 25 Parity (1(1 hits) Convolutional 1/2 78 On 1 S burst 2 4Parit' t63 hits) 6 Convolutional 1/2 last associated signalling on TCH/F and /II TCH/8. SACCH; BCCH, PAGCH On I access burst On 5 half bursts 184 20 m s speech slice. D e t e c t i o n o f such e r r o r s i s therefore i m p o r t a n t a n d allows the bad b l o c k t o be replaced b y s o m e k i n g less d i s t u r b i n g such as The c o n v o l u t i o n a l code consists i n a d d i n g 4 bits (set t o " I n t o the 3.6 RACH N. handover access) Coding 456 TCH/F2.4 SCH • a m o n g these 1 8 2 bits. 5 0 a r e additionally protected b y a detection code a d d i n u 3 redundancy bits ( w h i c h are themselves protected b y t h e a b o v e c o n v o l u t i o n a l c o d e ) . T h e s e 5 0 h i t s are the category l a hits; the other 132 hits are category l b bits; Fire code 224/154 Convolutional 1/2 456 On 4 full bursts Table 4.7 - In erleaving and c )ling for the different transmission modes initial 1 8 5 b i t s e q u e n c e . a n d a p p l y i n g t w o c o n v o l u t i o n s w h o s e polynomials a r e r e s p e c t i v e l y D 4 + D3 + I a n d D 4 + D3 + D + I . N o puncturing i s applied, a n d t h e result i s c o m p o s e d o f t w i c e I N b i t s . i.e.. 378 bits. T h e e f f i c i e n c y f o r class I b i t s i s j u s t b e l o w 0 . 5 . T h e 7 5 n o n protected h i t s m u s t o f course b e a d d e d t o these 3 7 8 b i t s . l e a d i n g t o a coded b l o c k length o f 456 bits. _ Interleaving Blocks of full rate speech are interleaved on 8 bursts: the .156 bits of one block are split into 5 groups o f 57 bits, each one earned by a different b u r s t . A b u r s t t h e r e f o r e contains t h e c o n t r i b u t i o n o f I w o successive speech b l o c k s A a n d B . I n o r d e r t o d e s t r o y t h e p r o x i m i t y relations between successive bits. b i t s o f b l o c k A u s e the e v e n p o s i t i o n s inside the burst and bits o f block 13 the o d d positions. _ 248 THE GSM SYSTEM 8 448 even bits of burst N I 9 449 even bits of burst N+1 I() 450 even hits of burst N+2 3 I 1 451 even hits of burst N+3 4 12 5 13 452 451 odd bits of burst N+5 6 14 454 odd bits of burst N+6 455 — — 57 columns— — — > odd bits or burst N+7 7 15 249 I I II I< \ lilt) IN II.WI' WE odd bits of burst N+4 clear text: O l O O I O I ciphering sequence: 0 0 I 0 ciphered text: (1 I I ( I 0 I I 1 0 0 1 1 [ 0 0 1 1 1 I I t) I 0... I I... Table 4.9 - Ciphering and deciphering mechanism An -exclusive or" operation is perlonned hemeen the clew' \ I and the ciphering sequence. Table 4.8 - Interleaving algorithm for full rate speech A speech block is split into 8 groups of 57 bits as shown. and each of these groups is carried in a separate burst. The major drawback o f interleaving is the corresponding delay: transmission time from the first burst to the last one in a block, taking into account one additional burst for the SACCH, is equal to (9x8)-7 = 65 burst periods, i.e., about 37.5 ms. Interleaving proceeds as shown i n table 4.8. The left column indicates the position o f the bits, number in order 0 to 455, in coded blocks, and the right column indicates their position in bursts. A burst therefore carries 116 bits of coded data as follows: • 5 7 bits from a given block B (odd bit positions); • 1 bit indicating whether this half-burst contains user data or la used in fast associated signalling mode; • 5 7 bits from block B+1 (even bit positions); • 1 bit indicating whether this second half-burst contains user data or is used in fast associated signalling mode. Signalling blocks have the same length as coded speech blocks, and a block in fast associated signalling mode matches exactly the place of a speech block. 4.3.3. CIPHERING Among the various advantages of a digital transmission system, the easy protection o f data against a non-authorised third party i s an important feature for the user. Such protection has been introduced in GSM by means of transmission ciphering. An important point is that the ciphering method does not depend on the type of data to be transmitted (speech, user data. signalling). but is applied only to normal bursts. Ciphering is achieved by performing an -exclusive or" operation between a pseudo-random bit sequence and 114 useful bits of a normal burst, i.e., all information bits except the two stealing flags. The pseudorandom sequence is derived from the burst number (known through synchronisation mechanisms) and a session key established previously through signalling means (see Chapter 7). Deciphering follows exactly the same operation. since "exclusive-oring" I tw.ce with the same data leads back to the original value. Table 4.9 shows an example of such an operation. The algorithm used to generate the pseudo-random sequence is called " A 5 - i n the Specifications. The cipher specialists claim that protection would be just as good if the AS specifications were known by everybody, because of the type of algorithm and because the key can he changed f r o m communication t o communication. However, a s a n additional security step. the full specification is not included in the public specifications. It is distributed under certain conditions by the association of European operators which have signed the GSM MoU. The algorithm —1-rfairly simple and can easily be implemented in a single VLSI chip by manufacturers of mobile stations and/or base stations. 4.3.4. MODULATION We arrive now at the last o f the steps. the modulation and the demodulation. We enter a reahn where inzahemztties reign even more ruthlessly than in the previous sections. In this section we will present the modulation used on the i n t e r f a c e of GSM. and an example of how demodulation can be done. For any expert i n signal processing. and maybe for others, to say that the modulation is GMSK with BT = 0.3. with a modulation rate o f 270 5 / 6 k bands. and that a possible demodulation would use a Viterbi akorithm. is the end of the story. We 250 T H E GSM SYSTEM 251 will elaborate slightly, in particular for people who are not experts in the domain but who want t o understand what hides behind this jargon. However, this is not so easy, and we apologise that equations appear in this section. I I n general and mathematical terms. for any modulation fitting in a small bandwidth, the electric field generated at a gi\ien instant may he represented by: From a very general point of view, radio emission is based on the generation o f a propagating electromagnetic field. This field will be detected through its effect on remote conductors. The emitter controls certain characteristics of the field (amplitude, frequency, etc.), following pre-determined rules, These characteristics will more or less be kept during propagation; and the task of the receiver is to determine the most probable characteristics imposed by the emitter by looking at the detected field. coo represents the carrier angular frequency (i.e.. 2tt times the frequency). which depends on the channel and on time. a(t) and (p(r) are respectively the signal amplitude and the signal phases. Both a(t) and tp(t) must vary fairly slowly (compared with w„ t), in order for the resulting spectrum to occupy the allocated frequency spectrun? or less. Modulation is the function which imposes the characteristics to the electromagnetic field based on a set of rules and the data to be transmitted (which "modulates" emission). In the case of GSM it is the phase of the electromagnetic field w h i c h carries t he information. Demodulation studies the received signal and tries to determine the data used by the emitter to modulate the field. It is usual to distinguish the modulation and demodulation on one hand, and the emission and reception on the other. The first processes transform binary data to and from a low frequency modulated signal, and the second pair o f processes transform this low frequency modulated signal to and from the electromagnetic field. This distinction corresponds to the use of different technologies. However, the border depends on implementation choices and there is no absolute functional meaning which would describe this border independently from the technical state of the art. The description given in this book will therefore group those two functions as a transformation between data on one side and the electromagnetic field on the other side. 4.3.4.1. T h e Modulated Signal Let us start with some generalities concerning modulation. The basic band available for GSM contains two sub-bands around 900 MHz. The multiple access technique calls on frequency multiplexing, the central frequencies being only 200 kHz apart: the spectrum o f the modulated signal around a given carrier determines how it interferes with signals in adjacent bands. E ( 0 = a ( o c o s ( u ) d +tp(i)) The modulation definition must give the relation between the data to transmit and the emitted field EU). In GSM. wo i s not a function of data, but derives from the multiple access rules. The amplitude a(t) does not depend on the data either. The amplitude during the active part of a burst depends on the emission power and may vary according to the power control algorithms, but not with the contents of data. At both ends of a burst, a(t) must follow a given ramping curve in order to avoid spurious emissions due to sharp changes between emission and silence. The G S M modulation scheme i s therefore a "constant envelope" modulation scheme, i n so 1 4 only the central part o f the burst i s concerned. The contents of data only impact the signal phase qv). More precisely, t h e modulation chosen i n G S M i s G M S K (Gaussian Minimum Shift Keying). with BT = 0.3. The formula giving the phase a t a given instant relative t o a n infinite b i t stream ... representing t he data i s ( B T appears o n l y i n the computation of parameter a n tp(t ) = + k,(1)( t— iT) with the following definitions: {k, =1 i f d, = d k, = —1 o f d 4',d c13(xl ) = —II(Mx + —I i - G( 1.— H 1 1 - ) The signal phase is tun an tihNointe concept, nut depends On a coin onion R. define the origin. 252 T H E 153 Ha, GSM SYSTEM T = 48 /13 with 11171; = 0441684 2710.3 a G(.v) = 2Ea e +dt V 2 a e m 2a- No may take any value; it is not specified and is therefore unknown to the receiver. To be precise, the formula of (1)(t) is obtained as the convolution o f a ramp of width one bit period and o f height ,t/2 (an M S K ramp) b y a Gaussian function,of parameter s, hence the name Gaussian M S K . There are many formulas t o express function 0 . We have chosen this one (though it does not seem very simple) because it contains no double integral. V 1bit duration The function 4)(t) is shown in figure 4.28. Basically, it consists of a it/2 step, smoothed in order to have a more narrow spectrum than if the step was steeper (if the step was instantaneous, the modulation would be OQPSK—offset quadrature phase shift keying—, and the spectrum larger; the MSK corresponds to a ramp o f width one bit time). This reduction o f the frequency spectrum has some counter-effects, such as intersymbol interference...the signal at a given moment does not depend 1bit changed Figure 4.29 — Effect of one bit on the modulated signal — Two bit sequences differing by a single bit are used to modulate a carrier at a frequency of 0.25 times the baud rate one cycle every four hitsi. using GMSK and the GSM step function cDtti: t h e effect of the change is negligible outside a w indow of 3 hit periods. on a single data bit, or correlatively each data hit influences the signal during a period exceeding the hit duration. Theoretically, a modulating symbol ki influences the signal during an infinite period. Practically. though. this influence becomt negligible outside a period lasting 3T. since: 3T 3 T ( 1 ) ( — - ) = it — —(13(-) = 0.003. 7 -3T -2T -T 0 T 2T 3T In order to illustrate this. figure 4.29 shows modulated signals for two modulating sequences differing by a single hit. Figure 4.28 — 0(t) in the GMSK modulation (Nu is basically a "rounded" slen. whose room is ,:nr.••.(1,..,”.• 1 a value very small compared to — . ,,, 254 T H E 255 VI Ili I4:\1)10 INTIt121. \ GSM SYSTEM This modulation is basically a frequency modulation. This can be visualised easily by taking the examples of two series of information bits d, , one of constant value and another one of alternate bit values. dB When d, is constant (i.e., all d, are equal to 0, or all d1 are equal to 1), all the ki are equal to I. A particular property of GMSK is that the phase of the resulting signal varies then linearly with time: 9(0 = ( p + t —IT) = ( p+ p Et 2T This is a sine wave of frequency j = t o 1 +) . 2t 4 T Similarly, a sequence of alternate modulating bits (0 1 0 1 0...) results in all kiequal to 0, and in a signal whose phase may be written as: -500 - 4 0 0 - 3 0 0 -200 - 1 0 0 0 1 0 0 200 300 400 500 kHz cp(t)= yo — — i T ) Figure 4.30 - (i SISK modulation spsietruin = (Po — 2 T This is a sine wave of frequency f2 = (—w° - —1. 2 rt 4 T ) The GMSK modulation has been chosen as a compromise between a fairly high spectrum efficiency (for radio waves...) o f the order of 1bit/Hertz, and a reasonable demodulation complexity. The theoretical modulation spectrum (calculated on an infinite random sequence of modulating bits) is shown in figure 4.30. It can be noted that the choice of a channel separation o f 200 kHz results i n a non-negligible overlap between the spectrum in adjacent frequency slots, quite higher than that __usually tolerated in radio. This source of interference can be limited by careful frequency planning, aiming at separating geographically the usage of adjacent frequencies. Since emission proceeds b y independent bursts a n d n o t continuously, the behaviour at each end of a burst must be specified. The number of modulating information bits in a normal burst is equal to 142. By convention, the modulation scheme considers that this sequence is preceded and-followed by an infinity of "I"s. The signal amplitude for a normal burst must follow the pattern shown in figure 4.23 (page 2331. This amplitude falls to 0 outside the time window allocated to a burst. The spectrum of the G\ISK modulation used in GSM is shown here for two adjacent central frequencies separated h 2 1 ) kl The overlap is not negligible and frequency planning MUNI take the cited of adjacent channel interference into acctiunk 4.3.4.2. T h e Modulator A typical implementation of the modulator includes Iwo successive stages. The first stage. YY likth is often referred to as the modulator by itself. crates a signal defined as above. but with a fixed small value of co„ (e.g.. 7 2 MHz). T h e second stage. called frequency transposition. transforms this intermediate signal in order to transpose it at the correct burst frequency and t o adjust its power t o the desired level before delivering it to the antenna. Frequency transposition relies on the fact that a signal may keep its phase properties while being transposed in frequency by multiplying it by a sinusoid function at a proper frequency and then filtering the result to eliminate the unwanted parts. The equation may he written as: cos(2mfer + j(t))•cos( 2ft( - f , II I= cos( 2rej;,/ + » - cos( 2g1f„ 2 LI/ + rill) 256 T H E GSM SYSTEM The choice of frequency dictated by the multiple access algorithm is introduced during this transposition. 4.3.4.3. T h e D e m o d u l a t o r After propagation through space, the field received by the remote antenna can be measured and an electric signal regenerated from it. However, the received field differs from the emitted field measured near the emission antenna, which itself is not quite equal to the theoretical emitted field. The differences include: • a variable attenuation, both because o f free space loss (depending on the distance between the emitter and the receiver through an inverse law) and also because o f environmental shadowing due t o the presence o f "masks" along the way (buildings, hills, etc.); • multipath propagation, when t h e signal i s reflected o r diffracted by different obstacles. This leads t o reception of several copies of the signal, slightly shifted in time one with another; • addition of spurious signals and noises, such as thermal noises, spurious emissions from other emitters, and most o f all interferences from other GSM emitters using either the same frequency band at the same moment (co-channel interference) or a n adjacent band (adjacent channel interference a t ± 200 kHz). The demodulator must estimate the most probable sequence of modulating data, given the received distorted signal. In order to help the demodulator to perform this task, a portion of each burst results from the modulation o f a pre-determined sequence which i s known b y the receiver: the training sequence. This sequence allows the receiver to estimate the distortion of the signal due to propagation, in particular as -•far as multipath is concerned. There exist numerous demodulation algorithms. The Specifications do not impose one such algorithm, but impose global performance figures measured after correction of the errors by channel decoding. Data about these minimum performances are given f o r a variety o f conditions concerning the environment (urban, suburban, rural, ..., which determines the average nature of the obstacles), the speed, the transmission mode, and so on. One important constraint is that the algorithm must be able to cope with two multipattis of equal power received at an interval of up to 16 ps, i.e., almost four bit periods. In such a situation, the amount of THE RADIO INTERFACE 2 5 7 `intersymbol interference is dramatically it3creased compared to what is introduced by the modulation itself. S i p l e demodulation techniques cannot cope with such intersymbol interference and equalisation i s required. An example of a demodulation technique with equalisation, chosen among many acceptable for GSM. is given hereafter. It describes the socalled Viterbi method. One of its advantages is that it makes use of an algorithm which is also fit f o r maximum likelihood decoding o f the convolutional codes, a n d s o opens t h e possibility t o implement demodulation and decoding with common parts. Viterbi demodulation is a maximum likelihood technique, that is to say, a technique which finds the most probable emitted sequence, taking into account some assumptions on the possible signals. and on the noise statistics. I t relies on the knowledge o f a finite set o f possible signal shapes received during o n e b i t period. Because o f intersymbol interference, the received signal depends on more than one modulating bit. I f this signal depends on 3 bits, S different shapes have t o be considered. Since the complexity o f the Viterbi algorithm increases exponentially with the number of bits taken into account. this number rarely exceeds 5. A bit period in GSM lasting about 3.7 ps. this figure of 5 is enough to take into account two paths separated by l b ps. The demodulator has first to evaluate and to store the 32 reference signal shapes associated with the 32 combinations of 5 successive hits. This is done on a burst basis by an analysis o f the part o f the signal corresponding to the training sequence. The assumption about the noise is that it is Gaussian and white. Now, let us take a candidate original sequence, and reconstruct the corresponding signal using the reference signal shapes. At a given instant. the probability that the noise explains the difference between the reconstructed signal and the received one is an exponential of minus the square sif this difference. The overall probability that the noise explains the differences all along the part to demodulate is the product o f t h e _ _ elementary probabilities. The result i s that the probability o f some original sequence given the received signal decreases with the integral of the squared difference between the received signal and the reconstructed signal. I n GSM the demodulation i s done on a burst basis, and the integral is computed on a finite time interval corresponding ton bits. The next point is that, an integral being a sum, each segment in time, in particular each one-bit period, contributes additively t o the function Ito minimise. Each o n e -bit-period segment depends b y 258 T H E GSM SYSTEM assumption on only 5 of the bits we are looking for. then we can write the value ' to minimise as a sum TI Ili RADIO INTERFACE 2 5 9 TS GSM 05.02 describes the channels in terms of the time aspect •and of the frequency hopping characteristics. I t also gives the digital structure of bursts. TS G S M 05.03 specifies t h e different error correction and —detection codes applicable to different uses of the channels. where ri(c/j, ai+1, aj+2, aj+3, c5+4) represents the contribution to the integral of the one-bit-period part of the signal influenced by bits a j to a +4, as computed from the reference elementary signals and the received signal. The aim consists in finding the a; values leading to the minimum of this function. Each one o f these a; appears i n 5 different terms (the functions corresponding to 5 successive instants). In order to solve that problem, a mathematical method consists in modelling i t b y a graph o f 16// points whose co-ordinates are (i,±1,±1,±I,±1). The "±1" values represent the two possible states for each modulating bit. ' is interpreted as a distance between the starting point 1=0 and the end point i=n-1, and the algorithm has to find the path through the graph which minimises this distance. The distance between two points in the graph is one of the terms of •, that is to say: ,a,,, ,a/÷2,a,+3,a,+4) for possible successive points, i . e . , p o i n t s o f t h e f o r m (i,ni,a,+1,ai+2,a,.+3) and (i+1,ai+1,ai+,,a1+3,ai+4) and is infinite for any other couples o f points (since they would not correspond t o a potential sequence). Looking for the shortest path is then equivalent to minimising the expression called . above. Finding shortest paths in a graph is a classical problem, o f wider application than modulation, and efficient solutions exist. W i t h t h i s presentation, t h e Viterbi demodulation algorithm appears as an application of a classical shortest path algorithm. SPECIFICATIONS REFERENCE Within the Specifications, the 05 Series is devoted to the subject of transmission On the radio interface. TS GSM 05.01 is a general description, introducing the major concepts. TS G S M 05.04 is a 3 page document specifying the GMSK modulation. TS GSM 05.05 is related to "radio transmission and reception". The bulk of the TS is devoted to the transmission performances. It deals also with the frequency aspects, with the emission power (including the burst envelope), and with the electromagnetic compatibility problems. This is were the frequency usage is defined, including the limitations concerning the energy leaking out of the right frequency band (spurious emissions). Conversely, the specifications request receivers to accept some level o f out o f band energy (sensitivity). These topics are not addressed in this hook (it is thick enough), but are nevertheless very important since they have impacts on the cost of the radio equipments, especially the mobile station. TS GSM 05.08 and TS GSM 05.10 are related a little to the topics in this chapter, but their contents are really dealt with in Chapter 6 (Radio • Resource Management). Another TS worth noting outside the 05 seriesiis TS GSM 04.03, which defines the channels from a usage point of vieW, and tackles a little the notion of cell channel configuration. I HE USM SYSTEM SIC iNALLING TRANSFlit / 6 1 :011 '4 0 1 : , 4 1 1 } p I SIGNALLING 5.1. The Needs -./b. 541. Contiguoyi Entities r. ,1,3,ProtocoilAt4frorldng 5.2. Linking ' " 1 5.2.1. Structuring in Frames .5.2.2.•Segmentakoiand Re-Assembly 5.1,p.-Erref1jetgttPri, and Correction Mult1p4Attg,-, 1.4113 anclhAPThn frames: a Summary , 5.2.7. RLP ChectEdsties 5.3. Networking:. 5.3.1. Networking in the BSS 5.3.2. Networking in the NSS 2 5.3.3. Networking for Supplementary Services Management2 5.3.4. Networking for Point-to-Point Short Messages 3 0 Specifications Reference SIGNALLING TRANSFER In the two previous chapters the focus has been put Oil the transmission functions close t o the physical medium. and, we mainly studied the means to transport user information. But user information is not the only thing to be transported in a complex network such as GSM. Most functions performed by such a network are distributed over several distant machines, and information exchanges are needed to co-ordinate what is done in these machines. We will address these exchanges from two points o f view. One is simply t o present what these -exchanges signify, why messages are sent, and what reactionsthey induce. This will be the subject of Chapters 6. 7 and 8. The other point o f view is simply how the messages are transported from one point i n the network t o another. These transmission aspects call f o r a number o f additional functions in comparison to those presented .11I t h etwo previous chapters. and will be the topic of this chapter. In all cases, the information needed for the co-operation of distinct entities, i n other words the signalling information. is organised into messages. The sending of a message is triggered by some event. and its reception triggers a chain o f events. A typical elementary message consists o f some "message type", which indicates what reaction the message will trigger in the recipient. and some qualifying information. under the form of mandatory or optional parameters. f O n e o f the tasks o f the transmission protocols i s t o provide :Message delimitation out o f bit streams. Another is to guarantee a very low level o f undetected errors, since such errors can have important consequences. e.g.. changing the meaning of a message into another one. —__Jhese frictions are part of the link layer functions. We will address in 262 T H E GSM SYSTEM this chapter a variety of link layer protocols, akin to the famous HDLC, including the link protocol adapted to the GSM radio interface. Another aspect o f signalling transfer is the organisation o f the message flows, and their routing: I f the exchanges between machines were to be done by speech, a network such as GSM would appear as a very noisy crowd, a source of an immense cacophony. Everybody would be speaking to almost everybody else at the same time, using other people as intermediaries to overcome the problem o f distance. There is then some need for organisation, independent from the actual significance of the exchanged information. Two main aspects can be identified. The first is the routing problem, i.e., how messages are passed from one point to another until they reach their final destination. The second is how to use references to handle several dialogues in parallel. These are the main aspects of the network layer, which will be the topic of the second part of this chapter. There we will see how messages are carried between the mobile station and the MSC. We w i l l also visit the realm o f the Signalling System Number 7 , a packet data network designed for signalling exchanges. In fact, our subject-matter is not limited exclusively to signalling information. Although most user information i n GSM is o f a circuit nature, there are a few user services which are basically of a non-circuit nature, such as short messages. Such services are very much akin to signalling messages i n so far as they both require the same kind of transport mechanism. They will then be studied here. Another special case is the RLP, the Radio Link Protocol which was presented at the end of Chapter 3. Because of its relationship with link layer protocols, it will be presented in a bit more detail here. Though the link and network protocols form the backbone of any system composed o f co-operating communicating machines, their detailed understanding is not a pre-requisite for the understanding of the next chapters, which deal with the co-operation itself. siGNALLINGTRANsmt 2 6 3 One may then wonder why there should be so many different methods when all of e m basically fulfil similar purposes? There is no simple answer to this t estion, but some sensible reasons do exist. One of them is the search for' optimisation, which applies in particular for the radio interface, on which traffic load is critical. Another reason is the . reuse of existing standards. For instance, the CCITT Signalling System / number 7 (SS7) forms the basis for signalling exchanges between the machines o f the Network and Switching Sub-system (NSS), since i t would have been very uneconomical to choose anything else in a switch environment. But, beside these arguments. the historical background must also be recalled. Each interface was effectively designed b y a different standardisation group or sub-group. with different antecedents, and this situation certainly contributed to the variety of protocols one can find in GSM. When trying to understand the needs for signalling transportation, 'let us first remark that each entity in the system may communicate with many others. There are cases when the two communicating entities are contiguous, i.e., directly linked through a physical interface. but this is not the general rule. We will start by listing the communication needs between contiguous entities, before proceeding to the relaying cases. 5.1.1. CONTIGUOUS ENTITIES Starting with the mobile station and unfolding the transmission chain, signalling exchanges within the scope of GSM are required on all interfaces, i.e., between: • M o b i l e Station (MS) and Base Transceiver Station (BTS); • B T S and Base Station Controller (BSC). in particular to enable the BSC to control the BTS; • B S C and Mobile services Switching Centre (MSC). for example for co-operation in the area o f communication establishment and handover: 5.1. THE NEEDS Both signalling messages and short messages require a packet switched system to transfer them. Packets can be stored, combined. segmented, multiplexed, and tortured in many ways along their route. They can be carried over interfaces in a variety o f ways, and this is actually what happens in GSM, where virtually each segment along the transmission path bears its own stack of protocols. • M S C and a point o f entry into an external network, for the establishment of communications. Looking at the internal architecture ()I' the 'nubile station leads to two additional cases, between: • Terminal Adapter ( TA I and Mobile Termination ( M T ) , i n particular for the establishment of communications: 264 T I I E GSM SYSTEM • S I M and Mobile Equipment (ME), f o r the exchange o f subscriber related data. In addition, exchanges must take place between NSS entities (MSC/VLR, GMSC, HLR/AuC, EIR). These entities may be physically interconnected by direct links. This may make sense for neighbouring MSCs for instance, but for an MSC to have direct lines to all HLRs in foreign networks would be very uneconomical, and administratively complex. So, most of the time, NSS entities are connected to intermediate nodes which are part of the general SS7 network. Messages between any two pairs of NSS entities are transported through this network, even when the communicating entities are in different countries. In the canonical model, we w i l l consider that each NSS equipment has no physical interface with other NSS entities, but instead with one or several entry points in the SS7 network. Another source of signalling needs is the Operation Sub-System (OSS). The OSS is a distributed network of its own, and as such it needs internal exchanges. This topic is outside the scope of the Specifications and operators have the freedom to make independent choices for the transfer of signalling inside their own OSS. However, OSS equipment must b e able t o give orders t o the G S M transmission machines (subscription management, configuration), and t o receive information from them (observations, alarm reports, charging records). Hence the BSC, MSC/VLR, GMSC and HLR/AuC have to be in contact with an entry point into the OSS. There again, operators are to a large extent free to make their own specifications in that domain, but the subject is tackled a little in the Specifications. More instances of direct signalling exchanges may be found, such as the HLR-AuC interface for instance, but they are not specified by the Specifications. - 5.1.2. RELAYING There are a number of cases where signalling messages must be transported between distant machines, i.e., where i t appears more economical not to install direct communication means, but instead to use intermediate nodes for transit. In such cases, the intervening nodes act as simple relays, which forward the messages onwards. These relays may have to adapt the messages to the transportation requirements of the different interfaces. They may also have to choose the output direction, SIGNALLING TRA 2 6 5 depending on an address provided by the sender of the message: this is the routing function. All these functions pertain to the network layer of the ISO layering model. GSM often uses the term "transparein- in this context, in a way which requires some explanation, since i t might be misleading. This concept of "transparent" has already been encountered in Chapter 3. and is used in a similar way for signalling, referring to the transmission of some data by an equipment without disturbing it. and in fact without having to even understand the meaning of the transported information. It can therefore he said that an equipment is "transparent f o r such data. in the same way as a transparent piece of glass will let light through without stopping it. The comparison can he pushed further: light can change its course when going through some transparent object. depending on its wavelength, and the direction of the message can be influenced based on its address. The way in which the Speed' ications may lead to contusion is when they refer t o a message being transparent. rather than t h e equipment. I n this case, what is meant is that the message is really "opaque" to the equipment! Even though the "transparent- concept and image are useful in understanding protocols. it must he treated with care because of the twist of terms in the SperUi(miems. Let us now proceed with the requirements for relaying between GSM entities. Most o f the exchanges between distant GSM entities involve the mobile station at one end: MS-13SC for the management o f radio resources, MS-MSC/VLR f o r location a n d communication management and MS-HLR f o r the maiia ement o f supplementary services. The transmission chain is linear and the BTS. I3SC and MSC act as relays for these signalling flows. The case of signalling exchanges between NSS entities has• already been touched upon in the previous section, and there all the nodes in the SS7 network act as intermediate relays for S57 messages. Another area where relaying is needed is the transfer o f short messages. Basically. short messages are exchanged between the mobile — station and a Short Message Service Centre (SCI-SC'). The SM-SC is in contact with one or several MSC% in the NSS. which act as gateways. A mobile terminating short message must then transit through this gateway. the visited MSC and the BSS (13SC + BTS), before reaching the mobile stintion. A . mobile originating short message will experience the reverse path. In the O&M domain, the BTS is not in direct contact with OSS entities, but exchanges between BTS and 055 transit through the BSC. which therefore also acts as a relay for these messages. In addition to this list of exchanges, all pertaining to signalling and to short message services, there is one case o f circuit-type user data where communication protocols are used inside G S M above the transmission functions o f the physical layer. This concerns the RLP, which has been introduced i n Chapter 3 when dealing with "nontransparent" (NT) connections. RLP frames are exchanged between the Mobile Termination, inside the mobile station, and the InterWorking Function (IWF) between GSM and external networks. RLP frames therefore cross the BTS, BSC, TRAU and MSC. But these entities do not in this case act as active relays: they are only the "transparent" support of an established circuit, so that for NT data the mobile station and IWF can to some extent be considered as contiguous entities with the meaning used here, and RLP belongs indeed to the same family as other link layer protocols used on single-hop interfaces. 267 SIGNALLING IltANtil+.1. THE GSM SYSTEM layer N message semantics layer N 1 layer (N-1) t ' A Intermediate Node (a) relaying B 41( interworking at the semantics level layer 5.1.3. PROTOCOL INTERWORKING layer (N-1) 1 The reader may wonder why logical relations between entities, such as e.g., the HLR and the SIM for maintaining consistency of their respective data, were not treated as such in the previous list of signalling exchanges. This situation corresponds t o cases where no signalling message is actually transmitted by one entity (say A) to another (say B). This does not mean that no information is exchanged between A and B, but that this information is transported as components o f messages, and not as complete messages. The difference seems subtle, but it is in fact important in terms of specifications. Figure 5.1 shows how this difference can be modelled. When messages are transferred without being tampered with between A and B, the intermediate entities act as carriers (they are transparent!): this is message relaying. In the other case, the intervening entities are involved to a much higher degree and one then speaks of protocol interworking. A first message. transmitted by A , triggers the ..sending by an intervening equipment of one or more messages conveying part of the information carried by the original message towards a third destination, and so on. There is some information transfer between A and B, but only at a semantic level. The correspondence between the content of the successive messages is then much more elusive than the identity of bit patterns. This can only be studied when the behaviour o f the intermediate nodes is looiced at. This will indeed be the subject of the next chapters of this book, where signalling functions will be described topic by topic and will take a complete system view. A Intermediate Node (b) interworking B physical entity P r o t o c o l Figure 5.1 - ttela)ing versus iniemorking In case (a). messages of laver N Iransil belueen A anki H. and the intermedtaie node zu.a. as a rela • i.e.. it is not i n v o k e d in the semanne, o f ihe me.sare. In case lb). the intermediate node k an M i e n \ orking Function. It is concerned with the semamics of the I a) er N message.. from :\ and issues different mes‘ages tow aid. II and vice ‘ersa. For the time being, let us start by looking at the basic methods which enable messages to be transported b‘ each individual protocol. The basic transportation means between cool; .guous entities w i l l first b e looked at, illustrating various implementations for the linking aspects (ISO link layer functions). Then the relaying aspects(IS() network layer functions) will be studied. both within the BSS f r o m the mobile station up to the MSC) and within the NSS. Leto T H E tilt INAI T I O N S I - 1 R GSM SYSTEM 5.2. LINKING The link protocols used in GSM are not the same on all interfaces. The ones which will now be described are summarised in table 5.1. They represent the major protocols used by GSM signalling in the linking area. In the mobile station itself, the interfaces between S I M and Mobile Equipment, a s w e l l a s between Terminal Equipment and Mobile Termination, should for completion be cited as well. In order not to be drawn into too many details, they will not be studied here. These three protocols (LAPDm, LAPD and MTP 2) have very similar functionality. The presentation w i l l therefore be done on a functional basis, along with explanations of the differences between the cases. The reader will therefore be able to gain a better knowledge of their functions, which should help in the understanding of the individual protocol's behaviour. For a detailed interface-per-interface specification, the reader is referred to the Specifications, as well as to relevant ETSI and CCITT recommendations. I n the case o f the Message Transfer Part (MTP) in particular, GSM has not introduced any new functionalities, but has used the protocol as defined in the ETSI specifications, themselves referring extensively to the CCITT recommendations of the Q series. The examples i n this chapter w i l l mostly b e taken f r o m the LAPDm protocol—and to some extent from the Link Access Protocol for the ISDN "D" channel (LAPD) protocol from which it derives—which is, among the examples shown, the link protocol with the most original features. In all cases except for the radio interface, signalling messages are sent over plain 64 kbit/s circuits. But between MS and BTS, the physical medium is very specific to GSM. On this interface, the transport of pointto-point messages can be done in two ways, as explained in the previous chapter: L i n k protocol MS-BTS L A P D m ( G S M specific) BTS-BSC L A P D (adapted from ISDN) BSC-MSC M S C N L R / H L R - S S 7 network MTP, level 2 (SS7 protocol) Table 5.1 - Link protocols on GSM interfaces Despite their apparent variety, the link protocols used on the G S M interlaces have similar functionality. 6 9 • using the main channel. if necessary by preempting the resource (Fast Associated Signalling). This stealing method typically consists of not transmitting one block of user data (representing 20 ms). and using the corresponding resource t o send a signalling message instead; • using the Slow Associated Control Channel (SAC'Cli). Leet us now consider how the data is structured on the link layer to be transported on these physical media. 5.2.1. STRUCTURING IN FRAMES The prime functionality o f a l i n k layer i s t o structure the information to be transmitted on the channel in units bigger than a single bit. The resulting atomic units are the basic structure on which all link layer functions work. In the signalling world. such a unit is called a frame. The whole issue consists in including sufficient information in the bit stream so that the receiver is able to find out the beginning and t h 4 end of each frame. Both LAPD and MTP 2 are the heirs of 11DI.0 in this area, whereas LAPDm makes use of the synchronisation scheme of the radio interface to convey information on frame limits. In HDLC, frames start and end with an eight-bit long pattern called a flag (see figure 5.2). To prevent false starts or ends, a mechanism ("0" bit insertion) is introduced to disguise the flag pattern when it occurs inside the data. The resulting hit stream only contains the "legal" flags. flag (frame end) flag (frame start) 01111110 I Interface 2 frame contents 01111110 I V V last bit sent first bit sent Figure 5.2 111 ) 1 . 0 frame Hags Frames start and stop ss lib a defined pattern called a flag. The figure applies both l o t I_API) and W I ' 2. Between transmitter and recek cr. a i s added to each sequence o f 5 consecutive -I6's in the data. in order to disguise the nag pattern ‘ ‘ hen it appears inside the 270 T H E The flag mechanism allows frames to be of variable lengths, without even the need to indicate the actual length value inside the frame. The same flag can be used to indicate both the end of one frame and the start of the next one. In LAPDm, the use of flags (and the corresponding octet waste) can be avoided thanks to the ready-made blocks of the physical layer. In order to benefit from this advantage (and also for error detection reasons), the choice was made to fit each frame into a single physical block, which is 23 octets long. As a consequence, a LAPDm frame has a maximum length of 23 octets on all TCHs, and of 21 octets on the SACCHs (this difference comes from a specific usage of 2 octets per SACCH block, for timing advance and transmission power control, as will be explained in Chapter 6). But the effective information length may be smaller than this maximum value. A length indicator is therefore included in each frame. Unused octets are filled with a default value, which happens to be "00101011". This value has been chosen to minimise the probability that a frame with little information (and hence many fill octets) results in a burst similar to the Frequency Correction Channel (FCCH) bursts, which would disturb mobile stations attempting to synchronise. For historical reasons, the value "11111111" may still be used as a filling pattern for uplink transmission by mobile stations. 271 \1.1.1\GTR. \\*FR GSNI SYSTEM upper layer message 1 segmentation Ill H frames to be sent 1 1 1 direction of transmission frames received 111 F7, re-assembly upper layer message n h e a d e r and trailer of eat* link frame 5.2.2. SEGMENTATION AND RE-ASSEMBLY The maximum length of frames may be limited as a result of lower layer transmission constraints, o r m o r e generally t o ease t h e dimensioning of buffers in the system. When the maximum length of a signalling message exceeds the maximum length allowed for frames, this message must be segmented and sent over several frames. Conversely, the message must be re-assembled at the receiver end. To do this, the receiver must receive enough information to know how to reconstruct the messages, and this causes additional overhead in the protocol. When it is not foreseen that the maximum length of signalling messages will exceed the maximum length of frames. then the segmentation and re-assembly mechanisms are usually dispensed with. This is actually the case on the A interface. The maximum length of frames on the A interface is limited to 272 octets of information (plus 6 octets for frame control, excluding Ilags). This maximum value has been inherited from the SS7 world, where it replaces an older value of 62 2 t i l l bits Figure 5.3 — Segmentation and re-assembl) o f messages in I.:Wpm The "more" bit or 1.:\ I'Dm frames enables Ow receiver to reciiiiNirtici l i t e o r i g i n a l m e s s a g e e t / I I C O t e l l a t i l l g t h e CkblItelltS O f f r a m e s until there are no more frames for this me..; 22e ("more" bit set to Ill. information octets. The value is enough to accommodate most signalling needs, and no segmentation is lob:seen at this level. AN a consequence. many GS\I protocols have inherited an tipper bound for its message length which stems directly from this value. Taking this constraint into account. Mere is no need to define a segmentation and re-assembly facility on the Abis interlace eillicr• The length o f the A h i s L A N ) Innics i s simply limited t o 261 octets texcluding flags), which corresponds t o 260 octets o f upper layer information. THE GSM SYSTEM Things arc different on the radio interface. where the maximum length of a frame is either 21 or 23 octets. Such a length is clearly not sufficient for most signalling needs. As a consequence, a segmentation and re-assembly facility is defined in LAPDm. It makes use of a so-called "more" bit, which distinguishes the last frame o f a message from other frames, as shown in figure 5.3. Thanks to this mechanism, there is no intrinsic limitation on the length o f radio path messages, and the only constraint comes from the need to transfer such messages on the other interfaces, hence the 2 6 0 octets mentioned i n th e radio interface specifications. *)1(;\ \I 11 \ i i l l t 273 and a counter is incremented and decremented according to the validity of each block. A link failure is reported to the upper layers ‘‘ hen the counter (whose initial value R.lino J./.\ i n i E o t 7 is set by the operators reaches zero. Next collies the frame acknowledgement and repetition function. which supports a very high performance in terms of removing remaining errors. None o f the three protocols studied Aire makes use o f forward error correction capabilities (such features are usual' considered to be in the physical layer). All three use IDLC-like mechanisms for backi‘ard error correction, with a choice between two modes: 5.2.3. ERROR DETECTIONAND CORRECTION • t h e non-acknowledged mode. in which frames are transmitted once. whatever the outcome at the receiver end: The next important functionality of a link layer is to improve the quality of transmission, by detecting frames which have been subject to transmission error, and then possibly asking for repetition of the frames in error. • t h e acknowledged m o d ' . ensuring correction o f erroneous frames by repetition. As far as error detection is concerned, both LAPD and MTP 2 use the HDLC scheme which consists in adding 16 redundancy bits (called FCS, or Frame Check Sequence in LAPD) to each frame, according to a coding scheme chosen for its error detection characteristics. In both protocols, the same generator polynomial is used for calculating the 16bit signature: .061-.02i- x5+1. On the radio path, LAPDm can dispense with such a mechanism, thanks to the error detecting performance of the transmission coding scheme offered by the physical layer. Error detection serves two purposes. The first one is to provide enough information on the likelihood of residual errors in a frame, so that repetition of the frame can be asked for. The second one is to monitor the quality of the link, in order to trigger the relevant alarms when the error rate exceeds some given threshold. Let us consider how monitoring is done in GSM, before describing how the repetition mechanism works. Link quality monitoring on SS7 links is done in such a way that the link is declared out o f order when the error rate exceeds some value (typically, frame error rate ?_ 4.10-3). Let us remark that a SS7 link is never inactive when in service: specific filling frames are transmitted when nothing else needs to be transmitted. Error counting is therefore reliable and watchdog timers can b e defined t o monitor the link behaviour. On the radio path, a similar situation (regular transmission) exists on the SACCH. This channel is then used for quality monitoring, The two modes coexist in all three protocols. One may wonder what i s th e use o f the n o n -acknowledged mode. i f better error performance is available using the acknowledged mode. lint the choice is not completely black-and-white. F o r example. f o r t h e recurrent measurement messages sent by the mobile station over the radio and Abis interfaces, the non-acknowledged mode i s more adequate than the acknowledged one. Not only does a loss not cause any great damage. but furthermore, it is better for the receiver to get a new (and up-to-date) measurement report in this case rather than to receive a repeated (and obsolete) message. So measurement reports are actually transmitted using the non-acknowledged mode on. the radio and Abis interfaces. A few other messages are also sent unat'knowledged. such as the answer to a handover access (e.g.. the kit.3-RR puysicAt, r o w a v r i o N messages) and the general information sent by the network to mobile stations on the SACCH (e.g., the RIt.3-RR SYSTEM INFORMATION T i m 5 AND 6 messages—all these messages come from procedures cited in Chapter 6). not to mention all the messages sent on a point-to-nitiltipoint channel (BCCH or PAGC'H ). Apart t h e s e messases. the vast majority uL messages sent on dedicated channels are sent in acknowledgedmode. Acknowledgement and repetition i s based o n a cyclic frame numbering, which enables the recipient t o detect possible repetitions and/or losses of frames. and to acknowledge specific frames. In I.API) and LAPDm, the acknowledgement is done h'. the receiver transmitting the number of the next expected frame to the sender. i n an indicator 274 THEGSM SYSTEM Receiver supervision timer ! 0 acknowledged, 1expected Receiver Sender receiving window sending window 6 1 275 SIli\ \ I I 1 \ (i MANS] Hi 4 3 6 5 timer expiry (1) 1 timer expiry (2) 2 4c- 1 6 5 0 acknowledged, expected 6 5 2 acknowledged, 3 expected 7 4 3 70 4 7 0 4 7 0 6 1 2 Figure 5.4—Repetition mechanism Repetitionlostriggered by the sender in two cases: - when it receives an acknowledgement for a frame which is not the last one sent; - when it does not receive an acknowledgement alter a certain amount of time. 6 7 0 , 1 ack 2 2 Figure 5.5 - Windowvmechanism for acknowledgement Windows represent a sliding -.et of contiguom frame. ‘.1.1tiehcan he either: called N(R). The mechanism is shown in figure 5.4. If the numbering of frames is modulo 8, a receiver expecting frame 2 indicates implicitly that frames 1, 0, 7, 6, ... have been correctly received. In MTP 2, a similar goal is achieved by transmitting the number of the last frame correctly received to the sender. In all cases, the sender is then able to repeat nonacknowledged frames, i f any. However, the total number of repetitions is limited in order to avoid infinite loops when a serious problem occurs. Frames must be kept by the sender until they are acknowledged, so as to be available i f repetition is needed. In order to limit the number of corresponding buffers, as well as to avoid numbering ambiguities, the concept o f a window is used in LAPD and LAPDm. The size of the sending window determines the number of frames which can at any given moment be sent and not yet acknowledged. This window size K must be sent a n d n o t v e t 1 , l e d e r d I • e i i t l i i i g w i n t i t i n t. l i r —accepted for reception t receiving U indow1at a given moment. sufficiently high to enable b y anticipationthesender without mailing for acknowledgement delay. Figure 5.5 slums how the sliding \ \ intim\ mechanism works on the sencicr side as \\ 1 1 1 1 rile rC,C1 1:1 .11.1C. The numbering cycle of I.APD (and M i r 2) is I 28. On I .AI'Dm. a ,mailer numbering cycle t8) \\ as chosen, to reduce the sire of the frame header. In-LAPD. the \\intim\ size can be configured. But in the case of LAPDm, the window size is set equal to I . This choice has been justified by the expected simplification of the protocol implementation: a \\ halo \\ size of 1 corresponds to a simple send-anti-\\ ait protocol. In the case of the small dedicated channels used for signalling (IC1118s). performance Joes not suffer any degradation from this simplification. because the organisation of the channel is or a basically alternate nature. \\ ith enough nine between a week oil Irani,' and the nest ,1111111W0111101.1111111\to 1.1111111 u T H E si(r..-utisa; TRANsi•m< CiSM S Y S T E M 277 SAam Up numbered frame transmission Figure 5.6—Acknowledged mode setting procedures Figure 5.7—Acknowledged mode release procedures The acknowledged mode is set-up on a link by an exchange of two commands, enabling contexts to be reset on each side of the link for a given link flow. Acknowledged mode transmission is normally terminated through a two-way exchange. enabling contexts to he released on each side of the link for a giN en link flow. the answering frame. A s for other channels, transmission will incur additional delay when several frames are to be sent in a row because of this window size of 1. For instance, the signalling capacity of full-rate channel (TCH/F) will be at best one frame every 120 ms, i.e., no better than twice the capacity o f a TCH/8. However, signalling applications suffer little from this restriction, since signalling frames usually become ready one at a time, the major exception being when a message is segmented over several frames. At any time, an unacknowledged frame o f information m a y b e sent. When no frame is pending and transmission should occur. then - f i l l frames" are sent, which consist o f an unacknowledged mode information frame ( " U l " frame, for Unnumbered Information) whose information length is null. In order to initialise the contexts on both sides of an interface for a transmission in acknowledged mode, a simple procedure is used in LAPD and LAPDm. It consists o f two messages, as shown in figure 5.6. In LAPD, the exchange of upper layer information can only take place after such an exchange. But i n the case o f LAPDm, the S A B M (Set Asynchronous Balanced Mode, a term inherited from HDLC jargon!) frame is able to carry a "piggybacked" message repeated in the UA answer (Unnumbered Acknowledge). i n order t o solve potential contention, as explained in Chapter 6. In a similar way to the setting up of transmission in acknowledged mode, the normal release o f a link i s performed through a simple procedure, as shown in figure 5.7. No piggybacking is allowed on the corresponding frames in any of the link protocols. -5-.14. MULTIPLEXING of Thus far, the flow o f messages has been considered to be a single one, i.e., a succession o f frames delivered one after the other by a single source. However the l i n k layer offers the possibility o f multiplexing independent flows on the same channels. These (lows are independent in so far as the ordering o f frames between them is not guaranteed and the window mechanism applies to each flow in a separate manner. In order to distinguish the flows f r o m one another. an address i s inserted i n each frame. Such a mechanism is essential on point-to-1111110point links. such as in the I S D N L A P D . which has been designed f o r user installations with one line and several terminals. It has been kept i n the Abis Though the TA C H s (i.e.. the dedicated channels on the radio path. see SIGNALLING TRANSFER THE GSM SYSTEM TCH/F TCH/8 SACCH "SAPI" Type of 11o‘. signalling (SAPI 0) ack. mode ack. mode non-ack. mode ' short messages (SAPI 3)- — ack. mode ack. mode 0 62 63 Radio signalling Operation and maintenance Layer 2 management Tab e 5.2 — SAPIs on radio channels Table 5.3 — Different flows on the :This interface All channels are not suitable to all combinations of the two SAPIs, SAPI 0 for signalling and SAPI 3 for short messages. In addition to the radio signalling procedures. the Abis interface also carries a flow dedicated to the operation anti maintenance of WISs. as well as a layer 2 management flim directly Mho lied from ISDN acct... Chapter 4 ) are point-to-point links on the radio interface, the link multiplexing mechanism is also provided in LAPDm. On t h e radio interface, t w o independent f l o w s c a n exist simultaneously. The first one is devoted to the transfer o f signalling messages, and the second one to the short message services. These two flows are distinguished by a link identifier called the SAPI (Service Access Point Identifier), This term is not quite proper OSI terminology since a SAPI defines the point of entry from upper layers into the link layer on each side of the interface, and the correct OSI terminology for link identifiers is DLCI (Data Link Connection Identifier). However, in GSM, the "SAPI" refers to the link identifier transmitted in the protocol, and this usage was inherited directly from ISDN access protocols. Nothing else but history can explain the choice of such a mechanism to discriminate functional flows, since the creation of separate layer 2 links was b y far not necessary. The LAPDm "SARI" can take values 0 (signalling) and 3 (short messages). Hence the reference to SAPI 0 and SAPI 3 which appears here and there i n t h e Specifications. All combinations of SAPIs on channels are not possible, and table 5.2 lists the allowed possibilities, together with an indication o f the modes (acknowledged or non-acknowledged) associated with each SAPI on each channel type. The only case in LAPDm where two real independent flows have -16 be transmitted simultaneously with acknowledgement and repetition is on the TCH/8. The choices were dictated both by a will to reduce the number of cases and also to prevent the use of the TCH/F for non-urgent matters, since user data might he disturbed b y pre-emption. As a consequence. the transmision of short messages is fairly slow. The link rate gives a maximum throughput of roughly 80 octets per second on the TCH/8, i.e., the equivalent of a 600 bit/s circuit. We will see in the next section that upper layers further reduce this rate of transfer. Multiplexing on the Abis interface has two facets. The first one corresponds to the distinction between different functions. It is performed 279 in a similar way as on the radio interface. The "SAPI- vat LICS on this interface are listed in table 5.3. SAPI 0 is used for all messages coming from or going to the radio interface. The other application of multiplexing is to provide different links ending in different pieces of equipment inside the BTS: the TRXs (Transmitter and Receiver). This topic w i l l he revisited in the networking section. The discrimination makes use o f another field o f the L A P D link layer address. the T E l (Terminal Equipment Identity). The dynamic management of the TEls is one of the functions of the messages of SAPI 63. No multiplexing is done at the link level (NITP 2) on the A interface. 5.2.5. F L o w CoNTRot, The last issue to he dealt with as far as the link layer is concerned is flow control. When a single link is considered. the usual assumption is that t h e processing a n d hollering capability o f t h e receiver i s dimensioned to cope with the maximum throughput of the link. However. resources are often shared between different flow's, with a handling capacity lower than the sum o f the maximum capacit f o r each o f the flows. One of the goals of congestion control is then to control each of the flows so that an overload on some pail of the system does not cause the overall system capacity to crash to zero, hut instead to try achieving the maximum throughput possible. Bottlenecks may he fairl r e m o t e from the actual sources o r flows. and i t i s essential that congestion situations are reported backwards to control the input load. and ultimately to require the source to slow down. Flow control on each of the segments along the transmission chain is one of the tools which helps to keep the throughput under control. 2S I SIlIN \I 11\0 I It Flow control is in some way provided naturally by HDLC-like protocols, simply by delaying the transmission of the acknowledgements. But this control is only marginal, because i f the delay becomes too long the sender will repeat the frame, which may have counter-effects in congested situations. A n additional mechanism can also b e used, consisting i n a "stop-and-go" control using two commands. Such a mechanism is provided in LAPD and in MTP 2, but not in LAPDm. 0 to 260 octets address control i n f o r m a t i o n N(S) 5.2.6. L A P D AND LAPDm FRAMES: A SUMMARY As an application o f the mechanisms presented in the previous sections, the list o f frames used in LAPD and LAPDm will now be briefly presented, as well as their structure. This review is by no means detailed, since GSM has introduced few specific modifications in this domain apart from the differences between LAPDm and LAPD which have already been introduced. Table 5.4 lists the frame types identified in the two protocols, together with their respective roles. The respective structures o f a LAPD and a LAPDm frame are shown in figure 5.8. The address field contains the so-called "SAN", and in addition, for LAPD only, the address of the destination terminal, since the interface is point-to-multipoint. The control field contains the type of frame meaning role SABM DISC UA DM UI Set Asynch. Balanced Mode Disconnect Unnumbered Ack. Disconnected Mode Unnumbered Infomiation first frame to set-up acknowledged mode first frame to release tick. mode Lick. to e.g.. the above two frames response indicating disconnected mode information frame (non-ack. mode) I Information information frame (ack. mode) RR Receive Ready RNIe RE! HIM** Receive Not Ready Reject FRaMe Reject "you may go on" (flow control), also used for acknowledgement "you should stop" (flow control) negative acknowledgement error hack-reporting Table 5.4 - LAPD and LAPDin frame types Frames belong to three families: unnumbered frames, information transfer frames and supervisory frames. 'l'he "RNR- frame (for flow cannot can be ignored in LAPDm." **The "I:12MR" frame is not used in LAPDm. SAPI MA) LAP D TEI 0 to 21 octets add. cont. information N(S) j LAPDm SAPI Figure 5.5 -1...\11) and I.A1,1)1111r;iine structure LAPDm frames are dab. cd from A P I ) 11%tifies. ‘vithout the ' S . and i t h shorter address and control fields. LAPINn frames do not need flats for s)nehronisaiion. frame, as well as the number of the frame Isending side) and the number of the next expected frame ireceiving side) fur numbered informationcarrying frames. 5.2.7. R L P CHARACTERISTICS Before moving to the next laver up. one should recall that there exists a link protocol in GSM i s not concerned with the transport of messages such as signalling messages or short messages. It is the RI-I'. which has been introduced in Chapter 3 and which concerns the transfer of user information between the mobile station and the I W I T h e major functions of a link protocol. as described in the previous sections. apply 282 T H E GSM SYSTEM for RLP as well, which is yet another HDLC-like protocol. Let us now briefly consider its characteristics. • structuring in frames: RLP frames consist of 240 bits, without a need for frame delimitation, since the alignment provided by the radio transmission scheme is used, as explained in the two preceding chapters for LAPDm; • segmentation and re-assembly: none is provided; • error detection and repetition: a 16-bit frame check sequence is included in each frame. As in LAPD or LAPDm, information can be sent either in acknowledged or non-acknowledged mode. The latter is initialised through an SABM/UA exchange and released through a D I S C / U A exchange. T h e repetition mechanism i n acknowledged mode i s similar t o LAPD or LAPDin, using a numbering scheme modulo 64 and a window size which is usually set to 61 but can be adjusted to lower values. This window size as well as other parameters such as the maximum number of re-transmissions, the timer value for deeming a frame lost and the delay before acknowledging a frame are negotiable through a procedure consisting o f an exchange o f t w o specific " X I D " frames (eXchange IDentification). In addition for the LAPDm protocol, an "SREJ" frame (Selective REJect) enables the receiver to request the repetition of a single frame: • multiplexing: not provided: • I l o w control: RR and RNR frames are used to indicate the capacity of the receiver to accept more frames or not. A peculiarity of the RLP protocol is the possibility to piggyback information on supervisory frames, an improvement compared to HDLC: The same frames (say from A to B) can be used to carry information from A to B as well as control fields related to the B to A information flow. A partial application of this possibility has already been encountered with the N(R) indicator contained in the control field of LAPD (and LAPDm) information frames, used to acknowledge the information frames sent in the other direction. But RCP goes further, enabling information to be carried on all supervisory frames such as REJ, SREJ. RR or RNR. The frame families listed in table 5.4 (see page 280) for LAPD and LAPDm also apply to RLP. The only difference is the addition of the "SREJ" frame and of three unnumbered frames: the X I D frame (which SIGNALLING TRANSFER 283 exists in ISDN LAPD. but is not used in the Abis LAPD). the TEST frame (used back and forth to test the link) and the NULL frame (used as a filling frame). 5.3. NETWORKING The link protocols described above enable the exchange of frames between two entities which are directly interconnected through some physical medium. Now. there are a number o f application protocols which involve t w o entities n o t directly interconnected. Additional transmission functions. described in this section. are needed to provide these application protocols with end-to-end connections for the transfer of the corresponding messages. These connections make u s e o f successive elementary links, as described in the previous section. along a route between the two extremities. Conversely. an elementary link is used for a number of network connections in parallel. with possibly different origins and ends. For example. call control messiiges originating from the mobile station must be routed t o the NISC. whereas radio resource management messages generated b y the same mobile station must terminate their course in the BSC. even though both flows of messages use the same signalling link on the radio path and on the Abis interface. One of the networking functions is the routing. i.e.. the choice of the successive segments composing the route. I n this area two broad techniques are used and we will give examples of both o f them. With datagrams, each message is routed following anal \ %is on its arrival. With virtual circuits, a route is established for some time. f o r the use o f complete dialogues: the route is established by the first message. and the following messages just follow the same route. Another function we will see here is closet related, and consists lit the possibility o f having several independent connections tlyisting i n parallel between two entities. This can he used between contiguous entities a s w e l l a s distant ones. T h e connections correspond t o independent application dialogues. f o r instance to the management o f different user communications. A concept common to all the aspects of networking is the address. which will be a leitmotiv of this section. The network protocols add tags to the messages to discriminate between the different flows. These tags can be addresses identifying the origin or Ihe,destination. or connection references, or route references. They are used to choose a route. i.e.. to 264 forward the message onto the next appropriate segment, or to distribute it to the right application software. Let us now consider how this issue is tackled i n the different subsystems, starting with the BSS. 185 SIG \ I . I I N G T i t \ NSI I IL THE GSM SYSTEM PD function Originidestination CC, SS call control management and supplementarp services management NIS from/to MS(' yanyl MM localion imumgemenl XIS Iromlo N1SC/V1.1( securil n i a n a g e m e t i l 5.3.1. NETWORKING IN THE BSS 5.3.1.1. T h e Mobile Station Point of View From the point of view of a mobile station, the origin or destination of the messages depends on the application protocol. The mobile station addresses different network functional entities, and these addresses are then used by the network to actually route the message to the appropriate equipment (BSC, MSC or HLR). Moreover, several protocols are handled between t h e mobile station a n d MSC. I n addition, several user communications may have to be managed in parallel between the mobile station and the MSC, as we have seen i n Chapter 1, e.g., for the indication of a new incoming call when another already exists. The identification of the link on the radio interface gives the ability to distinguish signalling messages from short messages for any single mobile station, but this distinction i s not enough t o determine the application protocol a message pertains to. It must be complemented by a networking address. T h i s function i s fulfilled b y t h e Protocol Discriminator (or PD). Several PDs are defined (see table 5.5), which were originally introduced just as a functional partition of the messages. As a consequence of the GSM functional architecture, this partition also corresponds to an entity on the infrastructure side, and this is why we can consider t he protocol discriminator a s a n address. T h e case o f supplementary services (SS Protocol Discriminator) raises specific problems which will be dealt with in a dedicated section. The B T S does n o t appear i n t h e l i s t o f partners o n the infrastructure side (third column of table 5.5). This reflects the fact that the mobile station does not have a dialogue with the BTS for reasons other than for link management. There is however a single exception to this rule: during handover, one message is sent directly from the BTS to mobile station, f o r the sake o f speed (see the handover execution procedure in Chapter 6). The protocol discriminator is specified as part o f the application protocols (in TS GSM 04.08), though it belongs to a sublayer common to several protocols. The PD is inserted by the originator. In the mobile station to infrastructure direction it is used by the BSC to decide whether RR radio resource management XIS tionno BSC Table 5.5 —Protocolyser.nimators on the lathy, int,1 Lye,. Three protocol discriminators are defined. which correspond to an originffideslinUlion on the Mir:Ishii:lure side. it is the destination (RR) or i f the message has to be forwarded to the MSC (all other cases). The PD o f received messages is used by the mobile station and the MSC to distribute them t o the right software module. Now, this is not sufficient to discriminate him\ een CC and SS messages pertaining t o different user communications. T h e t e r m transaction is used in this context, each transaction corresponding to a communication. I n fact, transactions also exist which are used f o r supplementary service management. In all case', distinct transactions can exist i n parallel. Messages pertaining t o different transactions arc distinguished by a Transaction Identifier (TI). The TI is inserted by the I originating entity (MS or MSC). and is used by the other to relate the message to the right context. The specifications related to the TI are part of the specifications of the RR protocol. which appears in this case as below the MM and CC protocols. 5.3.1.2. A b i s Interface The messages exchanged over the :this interface signalling links have many different origins and destinations. From a functional point of view, a first split identifies the messages between IITS and 1.3SC ( for BTS control) from all other messages transmitted between the mobile station and any infrastructure entity beyond the BSC' (including between the mobile station and the BSC). Going a step further. dieresis a need to UNfinguish the different mobile stations, which is done by or stinguishing thedifferent radio channels (the relationship between mobile stations and radio channels is managed by the BSC. not by the BTS). The specifications have added another dimension to this functional iew, which is the explicit discrimination on the Allis interface hem een DIEGSM SYSTEM sub-entities of the BTS, the TRXs (Transceiver and Receiver). The notion of the TRX has already been presented in Chapter 4, in the context of the organisation o f the radio channels i n a cell. I t was then an abstract concept, which becomes concrete when the Abis protocols are looked at. Each TRX in a BTS corresponds to one or several signalling links, as defined i n t h e section dealing w i t h "linking". These links are distinguished by "TEIs" (Terminal Equipment Identities). The TRX a message pertains t o is not identified b y a tag i n the message, but implicitly by the signalling link which conveys it. This small detail forbids any implementation where a link is shared between different TRXs, and so imposes the concrete existence of TRXs. Still, on each given TRX-BSC link, there remains the need to distinguish the general management messages f r o m the messages pertaining to specific radio channels, and for the latter to discriminate between the different channels a TRX manages. In order to reach these goals, each message o n t h e A b i s interface carries a "message discriminator", together w i t h complementary data, giving enough information to know what to do with the message in the BSC (for uplink messages) or in the BTS (for downlink messages). Table 5.6 lists the different message discriminators, the complementary data, and their use. For instance, the "radio link layer management" discriminator is used for the messages which come from o r are t o be transmitted t o mobile stations. Such messages on the Abis interface also carry a reference to a radio channel (which determines the mobile station), and to the link which is used on this channel. The channel indication contains the type of channel (TACH/F, TACH/8, BCCH, etc.), the time slot number and, when appropriate, the sub-TN of the channel. Note that this information • Message discriminator + complementary data Radio link layer management + channel reference + radio link reference 2N7 SIGNALLING 'FIZANSI,LI: Communication end nodes Use MS-BSC or beyond Relay of radio path ' ' messages From BSC to liTS From IVES to BSC Use ESTABLISH REQUEST ES1ABLISII INDICAIn IN ESTABLISH CONFIRM link establishment DATA REQUEST DA"I A INDICATION acknowledged info. Hansfer UNIT DATA REQUEST UNIT DATA INnicAmrs: non-acknowledged information transfer RELEASE REQUEST RII.LkAsE INDa Atlir. RIII.IASE CoNFIRM link release ERROR INDICATION link error notification Table 5.7 - Abis Radio link managem Ili messaees The messages relayed by the BTS for communication between the BSC' tor heyondi and mobile station are carried on the Ahis interface in at envelope containing the p c of message as shown here. as well as information on the radio path channel and link. is sufficient to determine the channel only in association with the TRX identification carried a t the l i n k level. T h e radio l i n k information indicates the LAPDm link on which the message is to be sent or has been received, and discriminates between "SAPIs" 0 or 3. and between the TCH and the SACCH when applicable. For messages for which the BTS acts as a "transparent'. relay (the radio link layer management messages). the additional envelope used to carry these messages o n the A b i s interface i s very similar t o t h e , primitives between a network layer and a link layer in the ISO model. The corresponding types o f messages are shown in table 5.7. This is presented here as an example of pure relaying protocol. The BTS can be considered as a remote radio link layer entity of the IISC. 5.3.1.3. A Interface The A i n t e r f a c e i s used f o r messaoes b e m c c t i B S C Dedicated channel management + channel reference BTS-BSC Interworking for a given TACH Common channel management + channel reference BTS-BSC Interworking for a given BCCH or PAGCH/RACH TRX management BTS-BSC Control of TRX status Table 5.6—Message discriminators on the Abis interface Four different messagegroups are defined between BTS and BSC, the first of which corresponds to messages transmitted or received "transparently- on the radii) interface by the BTS. \ a s well as for messages to and from the mobile station (using the ( o r NI M protocol discriminator as defined i n table 5.51. These Iwo Bows are rckITed to respectively as BSS M Al' (13SS Management Pout and I YEA P (Direct Transfer Application Pam. I n addition t o this distinction. BSSMAP messages which are of a general nature for the BSC most he distinguished from messages specific t o a connection with a 'nubile station, and the latter must he identified between themselves. For this last function, a virtual circuit approach w a s chosen. \ \ ith independent connections established and released. The choice was t o use an SS7 protocol, the SCCP (Signalling Connection Control Part I. which p r o v i d e s this function i n nitilition t o many others. ( h : ilisiinct ion h e m een TIlli GSN1 SYSTEM SIC N A 1 . 1 . 1 \ G I t . \ \ BSSMAP D T A P BSC MSC VLR distribution, layer, SCCP MTP 3 MTP 2 MTP 1 Figure 5.9—Signalling message transport on the A interface TheA interface makesuse of theSS7N1TP-SCCPstack of protocols to transfer signalling messagesbetweenBSC (or different mobile stations) and MSC. •BSSMAP and DTAP is fulfilled by a small "distribution" protocol on top, as shown in figure 5.9. SCCP is not a GSM-specific protocol, although GSM is one of the first implementations making use o f SCCP, together with intelligenl networks. SCCP is part of the signalling system n°7 stack of protocols] and when used appears just above the MTP protocol. However, SCCP ii not the basic network protocol in SS7. In fact, MTP includes more than the link protocol (MTP 2) described in the previous section. It is the combination of SCCP and MTP 3 (i.e., the network layer of MTP) which offers a networking service on the A interface. Let us now consider briefly the characteristics of these protocols. In a similar fashion to our -earlier description of the MTP 2, details of the protocols will not be dealt with here, and the reader is referred to the SS7 specifications. MTP 3 MTP 3 includes several aspects: one of them is the management of the SS7 network (management of traffic, channels and routes), another being the actual routing of messages inside an SS7 network. In fact, MTP is used on the A interface to transfer messages between two contiguous entities (the BSC and the MSC), and then these networking functions are of no use. But M I T 3 also fulfils some functions of local application. They are related to the possibility, which is offered on the A interface. to install several physical links rather that a single one to transport the messages ( a set o f such links i s ()ailed a " l i n k s e r i n t h e SS7 terminology). The main interest in so doing is the redundancy. which enables the network to keep things going i f one link breaks. NITP 3 provides some functions t o manage linksets. a s f o r instance t h e possibility to split the signalling load between several licks (this is "load sharing"); and the possibility to send messages on a fallback link instead of the "normal" one used within a linkset. e.g.. 11w maintenance purposes. without loosing the sequencing o f messages. as well as die reverse operation. SCCP As all SS7 protocol, SCCP has many functions. Only a small part of them is used on the A interface, and we will see other SCCP functions when dealing with the NSS. SCCP defines several classes of services. of which o n l y t w o are used o n t h e G S M A interface: t h e basic connectionless mode (which in the case of the A interface equates more or less to no added function!) and a connection-oriented mode without such goodies as flow control. expedited data. or reset. These modes are referred to respectively as the class 0 and class 2 SCCP services. The class 0 mode is used on the A interface for the messages not directly related to a single mobile station. such as reset or overload indication. The class 2 mode enables separate independent connections to he set-up. a function used on the A interface for distinguishing transactions with individual mobile stations. Such SCCP connections are established when the need appears, either by the I3SC (this is the general case when a signalling transaction such as a location updating or call set-up_ is started) or by the MSC (in the single case when handover needs to he performed to a new cell). They are released when the need disappears. The BSC manages a context per connection. where it stores the BTS and the radio channel used t o communicate with thfe reit. \ ant mobile station. This enables the BSC to make a one-to-one correspondence between messages from/to the BTS and using a given TACH on one side. and to/from the MSC and using a given SCCP connection on the other side. A similar context is kept by the MSC. where each SCCP connection is related to the identification of a mobile station. DTAP/BSSMAP Discrimination Still, the use of individual per-mobile station connections does not itismibintiiii ! h . , . \ answer all the needs 0 : I ;YU T H E BSSMAP messages (between BSC and MSC) and D TA P messages (between MS and MSC) may refer to a given radio connection, but their handling i s different a n d m u s t b e distinguished. A n additional "distribution function" has been introduced on top of SCCP to this avail. It consists basically in adding a small header to the application message ' before transferring it on a given SCCP connection. All messages on the A • interface therefore bear a discrimination flag, indicating whether the message is a BSSMAP or a DTAP message. As explained earlier, this usage means that the BSC acts as a transit node between the mobile' station and MSC f o r DTAP messages, whereas BSS MAP messages originate or end in the BSC. In addition t o the discrimination flag, D TA P messages carry information on the type of link on which they are to be (or have been) transported o n t h e radio interface. T h i s "Data L i n k Connection Identification" or DLCI matches the link identifier used on the Abis interface for the same purpose. Its main role is to distinguish what is related to the short message services from the rest. 291 suiNAt.t.usc.rit.‘xsukt4 GSM SYSTEM functions of MAP/E: its sole role is to carry signalling messages in a noninterfering way between the mobile station and the anchor MSC. The two . MAKE messages which are involved, PROCESS f u s s sioNAti.iNG and FORWARDACCESS SIGNALLING, a r e shown f i g u r e 5 . 1 0 . T h e i r information field contains the message exchanged between the anchor MSC and the mobile station, with the same contents as when i t i s —Ifinsported on the A interface as a DTAP message. or on the :This interface as a Radio link layer managtment message. for on the radio interface itself. The distinction between the message flows pertaining to different mobile stations is not done by using SCCP. as on the A interface. The SCCP protocol is part of the lower layers used by MAP/E. hut only in connectionless mode, as explained in the next section. The discrimination function is fulfilled by the TCAP protocol. as explained in general for the NSS in the next but one section. The relay MSC has then to maintain a fontext for each connection with a mobile station, and to translate back and forth between SCCP references (towards the BSC') and TCAP transaction references (towards the anchor MSC). 5.3.1.4. Networking on MAP/E The previous sections have described how the different BSS., interfaces are specified i n order t o transport signalling application messages between MS, BTS, BSC and MSC. However, there is yet another interface which messages between mobile station and MSC must cross in some conditions: this is the interface between the anchor MSC and the relay MSC. These concepts have not yet been introduced, and will be dealt with in Chapter 6. A short explanation is then needed. When a call undergoes a handover between two cells connected to two different MSCs, the entry point of the call in GSM is not changed. The MSC in charge 'of the call entry point is then stable throughout the call, and is., called the anchor MSC, whereas the other MSC, the one in charge of then .yMSC.For d isa b o rtm clh w e the management of the call, upper layer messages must transit all the way from/to the mobile station to/from the anchor MSC. with the relay MSC acting as an intermediate node. An additional segment of transportation is therefore required in this case, and will be dealt with here. In the Specifications, this transportation function is not identified as a separate protocol, but is specified as part of the MAP, more exactly of the MAP/E protocol (the MAP/E protocol i s the one between neighbour MSCs, and i s concerned w i t h a l l aspects o f inter-MSC handovers). As such, it obeys the rules for the transportation of MAP messages which will be described in the next section. However, the nature o f the transportation function is very different from the other Anchor MSC MSC VIP I S PROCESS ACCESS SIGNALLING FORWARD ACCESS SIGNALLING m e s s a g e from MS message to MS Relay MSC MSC VLR message I rom MS message to MS \ \ * \ 855 Figure 5.10 - Tran‘pori inechankm on \lArTh. Oncea handoverhaNoccurred. radio inc‘.agesMUNIby rely cd itin.p;ffinnl>In therelay MSC to enure then tralhonh d h . 292 relay MSC anchor MSC MSC VLR MSC VLR, BSC 1Q 3 rit.ANsi,FR THE GSM SYSTEM iini_02essit-stera; xtrtr:App t . ! • LITP 2 it I r I d 3 1 T h a r r i I AT P 1 radio interface Abis interface A interface M M T P T P 1 E interface mi Figure 5.11 - Signalling transport structure in the BSS The structure of protocols used between mobile station and anchor MSC for the transport of signalling messages shows a large variety. — a IIt I 5.3.1.5. C o n n e c t i o n s in the BSS: a S u m m a r y I Two figures are provided as a summary of the message flow on the • BSS interfaces. Figure 5.11 shows the overall structure of the networking protocols. Figure 5 . 1 2 ( o n t h e previous page) summarises the organisation of the message flows from MS to MSC, the different links, transactions, connections. a n d t h e i r identifiers. discriminators. "' addresses, references.... Figure 5.12 - Connection identifiers in the BSS (Olt The structure of protocols on the BSS interfaces leads to a hierarchical organisation of message flows shown as nested boxes, each with its own tag. Several mobile stations are shown in this example. The first one of them (MS a) is the only one where all the levels of detail are given: this mobile station has two calls in progress (T1=a and bon PD=CC, on SAPID), and one short message transaction (TI=a on SAP1=.3). It uses the same TRX as mobile station b. hut they use different SCCP connections on the A interface, and different TC.AP contexts on the F. interface. gjj I c 2 0 t .1 (7, .12Lgi ULNA 2 2 2 PD: protocol discriminator for RIL3 prolocois TI: transaction identifier for RIL3-CC protocol DLCI: data link connection icientilier SAPI: service access point identifier on the radio interface TEL terminal equipment identifier on the Abis inzefface 294 T U E 295 SIGNALLING TRANSFER GSM SYSTEM 5.3.2. NETWORKING IN THE NSS national network The methods of carrying messages between NSS machines are in some way much more standardised than on the BSS interfaces, since in all cases the SS7 signalling network standards are used. But roaming imposes that messages be exchanged between entities belonging to different networks, operated b y different companies i n different countries. The addressing and routing scheme is therefore o f prime importance here. (country of A) international SS7 network national network (country of B) I 5.3.2.1. T h e SS7 Network Protocols A fundamental point to bear in mind about SS7 is that there are two network levels. The lower level, used typically to build national networks, is based on the MTP 3 networking protocol, whereas the othei level, used for interconnecting all the national networks in a single global SS7 network, is based on SCCP. The rationale behind such a choice it simple: it is much easier to hold and maintain routing tables covering one national network (i.e., tables for addressing entities under the control of one operator), than many networks including foreign ones. To these two levels correspond two levels of addressing. The M112 J ,4 address, or "Signalling Point Code" (SPC) only holds relevance within a x i limited scope, such as one SS7 national network. Within this scope, the 414. networking functions of MTP 3 are able to route messages with the S as the destination address. M T P 3 follows basically th e datag approach, that is to say that each message contains the SPC of the enti to which it is directed. At the other level, we find the concept of the "global title". This versatile addressing scheme, at a higher level than the SPC, enables th4 identification of any SS7 point in the world. It is used in SCCP, where • provides in GSM the addressing capacity needed for the transport MAP messages between NSS entities. The addressing scheme o f SCCP therelbre warrants a quick description. SCCP messages can be addressed in two ways: i f they are bound for a destination in the same national network, then nothing really has to be added to the information carried by the MTP. However, in this case the SPC is included as the address at the SCCP level. Now, i f the message is bound for another national network, a global title must be used. The global title may be a number with no direct relationship with SS7, such as a number typed in by a user, a PSTN number (E. I 64), a data number ( X.1 2 I L or a GSM subscriber identity (an -4 global title (for B) GA derives SPC(GB) A derives rSPC(GA) • GB derives SPC(B) Signalling'', point Signalling transfer point (STP) A Gateway SCCP function Figure 5.13 G l o b a l title translation When a global title is used as an address in SCUP. it must be translated into ;to actual routing addresses lor the NH? within each network through winch the message transits. IMSI, specified in E.? I 2. see Chapter 71. etc. It d o s not contain explicit information on the way to route the message. and an SCCP translation-- — function is therefore required to derive the relevant MTP address from the global title, at least at each network border. Figure 5.13 shows an example of how a global title is translated into signalling point codes by SCCP gateway functions along the way, and also h‘ the originating node which must identify the SPC o f a relevant gateway within its own national network. In GSM the SCCP address, whether a global title or an SPC. includes a sub-address. called the sub-, \ stem number. which identifies the type of the target entity (I Hit. VI . MS('. FIR ). 296 T H E SRIN:MA.1\G 'IRAN:SI-lit GSM SYSTEM How do the two levels interwork? In addition to the national SS7 networks there exists a n international SS7 network, within which messages are transported using the MTP protocols. Typically within a national network, a few nodes are, in addition, part of the international network and can act as gateway nodes. When a message is bound for some external network, the originating SCCP entity determines from the global title the most suitable gateway node in the national network. The SPC of this node will be the MTP address o f the message. When the SCCP function i n the gateway node receives the message, i t will determine (still from the global title) the appropriate international node to enter in the target network and will forward the message to this node, using its international SPC as the MTP address. Then, the new transit node will analyse the global title, determine the final point, and forward it using the corresponding national SPC as the MTP address. Used in its connectionless mode as for the GSM application, SCCP is a typicil datagram networking protocol: each message contains an SCCP addreit such as a global title, and the scenario described above is repeated again for each message. It should b e noted that t he SPC address c a n have some international meaning when used between transit nodes. I n fact these nodes have two different SPCs, one for use within the national network, the other for messages with other networks. The numbering plan far international SPCs is defined in Recommendation CCITT Q.708. A i international SPC consists of 14 bits, and includes a geographical zone indicator, a network indicator within this zone and a point indicator within this network. Most often, international SPCs are shown as decinial. representations of these three indicators, e.g., 5-010-3 for an Australiat SPC. But such addresses are only significant within the international SS7 network, and are not relevant for the entities of a GSM network (except if some of them have direct access to the international network). h i The use of SCCP for the transport of MAP messages between NSS entities brings little else than the global addressing facility. In particular the capacity of SCCP to manage independent connections, which is used ..on. the A interface, is not used within the NSS. A l l messages use the connectionless mode. 5.3.2.2. T C A P On top of the services provided by SCCP, the MAP makes use of an additional protocol, the "Transaction Capabilities Application Part" (TCAP). TCAP is not usually presented as a transmission protocol (as i t name shows, it is considered as an application protocol). However, TCAP provides the means to distinguish independent message flows, and as 297 MAP r component sub-layer invoke - return result return error reject r transaction sub-layer r- begin. continue. end. abort. unit-data Figure 5.14 T C A P modelling TCAP provides the neeessar Correlation between individual operations. as well as hemcell the structured e \ change, building up a tran.actiun. such is close in functionally to many other protocols we have seen so far. Moreover, i t i s part o f the common platform o f mandatory layers supporting all the MAP protocols. There is no need to describe the functions of TCAP in detail to understand how GSM works. However. the M A P specification refers heavily to TCAP, so that a presentation of the vocabulary and principles is useful for any reader who will he involved with the actual N1AP specification. TCAP is best modelled in two sub-layers: a -transavtion" sublayer, on top of which sits a -component: sub layer. UN piClinCli in figure 5.14. The service offered by the transaction sub-layer consists i n the management of transactions. or dialogues. on an end-to-end basis. This is another example of a virtual-circuit approach for distinguishing several independent flows using in parallel the same transmission means. T C \ P adds a transaction indicator to each message. This indicator enables the Kier end to relate all the messages of the same transaction to a single context. Thanks to TCAP, the MAP protocols need not bother about the means t o link together the different exchanges concerning e.g.. the location updating of a given mobile station. They rely for this function entirely on the transaction capability of TCAP. In particul.n. there ate a 0 St( ;N:1 I .1.ISai TR AN SIT.I2 THE GSM SYSTEM 299 5.3.3. NETWORKING FOR SUPPLEMENTARY - - SERVICES MANAGEMENT number of cases in MAP where important data, such as the identity of the GSM subscriber, is sent only once within a dialogue. The information is then implicit for the rest o f the dialogue. I t is then important when reading the specification of MAP to note in the procedures the beginning of a dialogue (a message sent with TCAP primitive TC-Begin), how it goes on (message sent with TCAP primitive TC-Continue) and its end (message sent with TCAP primitive TC-End). In addition to the transaction management, TCAP also manages the correlation o f individual commands/responses within a dialogue. These operations are managed in the "component" sub-layer. The correlation between a request issued by a MAP entity and the answer to this request is not managed by MAP, but is part of the service offered by TCAP. It is in fact the most important feature o f TCAP. As a consequence, MAP often does not specify a message name for an answer to a request: this answer will simply be contained in the "RETURN RESULT" or "RETURN ERROR" linked by TCAP to the "INVOKE" component containing the initial message. Even when a message name appears i n the textual description of TS GSM 09.02 (MAP), the contents of the message can be found only in relation with the operation name. For instance, the contents o f the "RADIO CHANNEL ACKNOWLEDGEMENT" message cannot be found if the reader does not know that this message is an answer to the "PERFORM HANDOVER" message, and as such is carried in a "RETURN RESULT" component o f the TCAP "PerformHandover" operation. This kind of knowledge is important to understand the MAP specification, and can be found in section 7.2 of TS GSM 09.02. In the rest of this book, we have chosen to use TCAP operation names to refer to the signalling exchanges carried by the tviAP/x protocols. This will enable the reader to have a non-ambiguous reference to the actual information exchanged by the protocol. i • TCAP provides also a grouping/degrouping function. Several 7r E-' operations pertaining to the same dialogue can be grouped inside a TCAP ;-• message For instance, a given TCAP message may contain the result of an operation (e.g., the positive acknowledgement o f a subscriber authentication) while invoking another (e.g., requiring the start o f ciphering). A last point worth noting is that the syntax used for the description of TCAP messages and parameters (and hence used i n the MAP specification) i s ASN. I ("Abstract Syntax Notation I " ) , which is specified in CCITT Recommendations X.208 and X.209. ti One case of signalling message transmission has been left aside so far: the transmission o f supplementary services management messages between the mobile station and HLR. Let us first recall that. even though a supplementary service i s in general fulfilled by the MSC/VIA. its management (e.g., activation/deactivation of the service. interrogation of its status, etc.) is performed by the HLR of the subscriber. It is the HLR which is in charge of eventually modifying the context in the MSC/VLR to keep it in line with the service state in the HLR. This choice of having a single point o f control ( H L R ) ensures the consistency o f data throughout the network. We must consider two legs for the dialogue between the mobile station and the HLR: one from MS to MSC. which makes use o f the mechanisms described in the BSS section. and another in the NSS. using MTP, SCCP, TCAP as described in the previous sections. The disparity between the two worlds made things difficult, at the conceptual level as well as for the intervening MSC/VLR. The conceptual approach in this domain is somewhat fuzzy, and the vision presented here reflects the view of the authors. On t h e radio interface, supplementary services management information can be carried either in standalone messages. or be part o f some call control messages. Here we find the first fuzzy conceptual point. In our view, the call control protocol can he considered as a carrier for supplementary services messages. in the Saint.: way as e.g.. the messages belonging to the MM protocol between MS and MSC are carried on the Abis interface. The comparison w i t h the A b i s interface i s actually interesting, because some Abis messages from the BSC are also "mixed blood" messages, in so far as they contain an order for the BTS as well as amessage to be carried forward. But the radio interface is not the only area in which supplementary.. services raise architectural questions. The NISC/VI.R is another fuzzy area. It acts as a gateway betweih the MS- MSC/VER protocol and the MAP protocol towards the HI.R. As such. it rust perform some analysis of the messages. he it only to know whether messages received from the MLR must he transmitted towards the mobile station o r not. T h e corresponding information can typically he found in the message type. which is for instance the only information which provides the ability to distinguish between messages belonging to different mAtIN protocols. SIGNALLING TRANSF ) J 11:Ni MSC VLFI Lir" „, • : vg.• , rApi•it,,MAP/14" .0.. A l l . 0 er• a i r RIL3-CC 3 0 1 CC), or inside messages using the SS protocol discriminator. Between the MSC and the HLR. it makes use of all the SS7 protocol stack, and its messages are distinguished from other MAP message simpl> b y the message type. Messages pertaining to dill'' N I S s are distinguished on the MAP leg by the use of TCAP transact! s . on the A-interface by the use of SCCP connections, and finally be“veen the BSC and the mobile station by the radio channel. Several SS management transactions can exist in parallel. They are distinguished on the BSS leg by different Tls. and on the MAP leg as different TCAP transactions. I t falls on the MSC/VLR to make all the needed translations! TCAP SCCP MTP Figure 5,15 —Protocol stack for supplementary services management cli J. The actual application protocol between MS and HLR (to he described in Chapter 8).1'1 relies on two different transportation protocol sets. y a one between MS and MSC, the other inside the NSS. ' 1 4 addition to this analysis for routing, the specification is such that the: MSC/VLR checks most of the syntax o f the messages, and may reject them if they do not comply with the specifications. For these reasons, it difficult t o see t h e MSC/VLR a s a pure transit node f o r i l i t • supplementary service management messages. B u t i t cannot considered either that the MSC/VLR makes use of the semantics of t h e . messages: it only deals with them as objects to be transported or rejected,: Taking all these points into consideration, we prefer here to present ' MSCNLR only as a relay on the path between the mobile station " HLR as far as supplementary services management is concerned, and! consider the application protocol (MAP/1) as an MS/HLR protocol, even though it is not carried by SS7 all the way. This approach puts the stress more on the functional aspects (in particular on the fact that the dialogue is between the mobile station and the HLR), rather than on marginal issues such as syntax analysis. ,.Figure 5.15 shows the overall picture: the MAP/I protocol is an, application protocol between the mobile station and the HLR. Between the mobile station and the MSC, its messages are carried as encapsulated messages either inside call control messages (protocol discriminator. 5.3.4. NETWORKING FOR POINT-TO-POINT SHORT MESSAGES The short message services i s t h e second domain where networking encompasses both the BSS and the NSS \solids. In tact it encompasses even more. as GSM interworks for these functions with external networks. As explained in Chapter 2. the role of GSM (or short messages is to transport them between the mobile station and a Short Message Service Centre (SM-SC), the latter being out of the scope of the Specifications, and possibly outside the control of GSM network operators. The S\1-SC is connected to one or more \1SCs which act as ,gateways between the GSM world and the SM-SC. The corresponding functions are called SMS-GMSC in the case of Mobile Terminating short messages and SNii1WMSC (nterWorking MSC) in the case o f Mobile Originating short messages. We will call them both the SMS-gateway for short. Tlw stack of protocols between such gateways and the S\1-SC is left open. What the Specifications specify is a set of transportation means for conveying the short messages between the mobile station and the SN1S-gatewa. The transportation involves potentially t w o domains: t h e M S t o N I SC segment, and the segment between MSC and SNIS-gate". as. The protocol architecture for short message transport inside GSM is shown in figure 5.16. The SM-TP ("Short Message Transport Protocol-) specified between the mobile station and the S\1-SC' is i n fact an end-to-end protocol including some features o f an application protocol. and will therefore be dealt with in Chapter 8. together with other short message management functions. We will concentrate here on the relaying o f the short message between GSM entities. SR THE GSM SYSTEM P i 303 \ 5.3.4.1. T h e BSS Leg The lower layers used to convey the messages from the mobile station to the MSC have been described in the section about linking. On the radio path, they include an acknowledged-mode SAPI 3 link on a TCH/8 or on an SACCH. This link is then relayed up to the MSC using the relay protocol on the Abis interface and DTAP on the A interlace. ism-so I S — S a... as. a . ..... a a m ...le ass. ...s. , , . / . . . . . . e . , . . . . . . - e . . . . . - - . . . . • • • e , . . i n , - - . 0 0 . - a r : . . . . . ...111. • • • • . . . W. SM-TP - _11.1.0_4;stfiWittl, SMS-gateway . r r MSC VLR Figure 5.16- Protocol architecture for the transport of short messages in GSM The relay protocols co-operate to convey a short message between the MS and the point of interconnection of the SM-SC in GSM: MAP/H interconnects with SM-RP, itself relying on SM-CP for transport between MS and MSC. The stack of protocols between MSC and SMS-gateway MAP/H and the usual stack underlying MAP protocols, i.e., (from the bottom up) MTP. SCCP and TCAP. This interface is the most greedy in terms of message overhead, and the maximum length of short messages ' in GSM (140 octets, sufficient coding for 160 7-bit ASCII characters) is then a direct consequence of the MTP maximum frame length (see page 270). We will now consider the BSS leg (from mobile station to MSC) and the NSS leg (from MSC to SMS-gateway) in turn. Since on both legs the Mobile Originating and Mobile Terminating short messages use similar transport mechanisms but in opposite directions, only one of the directions w i l l be described i n detail (the Mobile Originating short message). As a last point in the description of each leg we will identify the differences for the Mobile Terminating case. We now consider the short message "Control Protocol" ISM-('P), which has very little added value ( i f any?) compared t o the service offered by the underlying layers. It is a very simple protocol. consisting of a command/answer procedure with three message types. as shown in figure5.17, and without even a reference to correlate them since its mode of operation is basically of an alternate nature ("send-and- \\ airi. The sts.tCPCP-DATA message is the only one carrying upper layer information. which does not necessarily include the short message itself. but could be some upper layer acknowledgement or error report. Each message of the SM-CP protocol includes a protocol discriminator specific f o r short messages, and a transaction identifier. This last field could enable a mobile station to manage several parallel short message transactions at .the same time. This would then be the only real added a l u e o f the protocol, but the Specifications do not indicate clearly whether it can be used as such. The next protocol on our list is the SM-RP (Short Message Relay protocol), whose functions include the management o f references and MSC VLR 01 CP-DATA RP•message MSC VLR 4 CP-At K or 4 CP-E#OR (cause) Figure 5.17 •--11w SM-('P protocol between XIS and NISC The SN1-CP protocol M r s . an 1 1 1 0 d i O f i l i V i a l 1 0 11 Iselldillg Of the message. then \‘ ail mg lor the dchntmlalgenran with the po,:ibility of the nte,satte being repeated once if a tuna e 305 SIG\ \ I I l \ r i addressing. This protocol is in fact the BSS part of . . protocol for the networking functions between MS and SM-SC. The SM-RP protocol interworks with the MAP/H protocol in the MSC/VLR. Three messages are defined in the SM-RP protocol, one to carry the message (RP-DATA), one for transporting the acknowledgement (RPAcK) and one for indicating an error (RP-ERROR). Though the message list is similar to the one o f SM-CP, the functions fulfilled are quite different. The messages are correlated together through a one-octet message reference generated by this protocol, enabling the sending or receiving o f different messages in parallel at this level. Addressing is performed by including in the SM-RP RP-DATA message the destination address (for an MO message), or the origin address (for a MT message), i.e., in all cases the address of the SM-SC. There are a few differences between the messages for M o b i l Originating messages, and for Mobile Terminating messages. One of them is the inclusion in the mobile terminating SM-RP RP-DATA messagfr of a priority indicator. However, since this priority indicator is not transported in the MAP, it is quite useless unless the MSC tampers with ft upper layers to find the relevant priority information. 5.3.4.2. T h e NSS Leg Between the MSC and the GMSC, the transport of short messages is performed by the same means as signalling messages, i.e., using the SS7 stack of protocols supporting MAP (MTP, SCCP, TCAP), on top of which is added one o f the MAP protocols, the MAP/H. Taking into account the underlying SS7 layers. the MAP/H provides the same functions as SM-RP in the BSS leg. It contains three messages, which can be mapped onto the three SM-RP messages, as indicated in figure 5.18. 5.3.4.3. T h e Relay in the MSC/VLR • A s was the case for the protocol supporting supplementary service management. the MSC/VLR is the translator between two worlds. The key points o f the translation are summarised in figure 5.18. As far as upper layers arc concerned, the combination of the SM-RP protocol on the BSS leg, the MAP/H protocol on the NSS leg, and the relay in the MSC/VLR, can all be considered as a single network protocol, providing the routing between MS and GMSC (and from then on to the Short Message Service Centre), as well as the possibility to deal with several SMS-gateway MSC VLR RP-DATA Forward Short M e s s a g s , message ref. (TCAP component reference) originator TP-message sm-RP-0A1 L sm-RP-UI RP-ACK message ref. RP-ERROR message rel. < Forwarding Acknowledgment ( T C A P component reference) Forwarding Error 1< ( T C A P component reference) cause error type Figure 5.18 —Short message relaying \\ ilh SX1-1ZU and NIAP/I Messages from S\1-R1' and NIAPIII nap onto nn: another. relaying both the short message and its ackno \\ ledgement between the MS and the NIS(' to \\ hick the 551-SC is connected. messages in parallel. This I 3e how these protocols will he considered in the description o f the upper layer protocols for the short message services, in Chapter 8. SPECIFICATIONS R E l ( EN The link protocols used in GSM are specified i n the follo\\ ing specifications: • f o r the radio interface. \ I'Din is cnt.re.y specified in TS GSM 04.05 (general aspects mkt TS GSM 04.06 (detailed protocol specification): • f o r the Ahis interface. the reference Ibr LAI'l) is the C l ' I • 1 7 5 46-20 recommendation. itself largely based o n the ( ' ( ' I I T Q.920 and Q.92I recommendations. TS GSM 08.56 contztins 306 T H E GSNI SYSTEM the information specific t o the use o f LAPD i n the Abis interface context, with reference to T/S -16-20; • f o r the A interface, the reference for MTP is contained in the CCITT recommendations Q.702 (physical layer), Q.703 (link layer), Q.704 (network layer) and Q.707 (test and maintenance). TS G S M 08.06 gives a point-by-point analysis o f the applicability o f these recommendations t o the A interface context, based on the CCITT recommendations; • t h e full specification of the RLP is given in TS GSM 04.22. l As for BSS networking, the notiun o f the protocol discriminator can be found in TS GSM 04.07 and 04.08. There is no clear explanatica of the PD as an address, though a few sentences on the subject can he found in TS GSM 08.08. TS GSM 08.58 specifies the protocol to transfer messages on the Abis interface. The reference for SCCP is contained is the CCITT recommendations Q.711 (primitives), Q.712 (messages), Q.713 (formats and codes) and Q.714 (procedures), based on s preliminary version for the CCITT blue book. TS GSM 08.06 specifics the simplifications which can be made in the A interface context. It also specifies the discrimination between BSSMAP and DTAP messages, sitt contains the complete specification of DTAP messages. as far as their transport on the A interface is concerned. The transportation layers between NSS machines are entirely. specified by recommendations outside the Specifications. MTP and SCCP are specified in CCITT recommendations in the Q.70x and Q.712 series as listed above, as well as in ETSI pr-ETS 300-008 (MTP) and ETSI pr-ETS 300-009 (SCCP). T C A P i s specified i n the c a r r . recommendations Q.771 to 775, as well as in CEPT T/S 43/BB. The special case-of supplementary services management messages is treated in TS GSM 04.80 (which also contains the full semanticsof these messages) for the mobile station to MSC segment, and in TS GSM 09.02 for the MAP part. The general architecture of protocols for the transport of point-topoint short messages is specified in TS GSM 03.40. where an example of a protocol stack for the connection of an SC to a GSM MSC can also k found. T S G S M 04.11 specifies both the SM-RP and the SM-CP protocols, which are used to carry short messages between the mobik station and the MSC. The MAP/H protocol is specified in MAP, that is to say by TS GSM 09.02 (section 5.131. I c 308 THE GSM SYSTEM RADI • S O JaCE +` t ) , r , IjADR) RF:SOrk(1:, \ I A \ AGEMEN I 309 RADIO RESOURCE MANAGEMENT 14; 0.1. RR FunctionSialo •••• 6.1.1. T h e Concept of RR-Session .4, 6.1.2. ,Initial'satipn Tmg-typ_ion Management fiandAyseiryreParation 6,1.5. P o w e r coptrol and Timing Advance 6.1.6. Radio. (panel Management 6.2.' Architecture aud.PrOtoco& 6.3. RR Procedures: ' ' 6.3.1. Initialitocedures: Access and Initial Assignment 63.2. Paging Procedures • 63.3. • Procedures for Transmission Mode and Cipher Mode Management 6.3.4. Handover Execution 6.3.5. C a l l Re-Establishment 6.3.6. R R -Session Release 6.3.7. L o a d Management Procedures 6.3.8. S A C C H Procedures 6.3.9. Frequency Redefinition 6.3.10. General Information Broadcasting Specifications reference As a telecommunication system, G S M enables i t s users t o communicate through transmission paths o f various characteristics. as explained in the previous chapters. However these transmission paths are not reserved once for all between any two pairs of users. They are set up on demand, and only for the time necessary for a given communication. This requires exchanges of information. not only between the users and the network they are in direct contact with, but also between machines within the network. This and the two next chapters are devoted to the description o f these information exchanges. which enable distant participants, users and machines. t o act together t o provide the communication services for which the networks are designed. This technological field i s known as signalling. and under this name, its reputation has not yet grown as wide as other fields such as modulation, signal processing. anti oilier transmission techniques. Many people consider it simply as a branch of so.t\1/4 are engineering. though it Is • in fact at the centre of the design of complex systems. \\ here tasks can he executed only through the co-operation of distinct machines. This is the case of telecommunications. where machines are b) essence distinct and distant. The study of GSM signalling is therefore ()I' prime importance to understand how the system operates. and the reader should not he surprised that half of the book is devoted to the signalling nth: I re.ianges. Signalling is often the justaposition of mans simple and none ur less inter-dependent procedures. and its com(lexit) stems mainly from the number and the di .yersity o f small issues. We ha‘c :dread) seen in 310 T H E GSM SYSTEM ItA1)10 RESul RCN \ I \ NM 3 1 I 4 Chapter 2, concerning the architecture, that the basic methodology to tackle such issues is "divide and conquer". Pervasive in the specifications of GSM signalling is a split in three functional domains: Radio Resource, Management, Mobility Management and Communication Management.; The management of the calls is the upper plane in this organisation, and. deals with the establishment and the release o f end-to-end transmission, paths, through the GSM domain and b y interworking with external networks, t o support user t o user communications. Communication,, Management, as a functional plane, relies on the Mobility Management functions for dealing with the mobility o f the users and with security..; related functions. T h e management o f the radio resources groups' r' functions specific to the radio interface. ir4 A major difference between a radio mobile telecommunication; • network and a network with fixed links is the management of the aceest resources. In a fixed system, a dedicated communication medium exists;`:' continuously between the user terminal and the infrastructure, ready to ft.,: used when a call needs to be established. On the contrary, in a cellelar' system like GSM, a dedicated channel over the radio interface is provided: to the mobile stations only on demand and for the duration of the c a r under the control of the infrastructure. This calls for functions which bearr no equivalent i n ISDN f o r instance. Even i f 64 kbit/s channels Itrei allocated dynamically in the case of an ISDN multi-terminal installatiorC this resource management is rather limited compared to the one in a full GSM cell. Moreover, in ISDN, a signalling channel is always ready foci use by any terminal. The matter is quite different in GSM, where 0 1 ' signalling capabilities offered to a mobile station in idle mode, that ittiP ; say when not allocated a radio channel to its private usage, are limiter:1ln the absolute minimum. The consequence is that a host of new procedtmi are needed. whatever the. user's movements. These functions form a well defined area, which we presented i n Chapter 2 as the R R (Radio Resource management) 4ictional plane. They are spread among four entities in the canonical GS architecture: the mobile station (01course!). the two base station sub-system components (BI S and BSC). and a small part in the MSC. All the higher layer functions, described in Chapters 7 and 8. are basically managed directly between mobile station and MSC, the base station sub-system ( BSS) acting for these functions as a single complex transmission system. The spread o f the radio resource inanagcnient functions implies the existence o f signalling procedures hem een the involved infrastructure machines: this is the purpose o f the signalling protocols on the A (BSC-MS0 and .ibis ( BSC-I3TS .nter:iices. which will therefore be described in this chapter. The chapter is basically composed of two parts. Alter preliminary architecture considerations needed to introduce sonic specific concepts. the major requirements which drove the design of the RR protocols ‘‘ ill be looked at. Then. alter a section to present the proton)] architecture. the various procedures needed to ltdfi I these requirements will be developed. Preliminary Architecture Considerations A section i n the middle o f this chapter will he devoted to the architecture and the protocols i n the Radio Resource management doma . However, s o m e basic notions a r e necessary f o r t h e under .andirlIR of the first part o f the chapter. I t concerns mainly the notion of anchor MSC. Besides dynamic channel allocation, another feature of GSM cellular systems i n general) compared w i t h fixed networks i s the handover. The problem consists in providing a dedicated channel frail mobile station t o MSC at every moment during a call, despite movements o f the user. This calls f o r a complex measurement and decision process to trigger the transfer of the communication at the right moment and toward the right cell. In a cellular system, the handover process is very important, since it impacts significantly the quality of the communications as perceived b y the users, as well as the spectral efficiency. The major roles are played in this chapter by the components of the BSS, the BTS and the BSC. The MSC intervenes a Mlle. to deal with handovers between cells managed by different BSCs. Some handovers may even transfer the mobile station from a cell within one MSC area to a cell in another MSC area, thus involving two MSCs. The roles of the two MSCs are different. In no case does the MSC in charge o f the communication relinquish its control to the new NISC. This NISC is called the anchor MSC for the Connection. This is an important design. choice of GSM, with numerous consequences on the procedures. Several arguments justify this choice: a compelling one is the charging problem. since toll ticketing is much simpled when one MSC follows a call from its beginning to its end. This chapter w i l l be devoted to these topics, that is to say to the functions required t o co-ordinate t h e mobile stations a n d the infrastructure so as to provide the suitable transmission means over the radio interface, whatever the telecommunication service requires, and A consequence of this architectural choice is that alter an inter--.....,2111C handover, two MSCs (and al most Iwo) May he invoked in the connection. The transmission chain hem een the mobile station and the interworking point with external networks is then composed of a 131-S. a THE GSM SYSTEM relay MSC RADIO RESOURCE MANAGEMENT 3 I3 the main properties of the transmission chain, such as whether signalling. speech or data is transported. and whether ciphering is applied or not. anchor MSC The next issue w i l l be o n how handovers are decided. The handover execution itself will he treated mainly in the procedural section. --Addressing the handover preparation issue w i l l take u s deep into considerations about the measurements performed by the mobile station and the base station. These measurements are the basic information upon which handovers can be decided. BSC M S C VLR M S C VLR Next, two ancillary functions o f the transmission over the radio interface will be looked at. the management o f the transmission pOWCr and of the timing advance. Figure 6.1 — The concepts of anchor and relay MSCs The transmission chain may involve two (and at most two) MSCs: the anchor MSC in charge of the communication management and the relay MSC in charge of the BSS with which the mobile station is in contact. Et)* BSC and either two MSCs, a relay MSC and the anchor MSC, or of a single MSC (see Figure 6.1). To ease the problem o f terminology, a convenient approach is to consider the notions o f relay MSC and of anchor MSC as functional, and t o admit that when there is only one MSC, it is at the same time the relay MSC and the anchor MSC. Thus the term relay MSC will be used to refer to the MSC in direct contact with the BSC, even i f it is also the anchor MSC; and the term anchor MSC will be used to refer to the MSC in charge of upper layer treatments, eveniftsa . C S ryM e h lo 6.1. R R FUNCTIONS In this first half of this long chapter. we will study the different .Radio Resource management aspects from the requirement side. While implementation issues w i l l often be addressed, i n particular for the distribution of the tasks between the involved machines, the details of the signalling procedure will be entirely in the second part. The functions covered by the management of radio resources are centred on the management of transmission paths over the radio interface, and more exactly between the mobile station and the anchor MSC. To develop these functions, we will use the concept of RR-session, which will be presented first. After a small passage concerning the access and the paging. through which things start, we will deal with the handling of Finally we will deal with the management of the radio channels on the radio interface as a whole set. The two main facets are the handling of the configuration of the radio channels in each cell. and the allocation strategy of the dedicated radio channels (TACHPis and TACIT/Fs). 6.1.1. THE CONCEPTOF RR-SESSION Most o f the functions in the Radio Resource management plane relate to the management of the transmission between the mobile station and the anchor M S C . F o r each mobile station engaged i n a communication, there exists a transmission path. as well as a signalling path, between itself and the anchor MSC. As seen from a mobile station. such a path is set up when it enters the dedicated mode 01.e.. when it leaves the idle mode), and is released when the mobile statio0 goes hack to idle mode. In the infrastructure, a transmission path exists for all this period, but can be thoroughly modilied. especially by handovers. We will refer to what is managed during this period of time as an RR-session. As a minimum, an RR-session must include means to transmit signalling between the mobile station and the anchor MSC through the 1313, BSC and possibly a relay MSC. including a dedicated radio ehannel._.. references to manage it on both the BTS-BSC interface and the BSCMSC interface, and means in the BSS to monitor the radio connection and take handover decisions when necessary. This minimum set silt] ices only in the cases where the transfer ()I' circuit-type user data i s not required, such as f o r location updating, short message transfer o r supplementary service management. When circuit-type user data needs to 314 M E O W SYSTEM SAPID L A P D link 1c o n n e c t i o n 4,-C5 i B SCCP connection • 315 R:\1310 I r E S O L I t C h S l A N A G I A I L V I BSC • MSC VLF, radio channel = = - - - -fa circuit terrestrial circuit TRAU-= terrestrial (16 kbit/s) ( 6 4 kbit/s) CM transactions k anchor MSC An RR-session contains both the signalling resources between mobile station and anchor MSC, including a dedicated channel on the radio path, as well as the user data circuits if need be. be transmitted, then a complete circuit connection between mobile station and anchor MSC is also part of the RR-session, as shown in figure 6.2. For instance, a speech call requires the use of a signalling connection as well as a speech-carrying connection between mobile station and MSC. This last connection makes use of dedicated resources such as the speech transcoder transforming the GSM-specific speech representation into the 64 kbit/s representation used in fixed networks. An RR-session has many different characteristics which have to be managed by procedural means. First, two different kinds o f dedicated channels exist on the radio interface: they have been referred to is TACH/8 and TACH/F in Chapter 4 (there will be three when "half-rate" channels are included: TACH/8, TACH/H and TACH/F). Second, when a circuit for user data is present, it can be used according to different transmission modes. Finally, some other less important transmission peculiarities characterise RR-sessions. An example is whether ciphering is applied or not. All these characteristics may change during the lifetime of an RR-session. In the Specifications, the term "RR-connection" refers to what is managed during the period o f connection to a given BSC. A change of BSC (e.g.. at inter-BSC handover) entails a change of the RR-connection. The Specifications do not have a specific term covering what is managed for the whole period where the mobile station is in dedicated mode between two periods of idle mode, i.e., what we call here an RR-session. Figure 6.3 illustrates the concepts of RR-connection and RR-session, and their relationship. The figure also shows that an RR-session can be used for several calls in succession or in parallel, or more generally several CM-transactions ( C M f o r Communication Management), as will be R -session relay MSC I BSC RR-connectIon BTS Figure 6.2 — Contents of an RR-session R A A _ I A I I inter-MSC handover inter-BSC handover intra-BSC inter-BTS handover or change of channel on same BTS Figure 6.3 - The concepts of RR-session and RR-connection From the concept of radio connection (bottom line) to the one of RR-session trop line . different levels of transition awareness ma> he defined. The .Specificarions use the concept of RR-connection. ‘k hid' corresponds to the BSC ie‘‘. CM-transactions may run in parallel or in tandem during the lifetime of ;in RR-session. as shim it on the top of the described in Chapter 8. The beginnings and ends o f CM-transactions relate to the usage of the inmsmission. and are compleielc independent from RR-connections, whose succession relates to the movement of the mobile stations. The RR-session is the bond between the two domains o f radio resource management and communication management. It represents the views of the mobile station and of the anchor NISC. An RR-session starts- - - when the mobile station goes to dedicated mode (the access, when the initial assignment o f dedicated channels is performed). and disappears when the mobile station goes back to idle mode. The l i f e o f a n R R -connection i s punctuated b y infra-13S(' handovers and changes o f radio channels. and this defines another subdivision i n the lifetime o f a RR-session. A t the lower level. the channel connection corresponds to the continuous usage o f the same radio channel by the same mobile station. :\ channel connection starts either through an initial assignment. a subsequent assignment (a change of channel done. e.2.. because the allocated channel is no more o f the 316 THE GSM SYSTEM It • 1W) REM)) '12('E NINNAC; ii i v c r 3 17 sometimes in addition a relay MSC. as shown in figure 6.4. Each of these machines maintains some context related to the RR-session. When a handover occurs, the configuration changes. some contexts must be erased and others must be created i n other machines. Based on the corresponding configurations. t h e functional architecture o f G S M distinguishes three kinds of handover. In an intra-BSC handover. only the radio channel, the context in the BTS and possibly the BTS are changed. In an (intra-MSC) inter-BSC handover, the BSC is changed in addition to the radio channel and the BTS. Finally, in an inter-MSC handover. the mlay MSC is either created, replaced or suppressed. In all cases, the *anchor MSC remains in place throughout all the life of the RR-session. It is indeed the only machine sure to be a constant, and the context of the RR-session in the anchor MSC is indeed the anchorage point o f the session. 6.1.2. INITIALISATION Figure 6.4 — Configurations changes for an RR-session During its lifetime, an RR-session may go through different transmission configurations.If the initial configuration is for example the left one, it may be changed to 1 any one of the others shown. The anchor MSC remains in charge of the upper layers.,, needed type) or an incoming handover of any kind. It disappears either ..through the release of the RR-session, the assignment of another channel or an outgoing handover. Channel connections represent the view of BTSs. B y design choice, a BTS considers a change o f radio channel within the same cell as two independent channel connections. When a channel connection is cleared, the related data is wiped out in the BTS, regardless of whether the mobile station is allocated a new channel of the same BTS or in another one. There are very few stable characteristics of the RR-session beside the corresponding contexts in the mobile station and in the anchor MSC, especially when one recalls that the physical path of the transmission may change thoroughly when a handover occurs. At a given moment in time, a given RR-session is managed by one BTS, one BSC, an anchor MSC and A mobile station has two widely different operating modes, the idle mode, when it is not engaged in a connection with the infrastructure. and the dedicated mode, when a f u l l duplex channel enables actual communications to take place. The transition from idle to dedicated mode is the first step of the initialisation of an RR-session. and is called the access. As part of an initialisation process. it has many functional aspects. and this section will address only some aspects specific to the access issue. The full-blown detailed procedure itself will be presented much later in the chapter. after all the relevant concepts w i l l have been introduced. Access may be triggered either to fulfil a need expressed first on the mobile station side (e.g., a call originated by the user of the mobile station, but also a location updating). or on the infrastructure side (c.a.. a call toward the GSM subscriber). In all cases the access procedure is the same, and is initiated by the mobile station. When the network desires the establishment o f an R R -session. i t pages the mobile station which constitutes a request for it to access. To this a\ ad, when normal service is provided, the mobile station listens in idle mode to a paging sub-channel. part of the PAGCH. I f a message on this sub-channel indicates that its subscriber is paged, the mobile station starts the access procedure. as it does when the user requests it. We will then deal separately i t h the two aspects, the access proper, and the paging. 6.1.2.1. Access The mobile station manifests its w i l l t o access b y sending a messaize on the random access channel (R A{'1-I). which is answered by an 318 T H E GSM SYSTEM RADIO RESOUlteli. MANAGENIENT initial assignment message on the paging and access grant channel (PAGCH), carrying the description of the allocated dedicated channel4*1 The mobile station provides very little information in its access requeg (the message has only 8 bits of contents). In particular the mobile station does not give its identity at this moment, nor the detailed reason for the. access. Another particularity of interest is that the access on the RACH k not regulated. The consequence is that two mobile stations may semi access requests simultaneously, which would result most often in neither being received by the BTS. Most o f the complexity o f the access procedure (repetition of the attempt, resolution of access to the s a c channel by two mobile stations) comes from fixing these problems. Jowl The access ends with the allocation of a radio channel for the' of the mobile station. This is called the initial channel assignmen—' referred to in the Specifications as immediate assignment. The accesi procedure, though limited to the means needed for the transition between the two modes, makes exclusive use o f two specific channels, tl PAGCH and the RACH. From a more general point o f view, the access function is the initiation of an RR session. As such it includes the initialisation of all the contexts and all the recurrent functions which are part of the RR session. The access will therefore be revisited in many of the functional sectiont, such as those dealing with the management of the timing advance, with the transmission power control (which must be started during the access): and with the channel allocation. The detailed description of the access procedure, in the last part of this chapter, will provide the synthesis of i[3: these different aspects. Lrt. 6.1.2.2. Paging and Discontinuous Reception Compared to the other functions described in this chapter, paging is somewhat particular, because it is not directly linked to an RR-session: What is indeed the relationship between paging and radio resource inanagement? Since RR-sessions are only established at the initiative of the mobile station, the network infrastructure needs some means to trigger such an establishment; this role is indeed fulfilled by the paging procedure. But there is no common reference, no clear relationship between the paging message and the ensuing RR-session establishment. The only clue for the network is the indication by the mobile station in the first dedicated message that the RR-session has been established in response to some unspecified paging indication. Paging is in some sense closer to Mobility Management functions than to Radio Resources management functions, as it serves to locate a 3 1 9 mobile station within a whole location area. The grouping of the paging function with RR management. which is also the one followed in the Specifications, reflects the relationship which exist between paging and a number of true RR functions. For instance, paging messages and initial assignment messages share the same channel (the PAGCH). This approach is also sensible from a pragmatic point of view since the .main job for paging is done by the BSS. which is otherwise only concerned With RR functions. 1 • H o w does paging start? When ay incoming call arrives, the MSC/VLR requests the BSS to perform arpaging in some of the cells of the BSS. The MSC provides the concerned BSC(s) with the identity of the mobile subscriber to be paged and the list of cells in which the paging should be issued. The BSC is in charge of managing the PAGCH, i.e.. the grouping and scheduling of paging messages as well as initial assignment messages. This scheduling may be more or 14ss optimised. depending on the manufacturer. The Specifications describe a framework for such an operation, but the operator/manufacturer can choose how often to repeat unanswered paging messages. whether t o send initial assignment messages also on those parts o f the PAGCH which correspond to the paging sub-channels. etc. The split of functions between BTS and BSC with regard to paging also allows some flexibility and is somewhat manufacturer-dependent. Typically, the high-level tasks, such as priority decisions. rest with the BSC. Sophisticated approaches may take the system load into account tbr the respective priority of paging and initial assignment messages. Another aspect o f paging also tackled by the BSS concerns the concept of discontinuous reception. For the sake of battery consumption in handheld mobile stations. it is important to minimise the amount of information the mobile station has to receive. demodulate and analyse while in idle mode. To this avail. the downlink common.control channel can be divided into several paging syb-channels. and all.paging messages pertaining to a given subscriber are normalls sent on. the sank• subchannel. The sub-channel organisation can a n o n a' cell basis. but broadcast information allows the mobile stations to determine it. Such a scheme allows mobile stations to restrict their monitoring o f paging messages t o t h e i r o w n paging s u b -channel. (heath\ increasing significantly the lifetime o f their battery. at the expense o f a small increase in delay for the setting up of incoming calls. Such a feature is rated to as discontinuous reception, or DRX. and is not to be confused with discontinuous transmission ( D T X ) w i t h which i t hears n o relationship except the similar names. Mobile subscribers are assigned to paging sub-channels in a pre-determined way taking into account the three last digits o f their international mobile subscriber identity (the 320 T I i C1SM sYsTiAi IMSI), so that the knowledge of the PAGCH configuration is enough for each mobile station to determine which blocks of which CCCH curia unit it should listen to. • :.? The PAGCH follows a 51 x 8 BP cycle, where BP denotes a bu period, using 9 blocks per cycle for a PAGCH/F (the PAGCH of capacity) and 3 blocks per cycle for a PAGCH/I' (the PAGCH o f ' ' capacity), as described i n Chapter 4. A certain (parameter-conn number of these blocks belong to some paging sub-channel, the 8 being reserved to initial assignment messages. This number may range. from 2 to 9 for a PAGCH/F and from 1 to 3 for a PAGCH/T. A p: sub-channel is (almost) cyclic, with a cycle ranging from 2 to 9 51 x 8 BP (that is to say from 0.95 second to 4.25 seconds), here agaita under the control of a parameter. Hence, on a given PAGCH/F, thereat, be from 4 to 81 paging sub-channels (2 to 27 for a PAGCH/T). Theitirty; parameters describe the PAGCH configuration and are broadcast °tribe BCCH. The choice o f these parameters is the operator's. The secciaL parameter (linked to the cycle of paging sub-channels) corresponds.to;eit compromise between the access time and the power consumption okthei0 mobile stations. T h e first one was introduced solely t o enablittbi development of very simple PAGCH scheduling algorithms which c l o 4 , use the possibility to send initial assignment indications on a paging euh4 channel. In such cases, the choice of the parameter is related to the rittio±: between the paging load and the initial assignment load. Othenvise;the"!' parameter is set so that any PAGCH block belongs to some paging stab; channel. A small detail is that the interval between two successive paging sub-blocks d ;r4-7 the same sub-group is constant, except (in some combinations) once every 3.5 o n . _ when the numbering scheme goes through 0. This happens when the paging sub-channel "cycle" is not a divider of the numbering cycle (lasting 2048 x 26 x 51 x 8 BP), i.e., tar cycles of 3, 5, 6, 7 or 9 times 5 I x 8 BR Procedural Requirements for Paging The procedural requirements f o r paging management include means f o r the MSC t o require a given subscriber to be paged, a mechanism for the BSC to control the sending of this paging message (or alternatively to provide the BTS with the relevant data to build and schedule the paging messages) and a way t o indicate the PAGCH configuration to all mobile stations. In addition, some means to configure the PAGCH are required in the operation sub-system (OSS), as part of the more general techniques for controlling the cell configuration. This is dealt with in Chapter 9. 321 k ADB RP:tic R 1:(1. IANAGIAIF \ 6.1.3. TRANSMISSION MANAGEMENT The life and deeds of the RR-sessions have only been sketched in this previous section. We will now see in more detail what constitutes the management of the transmission characteristics of an RR-session. These characteristics are chosen depending on the service to be provided. They are decided by the anchor MSC, but transitions are co-ordinated by the BSC. The most obvious area of co-ordination relates to the type of data which must be carried. On this depends the existence or not of terrestrial circuits, the type of the radio channel and the transmission mode (speech. or data with some data rate). Beside these, the BSS has also to manage the cipher mode, as well as the discontinuous transmission modes. We will take each of these aspects in turn, to examine what is to be done. — C U L Transmission Mode Management We group under t h e t e r m transmission m o d e t h e m a i n transmission characteristics. More precisely, the concept of transmission mode refers to the way the GSM transmission chain is used for carrying circuit user information, from the mobile station t o the point o f interconnection with partner networks. The set of possible transmission modes differs depending on the type o f channel used on the radio interface. The transmission modes have already been described in detail in terms of their "physical layer" characteristics in Chapters 3 and 4. Table 6.1 summarises which transmission modes exist on each radio TACH/8 T ACHIF TM71111 signalling only signalling only spee,.11 data 3.6 O n : . data 6 kW% data 12 kbit/s. transparent data 12 kbit/s. non-transparent siguallim; i 'illy data . “ , 4bit A data 6 kbit•A. it oucparem aura h Oil .. Thni-lranspareni Table 6.1 — Transmission modes used on the radio channels The only transmission mode on a TACIT/8 corresponds to signalling only. whereas the "full-rate" channels may carry 6 different modes. and the "half-rate" channels ss ill presumably carry the 5 modes listed. The "transparent" and "nun-transparent modes tsee Chapter 31 use the same channel iodine on the radio path, THE GSM SYSTEM channel type. The transmission modes which were oefined on half-rate channels are not in the phase I Specifications, but are also shown in this table. The "signalling only" mode corresponds to the non-usage of their channel t o carry circuit-type user data. The transport o f signalline information is a capability which exists in all transmission modes, once' an R R -session has been established. There are even cases wheretit, represents the only transmission need; for instance, at the beginning o f t call (before the conversation phase), for transmission of user data other than circuit-type (short message transmission), for location updating* for supplementary service.management. t t a " The existence o f a f u l l transmission path including teffes circuits between the BTS and the MSC/IWF, and the inclusion in path of a transcoder and rate adaptor unit (TRAU), depend on the m for instance, there are none of these in "signalling only" mode, In general, t h e transmission mode i s chosen b y the M S c “ . : depending on the end-to-end service. When the RR-session is initially t: established, the MSC does not intervene in the process before the point. when it knows exactly the transmission needs to fulfil; up to this po4k the Specifications impose the mode "signalling only". The channel can IS chosen (by the BSC) to be of any type, though it is typically a TACH/It (see page 355 for the allocation strategies which can be followed by the BSC). Later during the lifetime of the RR-session, the channel type and the transmission mode may change; these changes are decided by the MSC, in order to adapt the transmission media to the needs of the users."I ift Although the decision lies with the MSC, i t is the BSC which chooses the exact channel (of the requested type) and is in charge of tiiordinating the different machines, including the mobile station. The only exception t o this rule i s the establishment o f the terrestrial circuit (between BSC—or TRAU--and MSC), which is always done by the MSCs. This exception stems from no specific reason but the weight of history. Switches have always been i n charge o f establishing their surrounding circuits. This situation actually raises some problems in phase 2, and it would probably have been a better choice to let the BSC in total control o f the whole transmission chain. The management of the terrestrial circuit will be addressed in a hit more detail in the next section. Procedural Needs for Transmission Mode Management The procedural organisation o f transmission mode management has two facets. First the MSC must be able to indicate the need for a RADIO RESOURCE MANA(iEMENT , 11 change i n t h e transmission m o d e a t a n y m o m e n t d u r i n g an R R -session. Second, the B S C m u s t have means t o co-ordinate the m o b i l e station, the BTS and the TRAU for fulfilling the MSC request. This last aspect is split into two cases as seen from the BSC and the mobile station. whether the type (full rate or eighth rate) o f the existing radio channel fits the requirements or not. A.typical case when it does not is at the beginning of an RR-session established for the purpose of establishing a call, i f the existing channel is a TACH/8. In such a case, the radio channel must he changed. The procedure to change the radio channel used by an RRsession without changing the cell is called a subsequent assignment. I f the type of the existing channel is appropriate, but the transmission mode is not, the procedure between the mobile station and the BSC is a mode modification. It should be noted that the handover procedure can be used c h a n g i n g the transmission mode, including the type of channel. Reciprocally. an intra-cell change of channel is often called a handover, i f triggered by quality considerations and not by the adaptation to the needs o f the service. A small digression on ihe terms is useful ,u this stage. This last usage of the term hando‘er as to he understood Iron) ihe point of view of why it is done. However, from We point of vio\ of what is done as seen hrrvccn Ole BSC and the mobile station, there is indeed no imference between a subsequent ;is.igninent and an infra-cell "handover. The accepted use of the term handover thus depend, (in the context. • 6.1.3.2. Terrestrial Channel Management An RR-session may or may not include a full circuit between the mobile station and the MSC to carry user information. This circuit is not always present; for instance, i f the RR-connection is used kw location updating, such a circuit is of no Use. When it is present, it includes a radio channel (a TACH/F. or in the future a TACH/H), and terrestrial circuits connecting the I3TS with the anchor MSC via the BSC and a TRAU i f applicable. The terrestrial circuit from BTS to BSC ( possibly via the ' M A U I is in a one-to-one relationship with the radio channel, and is then dealt with lw the liSC as part of the radio channel management. On the A interface. the circuit (from the liSC or TRAL to the MSC) is allocated to an RR-session by the relay MS('. The indication that a circuit has been allocated is given to the BSC' through signalling 324 TILEGSM SYSTEM i R A D I O RESOURCF. MANACIFINIENT 325 handover, and i s i n particular entirely changed f o r an inter-MSC handover. The establishment o f the new path. and the release o f the previous one are pan of the handover execution procedure. 6.1.33. Cipher Mode Management tr The transmission over the radioltpath has a few characteristics which must be managed independently from the type of transported data. The first one is the cipher mode: the transmission may be ciphered or not, as desired by the MSC, according to some criteria vitirch depend on the Figure 6.5—Channel allocation on the terrestrial interfaces When the TRAU is situated on the MSC-side of the BSC, / O l t the MSC first chooses a circuit towards a TRAU, then a signalling exchange 11, W. takes place on the A interface, and finally the BSC controls the set-up of circuittwit between BTS and TRAU. means, enabling the BSC to then connect it to the radio part of the path through its switching matrix (see figure 6.5). D e g Finally, the circuit between the anchor MSC and the relay MSCrif they are different, i s established i n the canonical architecture using standard PSTN or ISDN methods, The establishment is initiated by the ;1 anchor MSC, using the same procedural means it would use to establishi 4.g call. 1 4 '1 The existence of a remote transcoder and rate adaptation unit aloes the BTS to relay MSC path makes this picture somewhat more cowl, ;rt.t. If no BSC-controlled switching matrix exists between the TRAU and tb MSC, then a one-to-one relationship should exist between a s transmission resource in the TRAU and an MSC-TRAU circuit. Thus, it my is in that case the MSC which implicitly chooses the transcoding device by choosing the terrestrial circuit. This situation bears no consequenceif 4 the TRAU devices are all equivalent: otherwise. a potential problem exists, since it is the BSS which is in charge of the TRAU, and of the consistency between what the TRAU does and what is done on the radio interface. Thus, the TRAU is in some way managed both by the MSC and the BSS, and in practice, the architectural choices o f GSM make it difficult to have distinct types of TRAU. Procedural Needs for Terrestrial Channel Management The terrestrial part of the transmission path o f an RR-session is established when a TCH/F is requested by the anchor MSC. This is a part of the subsequent assignment procedure. It is obviously modified at each operator. An RR-session is always started in "signalling only" mode, and always in clear text (i.e., not enciphered). The latter is a necessary reqiiirement, since the R R -session setup i s performed without the network knowing the subscriber identity, and therefore ciphering with a __user-related key cannot be applied. Means for transition from clear text to ciphered transmission on an existing RR-session are therefore required. The Specifications do not provide explicitly for the reverse transition. i.e., from ciphered mode to non-ciphered mode. No need was identified for such a transition. Similarly, the Specifications do not cater for the change of ciphering k e y while i n ciphered mode. However. the existing procedures could be used in the future to support these transitions. should . the need arise. Similarly t o t h e transmission mode management case, t h e transition from clear text transmission to ciphered mode is decided upon by the MSC, the BSC being in charge of co-ordinating the actual change. Cipher mode management impacts the mobile station and BTS. Procedural needs include the means for the MSC to provide the ciphering parameters (the mode, and the user ciphering key Kr if needed) to the BSC for an incoming handover, the means for the MSC to require a change of the current mode from non-ciphered to ciphered. and the means for the BSC to co-ordinate the transition in the BTS and the mobile station. This last aspect is of primary importance, since messages sent in one mode will not he understood by the peer entity if this entity is set in the other mode, resulting in an unrecoverable loss o f connection. The way to synchronise mobile station and BTS will he described in the procedural section. 6.1.3.4. Discontinuous Transmission Discontinuous transmission (DTX) has been described in Chapters 3and 4. When discontinuous transmission is applied, actual transmission 326 ".ADIO RESOURCE MANAGENIENT THE GSM SYSTEM 3 2 7 BTS is obviously concerned; but derives its behaviour dynaMically from data coming from the mobile station (uplink) and from the TRAU (where it exists) and MSC/IWF (downlink). Whether discontinuous transmission should be applied or not is, there again, decided upon by the MSC, and the execution is controlled by the BSC. A GSM BSS must indeed be able to cope with discontinuous transmission, whatever the strategy o f the corresponding MSC. ._4 The choice of the strategy for applying discontinuous transmission is one of the many configuration parameters on which operators may play to optimise their network. Several considerations must be taken into Account in this strategy. For instance. GSM mobile to mobile calls suffer a loss in quality when discontinuous transmission is applied to both radio segments; experts refer to this phenomenon as "double clipping". The ()Orator may therefore well choose not to apply downlink discontinuous transmission for such MS-to-MS calls, if they can be identified. The downlink discontinuous transmission mode can he changed when the transmission mode changes; these are indeed the only instants at which t h e Specifications a l l o w a change i n t h e discontinuous transmission mode, since no other needs were identified. For the uplink discontinuous transmission mode. the network can at any moment either force the mobile stations in communication to use it, forbid them to do so or leave the choice open. downlink DTX u p l i n k DTX 04 : Figure 6.6 — Procedural needs for discontinuous transmission management 4 ) . The use or not of discontinuous transmission needs only be notified to the transmitting end; the receiving unit does not need to know beforehand. .6n-the radio path is reduced to a minimum when it is detected that the user data does n o t contain meaningful information (during speech silences for instance). This feature is optional and must therefore be managed. Moreover, discontinuous transmission m a y b e applied independently t o each direction, so that the control o f discontinuous transmission must take into account two components: the uplink mode and the downlink mode. Discontinuous transmission i s o n l y relevant t o some o f the transmission modes, speech and non-transparent data, simply because in the other cases it is difficult to assess when user data transmission can be suspended without degrading the service. The discontinuous transmission mode affects the operation of the mobile station and of the TRAT 111P Procedural needs f o r discontinuous transmission management include the means f o r the MSC' t o indicate whether discontinuous transmission should be applied lbr uplink and for downlink. and means (or the BSC to configure the mobile station and the TRAU via the i n s (see figure 6.6). 6.1.4. HANDOVER PREPARATION The possibility to change the cell during an RR-session is a very important function i n a cellular system. and the major source o f complexity in the Radio Resource management plane. We have already touched upon some of the aspects related to the execution of a handover. i.e..; what has to be modified when a handover occurs. However. this repr,esents but the tip of the iceberg.. The process which precedes it. the handover preparation, may he thought o f as a "behind the scenes" activity, yet it is the most important topic, which conditions both spectral efficiency and the quality of service as perceived by the users. The decision to trigger a handover, and the choice of target cell, are based on a number of parameters, and various reasons may trigger this 328 m r : GSM SYSTEM decision. These reasons will be studied first, then a description of the parameters affecting the choice w i l l follow. Among these parameterks, radio measurements performed by the mobile station and by the BTS are of foremost importance, and will be looked at in detail. The functional distribution of functions will be addressed last. 6.1.4.1. Handover Purposes At first sight, the aim o f handovers is to avoid loosing a car progress when the mobile station leaves the radio coverage area of cell in charge. Such cut-offs are very badly perceived by the users, have an important weight in the overall perception o f the quality service by the subscribers. We shall call this type of handover "res handover", where a high probability exists that the call will be lost if cell is not changed. An extreme form of the rescue handover is the cal establishment, which is an attempt for the mobile station to salvage the connection after an effective loss of communication with the serving In other cases, it may be of interest to change the serving cell given mobile station even i f the transmission quality is still adeq This m a y happen when t h e global interference level would significantly improved i f the mobile station would be in contact another cell. Computations and simulations show indeed that there, usually a "best cell" from the point o f view o f interference. Thzi statement is especially true when power control is being used, since the cell corresponding to the minimum path loss will minimise the mob& station transmission power (ordered by the BTS), thereby statisticaiii minimising the overall interference level. A handover triggered with die goal to optimise the interference level and not for the sake of the ongoing communication shall be referred to here as a "confinement handover". Such handovers result in a "confinement" in optimal geographical are" of the mobile stations which have a connection in a given cell, preventing them from wandering out of the optimal cellular coverage even if theig,.. connections are still o f adequate quality. Confinement handovers may potentially conflict with local optimisation o f the transmission quality, and should not be performed toward cells for which transmission quality is not correct. A third kind of handover is referred to as the "traffic handover". It may happen that a cell is congested whereas neighbour cells are not. Such a situation happens typically when specific events lead to a very local geographical peak: fairs, sport events, and so on. Because the actual coverage of neighbour cells often overlap a lot, handing over some calls from one cell t o a less congested one could temporarily improvg congested situations. This kind of handover must be handled with greit itAinoREsot.ipical basis liar BSIC allocation ihi. It has been explained above how the mobile station is able recognise that the frequency i t monitors does correspond to a beat channel. But there could be configurations where the mobile station able to capture more than one beacon channel using a given frequency This may happen when frequency planning must be done with very feV: frequencies, or at national boundaries. In order for the mobile station to. be able t o discriminate between the cells transmitting their beacon channels on the same frequency, a mechanism based on the BSIC (B Station Identity Code, one o f the major misnomers in GSM) has introduced and warrants a detailed explanation. frequent along national boundaries. Whereas inside a countr t h e frequency allocation of different operators are disjoint. two public land mobile network (PLAN) operators on each side of the border will have some frequencies in common. Frequency usage co-ordination Jet‘s een l l operators helps. but cannot he enforced. In most cases a mobile station will still b e i n a position t o receive the same beacon frequency transmitted by two base stations o f different PLMNs. For all these reasons, a scheme allowing to distinguish cells using the SOW(' brawn frequency was deemed necessary. This is the role of the USW. a 6 -bit code word broadcast on the SCH of every cell. The BSIC (Base Station Identity Code), which applies more properly to a cell, is not an unambiguous identification of a base station. Mali), cells bear the same BSIC, and moreover the normal practice is to allocate the same BSIC to neighbouring cells. So what is it? The BSIC is in fact a "colour code" (by reference to map colouring), allowing mobile stations to distinguish cells which transmit their beacon channel on the same frequency (see figure 6.10). For instance, when the radio spectrum available t o a given operator is limited to, say. 2 MHz, frequency planning must cope with at most 10 frequencies. The best beacon frequency allocation scheme may not be able t o avoid overlapping coverage in this case, and a mobile station will in some cases receive two beacon channels with the game frequency. A similar situation is also The BSIC intervenes in different cases, all related to the distinction between cells using the same beacon frequency: • i n order to perform neighbour cells measurements. the mobile station is provided with the list of frequencies to monitor. In the reporting message. the mobile station is required to indicate the BSIC f o r each beacon frequency o n which i t reports measurements. This implies that the mobile station has decoded the SCH of the beacon channels it measures. but this is not an additional constraint, since i t must already be done for presynchronisation. The availability of the BSICs for measurement processing at the BSS enables i t t o check which cell has effectively been measured in the case of ambiguities. 338 THE GSM SYSTEM KANO KiiSUt Itt T \I \ N \ II \ I 3 3 t ) • t o prevent measurement reporting f o r cells toward which handover is precluded, a mechanism allows the n e t w o r k * indicate a subset of BSICs for which reporting should not be done. This screening mechanism is explained later. • w h e n a mobile station sends an access burst on the RACH o1 a cell, there is a risk that this random access be received by another cell than the one it was aimed at, i f these cells use the same RACH frequencies and if their TDMA synchronisation not fall too far apart. In order to avoid such spurious recepti the RACH coded burst is "exclusive-ored" with the BSIC,' that only the right cell has a chance o f decoding the successfully, based on the redundancy added to useful bits checking correct decoding. or 1of 5 2 or 6 MI 3 o, 7 • w h e n a mobile station in idle mode monitors neighbour cel regularly reads their BSICs; this provides the mobile station with a quick way to check whether the monitored cell is still the': same or not. Such a check could also be achieved by decocting the broadcast messages containing the full cell identity, but an greater cost. • The whole issue of BSIC planning consists in allocating different BSICs to cells using the same beacon frequency and between w overlapping coverage areas may exist. Inside a PLMN, BSIC plannin fairly easy; the matter is different at borders, when some co-ordinatt necessary. In order to help this co-ordination, a tentative allocation of first three bits of the BSIC has been introduced on a country bas' shown in figure 6.11. This three-bit part has been named the "P colour code" (or NCC, for Network Colour Code), a term which has widely misunderstood. Its value is not a definite attribute of a PLMN, a possibility for use near boundaries. and which can be overruled. bilateral o r multi-lateral agreements between the concerned PL Indeed, nothing prevents all 64 values of the BSIC (including all 8 values of The "PLMN colour code") being used inside a PLMN when far away from any international boundary. Moreover, two PLMNs in the same country usually have disjoint frequency allocations, and can use without risk the same BSICs (and the same "PLMN colour codes"!). The first 3 -bit part o f the BSIC is also the basis for screening measurement reporting. When two cells o f different PLMNs using the same beacon frequency may potentially overlap near a border, it is normally best for a BSS to indicate to mobile stations that they should not report measurements concerning cells o f the other PLMN, since these measurements would be worthless (since inter-PLMN handover is usually not performed), and might prevent the measurements of some cells in the Figure 6.11 — A European PLAIN code !nap This tentative allocation of the "I'LNIN code" b iountr) is proposed to ease allocation of the BSIC near international hounilarie.: but it does not represent a requirement. right PLMN to be reported, since the number of reported measurements is limited to 6 neighbour cells. To that end, the inFormation broadcast by each BTS includes an 8-bit screening indication, with one hit indicating for each of the 8 possible three-hit patterns whether cells with a BSIC' starting with this pattern must he reported on. This screening indication is the P1.1/A' PERMITTED indication. The PIMA; PERMITTED information happens to be broadcast on the BCCH. A current misunderstanding is that it influences cell selection (its name seems to carry such information indeed). This is not so: this information impacts only the measurement reporting and nothing else. The Measurement Period We will now look in some details to the measurement process itself. The various characteristics t o he measured (reception le\ el o r reception quality) are continuous by nature, and affected 11) a significant 340 T H E GSM SYSTEM industrial noise, Rayleigh fading, interference, and also due to the change of frequency introduced by frequency hopping. To be usable for decisionmaking, the measurements need to be filtered (e.g., averaged). The firl; step of the filtering is done in the mobile station. I t is quite simple: it consists in taking the average value of each measured parameter f o r k duration of the reporting period, which will be described in the following paragraphs. The other point worth noting is that what is averaged is di raw bit error rate for quality measurements, and the logarithm o f ' reception level, or, i n other terms, the reception level expressed decibels. A measurement report pertains to a given measurement period, is to say the period during which the measurements were done., duration of the measurement period is always equal to the periodicity, message transmission on the SACCH (i.e., 480 ms on the TACH/F," around 471 ms f o r the TACH/8). On the TACH/F, the start o f ' measurement period starts a fixed time before the start of trans But on the TACH/8, things are different. The absolute position of measurement period i s the same f o r all TACH/8s o f the same' (timeslot number, indicating the position in the 8 BP cycle). Its relatic the time of transmission of SACCH messages therefore varies, depe on the T N . I n addition, the position o f the measurement peri different for TACH/8s combined with common channels and for which are n o t . I n a l l cases however, t h e uplink a n d d o measurement periods are simultaneous. 4 A small subtlety in the definition of the measurement period coats from the quality measurement. With the interleaving scheme for spent, there is one speech block spread over two successive measurement periods in the case of a TACH/F. Quality measurements for this block belong to the second period. This problem should not have arisen S i b s TACH/8, where the aim was to design the measurement period carefully so that blocks are all entirely within a measurement period. However, there is one exception to this rule, since the third TACH/8 of TN 4 does not comply with this aim. •—• A simultaneity problem arises between the measurements done by the mobile station and those done by the BTS. The second ones we known by the BTS shortly after the end o f the measurement period. whereas the measurement report from the mobile station for the same period is delayed by the message transmission delay, which amounts to roughly half a second on a TACH/F. In order to provide the BSC with reports which match, the BTS must then buffer its measurements until it has received the message from the mobile station. As seen by the BSC measurements are received roughly one second after the start o f the measurement period in the case of a TACH/F, and between a bit more n.von) nEsot nut: \IAN:vit. vl \ r 3 4 1 than half a second and close to one second for the TACH/8. depending on their position in the 102 BP cycle. The shorter delay in the case o f a TACH/8 comes from the fact that each SACCH message is sent over 4 consecutive bursts, and not spread as for the T A M E The Interaction with Discontinuous Transmission The accuracy of measurements concerning the ongoing connection raises two problems. one in relation to discontinuous transmission, the other linked with power control and frequency hopping when using the beacon frequency. When discontinuous transmission is applied, some slots belonging to a channel may not be used for effective transmission. This is indeed the goal of discontinuous transmission. But then measurements on these slots will obviously report a low reception level. and a bad quality. Even more annoying, one could imagine a measurement period, or a succession of periods, with no transmission at all. hence some difficulties for the processes which rely o n these measurements. To circumvent these problems, the Specifiewions impose that at least 12 bursts are sent within each reporting period. These bursts amount to the systematic use of the SACCH for effective transmission (4 bursts constituting a coding block). and 8 bursts on the TCH itself. On a TCH/8. which corresponds to exactly 8 bursts per measurement period, this leads to the consequence that discontinuous transmission is not applied. On a TCH/F, this means that at least one block per measurement period must be sent (a block being interleaved over 8 half-bursts). For speech, this block contains a silence descriptqr frame (SID frame) refreshing the comfort noise characteristics. In addition to this minimum transmission rule. the .Skuilications require the B T S a n d t h e mobile station t o report t w o sets ()1' measurements concerning the connection: "full- measurements, done on all slots which may be used for transmission in the reporting period. and -sub- measurements. dune onl \ o n the manditiorily srnl bursts and blocks. On a TACI I/F. the second set is less accurate than the first one wHen discontinuous transmission is not used. because averagin:2 is done on a smaller set.(for instance reception level is averaged on 12 bursts instead o f 100 bursts,. O n a VACIUS. the t w o measurements are evidently identical. hut are nevertheless both sent for uniformity. Finally. both the BTS and mobile station report for each measurement period whether discontinuous transmission was effectively used or not (or in another terms, whether all bursts were effectively transmitted). thus enabling the processes using the measurement t o discard the "full-. measurements provided by the other end when applicable. .142 T H E GSM SYSTEM The PWRC Indication Another problem with measurements (and the last point on the topic) is in fact a consequence of a combination of different independent details of the Specifications. First, it is allowed that a frequency hopping TACH uses the beacon frequency as one of its frequencies (of course, not on TN 0). Second, power control can be applied on the downlink TAC1L Third, the beacon frequency must b e transmitted w i t h a constant transmission power, because o f the measurements performed by mobile stations o f neighbour cells. T h e result f o r t h e channels undet consideration is that power control applies only to a subset of the b whereas other bursts (those using the beacon frequency) are sent wi fixed transmission power. T h i s leads t o inaccurate reception 1 measurements. In order to alleviate this problem, the mobile statt requested in such cases not to take into account the slots falling on beacon frequency in the reception level estimation. This is controlled an indicator, t h e PWRC indicator (originally called power con indicator, a misnomer), sent on a connection basis to the mobile state This indicator should be set i f the following conditions are all met:' channel hops o n a t least t w o different frequencies, one o f th frequencies is the beacon frequency, and downlink transmission po, control is in use. It \ D I U Itl•siti It( \ I \ s i t s I 1 4 3 although to a lesser extent. battery lice for the mobile station. When one side is received too well by the other. it becomes advantageous to reduce its transmission power by such an amount as to keep a similar quality level on the communication, while decreasing the interference caused on other calls in surrounding areas. In a system such as GSM. the full gain in spectral efficiency can be obtained with a small power range. and the 20d3B minimum case specified b y the Specifications i s more than sufficient for this purpose. In GSM, both uplink and downlink power control may be applied independently o n e f r o m t h e other: furthermore they a r e applied independently with each mobile station. The range specified b y the Specifications for uplink power control lies between 20 and 30 dB, by steps of 2 dB, depending on the mobile station power class. An example is given in figure 6.12. The range used for downlink power control is manufacturer dependent and may he up to 30 dB. also by steps of 2 dB. The control o f the transmission power is a network option. i.e., the operator may choose to apply it or not, in one direction or in both. A l l mobile stations, though, must support the feature. thereby allowing power control to be really efficient when utilised. /Power control on both directions is managed by the I3SS. The transmission power o f the mobile station is chosen by the M S . and 6.1.5. POWER CONTROL AND TIMING ADVANCE 'at Most o f the functions needed f o r transmission over the ramo interface correspond to transformations o f the signals representing data to transmit. They were the object of Chapter 4. Two functions,;‘ management o f the transmission power and o f the timing advance,* somewhere between these pure transmission functions, and transmissirii: management functions. As the transmission functions, they are perform/ continuously and relate heavily to physical features, but they share 11g: use o f signalling means with the transmission management functions: They are studied here, because of this usage of signalling means, and also —because they are, as we will see, deeply entangled with the procedures managing R R -sessions. This comes from the initialisation o f these processes, which must take place any time the channel is changed. 6.1.5.1. P o w e r Control Power control refers to the possibility to modify within some range the transmission power on the radio, both (but independently) for the mobile station and the base station. Power control shares a common goal with discontinuous transmission: improving spectral efficiency and. A power level (dB) • 39 8W • • A • • • • • • range: 26 dB • 20 m W 13 0 2 • • • • 15 power control step Figure 6.12 - Power control step. for a clam. 2 ( iSM,HM mobile sunion Power can be controlled M Niel). o f 2 dB on a ranre coin_ from W (13 ( M i l l i to the m a \ imam 515 nim N a - (Min , 344 T H E GSN1 SYSTEM ADIO RESol RUE MANAGEMI:VI commands t o regulate i t are issued t o the mobile station. The BSS computes the required MS transmission power through reception levej measurements performed b y the BTS, taking into account the MS: maximum transmission power as well as quality measurements doneiby the BTS; this last parameter helps to ensure that transmission quality3* kept above some acceptance threshold. For the downlink direction„ttie BTS transmission power i s also computed b y t h e BSS f o r P a connection, based on the measurements performed by the mobile statigiti C and reported regularly to the BTS. ark! Inside the BSS, the split between BTS and BSC is basically.* option for the manufacturer. The specification o f the Abis interface;, found in the Specifications is basically adapted to the implementation, power control in the BSC, but implementation in the BTS is possible.11V detailed procedural means for the latter case are not specified tally. Specifications, but a number of "place-holders" in the BTS-BSC protqcoi allow manufacturers to specify and implement part or all o f the poll! 7004. control management in the BTS. , 5 v ; , . GSM, the initial power level to be used by a mobile station for the first messages sent on the new dedicated channel is fixed on a cell-per-cell basis, and is the same level as used for sending random access bursts. The value of this level is broadcast on the BC'CH. to be known by all mobile stations before any access attempt. A mobile station whose maximum power level is below the broadcast value shall simplv use its maximum -flower level instead. F o r subsequent channel connections. the M S !transmission power t o be used when accessing the new channel i s specified by the BSC, either using a default cell value (typically the case for an incoming handover between different BSCs) o r based on the wledge concerning the previous connection (this capability could be d, e.g., at subsequent assignment). At the start o f a connection, the initial value o f the transmithltlill power (both for mobile station and for BTS) is chosen by the BSC. I t i l h e l k „t, case o f an initial assignment, the information available to choose dtia power is at best very small: it consists in the reception level of a sines tt.. access burst, which i s necessarily o f limited accuracy. Therefore, La A Except at the start of a channel connection, a command to change the transmission power does not trigger an immediate transition tp the ordered value in the mobile station. The maximum variation speed is of 2dB each 60 ms (see Figure 6.13). This means in particular that a high jump (more than 12 (113) will not he terminated when the next cominand arrives. The basic procedural requirements for power control, as shown in figure 6.14, include the following: change of values transmission level (dBm) :q1 sic initial 39 values 13 ( -4 T-11 commands: 19 dBm 6 1 7 dBm 3 7 time ms intervals) 0 4 dBm 3 5 • - • --> dBm measurements Figure 6.13 - Transmission power adapt at ion The transmission power is adjusted by steps of 2 (113. recurring not more often than every 60 ms. A high jump in the power control commands will therefore he answered eraduallv. 345 Figure 6.14.- Procedural ten.s for pow et control Mechanisms to report radio measurement,. In set-up the initial power ;due. ,,, 346 THE GSM SYSTEM • some means for the mobile station to transmit measurement (the same S A C C H procedures a r e u s e d f o r handoveli, preparation) up to the BSC, even though some measurementr may stop their journey at the BTS to be "pre-processed"; ' A c e • some means for the BSC to command MS and possibly B transmission power; • some means for the BSC to indicate to the mobile station initial power level value to be used at initial assignment, as as at each subsequent channel transition; • some means for the BSC to indicate to the BTS the power level value when a channel connection is initialised. 6.1.5.2. T i m i n g Advance The time division multiplexing scheme used on the radio path GSM is such that the BTS must receive signals coming from diff mobile stations very close to each other. I n order to reach this despite the propagation delay incurred by the return trip from BTS, mobile station, and taking into account that guard times between have been chosen very small f o r spectral efficiency, a mechanism compensate for the propagation delay is necessary. To this avail, mobile station advances its transmission time relative t o its schedule, which i s derived from the reception o f bursts, by a indicated by the infrastructure, the timing advance. Once a dedicated connection has been established, the continuously measures the time offset between its own burst schedule the reception schedule o f mobile station bursts. Based o n measurements, it is able to provide the mobile station with the requ timing advance and does that on the SACCH at the rate o f twice p f second. The timing advance can take values from 0 to 233 ps, which is enough to cope with cells having a radius of up to 35 km without any other special scheme and given the speed of light. This limit comes from coding considerations (the timing advance is coded between 0 and 63, with the bit period as the unit, hence 233 ps), but there are mak important hidden limitations. A first point is the guard time for access' bursts, which in practice limits to about 220 ps the possibility for the It:\ DIO RESOURIT \ I \ \ I I . \ 3 . 4 7 inititg..`propagation time measurement. The other point is that some minimum time is needed between the end of the reception of a downlink burst and the beginning of the transmission of the next uplink burst. in order to allow the implementation o f mobile stations with the same frequency synthesiser for emission and reception. Even in rural or low-density areas, good coverage quality will in practice require cell radii smaller than 35 kin. I however, there are rases when larger cells would be useful. This holds i n part; F o r the coverage o f inshore coastal areas, where high antennas (e.g.. o n lighthouses) could be in sight o f boats more than 35 km away on sea. Such uses are indeed possible. at the expense of the number of channels per MHz, The trick consists in obtaining a huge guard time (more than 580 ps) by using only every second burst. I n such cases. only the channels of even TNs can be used (since TN 0 must be used for the BCCH, odd TNs will not do). This feature requires a specific reception processing in the BTSs. Upon establishment o f a new dedicated connection, the timing advance control process must be initialised. This happens at each initial assignment, subsequent assignment or handover. Depending on the case. the mobile station and the infrastructure do not always have the same amount o f information t o assess the new timing advance, and the initialisation method varies accordingly. The different cases will now he examined one by one. Both Mobile Station and Network nut assess the Timing Advance Beforehand This s'ts l. ilion happens upon subsequent assignment. The mobile station simp : uses on the new channel the old value o f the timing advance which was ordered by the BTS on the previous channel, On the infrastructure side, the BTS device in charge of the old channel.is aware of the last ordered timing ad% mice. but there is no communication means betv.een BTSs. The BTS is indeed not aware that the old and the no\ channel cotifiection concern the same mobile station. The 13S(' is kept informed of the last ordered timing advance value, and it is therefore able to transmit i t to the new BTS device when activating it. The timing advance control process just resumes on the new channel \\ ith the same values as on the previous channel. There exists another case where both mobile station and network are able t o assess the timing advance beforehand. This happens at 348 T H E RAN() Riisot Ruh mANA( GSM SYSTEM handover between synchronised cells which are collocated. However, the Specifications treat this case as i f the cells were synchronised, but,not collocated (see next paragraph). A more efficient scheme could have beck to give an indication to the mobile station that the old and new cells ate collocated, and such an improvement may indeed be made in the fat h) Only the M o h 4 Station can assess the New Timing Advance Beforehand 349 the difference in propagation times. and therefore to calculate the new timing advance to he applied. The indication that two cells are synchronised is therefore enough for the mobile station to assess the new timilig advance. as follows (based on the notations of figure 6.15): TA2 = TA I 2 (prop1 - op;21 This case arises when handover i s performed between ITS synchronised cells which are not necessarily at the same location.ntibi mobile station is then able to measure the difference between arrival times of bursts coming from the two BTSs. It must indeed do so for r i g synchronisation requirements (see page 333). This arrival time offset tail combination o f the transmission time offset between the two BT (which does not depend on the location of the mobile station), and of tilt two propagation times, as shown in figure 6.15. Therefore, if the mobile station is given the transmission time offset between the two BTU (which is zero, by definition, for synchronised cells), it is able to d a i i On the infrastructure side, the new timing advance cannot he computed, except in the case o f collocated and synchronised tells. A possibility would have been to wait for the mobile station to indicate this value (we will see later that the mobile station indicates hack the value of the timing advance it uses. at least once per second). Nevertheless. to allow for a possible slightly faster initialisation on the new channel. the handover procedure between synchronised cells •includes t h e 111C.Ift, for the new BTS to assess the propagation time with the mobile station. To this avail, the mobile station starts transmission on the new channel with a few access bursts sent with a null timing advance. before switching to normal transmission. ate c)Neither Mobile Station nor Network are able 10 assess the New Tinting Advance Bcjinvhancl timeoffset n prop2 prowl timedifferenceexperiencedby MS Figure 6.15 - Time offset between two BTSs Asseen by the mobile station, the arrival time difference between bursts coming from two BTSs is made up by the difference of the propagation times, plus the offset between clocks of the two BTSs. At initial assignment. or at handover between two cells which are not synchronised, no information can he used by either side to predict the timing advance. Signalling messages are different. hut the timing ads once initialisation process is very similar in both cases. The mobile station is forbidden t o transmit normal bursts until i t knows the new timing advance to apply. Because the BTS (whichis the emit). which decides on the tinning advance) must receive something from the mobile station in order to assess the propagation time, the mobile station is required to send access bursts to the BTS with a null liming-advance. When the BIS receives such a burst. the reception instant is a measure of the double propagation time and the BTS can derive the value of the Inning advance. which it sends to the mobile station in a signalling message. From the moment i t receives this message. the mobile station is able to start correctly transmitting normal bursts. T h i s exchange lengthens t h e duration o f the handover procedure between asynchronous cells compared t o the synchronised case described earlier. Moreover. the "asynchronous" handover leads to a longer communication interruption than its "synchronous" parent. 350 51)10 k ( ( vs.v(iitvirst THE GSM SYSTEM 6.1.6. RADIO CHANNEL MANAGEMENT . 1 . 4 1 1 •mot! So far, we have seen how individual RR-sessions are manag Though R R -sessions are independent, they share the same pool; i t resources, in particular the radio channels. In this section we will locitag the management of the radio channels in a cell as a whole, dealing with such problems as the configuration o f the channels and the chitood allocation strategy. The management o f the set o f channels to be used in each includes two main aspects. On one side, the set of channels of each must be determined and the machines need to be configured accor This "long-term" aspect is the cell channel configuration manage On the other side, the channels g o through allocation/release c following the communication needs of the mobile stations. This "s term" aspect i s the dedicated channel allocation management. channel configuration management and channel allocation are .,ti responsibility o f the BSC. The MSC only intervenes to indicate w type o f channel a given communication requires, whereas the B'!! executes different related tasks, but always under control o f the B A Both areas have strong impact on the procedures used over the r a f t interface and the Abis interface and will be described here. 6.1.6.1. C e l l Channel Configuration The channel configuration of a cell is the list of channels defined* a given time to be used in the cell. A typical cell configuration inclu set o f common channels to support mobile stations in idle mode sod initial mobile station access (a BCCH, a PAGCH and a RACH) and aleit of traffic channels (TACHs o f various rates, including what ..01; Specifications call TCH/F, TCH/H and SDCCH) for carrying sign and user data. The channel configuration of a cell may change in These changes m a y have various degrees o f impact o n traffic management, i.e., on the allocation and release o f channels used for specific communications. On one side, some modifications are related to the evolution of the whole network, f o r instance a capacity extension t o cope w i t h * increasing traffic density. Such changes are clearly within the scope$ network operation. However, because network operators appreciate the possibility t o handle such changes without disturbing the existing ongoing communications, mechanisms have been introduced in the traffic • 3 5 I management area for this purpose: these mechanisms will be described here. Configuration of the Access Channels Depending on the f u l l spectrum capacity o f the cell. usually estimated in terms of number of frequency slots. the capacity requirement for access channels (RACH, PAGCH) will vary. The Specifications cater for five different access channel capacities for a given cell. which are .summarised in table 6.2. if hese access channel capacities correspond in terms of radio consumption to a range from the equivalent of a half-rate traffic channel to the equivalent of 4 full rate traffic channels. Since the mobile stations are assumed to be able to listen to only one unit of spectrum usage (the equivalent of a TACH/F, i.e.. one slot every 8 time slots) at a given moment. they must he distributed among I to 4 population groups, depending on the capacity of the access channel structure. Mobile stations find the information about the applicable structure in the messages broadcast on the BCCH. These messages are sent on every carrier unit used for common control channels tot' TN 0. 2. 4 and 6 as applicable). The access channel configuration may change in time. and detailed procedures are provided in the .Speri/ications to cope with such dy mimic changes and ensure that the transition period between t w o stable configurations is as limited as possible. This impacts mainly the listening to the PAGCH in idle mode. number (ir kActi \ IS groups Jturst rate etti I capacity teguiv_ Ha 4 i hor•i. per wcotid I T M '11/1: I I,‘(;( it message talk, 'WI ,....Ilki , 1/'' I I 14.- 1 I 216.7 3s.2 433..4 76.5 65(1 I14.7 566.7 152.9 _ I 4 Table 6.2 A c c e . s channel capaciCe, Depending on the capacitc to he offered 1.1 a given cell, the access channel,: a cn be configured in set vial differeni sires. 152 T H E (3551 SYSTEM 12 1010 kES111 12('E NIANA(11.%11.. \ 1 The corresponding procedural aspects w i l l be covered i n • the section dealing with paging procedures. • i t • Organisation o f the PAGCH •): On each CCCH "unit" to which the mobile stations are able)) listen, the downlink paging and access grant channel is organised initwo parts: • several "paging sub-channels", in a one-to-one relationship with sub-populations of mobile stations, on which initial assigrunini messages can also be sent. ' a t if • possibly a sub-channel reserved exclusively f o r assignrmin ••30q messages; J004 This PAGCH configuration is indicated to the mobile stations in .„ messages broadcast o n the BCCH, i n order f o r mobile stationsAn determine where to listen for their own paging messages. The PAGCH configuration may change dynamically, and a mechanism is definedIn the Specifications to enable such a change while avoiding the risktoe mobile stations to lose paging messages during such changes. Both the corresponding contents of the broadcast messages and the allocation Of .::g> : paging messages to sub-channels are controlled by the BSC. - J c o n .istb. These procedural aspects w i l l be described respectively i n ;Abp sections dealing with general information broadcasting and with paging procedures. J at : I Traffic Channels Configuration Another point i n t h e area o f c e l l channel config management is the possibility to modify dynamically the set of traffic channels to meet the demand more closely. For example, the resource -Used at a given moment in time for a TACH/F can also be used for 8 TACH/8 at some other time. This kind of choice can be under control of the operation and maintenance sub-system (as a result of medium or longterm traffic analyses), or alternatively may be fully implemented in the BSC, so that the allocation process o f the BSC would perform do conversion when needed. For instance, if a TACH/8 is needed when nage is free but if a TACH/F is available, the latter could be converted WM, pool o f 8 TACH/8s instead o f rejecting the request. The Specification leave complete freedom t o the operator/manufacturer to choose the implementation anywhere between these two extremes. 353 These functions do not impact directly. the communications in progress, and are totally internal to the BSC. Ilence they are not visible on the connection management protocols. Changes in the Frequency Configuration The previous paragraphs have dealt with changes in the functional configuration o f channels, within a given pool o f time/frequency resources. But that is not the end o f the story: the frequency slots allocated to a cell may change dynamically in time. even though this is presumably not a very frequent event. In the case when only single-frequency channels are used, a change in the frequency allocation of the cell will impact the communications making use of any suppressed frequencies, but each such communication can be handled independently from the others. However. when frequency hopping is employed, a frequency is used in a (en tight!) co-ordinated way by several connections in the cell at the same time. Any change in frequency affecting a given connection must b e precisely coupled wish similar changes to other connections in order to keep the non-interfering properties of the channels. These changes must happen in a synchronous way. With this objective, specific mechanisms have been introduced in the Specifications t o enable the synchronised n u o f the frequency allocation of many channels. These mechanisms include the possibility t o order a precisely timed change of the frequency parameters to the mobile stations and to the43TS for all impacted connections, as well as to have precisely timed chgrinel assignments ( w h e t h e r i n i t i a l assignments. subsequent assignments or handovers). The indication of the instant of change relies on the cyclic numbering scheme of TDMA slots, which has a period of about 3 and a half hours and allows an accuracy of microseconds. As seen from the mobile station. these changes appear simply as changes of channels to be performed at a defined instant. On the l3SS side, the matter is somewhaltmore complex. The aim is to synchronise the behaviour of several m o t e stations and o f the HIS, using signalling means which are by nature asynchronous and subject to losses due to transmission errors. The operation must he performed in several steps. First, the general decision to perform a frequency change comes from the operation and maintenance sub-system, for such reasons as the setting up o f new hardware, o r the need f o r removing some f o r maintenance, o r due t o observations o f unplanned interference. The 354 \itiot(F)att 1 . \ x . v. t y THE G5N1 SYSTEM decision to modify the frequency organisation of a cell is notified to the BSC, which is then in charge of co-ordinating mobile stations and BTU, to reach the new coherent configuration. J C; f The first step f o r the BSC is then to determine the transition instant. This instant must not be too far away, to avoid introducing ambiguities from the cyclic numbering scheme. However, it must not pe too close, in order to ensure that all concerned mobile stations have ei received the command or have had their ongoing connection releas The concerned mobile stations are those engaged in transmission channel whose frequency parameters are affected by the change, those to which such a channel has been allocated before the a transition time: a timed transition order must be sent to them. There remains the case of the new allocation of channels befo actual transition time: such procedures are started, but with an indiC. that the mobile station will go on the new channel only at the trans' time. Thus, all mobile stations involved in a connection on one of, impacted channels, as well as the BTS, perform the transition when transition time occurs, and a normal situation is restored. 1 To summarise, the change of frequency configuration impacts all the assignment procedures, and require a specific procedure to deal id* existing connections, called the frequency redefinition procedure. corresponding details will be found in the sections dealing with eachal these procedures. ; O S 6.1.6.2. Dedicated Channel Allocation N i t .314 The second component of the management of the radio channels seen as a set is how the dedicated channels (TACH/8 and TACH/F)", chosen when allocated to an RR-session. As seen by the infrastructure. dedicated channels are at a given moment either allocated to the use of a _mobile station. or part of a pool ()I' idle channels from which a channel is drawn when a new need appears. To summarise the previous sections. such a new need may appear in three different conditions: • a t initial channel assignment. when a mobile station in idle mode has some communication needs, for instance because the user wants to set up a call, or because location updating must be •ft, performed: .e.;• • a t subsequent assignment, when what the communication needs does not correspond any more to the type o f channel itAls allocated, for instance when a TACH/8 was allocated at initial assitmment and a speech call needs to be connected: 3 5 5 • a t handover. when the movements of the user or the variations of the interfering level result in a situation where the connection would be better on another channel. often through another cell. Allocation Strategies From the mobile station point o f view, these various kinds o f channel assignments are simply orders to start transmission and reception on specified channels. From the infrastructure point o f view. the allocation of a dedicated radio channel involves two steps, first the choice of the channel to use, second the actual transition. The choice o f the allocated channel lies entirely within the responsibility o f the BSC. Sophisticated algorithms can be esigned is to try maximising the total amount of traffic which can be ser d with a given amount of resources. While maintaining a reasonable fairness level in the granting of requests. Allocation optimisation includes several aspects. A firs) one is related to the relevance of the type of channel which is allocated for the effective need. This leads to a real pr obl emin the case o f initial assignment, since very little information is available at the BSC to choose the type of channel (the mobile station gives only a rough description for its reason to access i n the initial access request message). A typical example concerns the setting up of a call: a TACH/F will he required to transmit user data, but a TAC7H/8 using 8 times less resource, \\ mild he enough until t h e correspondents have begun conversing. Several strategies can be chosen. which can be grouped under the three following categories, as also shown in figure 6.16: • Ve r y Early Assignment consists in allocating a '1AC11/1: at initial assignment. when i t i s probable that the requested connection will need such a channel: • Early Assignment consists in allocatin.:, a TA C I r s iniiiall then subsequently allocating a TA('11/12 as soon as it is knov, n for sure that this type of channel will he required: • O f f Air Call Set Up tOA('Sl,t consists in allocating a'FAt'IIiS initially. then waiting until the called party has answelvd heron: attempting the subsequent assignment of a TA('II/F. These different methods each have their pros and cons. )) filch fomented many debatesover the specification years of GSNI. OACSU differs from the two other methods in that it provides the users with a different grade o f service. The correct channel may he allocated a noticeable amount of time after the called party has answered, denending THEGSM SYSTEM 356 access f u l l VEA EA V L. " • TACH/E • . \ I A N . \(ir. \ I 3 5 7 Whatever the allocation strategy. there are cases where no adequate resource is available when needed. The network may then apply one of two strategies: either the request is rejected. relying on its originator to possibly retry later. o r the request is put aside to he served when a suitable channel becomes free. The latter strategy i s referred t o as "queuing", although there is not necessarily a queue in the strictest sense of the word. called party answer call request i n f o r m a t i o n V A n n ) RESHI ;),' Queuing 'TACH/F4 The interest of queuing varies with the conditions in which it is applied. At initial assignment. the repetition scheme put in place to cope with losses due to collisions or bad propagation conditions reduces the interest of queuing to nothing. Using queuing at initial assignment could even have adverse effects, since a request which is not answered in a 1ahort time will be repeated by the mobile station (the normal reason for noanswer being transmission loss), and this could lead to very inefficient multiple assignments. TACH/E OACSU Figure 6.16—Assignment strategies at call set-up it Very Early Asi,ignment (VEA), Early Assignment (EA) and Off-Air Call Set Up (OACSU) represent three different allocation strategies, between which operators may choose. Queuing could of course have been made possible at initial assignment. b) introducing a "please wait" acknowledgement to the 'while station request. but this has Dm ken introduced for simplicity reasons. The a Art IMM. 1110k element appearing in a messageanswering a request for access should not he mistaken l'or suchascheme: it is in fact a temporary rejection preventing the mobile station from making no) attempts for some time. some announcement to the called party, who may otherwise wonder why the phone rang at all! The grade of service which results is considered by, many operators as unacceptable, though opinions diverge on the issae: On the other hand, OACSU is certainly the most efficient of the three schemes in terms of resource usage. The dilemma lies (as often) b e t a efficiency and user comfort. The major interest o f queuing i s indeed limn(' in the rase ()I' subsequent assignments. The only resulting drawback is the lengthening of the call set-up time (this is perceived by the calling party when earls: assignment is used, and by both parties when OA( i s used). However. this is to be balanced with a rejection o f the call. probabh more illperceived by the subscribers! as The main drawback of early assignment compared to very early assignment consists in an increased call set-up time, with no real gain it terms of channel usage in the case when a TACH/F is actually needed (t should be remembered that a measure of channel usage must take into _account. not only the channel size (spectrum consumption), but also the usage duration, and that signalling exchanges are quicker on a TACH/F than on a TACH/8). On the other hand, very early assignment is ver) inefficient if a TACH/F is allocated when not necessary, e.g., for location updating. In this case, the amount of time during which the connection must b e kept i s mainly determined b y the duration o f signalling exchanges between infrastructure entities, and one cannot expect a significant reduction b y using a larger channel. Therefore, very early assignment is of interest only i f enough information on the use of the required connection is available to the network before initial assignmem (which is not the case in the phase I version of GSM). For handover, the picture is not so clear. If the handover is decided to salvage a rapidly degrading situation. queuing cannot help much since the connection w i l l most probably be lost i f the channel cannot he allocated right away: but i f the connection is not in immediate jeoparth, and the handover decision jUsi stems from general optimisation reasons, queuing can be a source of improvement. Thus. a correct use of queuing at handover requires that the handover process distinguishes several cases. 1 I Fairness between the treatment o f requests is to be sought when applying queuing. When no other consideration intervenes, the order of service is usually first-come. first-served: (here ‘‘e have the classic queue. More sophisticated schemes can also he devised. granting requests on the 358 THE GSM SYSTEM priority than re-assignments, and emergency calls are considered more important than other communications. 4 Even without queuing, there are ways t o bias the grantineof channel requests in congested situations. For instance, depending on the state o f channel congestion, i t could be worth rejecting the channel requests for outgoing calls to privilege those for incoming calls (a mobile user is more likely to answer than a fixed one i f the mobile station is switched on, and the end-to-end circuit is almost entirely established already in the case of an incoming call). An even more drastic approach consists i n forcefully terminating a connection estimated o f a low importance, in order to reuse the resource for some needs deemed more important. This approach is referred to as pre-emption, and should be used with great care, given the negative impact on the preempted user. For all the above reasons, some mechanisms have been introduced in the Specifications t o enable priority strategies f o r radio channel allocation in the BSC. These mechanisms consist in conveying categ4. priority and pre-emption indicators, which can be used to influence,the allocation decisions i n cases o f congestion. The Specifications do tot describe how to use these indications, since the allocation schemes we left to the operators or manufacturers. Only their transport on interfacern specified, and their actual use is indeed a source of difference betwec. ormi the products of different manufacturers. Minimum indications relative to the purpose o f the access:en provided by the mobile-station to the BSC at the beginning of the access procedure, to allow some priority mechanism for the initial allocatrk Similarly, the MSC may provide some information when it reque change of channel type (subsequent assignment procedure). In the c a s e handover the requesting BSC can also provide the MSC with soa information about the reason for handover, and this reason can be c up to the target BSC. To summarise, a few things here and there in the procedures all sophisticated allocation algorithms. with queuing. priorities, b u t no -constraints are put by the Specifications on the infrastructure equipment in this area. It is up to the manufacturers to include such functions, or for operators to request them. Interference Considerations The actual channel to be allocated may also be chosen with the optimisation of the transmission performance in mind. This requires the )10.Y \ I\ \ I 3 5 ( ) BSC to have some knowledge beforehand about the performance o f a connection for each of the free channels. Performance depends on ninny factors, most of them difficult to assess before the effective usage of the channel. However, there i s one which i s accessible: the level o f interference in the uplink direction. When a channel is not allocated to any connection, the level received on the channel gives an idea of the 1evel of interference and noise. GSM opens the possibility for the BTS to measure this reception level on all unallocated channels. and to transmit them regularly to the BSC. The BSC can then take this information into account i n order t o allocate a free channel o f minimum uplink interference level, or to decide on an infra-cell handover if it is noticed that an active channel suffers a higher level of uplink inteitterence level than the free one. It should be noted that this feature is in most cases of secondary interest. The concern of an operator is to obtain a system able to support a maximum capacity: thus the situation to optimiw is congestion. I f all the channels o f a cell are used a t congestion land this i s the usual assumption), the allocation o f the channels starting w i t h the less interfered changes nothing to the eventual congestion situation. The only gain is the improvement of the average performance when far from the ongestion state. . However, the assumption that a l l the channels are used a t congestion does not always hold. One can imagine a cellular planning where the number of channels allocated to a cell is too high. in the sense that i f all channels are used i n all cells the overall quality o f the —eblftlations is not acceptable. With such an approach congestion happen; before all channels are allocated. A first consequence is 111;11 the 13SCs must take care not to start additional connections when the congestion Mate is reached, even if channels are available. Another consequence is that taking uplink interference levels into account is then meaningful. What is obtained is a kind o f automatic channel planning between the cells: the use of a channel in a cell 1‘ ill result in some interference level in other cells. thus preventing the latter t o EN: interfered (and then interfering) channels. This is essentially a dynamic channel allocation method: it can be easily shown by taking the e ; twine assumption that all the channels are allocated to all cells. Without going to this extreme. small over-dimensioning of the cell capacity plus this dy mimic channel allocation method can he useful to cope with an abnormal distribution of ehe traffic between cells. I f a cell is overloaded. but i f the cells using interfering channels are not. the congested cell can w. ith this approach use more channels than if all cells were equally overloaded. 360 m i v GSM SYSTEM Pwin tots( ll Rri: 1 1 ARFCN . sI 3 6 1 Radio Channel Description 1 A side issue related to channel allocation is the way radio channels are described. The problem lies with frequency hopping. As explained in detail in Chapter 4, channel characteristics include time and frequency parameters. The time characteristics are not problematic; i n the time domain, there are only 8 types o f TACH/F and 68 types o f TACH/8, depending on the type of channel and the offset relatively to the reference clock. One octet is therefore enough to code which of these 76 families of channels is being referenced (92 if half-rate channels were included). In the frequency domain, things are somewhat more complicated. In the case of single-frequency channels, the number of different cases is 124 for GSM900 and 374 for DCS1800. In order to cope with some evolution, this frequency is coded on 10 bits, leading to a total of 18 bits to encode any single-frequency channel. However, when frequency hopping i s being used, the number o f combinations explodes. For GSM900 alone, a rough assessment shows that there are about 1 1 23 Cell allocation 4 2 10 V I 12 5 12 L I L L 34 5 1 1 4 12 1 3 : 3 Mobile allocation Figure 6.17 - Cell and Mobile Allocation The cell allocation is the portion of the total resource u:ahle in a cell. with regard to which catch mobile allocation may be defined. 66,141,633,339,297,631,280,564,218,.199,442,383,724,544 different possible hopping sequences compliant with the Specifications (including sitigle frequency cases). This value needs 135 bits to be written in binary format. The matter is exponentially worse for DCS1800. This causes a problem because of the size of signalling messages, which is particularly critical on the PAGCH on which all initial assignment messages are sent. Some kind of reduction was therefore desirable. But which reduction? First, what is needed to describe a hopping channel? The list of frequencies used by the channel, obviously. This frequency list, called the MOBILEALLOCATION in the Specifications, is evidently the main source of length. The description also contains two other parameters used for the computation of the hopping sequence: • t h e Mobile Allocation Index Offset (MAIO), of which there arc as many possible values as there arc frequencies in the list hence its name since i t describes the starting point for the hopping recurrence function: and • t h e Hopping Sequence Number (HSN), which can assume 64 different values. These two parameters fit on at most 13 hits. The real coding problem lies with the coding of the frequency list A first simplification consists in noting that only a set (i.e., an unordered list) of frequencies is needed to define a channel. The hopping sequence generation algorithm refers to a list o f frequencies. but the ordering is implicitly defined based on the respective value o f the frequencies in Hertz. Therefore only sets are required. reducing the eodMg requirement to 124 bits. This is still too much compared with the constraints of the PAGCH. Two approaches were possible: either restrict the number o f possibilities, or design an efficient and sophisticated signalli mi: scheme. Restricting the number of possibilities was inescapable fin' I)('S Im10. The maximum number of frequencies used I)) zin given channel Y :tries from 32 to 64 depending on the frequency range in which the channels are spread. B u t there i s still w i t h this limitation more frequency sequences in DC'S I 800 than in GSN.1900. There remains the m p h l i c a t e d signallin;2 approach. A l l , limincls in a given cell use only those.frequencies t . Mich an allocated lo Ill: cell by cellular planning. I f this usually rather short list can be biordcast to mobile stations independently 110111 ally allocation, the description n t a given channel in a known cell needs only to indicate \‘ hid: of :IR...L. cell frequencies are being used. This "R‘o-step- mechanism :see figure t1.171 leads to the concept of cell allocation. or cell channel description. v. hich are the terms used i n I s L ieSperificwionA t o designate the set o f all frequencies which are used in a given cell. (In fait. the main use of the cell allocation is for the initial assignment. so the cell allocation can he 362 THE (ISM SYSTEM allocated as initial channels.) The coding of the cell allocation uses 16 octets in GSM900 (where it includes a bit map for all 124 frequencies), and up to twice as much in DCS1800 (where the coding algorithm is` much more sophisticated). This cell allocation is broadcast regularly on the BCCH. When the BSC sends a channel allocation message to the mobile station (whether at initial assignment, subsequent assignment '• handover), it is able to encode efficiently the channel frequency list a y ; subset of the cell frequency list. I f the latter includes n frequencies, ttpe encoding of all possible subsets needs theoretically only n bits. The allocation in GSM900 is limited to less than 64 frequencies (since the map in the mobile allocation is limited to 64 bits). This is not really constraint, since all frequencies cannot be used in all cells. In the vast majority o f cases, the cell allocation will not con more than 32 frequencies; indeed, the gain brought by frequency hopputt increases very little with higher numbers. Some encoding schemes h a l been introduced in the Specifications to encode efficiently the frequency set when the cell allocation consists i n such a small number of frequencies. For historical reasons, two such schemes can be found: one for GSM900, one for DCS1800. The first scheme is reflected in what S called the FREQUENCY CHANNELSEQUENCE element in the Specificatio This element allows to bypass the two-step approach described ahoy and is used at handover. Afterwards, another scheme was designed cope with more frequencies, whilst maintaining compatibility with first one. The latter scheme is indeed able to cope with 1024 diffe frequencies, leaving some room for extensions in the future of GSM related systems... 6.2. ARCHITECTURE AND PROTOCOLS The radio resource management functions are mainly dealt with by the BSS, and in particular by the BSC which acts as the orchestra conductor for these functions. It directs the mobile station as well as the infrastructure machines involved, which are more or less slave (in the RR field), i.e., the BTS (and TRAU) on one side, and the MSCs on the other. The BTS and the T R A U are indeed the main performers i n the transmission chain, and they must as such be controlled by the Radio Resource functions, in fact by the BSC. The BSC is little involved in the transmission functions, b u t takes care o f the consistency o f the transmission chain, whether for its different properties (transmission • " • DI( I RESOl N I A N A C I • M I A 363 mode, cipher mode, ...) or for the quality o f transmission thandover preparation and execution co-ordination). If it is the anchor MSC which decides which properties o f the transmission chain are desirable to fulfil the service. the role of the MSC for radio resource management is limited t o some o f the handover aspects. The anchor MSC is in charge o f performing subsequent interMSC handovers when so decided by the serving BSC. The relay. MSC is in charge of the handovers when between two cells of different 13SCs under its control, and of the circuits between itself and its 13SCs. For all other functions, the relay MSC actsa s a transit node fur signalling exchanges between the anchor MSC and the BSC or the mobile station. Functionally the relay MSC functions are all within the Radio Resource management realm. This general description of the roles of the different nodes has to be refined for the handover function. Handover preparation makes use of a number of pieces of information, described in the above sections. and which originate from different sources. The 131'S is the infrastructure iliW-point for all measurement information (both for reports from the mobile stations and for its own measurements). The I3SC is the warden of all frequetncy planning and cell layout data. Information about traffic is spread between the BSC and the MSCs. Thus, whatever functional split is chosen, some real-time information transfer is needed bet een these entities in order to obtain a coherent handover strategy. The basic split lies between the BSS (BSC + i n s ) and the MSCs. The general rule puts the BSS in charge o f the management o f radio resources and o f the decision to perform a handover on a given RRsession. The split was not clear from the start i n the standardising process, and triggered many a discussion. The main problem between the MSCs and the BSS is taking traffic into account (either to influence cell choice, or for traffic handover). One could distinguish two approaches: either to have the MSCs indicate to its BSCs the level o f traffic o f surrounding BSCs. o r t o let the \1SCs intervene i n the hdtltitt er_ algorithm to ponder radio criteria coming from the BSS with traffic considerations. The second approach was finally the one chosen. even though it may not be the simplest one. The intervention of the \ISCs in the handover process blurs somewhat the functional border IVtween BSC and MSCs, and their respective role is not easy to describe. especially for traffic handovers. This is nevertheless what we will attemong'.do now. First o f all, when a BS(' decides thai an outgoing handover is necessary, it will indicate to the relay MSC one or more target cells, possibly managed by different BSCs and MSCs. Rescue handovers might call for several target cells, whereas confinement handovers obviously 364 THE OSM SYSTEM may just try them one after the other, in the indicated order. Or it may choose among these ranked possibilities, taking into account its oust traffic data. For a cell controlled b y another MSC, the relay MSd forwards the request to the anchor MSC; at this level only one target cell is proposed at a time. The choice between several targets is thee •10 function of the relay MSC not of the anchor MSC. Another possibility f o r traffic handovers also exists i n Specifications, allowing the relay MSC to force the BSC to hand over,* portion o f the traffic from a cell, with the BSC being i n charge rof choosing which connections to hand over and to which cells; this is called the "candidate enquiry procedure". There also this possibility is not opal to the anchor MSC. In all cases, the conflict between confinement criteria and traftt2 criteria is obvious, and is not solved since data related to these criteria ire under the control o f separate nodes. I n fact, there exists in. the Specifications a means to group all data in a single place. It consists 'in transporting up to the relay MSC all the raw measurements coming finis the mobile station and the BTS, and leave everything to the relay MSC for decision and choice. However, this possibility is not implemented by any manufacturer, and will disappear in phase 2, since the incurred load on signalling links and on the MSC processors would be formidable. Aolit The functional split inside the BSS was also an area o f I A ' debates. The split can, there again, be done in different ways. One of the options is to group all the processing in the BSC, with all measurements being transmitted by the BTS without it performing any computation on them. The advantage of this solution is the centralisation of all data, so that handover decisions are taken based on the best data available. The drawbacks are the high signalling load on the Abis interface, since the incurred load of 2 messages per connection and per second represents by far the dominant traffic o n this interface, and also a n important_ requirement f o r high computational power i n th e BSCs. Another approach, referred to as "pre-processing" in the Specifications, consists in - - - —Felling the BTS do an important part of the job, thereby relieving the Abis interface froM the major part o f its traffic and decentralising the computing load. Pre-processing i s a n accepted alternative i n the Specifications, but the exact functional split between BTS and BSC in this case is not specified. Messages and information elements exist to cater for this function, but their semantics are left open to operators and manufacturers. Pre-processing is often thought of as a way to reduce load on the Abis interface simply by making the BTS perform some averaging on the measurements. Such a simple approach could slow down the handover decision process significantly, which can be a source of inefficiency, in R A D I O REsorRcE NIANAt 3 6 5 particular in a small-cell environment. A correct usage of pm-processing introduces sophisticated algorithms in the BTS, including some part o f the decision-making process. This is the main reason why details are not specified. It would indeed require too great a workload before a scheme could be shown and accepted by the specification committees. • Alt these different options related t o the split o f handover preparation functions within the infrastructure are taken into account in the Specifications. More details about the relevant procedures can he found in the SACCH procedural section. as well as in the handover execufron section. • Protocols Independently o f t h e infrastructure architecture, t h e implementation o f the RR functions requires some kind o f protocol between the mobile station and the network. On the network side. the interlocutor (or peer entity) of the mobile station for this protocol is the relay MSC a n c h o r MSC B' BSC MSCVLR M S C RIL3-RR - - r : - L. • RR BSSMAP M A N E •••• TCAP AO. ••••• a m . N O . ••••• e a . LAPDm I a m . SCCP SCCP MTP APD Figure 6.IS R R protocol architecture Protocols for RR inanagenicni ;ire needed on man> interface‘. including the A interface A h i p and the NISC-NISC inierface I : \ I / I ' . 1 . The main co-ordinator k the liSC. vLR 366 T H E GSM SYSTENI BSC (in fact, a small part of the signalling is also handled by the BTS for efficiency reasons). This protocol will be denoted RIL3-RR. The functional distribution between infrastructure entities calls for other protocols on terrestrial links: one between BTS and BSC, one between BSC and relay MSC, and one between relay MSC and anchor MSC. The first one, on the Abis interface, is used f o r the BSC to configure the transmission path and for the BTS to report measurements to the BSC. It has no official name in the Specifications (experts refer simply to the 08.58 protocol, from the number o f the corresponding Technical Specification), and will be here referred to as RSM (Radio Subsystem Management). The protocol between BSC and relay MSC, on the A interface, is used to carry the requests for initial connection establishment, as well as for any change i n the connection attributes according to upper layer requirements. I t is used also for handling handovers between the relay MSC and the BSC. This BSC-MSC protocol is called the BSSMAP protocol (BSS Management Application Part). R:\1111) l i S O I k r t . NIAls..-\( N 3 6 7 mobile station. Next w e d i l l present the paging procedure. which precedes the access when the requirement fur a connection comes from the :infrastructure side. The following sections will be devoted to what happens during the life of an RR session. One aspect concerns the change of some characteristics such as the type of channel or the ciphering mode. when requested by the anchor MSC. Another aspect discussed is the execution o f a handover, w i t h a l l i t s variants, including call r e establishment, presented here as a last-resort kind of handover. The study of the main adventures o f an RR-session w i l l end with the release procedure. We will then discuss a number of ill-assorted procedures. such as the procedural handling of the signal measurements, the timing advance and the transmission power control, which is done on the SACCH: the frequency redefinition procedures. which are rather complex mechanisms to cope with a change i n the allocated frequencies i n a cell when frequency hopping i s used; and finally the broadcasting o f various information on the BCCH. The last protocol, between two MSCs of adjacent coverage areas and supporting the exchanges between relay MSC and anchor MSC, is part o f the MAP and will be referenced MAP/E protocol. Figure 6.18 shows the machines involved i n radio resource management and the protocols between them. In all cases, in this chapter and the following ones, what normally happens will be presented with some details, without going, however. into the internal structures of the signalling messages. The rarer cases. involving failures or collisions between events, will at best be hinted at, though they are maybe the most important aspects to have in mind when designing signalling protocols. But even a minimum attempt to correctly cover this subject would need a far larger text than can he included here. 6.3. RR PROCEDURES - H A . INITIAL PROCEDURES:ACCESS AND INITIAL ASSIGNMENT In the first part of this chapter, we looked at the different tasb needed for the management of the radio resources from an "object" point of view. We have studied how for instance the channels are managed, or _ how timing advance is controlled, but all these topics were seen rather independently. In this second part, we will look at the details of what happens at different moments, combining all these independent aspects. Most procedures in the RR area are concerned with several functional aspects simultaneously, and involve, o r may involve, all machines between the mobile station and the MSC. We will then revisit a number of the topics we have seen in the previous sections, but with stress on the temporal relationships. The purpose of this section is to describe the procedures enabling the transition between the two major states of a mobile station. i.e.. "idle" mode (where the mobile station is everything but idle. but refrains f l o o r , . any active transmission towards the infrastructure). to "dedicated- mode where the mobile station is actively transmitting on a channel allocated for its own use. The transition corresponds to the establishment of an RRsession (see page 313). We will start at the beginning, that is to say with the access procedure, where the RR session is created. This will be the occasion to look in detail at the use of the RACH for the initial contact from the The initial assignment procedure is always triggered upon the request of the mobile station. for one of three major reasons: • t o perform location updating: • t o answer to a paging: or 368 '?xhio nrsut kcy xus:sm mxtr.N.r THE GSM SYSTEM RACH PAGCH (or IMMEDIATE ASSIGNMENT EXTENDED) TAC H 6 9 mobile stations w i l l need t o communicate, and therefore this first message from the mobile station cannot be scheduled to avoid the simultaneous transmission of more than one mobile station (a collision). This is the major problem o f random access schemes, and the channel name (Random Access Channel, or RACH) indeed expresses the f a d that mobile stations transmit independently from one another. CHANNEL REQUEST IMMEDIATE ASSIGNMENT 3 link establishment initial message Figure 6.19 — Initial access procedure The transition from "idle" to "dedicated" mode is always triggered by the MS, through anRIL3-RRCHANNELREQUESTmessagesent on the randomaccess channel. Only when the signalling link layer hasbeen established and an "initial message" sent on the new dedicated channel does the network know the identity of the MS. 04: • a s a result o f a user's request, i.e., for an outgoing call, a supplementary service management request, or the sending of a short message. In all cases, the access procedure is the same (see figure 6.19) In broad terms, this procedure starts with an RIL3-RR CHANNEL REQUEST message sent on the RACH; the answer from the network is conveyed in an RIL3-RR IMMEDIATE ASSIGNMENT ( o r RIL3-RR IMMEDIATE ASSIGNMENT EXTENDED) message sent on the Paging and Access Giant Channel (PAGCH), conveying the description of the channel allocated to the mobile station; finally, the mobile station establishes the link layer for the transfer of signalling on the newly allocated channel, and sends a first signalling message on this channel (the "initial message"), conveying the subscriber's identity and the reason why it requests a connection. This basic canvas appears very simple a t first view; b u t each o f the corresponding steps reveals some complexity when studied more closely. 6.3.1.1. Random Access The channel request message is a curious animal indeed, deserves some attention. The network has no method of knowing w Of course, mobile stations do not transmit at any time, but follow the slotting of time imposed by the TDMA scheme (a RACH/F uses only one slot every 8 burst periods). Collisions may therefore he studiedrlot by slot. When two mobile stations transmit during the same slot, two things may happen: either one of the bursts is received by the BTS at a level significantly higher than the other one, allowing its correct decoding (this is called a "capture"), or none is received correctly. Collisions are therefore a source of message loss, which increase with traffic. In order to provide a satisfying rate of success for access attempts, repetitions oust be used. The repetition scheme cannot be too simple. otherwise its effect on the throughput in a high load situation can be disastrous. and n indeed lead to a complete deadlock situation. This kind of problem has been thoroughly studied in the field of random access techniques, an area of interest for many networks using shared resoutwes (Local Area Networks, Packet Radio....). GSM offers an example of one of the bestknown(and simplest) random access schemes. with the R A C b r i n g an I Ia "A -d s e fh n to lic p rl. When a request has not been answered. the mobile station will repeat it. I f two mobile stations whose attempts have collided would choose to repeat them some given constant time afterwards. their requests would collide again. Repetitions on the RACH are therefore performed after a "random" interval to avoid this phenomenon. As in all Aloha protocols, this re-transmission strategy is not enough to escape from collapse when the traffic goes over a given threshold. In practice, the offered load (in terms of number of requests. which can lead to one or more messages sent) on the RACH should not exceed about a quarter of the total sending opportunities (number of slots on the diann,•11. In older to control this load. (.SNI uses three t.. Ill I....Crent moans. corresponding to- three different kinds of overload. The collapsing threshold depends on the number of repetitious and the average thfie between them: one way to make the RACH robust to a higher load is to spread these repetitions further apart. and/or io reduce the number of repetitions. O f course. this may he detrimental it) the quality of service. respectively in terms of delay or in terms of success probability. Such methods cannot be pushed very tar and are not sufficient to control very high loads. It is nevertheless useful in cases of small and temporary overload. In GSM. both the number of repetitions 370 GSNI SYS'I parameter resulting value TX-WM4Y random scheduling of each attempt over 3 to it) slots MAX RETRANS up to I . 2. 4 or 7 repetitions a n i m a Table 6.3 - RACH repetitions control parameters Both the time interval between re transmissions of a random request and the maximum number of such repetitions are controlled by parameters. and t h e intervals between them are controlled through parameters broadcast regularly on the BCCH on a cell-per-cell basis. Since i t is important to have a short delay between the moment when the BSC decides to change these parameters and the moment when the mobile stations act on them, i t has been decided to send them in all BCCH messages, i.e., 4 times per second. The scheme is controlled by two broadcast parameters, the average time between repetitions (TX-INTEGER), and the maximum number o f allowed repetitions (MAX RETRANS, see table 6.3). These parameters should be controlled through a feedback loop taking into account the observed throughput. I t should be noted that Aloha control is in fact not absolutely necessary: the values can be set to a constant choice representing some compromise between throughput and delay without jeopardising the system. As already mentioned, this mode of control can only cope with brief (in the order of one second) traffic peaks or sustained marginal overload. Before studying further the other means to control the load on the RACH, it is worth noticing that this channel constitutes only the first link in a chain of resources, and is not necessarily (he bottleneck of the system in congested situations. There are indeed other candidates for such a role: the PAGCH, which offers a limited capacity for carrying both initial assignment and paging messages, is one of them. Channel allocation is another, since the pool of available channels is also limited in each cell. An efficient overload control takes all these factors into account and tries _ to cut the traffic at the source (i.e.. on the RACH) in cases of congestion. • This means that overload control mechanisms must not necessarily try to maximise the RACH throughput, but must limit it to the maximum traffic the whole chain can swallow. Having said that, the second way of "controlling" the load on the RACH consists of rejecting the requests with a message forbidding the mobile station to access the channel for some specified length o f time. This mechanism prevents any further repetitions by the mobile station, either through its automatic repetition scheme (controlled as explained I< m o w v.a ' he ut burst. Three 01 them indicate the reasilli for ATVS`, alai lilt' S i : I l e : I S a rand11111 d i s c r i m i n a t o r . This capacity is obviously insufficient to carry all the information the mobile station would want to transmit, such as the subscriber's identity, the reason for requesting a channel. the characteristics o f the mobile equipment.... A l l o f this information is in fact included in the "initial message" which will be the first information transmitted on the dedicated channel. once allocated. But the most critical usage o f a discrimination between random Gk. access attempts is not to irovitie information to the network. : \ given Mobile station must he able to correlate an initial .issiiinmeni troin the networkir"(kthiswith its own request. \+ills as "• it N i)L'If C a n I 1 1 1(11:-111;Cy he t h e mobile station. reducing drastically the probabilit‘ that t w o mobile stallions send identical messages during the same .lot. which may in case of capture lead w an ambiguit) as to which of the two requests is being granted. Three hits remain. which are used as shown i n figure 6.20 to provide a minimum indication of the reason for accessing the net ork. This first rough indication mac be useful for discriminating rejections in case o f congestion. and also to choose the best type o f channel to allocate. 374 T H E GSM SYSTEM R.\ I)I( I RE S' n Itt I \ I : \\,\l.1 \1.1\ \ I 375 6.3.1.3. T h e Initial Channel Assignment The a c t i v a t i o n process requires t h e B T S t o prepare f o r the access o f the m o b i l e station o n the n e w l y allocated channel. T h e t i m i n g advance A f t e r the B T S has c o r r e c t l y decoded a channel request. i t indicates it t o t h e B S C t h r o u g h a n RSM CHANNEL REQUIRED message, w i t h o n e important piece o f additional i n f o r m a t i o n : an estimate o f the transmission delay (this i n d i c a t i o n is critical t o initialise the t i m i n g advance control). A field o f u n s p e c i f i e d c o n t e n t ( P H Y S I C A L INFORMATION) a l l o w s t h e manufacturer to add m o r e i n f o r m a t i o n , such as•the reception level. is i n i t i a l i s e d b a s e d o n t h e t r a n s m i s s i o n d e l a y estimate w h i c h t h e B S C indicates ( b a c k ! ) t o the B T S . E v e n t h o u g h this estimate has been i n i t i a l l y calculated b y the B T S . t h i s g o i n g h a c k a n d f o r t h between TUTS and BSC' In n o r m a l l o a d s i t u a t i o n s , t h e B S C t h e n c h o o s e s a f r e e c h a n n e l ( TA C H / 8 o r TA C H / F ) , activates i t i n t h e B T S , a n d , w h e n t h e B T S has acknowledged t h i s a c t i v a t i o n , b u i l d s a n i n i t i a l assignment message t o be sent on the PA G C H . B AI T BSC The i n i t i a l assignment i n d i c a t i o n sent t o the m o b i l e station o n t h e PA G C H c o n t a i n s t h e d e s c r i p t i o n o f t h e a l l o c a t e d c h a n n e l , t h e i n i t i a l l i m i n g advance to he applied. the i n i t i a l m a \ nimin transmission p o w e r. as well as a reference allowing all the mobile stations expecting such a message to know whether they are being addressed or not. This last point is worth some more explanation. Addressing is clone by including in the initial assignment indication the exact contents o f the R 10-RR CHANNEL REQUEST message v\ hick is being answered, plus the t i m e reference o f the slot i n \ \ hick i t w a s received (such a time reference exists thanks t o TDNIA I. This allows mobile stations t o check whether they are actually concerned b y each initial assignment. h> comparing these values " i t h t h e ones the). have fore(' when sending the I31 1.3-13R LIIANNI11. Rli(lt lSF message. as shown m figure 6.21. CHANNEL REQUEST CHANNEL REQUIRED )1, (frame number) frame number I delay 4 channel activation IMMEDIATE ASSIGNMENT 1< is necessary. s i n c e t h e B T S h a s n o m e a n s t o c o r r e l a t e t h e messages received o n the R A C ' H w i t h corresponding channel assignments. (or IMMEDIATE ASSIGNMENT EXTENDED) frame number =? , Figure 6.21— Initial assignment procedure After the activation handshake on the Abis interface, the BSC prepares an initial assignment indication containing the 8 bit discriminator as received in the correctly decoded RIL3-RRCHANNELREQUEST message, as well as the frame number in which it was received. This enables the mobile station to check whether it is concerned Besides. answers to Rii.3-Rit c i t A y s i i . Rticn ESI messages call be sent in any block o f the PAGCH. even on the paging sub-C11:1111lik. A s a consequence. once a mobile station has made an access attempt. it should monitor the whole PAGCH ( o f the same TimesIM Ntutlher as lite R A C I it used f o r access) f o r an ans‘‘er f r o m the network. Furthermore. the BCCH messages must be decoded continuously during this period. i n order for the mobile station to set the R A O I control parameter \attics in real-time. This phase is very constraining 16r mobile stations in terms of reception ( 4 0 bursts every 51 x N burst periods), almost comparable Itl T:\CH/F reception. Let us examine some side issues. I t mars p e n that :he .eactIL in in the intra‘tructure to an r it 3 los tit \ ssi . . s L r. 4 . ;-, : . • j k l i t i o n limn the mobile station. In stkii inetticient situations .1 s l a t tot; lira' be allocated a channel R' ice tor e'en more tiniest. since the mil astrudidre ha. w. means in knowing whether alt nit.3-nis ('II \ \ \ u RI Qt I s I is the repetition of a picsiim...ric or not. The mobile station will use the channel allocated in Ilk, lost initial assir.n11k inessaic it decodes. and the other ones wilt ha‘e been blocked for a few second-. III %din. Nevertheless. the Spectlicinium require that the mobile station he able to accept the network answer to any of its last three RIL3-HR ('II \\\EI RI.Q1 Ls] messages. in order to get service from such not-so-efficient BSS equipment. In congested situations. when no channel is free for allocation. the BSC nias choose not to answer to an Rit.3-nn unAssitt. REQVI:SI• message. or to send back a rejection indication. The first choice is not very efficient. since the mobile station will 376 TiniCiSX1 SYSTEN1 rs um, RI l o I \ N m.1 511 '.1 REJECTmessage, containing a lime indication during which the mobile station is forbidden to make any inure attempts on the RACI I (the it Arr ram- \)Io\ parameter). If the overload situation does not concern the RACI I. the value of the it 10 nn/c1IION can be null, otherwise it is a useful mechanism to help reduce RAO I load !see page 370). 377 MS 2 The PAGCH is an important potentiall bottleneck of the is) stein. In order to improve its efficiency, both initial assignment indications and channel request rejection indications can be grouped together to form messages. There are two assignments in an RIL3-RR IMMEDIATE. ASSIGNMENT EXTENDED message. and up to four rejections in an RIL3- RRIMMEDIATEASSIGNMENTREJEcr message. Though this point is unclear in the Specifications, the original intention was for the BTS to perform the grouping on (he basis of the individual indications provided by the BSC. The BSC has the possibility to provide the BTS with immediate assignment indications which are not ready-to-send messages (using the RSMIMMEDIATEASSIGNCOMMANDmessage). The BTS must then build the corresponding RIL3-RR messagesThe fact is that the BSC can also perform the grouping and build the messages, and can provide ready-to-sendmessages to the BTS (this is, by the way, the only possibility for assignment rejection messages). However. nothing really precludes the BTS to un-build these messages and to build others grouping the requests differently. The samedebate exists for the paging indications, 11 •648/4,---.,„ -CZ 12 S A e m In no case does the BSC schedule the transmission of the assignment indications: this is much easier for the BTS. The counterpart is that, despite grouping. congestion may happen. This is resolved simply by the BTS, which dropsmessages which are in excess of the achievable throughput. While this resolves the BTS congestion problem. for the BSC it results in TACHs allocated but not "assigned". To avoid a worsening of the congestion situation through this effect, the BTS can indicate the non-sending of amessage with an RSMDELETEINDICATIONmessage. A second effect of the message is to indicate the overload to the BSC. K -t-> leaves stays 6.3.1.4. T h e Initial Message _ Once it has received an initial assignment indication, the addressed mobile station modifies its reception-and transmission configuration to adapt it to the frequency and time characteristics of the new channel. In phase 1, this new channel can be a TACH/8 or a TACH/F. always in "signalling only" mode (see page 321). The transmission level is set to a value broadcast on the BCCH (or to the maximum transmission level of the mobile station, whichever the smaller), and the transmission starts with the timing advance value specified by the IISC'. The first thing the mobile station does on the new channel is to transmit a link layer SABM frame for SAPI 0, i.e.. the frame used to establish in acknowledged mode the link layer connection for signalling messages. In standard HDLC protocols, an SABM frame does not carry any information other than the one necessary for the link layer level. In GSM, the SABM frame sent within the initial access procedure contains a signalling message, the "initial message". The reasons for departing from standard usage are twofold. The first reason is efficiency. though this was Figure h.22 - Contention resolution at liiik establishment In rare cases. more than one mobile station ma) I n d useli On the .tone c h a n n e l . The transfer or a non-tunhiguous initial message in the s X I -I : \ exchange allows each mobile station to knov. ‘‘Nether the channel is for its own usc or not. • not the leading ruse. The other reason comes from the fact that the reference used in the initial assignment to address the mobile station is ' — not fully unambiguous. It i n d e e d happen (though this is a rare case. around I C s i n h i g h l o a d situations( t h a t t w o mobile stations simultaneously send kit.3-RR c1lANNE1. REQUFSI- messages with exactly the same contents. and that one o f them i s correctly received and answered b y t h e BSS. T h e ensuing channel assignment w i l l h e understood by both mobile stations as their own. and both mobile stations will access the "dedicated" channel. Therefore. until the point where mobile stations identify themselves in a non-ambiguous wax'. there is not a 100% guarantee that a single mobile station will access the channel. 378 THE GSM SYSTEM As a consequence of this situation, potential collisions must be detected as soon as possible on the new channel. and the SABM-UA including ("piggybacking") exchange provides this facility by unequivocal information on these link layer frames. Mobile stations check the piggybacked contents of the UA frame. If one mobile station receives a UA containing something different from the contents of the SABM it sent, it must leave the channel and start the access procedure all over again, thereby enabling the "right" mobile station to stay There is usualls serve the purpose of collision detection. However, the efficiency criterion was also taken into account and the SABM includes more than just this identity, and includes a full "initial message". The "initial message" comes in four different brands. depending on table All these 6.5). messages contain an identity of the mobile station: the classmark, a field indicating some key characteristics of the mobile equipment. including the maximum transmission power: and RIders sdw a the reitsons for lending hte original kit.-kk catast kEatest macoige. or shouldit eht het RIs-KI PAGNe RESPOsSE message? I would som fair to choose the ent donk r a c e . k e s t o s s message fi for instance the request from the mobile station concerned pened location updating. but no priorits scheme is delined ni left open for mobile station manufacturers. the Sea,t and the whole matter is Once na "initial message" has been received by the BTS and sent One obvious way to obtain an unambiguous SABM content is to use an identity unique to the mobile station. This would be enough to the access was triggered (see NaRRomise: What should the initial message eh i n such acane? Shopld it he sonsege undisturbed on its own channel (see figure 6.22). the reason why bat one c h e a t may arse. when. after having som a Kil station receites a paging indication and then the answer t i l complementary information making the reason for access more precise when need be. All but the first of those messages belong to the RIL3-MM protocol. and will be explained in Chapter .7 The first belongs to the RIL3-RR protocol, but could have as well been put in RIL3-MM. If there had been enough room, these messages would have been properly formatted. with a part for the RIL3-RR protocol (including the classmark). and the rest in an RIL3-MM part. back without a n modification inside the CA frames. it is passed to the BSC n i an RSM ESTABLISH INDICATION message. A t this point. hte mobile station classmark is stored for further use reg.. to choose the power control loop parameters). and the BSC then sets up an SCCP connection This si done through an seep towards the MSC asce Chapter 5. CONNECTION REQUEST message. on which the initial message man optionally be piggsbacked. O n then does the N S C become aware of the with contact the mobile station. The initial message. piggybacked or sent after the S C ( P connection establishment. is carried N i a BSSSLAP COMPLETE LAVER 3 ISFORMATION message. independent from its protocol aRR or A N . It contains enough informatio n for the MSC to trigger required actions n i the upper livers aNd. (C. ... bu this c o m e s o u t o ft h e scope of the a c c e s s procedure. "ishen the secess procedure ands. the RR-session is fully established with a complete signaling path between the mobile station and the MSC. With the establishment of the S C ( P connection. the AIS(' takes control of the decisions concerning the transmission characteristics ol the R E - s i s s i o n . a n d t o BSs ir al o s h i l l m o n i t o r i n o the rrads transmission and performing handover decisions initi al m e s s a g e R e a s o n for access K U . Response to a paging 3.RK MAGISGRESPOSSI RIL3VSILOGSTIOS TPOSTISGREOCEST Normal location updating. periodic location updating. "IMSI attach" 6.3.1.5. The Mobile Station C l a s s m a r k Atobile stations differ maximum transmission p o s e r important for characteristics me inflasrur when the sl mood to mar the he sation is eneaged Because the equipment of the user mas be changed withou IMSI detach operator ( s u b s c r i p t i o n is linked ot RIL3-AIST CSSERVICE REOLEST All o t h e r cases (call set-up, short message transmission. s u p p l e m e n t a r y service management. mans and the s e r i c e s ).. Table 6.5 - Possible initial messages Four different types of signalling messages may be used as the "initial messag the Stal. nor to the mobile this indication must be given ni the beginning of This is the purpose o f the m o b i l e station classmark. the c l a s s m a k a r e s h o w n in table 6.6. There exists in fast. for e f fi c i e n t reasons. "mobile station classmark ape I" ni the spendisationg the to as e "mobile station c l a s s m a r t a m rol classmark m ohe retoried 380 TILE CISM SYSTEM parameter contained in the classmark ('lass revision level I — RF power capability encryption algorithm frequency capability short message capability ;tiI " 1 2 1 , S i 11 R CI \ ( I f \ I I . \ I GSM900 I it S 1 Son Ni W I \\ SW ;w rt 4 2W 11.5 W Table 6.6 — Full contents of the mobile station classmark Table 6,7 - Mobile station pinker cla%.e. The classmark identifies those characteristics of the mobile equipment which are needed by the infrastructure during a connection. Five classes are defined For (iSN1900. and two for DC'S I s(x). corresponding to the maximum transmission pinker (il' the mohile station, used when the message is piggy-backed). The shortened classmark is also used in the RIL3-MM imst DETACH message, for no obvious reasons. The revision level i s used f o r upward compatibility handling between successive phases o f t he Specifications. M o b i l e stations developed and type approved according to the phase 1 Specifications should set this value to 000. In the future, other values will be allocated, so that the infrastructure will know which level o f upgrade is used by each mobile station. The RF power capability, often referred to as the transmission power class, o r even as the class, refers to the maximum power the mobile station is able to transmit. This information is used for power control and handover preparation. Power classes are not defined in the same way for GSM900 and DCS1800, as shown in table 6.7. Mobile stations of class I for GSM900 may possibly never he developed; indeed the typical classes in GSM900 are class 2 for portable or vehicle-mounted equipment, and class 4 for handheld. Class 5 handheld stations will most certainly experience strong limitations in coverage and will be destined .fartilban areas. The typical class of DCS1800 mobiles is class 1. The encryption algorithm indicates which ciphering algorithm (if any) is implemented in the mobile station. In phase I , there is but one choice: all mobile stations must implement the A5 algorithm specified by the GSM MoU. This field will enable the BSC to cope with future mobile stations with other ciphering capabilities, b y being able t o choose whether to cipher or not, and if yes with which algorithm. The frequency capability is also present for future use. It enables the network to cope with mobile stations having different capabilities in terms of frequency bands. A phase I GSM900 mobile station must be able to cope with frequencies anywhere in the 2 x 25 MHz hand allocated from the start to GSM. A n extension o f this hand to. say. 35 MHz is under study, but all mobile stations will not h e able to cope with the extension. Therefore. the classmark will enable the 13SC' to distinguish the different populations of mobile stations and allocate channels lo each mobile station according to its own frequency capability. Last o f all, the short message capability indicates whether the mobile equipment is able to deal with short messages. Such an indication is not strictly necessary i n the classmark. since a mobile station not equipped to deal with short messages ( i f any!) can always infomi the network by rejecting a LAPDm SABM frame on SAPI 3. But it is more efficient to know the capability of the mobile station from the start. since it enables the network to eschew the transmission o f short messages between the short message service centre and the VNISC. We have seen that the classmark is sent by the mobile station in the initial message. at the beginning of the RR-session. The network does not store the classmark between RR-sessions. since the user eau change his,. equipment. Now. it may happen that the classmark changes during the ._,.....12.1-session. This is not a common event. but an example is a mobile e iipment composed o f a handheld part and a vehicle-mounted part inetuding a n R F transmitter. w i t h t he possibility t o connect and disconnect the two parts during a communication. Then the power class changes. and the new value HUN he provided to the network. To achieve this. a procedure is included in the RR plane. The indijoion from the mobile station is carried in an kii.3-kit et..xssmAki: t i I.N.Ar• message and the BSC can forward the indication to the relay MSC with a liSSMAP CLASSMARK UPDATE message. I f the relay MSC is not the anchor MSC, the chain stops there in phase I . since the MAP/F. protocol does not support this functionality. THE GSM SYSTEM 6.3.2. PAGING PROCEDURES A little has to be said on the paging procedures. When a call to a subscriber reaches a MSC through which the subscriber is deemed reachable, the MSC determines the location area where the mobile station is registered and sends a BSSMAP PAGING message t o all the BSCs controlling cells i n this location area. T h e message contains the subscriber identity to page with (it could be a temporary mobile station identity, called the TMSI o r the full international mobile subscriber identity, the IMSI), the IMSI to determine the paging sub-channel (to cope with discontinuous..reception), and the list o f cells in which the paging must be issued. The BSC in turn sends an RSM PAGING COMMAND to the BTS device in charge of the PAGCH of suitable TN (determined by the BSC, from the IMSI and the common channel configuration), for each cell in the list. This message contains the number of the paging sub-channel, which is computed by the BSC, as well as the TN of the PAGCH. The BTS in turn possibly packs some paging requests together and sends the resulting messages on the correct paging sub-channel. A s indicated in the functional section, more sophisticated approaches are possible. Pagingmessages come in three brands (called1211.3-RRPACINGREQUESTTYPE TYPE2 andTYPE31. adapted to the size of the identity used for paging. Type I can carry two identities of whichever sort, whereas type 3 can carry four TMSIs. and type 2 two TMSIs and one identity of whichever sort. A topic of interest is the repetition policy. In other areas. GSM is specified s o as t o provide a correct quality o f service when the transmission quality is far from perfect. To be consistent. the paging indication should not be sent only once in each cell. A typical value (not specified) would be to send it three times. No repetition mechanism is described i n the Specifications. and this leaves some obscurity as to which machine takes care o f the repetition: the N1SC, the BSC or the BTS'? To choose the BTS has some advantages: it can optimise the use of the PAGCH, and in particular it may repeat paging requests more than the minimum required if the channel has some room left. On the other hand, if the MSC is in charge of managing the repetition process. it will wait some time before requesting repetitions, thus avoiding useless repetitions in many cells in cases o f successful mobile station access. Neither the BTS nor the BSC are capable of such monitoring, since they are not able to relate the mobile station answer to the paging. This seems obvious when the mobile station has accessed another cell (the MSC does not indicate it back to the BSCs from which it requested paging. it would be too big a procedure). but the same is also true when the mobile station It ) l a s ( )1 R ( I \ 1 1 \ N , \ I 1 g 3 accesses through a cell under their control. :\ sensible scheme seems to pnivide two levels of repetitions, one in the MSC. with a long period, to cope For instance with short reception interruption (such as a change of cell, or the mobile station passing through a tunnel): and a short term repetition i n t h e B T S , when load allows. coping w i t h mediocre propagation conditions. The Page Mode The PAGCH configuration may change in time to adapt to the traffic distribution. Although this configuration is managed by the 13SC. it is usually the OSS which decides on such a change. either automatically: ---trm-the basis o f traffic observations o r Through the command o f an operator. The configuration of the PAGC1.4 in a given cell must he known by all mobile stations camped on this 411. and is therefore part o f the broadcast information. When the configtf)ations changes. the BSC must co-ordinate the change of the broadcast information and the scheduling of paging messages. It is not possible to control yery precisely the time at which each 'nubile station will decode the neyy broadcast parameter. and hence switch to the new configuration. The corresponding transitional period May last up to a few seconds..In order to avoid the loss of paging messages during t h i s "fuzzy'' period. a special feature has been incorporated i n the Specificathms: the page mode. The page inode indicates to mobile stations in exactly which part o f the PAUCII their own paging messages may he sent. The three values of the page mode are the following: the -normal- page mode corresponds t o the ha sic scheme. Paging messages are sent only on the sub-channel as defined by the PAGCH configuration and the !NISI: • t h e -full- page mode has been designed to cope \\hit a dynamic change i n the PAGCII configuration. When this 'node i s indict/4rd to mobile stations of a sub-group. then it means that paging messages for the subscriber of this sub-group nia\ be sent anywhere on the PAOLI I of the saute ti meAhir: .• T h e "next-but-one- page mode has been introduced f o r sophisticated scheduling algorithms. It enables the I3SS to send additional paging messages for subscribers in a giN en sub-group in another paging sub-channel. This feature ma> he useful in a situation of temporary overload on some of the sub-channels. or to free a block to send an initial assignment message. Based on the tact that the definition o f paging sub-channels includes an implicit numbering o f these sub-channels, the -ite t -but-onemode indicates that nag lm] 384 T i look) l a y )1 Hi GSM SYSTliN1 102 x 8 BP cycle =EL/TN0 > paging I M S I subgroup m o d . 8 . 02 2 03 3 20 4 21 5 22 6 23 7 normal mode . ; 8 5 Transmission modes, a s f a r a s r a d i o resource management i s concerned, involve 7 different types i n phase I a n d are expected t o expand to 12 in phase 2 (see table 6.1. page 321 k The mode is one of the properties of the transmission chain. Other variable characteristics of Ibis chain include the cipher mode (whether transmission i s ciphered o r in clear text). as well as both uplink and discontinuous transmission modes. trEn1L-7-1t0 1 ; I A N N 6.3.3. PROCEDURES FOR TRANSMISSION MODE AM) CIPHER MODE MANAGENIEM TN 2 00 01 \ 0 I "full" mode "next-but-one" mode + next I I At initial assignment. the transmission mode is chosen In the BSC. It consists necessarily of one of the "signalling only- modes. in clear text. The channel m a y be a TA C H / 8 o r a TA C H / F. Aftemards during the lifetime o f the RR-session. the choice o f the transmission mode depends on the communication needs and is done by the \ISC. which can request a change o f the transmission mode at any time while a connection i s established. I t does so through an "assignment" procedure \ \ hick does not necessarily result in a radio channel assignment I The basic procedure consists o f a USSMAP ASSIONNIENT REQUEST message. describing the transmission characteristics a s d e s i r e d b y t h e N I S ( * . a n d t h e corresponding liSSMAP ASSIGNNIENT COMPI.FIF ackno lodgement. a s shown i n figure 6.25. A negative answer i s also possible i n case o f problems. and i s carried i n a 13stixtA A s s i ( i N x i i N f m e s s a g e . Figure 6.23 - Different page modes In addition to the "normal" paging mode, the "lull" page mode and the "next-but-one" page mode are useful in transition situations. or to compensate for partial congestion situations. normally paged in sub-channel n (the sub-channel where the page mode indication was sent) will in addition he found in the next block o f the paging sub-channel n+2, modulo the total number o f paging sub-channels on the current tinseslot. The example of figure 6.23 will help to understand this feature. It is worth noting that the mobile station is free to listen to more than the required minimum. This may indeed represent a simplification for vehicle-mounted mobile stations, for which battery consumption is not at stake. In cases where the load on the PAGCH is very low, it might even be interesting for the BSS to send the paging messages in more blocks than those of the required paging sub-channel. This will allow a speed-up of their reception for mobile stations which listen to more than just their own paging sub-channel. BSC MSC VLR ASSIGNMENT REQUEST change of transmission mode ASSIGNMENT COMPLETE Figure 6.24 -- Change or the tyammis,don mode I)) the NISL The MSC may change the transink,iim mode or an R12-conneciion at an Bole. by running the -assignment- procedure um ands the BSC. which then takes charce c i i n n o l l i t r a n s i m . . i . i i i oho, ith• 386 G S M SYSTEM NH I itl.Sc )1 R(l \I AX mil %n \ 1 3 3 7 A n o t h e r special case i s w h e n the a l l o c a t i o n cannot proceed i m m e d i a t e l y, but the request is p u t i n a queue; t h e M S C can be w a r n e d o f the situation by a BSSMAP QUEUING INDICATION message ( t h e c o m p l e t i o n o r f a i l u r e TRAU B T i n d i c a t i o n i s e v e n t u a l l y sent later). T h e request also includes the i d e n t i t y o f the n e w l y allocated terrestrial c i r c u i t between t h e B S C a n d t h e M S C , when required, i.e., w h e n the m o d e i s changed f r o m " s i g n a l l i n g o n l y " to BSC another one. r< The a c t i o n o f t h e B S C w h e n r e c e i v i n g a BSSMAP ASSIGNMENT MODE MODIFY REQUEST ml >m2 REQUEST m e s s a g e d e p e n d s o n t h e c o m p a r i s o n b e t w e e n t h e e x i s t i n g transmission m o d e and the required one: n / i n i a ji2 • i f b o t h m o d e s are t h e s a m e , t h e B S C j u s t s e n d s b a c k t h e MODE MODIFY ACKNOWLEDGE BSSMAP ASSIGNMENT COMPLETE message b a c k t o t h e M S C w i t h o u t any other action; request; • i f the new mode requires a channel of a type different from the one i n use, t he B S C performs a subsequent assignment procedure, i.e., it transfers the connection to a channel of the required type before acknowledging the MSC request. I in parallel • i f both modes d i ff e r b y t h e t y p e o f information t o b e transmitted, but use the same type o f channel, the B S C performs a " m o d e m o d i f y " procedure before acknowledging the M S C BTSTRAY configuration CHANNEL MODE MODIFY mi> m2 CHANNStMODE MODIFY ACKNOWLEDGE MS configuration ml > m2 change•of ccnfig,Jraton from mode 1 to mode 2 / i n -band control The t w o last cases w i l l n o w be studied i n more detail. Figure 6.25 -.'1 he "mode modil) procedure 6.3.3.1. T h e Mode Modify Procedure As for any procedure affecting the transmission mode. the mode modify procedure includes t w o parts: t h e configuration o f the transmission devices on the infrastructure side (BTS, TRAU and BSC). and the configuration of the mobile station. No means are provided to — — s y n e t t t onisi these t w o parts i n a precise w a y. resulting u s u a l l y i n a short period of time during which the whole configuration is inconsistent. The Specifications do not specify in which order the two configuration steps should be managed. and this choice may have an impact on the period of inconsistency. In the following example. it has been hypothesised that the two tasks are run in parallel. The BSC triggers the reconfiguration of the BTS and the TRAU by sending an RSM MODE MODIFY REQUEST to the BTS. Following reception of this message, the BTS modifies its coding and decoding algorithms and changes the in-band mode information in the BTS-TRAU frames. The TRAU, i n turn, reacts by modifying its data processing (speech The BSC is in charge of configuring both the WI 'S and the awhile ,fatiou. hut the ogler in %%hick filen. stens .hould he poriormed i. nor .pecn [e d. The I N A I i s configured through in-hand inio:inaiion lion] the 131 S. coding o r data rate 1 1 the nekk tootle i s speech. then sYnchronisation bet \\cel t anti I l l s is needed. When the chain r — ' - - ready (the Spec" 11 nil lOn.\ d o not specify \ \ haler this includes the synchronisation with the I R A ( ' or not). the BTS ans‘‘ers the BSC by sending back an Its \I NI( Mtn NI(M)11:1. NCI< NOWLEDGE message. In parallel. the BSC triggers the reconfiguration o f the mobile station b y s e n d i n g a n RI1.3-RR L U A N N E ! . W O E I l o D I I Y m e s s a g e containing the new mode Io he applied (see also figure 6.25). When the mobile station receives t h e order. i t modifies i t s channel/source coding/decoding according to the nok requirements and mist\ ers %kith an - S/93 T H E GSM SYSTEM 389 12.\ R I : S O L I : U T NIANXIIIAlliN I RIL3-RR CHANNEL MODE MODIFY ACKNOWLEDGE message t o t h e B S C via the BTS. It is worth noting that the connection of higher level devices (microphone and loudspeaker for speech, or terminal adapter for data) is not controlled by this procedure, but by a call control procedure (see Chapter 8). A third action needs t o b e performed b y the BSC: circuit switching. I t may happen that nothing needs to be done, for instance when a terrestrial circuit is already established and satisfies the needs. But i f the BSC-MSC terrestrial circuit needs to be reset or altered, the BSC must connect it to the correct BTS-BSC circuit (in correspondence with the TACH/F). The most complex case arises when the TRAU is remote, and must be changed to another one, adapted to the new mode. This is in fact only possible when BSC-controlled switching facilities exist between both the BTS and TRAU and between the TRAU and BSC. In this case, the BSC must switch the connections before ordering the reconfiguration o f the BTS. In the other cases, BSC switching can be done in parallel with the other actions. TRAU 7S. BSC CHANNEL ACTIVATION / ch. CHANNEL ACTIVATION ACKNOWLEDGE ASSIGNMENT COMMAND - old cannel ERRORINDICATION > , new L hE 6.3.3.2. T h e Subsequent Assignment Procedure When a change of the radio channel is required in addition to the above-described "mode modify" procedure, the procedure is somewhat more complex, because the change o f channel implies a break i n the signalling carrying capability between mobile station and infrastructure. Besides, let us recall that the BTS devices in charge of transmission on a given TACH are independent and do not communicate. Therefore, the whole control o f the operation i s centralised i n the BSC, and the operation itself is quite similar to a handover. A transfer of channel first starts with the setting up of the new path in the infrastructure. This includes the allocation of a new radio channel (with all the priority and queuing management aspects described on page 357, to cope with congestion), the activation of the corresponding BTS device, and possibly the allocation o f a T R A U and the switching necessary to connect all these terrestrial segments. The activation of the BTS is ordered by the BSC through a simple request/acknowledgement procedure, as shown in figure 6.26. The RSM CHANNEL ACTIVATION message contains all the information specifying the transmission mode, including the basic transmission mode (among those listed in table 6.1), the cipher mode and the downlink discontinuous transmission mode. An uplink discontinuous transmission mode is also channel ASSIGNMENT COMPLETE conlipration of the i c j equipment for / i n - b a n d control the n e w c h a n n e l Figure 6.26 • ActiN alum of a new channel iu die After the activation handshake on the Ahis interface. the BSC orders the mohile aaliun . t o change channel by an kii.3-104 AsskiNxtkvi t t A i m xi) inc..age. which is in general not acknowledged11% the mobile sou ion on the old eh:newt 4 T h e completion of the procedure is done on the new channel after establishment of the full signalling link. sent, though one may %yonder what it may he used for. and moreover what its meaning is. since in some cases the network leaves the choice to the mobile station. In addition, it contains the information needed by the mobile station for access (see the initialisation o f the timing advance, page 346) and the first power control settings. The 13TS. upon reception of this message. starts the i n -band information exchanges with the TRAU, t o set the basic transmission mode and the discontinuous transmission modes; this is the point where synchronisation with the TRAU starts. Once the BTS and the TRAU are activated (more exactly when the BSC h a s r e c e i v e d t h e R S M C H A N N E L A C H VAT I O N ACKNOWLFIRMI 390 I E GSM SYSTEM message), the BSC orders the mobile station to perform the transfer of channel, through an RIL3-RR ASSIGNMENT COMMAND message. The previous path, including the signalling connection, is not released by the infrastructure at this moment. This allows the mobile station to go back to the previous channel should access fail on the new channel f o r any reason. In the case o f a timed assignment (see page 353), the mobile station stays on the old channel until the instant of change indicated by the infrastructure. Otherwise, the mobile station performs the transfer immediately after reception o f the RIL3-RR ASSIGNMENT COMMAND message, not even acknowledging the corresponding frame at layer 2. This lack of acknowledgement results in a repetition of the message by the BTS, until it decides that a link layer failure has happened, of which the BSC is advised. The BSC does not act on this indication, since it knows that an assignment is in progress. In case o f return to the old channel, the mobile station starts re-transmitting on this channel by performing a link establishment, which resets all contexts irrespective of what happened in the link layer process in the BTS. The transmission mode used after return on the old channel is the one that was used on it, not the one asked for in the assignment message. Whether in the case of a successful procedure or of return to the old channel, the interruption of the link layer may result in leaving a message sent by the mobile station in a non-acknowledged state. This situation is handled by the upper layers, as will be explained below. Once the mobile station has changed its settings t o the ones corresponding to the new channel, it starts transmission and reception according to the transmission mode indicated in the RIL3-RR ASSIGNMENT COMMAND message. I t must also establish a new signalling link in acknowledged mode. The Specifications do not say whether this link layer establishment should be done first (before the transmission of any user data) or not. Anyway, once this link is established, the mobile station sends an RIL3-RR ASSIGNMENT COMPLETE message to the BSC before _ _ s l y . other message. Then, all the messages waiting for transmission can be sent: first the ones already sent on the previous channel but still not acknowledged, then the ones which have arisen during the procedure. The same applies in case of return to the old channel. except that the first message is an RIL3-RR ASSIGNMENT FAILURE. A message sent by the mobile station before link interruption and not acknowledged then cannot be lost, but may be duplicated. R1L3-RR messages are considered immune t o duplication, and therefore no mechanism has been introduced to cope with such duplications. This is worth remembering when new procedures are added in the future! The RADIO RitSoLIWIc NIAN.Vir 3 9 I way in which the BSC is implemented has also sonic impact. A "careful" BSC will postpone the sending of an itit.3-RR AssRrsixtiAt commAND (or an RIL3-RR HANDOVER COMMAND) w h i l e i t i s w a i t i n g f o r the a n s w e r t o another RR command (e.g., for the start of ciphering!). In t h e case o f upper layers. e.g.. call control o r mobility management. message duplication could in some cases he harmful. In order to avoid problems, correction i s obtained b y detection and suppression of the duplicated message. Detection is obtained through the use of a simple numbering scheme, the I-bit sequence number referred to as N(SD) in the Specifications. This number is included in each upper --layer-message. and is changed for each new message. Since messages are sent and acknowledged one at a time (window size I ). this scheme is enough for the infrastructure to detect duplication. When it receives two successive messages with the same N(SD). it discards the second one. This task is performed at the RR level, but not in the BSC. It is one of the tasks o f the anchor MSC. which i s the destination o f upper layer messages sent b y the mobile station. The reason why i t cannot he performed in the BSC is that it must he done at a point which remains stable when the transmission chain changes. and the reader will recall that the only stable part of an RR-session lies in the anchor MSC. 6.3.3.3. T h e Change of the Cipher \lode During the lifetime of an RR-session. the cipher mode may change on the radio interface. This procedure is not all that easy. As explained in Chapter 4. the cipher mode is applied to all transmitted information. including signalling messages. Thus. a change of the cipher nuide entails a signalling break, N,‘ ith a possibility of message loss. In the ease of subsequent assignment and handover. the problem raised by the signalling break is solved through a lid! re-establishment of the signalling link. This is full-proof against call loss onl) because b o t h - - — the old and the new channels are available during the critical period. thus allowing the return to the old channel in case of problems on the new one. In the case of a cipher 'node change, it would be too costly to require the BTS to perform reception in both modes (ciphered and non-ciphered) simultaneously. and a different solution was adopted. I t consists o f dividing the procedure into three steps instead of two: • step I : the BTS is configured to transmit according to the old mode. and receive according to the new mode: • step 2: the mobile station is configured to the full (transmission 392 T H E (iSNI SYSTEM \ DI() fins' n I I M . \ \ N I 3 1 J . 1 triggqing step 2 to be repeated a number of limes by the infrastructure. until received. Similarly. from step 2 t o step 3. only M S t o vansmission is correct, But this is sufficient for the mobile station to repeat the uplink acknowledgement message required after step 2. to trigger step 3. In no case does a single message loss jeopardise the whole connection. (non ciphered) non ciphered) A 0 / deciphers the received now but sends in clear 0 deciphers the received flow 1 and ciphers for sending V (ciphered) A deciphers the received flow and ciphers , for sending C repetition mechanism on the link layer C Figure 6.27 — Cipher mode change: the 3 steps In order to cope with an interruption in the signalling link when a change of the cipher mode occurs, the procedure is split into three steps, so that only one direction of the transmission will he in ❑ critical +tate at each transition. The example shown here represents a transition from clear text to ciphered mode. and reception) new mode; • step 3: the BTS is configured to the full new mode. It can be accepted that steps I and 2 be inverted, but this results in an increased frame loss probability, since in this case both uplink and downlink messages would be lost between step 1 and step 2, whereas only downlink messages are lost in that phase with the given order. The critical period is split in two with such a mechanism. From the first to the second step, BTS to MS transmission functions correctly, but not i n the other direction. This is sufficient for a downlink message The procedure, shown in figure 6.27. relies Ilea i l y on the link layer mechanisms. namely the repetition o f messages after a given timeout period, in the absence o f an acknowledgement from the other side. Though not obvious, it can he shown through a careful study that. even with a window size greater than I . and/or messages waiting to he sent on the mobile station side, the procedure terminates correctly even if the radio conditions are not perfect. Vet it should he noted at this point that the cipher mode change procedure can by itself GLUM: the loss of a frame, and transmission is therefore marginally more sensitive to errors at the time of a cipher mode change. Because of the strong requirement on the order of the three steps listed above, and because ciphering is implemented by the BTS on the infrastructure side. the procedure is not managed by the BSC, but by the BTS. The BSC transmits a single order to the 13TS. which then runs the procedure, including its own configuration as u.ell as the one for the mobile station. As a matter of fact. the decision to change the cipher mode is taken by the \•ISC. and results in a cascade of messages from MSC to BSC. then from BSC' to BTS. and lastly from 131S to MS on the radio interface. The whole chain is shown in figure 6.28. The BSSMAP CIPHER NIODE C'emNIAND message indicates the new requested mode, After having extracted the new parameters from this message. t h e B S C b u i l d s u p ; i n it11.3-RII L i P l I l l < I N t i M t n * : LONINIAND message targeted at the mobile station ;41d encapsulates it in an RSNI ENCRYPTION COMMAND message sent to the BTS. The BTS then configures its reception to the new mode and sendsthe encapsulated 811.3-kit ufklirRiNi: Nun* cum \ LAND message to the mobile station using the old mode. When receiving it. the mobile station sets its configuration to the net\ mode and puts an RI1.3-RR CIPHERING MODECOMPLETE message in the sending queue. This message is not necessarily the first one to he sent in the new mode. as another one may be there, and the Itni-Rk Ctilimust; MODE COMPLETE message has no particular priority. T h e m a i n reason f o r t h e existence o f t h i s ackno,wledging message is to ensure that at least one layer 3 message is sent at this moment. in order to trigger the switching to the new mode in 394 12A DU) iii sot Id'I. N1ANAl II \ II \ THE GSM SYSTEM / \ I . BSC 8,4 MSC VLR 31)5 It i s not possible t o change the cipher mode when changing the channel (assignment o r handover) f o r unclear reasons ( l a i n phases may indeed remove this constraint). A mobile station developed according to the phase I Specificittion.v must apply the same cipher mode on the new channel as ss as used on the previous one, :111(1 this situation can lead to problems in ease Of collision between a channel iransier and a change o f the cipher mode. I n order to avoid such l p ro h...ems. a sequential approach between these two procedures must be sought by the BSC. CIPHER MODE ENCRYPTION e d C O M M A N D Discontinuous Transmission Modes C04—---OMMAND 0 CIPHERING MODE ÷ a ® CIPHERING MODE COMPLETE ( d a t a indication) \h. CIPHER MODE C 2 1 I P L E - _ 11 0 . Figure 6.28 -The cipher mode setting procedure The procedure is initiated by the MSC. but all the synchronisation management is done by the BTS, which is in charge of ciphering/deciphering. The grey area corresponds to this management, as detailed in figure 6.27. the BTS. It is when receiving any correctly decoded message (in the new mode), which implies that the mobile station has indeed correctly switched to the new mode, that the BTS switches fully to the new mode. E i t h e r t h e n o r a f t e r w a r d s , t h e R I L 3 - R R CIPHERING M O D E COMPLETE message is forwarded to the BSC, which translates it into a BSSMAP CIPHER MODE COMPLETE message to indicate to the MSC that its request has been fulfilled. The case when the MSC requires a new mode which is already the one in place has not been dealt with. This omission is purposeful, since in this case the Specifications impose that the whole procedure be run anyway. This requirement comes from the second meaning of the RIORRCIPHERING MODE COMMAND message, which is used to acknowledge an RIL3-MM CM SERVICE REQUEST (this will be explained in Chapter 7). The cipher mode command procedure is then strictly speaking not only spread over the RR layer and the link layer, but also on the MM layer. The Specifications do not indicate clearly whether it is allowed to change the discontinuous transmission modes during the lifetime of an RR-connection. The need can he identified, at least for the downlink discontinuous transmission mode, since this mode may depend on the correspondent and a single R R -session can h e used f o r several communications in succession. It should be noted as a preliminary remark that there is no need for the receiver to know forehandhe whether the sender applies discontinuous transmission or not. Therefore. no specific procedure is required for the indication o f the downlink discontinuous transmission mode to the mobile station or of the uplink discontinuous transmission mode to the infrastructure transmission devices. Means exist for the other cases, and will now be described. As f a r as the downlink discontinuous transmission mode i s concerned, it must be ordered connection by connection by the MSC to the. BSC, which configures the BTS. itself configuring the TRAU. It must be recalled at this stage that, on the transmitting side. discontinuous transmission does not only result in some frames not being sent. but it also modifies the speech coding algorithm (e.g.. sending of comfort noise frames is done only in discontinuous transmission mode). The initial command i s issued through a downlink discontinuous transmission indicator included in the messages used for the management of the basic transmission mode from the \ISC' tussm \ ssit;Nxii.m (t(il)t Fst ) and_ for channel activation toward the BTS IRSM ClIANNEI.MTIYATION I. The TRAU. in turn, is configured through an in-hand indicator set by the BTS. Means to change the downlink discontinuous transmission mode on an established RR-connection also exist, since the sante indicator is also included i n the RSM MODE MODIFY REQUEST. Besides. the iissNIAP ASSIGNMENT REQUEST message can he used by the MSC to trigger a discontinuous transmission mode change on the infrastructure side. As already mentioned, the mobile station need not be warned o f such a 396 T I Ili osm sysnim change, and no procedure has been defined on the radio interface for this purpose. The mobile station can be ordered to use the discontinuous transmission mode in the uplink direction as a cell option. This view is consistent with the consideration of discontinuous transmission as a means to improve spectral efficiency, but it does not take into account mobile stations or connections at an individual level. The cell options are regularly broadcast on the BCCH for mobile stations in idle mode, and they are also part of the general information sent to mobile stations on.their SACCH when they are in dedicated mode. The uplink discontinuous transmission mode could thus at first sight be set on a connection basis by this slow signalling method, and changed in the same way. However, the specification of the RSM protocol requires to set the general information messagessent on the SACCH on a transmitter/receiver (TRX) basis. This state of things is even more fuzzy because of the presence of an uplink discontinuous transmission indicator (besides the downlink one) in theCHANNELMODEinformation element, which is contained in the messages for the management of the basic transmission !node between BSCand BTS. The BTS has no need to know whether uplink discontinuous transmission is used or not; a possible justification for the presence of this indicator would be to ask the BTS to modify the RIL3-RRSYSTEMINFORMATION6 it sends to the concerned mobile station. A point against this interpretation is that the information in the SYSTEM INFORMATIONTYPE6messagehas three values (DTX must be applied, must not be applied or may be applied), whereas the information in theCHANNELMODEinformation element is two-valued. I t : \ U l u 10.11 It RCF. \ I \ \ \ ; I \ II v r 3 9 7 some application. When a connection is established. the cell has been chosen by the mobile station. While in connected mode. the cell is chosen by the network. The two cell choice algorithms are different. and hence may lead to different results in many eases. When this happens. and the initial channel is a TACH/8 whereas the needed channel i a TACH/F, a handover directly from the TACH/8 to a TACH/F in the right cell is quicker than performing a subsequent assignment first. and a handover next. This way of proceeding is called a "directed retry- (from the name of a similar action i n analogue networks). and can he used also for congestion cases. The handover procedure can be executed for different reasons. explained within the scope of the handover preparation function. But in all cases, the decision to attempt the handover of a given mobile station is taken by the BSC. Once so done, and once the new cell for a list o f candidate new cells) has been chosen. the actual transfer must be coordinated between the mobile station and the machines managing the old cell (BTS-old) and the new cell (BTS-new). The handover procedure comes in several varieties. according to two main criteria. 6.3.4. HANDOVER EXECUTION The handover execution procedure enables t h e network t o command a mobile station in dedicated mode to go onto another channel in another cell. The handover execution procedure is very close to the subsequent assignment procedure. The fundamental difference i s the change of cell. The handover execution procedure differs mainly from the subsequent assignment procedure by the timing advance management, by the need to transmit some data specific to the new cell, and by a few limitations. Basically the procedure was designed not to constrain the type or the mode of the new channel. As for the assignment procedure, the type 4.tad-1-riode of channel before the handover, or the type and mode of the channel after the handover can be anything. Or almost... I f the indicative value o f timer T3124 (a timer used on the mobile station side for handover between non-synchronised cells) as defined in TS GSM 04.08 is applied by mobile stations in all cases, handover toward a TACH/8 does not run properly; it is therefore not part of the test requirements for type approval of phase I mobile stations. This will be corrected in phase 2. Yet a handover from a TACH/8 toward a TACH/F is possible. This has The first criterion i s related to the timing advance issue. and impacts only the "incoming- part o f the radio interface procedure. between the mobile station and BTS-new. As expressed in the functional description of the timing advance management (see page 3-I6). two cases .7:can be distinguished: • t h e mobile station is able to compute the new timing advance up he applied with BTS-new because the old and the new cells are synchronised (synchronous hand( well: • t h e timing advance must he initialised both at the mobile station and at 131'S-new during the handover procedure (asynchronous handover). In a way, the subsequent assignment pri1/4.eslute Can he as a third case. in which the timing advance is not changed. The second criterion concerns the location of the switching point in the infrastructure. This location impacts heavily the procedures to he used between infrastructure entities: the procedure on the radio path is not impacted by such a consideration. except to distinguish the special case of intra-cell handover which uses the same procedure as for subsequent channel assignment. 398 399 TF1F. GSM SYSTEM RADIO RESOURCE NIANAGIAIENT IASC VIA must be recalled that the anchor MSC, which originally established the RR-session, may be the same as MSC-old or MSC-new. or both, but may also be different. I f the call already goes through two MSCs. the m e l t MSC and MSC-old are distinct. If a handover is needed toward a cell of a new MSC different from MSC-old, the switching point is the anchor MSC. Figure 6.29 shows the most complex case, when all entities are physically distinct. Whether synchronous o r asynchronous. whether inter- o r intraMSC, and whether inter- or intra-BSC. the execution o f handover is composed of two main phases: MSC VIA •rat., NLAi',;• •, arr.' • i n a first phase, BSC-old triggers a set o f events with the purpose o f establishing the future communication path. Once this i s done, this phase terminates with the sending o f a handover command to the mobile station; • i n a second phase. the mobile station accesses the new channel. This access triggers the switch ()I' paths in the infrastructure. and the release of the old path. 6.3.4.1. T h e Set-Up of the New Path BSC: intra-BSC handover • 2 M S C : inter-BSS, intra-MSC handover 3 anchor MSC : inter-MSC handover (when MSC-old = anchor MSC), or subsequent inter-MSC handover (otherwise). Figure 6.29 — Position of the switching point at handover The switching point depends on the relative position in the infrastructure hierarchy of the BSCs controlling the two cells. In all cases, the anchor MSC remains involved in the communication path, and there can he hut only one other MSC in the path. In order to describe the different cases, the suffix "old" shall be used to refer to all functional entities along the communication path before the handover, and "new" shall be used for the path after handover. BTS-old, BSC-old and (transit) MSC-old represent the machines i n charge of the old cell, and BTS-new, BSC-new and (transit) MSC-new the machines i n charge o f the new cell. I t may happen that BSCold = BSC-new, or MSC-old = MSC-new. Regarding the relay MSC, it Once the decision o f handing over a given communication has been taken by BSC-old. this must be indicated to the switching point. The latter must in turn establish the terrestrial resources. i f need he. up to BSC-new. signal to it to allocate a radio resource and more generally provide all impacted machines with the information they need for the handover and the future management of the connection. This information includes: • t h e transmission mode, used t o choose and configure the transmission path i n an appropriate way. including the new radio channel: • t h e cipher mode: • t h e identity o f the origin cell. used to determine whether the handover can he done in a synchronous or an asynchronous way: • t h e mobile station classmark. used for future man gement of the connection. Once BSC-new is aware of all this information. it is in a position to allocate the new Channel. to build the R11.3-RR 11ANDOVER (.'UNHAND message and to transmit i t to the switching point, which in turn will I r NO I N t i \ I \ \ \tri \ \ 4113 From the Switching Point to BSC-new BSC-old Switching point BSC-new The purpose o f this step is t o establish the signalling pathway between t h e switching p o i n t a n d BSC-new, t o prepare f o r t h e establishment of the circuit i f need be and to provide the machines aloftg the new path with the information they need. The events at this stage depend on the relative position of the switching point and BSC-new, as shown in figure 6.32: A handover is required>. Start of path establishment a. BSC-new is the switching point (BSC-old = BSC-new): Handover command to MS M End of path est.. and handover command to S this step is internal. The BSC is aware o f all the relevant information since i t already manages the current context o f the connection. No terrestrial circuit needs to be allocated apart from the Abis circuit linked to the new radio channel. h. MSC-new is the switching point (BSC-old # BSC-new, MSCold = MSC-new): when receiving the indication that a handover is required. MSCnew establishes an SCCP connection towards BSC-new. This connection w i l l be used throughout the life o f the new RRconnection. M S C -new then transmits a BSSMAP HANDOVER REQUEST message to BSC-new, including the information on the cells (both origin and target cells), the transmission mode (derived from the present needs, hence possibly differing f r o m the characteristics of the connection established through the old cell). the cipher mode (which must be the same as in the old cell, since the mobile station will assume so). the classmark and finally the reference o f the terrestrial channel between MSC-new and BSCnew if need be. c. Anchor MSC is the switching point, and differs from MSCnew: this corresponds to the most complex case. The tasks of the anchor MSC cannot be done in a single step as in the previous case, because the new communication path between the anchor MSC and BSC-new (through MSC-new) may transit via the PSTN or the ISDN. Standard inter-switch procedures must therefore be used, which are part of TUP or ISUP (or national variants of them). These protocols do not provide the means to convey the relevant GSM information. They are therefore only used for circuit set-up and M A P / E procedures are used f o r the specific handover signalling needs. S anchor MSC ' M S C -new B S C - n e w (internal) a ) BSSMAP HANDOVER REQUEST b ) MAP/E PERFORM HANDOVER BSSMAP HANDOVER REQUEST c) Figure 6.32 S t a r t of path establishment At hanklo‘e, git The sea path is established starting H. ipplicahle ith the channel on the rick\ A interlace. It is triggered 11) a message coming front ilk anchor NISC. Oril i n the next step mill the actual circuit h e M e e l l M O relay M M . he established it dppliedble. The anchor NIS(' p ro%ides the required information to NIS( ' new through a mxPitr. ithRickm o u v i i k message: receiving it. NISC-new establishes an SC(.'11 connection B S C new, allocates if need he ,r circuit on the A interface and transmits a BSS\I:\P HANN) i t it hytTst message to BSC-nev coma/fling the information Ricci ed i n the x l x i . T message. as explained in case b. From BSC-new Back to the Switching Point At this point BSC-nev [mist try to allocate the radio channel. This .1procedure results i n either a posits Ls o r it negative .insv cr. Unless BSC-old Decision A handover is required ) Handover 4 command to MS BSC-old BSC-1 Switching point Switching point Start of path establishment 1 0 . End of path est., and handover command to MS Switching point BSC-old BSC-new A handover is required Start of path establishment Handover command to MS End of path est., and handover command to MS Allocation and activation of radio channel B S C - n e w The MS has accessed cell B 4 — BSC-old Release of path a) Figure 6.30 - Handover execution sequence M S -old anchor MSC (internal) b) BSSMAP HANDOVER REQUIRED c) BSSMAP HANDOVER REQUIRED Once the handover decision taken, the set-up of the new path unfolds in four steps, to prepare for mobile station access. C MAP/E PERFORM SUBSEQUENT HANDOVER )0' convey it to the mobile station along the old path. as shown in figure 6.30. Let us study these steps and the respective variations in detail. Figure 6.31 — I lantImer requirement from the , . w i n g WS( t o the . ‘ ‘ itching point From BSC-old to the Switching Point When applicable (depending On the relative poNit ion o f the s u . k . u n g the liSSNIAP IIANIX)VERRI.QUIRED message n a y he folloued hy a mAilv pa r i , „ < \ si \ I I \ i „ , \ • 1 R wes‘agc. The purpose of this exchange is the transmission of the information that a handover is needed, and toward which cell (or cells). The different cases depend on the nature of the switching point (see figure 6.31): - a. BSC-old is the switching point (BSC-old = BSC-new): c. Anchor MSC is the switching point. and diners from MSC-old: this step is internal and does not raise any problem. the behaviour of BSC-old is the same as in the previous case. but the behaviour o f MSC-old when recei% ing the tissm,\P_ HANDOVER REQUIRED message i s different. I t translates t h e message in a MAP/E PERFORM SUBSEQUENT HANDovER message towards the anchor MSC. Both messages have similar contents. b. MSC-old is the switching point (BSC-old t BSC-new, MSC- o l d = MSC-new): in all cases when the target cell is not under its control, BSC-old sends a BSSMAP HANDOVER REQUIRED message t o MSC-old, containing the identities of the target cell(s) and of the origin cell. t 404 THE GSM SYSTEM RADIO RI SOURCE otherwise specified, queuing should not be applied. because other machines are waiting for the answer, and timers are running. Negative cases will be dealt with further on. As for the "happy-ending" case, when a channel can be activated and the corresponding device in BTS-new is ready for mobile station access (an exchange of RSM CHANNEL ACTIVATION and RSM CHANNEL ACTIVATION ACKNOWLEDGE takes place BSC-old point Start of path A handover is required establishment MS anchor MSC BSC-new ó BSSMAP (internal al HANDOVER REQUEST ACKNOWLEDGE at this point in time, both terrestrial paths are set-up. towards the old and the new BTSs. b) BSSMAP MAP/E PERFORM in a bssmap handover request acknowledge message. Nothing else to MS MSC-new a. BSC-new is the switching point: B S C - n e w e n c a p s u l a t e s the r i l 3 - r r h a n d o v e r c o m m a n d m e s s a g e handover command 4 There again, different cases arise depending on the respective positions of BSC-new and the switching point. as shown in figure 6.33. h. MSC-new is the switching point: - End of path est., a n d H a n d o v e r c o m m a n d to new which builds this message (which will be eventually sent by BSCold), and thus decides for instance whether the handover will be synchronous or asynchronous, chooses the handover reference, and the initial MS transmission power. In fact, one may consider that BSC-new is in charge of the mobile station from this very moment. BSC-new Switch ing on the (new) Abis interface), BSC-new builds up the RIL.3-RR HANDOVER COMMAND message and transmits it to the mobile station. via the switching point and the old resources. It should be stressed that it is BSC- 105 ALASAGESH.ST H A N D O V E R ACK € HANDOVER REQUEST ACKNOWLEDGE C) needs to be done at this stage, since the terrestrial path is already completely established. c. The anchor MSC is the switching point, and differs from MSC- Figure 6.33 new: BSC-new acts as in case b above. When receiving the BSSMAP HANDOVER REQUEST ACKNOWLEDGE message, MSC-new inserts the included RIL3-RR HANDOVER COMMAND message in a new envelope, the MAP/E PERFORM HANDOVER RESULT message. This message contains a telephony-like number (provided by MSCnew) to allow the anchor MSC to set up a circuit through normal ISUP or TUP means. This h a n d o v e r n u m b e r is allocated solely for the anchor MSC to establish the circuit with MSC-new, and serves as a reference for MSC-new to link the context with the incoming circuit. The same MAP/E exchange (PERFORM HANDOVER and its RESULT) serves both purposes of providing the information needed for circuit establishment and carrying the RIL3RR HANDOVER COMMAND message back. ready to be sent to the mobile station along the old path. Upon receipt of the MAP/E PERFORM HANDOVER RESULT message. the anchor MSC is able 10 Once B S C . n e w (and c o r r e s p o n d i n g r e s o u r s which then triggers in turn and fi applicable the set-up of hte circuit b e r c e n the anchor A S C and the new retas s i S t " w h e n the sAP/E PeRFORStEAsposeR Rest or message has been received. set-up the communication with ASC-new. through. e.g.. the 1111 a n d ACM m e s s a g e s o f I S U P. Depending upon implementation choices for the switching machine. the paths can be at this stage linked in a w o - w a y conference bridge. a one-way conference bridge or not linked at all. In the first two cases. downlink user data is transmitted to the mobile station via both 406 41111,.. GSM SN'STIAI os BSC-old Ahangover is required Start of path establishment Handover command to MS End of path est.. and handover command to MS BSC-old Handover performance can be improved by the insertion at the BSC of aconference bridge, either one-way (downlink) or two-way. sothat both paths of an Mira-BSC handover maybe connected in parallel to the same connection towards the MSC. BTSs. In the first case only, the two uplink flows are combined into a single one towards the other correspondent (see figure 6.34). It is feasible to insert such a conference bridge only in some cases. for instance when the transmission mode is speech and the transcoders on the old and new paths are different (i.e., speech is carried at 64 kbit/s at the switching point). From the Switching Point to the Mobile Station The last step o f the first phase o f handover execution consists simply in sending the RIO-RR HANDOVER COMMAND message to the mobile station, as shown in figure 6.35. according to the three following cases: a. BSC-old is the switching point (BSC-old = BSC-new): b. MSC-old is the switching point (MSC-old = anchor MSC): c. the anchor MSC is the switching point, and differs from MSCold. BSC-new Switching point MSC VLR Figure 6.34—Conference bridge in the BSC for handover -107 1:ADIt I I : 1 5 R u t NI. . \ t i l . N 11 : : \ I 4 MSC-old anchor MSC a) ( i n t e r n a l ) BSSMAP HANDOVER *COMMAND b) 4 c) BSSMAP HANDOVER COMMAND MAP/E PERFORM SUBSEQUENT 4 HANDOVER ACK Figure 6.35 Sending back of the c o s i s i i ‘ s n As a last step in the network before mobile siation access on the new channel. the Hit. 3-104 ins:D(1\TR cintst i n e . . a g c is sent to the n u b i l e station. The RIL3-88 ii.ANDovklt ('omm.\ NI) message is carried °Ye): the_ different interfaces in a variety of different envelopes. as shown in table 6.8, which also summarises the transfers already mentioned i n the previous paragraphs. The RIO-RR HANDOVER COMMAND message identifies the new cell only through its beacon' frequency and its BS1C. This is sufficient for the mobile station. and the full cell identity will be read later by the mobile station on the SACCH sent by BTS-new. I S A i n t ) K 1 m i t IC( I t NI V \ v.i..\11..N.1 uu:uSMSISIEN Interface Encapsulating message between BSC-new and MSC-new BSSMAP HANDOVER REQUEST ACKNOWLEDGE between MSC-new and anchor MSC M A N E PERFORM I IANDOVER RESULT between anchor MSC and MSC-old M A N E PERFORM SUBSEQUENT HANDOVER RESULT between MSC-old and BSC-old BSSMAP HANDOVER COMMAND Table 6.8 —The transfer of the R L3-RR HANDOVER COMMAND message subsequent handover failure HANDOVER REQUIRED REJECT The message aimed at the mobile station contains everything the mobile station may need to access the new channel, and is carried unaltered in a variety of encapsulating messages over the terrestrial interfaces. M S C VLR anchor MSC no radio resource availableas MSC VLR MSC VLR HANDOVER MSC-old MSC-new FAILURE BSC Not so Successful Alternatives BSC-new BSC-old A number of obstacles may block the smooth succession of events as described above. The main obstacle is the non-availability of radio or terrestrial resources. In this case, an unsuccessful indication is carried back from BSC-new. Considering the longest path, a BSSMAP HANDOVER FAILURE message is transmitted from BSC-new to MSC-new, which in turns triggers backward messages as shown in figure 6.36. Alternatively, a watchdog timer may expire in BSC-old, resulting in a similar situation. Two possibilities can be envisaged. Either a new handover attempt towards the same cell is performed after some time, o r a handover towards another cell is attempted. In the first case, the failure indication goes all the way back to BSC-old, which will re-initiate the handover process when i t decides so. A l l resources which have been allocated along the new path are released. The second case can also be treated this way. However, an alternative exists in the Specifications, whereby BSCold may provide MSC-old with an ordered list of suitable target cells (this applies as already mentioned mainly in the case o f rescue handovers, when staying in the old cell is not going to be a good alternative to the first target cell). This list is not conveyed to the anchor MSC if MSC-old differs from it and is the switching point. The failure indication, when reaching MSC-old, will therefore trigger a handover attempt towards the next cell in the list. Only when all cells have been tried in vain will BSCold be given the failure indication (and the full control of the connection back). This possibility of multiple cell choice in the BSSMAP HANDOVER REQUIRED message is an option for the BSC. Speed requirements tend to favour the multi-cell approach, whereas the optimisation of cell allocation favours the single cell approach, since only BSC-old is in a position to change the cell list according to up-to-date measurements. Figure 6.36 •- I landover failure at the new BSC The failure indication goes all the way hack to BSC-old as shown. so Mai the decision to retry or perform another action can be made. Alternatively. MSC-old may attempt handovers toward other cells if it is in possession of a list of candidate target cells. 6.3.4.2. M o b i l e Station Access and the Conclusion of the Procedure The mobile station is completely unaware o f the infrastructure processes and decisions until i t receives the itti.3-R1t HANDoviiit COMMAND message. As already mentioned. this message contains all the information characterising transmission on the new channel c e p t for the cipher mode which is assumed to remain the same as on the old channel), and the data needed for the access procedure. In pan icular. it indicates to the mobile station whether the synchronous or asynchronous handover procedure should be followed. In both cases. thanks to pre synchronisation, the mobile station is able to synchronise itself quickly on the new channel and starts reception immediately. It will actually receive speech or data from this point, i f applicable and i f the switching point uses a conference bridge. —r I k r USM SYSTEM R A W ( R I - . S ( 11 \ \ c 411 ;11: \ The RII.3-RR PHYSICAL INEORNIATION message i s the o n l y case i n MS BTS-new BSC-new M S C -new a n c h o r MSC B.SC • HANDOVER ACCESS justified b y p e r f o r m a n c e r e q u i r e m e n k . F o r e ff i c i e n c y reasons. t h e P i L 3 RP PHYSICAL INFORMATION message m a y he sent several times i n a r o w. MSC VLR MSC VLR (HANDOVER DETECTION) ( H A N D O V E R DETECT) PHYSICAL INFORMATION until the reception o f normal bursts f r o m the m o b i l e station makes it clear to B T S - n e w that i t has received the message. T h i s w o u l d n o t have been so easy i f the RIL3-RR PHYSICAL INFORMATION message had been sent b y the B S C . The B T S m a y as an o p t i o n indicate t o the B S C that i t has received SABM UA HANDOVER COMPLETE the Specifications w h e r e a message a b o v e t h e l i n k l a y e r i s sent a s a n autonomous decision b y the B T S . T h i s departure f r o m the getteral rule is >. HANDOVER COMPLETE S E N D END S I G N A L adequate RIL3-RR HANDOVER ACCESS b u r s t s o n t h e a l l o c a t e d c h a n n e l . through a n RSM HANDOVER DETECTION message; B S C - n e w m a y i n t u r n pass t h e i n d i c a t i o n o n t o M S C - n e w t h r o u g h a B S S \ I A P ILANDOVER DETECT message. T h i s m e c h a n i s m a l l o w s the s w i t c h i n g p o i n t (except i n the c a s e w h e n i t i s t h e a n c h o r M S C . because t h e i n f o r m a t i o n i s n o t carried b y the M A P / E p r o t o c o l ) t o switch the c o m m u n i c a t i o n path at this moment w i t h o u t w a i t i n g f o r the Hill c o m p l e t i o n o f the procedure. When i t is i n normal transmission mode. the m o b i l e station sets the Figure 6.37 —Access in the case of an asynchronous handover Only following reception of RIL3-RR PHYSICAL INFORMATION messages does the mobile station switch to normal transmission mode with the timing advance as indicated, and sends an RIL3-RR HANDOVER COMPLETE message after having established the SAM 0 signalling link on the new dedicated channel. A s f a r as t r a n s m i s s i o n f r o m t h e m o b i l e s t a t i o n i s concerned, t h e type o f handover intervenes. I n t h e case o f a synchronous handover, t h e m o b i l e s t a t i o n f i r s t s e n d s a f e w access b u r s t s ( t h e PiL3-RR HANDOVER ACCESS m e s s a g e ) , t h e n s t a r t s n o r m a l t r a n s m i s s i o n b y a p p l y i n g t h e computed t i m i n g advance. I f the h a n d o v e r i s a n a s y n c h r o n o u s o n e (see figure 6 . 3 7 ) , the m o b i l e station continues to send access bursts u n t i l i t has received a n RIL3-RR PHYSICAL INFORMATION message f r o m B T S - n e w, c o n v e y i n g t h e a c t u a l t i m i n g a d v a n c e t o a p p l y. O n l y t h e n d o e s i t s t a r t normal t r a n s m i s s i o n . I n b o t h c a s e s , t h e R I L 3 - R R HANDOVER ACCESS message o n l y contains a n 8 - b i t h a n d o v e r reference. T h i s message i s t h e o n l y case w h e r e short access bursts are used o n a dedicated channel. T h e handover reference ( n o t to be confused w i t h the handover number) is part o f the d a t a t r a n s m i t t e d t o t h e m o b i l e s t a t i o n i n t h e RIL3-RR HANDOVER COMMAND message a n d can b e used b y B T S - n e w as an a d d i t i o n a l check that the accessing m o b i l e station is indeed the expected one. link l a y e r t o a c k n o w l e d g e d m o d e f o r s i g n a l l i n g messages b y sending an S A B M f r a m e answered b y a V A frame. T h e m o b i l e station then sends an RIL3-RR HANDOVER C o m I T L I E message. w h i c h w i l l h e c a r r i e d b y t h e infrastructure u p t o t h e s w i t c h i n g p o i n t . w h e n a p p l i c a b l e t h r o u g h a BSSMAP HANDOVER CONIPLEIE message f r o m B S C - n o k t o M S C - n e w and through a NiAP/F. SEND END siGNAL f r o n t \ I S C - n e w t o anchor M S C . T h e switching p o i n t w i l l release t h e p r e v i o u , p a t h b y s e n d i n g a p p r o p r i a t e messages (NIAP/E SEND END SIGNAL R E M I T f r o m anchor M S C t o \ 1 S ( ' . old. and BSSNIAP CELLAR CONIMAND f r o m M S C - o l d t o B S C - o l d ) . relayed up t o B S C - o l d w h i c h releases t h e p r e v i o u s r a d i o channel h e l d u p u n t i l this p o i n t . T h e release o f resources w h i c h t h e n t a k e s p l a c e o n t h e A interface, the A b i s interface a n d at t h e H I S have n o specific differences compared to an R R -session termination. The s e n d i n g o f t h e h a n d o v e r c o m p l e t e i n d i c a t i o n t r i g g e r s t h e switching o f p a t h s h e m een t h e o l d a n d t h e i l e x \ t h i s h a s Uut already been d o n e (e.g.. u p o n access detection o n the ne \1/4 channel). T h e question o f t h e n e c e s s i t y o f t h e h a n d o k er c o m p l e t e i n d i c a t i o n c a n h e raised at t h i s p o i n t : w h y i s there a need o f a t w o -stage mechanism? T h e difference betw ee n t h e access a n d t h e c o m p l e t i o n i s M a t o n l y t h e l a t t e r triggers the release o f the previous channel. O n l y when i t sends the M I . 3 : R R HANDOVER COMPLETE message does t h e m o b i l e station abandon a l l possibility o f returning t o t h e o l d channel. T h e f i r s t stage w a s added t o shorten the interruption time. The returri+on the o l d channel i l l case o f problems is s i m i l a r to the subsequent a s s i g n m e n t c a s e : o n l y t h e n a m e o f t h e message changes: 413 THE. GSM SYSTEM 12ADIO RES( It N I A N . V N I lcN T RIL3—RR ASSIGNMENT FAILURE in o n e case, RIL3—RR HANDOVER FAILURE more efficient in these cases. In sonic systems crafted for a microcell or pico-cell environment, all handovers are triggered by the mobile station alone. However, network handover control has many advantages when it can be applied: this sterns from the obvious fact that the network has a much better understanding of the general situation than any single mobile station. Call re-establishment may then be considered as a kind of mobile station triggered handover, but limited to the extreme case o f rescue handover when communication with the current cell is effectively lost. One could foresee that the importance of this feature will grow in the future. For instance, it can be imagined that in some environments the procedure w i l l b e triggered sooner. s o as t o improve the system performances where network triggered handover will have shown its in the other. When this unsuccessful outcome arrives, BSC-old is advised and transmits the information up to the MSC-old i f applicable, through a, BSSMAP HANDOVER FAILURE message. The MAP/E protocol introduces some limitations. When the anchor MSC is different from MSC-old, there is no means o f passing on this information between MSC-old and the anchor MSC, since no message exists on MAP/E for this purpose. The only way for the anchor MSC to react in this case is through timer expiry upon n o n -reception o f a message f r o m BSC-new indicating the completion of the handover. Whatever the means by which it recognised a failure condition, the switching point releases the new path, using normal release procedures, and it is up to BSC-old to decide what action to perform, e.g., make another handover attempt. An intra-BSC handover is usually performed autonomously by the BSC. A s explained in the handover preparation section, i t is a BS$ implementation option not to involve the MSC (the one in charge of tit BSC) at all in the decision when the best cell, as seen from the BSC, la also under the control of the same BSC. In this case, the whole handovp will unfold without any knowledge of the MSC. In order to advise it a handover has occurred successfully, a BSSMAP HANDOVER PERFO message is sent from the BSC to the relay MSC, possibly relayed h t MAP/E NOTE INTERNAL HANDOVER message from the relay MSC to anchor MSC when they differ. This message may also be sent in the c a l of a handover internal to the relay MSC (case when MSC-new = MSG old # anchor MSC). The sending or not of this MAP/E message depends on Operation and Maintenance requirements. l o f 6.3.5. CALL RE-ESTABLISHMENT In a radio mobile environment, there is always some risk that 2 connection will be suddenly cut. This may happen because of a brutel propagation loss, due to obstacles such as bridges, tunnels, or simply buildings in the case of handhelds. But another cell could often be used to continue the communication either immediately or after a very short time (think about a short tunnel through a hill). The handover preparation and execution are a means of limiting the occurrences of call loss, but they cannot suppress them totally. In the future, when cells become smaller and smaller, the risks will increase; The performance achieved b y handover algorithms running on the network side will then lessen, in prediction accuracy as well as in reaction time, and connection loss probability will increase. The mobile station has in fact some ways to determine that a handover is needed, and may be limits. Despite these considerations. the call re-establishment procedure is a poor relative in the GSM procedure tribe. and has serious limitations in phase 1. Let us see what there is of it. As a general point, it should be noted that the call re-establishment procedure is not a full-blown RR procedure. We will see in this section messages from the RIL3-MM protocol. Only because of its kinship with handover do we present the re-establishment procedure in this chapter and not in Chapter 7. - - - - Call re-establishment has two parts. The mobile station has the leading role in the first one, which is very close to an access procedure. The second part is the network's, and consists in the recovering of upper layer contexts. The closeness of call re-establishment with initial access is quite normal, because the mobile station has t o start from scratch. The differences are important, and come mainly front the requirement for speed: when a connection is lost. a tinier starts ticking in the anchor MSC, and at its expiry everything related to the moribund connection is erased. As a consequence, any fraction o f second lost in the call reestablishment procedure increases the risk of total loss. t The first 'sue is to determine the new cell. the speed requirement— I limits the choice o the neighbouring cells already known and with which the mobile stati n is pre-synchronised. because finding new cells and getting synchronisation may take seconds. The selection rule is to just select the one with the highest signal strength. The parameters to compute the different radio criteria (the CI's of the possible cells, which will be described in Chapter 7) as for idle-mode cell selection are not known. and receiving them takes time. The Specifications nevertheless require the mobile station to check the radio criterion for the chosen cell. This constraint forces the mobile station to wait for the reception of a BCCH K A I ) 1 1 1 1 1 : 1 ( Pt I<1 I k o l a 31.3 I LIM toute when need ,..... The M I f m a y even choose t o perform a n authentication despite the incur. d delay. The BSC then performs the needed procedures within t h e B T S a n d w i t h the mobile station (subsequent assignment, ciphering start, mode modification....). Only then can an RIL3-MM CM SERVICEACCEPT message be sent to the mobile station, and the end to end communication resumed. Let us note that means for the network to reject the request have been foreseen. The Rit.3mm CM SERVICE REJECT serves this purpose (in particular with the cause "call cannot be identified". which is likely to he used a lot). message containing the required parameters, and this ..an take up to three quarters o f a second ( i f the first decoding attempt succeeds!). Another formality has to be checked: the chosen cell must not be barred, and call re-establishment must be allowed in it. The corresponding indications are part of all BCCH messages, so that this checking does not add any more delay. After having received the required BCCH message, and checked the radio criteria and authorisations, the mobile station sends an access request on the RACH. Though not specified in the Specifications (as many other small details Of the procedure), it seems that it is allowed to use the RACH on T N 0, even i f there are others. The access request indicates the reason for access (i.e., call re-establishment), so that the network is aware o f the critical nature o f the request. The required channel is not indicated, but the network can easily play safe and allocate a TACH/F. A number of other fuzzy points exist in the Specifications. They include in particular, the recovery o f several CM-transactions and the corresponding transaction identifiers (see Chapters 5 and 8). and. more important, all the cases of collision, when the connection loss happened during an ongoing procedure in any layer. Another issue is the release of the old path. As far as can he analysed, the anchor MSC can and should release it once aware of the re-establishment attempt. even if in the same BSC (or the same cell). One wonders if call re-establishment will really he used in phase I. This is hopefully an area in which the future phases of the Specifications will bring improvements. The initial message i s a n R I O -Mm CM RE-ESTABLISHMENT REQUEST. Its information contents are minimal: the subscriber identity and the classmark. The mobile station does not volunteer anything else, and the network has to use this to find out everything about the lost connection! Among the conspicuous missing data known directly or indirectly by the mobile station are the cell with which the connection was lost, the identity of the anchor MSC and the required type and mode of the channel. 6.3.6. R R -SESSION RELEAsE In any case, the only case catered for by phase I protocols is when the new cell and the previous ones are managed by the same MSC. The anchor M S C i s then implicitly determined. Moreover, even i f the previous cell was known, it would be to no avail since no mechanisms have been included to recover the RR-session except when the new MSC is the anchor MSC. A n inference i s that call re-establishment is impossible when there is a relay MSC. When all needs for an RR-session have disappeared. for instance because a location updating procedure has ended. because a call i s terminated, or because of a failure. the mobile station must go back to idle mode and the resources must he released, in order to he free for allocation for other needs. This release mechanism is done through a socalled "normal release" procedure. \1/4 Inch is always triggered t h e anchor MSC. If distinct from the relay MSC. the anchor MSC releases the RRsession by sending a MAIIE SENO LND SIGNAL RESUL'I to the relaN MSC_ on one hand, and by releasing the circuit if prescm, througlilsl release procedures. sent from The next step is the BSS\IAP ('LEAR commAND message relay MSC to BSC. This message may be piggybacked on an MVP RELEASE message releasing the SCCP connection. In Mi. case. the BSC acknowledgement (BSSMAP CLEAR CONIPLETE message) must h e piggybacked on the SCCP RELEASE cOMPinE message. The clearing actions of the BSC can take place in parallel with the sending of the BSSMAP CLEAR COMPLETE message. since the Specifications do not The fact that the required type o f channel is not given by the mobile station has no clear explanation. This is a source of delay, because the BSC has to wait for the indication from the MSC to allocate the right type of channel. Unless, as already mentioned, i f the BSC gambles and initially allocates a TACH/F. The recovery o f the contexts is then entirely an MSC issue, and must be done with only the subscriber identity to start with. From this the MSC must find the old context ( i f it still exists—it could have been erased after a timer expiry, or simply because the correspondent was not patient enough). Then the MSC performs an assignment procedure and possibly a ciphering start procedure, telling the BSC the type of channel required, the mode, and so on, and allocating the BSC-MSC terrestrial pt 2.F' - r I .11 I. \ I)II) lastiiIiReI. \IA \ TI lli GSM SYSTEM U relay MSC a n c h o r MSC 7C 7 C MSSVLF, M S C VLSI SEND END CLEAR S I G N A L RESULT COMMAND CHANNEL RELEASE CLEAR COMPLETE link disconnection: UA RELEASE INDICATION RF CHANNEL RELEASE RF CHANNEL RELEASE ACKNOWLEDGE> \ II \ -117 or when the interference level is too high. One role of handover is to cope with such cases, but this is not always possible. e.g.. when the user has gone out o f coverage i n an underground car park for instance). o r switched o ff his mobile station in the middle o f an RR-session. Such cases must be detected, i n order f o r the infrastructure t o free the corresponding resources. The mechanism specified in the S'peenIcations l'or this purpose consists in having both the mobile station and the BTS monitor the message loss rate on the SACCII. Let us recall that messages are sent regular) 'about twice per second) in both directions of the --•-....1ACcH. throughout the life o: ta connection. A ptm I n t error detection mechanism has been included in signalling messages. enabling the receiver to estimate message loss. This estimation is done through a counter. incremented in case o f a correctly received block and decremented i n the other case (see figure 6.39). When the counter reaches a minimum threshold. the link is considered as broken. On the mobile station side. this et cm leads to a return to idle mode she mobile station may subsequends WICMIM a reestablishment. see page 412). The infrastructure is able t o adjust the mobile station behaviour. in order to allow fine tuning on a cell basis. althou01 it seems minket> that it %%il be necessary to manage differences between mobile stations. The relevant parameters are sent regularly on the SACCI I as well as on the W i l l . Setting them to satisfying values must b e clone through field e r a i m e n t . i l l operational networks. O n the infrastructure side. the counter is in the BTS and the failure indication is gis en to the BSC in :In Ks \I (i)NNECTI( rs FAILURE 1•DIC m e s s a g e . Figure 6.38 —The normal release procedure of an RR-session Normal release is always triggered by the anchor MSC. but the BSC manages the return of the mobile station to idle mode before releasing the BSS resources. A impose any specific order between these two actions. Once the BSC has ordered the mobile station to go back to idle mode through an RIL3-RR CHANNEL RELEASE message, the mobile station disconnects the signalling link, and this event is reported by the BTS to the BSC through the RSM RELEASE INDICATION message. The Specifications include a number of timers and repetitions in order to ensure that whatever frame losses may occur during this period the mobile station eventually goes back to idle mode and stops using the channels. This is of prime importance to avoid -allocating a channel to a mobile station when another mobile station may still transmit on this same channel. Only when the BSC is sure that the mobile station has left will it de-activate the BTS device, through the RSM radio link counter RADIO LINK TIMEOUT 4 — o r—, Cl O O = SACCH blocks decoded expected. but not decoded link deemed broken RF C H A N N E L RELEASE / R S M R F C H A N N E L RELEASE ACKNOWLEDGE exchange. The corresponding radio channel is then considered as part of the pool of free channels by the BSC. The whole procedure is illustrated in figure 6.38. An R R -session may also be released i n other conditions, f o r instance when the infrastructure has lost actual contact with the mobile station. Such a situation may arise when propagation conditions are bad, Figure 6.39 - SACCI I counter for link management A counter enables each receiver to estimate the frame loss rate on the SACO! (downlink for the mobile station. uplink for the infrastructure). When this counter reaches O. the link is deemed broken and :iction. are taken lo release the resource.. 110'. liSN1 S1 S W h i l e the detection o f transmission loss s i m p l y (riggers the m o b i l e station t o abandon t h e c o n n e c t i o n a n d r e t u r n t o i d l e m o d e . t h i n g s are a little bit m o r e c o m p l e x o n the infrastructure side. Since both directions o f transmission m a y experience d i ff e r e n t qualities. reaching the threshold i n the u p l i n k d i r e c t i o n d o e s n o t necessarily i m p l y t h a t t h e m o b i l e station also experiences a break o f the l i n k . T h e infrastructure m u s t nevertheless make sure that the m o b i l e station leaves t h e channel before d e e m i n g the channel t o b e f r e e . T h e B S C t h e r e f o r e c o m m a n d s t h e B T S t o s t o p transmission o f d o w n l i n k S A C C H f r a m e s ' ( b y a n R S M D E A C T I VAT E SACCH message), so that the m o b i l e station c o u n t e r w i l l i n e x o r a b l y reach the m i n i m u m t h r e s h o l d a f t e r s o m e g i v e n t i m e . T h e m o n i t o r i n g o f the u p l i n k channel. m a y t a k e p l a c e e i t h e r i n t h e B T S o r i n t h e B S C . I n t h e first c a s e , t h e B T S m a y r e p o r t t h e l o s s o f t h e l i n k t h r o u g h a n R S M CONNECTION FAILURE INDICATION message to the B S C . Once t h e l i n k f a i l u r e i s dete c te d a n d n o t i f i e d t o t h e r e l a y M S C through a BSSMAP C L E A R REQUEST m e s s a g e , t h e s a m e A i n t e r f a c e exchange takes place as i n the case o f normal connection release. i.e.. the M S C sends a BSSMAP CLEAR COMMAND message, answered as expected by a BSSMAP CLEAR COMPLETE message f r o m the BSC. 6.3.7. LOAD MANAGEMENT PROCEDURES A f e w procedures i n the R R p l a n e a l l o w t h e M S C and the B S C t o deal w i t h o v e r l o a d s i t u a t i o n s . T h e y i n c l u d e m e a n s t o e x c h a n g e information between machines so that each one gets the information i t needs a b o u t the c u r r e n t load situation: a n d means t o act so as to l i m i t the effect o f the overload. Procedures d e a l i n g w i t h l o a d management appear in t w o main areas: R A C H and PA G C I I load. and T C H load. 6.3.7.1. L o a d on Common Channels Some i n f o r m a t i o n c o n c e r n i n g t h e l o a d o n t h e R A C H a n d o n t h e . E A G C H c o u l d b e i n f e r r e d by., t h e B S C f r o m t h e requests i t sends t o . o r receives f r o m , the B T S . S t i l l . the B T S i s i n a better position t o assess the exact load on these channels. A message, RSM CCCI I LOAD INDICATION, has thus been introduced in t h e RSM p r o t o c o l , t o enable t h e B T S t o send s o m e i n f o r m a t i o n about the R A C H and P A G C H loads to the B S C . T h e c o n d i t i o n s f o r sending this message a r e set t h r o u g h t h e O p e r a t i o n S u b -System; i t c a n b e r e g u l a r l y sent, o r o n l y w h e n t h e l o a d o n o n e o f t h e c h a n n e l s i s a b o v e s o m e threshold. T h e message pertains to a single pair R A C H / PA G C H . k % D u ) k i •Nt )1 1 he too short in case of frequency hopping t o send a complete description o f the channel frequency characteristics. Therefore. as explained on page .'ibl . part or the channel description information. applicable to all initial assignmenis. is tiroadcast regularly (once per second), in the ( / i t f 7/.1x ‘, H. /4.)cm/innx ammeter. In most cases. this information will correspond to the list or all frequencies which might be used for ll,'Lli,:alell channels i n the cell. although it need only contain the list in all frequencies \\ hich might be used for itillia/ channel assignment in the cell. This list is critical in.1 Ormation for ae...ess. since the oilier frequent:\ parameters of the allocated channel cannot be understood II\ the mobile station without it. Iltmever. a mobile station could pellet:it \ start the access procedure t i.e.. send an Itlt.3-kl: (AR \ \Fr- Ithkli Ils I- message) before decoding the corresponding IIC('I I block, and therefore wait until it has done so after having received the nem ork initial (issignment, before actually accessing the dedicated channel. The Spectlicafina) are not clear on this point. and it seems a good ‘‘ay to impro\ e the perk:nuance of call re-establishment. 6.3.10.4. Information for Mobile Stations in dedicated Mode Strangely enough, part of the information broadcast on the BCCH has no application other than after access, i.e., for mobile stations in dedicated mode. This information, given again on the SACCH. includes parameters to control the reporting o f measurements, in particular the BSIC screening information included in the PLMN PERMITTED parameter. to which we will come back in Chapter 7. I t also includes the "power control indicator" which we encountered i n the section dealing with measurements, as well as the indication whether mobile stations are obliged, forbidden or permitted to use uplink discontinuous transmission. These three indications are included in a CELLOPTIONS parameter. This information bears no real timing constraint; its absence would indeed little affect system performance. It is sent in every fourth message. 6.3.10.5. Cell Identity The final piece of information found on the BCCH is the complete CELL IDENTITY, which is sent in every fourth message. This element has no direct usage as far as the contents o f the Specifications can be analysed, although i t seems mandatory because o f general radio regulations. Beside, it may. be useful for network testing purposes. Occurruncy E d o tit 11 possible inussaUrs: s l s I I \ I I \ I I \ I I I I \ I 11 ' ! ' It ..'III .0 oft or 2. z. i l l 214s 4 I. 2. 3. 4 or 2bis I. 2. 3. 4 or 2bis 7 Table 6. ) -- Scheduling of BM I messages The tOur types of BCC!l messagesare broadcast usutg a pClit‘ti iii S occurrences. corresponding to ;1duration of about 2 second,. management. Some parameters must be managed dynamically. and their values may change as the result o f local observation. This is the case mainly for the RACH control parameters. As far as the access class is l yS. oS n .I v or by both: concerned, it can he controlled by B S oCn O this is a choice of implementation. 6.3.10.6. Message Scheduling and Contents SPECIFICATIONS REFERENCE The different items described above are grouped in four different messages for GSM900, bearing the non-informative names o f SYSTEM INFORMATION TYPE I to 4. An additional RIL3-RR SYSTEM INFORMATION TYPE 2RIS has been defined for DCSI ROO, in order to cope with the extra length o f the list o f frequencies coming from the number of available frequencies. All these messages are sent according to an K x (51 x 8) BP cycle, which includes 8 message occurrences for a duration o f about 2 5.ecclads. The scheduling is performed according to the description o f table 6.9. RIL3-RR SYSTEM INFORMATION TYPE I and 2 are sent at least once every 2 seconds, and TYPE 3 and 4 are sent at least every second. The concept of radio resource management as a specific area is introduced basically in 'I'S GS:11 04.07 and in 'I'S GSNI 04.08 (the radio interlace application protocols specification). at the start of section 3. The broadcast information i s tilled b y the BSC. Most o f the parameters pertain to system configuration, and as such are set by the OSS which indicates them to the BSC as part of the general configuration Functional descriptions can be found on some topics in the 03 series. TS GSNI 03.09 deals with the handover function (almost only from the NISC point o f view). and I S (;S NI 03.13 deals w i t h discontinuous reception. Power control. measurement reporting and handoyer preparation are described i n detail i n T S GSNI 05.08. a d v a n c e and synchronisation aspects arc dealt with in 'I'S GSM 05.10. where the different kinds of handover are described. TS GSN1 04.04 describes the contents of the 1.1-header in the SACCH messages. 430 T H E GSM SYSTEM The bulk o f the specifications i s i n fact i n t h e interface specifications, that is to say in: • T S GSM 04.08, section 3, for the RIL3-RR protocol (a part of section 4 is also relevant, since call re-establishment is dealt with in TS GSM 04.08 as part of the mobility management), section 9.5.1 for the description of the messages, and sections 10.5.1 and 10.5.2 for the coding of their information elements: • T S GSM 08.58 for the RSM protocol: • T S GSM 08.08 for the BSSMAP protocol: and • T S GSM 09.02 for the MAP/E protocol. The relevant sections are mainly 5.5 (which deals with handover), plus bits in section 5.7 (operation and maintenance). For those interested in the MAP/VLR interface (the "B" interface, not addressed in this chapter), section 5.15 (paging and search procedures) is also relevant. 432 THE GSM SYSTEM ‘101{11 .I I 1 \ \ ) 1•.(' V1211 \ I \ \ \ 1 1 \ • " ' j •!•v•• MOI3ILkTt MANAGEMEN MOBILITY AND SECCA01.1 MANAGEMENT •The radio resource mari;:perrierr Sipildlinf • H M I w 12. Q ' t a r n t y a r n ( ) a 7 lifir I Sec b dcz • •, . 433 434 T I IE. GSM SYSTEM and is concerned with getting and updating the location information needed to route calls toward a GSM subscriber who can move between cells or even networks. The second group relates t o the management o f the security features of GSM, that is to say the protective measures against fraud or eavesdropping on the radio interface. Both groups share common aspects when the implementation is looked at. They involve the same equipments, and they interact in some procedures. In both cases, the Subscriber Identity Module (SIM) and the Home Location Register (HLR) play an important role. Some little related functions are added t o these t w o groups, according to the modelling o f the Specifications. They are dealt with rapidly in the last section of this chapter. 7.1. LOCATION MANAGEMENT The mobility of the subscribers has major technical consequences in the infrastructure, but has also the important consequence that the service provided to a given subscriber changes as he moves, because of radio propagation ( h e m a y move o u t o f coverage): because h i s subscription may be limited geographically; and because he may be served by different networks, providing different services. In idle mode, the mobile station must choose one cell from which it expects to receive call attempts towards the subscriber. To this avail, the mobile station listens to the Paging and Access Grant Channel (PAGCH). It is said to be camping on this cell. The way the mobile station should choose which cell or which network it will camp on depends a lot on these service considerations. W e w i l l then present t h e mobility management functions starting with an overview. including the various factors influencing the service provided to the subscribers. Some are of administrative nature, another is propagation, and still others come from the system behaviour, such as congestion control. This will allow us to present the way the mobile station chooses (or helps the user to choose) between networks and between cells, while explaining the rationale behind these design choices. Only then will we tackle the infrastructure side, that is to say how the infrastructure keeps track of the subscribers' location. \IUIiIIJ I \ \ \ 1) x l ( l I n n \ \ \c .1 \ 11 \ I - 1 3 5 FAcroRs DETER\IINING THE SERVICE In the fixed telephone system. the service as i t appears t o a subscriber depends on which network the subscriber's telephone line is connected to. and hence on the location. For instance. the \\ay a called number is entered, the price of the communication. the additional services that may be available. a l l depend o n the location. Fixed network operators are working toss aids harmonisation. but the process is very lengthy. The consequences of these differences are however minor in the fixed system: in a system where the users move. the situation is quite different. GSM has been designed to enable an international coverage, for instance users of the European GSM900 will be offered a European-wide system area. A subscriber will I 3Cable to get full access 10 the service from many countries in Europe with a single subscription. However. in order to adapt the system to various types o f users. several levels o f service may h e offered on a geographical basis. For example. GSM operators may offer their customers different subscription choices, ranging from a (cheap) subscription limited t o part o f the country (regional subscription). to a (more expensive) subscription encompassing the whole area covered h> GSN,19(10 networks. This can go even further with SIM-roaming. I f suitable agreements exist. the subscription may extend to networks of other types. Since the SINI interface is common to all GSM-based systems. the subscriber of a GSNI900 network may obtain n r s anetwork. and reciprocally. pro‘ ided he uses an m fo ic rv e IMin s adapted mobile equipment. In order to manage this flexibility. subscription in (iSNI is defined around the leading concept of PlAIN ("Public Land mobile Network"). This concept will he developed together with the description of the other administraVve aspects. Subscription on the one hand. and coverage linthations on the other. impact the Nen ices a user has access to `.\ hen he 'no( es. A l i p + rough division distinguishes three levels: • t h e "normal service-. where a 1.1),:r can he called and can call. using all the services he has subscribed to tat Last those the serving network can pros ide): • t h e "limited service-. where the only possibilii‘ len lo the user is making einertiency ( . c a tls)1.1.111.. i n an area \\ ithin Coverage but not included in the subscription entitlement): and • t h e "no service" ease. typically when the user is completely 0111 t i coverage of any compatible network. THE (iSNI SN'STF.N1 Let us see in more detail what determines the level of service given to a subscriber. 7.1.1.1. A d m i n i s t r a t i v e Aspects "HUI I I \ \ \ I CI HI I \ \ I \ \ \ I 437 The a i r interface and the inter-MA.1N interface are the only standardisation requirements which are required to provide MS-roaming. For SIM-roaming. the standardisation of the inter-PLMN interface is still needed, but the only other requirement is a common interlace between the SIM and the mobile equipment S Nil-NIF interface). The Notion o f PLMN Subscription The development o f the Technical Specifications o f GSM was concurrent with the organisation o f a European-wide service b y the would-be GSM operators. Most o f the administrative features o f the system have been deeply influenced b y the context o f European Telecommunications, and a number of these features have left their marks in the GSM Technical Specifications. The European GSM system is divided into a number of separate operational networks, each being operated to a large extent independently from the others. Each of these networks is called a PLMN (Public Land Mobile Network—the term is much more generic than its specific usage in the Specifications). One of the restrictions, probably derived from the organisation o f CEPT, is that the commercial coverage area o f each PLMN is confined within the borders of one country. PLMNS of different countries may nevertheless overlap a little in border areas (radio waves have n o respect f o r political borders). Most countries have several PLMNs, whose coverage areas overlap partly or completely: competition between operators i s the rule o f the game. Presently, licences f o r operating GSM900 or DCS1800 in Europe have been granted to typically two or even three operators per country. The operator may be a private company, a public company or an administration. The total number of European operators holding a GSM licence is in the order of 25 in 1992. A GSM customer has a subscription relationship with a single PLMN. This specific PLMN is called the home PLMN of the subscriber. Service can he obtained from other PLNINs, depending among other conditions on subscription. I n the Specifications, the term o f visited PLMN (or VPLIMN) is sometimes used to refer to a PLN1N other than the home PLMN. In order to remove ambiguities, this term will he used in this book only when it is relevant to mention explicitly that the PLMN referred to is not the home PL\IN of the subscriber. In other occasions, i.e., when the relationship with subscription is not role ‘ ant. the term -PLMN-, or "serving PLN1N- will he used. Subscription information includes the set of services chosen by the user. as well as regional or international entitlenients. Emergency call is the only service which is available anywhere in the system. whatever the subscription conditions. In fact. this service mai even tin most PLMNS) he open to anonymous calls. i.e.. calls for which no subscriber identity is mentioned. I n this case. the S I M i s not necessary. and a mobile equipment without SIM may he used for emergency calls. Access to normal service i s o f course a different mattcrt the home PLMN. the visited PLMN and subscription entitlements all have a role iv play. P L M N Accessibility Roaming It should be noted that the grouping o f several operationally independent PLMNS in a single system open to roaming, in which the users can move and keep access to the service, is only possible i f some conditions are met. First, the PLMNs must communicate between themselves. This requires standardised means of communication between PLMNs. Second, a subscriber must have a piece of equipment enabling him to access the different networks. As explained in the first chapter, GSM is designed to support MS-roaming, where the piece of equipment is the mobile station itself (thanks t o the standardised GSM900 o r DCS1800 radio interfaces), and moreover opens the door t o SIMroaming, where the piece of equipment is the SIM only. We w i l l n o w look more i n detail a t the different conditions governing access t o a given PLNIN. taking into account roaming_ agreements and subscription limitations. To do this, \‘‘.• k‘ ill follow a subscriber called Alan. Access to the Home PLMN Depending on his subscription, Alan can access normal service in the whole area covered by his home PLMN. or only in a part of it. The last case is referred to as a regional subscription. Presently, the home PLMN is the only one in which a regional)) limited access is possible. on a subscription basis. There are no technical problems in doing otherwise. 4,58 T H E GSM SYSTEM although the way the mobile station selects cells i n the phase I Specifications takes this restriction into account. This will be changed in phase 2, and regional subscription over several PLMNs may be offered, it the commercial interest is worth the complexity o f the administrative steps. According to the Specifications, the regional limits for subscription zones are constrained by the requirement that they must not include parts of VLR areas. A VLR relates to one or several MSCs, each controlling a number of cells. The coverage areas of all these cells form the VLR area. It would have been more flexible for operators to enable the management of regional subscription on a smaller basis, and indeed the restriction would be very easily alleviated by a change of the inter-location register protocol. In this case the subscription zones could consist o f location areas (introduced in Chapter 1). It might even have been better on a cell per cell basis, because location areas should be designed so as to balance location updating traffic with paging traffic, and this traffic optimum would have been easier to reach without mixing in administrative aspects such as the boundaries o f regional subscription. However, there are no simple means to go lower than the location area level without a large increase in the technical complexity. Because the list of cells composing a subscription area constantly evolves with network extension and reconfiguration, the mobile station cannot know this list beforehand. Only Alan's home PLMN is aware o f this information. The mobile station has to learn in real time whether normal service can be provided or not in a given cell, by some enquiry means. This would be very costly in terms of signalling, unless done at the same time as location updating, and this is possible i f and only i f subscription area borders are also location area borders. This was the choice, and the location updating procedure has also the function o f verifying subscription entitlement in the location area. Access to PLMNs of the Same Country The rules for roaming in PLMNs of the same country as the home PLMN are one of the aspects of the Specifications which had continued to evolve during the elaboration of the standard. GSM900 phase 1 and DCS1800 phase I offer different views, not to mention later phases. In GSM900 phase I , all PLMNs other than the home PLMN are treated on the same basis for selection, independently from their country. Access to them may be allowed or not depending among other conditions on subscription choices, but always on the basis o f "everywhere" or "nowhere" within the PLMN coverage area. M 0 1 3 11 . 111 A N D N I A ' t R I I I \ 1 A \ A ( i l . \ I l i \ I t i t ) When DCS o00 phase 1 was standardised. there was a strong will to introduce some more controlled form of competition between operators of the same country. Operators asked f o r a mechanism b y which subscribers of other PLMNs of the same country could be tolerated in a (typically low-density) area where a single PLMN provides coverage, whereas these same subscribers could he barred in other (typically highdensity) areas where several P L M N s provide coverage. Such a mechanism would allow each operator t o install onliv part o f the infrastructure needed f o r f u l l coverage o f l o w -traffic: areas. whilst ---Foviding overall a f u l l nation-wide service f o r customers. T h i s mechanism was introduced in DCS 1800 as early as phase I. and is called "national roaming". It allows users to access parts of PLMNs in the same country, these parts being chosen by agreements between operators and not b y subscription. Relevant areas may i n fact evolve w i t h the deployment of the networks. without the subscriber being directly aware of these chinges. As f & regional subscription, national roaming is performed on a location area basis. for the same reasons. The !nubile station must learn by location updating attempts which location areas are acceptable or not. In future phases. the mechanism of national roaming will be part of both GSM900 and DCS 1800. PLMNs of Other Countries Access to PLMNs in countries other than the home PLMN country is possible for subscribers entitled to it by subscription. If so. the access is possible i n the whole P L M N area. This feature i s referred t o as "international roaming". Since new PLMNs can be created any time. the mobile station does not keep a list o f the subscribed-to PLMNs. I t will learn i f a PLMN accepts the subscriber or not by attempting location updatings. Mobile Station Constraints Because the service offered to Alan may depend on the PLMN used for access, an important feature o f the mobile station is how i t chooses, or how i t helps Alan in choosing, the serving PLMN when several PLMNs are possible. Moreover, a number of operational details may change when the serving PLMN changes. such as cost and local dialling format. Therefore. Alan needs to be aware (and. i f possible. in control of the choice) of the serving PLMN. 440 T H E GSM SYSTEM PLMN selection is one of the subjects which have been under discussion even after the freezing of phase I. There are some differences between GSM900 and DCS1800, and there are much more important differences between phase 1 and phase 2. Thewhole issue arises from the conflict between two opposite aims: quick response of the mobile station to a change in the configuration of available PLMNs on one hand, and battery life on the other hand. Any solution reflects a compromise between these two ends. Quick MS Response to PLMN Availability ... If power consumption was not at stake, the solution would be to let the mobile station explore continuously the whole GSM900 (or DCS1800) spectrum for BCCHs. It would then detect as soon as possible any new PLMN, and take it into account in its selection algorithm. ... Versus Low Power Consumption The monitoring of the radio environment for beacon frequencies is an operation that costs power consumption. Now battery life is a very important aspect o f a handheld mobile station: no subscriber would accept such a device i f refuelling was needed every second hour... Therefore a trend is to try to limit as far as possible the search for neighbouring cells done by the mobile station. The scheme adopted for GSM900 phase I is an extreme example of this method: the mobile station in normal service only monitors the cells in the same PLMN and in the neighbourhood of its serving cell. To this end, the serving cell broadcasts the list of the beacon frequencies used by neighbouring cells. This scheme is very efficient for limiting power consumption, but not for finding alternative PLMNs when the user moves into an overlapping area. The resulting behaviour is not optimum as seen by the PLMN operators: they would like the mobile station to make the right choice, in particular when the home PLMN becomes available for a mobile station who was being served by another PLMN. •In the cases where the list of frequencies used by neighbouring cells is not available, or more generally when a PLMN selection needs to take place, the mobile station has to search the whole spectrum. 7.1.1.2. Radio Considerations Administrative considerations are not, b y far, the only factor determining the service a given cell may provide to a given subscriber. A very important aspect is that the transmission between the base station 'Ruin. AND sEct RH). \ 4 4 1 and the mobile station must offer a good quality. Radio propagation considerations must then he within the criterion for the choice o f the serving cell. In idle mode, the only thing the mobile station has to do is to listen to the information broadcast by the cell it camps on (including the paging requests). If we were to take only this into account, then the best cell to choose would be simply the cell which has the best reception level or quality. But the rule in GSM is that when a mobile station wants to exchange information with the network. e.g.. to set-up a call at the user's request, or to answer a paging, it must do it in the cell it is camping on. It could have been different: one could imagine that the mobile station selects the cell to communicate with at the very last moment. Of course, nothing precludes the network to perform a handover very soon after receiving a call request. This mechanism. called "directed retry". is however far from systematic in GSM networks: in the general case. the call will stay for some time in the same cell as was selected by the mechanism in idle mode. Because of this choice, the camped-on cell should also be as close as possible to the best cell in which a potential connection will be set up. As a consequence, the quality of reception by the mobile station must not be the only parameter taken into account, but also the quality ()I' reception by the base station. This cannot be measured directly in idle mode. since the mobile station is not transmitting. but this can be derived from reception measurements and from the maximum power the mobile station can use for transmission. A criterion used t o choose a cell i n idle mode combines the reception level o f the mobile station on the beacon frequency. the maximum transmission power o f the mobile station, and several parameters depending on the cell (and broadcast on the BCCH). The exact algorithm is described later on (see the description o f the C I criterion, page 453). 7.1.1.3. System Load Control Congestion is a risk which exists in any telecommunication system. Mobility changes slightly the bases of the problem. The traffic variations are of bigger amplitude, since they come not only from the change of traffic per subscriber, but also from the movements of the subscribers. For instance a sport event may see a huge concentration o f mobile subscribers in one small place at the same moment. Another point is that, because subscribers are not physically linked to a cell, they may "move" from a cm-met:ted c e l l t o a n o t h e r i t ' t h e , : o r o n d o n e c a n o r o v i d e l h e 442 MOBILITY AND SliC1:121T1' THE GSM SYSTEM service. Another aspect to look at in connection with cell selection is then the way in which a network can control the traffic distribution among cells. Two mechanisms exist, one of which impacts cell selection. The network can bar completely a cell against access by all normal subscribers. This "barred" status is indicated in the information broadcast by the cell, in the CELL_BAR_ACCESS flag described in Chapter 6. Such a mechanism is used when a BTS is unable to operate properly, e.g., for maintenance purposes. It may also prove handy when the operator sets up new cells and performs tests on these cells before opening them for normal operation. Test mobiles (which ignore the "barred" status, and do not then conform to the Specifications) are able to establish connections with such cells for test purposes. Another potential application of cell barring concerns cells restricted to handover access. Cell barring is an all or nothing control mode, which must be taken into account by the mobile stations for cell selection in idle mode. GSM also includes a subtler mechanism, called the access class mechanism, which allows selective access of certain mobile stations to certain cells. Its purpose is to cope with abnormally high traffic load or emergency situations, but it does not influence cell selection. For example, a mobile station may perfectly select and camp on a cell which will not at this very moment accept a connection request from this particular mobile station (even for location updating purposes) because of a temporary high load. Allowing such a cell to be nevertheless selected avoids the congestion situation to spread in neighbour cells. This is an example of a choice in the specifications where a global optimum, evaluated on several cells, is favoured against the local improvement. The topic o f access class has already been developed in the appropriate place, in the Radio Resource management chapter. 4 3 when a call toward him has to he established. Cells are grouped i n location areas, and a mobile station is typically paged only in the cells of one location area when an incoming call arrives (see the basic concepts of location management, in Chapter 1). Therefore. the mobile station must inform the system of the location area in which the subscriber should be paged. It does so by a location updating procedure. The network. on the other hand, must store the present location area of each subscriber: this storage is inside location registers as will he detailed together with the description of the location updating procedure. Each change of location area puts an extra load, not only on the radio path, but also on the infrastructure equipments; t h e c e l l selection mechanism therefore includes some features to limit the number of location updatings. Location Areas For obvious technical reasons, the Specifications impose that each location area be a subset of the cells of a single PLMN. In fact. because of the way a mobile terminating call is mitten in the network. a location area must include cells managed by a single MSC (see figure 7.1). This restriction to an MSC could have been avoided. but only at the price of complex procedures. Within these constraints, the operator has complete' PLMN area MSC area M S C area location I l o c a t i o n area a r e a 7.1.1.4. P a g i n g and Location Areas Finally, an important point to take into consideration for the PLMN aiid eel' selection is that the network must be able to route calls toward the subscriber. The infrastructure must know some minimum information concerning the location of the subscriber to do so. This information can be provided only by the mobile station, and the service provided to the user depends on the consistency between the location currently assessed by the infrastructure and the cell chosen by the mobile station. It is then necessary to look in general terms at how the infrastructure deals with calls toward GSM subscribers. In order to avoid a waste of signalling, the system is so designed that a subscriber is only looked for (paged) in a few cells of the system 4 location I l o c a t i o n area a r e a I Figure 7.1 L o c a t i o n area vs. \1S(' and PLNIN area 1 A location area may only contain cell,. or a \ 1 S ( ' . in a .inch \ IN. 444 - H I E GSM SYS•1•1.:M \ H freedom to allocate cells to location areas. The goal of this operation is to minimise resource consumption, taking into account the signalling load on the radio path (both from paging and location updating) as well as the processing load of the equipments. Location Updating If a mobile station wants to obtain normal service from a cell, and in particular to receive calls, it must make sure its subscriber (represented by the SIM) is registered in the location area of this cell. The registration state of the subscriber, except in network failure cases or after some very long inactivity period, can only be changed at the initiative of the mobile station. The outcome of the last registration attempt is stored in the SIM, as well as the identity o f the location area. I f this storage indicates success, normal service is assumed automatically i f the mobile station camps on a cell of the same location area. The situation is different i f the mobile station is switched on in a different location area than the one where i t was last successfully registered; when the mobile station moves into a place where a cell from another location area is better suited; or when the mobile station tries to get normal service in a different PLMN. In all these cases, the mobile station must attempt to register the subscriber by performing a location updating procedure prior to camping on the cell in normal service. Status after Location Updating Several outcomes of location updating are possible. As we have seen, the location updating procedure does more than just tell the network where the subscriber is. It also provides the network with a mechanism to tell the mobile station whether the cell can grant normal service, taking into account subscription limitations, national roaming restrictions, and so on. The best outcome of the procedure is when the procedure is run correctly and the network indicates that normal service is possible. The subscriber is then registered correctly in the network, and the mobile station is allowed to camp on the cell in normal service mode. Several outcomes are possible when th e registration i s n o t successful. We must distinguish the cases where the procedure is run correctly (i.e., the network provides a meaningful answer, possibly denying normal service) from the others (i.e., the network does not answer or answers that it cannot give a meaningful answer). It is worth mentioning that the Specifications put all these cases together in an M I ' 1'1'1 1 ) s r LI v 1 \II xi 4 4 5 "abnormal cases" category. but there is quite d i n e r e n c e between no answer and a had answer! Meaningful negative outcomes c a n h e o f t h r e e k i n d s , corresponding to the different limitations presented in the administrative aspects section: • t h e c e l l may belong t o a P L M N n o t supported b y t he subscription. Mechanisms are specified ("forbidden PLMNs" list, see further on) so that cells of this PLAN will not be tried anew, except on explicit request from the user. A new PLMN (and hence a new cell) will he looked for if n o r a l service is sought. • t h e cell can be in a location area not suitable because of regional subscription. The rule is then that the mobile station must stay in the cell (and in the home PLNIN), but only limited service can be provided to the user. There are some reasons for this choice: because o f the implicit assumption that, when a subscriber is only entitled to regional subscription in the home PLNIN, access to other PLNINs or the sank' country is most probably not allowed by the operator: therefore looking for other PLMNs is not essential. However, this choice leads to a peculiar behaviour of the mobile station when the subscription covers a part of the home PLMN aiff/ foreign PLNINs. In such eases. the mobile station will look fur these other PLNINs only when leaving the coverage o f the home PLNI N. not when leaving the subscribed-to part of it. • i n DCSI800, the cell can be in a location area which does not accept roamers of another PLNIN of the same country. As with non-subscribed-to PLAINs, mechanisms are included to prevent further attempts i n the cells o f the same location area. I n addition, the mobile station will look immediately to sec if the home PL MN is available. An important point is that in sonic cases tit reject collies from Ihr home PLMN or from a PL\IN which is connected to the home PLAN), the subscriber is effectively ®istered in the HLR: it is marked in the network as not being in a state where calls to his destination can he established. Such calls are either directed t o an announcement, o r forwarded t o another destination i f die subscriber has !activated call forwarding unconditional o r call forwarding on not reachable. I n the other cases, the subscriber keeps the same HLR registration state he had before the attempt. However he is considered as ®istered by the mobile station, since there is no indication from the network dhow what happened with the HLR state. 44b T H E GSM SYSTENI \ H M I 7.1.2. CELL AND PLMN SELECTION The cell and P L M N selection mechanisms are almost totally specified in the Specifications. We have seen in the previous section the different factors t o take i n t o account, a s w e l l a s some general mechanisms. Let us see now the detailed process. 7.1.2.1. P L M N Selection Though the PLMN selection process (as well as cell selection) affects only the mobile station, it is specified in a fair amount of detail in the Specifications. One of the reason is that the SIM intervenes. and the SIM-ME interface has to be specified. Other reasons are to harmonise somewhat the behaviour o f the mobile station, s o that i t can b e predictable when a user changes his equipment, and also to avoid the possibility that some implementations bias the choice of PLMN and thus the competition between operators. However, the Specifications are not totally constraining, and have to b e completed f o r some marginal cases b y th e mobile station manufacturers. W e w i l l mention these latitudes, w i t h some o f the possibilities. What the mobile station really has to do is to select a cell. II\ I ) ( R I I \ I V\ . 1 \ ! 4 4 7 There is however an important difference between the selection o f the serving PLNIN. and the selection of the see ing cell. The first is under control of the user. whereas the second is fully automatic. The PLMN selection is important for the user when the service finall . obtained is normal service. I n the limited service case. the choice i s o f little importance. and in the case where no PLNIN can be found the selection of a PLNIN is o f no immediate application. W e will look at the three eases in turn, before summarising the user's view. When the mobile station receives one o f the negative answers listed above, it may look for other PLMNs. I f none of the found PLMNs are able to provide normal service, the mobile station goes to the limited service state. Cases of procedural failure are even more complex. Details can be found in the Specifications and will not be dealt with here. Basically. the mobile station w i l l not t r y other cells (unless the radio criteria so determine) despite the lack of knowledge about the service that can be provided to the user: this is an important point. There is a first phase during which the mobile station will try several times to get an answer from the network. During this phase, the mobile station will behave as if it was granted the same service as before. I f all attempts fail during this phase, the mobile station enters a special state, in which i t does not assume any registration state in the network. I t tries now and then a location updating procedure, to get out o f this unsatisfying state. The mobile station does not reject a request from the user, but this request triggers the mobile station to start a location updating. and the mobile station will eventually reject the user's request i f the network still does not answer positively. ! • T h e Normal Case If already in normal service mode the mobile station t i l ) looks for the cells in the serving PLNIN. independently from other PLNINs. A change of PLNIN can occur at only two occasions: when the user decides so. and when the mobile station finds out that the serving PLNIN can no longer provide normal service (e.g.. because the mobile station is leaving the PLMN co \ Clage area). In those cases. the mobile station Mill search for cells in thek1/4hole spectrum. to find ‘‘lbch PI.NINs cover the location. Access to some of these PLAINs is then tried. according to the PLAIN selection method. 'Iwo methods are provided for the choice of the PLNIN to try. the choice being left t o the user: the manual mode and the automatic mode. Several aspects are common to both modes. For example. the home PLMN is set as the PLNIN t o try at switch-on. independently front previous history (even if the mobile station kno‘t,S.11.0111 the SIM. that the subscriber is currently registered in another PLMN). Another common aspect between the manual and automatic modes i s the f i n -bidden" PLI4Ns list—the first one in a series of 3 l'1.MN lists to juggle with! The actual list o f PLMNs accessible t o a subscriber according t o h i s subscription may change as nok PLNINs are opened tor closed!) f o r service. The list of these PLNINs cannot he stored once and for all (for instance in the SIM). Instead. it has been chosen to build tl nautically a list of PLNINs which are emit accessible according to subscription. This list is updated according to the result of access attempts performed by the mobile station. and is stored in a non-volatile memory in the SIM (and hence is not lost when the mobile station is switched MTh The list is limited by the Specifications to -I entries. but nothing precludes a longer list to be stored in the ME (therefore possibly lost at s‘‘itch-orli. This so-called /in:bidden P/MNIN list includes PI.NINs which are not subscribed to. They are nut really "forbidden-. since they could be used for emergency calls. and i n manual mode the user is perfectly authorised to select one o f them. When a mobile station attempts to update its location in a Pl.\1N to which the user's subsotiption does not 448 THE MOBILITY AND SECURITY MANAGEMEST GSM SYSTEM authorise access, the network will tell the mobile station so, and the PLMN identity will be put in the list, ejecting fi needed the oldest entry. The forbidden PLMNs list is used for PLMN selection both in manual and automatic mode, though in different ways. Another list, this time stored in the mobile equipment (not in the Since. at switch-on time. 440 the mobile first looks for the home P L M . a n d d o e s not (in m a n u a l m o d e ) t a k e i n t o a c c o u n t the P L . M N the s u b s c r i b e r is registered in, s o m e situations c a n b e c o m e quite i r k s o m e when the user stavs in a foreign country for some time. Automatic mode is in such cases more appropriate. SIM) intervenes a little in the PLMN selection process in automatic Once the user has selected a PLMN. the mobile station will attempt mode. It contains the identities o f the location areas which have rejected access before because of national roaming limitations. Cells belonging to the location areas in this list can no longer be candidates for selection. and a PLMN for which all the found cells are in this category cannot be to get normal service on this PLMN. Several outcomes are possible, as tried in automatic mode. This list is filled as a result of location updating r e j e c t i o n s w i t h a suitable c a u s e (but the P L M N is not put in the f o r b i d d e n PLMN list), and is erased when the mobile equipment is switched off, or the SIM removed. we have seen. It may well succeed: fine! The P l A N mar also be indicated as not allowed by subscription. and will therefore join the forbidden PLMNs list. Unfortunately. the other cases are not so simple. and not always clearly specified. In manual mode. these cases can be treated generally by letting the mobile station do its best to get normal service (possibly by trying several cells in different areas). After some time, the PLMN list is presented again. including the PLMN on which the a t t e m o t f a i l e d : it is u p t o t h e u s e r t o c h o o s e a n o t h e r one... If e v e r y t h i n g fails. t h e m o b i l e s t a t i o n s h o u l d at s o m e s t a v e r e a c h the "limited service" Manual Mode d e s c r i b e d b e l o w. In manual mode the list of PLMNs the mobile station has found as Automatic Mode potential candidates for providing normal service is presented to the user, whether or not they are in the forbidden PLMNs list. This list of found PLMNs is displayed using explicit names (such as D1-Telekom or D2 In automatic mode. the mobile station will choose which PLMNs Privat for the German PLMNs, DK TDK-Mobil or DK Sonofon for the to try all by Danish PLMNs. another list of PLMNs. the preferred PLMNs list. which is stored in a etc.) and with an explicit mention telling the user itself. The automatic mode is based on the caistence of whether the PLMN is in the forbidden PLMNs list or not. Normally the user then chooses one of the PLMNs of the found PLMNs list as the non-volatile memory in the SIM. This list. capable of holding at least 8 entries. includes a number of PI.MN identities in order of preference and s e l e c t e d PLMN. si under the control of the user. The most preferred is usually the home The explicit names of the PIMN cannot be found in the Specifications, though they are part of the mobile station specifications. The list is distributed by the GSM MoU, and regularly updated. There is no requirement on the undate of already manufactured (and sold) mobile stations, or on the way in which they should visualise the names of new PLMNs who have been introduced meanwhile. Typically. if the knowledge of the network commercial name is unknown, the mobile -Station will display the country initial together with the numerical network code, for instance "DK Network 05" The user can choose any PLMN in the found PLMNs list, even if it is also part of the forbidden PLMNs list. This possibility might be useful after a change of subscription category: for example, if Alan requests his subscription to be changed from "home PLMN only" to "all PLMNs" while he roams abroad, he might want to force his mobile station to attempt location updating on a network previously known as "forbidden" in order to unlock the situation. PLMN. but it is nowhere specified whether the home PLMN should appear at first rank in the list or not. whether the user may choose to have i this ni list with a different PLMN as "top of the list" and what happens n case. The list may originally be filled in by the home PLMN operator. during the SIM personalisation process. al can afterwards be modified at will by the user through a mechanism to be specitied by the mobile station manufacturer. No automatic moditication of the list can take place. When a PLMN selection takes place in automatic PLMNs are tried starting with the first PLMN in the mode. the list o fwreferred PLMNs which is in the list of found PLMNs and not in the forbidden PLMNs list. The treatment of one of the failure cases is clean. fi the PLMN is indicated as not allowed by subscription , its identity is put in the forbidden PLMNs list and then cannot be automatically selected again, except in the (rare) case where more than 4 PLMNs are found. none being allowed by subscription: in that case. the mobile station may rewrite the forbidden PLMNs list n i a cyclic manner! nI hte other failing 450 T u g cism sysTEm cases, the mobile station must try the other possible PLMNs, in order of preference for those in the preferred PLMNs list. It is not clear what is the status o f PLMNs found by the mobile station but not included in the preferred PLMNs list; the most literal interpretation is that they cannot be selected. This raises different problems, for instance in the case where the list is empty. It seems more logical for the mobile station to also include them in the choice process, according to some ordering rule, for instance randomly. The situation where none o f the found PLMNs can provide normal service is not clearly specified in the Specifications either, but it seems logical that the mobile station should go to the limited service state in that case. The mobile station should keep some record of its attempts to avoid a deadly loop in which it will stubbornly attempt to get service from a PLMN without success when another would grant normal service. As in manual mode, the user can force PLMN selection at any moment. The process is then the same as described above when the serving PLMN ceases to provide normal service, except that the serving PLMN is in the list of found PLMNs. The mobile station will then stay as the selected PLMN, except if a PLMN with higher preference rating has appeared since the last PLMN selection. \ 1( t \ \ 1 ) sit a itri \INN \ii1.\11 \ I 4 5 1 location area is not forbidden either). I n manual mode, one behaviour would be to ask the user every time a PLMN pops up: this could however lead to annoying situations in border areas where PLNINs may appear and disappear very frequently. Another approach for the manual mode could be to stay in limited service mode as long as sonic PLMN chosen by the user stays unavailable, whatever other PLMNs pop up. which is not very much moreoptisfactory. It is hoped that mobile station manufacturers will find a satisfying compromise. When the home PLMN is available, the limited service mode can happen only when the user has a regional subscription and is in a part of the home PLMN area where he is not entitled to normal service. In this case, the Specifications impose a cell selection mechanism identical to the ----normal service state, i.e.. limited to the cells of the PLMN. The mobile station tries each new location area of the home PLMN that pops up so as to find one granting normal service, but will go on ano1/41;er PLMN (on which the subscriber may be entitled to service) only wirer leaving the home PLMN coverage area, or when asked by the user. The mobile station does not store any list of the location areas in which the user is or is not entitled to normal service. Protection against trying again and again the same location area is obtained by the hysteresis mechanism for the change of location area. as in normal service mode (see further on). The Limited Service Case In limited service mode, the only service available is emergency call. The theory is that all PLMNs can provide this service. The purpose of selecting a PLMN is then fairly limited. It could however prove useful for two main reasons: in border areas, the user might want to control the choice of the country where a potential emergency call would be routed; besides, the user could choose a PLMN not for immediate use, but to be the one used for normal service as soon as the mobile station can find a cell of this PLMN. The Specifications specify that the choice o f the cell in limited service mode is done generally independently from the PLMN or the location area. This general rule admits a noticeable exception. when the home PLMN is available for limited service, and no other PLMN is available for normal service provision. In parallel, the mobile station monitors continuously (though possibly at a slow rate to save its battery) the 30 strongest carriers i t receives for new PLMNs. The phase 1 Specifications leave open the behaviour of the mobile station, especially in manual mode, when a new PLMN is found. In automatic mode, the mobile station will perform a PLMN selection and try to get normal service on a newly found PLMN which is not i n the forbidden PLMNs list (and f o r DCS 1800 i f the The No Service Case If no cell can he found at all, the 124 carriers in the hand (374 for DCS1800) must be monitored, not just the 30 strongest carriers found as in the other cases. Otherwise. the mobile station behaves as in the Ihnited service case. except that it cannot accept even emergency calls. What the Users see The mobile station must indicate the service state to the user. Different levels o f precision seem allowed. Typically ihe information provided b y the mobile station t o the user distinguishes the normal service, the limited service and the no service cases. We will see further along that in cases of signalling failure an additional state exists, but it is not necessarily indicated to the user. In addition, the mobile station must be able to indicate to the user. possibly on request. the serving PLMN when in normal service state. The pLNIN identity is given in clear text when possible. including country initials and network name. For the users. PLNIN selection is first a choice between manual mode and automatic mode. Some control must he provided to allow the 452 THE (iSN1 SYSTIEN1 user to know whether the mode is set to automatic or manual, and to change the mode. In automatic mode, the user usually does not intervene i n the selection mechanism. He is however able to direct the process by two actions: he controls the preferred PLMNs list and he can ask for a forced PLMN selection at any time, in which case the mobile station will search the whole spectrum and select a PLMN, possibly the same as before. In order to enable the user to control the preferred PLMNs list, the mobile station must provide commands to display the list and to edit it. Whether a modification of this list, is taken immediately into account (e.g., by a forced P L M N selection) o r n o t i s l e f t open t o mobile station manufacturers: In manual mode, the user has a total control, but is solicited in each case, maybe too often in some situations. He is asked for a PLMN choice in several instances. This happens after switch-on (or SIM activation) i f the home PLMN is not available. This happens also when the mobile station moves out of coverage of the serving PLMN. In addition, this may happen in DCS1800 when the mobile station was in a visited PLMN of the home country, and cannot find any more a cell that accepts it. In this case, a forced PLMN selection happens, and hence the user will be asked to choose a new PLMN. Finally, a prompt to select a PLMN may appear when not in normal service mode, and a new PLMN is found. A s in automatic mode, the user can in addition force a PLMN selection at any time. Whatever the triggering event, the list o f all found PLMNs is presented to the user, including those in the forbidden PLMNs list. though with a distinguishing mark. The order of presentation was a topic raising some discussion between operators, since it was felt that this order may influence the user's choice. The issue was settled on the specification that the order should be random. 7.1.2.2. C e l l Selection As explained in the requirements, only cells can be chosen with which transmission will be a priori of at least minimum performance, and the cell choice should aim at maximising the transmission quality. The radio criteria therefore play the foremost part in cell selection. Thus. before describing the actual cell selection algorithm, we will first study them. \1011 I I l l \ \ C l I:111 \ I \ \ \(.1 \ I \ I 453 Radio Criteria In order to maximise transmission quality. a criterion has been defined, which takes into account the level of the signal received by the mobile station on the beacon frequency. the maximum transmission power of the mobile station and some parameters specific to the cell. This criterion is named Cl. (There is no C2 in the phase I Sperifinitions, but it will appear in phase 2.) Description of the CI Criterion CI is defined as follows: Cl := (A —Nlax.(13,0fi A := Received Level Average — pl 13:= p2 — Maximum RF power of the mobile station pre,...t1 in.11)' The I w o parameters / 1 / a n d p 2 --called respectively RATELAccEssmtv and .11.V TVPII IC:VAX...CCU in the Specified/ions—are broadcast by the cell. The first one can take a value between --I In (11311, and —48 ( i m . the second hem een 13 Mint and 43 dlim in (;S\1`00 tin DCS1800. the range is different). The second parameter has another independent usage: i t represents the maximum transmission power a mobile station is allowed to use on the RACI I. Because the range of the mobile station maximum transmission power is 29 to 43 i.113m. only this sub-range o f the second parameter is useful. whether for C / o r as a maximum transmission pox\ er on the R:\C•fl. CI is used as follows. When looking for cells. either when looking for neighbour cells in normal service mode. or when searching PLNINs. only cells o f positive C I (calculated from the p l and p2 broadcast by each cell) are taken into account. When a choice between cells has to be made, the cell o f best C I is chosen among those Nub. alem for other criteria. As a consequence. ('1 determines Iwo things: • t h e coverage limit o f each cell taken in isolation, in the sense that outside the area where CI is positive, the cell does not exist for the mobile stations: • t h e boundary between two adjacent cells for selection in idle mode, determined as the locus where ('I=C'I T h e boundaries 454 B A B A Figure 7.3 - Impact of cEll area in which cell A can be the serving one * a r e a in which cell B can b e the serving o n e The boundary between two cellsA andB f o r a m o b i l e s t a t i o n to p e r t o r m l o c a A a l A a n d s b e l o n g to different location afeds. and the m o b i l e station is registered in the (b) A and B a r e in the same location a r e a Figure 7.2 - Cell boundaries according to C/ (c) A and B belong to different location areas. Thefigure shows the boundariesof two cells Aand B, and the mobile station is registered ni the location area of A according to hte values CIA and CIg of the CI cell selection criterion. The dashed line is the locus where CIA = C/B Since C/ depends on the mobile station maximum power, these boundaries This is acceptable if the cells are equivalent. differ from one mobile station class to another. with all adjacent cells determine a second cell limit, usually inside the area delimited by C/=0. Figure 7.2 shows an example of two cells. with their C1=0 limits, and the line of equal C / ' s . Two points are important to keep in mind with these limits. Because the maximum transmission power of the mobile station intervenes in C1. the limits are different for different mobile * A c o n s e q u e n c e of this s p e c i fi c a t i o n is that the station classes. Second, that other cell limits exist. the ones determined by the selection of the cell for handover. It is up to the operator to choose p/ and »? to obtain the correct compromise between cells boundaries, traffic adiacent cells belonging to different and quality of transmission for the different classes of mobile stations, as well as consistency with the handover algorithms and parameters. of propagation into a geographical hosteresis. location boundaries b e t c e areas is not the whether the mobile station goes from one cell to the other or the contrary The hysteresis in terms of received level is t r a n s f o r m e d be the variations Taking the same cell configuration example as before. figure 3.7 shous the cell boundaries is obtained when such handicaps are used. We have three different b o u n d a r i e s for a g i v e n m o b i l e s t a t i o n class: o n e for m o b i l e s t a t i o n s going Criteria O t h e r than CI Because of its radio-electric nature. C / varies quickly and to some extent randomly around a mean value depending on the location a n don the m o v e m e n t o f the m o b i l e station. but not otherwise. for instance between two cells belonging to different location areas. This is why the comparison between CT's is modified to be biased against the -gells on which camping on must be preceded by a location updating. 'This is obtained by a handicap. added to the CT of these cells. The value of this handicap is not fixed. and is broadcast by each cell. It is called CELL RESELECT MYSTERESTs. though it is not strictly speaking a hesteresis value (the real hysteresis si the sum of the values in the m o cells). This m e a n s that the m o b i l e station would often change between cells if C / were the only selection criterion. from cell A to cell B. one for mobile stations going from cell B to cell 1. and a third for mobile stations which do not a p p t the bias de.g. @reigners not entitled n i the PLMS by subscriptions. Mobile stations located in the area between lines ad and cocan be attached ot either of the cells, depending on the directioß they come from 456 p Ili GSM SYSTEM N10131E11.1 . \ \ I ) SECt RITY NI. \ N.V.& \ IFNI The Cell Selection Algorithm The different requirements the cell selection algorithm has to meet have been detailed in the previous paragraphs. This algorithm is specified in the Specifications, and we will present it, as a summary of the points seen so far. The aim of cell selection can be summarised as follows: in order to get normal service, the mobile station must camp on one o f the cells fulfilling the following conditions: • a S I M must be inserted, and the corresponding subscriber registered in the location area the cell belongs to; • criterion C/ for the cell must be higher than 0; • t h e cell must not be barred. Among the cells which comply to these three requirements, the chosen cell must fulfil the two additional conditions: • t h e cell's C/ must be higher than the C/ of any other cell found by the mobile station in the same location area; • t h e cell's C/ must be higher than the C/ of any other cell found by the mobile station i n different location areas o f the same PLMN, corrected by the applicable handicap factor. As can be noted, potentially better cells in PLMNs other than the one the mobile station is registered in are not taken into account. This corresponds to the decision that PLMN selection is triggered only by the user, or when the mobile station leaves the coverage o f the selected PLMN. Cell Selection in Normal Service State Having this goal in mind. the behaviour of the mobile station can be easily derived. Let us start with the mobile station being in normal service state (hopefully the likeliest state). In this state, the mobile station receives a list o f frequencies (broadcast by the serving cell) indicating where to look for the beacon channels of the neighbouring cells of the same PLMN. The mobile station must then find these beacon channels one by one and get their synchronisation information in order to decode some o f the broadcast information they carry; this information enables the mobile station to check the PLMN, to know if the cell is barred or not and to obtain the identity of the location area the cell belongs to, as well as t o get the various radio parameters so as t o compute C l . Only i 4 5 7 acceptable cells (i.e.. non-barred and of positive ('/1 of the right PLMN will be taken into account. The mobile station can then compare the CI of these cells with the CI of the current cell. All this process takes place in parallel with the periodic reception of the paging channel on the current cell. If the mobile station finds a better cell in the same location area, it changes to this cell and goes on with the process o f listening to the paging channel (of the new cell) while monitoring the beacon channels in the new list. However. i f the mobile station finds a better cell i n a different location area o f the same PLMN. having taken the location updating bias into account. it changes to this cell. At this very moment, the mobile station is no more—strictly speaking—in the normal service state: calls to it will in general not reach :their destination. The mobile station tries immediately a location updating procedure t o warn the network o f itsanew location. Most o f the time, the mobile station is granted normarservice by the network. and is hack to the normal service state in the new location area after a few seconds. Cell Selection at Switch-on Thne Let u s n o w describe other cases. A n important one i s the initialisation: how to get normal service in the first place alter switch-on. The first PLMN to try ( i f found) is the home PLMN. according to the Specifications. The mobile station must search for non-barred and C I positive cells within this PLMN. Without any information. the mobile station has to search the whole spectrum for beacon channels. This can he a lengthy operation. in particular in DCS IMO where 374 frequencies are supported. The Specifications include a mechanism to help the mobile station in such conditions: a list of frequencies to look for can he stored in a nonvolatile memory (in the SIM). The Specifications are however not clear about which frequencies the mobile station should put in that list—Ise-he consistent with the imposed search for the home PL.NIN at switch-on time, it should be the neighbouring cell frequencies broadcast by the last cell of the home PLMN on which the mobile station camped ( whether for normal service o r not, and whether o r not the mobile station was registered on other PLMNs meanwhile). Whatever the case. the search results in a list of acceptable cells, with their C/. I f this list is not empty. the mobile station chooses the cell of best Cl. Also stored in a non-volatile memory is the identity of the location area (if any) i n which the mobile station knows it is registered. If the chosen cell belongs to that particular location area. the mobile 458 TIIE GSM SYSTEM presence to the network (see the description of the "INISI attach/detach" mechanism, page 474). Otherwise, i f the chosen cell is in a different location area, the mobile station camps on the cell and starts a location updating procedure immediately. If no acceptable cell o f the home PLMN is found, the mobile station acts as if it were leaving the home PLMN coverage area. Cell Selection at PLMN Change Another special case is when the mobile station has to look for the available PLMNs, for instance because the mobile station has moved out of the coverage of the PLMN previously providing service, or because of a forced PLMN selection by the user. The process is similar to the switch-on case, except that the mobile station has n o information whatsoever about which frequencies to search: it must search the whole spectrum. The mobile station proceeds in two steps: first, it searches all GSM (or DCS 1800) carriers, then i t selects the 30 strongest ones to obtain the information they broadcast: the PLMN to which they belong, whether they are barred or not for access, and the parameters controlling Cl. The mobile station can then establish the list o f the acceptable PLMNs, that is to say those for which it has found at least one acceptable cell. This will result i n a found PLAIN list as described some pages before. When one of the PLMNs in the list is chosen, the mobile station will access the acceptable cell of best CI within those previously found in this PLMN and request a location updating. Cell Selection in Limited Service Mode Now we have to address the cases where normal service cannot be granted, but limited service is possible. This happens when the subscriber is not entitled to normal service in any of the found PLMNs. I f the home PLMN i s acceptable, i.e., i f access to the home PLMN i s locally -prevented by subscription rather than radio propagation, the cell selection is the same as for the normal service state. Other Re. the mobile station selects the acceptable cell o f best C l , irrespective o f the PLMN or the location area o f the cells, and hence without applying the location updating bias. The mobile station searches continuously the whole spectrum for new cells, in order to find an acceptable PLMN as soon as possible. When such a PLMN is found, it may be selected and then the mobile station will try to get normal service on this PLMN. \ E m i l 111 \ \ I ) l < 1 1 1 \ l \ \ Milt111 4 5 i ) 7.1.3. ARCHITECTI Selecting a PLMN and a cell is but one side of the management of mobility. The goal o f the network as far as location management is concerned is to prepare for the routing of calls toward the subscribers. taking account o f their movements. To that end. the network must memorise for each subscriber (very precisely for each SIM) whether he is known to be in 501Ele phlie or not (he is said to be registered). and if so. in which location area. This information is retrieved when a call toward the user must he set up. as will be explained in Chapter 8. Because a location area is mandatorily in GSM k‘ holly included in a single MSC area. the stored information is sufficient for routing the call up to the MSC which will he in charge o f the communication. Furthermore, the knowledge o f the precise location area (there may be several location areas within a n M S C area a l l o w s t o restrict the paging t o the corresponding cells. A simple solution to the basic location management issue could consist in storing in a database the identity of each subscriber together with an indication on whether or not he is registered. and if so. where to find him. Indeed. the canonical architecture o f GSM identifies such a database. the Home Location Register i IILR ). This function is separated from the routing function itself, which consists in choosing and reserving circuits to obtain a continuous connection between users that desire to be in communication. In CiSNI. the main actor on the mobile user side for routing and communication management functions is the Mobile services Switching Centre (the MSC). But the canonical architecture is a bit more complex. and before describing it. i t is interesting to look at t o the reasons why. Every telecommunication system includes a database containing a variety o f information concerning each subscriber. such a s t h e subscription limitations. the services subscribed Mr. the states o f the supplementary service : k i n ation. o r the throrthation needed f u r the management of the charging inlbrmation. In GSM. the same information exists. plus some which is specific such as the information related to the confidentiality. functions. In a fixed network. each subscriber is connected t o one local switch. for a long time. Every call involving this subscriber. whether an originating or a terminating call, goes through this switch. This is then the natural place for the storage of the subscriber related information. In a system dealing with moving subscribers, there is no such natural place for the storage of subscriber parameters. However, the two kinds of data to 460 T H E GSN1 SN'STENI N U M M I ! ) \ N I ) Slitl.1611 \ I \ \ \GI, \II \ 1 4 6 1 be stored (location information and subscriber data) call for a common storage solution. This is the choice made in GSM. and the HLR is the database for both sets of information. location information in the networClilor instance. it has to tell the old MSC/VLR to erase a subscriber record ‘‘. hen this subscriber is registered under a new MSC/VLR. If location information is needed only for the establishment o f mobile terminating calls, the rest of the information is needed at various moments during any call. Basically, i t is the visited MSC, the one in charge of a mobile subscriber engaged in a call, which needs these pieces of information. Then an important signalling load would result i f the MSC had to interrogate the H L R each time i t needs some piece o f information. To avoid this signalling load, the data record of a subscriber is copied i n a database close t o the MSC while this subscriber is registered in a location area controlled by the MSC. The database, the VLR (Visitor Location Register), will be ignored for the moment as an entity separate from the MSC: we will speak o f an MSC/VLR. The distinction will be dealt with later, and our approach explained. Functionally, we could say that there is a unique HLR function system-wise, possibly distributed on several equipments. In practice, he it only for operation reasons, one HLR function is implemented in each PLMN. There again, the HLR function for a PLMN can be implemented in a single equipment o r distributed among several equipments. Both approaches are allowed, and used. Note that the usage of the term HLR may refer to the function. possibly encompassing several equipments, or to a single equipment. In most o f the cases, this ambiguity causes no understanding problem. This capacity of temporary storage in the MSC/VLR allows some distribution o f the function o f location information storage. The HLR needs only store information concerning the MSC/VLR in which area where the subscriber currently is. The identity of the precise location area is stored i n the MSCNLR, together with a copy o f the remaining subscriber related information. This i s sufficient t o get the routing information needed to deal with an incoming call, and this is all the HLR needs. This temporary storage in the MSC/VLR introduces new functions. The subscriber information has to be copied when the subscriber enters a new MSC/VLR area. Conversely, the corresponding record has to be erased in the previous MSC/VLR in which the subscriber was registered. Some mechanisms are needed for maintaining the consistency between what is stored in the HLR and what is stored in the MSC/VLRs, including the case of a failure resulting in a loss of stored information. 7.1.3.1. F u n c t i o n s As defined in the Specifications. the HLR is basically an intelligent database used t o store the location information and th e subscriber related information needed for providing the telecommunication services. The HLR has no switching capability. It is connected to the other entities o f the network and switching subsystem (NSS) through signalling means. as discussed in Chapter 2. The HLR is not a simple database which can accept only "store" or "retrieve" orders. In fact, the HLR completely manages the LEL The H L R has many different roles. What concerns us in this section is the management o f the subscriber mobility. Functions of the HLR related to. e.g., the management of the confidentiality data, or the management o f t h e supplementary services shall h e described respectively further on in this chapter, and in Chapter S. The Visitor Location Register While w e have introduced the term V L R . the corresponding concept has been somewhat masked. Nlost readers will have also noted that the VLR was not even allocated an icon. We have to answer for this rather off-hand treatment. In the canonical GSM architecture. what has been referred up to now as the MSC/VLR consists o f two disjoint functional entities, the MSC' itself and a database. the VI,R. The MSC is defined as the switching function in charge of the management of the calls. and the VLR as the database where subscriber information are temporarily stored for those subscribers which are registered under a MSC connected to the \TR. A VLR can manage the subscriber data for one or se\ end NIS('.. and can be an equipment ph)sically distinct from a \1S('. The reason why the two functions arc split is not sin much because of this possible implementation in distinct equipments. but because of the option to have a VLR for more than one \ISC. The point to analyse is then why should such a choice he taken. From an architectural point of three different vantage point,: the V L R c a n 1-1•.: ! , : l n h o r n • a f i r s t approach is to consider the set of the HLR and VLR as a 4 s i n g l e distributed database, The distinction between the 1I.1< 462 T I NU/1111.111 \ \ U 51.(Il k I I 1 \ I \ N \ (INNIEN I III (ISM s Y s Tr i o the interrogation o f a centralised I -11.R are the first bricks f o r the construction of a full-blown intelligent network structure. parts and t h e V L R parts becomes a matter o f internal architecture o f this database. This would correspond t o an approach where the VLR is introduced solely for signalling load distribution, and a V L R would naturally be serving several MSCs; • t h e opposite point of view is to consider the VLR as fulfilling a set of ancillary tasks to the MSC, including the management of the visited subscriber database and the corresponding dialogues with the HLR. It is then naturally a part of the MSC, and there are little reason for having a VLR connected to several MSCs; This situation is the main reason why the authors o f this hook decided not to describe the VLR as an entity separated from the MSC. This position is supported by the fact that up to now all the switch manufacturers have chosen to develop a combined NISC/VLR, and none offer the possibility to physically split them up. Let us see rapidly some of the points of the Specifications among the casualties caused by our approach. I f separated. NISC and VLR are connected through the SS7 signalling network. and the signalling procedures for the corresponding dialogue are specified in the MAP (they constitute the MAP/B protocol). The MAP/B protocol represents a major part of the MAP specilieation. at least as far as the number of pages is concerned. Because of the reasons explained above, it will not be treated here. • a third approach is to consider the VLR as a truly independent entity, having tasks of its own, with added value compared to the natural roles of the HLR and the MSC. The exact philosophy of the functional split between the VLR and the HLR on one side, and between the MSC and the VLR on the other, can be determined by looking in details of the corresponding protocols. In so doing, it becomes apparent that the cut between the MSC/VLR and the HLR can be considered as a minimum, whereas the cut between the VLR/HLR and the MSC seems not to follow strongly directive lines. I f the VLR/HLR protocol can be easily presented as an MSC/HLR protocol, it is obviously impossible to do so with the MSC/VLR protocol (the proof of the first statement will be found in this book, whereas it would be too long to justify the second: the interested reader can look to the Specifications to make up his mind). This militates strongly f o r the second approach, that we have followed in this book. It is often invoked as a counter argument that the split between the VLR and the MSC is related t o the Intelligent Network (INI) approach f o r building switch equipments. When looked at closer, though, it is obvious that the two philosophies are somewhat different: in an IN approach, the MSC would have no high level function and the VLR would be where all the complex _protocols are dealt with. This is very different in the .Specifications: the MSC deals indeed with most o f the complex protocols. The V L R is neither a pure database (in which case, the VLR-MSC interface would be very close to an interface between a MSC and a HLR i f temporary storage was not used), nor the true call manager controlling a rather dumb switch (this would be the I N approach). Implementations o f a GSM network based o n a n I N approach are nevertheless possible and implemented, but the split between the "durhb" Switching Service Point and the "intelligent" Service Control Point is not based along a MSCVLR line. While the architectural split as described in the Specifications is for these reasons certainly not the last say in the domain, GSM is considered as one of the first "intelligent" networks and concepts such as 463 The Mobile Station The mobile station holds a starring role i n location managiement, for obvious reasons. It is at the origin of the location information. a n d i n f a c t deals w i t h problems posed by mobility almost entirely on its own in idle mode. An important point which may have been less evident up to now is the respective role of the SIM and o f the rest o f the mobile station, known in the Specifications as the mobile equipment. , 0 The assumption in the .S'pecificatins is that the mohile equipment does not hold in a non-s °tattle memory any information specific to its user. Still, mobile station manufacturers are not prevented from doing so, hut the system can work properly without such a mentor\ . The converse point is that the SINI hold; this information. and i n particular firrii7 related to the mobility management. whether for handling location or security related information. I The mobile equipment contains however some information o f some temporal scope. such us the list o f forbidden location areas for national roaming. or lists o f beacon frequencies for different PLMNs. This information is lost when the mobile equipment is switched off. The choice o f what is i n the SIM was a compromise between memory consumption (a scarce resource for a SIM) and keeping as much as possible potentially useful information over a switched-off period. MOBILITY AND SECURITY MANAGEMENT 4 6 5 Tin' SIM 1 GSF:TI RIL3-MM The SIM has already been presented in the very first chapters. What is of interest here is the information it contains in relationship with location management. In this area, t h e S I M i s b u t a passive information container. so listing the relevant fields will give us all the functional description we need: • T h e update status: • A location area identity: These fields usually contain the result of the last location updating attempt. and the location area where it was done: their main purpose is to avoid a location initialing attempt in sonic cases when the cell selected alter switch-on i n the smile location area. • A list of beacon frequencies ("BCCH information''): We have already met this list. The Specifications are not clear about which PLAIN it relates to. In order for this list to achieve its aim. i.e.. 10 speed up the initialisation time after switch-on. it should pertain to die home PLMN. Another interpretation is that it should be the last list received from the serving network. • TheThrbiddeti PLMNs list ("forbidden PLMNs"); The function of this list was described in the section of this chapter relative to PLMN selection. It should he noted that it is an ordered list. with entries sorted by order of introduction, because of the requirement to replace the oldest entry when the list is full. • T h e prdiTred PLMN list. MAP/D H L R r - - / MSC VLR [ SIM-ME • GSM' Figure 7.4 — Location Management Protocols MM protocols involve the location registers (VLR and HLR). which communicate between themselves and with the mobile station, in which the SIM plays a prominent part. the subscribers, mainly location information with the aim of setting up calls toward these subscribers. This will be dealt with in Chapter 8. The protocol between the HLR and the MSC/VLRs is 'supported through the world-wide signalling network, the signalling system n°7 (SS7), as described i n Chapter 5. The application protocol f o r the dialogues between a HLR and a .MSC/VLR is part of the MAP (Mobile Application Part). In this book, we will call it the MAP/D protocol. The function o f this list was also described whch dealing with PLMN selection. The MS-MSC protocol is called the RIL3-MM protocol (for Radio Interface Layer 3, Mobility Management). It uses the MS-MSC signalling connection provided by the RR layer, as seen in Chapters 5 and 6. 7.1.3.2. P r o t o c o l s The SIM-ME protocol is limited to read and write commands as far as we are concerned here. and corresponding messages will not be cited in the text. The protocols belonging functionally to the Mobility Management •plane are the one held between the H L R and the MSC/VLRs, the MSC/VLR and the mobile station, and between the mobile equipment and the SIM. To allow full roaming. it is of utmost importance that every I I l k be able to exchange information with every MSC/VLR throughout all the PLMNs o f the system (and possibly with switches from other types of networks if SIM-roasting is implemented). A HLR must also be able to dialogue w ith all the entities that want to get information about Figure 7.4 summarises the simple architecture o f the protocols needed for location management. 7.1.4. THE LOCATION UPDATING PROCEDURES The main procedure o f interest for location management is the location updating procedure. which is triggered by a mobile station to I n n v 3LIt. U K ! i t t V I A N A C i 6 M E N 1 467 update the location•data o f its subscriber. For various s o n s . slightly modified versions o f this procedure are used f o r different related purposes. These variations are also described in this section. 7.1.4.1. T h e Basic Procedures new MSC The location information is stored in two different places in the ( lSNI infrastructure. the HLR and the visited MSC/N11.12. In fact the same information is known in three different places in the system. the mobile station (and more. explicitly the S I M ) being the third place. This information may change. and various procedures are needed to keep the consistency betweenithe three entities. The normal reason for a change is %%lien the mobile station decides that the location area best fit to serve its subscriber must be changed. Then the mobile station notifies the MSC/VLR to which the new cell belongs. This MSC/VI.R may he the same as before. i f it controls both ihe previous and the new location area. or a new MSC/VLR. In the latter case. the MSC/VLR notifies in turn the HLR. which notifies the previous \ISC/VLR. There are other cases where an inconsistency may appear, for instance when stored information is lost in the MSC/VLR or the HLR as a result o f some hardware or software failure. Then procedures may be run to correct the failed database using information in other equipments. 1 3 4' 4 2 old MSC In order t o cover all theses cases. the following elementary procedures have been specified (see figure 7.5): new MSCI a --T 5 • Updating or the MSC/VLR storage at the request of the mobile station; • Updating of the HLR storage at the request ofthe MSC/VLR; Figure 7.5 — Main elementary location updating procedures a • Cancellation o f a subscriber record i n a MSC/VLR a t the request of the 1-ILR: a The Mobile Station to MSC Location Updating Procedure This procedure is part of the RII3-NINI protocol. It requires a radio connection. as l i t ' any dialogue between the mobile station and the network. The establishment o f such a connection is a function o f the Radio Resource Management functions. described i n Chapter 6. This establishment has almost nothing specific t o the location updating procedure. In order to change from the Old situation (shown on top) to the new situation (bottom). the mobile station takes the initiative of location updating (I). hut the HLR. upon subsequent request from the new MSC/VLR (2), takes care of cancelling the old record in the previous MSC/VLR (3. 4') in parallel with confirming the updating in the new MSC/VLR (4). which in tom acknowledges the mobile station request (5). The M S -MSC location updating procedure i s basically very simple. and consists o f a request (a location updating request) and an answer. The request is carried by the RIL.3-MM LOCATION UPDATING REQUEST message. T h i s message contains mainly the information necessary to identify the subscriber. The NISCIVIR nitt a n s ‘ ‘ e r autonomously • s o m e cases, o r alternatively mats have t o update the III.R first. wan the procedure described in the next section. There is one case when the NISCIVLR cannot do otherwise than ans‘‘er on its own: when it cannot reach the II.R for lack of an \ roaming agreement between the two operators. This case is not addressed in the .N/Ictificutitms. though it can happen. The answer of the N1SC/VLR is necessarily negative and must be chosen CO ensure that the mobile station ‘‘ ill search other PLNINs (for instance. sending the cause "PLNIN not allim ed.). The usual "normal case- w hen the NISCIVLR can answer on its own is when the subscriber is already registered in the database of the N15C/VL.R. The response in that case is usually positive. and can he negative only in case of national roaming restriction (cause "location area not allowed f o r national roaming-. used only i n I)CS Iti(Hli: regional subscription cannot lead to a negative NISC/VI.R answer without IILR invhlvement, since the MAP restricts regional subscription to he offered on a per-N'I.R basis. National roaming then merits some attention. The rule is that if the mobile station belongs to an unwanted PLMN. and i f the requested location area is restricted. the NISC/VI.R is entitled tO directly answer negatively. Most of the time the NIS('/V1_1:2 will not be able to contact the M i t and this would he a particular case o f the situation mentioned above, with the difference that the cause sent to the mobile station is specified. When the NISCIVLR needs to contact the U R of the subscriber, it must first Rums which 1-ILI( is concerned. Subscribers are identified for the internal business o f GSNI h a number. the INISI (International Mobile Subscriber Identity). This number is provided by the mobile station anytime it accesses the nek\ ork (the number is not always given directly. see the notion of TNISI. page 454. The INISI is so specified that the NISC/VLR is able to clerk e the identity o f the subscriber's home PLMN. and possibly more information on the 1-112 equipment Acharge of the subscriber. Figure 7.6 shows the IMSI structure. With the help of the reit:\ ant translation tables. the NISC/VLR is then able to derive the S57 address to which the location updating request must be sent. I n practice. the IU.R can usuall h e identified 11.1) looking at the most significant digits o f the IMSI i n n s the mobile country code and mobile network code. However. ibis possibility is usually only used inside the home PI.N1N country. P T \ IN. of other countries route their messages using the !NISI as a global title. towards a gateway entity in the home PINES s m o l t . There the global title can be translated in the Signalling Piiint Code o f the right equipment i n the right PLNIN. as explained in Chapter 5. / 4Q, f / t Date ads MCC , MNC (3 digits) (2 digits) 15 digits or less Figure 7.6 - The structure of an IMSI An international Mobile Subscriber Identity consists of three parts: the Mobile Country Code (MCC), identifying a country: the Mobile Network Code (1NC). identifying a PLMN within this country: and the Mobile Subscriber Identification Number. identifying a subscriber within this PLMN, using no more than I0 digits. Before giving the answer to the mobile station, in an RIL3-MM LOCATION UPDATING ACCEPT message o r a n RIL3-MM LOCATION UPDATING REJECT message. the MSC/VLR may proceed to some actions like authentication or ciphering setting. These procedures are detailed in the second part of this chapter. The answer provided by the MSC/VLR. possibly after contact with the HLR. is either the RII.3-MM LOCATION UPDATING ACCEPT, thus indicating that the subscriber is effectively registered in the new location area as required. o r RI1.3-MM LOCATION UPDATING REJECT, with a suitable cause. such as "PLMN not allowed" o r "location area not allowed". The first indicates that the subscriber has no subscription entitlement for service in the visited PLMN. whereas the second, which can he used only in the home PLMN. indicates the subscriber has no subscription entitlement for service in the location area. The HLR does not indicate which cause to use (the M A P provides only one cause "roaming not allowed"). Though not indicated in the Specifications. the cause in the RIL3-MM message should he "location area not allowed" if (he visited N1SC/VLR belongs t o the home PLMN. and "PLMN not allowed- otherwise. The mobile station reacts on the receipt of a negative answer as explained in the section on cell and PLMN selection. , , , . • s t , ; • / “ . . 11 . 1 . V I I : I N I The MSC/VLR to HLR Location Updating F • e d u r e This procedure is run when a mobile station asks for registration under a new MSC/VL.R. It can also he run when the HLR has suffered a failure. and request the MSC/VLR for a confirmation of the subscriber location as soon as the mobile station is in contact. The request is conveyed in a \LAND EI'DATE LOCATION message. This message carries (among other information) the subscriber identity and enough information for the I ILR to knots how to find routing data for the setting up of a mobile terminating call. i.e.. the SS7 address of the MSC/VLR. It does not convey the identity of the precise location area. As an option supported by the standard, the routing data can be included at this point, but we will see in Chapter S that this is not what is usually done. The FILR determines >> holier t o accept t o register the mobile station in the new MSC'/VLR or not by consideration of the subscription limitations of the user. If the subscriber is entitled to normal service in the requested VI..R area. the answer is positive. the IILK updates its memory and triggers a location cancellation procedure w i t h t h e previous AltiC/V1,12 ( i f the mobile station ttas acivall> registered. and in another MSC/V1,10. If the subscriber cannot be granted normal service. the HLR ®isters the subscriber: i t updates its memory t o indicate that the location o f the mobile station is unknomi. and i t triggers a location cancellation procedure if applicable. The anstter from the HLR is carried to the NISC/V1.2 in a Tw i n t ( i t \ Ti n \ REst LT message. If the requesting MSC/VI.R recei>cs a 'legal' \ e answer. it erases all information relative to the subscriber. Otherwise and i f necessary, it enters the subscriber in its database. Normally. the 111.R teill then provide Ihr iitbscriber information the MSC: VI.R needs. 'Ibis is done by sending a w i t ) INSERT SUBSCRIBER DA'l A message. \\ hich is acknowledged by the M S C / V 1 . 1 ( t h r o u g h a M A P / ) INSERT S t BSCRIBER D A T A R E S U LT message. The same procedure ma> also he used when some change in the subscriber's information stored in the I11.12 has occurred, for instance at the request o f the subscriber. Ahernati> ely. depending on the nature of the modification. a procedure composed o f t h e NIAP/D DELETE SUBSCRIBER D ATA a n d t h e \ I A N ! ) D E T L I T SUBSCRIBER D ATA RESULT messages may be used. The III,R to RISC/1'/.R Location Cancellation Procedure Location cancellation from III.R to NISCIVI.R is a very simple procedure, composed of a request carried b> a i„,1 \I' I) 'As. cEt. — LOCATION 471 message and the x s p o n d i n g acknowledgemerit carried in a MAP/D CANCEL LOCATION RESULT message. The HLR does not wait for this acknowledgement to confirm location updating to the other MSC/VLR who triggered the change of location. The MSCIVLR to HLR Deregistration Procedure A MAP/D procedure opens the possibility for the MSC/VLR to ask the H L R f o r the deregistration o f a given subscriber. There i s n o identified case in GSM for this procedure, and it is indicated as not used in a laconic sentence in the first page of TS GSM 09.02. This procedure is the simple order/acknowledge procedure. The request i s carried b y a NIAP/D REREGISTER MOBILE SUBSCRIBER message, and acknowledged by a NIMID REREGISTER MOBILE SUBSCRIBER RESULT message. 7.1.4.2. P e r i o d i c L o c a t i o n U p d a t i n g , and Database Failure Recovery An HLR. or an \ISC/VLR. may suffer a failure such that a part of its database is damaged. These equipments are usually implemented in a secure way with hack-up systems. allowing the database to be restored to sonic consistent state. However, there are cases when the restored database is no longer up-to-date. e.g.. because updates having occurred heti\ Lien the last back-up and the failure have been lost. Since such conditions may affect the possibility to set up calls correctly. a number of mechanisms arc described in the Specifications to improve the situation. The general mechanism is as follows: in a first step. the affected equipment will mark the uncertain information as such. Then it will warn other network entities (those with which it knows to be sharing information) about the unclear status of its information contents. Consequently. these entities will also mark the corresponding records for checking. To avoid overloading the signalling system. no attempt is made to restore the consistency right away: the recovering equipment would probably not survive a high peak in signalling load! A given subscriber record (say Christian's) is corrected only when some event happens concerning this subscriber. such as a radio contact initiated by the mobile station. or a mobile terminating call attempt. I f after some time nothing has happened. Christian is deregistered. Obviously. the best place to stay informed about the subscriber position is the mobile station. and a radio contact is•at the end of the day .4 1 .1 rejection with cause .).451 unknown in VLR". Because this will also (rigger a location updating from MSC/VLR to HLR, consistency will be reached anew. A milder case is when the mobile station calls from a location area which is different from the one in which the MSC/VLR thinks it is registered. In this case, the MSC/VLR simply corrects its record. the only sale wan to restore consistency. Periodic 'neattt updating has thus been introduced to ensure regular radio contact (it the initiative of the mobile station. Basically. periodic location updating refers to the requirement that the mobile station contacts regularly the network when in normal service. This is done automatically b) the mobile station. and takes the form of a location updating procedure. The period is under contml of the network which broadcasts its value. The possibilities range from 6 minutes (in which case the network is probably completely overwhelmed by location updating procedures!) to slightly more than 2-1 hours. In addition, the infinite value is included: i n other words. the network can suppress completely periodic location updating i f so wished. The choice o f periodicity is the operator's. and is typically a trade-off between the quickness of recovery alter a failure (therefore the time when it may be impossible to serve mobile terminating calls for some subscribers) and the traffic load due to periodic location updating. This clearly depends a lot on the reliability o f the MSC/VLR and Ill.):: i f the mean time between failures is ver) long. there is no point in incurring the load and cost of periodic updating (except if implicit detach is also used. see page -1761. On the other hand. the periodicity can be increased by the operator alter a failure. In the case of a mobile terminating call, the MSC/VLR May notice a problem i f it receives a request from the HLR Foncerning a mobile station which is not in its table though the HLR obviously thinks so. This may happen either because of an MSC/VLR failure or because of an HLR failure. In all cases. the MSC/VLR enters the subscriber in its tables, and asks the H L R f o r the subscriber information v i a a MAP/D SEND PARAMETERS message, w h i c h i s answered w i t h a MAP/D SEND PARAMETERS R E M I T message f r o m the H L R . T h e location area information is in this case still missing. Henceforth, and until actual contact with the mobile station. the paging is done in all cells o f all location areas controlled by this MSC/VLR. Note that i f the error was in the III.R. that is to say i f the mobile station i f effectively not under the NISCAIR as the H L R knows. the information i n the network i s inconsistent: after a mobile terminating call attempt. the subscriber is registered i n two VISC/VLRs (MSC/VLR—I where i t is really, and NISC/\'LR-2 where the HLR imagines i t is). and the HLR stores the identity of the wrong one. The error will he corrected in the HLR and in \ISC/\'LR-2 when a location updating procedure is performed by the mobile station. However, MSC/VLR-I will remain in error by keeping the subscriber in its database. The cleaning-up can only be done by an internal mechanism in the MSC/VLR, erasing all records of subscribers which have been inactive for more than some long time. e.g.. a month. Let us now see in more details how the location registers cope with losses %)f information after failure. and how in general location data is recovered. MSCIVI.I? Failure When a NISC/VLR suffers a database failure. it first restores its state to some pre% iousl saved state. and alter this recovery marks all its records as to be checked with the mobile station and \\ ith the HLR. Then it sends a mAP/i) ithslit message to all IILRs for which it still has a subscriber i n its tables. The NISC/VLR will cs emu:illy notice that a mobile station is missing i n its records when some service must be provided to the subscriber. and at the latest when the mobile station performs a periodic location updating. In the reserse situation (i.e., a mobile station still recorded when it should not he). the correction may take longer. In the case of a request from the mobile station (e.g.. a call set-up, but not a location updating of some sort). the NISCAIR will notice for instance that the corresponding subscriber is not in its table though the mobile station thinks so. The NISC/VI.R reacts then by requesting the mobile station t o perform a location updating. This is done with a t HLR Failure Restoration of the HLR is a bit more complex, because the HLR is not necessarily contacted in the case of a periodic location updating or a mobile originating call. To enforce this contact, the HLR sends a MAP/D RESET message to all MSC/VLRs in which at least one o f the HLR subscribers is known to be located. as indicated by the salvaged records. The NISC/VLRs will mark all corresponding records as to be checked with the HLR: the next radio contact will then trigger a location updating procedure from MSC/VLR to HLR. thereby correcting the HLR records. - In the case of a mobile terminating call, as we have already seen two paragraphs above, the interrogated visited MSC/VLR aligns its database with the (recovered and unchecked) HLR data. At least, this is all that the MAP/D protocol. as specified. enabler t o do. However. HLR restoration methods described in TS GSM 03.6, include a different mechanism. by which the HLR would indicate to the MSC/VLR that the internigation concerns a mobile station for which data is unsure: the NISCRIR would in that case know whether it is better to take the HLRas a reference o r to rely on its own state and answer accordingly. Ilowe+Cr. this mechanism has not been implemented in the phase I MAP protocols. 7.1.4.3. T h e IMSI Attach and Detach Procedures The location updating procedure has two very close siblings: the periodic location updating procedure, which has just been described, and WC LI/Si attach procedure. On the radio path. both procedures are almost identical to a location updating procedure. They differ from it almost only by the events that trigger them. These events are such that these 'location updating- procedures usually appeala s a request to he registered in the location area \\ here the subscriber is a/ready registered. Hence in most of the cases. the HLR is not concerned with these procedures. Ihe strange names of the /MS/ (mach and /MS/ detach procedures are due to the accidents o f the standardisation. and are o f little use to understand their meaning. The best way to understand these proiedures is 10 explain what purpose s a t i s f y . When a mobile station i s switched o f f (or when the SIM i s reino\ cd 11) the users. the calls toward the corresponding subscriber can no inure be completed. Important resources are then consumed f o r nothing: as i l l he seen in Chapter 8. a eircitit is established between the caller and the MSC in charge of the called mobile station. and the paging procedure is performed. all to no avail. Worse ifrom one point of view). the csiahlished circuit is not paid for. As ++ ill he seen \( hen dealing with die routing o f calls. the establishment 01 )tle first part o f the circuit (before III.R interrogation) cannot he avoided. If is a different story with the second portion. bet"ecn the point h e r e I1LR interrogation is done and the visited NISC. To allex iate this useless load (and costs. the /MS/ detach and /MS/ (mach mechanism has been introduced. Basically. the subscriber's record in the MS('/VI.It contains a binary information indicating whether it is useful o r not t o try t o complete a call toward this subscriber. This information makes it at least possible to economise on paging. It may also pre cot the establishment of a part of the call. The /MS/ detach procedure will set this hit to -not useful to Irv-. whereas the WS/ attach procedure will (I() the reverse. The mobile station triggers an 11151 detach when it goes inactive.' I either a location updating procedure ( i f in a new location area) or an /MS/ attach procedure when it comes back (in the same location area). The M A P specifications include t w o different ways f o r the management of this feature in the infrastructure:_either the information is only stored in the MSC/VLR (the mobile station staying registered in this MSC/VLR as far as the HLR is concerned), or the subscriber is simply deregistered in the HLR and its record cancelled in the MSC/VLR. In tact. only the first option is allowed for GSM, since the deregistration procedure is not used. With this first option, paging can obviously be prevented. The second pan of the circuit establishment may also be prevented, but not so obviously. Moreover. even when possible, it is an option to prevent it or not. A s w i l l be seen i n Chapter 8, the basic scenario o f a mobile terminating call set-up attempt requires an interrogation o f the visited NISCNLR b y the H L R before the latter provides the information necessary for the continuation o f the routing. This phase allows the visited MSC/VLR to reject the call on the basis of the attach status before the costly set up of the traffic circuit. I f it does so, call forwarding i f applied can potentially he controlled by the HLR. Another possibility is that the visited MSC/VLR accepts the call, and applies the call forwarding itself if required. "lo complete the option list, the support of the attach/detach feature is a network option. It is allowed that a PLMN, or a portion thereof, does not provide this facility. This is indicated to the mobile stations on a cell basis in the broadcast information. The choice between all these options and sub-options lies x+ ith the visited MSC/VLR. since the HLR does not intervene. I f attach is indicated as supported in the current cell. the /M.S./ detach procedure is used by the mobile station to indicate that it (more exactly the SIM) will go inactive. The /MS/ detach procedure is an example of a procedure reduced to its bare hones: it consists of a single message from the mobile station to the visited MSC/VLR. the RII.3-xixt INISI DETACH message. This nie,age is not acknowledged. simply because it has been considered that the mobile station is typically switched off. or more generally not in a position t o receive an answer from the network. The IMSI detach procedure must use a radio connection. as any RIL3-MM procedure. This connection is either established for the purpose of the detach, or may preexist. T h e connection c a n h e abandoned b y t h e mobile station immediately after the sending of the k11.3-MM IMSI DETACH message. The mobile station keeps no track of having asked for a detach (for instance 1.)\ storage in the t h e state of the attach/detach information in the netv (irk is not monitored by ille mobile station. If attach is indicated as supported in the cell the m(1 s t a t i o n has chosen at switch-on (or SIM insertion). and if the mobile station knows the subscriber is already registered in the same location area, it starts an N S ! attach procedure. that is to say (except for a negligible detail) a location updating procedure. It should be noted that the attach procedure may happen even i f there was no request for detach beforehand (because the network did not require it at the time. or simply because it was not physically possible. e.g.. in case of a loss of coverage before switch-off). This is consistent with the specification that the mobile station does not monitor the attach/detach status. The advantagi a network time-controlled detach is that the detaching may happen even in the cases where the mobile station is physically unable to send an R1L3-MM IMSI DETACH message. This feature is not mentioned anywhere i n the Specifications. However it would work without side effects, and i t is known that a number of operators intend to use it. 7.2. SECURITY MANAGEMENT These procedures would then Rave been better called "Subscriber deactivation'' and "Subscriber reactivation-. From a v e r y slightly different point of view. these procedures are very close functionally to the call forwarding supplement:if\ servifes in the case where the mobile .station is not ®istered. The /.1/S/ detach procedure can indeed be understood as an automatic activation Of an unconditional call forwarding (toward an announcement. or a specific number). and the IMSI attach as the corresponding deactivation procedure. It can therefore be said that, from the point of view of the purpose of these procedures. of their why, they have little to do with ilw location updating. but are akin to call forwarding. From the point of view of the mechanisms they use. that is to say from their how, these procedures are close siblings to the location updating procedure. Radio transmission is by nature more prone to eavesdropping and fraud than fixed wire transmission. Listening to communications is easy and does not require access t o special locations. Impersonating a registered user (and therefore having him foot the bill!) can also he very easy i f specific protection means are not provided. Analogue systems have indeed suffered from such problems during the 80's. GSM had to bring significant improvements in these matters. 4 The security-related !Unctions o f GSM aim at two goals: first, protecting the network against unauthorised access (and at the same time protecting the users from fraudulent impersonations); second, protecting the privacy of the users. 7.1.4.4. A u t o m a t i c Detach by the Network As explained above. the attach status can be modified b y the mobile station. through the /MS/ ( m c h and /MS/ detach procedures. A past debate i n t h e G S M committees was whether a n additional mechanism may be added. consisting in putting a subscriber in the detach state, or deregistering him. it' nothing has been heard of him for more than a given period. The debate ended more or less against this feature. It can be nevertheless used. since i t requires only existing signalling procedures and otherwise affects only the NISC/VLR. Thanks to periodic updating. the mobile station o f a subscriber when switched-on MUM indeed Contact the network at least once every so often. Thanks also to the non-storage of the detach status in the SIM, the mobile station will always make itself known when re-activated: there is no period after switch-on during which the mobile station cannot be called because of a forceful detach by the network. 7.2.1. THE NEEDS Preventing unauthorised accesses i s achieved b y means o f authentication. i.e.. b y a secure check that the subscriber identity provided by the mobile station corresponds to the inserted SIM. From the point of view of the operator. this function is of paramount importance, in particular in conjunction with international roaming, where the visited network does not control the subscriber's record... and his ability to pay. Preserving the privacy o f the users is achieved through different means. Transmission can h e ciphered t o prevent eavesdropping o f communications on the radio path. Most o f the signalling can also be protected in the same way. preventing third parties from knowing who is being called. for instance. Finally, the replacement o f the subscriber's identity by a temporary alias is another mechanism to convince third parties that listening o n the radio path i s useless f o r tracing GSM subscribers. Since most of the calls involving a GSM user go through the fixed network. the designers of GSM did not aim at a level of security • •., • • I 1 I a i n n e 1 l i t l i V I h N I 479 much higher than that of the fixed trunk network. Mechii m s to ensure privacy have only been introduced f o r the radio path. Within the infrastructure. communications are transmitted in clear text. as they are in the PSTN. It is important to note at this stage that all the security mechanisms of GSM are under sole control o f the operators: the users have no possibility to affect whether authentication. encryption. etc. are applied or not. Moreover. users are not necessarily aware of what security features are used. Conversely. these security services are not usually subscribedfor. The Specifications leave a lot of flexibility to apply them in various conditions. Some harmonisation is however desirable. and is settled for the GSM90(1 operators for instance by discussions within the GSM MoU. 7.2.2. THE FUNCTIONS 7.2.2.1. A u t h e n t i c a t i o n 4 A simple autfrmication method is the use i)1' a password (or a PIN code- -Personal Identity Number). The level o f protection achieved by such a method is ytty low in a radio environment. since listening once to this personal code is enough to break the protection. GSM does make use of a PIN code in conjunction with the SIN!: this PIN code is checked locally by the SIM itself. vv ithout transmission on the radio interface. But in addition. GSM uses a more sophisticated method. consisting i n a layman's words i n asking a question that only the right subscriber equipment tin that case. the SIN1) may ansx‘er. The crux in this method is that a huge number of such questions exist. and that it is therefore very unlikely that the same question viould be used ILL ice. More precisely. the question takes the guise of a number. called RAi\D in the Specification.). LL hose L aloe is draLL n randomly between 0 and 2125-I !something like a few millions o f milliards o f milliards of milliards o f milliards: I. T h e ansLyer. c a l l e d S R E S i n t h e .Vecifinethitts----i.i.:.. Signed RkStilt i n cry plographic terminology—is obtained as the outcome of a computation involving a secret parameter specific to the user. and called Ki in the Spcciinco(ions (see figure 7.7). The secrecy o f K i i s the cornerstone o n Ly hich a l l t h e security mechanisms arc based. We is ill sec that it is stored in a very protected »av: f o r instance a subscriber cannot knoLL h i s K i . The algorithm describing the computation i s referred t o as algorithm A l i n the .1. pet-Mut:lit tm. b u t i t s s p e c i f i c a t i o n c a n n o t b e f o u n d t h e r e . I n f a c t . t h e design choices i t i b o t h i n t h e mobil,: station and i ❑ th e infrastructure. alio%) t o he opomor-dependent Li.hile allowing full Figure 7.7 - The :unbent ication computai ion \ uthernication is performed h) requiring the correct answer to the lolluwing riddle: what signed response SRES are )ott able to derive (runt the input challenge RAND. he applt ing the A3 algorithm with your personal (secret) key A7? inter-PLNIN roaming. Operators can therefore choose the A3 applicable to their ossn subscribers independently from other operators. Such algorithms are usually kept secret (belt and braces are never too much in this domain!). In order to obtain to desired security level. A3 should be what cry ptography experts refer to as a one-way (or trap-door) function. This means that the computation of SRES from Ki and RAND should be easy, k1hereas the computation of Ki knowing RAND and SRES should be as complex as possible. I t i s indeed this level o f complexity which determines which security level has been achieved. Even w i t h the knowledge o f several pairs (RAND, SRES) pertaining t o the same subscriber (i.e.. the same Ki). the computation should remain highly complex. Beyond this requirement. the only constraint imposed on A3 is the size of the input parameter iR.t.VD is 128 bits long) and the size of the output parameter (SRE.S must be 32 bits long). K i cari:• deed be of any format and length: here again the design choices leave t1 operator with a maximum flexibility. Only if Ki would be transported in the network (see page 4881 would it be constrained to a maximum length of 128 hits. At first view. it may surprise the reader to learn of the possibility for each operator t o choose A 3 independently. given the general specification philosophy of CSNI. Special elThrts were necessary to cover the case of international roaming. where the specification of a single A3 algorithm could appear an easy solution. However, several reasons justify the approach. One of them is the administrative complexity linked to the specification and distribution o f cryptographic algorithms. especially when they are t o cross borders. A s will' he explained later on, the algorithm used for ciphering in (iSNI is unique. and its specifications are managed in a totally different way from the other Specnicatioris. The management o f a single A 3 algorithm would have been even more complex. since authentication i s more sensitive than communication ciphering (the consequences o f a "broken" algorithm are more farreaching in the case of authentication). The management of A3 is much simpler if controlled by a single operator. :Willer reason is the existence of algorithms fit f o r authentication and Arcady implemented in smart cards. but possibly not open for sharing. A [Uniting factor being the smart card memory capacii>. the choke o f having an operator-dependent A3 algorithm enables telecommunication operators to use a single algorithm for. e.g.. GSM SIM and pay-phone access. frame number (22 bits) K c (64 bits) frame number (22 bi s) K c (64 bits) A5 I A5 it Si S 2 S i S 2 (114 bits) ( 1 1 4 bits) ( 1 1 4 bits) ( 1 1 4 ciphering deciphering MS bits) deciphering ciphering BTS Figure 7.8 — Ciphering and deciphering .\5 derives a ciphering sequence of 114 bits for each burst independently, taking into account the frame number and the ciphering key Ke. 7.2.2.2. E n c r y p t i o n ciphered to non-ciphered mode (and vice versa) belongs to the radio resource management functions and has been described in Chapter 6. Obtaining a good protection against unauthorised listening is not an easy matter with analog transmission. but digital transmission permits an excellent level o f protection with relatively simple means. thanks to digital cryptography methods. This has been taken advantage of in GSM, where the position o f the encryption and deciphering processes in the transmission chain allow a single method to be used for protection of transmitted data in dedicated mode. whether user information (speech, data. ...). user-related signalling (e.g.. the messages carrying the called phone numbers) or even system-related signalling (e.g.. the messages carrying radio measurement results to prepare for handover). This choice is not the result of a paranoiac approach. hut is justified by its simplicity. Only two cases need to be distinguished: either transmission is protected, and everything is sent enciphered. or transmission is not protected, and everything is sent in clear text. The actual procedure for changing from Both ciphering and deciphering are performed by applying an "exclusive-or" operation between the 114 "coded" bits o f a radio burst and a I I4,bits ciphering sequence generated by a specific algorithm, called AS. as described in Chapter 4. In order to derive the ciphering sequence for each burst, A5 performs a computation with two inputs: one is the frame number and the other is a key (named Kc) agreed between mobile station and network (see figure 7.8). The uplink and downlink directions use two different sequences: for each burst, one sequence is used for ciphering in the mobile station and for deciphering in the BTS, whereas another one is used for ciphering in the BTS and deciphering in the mobile station. For all types o f radio channels, the frame number changes from burst to burst, so that each burst of a given communication in the same direction uses a different ciphering sequence. The successive values of the frame number depends on the time organisation of each channel. The t • .• A I a I .VIPI / .s1J L I V i L I N I nine organisation of a TACH/E. exposed in Chapter 4. sh ; for example that the frame number is not always incremented by one every burst. As far as its representation is concerned. the frame number is coded :IS the concatenation of threes attics. called respectively TI. T3 and T2 (in that order) and amounting to 22 bits all together. What these three values actually represent is in fact meaningless as far as ciphering is concerned. The resulting cycle. the hyperframe. is a hit less than 3 and a hall hours long. and determines a periodic return o f the ciphering sequence. should the communication last this long. However. the key Kc is controlled b y signalling means and changes typically a t each communication. This key is not publicised. but. since it is often changed, it does not need as strong a protection as Ki: for instance. Ku can he read freely from the SIM. Algorithm A5 must be specified at the international level. since for achieving MS-roaming it must be implemented ss ithin every base station las well as in any mobile equipment). \leans to cater for several AS algorithms (e.g.. to cope with some regulation restrictions with regards to export outside Europe) hake been introduced i n the Specifications partially in ',bast: I . and definitely for phase 2 r ['or the time being, a single A 5 algorithm has been specified f o r use i n all countries. Its specification cannot however he found in ihe Specification). for security ieasons. This algorithm is the property of the GSNI a n d is tightly cops right protected. Its external specifications are however public. and it can he described as a black hog taking a 22-hit long parameter (the frame number) and a 64-bit long parameter We) to produce two 114-hit long sequences. As for the authentication algorithm A3. the level of protection offered by .\5 is determined by the comple.xits of the reverse calculation, i.e.. the computation o f A'1' know ing two 114-bits ciphering sequences and the frame number. Key Management The key Kr must he agreed by the mobile station and the network prior to the start of encryption. The choice in GSN1 is to compute the key k r independently from the effective start o f encryption. during the authentication process. Kr is then stored in a non-volatile memory inside Ilse SIM. so as to he remembered even after a switched-off phase. This "dormant" key is also stored in the visited N1SC/V1..R on the network side, and i s ready t o b e used f o r a start o f encryption. When authentication happens while the transmission is ciphered. then the active key Kr being used for ciphering/deciphering is not affected. hut the new —dormant" key is stored, and is reserved for use at the next occurrence of Ki R A N D C r 4 <----- sue, 4 4 T A RAND KI R A N D 8 g i r e4 4 A8 a a Kc S 1 Kc )1( MS network Figure 7.9—Kc computation Each time a mobile station is authenticated. the 'nubile station and the network also compute the ciphering key Kc by running algorithm AN with the same inputs RAND and Ki asfor the computation of SRESthrough algorithm A3. a transition between clear mode and cipher mode. Hence the terminology "dormant" key (versus "active" key) introduced in this book. The algorithm used to compute Kr from RAND (the same one used for authentication) and Ki is called A8 in the Specifications (see figure 7.9). Similarly to A3 (the authentication algorithm computing SRES from /?,4,\D and Ki). A8 is not specified in the ETSI Specifications, but can be chosen independently by each operator. Both algorithms could in fact he implemented as t i single computation. F o r instance. they could h e implemented as a single algorithm whose output consists of 96 bits: 32 hits to form 5/?/:S. and 64 hits to form Kr. Care must he taken that the knowledge of RAND and •S'RES does not give away too much information about Kr. It is worth noting that the length of the significaw a r t of key Ke output by algorithm A8 is fixed by the group of signatu!,.,s of the GSM MEM. and may he less than the maximum 64 bits. I n that case, the significant hits are complemented with zeroes. so that the format always uses the full 64 hits. As far as A5 is concerned, all patterns of 64 hits are possible and meaningful: this mechanism allows the level o f security to he increased in the future if needed. without any change to A5 (therefore without any change ()I' the mobile equipments) by increasing the number of significant digits within the limit of 64. Since A3 and AN are always running together. and in most cases are implemented as a single algorithm. they w i l l always be treated together in the rest of this chapter. and referred to as A3/A8. They are indeed so intermixed that authentication and "dormant" key management cannot he dealt with independently. No other " A x" algorithms are to he Mund in the Specs. A I. A2, A4. and so on. were place-holders during the design of the system. and disappeared eventually. but the terminology of the three survivors A3, A5 and AN was not changed. 7.2.2.3. U s e r Identity Protection Encryption is very efficient for confidentiality. but cannot he used to protect every single exchange on the radio path. Ciphering with K r applies only when the network knows the identity 01 the subscriber it is talking to. Obviously. ciphering cannot he applied to common channels, such as the BCCH. which is received simultaneously b y all mobile stations in the cell and in neighbouring cells, or else it could he applied with a key known to all mobile stations, and therefore quite useless as a security mechanism! When a mobile station moves t o a dedicated channel, there is some "bootstrap- period during which the network does not vet know the identity o f the subscriber. say Charles. and therefore cannot cipher. This has a major consequence: all the signalling exchanges up t o and including t h e first message carrying a non-ambiguous subscriber identity must he sent in clear. A third-party could at this stage listen to this identity. and know where Charles roams at this particular moment. This is considered harmful to Charles' privacy, and a specific function has been introduced in GSM to cater for such confidentiality. Protection i s obtained b y using an identity alias, the TMSI (Temporary Mobile Subscriber Identity). which is used instead o f the subscriber identity (the IMSI) when possible. This alias must be agreed before-hand between the mobile station and the network, during protected (ciphered) signalling procedures. Since this confidentiality feature is totally independent • l e other security functions, the description of the corresponding mecricinisms w i l l b e d e a l t w i t h separately f r o m authentication/ciphering later on in this chapter. 7.2.3. ARCHITECTURE AND PROTOCOLS The actors and protocols involved i n security management are almost the same as f o r location management, and this justifies their inclusion i n t h e same functional plane. However, f o r security management. the staring roles are displaced, and must be attributed to the SIM on the mobile station side, and the Authentication Centre (AuC), which can be seen as a part of the HLR, on the network side. The SIM and the AuC are the repositories of the key K i o f the subscriber. They do not transmit these keys. but perform the A3 and AS computations themselves. As far as authentication and setting the key Kr are concerned. all other involved equipments are intermediaries. The AuC is not involved in other functions than the ones just listed, concerning the GSM radio path security management. The AuC may be implemented as a separate machine or as modules of the HLR. The main reason f o r the distinction between A u C and H L R i n the Specifications is to sensitise operators and manufacturers to the security issue. As mentioned earlier, all the security mechanisms described in this chapter rely on the secrecy o f Ki. The AuC is a means to build an additional layer of protection around the Ki's. The SIM takes responsibility for most of the security functions on the mobile station side. It stores Ki, it implements the operator-dependent A3/A8 and it also stores the "dormant" Kc. The existence of the SIM as a separate physical piece from the mobile equipment is indeed one of the elements enabling flexibility i n the choice o f A3/A8. The mobile equipment manufacturers need not be aware of the specifications of these algorithms for any operators. The SIM manufacturers, on the other hand, must implement potentially different algorithms f o r each o f their operator-customers. b u t competition, mass-market production a n d distribution issues are totally different compared w i t h the mobile equipment market. The SIM protects completely Ki against reading. The smart card technology, introduced some time before GSM to produce these tiny electronic safes, was exactly fitting for this purpose. The only access to Ki happens during the initial personalisation phase of the SIM, when Ki is written in the SIM. This phase happens under the fig : o n t r o l o f the operator. Later on. Ki is only accessed internally within the SIM when it has to compute SRES and Kc: a procedure on the SIM-ME interface allows the mobile equipment to send a value RAND and to get in return, typically a few tens o f milliseconds later. the corresponding SRES and KC. Another advantage of SIM-storage for Ki lies with the possibility, if security requires (e.g.. as a regular measure. or i f it turns out that the chosen A3/A8 is not as secure as expected) to issue a new SIM on a persubscriber basis. The MSC/VLR plays several small roles. It initiates authentication; it decides to switch to ciphered mode: it checks the SRES provided by the SIM (through the mobile station) with the one provided by the AuC (through the HI.R): it stores the -dormant" Kc on the network side: and it manages the TNISI. Ciphering i s a transmission function. and as such i t involves transmission equipments (the I T S for instance). and the radio resource management protocols (and the RS('). These aspects were tackled in the respective sections. and will not be revisited here. This chapter deals only with the decision to go to ciphered mode. as well as with the management of the needed parameters. The security management functions at this level are supported by the same protocols (plus some others) as seen for location management. The R I D -MM protocol supports the dialogue between the mobile station and the MSC/VLR. whereas NIAP/D is used between the MSC/VLR and the 111.R. The SIM-ME protocol in the area o f security management uses more than just read or write commands. Others are added to' provide RAND and to request for an A3/A8 computation. as well as for getting back SRES. The additional protocols compared to those used in the location management area include the protocol beR‘een HLR and AuC. which is only indicatively specified in the Speed/cut/Dm. as part o f the general Operation and Maintenance GSN1 protocol. Last. a small additional protocol is introduced bemeen NISC/VI.Rs to enable an NISC/VLR to ask another one for the identity and subscription data of a user. before access to the HLR. This protocol is used upon access of a mobile statio identified by a TMSI with reference t o another N1SC/VLR. I t is the MAPIG protocol and is limned to one operation. RIL3-MM M A P / D /1\ MSC VLR SIM-ME r MAP/G MSC VLR A u C Figure 7.10—Security management protocols Security management is coupled with location management. and makes use of the same protocols. with two additions: a small protocol to transfer subscriber data between MSC/VLRs (MAP/Ci). and the connection of the AuC to the HLR. The protocol structure is summarised in Figure 7.10. 7.2.4. THE SIGNALLING MECHANISMS 7.2.4.1. A u t h e n t i c a t i o n a n d E n c r y p t i o n K e y M a n a g e m e n t As explained i n the preceding section, the computation o f the signed response f o r authentication and o f the ciphering key K r are performed simultaneously. based on the same inputs Ki and RAN!), the latter changing every time. The signalling mechanisms used for managing the corresponding data in the network are strongly coupled and will therefore he described together. The key Ki is a subscriber parameter. As such, it is stored in the 1-11_12. more precisely in the AuC. The authentication and key setting procedure. on the other hand. is controlled by the visited MSC/VLR, which decides when to run this procedure: e.g., at'call set-up, location updating, etc. There are then two different procedural aspects, distinct in time as w e w i l l see: the real-time authentication and k e y setting procedure between the mobile station and the MSC/VLR. and the ,„,.„ , , „ A . , procedure for transporting security related data bow! I E R / A u C and NISC/VER. The MS-A1SC Procedure The authentication procedure between visited MSC/VLR and mobile station consists of two messages: the 1:11.3-XIM AUTHENTICATION REQUEST message from visited MSC/VLR to mobile station, transporting R A N D . a n d t h e c o r r e s p o n d i n g R 11 3 - \ 1 \ t AUTHENTICATION RESPONSE answer from the mobile station. giving SRI'S for checking in return. At the reception o f the Rit.3-Nnt AUTHENTICATION REQUEST message. the mobile equipment sends to the SINI a RUN GSM ALGORITHM message, containing I t a n d immediately after a O n RESPONSE message \\ hose :ins \\et. contains SRLS and /Cc sItt.. is sent back to the network i n t h e k11.3-MNI AUTHENTICATION RESPONSE message. whereas Kr is written back in the SINI at the right place. and so kept for further use. An interesting procedural detail concerns the sequence number of Kr. also called ( ' K S N (Ciphering K e y Sequence Number) i n the Specifications. T o cope w i t h possible inconsistencies h e m een the dormant Kr on the infrastructuri.• side, and the one on the mobile station side. a sequence number is associated with it. This number is provided by the M S C / V L R i n the Rit.3-x1m A t THENTICATIoN REQUEST message. and is stored together with the dormant K r in the SRI and in the subscriber record in the MSC/VI.R. This number is given hack to the NISC/VLR in the initial message in the access procedure. so that it can he checked. If not consistent. the MSC/VLR knows that an authentication procedure is needed before ordering the ciphered mode. Value U o f the sequence number corresponds to "no Kr allocated 489 another subscriber s'arrieter among many others, t o be stored and provided to other equipments when perchance requested. Ay! there's the rub: this scheme implies that the key K i circulates through the SS7 network. introducing a weakness i n the system security, because interception cannot b e precluded. Moreover, fiternational roaming requires agreement between operators as far as A3/A8 is concerned. Either a single A3/A8 algorithm 'is standardised on a multilateral basis, or each operator must undertake to provide the specifications o f its own A3/AS algorithm to all others, the most cumbersome aspect being the implementation of them all in each visited MSC/VLR! The second solution overcomes both the security breach and the roaming problem. by having the computation performed in the HLR, in the AuC precisely. No need to transfer Ki any more, no need to divulge A3/AN specifications either! However, signalling means must be devised to transfer the result o f the computation from H L R t o the visited \ISC'/\'LR. I n order to avoid such a transfer every time the visited \ISC/VLR decides that it must authenticate, the computation is done in advance. For each computation. the AuC must draw a value for RAND and apply A3/A8. The result is a triplet of values: (RAND, SRES, Kr) to he sent to the visited MSC/VLR. The visited MSC/VLR stores a reserve of a few such triplets per subscriber, in which it can draw at need. This reserve is first established when the subscriber first registers in the visited NISC/VER: it is part of the subscriber data provided by the HLR in the \IAP/I) INSERT SUBSCRIBER DATA message. Tight security requires that a triplet he used only once. As a consequence. when the reserve falls below sonic threshold. the visited MSC/VLR asks the AuC. through the HLR, for more triplets. The only accepted exception to this "throw away after use" rule is when a communication failure has occurred between the visited NISC/VLR and the HLR. The triplet replenishing procedure consists in two messages: the MAP/D SEND PARAMETERS message and its answer. the MAP/D SEND PARAMETERS RESULT message. The MSC-HLR Procedures The computation of .SRI'S and K r on the network side. requiring A the knowledge of both Ki and A3/A8. must he performed so that its result is made avqilable in the visited MSC/VLR. Two options are allowed in the Specifications: either th e computation i s done i n t h e visited MSC/VLR. or in the AuC. The first possibility (computation i n the visited MSC/VLR) requires the visited network to cope with different A3/A8 algorithms depending on the home PLNIN operator. In this scenario. the HLR has no specific role for this function. and the AuC does not exist: K i is just 7.2.4.2. U s e r Identity Protection The Temporary Mobile Subscriber Identity (TMSI) is an alias for the subscriber identity used in order to avoid sending the IMSI in clear on —the radio path. The TMSI is allocated by the network on a location area basis: at a given moment, it refers non-ambiguously to-a subscriber when used i n conjunction w i t h the location area identity (LAI). Strictly speaking. the term TMSI should be used to refer to the full digit string composed of the LAI and of the digit string allocated at a given moment to a certain mobile station in this location area (which shall be called here "TNISI-code" or TIC). However. in the Specifications. TMSI is more -010 LOCATIONUPDATING L O C A T I O N UPDATING REQUEST R E Q U E S T LOCATION UPDATING ACCEPT 4 Or LOCATION UPDATING ACCEPT TMSI REALLOCATION COMMAND . TMSI REALLOCATION COMPLETE > T M S I REALLOCATION C O M P L E T E l'iL,tire 7.11 I N I S I allocation \ new TMSI can he allocated hi a mobile station throueli 3 ‘tandolone procedure consistim2 id two messages. However. a more economical sequence is allowed w hen TNISI allocation is perhirmed in conjunction w ith a successful location updating procedure. often used to refer to the TIC. hence with ambiguity when the context of the location area is not clear. On the radio path. most connections are set up in the location area in which the mobile station is registered. The TIC is in such cases sufficient to identify the subscriber non-ambiguously, the LAI being implicit]) equal t o the one the cell belongs to. The only exception to this rule is when the mobile station must perthrm a location updating attempt i n a cell o f a new location area: the f u l l TMSI (including the "old- LAI) must then be used, The length of the TIC is 4 octets. whereas 01 IMSI consists of up to 15 digits. coded in 9 octets including the length indicator. The short length of the TIC allows spectrum saying on the radio path when it can be used alone. This is especially the case for paging messages. where more than twice as man mobile stations may be paged in a single message when using the TIC. which cannot he ambiguous in these messages. But the most interesting area of study concerns the management of TMSIs, i.e.. how and when they are allocated and released. In the network. TMS1s are managed by the MSC/VLR (by the VLR in the fullblown canonical architecture). A TMSI is first allocated to a mobile. station the first time it registers in the location area. and is released when the mobile station leaves it. A standalone allocation procedure uses two messages: t h e R11.3-NINI TNISI REALLOCATION ( ' O \ t M A N D message f r o g MSC/VLR t o mobile station and the corresponding RiL3-mm TMSI REALLOCATION CO: 2 T E acknowledgement. However, when the TMSI allocation is performed just after a successful location updating (which is usually the case), the allocation message can be "combined" with the RIL3-NINI LOCATION UPDATING ACCEPT message f r o m t h e n e t w o r k : t h e new TMSI is then part o f the RIL3-MM LOCATION UPDATING ACCEPT message and the RIL3-MM TMSI REALLOCATION COMMAND is dispensed of. but not the acknowledgement. The full sequence o f 4 messages is however allowed as well. The corresponding sequences are shown in figure 7.1 1. TMSI cancellation i s usually implicit. I n the mobile station, cancellation i s automatic upon allocation o f a new TMSI, o r upon location updating acceptance in a new location area. Explicit cancellation can also be done. b y sending the IMSI i n the RIL3-MM LOCATION 131).VIING AccEPT message: this is to he understood by the mobile station as a cancellation of the previous TMSI. Another feature o f the TMSI allocation procedure enables the network t o use a short version o f the RIL3-MM TMSI REALLOCATION (UNMAN!) message in order to allocate a TMSI having the same TIC part (the 4 octets meaningful within a location area) as the previous one. This is achieved by not mentioning any identity in the RIL3-MM allocation message (whether TMSI REALLOCATION COMMAND o r LOCATION LPDATING ACCEPT), The gain brought by this feature seems however negligible. Table 7.1 shows a summary o f how to interpret the allocation message depending on its contents as f a r as the MOBILE IDENTITY parameter is concerned. The T M S I i s stored i n the subscriber's record held b y the NISCAIR. but not i n the HLR. When the record is destroyed, f o r example upon location cancellation from the HLR, the TMSI is then cancelled implicitly in the infrastructure. This raises some problems when database failures are considered. For instance, some situations may result Identity used in the message New TMSI none LAI (new) + TIC (old) TIC (new) LAI (new) + TIC (new) IMSI none Table 7.1 - Rules for deriving a new TMSI upon allocation Depending on the (Invents of the message sent by the network. a mobile station can derive the new value—if any- - o f its TMSI. . in the mobile station using a TMSI which is not allocate(' i t s subscriber any more, or worse which is allocated to another subscriber. In order le cope w i t h such situations. some tools have been included i n the Specifications. In particular. a procedure allows the network to ask the mobile station for its full 1MSI. Typical usage of such a procedure is when the 'TMSI by which the mobile station identifies itself is not known in the MSC/VLR. or. i f known. after the failure of authentication. which may reflect a TMSI discrepancy between mobile station and network. The identification procedure consists of two messages. the 1211.3- NINI mtiQuusT message sent by the network and the corresponding It 11.3-MM IDENTITY RESPONSE arISWer by the mobile station. The procedure is in lito more general than a request f u r the INISI. since other types o f identity inav be asked for. When a database failure has occurred in the N1SC/VLR. all the TMSIs stored in the salvaged records are dubious. In this situation. when an incoming call arrives, the MSC/VLR pages the mobile station throughout its ❑rea ( instead of just in one location area) using the TMSI instead of the TMSI. A side effect o f the TNIS1 is its usage in an itii.3-NiNt tocKrioN UPDATING REQUEST message sent i n a location area managed by an MSC/VLR other than the one which allocated the TNISI. We have seen that in this case the mobile station gives the full TNISI, and not only the TIC ❑s for a call setup or a response to paging. While the full TMSI is unambiguous. it does not indicate the HLR or even the home PLMN as the IMS1 does. But it does give the indication of the MSC/VLR which allocated the INISI. and which then presumable knows the corresponding IMSL The new NISC/VL.R has two possibilities. The first is to ask the mobile station for its T h i s is simple. but this is a breach in the protection of the subscriber identity. The other possibility is to request the IMS1 from the previous N1SC/VLR. as indicated by the LAI part of the INISI. This is supported by the N1AP/G procedure (the only one in this protocol). consisting o f a mApA; SEND PARANIFTERS message and the answer, the NtArAi stiND PARAmErliRs REsal:r message. The first message includes the list of the desired data. i.e.. the !NISI and optionally authentication triplets. The answering message includes the requested data. An authentication triplet allows the new \ISC/VLR to perform the authentication before the HLR answer. in order to gain time. • . A l I iviniNAtitAttN I 493 7.3. MISCELLANEOUS MM FUNCTIONS As explained a t the beginning o f the chapter, the Mobility Management realm includes more than just location management and security management, at least i f all the procedures i n the RIL3-MM protocol are taken into account. Even though this chapter might be a debatable location for some of these other functions, it was felt best to keep their description here, so as not to wander too far away from the Specifications description. It should not be forgotten that the goal of a structured protocol modelling is intended as a help for understanding inure than as an implementation guideline. A protocol architecture can therefore he to a great extent subjective. There are four MM miscellaneous functions not yet described. All of them concern only the mobile station and the MSC/VLR. In fact only two of them involve procedural exchanges (in the RIL3-MM protocol), the t w o remaining ones being specifications o f the mobile station behaviour. These procedures or specifications are related in so far as they use a common modelling concept, the MM-connection. This should not be mixed with the concept o f CM-transaction (CM for Communication Management). which refers to a transaction in the upper layer, the Call Control plane. A CM-transaction corresponds to a call transaction (RIL3CC protocol). to a Short Message transaction, or to a Supplementary Service management session. A CM-transaction corresponds then to all the activities described in the next chapter. After the presentation of these functions. we will discuss how useful this modelling is. Generic Mobile Originating CM-Transaction Establishment In ISDN call set-up procedures, the first message on the originating access interface is the SETUP message. This message contains a lot of intormation, including the calling number. In GSM, privacy on the radio path requests that this first message be ciphered. However, because the decision to cipher lies with the infrastructure, a preliminary message from the mobile station is necessary to give enough information to the network for it to decide whether to apply ciphering (and authentication) or not. An alternative would have been to cipher systematically. Because this preliminary message does not exist in ISDN, because of the will to keep call control as little adulterated as possible by the specific aspects of radio transmission, and because the same need exists for other kinds o f services such a s short message transfer o r supplementary services, a generic mobile originating a b l i s h m e n t procedure was introduced as part o f the RIL3-MM protocol. When initiated b y the mobile station. a call. a short message session or. a supplementary service management session all must use the generic establishment procedure. even i f transmission means a r e alNaidy established and used in ciphered mode. This allows the infrastructure to perform authentication, and/or to go to the ciphered mode before any further progress of the session. There is no equivalent t o the mobile station-originated generic establishment procedure i f the session is initiated from the,network side, simply because the network chooses t o apply authentthation and/or ciphering before starting any upper layer procedure. The generic mobile station originated establishment procedure consists basically in a preliminary message sent by the mobile station. the R11.3-NIM CM SERVICE REQUEST message. and a signalling sequence in answer from the MSC. The RIL3-NINI CM SERVICE REQui-sr message may he an initial message. as described in Chapter 6. The reactions of the MSC c a n b e t o s t a r t ❑ n authentication procedure (RIL3-MM At I I IENTICATION REQUEST message). o r to answer positively the request. This can be done by sending an ( N i SERVICE ACCEPT message, which seems normal. or by starting a ciphering mode setting procedure (RII.3-RR CIPHERING M O D E CONINIAND message). w h i c h i s a c u r i o u s specification. Still another possibiliiv for the MSC is to reject the request by sending an it11.3-mm s i t v i c F . kii.wri message. I f the answer is positive. the mobile station may start t o initiate the procedure v hich justified the pre ions actions. for instance by sending an 811.3-cc sauP message. The o n l v reason f o r "mixing u p - the generic establishment procedure w i t h t h e cipher 'node setting procedure (from another protocol!) lies i n the v ill t o reduce the number o f messages. A shortcoming o f this method is however the ambiguity thus introduced: there is no means to distinguish an itti..3-RR i i d t i v i MODE COMMAND message acknowledging a CM-transaction establishment and one which does not. -Collision- cases may well occur where the MSC wishes to start a ciphered session at the same time as the mobile station initiates a CM-transaction f o r another purpose. T h i s situation w i l l l i k e l y b e Unproved in phase .2. Another important point to note is the lack of connection reference in ❑n RIL3-MM ('Ni SERVICE REQUEST message and in the corresponding acknowledgement. This leads to ambiguity i f two generic establishment procedures are rug i parallel. Hence, this i s forbidden b y the Specifications. We will come back to these issues at the end o f the chapter. Upper Layer Synchronisation The Specifications require that CM-transactions cannot be initiated while a location updating procedure is running. The requirement is even more stringent in GSM phase 1: the mobile station must go back to idle mode before setting up a new RR-connection aimed at supporting a CMtransaction. "This requirement arises from the need f o r a subscriber to be correctly registered with the network before accessing any service. In other words, the rationale requires that no other Mobility Management or Call control procedure be started during a location updating procedure in a location area different from the one in which the mobile station was previously registered, at least u p t o the reception o f the RIL3-MM LOCATION UPDATING ACCEPT message. An alternative would have been to allow the mobile station to anticipate the closing o f the location updating procedure and to require the MSCNLR to store the service request until (and if) location updating is successful with the HLR. This is not allowed in the phase I Specifications. A milder position could have been to allow the mobile station to send a request just after the reception of the it 11.3-NiNt LOCATION UPDATING ACCEPT message, or to anticipate in the case o f a periodic location updating. But even this is forbidden in phase I. The only reason for such a drastic approach was the simplicity of the MSCNLR. Whatever the shortcomings. the specification must be implemented as it is. The impact of this synchronisation function, modelled in the MM plane in the Specifications. lies only with the mobile station and does not involve any procedure. Infrastructure Activity Monitoring A third function modelled within the M M plane and also not involving any protocol procedure i s a watchdog f o r MSC signalling activity. Radio channel release is indeed a privilege of the infrastructure. In case o f failure, the mobile station may then find itself in a difficult situation. with an unused dedicated channel which it is not allowed to release explicitly. In such cases. the mobile station must g‘ A c k to idle mode autonomously. This requires that the mobile station continuously checks i f there is—to its knowledge—a CM-transaction in progress. In the opposite case. the mobile station waits some time and decides to go hack to idle mode i f nothing happened in-between, without sending any message t o the network. T h e corresponding watchdog function i s modelled in the ,Specifications through a timer called T3240. This timer -could just as well be part o f the RR plane. but is specified in the NIM plane ()I' the Specifications. MM-location a e a C C S S etc... A s a t e r . . . 7 - * * . F M ) MM-security 7.7.7C) 1 7 = 1 7:7) RR Re-Eslablishinem When a mobile station, being provided w i t h some service, suddenly looses contact with the infrastructure. there is a possibility to resume this contact. for instance in another cell. Such cases may happen for example in configurations where the handover procedure proves to he too slow. A salvaging attempt by the mobile station has been introduced in the Specifications. and has been modelled in the M M plane. This is called the re-establishment procedure. A l l mobile stations must support this procedure. but it is optional on the network side. I f it is supported, call contexts must he kept a little while after the contact loss. to allow potential re-establishment to he effective. This feature is very close to what is called mobile station-triggered handover in other systems. It fulfils the same requirement as handover, but i n a much less controlled way. though w i t h possibly a better efficiency i n configurations where propagation loss i s very steep. Because of this analogy. the re-establishment procedure has been studied in Chapter 6. The initial message to request a re-establishment is the It.3-\t\1 CM RE-ESTABLISIINIENT REQUEST. It is then worth noting that the acceptance and the rejection by the network uses the same message as for the generic CM-establishment procedure: respectively the RiL3-MM ( T I SERVICE ACCEVt' message and the 811.3-xtxt CNI SERVICE REJECT. T h e re-establishment procedure is then deeply entangled with the generic CMtransaction establishment procedure. a point which renders difficult the evolution of the re-establishment procedure. Modelling t The four miscellaneous functions described above may seem quite ill-assorted from an architectural point of view. The two last ones would lit better in the Radio Resource plane for instance. There is however a common denominator between the four functions in the Specifications: Figure 7.12 - Alternative MM twiddling GSM NINI functions could he modelled in two blocks: a location management block on the same level as call control. etc. and an intermediate layer providing service to the location management block as well as to other upper layer blocks. the concept o f MM-connection. This pure modelling concept was not used here because i t i s not necessary and does not i n fact help understanding. I n fact the notion o f M M -connection does not appear concretely i n the protocols. There i s a .one-to-one correspondence between CM-transactions and M M -connections, which renders useless the insertion o f a C M -connection identifier i n the messages. M M connections are explicitly established only when initiated from the mobile station side (by the generic Mobile Originating establishment presented above), b u t n o t w h e n initiated f r o m t h e M S C ( t h e - - - - e s t a b l i s h m e n t i s implicitly done b y the establishment o f the C M transaction). Moreover it is never explicitly released. _ An alternative t o t h e modelling approach u s e d i n t h e Specifications, possibly reflecting the different roles o f the M M plane more accurately. would consist in considering the MM functions in two separate groups (see figure 7.12). On one side, a location management functional block which would stand in parallel and at the same level as call control, supplementary service management, etc. A location updating connection would then be considered the same way as a CM-transaction. Below such blocks, and above the RR layer, a second functional M M block would provide security related services to the upper layers, namely authentication. ciphering key management, as well as the generic Mobile Originating establishment procedure and a synchronisation function whose sole aim is to forbid the establishment of upper layer procedures while a location management connection is in progress. SPECIFICATIONS REFERENCE 'Hie problems raised b y inter-operator roaming are exposed from the service point o f view in Ts G S M 02.11. The main part o f this short document concerns the choice of PLAN. The technical aspects o f location updating i n the NSS are the subject o f 'I'S G S M 03.12. T h i s specification m a i n l y introduces the procedures included in the M A P. It i s worth noting that the general aspects o f location updating, PI AIN a n d c e l l selection i n t h e mobile station w i l l b e presented i n r,'neral terms in the future TS GSM 0 3 . 1 , unfortunately not present in die phase I . yerifiediions. Though requiring some caution. because o f the functional differences between phase I and phase 2. tins document Lan he useful. The details o f the cell choice ❑lgorithms. including measurement considerations are addressed in TS (;SM1 05.08. section b. The signalling aspects o f the M M protocol between the mobile 1,ition and the infrastructure (dealing with all the topics addressed ill this hapter. location updating as well as security management) arc dealt with ill 'IS G S M 04.08. section 4. A v e r y g o o d synthesis o f t h e general scheme f o r security management can he found in 'I'S GSM 03.20. The M A P protocols in general ore the subject o f 'I'S GSNI 09.02. ()1 relevance f o r t h i s chapter. o n e c a n c i t e section 5 . 2 (location registration/cancellation). section 5.8 (fault recovery of location registers, a subject introduced i n T S G S M 03.07). section 5.11 (management o f security r e l a t e d functions). a n d s e c t i o n 5 . 1 5 ( p a g i n g a n d search procedures). COMMUNICATION MANAGEMENT COMMUNICATION MANAGEMENT 503 8.0.1. The Communication 507 8,0.2. Management Functions • 510 8.1. Call Control 510 8.1.1. The Routing of Mobile Terminating Calls 528 8.1.2. Architecture 8.1.3. The Mobile Originating Call Establishment Procedure530 8.1.4. The Mobile Terminating Call Establishment Procedure539 543 8.1.5. The Interrogation Procedures 545 8.1.6. Call Release 547 8.1.7. In-Call Functions 8.2. Supplementary services management 8.2.1. Architecture 8.2.2. Procedures 552 553 554 8.3. Short messages 8.3.1. Architecture 8.3.2. Mobile Originating Short Messages 8.3.3. Mobile Terminating Short Messages 556 557 559 560 Specifications Reference 564 Managing calls, that is to say principally establishing and releasing transmission paths through meshed networks, is not by far a new topic. The development o f fixed networks, mainly the PSTN, and the ISDN nowadays, has seen an important evolution o f signalling techniques. Public cellular networks in general, and GSM in particular, are basically access networks for these general telecommunication systems. As such, the design o f their communication management signalling i s very dependent on existing techniques. and offers few novelties. For instance, the signalling exchanges at the interface between GSM and the external networks are imposed b y these networks. The influence goes even further, and th e communication management signalling procedures defined between mobile stations and t h e G S M infrastructure are simplified and somewhat adapted copies o f that specified f o r ISDN access. A part of what is described in this chapter is therefore not really specific to cellular networks, and could equally apply to fixed terminals. However, it is part of the Specifications, and this description is included to complete the full picture of the system. Still, call control is not inherited totally from ISDN. Cellular systems bring some problems of their own, due to the mobility of users and to the fact that there is no fixed link between each user installation and the infrastructure. The central issue is the establishment o f calls toward users that move around. One facet of the problem is the way in which the system must follow the movement of users between calls, in order to readily find them when need he. This has already been treated in Chapter 7 (Mobility Management). The other facet is the routing o f the u n o - A 111/4).4 I V I A N A U t r V I E N I call through networks to a GSM user. This js part of cal! n i t r o l and will be detailed in this chapter. • The technical issues related to the establishment of calls toward mobile users i s the reason f o r a number o f novel approaches f o r telecommunication networks. One of these important new concepts is the notion o f the GMSC (Gateway MSC, an i l l -coined term from the conceptual point4of view, since the "GMSC"Iimetioit has nothing to do with the MSC jimoion, even i f they are often grouped i n the same equipment), which is the central actor in call routing. This notion is deemed to be important in the future, not only for mobile networks, but for all public telecommunication networks. Their future lies undoubtedly \\ till the development of Personal Communication. i.e., the concept that calls arc not directed to locations, as it is the case in PSTN and ISDN at the moment. but t o people. w h o can he reached through different telecommunication means. according t o their present whereabouts. Mobility is an essential point for Personal Communication. and the GSM technology provides it' not a ready-for-use solution, at least a case of pioneer work in the subject. Another area where the signalling protocols developed for ISDN do not fit all the needs of GSM is the management of the radio interface, with the specific problems raised by the complex transmission scheme over the radio path. and b y the movements o f the users during a communication. These issues have been dealt with in Chapter 6 I Radio Resource Management). A s far as this chapter is concerned, the path between the MSC and the mobile station is considered to he a simple fixed link. This is again an application o f the "divide and conquer" approach. This kind of presentation allows a more consistent description, 503 not going this wf . n d that between different subjects (on the other hand, the complete picture of how a call is established in GSM can only be grasped by considering both points of view). Thanks to this approach, illustrated in figure 8.1, the substance of this chapter can really focus on the notion of end-to-end communication. The key "objects" that call control signalling deals with are communications, and we will organise the presentation in this chapter on how communications come into and out o f existence, and how they change. We will first try to define what a communication is, and what are its attributes, i n a n attempt t o identify the k e y functions o f call management. 8.0.1. THE COMMUNICATION Basically a communication is a temporary relationship between telecommunication users, for the purpose of exchanging information. A communication makes use o f a transmission chain established through networks between users. The only communications of interest here are the ones involving at least one GSM user. A communication is by essence a "distributed" object. which exists o v e r distances. Many o f i t s characteristics must be managed co-operatively by the different machines which appear along the transmission path, This management concerns both static attributes, set at the beginning o f the communication, and dynamic ones. which are modified during the lifetime of the call, usually at the request of the user. 8.0.1.1. T h e Users Gs m ISDN-like call control + Mobility Management + Radio Resource Management Figure 5.1 G S M functional split The split of signalling aspects into three major domains enables a description of a communication ‘v Mout taking into account the management of user's mobility and of radio resources. The first thing to be considered in a communication link are the users, or "parties". The basic communication case involves two users, but GSM will provide (in phase 2) the possibility to have communications involving more than two users. Though not a phase I capability, we will address this "multi-party" possibility, at least at the functional level, principally because this allows a wider approach to the concepts of call control. w h i c h c o u l d e a s i l y h e biased w h e n o n l y two-party communications are considered. A call is always started between only two users. As will he seen. the multi-party possibility allows the option of adding or removing users at any moment once the communication is fully established, making the attribute "users" a dynamjc one. ...... t.. But users are not functionally identical. One o f the i. s has a special status: the one who originally requested the establishment of the call. This user is called the "calling party". The other one is the "called paily:". There i s here a difficulty i n terminology.. Because o f the possibility of call forwarding. the subscriber whose number is dialled by the calling party is not necessarily the "called party" once the call is fully established. Let us recall the example of Chapter I. where Bjorn wanted to call Nina. but got I lans instead because Nina had requested her calls to he forwarded. The use of the term "called party" is therefore ambiguous. at least in the establishinent phase. Unfortunately. no established terms are used to distinguish the two notions. Where need he we will use the terms "forwarded-to party". or "connected-to party". Still to make things somewhat more complex. additional users can he added by either of the existing parties. New ones are then -called parties". but the notion o f "calling party" is relative to each called party. Well. it is hoped that the context will he sufficient to understand what these terms mean! The asymmetry between users is significant for the whole duration of the communication. For instance. in fixed networks. the call will he released at once i f the calling user hangs up, but will remain operational for a few seconds if the called user hangs up. enabling him. for example. to resume the call on another terminal. Another example of asymmetry is charging. But the asymmetry is obviously most important for signalling during the establishment phase. Because our purpose here is to focus on ISM users. two-party communication can be split into three categories: • calls where the GSM user is the calling party. These are the "mobile originating calls". or more simply "MO calls": • c a l l s where the GSM user is the called and/or the connected-to party. These are the "mobile terminating calls". or more simply "MT calls": • c a l l s 'where the two parties are GSM users. The last type o f call hears n o specific name. and being a combination of the other two. will not be considered separately. 8.0.1.2. T h e Partner Network Nlore generally. another important characteristic o f t h e l'onlinUn 'Cation is the network to which the other party (other than the itiNI user under consideration) belongs. This network will be referred to as the terminating network, for an MO communication. or the oriuinating t i , . I Et../il A l / N 4 N A U i i N i t i N 505 network, f o r a n N i . :ommunication. G S M i s designed t o allow communications between its users and users from a variety of external networks, such a s the PSTN, the ISDN, CSPDNs and PSPDNs. Establishment procedures in particular, and communication management in general will he heavily influenced by the nature of the other network. We will concentrate on the PSTN and ISDN cases, since they represent the most important part of the traffic by far. Also important when involved in a communication link are transit networks. Many different network configurations can be imagined. There again, f o r practical reasons. w e w i l l address mainly simple cases, typically when two networks only are involved, one being a GSM PLMN, the other the PSTN or the ISDN. 8.0.1.3. T h e Type of Service Continuing with communication characteristics, the type of service which is provided during the communication is worth some attention. The "type of service" is a concept which is concerned with the type of information exchanged between the end users (speech, fax, data, ...), but it encompasses more than that. The "service" i s more generally a subscription a n d management concept. I t s p r i m e characteristics correspond to the capability of the transmission path which is established between th e users. These "bearer capabilities" correspond t o the transmission as described in Chapter 3, that is to say (from the GSM point o f view) the connection type between the terminal adaptation functions inside the mobile station (the TA F ) and the interworking functions at the border with external networks (the IWF). GSM can deal with many connection types. and these have been summarised i n Chapter 3. Some services call for two bearer capabilities, so that the service can alternate between them during a call. "Alternate" services make use of the speech bearer capability and a data bearer capability (other cases could he imagined, but are not to be found in GSM). The interest of these services is that the call can be started in speech mode, e.g., for the users to agree orally on some details, before the data transmission is started. Such services require the system to be able to switch between the two bearer capabilities during a communication. This is not a simple function, and this is why the network must know that the parties desire an alternate service during the call establishment phase, so that i t can reserve the relevant devices. The alternate operation is an attribute of the service, and not a facility available f o r any speech o r data communication. The procedure which enables the toggling between two bearer cal i l i t i e s is the "in-call modification- procedure. For some services, the description o f the capabilities o f the terminals used by the end parties is relevant. These services are the "teleservices-. Along with the bearer capability (or capabilities), their description may include the "low layer compatibility- (LLC) and the "high layer compatibility- H L C ) o f the terminals. The " l o w layer compatibility- includes parameters f o r the basic transmission from terminal to terminal. such as the minimum duration of the stop signal in the case o f asynchronous data. whereas the "high layer compatibilityincludes characteristics related to the terminal to terminal protocol. such as w i d e x protocols. o r videotex protocols. These attributes are not specific to GSM. and the corresponding parameters are directly inherited from ISDN. Moreover. they are not usually necessary for the intervening networks, which obtain sufficient control information from the hearer capability. However. the indication of these attributes is the first phase of the negotiation between t h e terminals, and t h i s allows a n early compatibility check, which may result in the termination of the call i t the terminals ate unable t o communicate together correctly. Important teleservices are speech (basic and emergency calls). fax and videotex. Other services do not include anything concerning the terminals. They are the "hearer services... and their description as far as information exchanges arc concerned i s limited t o the bearer capabilities. The difference between a bearer capability and a hearer service is that the latter is used in the realm of subscription whereas the former is restricted to transmission capabilities. Services must he subscribed to individually. and supplementary services are linked to the basic services (i.e.. bearer or Rh:services). For instance. the forwarded-to number used for speech may differ f r o mthe one used for•fax. 5.0.1.4. A c t i v e and Held States . 1 , 11111 i 11111 . . 111 . 111 . 1 1 1 1 1 1 1 1 1 1 , " . 1 1 1 1 . \ Ve r t : m o s t l y o f a s t a t i c •..In die iind,e, t,l , Ailed ponies and the choice of hearer ,,, ( h . 1 1 1 .111,111.11e ‘ c l ‘ t e k " , . M o t h e r dyllanni aspt'tI Or a Ihc iii.,•••witos 1111 one 01 the other of the pamec to icinpoiaith lea o m m o i f i c a l i o n (1 e s t a b l i s h anoilicr one) mid then to Jesuitic it. Such possibilities are referred to as -to put the call NAV .flid then to t i t s les s s ill he available in phase 2. and -tic nisi h c i e to complete the picture. Horn the cominimicalioa point •,1 i t ' ‘ ‘ . ' A n s i \ Nialt'• exist for each of the kiwis ed part t'': the ...ill can he either held or active. If. for an reason. one of the tss,, parties is in the held state, user data is not transmitted, but all of the transmission resources are retained' -AllOW a fast resumption of transmission when so desired by the party who asked for the held state. Of course, the charge counter is also kept running... Because there is the possibility to establish or to answer another call while a first one is on hold, a given user can be part o f several communications at the same time. However, only one communication can be active at a given moment for a given user. all others being on hold (except f o r short messages, which are transmitted l i k e signalling messages and hence can be sent and received in parallel with other basic services). In the future, with half rate channels. it could be imagined that two communications are active at the same time with a single terminal. each on a different half rate channel. This is something that is foreseen, and for which some procedures are already documented in the phase I Specifications. 8.0.2. MANAGEMENT FUNCTIONS This panorama o f the attributes o f communication has already hinted at most o f the operations that can affect a communication. Foremost is the establishment of the communication, which must result in the setting of all the attributes, and in the establishment of the end-to-end transmission path (this description excludes the Short Messages, f o r which there is no establishment of a circuit-like transmission path. Short messages are addressed in a separate section). Most attributes of a communication are fixed between the calling party and the infrastructure machines, more precisely with the MSC. during the establishment phase. Some of the attributes are chosen solely by the calling user, such as the directory number of the called party or the type of service. Other attributes are fixed by the infrastructure alone, such as the actual connected-to party. which depends on forwarding conditions managed by the infrastructure under the control o f the called party. Finally. sonic attributes are negotiated between the user (or the mobile station) and the MSC. An example of such negotiation will in the future he the choice of the type of channel on the radio path, when the required service may he accommodated on either a full or a half rate channel. Then comes the setting of the transmission path. Since GSM is mainly. an access system. most of the job is done by external networks. For a Mobile Originating call. the MSC must analyse the called number :Ind the requested service in order to choose the external network towards which the call will he routed. It has then to reserve a suitable link toward t. this network and to proceed with the relevant signalling. aces. . n g to the access rules of the chosen external network. Afterwards. the MSC (and the GSM user) w i l l just wait and react according t o the signalling information available about the progress of the call. In all standard telecommunication networks, the transmission path is set up segment per segment. from caller to called user. Each node receiving the call establishment request analyses the called number. then determines the next node. reserves a link toward this node. and passes on the ball. The path is built in this way until it reaches the node in charge of the calle6i-party. Routing tables are used in each node. giving the rules for translating the number i n a link toward the next node. This simple technique works well because the called party is in a fixed location. The stun is somewhat different when the called user moves around, as a GSM subscriber does. Then the called number is not sufficient t o determine where to route the call. and a more sophisticated approach is necessary. The routing of MT calls is a complex matter that warrants a detailed study and a whole section will be devoted to this subject later. Once the transmission path is established and the end user has accepted the call. the call is "connected". meaning that the transmission path is fully available for the transmission of data. at least between the TAF and the IWF. In the case of speech. this means that end users can hear each other, but in the case of data. further establishment steps have to be taken before user data can truly be exchanged end-to-end. We will conic back to this issue. During the "connected" phase. also called the "active- phase of the call. a small number of functions are provided. They concern mainly the management of alternate services and (in phase 2) the management of multiple calls. During an alternate service communication. the GSM user can request a toggling of the bearer capability (from speech to data, or vice versa). This request must be transmitted to the MSC, which then controls the changes along the transmission path (e.g.. in the IWF) to meet the new requirement. Multiple call management corresponds to multiparty calls. or the ability to resume calls or put them on hold. Here again. the MSC is the director, and signalling between the mobile station and the MSC enables a co-ordination of the operations. In-call functions also include the transmission o f digital audio tones (DTMF tones. f o r Dual Tone Multi Frequency). These tones, usually generated by pressing the keys on a modern telephone set, are used for the control of, e.g.. voice mailboxes, answering machines, etc. Though DTMF transmission is not really part of the management of the communication. being more akin to a transmission facility (providing a I A S AI AN A t i l t M E N T 509 data transmission capal. . y of a few tens bit/s!), it has been chosen to support it in GSM by using signalling exchanges in the Call Control area. Let us recall that the speech coder in GSM has been optimised for speech signals. and DTMF tones going through a coder-decoder ordeal will not meet the quality required by external networks. This is why signalling means are used f o r conveying D T M F signals on the M S t o MSC segment. The last function needed f o r circuit call management i s an elementary. but very important one, the release of the call. This can be initiated by any one of the users, and is controlled by signalling between the users, the mobile station and the MSC. We have s o f a r left t o one side t w o functional areas, the management o f Supplementary Services and the transfer o f Short Messages, and these will now be introduced. GSM enables its subscribers to inspect and modify the status or the parameters o f their supplementary services, b y using their mobile station and transmitting over the radio path. This can be done at any time, irrespective of whether a call is in progress or not. Some supplementary services can be activated or deactivated at the subscriber's request, such as for instance the call barring services. Another example is the call forwarding services, where the forwarded-to numbers can be changed at will by the subscriber. In relation to these modifications, a password can he set by the user, to improve his level o f security. For instance, a password set in connection with the barring of all outgoing international calls enables the GSM subscriber to lend his SIM without having to fear that international calls will be charged to him. All these supplementary service management functions are done between the user and his home location register (HLR), with the use of the mobile station. They call for a few signalling procedures. It should be noted that these management functions represent only one facet o f the signalling related to supplementary services. The rest relates to the actual impact of the active services on call handling, introducing variations of the call establishment procedure. As such, they are described later in the section dealing with call establishment. Short m e s s a g e c o m m u n i c a t i o n d i f f e r s f r o m c i r c u i t communication in the fact that the establishment, the transmission of user data and the release are all done in one procedure (at the communication management level). This is close to a datagrarp technique. There is no establishment o f a dedicated transmission path, since only signalling means are used. Another difference comes from a special function added for improving the quality o f service when the called party cannot be reached: the undelivered messages are kept, and another attempt to send 3 , ! can be initiated when the destination subscriber becomes rc ' a b l e . This mechanism calls for signalling exchanges between networn entities to alert the short message service centre. national destination code country code Nfly This completes o u r l i s t o f the communication management functions, and we can now look in more details to each of these issues. CC NDC subscriber number +44 802 UK GSM number (Cellnet) 8.I.- CALL CONTROL +44 385 UK GSM number (Vodaf one) +44 956 UK DCS 1800 number (Mercury P.C.) Before describing the architecture and the procedures involved in call control, we will analyse how calls to a mobile subscriber can be routed. since this aspect underlies much of the network architecture, and is of foremost importance in understanding many aspects of the system, such as. e.g.. how to call a GSM user or charging. +44 973 UK DCS 1800 number (Hutchinson Microtel) +358 40 Finnish GSM number (Telecom Finland) +358 50 Finnish GSM number (0Y Radiolinja) 8.1.1. THE ROUTING OF MOBILE TERMINATING CALLS Figure 8.3 — The structure of a GSM directory number A GSM "MSISDN" looks like a standard PSTN or ISDN number, but the knowledge of the "National Destination Code" (NDC) identifies an operator within a country, and not an area code. The first digits following the NDC are used to identify the relevant subscriber's HLR within the home PLMN. For a Mobile Terminating call. the number given by the cal t:. ing party does not refer to a telephone line or a location: but points a record in some HLR. The first digits of a GSM directory number are sufficient to indicate that the number is a GSM number. and furthermore to designate the operator with which the subscription i s held. The structure of the GSM directory number. also called "MSISDN" because it is part o f the same numbering plan as ISDN numbers, is defined in 4 i l \ I * ' -- + i 3 .: 5S i 8 p i o i interrogation n t 40 123456it CCITT Recommendation E.164 and is shown in figure 8.3. The HLR holding the record of the subscriber can be determined by the analysis of the first digits of this number. The HLR record contains information necessary for finding the final destination o f the call, i.e.. the MSC where the GSM user is currently visiting. As a consequence. the final routing can be done only after the interrogation of the HLR. This splits the call establishment into two parts: before the interrogation, and after the interrogation. This corresponds also to a clear division of the call route into two parts: from the call originating point t o the interrogation point, and the rest, as represented in figure 8.2. 2 routing to actual location Fjpire S. —The two Ilan, Of a mobile terminating call route The call route consists or two parts• the first part is based on the called d i r e c t o r n u m b e r . a n d c o n t i n u e s t o I l k ' 1101111 W h e r e t h e :1011,1110C:111011 01 I h r c a l l r , l t ; s x t s u b s . i . d . c n M b ' .1,- , . a w n I A I I , . e s ! A the with.. I What follows applies to Mobile Terminating calls requiring the establishment o f a circuit. The routing o f Mobile Terminating Short \less:nies is similar. but hears a few differences and will be described separately in the section dealing with Short Message Services. Another point is that what is specified in GSM applies mainly, i f not only, to the This simple description raises several interesting points. First, what is exactly a switch caps o f the GMSC function? Second, where does the routing number come from? Third, who pays for what? The last question being the key to the first one, let us start with the charging issue. HLR directory number routing number kt ....," G M S 8.1.1.1. W h o pays What? C routing number I ih,...1 l )/ t directory number MSC VLR Figure 5.4 ' l i e F O I C Or the C M S ( ' A inithile lerminaling call i . I ir4 mined iin‘aril. a (iNISC. i.e.. a ' , W i t c h a b l e I n i n t e r r o g a t e I l l e m o b i l e s u b s c r i b e r s t o k n o \1/4 1/2\ I l e r C t r i : I I : W a l l y R a l i k ' I h e c a l l . cases where the second part of the route goes through the PSTN or the ISDN,' not through a packet-switched or circuit-switched data network (PSPDN or CSPDN). A n M T call coming from these latter networks must then either enter directly GSM before reaching the interrogation point. or use the PSTN or the ISDN as a transit network. We will then restrict our presentation t o M T cirt4oit calls involving only general networks, such as the PSTN. ISDN or GSM. The first part of the routing is done only with the information that can be derived from the called number (the MSISDN) independently from the called party location. This routing is done as for apy ISDN number, with tables in each o f the intervening switches. The routing tables are normally set so as to reach rapidly the switch which is able to interrogate the corresponding HLR. Not only does such a switch include the software necessary for running the interrogation procedures, but it also holds a table which relates an MSISDN with the corresponding HLR. This function is referred to as the GMSC (Gateway MSC) function, •and its role is shown in figure S A The interrogation o f the HLR is a simple request-answer procedure as seen from the GMSC. The answer contains the identity o f the subscriber (for billing purposes), and the information for the next routing step. This information is basically a routing number pointing either to the called GSM subscriber i n his current location. or to a third user in the case of call forwarding. The charging and tariffing policies are to a large extent outside the realm of the Specilicai inns. They are an operators' issue, dealt with in the GSM M o U meetings. However. t h e answer has some technical consequences. and s o i s worth some consideration. T h e charging principles which follow have been extracted f r o m various public conferences of the GSM MoU operators. In the case of a non-forwarded call, two parties are involved and may be charged: the calling party (say Woldemar, a German PSTN subscriber) and the called GSM subscriber (say Peter, who is a subscriber in the Netherlands. but who happens to be currently travelling in Spain and obtains service from the Spanish TELEFONICA GSM network). Several networks are entitled to demand some part o f the call charge. These are the German PSTN, the transit PSTNs or ISDNs along the way, and the PLMN of TELEFONICA which will establish the final segment of the path. Woldemar will receive his bill from the German PSTN, and Peter. i f any charge is levied on him, will be billed by his home PLMN, i.e.. in the Netherlands. and in Dutch currency. The other networks are not able to bill these subscribers directly. The matter is then not simple, and accounting transfer between some of the involved,networks will be necessary. as well as specific agreements between the operators. Some practical considerations must be taken into account and some rules must he fixed to simplify the combinations. The total cost of a call depends obviously on the location o f the GSM subscriber. It is also clear that the calling party would like to know in advance how much he will be charged for the call. Moreover, it seems that GSM subscribers are not willing to have anybody calling them know where they are located, even with little accuracy. These arguments lead to the principle that the charge levied on the calling party is independent from the actual location o f the called party. This philosophy is in line with the similar problem of a forwarded call in the PSTN: in this case, the calling party is usually levied the same charge as i f the call was not forwarded, and the party which asked for the forwarding is charged for the complement. • T h i s principle still leaves room for different solutions. In a first xtreme solution, the rate of the charge levied to the calling party can be ,designed so that on average it covers the cost of the call, wherever the . 11 - 4 1 . N t i t A l u m M L N I C AT I O N t location o f the called party (and including the charge L t h e radio segment): and the called party never pays anything. Another extreme. and opposite. solution consists in charging the calling party with only the minimum charge the minimum being taken on all possible locations of the called party). and hiac the called party pay systematically for the rest ()I' the cost. But ley) ing a charge on the called party is not particularly easy. since it means raising charges on calls that the user would sometimes haye—pfcleitped not to recciyta,- In fixed networks and in most cellular networks in operation today. the entire charge is supported by the calling subscriber (except in the case M. forwarding. as explained above). I t seems then good practice to do the same when the subscriber is located in 11k home PLMN fin our example i f Peter was still in the Netherlands), i.e.. t o charge onl) the originator o f the call. This will still require compensation mechanisms between operators. but in a more deterministic way. Let us consider several examples. to understand better how this principle applies. Starting with the simplest case. let us see what happens when Waldemar (using the PSTN ) calls a German GSNI subscriber. say Frieder. who is roaming in his home PLMN. Figure 8.5 illustrates the PSTN e GSM' Home PLMN MANAGEMENT case. Woldemar is the t . y one charged. The amount of compensation paid by the German PSTN to Frieder's home PLMN, if any, is a matter of agreement between the operators. It seems logical that this compensation should cover the costs concerning the "radio" segment (MSC to MS) at least. and possibly more depending on the point o f entry in the GSM network. However, since for Mobile Originating calls the situation is reversed (the charge is recovered by the GSM operator, who should give some money to the German PSTN for routing the call toward the called user). there is not necessarily a positive flow of money at the end of the month from the PSTN operator to the GSM operator! In any case, the German telecommunication networks taken as a whole cannot obtain any more money for a Mobile Terminating call than what is levied on the calling user by his PSTN. I.et us now offer a trip to Frieder, say in Portugal. Woldemar still dials the same number to call Frieder (i.e., a German number), and the interrogation point (the GMSC) automatically re-routes the call from Germany to Portugal. without Woldemar being aware of it (principle of location confidentiality). He should then pay the same charges as in the previous case. But more networks have been involved in the call, which now includes an international leg. as shown in figure 8.6. The standard compensation mechanisms i n u s e between f i x e d networks f o r PSTN GSM' Visited PLMN. 4 transit networks Gam' Home PLMN Figure 8.6—NIT call 1(m arc! a roaming subscriber Figure 8.5 -- Routing of an \ IT call within one country When the called (.551 user roams in his home PI .\I\. a „III from a I.; ),..1 user in the same country is simply routed to the home Pl. NIN t h e PS IN. 515 the calling party is unaware of the actual location of the called subscriber (for him. everything is the same as in figure 8.51. The c0.1 Of the international leg can he recovered h) le‘ i n g a chare t h e called GSM user. ... international routing will apply to the German PSTN, so, t t the transit networks get their share. In such cases, an additional charge to cover the international leg will have to be raised, and it can he levied on nobody but Frieder if the principle of location confidentiality is retained. Let us now take the most complex case. where the originating network, the home PLMN of the called user and the visited PLMN are all potentially i n different countries. T h i s corresponds t o o u r original example, w h e n Woldemar calls Peter (diSNI subscriber o f t h e Netherlands). himself roaming i n Spain. I n this case, Woldemar is tlialling•an international number starling with 31 (the country code for the Netherlands). and as such is prepared to pay,the cost of an international call towards the Netherlands. and possibly a;little more if he knows that the number he is calling is the one o f a mobile subscriber (and i f the German PSTN also knows this information. which i s not b y far an obvious matter). Depending on the location of the interrogation point (the GMSC able to interrogate Peter's HLR in the Netherlands to know where he is located), the call may experience one or several international legs, as shown in figure 8.7. Besides the German PSTN. which gets the money G s rifr Home PLMN transit networks PSTN Gsm. Visited PLMN. possible positions e of the interrogating point corresponding routes Figure 8.7 — c a l l involving three countries Depending on the location of the interrogating point, the call may he routed through one or several international leg.. without the calling user being aware' of it. , , , / , I \ „ V L.0,11.11'. I J1/ from Woldemar, Petei i o n i c PLMN may charge him for some amount linked with his current location, but i t is the GMSC which holds the relevant information on the routing cost of the second segment of the call (from GMSC to VMSC). As for the other networks through which the call i s routed (including the visited PLMN), they need t o get some compensation for the costs incurred. All this leads obviously t o complex transfer mechanisms. The GNISC appears as an important participant, and the network to which it belongs heavily influences the complexity of the problem. In fact things can be simplified if some choices are made concerning the position of the GMSC. Let us examine this issue in more detail. 8.1.1.2. T h e GMSC Function The GMSC function requires only a switching capacity, and special software (including the ability to establish a toll ticket for the second branch o f the call). It must hold a table linking MSISDNs with HLRs. There is no reason why the table should be complete. Indeed, a given GMSC may f u l f i l i t s function only f o r the subscribers of some Home PLMNs, or even for a single Home PLMN. As will be seen, a simplification is to have the GMSCs specific to one PLMN. GMSC By nature, the GMSC function is independent from the radio access function provided by a PLMN, and can be implemented as a part of any network through which the call is routed, typically the PSTN or the ISDN. As specified in the Specifications, the GMSC function. can be implemented in any switch of the PSTN, the ISDN, or directly connected to those networks. From a service point of view, the deater the density of GMSCs the better. Indeed, the closer the interrogation point to the calling party. the more efficient is the routing. The epitome of what should be avoided is what is called in technical jargon a "trombone", that is to say the case where the second branch begins by backtracking the path of the first branch. This is obviously not efficient, and is prevented i f GMSCs are everywhere. This ideal view is not so ideal when taking into account the charging considerations. The billing record established by the GMSC covers the second segment of the call (from GMSC to called party). This record must be transferred towards the home PLMN o f the called subscriber, in order to bill him the amount not covered by the fixed , v o t i • a u n i l . A I !UN M A N A C i l i N I E N T charge of the calling subscriber. A first problem is tha t i s transfer of billing records must be organised with all Home PLMNs. for which the GMSC is able to interrogate the HLR. A second problem is that the charging mechanism for the second branch depends on the position of the GMSC. If the GMM • is not in the home country of the called party. a a r t of the charge lior the second branch may have to go to the calling parr if we want to follow the principles developed in the previous section. An important simplification for charging management is obtained if the first leg:bit:ay% charged to the calling party and the second branch always charged t o t h e called party. N o w. applying t h i s principle i s i n contradiction with the search for an efficient routing. As explained in the previous paragraphs. in a network configuration optimised for routing, there should always be a GMSC close to the calling user. and he would nut pay for, e.g.. the international path if the called user is a foreigner. To solve this dilemma. the choice was made in favour o f simplifying the accounting procedure. and to forget routing efficiency. To achieve this aim the interrogation function for the subscribers o f a given PLMN is performed only by switches of this PLMN. The GMSC is always in the home PLMN (and hence i n the home PLMN country). and the two segments of the call can he considered as "from calling party to home PLMN", and "from home PLMN to visited PLMN". and each charged to one end of the route. The resulting principles appear in figure 8.8. Let us consider the consequences of this approach. I f the called party is in his home PLMN (Peter is in the Netherlands), the second branch of the call is reduced to a minimum. and this is consistent with not levying the called user in this case I Woldemar pays the whole amount, and Peter is not charged). Conversely, when the called party is roaming abroad (Peter is in Spain). he will pay for the international leg between his home PLMN t o his actual loCation. This also does not raise an acceptance problem. because this rule is simple to understand, and the charge is easy to predict (it consists in the price of an international call between the home PLMN country and the visited country). Moreover, a GSM subscriber has the possibility to ask that incoming calls are not completed when he is roaming abroad: to do that, he must activate the supplementary service "Barring of Incoming Calls when roaming outside the home PLMN country" (131C-roam). Call charge transfer is much simplified. The same network (the home PLMN of the called subscriber) generates the charging information concerning the second portion o f the route. collects the corresponding charges i f any from its subscriber, and receives the invoice of the other intervening networks further down the call route. for compensation. 519 Home PLMN 7 . GMSC c (international leg of the call) Visited PLMN MSC VLR pays for: a b e d ® . Gem' pays for c (to the Home PLMN) Figure 8.8 — Charging principles of mobile terminating calls The charging principles retained by GSM MoU operators are based on the following principle: the originator of the call is charged as if the called GSM subscriber was in his home PLMN. The international leg. if any. is paid by the called GSM user. Conversely the GSM subscriber is not charged for incoming calls when located in his home PLMN. The originator is unaware of the actual location of the GSM subscriber, and will always know how much she will pay. The only problem with this approach is that "tromboning" is not suppressed. Routing optimisation can still be achieved on a national basis, i f the density of GMSCs is high, but this is not possible when the calling user and the home PLMN are in different countries. The lack of efficiency is obvious when the called GSM subscriber is roaming in a foreign country A and is called by somebody from country A, possibly a few meters away. Then the call is routed to the home PLMN country and NICATION MANACitiMENT country A c o u n t r y trombone 5 2 1 it from his end, since alt p r o b l e m s of subscriber location do not exist front the mobile user to tixed user. More generally speaking, the lack of efficiency (and therefore the additional cost) in certain situations such as the one described represents the price to pay for location confidentiality in an international environment where billing and accounting i s a complex and constraining issue. B Home PCMN GMSC HLR MSC WR Visited PLMN bigure 8.9 — The "tromboning- effect When the originator of the call and the mobile .ubscriher happen to be geographicalls close to each other. a "trombone" may appear in the routing of the call \\hen the home PLMN of the mobile subscriber is in another country. It consists in routing the call back and forth between two countries. leading to two international legs instead of a national tor even local I call. back, as shown in figure 8.9. Two international calls are then established where a local route would have been sufficient (this is the problem of "tromboning-). -flit: two international calls are being paid for (one by the calling party, the other by the called party). and only in a small number of cases will the calling user realise the actual situation, since he remains unaware of the location o f the called user i f the latter does not tell him explicitly. Of course. the example shown represents an extreme and fairly marginal case. I f both users realise the situation. they can both save on their hills by having the mobile user immediately end the call, and restart To conclude this long discussion, let us just recall the result: the GMSC function, though designed to be used in a versatile way. will be at least i n the first g a r s o f GSM always co-located with an MSC. and limited to the interrogation of the HLRs inside the same PLMN. 8.1.1.3. Where does the Routing Information come from? The answer to the question of where the routing information conies from is strongly linked to how subscribers are identified in GSM, as well as i n I S D N and PSTN, f o r routing purposes. W e have already encountered the MSISDN, which is the "directory number" used to call GSM subscribers. The MSISDN is part o f the E.I64 numbering plan. Also part o f this numbering plan i s the MSRN, o r Mobile Station Roaming Number. which is the routing number (see page 512) used on the second leg of an incoming call between the GMSC to visited MSC. The N1SRN is not visible to GSM users or to calling parties, but is used solely between infrastructure machines. It is not allocated permanently to a subscriber. and it is geographically integrated into the numbering plan of the fixed networks, since its purpose is for routing towards the visited MSC. A third type o f identity, introduced in Chapter 7, is the IMSI, which is used as the main subscriber key in the GSM location databases. Both MSISDN and IMSI contain an identification of the country and of the network within this country. Table 8.1 (overleaf) gives a few examples of the correspondence between the "country code" (CC) of the international telephone (or ISDN) numbers and the "mobile country code" (MCC) of the IMSI. The problem now is how the GMSC gets the MSRN to point to the N1SC where the subscriber is located. The method by which the system originally determined the location of each subscriber has been studied in detail in Chapter 7. To sum up. the HLR stores within each subscriber record some location information. including at least an address of the visited \ISC/VLR which can he used by the SS7 signalling. The HLR record may also include an MSRN usable for the routing of the second branch of the call. i f the visited MSC/VLR has provided such a roaming number when updating the location information in the HLR. In this case, the ans e r to the interrogation can he readily given. hut when the record NISISDN MCC CC MSISON > IMSI, VLR number, ... N1NC (static) HLR, Country ('l' MCC Denmark , 4 5 238 Finland' ; :7,i; 244 France 33 108 German \ .. .4n 262 hal?. 3ii Ill The Netherlands 31 204 Nor„ :o .47 242 Portugal 351 268 Spain 3 4.) 0 9 MSISDN :1/ GM89,.. •' 12( I T V MSRN MSC VLR 214 4 Snetlen 0 tiv itterlatel 46 240 41 218 1F. 44 134 Table 8. ( . 4 wresyt Indeilie between CC and MCC It r a few European countries The cousin rode Ill.') lomun a, subscribers in the telephone numbering plan is defined in Recommendation c c r r t E164. whereas the mobile country code l \IC('I used in the lira lc k digits or the I NISI is del ined in Recommendation CCITT E.212. is limited t o the visited NISC/VI.R address. the H L R has first t o interrogate (using this address) the visited NISC/VLR to get the routing information. The flow o f information is shown i n figure 8.10. When receiving the message. the visited MSC/VLR chooses the roaming number From a pool o f free numbers. and links it temporarily with the I MS1. When the call eventually reaches the visited MSC. using the roaming number as the address. the NISC can retrieve the IMS1 from its records. and can go ahead with the establishment of the call towards the mobile station. The roaming number can then be freed as soon as the call is fully established. The m o scenarios f o r the provision o r the MSRN (available continuously at the I II.R or provided by interrogation at each incoming call) triggered a long debate in the specification committees. The issue is now settled in favour of the provision o f the \ISRN on a call per call basis. though both solutions still appear in the phase I Specifications. MSRN> IMSLi (temporary) Figure 8.10 —The provision of the MSRN The HLR requires the visited location register to provide an MSRN which will he used to route an incoming call towards the correct MSC. One o f the fundamental arguments is the consumption o f directory numbers. I f the MSRN is allocated to a subscriber for the whole duration of its stay i n the VMSC. the quantity o f numbers reserved for this purpose in each MSC must be about equal to the number of subscribers ho may be registered at the same time under this MSC. By comparison, a call per call allocation requires a quantity of numbers which are about equal to the number of simultaneous call establishments by the MSC, a much smaller value. This advantage is amplified when the multi service problem is taken into account (see next section). In the following, only the call per call allocation of the MSRN is considered. 8.1.1.4. T h e Problem of Multi Service GSM is designed to provide different telecommunication services. Most o f them. such as speech. fax, o r data bearer services, can be provided through the PSTN or the ISDN. A GSM mobile station, or an CONINIUNICATION MANAGEMENT ISDN terminal. is designed so that the user can indicate' , i c h service it requires amongst the supported ones. This is not so easy for a calling party using a PSTN access line. The problem is for instance how to make the network understand that a hits call is requested instead of a speech call? In addition. even if the information is provided by a GSM or ISDN calling party, an intervening PSTN is unable to carry it. The problem exists if any part of the route is via the PSTN. 525 MSISDN > I M S I . VLR number, GSM bearer capability, ... (static) A solution exists i f the first branch does not include ;Inv analogue PS41,t—segment. [ h i s applk's i f the call is issued from GSNI o r from ISDN, and i f the (IMSC can be reached via ISDN using SS7. Then the information can he transferred up to the GNISC'. and from then on to the FILR, and thence t o the VMSC (in substance bypassing a possible passage of the second branch through the PSTN). The problem remains when the only information received by the GMSC is the culled number (the MSISDN). Two solutions o f general application have been put on the table. and were the source of intensive debates. The first solution consists in letting the sen ice he chosen by the called party. The message setting the call from the network to the mobile station does not specify the service, and the mobile station indicates it in return. This solution imposes the requirement that the service is set by the user in the mobile station /wfine the actual start of the communication. A typical scenario to send a fax to a GSM subscriber is then first to phone him (speech communication). asking him to set the mobile station so that the next call will be treated as a fax call: then hang up and re-dial to establish the fax call. This solution has minimal impact on the network. but is not very convenient f o r the users. This i s w h y a n alternative solution was proposed, consisting i n providing a GSM subscriber with as many MSISDNs as services for which he wishes to receive incoming calls (for instance a speech number and a fax number). The service can then be chosen by the calling party. by using the right number. The relationship between numbers and services is held in the HLR. The next issue is to convey the information t o the VMSC. This i s simply done i n the procedure used to get the roaming number, which allows once again to bypass the networks intervening between the GMSC and the MSC/VLR. The HLR sends the reference of the requested service to the MSC, which stores it against the provided roaming number. When eventually a call with this roaming number reaches the VMSC, it proceeds with the call establishment for the requested service. This scheme is shown in figure 8.11. Its only drawback is the consumption of directory numbers for the MSISDNs, which can he unacceptable in some countries. On the other hand, it enables the provision of a service no worse than that which PSTN subscribers are familiar with. IP•7 111 / 4 GMSC MSISDN i t . M 1 1 S R N MSC VLR MSRN > GSM bearer capability (temporary) Figure 8.1 I - Routing of service information When an incoming call arrives from the PSTN. the only way to convey the information on the service (based on the dialled directory number) up to the visited MSC/VLR is to carry a service profile (bearer capability) between the HLR and visited MSC/VLR. This service profile is stored in the HLR in association with each MSISDN. The two schemes are provided for and both must be supported by all NISC/VLRs. The choice of the scheme is done by the Home PLMN of the called party. When no service information is provided by the HLR, the MSC/VLR can either choose the service or let the mobile station decide. The first approach i s acceptable only i n the case when the subscription is limited to one service. 8.1.1.5. D a t a Services Calls originating from a PSPDN or a CSPDN and directed towards a GSM subscriber raise specific problems. The most important issue is related to numbering. While PSTN. ISDN and GSM use E.164 numbers, public data networks (PDNsi use a different numbering plan. specified by •('ITT Rec. N. 121. GSM and ISDN :dims f o r mobile originatitrE a i l s to s' addressed b y X.121 numbers. hut t s u p p o r t o f E.164 numbers by '1)1Ns w i l l become mandatory i n a f e w years (this is planned for in the ‘Cginning o f 1997). Several cases can he envisaged. The GSM subscriber may be given an N. I21 number. to he used to :ill h i m from PDNs. This number most he allocated so that the PDNs •.ill route calls towards GSM. This can be obtained by defining a GSM .MN as a "PDN". in substance by allocating a data nen\ ork code to i t •S \1 PI a c T l i n i r s t part o f the N. 12 I number of its subscribers. l i e call is routed through PDNs directly t o an interworking function of he target P L M N . which acts as a GMSC. Roaming may he dealt with by 'sing ISDN/PSTN routes. It should be noted however that i f the called itiM user is served by his home PI AIN a cireffil can he set up without milk) modems. whereas such modems may he necessary i f the roaming equirements force the call to go via the PSTN. A variant consists in defining i n the PDNs the PLMN as part of S I)N. The data network code then refers to the ISDN. and the calls are Aimed first t o the ISDN (and more precisely to a packet handler in the •ase o f a call f r o m a PSPDN I. which has then to translate the X.I21 itimher into an E.164 number for routing the call to any GMSC for the arget P L M N and then to the suitable MSC/IWF. Roaming is dealt with iiirmally. Whether audio modems are used o r not is determined by the •SDN. This determination may he difficult when the user is roaming. and he default solution is to use audio modems. Another case is when the PDN does support E.164 numbers. The gating depends on the capacity o f the PDN to determine that the number efers to a GSM user. The basic assumption is that it does not, and the :all is routed to ISDN. which will then route it to a suitable GMSC. If the -can further analyse the number. the route can be routed through (3DNs directly to an IWF which may act as a GMSC. It i s not foreseen t o have an H L R interrogating function in the middle o f the PDN leg. This means that tromboning routes will not be ivoided. This is not too much o f a problem. as we have seen that this ;couid save only a part of the route within the home country. 5.1.1.6. T h e Impact of Call Forwarding The category o f supplementary s e e ices which have the greatest influence on call routing are the forwarding facilities. There are various reasons w h i c h can lead t h e infrastructure t o forward a call: and the process differs accordingly. There are Iwo main eases: • either the HLR is• e to decide to forward the call, in which case it sends hack the forwarded-to number to the GMSC as a result of the interrogation: • o r forwarding can only be decided by the visited MSC/VLR in charge of the subscriber. The first case is obviously more efficient for routing, and should be favoured as much as possible. Let us now study the reasons to forward a call and see how they fit into these two broad categories. In the case of an unconditional call forwarding (CFU), the HLR. knowing the status of the supplementary services, has no difficulty to do it h' itself. However, a call forwarding done because the end party is found to be busy (Call Forwarding on Busy. CFR) can only be performed lw the VMS('/VLR in the present slate of the Specifications. The route is first established toward the VMSC/VLR, and only then to the forwardedto party. Two types of conditional call forwarding exist: Call Forwarding on Not Reachable (CFNRy) and Call Forwarding on No Reply (CFNRy). The first case can be treated by the network in several ways, according to the method by which knowledge that the mobile station is not reachable ---is-tthtained. In some cases, for instance because the mobile station last tried to register from a geographical area where the subscriber is not entitled service by subscription, the HLR knows o f the situation (see Chapter 7). and performs the forwarding itself. In other cases the situation may he known only after an effective (and unsuccessful) attempt to contact the mobile station over the radio path. This can be done only by the MSC/VLR. Moreover this is done after the route from the GMSC to the MSC/VLR has been established, and then the forwarding is done by the MSC/VLR. Another case,for Call Forwarding on Not Reachable happens when the mobile station has been able to indicate that i t was going to be switched o f f (this i s the " I M S I detach" procedure, described i n Chapter 7). or i f the subscriber has been inactive for a long time and "implicit detach" i s implemented b y t h e visited P L M N . T h e "unreachable" indication is stored in the VLR, but not in the HLR. It will become possible in phase 2 to send back this information to the HLR during the interrogation procedure. but this is not the case in phase I and forwarding must also in this case be carried out by the MSC/VLR. Call Forwarding on No Reply is triggered when the mobile station has been successfully reached on the radio path, but when no answer from the user has been received after some time. CFNRy is therefore a l Ly s triggered by the VMSC. 3Z.7 Whether the forwarding i s done b y t h e H L R o r by 1). v i s i t e d MSC'/VLR i s o f no importance f o r the c a l l i n g part•. s‘hose lila is not .11Teeted by the actual routing o f the call beyond the CINISC'. hut it can make a big difference for the called party. w h o is billed for the remainder tl' the route. The most serious case is 'when the visited NISC of the called party i s 111 a comm.> o t h e r than t h e h o m e country, and when the ww airded-to user is in the home country. The difference is then between .1 nat tonal. or even local call. and two international calls! Users who want sa ve on their bills ‘s ill probably prefer t o activate 'neonditional Call fors`•:u'ding in this case. 8.1 .2. ARCHITE(TIIRE So far (Ac ha se presented in general terms the functions for call inailagement. I n t h e 1011owing w e w i l l describe more precisely the signalling procedures. and for this objective i t is necessary to indicate f i r s t t h e different entities and protocols w h i c h are imolvcd i n these procedures. The main actors of Call Control are I he user and the mobile station •In o n e side, and the Network and Switching Sub-system !NSS) and the ext tonal network on the other. The functional entities of the NSS which are involved in Call Control include the NI SC/VLR and M T. the GMSC ant. I the M i t The main one is the M S C / N t l t , which deals with all the cot thimnications. mobile terminating or on o f which one end is a ISNI m o b i l e station w i t h i n i t s r a d i o coverage. T h e NISC i s the unection point betv.een the mobile stat ions on one side. through the IISS, and the external networks on the other. As such the MSC is the Edion \\Acre the call control procedures held with the mobile station interskcirk I n n those held w i t h the external network. I n addition. for NI obile Te t r i n a t i n g calls. the H L R a n d the GMSC intervene. They -)inniunicate with each other and w i t h the VLR/MSC through the SS7 network. 11! innalling protocols on the C a l l Control scene are shown in igure 8. I 20'hey include the following: • t h e M S -MSC protocol enables user requests to be conveyed between t h e m o b i l e station a n d t h e network and service provision to he co-ordinated het ween them: this protocol will be referred to as R111,3-CC. It acts as a relay of the man-machine protocol held between the user and the mobile station: • t h e GMSC-111.12 protocol enables the interrogation of the IILR by the GMSC' t o get routing information f o r incoming call establishment: this protocol t . referred to as m A r c : HLR / MAP,'D RIL3-CC \MAP/C GMSC / MSC V L R Figure 8.12 — Prot kill. in the CC domain The key infrastructure emit f o r Call Control is the MSC. which is in contact ‘‘ ith the mobile station. Rut torso of the complexity comes from the routing of calls toward mobile stations. o additional protocols ben.eett the Gateway MSC and HLR, as well as between the H I S and MSC (the latter appears also in the MM domain). • t h e M A N ) protocol between the HLR and the MSC/VLR, which is mainly a protocol for Mobility Management (and as such has already been cited in Chapter 7). also serves to convey call-related information for incoming calls. The protocols with the external world must be added to these internal GSNI protocols: the user-MS protocol on one side, protocols on the interface with external networks on the other side. On the mobile station side. the man-machine interface ( M M I ) between the user and the mobile station has already been mentioned. It is to a large extent not specified by the Specifications, leaving each mobile station manufacturer free to develop user-friendly means for this purpose. On the MSC side, the protocols with external networks are specific to each of these networks. The functionalities are always more or less the same. hut variations exist according to the type of network, and also to the specific implementation adopted by each country. The basic examples are the TUP !Telephone User Part) and ISUP (ISDN User Part). o r Belt national v a r i a t i o n s thereof. w h i c h a r e t h e standard ( 1 . . i t c a l l management protocols used in an 557 environment. The TUP or ISUP is in direct interworking with the RIL.3-CC protocol in the MSC. Similarly. the GMSC interfaces with the PSTN o r ISDN for which it performs its gateway r o l e . t h r o u g h the call management protocol specific t o the relevant e x t e r n a l network. I n t h e description o f procedures which follows. the I S IA' will he taken as an example for interworking. A last a tea where relevant signalling protocols exist is between the mohtle-srmirrrr mid the-IWI: or the other terminal. For most data services. the establishinent o f the transmission circuit is but the first phase of the call establishment. Additional negotiations a n d synchronisation are performed " i n -hand", that i s t o say using the establishment circuit to convey the tnessages necessary f o r the dialogue. This phase is of little concern to t h e intervening entities. and hi GSM. but some aspects have an impact on the. I W F. so we will briellx address these protocols. 8.1.3. T H E MOBILE ORIGINATING CALL ESTABLISHMENT PROCEDURE As seen by the calling user. the establishment of a communication follows a nu tuber of steps. which he perceives through displayed or audio information. T h e rough sequencing for basic telephony is well known by ❑nybody. A in originating call starts by the user lifting the receiver, an action w h i c h the network answers w i t h some tone. Then the called number is keyed in. Once the last digit is entered. there happens in some cases n o t h i n g for some time: in other cases. a waiting tone can be heard. Eventually, a n answer comes. I t can be a busy tone. indicating that the called p a r t y i s already engaged i n a communication. an announcement stating for vchich reason the network was unable to fulfil the request (e.g., congestion, o r non-existent number). or. sometimes. a tone indicating that the called p a r t y is being alerted. Then. in the last case. it may happen after a w h i l e that the other party lifts the receiver. and this is perceived by the speech p a t h being connected. thus allowing the person to person discussion t o take place. Cell v i m telephony does not introduce many modifications to this basic scheme. The main distinction is that the number is keyed in before establishing the contact with the network. It is in addition displayed. and a very u s e f u l consequence is that i t can be modified i f need he before transmissi on. Another difference. applicable to multi-service networks such as G S M . is that some additional information may be exchanged COMMUNICATION MANAGEMENT 33' between the user and the riLiwork at the beginning of the,call, such as the type of service or in the future the type of channel. In most of the cases, these issues will he dealt with automatically by the mobile station, with default or implicit values. For all these interactions, the mobile station intervenes between the user. with which i t exchanges messages according to a man-machine protocol. and the infrastructure. with which it exchanges information by electronic means. As far as Call Control is concerned, these two flows of irdimnation are for the important parts in a one to one correspondence with each other. The mobile station acts as a protocol translator. I t receives orders from the user in the form of key pressing for instance, and translates them into signalling messages for the network according to the RIL3-CC protocol. In the other direction, the MSC provides its answers or indication to the mobile station in the form of RIL3-CC messages, and the mobile station translates them into signals understandable by the user, such as audio signals (tones) or visual signals (lights, alphanumerical This modern approach, which is also the one of the ISDN (from which the RIL3-CC protocol has been derived), is different from the one still used in the PSTN. In the PSTN, all the information provided to the user are in the form o f audio signals (tones, announcements), and are generated b y the network. There i s no digital message to be found coming from the network to the telephone set. Inside the PSTN, the transmission of the information ultimately destined to a user may be done in the guise of messages (digitally encoded), in which case the translation in tones or announcements is done at the end, by the switch in charge of the user line. Yet there are still old machines not supporting this scheme, and then the tones and announcements can come from switches far away 'rom the user, the role o f the local switch being limited t o their transmission, a s d u r i n g t h e actual communication phase. T h e consequence for GSM (and f o r ISDN) is that an escape mechanism ("progress" information) must be provided so that the dialogue between the user and the network can be done in the old fashioned way. There is yet another approach, which is specified (besides the "functional" protocol described above) in the ISDN terminal to network protocol for simple terminals, and is called the "stimulus" mode. In this mode, the network has the full detailed control of the signals (display or audio) provided to the terminal, and instructs it through simple messages, directly giving the signals to be provided to the user. This approach is not used in GSM. but some traces of it can be found, in the guise o f the When the M S C receives the setup message. it analyses r e q u e s t . and cheeks whether it can accept it. Whether it is accepted deF....ods on the capacity o f the NISC/VI.R t o provide this service t i n a compatible way with the mobile station capacity). on Ansgar's subscription characteristics (this i s determined l o c a l l y thanks t o the subscriber information record seal by the I ILR (luring location updating. and stored in the MSC/V1.10. and on the availability o f resources Omen, orking devices. a free circuit with the external n e t w o r k . etc.). I f sonic o f these checks fail, the call establishment i s aborted b ) sending a n Rit.3-n.' RELEASE COUPLET I essage t o t h e n i o h i lc station. before t h e release 11t t h e l o w e r layer connections returning the mobile station to the idle mode. I f everything is all r i g h t , the M S C o n o n e hand starts t h e establishment through the network ( f o r instance b y sending a n I s c P INITIAL ADDRESS message ( IA NI) in the case o f a communication w i t h the ISDN). and on the other hand sends an RII.3-cc ( ' A l t , itRociTors:() message to the mobile station, which indicates simply that the request has passed the MSC tests. and that the M S C is proceeding with it. More pragmatically. it tells Ansgar ( i f so indicated by some signal on the man-machine interface) to he patient. Sooner o r later. the M S C w i l l receive f r o m the external w o r l d a report o f the requested call establishment. as seen by the switch in charge of the called party. Such a report may indicate that Remi is being alerted (alerting indication). or that the call establishment is aborted (call release indication) because i t f a i l e d f o r sonic reason (congestion. o r because Rem' i s already busy. o r not reachable. etc.). I n the case o f (SUP. the report o f a Isitccessful alerting takes t h e f o r m o f a n ISUP ADDRESS c(MPLETE message ( A C M ) . a n d a f a i l u r e i s indicated b y a n ISUP ItELNASE message. T h e M S C r e a c t s respectively b y p a s s i n g t h e information t o A n s g a r t R i t S - C c ALERTING message. translated f o r instance to an alerting tone by the mobile station). or by aborting the call establishment. I f ❑n Rii.3-cC ('ALL PROCEEDING message has been sent before. the abortion is done by sending an RiL3-CC DISCONNECT message. which w i l l be answered b y the mobile station with an RIL3-CC RELEASE message. i t s e l f acknowledged b y t h e M S C w i t h a n RIL3-CC RELEASE compLETE message (explanations on these exchanges will be found in the section concerned w i t h c a l l release). O n l y t h e n c a n t h e l o w e r layer connections he released. i f not used for some other context. Some t i m e c a n t h e n elapse before t h e e n d -to-end answer, i.e., before the acceptance o f Remi ( l i f t i n g the receiver in the case o f plain telephony). A t this stage the mobile station (and Ansgar) is still waiting for the result o f the request. The ball is on R e m r s side. who may answer, or not. In the case o f a no answer. the call is aborted by the network after some time (say 3 minus p r o v i d e d Ansgar has not decided to abort it himself. usually by pressing an "end" key. The acceptance o f the call by Remi leads t o an ISUP ANSWER message being received b y the MSC. When this happens. the call is connected-through, that is to say that the transmission path i s completed between the t w o end users, and this is indicated to Ansgar through an RIL3-CC CONNECT message. The mobile station reacts first by stopping the alerting indication i f any, secondly by answering t h e n e t w o r k w i t h a n RIL3-CC CONNECT ACKNOWLEDGE message. and thirdly by connecting the circuit transmission on the radio path w i t h the suitable terminal. I n the case o f speech, this consists o f completing the speech path t o the microphone and to the loudspeaker. The c a l l then enters t h e connected phase ( " i s connected" i n short), charging starts a n d effective bi-directional transmission i s provided between the two users in the case o f speech. or between the terminals, or the terminal on the mobile side and the I W F in the cases o f data calls. In the last vase. the establishment procedure goes on in-band. Let us now sec some o f the variants from this basic scenario. 8.1.3.2. Automatic Answering When the called party is a machine, which is often the case for data calls, some of the steps can he merged.. For instance. the connect-through can happen without any alerting indication. In some even more expedient eases. the connect-through can happen as an immediate answer to the request. In the latter case. the R)L3-CC CONNECT message is sent directly as an answer to the RIL3-CC SETUP message. In the former case, it directly follows the RIL3-CC CALI. PROCEEDING message. 8.1.3.3. Entered Number One o f t h e p r o b l e m s s p e c i f i c t o a n international m o b i l e communication network like GSM. where each user can establish calls from different countries, is the format o f the called number. A s PSTN users. w e usually k n o w three basic types o f formats: a local format; which refers to one destination only in some regional area; an interurban or national format. used when the destination is in the same country but in a different region. and an international format used to call abroad. The regional area (area code: 123) for instance i n t h e Frencl ' S T N , k e y i n g a n international c a l l w i t h country code 33 (France) from a PSTN telephone set results in a call to the international directory enquiries for the country whose code followed 33. A different treatment i s required f r o m the M S C . T h i s m i n i m u m specification enables the user to enter the number of a specific destination totally independently from his location: it is therefore advisable for GSM subscribers who travel abroad to always store their abbreviated dialling numbers i n this format (e.g., + 4 4 701 1234...), instead o f a national format. e As already mentioned, this format is only one among many that can he supported by the RIL3-CC protocol. The translation between the number in the man-machine protocol (the number keyed by the user), and the n u m b e r sent i n t h e R I L 3 - C C message i s n o t specified b y t h e Specific-at/mu, n o r is the w a y the M S C must interpret the numbers i t receives. T h i s i s unfortunate, because what the user has t o k e y does depend on the terminal as well as on the P L M N operator. Most probably, sonic standardisation w i l l appear. Though other schemes cannot at this date be precluded. the most widespread format for user dialling (besides the " + " k e y format) seems t o b e t h e national format o f the visited country. usually without prefixes. t o h e used f o r numbers w i t h i n the country. 54321 national area (country code: 1) 0 123 54321 worldwide area 00 1 1 2 3 54321 Figure 5.14 D i a l l i n g number formats Depending on the relati‘c positions of the calling and called users. I'S'I'N numbers can usually be entered in three different way,. which are ‘listinguished bs different prefixes. 8.1.3.4. O f f Air Call Setup In the basic call set-up scenario. the transmission path between the \ISC and the MS is fully operational for the requested service before the MSC is aware that both parties are ready to converse. The call can be connected-through immediately. A variant is possible, called O f f Air Call Setup (OACSU). where the allocation o f the suitable radio resource is delayed as much as possible, in an attempt to save radio resources (see Chapter 6). In this case, the MSC has first to request the BSS to establish the c i r c u i t connection w i t h t h e r i g h t transmission m o d e , b e f o r e connecting t h e communication. F o r t h e called party, t h e instant o f connection is not delayed, but the connection may be temporarily routed to an announcement machine until the actual connect-through happens. I'utinetiuit i s done b y the use o f prefixes. Figure 8.14 shows the three (urinals i n a n example using the standardised (hut not y e t universally .mplied) prefixes 0 and 00. The way in which a number is dialled in the PSTN then depends on We location o f the calling party. T h i s is quite unfortunate i n a mobile ,livironment. To avoid the difficult task o f requiring the user to know in nich telephone area he currently is. and how numbers are to be entered, the issue must be treated differently i n a cellular network. T h e ISDN protocol, a n d Ra.3-C.C. support m a n y different n u m b e r formats, b y qualifving the digit string by two indicators. the "type o f number ( T O N ) nd t h e " N u m b e r i n g P l a n I d e n t i f i c a t ion ( N PI H o w e v e r . t h e Sjn'cifica/nis are not very informative on the subject o f what the user h e n t e r . There is however one mechanism that must mandatorily be sus to upported, concerning thee international format. I n order not t o ask the user to know all the prefixes in all the countries where he wishes to roam. an international number can he entered by pressing the " + - key followed by the country code, etc. Moreover. a switch must be able t o correctly treat a n u m b e r presented a s international. e v e n i f i t addresses a destination in the same country. This behaviour is by no means' obvious: 1 I 8.1.3.5. Progress The basic set-up scenario assumes that signalling messages are received f r o m t h e originating network. N o w , t h i s i s n o t a l w a y s guaranteed. since an important proportion o f the PSTN equipment are not using the advanced protocols. I t may then happen that the indications THU CISM 5' S E L L meet-fling: the progress o f t h e c a l l c o m e a s :nano t . .y.s o r mouncements. This case i s foreseen and coped with i n the ISUP coliwoi. and the MSC will know when the call establishment must go on I the old l'ashioned way. The MSC reacts by performing a kind of oneas ihrough_connection i n advance to the normal point. so that the peed) path from the called party to the (ISM user is established and the me. and announcements can be heard. (Nete that this is not easy with ato services except a s alternate speech/data services.) T h i s early .innection is commanded by way of the RH .3-cc mu( ;HUSS message, or be Ivo icumss information element included in messages such as RIL3-CC •\I I POLWITDINIG or Al ER"rimi. depending on the moment it is known ,at the call is not ISDN-compatible all the way. In addition. i f the RIO1,Kixitalss message i s used. there are no more messages for call astahlishment. except the Rii.34'(' coNNE(t message used to indicate the Iwo-way through-connection. SI .3.6. Additional In-Band Establishment for Data Services In the case o f data calls. the connect-through is not sufficient to a l l ) ' user-to-user information exchange. and n o t even terminal-toterminal exchanges w i t h some cases o f interworking. The connectthrough can be understood as the establishment of the lower protocols in the transmission plane. but other data specific protocols must also be initiated. The next step is the confirmation of the connect-through between the terminal adaptation functions in the mobile station (the TAF) and the (WV T h e Specifications d o n o t state clearly w h e n t h e V . I 10 synchronisation pattern is established between the rate adaptor (TRAU) and the IWF. It can he imagined that it is when the transmission path is established between the IWF and the mobile station, that is to say before the connect messages exchange. After this exchange a small procedure lakes place between the TAF and IWF. as follows. First one end sends data and auxiliary information all set to "I's". and then the other answers in the same way. When this is completed. the transmission path is used by higher layer protocols. for instance to establish a link layer in the case of an N.25 connection. This procedure raises some problems. There is no clear indication i n the phase I Specifications on what i s transported before the connect-through. that is to say just before the above mentioned cxample. Unfortunately it can well he an all " I ' s - pattern! The second point is that this detail in the Specifications ruins the possibility of a interworking function between GSM and ISDN for T connections the synchronisation procedure described above does not exist in ISDN). COMMUNICATION MANAGEMENT 5 3 9 As egplained in Chapter 3, we possibility o f such a "null" IWF with [SDPwas an argument for the peculiar management of the El, E2 and E3 bits in the V.110 frame. Once the existence o f a V. ! 11)-like connection is confirmed, the RLP protocol MUM be initialised in the NT cases. This is done by an exchange of SABM and UA frames, as explained in Chapter 5. The next step involves the modern control signals. and is done in the same way as between a terminal and a modem. The next. and last steps, depend on the service and are normally carried between the terminals themselves. They correspond t o the establishment of the upper layer protocols for transmission, or even to further call establishment as with the double-numbering cases we have seen for PSPDN or CSPDN access. They do not generally affect GSM, but there are some exceptions. For instance, the IWF is involved in the procedure in the case of facsimile, which consists mainly in a negotiation concerning various characteristics of the transmission, and in particular the modem speed. Another example is for dedicated PAD or packet access. where the establishment procedure through the PSPDN is taken bare of by the IWF. It should be noted that part o f the procedures presented in the previous paragraphs duplicate the basic call establishment, or could be supported by it. The trend is to limit them. at least f o r the user, as exemplified by the introduction of the dedicated PAD and packet access services. 8.1.4. THE MOBILE TERMINATING CALL ESTABLISHMENT PROCEDURE Let us now consider the case of Mobile Terminating calls in the same way as we just did for Mobile Originating calls, i.e., from the point of view of the user to network interface. At this stage, the call has just reached the Visited MSC and we are concerned with what happens from this moment on. In usual telephony, the interaction between a called user and his handset is very simple: the latter rings, the former then lifts the receiver and communication ensues. I n cellular telephony, the behaviour i s tiasically the same. Some sophistication appears if additional facilities are invoked. For instance the directory number of the calling party may be displayed. as i n ISDN. allowing the called party the option o f not responding to unwanted calls. _) From a signalling point of view, a Mobile Terminating ea . a c h e s ie visited MSC through one of its interfaces with external networks. I f ;III' is used on this interface, this event corresponds to the reception of a ISUP INITIAL A l M VSS message ( I A Nil. From the contents o f this less:we. and from the data stored previously (during the interrogation base a n d which can he linked with the incoming call through the gamin;_ number. the NISC/VL.R can derive all the information it needs, tich as the I MSI. the requested type of service.... 1.ei us see what most ;hen happens :diem ands. and then study the variants of a call from an SI)N user called Carlo to a GSM subscriber called Jan. The signalling Achatigji&responding to the basic successful mobile terminating call ..stablishment is shown in figure K.I5. If Jan is not known to be already engaged in a communication, the next step consists in "paging t h e mobile station. that is to say in a few words, to find whether the mobile station is actively in coverage. and to ask it to establish a signalling connection with the MSC'. When this and other ancillary tasks are done (see Chapters 6 and 7). an 811.3-CC SETUP message is sent to the mobile station. indicating many details concerning the call, which include the type o f service required and. i f applicable. Carlo's directory number. The mobile station checks if it can deal with the type of service. and i f not rejects the setup by an RIL3-CC RELEASE coMPLETE message. Otherwise. the mobile station answers with an RIL3cc CALL coismitatFu message. and starts to alert Jan. by a display or a ringing tone. The RiI.3-Cc CA1.1. CONFIRMED message is also the vehicle for the choice of parameters which can be decided upon by the mobile station. Two eases are of interest. First there is the future case of mobile stations able to deal with both half rate or full rate channels for the same service: the message then conveys the choice of the mobile station. The other case is when it is the called party which chooses the type of service. We have seen that this may he the case with a call originating from the PSTN, and when a single directory number is allocated to a mobile user who desires access to different services (speech and fax for instance). Then the R1L3SyrLIP message does not contain any service indication (the MSC knows by the subscription record that the subscriber may receive calls for different services). and this information is provided by the mobile station in the RiL3-CC CAI.I. CONFIRMED message. The following step is the start of user alerting. There is however a problem here. The allocated radio channel at this stage is not necessarily suitable for the communication. I f the mobile station is alerted, Jan may answer (this is the very purpose of alerting!), and i f at this moment the MSC VLR SETUP CALL CONFIRMED ringing ALERTING user answer: CON E Figure 8.15 — Basic MT Call Establishment sequence After the negotiation phase (SEICP, CALL CONFIRM). the mobile station sends back the ALERTING message while ringing the user. The sending of the CONNEC'T message happens only once the called user has answered the call, triggering full connect-through. channel is still not suitable, the communication cannot be established readily and Jan will not be happy. On the other hand, it is useful, for saving spectrum, to delay the allocation o f the full rate traffic channel (TACH/F) as much as possible: this is the notion of OACSU, a channel allocation strategy which we already encountered with mobile originating call establishment. Two schemes are possible, and are left for choice by the operator. Either the start o f alerting is postponed until a suitable channel is allocated (early assignment), or alerting is started immediately so as to delay the allocation of a suitable channel (off-air call set-up). The mobile station must know which of the two schemes is used, in order to decide whether to start user alerting or not. This information is conveyed by an indication i n the RIL3-CC SETUP message (a SIGNAL infomiation element, hijacked from its customary usage in ISDN, where it is used for the operation o f the stimulus mode). When the SIGNAL information element is absent, the mobile station only alerts the user once a suitable channel i s allocated. Conversely, t h e presence o f this inf9rmation element indicates an o ff -air call set-up. where the suitable I I e l . . . . . . . • U M : M U N K : A I ION N I A N A U L M L N I JYJ instnission path will only be established after the user has ace', ( 2 d the II (and hence after his being alerted). 8.1.4.2. Automatic Ans‘vering Once the appropriate alerting signal has been started. the mobile at ion sends all Hit.3-cr m e s s a g e . A t the MSC. the alerting art.( k reflected towards the network from u. Inch the call collies. in the Ise of ISM) with an 1St I' ADDRESS c o m i t r i T message (ACM). When IT-air call set-tip is used. this message is sent as soon as the 811.3-ct vi I (4( rsa-IPmili message is received from the mobile station: otherwise. is sent when the A t i l t i (.; message is received from the mbi le station. The next step is the acceptance of the call by the GSM user, which ',livens in the case or speech when Jan lifts the receiver or presses some .ev. This action is translated into an 10.3-cr coNNErr message. At the eceipl of tins message. the MSC joins the network transmission path and he access path. and the end-to-end transmission may start. In the case of a data call involving a separate Terminal Equipment ('FE) connected to the Mobile Termination (MT), the RIO-CC SETUP message is echoed on the interface between the M T and the TE. This interface is not specified within GSM. but uses existing standards such as the V.25bis or the X.21 access methods. The RIL3-CC messages are then relayed by signals on this interface. Typically, in this case, the incoming call signal is answered immediately by the TE, requesting the connection of the line. This is an example of automatic answering, which can also of course h e implemented i n a n integrated mobile station. I n these situations, the mobile station is allowed to skip the RIL3-CC ALERTING message and i t directly sends the Rit.3-Cc CONNECT message t o the network alter RII.3-CC CALI. CONFIRMED message. Let us see now some variants from this general scheme. walking through the procedure ruin the start onwards. 8.1.4.3. O t h e r Points 8.1.4.1. C a l l Waiting When the MSC receives an incoming call from the external world. it checks whether by i l l chance the called user is already engaged in another communication. I f the answer is yes and the subscriber has not activated the Call Waiting facility (CW). or i f for any other reason (e.g., too many calls on hold already) call waiting cannot be applied, the request is rejected or forwarded. Conversely, i f Call Waiting is active, the MSC proceeds directly with the setting up of the call. using the signalling means which already exist between the MSC and the MS. The paging phase and the establishment o f the lower layers between the mobile station and the MSC can then he skipped. and the procedure begins with the Rit.34x7 SETUP message. A s explained i n Chapter 5. distinct call control transactions are identified b y a Transaction Identifier ( T I ) inclUded i n each itti.3-CC message. This number enables the mobile station t o distinguish t h e messages pertaining t o different communications. Once having received the RIL3-CC SETUP message for this new call, the mobile station behaves at the call control level exactly as for the- first call. and it is up to the user to release one of the calls or jug;!le between them if Call Hold is available. As for a mobile originating call set-up, i t may happen that the originating network does n o t indicate t h e progress o f t h e c a l l establishment through messages, but by in-band tones. The same escape mechanisms we have seen for Mobile Originating calls can be used for Mobile Terminating calls. In this case, the PROGRESS information element is included i n the RIL3-CC SETUP message. This commands an early connect-through, but does not impact otherwise the running of the call —establishment procedures. Data calls need additional in-band establishment procedures, as for mobile originating calls, and there are little differences, apart from the point that most procedures are initiated from the network side rather than from the mobile station side. 8.1.5. THE INTERROGATION PROCEDURES As studied in detail in previous sections o f this chapter, a call toward a GSM subscriber must go through a complex routing procedure, involving the interrogation o f the HLR, before i t reaches the visited MSC. There is no such equivalent for mobile originating calls. We have studied the general issue of MT call routing. Let us now describe in detail the procedures involved when a call reaches a GMSC, taking as usual the COMMUNICATION MANAGEMENT 5 4 5 ;4-1 iSUP message names as examples lbr the signalling interworKmg. The nasie flow is shown in figure 8.16. The i n e o r n i n g I S M ' INITIAL ADDRESS Message contains at least the i s i s m o f the called GSM subscriber, and it can. in addition, include the type o f required service i f the call was issued in the ISDN. The °Nis(' derives f r o m t h e M S I S D N a n SS7 identification o f the corresponding 1-11.12. and sends it a \IAN(' Slink) ROUTING INFORMATION tuesc,ige. containing the MSISDN and the service indication if available. Ai Ore •re713ilit of this message. the 111.12 examines the data record of the subscriber. and takes different courses of action. depending on what i t finds in the record. I t may answer the CINISC directly in the following cases: • t h e call cannot be routed to a destination: this situation may happen when, e.g.. the subscriber has not paid his bill and is temporarily suspended (this i s called "operator determined barring")• or when the subscriber is known not to he reachable but has not activated forwarding: r - HLR 4 ▪ 4 , 0 0 Is 0 6t. < c • t‘'‘ .5) • t h e call cannot be delivered because the subscriber has activated the Barring o f All Incoming Calls (BAIC), or the Barring of Incoming Calls when Roaming (BIC-roam) and he is known to lie roaming abroad: • t h e HLR knows that the call must be forwarded (e.g., when Call Forwarding Unconditional is active): • a re-routing number (MSRN) is readily available at the HLR; as explained earlier, this possibility was left in some Specifications hut is not used in practice. and will disappear in phase 2. A negative answer from the HLR comes back to the GMSC with an error message to the MAP/C SEND ROUTING INFORMATION message, whilst a positive answer includes the forwarded-to number or the MSRN in a NIAP/C SEND ROUTING INFORMATION RESULT message. In the alternative and more usual scenario, the HLR knows only some part o f the identification (SS7 address o r global title) o f the MSC/VLR visited by the subscriber. To get the MSRN, the HLR sends a mAP/D PROVIDE ROAMING NUMBER message to the MSC/VLR. This message contains a variety o f information, including the IMSI o f the subscriber. and i n th e case o f multi-numbering, the G S M bearer capability needed for the required service. I f the original JAM message contained service information, the high layer part of it (relevant for end terminals) can also be transported to the GMSC through the MAP/D PROVIDE ROAMING NUMBER message. The answer o f the VMSC, i f positive. may contain a forwarded-to number (if the MSC/VLR knows already that the subscriber cannot be reached), but more often an MSRN. This i s i n c l u d e d i n a M A P / D PROVIDE ROAMING N U M B E R R E S U LT message. The receipt of this message by the MLR, is translated into a mAP/c SEND ROUTING INFORMATION RESULT message sent t o w a r d s t h e GMSC. which can then proceed with the call establishment towards the actual destination. GMSC 2_4.14 !?‹. IAM 6 [ N I MSC VLR Figure S.16— Routing or an NIT call from GMSC to VMSC The interrogation of the 111.8 by the CINISC. which usually triggers the request of a roaming number from I ILR to VLR, provides the ONISC with a rotilitW number used in nillire m e s s a g e s . the flow of identities in the‘e inc..ages n in figure S. W. and the exchange also enables Ihe muster of sea ice inhumation. a. shown in figure S. I 1. when the origin:1(1ml or trun,it net‘t olio. are not able to generate or transport it. 8.1.6. CALL RELEASE We have seen how communications are routed and established. Means to end them must be provided. We will now study the signalling means associated with these call releases, which are not specific to GSM. A communication can be stopped at any moment by one of the two users. .A user will indicate that he wishes t o terminate the call b y replacing the telephone receiver, or by pressing an "end" key, etc. In the case of a GSM mobile station, such an action is translated by the mobile station into an RiL3-CC DISCONNECT message (see figure 8.17). The result THE GSNI SYSTENI C O W J I C A T I O N MANAGEMENT s that the call is "disconnected-. in the sense that the other party is unified of the complete termination of the call (the MSC sends an iSuP stEl.l(ASE message to achieve this in the case o f an ISDN call) and the :nil-to-end connection i s terminated. The call i s however not fully released at this stage. The local context between the MSC' and the mobile station is kept. enabling the completion o f 'side tasks. such as charge indication. When the MSC determines that the call has no more reason to exist. it sends all Pit..;-ct. m e s s a g e to 1k. mobile station. which ans‘‘ers back with an Pit.3-cc PidriAsk comPt.FTE message. Only then aro the..laieLes....ctuinections released (unless they are used for something else in parallel) and the mobile station returns to the idle mode. with a cause field giving some more explanation to cope with abnormal termination. The mobile station should answer by an RIL3-CC RELEASE message. acknowledged b y an Rii..3-CC RELEASE COMPLETE message from the network. Between MS and MSC, the procedure is symmetrical with the one used when the release is initiated from the mobile station side. As in the case of call establishment, this procedure is not suitable for the situation when t h e PSTN i s involved. Explanations f o r t h e disconnection are then given it' needed as tones or announcements. To cope with this case, the RIL3-CC DISCONNECT message may include a PROGRESS indicator. in which case the mobile station may keep the audio connection, and the release will be completed afterwards at the network initiative. In this case the DISCONNECT and the RELEASE messages both come from the network. In f a i t the phase I procedure described i n the Specifications requires the MS(' to send the RII.3-Cr RELEASE message straight away after reception o f the itii.3-CC DiscoNNECE message. This is consistent with the fact that there exists in phase I n o possibility for the user to exchange signalling data during the period between disconnection and release. The phase I release procedure could then have been designed in a simpler way, e.g.. using two messages o n l y The choice o f a more complex three-way handshake procedure was done in order not to block future possibilities. The case of a release triggered by the other user is done in a similar way: the MSC receives an istIP RELEASE message. which causes the sending of an RIL3-CC DISCONNECT message toward the mobile station, 5 4 7 -8:1,7. I N -CALL FUNCTIONS We have seen so far how calls are routed, established and released. Between the establishment and the release phase, the call is deemed "active". During this period, several events may happen which require some additional functions on the Call Control level: for instance, the type of service may change within the same call, the user may want to juggle between several calls or may want to press some keys to send control signals to the other party (e.g., his voice mailbox). Let us consider these events in turn and describe the corresponding Call Control procedures. 8.1.7.1. Alternate Services 7 c 0 MSC VIA MSC VLR isScityw k cT NEG1 0k€•.c.0". • • ! ! E . ) , 4 ”cf1,-5-.• S R e i t f61.... RELEASE OEM/PEE re ,(IrrnsE Of+111 r Yr 4 release by the GSM user 4 release by the other end Figure S.I7 r e l e a s e The release procedure is a three-way handshake procedure. A few of the services proposed in GSM allow the user to toggle between two transmission modes, speech and a data mode. An important example (the only phase 1 case, other alternate services being part o f phase 2 ) i s th e alternate speech/fax. Some signalling means are introduced i n the Specifications t o co-ordinate the transition o f all machines along the transmission chain from one mode to another. The corresponding procedures w i l l h e f u l l y used mainly i n phase 2 implementations, where they will he corrected and enhanced. We will therefore only briefly describe them. looking at the case of a mobile to mobile fax call. For the fax application, the toggling o f transmission modes is always initiated by the user, through some man-machine command, and applies only to his end of the call. This is translated by the mobile station in an RII.$-CC MODIFY message. aimed at the MSC. A t this stage the mobil, • •.• • L.L,.% A I L;NILA I I O N M A N A U E M E N T MSC VLR M S C VLR MOD, Y Reou:sTIP" DM.L: MCDOWL"E I , ows,,ne w.o131r1 0 change of transmission mode i n the BSS (ordered by the MSC) Figure 8.18 — In-call modification procedure The change of mode within a call is controlled by a call control procedure. which in turns triggers the relevant actions on the radio resource level. to change the connection type to the appropriate mode. ) 4 9 Why was it so specific..? Alternate fax/speech service is by nature designed tor interworking with the PSTN. Clearly. indicating a mode change to another extremity in a PSTN, or even simply through the PSTN. is not possible. This comes simply from the PSTN not being equipped for this kind o f signalling. There is then in these cases no alternative to letting the users cope with the co-ordination themselves. Things can he different i n the case o f a full ISDN compatible call ( through ISDN. and terminating for instance in the ISDN or in GSM): !SUP supports this function, and the information indicating that one end has asked for changing the transmission mode is carried by an ISUP INAIL MODIFICATION REQUEST message. Figure 8.18 illustrates this case. The receipt of this message at the other end triggers the sending, by the MSC. of an RIL3-CC MODIFY message, and an order to the BSC to change the transmission mode for this MSC-MS segment. Eventually, the mobile station will answer with an RIL3-CC MODIFY COMPLETE message, the BSC will indicate the correct completion of the command, and then the MSC can report this successful outcome to the other end, with an ISUP CALL. NIODIFICATION COMPLETE message. There is necessarily a (short) period during which the end-to-end transmission path is not consistent, before transmission can resume in the new mode. Let us recall that the I A procedure as described here will be in use only in phase 2, though it is for its major part specified already. 8.1.7.2. Multiple Call Handling As a preliminary to this section, we would like to remictd the reader that a l l cases o f multiple calls are also phase 2 features. Their specifications were not frozen at the time of writing. We will therefore not go to any level of detail, but briefly *describe the procedures which enable the users to juggle between several calls in parallel. node. When the MSC serving the initiating mobile station receives the message, it changes i f applicable the interworking function. and then it orders the change o f mode t o the BSC. which i s i n charge o f the transmission aspects on die MS to MSC segment. It is the BSC which will order the effective transmission mode change to the mobile station (this is outside the scope o f Call Control, and has been described in Chapter 6) and the BTS. When the change o f transmission mode is successfully indicated hack to the MSC. it sends to the mobile station an R1t.3-cc MODIFY COMPLETE message. Then. when the new mode is data, the f u r t h e r r e -connection procedures needed b y t h e service ( synchronisation. modem commands....) are run. We have already met one case o f multiple calls, which is call waiting. when an RtL3-CC SETUP message is sent to a mobile station fully in another communication. As seen in that example, and true in engaged all other cases. dealing with multiple calls involves only the mobile station and the serving MSC. Beyond this point in the network, the calls are all independent, and managed as such. The related functions are then essentially local to GSM. _ A t the other end. the same procedure should take place, triggered by the other user. The important point is that the intervening system does not take care of the synchronisation of the two changes. In some cases, the other end will be aware of the change from fax to speech by hearing or detecting the start of the modem tone exchange. and will only then ask for the mode modification on his/its side. The basic issue raised by the need to exchange signalling messages pertaining to several calls in parallel is solved, as already mentioned, by the notion of parallel Call Control transactions, the messages of which are distinguished b y a Transaction Identifier. T h i s mechanism enables signalling activities to take place for several calls, and this can be done even when one of them is fully connected (through the fast associated C O M M U N I C AT I O N M A N A G E M E N T usvt svcri:Al 7c1' MSC VLR b)I MSC VLR MSC VLR MSC VLR portion on hold fully connected portion_ r lully connegJeci c:iii 7 RETRIEVE HOLD RstRI OWVELF oGE K N , --7417:0c° LeoGs p.co NOTIFY NOTtry fully connected call IWIy cOnnucled Iwrl C^ p o r t i o n 0 11 N A ! J Figure S.In - call hold Figure 8.20 — Retrieving a call on hold Either end of a call can be put on hold independentl f r o m the other. I lere. user A puts his end of a mobile to mobile call on hold. The user may resume end-to-end communication on a call previously put on hold, using for this purpose the retrieve procedure. sinalling scheme). However. the only cases where this currently is allowed is the rejection o f a newly arrived call (call waiting) or the release of another one on hold. In all other cases the user can intervene in .1 call only i f all others arc on hold. the other extremity may be notified in a full ISDN/GSM environment, using in the GSM radio part, an RIL3-CC NOTIFY message. The procedure is shown in figure 8.20. Putting calls on hold. and retrieving them, represent the two major actions in the domain o f multiple calls. Putting his own end o f the communication on hold is done at the request o f the user. This action triggers the procedure shown i n figure 8.19. First. the mobile station sends an Rit.3-CC lump message to the MSC. A t the receipt o f the message. the MSC acknowledges the new state with an RII.3-CC HOLD AcKNowt.EDGE message, and i f possible warns the other party. In an ISDN environment. this is done with an itivP SUSPEND message. which would he. in the case of a GSM user at the other end. translated into an kti.3-cc NOTIFY message. These messages are only indicative, and do not entail any state change at the other end. It should he noted again that the hold condition pertains to a single end of the communication. so that a call may he in one of four states. depending on whether none. one or both extremities are on hold. The converse procedure consists in resuming full connection on the end o f the communication which was put o n hold. I t i s called the -retrieve- procedure. and can he done only i f no other call is connected. 11consists in the evehange o f an gn .3-c(' grtgikvT message and the 1•.. . • , • 1 1 1 , ' . ‘ 1 1 W 1 ; i , m o s s a t t e . T h e r e a g a i n Once all existing calls are on hold, the user can retrieve any one of them, release any one o f them, establish a new one ( i f the allowed maximum number o f communications i n parallel is not reached), o r --atimver. a pending Mobile Terminating call. This last procedure is not done b y a retrieve procedure, but b y sending an RIL3-CC CONNECT message. as described in the Mobile Terminating call case. Another action which can be clone in the connected state is to merge calls into a multi-party conference. More exactly, a user can ask for the conferencing of his calls. At the other ends, the calls have nothing specific from a management point of view. The calls to merge must exist when the conferencing is asked for. The first step will then be to merge two calls. the active one, and a held one (there can be no more than one held call i f another is active). Other calls can be added afterwards at the same end. by putting the multi-party on hold, establishing a new call, and conferencing the new call and the multi-party. Calls in multi-party can be disconnected temporarily, or released independently. The multiparty state disappears when reduced to one call. 553 COMMUNICATION MANACIEMIENT Tr a n s m i s s i o n o f A u d i o To n e s In the PSTN. dual tone multi-frequency refers to the transmission dual-frequenc tones generated on the speech path by pressing the minber keys of the telephone set (or b a separate tone generator with ,Itl-fashioned sets!). T h e specification o f the G S M mobile station Nuests Mat l ) i \ 11 ; tones are not generated by the mobile station. but ustedd hs t h e MSC, t o avoid coin_ through the k b i t t s speech twodoth Instead signalling roc.,,age. ;tic scut as the result o f pressing the mobile station side. ;Ind tones arc generated 1-1‘ the \ a t tie reception o f such messages. Note that t h i s applies o n l y f o r ransmission from the mobile station. Nothing precludes sending DTMF .ones to a mobile station. and the w i l l go through the I 3 o n / . encoding. \s a consequence. there is no guarantee that a l i f \ l f : receiver listening to 'he loudspeaker of the mobile station will re,...o!.tnise the tones. The procedure is not that simple. because the period during which .1 tone is generated is under the control o f the user. and because it was .'onsidered necessary that t h e M S C ackno‘v ledges these messages 'though the probability o f all RIL3-CC message being lost without the :all being cut is very small. thanks to the lower layer protocols). As the result. the sending of one tone requires lour RIL3-CC messages. and this does not include the layer 2 ackrmledgements. The generation of a tone in started by an RI1.3-CC START DIXIE message. which is acknowledged by the MSC with an RII.3-CC START inmi, ACKNOWLEDGE message (or alternately rejected with an RI1.3-C(' START DTMF REJECT message). It is stopped by the sending of an RIL3-CC 5T01, trrxiF message. acknowledged predictably by an RIL3-Ce STOP DTMF ACKNOWLEDGE message. This is repeated for each tone. resulting in a total of 40 messages for the sending of a ten-digit number. the forwarded-to number. Signalling means are provided to support their modification. or to check their value. by the subscriber and using the mobile station to access the GSM network. They will now be described. 8.2.1. ARCHITECTURE The signalling requirements f o r supplementary services management actively involve only Iwo entities: the mobile station and the I II.R. as shown in figure 8.21. The signalling exchanges are grouped in a special protocol, dubbed here the MAP/I, which differs from the other MAP protocol in that one of the two protagonists of the protocol is the mobile station, which is not directly connected to the CCITT signalling system number 7 network (SS7 network). Between the mobile station and MSC/VLR, the MAP/I messages are then not carried by SS7, but are piggybacked o n RIL3-CC messages encapsulated i n a FACILITY information element, either as stand-alone information ( i n RIL3-CC FACILITY messages, which may be carried in fact under a special protocol discriminator. SS), o r as information carried i n some other RIL3-CC messages such as SETUP or ALERTING. The exact meaning of a FACILITY message or of a FACILITY information element coming from the mobile station is determined by information inside the message or the element. They are only carriers, of limited semantic value. In some ,cases (as for the management of a multi-party conference) the meaning of the facility is local, and relates to the call. In this case the MSC/VLR treats these messages by itself. In the other cases. th'e message is forwarded by the MAP/I HLR 2 71(j, 8.2. SUPPLEMENTARY SERVICES - MANAGEMENT In the previous sections. a number of supplementary services have been described as far as their interaction with call handling is concerned (e.g.. call forwarding. call waiting or call hold). Most of these facilities can be activated (allowing their use o r their effect, and deactivated (preventing their use or 'stopping their effect) by the subscriber, and some have parameters which have to be set. or may be changed. such as, e.g.. MSC VLR Figure /4.2I - Protocols for supplementary services management Userscanchangeand check the status of their supplementary services through GSM; theMSC/VLR only acts as a relay for the corresponding MAP/I protocol. loch is handled iwt,1 . COMMUNICATION MANAGEMENT 555 GSM Sl'STFM 554 supplementary all Ion% aiding on lilts> applied to all basic .01.‘ ices BJ11111(2 of All Outgoing Calls applied to all basic .cr% iees General deactkation iTTETT'alllrarlitg . m i r e . Change of passuord for all supplementary ',en ice. requiring a passw ore Parameters for activation Call Forwarding t 'nconditional. on Busy. on Not Reachable REGISTER SS forwarded-to-number, basic service ('all Forwarding 'In No Reply kit ISTER SS forwarded-to-number. basic service, condition time ('all Barring sci is. s TESS basic service services :*2 I i f o r w ar(led-to-ntinillerr: I In' SE \ I Supplementary service (....11 for \\arding unconditional applied to telephony Type of activation message Sequence of keys n I it I • 10 1111111be td it IN 31 ip.i.,,,‘,1,11"hd xth ' . 1 M r 11 ) : 1 • • • • ‘ ‘ O r d l a l s r s n l old possw ord Inc\\ Pirsswordlilic l'ass"ordINIsENDI Table 8.3 —Activation of supplementary services For both call fin-warding and call barring, a single "activation" action is defined, which requires specific parameters depending on the service. Table 8.2 G e n e r i c man-machine interlace commands Asa fall-back solution. in particular for public ( N M phone". a generic method of entering supplcmcntan ‘ e n ices control orders on the kelNoard of a mobile station has been •dandardked. But most mobile stations offer more user-friendb means. done i n parallel with any other process, including a communication which is fully operational. Messages pertaining to a call independent facility management are then distinguished from call related messages by a separate Protocol Discriminator, and among themselves by different Transaction Identifiers if several transactions exist. MSC/VI.R to the III.R. In the other direction things are simpler for the NISC/VLR, since all MAP/I messages are f o n \ arded t o the mobile •--.. Facility messages pertain to different kind of operations, which can he divided into several classes. First we find the messages enabling the activation or deactivation o f facilities. There are several o f them, with marginal differences. Activation (as a general term) makes use o f the station. NIAP/I ACTIVATE SS, REGISTER SS o r INVOKE SS messages. R e c i p r o c a l l y, 8.2.2. P R O C E D U R E S Facility management can he done at any moment. at the request of the user. A generic. but not very user-friendly. man-machine protocol can he found in the Specificatirms. It specifies how commands can be issued b) the user on his mobile station keyboard. This protocol only uses the 12 basic telephony kQ i the 10 digits. plus a n d a n d examples of it are gi‘ en in table 8.2. The main advantage of this generic method is that it can he used for new facilities in the future without having to modify the nuthile stations. However, it is likely that mobile station manufacturers will add more friendly methods. to help the users to control efficiently the different facilities. which can, once mastered. bring a real improvement in user comfort. Whatever the method used b) the subscriber to issue the control commands. the mobile station will generate signalling messages t o be ng o connection already exists between the sent to the network. I f a signal]; deactivation (still a s a general term) makes use o f t h e MAP/I DEAcrivATE SS or ERASE SS messages. The choice of the message to use depends on the facility. Such a message contains a reference to the facility to activate or deactivate, a reference to the basic service (speech, tax. Short Message....) to which it is linked, and, for activation, various parameters depending on the facility. Table 8.3 indicates, for some of the major supplementary services including the phase 1 services, what type of activation parameters are relevant. A second group of messages enables the user (let us call him Finn) to enquire about the status o f his facilities. I t consists o f the MAP/I INTERROGATE SS message and its answer. A n interrogation message pertains to a single facility, which is referred to in the message. The answer contains the value of the different parameters as set in the HLR for this facility. As an exception to the general rule that MAP/I messages are handled between the mobile station and HLR, the Specifications consider t h a t t h e MAP/I INTERROWVIT S S t-nnec..,.. Tin,. OMNI SISTFM bile s t a t i o n a n d N I S C / V L K w h e n i t p e r t a i n s t o c o n d i t i o n a l c a l l :tilling. F i n a l l y w e c o n s i d e r t h e h a n d l i n g o f passwords. A p a s s w o r d c a n e e l b e associated w i t h each f a c i l i t y. as an a d d i t i o n a l p r o t e c t i o n f o r the ,xcriher against a m o d i f i c a t i o n o f its status O r parameters b y a t h i r d .'oil. T h e NIAP/I REGISTER PASSWORD message e n a b l e s F i n n t o s e t , Inge o r s u p p r e s s a p a s s w o r d i n r e l a t i o n a g i v e n f a c i l i t y. nversote, the %1 w i t -Gm. p A s s w o R t ) message and its a c k n o w l e d g e m e n t nvev respectively a request from the 111_12 to provide the password and password itself. as g i v e n b y the user. T h i s is o f course i n v o k e d b y the A2 when LI request to tamper with the concerned facility is received. A l l t h e messages seen s o f a r relate t o t h e cases w h e r e t h e m o b i l e Aim) understands exactly what is going on. I f facility management immands are entered as a sequence o f digits. # and * . a MAP/I PROCESS x s T R Tc r t n t h o s s D ATA m e s s a g e i s s e n t b y t h e m o b i l e s t a t i o n . I f a COMM I C AT I O N MANAGEMENT 557 transmission may take place even i f the mobile station is already in full circuit communication. A short message communication is limited to one message, or in other words the transmission of one message is a communication all by itself. The service is then asymmetric, and Mobile Originating Short Message transmission is considered as a different service from Mobile Terminating Short Message transmission. This does not prevent a real dialogue. but the different messages are considered to be independent by the system. The transmission of a message is always relayed by a Short Message Service Centre (SM-SC), considered to be outside GSM' . The consequence is that the transfer of a short message always takes place between a mobile station and some SM-SC from the point of view of the GSM infrastructure. However, for the user, the message has also an ultimate destination or origin, identified by some field in the message, but relevant only for the user and the SM-SC not for the GSM infrastructure. iy:word i s needed. i t has t o be included with the request. Status iecking i s n o t p o s s i b l e t h i s w a y. T h i s m e c h a n i s m enables operators t o nroduce new services between the user and the home network. without ie need f o r e x i s t i n g m o b i l e stations o r other n e t w o r k s t o handle t h e m i n specific w a y. A s such. i t c o n t r i b u t e s t o the u p w a r d c o m p a t i b i l i t y o f the stem. The MAP/I FORWARD SS NOTIFICATION message i s used w h e n the 11.12 detects that an activation o f a supplementary service requires leactivation of another service in order to avoid conflicts in the operation if the two services. The last MAP/I message. FORWARD CHECK SS INDICATION, is used tt the initiative of the HLR. in cases when some failure may have resulted n a n erroneous state o f the f a c i l i t i e s f o r the subscriber. T h e subscriber is n essence asked to check f o r h i m s e l f i f all is correct. 4.3. SHORT MESSAGES A l l c o m m u n i c a t i o n s o f a c i r c u i t n a t u r e , s u c h a s s p e e c h o r data transfer. are established. released and generally managed b y the procedures described i n the previous sections. Let us now see how :ommtmications o f another nature are treated in GSM. The only GSM ,ervices not requiring the end-to-end establishment of a traffic path are the Short message services. A s a consequence. s h o r t message 8.3.1. ARCHITECTURE As explained i n Chapter 2 , the point-to-point short message services defined in GSM enable the transfer of short messages between the mobile station and a short message service centre which is in contact with GSM networks through specific MSCs called SMS-GMSC (for Mobile Terminating Short Messages) o r SMS-IWMSC (for Mobile Originating Short Messages), referred hereafter, as in Chapter 5, both as "SMS-gateway". The protocols involved in SMS management are shown in figure 8.22. They include the following: • t h e mobile station to SM-SC protocol enables the transport of short messages, whether from or to the mobile station. This protocol is referred to as SM-TP (Short Message Transport Protocol): this protocol relies on underlying protocols which have been described in Chapter 5; • t h e protocol between the SMS-gateway and HLR enables the SMS-gateway to interrogate the HLR in search of the address of the subscriber when reachable; it is part of the MAP/C protocol already mentioned for the interrogation of the HLR by a GMSC within the standard Call Control procedure; The detailed functions of the SM-SC. and its interfaces are out of the scope of the Spei if/cations. This does not however preclude a GSM operator to operate an S\t-S( i h u , 1 . , ;,. LiVILIN I J J V 8.3.2. MOBILE OR. JINATING SHORT MESSAGES HLR \ MAP/D. z I - - - - - - - • When a GSM user wants to send a short message, he must as a minimum type i n i t s contents, t h e identification o f --the ultimate destination. and the directory number of the service centre which must deal with the message. Then, by some man-machine command, he can request the transfer of the message. MAP/C It is worth noting that there is no specification concerning the manmachine aspects o f Short Message handling. Quite certainly, various solutions will be presented by the manufacturers. The simpler mobile stations w i l l necessarily include a small display and a more o r less sophisticated keyboard, which may b e sufficient t o support Short Message entering, e.g., with escape mechanisms o r function keys to emulate a full alphabetical keyboard. The syntax to be applied to the message contents is irrelevant to GSM, but may depend on the service centre for specific applications and in this case will need to be known by the user. SMS-gateway MAPIH MSC VLR SM-RP, SM-CP SM•TP SM-SC Figure 8.22 Protocols for Short MeY:age Transfer The lower layer protocols (shown here as arrow, Of lighter shade' and the SM-TP enable the delivery of short messages between the mobile station and SM-SC either in real-time or ;1, mum a, the user becomes reachable. through the help of kaolin:ikon stored in the 111 It. • t h e protocol between MSC and HLR. as \\ ell as the protocol between HLR and SMS-gateway. enable the alerting of the SMSC when a mobile station has missed a message while it was out o f reach but has subsequently become reachable. This function must also he supported on the interface between the SMS-gateway and the SM-SC. b u t the protocols o n this interface are not defined in the Speci.ficatiems. Of course. application protocols are needed on top of the SM-TP (to format the user information for instance), hut these protocols are there again left to the choice o f Operators. and are out o f the scope o f the Specifications. We will now describe the functions o f these protocols in turn. starting with the case of Mobile Originating short messages. Short Message transmission requires the setting up of a signalling connection between the mobile station and the MSC i f none currently exists. This is done as for any other communication. The transfer of the message itself requires t h e establishment o f a special l i n k layer connection on the radio path. the SAPI 3 connection (see Chapter 5), and the use o f specific message transfer protocols. On top o f the specified protocol stack is the so called Transport Layer protocol, which consists in the case of a Mobile Originating message of a single message, the SM-TP SNIS-SUBMIT message. Lower layers deal with the delivery acknowledge, which indicates only that the SM-SC has received the message. There is no support at this level of an automatic acknowledge to indicate when the message has reached its ultimate destination. If the application in the SMSC supports this kind of service, the end-to-end acknowledgement will probably be sent in an independent short message. In any case, what the SM-SC does with the message is outside the scope of the Specifications, and no literature is available on the subject at the time of writing. What the user can do as offered by the SM-TP protocol is to set a period of validity for the message, after which the service centre will not try to deliver the message any more but destroy it. There are also some provisions in the format of the destination address to ask for the transfer from the SM-SC to a variety of directions, such as fax machines, teletex machines. other message handling facilities, ..., and GSM subscribers. dl THE (ISNI SISTI.A1 .3.3. MoBILE TERmINATINGSHORT MESSAGLS A short message addressed t o a G S M subscriber must first b e tiled from the sender to a Short Message Service Centre. and from then routed to the actual destination. The \‘,1 t h e message first reaches the \ I -S(' i s once again out o f scope o f the Speci/ications.iThere also a Het \ o f solutions can h e imagined t o enable P S T N tisers t o send Icssages towards G S M users. using human operators o r inter-working nth other services such as videotex. to When the SNI-SC has a message to .end to some ( GS\1 subscriber. builds a SM—li' sMs-Dimixtit message. containing various pieces o f ilormation f o r the benefit o f the recipient. This information includes in ;Irlicular the user content, the identification o f the original sender. and a me-stamp indicating when the message was received b y the SM-SC. Mnilarly w i t h t h e M o b i l e Originating ease. t h e SNI-TP SNIS-DELIVER lessage will he transferred on various interfaces. using the capabilities of \‘‘.- 1,1)er protocols described in Chapter 5. in particular to convey the ckno‘\ ledgement back to the SNI-SC. Before the SM-TI' s m s - D i d . i v i t message can reach its destination the m o h i l e station). i t s actual r o u t i n g m u s t b e derived u s i n g t h e nierroi:ation functions o f M A P / C . T h i s i s achieved i n the f o l l o w i n g (limner. The SM-SC conveys the short message to an SNIS-gateway to Inch the service centre is connected. which it chooses depending on the ubscriher it wants to reach, since most often a gateway w i l l he able to lea( " A i o n l y some o r the subscribers ( f o r instance those o f some ountrv. o r some operator). T h e subscriber i s then identified b y h i s lirectory number (the same MSISDN as for telephony. typically). entered irioinallv b y t h e originator o f the message. T h i s enables t h e S M S !ate‘‘av to identify the relevant HLR and interrogate it. The interrogation .s done by sending a special message. the NI ANC SEND ROUTING INFO FOR ,110RT MESSAGE message. This is answered by either the corresponding Si.w/c SEND ROUTING INFO FOR SHORT MESSAGE REsurr message. which contains an SS7 address pertaining to the MSC/VLR where the subscriber isiting, or by a rejection message i f the subscriber is known not to he reachable at this instant. There is no need for a specific roaming number, rs f o r circuit calls, since the short message uses o n l y SS7 signalling means to be transported to the visited MSC. The SMS-gateway makes use o f the SS7 address to forward the message to the relevant MSC. which delivers it to the mobile station after „ A m ( i f need he a signalling connection. as f o r the mobile originating e. The delivery to the mobile station does not involve the user. The • COMMUNICATION MANAGEMENT 561 message can he stored until the user decides to discard i t after reading. More precisely. it is stored in the SIM. and then can be kept in storage even alter the mobile station has been switched o ff , o r even read on another m o b i l e station. T h e l i m i t e d m e m o r y capacity o f the S f M , however. raises a small problem when for any reason the memory is full. In phase I . a message delivered to a SIM with no free memory could be lost. In phase 2. a mechanism has been specified to enable a crude sort of flow control by the mobile station. which will then he able to indicate to the net o r k when the memory is full. or conversely when it is hack to a state where messages can he accepted again. An important variation f r o m this basic scenario corresponds t o cases %%hen the mobile station cannot he reached. To provide a satisfying quality o f service, the message is not lost i n these cases, and steps are taken so that this message, and possibly following ones, are kept and delivered to the subscriber as soon as possible. Since the message, i f not acknowledged. is still stored for some time in the SM-SC, it can be sent anew as soon as the subscriber resumes contact with the network. This requires the GSM network to store the lack o f delivery condition and the address o f the SM-SC, and t o start a procedure t o "alert" the SM-SC when the subscriber pops u p again. The H L R i s obviously the "focal point" in such mechanisms. Let us describe how they work, by looking at the different situations in which deliver). fails. Three different kinds o f non-reachability can he identified, similar to what we have seen with circuit calls. The HLR- can know beforehand that the subscriber is not reachable for the moment, the V L R can know it, but not the H L R . and finally i t can be discovered after failure o f the effective attempt t o d e l i v e r t h e message b y t h e M S C / V L R . W h e n interrogated b y an SMS-gateway, the H L R may know immediately that delivery cannot take place, because it already holds a non-empty list o f ------s-erviee- centres which have not succeeded i n transmitting messages, and waiting to be alerted. It then adds i f possible the new S M - S e identity to this list. I n such situations, i t w i l l usually indicate the problem t o the SMS-gateway with a negative answer to the MAP/C SEND ROUTING INFO FOR SHORT MESSAGE message. The same course is taken when the H L R knows that t h e subscriber i s n o t reachable, f o r instance because the subscriber is not entitled to get service in the geographical area where he is currently located. There is however one exception to this rejection at the HLR level. A priority indication is linked with each message, and it is used t o bypass the straightforward rejection o f the H L R when some messages are still undelivered. The H L R w i l l answer positively i f the message is o f high priority, and the potential delivery problem, i f it still exists, will he detected by the M S C N L R . (ION MANAGEMENT 1111% • I ' ' ' 563 " IMSI > S M - S C address HLR IMSI 49( %474 4/ cc, SMS gateway error reporting S M S gateway '1/ /at I MSC VLR MSC VLR L - , IMSI > . m e s s a g e waiting IMSI i message waiting 1-1L.kireK.2.3 i n c , , i t g e tailed Jeli\erc management \Viten the MS(' is unable to deliver a short message to the mobile NialiOn. ilos state of things is stored in,both NiSC/VI.R and I II.R. For irilieritv retries Figure 8.24 - Service Centre alerting when the m o b i l e station reappears. In the cases where the MSC/VLR is given the message but is not M k to deliver it. a failure indication is first sent to the SMS-gateway, as m answer t o t h e mAP/ii FoRwARo s i t o RT MESSAGE message. T h e .tatewav then sends on one hand a negative report to the SM-SC, and on ihe other a MAP/C' SET NIESS:KIE WAITING DATA message t o the H L R , AHeti acknowledges the updating o f its table by a NIAP/C SET MESSAGE AITING DATA RESULT message. This state of affairs is stored by both the \ I S C / V L R and the H L R in the subscriber record. In addition, as already mentioned. the H L R maintains for each subscriber a list o f addresses for the S M - S C h o l d i n g messages i n w a i t . T h e sequence o f events i s represented in figure 8.23. Eventually. the subjcriher surfaces again. This may he known for instance b y a contact w i t h the M S C / V L R where the subscriber was located (e.g., a mobile originating call attempt). When such an event happens. thanks to the stored indicaetion of a previous delivery failure, the NISC/VLR notifies the H L R with a MAP/D NOTE Nis PRESENT message. When a subscriber becomes reachable, the HLR alerts all the service centres which are known to hold messages not delivered to the given subscriber. The mobile station may also reappear within coverage o f another MSC, in which case the H L R w i l l be directly aware o f this state o f things, thanks t o t h e m o b i l i t y management procedures w h i c h h a v e b e e n described in Chapter 7. In any case, the HLR then sends an indication o f the subscriber's reappearance t o a l l the SM-SCs whose identities are stored as holding a message f o r this subscriber. T h i s i s achieved b y sending a MAP/C ALERT SERVICE CENTRE message to the suitable SMSgateway for each service centre. The whole sequence as described above is shown i n figure 8.24. T h e SMS-gateway w i l l convey the relevant Information to the service centre, to trigger a new transfer attempt. The alerting mechanism must be supported by all MSe/VLRs, but it is an operator's option to store the list o f SCs in the H L R and to alert them. SPECIFICATIONS REFERENCE The only GSM TS that can be cited for a general presentation of the call management mechanisms is TS GSM 03.01. Network firnctions, hut the subject is only partially treated. The interface betty' M S C ' s or GMSCs and external networks are refene.d to in TSs GSM u9.013 to 09.07, for the various categories o f external networks. O f particular relevance are TS GSM 09.03 and TS GSM 09.07 which deal with the signalling aspects of the interface with the PSTN or the ISDN; and TS G S M 09.11, which deals with the signalling interworking for supplementary services. For inure details the reader must jump immediately to the detailed interface specifications. Short message services are dealt with in TS GSM 03.40, which covers all the higher layer aspects of the service. The man-machine interface betkseen the mobile station and the fiscr ic delft with in-the-02. Series. TS GSM 02.30 is devoted to the subject. and pieces can also he found in TS GSM 02.07. dealing with mohile station features. All the Sperdinthems cited hereafter are deeply technical, and contain little general presentation to aid an understanding ()I' the subject. The interfaces for call management between a mobile station and a Terminal equipment are referred to in TSs GSM 07.01, 07.02 and 07.03, of which the main part is devoted to transmission aspects on the intelice. Generalities concerning t h e modelling aspects o f the .RIL3 protocols can b e found i n T S G S M 04.07. T h e R I L 3 protocols themselves are specified in TS GSM 04.08. where section 5 is devoted to the CC protocol. Section 6 deals with call control for packet mode data, but i s o f n o application ( i t i s even removed l O r phase 2 ) . The supplementary services aspects o f Call Control are not included in TS GSM 04.08, b u t are dealt w i t h i n 'I'S G S M 04.10 and i n the Specifications of the 04.8x series. including the coding of messages in TS GSM 04.80. I n phase I . only the call barring and call forwarding facilities are fully included. and they are specified respectively i n TS GSM 03.82 and 03.88 (for the general technical aspects) and in TS GSM 04.82 and 04.88 (for the detailed specification o f respectively the call forwarding and the call barring service implementation). More generally in phase 2. the implementation of the facilities described on the service level i n TS GSM 02.8x (x ranging from I t o 8) is described in the corresponding TS GSM 03.8x and 1/4.8x. The details o f the MAP protocols are to he found in TS GSM 09.02. more particularly in section 5.4 (retrieval of subscriber parameters during call set-up). isection 5.3 (handling of supplementary services) and section 5.13 (support of short message services). N r w o K s1ANAGEMtjNT 5 6 7 9 NETWORK MANAGEMENT 9.1. Subscriber Management 9.1.1. Subscription Administration 9.1.2. Billing and Accounting 9.2. Maintenance 9.3. Mobile Station Management 9.4. System Engineering and Operation 9.4.1. Cellular Planning 9.4.2. Cell Configuration 9.4.3. Network Engineering .14 9.4.4. Observations 9.4.5. Network Change Control 9.5. Architecture and Protocols 9.5.1. Management Network Arc nee u 9.5.2.. Operation and Maintenance in the Traffic Handling Protocols 9.5.3. The BTS Management protocol 9,15.4. The GSM Q3 Protocol Specifications Reference NETWORK MANAGEMENT 568 569 572 578 585 591 593 613 622 627 628 633 633 637 638 640: 646 The picture of the GSM communications system as presented so Gtr is a functional one. We have seen the different actions performed by the traffic handling machines i n order t o provide communications services to the users. We have described in a quite abstract way machines, transmission lines and users. Now. for a network operator, users and machines are more than abstractions. A real network contains many machines, whose location and capacity must be chosen. Machines must be procured, installed, linked to one another, and configured so as to be part of a consistent and cost-effective network. Machines and links can suffer failures, and must b e repaired. Subscribers must be sought. corresponding data must be entered in the system and money must be recovered from these subscribers. These different activities are an integral part o f telecommunications systems. Because o f the ever-increasing complexity o f telecommunications networks and because o f the search for cost-effectiveness. these tasks are supported more and more by machines introduced in the system for this purpose. The focus in this chapter is given to the point o f view o f the operator. T h e different operation, maintenance a n d subscriber administration tasks will be developed, with a stress on the aspects done by electronic machines. whether by the traffic handling machines or by other machines devoted to these tasks. The latter constitute the OSS, the Operation Sub-System. Most o f the concepts presented here are o f general application to telecommunications systems. In each case, GSM will he taken as an example, and some importance will be given to its particularities, which are often those of cellular mobile communications systems. I:I kU)Ith NIANAUENIENT TI Ili (1551 SVS.11-.N1 OS The second area is maintenance. that is to say the functions aiming it maintaining a satisfactory level of functioning in the system despite the unavoidable failures. We have included in this domain the management !it the mobile stations (with the related machines and procedures), and their type approval, which can i n a way be presented as preventive 9.3. SUBSCRIBER MANAGEMENT The relation o f an operator with its subscribers has two main facets. First a commercial dialogue must he initiated with them which will lead to a subscription relationship being established. A subscription includes an entitlement to obtain service from the network. and this requires actions on the network equipment so that the subscribers are recognised as such. The second facet is billing and accounting. The operator must calculate the call charges. and—a vital issue for operators—must recover the money. 1w. among others means. billing its own subscribers. The 9 9.1.1. SUBSCRIPTION ADMINISTRATION Access to GSM services is conditioned by the user being known to the system, which i s the purpose o f subscription. From the traffic handling point o f view, a subscription is materialised by a subscriber identity module (SIM) and corresponding entries in databases (in the HLR and in the authentication centre. AuC), and is identified by an IMSI (International Mobile Subscriber Identity). Means are needed i n the system to create, upgrade and cancel subscription data. maintenance. The third, last, and longest part is devoted to operation. This covers the "piloting" of the network by the operator personnel. This includes the engineering choices to build, deploy and improve the network: the means of knowing i f the level of performance is as expected: and the means to modify the network and its parameters. The description o f the GSM system engineering part is developed using a number o f examples. In particular cellular planning w i l l b e addressed. and this w i l l b e the opportunity t o tackle topics such as radio propagation and spectral Specifications. 6 subscriber mobility, and _yen more the roaming capability, somewhat complicate the issue, compared with wireline systems. With roaming it is possible that some service is provided entirely by one PLMN, and paid for by the subscriber to another. Inter-PLMN billing and accounting is then a very important topic, which determines, more than the technical aspects. the possibility to provide roaming to subscribers. Some stress will then be put o n the transfer o f charging information, between machines. and between networks. The first area presented i n this chapter. subscriber mati..gement, ncompasses the handling of the subscriber data needed in particular by he traffic handling machines. and the charging aspects. The study of this rca w i l l lead us t o present the relationship between customers and gyrators. and the concept of service providers. We will also look at the lad ()I' the Operation Sub-System related to all these aspects. efficiency. The last part o f the chapter will present the architecture o f the operation Sub-System. with a focus on the part devoted to infrastructure equipment maintenance and operation. This will be the opportunity to address t h e concept o f T M N (Telecommunications Management Network), both for its architectural aspects. and for its protocol design approach, a n application o f which c a n b e found i n t h e G S M 5 _ Some additional data related to the subscriber, such as his name and address, are needed for commercial aspects, and in particular for recovering the charges. The Specifications do not cover these aspects, and the implementation will differ from one operator to another. The HLR may. in addition to its canonical functions, be the repository of this data, or they may be kept in another database, with typically the IMSI serving as the common reference. This choice depends in fact less on technical considerations than o n the commercial organisation set u p b y the operator. A commercial structure which i s becoming more and more common is based on the concept of service providers. Service providers are companies, usually distinct f r o m t h e operator, w h o take the responsibility f o r the commercial contacts w i t h the customers. This includes typically the establishment of the subscription, the establishment and the dispatch of the bills towards the subscribers, and the recovery of the money. A subscriber who goes through a service provider has no direct contact with the operator. Usually, the operator bills globally the service provider for all the charges related to the subscribers managed by the latter, and provides the service provider with toll ticket information necessary to produce the individual bills. Looking to practical cases, such as existing cellular networks or the organisation o f the GSM operators. there are i n general many more service providers than operators, and a service provider may deal with several competing operators. Some operators retail exclusively through service providers. At the other extreme, some operators deal with all the NETWORK MANAGEMENT 571 iiiminercial aspects themselves. Mixed cases can also be limn w h e r e ‘oth a direct sales network belonging to the operator and a number of ervice providers coexist. Figure 9.1 pictures the relationships between lie operator, the service provider and the customer. -like intervention o f service providers splits the work between a totwork operating company. operating a telecommunication equipment, intl companies in charge of the commercial network. The subscriber data then spread over. for instance. a database held by the service provider including name and billing address in particular) and a database held by he operator. The latter is typically implemented in a database separate on) iiii‘111.61t. and it holds all the subscriber information not necessary or d i e provision o f the telecommunication service. i n particular nformio ion such as "how to bill- the subscriber (for instance from which Device provider each subscriber depends). In the killowing we will assume that there is a subdivision of tasks 'cm col operator and service provider. This i s t o be understood as overing also the case when the operator manages a l l commercial elation,: li)\ itself. i.e.. the service pro\ ider is the operating company Creating a subscription entails diffecint tasks. One is the contact Aits. the would-be customer. to establish \\ ith him or her the subscription iiontract. The other is the initialisation of the different databases and of Ole SINT Because customers like to subscribe in a one-stop shopping approach. leaving the sales agency v, ith their SINI. it is useful for an tperator to prepare as many things as can he in ;atomic o f the actual subscription. In a typical scenario. the operator prepares the AuC and the data in the SIMs before the actual contact \\ ith the future subscriber, and ma m e n prepare incomplete corresponding records in the HLR. The service provider have then at their disposal a number o f SIMs which already contain a stored IMSI. but which cannot be used at this stage for i‘ettiiii, service (by lack of a proper initialisation in the HLR). When the subscription is created. the service provider enters the data related to the customer in its own database and ascribes an IMSI and a SIM to the new subscriber. In this scenario, the service provider need not perform any action on the "ready-to-use" SIM. except possibly printing the subscriber name on it. I n some sales networks. the customer may choose his directory number (the NISISDN) among a proposed list of numbers. Another scenario. although less attractive to the customer, consists in completing the SIM personalisation process (including the storage of the IMSI and secret key Ki) once the subscriber is known. In such a the subscriber must wait a few days to receive a SIM, e.g.. by mail.glhe main advantage of this scenario is to remove the need for the operator to have a real-time subscription management network. operator HLR information c h a r g i n g information service provider subscriptiont billing subscription \ir billing indirect sales through service providers direct sales by the operator Figure 9.1 — Operators and Service providers The operator may choose to deal with customers directly, or alternatively to entrust service providers with customer care. Many operators choose a mixture of both solutions. • I n both cases, some steps must be taken to properly initialise the HLR record. This includes the description of the subscribed-to services, the allocation of one or several MSISDNs, and the enabling of the IMSI for actual service. Finally some tests have to he conducted to check all the operations were done correctly, so that the subscriber can actually access the services he or she asked for. The service provider must have means. whether in real-time or not, to trigger the initialisation of the HLR record. usually under the control of the operator. ink I ism s)Si F.N1 There are other aspects to subscription management. Sti, _Tiber no may be modified after the initial subscription. Sonic modifications a change of address) will impact only the service provider database. 'Mors. such a s a change i n the operational characteristics o f the ihscription (subscription cancellation, extension to new services, change regional limitations....) also impact the I II.R. In all cases. if a service :odder intervenes. it will take in charge the contact with the subscribers ir these actions also. I he role of the sen ice provider for customer care is not limited to line new subscriptions and changing them. Beside billing (which will be :scribed in the next section). the service provider must perform tasks Ich as the after-sales assistance to customers. For instance. the report of 'NI SIMs must he handled quickly and replacements issued with another \1S1, while the IMSI of the potentially stolen SIM is deactivated in the II.R. A first level of protection against misuse of a stolen SIM consists blocking the SIM itself. i.e.. forbidding any access to the system when strong PIN code has been entered three times ill a row. Of course, the bsentininded but rightful owner of the SIM might occasionally fall in he trap himself. To cope with such cases. the Specifications define a PIN nblocking Key (PUK), different for every SIM and which can be used .) unblock the SIM. Depending o n the operator's choice, the PIN inblocking key can be provided to the subscriber or alternatively made ivailable only to the service provider. In the last (and more secure) case, he subscriber will need some assistance from his service provider to mblock his SIM. The commercial activity requires a number o f points o f sales Tread all over the covered area. A technical approach to support the operations described so far, with a high level of automation. is to equip he points o f sales with computer terminals able to read SIMs. These erminals are connected through some data network to machines which ire able on one hand to update the service provider database, and on the ether to provide contact with the HLR machines. 9.1.2. B I L L I N G A N D A c c o u N T I N G A subscription relationship enables access t o service. T h e (mini:Tart is the payment of some periodic charge. the charges for calls .nut other services, and possibly a fixed initial subscription fee. The collection of the corresponding revenue is vital for the operator. as well is t h e accuracy o f the services charges. T h i s requires recording mechanisms in the traffic handling part of the system. so that sufficient iota vccorded for each chargeable service. Calls are obviously the first outer o f traffic-dependent charges. b u t supplementary services NE:rwoRk NIANAGENIENT 573 charging information Home PLMN + invoice (service provider, if any) bill Visited PLMN /access to service Figure 9.2 — The "transferred account procedure" The body responsible for paying the visited network is the operating company t%ill) which the customer holds his subscription. This way, a subscriber receives a single bill for all his calls including those made when roaming. management transactions may also be billed for instance, depending on the operator's policy. The GSM MoU operators have decided so far not to charge for location updating. The tariffing principles for the different transactions (calls, short messages, supplementary services, ...) are being discussed at the international level, but the tariff levels are left open for every operator (or even in some countries for every service provider) to decide upon. We have seen in Chapter 8 that toll tickets are created by the MSC/VLRs and b y the GMSCs. To l l tickets are individual records generated for each call and containing all the information necessary to calculate the call charges. Because o f roaming, the MSC/VLR tickets pertain to subscribers which may belong to networks other than the one corresponding to the issuing machine. Each ticket includes the IMSI of the to-be-billed subscriber. The IMSI is the basis for forwarding the ticket to the right place. In the GSM MoU. the rule is that the charges are collected from a subscriber by the operator he holds a subscription with. The responsibility for paying the visited network operator lies with the home network operator o f the subscriber: this is the concept o f the "transferred account procedure'', illustrated in figure 9.2. A first task for I o n N t i \ I 5 ' `+11A1 he \ kited network is then to sort the toll tickets and forward uk. in to the .•orrect subscription n e t w o r k s . T h e t i c k e t s pertaining t o r o a m i n g subscribers m u s t b e g r o u p e d 11) n e t w o r k . T h e > a r e u s e d f o r t h e .•siablishment o f a global bill for each o f the other operators. and they are scin it' the same operators for their internal accounting. the tickets for the home subscribers. \\.liether internally generated ,r coming f r o m other operators. must then h e sorted o n a subscriber HaHs. The operator is in chargi.• o f the processing tip to this point. In a H i l t ;1 service p r o \ ider. a global b i l l i s sent t o t h e SerViCe ;Irk Is tiler. H r all the calls of the corresponding subscribers. and the tickets are sent to the service providers to allow them to establish the individual bills. Variants o f this scenario m a y he envisaged. where the operator establishes the individual customer bills for the service provider: this can happen when. e.g.. the tariffs are fixed b y some national authority. The actual dispatch o f the invoices to the customers is however always the task o f the service provider, as is the recovery of the charges. All these transfers can be automated. Machines ( i n the operator part o f the system) can collect and dispatch the toll tickets. and possibly establish the h i l l s f o r the other operators o r f o r the service providers. Then. means are required to transfer data between the A t c h i n g offices and these machines. between these machines and those '"of the service pmviders, and between such machines f o r different operators. Often at the beginning the data transfers w i l l be done by magnetic tapes, written by one machine and transported to he read by another. Real time is not really a constraint in this field (except for a real time advice o f charge to the subscriber: the difficulty i n case o f roaming is the reason w h y this service w i l l not be provided initially). Electronic data interchange w i l l shortly replace tape transfer. especially between those networks where roaming leads to substantial traffic. The Specificadons provide a method f o r the real-time transfer of toll tickets between networks held d i f f e r e n t operators. T h e M A P / C includes messages f o r carrying t o l l tickets on a call basis between an NISC/VI.K and a HLR. This involves m o messages. the NI ANC REGISTER t RHINO INFORNIATION and the corresponding acknowledgement. The contents of the message gives an example of what can he a toll ticket, and we w i l l come hack to this topic in a few paragraphs. I t should be noted that there is no corresponding procedure between GNISC and HLR. This is in line with the approach developed in Chapter 8. that a GMSC deals kink w i t h calls toward subscribers o f the network to which the GMSC belongs. The use of the NIAP/C facility to send toll tickets hack to the home network is not mandatory. It will be little used in practice. since the task of valorising t h e t o l l tickets i s t y p i c a l l y performed b y specialised . •... • , , , , , , ‘ , J L . . . • I L . . 1 I /3 GK' SC Home PLMN e z e t i charging I centre (MAP/C) billing centre Service Provider 1 . 4 centre charging MSC VIII tape or electronic data transfer Visited PLMN optional transfer of individual records Figure 9.3 —Transfer of billing data The transfer of billing data between networks for roaming subscribers, and between operator and service provider, calls for an architecture of interconnected machines. charging centres separate from the MSCs and GMSCs. The inter-PLMN transfers o f charging data are then performed between such centres, by tape or electronic transfer through means other than the SS7 network. Figure 9.3 presents a general view o f how the different machines interact in the administration o f subscriber charges. with the assumption that there is a split between a network operator and a service provider. We have described so far how GSM operators are able to exchange toll tickets between them. The transfer of money involves more than just GSM operators, and the transfer of the billing data is but the start o f the operation. T h e f i l e s transferred between t w o operators include t h e individual call records f o r a given period o f time (e.g., a few days). and enable the home operator to charge its own subscribers. They indicate the applicable exchange rate f o r each call and the inclusion o f value-added tax o r not depending o n national regulations. and they give the total l i l t i . / 7 ) . , 1 , 11 . 1 1 1 . t i .110L1111 of all call charges they contain. These files serve as a s i s for stablishing invoices between operators. The payment of invoices may be done directly. on a bilateral basis. lowever. since ibis would give rise to a lot o f transfers even when the ;milting balance is small. many operators prefer 10 settle these accounts hrongh an international "clearing house". in charge ()I' receiving all the greed amounts to be transferred and combining them so as to minimise he number of actual more transfers. This concept of a clearing house is ,nead‘ ❑pplied within ('LI'T. where operators of the wireline networks cute their accounts in this way. 9.1.2.1. An Example of Toll Ticket The structure o f the MAP/C REGISTER CHARGING INFORMATION message is an example o f the information which may compose a toll ticket issued by an MSC/VLR. Though this message may in most cases not be used. the contents of the files exchanged between operators will :ontain charge records which have fairly similar contents. Since the specification of the MAP message is public. it can be reviewed here as an example. to illustrate many different aspects seen in this chapter and in Chapter 8. Though the MAP procedure does not cover the transfer of accounting from the GMSC. the toll ticket structure covers all cases, including the re-routed part o f mobile terminating calls from GMSC onwards. First. the message contains identities: the IMSI Ito identify the subscriber) and the identity of the MSC that generated the ticket. Then we find a call reference, which will enable cross referencing. Next i s included the "type" o f the call. The following cases, correspond to tickets produced by a visited MSC/VLR: • a mobile originating call; • a mobile terminating call forwarded by the MSC/VLR; the toll ticket then concerns the portion from the MSC/VLR toward the 10rwarded-to destination. • a non-forwarded mobile terminating call; this is not foreseen to he charged in networks of GSNI MoU operators. but is included to keep the flexibility to do so, as well ❑s for other purposes such as statistics. etc. In addition. two cases correspond to tickets issued by a GMSC: I WORK MANAUEMbNii 5 7 7 • ❑ mobile terminatint call forwarded by the GMSC; the toll ticket then concerns the portion from the GMSC toward the forwarded-to destination. • t h e re-routed leg of a mobile terminating call, from the GMSC to the visited MSC/VLR. Then we find the "status" of the call, that is to say whether it was successful or not, and in the second case the reason for failure. While toll tickets for successful calls are vital to the operator, the establishment of call records for unsuccessful calls usually serves no other purpose than for statistical analysis, and is not provided by all machines. The next part o f the information i s the main data f o r the computation o f the charge. i n addition t o the MSC identity which provides the ability to determine one end o f the call. This includes the date and time of the beginning of the call (to cope with tariffs depending on time of the day), and its duration: the nature of the service provided (speech. data, short messages, supplementary service management transaction, with, in the last cases, more details on the nature o f the transaction); and the called number (pertaining to the other end o f the concerned segment o f the call). To cope with packet services, a field includes the volume of transmitted data. So as t o enable the subscription operator t o provide detailed billing. the calling party number (for mobile terminating calls) or the called party number (for mobile originating calls) may be included. Finally, the record includes the charge which will be collected by the originating operator on the subscription operator. The usual unit is the ---'513-ectal D r a w i n g R i g h t " ( S D R ). a "currency" u s e d between telecommunication operators for inter-network accountings. - It should be noted that this structure has been designed only for billing and accounting purposes, and covers only the information needed for transfer between networks. Recording user activity is needed for other reasons within the network, such as network statistics. Since it would be inefficient to have a double recording mechanism, records created by MSCs o r GMSCs and dealt with i n the OSS usually contain more information. In particular other activities such as location updatings may he recorded, and the records may include information such as the mobile equipment identity (IMEI) or the identity of the cell where the transaction was initialised. IN i cS URI. MANAGEMENT 9.2. MAINTENANCE • V e r y rare are the systems which never fail. and telecommunication skims like GSM are not among them. A very important task for an operator is therefore to maintain the system in a state where the quality of ser\ ice offered to the subscribers is acceptable. Maintenance includes the techniques aiming at minimising the loss of service quality incurred by a railiiro.-44)444wans to detect such failures and to report them accurately to the right person, and finally the Mean', to restore the slate of the network. Failure causes are numerous. Electronic components have a finite -though long—lifetime. The hardware of all> piece of equipment include thousands of components and a \\ hole PI .\4N vi ill include tens of thousands o f such pieces. Fven with a component lifetime o f 10 000 )ears. several failures w i l l occur per day i n such a network! Other problems may come from the external world: the power supply may fail, communication lines may he cut by accident. thunder strikes regularly... lIniors are also a n important source o f dysfunctioning. Zero-error software is long and costly to develop and in complex areas such as GSM not obviously achievable. The co\ erage o f tests never represents one hundred percent of the running conditions. since the most marginal cases are often skipped. Failures tend t o occur f o r instance i n overload situations. where the system is pushed to its limits. and such situations are difficult to mimic on a test bed. Maintenance is faced with this large ariety of failure sources. ❑nd the increase of system complexity makes automation and specific software necessary to assist the maintenance teams in their task. Observing what happens in connection with ❑ failure hints at the different facets of maintenance. It \\ ill serve as our guideline to describe them. What happens brfore the failure is the subject of preventive, or proactive. maintenance. The aim is to minimise failure occurrences. In fact. this step merges with the next one. since preventive tests (such as test loops. for instance) nay also serve to detect that a failure has occurred. There are other ways to detect failures. such as watch-dogs. that is to say independent devices which monitor some aspect o f the activity o f a processor. o r temperature detectors. ... Once a failure is detected, a number of immediate automatic measures are taken locally by machines directly connected to the one suffering the failure. The aim is to limit the immediate effects of the failure. Examples of such emergency measures include software reset. the taking over of another piece o f equipment in a redundant design or the stopping of the machine if material danger exists. These first-aid measures do not preclude repair. and to this avail the failure must be identified ❑nd located in a precise manner. The actual 579 replacement may involve.,,Ilythe faulty module, or imply a modification of all similar modules in the system. Let us look at some of these points in inore detail. 9.2.0.1, Minimising Failure Occurrences Obviously ihe best method to limit maintenance costs is to use reliable equipment. During the procurement activity, an operator will consider different manufacturers and choose the best one according to specific criteria. This criteria may include the way in which quality insurance is performed by the manufticturer. In addition, the operator may also submit the pieces of equipment to its own set of tests before formally accepting them. Usually, a minimum set of tests is also passed on each single piece of equipment during the installation phase, before going into commercial operation. Cellular networks experience a difficulty in this area. When for instance a new base station is introduced in a network, a normal practice would he to test it in a fully operational configuration while preventing commercial traffic until the machine has been fully tested. But radio V,aves cannot be restricted to the needs o f the test, and the new. cell impacts on the neighbour cells already in operation. The tested cell must be barred. to prevent the mobile stations of the customers to camp on it. and handovers towards this cell must be forbidden for these customers. Testing the new cell requires the operator to have special mobile stations, which (in addition to tracing facilities) will be able to camp on a barred cell. Modifications in the existing part of the network are necessary to test the new cell i n a f u l l y operational configuration, since handover parameters in neighbour cells are then different whether or not the new cell has to be taken into account. I f the handover configuration had to be tested. means would he needed for BSCs and MSCs to distinguish test mobile stations from others. No such means exist for the moment, and, in short. the very final test will be done by the users themselves! 9.2.0.2. Minimising the Effect of a Failure Failures can have very different effects on the service perceived by the users. The failure of some component involved in the transmission of a single radio channel or terrestrial circuit will only marginally degrade the capacity and hence the quality o f service i n terms o f blocking probability. On the other hand, the fitilure of a switching matrix handling hundreds or thousands of circuits (which can be on a single chip) may result in the suppression of service over a whole geographical area. 5;SO The key method to limit such catastrophic events is re i d a n c y. Where II interchangeable devices are used in parallel. n+p are installed. I'or instance, i f a c e l l needs I Transmitter/Receiver (TRXs) t o .tecommodate the expected traffic:. 5 will typically he installed. Another example of redundancy concerns the central chain of switching machines such as MC's and MSCs. which are usually doubled. 'the use of the redundant parts belongs to the domain of immediate defence. and i s triggered locally b y failure detection means. Several techniques exist t o minimise the impact o f the transition on existing traffic. and we will rum consider a few examples. In signalling system n°7. a signalling relationship between two adjacent points can consist of several links, and redundancy ensures that the required traffic load can he supported when one o f the links is missing. I n normal operation. the traffic i s spread o n a l l links. I f acknowledgement and repetition protocols detect that one link is faulty, the corresponding load is reported on the other links. There is no loss of ongoing traffic, and capacity is maintained. Another example is the hot-standby method, used typically for central processors. Two similar machines (A and B) are installed instead of one. and they run in parallel. Only one of them (say A) is actually in charge at one given instant. and provides its results to the outside world. When a failure is detected. the roles of the processors are swapped, and the outputs of the system are now provided by B. which then becomes the leading processor. At its highest level of sophistication. the hot-standby technique enables a transition without hiatus, and damages are limited to what occurred between the failure and the swap. Several degrees of standby exist. from cold to hot. At the bottom of the scale, the "spare" device is not running, and is started only when the fault is identified. In this case, the swap involves :A loss of context (e.g., all ongoing communications are aborted). but traffic capacity can be maintained. A similar case exists when one of a set of ) TRXs of a BTS fails. Typically, the ongoing communications using this TRX are lost, and a reconfiguration at the BSC level is required to restore the capacity to nominal values by using the spare TRX. A last example in this domain is the case of A interface circuits between the BSC and the MSC. When such a circuit fails, the communication using it, if any, is typically lost. In order to circumscribe the effect o f the failure, the circuit is then marked as "blocked" until repaired, i.e., tagged as not fit for allocation in the MSC. I f the problem has heeii detected by the BSC. it indicates it to the MSC by signalling means. 061 Fault Detection So far, we have briefly discussed the measures taken once a fault has been detected. How failures are detected is an important area o f maintenance, and we will illustrate the matter by a few examples. A number of detectors are incorporated in machines for the sole purpose o f detecting failures. This is the case of temperature detectors, which may indicate faulty componehts consuming too much power, or failed fans. The power supply may be monitored by voltage or current measurement devices, triggering alarms if some thresholds are exceeded. Contacts may be used to detect board removals, door openings, and so on. In all these cases, the fault is detected locally. In the case o f hot-standby, a discrepancy between the results provided by the two parallel modules is the symptom of a failure. The problem i s t o decide which i s the faulty one! Usually, a separate monitoring device will ask the modules to perform specific tasks and will compare the results to an expected pattern. A majority decision can also he applied i f there are more than two parallel processors (e.g., i f the modules are tripled). Faults can also be detected remotely. For instance, when two machines communicate, they monitor each other. The lack of respect of some protocol rules, detected by the receiving end, may be the symptom of a failure o r o f a software fault i n the other entity. Similarly, transmission circuits can be monitored i f one end loops back what it receives when the circuit is not engaged. The other end can thus verify the continuity and consistency of the connection. Regular and automatic testing i s an important fault detection method. An advanced example of such tests is the use of a test mobile equipment in a BTS. Such a test mobile may be able to perform all the normal activities o f a mobile station, initiating calls and so on. But in addition, it will analyse the way in which the network reacts, and check whether this behaviour conforms to some pre-established rules. I f such a mobile station uses an antenna, it will also detect radio problems such as a decrease in the BTS transmission power, a fault difficult to detect otherwise. There is quite a variety of tests which can be performed. They are done in some cases routinely, in particular i f running them does not take up resources which could be used for traffic; in other cases, they are controlled through the OSS. A source of information which must not be neglected is the users themselves. User complaints, once analysed, often reveal faults that have escaped internal checks. Most operators have a free-phone number dedicated to the reception of customer complaints. In addition, a panel of • W O R K MANAGEMENT THE GSM SYSTEM s82 voluntary subscribers or operator staff may regularly perform a postman's round and report strange events. A last example of failure detection is the analysis of global observations on the system. For instance, abnormally low traffic in some area may reveal a fault. Such high level detection can sometimes he helpful to detect failures affecting other failure detection mechanisms! Generating Alarms Fault detection can be done in a large number of different ways, including local and iremote testing as w e l l a s analysis o f various observations. However, failures o f a similar nature can be detected by. very different means, and reciprocally some analysis has to be performed to dispatch the failure information to the right place. In all cases, the., original problems are translated into a general currency the alarm. Art alarm i s a message sent t o some central facility f o r analysis. It.; 4; porissi generated at some point in the system to indicate a failure, and it r e the local analysis of the failure after immediate defence actions have been t taken. ( 5 1 . Locating a Failure Alarms are but symptoms. which do not necessarily point to the exact illness. The more global the symptom. the less precise the identification of the failure location. For instance, a lack of coverage in some area can be caused by a number of problems, including failures in Wiss. BSCs and MSCs. A fair assumption is to consider that the failure happened at a single point, and can he pinpointed to a single piece of equipment. But a failure can also he an inconsistency between twoltelf operating modules or machines. The aim of the diagnosis will then belt) determine the site of the faulty component. and if possible the moduleI Its idreplace. Several techniques can be used for this purpose? and many of them can be done cost-effectively through a management network. The M S S themselves m a y c a r r y sufficient information. I n sophistic:OS implementations the analysis may be done with the help of a dataliMe where similar faults which happened in the past history of the system ea be found. and even with the help of an expert system. If the first analysis insufficient, complementary information can be obtained remotely by performing active tests on the machine. 583 Repair Once the illness has been identified. a cure must be prescribed. It may he a hardware repair. In this case, a maintenance engineer will go on site a replacement unit, run necessary checks, replace the (presumed) faulty module by the new one and run again some tests to check that the symptoms have disappeared, before putting the machine back to full operation. Ideally, the replacement and tests can b e done without interrupting the handling o f traffic through the concerned part o f the network. It is useful for the maintenance engineer to communicate with the management network to order some tests or to re-enable operation. Supporting such communications may indeed be one of the functions of the data network interconnecting traffic handling machines with the management network. In order for such replacements to be safe and efficient, consistency between hardware and software versions must be maintained. To this purpose, a database of the actual versions of all modules must be kept updated, and each module must carry an easy-to-get reference to its version. This is once again a function where management machines may improve efficiency. Other kinds of corrective measures may apply after a failure. If the problem i s a software e r r o r, t h e r e a l correction requires new—corrected—software t o be loaded. I n the meantime, software patches may be required to reduce the effect of the failure. Alternatively, modifications of the configuration are sometimes used to avoid going through the dubious routines. Similarly, errors imputed to inconsistencies between co-operating machines can b e corrected b y changing the configuration. In all cases a precise location of the fault is of foremost importance to minimise maintenance costs and the time during which service is either degraded or critically sensitive to a second failure. 9.2.0.3. M a i n t e n a n c e and the M a n a g e m e n t N e t w o r k The maintenance of a complex system such as GSM can represent an important part o f the running costs. A computerised system communicating with all traffic handling machines contributes to lower these costs. Although fault detection is mainly a function spread among the traffic handling equipment, the network management system can fulfil a number of useful functions such as alarm centralisation, remote testing, alarm analysis t o determine the nature o f the fault, failure database. trend analysis etr HIV (i!(51 s1 Li I liXt NFTWORK %%NAGE:M.:NT 585 9.3. MOBILE STATION MANAGEMENT remote commands II BSC O M C access to central facilities Figure 9.4 - Remote access to central maintenance facilities It is useful for maintenance staff to be able to he in contact tt ith the central maintenance facilities (e.g.. OMCs) \\bile on site. ;aid have for instance remote tests ordered by the ONUS to help localise failures. The mobile stations are often not the property o f the network operator. but they intervene heavily in the quality o f the service as perceived by the users. A badly designed or damaged mobile station may not only degrade the quality of the service for its user, but also degrade the quality of service for other users, for instance by being the source of unacceptable interference. The maintenance activity must take the mobile stations into account, since a symptom detected within the infrastructure may come from the failure o f a mobile station and n o t o f the infrastructure equipment. The correct functioning of the mobile stations is then a concern f o r the operators. However, the fact that the mobile stations are owned by the users and do not have any direct link with the subscription, and the requirement f p r a free mobile stations market, significantly limit the actions the operators can take. A first approach is curative, and consists of detecting failed mobile stations, to indicate the problem to the subscribers using them, and to list them so as to bar service. The second approach is preventive: the type approval procedure gives some control to the operators, through the regulation authorities, on the mobile stations which may be used on their networks. 9.3.0.1. Ty p e A p p r o v a l An integrated system to support maintenance activities will then include central facilities from where the whole network is monitored. Fitch of the machines in the network is linked to one of these facilities. Through these links. alarms can be forwarded. and remote tests to localise failures can be commanded. Another important feature i s t o allow communications between the personnel in charge of repairs on the site and a central maintenance facility. This facilitates the administration, provides the ability to command tests involving more machines than the ones on the site, permits access to the failure database to assist the diaonosis, and so on. Figure 9.4 shows an example of remote access to central facilities from a site. The maintenance functions are usually implemented together with a number of operation functions. For instance, Failure localisation may require a detailed knowledge o f the network structure. T h e implementation aspects o f the maintenance support functions will then be looked at after the operation functional aspects have been developed. In many countries. t h e connection o f a telecommunications terminal to a public network requires that this terminal be type-approved by a regulatory body. Until recently, such procedures were almost uniquely a national matter. For GSM, things are different. Full MSroaming. made possible b y the technical specifications, can only be achieved i n the real world i f common regulations are used b y all countries in the domains of type approval. marking, free circulation and use of mobile stations. Well aware of this goal. the European Community has produced directives concerning the mutual recognition o f type approval o f telecommunications terminal equipment. The first step in this process was a directive produced in 1986. whereby test reports produced by accredited test laboratories are mutually recognised, so as to avoid the need for each country to reproduce the tests for granting type approval. A further step was achieved in 1991 with directive 91/263 introducing the mutual recognition. not only of test reports, but of type approval itself. NETWORK MANAGEMENT 1E (ISM SYSTENI S86 user safe!). 1...1: ,alet o f employees of public telecommunications note Mks operators. t o electromagnetic: compatibility requirements in so far as they are specific to terminal equipment.. Id' Protection of the public telecommunications nemork from harm: eI et locti‘l: use of the radio frequenc.s sprctrmn. ‘‘ here appropriate: it) interworking of terminal equipment \\int public telecommunications netviork equipment for the purpose of establishing. modifying. charging 101. holding and clearing real or virtual connection: interworking of terminal equipment l a the public telecommunications network. in justified cases. table 9.1 - Essential requirements to he met for t?pe approval 587 mechanism, k n o w n a s a N E T ( N o n r i e s Europeennes d e Telecommunications). The NET involved an approval at the CEPT level. without any involvement of the European Community. The NET relevant to GSM is known as NET 10. In the GSM area, NET 10 is replaced by two CTRS, CTRs 5 and 9. CTR 5 is concerned with the interconnection of the mobile station to a GS\1 network and the non-disturbance of the latter by the mobile station. CTR 9 is specific to telephony and tests the end-to-end interworking capabilities of the mobile station. More CTRs related to the end-to-end interconnection for other services may be introduced i f these services become "justified cases". The coverage of the tests may differ when going from the NET regime to the CTR regime, but in both cases the main work is to produce a technical specification o f the tests, and t o develop test equipment capable o f performing the required tests. A test specification for GSM mobile stations is included in the Specifications, and this specification (TS GSM 11.10) represents the technical foundation for producing NET 10 and the TBRs. T h e D i r e c t ' s c t ) t / 2 6 3 o f t h e E u r o p e a n C O M M U n i t i e • slates, a l i s t of essential requirements ‘‘hich terminal equipment shall This directive defines essential requirements which must he met by the equipment. as reproduced verbatim in table 9.1. All these requirements are not enough to ensure that the equipment is functioning correctly for the service expecte'd by the person using it, but it certifies to a reasonable degree that the use of the equipment in a public network will not disturb the service provided to other users. In addition. services recognised a s o f prime importance may lead t o additional t e s t s o f t h e a b i l i t y t o reliably perform e n d -to-end interconnection: they are referred to as "justified cases- in the directive. Within t h e scope o f directive 91/263. -Common Technical Regulation,- tri-Rst are to he produced to describe the relevant checks. these ("Fits. whose approval lies mainly with LEVI' and the European Community (for member slates). include both technical and regulatory aspects. The technical contents. i.e.. the purpose and description of the tests. is referred to as a TBR (or -Technical Basis for Reaulation"); the TliRs are to he produced by the European Telecommunication Standard Institute. The existence o f the CI'R procedure being recent. the first type approvals granted to (ISM 'nubile stations are still based on a former TS GSM 11.10 covers most areas o f the Specifications affecting the mobile stations. The bulk of it concerns the signalling protocols. It must be stressed that i n this area, i n particular, there is no way to guarantee with a reasonable test time that a machine as complex as a GSM mobile station will behave correctly for every possible sequence of events. TS GSM 11.10 has been designed to guarantee to a reasonable level that the mobile station conforms to the set of GSM Specifications. The type approval tests must be conducted in a non-disputable Way. by accredited laboratories who are proven to be independent from operating and manufacturing companies. Although the actual connection of a mobile station under test with operational GSM networks may help to identify potential problems, such tests cannot be included in the type approval process. because no network can be considered as a neutral reference. The need for a reference test tool. recognised early in the process, led t o the development o f a single "system simulator" capable o f performing the whole set of tests included in TS GSM 11.10. However, commercial pressures to start GSM operation proved incompatible with the delay for the system simulator delivery. Hence, a reduced set of tests • was defined as a first priority, and a less sophisticated test equipment, S8 THE GSM SYSTEM NETWORK MANAGEMENT 589 (spare) serial number Figure 9.6—Structure of the IMEI The IMEI, ranges of which are allocated v o n the granting of type approval. references the mobile equipment type approval and the final assembly plant. and includes a serial number which is unique for each unit of a given type. structure shown i n figure 9.6. The Final Assembly Code (FAC) is intended to be used as an identification of the final assembly plant. Serial numbers are allocated by ranges to the manufacturer for inclusion in the produced mobile stations. Operators. through the permanent secretariat of the GSM MoU, are notified of the valid IMEls of type approved mobile stations they can expect on their networks. Figure 9.5 —The test equipment for interim type approval (by courtesy of Rohde & Schwarz.) This equipment is used to type approve mobile stations on an interim basis; before the full system simulator and full type approval procedure become available. It performs sonic 140 official 1TA tests. mainly in the areas of radio and protocol conformance, and more test sequences can he programmed on the machine. shown in figure 9.5, was developed to cover the initial needs, before the full system simulator becomes operational. The resulting "Interim Type Approval" (ITA) has only limited validity in time. Mobile stations which have heen granted interim type approval will be subject to pass the full type ❑pproval tests as soon as suitable test equipment and software becomes available. The granting of type approval to a mobile station type results in the allocation of a 6-digit Type Approval Code (TAG). which is part of a (,; i ( l o n i i t y number t h e IMFI). The I M E I has the 9.3.0.2. Mobile Equipment Management When maintenance activities detect a problem which are attributed to a mobile station, the network must determine the IMEI of this mobile station. The procedures on the radio interface are such that the mobile station does not volunteer its IMEI. The network must ask for it, through the RIL3-MM IDENTITY REQUEST message. Depending o n t h e implementation, this request can be systematic, at the beginning of each RR-session (but this creates a lot of additional signalling traffic), or can be done at certain occasions only, e.g., at each location updating and for a sample of call attempts. The knowledge by the operator of the IMEI of the (presumed faulty) mobile station will help to produce statistics and to determine the origin of the fault, in order to take corrective measures. In the worst case, all mobile stations of a given type and series might have to he removed from the market, i f the defaults are important enough to justify a retrofit. An advisory group consisting of GSM MoU members has been set up to exchange information on problems identified during commercial operation; these problems may reflect a need for new tests in 'HIE \ I 5 ' 1/4; I I \ I NETWORK MANAGEMENT (granting o f n e w t y p 2...,..,...,,,,..A approvals) PLMN Common EIR Visited PLMN EIR PLMN MSC VLSI ass update of IMEI status ''•• interrogation of IMEI status Figure 9.7 - The Equipment Idennt> Register (FIR) The status o f I M E l s is stored in EIRs. to he :pecked 11). the NISCs at any time. Operators 'nay interconnect their F I R . to update the stain,: u1 IMEIs. the type approval process. which would then he discussed with the appropriate authorities. The nest problem for an operator is to bar mobile stations which do not ,perate properly. i.e.. to refuse to grant them service (except possibly for emergency calls). This can also be applied to stolen mobile equipment (subscription barring is insufficient i n this case, since the equipment can be used with another. legitimate. S I NI). To achieve this, IMEIs for which problems have been detected must be recorded in a database. This database. referred to as the Equipment Identity Register EIRIEIRI in the Specificarions. is updated though the OSS. It can be accessed by the MSC/VLIts to check the status of a particular IMEI. This access is supported in the Specificolions by MAP procedures. Figure 9.7 shows ate position o f the EIR i n the system architecture. A s an example o f a realistic configuration. it shows two levels of EIRs. one at the individual network level. and one common to all networks. The interconnection of EIRs between themselves is not specified in the phase I Specifications, and is a matter for operators to agree on. The information stored in the ElR can be to a large extent operator dependent. However. the control, of. e.g.. stolen mobile stations, will 5 9 1 only become effective it operators with a roaming agreement have an agreed I M E I checking policy. This i s obtained f o r instance b y an harmonisation at the level of the GSM MoU. Within this group, operators have agreed to use three levels f o r the status o f the IMEI, and the .S.pecifications define three corresponding "lists" to be stored in EIRs. The while list includes the ranges of IMEIs allocated to type approved mobile equipment. Consequently, an I M E I not i n the white list does not correspond t o a valid type-approved mobile station. The black list includes the list of the IMEIs for stations which need to be barred, either because they have been stolen o r because o f severe malfunctions. The ,ercy list. as its name implies. is intermediate between the white and black lists, and includes the IMEIs o f faulty stations whose fault is not important enough to justify plain barring. The grey list can also be used as a iemporary buffer before authorities confirm or impose black-listing. The use o f a grey listed mobile station is reported to the maintenance system, together with the IMSI of the subscriber which used it. This can he used to notify the subscriber, or to trace malfunctions in the network. To detect persona non grata mobile stations the network has first to know the IMEI and then to check it in the EIR. A thorough IMEI checking is costly in signalling, because i t must be applied at every establishment of an RR-session. since the user may change his equipment at any moment. The traffic within the network can be minimised by storing in the visited MSC/VLR the IMEI of the equipment in use with a given SIM. in order to interrogate the EIR less often. The MAP protocol to support the exchanges between the EIR and the MSC/VLRs is called in this book the MAP/F protocol. This protocol consists in fact o f a single request-response procedure, using the MAP/F CIIEC'K IMEI message and its result. 9.4. SYSTEM ENGINEERING AND OPERATION The architecture of GSM as presented in all the previous chapters is an abstract functional architecture. Now, a network in operation is composed of actual machines. The operator must choose how many of each machine type to order, with which capacity, where to install them, to which o t h e r equipments t o connect them, e t c . Similarly, t h e 0)2 s Y s - i NETWORK NiANAGEMENT itNt cost o f the system (including investment costs and recurrent charges), while maintaining a balance between the traffic handling capacity at a given level of service quality and the number of subscribers. When all the uncertainties o f market forecast are borne t o mind, i n particular when several operators are i n competition, i t w i l l b e agreed that this i s n o simple task! .elecommunication l i n k s between machines must he dimensioned and provided for. I n addition. a number o f parameters must be set, as f o r instance t h e r a d i o channel configuration i n each c e l l (including t h e frequencies used and the number o f channels f o r each channel type), liando) cr parameters. priority levels. and so on. All these aspects are covered under the umbrella term "system cligineering-. N o t o n l y d o e s system engineering a i m a t h a v i n g a A way out is to refine the system engineering while in operation. This requires first the means to monitor the system, to know i f the level of performance i s adequate. T h i s i s t h e observation function. T h e engineering of the system can then be re-assessed with the gathered data, and decisions t o m o d i f y t h e network can b e made accordingly. A n important topic is then how to modify a network in operation. The is the subject o f n e t w o r k change control. Observation and network change control involve actions of the traffic handling machines, and are to a large extent performed automatically. These implementation aspects w i l l b e dealt w i t h together w i t h t h e maintenance implementation, i n t h e architecture a n d protocols section. T h e integration o f t h e n e t w o r k engineering t a s k s w i t h o t h e r o p e r a t i o n t a s k s i s n o t u s u a l i n telecommunications systems a t t h e date o f writing: m a n y operating companies use separate teams and tools f o r these tasks. However, the complexity of these tasks in cellular networks will be an incentive for this integration, and w e w i l l see i n the near future management networks handling the complete operation chain, f r o m observation gathering t o network reconfiguration through observation analysis. consistent a n d o p e r a t i o n a l s y s t e m . b u t a l s o t o o b t a i n a c o s t - e ff e c t i v e •vslein providiug an acceptable quality Of service to subscribers. The m a i n issue o f cost e ff e c t i eness i s t o adapt t h e network configuration t o the traffic. The system must he dimensioned to support the expected traffic w i t h the required quality o f service (probability o f :all loss. o f call establishment failure. etc.). but over-dimensioning must be avoided to avoid wasted cost. The right choice of equipment location, of e q u i p m e n t d i m e n s i o n i n g a n d o f l i n k d i m e n s i o n i n g a r e a l l interdependent. S y s t e m e n g i n e e r i n g a p p e a r s t h e n a s a g i g a n t i c optimisation problem. a i m i n g a t m i n i m i s i n g costs under t r a ff i c a n d quality of service constraints. The aim o f the first o f the following sections is to set down some basics for the system engineering o f GSM. W e w i l l concentrate on two main issues. One o f them is general to all telecommunications systems, and concerns switching equipments a n d links. T h e problem i s o f a topological nature. h consists in choosing sites. switching capacities and link capacities to support the traffic with a minimum cost. This issue will he dealt with in the next but one section. and not in a great level of detail. More emphasis w i l l b e put o n the other facet o f system engineering, cellular planning. which is specific t o cellular radio. and concerns the radio coverage. This area deals with the choice o f radio sites. with the attribution of frequencies in each cell and with many other parameters for cell selection and handover. The concept o f spectral efficiency plays a key role i n this domain. and w e w i l l present some o f the engineering points which impact the spectral efficiency o f GSM. Yet even the system engineering view presented above is static, and a hit narrow. Another dimension must he added. which is time. For any system. and in particular during the initial ramp up. the traffic does not remain constant. A n important subscriber growth at the start o f the system leads to a steep increase i n the traffic during the first years o f operation. One aspect o f cost is investment. and installing from the start a large capacity suitable f o r long-term traffic i s not cost-effective. The problem is then to optimise the tiepin) ment of the system across time, in conjunction \\nil the marketing pone). The aim is to minimise the overall 593 1 *-9A.1. CELLULAR PLANNING GSM is aimed at cellular networks of wide but dense coverage, and operators typically design their networks to ultimately reach a continuous coverage over a l l inhabited areas o f a country. T h e goal o f cellular planning is to choose the cell sites and a number o f system parameters such as the frequency allocation and the capacity o f the cells, in order to provide economically s u c h a continuous coverage, supporting t h e required traffic density. W e w i l l first examine more closely what is the problem. and then look at the different parameters to optimise. 9.4.1.1. G o a l s In a cellular environment, the traffic to support is best presented as a function o f the plane giving the local traffic density at each point. A s a first approximation. this function matches the population density. A n -important factor is the population penetration, which gives the ratio of the nuniher of subscribers over the total population in an area. s value varies. being higher in big towns than i n the countryside. The other intervening factor i s the traffic p e r subscriber. Taking a n average penetration factor of 55;, mid a traffic of 0.05 Erlang per subscriber, the large scale traffic density varies between ( i n non-populated areas) to, say. 100 I rIang per square kilometre. At smaller scales. the traffic density can reach much higher values. I f we go too far. we can obtain a traffic density o f 50000 Erlang per square kilometre. taking a square meter with one subscriber in it. whereas in the neXt square meter the density is 0 because there is nobody in it. The scale al which to measure the traffic !emir> intracadapted to the -sin. of the cells. and k‘e assume here that ith l itiM technology-. cells are not too small vi hen. say. they are at least 0.2 square kilometre. In reality. to complicate things even further. this traffic density series v ith time. Even it use do not consider its long-term evolution, the traffic pattern eNneriences short-term variations. according to a daily and ‘‘)•ekly cycle. t stall >. telecommunication operators dimension their networks according to the "peak hour. t rid tic. \\lien the system is closest to congestion. In a first approximation. a network supporting the peak hour traffic will support the traffic at any time. This is in fact not always true. since the geographical distribution o f traffic may he substantially Different from one hour to another. A s an example. the traffic around spoil Ileitis may be much larger during i‘eek-ends than during the week. but this happens when the total traffic of the whole city is much smaller than durimg the rush hour of week days. The first goal—or constraint- - o f cellular engineering is that each cell i s able t o support (for most o f the time) the traffic from the corresponding geographical lone. Nox‘; there are a lot o f solutions to such a problem. and an obvious one ctinsists in installing a very large number ()I' cells. The other goal of cellular engineering makes the picture less simple. and asks that the cost of the tadio infrastructure he minimised for a given traffic load. This cost has two aspects. and it is important to differentiate them. l'art o f the cost depends directly on the traffic. For instance, the total number o f installed speech channels can i n a first approximation be derived directly from the total expected traffic. whatever the distribution over different sites. This part of the cost can then be considered as a fixed one itir a given total traffic, and cannot he decreased otherwise than by reducing the cost of each equipment unit. L'sing the GSM terminology, it means for example that the cost of a 'URN can he neglected in the first step towards optimisation of the system parameters. The number of traffic channels (and of TRXs) d a d s at second order on cell planning because of the Erlang law. More precisely. more TRXs must be installed than just the minimum number to support the expected average traffic, allowing for normal statistical peaks. The analysis of this problem. embodied in the Erlang law. shows that the ratio between the number o f installed resources and the average traffic they can support decreases when the number o f resources increases. I n practice, this means that the total number of traffic channels installed in a network is the sum of the number of channels supporting the average traffic, plus some quantity which increases when the average capacity per site decreases (i.e., when the number of sites increases). The inclusion of this factor in the optimisation process directs the choice towards less cells of larger capacity. A The second aspect o f cost is not linked directly to the amount of traffic. but to the number of sites. Each individual site is a source of cost for the operator. independently from the traffic it supports. Not only does each site include equipments which are to a large extent independent from the traffic (e.g.. antennas). hut real estate costs and installation costs (e.g.. air conditioning. power supplies) are far from negligible compared with the costs of the electronic parts. Cellular planning will then attempt to minimise the total number o f sites, given the traffic distribution constraints. 9.4.1.2. R a d i o C o n s t r a i n t s The cost optimisation problem i s not an easy one. The cost considerations alone lead to having as few sites as possible, ideally a single (enormous) cell. This is however not feasible, for two reasons. The first one is the amount of available spectrum. The provision of a number of traffic channels in a cell requires some minimum amount of frequency, which depends o n the system. I n GSM, 25 k H z per simultaneous conversation is required with full-rate speech, for instance. The support of 3 million GSM subscribers, each of them generating 25 milli-Erlang at the peak hour, would require a spectrum allocation o f 1855 MHz, whereas 2 5 MHz a r e available f o r GSM900! Multiple cells a r e unavoidable for any big radio network. But spectrum scarceness is only one o f the reasons for cellular networks. A s explained in Chapter 1. the radio transmission range is limited by the maximum transmission power of mobile stations and by noise. Maximum cell size, and hence the total number of radio sites in the system, is therefore affected by two factors: the transmission range and ('SXI SI SI 1-.1I i9() NETWORK MANAGEMENT 597 the interference introduced by the reuse of a radio resource . s m a l l to accommodate the total taffic. Range The emission power o f mobile stations is limited t o 20 W for ISM900 vehicle-mounted stations, t o 2 W f o r (ISM900 handheld siatiairiThial to I W for I)( 'ti1800 handheld stations. On the other side, the sensitivity of base stations is limited because of noise factors, and in particular because o f the thermal noise introduced in the transmission chain. Because o f this combination o f an upper limit' for the emission power and a lower limit for the reception level. the propagation of radio waves intervenes to limit the coverage. Radio waves have different means to go front one joint to another. The first one is direct "line of sight". The power loss clue to propagation varies in this case with // ( ( I being the distance between transmitter and receiver). provided no obstacles are too close from the line of sight. Other possibilities include reflection and diffraction. as illustrated in figure 9.8. At 900 MHz and even more so at 1800 MHz. a reflection or a diffraction leads to an important propagation loss. On the other hand. since both the base station and the mobile station are fairly close to the ground, a lot of obstacles usually hinder the direct line of sight between them, and they can communicate only through reflected o r diffracted paths. A s a consequence. the mean loss increases rapidly with the distance in uneven areas such as the centre of towns. The power loss is modelled typically as an exponential of the distance t a . with typical values of a in the range 3 to 4. depending on the town. The matter is a little better in suburban areas, and even more so in flat rural areas. In the latter case, however, the cells are large because o f traffic, and the roundness o f the earth also intervenes to encumber the direct line o f sight, and even more so i f antennas are close to the ground. So far, we have only considered the average propagation loss. Another factor makes things worse. Users want t o use their mobile stations everywhere. including inside buildings. or even in underground car parks or deep basements. Even at short distance from the base station site. there are places such as these where the propagation loss is high enough to prevent communication. More generally, for a given distance, the propagation loss follows some statistical distribution. By definition, only in 50% of the cases will the propagation loss he less than or equal to the median propagation loss at a given distance from the base station. This value does not hold much interest for an operator, whose aim is to uuarantec a 90,7( o r 95(i coverage. measured over the places where a diffracted path reflected path direct line of sight Figure 9.8 - Propagation paths In most cases, propagation for land mobile radio systemsdoes not follow adirect line of sight (a), but encounters many obstacles which reflect the radio waves (b) or diffract them (c), leading in both cases to losses much higher than with direct line transmission. customers are likely to want service. As a result, an operator must take into account an important margin from the mean propagation loss. The spread o f the propagation loss around its median value depends on geography, on the height of the base station antenna and on .places where customers are expecting traffic. This last point is very tricky. since i t depends on the subjective point o f view o f the users, influenced in particular by the indications provided by the operator or the service provider. I f it is clear for users that they cannot expect coverage in underground trains, these places can be taken out from statistics; i f users know that they may have to move by a few meters when using their terminal indoor, the indoor part o f the statistics has to be corrected accordingly. A f e w years ago, w i t h o n l y vehicle-mounted mobile equipment, things were simpler, and measurements in the streets were sufficient to have an idea o f the propagation loss statistics. From this period we inherited the classical log-normal model, where the distribution of the propagation loss for a given distance is modelled as a Gaussian distribution when expressed in decibel. This is the model assumed in TIIE (IS\I SYSTEM 598 NETWORK MANAGEMENT 5 9 9 Interference • c h i n e c e i v e d leve.. .ilng Interference precludes the use of the same GSM radio channel for two simultaneous communications in the same cell. Furthermore, because the signal level does not become completely null until a long way off, interference exists between any simultaneous use o f the same radio resource. even i f from distant places. I n a cellular system, since the spectrum resources are by necessity reused in several places, the goal is to ensure that the interference caused by this reuse is negligible, or at least statistically acceptable. :14 M • Figure 9 . " • The propagation sl:ni.tival mi‘del The gee' le% el is proportional to the probahility. at a .,iven dikiance from the emiiiint kite. o f a gi% en p r o p a a t i o n loss. .1 he attenuation Nk ith diaanec Is t h e lo- n o r m a l st.indard deviation is S dR l o r all distances. and the median reception le\ el at i k m is - 9 2 dBm. figure 9.9. and the one we will assume in the rest of the chapter, though it can be debated at length if it is adapted to handheld dominated traffic. With the log-normal model. the spread can he measured by a single value. the standard deviation. usually expressed in decibels. As defined it mav van' with the distance: how.es er. the measurements (in the streets) show that i t can be assumed constant, at least f o r distances over a kilotnetre. The standard deviation cy is greater in towns than in rural areas. because of obstacles. :\ typical value of the spread in urban areas is cs = 4 tIll. With such a value, a 95,e; coverage of the cell can be reached with a propagation loss around the median reception level value at the cell radius. When all these factors are taken into account. the cell range in GtiN1900 is limited to a few tens of kilometres in rural areas, depending on the terrain and on the base station antenna height. In urban areas, this limit conies down to 5 to l e km. except for handheld stations, for which the limit is a few kilometres at best. The impediments o f propagation as explained i n the previous sections are clearly an asset f o r reaching this goal. Cellular systems would indeed not he so easy i f propagation did not imply a loss varying significantly quicker than d-2 ! An obvious way to reduce interference is to increase the distance between machines using t h e same radio resources. As a negative consequence, the area in which a given radio resource may be used only once is larger than a cell. I t extends to neighbouring cells as well. The minimum number of cells able to support a given amount of traffic T is then not equal to the ratio T/S (where S represents the traffic supported by a single cell using all the available spectrum). but to T/S divided by some value R called the average reuse factor. Correspondingly. the maximum number of resources in a cell is on average the total number of resources divided by R. As an exarnple, an operator using 12.5 MHz (half the primary band of GSM900, 62 x 8 fullrate speech channels) and a frequency plan based on a 7 reuse factor will handle at most 70 simultaneous calls per cell on average. II' the traffic density is high, this capacity l i m i t may lead the operator to build smaller cells than made necessary because of the range problem alone. For instance, a traffic density of 20 E/km2 (a rather high value. but not excessive for large cities) would lead the same operator as above to limit the cell size to 3 km2 in order to offer an acceptable quality of service. This corresponds to a cell radius of about 1 km. These elements show that a cell is limited in size, either because of the range or because of interference. I f range is the limiting factor, more frequencies are available than necessary to guarantee both a far-enough distance f o r reuse and the required quality o f service i n a cell o f maximum range. In the other case, on the contrary, the minimum reuse factor must be used and cell size must be reduced to accept the planned traffic. For a given system, it is the traffic density and the available amount o f spectrum which dictate the applicable case f o r each cell. Typically, a mature network will include both types of cell, the first in rural areas and the second in urban areas. NEI-WORK MANAGEMENT 1111 t 1 e S A I : N I : 111 * % 1 M t / 6 0 1 The analysis so far has shown that the two key factors for cost efficiency are the possible range, and the reuse factor in zone B. Little can he done for the range at the level of cellular planning, since this range is dictated by propagation conditions, mobile station emission power and BIS sensitivity. Conversely, the reuse factor is an important parameter. and it will take us into the domain of spectral efficiency. cost Erlang (log scale) range fry:a:c1 area 9.4.1.3. Spectral Efficiency B TRX part of cost propoitional to traffic Erlang/km2 (log scale) Fiume Q 10— Number of sites per lirlang as a function of the traffic density the evolution of the cost (per hrlangi a ith the traffic. density identifies three types of cell configuration: - in tone A. cells are limited by the range: in /one 13. cell, use the maximum capacity for a given reuse factor: - in tone C. cells become too small and the cost efficiency decreases. It is interesting to measure the investment as a function of traffic. It becomes even more interesting when noting that the operator's income is basically proportional t o the traffic. Part o f the investment i s also proportional to the traffic. This part contributes to a fixed portion of the cost per Erlang independently from cellular planning choices. However, another part o f the investment depends on the traffic density, and the resulting cost i n terms ()I' number o f sites per Erlang i s shown i n figure 9.10. Zone 13 in figure 9.10 corresponds to areas where cells use the maximum capacity. The number of cells in relation to the traffic is fixed and is dictated by the average reuse factor of the system. In zone A, the size o f cells is dictated by the range only. and the traffic per cell is proportional to the traffic density. Asa consequence. the number of sites per Erlang i s inversely proportional t o the traffic density. Zone C corresponds to cells which become too small. When the cell size has the same order of magnitude as the size o f obstacles to propagation, some problems appear (which will he dealt with later). A higher reuse factor is then required compared v‘ ith /one B. and die cost efficiency decreases. There are many definitions o f spectral efficiency. When the problem is considered from the cost point of view, the key factors are on one hand the cost of BTS equipment, and on the other the number of sites to install. which depends mainly on the capacity of one cell when it is limited by interference (and not by range), as in zone B of figure 9.10. While the main thing to optimise is the cost, which combines the two factors. the spectral efficiency we will study here is the second term, that is to say the number of sites to install with a given number of MHz and a given traffic density to support. The right unit to compare the spectral efficiency of different networks is then the number of communications per cell and per MHz. The figure combines two aspects: • t h e number of communications per MHz i f the network was composed of a single cell: for GSM900, this figure is less than 40 with the full-rate speech coder and less than 80 with the halfrate speech coder; • t h e reuse factor, which depends on the cellular planning o f each operator, a n d which i s constrained b y interference limitations. The reader may wonder at this stage what is the _reuse factor of GSM? Until now. we have been careful not to mention any value, except as examples. Why? Simply because there is no ready-made answer. The reuse factor depends o n a variety o f operational choices, and i t s evaluation with given assumptions is still not easy. In the past, a value of 9 circulated among the GSM arena. But this value does not take into account a number of features which lead to substantial improvements of this figure. Our goal is not here to give a definite value, but to analyse the influencing factors. Relative Interference Levels (C/I ratios) For a cell which is part of an operational system, it is possible to measure the relative interference levels caused by the environment during :I Communication. These levels are usually expressed as the ratio of the received signal level from the wanted source (carrier level: - ) to the interference received level (interlerence level: I t . o r CY:. and arc expressed i n d B . C / / ratios follow some statistics. which arc best visualised by showing their cumulative distribution. as in the examples of figure'). 1 I. The distribution in a given cell depends on the locations o f the mobile stations in communication with it. and on Ihr locations o f the inlet tering sources: hence i t depends o n cellular planning and o n frequency reuse. In order to ensure an acceptable quality o f service to subscribers. some objective must he put on the minimum ('// ratio. This object _ c a p h e expressed a s forbidden areas i n t h e cumulative ,Ikiiibution graph. For instance. a criterion can he that at least 9W% of the communications have a quality above some given threshold C//90, as represented in the figure. where we have chosen for this threshold a value of ( 1 1 i . which corresponds t o a transmission quality close t o the maximum with GSM run rate speech. ()tiler criteria can be thought of e.g.. a C//50). but usually a "worst case- criterion is the only one taken into account. There is in fact no real incentive t o provide a better quality than the minimum acceptable, because o f t h e consequential costs. T h e actual g o a l o f a telecommunication network is not to maximise quality. but in fact to minimise the cost. (and hence the quality!). whilst keeping the quality above some threshold. The quality criterion provides the ;think t o translate a C H distribution into a minimum reuse factor. The propagation model we presented in a previous section has the particularity that an homothetic modification of emitter positions. of ratio r relative to the reception point, changes the reception levels by a simple multiplication factor of r-cx (a being the value introduced earlier as the exponent ‘‘ ith which the level decreases with the distance). Then we can take the assumption that the cumulative distribution o f I (in d13) is simply shifted along the / axis when the reuse distance varies. With (I.= 3.5. the shift is of 2.5 dB for a multiplication of the reuse distance by 2. The minimum reuse distance can then be derived from the / distribution determined for another reuse distance. For instance. i f the C// cumulative distribution is as shown in figure 9.11. case b). the reuse distance is too small and must be increased to compensate 6.5 d13. i.e.. by a factor of 1.5 i f propagation varies with (/ ;.5 (this corresponds to an increase of the reuse factor of 2.3). Probctility 1.0 0.8 0.6 b 0.2 0.0 -10 0 7 10 2 0 3 0 4 0 5 0 C/1 (dB) Figure 9.11 - C// cumulative distribution 1he inuidative distribution of the C// statistics depends on cellular planning. ytialit o f service criteria lead to forbidden areas of the graph: for instance, the criterion "90C/ o f the calls must experience a C11 better than C/Nothere C/I90 is equal to 7 dB) is a "worst case" constraint represented by the forbidden grey area. i a The C// distribution is outside the area and respects the constraint, whereas. thi the C:1 distribution is not compatible with the quality of service criterion. In order to achieve the required quality, the reuse distance must be increased to shift the distrilmiiim. In the example shown. 6.5 dB must he gained for the I 0% worst cases. The statistical distribution o f C// could be derived from the statistics of C and from the statistics of I, i f these two distributions were independent. They are on the uplink (from mobile station.to base), not on the downlink. but some insight can be obtained by studying each of them. C' varies with propagation fluctuations and with the distance between mobile station and base station. / depends in particular on the distance between interfering cells, and hence on the reuse factor. These two distributions will now be studied in more detail, especially with regard to the features which influence them. The two directions (uplink, mobile to base and downlink, base to mobile) are not equivalent as f a r as the interference statistics are concerned. This state of things conies in particular from the fact that a I 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% • • a cell as a function of the "cation. Using this concept, a typical cell is shown in figure 9.12. A t j cell centre, all communications are held within the cell. but this probability fades away when the distance increases. the calls then being supported by neighbouring cells. What determines the cell area is not only propagation with the given base station (this i s true o n l y f o r isolated cells), but also propagation with the neighbouring base stations and handover criteria. Handover moves a communication from one cell to a neighbouring one, and thus influences the cell function as described above. Different handover algorithms result in different C distributions, and have a small impact on the / distribution. • Impact of Handover • • Figure 9.12 - An omnidirectional cell (using an antenna emitting evenly in all horizontal directions) viewed as a probabilistic function A cell "area" is best considered as a function of the location giving the probability with which a mobile station at this location will communicate with the relevant base station. The black clots indicate the location or the neighbouring base stations. mobile station receives interference front a small number of fixed sites (the base stations). whereas a base station i s being interfered b y a potentially great number o f mobile stations moving around inside the interfering cells. As a first approximation. however, we will ignore these differences and look how the features of the system influence the C and the / distribution. The C distribution is determined firstly by propagation statistics and secondly by the area in which the mobile stations are located while in communication with a given base station. The second point is interesting, and is not that obvious. What is exactly the geographical area of a cell? We have been careful not t o mention the boundary o f a cell as a measurable quantity. since there is no such simple border (not to mention the fractal dimension o f borders in general). There are places where a mobile station in communication can he in one or several cells depending on its past history. in particular depending on the direction in which it has moved to get there. Such overlapping areas are fundamental f o r the operation of any cellular system. The correct approach to describe a cell "area- is to consider the probability of being in communication with such In order to best visualise the impact o f handover on the traffic distribution between cells, figure 9.13 shows the cumulative distribution of C in three different configurations. The first curve represents the case 'Probability 1.0 0.8 0.6 a / b , 0.4 / d B / 0.2 0,0 -120 - 1 1 0 - 1 0 0 - 9 0 - 8 0 - 7 0 -60 - 5 0 - 4 0 received level C (dBm) Figure 9.13 - Impact of the handover strategy on the C distribution The cumulative distribution of the carrier level is shown here in three different cases: a) no handover (isolated cell): bi distance-controlled handover (cell with defined boundaries): CI handover based on the comparison of received levels (basic propagation model, reuse factor 4). V V / of a typical urban cell. it' it was isolated (no neighbourint i l l s ) . For simplicity, the distribution is limited to a distance ()I' twice the size of the cell as defined in the next case. The second curve corresponds to the same cell assuming that cell boundaries were geographically perfect: this case could he approached with a 11:111(101/2Cr criterion based on precise distance measurements. The third curve represents a more realistic situation. where the cell chosen b y the mobile station i s the one i t receives best. with an hysteresis of 6 d13. The curve shows that. in this ease. the H Y; ( ' value is improved by 4 dB compared t o the value computed is hen geographical allocation is applied. The handover strategy corresponding to the third curve consists in comparing received levels. T h i s i s an excellent handover strategy, referred to by some as "mobile station assisted handover" and supported by GSM. The example o f a handover algorithm provided in TS GSM 05.0S uses this principle. Since the choice o f the handover algorithm is left open to operators and developers. we have here a first case where operational choices influence spectral efficiency. Mobile station assisted handover has little influence o n the distribution o f the interference level. when no power control is used. Things are different when the two features are combined. and this will be considered alter sonic study of the impact of power control itself. Impact o f Power Control Handover strategy is not the only factor which influences the statistics of C. Another orte is power control. This feature is provided in GSM w i t h a dynamic range o f 20 d13 o r more. depending o n the maximum emission power of the mobile station (the mobile station power class). Its impact on the C' distribution is due to the reduction o f the transmission power in some cases. The reduction o f the transmission power depends on the level which would be received i f there was no power control. Typical cumulative distributions of C' are shown in figure O.14. Not only do they show the impact of using power control compared with not applying it. but they also show the impact o f different power Control strategies. Two parameters have to he chosen by the operator. One determines the reception level threshold above which power control • is applied, and the second is the maximum accepted emission level, . which call he chosen to be lower than the maximum emission level of the mobile stations. Tw o choices for these parameters are represented in curves h) And c) of figure 9.14. The parameter values should he chosen as part of the general cellular planning optimisation. Let us turn to the interference distribution. The interference level received on a given channel is the sum of the contributions from several • Probability 1.0 0.8 0.6 a 0.4 0.2 0.0 • . -120 • 1 1 0 - 1 0 0 - 9 0 - 8 0 - 7 0 - 6 0 - 5 0 - 4 0 received level C (dBm) Figure 9.14— Impact of power control on the C' distribution The cumulative distribution of the carrier level is shown here for three different cases: at no power control. maximum emission power; hl power control applied to all the population except the 10% worst received.' for which emission at their maximum power is tolerajed: ci as N. but the maximum authorised power is lower. sources. Before looking at this sum, we will study the statistics for one contribution, which is the same for the uplink and downlink. Figure 9.15, curve a). shows the cumulative distribution of the interference level from one communication using the same frequency in a cell a 3-reuse-factor away. with no use of any special feature. Having seen how power control influences the C distribution. we must now consider its effect on the / distribution. which is even more important. With power control, a large number of mobile stations in dedicated mode transmit with a power level below their maximum power capability. This leads to an important shift of the / cumulative distribution toward lower interference levels, as shown in figure 9.15. curve b). When power control is used in conjunction with "mobile station assisted handover", the improvement on the I distribution is even greater, when compared with distance-controlled handover using the same value of the power control parameters. This comes from the fact that a choice of cells based on the reception level will favour high reception levels, favouring situations where the decrease of power due to power control is Discontinuous t r a r nssion provides t h e a b i l i t y t o reduce transmission to a low activity cycle when the user is not effectively generating a traffic flow. For speech communications, this applies for typically 40% o f the time in each direction. The use of DTX in such conditions changes the interference distribution from curve b) to curve c) of figure 9.15. Probability 1.0 0.6 Impact of Frequency Hopping Let us now consider the last influencing factor, frequency hopping. Usually several sources contribute to the interference level experienced by a given connection. When frequency hopping is not used, the number of sources is fairly small (typically between 2 and 6 interferers). With such a small number of interferers, it requires only one to be sufficiently high for the connection to suffer a bad quality. With frequency hopping, things become quite different, at least in the mobile to base direction. It is 0.6 0.4 0.2 , / 0.0 ' -140 - 1 3 0 -120 - 1 1 0 - 1 0 0 - 9 0 received interference level , from a single source (dBm) Probability 1.0 7 1 0.8 j Figure 9.15 - Impact of power control and DTX on the / distribution the cumulative likiribiniOn of the interference le% el is shifted towards lower interference values ‘s lien power control or a r x is used: (a) without power control Or discontinuous transmission; (t)) with power control. without discontinuous transmission: lc) with power control. and transmission is effective 60ri o f the time. the largest. What has been called "confinement handover in Chapter 6 improves greatly the C// statistics when used together with power control. Impact o f Discontinuous Transmission (DTX) Another feature which has a great influence on spectral efficiency is discontinuous transmission (DTX). which has been presented in Chapter 4. DTX is an option controlled by the operator. and which may be used independently in the mobile to base and in the base to mobile directions. 0.6 0.4 j 0.2 0.0 = — -140 - 1 3 0 - 1 2 0 -110 - 1 0 0 1 -90 received interference level , from a single source (d8m) Figure 9.16 - Impact of frequencc hopping on the / distribution Because of the diversity of interferers which is generally assumed when frequency hopping is applied, a cumulative distribution o f t as shown applies to the bursts of any given connection. 611 possible «) choose the hopping sequences in a v ay such Ma! a c h mobile station in a cell interferes a little bit with many communicai,,,ns in other cells. This is called "interferer diversity and the corresponding choice of suitable hopping sequences is explained in Chapter 4. With this property, the probability to be interfered with by a strong interferer is not at the connection level. but at the burst level. I.et us make sonic simplistic assumptions to outline what happens iSNI, assuming that a burst is all good or all bad (this is obviously simplistic). The coding scheme used for speech is such that a speech frame iwtorreei i f at least 5 hursts•are received correctly among the 8 hursts carrying information for this frame. Let us decide that the quality Icy el aimed at for 90'4 o f the cases is a frame erasure rate o f 5%; the corresponding acceptable burst erasure rate is then 19%. Figure 9.16 shows the I distribution when six interfering cells are taken into account. The corresponding C' and C/1 distributions are those in figure 9.14, case b. and figure 9.11 respectively, and the distribution for a single interferer is the one given in figure 9.15. case b. Assuming frequency hopping with a huge number o f different interferers the I curve represents the distribution o f I f o r the bursts y l any connection. O n this curve, a percentage of 19% burst erasure is obtained for 1 equal to —103 dB, thus requiring C' to be above —96 dB in 90C; o f the cases (if we assume that the threshold between had and good bursts is 7 dB). This is 4 dB less than what is required without frequency hopping (-92 dB. as derived from figure 9.14). O f course, these computations cannot totally reflect the reality and the gain is somewhat smaller because the bursts are not all good or all bad, and because the number of interferers is not so large. Sectorisation We have seen before i n this chapter that the investment cost - depends on the number of sites to be installed for supporting a given traffic. A way to reduce cost is then to use the same site for several cells. Before going further, it is important to note that the word cell is used here (and in GSM in general) to refer to the unit o f choice as seen by the mobile station. A cell corresponds to one beacon frequency on which an SCH and an FCCH are broadcast. In the North-American literature in particular, the term cell is used to refer to the area covered by all antennas of a given site, and the term sector is used to refer to what we call "cell" here When a site is used for a single cell. it is usually at the centre of the cell area, and the radio antennas are omnidirectional in the horizontal plane. Now, it is possible to use directional antennas. which cover only a sector in the horizontal n e . With such antennas, several cells can be covered from the same site (3 o f them typically). The impact on the spectral efficiency when measured in terms of traffic per cell is negative. Though there is a gain on one hand because interference is limited to a sector, o n the other hand the ratio between the maximum distance between the BTS and a mobile station in the cell and the square root of the cell surface is doubled, which worsens the C statistics fora given cell area. These two effects can at best compensate for each other, but in practical situations the negative side wins. A reuse factor of 9 with 3sector sites (hence with a reuse pattern of 3 sites) corresponds roughly to a reuse factor o f 7 with omnidirectional cells. Now, i f the spectral efficiency is evaluated in terms of traffic per site, there is a gain factor equal to the number of cells per site, and the balance is positive. In the previous example 3 sites have to he installed instead of 7, resulting in a substantial economy for the operator. Conclusion: GSM Spectral Efficiency All the factors introduced so far can be used in computations aiming at minimising the investment cost for a given quality threshold. Our purpose here is not to push such computations to their limit, but to show how the different factors influence the result. It should now be clear that giving a value for the reuse factor of GSM requires first to state the choices about all the identified impacting features. The value o f 9 determined in the past does not take power control and mobile station assisted handover into account, n o r discontinuous transmission, n o r interferer diversity brought about b y frequency hopping. I f all these features are used, the reuse factor becomes much better, maybe below 5, making GSM with full-rate speech a system o f high performance, and with half-rate speech the best system yet developed in terms of spectral efficiency. 9.4.1.4. Ta s k s o f Ce llular P l a n n i n g Let us now come back t o cellular planning, and look at the different steps an operator must perform to make a cell plan. The basis of the optimisation process is the assessment of the traffic density which is aimed at. This determines, together with the available spectrum and the expected spectral efficiency, the cell sizes and capacities. A t this point, the operator must choose whether the cells are omnidirectional o r sectored, and in the latter case the number o f sectors covered from a given site (from 2 to 6. very often 3) and the directions of the antennas. These different data determine the density of sites. The next i ) , and not the least one, is to find the sites as best as possible to cover tne required area. In areas where planning is constrained by coverage. the choice of the sites aims at minimising the uncovered area. hi areas where planning is constrained by interference, the coverage at the street level is usually not a problem. However, an over-dimension of such an outdoor coverage is useful to improve indoor coverage. A compromise has to he found between improving the coverage o f indoor premises. and limiting the interference level generated outside the cell. The goal is not so much to have a geographically extended coverage than to match the covered area with the planned cell. Next. the following parameters must he chosen: • t h e frequency allocation. including frequency hopping considerations: • t h e power control parameters: • t h e handover parameters, chosen to fit exactly the cell to the expected traffic. In areas limited b y t h e range. the goal i s simply that n o dvsfunctioning is introduced by an incorrect choice of parameters (for instance, it must be ensured that a connection handed over to another cell will not he immediately handed back). In interference-limited areas, in addition, these parameters must h e chosen s o that interference i s minimised and that the cumulative distribution of C// in each cell respects the quality criterion (e.g., nek of the traffic above a given threshold). As already noted, an important factor is the cell "boundaries", which are determined by the handover parameters. Different choices will change the distribution of the traffic between cells, and change the conformation of the cells. It should be noted that these operations are inter-dependent, and that the global optimisation problem is a very complex one. As already mentioned at the start of the system engineering section, medium- and long-term evolution also intervenes, as well as financial aspects. A last step, of marginal impact on the spectral efficiency. is the choice of the parameters controlling how a mobile station in idle mode chooses the serving cell (for instance the parameter CI described in Chapter 7). The goal is usually to have the cells as defined by the idle mode selection as close as possible as the cells determined by the handover criteria. The idle mode selection parameters also influence the quality o f service, in so far as they act to prevent access when (plain) is so had that a call will be dropped or handed over immediately. A correct setting of the cell selection parameters ‘‘ ill therefore delay the first 11;111(1m:cc of a communication. Similarly, the choice o f the beacon frequencies is part o f cellular planning, but does not impact spectral efficiency. The beacon frequencies are chosen so as to obtain as wide a radio separation as pot. ' e between two cells using the same beacon frequency. 'Ihe -reuse- factor can be as hie.. as the number of frequencies available to the network. The deployment o f a network is typically an on-going process. Cells are introduced progressively and t h e capacity i s gradually increased. Cellular planning is then a continuous activity throughout the life of the system. I f some o f the parameters are difficult to change in time (such as the location of existing sites), many others can be modified by remote control. enabling the operator to re-configure part of a network with m i n i m u m — i f a n y —service interruption. S u c h changeable characteristics include power control and handover parameters. With some ICIS implementations. a remote change of frequencies is possible, enabling an operator to change the entire frequency plan overnight, for instance. The general optimisation is very complex and requires automation and centralised computations. System observation and performance analyses need t o b e done regularly, leading t o new configuration attempts. By such computer-assisted trial and error methods, the cell plan ----approaches its optimum. thus leading to cost reduction 'or conversely traffic increase. 9.4.2. CELL CONFIGURATION In this section and in the next, devoted to network engineering, the focus is given to the dimensioning aspects, that is to say how to choose the number of transmission devices, the switching capacity, and so on, of the different pieces o f equipment. These activities require some knowledge about the behaviour of the subscribers: how many there are; where they are when communicating; their movements; when do they attempt calls; what i s the duration o f a call; and so on. A l l this information is of a statistical nature. It is part of a traffic model, and any network engineering activity implies such a model. So, before looking to the detailed points, we will present a typical traffic model, which will be used afterwards for providing examples. Traffic Model A traffic model is a mixture between observations in the field, either on the target network, or on functionally close networks, and of assumptions. some of them quite arbitrary. The model presented in this section is by no means the ultimate traffic model for a GSM network. The figures most often found in the literature come originally from analog cellular nets arks. and may he quite obsolete already now or at least in a few years time. It should be understood that the goal here is t i t o replace the traffic modelling work which i s the first step o f n e t w o r k engineering. It is rather to give sonic general ideas on the process. and we ask for the understanding o f the readers who might consider some of these figures inadequate. A general point first. A l l traffic data varies i n time. and the statistical characterisation ()I' such phenomena is not easy. A general evolution exists on a time scales of ears. This will not be our concerti hero._Other variations exist at the time scale o f the month (seasonal ariations) and at the time scale of the hour ((tail) variations). : \ network is designed to cope with a peak traffic situation. The traffic model is given for a peak hour. chosen by consideration o f quality o f service. variations also exist at a small time scale. over minutes or less. They are considered as statistical noise. and averaged. In a cellular system there are two main categories of traffic data. The first batch relates to the communications, and the second to the users movements. ConlffillIliCati011s Beside the number of subscribers. two values are used to estimate the communication traffic. One is the traffic per subscriber. defined as the average probability that a given user is engaged in a conversation at a given moment during the peak hour. As for all occupancy averages, the unit is the Erlang. The value used in GSM studies around 1988 was 0.025 Erlang, a value observed with cellular subscribers at this time. This value will necessarily increase with the success ;of cellular telephony. We will take here, arbitrarily. 0.05 Erlang. The other basic data is the mean duration of an effective communication. The usual value is 120 seconds, i.e.. 2 minutes. Thettraffic evaluation in terms of Erlang is well adapted to transmission. when the number of channels or of circuits must be chosen. In switching arenas. where the stress is more on the signa11 ing load than on transmission line occupancy. another related value i used to assess traffic: the number of call attempts per hour, measured in BHCA (Busy Hour Call Attempts). This includes all attempts. whether coming from or targeted to mobile stations. and whether successful or net. This figure in particularly useful to estimate the load in terms o f signalling processors. To derive the number o f BHCA p e r subscri4r, more parameters are needed in addition to the average call duration and the traffic per subscriber. I t i s usually assumed that i n public' cellular telephony there are more effective communications sent from the mobile station than towards it. This is because people rarely give their cellular phone number as the first number to try to reach them. Typical values are 6054 o r 70% o f effecti,— calls being mobile originating calls. We will take here 60%. Now comes the relationship between call attempts and effective communications. For mobile originating calls, the PSTN figures will do. For 100% call attempts, 60% succeed, 35% fail by lack of answer, and 5(4 fail before alerting starts. For calls toward mobile stations, a first point is the probability for the mobile station to answer when called during the peak hour. that is t o say t o be switched-on and within coverage. We will assume a 60c/i figure for this probability. We will also deem I this is an approximation) that the user always answer if his mobile station indicates a call arrival. With such hypotheses, there are per subscriber 1.5 mobile originating BHCA, 0.6 mobile terminating and successful BHCA. and 0.4 mobile terminating BHCA towards non reachable 'nubile stations. User Mobility Handover rates and location updating rates depend o n t h e movements of the users. The estimation of this signalling load must be based on statistics concerning these movements. Little is provided in the general literature t o q u a n t i f y t h e movements o f m o b i l e telecommunications users. Moreover, the rapid increase of the relative proportion of handheld mobile terminals compared with vehicle-mounted equipment makes past experience rapidly obsolete. To give an idea on the order o f magnitude, we can make very simple assumptions. First we will take the assumption that the speed of 70% of the users is zero, and that the speed of the other 30% is 30 km/h. Then. we will assume an average cell diameter of 3 km, and translate this into a mean lifetime in a cell for the moving users of 4.5 minutes, that is to say an average of around one handover every two communications. A related point is the location updating trafficiDifferent reasons may lead to location updatings, as explained in Chapter 7: movements of users between cells, switch-on and off, periodic updating. I f the two last terms can be considered roughly proportional to the traffic in the cell (within a given traffic model), the first one varies from 0 to a high value depending o n the proportion o f the boundary o f the cell which corresponds to a boundary between location areas. We will make the simple assumption that the location areas are big enough, and their boundaries carefully chosen so that we will neglect the location updating traffic due to movements. We will make the assumption that the period for periodic location updating is half an hour, and we will neglect the traffic clue to switching on and off. - • T h e aim of this presentation was to give some insight 0,1 the issues. For the next sections. we will not go too much into the fin a r t s of the model. The main parts of the traffic can be described with a few figures, related to a number of events during the peak hour as follows: 1.5 Number of mobile originating calls - successful - unsuccessful Number of mobile terminating calls 1.0 - successful - mobile station not reachable 2.0 Number of location updating -• 1.0 0.5 0.6 (1.4 9.4.2.1. Dimensioning of Radio Channels With cellular planning. the main topic is radio coverage. With cell configuration. the main issue is to choose the channel configuration in each cell, given the traffic i t must support. A s explained i n detail in Chapter 4, a cell includes a set o f common channels and dedicated channels. We will look at both aspects in turn. Dedicated Channels Several dedicated channel types exist in GSM. In phase 1, the radio capacity must be distributed among TACH/Fs (full-rate traffic channels) and TACH/8s (channels eight times smaller, used for signalling). Some machines enable a dynamic distribution between these types of channels (one TACH/F = 8 TACH/8), but in other cases the allocation is more Static. and can only he changed by some intervention on the configuration parameters. The number of TACH/F required in a cell can be directly evaluated from the planned traffic in the cell, through the application of the "Erlang B formula". An important parameter is the planned probability that all the channels arc used at a given moment during the peak hour. This blocking probability measures the probability that a call attempt fails by lack of a lice radio channel. The Erlang B formula relates the average channel occupancy (in Erlang). the number o f channels (an integer) and the blocking probability. under the assumption that the instants o f call establishments and the duration o f the calls follow Poisson processes. Table 9.2 gives the capacity in [Hang for some number o f channels (corresponding t o representative cell configurations) and a blocking factor of 2r4. a typically acceptable value. URN in cell 3 4 5 6 7 C'hanncls 7 14 22 30 37 45 53 Capacity 2.9 8.2 15 22 28 353 43 ratio 0.41 0.57 0.68 0.73 0.76 0.79 0.81 -fable 9.2 - Capacity (in Erlang) of a set of full rate traffic channels (Erlang B formula, 2% blocking) A blocking probability of, here. 2% results in an average load which is not proportional to the number of traffic channels. The needs in terms o f TACH/8 capacity come from two main signalling requirements (not including short message transmission needs): communication establishment attempts and location updating attempts. The dimensioning must be done for full-load conditions, that is to say when the cell is congested and all TACH/F are used. The maximum useful rate o f communication establishment is then obtained when the - - congestion state is maintained, and is roughly proportional to the number —6r-TACH/Fs, given the mean duration o f calls and the ratio between successful and unsuccessful call attempts. I t is worth noting that the channel allocated to a mobile station upon access may not always be a TACH/8. as explained i n Chapter 6. I n particular, the Very Early Assignment strategy does n o t make use o f TACH/8s f o r c a l l establishment (but alternatively increases the number o f TACH/Fs required for a given amount of traffic). The second signalling flow impacting on TAO-I/8s is the location updating flow. As an example, with the traffic model presented in the Previous section, a cell of 30 traffic channels will support at congestion it2 Erlang and thus 440 subscribers, among which around 260 on average have a switched-on mobile station. The access rate is 0.27 access per second for successful communications including mobile originating and mobile terminating, plus 0.11 p e r second f o r unsuccessful mobile originating attempts, and 0.14 for location updatings. The total is close to 0.5 accesses per second. Assuming that any of these events makes use of a TACH/8 for 4 seconds, we obtain a TACH/8 average occupation of 2, which requires at least 6 TACI-I/8s to obtain a blocking rate less than 2%. Consequently, many parameters influence the dimensioning of the TACH/8s, such as the choice to use periodic location updating or not (and the corresponding period), the choice to use "IMSI attach/detach" or not and t he planning of location areas. 618 NETWORK MANAGEMENT THE GSNISYSTNNI Common Channels The (limensioning of the common channels has much in common will) the choice of the number of TACT I/8s. The 13E1'11. FUCH and SCH raise no engineering problem. On the other hand. the configuration of the PAGCI I and o f the RACH can he chosen among several cases. The common channels have to support two kinds of traffic. not proportional one to the other. The first is the access traffic. which is measured as a number of accesses per unit of time. which we havc already assessed to evaluate this needs for TACH/Ss. The second source of traffic is paging, measured as a number of pagings per unit of time. The number of paginf, messages in a given cell depends not on the traffic in the cell, but on the raffic in the location area. The lowest paging load corresponds t o the case o f a one-cell location area. and is comparable to the access load, On the other hand, in such a configuration the whole cell boundary is a location area boundary, and location updatings due to movements lead to an important location updating load. A t the other extreme allowed by the Specifications, a location area can include all cells connected to one MSC. This leads to an important increase of the paging load. I lowever. if the TMSIs (temporary identities shorter than the IMSI) are used and i f message scheduling is optimised on the downlink common channels. as explained in Chapter 6, the corresponding increase can be supported easily. There are several possibilities for the configuration of the common channels. a s described i n Chapter 6 . Depending o n th e capacity requirement for access. five configuration choices exist for the RACH and PAGCH. For each o f these configurations. table 9.3 shows the maximum access rate. calculated as the lowest o f the two following capacities: • t h e access capacity on the RACH. evaluated at about one fourth of the RACH burst rate (see the section dealing with the RACH throughput. in Chapter 4): • t h e assignment capacity. calculated here based o n t h e assumptions that each message contains two assignments, and that the channels are entirely used for assignments. This is an upper bound. as it is not very realistic not to count any paging! The table also shows the maximum paging rate for each case, based on the assumption that the channels are entirely used for paging, - and that paging is performed solely on TMSI (this enables the packing of up to four requests i n a single message). The maximum number o f pagings and the maximum number of accesses are mutually exclusive, CCCII capacity maximum RACH access rate I \ I II I i per s e ‘ o n d I maximum paging rate 619 maximum assignment rate maximum channel access rate paged per second) l a s ‘ i g n i n c n I s per second) (accesses per second) 29 50 25 25 I 54 152 76.5 54 ' 108 306 153 108 162 -155 229 162 217 611 306 217 i.m.v..e. 1/2' nuffilite.forion% Table 9.3 - Access chan lel capacities Depending on the capacity to be offered by a given cell, the access channels can be configured in several different sizes. "I the other half can be used only for 4 TACH/8. and the choice of common channel configuration has to be done taking the two factors into account. As an example, with the same assumptions as for the computations for the TACH/8 requirements, a cell o f 30 traffic channels experiences about 0.5 accesses per second. For pagings we have to assume some size of location area, say 20 such cells, and the paging repetition rate, for which we will take 3. This leads to around 3.5 pagings per second. Table 9.3 shows that with such values, even .the smaller common channel configuration supports these requirements with ample margins. 9.4.2.2. Other Cell Configuration Parameters Beside the dimensioning of channels, system engineering must also deal with other cell characteristics. A set of parameters, which has been described in Chapter 6, is used to control the access of mobile stations in a cell. They represent in fact one facet of a more general issue, which is the management o f congestion. T h e respective priorities given t o handovers. location updating, call establishments, etc. must also be managed by system engineering through BSC parameters. Another cell parameter to deal with is the BSIC. As described in Chapter 7. its allocation must take into account the beacon frequency Another aspect o f t h e relationship between b o u r c e l l monitoring and cell synchronisation is the initial search for Ex.CH bursts. This is a costly part of the monitoring. because the mobile station must scan all time slots to find such a burst. which is transmitted 5 times every 51 x 8 = 408 burst periods. i.e., once every 82 bursts on average. I f the mobile station was to have information on the relative synchronisation of the cells. the search would he sped up considerably. However, the phase 1 Specifications do not include any means to indicate to mobile stations in idle m o d e whether surrounding c e l l s a r e synchronised o r n o t . Syncluonisation o f base stations has therefore n o influence o n the process. I f t h i s application w e r e introduced. i t w o u l d require \ of ironis.atithi up to the 51 x S burst cycle, to a low accuracy (100 ps would not be a problem. since it is already the order of magnitude of the spread due to propagation limes). A last application of cell synchronisation is distance measurement. With pre-synchronisation. a mobile station is already able to know the difference in propagation times between its serving cell and the each of the neighbour cells it is pre-synchronised with. This information is not exploited in the phase I Specifications. but will he in phase 2. It can be used to several ends. Let us cite handover preparation based on distance measurements. such as i n the German C -network, o r geographical localisation of mobile stations. Phasing constraints for such applications are the same as for synchronous handover. But specific localisation applications may be more demanding. since the accuracy of the distance estimation derives immediately from the synchronisation accuracy. 9.4.3. NF;TwORK ENGINEERING Network engineering in general consists in choosing the place and capacity o f the various network nodes. the capacity o f the intervening transmission links, and a variety o f parameters. For GSM, the nodes under consideration are the BSCs and the MSCs. The HLR/AuC capacity will also he looked at. The links under consideration are those on the A and ,this interface, and those between the MSCs and the external networks. 9.4.3.1. The Abis Interface The capacity required on the Allis interface in terms o f traffic circuits is equal t o the number o f TACH/F o f the cell. The capacity requirement for signalling is dominated by the measurement messages. Without preprocessing in the BTS. there are around 2 messages per SC B n B' B ' T TS' i T I one single 2 Mbit/s multiplex Figure 9.17 - A drop-and-insert configuration Such configurations may he useful to reduce the number of 2 Mbit/s multiplexes necessary on the Abis interface, especially for small capacity cells, compared with a star configuration where each BTS requires at least one multiplex. TACH and per second. For instance, a cell o f 30 TACH/F and 1 2 TACH/8 w i l l generate a t congestion (i.e., when f u l l y loaded) 8 8 measurement messages per second. that is to say a load close to 16 kbit/s. Thus a medium cell will have little spare capacity with one 64 kbit/s circuit for signalling transfer with the BSC. Abis links can represent a substantial part of the running costs of a PLMN. One of the problems is that each BTS site requires a relatively small number o f circuits (in comparison with the usual figures inside public telecommunications networks s u c h a s t h e PSTN). T h e transmission unit being the 2 Mbit/s multiplex, economies can b e obtained i f the drop-and-insert connection method can be used at the BTS. This technique provides the ability to share a 2 Mbit/s multiplex between several BTS sites. and to decrease the number o f leased o r installed transmission links. Figure 9.17 shows an example of a drop:andinsert configuration for the Abis interface. 9.4.3.2. The BSC Two aspects o f the BSC have t o be looked at: the switching capacity. and the computation load. The switching capacity is measured as the total number of circuits with the BTSs and the MSC. On the A TIIE GSM SY STENI Area (-ell planning Dimensioning I Anti' control Objects Frequencies Beacon frequencies I lipping sequences Potter control parameters I holdover parameters ( s e l e c t i o n parameters BSI(' motion channels TAc'l I/8s Location areas Periodic location updating INISI :ttiaelViletach /ierloatl control parameters [able 9.4 -- (ell characteristics System engineer‘ must manage 'nail) different cell characteristics in a co-ordinated manlier, in order to ensure proper guanty of service in terms Of coverage, dimensioning and congestion control. allocation, and must be co-ordinated near the boundaries with PLMNs using common frequencies. Table 9.4 summarises the objects which system engineering must manage to optimise the network configuration f o r a given quality o f service to an expected population of users.. 9.4.2.3. Inter-Cell Synchronisation A last aspect o f system engineering is "time engineering": the synchronisation between cells. This has some impact on the quality of service. i n t h e area o f handover performances. T h e notion o f synchronisation is taken here at a general sense. It includes also the desynchronisation of the cells. as we will see that full synchronisation can be very detrimental t o some aspects of system performance. Best performance i s obtained when time bases i n neighbour cells are synchronised s o t h a t burst emissions a r e synchronous, b u t desynchronise(' so that in particular multiframes are not synchronous. NETWORK MANAGEMENT 6 2 1 The time division multiplexing scheme on the radio interface requires a complex time management, resulting in a series of clocks to be managed at the level of bursts, TDMA frames, multiframes, superframes and hyperframes. GSM design enables the clocks in different cells to be run independently. Even their frequencies in the long run may differ somewhat. each frequency being within a 5.10-8 tolerance. However, sonic performances are improved when burst emissions in neighbour cells are finely synchronised (and this requires equal long term clock frequencies). The main one is handover execution time. We will now look at what is exactly to be synchronised and to what accuracy. The interruption t i m e during a handover execution can b e improved (from roughly 200 ms down to 100 ms) i f the "synchronous" handover procedure is used, and this requires synchronised clocks. In this case. the only problem is timing advance. What needs to be synchronised is therefore the long term frequency and the phasing of the burst clock. Unphasing of higher rank clocking should not cause any problem, as long as the return propagation time in a cell does not exceed 577 ps (which would correspond to a cell radius of 90 km). However, the Specifications do not explicitly state whether mobile stations should behave correctly if asked to perform a synchronous handqver and only the burst clocks are in phase. This imposes an a l l -clock phasing between these cells, the drawbacks of which will be explained later. From the accuracy point of view. synchronous handovers require simply that the timing advance evaluation is correct with an accuracy of a few microseconds. Though not explicitly stated in the Specifications, this value comes from the duration of the guard time. The phasing accuracy between the two cells must - - - therefore be of the same order of magnitude. Synchronisation between cells, i f limited to bursts, can also be useful for pre-synchronisation (see Chapter 6). I t improves the search time for neighbour cells, though not in an obvious way. In fact, all-clock phasing i s t h e w o r s t imaginable case f o r p r e -synchronisation performance. Let us recall that the mobile station uses the long intervals in its traffic channel cycle to decode the bursts o f neighbour beacon frequencies. I f a l l S C H , F C C H a n d B C C H bursts a r e exactly synchronised in all cells, the time between reception of a burst of a given type on two neighbour cells will be maximised: about 1.3 seconds for an SCH or FCCH burst, and up to 3 seconds for a given BCCH burst. On the contrary. i f cells are carefully de-phased, simultaneity can be avoided, and the average time described above will be statistically reduced to half of this value. The best scheme for pre-synchronisation is when cell clocks are organised to minimise the probability of simultaneity between FCCH, SCH o r BCCH bursts i n two adjacent cells. This kind o f "offset" sygchronisation is o f course more complex to implement than an allcIrck phasing synchronisation. • , , , , , , . • • • , , , , , , LII...*111101 I • interface side. things are simple. The required number of cire s can be computed with the Erlang formula from the total planned traffic in the 4 cells managed by the I3SC. The blocking factor is chosen to be much smaller than the one which constrains the capacity of a cell. for instance it can be set at 0.2 'X,. With such a value and a total planned BSC capacity of 500 Ertang. the required number o f circuits on the A interface is roughly 550. From the Abis interface side. constraints on the BSC vary whether the total planned capacity is spread over a few large capacity cells, or on the conkrzy.on many small capacity ones. The number of circuits is equal indeed to the number of traffic channels in the cells. not to the traffic. Moreover, redundancy considerations may ask for additional circuits on a cell per cell basis. We will take two extreme examples. 11' the traffic is handled by a single cell. the number o f circuits is 513 (applying the Erlang law with 2(h blocking). and the required switching capacity, adding circuits from both directions. is 1060 circuits (that is to say with multiplexed 1 6 kbit/s circuits and a little redundancy. around 1 2 multiplexes of 2 Mhit/s each. 6 on each interface). At the other extreme, if we imagine that each cell generates only I Erlang of traffic, there are 500 cells connected to the BSC. and each cell has 7 TACH/Fs, out of which 4 are needed for this traffic. Assuming in addition a redundancy at the 2 Mbit/s level. the number of circuits is doubled. and the total number of circuits entering the 13SC from the BTSs is 4000: the total number of circuits entering the BSC is then 4550 (including the A and Abis interface circuits). This shows that the switching requirement for a given total planned traffic can vary in the ratio of 4. or reciprocally that the actual traffic handled by a BSC of some make can vary in the same proportions according to the average cell capacity. The location and capacity range o f the BSCs is a debated point. Sonic operators want small BSCs on the BTS sites. A t the other end, some operators want big BSCs on the MSC sites. even possibly a single BSC p e r MSC. Others want independent BSCs w i t h a capacity intermediate between a BTS's and an MSC's. and which can potentially he sited in any location. not necessatily with a BTS or an MSC. Various considerations will dictate the choicei!A BSC has three main functions: it acts as a circuit concentrator, and as such its position impacts the running costs of the transmission lines between BTSs and MSCs. A BSC is also an operation and maintenance agent: we will see that the BTSs are not linked directly to the OSS. but through their BSC. Finally, a BSC is where handovers are controlled. Bigger BSCs lead to a smaller number of handovers which must he handled by the MSC. and the bigger the BSC 623 the wider the knowle . ! concerning the traffic used t o decide on handovers. The A Interface I r. The capacity of the link between a BSC and an MSC has already been looked at. in so far as the circuits carrying user data are concerned. The signalling load on an A interface depends on the traffic handled by the BSC. The dominant factor comes from the messages exchanged durini2. call establishment. The analysis o f the protocols shows that a typical call establishment requires around six messages in each direction. A c a l l release corresponds t o . say, f i v e messages counting both directions. The traffic handled by a BSC o f 500 Erlang corresponds to 550 simultaneous communications at congestion, and with our traffic model to some 9 calls established and released per second, that is around 75 messages per second in each direction. Assuming an average message size of 25 octets, we obtain a load of 15 kbit/s. One 64 kbit/s link is then sufficient. including the location updating and paging traffic. However, the current practice in such a case is to install two links, for redundancy against failures. The MSC I The trend is to have MSCs of as high a capacity as possible with the present switch technology. A t the date o f writing, the order o f magnitude of an MSC capacity is one to several thousands of Erlang. For a network with a 5% penetration o f the population and 0.05 Erlang per subscriber. a 2000 Erlang MSC is suitable for an area with 800 000 inhabitants. This is commensurate with the present density o f PSTN switch locations. MSCs can then be sited in rather important towns, and _ w i l l cover a part o f the biggest towns, o r a medium town and the surrounding area. HLR and AuCs An HLR machine and the corresponding AuC must support a load which is dominated first by the GMSC-HLR and MSC-HLR exchanges which occur for each mobile terminating call attempt, and possibly by the authentication flow, when this procedure is done as often as once per call. I.et i t s take a numerical example t o s n o w t n e importance 0 1 authentication. Aiming at an HLR capacity of the order of half a million subscribers seems reasonable. This means that a typical na., t a t GSM network will have no more than a few HLR machines. This corresponds with our traffic model to 4(X) call attempts involving the corresponding mobile stations per second. An authentication triplet has 224 bits, say 250 to include i t i n a signalling message. The AtiC' must then generate I(H) 0(1(1 hits per second, and the HLR has to transmit this volume o f information to the MSC/VLRs. This figure climbs up if authentication is required at each location updating procedure. With the same HLR capacity. there are 200 mobile terminating call attempister second. each involving two messages to the HLR. and two messages from the I ILR. The order o f magnitude o f this traffic i s comparable to the one caused by authentication. Of course, the HLR must also deal with the signalling caused by location updatings when the subscriber changes from MSC/VLR area. but this load can he considered to be much smaller than the two previous ones. The NSS Coidiguralion Up to now, wee have focused on the Base Station Sub-system and on the dimensioning o f MSCs and HLRs. There is another part o f network engineering as far as the Network and Switching Sub-System is concerned, which is closer to fixed network engineering than any of the other items. I t deals with the network architecture for traffic and for signalling. The operator may or may not. depending on the terms of its licence. have the right to mesh its MSCs and CiMSCs and have its own transit exchanges. Similarly, the operator may have the right to set up its own signalling links between NSS machines, and have its own Signalling Transfer Points (STPs). When the operator has the freedom to set up its own traffic network and/or signalling network. it must choose the number and location of the transit exchanges and/or signalling transfer points, and it must optimise the configuration of the links between all the machines of the network. In all cases, i t must decide on the number and location o f the (INISCs (e.g.. i n the same machine as a n M S C o r not), o f the interworking functions with the fixed networks, of the gateway-MSCs for short messages, etc. This optimisation problem is not specific to GSM and is one of the tasks of every operator of a large network. 9.4.4. O B S E R VAT I O N S As seen in the eligineering section, the dimensioning of a system and the setting of its multitudinous parameters are based on a number of assumptions concerning propagation or the behaviour of the subscribers. Moreover, even the effect of some choices on the internal functioning of the system is not easy to predict. With a system in operation, this data, as well as the results in term o f quality o f service for instance, can be observed directly. The observation activity consists in gathering statistical data in the traffic handling machines, and in analysing them for different ends. One goal is to determine the bottlenecks in the system, so as to know where upgrading is needed. This allows an increase the network capacity in the most efficient way. More generally, one goal o f the analysis i s t o streamline the network so as to get the maximum capacity, and revenue, out of it. We have seen another purpose when dealing with maintenance: the a priori model f o r performance can be compared with effective results. A discrepancy can he the symptom of a dysfunctioning otherwise undetected. These different goals direct what should be observed. Observation can be quite load consuming, and it is important to select correctly what has to be observed, and when. Unfortunately, most o f the performance observations are interesting at the peak hour, i.e., at the moment when the machines are responding t o their greatest demand for traffic handling tasks and have little extra time to devote to observations. A way out is to make some observations on some machines only, changing them f o r instance each day. The amount of information which may be collected is very high. It can be classed roughly as follows. Traffic measurements aims at the evaluation o f the load on the various interfaces at the peak hour. The goal is to know the mean peak traffic on a group of circuits, on a cell, or for signalling at a given node. A finer knowledge is in fact needed, with the distribution of the load between for instance mobile originating calls, mobile terminating calls, location updatings, and so on. The traffic model expounded on in a previous section is a good example of the information sought. Quality o f service measurements aim at evaluatipg a number o f indicators o f the quality as perceived b y the users. This includes congestion rates. for instance the proportion of call attempts which failed because o f lack o f resources. I t includes also delay statistics, as f o r instance the call setup time. or tne new) ocui, cell ate uccistoit to make a handover and the effective transfer. and so on. Such ohs—v(100ns can include error rate statistics. in particular on the radio interfat Availability measurements aim at evaluating the failure rates of the components of the system. and the time between a failure and its repair. It is clear that observation is a function of all the machines of the system. f r o m t h e BUS (which i s t h e o n l y point where detailed measurements can be made on the radio interface) up to 055 machines, which must observe their own availability. The rant observation data is usually kept rather simple. such as the value of an event counter. Such raw data is stored in a local file. and transferred to a place where they can he annlyTed: In the past (not so long agoi. this was done simply by a local printout, analysed afterwards directly by human beings. Progress was 'nude when the transfer was done by magnetic tapes. and a first level of analysis w a s performed b y a computer. W i t h a n integrated telecommunication management network. electronic tile transfer is used. In.practice, a measurement is obtained via the OSS. Such a request includes the type o f data to gather. the machine which will do it, its schedule (it may he a regular observation. e.g.. every Monday from (, p.m. t o 8 p.m.. o r a point observation. at the request o f operator personnel or as a follow-up o f some event) and where to forward the result. This activity requires signalling functions between the OSS and all the machines in the network. implemented but with minimum disturbance for the traffic handling. Let us sec what can be( quilled. and the difficulties f o r performing the modification without traffic interruption. Three general categories o f modifications can be identified. Hardware Modifications Some changes affect the hardware. This includes the installation of a new piece of equipment. to increase the traffic capacity (e.g., a new radio site), the replacement of some machine by a more advanced one, or also to change some functionality such as for instance the introduction of half rate transcoders. Such changes cannot be done remotely. To avoid a disturbance of the traffic, the machine to be replaced, i f any, is put out o f operation —when the new one is put in. A central control can be helpful in the testing phase which precedes the effective operation o f the newly installed hardware. I t can also be o f interest to speed up the transition. The introduction of a new node in a network, and even more of a new cell in a cellular network, requires the changing o f a number o f parameters in neighbouring nodes. This can be commanded at the right moment through the management network, to minimise the period during which the network parameters are inconsistent. 9.4.5. NETWORK CHANGE CONTROL An example of hardware modifications is the installation of new switching nodes (MSCs o r BSCs). When this is done f o r coverage extension into an area not yet covered, there is no problem o f service interruption. The impact o f the machines already i n operation is the parameter modifications needed for them to take into account the network extension. The other case is when for traffic load reasons a new switching node is introduced to take over a part of the coverage of an existing node of the same nature. The modification has many more impacts, since transmission routes are modified. A network is not designed on paper and implemented once and for all. It evolves, to follow the progress of technology. to remedy errors or inefficiencies detected by maintenance or observations. or to follow the planned evolution o f the handled traffic. These modifications could be dime easily by stopping the network. installing the new configuration, testing i t and restarting the normal functioning. This approach, i f comfortable for the operator. is less so for the users. and technical means are introduced i n the system so as t o allow these changes t o be Another example of system extension is the addition of a new cell. One o f the difficult points is the testing o f the new cell prior to its opening to commercial operation. To forbid such a cell during the test period, it must be barred (this state is indicated in messages sent on the BCCH). Special mobile stations, disregarding this indication, are useful to test the cell. However. the complete new set of the parameters which control the choice o f cell ( i n idle mode as well as f o r handover preparation) cannot h e introduced and tested, since they must b e consistent with those in the neighbour cells already in operation. After the transfer of the results, the raw data must 20 through some statistical analysis. possibly compiling data from many sources, so as to provide meaningful and accurate indicators. This function is located in the OSS. Finally. the results have to be used. An example is the access to the statistics by operator personnel, through workstations. Software Modifications Another category o f change i s the replacement o f software modules. Software is much easier to replace than hardware, and some Norm are mu lutes are upihited regularly. for error correction or because of the progressive introduction o f new functionalities. W i t h digital technology. more and more functions arc easier to implement in software than i n hardware. including possibly transmission functions such as coding/decoding o r modulation/demodulation. L i s ) means t o change solm are modules c a l l t h e n provide a v e r y powerful evolution mechanism. for a better optimisation of the s t e i n . and hence for higher revenues tor the operator. The ()SS will need to monitor the subsequent performance o f the network to ensure that the new software does not subtly degrade the network performance. The replacement o f a software module w i t h minimal traffic interruption is possible with redundant machines. The new module is loaded and tested in the redundant part. I f the test fails. the old software can he re-loaded. or. best, kept on the machine to be restored to operation if need be. I f tests indicate all is right. a change of configuration puts the modified module in operation. and so on. The change of a software module does not intrinsically require an on-site action, as for hardware. Such changes can he done remotely (software downloading). through t h e management network. A f i l e transfer protocol is needed, as well as procedures to control the activation and deactivation of the modules. Parameter Modifications We have seen in the engineering section that a network can be tuned b y a variety o f parameters. affecting f o r instance the cellular planning. Some parameters in a given machine are simply the description of the surrounding parts of the system. and are modified when those parts are changed. Other parameters can he chosen between several values for a given system configuration. Some choices are arbitrary (e.g., PCM link numbering), and the only issue is to ensure the consistency between several machines: such parameters are rarely modified once set. In other cases. the choice o f value impacts in some way the behaviour of the network, and is done according to some optimisation criterion. I t is important that they can h e modified easily during operation, with minimum impact on the service. This allows the operator to dynamically refine the optimisati o f the system, when observations enable a better choice o f parameters. In wireline systems, and in the switching field, there are few such parameters (an example is the routing table's). On the other hand, cellular planning is a field full of them. Let us see some of them. and how they can be modified. The frequency plan has an important impact o n the system capacity. and i s difficult t o optimise before f u l l scale operation. Moreover. any addition o r removal o f a cell, o r any change o f the capacity o f one cell may change the optimal frequency plan for many cells. The possibility to change the frequency plan during operation is then an important asset for an operator, and even more if no on-site action is needed. There are several difficulties. The major one comes from the radio frequency part of the BTS, and more exactly from the coupling device, that is to say the piece of hardware which combines the output of several amplifiers to feed a single antenna. 900 or 1800 MHz signals cannot be simply added by connecting wires. The power emitted by one amplifier would be dissipated in the other ones: the antenna would radiate nothing. and (worse!) the amplifiers would simply burn. Special devices are needed to force the energy to go into the antenna (and overall not into the other amplifiers). Two major techniques exist. A hybrid coupling device has the advantage that it does not require special tuning depending on the emitted frequency. Changing the operation frequency remotely does not raise any problem. Unfortunately, such devices introduce an important power loss. proportional to the number o f combined outputs. Hybrid coupling i s typically unacceptable above 4 combined outputs. With TDMA. even a one-output cell is meaningful (with half-rate coders, this corresponds to around 15 voice channels), and hybrid coupling appears as a good solution for low to average density areas, but not to the cells of highest capacity. The other coupling technique, called cavity coupling, makes use o f tuned filters put between the amplifier output and the common point, to prevent the energy from the other amplifiers (which is at different frequencies) to go to the amplifier. The insertion loss is smaller than with the hybrid technique, and moreover increases slowly with the number of combined outputs. The main drawback of tuned filters (sometimes known as cavities) is that the frequency tuning involves moving mechanical parts. and is not an easy task. Remotely tuneable cavities. with small motors, now exist, but still seem to have reliability problems (as with most mechanical devices). Otherwise the frequency tuning requires the intervention of a screwdriver! Providing the combining method does not prevent a remotely and rapid change of the transmission frequencies. a second d , : u l t y comes from frequency hopping. With frequency hopping. the change o f one Frequency in the set allocated to a cell impacts many ( i f not all) traffic channels of the cell. This makes such a change difficult i f the goal is to have no impact on the existing traffic. We have seen in Chapter 6 that special procedures have been introduced for this purpose. Another way is simply t o stop the cell operation i n order to change the Frequency configuration. for instance at three o'clock in the morning. • --A-second example of cellular planning parameters for which traffic handling procedures enable a change with limited traffic interruption are those concerning the control channel configuration. The number o f PAOCHs in a cell ( I. 2. 3 or 4). or the organisation of the paging subchannels, can be modified during operation to match the access and paging load. We have seen in Chapter 6 how the paging mode can be handled to ensure that the mobiles stations in idle mode in the cell do not miss paging messages. The last example we choose to give here concerns the parameters controlling the handover preparation. We have seen how some of these parameters control the true borders of the cells. and hence the distribution of the traffic between adjacent cells. This is useful to refine the cellular planning, or even on a shorter time scale. to limit the effect of a local traffic overload. Handover preparation parameters can be modified at any moment, without more precaution than ensuring the consistency between the set of parameters used in each cell. These examples are only a few among many others. Priority mechanisms, access classes. emission power. barring status, location areas. etc.. are other parameters on which operators can play to drive their network and get the maximum out of it. 9.5. ARCHIJ"CTURE AND PROTOCOLS Maintenance, monitoring and system modification can be done from remote centralised machines to a large extent. The main requirement is to connect all the traffic handling machines to one or several operation and maintenance centres, from which commands can be sent, to change parameters, to download software, to start tests or observations, to gather the results of the tests or observations, or to receive alarms. This calls for machines devoted to these tasks, and for a network to link them between themselves as well as with the traffic handling machines and work stations from which the operating staff can act. These connections between the machines call f o r signalling protocols t o structure their ) 9 . 5 . 1 . MANAGEMENT NETWORK ARCHITECTURE The operation, maintenance and administration functions we have described a r e n o t r e a l l y G S M specific. T h e y e x i s t i n a n y telecommunications system. Historically, these tasks were first performed manually, with only local access to each machine. The next step was characterised by a proliferation o f ad hoc systems, for instance one for observations, another one f o r failure reporting f r o m transmission equipment. a third one f o r the configuration o f another type o f equipment. and so on. all o f these systems using different proprietary protocols. Such an approach can still be found in many networks. However, the Specifications include some specifications to guide the setup of an integrated management network, according to the Telecommunication Management Network concept (or TMN) developed by c o n , which aims at providing bases for integrated management networks. These GSM specifications are mainly a template, a conceptual framework underlying the development o f equipment by different manufacturers. When applied, this framework will make the integration of machines of different make much easier for the operator. An integrated management network is linked, on one side, to the traffic handling machines. and on the other side, to the operating staff \URN (L'4 Operation I tem SysFunctions r Data Communication Functions Nss r- Mediation Functions A BSS Figure 9.18 —TN1N categories of functions The TN1N concept distinguishes three main functions: the Operation S\ stem Functions (the centre of the management network), the Mediation Functions (adaptation functions for specific equipment) and the Data Communication Functions linking them. through workstations. Between these interfaces. TMN distinguishes three main categories of functions. as shown in figure 9.18: • t h e Operation System Functions, which are the management applications proper. connected t o workstations from which wide-scope applications c a n b e conducted b y operating personnel: • t h e Mediation Functions. which are the intermediaries between the operulitm system functions a n d t h e traffic handling machines. Mediation functions are introduced for concentration purposes but also for adaptation of the generic operation system functions to specific machines. Typically. operators introduce different mediation functions for every subsystem (the BSS or the NSS a n d for every manufacturer of the machines. In the TMN architecture. the mediation functions may or may not appear. Traffic handling machines can be connected directly to a machine implementing operation system functions: MANAGEMENT 635 • t h e Data Communication Functions, which are the transmission means used t o l i n k t h e operation system functions, t h e mediation functions, the traffic handling equipment and the workstations. These data communication functions can b e point-to-point leased lines, or switched links provided by an X.25 network o r an SS7 network. The choice o f the data communication structure is entirely the operator's, and does not impact on the possibility to harmonise the applicative functions. l he Specifications do not give much direction o f the structure of the management network. except on one point: the BTSs are connected to the management network only through their BSC. which then acts as a mediation function for all the BTSs under its control. The data connection between BSC and BTS for network management needs is specified in the Specifications, i n particular i n T S G S M 08.59 ( f o r the transport mechanism) and in TS GSM 12.21 (for the semantics). The Specifications introduce other architectural terms than the TMN terminology cited so far. In particular, the OMC, or Operation and Maintenance Centre, is a term used to describe the management network as seen by a given traffic handling machine. It may stand for a single piece of equipment responsible for operation and maintenance of a given part o f the network, or alternatively for a mediation device party o f a centralised management network. An OMC is typically considered to be in charge of a subsystem. such as the BSS (the term OMC-R, with "R" standing for "radio", is used by some manufacturers) or the NSS (OMCS. with "S" for "switching"). Besides the OMCs, the Specifications also refer to the NMC. or Network Management Centre, which is concerned with the management of a whole PLMN, and hence has a wider scope than a given OMC. In the rest of the chapter, we will stick to the TMN terminology, to be more general. 9.5.1.1. TMN Interfaces The goal o f TMN is to harmonise the applicative management protocols. For this objective. a common language must be used regardless of the type and make o f the equipment to manage. But the internal organisation o f each machine is different. It would be costly and nonefficient t o have remote operation system functions taking care o f irrelevant implementation characteristics. Therefore, the T M N model defines one or more Network Element Functions in each machine; these functions are the sole correspondents o f the rest o f the management network, and they are in charge of the dispatching of commands toward V J r traffic handling prott I s for operation and maintenance activities, the BTS management protocol on the Abis interface, as well as a template for a Q3 protocol specific to GSM. We will now describe-these three points in more detail. NEF q i , q2 N 1 F BSS or NSS machine Figure 9.19 —TMN reference points ppl icative protocols are defined at specific reference points (lb g ) defined between Network Element Functions (which are part of the machine to be managed), Mediation Functions, Operation System Functions, Workstations and the operating staff. the different specific modules o f the machine, as well as gathering information from these modules. The structure o f a Teleconimunication Management Network, based on the functions we have identified so far. determines several classes of applicative protocols. which are defined at different reference points. These "reference points" between functions are shown in figure 9.19. The major points at which exchanges are modelled by TMN are the q points. The more general protocols are found at the q3 reference point for connection of the Operation System Functions, whereas protocols of less generality, e.g., specific to some manufacturer. are found at the ql and q 2 reference points involving Network Element Functions and Mediation Functions. Other reference points than the q points are also defined in TMN, such as the,/ reference points between workstations and other network management entities: the g points between workstations and the operating staff: or the .v points between different management networks. The management exchanges between different PLMNs must not be neglected. They concern the accounting procedures as well as the exchange of information regarding mobile station equipment (e.g., blacklisted IMEIs). Most of the protocols on the reference points identified here are left for the operators to specify. and we will not deal with them further. But some protocol aspects are covered in the Specifications: the usage of 9.5.2. OPERATION AND MAINTENANCE IN THE TRAFFIC HANDLING PROTOCOLS The radio interface protocols are hardly impacted by operation and maintenance functions (the interrogation of the IMEI may be considered one of the few functions in the management domain on this interface). On the contrary, the internal protocols such as the BSSMAP (between BSC and MSC) and the MAP (between NSS entities) include a number of management functions. Such functions are grouped with the traffic handling protocols f o r two main reasons: either they correspond t o actions of limited scope impacting only the two ends of a liaison, or they are so entangled with traffic handling operations that they warrant separate handling by network management entities. The first category of functions includes the management of circuits between BSC and MSC, and between switching machines. Both the BSSMAP and the ISUP include procedures t o block and unblock circuits. i.e.. to inform the other end not to allocate a given circuit (e.g., because of a failure detected on this circuit), and conversely to set it back to normal use. This is indeed the role of the BSSMAP BLOCK, BLOCKING ACKNOWLEDGE, UNBLOCK a n d UNBLOCKING ACKNOWLEDGE messages which enable a BSS to advise the MSC about the status of its circuits. Another procedure for the local management of the A interface is the reset procedure. which enables one of the machines (either the BSC or the MSC) to restart and indicate it to the other end. In such cases, all communications i n progress are lost, making i t a last resource mechanism. The procedure makes use of the BSSMAP RESET and RESET ACKNOWLEDGE messages. A related procedure exists at the level of individual circuits: it uses the BSSMAP RESET CIRCUIT and RESET CIRCUIT ACKNOWLEDGE messages. It is used to release a possible connection using a given circuit, when normal means cannot be applied because some of the knowledge of the relationship between SCCP connections and terrestrial circuits has been lost. An example of a procedure belonging to the network management scope. but so entangled with the traffic handling protocols that it cannot lie handled by other means, is the trace procedure. The aim o f this procedure i s t o start a l o g o f a l l o r s o m e e v e n t s happei 1 d u r i n g t h e lifetime o f a c o m m u n i c a t i o n . T h i s m a y b e p a r t i c u l a r l y u s e f u l w h e n t h e mobile station i s k n o w n t o he d u b i o u s , o r t o h e l p the diagnosis o f some * s f u n c t i o n i n g . W h e n i t i s f e l t useful t o record b o t h r a d i o interface aspects ( c h a n n e l a l l o c a t i o n . h a n d o v e r s . e t c . ) a n d c o m m u n i c a t i o n management aspects, the 13SC a n d t h e M S C m u s t co-operate. S i m i l a r l y, when a relay M S C i s part o f the chain. i t i s party t o the trace generation. G S M a l s o p r o v i d e s m e a n s t o t r a c e a l l a c t i v i t i e s o f a g i v e n subscriber. This m a y. i n v o l v e several N I S C / V I . R s . d e p e n d i n g o n t h e m o v e m e n t s o f the g i v e n subscriber. T h e I I I . R a c t s i n t h i s case a s a c o - o r d i n a t o r. a n d signalling means are p r o v i d e d t o achieve this. All t h e entities i n v o l v e d i n t h e tracing o f events m u s t use a c o m m o n reference. s o M a i a l l l o g s c a n l a t e r o n b e analysed together b y the m a n a g e m e n t n e t w o r k . O n t h e A i n t e r f a c e . t h e t r a c e n u m b e r i s exchanged b y the BSSMAP TRACE INVOCATION message, either f r o m B S C to MSC or in the reverse direction. The corresponding MAP/E message (between r e l a y M S C a n d a n c h o r N I S C ) i s the M A N E TRACE SUBSCRIBER ACTIVITY message, while the triggering by the H L R o f the tracing activities in the MSC/VLRs is supported by the MAP/D ACTIVATE TRACE NI0DE and DEACTIVATE TRACE. MODE messages together with their respective results. The M A P p r o t o c o l s p r o v i d e t h e s u p p o r t f o r m a n y o t h e r management r e l a t e d f u n c t i o n s . T h e M A P / F p r o t o c o l e n a b l e s t h e MSC/VLRs to get the status of a given IMEI from the EIR, as explained in the mobile stations management section. We have also seen in Chapter 7 that procedures exist i n MAP/D t o deal w i t h the recovery o f information alter a database f a i l u r e a n d r e s t a r t . T h e s e m a i n t e n a n c e f u n c t i o n s a r e mixed with traffic handling functions, and would be less efficient i f managed independently b y the management n e t w o r k alone. Maintenance i t p p o r t e d by a number of procedures in the BTSM dealing w i t h e v e n t r e p o r t s (CHANGED STATE EVENT REPORT, FAILURE EVENT REPORT, STOP SENDING EVENT REPORT, RESTART SENDING EVENT REPORT' ), as well as procedures to manage tests, for fault localisation or preventive m a i n t e n a n c e ( S E T E R R O R PA R A M E T E R S , G E T E R R O R PARAMUERS. PERFORM TEST, TEST REPORT. SEND TEST REPORT, STOP The majority of the BTSM procedures are related to operation. No procedure deals explicitly with observations. These tasks are in fact included in the tests, together with those needed for maintenance purpose. The operation procedures mostly deal with the BTS configuration. This includes the configuration of the terrestrial links between the BSC and the BTS site. An example are the procedures to connect and disconnect links. Another procedure i n the same area i s more specific: t h e ESTABLISH TEI message is used to set the TEI (see Chapter 5) used by a given TRX for instance. The configuration o f the radio interface calls for a number o f procedures. A number of them, defined as part of the BTSM protocol, are related to the radio transmission parameters (SET RF PARAMETERS, SET HOPPING PARAMETERS. S E T CARRIER 0 FILLING). T h e S E T HOPPING PARAMETERS is a first example o f a procedure at the border of traffic handling and operation. This procedure can be used to change the timedomain parameters of a channel at a specific time: i t is designed to be used during a frequency redefinition, which also involves traffic handling procedures. Another BTSM procedure which i s used f o r frequency redefinition (and maybe only for that) is the GET GSM TIME procedure, which enables the BSC to know the current position o f the BTS clock within the 3.5 hours transmission cycle (the hyperframe cycle). A difficult area is the common channel configuration, and the contents of the signalling messages sent on the SACCH and the BCCH (the RIL3-RR SYSTEM INFORMATION TYPE I t o 6 messages). T h e r e i s a 9.5.3. THE BTS MANAGEMENT PROTOCOL The A b i s i n t e r f a c e appears a s a s i n g u l a r i t y f o r t h e operation a n d maintenance protocol structure. I t is the only interface between GSM traffic handling machines used t o transfer a l l O A M information pertaining t o a m a c h i n e . n a m e l y t h e B T S . B T S management procedures can b e f o u n d i n a p r o t o c o l s p e c i f i e d i n t h e S p e c i f i c a t i o n s ( i n T S G S M 12.21). hut of optional application. and which has no specific name: we will call it the I3TSM t for [ITS management). need for consistency between some of the contents of these messages and the configuration of the common channels. On the other hand, a part of the messages broadcast on the BCCH, for instance the RACH control parameters. is under control o f traffic handling functions. The split between the RSM and the BTSM protocol is as follows: the message contents are set by the BSC through RSM procedures (using the RSM BCC!! INFORMATION a n d RSM SACCH FILLING messages). T h e c o m m o n channel configuration (number of PAGCH/RACH, paging sub-channels The point here is not to detail all the procedures. but just to use these p r o t o c o l s t o e x e m p l i f y t h e f u n c t i o n s d e s c r i b e d i n t h e p r e v i o u s sect ions. The procedures are referred to in the BTSM Technical Specification by the name of the first message: we will do the same. structure....) is set through BTSM messages. The BSC it -sponsible for the overall consistency. Other sets of I3TSM procedures support the downloading of BTS software From the I3SC'. and different other aspects of operation. 9.5.4, T H E GSM Q3 PROTOCOL. It is now time to nave closer hi the management network itself. The Spccifientipm include a basis for a GSM 03 protocol. i.e.. a protocol for network management between Operation System functions and Mediation Functions (or some traffic handling machines). This basis is of general application t o any GSM network. and must be refined and tailored to each operator's needs. The conception of such a generic GSM Q3 protocol was aimed at minimising the needs for adaptation, while providing the flexibility for them. Two aspects of the GSM Q3 protocol are important to deal with. The first one is that protocols derived from this basis use a number of standardised data communication protocols on the application level, for instance for file transfer. The stack of protocols underlying the GSM Q3 application will therefore he shortly described. The second major aspect concerns the network modelling embodied in the GSM Q3 protocol. and we will then deal with this information model. 9.5.4.1. The GSM Q3 Protocol Stack As already mentioned, connections between the entities involved in network management may be supported by various network protocols, specific t o the networking choices (X.25, SS7. ...). I n the T M N environment, a whole 051 stack is used, and this includes a transport protocol, a session protocol, a presentation protocol a n d various application protocols (layer 7 i n the OSI stack). The latter are to be distinguished from the applicative protocol itself (the end user), which is GSM Q3. In. the GSM TMN application, these protocols are among CCITT or ISO standardised protocols. F TA M is a file transfer protocol used when large amounts o f information must b e exchanged, such as observation logs or software versions for downloading. In parallel with 17IAM, usual signalling exchanges make use o f a protocol defined between so-called ACSE (Association Control Service Elements). This protocol structures exchanges into lull transactions having a beginning, an end, and things in between. It also manages the existence of several such "associations" in parallel. Beside ACSE. the concept o f ROSE Q3 application I C M I P protocols ROSE underlying layers Figure 9.20—Application protocols stack for GSM Q3 CMIP provides the common procedural basis for the management of objects, andmakesuse of FTAM for large file transfers, aswell as of other CCITT defined protocols for structuring the dialogue betweenmanagement entities. (Remote Operations Service Elements) structures the basic exchanges into operations and their results. One may well compare inter-ACSE and inter-ROSE protocols' with the transaction and the component entities used in TCAP, and described in Chapter 5. The whole protocol stack underlying the GSM Q3 protocol is shown i n figure 9.20. O n top o f the inter-ACSE and inter-ROSE protocols, the CMIP (the Common Management Information Protocol) between CMISE (Common Management Information Service Elements) provides generic procedures for the management o f "objects" between distributed data bases. I t includes procedures f o r object creation o r deletion, for the modification o f their attributes, for actions directed to these objects and for the transfer of reports concerning these objects. The notion of object as managed by CMIP will be described in a later section; objects range from a full network to a software module, a transmission link or an alarm. It seems that the terms ROP(Remote Operations Protocol) andACP (Association Control Protocol), coined on the pair CMISE/CMIP. are not used. 9.5.4.2. T h e G S M Q 3 Procedures The procedures in the Q3 protocol always refer to objects. They create, delete or modify objects. they enable the management network to ask for "actions'. from or on the objects, and they enable objects to send "events", i.e.. information that something happened, or data. Actions and events are i n some way procedures. and they are also properties o f objects. T h e main part o f the protocol specification i s organised according to classes. describing for each of their attributes (which can be the sulziect o f attribute management procedures), and the specific procedures pertaining to the object of the class. the actions and events. The procedures, in terms o f exchanged messages, are described generically in the Specifications. The following aspects are covered: • t h e generic object procedures, coming from CM1P, include means t o create and delete objects i n the databases o f the network element functions. and to set and read the attributes of these objects: • other generic procedures include means for state handling and failure reporting; others c o v e r t h e basic f i l e transfer mechanisms using FTAM: • procedures specific to some of the common objects are defined for given classes of objects. These procedures include: the measurement o b j e c t procedures, f o r reporting measurements. forcing their execution or inhibiting them; • t h e test object procedures. for starting tests and reporting their result; • t h e software object procedures. to download software. 9.5.4.3. T h e G S M Q3 I n f o r m a t i o n M o d e l The general philosophy o f T M N i s that dialogues between management entities pertain to modelled abstract representations of the network to manaue. This representation is the basic information medium on which all management operations will take place. What the operating staff see on their workstation displays are not actual BSCs or HLRs, but text o r drawings representing them. w i t h a n important level o f abstraction: only those aspects relevant to network management will be considered (this is already a lot!). This representation implies that an information model, representing and abstracting the network, be defined and stored in a management database. This model must list the different components of the network. their relationships and their attributes. The concept of attribute by t o be taken in the widest sense, covering static aspects (localisation c. an equipment, manufacturer, ...), less static ones (software versions, parameters influence traffic handling functions, ...) as well as real-time information such as the operational state (correct, or in failure) or current observations. Network components are objects, a more general concept encompassing anything which is managed. We will see that the information model covers also objects such a s tests o r observations. This information model serves as a template f o r the GSM Q3 protocol. Actions done on the representation of a BSC in this model will be conveyed by messages up to the network element function in the BSC. This agent holds the part of the model representation which is directly relevant to it, and modifies it consistently. It must also (otherwise all this would be quite i n vain!) translate the modifications o f the abstract representation into concrete actions on the machine itself. The specification of a GSM Q3 protocol appears then as the design of an information model, more precisely as the conception o f object classes. The concepts of object, class, and inheritance, which we will see soon, are part o f a general approach to the design of the information model. These concepts are, both in name and in reality, very close to "object oriented programming" in the field of software languages. A class is a description o f the attributes, properties, and actions which are common to all objects belonging to the class. The objects themselves are instances o f classes, and are dynamically created t o build u p the representation of the network. The classes are chosen during the model design and may cover many things. Typical managed objects are sites, machines such as MSC/VLR, BSC, HLR, hardware modules inside these machines. transmission links, software packages, but also activities such as observations, tests and s o on. The management o f objects, as mentioned earlier, is supported by CMIP for its basic part. Additional operations can be added specifically to each object class. This concept of object class (e.g., BSC is an object class, each instance o f the class representing one concrete BSC) is very important for obtaining both the universality and the flexibility sought in management protocols. Universality is achieved by a suitable choice of objects, o f their attributes and o f their relationships. The detailed specifications o f the GSM canonical architecture give the ability to identify object classes which will apply to all GSM PLMNs, such as BTSs, radio channels, MSC/VLRs, HLRs, links t o PSTN o r ISDN, BSCs, remote TRAUs, AuCs. network management machines, and you name it. A number of attributes and properties of these classes can he introduced in the model, independently of the actual implementation. PLMN object TRX TRX of make A TRX of make B TRX of make C I coupling device hybrid coupling device. cavity coupling deviced , Another type r relationship which leads t o a different tree representation i s thL hierarchical dependencies o f objects inside a network. For instance, a TRX is part o f a BTS, itself part of the set of machines on a given site, itself part o f a BSS, and so on. This object hierarchy must obey some rules. For instance, only certain object classes can depend on another given one: a TRX or an antenna can depend on a BTS, but an MSC cannot. Reciprocally, certain object classes can only be found as children o f a given set o f parent classes: a power supply can depend on e.g.. a BSC or an MSC, but not on a software module. These relations define a tree whose nodes are object classes. This tree is called the containment tree. and is not to be confused with the inheritance trees we have just described. The containment tree between classes represents a generic template for the containment tree of indi'idual objects in the management database. An important use of the object containment tree is the unequivocal naming of objects. The name of an object can be built up by a succession of identifiers along a unique route in the containment tree, starting at the I A interface circuits BSC I power j supply Figure 9.21 — Inheritance trees The characteristics of an object class may he derived from another one, building up trees of inheritance relations between object classes. Abis interface circuits BTS site r Flexibility is obtained by the notion of inheritance, which covers the possibility to derive a class from a parent, using all of the attributes of the parent while adding specific operations or attributes, enhancing or adapting the inherited operations and attributes. By this method, it is for instance possible to derive from a generic BTS object class (describing the universal aspects of a BTS) a more specific class corresponding to the BTS o f a given manufacturer ( b y adding specific attributes o r operations). The inheritance method is also used for a stepwise approach to the design o f the classes. For instance BTS, BSC and MSC are machines, and as such share sonic properties. This commonality can be expressed by a class from which the specific BTS, BSC or MSC classes will be derived. This avoids the need to redefine several times the same properties or actions. The relations o f inheritance give the ability to represent the set of the classes by trees as shown in figure 9.21. In such trees. child classes are derived from the parent class. power isupply BTS 1Jipower Isupply coupling i devices TRX I • I antennas Figure 9.22 — Containment tree The hierarchical dependencies of objects in a network can be represented by a containment tree. used as a model for naming objects by their position in the tree (e.g.. TRX 3 within BTS 2 within BTS site 9 within BSC 56...). root and adding at each step an identifier relative to lb. r e v i o u s object encountered. With the example o f a containment tree shown in figure 9.22 (covering only a part of the classes needed for the representation of a GSM network), a BTS number would identify a given BTS among those connected to a given BSC. or even those of a given site, the site being identified among all sites connected to a given BSC. and the naming of a !VI'S inside a given PLMN would include its BSC number, site number and 13TS number, instead of just a global and not so meaningful number. Having thus set the scent. I'm the model. the Specifications list an impro‘sive number o f generic GSM object classes, with their generic attributes. There is no point here going into the very detail o f this information model. It can be found in TS GSM 12.20, and the phase 1 version of this specification represents a first attempt at defining a GSM generic inheritance tree and containment tree. Substantial work is still going on at the date o f writing. and the phase 2 Specifications will undoubtedly present many improvements in this area, with a detailed information model rendered possible by the extensive standardisation of the GSM system. SPECIFICATIONS REFERENCE The h u l k o f t h e G S M Specifications related t o network management is the 12 series. However. a number of other specifications also contain material related to the substance of this chapter. Let us list them according to the outline ()I' the chapter. The subscriber administration aspects are introduced in TS GSM 12.02. but the more technical description of the subscriber-related data to he handled by networks is given in TS GSM 12.05. As far as charging is concerned, the tariff structure and the requirements o n interactions between networks are listed in TS GSM 02.20. whereas the transferred account procedure is described in 'I'S GSM 02.21. Maintenance is the subject of the specifications in the 12.1x series: TS GSM 12.10 is concerned with mobile stations. TS GSM 12.11 with the BSS. TS GSM 12.13 with the MSC and TS GSM 12.14 with the I ILR and VLR. The control ()I' mobile equipment identities is described in TS GSM 02.16. and the structure of the IMEI is given in I S GSM 03.03. System opera i s dealt with from the procedural point of view, but obviously cellular engineering is a matter left to the operator. A report, TS GSM 03.30, is also available, though not part of the official list. I t addresses the radio network planning aspects. TS GSM 12.07 gives general requirements o n network performance control a n d configuration. The management, collection and transfer o f statistics generated by the traffic handling machines is described in TS GSM 12.04, and the data required for network change control is identified in TS GSM 12.06. The protocol dedicated t o operation, administration a n d maintenance on the Abis interface, which we have called BTSM, is specified in TS GSM 12.21. A number of traffic handling protocols also include relevant procedures, such as the A interface BSSMAP (specified in TS GSM 08.08) or the MAP (specified in TS GSM 09.02). A general overview o f the TMN principles can be found i n TS GSM 12.00. This specification, as well as T S G S M 12.01, i s an introduction to the generic GSM Q3 protocol. The containment tree and inheritance rules for the GSM application, as well as the procedures of the Q3 protocol, are defined in the bulkiest of the 12 series specifications, TS GSM 12.20. THE LIST OF THE GSM TECHNICAL SPECIFICATIONS The whole set of official phase I Specifications is structured in 12 series, one of which is empty (the 10 series, originally planned for service interworking specifications). At the end of each chapter in this book, we have introduced a section summarising the Technical Specifications applicable, domain by domain. The reader will now find here the list of the Specifications sorted by serial number, presented in such a way as to provide a quick reference. For each GSM Technical Specification, the entry lists its number, title. number o f pages and when relevant the existence o f a separate addition for the DCS1800 standard (A DCS). Below each Specification number, we have listed the parts o f a network concerned b y the Specification. Moreover, for each Specification,qt short summary of its . s c o p e is given, with a reference t o the chapters o f the book where • r e l e v a n t information can be found. SS series: Title of the series n u m b • e r of pages Title of the Specification — impact S u m m a r y and reference to relevant chapters of this book. number of pages (+ DCS) ItltI I n k ( LIS I Di' (iSM sPrictricA-rioNs o N N I N 1 S 11 . . \ 1 01 series: General 54 p, s t a t 6 5 MS D i s t i n g u i s h , c h i c l e -mounted. portable and handheld mobile o n s . and defines their transmission power capabilities. i M o b i l e station features Vo c a b u l a r y in a C3SM P L M N 51 p. I 3 p. Presents the method used in GSM to ensure the privacy of user data and of user location and to guarantee the veracity of subscriber identities. These methods must mandatorily be supported both by mobile station and network. See Chapter 7. Defines a classification of services. which is used in the (12 series. in terms of an implementation calendar. Principles o f telecommunication services supported b y Provision o f telecommunication services 276 p. IR p. Service accessibility MS S e t s up the requirements for the selection of a network by the mobile station and defines how subscribers are spread among classes for selective access in cases of congestion. See Chapter 7. Presents the description method used in TS GSNI 02.02 and 02.03 to define services. This method uses "attributes" ( i.e.. characteristics in teats of rate. connecting network. etc.1 in the same way as ISDN specifications. See Chapter I. . B e a r e r services supported b y a G S M P L M N :.; L i c e n s i n g 38 P. R e c o m m MS. NSS L i s t s the bearer services. that is to say the see ices limited to information transport, and describes their main characteristics. See Chapter I. Teleservices supported b y a G S M P L M N 7 3 2 +A DCS 2 p. n d s that no individual license be required for the transportation or use of type approved mobile stations. 4 p. Service directory P. ass R e c o g n i s e s the existence of directories. 18p. C i r c u l a t i o n o f mobile stations P. Recommends that Administrations make provisions for the free circulation and use of type approved mobile stations. International M S equipment identities t S u b s c r i b e r identity modules, functional characteristics 4 p, 5 p. MS. OSS Introduces the concept of IMEL as an identity for the mobile equipment. and its use to control the behaviour of mobile stations through white. grey and black lists of IMEIs. 3 p. Specifies that. except for short messages. the MS may use only one (hearer or tele-) service at a time. but may the two different services t i t alternate manner inside the same 01111111UlliCall011, if indicated as such at call establishment. Ty p e s o f m o b i l e stations 7 p. OSS D e s c r i b e s different geographical entitlements for subscriptions. 23 p. MS. NSS L i s t s the supplementary services to be described in the 02.8x series. that is to say the services nuidif ing the provision of bearer or teleservices. and provides a set of definitions applicable to them. See Chapter I. Simultaneous and alternate Use o f services e Subscription to the services o f a G S M P L M N MS. NSS L i s t s the teleservices, that is to say the services including terminal functions. and describes their main characteristics. See Chapter I. 4444.:: G e n e r a l on supplementary services 3 p. Recommends that there he operators in all CEPT countries, offering a Europe-wide service to subscribers. a GSM P L M N XIS. NSS IS p. Security aspects • phases in. the G S M PI .MN 02 series: S e r v i c e A s p e c t s DCS Gives a non-exhaustive list of features which nits he offered lucidly by the mohile station. For each of them. gives their status as mandatory or optional. and specifies restrictions on automatic calling facilities. Glossary of terms and talons ins used in the ( ISM Technical Specifications with their definitions and references when need he to the ('('I'I-I' recommendations. • S e r v i c e i m p l e m e n t a t i o n phases and possible f u r t h e r 1 o SIM D e f i n e s the concept of subscriber module, or SIN-I: as an IC-card r a cut-out thereof called theplug-in SIM. Describes the life cycle of a SIN-I. and lists the items for which a non-volatile stdrage must he provided in the SIX. See Chapter 7. II p I I p. .);•e C o l l e c t i o n charges 02.87 A d d i t i o n a l i r m a t i o n transfer supplementary services NSS. M S Defines the general principles of a 1.triff structure (access charges. call charges). taking into account the various touting eases for mobile terminating culls. See Chapter K. Defines a supplementary service enabling user-to-user signalling transfer. 18 p. d. M a n - m a c h i n e interface o f the m o b i l e station 03 series: Network Aspects MS i d e n t i f i e s a minimum level of requirements for the exchanges between the user and the mobile station. leak ing open the way in „hie') the) c a r r i e d out. Also specifies a "basic public nunmachine inlet tact." using only• a telephone key pad and which _ _ s h o u l d apply to public mobile phones. See Chapter s. 1 6,62Network architecture 60 p. o ff e r i n g supplementary services 663 Numbering, addressing and identification MS. NSS D e f i n e s the call barring services and the way in which they must he managed. See Chapter 8. Signalling requirements relating to r o u t i n g o f calls to m obile subscribers N o t e : O t h e r phase 2 SpecOications dealing w i t h supplementary services should also be cited here: 022.81 N u m b e r i d e n t i f i c a t i o n s u p p l e m e n t a r y services n.1:I Call c o m p l e t i o n supplementary services M u l t i party s u p p l e m e n t a r y service Defines the supplementary yen ices dealing with more than two users in a single communication. N S 02.85 Technical performance objectives 23 p. 14( Restoration procedures • Defines procedures for restoring the consistency between network data bases ( HLRs. VLRs) alter a failure. See Chapter 7. S II p. C o m m u ) i t ( ) I ' interest supplementary service Defines the closed user group supplementary sen ice. '1.8(1 11 p. BSS. NSS Specifies performances to he aimed at when implementing infrastructure equipments. in terms of probability of failure, processing delays and transmission delays. Resulting delay diagrams for speech and data transmission are included. See Chapter 3. Defines the call hold and call wailing suppiemeniur services. 02.84 14 p. Describes the different routing configurations for mobile terminating calls. based on the relative positions of the interrogating exchange. See Chapter S. Defines the supplementary serk lees dealing with the presentation of the calling .ind called numbers. (12.83 17 p. Specifies the structure of identities for mobile stations (subscription, equipment). location areas and infrastructure entities. Also specifies the numbering plan for GSM users. 18p. restriction supplementary services 10 p. Introduces the "canonical architecture" of a GSM network, with an emphasis on the switches ( MSCs) and data bases, and lists the interlaces between them. See Chapter 2. MS. NSS D e f i n e s the call ion\ arding set' ices and the \ kay in which they must he managed. See Chapter S. • ? C a l l •N e t w o r k f u n c t i o n s Lists some of the functions to he provided by the network to handle calls in a cellular environment. MS S p e c i f i e s the different supervisory tones which the mobile station may generale liar the user to indicate the progress of a call (busy tone. etc.). Sec (*haraer 8. •' 1 6 p. :• P r o c e d u r e s f o r call progress indications ti C a l l 1 636 p. Charging supplementary services Defines the .01‘ we of charge supplementary services. r . , •' • Organisation o f subscriber data NSS Summarises the data to be stored in the network data bases (HLRs, VI.Rst with the status of each parameter (mandatory/optional. permanent/temporary). see Chapters 7 and S. 18 p. 38 p. Hark10 VC!' procedure, NSS D e s c r i b e , procedures (lases I N ' H I 1 , C h , • \ ‘ t • t VA3 ' 1 1 c e l l s 4 11111 L ' r e l l t N 1 S C s . See C'hopici (1. Support of teletex in a GSM PLMN 31) p. GSM PI. \ IX connection types Defines the transmission ;nod:, inside ( iSNI. and identifies the adaptation !unctions requiwil for each transmission configuration. See Chapter 2. . _— Technical realisation of facsimile group 3 service — transparent MS, NSS Technical realisation of facsimile group 3 service — non-transparent 9 p. Describes the procedures I, ir updating subscriber location. See Chapter 7. 1%1W ci? Discontinuous reception (DRX ) in the GSM system +A DCS 4 p. t 8p. „.. NMS. MSC Describe, the procedure, to transport DTMF (Dual Tone Multi Frequency t signals generated I)) the pressing of a key on the mobile stationduring a call. See Chapter 5. Security-related network functions MS. IWF Specifics the adaptation functions necessary in GSM for supporting automatic facsimile transmission using an "NT connection". See Chapters 3 and 8. 03.50 Transmission planning aspects of the speech service in the GSM PLMN system 42 p. 30 p. Specifies transmission performances for speech in terms of delay, distortion, crosstalk, noise, echo, etc. based on CCITT and ETSI specifications. See Chapter 3. MS. BSS L i s t s the actions required hi enable mobile stations in idle mode to reduce their reception needs. and in particular describes the impacts of N o n the scheduling of paging messages. Sec Chapter 6. Support of DTNIF via the GSM system 37 p. NMS. I WE Specifies the adaptation functions necessary in GSM for supporting automatic facsimile transmission using a "T connection". See Chapters 3 and 8. Defines the terms used lin the specification of supplementary services. and summarises the parameters and entities involved for the provision of each supplementary service. See Chapter 8. Location registratitm procedures 16 p. Summarises the transmission mode aspects applicable to teletex. — p 'Technical rcalisnlion i i i suppletitcillary s a y lees — general aspects I I p. Summarises t.._ transmission mode aspects applicable to videotex. See Chapter 3. changes and the d a t a . Technical • 'isation of videotex 44R Describes the confidentitiln a n d authentication procedures used in GSM to fulfil the requirements expressed in TS 02.09. Lists the 03.70 Routing of calls to/from PDNs NSS Identifies the implications on GSM of the numbering scheme used in public data networks (PDNs), which differs from the one used in ISDN or GSM. See Chapter 8. 03.82 Technical realisation of call offering supplementary services MS, NSS Specifics the signalling flows necessary to support call forwarding facilities and the management of their status. See Chapter 8. 03.88 Technical realisation of call restriction supplementary services 12 p. 92 p. c o r r e w o n t l i n : ; i n f o r m a t i o n 11 , , \ ‘ ‘ h e M e e n e n t i t i e s . S e e C h a p t e r 7. Technical realisation of the short message service point-to-point MS, NSS Specifies the signalling flows necessary to support call barring facilities and the management of their status. See Chapter 8. NIS. NSS Specifics the transport ph ‘tocol hemeen mobile station and short message sen ice centre and models its interactions with the underlying pmlocol A i m i gives an example of a protocol stack for connecting a service centre to an NISC'. See Chapter 8. ••• T e c h n i c a l realisation of the short message service cell . broadcast tosi, BSS specifies the liwmat of cell broadcast shoo messages on the radio interface. 29 p. [ 0 4 series: MS —BS Interface and Protocols 908 p. 7p. 04.01 MS-BSS interface —general aspects and principles Introduces the general modelling principles applied in the 04 series to specify the protocols on the radio interface. 7 p. • • • • • • 1 1 !'V% G S M PLMN access reference configurations MS D e f i n e s a functional architecture of the mobile station and identifies the teterence points at%%Inchnelmirk access is specified.SeeChapter 2. 6 p. ••••• M S -BSS interface —channel structures and access %- capabilities \IS. 1455 Defines the t pr. of channels ou ttr radio path. Gives the requirements on their configiudoon from the mobile station and from the net.. ort, points of S e e ( 'Impter 4. 9 p. MS-BSS layer I — general requirements MS. liSS Models the interaction between the link layer and the physical layer. and specifies the format of physical blocks carrying the 19 p. parameters p m 0 . 0 ) 1 1 1 m l ii 4 , 0 1 . I 0 . . 1 1 ` 1 . Point-to-pt A short message service support on mobile radio interface 0D/ 72 p. MS. MSC Specifies the type of radio channel on which short messages can the transfer protocols between these two entities. See Chapter 5. Cell broadcast short message service support on mobile radio interface .4 1 I 4 p. MS. S p e c i f i e s the segmentation of cell broadcast short messages into BSS b l o c k s on the radio path. C a R a t e adaptation on MS-BSS interface 12 p. MS, S p e c i f i e s the different steps of rate adaptation (inspired from the TRAU I S D N rate adaptation schemes) which are used to support user data between the mobile station and rate adaptation equipment on the MS-BSS data link layer general aspects MS. BTS IN:fines the protocol model applicable to the radio interface link layer and lists the modes of operation (with or without acknowledgements) on each channel type. See Chapter 5. 24 p. MS-BSS data link layer specification \IS. IVES Specifics the data link layer protocol on the radio path (LAPDm). SeeChapter 5. 82 p. Mobile radio interface signalling layer 3 — general =r aspects \is. MS. Defines the protocol model ;utile:Mk to the three protocols RR, MSC M M and CC itecilied in IS ( iSN104.08. and to the management of Short Nle,,age, and Supplementary sea ices.SeeChapter 2. 62 p. Mobile radio interface layer 3— specification NIS.I3SS. Specifies completely three protocols: RIL3-RR (Radio Resource MSC Management. RI1.3-MM Mobility Management) and R1L3-CC (Call Control ). including the procedures and the coding of messages for each of them. Combinations of the three protocols (structured procedures) arc gi‘ enas examples. SeeChapters 5, 6, 7 and 8! 446 p. MS. MSC L . • , 4 1 be transported between the MSC and the mobile station, as well as n e h r o n i s a l i o n . S e c C h a p t e r 6. Mobile radio interface layer 3— supplementary services specification —general aspects '04.11 4 4 1 network side. Sec Chapter 3. Radio link protocol for data and telematic services on the MS-BSS interface MS, IWF 59 p. Specifies completely the link protocol (RLP) used to support data services on "NT connections- inside GSM. See Chapters 3 and 5. Mobile radio interface layer 3— supplementary services specification —formats and coding 38 p. MS. S p e c i f i c s the coding of messages concerning the management of MSC, supplementary services and their impact on calls. Sec Chapter 8. HLR + DCS 22 p. MS, Mobile radio interface layer 3— call offering supplementary services specification Specifies the procedures to manage the call forwarding services, MSC. MLR using messages defined in TS GSM 04.08 and TS GSM 04.80. See Chapter 8. 04.88 Mobile radio interface layer 3— call restriction supplementary services specification MS. MSC. HLR Specifies the procedures to manage the call baring services, using 34 p. 12 p. messagesdefined in TS GSM 04.08 andTS GSM 04.80. See Chapter 8. Lists the general principles of supplementary services management as seen on the radio path. including the specification of generic procedures and the management of passwords. Also 05 series: Physical Layer on the Radio Path concernssomephase 2 service,. See Chapter N. Physical layer on the radio path (general description) MS. BSS Introduces the contents of the 05 series. Summarises the channel types and the hierarchy of time cycles. See Chapter 4. 135 p. I1 p. +A DCS r sa77.3 Multiplexing' and multiple , access on the - radio r , 36 p. Comfort ns a s p e c t s for full rate speech traffic channels MS. BSS Specifies the time and frequency characteristics of the rao... channels and gives the structure of bursts. See Chapter 4. •-cmxi C h a n n e l coding 22 p. MS. 13TS Specifies the channel coding and inter leaving schemes for all radio channel types and nodes of transmission. See Chapter 4. 714.2Modulation -"• . , MS. TRAU • D i s c o n t i n u o u s transmission (DTX) for ruff rate speech 1 0 7 . . t r a f f i c channels Radio transmission and reception MS. I3TS Specifies the transmitter and receiver characteristics and their performances (transmission power and spectrum, phase accuracy, intermodulation. receiver sensitivity, etc.). See Chapter 4. Radio subsystem link control MS. BSS. Specifies the mechanisms necessary to support handover preparation, power control. radio link failure management and cell MSC selection. Lists the corresponding parameters. See Chapters 4 and 6. Radio subsystem synchronisation MS. BSS 19 p. +0 DCS 06.32 Voice activity detection MS TRAU Specifies the hit-exact algorithm for detecting the presence of speech at the output of the speech coder (with the aim to benefit from DTX. as specified in TS GSM 06.31). See Chapter 3. +e DCS r 6 p. 07 series: Terminal Adaptors for Mobile Stations 102 p. i 49 p. n MS P r e s e n t s the functions required in the mobile station to accommodate off-the-shelf terminal equipments. Also lists the values of the compatibility characteristics for all services defined in TS GSM 02.02 and (12,03. See Chapter 3. MS 8 p. Introduces the Speryieutions in the 06 series. See Chapter 3. MS MS, BTS, Specifies the way in which portions of the speech flow should be TRAU r e p l a c e d when identified as missing, See Chapter 3. 5 p. 17 p. Specifies the adaptation functions required in the mobile station to support terminals with an asynchronous interface. Includes the specification of the data link protocol used in connection with RLP (TS GSM 04.221 to replace the terminal link protocol. See Chapter 3. Terminal adaptation functions for services using synchronous bearer capabilities -. 93 p. MS. S p e c i f i e s the bit-exact algorithm used to encode speech as a 13 TRAU k b i t / s flow. See Chapter 3. Substitution and muting of lost frames for full rate speech traffic channels General on terminal adaptation functions for MSs Terminal adaptation functions for services using asynchronous bearer capabilities 06 series: Speech Coding Specification GSM full rate speech transcoding 37 p. 38 p. Specifies ihe initial acquisition of synchronisation by the mobile station, as well as the mechanism used to maintain a good synchronisation of mobile stations received by a given base station. Sec Chapters 4 and 6. Speech processing functions: general description 13 p. MS. BIS. Specifies how to reduce the amount of transmission when no TRAL! s p e e c h signal needs to he sent by the mobile station or by the network. Sec Chapters 3 and 6. MS, 09:5Specifies the (11ISK muiluhiuun i n CISN1, See Chapter 4. t i t 5.05 Specifies how the background noise should be evaluated and information transmitted for its regeneration in the absence of speech. See Chapter 3. 7t777717 • 3 p. 6 p. Specifies the adaptation functions required in the mobile station to support terminals with a synchronous interface. Includes the specification of the data link protocol used in connection with RLP (TS GSM 04.221 to replace the terminal link protocol. See Chapter 3. 36 p. 21 08 series: BS to MSC Interfaces General aspects on the BSS-MSC interface BSC-BTS 5 p. 17 p. 2 p. Specifies the structure of 64 kbil/s digital circuits on the A interlace. based on the CCITT Recommendations of the G series. See Chapter 5. 11SC. MS(' Signalling transport mechanism specification for the d BSS-MSC interface 29 p. 92 p. BSS-MSC layer 3 specification BSC-BTS O&M signalling transport Rate adaptation on the BSS-MSC interface Specifies how network management messages are transported over the Abis interface lbut not their semantics). See Chapter 5. 08.60 In-band control of remote transcoders and rate adaptors BTS. TRAU Specifies the format of speech and data frames between a BTS and the transcoder/rate adaptor when it is remote from the BTS, and describes how these frames are used by the BTS to control the configuration and behaviour of the remote transcoder/rate adaptor. See Chapter 3. 09 series: Network Interworking 5 p. NSS NSS. MS 4 p. BTS, I n t r o d u c e s the .Speruicatioris of the 08.5s series. and lists the I3SC r e a s o n s why the Abis interface was specified. 15 p. BSC-BTS interface principles 13TS. D e s c r i b e s the split of functions between 1315 and BSC, as well as BSC t h e protocol model on the Allis interface. See Chapter 6. :• BSC-Tlt N layer I: structure of physical circuits Specifies the structure of 16 and 6+ kbit/s digital circuits on the [its. B S C A l l i s k n e l l ACC. r e f e r r i n g e x t e n s i v e l y to the ( X I I I ReV0111111Clid.1111111S in the G series and to ( i TIT 1.460. See Chapter 5. 6 General network interworking scenarios 'FRAU, S p e c i f i e s the rate adaptation characteristics for the transport of IWI: d a t a services on 64 kbit/s channels in GSM. See Chapter 3. BSC-BTS interface. general aspects 80 p. BTS, BSC Specifies the protocol used litr co-ordinating the MSC and the base station sub-system in the area of radio resource management. See Chapter 6, BSC. NI SC I p. BTS. S p e c i f i e s the protocol used for relaying radio interface messages BSC o n the Abis interface and the protocol used for co-ordinating the BSC and the BTS in the area of radio resource management. See Chapter 6. Specifies the applicability of the CCITT signalling system n°7 lower layer protocols for the transfer of signalling messages on the A interface. Also specifies a distribution function on top for identifying Llifferent .lows or messages. See Chapter 5. BSC. MSC' I BSC-BTS layer 3 specification Defines the split of functions between the base station sub-system and the NIS( a n d describes the protocol stack applicable at the A interface. See Chapters 6.7.ant.L.8. BSS-MSC layer I specification specification Specifies the applicability of the link layer protocol used for ISDN access (LAPD) for the transfer of signalling messages on the Abis interlace. See Chapter 5. introduces ilk, speruictiiions in the 08.0x series. as specifications of the A interface, BSS-MSC interface — interface principles 2 3 p. NSS + DCS p. 28 p. 666 p. 8 p. Introduces the Sperifiethions of the 09 series. Mobile application part (MAP) specification 507 p. Specifies the application protocols between exchanges and data bases (MSCs. GMSCs. VLRs. HLRs, Ellis) for supporting call management. supplementary services management. short message transfer. location management, security management, radio resource management and mobile equipment management. Specifies the applicability of the CCITT signalling system N°7 protocols (SCCP. TCAP) to support these exchanges. See Chapters 5, 6, 7. 8 and 9! +A DCS Requirements on interworking between the ISDN or PSTN and the PLMN 6 p. Presents a few considerations on the interconnection of GSM with a public switched telephone network (PSTN) or ISDN. See Chapter 3. 12P. Intervvorking between the PLMN and the CSI- . NSS D e s c r i b e s the configurations for interconnecting (iSM with a circuit-, ached puhite data network. See Chapter 3. Intent:i)-king beIN‘cten the PLAIN and the PSPDN for 19 p. PA D access a . , S p e c i f i c : a t _ o f the SIM-ME interface SIM, ME Specifies the structure of the subscriber module as far as its operation in a mobile equipment is concerned. Specifies the application protocol between the SIM and a mobile equipment, n d the applicability of ISO standards on this interface. See Chapter 7. NSS D e s c r i b e s the interworking functions in ( iSM fin- interconnecting G S M h p u b l i c sts itched P a c k e t d a t a notN‘orks t h r o u g h a +A DCS a T h e GSM base station system: equipment specification 423 p. 1355 S p e c i f i e s a set of tests to check the conformity of the base station system (13TS + BSC) +A DCS p a c k e t assembler/disassemble!. (PAD). See Chapter ;. General requirements on interworking between the PLMN and the ISDN or PSTN 53 p. Mobile services switching centre (report) z. Detailed signalling interworking within the PLMN and 3 6 1 x with the PSTN/ISDN (report) Gives ihe correspondence between call control and 'nobility management messages between a mobile station and its MSC on one side (specified in Ts GSM 04.081. and MAP or telephone user part/ISDN user part messages on the other side. for typical signalling flows involving the MSC. See Chapters 7 and 8. Information element mapping between MS-BSS/BSSMSC signalling procedures and MAP Maps mobility management and call control messages used between a mobile station and its MSC (specified in TS GSM 04.(181 to messages transported between MSC' and VLR (specified in TS CISNI 09,02). and vice versa. •,!))?)..!:;i7F),:,,•.)3: V,c4m, t; (.t.-2 Signalling interworking for supplementary services 55 p. MSC S u m m a r i s e s the functions of an MSC, giving their status (mandatory/optional), and lists the relevant Specifications. NSS D e f i n e s the interworking functions for interconnecting GSM with ISDN or PSTN. Maps the GSM services characteristics (GSM bearer capabilities) to the ones used in ISDN. See Chapter 8. NSS 131 p. Home location register specification (report) 9 p. HLR S u m m a r i s e s the functions of the HLR and lists the relevant Specifications. F a l • Visitor location register specification (report) 13 p. VLR S u m m a r i s e s the functions of.the VLR and lists the relevant Specifications. 16 p. +8 System simulator specification 46 p. Specifies the equipment which is used to perform type approval tests on mobile stations, according to TS GSM 11.10. See Chapter 9. +A DCS DCS 12 series: Operation and Maintenance 794 p. 9 p. NSS M a p s parameters or messages transported on the radio interface in the area of supplementary services into messages to be transported within the signalling system if'7 network. and vice versa. See Chapter 8. Objectives and structure of network management OSS 11 series: Equipment and Type Approval Specification U. M o b i l e Mobilestation conformity specifications 523 p. MS S p e c i f i e s a complete set of tests and the associated methods of measurements to check the conformity of mobile stations to the Spcifinainns. Serves as a technical basis for mobile station type approval. See Chapter 9. + DCS 51 p. Defines the objectives of network management. Introduces the TMN (Telecommunication Management Network) concepts and the relevant definitions. See Chapter 9. Note: the phase 2 version of this specification is an excellent primer in the domain of network management. '01 Common aspects of GSM network management OSS Describes the GSM network management functional areas, interfaces and protocols. Subscriber, mobile equipment and services data administration NSS. OSS Lists the network management functions in the domain of subscriber management and mobile equipment management. 69 p. 17 p. t e r r y Security management 16 p. 14145. Describes the management aspects of security functions listed in NSS. OSS TS GSM 02.09. and introduces the requirement. for access control to the operator's equipment. ..:7?"":34 Performance data measurements BTS. 13SC 45 p. USS. D e s c r i b e . the administration. collection and transfer of statistics NSS, 11SS measured bs the transmission machines. Subscriber related event and call tlatil 27 p. 0145 S u m m a r i s e s the data which an operator must manage to configure a network. Operations and performance management ass 61 p. Defines the requirements to be fulfilled to control the performance of a network and to configure the system. 41 Maintenance provisions for operational integrity of Ms 7 p. MS I d e n t i t i e s ('aults of nubile stations which can affect the network and describes maintenance actions in detect and isolate them. '.:4411172 Maintenance of the base station subsystem 39 p. tiss. OSS Defines the functions related to maintenance actions to be provided by the BSS (faun diagnosis. recovery actions). Maintenance of the mobile services switching centre MSC. OSS 10 p. Defines the functions related to maintenance actions to he provided b) an MSC, in addition to those of a normal digital exchange. Maintenance of location registers 9 p. II.R. D e f i n e s the functions related to maintenance actions to be VI.R. p r o v i d e d by a I II.R or a VI.R. ass 4.:,711Tfitt.- ,t1:gPJ Network management procedures and messages BSS, D e s c r i b e s the procedures necessary to fulfil die network NSS. OSS management functions defined in the previous Speerfications of the 12 series. Includes the dellinition of object classes and their attributes. See I 'limner 1). Specifies a protocol for network management of the BTS as a superset of the procedures needed for operation of any one BTS by the BSC. See Chapter 9. Total: 5230 pages of official phase I Specifications. p. NSS. ()SS Describes the data to he generated and collected in elation with a given subscriber Ifor s t a l i s = l l i n g and tracing purposes). 4 0 : GSM network change control Network m' cement procedures and messages on the Abis interfats 342 p. 83 p. BIBLIOGRAPHY MOBILE COMMUNICATIONS IN GENERAL E.A. LEE, D.G. MESSERSCHMITT. Digital Communication. Kluwer Academic Publishers (Boston). 1988 J.D. PA R S O N S , D . J A R D I N E , J . G . G A R D I N E R , M o b i l e Coninnuileation Systems. Blackie (Glasgow), Halsted (New York), 1989 G. CALHOUN, Digital Cellular Radio, Artech House, 1988 GSM IN GENERAL Most o f the existing literature about G S M consists o f the proceedings of seminars or conferences. Two seminars, organised by the MoU and devoted to a general presentation o f GSM, were held i n Europe, at Hagen (FRG), October 1988, and in Budapest, October 1990. Besides, there are regularly. about once a year since 1985, a technical conference in Europe about digital mobile communications. These are the forums where most technical papers directly related t o GSM are published. A number of such papers, among the more recent, are listed in the bibliography hereinafter. The full list of these conferences is given here, for those interested in what was said on the subject in the first years: • D M R 1 (Nordic Conference o n Digital L a n d Mobile Radio Communication). Espoo (Finland), February 1985 (-20° C!) • D M R II. Stockholm. October 1986; • D M R C (International Conference on Digital Land Mobile Radio Communications). Venice (Italy). July 1987; • D M R III. Copenhagen. September 1988; • D M R IV. Oslo, June 1990: • M R C (Mobile Radio Conference), Nice (France), November 1991; • D M R V. Helsinki. December 1992 (not held when printed). 668 r n i . I o l t i L i o U R AV I I Y iSN1 s v s I NM CHAPTERS 1 AND 2 T. HAUG, "The GSM Programme, a pan-European effort". Proceedings of the Mobile Radio Conference. Nice, November 1991 A. FOXMAN. "Activities of the group of Signatories of the GSM MoU". Proceedings of the Mobile Radio Conference. Nice, November 1991 6 6 9 M. BERTELSMLI.., et al., "Design and Performance o f Punctured Convolutional Codes f o r the Pan-European Mobile Radio System". Proceedings of the third Nordic Seminar on Digital Land Mobile Radio Comnutnication, Copenhagen, September 1988 G.D. FORNEY. "The Viterbi Algorithm", Proceedings of the IEEE, 61, March 1973 A. 11/DDEN. "Development of the DCS 1800 standard", Proceedings of th 9 :sv1 IL rc.N m d a ilR b o M e A. K R A NTZIK, D . W O L F, "Statistische Eigenschaften v o n Fadingprozessen z u r Beschreibung e i n e s Landmobilfunkkanals", frequen:, vol.44, no.6, June 1990 13. MALLINDER. "An overview of the GSM system", Proceedings of thq third Nordic Seminar on Digital Land Mobile Radio Communication: Copenhagen. September 1988 G. CASTAGNA, A. COLAMONICO, R. FAILLI, R. MONTAGNA et al., "Italian experimental activity on the European digital land mobile system". CSELT Tech. Rep., vol. 17, no. 2, April 1989 CHAPTER 3 R. MONTAGNA. "The half rate speech voice codec f o r the GSM system". Proceedings of the Mobile Radio Conference, Nice, November 1991 I. DITTRICII ct al.. "Implementation of the GSM data services into the radio mobile r a d i o system". Proceedings o f the Mobile Radio Coqerence, Nice. November 1991 G. CRISP, A . EIZENHOFER, "Architectural aspects o f Data and Telematic Services in a GSM PLMN". Proceedings of the third Nordic Seminar on Digital Land Mobile Radio Communication, Copenhagen,- September 1988 T. MASENG, "On selection of system bit-rate in the mobile multipath channel", IEEE Trans. VT, Vol. 36, no. 2, May 1987 W. W. PETERSON, Error-Correcting Codes, M.I.T. Press, Cambridge, Mass. (1961, 1970) CHAPTERS 6, 7 AND 8 U. JANSSEN, P. BRUNE, "The Mobile Application Part protocol", Proceedings of GSM Seminar, Budapest, October 1990 M. B . PAUTET, M . M O U LY, "GSM protocol Architecture: Radio Subsystem Signalling", Proceedings o f VTC'91, Saint-Louis, Missouri, May 1991 A. CLAPTOIC1 e t al., "Supporting Facsimile i n the G S M PLMN", Proceedings of the .fourth Nordic Seminar on Digital Land Mobile Radio Communication. Oslo, June 1990 Chapter 6 P. V A R Y, R . H O F M A N N . "Sprachcodec f t i r d a s Europaisthe Funkfernsprechnetz-, Ft-mien:. vol. 42, no. 2-3, February 1988 I. B R I N I e t al., "European roaming related technical problems", Proceedings of the Mobile Radio Conference, Nice, November 1991 CHAPTER 4 Chapter 7 A. MALOBERT1. "Some aspects o f the G S M radio interface". Proceedings of the third Nordic Seminar on Digital Land Mobile Radio Conummicalion. Copenhagen, September 1988 B. CHATRAS, C . VERNHES, "The European mobile service: a n application o f the intelligent network concept", Proceedings o f the 1st international seminar on intelligent networks, Bordeaux, March 1989 M. B A L L A R D , D . V E R H U L S T, "ECR900 : l a radiotelephonie numerique europeenne d'Alcatel", Commutation & Transmission, no. 4, 1988 (171) ' M E i t i N 1 S VS I E M G. MAZZIOTTO. "The Subscriber Identity Module I... the European Digital Cellular System G S M a n d other Mobile Communication Systems". Proceedings o f X/1: International Su'it•hitng Symposium. Yokohama. October 1992 liApTER 9 R. HAGFDOOR N. "Confortnift Testing o f (ISM mobile stations". Proceedings of GS:11 Seminar. Budapest, October 1990 A. BERGMANN et al.. "Protocol conformance testing of a GSM mobile station". Proceedings of the Mobile Radio Conference, Nice. November 1991 R. KOSTER et al.. "ISO test methods and their applicability to the GSM 'nubile network system", Proceedings of the Mobile Radio Conference, Nice, November 1991 P. GUILLIER e t al.. "Implementation and operation o f the GSM network". Proccedinms of the Mobile Radio Conference, Nice. November 199 I II. PERSSON. "Performance of GSM options in a cellular environment", Proceedings of M o b i l e Radio Conference. Nice, November 1991 R. THOMAS. M. MOULY, et al., "Performance evaluation of common control channels in the European digital cellular system", Proceedings of. the Digital Mobile Radio Communications Conference, Venice, July 1987 I2. THOMAS et al.. "Influence of the moving of the mobile stations of a radio mobile cellular network", Proceedings of the third Nordic Seminar on Digital Land Mobile Radio Communication, Copenhagen, September 1988 D. VERHULST. "High performance cellular planning with Frequency Hopping", Proceedings o f the. fourth Nordic Seminar-on Digital Land Mobile Radio Communication. Oslo, June 1990 R. WYRWAS. J. C. CAMBELL, "Radio topology design with slow frequency hopping f o r interference limited digital cellular systems", Proceedings of I TC19/, Saint-Louis, Miss., May 1991 INDEX Signalling messages are referred to in this index by alphabetical order o f the le\ %tut mane. A separate message index, sorted by protocols, follows the general index. .1/,..%‘age /tames are i l l small capitals. and message field names are i l l italic small capitals. A A interface impact of the transcoder position 1 5 2 protocol stack 2 8 8 BSSMAP/DTAP distinction 2 8 9 terrestrial channel assignment 3 2 2 functional split for handover 3 6 3 circuit blocking 5 8 0 dimensioning 6 2 5 A? (authentication algorithm) 4 7 8 AS (ciphering algorithm) 2 4 9 , 4 8 1 , 4 8 2 AK (ciphering key computation) 4 8 3 Abis interface addressing of signalling messages 286 message discriminators 2 8 6 relaying of radio path messages 2 8 7 functional split for handover 3 6 4 RSM protocol 3 6 6 dimensioning 6 2 2 access concept of access procedure 1 9 2 RACH 1 9 3 format of an access burst 2 3 4 time delay for access bursts 2 3 6 general description 3 / 7 access channels 3 5 / capacity of common channels 351, 6/9 reasons for access 3 6 7 initial message 3 6 8 contents of the citAsmit. REQUEST message 3 7 3 on a new cell at handover 4 1 0 BCCH information 4 2 7 access class mechanism 3 7 1 list of special access classes 3 7 1 information sent on the BCCH 4 2 7 mobile station behaviour when the class is barred 3 7 2 , 4 4 2 access grant see initial channel assignment acknowledgement link layer mechanism 2 7 3 sliding window 2 7 4 ACSE, Association Control Service Element 6 4 0 ACTIVATE TRACE MODE (MAP/D) ACTIVATE SS ( M A P / I ) 5 6 5 3 8 5 Advice of Charge 6 5 AGCH, Access Grant Channel see PAGCH alarm generation 5 8 2 ALERT SERVICE CENTRE (vIAP/C) 5 6 3 ALERTING (RIL3-CC) 534,538, 542, 543, 553 allocation see dedicated channel allocation Aloha s e e random access alternate services procedures 5 4 7 anchor MSC networking between anchor and relay MSC 2 9 0 stable core of an RR-session 3 / 7 role to avoid duplicated messages 391 AoC. Advice of Charge 6 5 bb architecture methodology of description . \ I l (;sh.1 sub-systems CISNI functional planes / kIZ functions . 3 6 2 lot:alio') management lunciiiins 4 ' 1 locution management pi otocol, security protocols / . \ • •protocol architecture • . : • SN-IS protocol architecture OA NI protocol architect Ire .14:4:41s/o. .e• c h a n n e l ; i l k .::41,1).;iii, ‘, \sSII l \ \ ( W I sSIGNNIEN I DiSSSIAPI R E .00 ;-RR ssSONSILN I I I I I RED(11.3-1th ) . 0 ) 0 , 4 1 2 SSI(iNNINN I REQCEST (BSSNIAM as)ncluonous ;isynchronous data h a y AO iransformation handover ABC. Authentication Centre architectural description / 3 6 0 2 / / role authentic:ition introduction of the function 7 / . 4 7 description of the AtiC / 0 1 mechanism 4 N subscriber key Ki 4 7 S algorithm requirements 4 7 9 algorithm choice 4 N ( ) exchange of triplets 4 8 9 . 492 Authentication Centre At TIIENTICATION REQUES1 ( R I L 3 - \ I \ 11 4 N N 488 BAIC. Barring of All Incoming Calls service description 6 2 BAOC. Barring of All Outgoing Calls service description 6 2 barring services - description of BAIC description of BAOC 2 8 I RSNI) 6 3 9 beacon Irequenc> 225 einis,uni requirements ' 2 7 . 333 of surrounding cells 3 3 3 frequencies lo monitor 4 2 6 bearer capahilit> concept 1 3 2 use fur call control 5 0 5 determination by the HLR 5 4 5 hearer services list of bearer services 5 8 description 5 0 6 BHCA. Busy Hour Call Attempts 6 / 4 BIC-roam. Barring of Incoming Calls when ROAMing outside the home PLMN country service description 6 2 charging aspects 5 / 8 billing 5 7 2 black list (of INIER) 5 9 / BLOCK (BSSMAP) 6 3 7 BLOC'KINll ACKNOWLEDGE (BSSMAP) 6 3 7 At TIIIENTICATION RESIONSE(RIL3- \ 1 1 B 4 13CCI I frequene s e e hem on frequency BC('I I I \ I I '1t \I SIP C O M P E L ( 1: ( U S S M A N \ Sll Am! v u o m i l 192 207 371 4/3 use aii call re ,•,lablisliment gncn 4 2 4 ci n t , ;Ind Inc 4 2 4 to 429 message p e r i o d i c i t y ssic C O N I N I A N I ) IR113-1*() 62 62 62 555 145 145 description of Iii i a r n description 01 ROIL' description of 1301C-ex/IC aci k ;MI in procedure basic packet access basic P. \I) act es, 13(r1 I. Brolorti51 Control CHannel purpose time org.iiiisautin BOIC. Barring of Outgoing International Calls service description 6 2 1301C-cNI IC. Barring of Outgoing International Calls except those directed toward the home PLMN country service description 6 2 BP B u r s t P e r i o d . 5 7 7 ps rationale for its value / 9 9 35itowER I ' ( rsTROL (RSM) 4 2 1 liSC. Base Station Cont. J r architectural description 9 7 choice of the radio channel type 3 2 2 ierrestrial channel allocation 3 2 4 radio channel assigmnent 3 5 5 handover 3 6 3 scheduling of paging messages 3 8 2 transmission mode change 3 8 6 switching capability 3 8 8 collisions between procedures 391. 395 395 ciliuml of claw nlink dimensioning 623 13SIC, Base Station Identity Code definition 3 3 6 measurements process 3 3 7 111.MN colour code 3 3 8 use at handover 3 3 8 , 407 BSS. Base Station Sub-system architectural description 9 4 BSSMAP. BSS MAnagement Part BSSMAP/DTAP distinction on the A interface 2 9 0 introduction as an RR protocol 3 6 6 management aspects 6 3 7 message list 6 9 3 RTS. Base Transceiver Station architectural description 9 5 role for scheduling paging messages 382 role in the mode modify procedure 386 cipher mode change 3 9 3 sending of physical information messages 418/ management 635. 638 6 BTSM. BTS Management 639 burst definition /95 burst period (BP) /95 usage in GSM /96 numbering period 216 burst formatting 230 types of bursts 23/ to 238 format of a normal burst 232 guard time 232 133,735 amplitude profile format of an access burst 234 S burst (on the SCH) 236 F burst (on the FCCH) 237 full-rate speech burst 248 C C/I, Carrier over Interference ratio cumulative distribution CI definition comparison between cells in limited service mode CALL CONFIRNIED(RIL3-CC) call control call hold FA L L PROCEEDING (RIL3-CC) 602 453 453 458 540 see CC see HOLD 534, 535. 538- call re-establishment cell choice procedure call waiting camp (to camp on a cell) 4/ 3 413,496 see CW 434 CANCEL LOCATION (MAP/D) 470 CANDIDATE RESPONSE (BSSMAP) 419 cavity (coupling) 631 CBCH, Cell Broadcast CHannel 193,211 channel description 426 CC, Call Control 521 role of the CC sub-layer 115 protocol architecture 528 CC, Country Code 511 C C C H L O A D INDICATION (RSM) 418 cell impact of timing advance on cell size 346 very large cells 347 CELL J A R _ACCESS 425,442 CELL CHANNFIL DESCRIPTION 427 CELL IDENTITY 423 CELL OPTIONS 428 test of a new cell 579 cellular planning 593 to 613 maximum range 596 range vs. interference 5 9 9 , 612 probability distribution 6 0 4 parameters for cellular planning 6 2 0 CELL ALLOCATION 3 6 ) cell selection in cases of congestion 3 7 2 at call re-establishment 4 1 3 compatibility with paging 4 2 5 idle 'node requirement 4 3 4 aim in idle mode 4 4 1 impact of the registration status 4 4 5 vs. PLMN selection 4 4 7 algorithms 4 5 2 to 459 the CI criterion 4 5 3 hysteresis 4 5 5 O// 394 IER MODE C o M P L I M : t 5 1 . 5 1 , 1 cellular engineering goals 5 syslems 9 565 3 '7 InAior E u r o p e a n systems all main principles ( T B . C a l l Forwarding on mobile subscriber Busy seri.ice_descripiion handling t h e NISC/V1.1( 230 processing ciphering algorithm 2 4 9 . 4 8 1 3 2 7 380 39/ cipher ',ide change procedure setting of the cipher 'node al handover 402 cistern:II 55 specifications 4 8 2 to acknoss ledge a C M - t r a n s a c t i o n i v g istrat km p r o c e d u r e estahlishment T- N l i c . C a l l F o r w a r d i n g on mobile subscriber Not Reachable 4 9 n3 527 555 handling b y the M S C / V L R registration procedure ( T N R y . C a l l Forwarding on No Reply 64 service description 5 2 7 555 re gist r a t i o n p r o c e d u r e e t t •. C a l l F o r w a r d i n g U n c o n d i t i o n a l CIPHERING %HOE CompLETE (R11.3-1(R) 3 9 3 C K S N . Ciphering Key Sequence Number 488 classmark 379 379 to 382 contents CLASSMARK CHANGE (811.3- K R ) 3 8 1 ('LASSMARK uptyvtit ossmAp) 3 8 ! 2 CI.EAR COMMAND IRSSMAP) 4 1 1 , 4 1 5 , 4 1 8 111.R handling 5 registration procedure 2 5 7 CLEAR COMPLETE (135551AP) 4 1 5 , 4 1 8 e definition e 5 5 a l s o w a i n elhIMICI / 9 4 4 clearing house 7 5 I ' l l \ \ E L m l I V A T I O N ACKNOWLEIXIE 404 IRSNI) channel coding 2 general principles 2 2 8 4 f o r all t r a n s m i s s i o n m o d e s 2 for full-rate speech 4 2 service description 1 4 n 8 8 channel release after successful handover at the end of the RR-session I I I N M I RELEASE: (R113-12R) \ R N A uisT (H113-RR 3 6 8 . 3 7 , NH. M A X I M : 1 ) ( 1 * A I ) • charging for calls toward mobiles impact of roaming mechanisms transfer of data 513 5/5 57: c m i t 591 $9.4 S I M / I 1 ipill:R \1(11)1 c o N I M A N D I l i S S M A N 575 service description 6 service description conthirt noise / column!) channels /92 192 /92 /93 /93 206 to 211 208 PA G C H RACH time organisation combinations as f i x e d f r e q u e n c y c h a n n e l s 1 1 1 3 / 2 9 description on the B C C H dimensioning CSI SERVICE ACCEPT ( R I L 3 - M M ) •415,494,496 4 0 415,494,496 5 conference bridge queuing 0 4 4 0 378,394,494 3 5 3 4 9 4 6 4 1 , 642 C M I S E . C o m m o n Management Information Service Element 0 4 1 600 / impact o n services 4 3 5 indoor vs. outdoor 6 / 2 CP-ACK (SM-CP) . 1 0 3 CP-DATA (SM-CP) 3 0 3 CP-ERROR (SM-CP) 3 0 3 C S P D N , C i r c u i t Switched Public Data Network C S P D N - b u s e d services 5 access c o n f i g u r a t i o n s 6 1 4 7 C T R , C o m m o n Te c h n i c a l R e g u l a t i o n 5 8 6 C U G , C l o s e d User Group service description 6 6 C W , C a l l Wa i t i n g service description 6 4 impact on call establishment 5 4 2 3 2 D / transmission inside G S M / 6 6 to / 8 4 B T S - T R A U data blocks / 4 8 data s e r v i c e s description 5 0 5 2 in- b a n d c a l l e s t a b l i s h m e n t 7 5 8 3 7 0 7 1 3 cell s e l e c t i o n p r i n c i p l e s 3 7 2 , 4 4 1 5 5 3 8 D C S 1800, D i g i t a l C e l l u l a r S y s t e m a t 1 8 0 0 MHz introduction 2 frequency bands 2 9 1 7 CONNECT (RIL3-CC) 5 3 5 , 5 4 2 , 5 4 3 , 5 5 1 coding o f frequency lists 3 6 2 , 4 2 8 CONNECT ACKNOWLEDGE (RIL3-CC) p o w e r classes 3 national r o a m i n g 4 5 3 5 SCCP connection 2 3 8 / 9 0 , 4 DEACTIVATE SACCH (RSM) 4 DEACTIVATE TRACE MODE (MAP/D) DEACTIVATE SS (MAP/1) CONNECTION FAILURE INDICATION (RSM) connection types T connections NT connections containment (TMN concept) 8 3 9 417,4/8 generic establishment , automatic call answering 5 3 5 , 5 4 3 C M -transaction C M I P. C o m m o n M a n a g e m e n t Information 4 3 n u m b e r i n g issues 5 use o f access c l a s s e s RR-connection CNI SERVICE REOFEST ( R I L 3 - M M ) 9 6 242 charging end- t o -end data transmission 3 connection CM SERVICE REJECT ( R I L 3 - M M ) 3 coupling device e data 5 b o t t l e n e c k s f o r access 3 cost (of a network) 335 35/ 426 618 6/9 configurations pre-emption 5 CM RE-ESTABLISIINIENT REQUEST ( R I L 3 - M M ) Protocol 2 congestion i n t r o d u c t i o n o f the C M l a y e r e consequences o f continuous emission 379 CUG s 2 SCH 24/ coverage 5 6 426 role f o r channel coding cost ( f o r c u s t o m e r s ) 5 6 377 principles 4 414,496 411 -I/O 41n 375 374 BS1C C o l . R . C o n n e c t e d Line Identification Restriction terminating e 4 e COMPLETE LAYER 3 INFORMATION (BSSMAP) 6 CM-transaction e m o b i l e o r i g i n a t i n g vs. m o b i l e 4 e r o l e o r the C M l a y e r s communication C.M. C o m m u n i c a t i o n M a n a g e m e n t 7 convolutional codes C o l . P, C o n n e c t e d L i n e I d e n t i f i c a t i o n Presentation characteristics 6 s 2 7 2 capacity CL1R. C a l l i n g Line Identification Restriction Closed User Group e l l . \ \ N H . MODE MODIFY (RIL3-RIO 3 8 7 c l i A N N E L 5101E MODIFY ACKNOWLEDGE 3 8 6 Presentation service description introduction 1 ( ' L I P. C a l l i n g L i n e I d e n t i f i c a t i o n RR-session 3 / 3 s A c r i v A m 1 N (RSM) 3 8 8 , 4 0 4 IRII.3-RR I CLEAR REQUEST I ( S S M A P ) C OKI ROL CI I AA'NEI. DESCRIPTION p r o p e r t i e s o f the 51- c y c l e 0 s c o l o u r code at i n i t i a l c h a n n e l a s s i g n m e n t / 227 service d e s c r i p t i o n channel 4 ITCH 393.394.494 purpose 2 BCCH 4 CIPHERING MODE ('I A l VIAND ( R11.3- K R ) .en ice description handling b y the M S C / V L I Z 325 M O N I C Sla0011 2S ovation of GSM r a d i o path error detection i n l i n k p r o t o c o l s I r a n s i l i o n s belkkeell I n o d e s ( ' F Y I . C o n f e r e n c e Europeenne des P o s i e s e t Te l e c o m m u n i c a t i o n s contention resolution 71.477 i n t r o d u c t i o n as part o f signal II ill m1001011 codes error correction and detection on the ciphering introduction of the function 4 5 4 / 5 8 6 3 8 5 5 dedicated channel allocation requirements /77 s t r a t e g i e s 3 3 5 4 5 5 / 7 9 p e r f o r m a n c e s o f different strategics 6-15 356 queuing 3 priority mechanisms 5 7 3 5 8 671.1 INDEX ' I / I L ( S I / STSTEll dedicated channels 1(11/I: and I C I I/11 / 8 9 SACCII / 9 0 'I I/8 / 9 / ollser between uplink and downlink 200 intervals, for neighbour cells 371.334 measurements 335 properties of the 26-cycle 397 channel change at handover 4/6 nlease pnicedure 776dimensioning dedicated mode 44 concept /92 definition 315 RR session /46 dedicated PAD access 376 F INDICATION ( a S M ) 1)19.1.1 k St 711SCRIBER D ATA (SIAP/D) 4 7 0 density (of traffic) 3 5 9 1)11(1 I lS1 ER MOBILE SUBSCRIBER f m A i l i ” 471 directed retry description 3 9 7 usage 4 4 1 Ins(ovsErr tRa.3-cc) 5 3 4 , 5 4 5 discontinuous transmission s e e M X 1)1,CI. Data Link Connection Identifier use on the A interface 2 9 0 downlink / 9 3 (MT-and-insert 6 2 3 M N , Discontinuous Reception scheduling of paging messages 319 to 320 DTAP, Direct Transfer Application Part introduction 2 8 7 BSSMAP/DTAP distinction on the A interf4 2 9 0 DTMF. Dual ! n e Multi Frequency description 5 0 8 transmission on the radio path 5 5 2 I D i s c o n t i n u o u s Transmission introduction / 6 1 impact on the TRAU 1 6 4 uplink and downlink IYI'X 3 2 5 intimet on measurements 3 4 1 modes at channel activation 3 8 8 1)IX control procedures 3 9 5 (1)( impact on cellular planning duplex separation 9 ) , 7 duplication 390 of radio interface messages Fire code role for channel coding performances fixed networks functions offered to GSM flag stealing flag pattern for link layer frame delimitation flow control link layer mechanisms l'N. TDMA Frame Number definition E.164 5 2 1 I' 5 2 ' NA. Early Assignment 3 5 5 FIR. Equipment Identity Register 107.590 cmergene) calls definition n hen roaming 4 3 5 , 4 3 7 I.\ 'Had m•N o u i . 3 - m 5 3 3 cnci- m o n s e e ciphering t\ciol'IIONCOMMANDIRSND • 3 9 3 engineering goals of system engineering 5 9 2 cellular planning 5 9 3 to 613 goals of cellular engineering 593 to 595 equalisation 2 5 7 ERASE SS ( \ A i l i ) 5 5 5 Edang 13 formula 6 1 6 error detection and correction codes on the radio path 2 4 1 in link layer protocols 2 7 2 ESTABLISH INDICATION (RSM) 3 7 GET RESPONSE (MM- M E ) 86 G l o b a l System f o r M o b i l e GMSC. Gateway MSC 2 7 9 d e s c r i p t i o n o f the f u n c t i o n s position 5 1 7 2 / 5 i n t e r r o g a t i o n procedure . 5 52/ 4 4 G M S K , G a u s s i a n M i n i m u m Shift Keying 5 5 6 forwarding services description of CFU description of CFB creation description of CFNRy impact on call routing impact on the bill frame 64 H D L C frame delimitation link layer frame numbering 273 LAPD and LAPDm frames frequency frequency diversity 2R0 219 n u m b e r o f T R X s vs. f r e q u e n c i e s in a cell 227 change of the frequency allocation 353 timing of a frequency change 3 5 4 coding o f a frequency list 3 3 6 frequency redefinition 4 6 0 2 2 7 3 frequency hopping techniques also SFH 8 6 6 2 0 9 FREQUENCY REDEFINITION (RIL3-RR) 4 2 4 F r A m . File Transfer Access and Management 6 4 0 . Hamlet (citation f r o m ) 4 8 9 handover s e e also measurements definition 4 5 confinement handover 3 2 8 purposes of handover 3 2 8 rescue handover 3 2 8 3 3 2 2 3 8 9 3 0 parameters f o r h a n d o v e r decision 3 3 0 7 8 0 e / 36 119 29 232 347 H algorithms FREQUENCY CHANNEL SEQUENCE e 32 35 name and l o g o criteria 4 1 for GSM900 2 1 mobile station capability 3 s 2 s y s t e m traffic h a n d o v e r 1 frequency bands f o r D C S 1 800 29 526 o r i g i n a l system requirements 528 c h o i c e of the main radio access characteristics signalling architecture (figure) 180,282 GSM900 2/5 guard time 269 impact on cell size 269 concept o f link layer frame cell allocation 28 63 G S M , G l o b a l System for Mobile 63 c o m m u n i c a t i o n s description o f C F N R c r. Groupe Special M o b i l e ( G S M ) 62 o r g a n i s a t i o n impact o n spectral e ff i c i e n c y 0 /03 5/2 5/7 role T D M A frame 7 global title 269 a r c h i t e c t u r a l description R L P frames 2 see GSM 294 communications /90, 238 FORWARD SS NOTIFICATION (MAP/I) F in LAPDm frames 556 488 2 4 4 G E T PASSWORD (MAP/I) 291 c h a r a c t e r i s t i c s of the GSM modulation 1ORWARD CI IECK SS INDICATION (MAP/I) 249 to 258 556 f o r m u l a s 2 5 / FORWARD SHORT MESSAGE ( M A N N ) modulation spectrum 2 5 5 305.562 grey list ( o f I M E I s ) 5 9 / ETS. European Telecommunications 3/ Standard European Community type approval directives 5 8 5 filling pattern G 241 FORWARD ACCESS SIGNALLING (MAP/E) 9 F burst 2 3 7 FAC. Final Assembly Code 5 8 9 FACCH. Fast Associated Control CHannel see fast associated signalling FActirri (kiL3-CC) 5 5 3 facsimile reference configuration 1 3 9 transmission functions / 3 9 example of an alternate call 5 4 7 fading 2 1 9 failure detection 5 8 1 fast associated signalling definition 1 9 0 mechanism 2 3 7 fault detection 5 8 1 FCC I I. Frequency Correction CHannel ti me organisation 2 0 6 role for synchronisation 2 / 4 use to identify a beacon frequency 333 FCS. Frame Check Sequence 2 7 2 679 642 pre-synchronisation 3 3 3 assessment of the timing advance 3 4 8 synchronous vs. asynchronous handover 3 4 9 , 3 9 7 applicability of queuing 3 5 7 functional split between entities 3 6 3 pre-processing o f measurements 3 6 4 handover execution procedure 396 to 412 tun II tuu.3-cri 5 5 0 directed retry l(?99,87001.1)AckNowii.ix;y tkit 5 5 0 inter vs. inira-BSC handover home PI.NIN inter vs. infra-MSC' handover 3 9 8 definition 4 3 7 subsequent inter-MSC handover 3 9 8 search w MI national roaming 4 4 5 e ecution phases $ u u search at so itch-on 4 4 7 , 457 handover number 4 0 4 charging principles 5 / 4 use ot a conference bridge 4 0 5 hopping s e e ,S'FH sot; de target cell vs. list of cells 4 0 5 INN. Hopping Sequence Number access on the new cell 224 introduction release of the old path 4 4 ( 1 ) 1 properties of hopping sequences 224 mira• BSC handover 4 1 2 360 coding impact on spectral efficiency ITHialce 6 0 5 631 2/6, 482 hyperframe II D o v i i R ACCESS ( R 1 1 . 3 - R I O 4 1 0 hysteresis II NNI)DVIiR CANDIDATE. ENQUIRY triSSMAPI 455 for cell selection 4/9 -MAP II \ DDVER CDNIMAND HISSNIA19 I II NDDVER COMMAND (1411.3-KR) 391,399,404,407.409 IDENTITY RI:0'ES] (R11.3-NIND 4 9 2 , 5 8 9 ilysix(vER COMPLETE (RSSh1AP) 4 / 1 nmystrry RESPONSE (RiL3-Nut) 4 9 2 ANDOVERCONIPLETE (RI1.3-RR) 4 1 1 idle mode • DovvR ( i i s s m A N 4 1 1 concept 4 4 IANIX)VER DETE(TION (RSN1) 4 1 1 purpose / 9 2 II NNDDVER FAILURE ( IiSSMAI') 4 0 8 . 412 return upon tinerexpiry 4 9 6 HANIH)VER FAILURE(RIL:I-RR) 4 1 2 idle slot s e e also TACHIF NDDAERPERFORMED (RSSNIAP) 4 1 2 use for measurements 3 3 4 )\ ERREQuEsT (itssmAil 4 0 2 , 403 !MEI. International Mobile Equipment !IA DoVI,R REQUEST ACKNOWLEIXiE Identity (ItSSMAPI 4 0 4 . 4 0 8 structure 5 8 8 HANDoVER REQUIRED (1iSSNIAP) 4 0 0 , 4 0 8 checking 5 8 9 \ D o v ERREQUIRED REJECT (IISSMAP) bite. black and grey lists 5 9 1 409 IMNIEDIATEASSIGN CONINIANDIRSNO 3 7 6 I I DIX% High level Data Link Control 2 6 9 immediate assignment s e e history initial channel assignment of public radiocommunications 2 4 txtxtittATE AssiGN 'ENT (RIL3-RR) 3 6 8 oc(iSM ' S 5(16 IMMEDIATE ASSIGNMENT EXTENDED (R1L3I II.C. High Layer Compatibility RR) 3 6 8 , 3 7 6 I II.R. Home Location Register IMMEDIATE ASSIGNMENTREJECT (RIL3-RR) architectural description 371, 376 location updating status 4 4 5 !NISI. International Mobile Subscriber role for location Management 459.40'I Identity derived from the I MSI introduction 4 6 8 restoration of the database request on the radio path 4 9 2 derived from the MSISDN 4 . 4 : 7 y ; correspondence with NISISDNs 5 2 1 routing information recovery 5 2 1 !NISI attach/detach interrogation for call routing 5 4 3 purpose of the procedure 4 7 4 subscription management 5 7 / option. 4 7 5 dimensioning 6 2 5 forceful detach by the network 4 7 6 1101.D impact on call forwarding 5 2 7 description of the call hold service 6 2 378, 380, 475 oast DETACH IR11.3-NIND impact on call control 5 procedure 5 1 : ) , 7 , in-call nmdification proce K 506, 54.7 introduction 644 K c (ciphering key) inheritance (TMN concept) introduction 481 initial channel assignment computation and storage 482 definition 318 input in A5 radio channel type 482 32) length 349 484 initial timing advance exchange of triplets 374 489 description or the procedure 375 Ki double allocation authentication mechanism initial message 478 role 3 6 8 . 373 use to compute Kc 483 storage piggybacking on SABM 3 7 6 485 types and contents 3 7 8 L reception by the MSC 3 7 9 in generic connection establishment L2R-BOP. Layer 2 Relay - Bit Oriented 494 Protocol 1 8 2 INSERTSUBSCRIBER DATA (NIAP/D) L2R-COP, Layer 2 Relay - Character 470,489 Oriented Protocol / interface LAI GSM external interfaces 8 4 sending on the BCCH 4 2 5 concept 1 0 9 in combination with the TMSI 4 8 9 list of interfaces supporting LAPD. Link Access Protocol for the D signalling needs 2 6 3 channel interference error detection 2 7 2 impacts on cellular planning 5 9 9 acknowledgement mechanism 2 7 3 interferer diversity 2 2 / frame numbering 2 7 5 interleaving role of multiplexing 2 7 7 introduction 2 3 0 Bow control 2 8 0 general principles 2 3 8 frame types 2 8 0 thr data 2 3 9 LAPDm. Link Access Protocol for the Dm for all transmission modes 2 4 6 channel for full-rate speech 2 4 7 frame lengths 2 7 0 INTERROGATE SS (MAP/I) 5 5 5 segmentation and reassembly illIerSyMb01 interference mechanism 2 7 2 description 2 5 2 acknowledgement mechanism 2 7 3 need for equalisation 2 5 7 frame numbering 2 7 5 interworking multiplexing 2 7 8 protocol interworking 2 6 6 frame types 2 8 0 • f o r call control signalling 5 2 9 piggybacking 3 7 8 Interworking function s e e IWF change of channel 3 9 0 INVOKE SS (MAP/i) 5 5 5 limited service ISDN. Integrated Services Digital Network definition 4 3 5 influence on GSM 3 5 applicability 4 4 6 ISDN-based services 5 1 PLMN selection 4 5 0 access configurations for data services link layer /40 general role 2 6 2 529 ISUP. ISDN User Part GSM link layer protocols 2 6 8 588 ITA, Interim Type Approval segmentation and reassembly 2 7 0 IWF. InterWorking Function link quality monitoring 2 7 2 architectural description /02 acknowledged mode versus nontransmission role /33 acknowledged mode 2 7 3 538 with ISDN for T connections multiplexing 2 7 7 flow control 2 7 9 LLC. Low Layer Compatibility 5 0 6 location area definition use for subscription checking range impact on cell boundary storage in the SIM )1 ins 45 438 443 455 464 469, 491 443 444 466 niaimenance role 7 3 aims and techniques 5 7 8 c.11111111Thicat11)11 needs I \I( . Mobile Allocation Index Oftset introduction 2 2 4 hopping sequences 2 2 4 description 3 6 0 MAP, Mobile Application Part architectural description / / ' ) summary of MAP protocols / 2 1 role of TCAP 2 9 7 naming of messages 2 9 8 7 mArdm 4 6 3 NiApA. introduction 5 2 8 . 5 5 7 message list 6 9 3 NIAP/D kw location management 4 6 5 for security management 4 5 6 for call control 5 2 9 message list 6 9 3 MAP/E networking function 2 9 0 RR functions 3 6 6 message list 6 9 3 cintuk !MEI Ni..‘1)A-; introduction 4 St-.\n \ \ r i l t s \IIASIM:\ \ REPDICI(1211.3-NR) 5q/ 4 6 9 4 2 / cells 3 3 0 3 / 3 3 3 cells reported by a mobile station 336 337 role II the BSIC 340 measurement period synchronisation ()I' MS and WIN measurements 3 4 0 impact of DTX .4(?,422 71/ periodicit) on the SACCH o f IICIs\ Olk b e h a v i o u r mediation functions 6 3 4 midamble 2 . 3 3 MM. Mobility Management general nuroduction 4 4 introduction of the NINI layer / 1 2 role of the MAI layer 1 1 4 . 4 3 3 protocol architecture 4 6 4 protocols f o r location management 465 protocols for securit) management 487 MM-connection. Mobility Management connection modelling concept 4 9 7 MNC. Mobile Network Code IMSI structure 4 6 9 immix a c trim. 3 6 0 mobile originating call establishment procedure 5 3 0 mobile station architectural description 8 9 functional split for data services 92. 149 equipment capabilities (classmark) 379 display of PLMNs 4 4 8 . 4 5 1 originating call establishment 5 3 2 dialled number format 5 3 6 access to supplementary services 5 5 4 5 8 5 mobile station classnlark s e e clossmark 2 5/0 539 see MM MODE MODIFY ACKNOWLEDGE (RSA1) 3 8 7 !node modify procedure 3 MODE MODIFY REQUEST (RSM) 8 6 386.395 modem position for PSTN access audio modem types control signals in V.I10 control signals in c o n n e c t i o n s 3 neighbour cells measurements 3 3 / beacon frequencies or surrounding management 5 5 2 1 mobile terminating ca. interrogation point establishment procedure mobility management measoremeni • mobil.: station reporting NI 9 correspondence e i l h \ ISISDN 472 495 378. 379, -M'. 492 3 6 use M r h a n d m or d e c i s i o n t u I s t i l t \ I P I M !NG R E Q I Ts t I a l l . 3 - s n i 6 4 469 " 460 management aspects MAP/H 304 role mawAar) )0 mai MESSAGE 3 0 5 . 562 MAP/I 300 modelling 553 description 694 message list MCC. Mobile Country Code I M S I structure 1.1) anon ululating procedure role ',incomes tequirements . ,reiccionperiodic location updating belltre any CM-transaction , r t . es t kn,vresai 683 iNutA 1,.1.11 .1).\I I:NI 1)82 control signals in N T connections (RIO-CC MODIFY COMPLETE (RIL3-CC) /35 /37 /72 /78 /80 547 548 modulation introduction 230 description 249 to 258 general principles 250 intersymbol interference 252 modulator implementation 255 role of the demodulator 256 Viterbi demodulation 257 Moll. Memorandum of Understanding history 32 signatories 33 MSRN, Mobile Station Roaming Number definition 52/ allocation scenarios 522 provision at call set-up 523 call per call transfer 545 MT, Mobile Termination /49 MTP. Message Transfer Part introduction 2 6 8 maximum length of frames 2 7 0 error detection 2 7 2 acknowledgement mechanism 2 7 4 frame n u m b e r i n g 2 7 5 rims control 2 8 0 functions of the network layer 2 8 8 addressing scheme 2 9 4 multiframe 26 TDMA frame multiframe 215 51 TDMA frame multiframe 2 / 6 multipath Rayleigh fading 2 1 9 demodulation / 5 6 propagation 5 9 6 multiple access / 9 5 to 227 multiple calls description of MPTy 66 N NI P Ty. M u l t i p a r t y service description 6 6 national roaming 5 / definition 4 3 9 mobile SICM011 cell selection principles 4 4 5 MS POWER CONTROL IRSM) 4 2 / MSCNLR handling 4 6 8 MSC. Mobile services Switching Centre NCC. Network Colour Code 3 3 8 architectural description / 0 0 NDC. National Destination Code 5 1 / terrestrial channel assignment 3 2 2 neighbour cell terrestrial circuit allocation 3 2 3 measurement need 3 2 9 handover 3 6 3 measurements 3 3 / call r e - e s t a b l i s h m e n t 4 / 4 NIS procedure s e 5 e role for location management 4 6 5 role for communication management 529 dimensioning 6 2 5 MSC/VLR restoration of the database - 4 7 2 % I S D N . Mobile Station ISDN number definition 5 / 0 structure 5 1 / link to the IILR 5 / 1 . 5 / 2 correspondence with IMSI 5 2 / multiple numbering scheme 5 2 4 for routing short messages 5 6 0 allocation at subscription 5 7 0 NEIGH/W(7? CELLS' DESCRIPTION 4 2 6 NET, Norme Europeenne de• TelOcommunications NET II/ ' 5 8 7 network change control 5 9 3 , 628 to 632 network independent clocking s e e N1C network layer general role 2 6 2 relaying requirements in GSM 2 6 4 network management introduction of the functions 7 2 tasks 59/ NIC. Network Independent Clocking concept / 3 8 V. 110 mechanisms 1 7 4 NMC. Network Management Centre 6 3 5 non-transparent mode of transmission see NT name, 1101/S NoTE INTERNAL H A N I X W E R (MAP/E) Na YI 11 MS PRESENT ( MAP/D) N o I ITY (1411.3- C C ) 5 5 4 / 2 5 6 2 0 , 55/ NSS. Network and Switching Sub-System architectural description / 0 0 configuraiiim 6 2 6 NT connections - principles / 6 8 NT connection types / 7 V —LL:111.11.11i.ssion o r ;MX i t i a t ) J79 0 OA('SU, Off Air Call Set Up definition 3 5 5 impact On call control 5 3 7 OAM role of the OAM layer / observation lot a network) 5 9 3 . 627 Ode an die Freude ()MC. Operation and Maintenance Centre /06,635 operation role of network operation operator 567 tasks OSS. Operation Sub-System 105 architectural description 264 signalling needs 634 structure 42H .0A ( B S S M A P ) 420 OVERLOAD IRSM) P packet s e e also P.SRON. 8.25. X..0 GSM services for access to PSPDNs GSM interconnection configurations with a PSPDN / 4 2 basic packet access / 4 5 /46 dedicated packet access PAD. Packet Assembler Disassembles PAD access services 5 3 configurations for accessing a PSPDN through a l'A D / 4 3 PAGC11. Paging and Access Grant CHannel purpose different types 2 ( 1 7 nine organ kali( n 2 0 7 scheduling of mien', .nessages 3 2 0 organisation 3 . C . 426 immediate assignment message 3 6 8 AGC11 and PC11 pan Ulan 3 7 5 grouping of assignment messages 3 7 6 compatibility w ills cell selection 4 2 5 ca n a c i t > 6 / 8 paging seen as an RR function 3 1 8 impact of DR X 3 / ' ) to 320 \oh3 2 0 Paging calmcil t h e PAGCI 35/.6/9 3 7 9 382 383 426 382 382 collision with initial access repetition page mode PA C I C I 1 o r g a n i s a t i o n PAGING 11355\t:\19 PAGING COMNIAND IRSNI) PAGING REQUEST TYPE I T O 3 Otil.3-RR) 3 8 2 PAGING RESPONSE IRIL3-RR I 3 7 8 password for barring services 5 0 9 path loss role for handover preparation 3 2 9 PCII. Paging CHannel s e e PAGCH PD. Protocol Discriminator role 2 8 on the radio interface use for SS 5 peak hour (traffic) 5 PEREoRNI ( 4 2 8 5 5 3 9 4 NIAP/E) 4 0 3 r u z i r o m i St • IISEQU ENT }LANDOVER (NIAP/E) 401 2 peritxlic location updating phase 1 vs. phase 2 history 3 revision level 3 4 8 0 pi I YSICAL (RSSI) 4 2 1 PHYSICAL c o N T E N T REQUEST (RSM) 4 2 1 C O N F I R M 7 PLAN, Public Land Mr ' e Network introduction 39 concept 436 home PLMN vs. visited PLMN 437 commercial names 448 preferred PLMNs list 449 forbidden PLMNs list 447,464 t t \ PERMUTED 339.428 l'I.NIN screening 338 PI .MN selection needs 4 4 0 algorithms 4 4 6 to 452 111E111U:1i vs. automatic mode 4 4 7 plus + key for international numbers 5 3 6 power class 3 8 0 power control impact on measurements 3 4 2 control requirements for uplink and downlink 3 4 3 range and steps 3 4 3 initial power value 3 4 4 SACCH control mechanism 4 2 0 Abis interface procedures 4 2 1 impact on spectral efficiency 6 0 6 pre-processing (of measurements) 364.42/ pre-synchronisation common channels cycles 2 0 6 introduction 3 3 3 performance 6 2 / pre-emption fast associated signalling mechanism see stealing of dedicated channels 3 5 8 PREPROCESS CONFIGURE (RSNI) 4 2 PREPROCESSED MEASUREMENT RESULT ( RSM) 4 2 / 1 PROCESS ACCESS SIGNALLING ( M A N E ) 2 9 1 7 PHYSICAL INFORMATION (RIL3-RR) 410,411 piggybacking • on the LAPDin SA13N1/UA frames 276.376 in R U ' • 2 8 2 SCC'P connection establishment 379 at SCCP connection release 4 1 5 PIN code. Personal Identity Number code SIM blocking 5 7 2 PIN Unblocking Key 5 7 2 PROCESS UNSTRUCTURED SS DATA ( M A P / I ) 556 538 543,547 PROGRESS (RIL3-CC) PROGRESS propagation Rayleigh fading 2 1 9 multipath 5 9 6 power loss 5 9 6 protocol protocol versus interface distinction 6 2 3 6 PROVIDE ROAMING NI:MBER ( M A P / D ) 6 5 4 5 PSPDN. Packet Switched Public Data Network PSPDN-based services 5 3 access configurat ions . / 4 2 PSTN, Public Switched Telephone Network PSTN-based services 5 0 access configurations for data services 135 use for accessing a PSPDN from GSM /42 multiple numbering scheme 524 PUK, PIN Unblocking Key 572 punctured convolutional codes principles 242 example 243 Q Q3 (GSM Q3 protocol) queuing applicability for handover role, advantages and drawbacks QUEUING INDICATION (BSSMAP) 640 357 357 386 R RAO RA I RAI' -RA2 RACH, Random Access CHannel purpose different types time organisation access messages random access scheme throughput control of the repetitions WAIT INDICATION /09 multiplicity 2 protocol interworkine RR protocol architecture 3 6 5 location management protocols 4 6 4 security management protocols 4 8 7 CC protocol architecture 5 2 8 SS protocol architecture 5 5 3 SMS protocol architecture 5 5 7 protocol discriminator s e e PD use at call re-establishment load control parameters maximum transmission power capacity 172 175 178 177 193 208 208 368 369 369 370 376 414 427 453 6/8 687 ue.A radio channel for signalling common radio channels dedicated radio channels Lonfigurations in a cell transmission modes cell channel configuration description frequency charaderislies 192 192 2/2, 22.c 321 35a 3t1t, ROM link failure radii • link protocol k 'h 4/: Nee k 1 .11 _ _ l e i ' RA l r, A //Mtn/ / role 19/ radio link monitoring RAND. RANDom number anthem icat ion mechanism length authentication procedure c \ change of Iriplets 417 47s 479 48.s 489 /9.4 RACH collisions and capture 399 repetitions 369 access m e s s a g e range t o r c e l l s ) t;tie adoptation i Mermediale rates inside G S M 70 for T connections / 71 Mr NT connecions / 72 RAo 175 ISDN scheme 175 177 It A2 /78 14A I' 183 roles ()I' the l3TS and the TRAI. 219 l(;i Wig!) fading re-establishment see call re-establishment R e c o m m e n d a t i o n s s e e Sbectlicati5,, redundancy a equipment regional subscription 4 , 4 s impact on cell selection 4 4 5 ;1st ER CI IARGINO INFORMATION I MAPIC 551, 574, 5 -n kFt iis'l ER PASSWORD M A P / I ) R H EASE (RIL3-CC) 415 410 545 534, 546 KI.I.I..\SE U O N I P L I M 1 I R 11 . 3 - c 0 5 3 4 . 5 4 0 . 5 4 9 RELFAsE R.,;DicATioN I t t s w , RINET (13SSMAP) x Hi l lissNIAN RESET I I A 7 2 I liSSNIAP) , 4 7 6 6 3 3 3 7 7 au.sri*C11411 A C K N O W L I A X I E R S S M A P ) 637 419 419 HEN( it K I T : V i i i )N ( I i s s M A N RESol'IWE 14E1)1 I..ST I ItSSMAPI restoration 471 492 550 o f l o c a t i o n databases 1(1.1 TNISIs it1:iku.‘1. nth 3-11 W I R D \ ' P. . \ ( ' K \ U N I ! D U E I I t l l . 3- C C ) cruse laCI0/ 5 le isiim Ica el R E 9 9 , 3 ('I IANNEr. R H I. \ SE (RSNI I 5 5 0 601. 602, 61/ 8 0 4 1 6 141, CI IANNItl. RH I s SE ACKNOWLEDGE. (RSM) r a n d o m access mails1m SS (MAP/11 release of an RR-session of a radio channel of a communication 4 HESE' : \ ( ' K \ ( I ell fi IL019,49,(Wlet: m a n a g e m e n t ttEsF.T (NiAP/D) 4/6 R I D -CC 528 introduction 694 message list RIL3-NINI 465 for location management 486 for security management 694 message list RIL3-RR. Radio Interface Layer 3 Radio Resource management protocol .366 introduction 694 message list RLP. Radio Link Protocol 169 role in the NT connections 180 frame delimitation /81 link protocol conversion 266 as a link layer protocol 281 chai actedsties of the protocol roaming 46 definiiion mid principles 70 SINI-roaming 436 requirements 439 international roaming 439 national roaming 515 impact on call charges ROSE. Remote Operations Service 641 Element 304 k r - A c k (SM-Ill') 304 HP- D ATA (SNI-RP 1 304 RP-ERROR ISM- R N RR. Radio Resource management 43 general funclions 111 introduction of the RR layer 114 role of the RR layer 310 introduction of the functions 3/3 RR-session 314 RR-connection 365 protocol architecture RR-session definition 313 release 415 RSM. Radio Subsystem Management introduction as an RR protocol 366 695 message list RUN GSM ALGORITHM (SIM-ME) 488 S S burst 2 3 6 SAfiM. Set Asynchronous Balanced Mode role of the link layer SA BM frame 276 use ai initial channel assignment 376 SACCH, Slow Associated Control CHannel purpose /90 cycle of the TACH/F /99 cycle of the TACH/H 203 cycle of the TACH/8 204 rationale for a separate channel 331 use for measurement reporting 33/ role for link monitoring 417 roles 420 SYSTEM INFORMATION TYPE 5 A N D 6 472 423 use to carry short messages SACCII FILLING (RSSII 4 ' 3 , 6 3 9 S.API. Service Access Point Identifier Lin radio channels 278 use in LAPDm 278 use on the Abis interface 279 for short messages 559 SCCP. Signalling Connection Control Part use on the A interface 287, 289 addressing scheme 294 global title 294 SCCP gateway 295 use for MAP 296 SCH. Synchronisation CHannel purpose 192 time organisation 206 role for synchronisation 214 contents of a burst 236 SDCCH. Standalone Dedicated Control CHannel s e e TCH/8,TACHi8 SDR. Special Drawing Right 577 security functions description 7/.477 as part of the MM layer 434 protocols 487 protection against SIM theft 572 S F \ D END SIGNAL (MAP/E) 41/ AR:smut:Rs (MAPID) 473.489 . )492 ) (/G T M R A P D N E S SEND ROUTING INFO FOR SHORT MESSAGE ( M A P / C ) 5 6 IG T U O R D N E S F M IO T A N (MAP/C)544 service providers 5 6 9 services classification of user services 4 7 speech services 4 9 PSTN-based data services 5 0 ISDN-based services 5 / between two GSM users 5 2 PSPDN-based services 5 3 CSPDN-based service. 5 / 1 supplementary services 6 0 impact of mobility 4 3 5 description for a call 5 0 5 single vs. multiple MSISDN 5 2 3 SET MESSAGE WAITING DATA (NIAP/C) 5 6 2 SETUP (RIL3-CC) 5 3 3 , 5 3 5 . 540, 541,542, 553 SFH. Slow Frequency Hopping principles 2 / 9 hopping sequences 2 2 3 , 6 MAIO and HSN impact on measurements 2 2 4 3342 0 impact on spectral efficiency 6 short message services s e e .5.1,c)9 .S. SID frame. Silence Descriptor frame generation of comfort noise / 6 2 sending requirement siavi tt 532. ..4 5?-/1/ signalling types of radio channels 9 ( interfaces 2 / 6 . ? ) BSS connection identifiers 2 9 2 signalling transfer in the BSS 2 9 2 signalling only transmission mode 322 SIM, Subscriber Identity Module introduction 6 7 formats 6 8 concept of SIM-roaming 7 0 role in security functions 7 / architectural description 9 3 role for location management 4 4 4 forbidden PLMNs list 4 4 7 preferred 11..MNs list 4 4 9 list of beacon frequencies 4 5 7 storage of location information 4 storage of Kc 4 8 2 authentication computation 4 8 5 storage of Ki 4 8 5 storage of CKSN 4 5 storage of short messages 5 management at subscription . . 6 t : Jot NNI -LP. Short Message Control Protocol description 3 0 . ; 303 (11.11A I A C I L E R R I )1( 3 1 . ? / ( ) j . ? SAl RP. Short Message Relay Protocol descripl inn 3 1 1 ; 1W-1/A-1 A RP- U M IN 3 3 0 4 0 1 SNI-SC. Short Message Service Centre m i t t u r a l descripi i n / 557 connect ion to CiSNI 5514P, Short Message Transport Protocol 301 introduclion .557 description 559 sMS-SUBMIT 561) SNIS-DELIVER S\710, Special Mobile Group structure of the committee 3 2 SNIS, Short Message Services list of services role of the SMS management sub- SPC. Signalling Point . use Gtr NITP addresomg 51ruct urc or 'mei-minimal SPUs Spedlie:Mons history phase I l CISIls phase 2 for . e r e ices M i d l e a t u r e s lor d e s c i ipt ions for data transmission lily speech coding for radio transit] ission for link layer protocols for networking protocols F o r R R 111 ' 0 1 0 0 1 k 294 290 30 37 75 /2/ 184 185 258 305 306 429 498 498 564 For location management for security inanagemeni for call control for supplementary services 564 management 565 for short message services 646 for maintenance 647 for opera' ion spectral efficiency methods of improvement 1 / n00/46 tio lu a v e layer / 4 : I (LRCM.2 n sth d a ro lb ce impel of handover 6 0 5 impact of power control transfer on L Am P D 2 7 8 , ( 6 : () i6im IN93 f'Y to c a p message length impact of frequency hopping 0 role of MAP/H 3 0 4 role of the MSC/VIA< for transporting/fH speech list of services 4 9 short messages end-to-end Iransinission 1 2 8 to 132 mobile station capabiIils 3 \ ' / speech signal represcntillkm. / 2 9 communication management functions 64 khit/s representat hut / 3 / 557 transmission inside (15.51 / 5 4 to /66 concept of SMS-gatew shy a OSNI specific representation? protocol architecture ice /55 mobile originating 560 full-rate algorithm 1 5 6 mobile terminating structure of the coder 1 5 9 service centre alerting 5 6 ' milli:lure of the decoder 1 5 9 SMS-CB, Short Message Service Cell Broadcast comfort noise. SID frames 1 6 2 service definition v l contents of a speech block between Cell Broadcast Control Channel 2 1 1 FITS and TRAU / 6 6 sms-DEuvEtt (SNI-TP) 5 n 0 channel coding 2 4 7 SMS-MO/PP. Short Message Service interleaving 2 4 7 Mobile Originating Point to Point protection of bits 2 4 7 :q) service definition SRES. Signed RESult 55o iransmission procedure authentication mechanism 4 7 8 5515-MT/PP, Short Message Service length 4 8 0 Mobile Terminating Point to Point authentication procedure 4 8 8 ,C0 service definition exchange of triplets 4 8 9 5 6 11 transmission procedure SS. Supplementary Services management 5 5 i J s\6-st role of the SS sub-lamer 1 1 5 introduction of the iv i o n s 5 0 9 FAC/LITY information m e n t 5 5 3 protocol architecture 5 5 3 basic man-machine interface 5 5 4 SS7 network. CCITT Signalling System number 7 network role / 0 / role within the NSS / 0 3 . 264 protocol stack 2 8 8 network layer protocols 2 9 4 national and international networks. 296 CCITT references 3 0 6 START °TM!' (RIL3-CC) 5 5 2 STARTDTMFACKNOWLEDGE (RIL3-CC) 5 5 2 STARTDTMFREJECT (RIL3-CC) 5 5 2 stealing fast associated signalling 1 9 0 stealing flag in radio bursts 1 9 0 , 238 STOPDTMF(RIL3-CC) 5 5 2 STOPDTMFACKNOWLEDGE (RIL3-CC ) 5 5 2 STP. Signalling Transfer Point 6 2 6 subscription geographical area 4 3 5 home PLMN 4 3 7 regional subscription 4 3 7 subscription management role 7 2 , 569 architectural description / 0 6 subsequent channel assignment timing advance assessment 3 4 7 transmission made change 3 8 6 superfrante 2 / 6 supplementary services s e e also SS impact on a call 6 0 role of the SS sub-layer 1 / 5 transport of management information between MS and HLR 2 9 9 MAP/I and underlying protocols 3 0 0 synchronisation initial radio synchronisation 2 1 4 initial V.110 synchronisation 5 3 8 between cells 6 2 0 synchronous synchronous data flow / 3 6 handover 3 4 9 , 6 2 1 SYSTENIINFORMATION TYPE I "F04 (Rit.3RIO 4 2 2 , 4 2 8 , 639 SYSTEM INFORMATION TYPE 2BIS (RIL3-RR) -128 SYS')EN! INFORMATION TYPE 5 AND 6 1kii.3Kk 4 2 2 . 6 3 9 system simulator 5 8 7 T T connections principles 1 6(ii error performances modified V. I I 0 scheme 1 7 7 T connection types / 7 7 TAC. Type Approval Code "MACH. Traffic and Associated CHannels pools of TACH/F and /8 352 TACH/8 definition 204 different types 204 time organisation 204 to 206 neighbour cells measurements 335 allocation strategies 355 phase I handover restrictions 396 dimensioning 6/7 TACH/F definition /98 time organisation 1 9 8 to 202 neighbour cells measurements 3 ) 1 4 ) dimensioning ( ( . TACH/H definition 2 0 2 time organisation 2 0 2 TAF, Terminal Adaptation Function transmission role 1 3 3 rate adaptation / 7 9 TBR. Technical Basis for Regulation 5 8 6 TCAP, Transaction Capabilities Application Part transport of MAP messages 2 9 ( transaction sub-layer 2 9 7 ) component sub-layer TCH. Traffic CHannel purpose and types TCH/8, Traffic CHannel at Eighth rate 2 : .see also TACIP8 purpose / 9 / signalling capacity 2 7 5 TCH/F. Traffic CHannel at Full rate see also TACYNF purpose 189 transmission modes 228 signalling capacity 276 sending requirements 34/ TCH/H, Traffic CHannel at Half rate see also TACT/%11 purpose - - / 8 9 transmission modes 2 2 8 TDMA, Time Division Multiple Access 215 ,„. \ I, \ ( I t 1 1 - \II 4( It I I \1St REALLOCATION COMPLEIT.(R11.3-‘1\ I I I \ ISI l t k A l I .(1(' A I I( )N, TH. Terminal Equipment Identity role on the Abis interface teleservices list of teleservices 5/16 description terminal connection to a mobile station 1 - 1 9 terrestrial channel assignment management tests for quality control 7 • r r r n y r e approval Transaction Identifier role 2 S . 5 handling of multiple calls lime slot IUc defin it ion numbering / 9 7 TN timing advance purpose 2 0 1 role for handoyer preparation . : 2 9 impact on cell size requirements estimation of the access delay .),'11(211r0i101.is VS. ilSyllt2hroBoils handoyer control mechanism • 1 2 i t accuracy 6 2 1 TMN. Telecommunication Management Network / I Y, concept structure 63() reference points 041 stack of application protocols Temporary Mobile Subscriber Identity role within a location area allocation at location updating as a short identity cancellation storage in the network in conjunction with roaming W A / TN. Timeslot Numbe definition 1 9 8 definition of TN 0 2 0 7 choice for common channels 2 1 1 acquisition at synchronisation 2 1 4 loll ticket 5 7 3 contents 5 7 6 tones and announcements 5 , 3 / . 5 3 8 TRAcF. iNvocATIoN Mimi AP) 6 3 8 trace procedure 6 3 7 TRAcE st ;BSCRIBER ACTIVI1V (MA1111 6 3 8 traffic model 6 / 3 training cequenceintroduction 2 3 / for normal bursts 2 3 3 for access bursts 2 3 5 for S bursts on the SCII 2 3 7 transaction identifier s e e T/ transcoder . v c ' e TRAU transcoding speech transcoding in GSM / 5 4 t rails le nted account procedure 5 7 3 transmission architectural description of transmission functions 1 1 3 modelling principles / 2 6 transmission mode examples for speech / 2 9 data connection types / 6 7 on each radio channel type 3 2 1 RR-connection characteristics 3 2 / change triggered by the MSC 3 8 5 In a n d 'FRAU reconfiguration 3 8 6 mode modify procedure 3 8 6 mobile station reconfiguration 3 8 7 transmission power use for handover decision 3 3 0 initial value 3 4 4 power classes 3 8 / transparent mode of transmission see T connections use of the term in GSM 2 6 5 TRAU. Transcoder and Rate Adaptor Unit presentation 1 5 0 positioning / 5 / to 153 time alignment of speech frames / 6 3 contents of a speech block•between BTS and TRAU 1 6 6 UA. Unnumbered Acknowledge role of the UA frame 276 Ul. Unnumbered Information role of the UI frame 277 uNtiLock (BSSMAP) 637 UNBLOCKINGACKNOWLEIXIF (FISSMAP) 637 I 1.1).vn: LOCATION (MAP/DI 470 uplink 193 UPT. Universal Personal Telecommunications 70 U rate adaptation / 7 9 contents of a data block between BTS and TRAU / 8 4 terrestrial channel management with a remote TRAU 3 2 4 change of transmission mode 3 8 6 configuration for downlink DTX 3 9 5 trombone definition 5 / 7 l'or international calls to mobiles 5 / 9 TRX. Transmitter/Receiver definition 2 1 2 number of TRXs vs. frequencies in a cell 2 2 7 addressing on the Abis interface 2 8 6 TUP. Telephone User Part 5 2 9 type approval a controversial issue 7 4 goals 585 CTRs 5 8 6 IMEI allocation 5 8 8 X X X w w W v V V V ”41,1-/, 404,408 HANDOVER REQUIRED 400,408 HANDOVER REQUIRED REJECT 409 OVERLOAD 420 PAGING 382 QUEUING INDICATION 386 RESET 637 RESET ACKNOWLEDGE 637 637 RESET CIRCUIT RESET CIRCUIT ACKNOWLEDGE 637 RESOURCE INDICATION 419 RESOURCE REQUEST 4/9 TRACE INVOCATION 638 UNBLOCK 637 UNBLOCKING ACKNOWLEDGE 637 IIANDOVER REQUEST 4 0 2 , 4 0 3 HANDOVER REQUEST ACKNOWLEDGE ASSIGNMENT COMPLETE 385 ASSIGNMENT FAILURE 385 ASSIGNMENT REQUEST 385,395 BLOCK 637 BLOCKING ACKNOWLEDGE 637 CANDIDATE RESPONSE 4/9 CIPHER MODE COMMAND 393 CIPHER MODE COMPLETE 394 CLASSMARK UPDATE. 38/ CLEAR COMMAND 411,415,418 CLEAR COMPLETE 415,418 CLEAR REQUEST 418 COMPLETE LAYER 3 INFORMATION 379 HANDOVER CANDIDATE ENQUIRY 419 HANDOVER COMMAND 408 41/ HANDOVER COMPLETE HANDOVER DETECT 411 IIANDOVER FAILURE 408,412 HANDOVER PERFORMED 412 BSSMAP, BSS MAnagement Part M M M The following inde.v includes the messag description of the applicative protocols. This do specified for these protocols. The main exceptio protocol errors and the messages not used in phas MESSAGE I 1%1i:33/A011 I N L nk \RI) y ink I \ ( , 1 SI N I ) l' \It.1 \II I I . R S ('/111 K1\11.1 415. 494, 496 496 4/4, 4 0 . HOLD 5 5 5 4 . 5 ; 7 6 8 0 55/ 5 2 5 2 552 5 5 51111'1)10F ACKNOWLEIXil, l L A ( ' K N O W I . F. I X ; F. 552 r SART III slop DT \ sT 3 5 5 8 534. 546 comPLETE 5 3 4 . 540. 546 kyrkir.vr, 5 5 0 RFTRir.vr. AcKNowi.mcd 5 5 1 ) 533.535.540.54/.542.553 STARTHEMP 5 5 2 5 0 IN( x;kr.ss m yr !FY 4 5 5 . \111011..1 ('OMIT./'/'(, moDiry HOLD ACKNOWLEDGE. 5 I.\ SII•I L I ' (..\ A. PRucki.:Dimi 5 3 - 1 . 535. 53x c().\; r(-1 5 3 5 . 5 4 2 . 5 4 3 , 5 5 1 \ ,\('i; 5 3 5 Ins( H 5 3 4 . 545 5 5 3 4 , 538. 542. 5-13. :)53 VI.I CUNIARNIED ALLR v i It L 3 -CC, Radio Interface Layer 3 Call Control protocol REGISTER S S 556 555 T051 It EA LLOC'ATION COMPLETE s 556 ia.Gis I ER PASSWORD s pkockss t ) 555 555 IN l'ITItO)(ATE SS i TNI51RFALLOCATION COMMAND 556 (1E1 PASSWORD r LOCATION UPDATING REQUEST 3 7 8 , 556 I I I 0 \ \ k D SS N o 111ICA1 k LOLA I ION UPDATING REJECT 556 M R \V AltD CHECK SS INDIC.VFION INv()KI., S S INIS1 555 ERASE S S 9 9 3 ASSIGNMENT FAILURE 0 9 , 412 0 4 1 6 4 / 3 8 2 / 0 9 4 3 4 4 / / 2 1 MEASERENIENT REPORT 4 3 7 1 8 3 8 2 2 422,639 syti MINI INFORMATION TYPE 5 A N D 6 s1 511 51 INFORMATION TYPE 2 B I S 4 2 8 422,428,639 SYSTt NI INFORMATION TYPE I T O 4 IlIytilL,51. INFORMATION 4 1 0 , 4 1 1 ;INc;RESPONSE PAGiNc; REQUEST TYPE. I T O 3 376 IMNII M AT E ASSIGNMENT REJECT 3 7 / , 368,376 IMNIEDIVIE ASSIGNMENT 3 6 8 INI \ IFI )IATE ASSIGNMENT EXTENDED HANDOVER FAILURE II:\NDOVERC'OMPLETE 391. 399. 404. 407, 409 4 HANDOVER ACCESS IIANDOVER COMMAND 3 I T Ey( ENCY REDEFINITION I. \ SSMARK CHANGE MODE COMPLETE 393,394,494 ILK] NG MODE CONIMAND II.\ R E Q U E S T 3 6 8 . 372, 375 II \ \ N E I RELEASE (1 [ANNE'. MODE MODIFY 3 8 7 I:\\NE:I. MODE MODIFY ACKNOWLEDGE 388 3 ASSIGNMENT COMPLETE RIL3-RR, Radio Interface Layer 3 Radio Resource management protocol AssiGN IFNTcommAND 3 9 0 , 3 9 / 4 9 / 4 9 0 379.467.492 6 3 4 4 3 3 7 / 7 1 3 9 6 3 8 7 3 7 6 9 8 2 4 I l l Y S I C A L CONTEXT REQUEST 0 2 2 2 / 1 1 4 / 6 421 4 2 3 , ROlt k1,1) I Protocol RI, A r K SCI-RI', Short Message Relay (1'-ERROR ('P- D ATA ('I'-ACK SM-CP, Short Message Control Protocol s m -CI I FILLING 304 304 304 303 303 303 416 639 Ill- ( -I IANNEL RELEASE 4 / 6 RI: CI IANNEL RELEASE ACKNOWLEDGE RELEASE INDICATION PREPROCESS CONFIGURE 4 2 / PREPROCESSED MEASUREMENT RESULT 4 8 2 PI IYSICAL CONTEXT CONFIRM PAGING COMMAND 3 4 4 \ IS POWER CONTROL OVERLOAD / 417,418 404 NIODE MODIFY REQUEST 3 8 6 , 3 9 5 MODE MODIFY ACKNOWLEDGE IMMEDIATE ASSIGN COMMAND FIANDOVER DETECTION ESTABLISH INDICATION ENCRYPTION COMMAND DEI.ETE INDICATION :\ccuri- 469,491 4 DEACTIVATE SACCI I CI IANNEL REQUIRED 3 7 4 CONNECTION FAILURE INDICATION ('I IANNEL ACTIVATION 3 8 8 , 4 0 4 ('I IANNEL ACTIVATION ACKNOWLEDGE 418 BS POWER CONTROL C C C H LOAD INDICATION 639 42/ BCCH INFORMATION RSM, Radio Subsystu Management 378. 380, 475 492 I DENTFIN RESPONSE 492,589 IIE:N.I ITN REQUEST 555 378. 394, 494 555 ( ' \ l SERVICE REQl 'ES I (.\1si..0..icr.RFJEct /5, 494, 496 I'M SI V I C E AC:CIT.!' RE-ESTABLISIINIENT REQUEST 488 488 AC11\ A I L S S -162 Al 'I I IEN'OCATION RESPONSE AI 1 I IENTICATION REol EST RIL3-AIM, Radio. ' e r f a c e Layer 3 59/ M o b i l i t y managem,.,t protocol DEA('I IRATE SS NIAP/I \ Figure 3.3 - Data transmission planes Figure 3.1 - Layered approach Figure 3. 2 - Speech representations Figure 2.22 - MAP/C to MAP/I protocols 1 Figure 2.20 •- GSM signalling architecture Figure 2.21 - Stack of protocols on the SS7 interf Figure 2.18 - The functional planes of GSM Figure 2.19 - General protocol architecture of G Figure 2.16 - OSS organisation Figure 2.17 - Protocols versus Interfaces Figure 2.15 - Internal structure of the NSS Figure 2.14 - A GSM MSC (by courtesy of Matr Figure 2.13 - The external environment of the N Figure 2.12 - A GSM BSC (by courtesy of Matra Figure 2.11 - BSS components and interfaces Figure 2.10 - A DCS1800 BTS (by courtesy of N Figure 2.9 - A GSM BTS (by courtesy of Motoro Figure 2.7 - Mobile station functional architectur Figure 2.8 - The external environment of the BS Figure 2.6 - A GSM handheld mobile station (by Figure 2.5 - A GSM portable mobile station tby Figure 2.4 - GSM subsystem organisation Figure 2.3 - External interfaces of GSM Figure 2.2 - The three axes of the description Figure 2.1 - Two-dimensional view of a network Figure 1.8 - PLMN interfaces Figure 1.7 - The two types of SIMs Figure 1.5 - Ways of accessing a PSPDN from G Figure 1.6 - The puzzle of supplementary service Figure 1.4 - Cellular coverage representation Figure 1.3 - European GSM MoU signatories in Figure 1 . 2 - Partial view of the ETSI organisatio Figure 1.1 - The concept of cellular coverage INDEX OF F kir 1 1 9 2 1 2 0 7 2 9 2 0 0 8 2 8 1 1 6 Figure 4.21 • Frequency/TN groups 2 2 6 1-i,..uire 4.22 Sequence of operations from speech to radio stases... and back to speech229 2 2 2 Figure 4.2(1 Nsample of interfering cells o ith Slots Frequency Hopping Fig.tire 4.19 -- Typical amplitude variations due to Rayleigh fading 9 9 5 2 1 6 2 0 0 Figure 4.16 G S M primar hand 2 1 7 l'igure 4.17 • Carriers at the border of the ( iSNI hand 2 1 I 'gun' 4.18 - Slow Frequency I lopping in the time-frequency domain 2 1.igtire 4.15 • I lierarchy ul frames Figure 4.14 • Time organisation of a Cell Broadcast ('I lannel (CBC111 1•1:iire .1.12 B a s i t r i n n i o n channel pattern 2 0 Figure 4.13 . \ common channel pattern tot small capacits cells 2 0 Figure 4.10 ' l i m e drganisation of a B c c I I :111,1 a P.\GCIITI 1 4 . 1 1 Timeirganisittion of a R:\('11/1 2 0 2 2 Figure 4.7 T i m e o'rganisation of TACII.'s 2 I•nnire 4.8 t i m e organisation of the FCC11 and SCII 0 Figure 4.0 T i m e organisation of a BCC! I ;Hid a PA( i( 1111: 2 1:Hire 4.5 - Time organisation of TACI1/11, 2 tire 4.6 -• Cycle of 16 13Ps seen hs a mobile station I•igiire 4.3 • Time organisation of a -VAC' I I 1 9 Figure 4.4 - Cycle of 8 liPs seen by a nubile station close to the base Figure 4.2 -• Choice of \ :due of a burst period Figure 4.1 - A slot in the time anti frequenQ domain 3 Figure 5.11 - Signalling transport structure in the 0 Figure 5.10 - Transport mechanism on MAP/F. 183 1 Figure 5.9 - Signalling message transport on the 179 Figure 3.22 R a t e adaptation in CISNI I iglu:y.1.2.1 T h e split beR\ { I T S and I 2 Figure 5.8 - LAPD and LAPDm frame structure 175 Figure 3.21 - Rate adaptation in ISDN: R.•\ 9 0 Figure 5.7 - Acknowledged mode release proced 174 Figure 3.20 - Rate adaptation in ISDN: the three steps Figure 6.13 - Transmission power adaptation Figure 6.14 - Procedural needs for power control Figure (1.15 -• Time offset hem een (no WI'S, Figure 6.12 - Power control steps for a class 2 G Figure 6.10 - Choice of the BS1C Figure 6.11 - A European PLMN code map 'igure 6.9 - TACH/8 cycle vs. FCCH cycle Figure 6.7 - Nleasurement intervals available at t Figure 6.8 - Sliding measurement cycle Figure 6.6 - Procedural needs for discontinuous tr Figure 6.5 Channel allocation on the terrestrial i Figure 6.3 - The concepts of 12R-session and kRFigure 6.4 - Configurations changes for an RR•s Figure 6.1 - The concepts of anchor and relay VI Figure h.2 Contents of an RR-session Figure 5.18 - Short message relaying with SM-R Figure 5.17 - The SM-CP protocol between MS a Figure 5.16 - Protocol architecture for the transp Figure 5.15 - Protocol stack for supplementary s Figure 5.14 - TCAP modelling Figure 5.13 - Global title translation Figure 5.12 - Connection identifiers in the BSS Figure 5.6 - Acknowledged mode setting proced 165 173 Figurk.• 3.17 Transmission of speech frames I•igure Figure 3.19 -- The RAO -adaptation" function • f o r r.tie adaptation Figure 5.4 - Repetition mechanism Figure 5.5 - Window mechanism for acknowledg 157 A linear filter and the corresponding inverse idle, Figure 3.15 Figure 3.16 156 Figure 5.3 - Segmentation and re-assembly of me Figure 5.2 - HDLC frame flags Figure 5.1 - Relaying versus interworking 151 153 Figure 4.30 - GMSK modulation spectrum 146 148 Figure 4.29 - Effect of one hit on the modulated 144 145 Positions of the TRAI: Modelling of the 13 kbitis speech signal 1irurr 1.12 Dedicated ( iSM access to the PSPON ( SI'l )N i iters'onni•cl ion 1•igury 3.14 Mobile stations configuration, 6 Figure 4.27 - "19 burst" Interleaving scheme for 143 144 Figure 3.8 P S P D N packet access 8 Figure 4.26 - Delay of an access burst 141 Figure 3.7 Interconnection with ISDN 9 Figure 4.25 - Time mask for an access burst 139 1.i.,tire 3.5 -- An asynchronous data Ilov, Figure 3.6 - Reference configuration for fa v transmission Figure 4.28 - Flo in the GMSK modulation Figure 4.24 - Autocorrclation function of a GSM 136 I is ore 3.9 P A D access to a PSPDN inure 3,10 I S D N X.25 access 1.11 - B a s i c - packet (X.321 and PAD access in (iSM Figure 4.23 - Amplitude. ' d e of a "normal- bu 135 Fi._uire 3.4 Interconnection with the PS I'\ Routing of service information Figure 8.23 - Short message failed delivery manag Figure 8.24 - Service Centre alerting 401 403 405 406 407 Figure 6.31 - Handover requirement from the serving BSC t o the switching point Figure 6.33 - Start of path establishment at handover Figure 6.3.3 - End of path establishment at handover Figure 6.34 - Conference bridge in the BSC for handover Figure 6.35 - Sending back of the handover command 490 497 Figure 9.15 - Impact of power control and DTX on Figure 4.22 - Containment tree 511 512 of the (iMSC Figure 8.4 - The key roel Application protocols stack for i S M Figure 4.21 - Inheritance trees 510 Figure 8.3 • The two parts of a mobile terminating call route Figure 8.3 The structure of aGiSM directors number Figure 9.20 - 502 Figure 9.19 - TMN reference points Figure 9.18 - TMN categories of functions Figure 9.16 - Impact of frequency hopping on the Figure 9. 17 - A drop-and-insert configuration Figure .8 1. G.SM functional split Alternative AM modelling Figure 7.12 - 487 483 Figure 7.10 - Security management protocols Figure 7.11 - TMSI allocation • Figure 7.9 - K e computation Figure 9.13 - Impact of the handover strategy on t Figure 9.14 - Impact of power control on the C dis 481 Fieure 7.8 - Ciphering and deciphering An omnidirectional cell 479 Figure 9.11 - C/I cumulative distribution Figure 9.12 - 469 467 The propagation statistical model Figure 9.10 - Number of sites per Erlang as a func 465 Figure 9.9 - Figure 9.7 - The Equipment Identity Register (EIR Figure 9.8 - Propagation paths Figure 7.6 - The structure of an IMSI Figure 7.7- The authentication computation Figure 7.4 - Location Management Protocols Figure 7.5 - Main elementary location updating procedures 455 443 454 Figure 7.1 - Location area vs. MSC and PLMN areas Figure 7.2 - Cell boundaries according to CI Figure 7.3 - Impact of cOs. RESTARCT MSTERESAS on cell boundaries - . . _ F i g u r e 9.6 - Structure of the IMEI Figure 9.4 - Remote access to central maintenance Figure 9.3 - Transfer of billing data 417 Figure 6.39 - SACCH counter for link management 423 416 C o n t e n t s of an S A C C H block 410 Figure 6.38 - The normal release procedure of an RR-session Figure 6 . 4 0 - Figure 9.1 - Operators and Service providers Figure 9.2 - The "transferred account procedure" 409 Figure 6.36 - Handover failure at the new BSC Figure 6.37 … Access in the case of an a s v n c h r o n o u s h a n d o v e r Figure 8.21 - Protocols for supplementary services Figure 8.22 - Protocols for Short Message Transfer Figure 8.20 - Retrieving a call on hold Figure 8.18 - In-call modification procedure Figure 8.19 - Call hold 400 Figure 6.30 - Handover erecution sequence Figure 8.17 - Call release 394 398 Figure 6.29 - Position of the switching point at handover Figure 8.15 - Basic MT Call Establishment sequen Figure 8. 16 - Routing of an MT call from GMSC t Figure 8, 14 - Dialling number formats Figure 8. 13. - Mobile originating call establishment Figure 8.12 - Protocols in the CC domain Figure 8.11 - Figure X.8 - Charging principles of mobile termina Figure 8.9 - The "tromboning" effect Figure .8 10 - The provision of the MSRN Figure 8.7 - MT call involving three countries Figure 8.6 - MT call toward a roaming subscriber 392 389 FIGU Figure X.5 - Routing of an A . call within one cou I N D E X OF Figure 6.27 - Cipher mode change: the 3 steps Figure 6.28 - The cipher mode setting procedure Activation of a new channel in the BTS Figure 6.26 - 387 385 The Feare-6.25 "mode moditi Figure 6.24 - C h a n e e of the t r a n s m i s s i o n mode be the A I S 377 Initial assignment procedure C o n t e n t o n r e s o l u t o n a l m s e s D a s m e n 384 374 Figure 6.23 • Different page modes • Figure 6.21 reure 6.22 373 logare 6.20 contents of an access burst on the R . 'Il 368 Figure 6. 19 - Initial access procedure lIselal 365 Figure 6. 18 - - RR protocol architecture 356 361 Assignment strategies at call set-up GSM SYSTEM Figure 6.17 - Cell and Nobile Allocation Figure 6.16 THE E U R O P MEC,A 53110 ,w.s.,:,..es N° 2871 I c Fr E