WILEY Beyond 3G User Manual

Beyond 3G – Bringing Networks, Terminals and the Web Together
Beyond3G–BringingNetworks,TerminalsandtheWebTogether:LTE, WiMAX, IMS, 4G Devices and the Mobile Web 2.0
MartinSauter © 2009JohnWiley&Sons,Ltd. ISBN: 978-0-470-75188-6
Beyond 3G – Bringing Networks, Terminals and the Web Together
LTE, WiMAX, IMS, 4G Devices and the Mobile Web 2.0
Martin Sauter
Nortel, Germany
A John Wiley and Sons, Ltd, Publication
This edition first published 2009 # 2009 John Wiley & Sons Ltd
Registered office
John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, United Kingdom
For details of our global editorial offices, for customer services and for information about how to apply for permission to reuse the copyright material in this book please see our website at www.wiley.com.
The right of the author to be identified as the author of this work has been asserted in accordance with the Copyright, Designs and Patents Act 1988.
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by the UK Copyright, Designs and Patents Act 1988, without the prior permission of the publisher.
Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books.
Designations used by companies to distinguish their products are often claimed as trademarks. All brand names and product names used in this book are trade names, service marks, trademarks or registered trademarks of their respective owners. The publisher is not associated with any product or vendor mentioned in this book. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold on the understanding that the publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought.
Library of Congress Cataloging-in-Publication Data
Sauter, Martin.
Beyond 3G : bringing networks, terminals and the Web together / Martin Sauter.
p. cm. Includes bibliographical references and index. ISBN 978-0-470-75188-6 (cloth)
1. Wireless Internet. 2. Smartphones. 3. Mobile computing. I. Title. TK5103.4885.S38 2009
621.382—dc22
A catalogue record for this book is available from the British Library.
ISBN 978-0-470-75188-6 (H/B)
Set in 10/12pt Times by Integra Software Services Pvt. Ltd. Pondicherry, India Printed and bound in Great Britain by Antony Rowe Ltd.
2008047071
Contents
Preface xi
1 Evolution from 2G over 3G to 4G 1
1.1 First Half of the 1990s – Voice-centric Communication 1
1.2 Between 1995 and 2000: the Rise of Mobility and the Internet 2
1.3 Between 2000 and 2005: Dot Com Burst, Web 2.0, Mobile Internet 2
1.4 From 2005 to today: Global Coverage, VoIP and Mobile Broadband 4
1.5 The Future – the Need for Beyond 3G Systems 5
1.6 All Over IP 8
1.7 Summary 11 References 11
2 Beyond 3G Network Architectures 13
2.1 Overview 13
2.2 UMTS, HSPA and HSPAþ 14
2.2.1 Introduction 14
2.2.2 Network Architecture 14
2.2.3 Air Interface and Radi o Network 23
2.2.4 HSPA (HSDPA and HSUPA) 31
2.2.5 HSPAþ and other Improvements: Competition for LTE 36
2.3 LTE 45
2.3.1 Introduction 45
2.3.2 Network Architecture 46
2.3.3 Air Interface and Radi o Network 51
2.3.4 Basic Procedures 65
2.3.5 Summary and Comparison with HSPA 68
2.3.6 LTE-Advanced 69
2.4 802.16 WiMAX 70
2.4.1 Intr oduction 70
2.4.2 Net work Architecture 70
vi Contents
2.4.3 The 802.16d Air Interface and Radio Network 76
2.4.4 The 802.16e Air Interface and Radio Network 79
2.4.5 Basic Procedures 83
2.4.6 Summ ary and Comparison with HSPA and LTE 85
2.4.7 802.16 m: Complying with IMT-Advanced 86
2.4.8 802.16 j: Mobile Multihop Relay 87
2.5 802.11 Wi-Fi 88
2.5.1 Introduction 88
2.5.2 Network Architecture 89
2.5.3 The Air Interface – from 802.11b to 802.11n 92
2.5.4 Air Interface and Resource Management 97
2.5.5 Basic Procedures 101
2.5.6 Wi-Fi Security 101
2.5.7 Quality of Service: 802.11e 103
2.5.8 Summary 104
References 105
3 Network Capacity and Usage Scenarios 107
3.1 Usage in Developed Markets and Emerging Economies 107
3.2 How to Control Mobile Usage 108
3.2.1 Per Minute Charging 109
3.2.2 Volume Charging 109
3.2.3 Split Charging 109
3.2.4 Small-screen Flat Rates 109
3.2.5 Strategies to Inform Users When Their Subscribed Data
Volume is Used Up 110
3.2.6 Mobile Internet Access and Prepaid 110
3.3 Measuring Mobile Usage from a Financial Point of View 111
3.4 Cell Capacity in Downlink 112
3.5 Current and Future Frequency Bands for Cellular Wireless 117
3.6 Cell Capacity in Uplink 118
3.7 Per-user Throughput in Downlink 120
3.8 Per-user Throughput in the Uplink 125
3.9 Traffic Estimation Per User 127
3.10 Overall Wireless Network Capacity 129
3.11 Network Capacity for Train Routes, Highways and Remote Areas 133
3.12 When will GSM be Switched Off? 135
3.13 Cellular Network VoIP Capacity 136
3.14 Wi-Fi VoIP Capacity 140
3.15 Wi-Fi and Interference 141
3.16 Wi-Fi Capacity in Combination with DSL and Fibre 143
3.17 Backhaul for Wireless Networks 148
3.18 A Hybrid Cellular/Wi-Fi Network for the Future 153
References 155
Contents vii
4 Voice over Wireless 157
4.1 Circuit-switched Mobile Voice Telephony 158
4.1.1 Circuit Switching 158
4.1.2 A Voice-optimized Radio Network 159
4.1.3 The Pros of Circuit Switching 159
4.2 Packet-switched Voice Telephony 159
4.2.1 Net work and Applications are Separate in Packet-switched
Networks 160
4.2.2 Wireless Network Architecture for Transporting IP packets 160
4.2.3 Ben efits of Migrating Voice Telephony to IP 162
4.2.4 Voice Telephony Evolution and Service Integration 162
4.2.5 Voice Telephony over IP: the End of the Operator Monopoly 163
4.3 SIP Telephony over Fixed and Wireless Networks 164
4.3.1 SIP Registration 164
4.3.2 Establishing a SIP Call Between Two SIP Subscribers 167
4.3.3 Session Description 169
4.3.4 The Real-time Transfer Protocol 171
4.3.5 Establishing a SIP Call Between a SIP and a PSTN Subscriber 172
4.3.6 Proprietary Components of a SIP System 174
4.3.7 Net work Address Translation and SIP 175
4.4 Voice and Related Applications over IMS 176
4.4.1 IMS Basic Architecture 179
4.4.2 The P-CSCF 181
4.4.3 The S-CSCF and Application Servers 182
4.4.4 The I-CSCF and the HSS 184
4.4.5 Media Resource Functions 186
4.4.6 User Identities, Subscription Profiles and Filter Criteria 188
4.4.7 IMS Registration Process 190
4.4.8 IMS Session Establishment 194
4.4.9 Voice Telephony Interworking with Circuit-switched Networks 199
4.4.10 Push-to-talk, Presence and Instant Messaging 203
4.4.11 Voice Call Continuity 206
4.4.12 IMS with Wireless LAN Hotspots and Private Wi-Fi
Networks 209
4.4.13 IMS and TISPAN 213
4.4.14 IMS on the Mobile Device 216
4.4.15 Challenges for IMS Rollouts 219
4.4.16 Opportunities for IMS Rollouts 222
4.5 Voice over DSL and Cable with Femtocells 224
4.5.1 Femto cells from the Network Operator’s Point of View 226
4.5.2 Femto cells from the User’s Point of View 227
4.5.3 Conc lusion 228
4.6 Unlicensed Mobile Access and Generic Access Network 228
4.6.1 Tec hnical Background 229
4.6.2 Adva ntages, Disadvantages and Pricing Strategies 231
References 232
viii Contents
5 Evolution of Mobile Devices and Operating Systems 235
5.1 Introduction 235
5.1.1 The ARM Architecture 237
5.1.2 The x86 Architecture for Mobile Devices 238
5.1.3 Fro m Hardware to Software 238
5.2 The ARM Architecture for Voice-optimized Devices 238
5.3 The ARM Architecture for Multimedia Devices 241
5.4 The x86 Architecture for Multimedia Devices 244
5.5 Hardware Evolution 247
5.5.1 Chipset 247
5.5.2 Process Shrinking 248
5.5.3 Displays and Batteries 249
5.5.4 Oth er Additional Functionalities 250
5.6 Multimode, Multifrequency Terminals 252
5.7 Wireless Notebook Connectivity 255
5.8 Impact of Hardware Evolution on Future Data Traffic 255
5.9 The Impact of Hardware Evolution on Networks and Applications 257
5.10 Mobile Operating Systems and APIs 258
5.10.1 Java and BREW 258
5.10.2 BREW 259
5.10.3 Symbian/S60 260
5.10.4 Windows Mobile 262
5.10.5 Linux: Maemo, Android and Others 262
5.10.6 Fracturization 265
5.10.7 Operating System Tasks 265
References 271
6 Mobile Web 2.0, Applications and Owners 273
6.1 Overview 273
6.2 (Mobile) Web 1.0 – How Everything Started 274
6.3 Web 2.0 – Empowering the User 275
6.4 Web 2.0 from the User’s Point of View 275
6.4.1 Blogs 276
6.4.2 Media Sharing 277
6.4.3 Pod casting 277
6.4.4 Adva nced Search 277
6.4.5 User Recommendation 278
6.4.6 Wikis – Collective Writing 278
6.4.7 Social Networking Sites 279
6.4.8 Web Applications 280
6.4.9 Ma shups 280
6.4.10 Virtual Worlds 281
6.4.11 Long-tail Economics 281
6.5 The Ideas Behind Web 2.0 282
6.5.1 The Web as a Platform 282
6.5.2 Harne ssing Collective Intelligence 283
Contents ix
6.5.3 Data is the Next Intel Inside 284
6.5.4 End of the Software Release Cycle 284
6.5.5 Lightweight Programming Models 285
6.5.6 Software above the Level of a Single Device 285
6.5.7 Rich User Experience 285
6.6 Discovering the Fabrics of Web 2.0 286
6.6.1 Aggregation 286
6.6.2 AJAX 289
6.6.3 Tag ging and Folksonomy 290
6.6.4 Open Application Programming Interfaces 293
6.6.5 Open Source 295
6.7 Mobile Web 2.0 – Evolution and Revolution of Web 2.0 296
6.7.1 The Seven Principles of Web 2.0 in the Mobile World 296
6.7.2 Adva ntages of Connected Mobile Devices 301
6.7.3 Offline Web Applications 304
6.7.4 The Mobile Web, 2D Barcodes and Image Recognition 308
6.7.5 Wa lled Gardens, Mobile Web 2.0 and the Long Tail 310
6.7.6 Web Page Adaptation for Mobile Devices 311
6.8 (Mobile) Web 2.0 and Privacy 317
6.8.1 On- page Cookies 318
6.8.2 Inter-site Cookies 320
6.8.3 Flash Shared Objects 320
6.8.4 Site Information Sharing, Social Distribution 321
6.8.5 Session Tracking 322
6.9 Mobile Applications 322
6.9.1 Web Browsing 323
6.9.2 Audio 324
6.9.3 Media Sharing 328
6.9.4 Video and TV 330
6.9.5 Voice and Video Telephony 332
6.9.6 Widgets 333
6.9.7 Social Media 335
6.9.8 Microblogging 335
6.9.9 Loc ation 338
6.9.10 Shopping 340
6.9.11 Mobile Web Servers 341
References 343
7 Conclusion 345
Index 349
Preface
In recent years, cellular voice networks have transformed into powerful packet-switched access networks for both voice communication and Internet access. Current 3.5G net­works such as UMTS/HSDPA and CDMA 1xEvDO now deliver bandwidths of several megabits per second to individual users, and mobile access to the Internet from handheld devices and notebooks is no longer perceived as slower than a DSL or cable connection. Bandwidth and capacity demands, however, keep rising because of the increasing num­ber of people using the networks and due to new bandwidth-intensive applications such as video streaming and mobile Internet access from notebooks. Thus, network manu­facturers and network operators need to find ways to increase capacity and performance while reducing cost.
In the past, network evolution mainly involved designing access networks with more bandwidth and capacity. As we go beyond 3G network architectures, there is now also an accelerated evolution of core networks and, most importantly, user devices and applica­tions. This evolution follows the trends that are already in full swing in the ‘fixed-line’ Internet world today. Circuit-switched voice telephony is being replaced by voice over IP technologies and Web 2.0 has empowered consumers to become creators and to share their own information with a worldwide audience. In the future, wireless networks will have a major impact on this trend, as mobile phones are an ideal tool for creating and consuming content. The majority of mobile phones today have advanced camera and video capabilities, and together with fast wireless access technologies, it becomes possible to share information with others instantly.
While all these trends are already occurring, few resources are available that describe them from a technical perspective. This book therefore aims to introduce the technology behind this evolution. Chapter 1 gives an overview of how mobile networks have evolved in the past and what trends are emerging today. Chapter 2 then takes a look at radio access technologies such as LTE, HSPA+, WiMAX and the evolution of the Wi-Fi standard. Despite the many enhancements next-generation radio systems will bring, bandwidth on the air interface is still the limiting factor. Chapter 3 takes a look at the performance of next-generation systems in comparison to today’s networks, shows where the limits are and discusses how Wi-Fi can help to ensure future networks can meet the rising demand for bandwidth and integrated home networking. Voice over IP is
xii Preface
already widely used in fixed line networks today and ‘Beyond 3G’ networks have enough capacity and performance to bring about this cha nge in the wireless world as well. Chapter 4 thus focuses on Voice over IP architectures, such as the IP Multimedia Subsystem (IMS) and the Session Initiation Protocol (SIP) and discusses the impacts of these systems on future voice and multimedia communication. Just as important as wireless networks are the mobile devices using them, and Chapter 5 gives an overview of current mobile device architectures and their evolution. Finally, mobile devices are only as useful as the applications running on them. So Chapt er 6 discusses how ‘mobile Web 2.0’ applications will change the way we communicate in the future.
No book is written in isolation and many of the ideas that have gone into this manu­script are the result of countless conversations over the years with people from all across the industry. Specifically, I would like to thank Debby Maxwell, Prashant John, Kevin Wriston, Peter van den Broek and John Edwards for the many insights they have provided to me over the years in their areas of expertise and for their generous help with reviewing the manuscript. A special thank-you goes to Berenike for her love, her passion for life and for inspiring me to always go one step further. And last but not least I would like to thank Mark Hammond, Sarah Tilley, Sarah Hinton and Katharine Unwin of John Wiley and Sons for the invaluable advice they gave me throughout this project.
1
Evolution from 2G over 3G to 4G
In the past 15 years, fixed line and wireless telecommunication as well as the Internet have developed both very quickly and very slowly depending on how one looks at the domain. To set current and future developments into perspective, the first chapter of this book gives a short overview of major events that have sh aped these three sectors in the previous one-and-a-half decades. While the majority of the developm ents described below took place in most high-tech countries, local factors and national regulation delayed or accelerated events. Therefore, the time frame is split up into a number of periods and specific dates are only given for country-specific examples.
1.1 First Half of the 1990s – Voice-centric Communication
Fifteen years ago, in 1993, Internet access was not widespread and most users were either studying or working at universities or in a few select companies in the IT industry. At this time, whole universities were connected to the Internet with a data rate of 9.6 kbit/s. Users had computers at home but dial-up to the university network was not yet widely used. Distributed bulletin board networks such as the Fidonet [1] were in widespread use by the few people who were online then.
It can therefore be said that telecommunication 15 years ago was mainly voice-centric from a mass market point of view. An online telecom news magazine [2] gives a number of interesting figures on pricing around that time, when the telecom monopolies where still in place in most European countries. A 10 min ‘long-distance’ call in Germany during office hours, for example, cost E3.25.
On the wireless side, first-generation analog networks had been in place for a number of years, but their use was even more expensive and mobile devices were bulky and unaffordable except for business users. In 1992, GSM networks had been launched in a number of European countries, but only few people noticed the launch of these networks.
Beyond3G–BringingNetworks,TerminalsandtheWebTogether:LTE, WiMAX, IMS, 4G Devices and the Mobile Web 2.0
MartinSauter © 2009JohnWiley&Sons,Ltd. ISBN: 978-0-470-75188-6
2 Beyond 3G – Bringing Networks, Terminals and the Web Together
1.2 Between 1995 and 2000: the Rise of Mobility and the Internet
Around 1998, telecom monopolies came to an end in many countries in Europe. At the time, many alternative operators were preparing themselves for the end of the monopoly and prices went down significantly in the first weeks and months after the new regulation came into effect. As a result, the cost of the 10 min long-distance call quickly fell to only a fraction of the former price. This trend continues today and the current price is in the range of a few cents. Also, European and even intercontinental phone calls to many countries, like the USA and other industrialized countries, can be made at a similar cost.
At around the same time, another important milestone was reached. About 5 years after the start of GSM mobile networks, tariffs for mobile phone calls and mobile phone prices had reached a level that stimulated mass market adoption. While the use of a mobile phone was perceived as a luxury and mainly for business purposes in the first years of GSM, adoption quickly accelerated at the end of the decade and the mobile phone was quickly transformed from a high-price business device to an indispensable communication tool for most people.
Fixed line modem technology had also evolved somewhat during that time, and modems with speeds of 30–56 kbit/s were slowly being adopted by students and other computer users for Internet access either via the university or via private Internet dial-up service providers. Around this time, text-based communication also started to evolve and Web browsers appeared that could show Web pages with graphic al content. Also, e-mail leapt beyond its educational origin. Content on the Internet at the time was mostly published by big news and IT organizations and was very much a top-down distribution model, with the user mainly being a consumer of information. Today, this model is known as Web 1.0.
While voice calls over mobile networks quickly became a success, mobile Internet access was still in its infancy. At the time, GSM networks allowed data rates of 9.6 and
14.4 kbit/s over circuit-switched connections. Few people at the time made use of mobile data, however, mainly due to high costs and missing applications and devices. Nevertheless, the end of the decade saw the first mobile data applications such as Web browsers and mobile e-mail on devices such as Personal Digital Assistants (PDAs), which could communicate with mobile phones via an infrared port.
1.3 Between 2000 and 2005: Dot Com Burst, Web 2.0, Mobile Internet
Developments continued and even accelerated in all three sectors despite the dot com burst in 2001, which sent both the telecoms and the Internet industry into a downward spiral for severa l years. Despite this downturn, a number of new important developments took place during this period.
One of the major breakthroughs during this period was the rise of Internet access via Digital Subscriber Lines (DSL) and TV cable modems. These quickly replaced dial-up connections as they became affordable and offered speeds of 1 MBit/s and higher. Compared with the 56 kbit/s analog modem connections, the download times for web pages with graphical content and larger files improved significantly. At the end of this period, the majority of people in many countries had access to broadband Internet that allowed them to view more and more complex Web pages. Also, new
Evolution from 2G over 3G to 4G 3
forms of communication like Blogs and Wikis appeared, which quickly revolutio­nized the creator–consumer imbalance. Suddenly, users were no longer only consu­mers of content, but could also be creators for a worldwide audience. This is one of the main propert ies of what is popularly called Web 2.0 and will be further discussed later on in this book.
In the fixed line telephony world, prices for national and international calls continued to decline. Towards the end of this period, initial attempts were also made to use the Internet for transporting voice calls. Early adopters discovered the use of Internet telephony to make phone calls over the Internet via their DSL lines. Proprietary programs like Skype suddenly allowed users to call any Skype subscriber in the world for free, in many cases with superior voice quality. ‘Free’ in this regard is a relative term, however, since both parties in the call have to pay for access to the Internet, so telecom operators still benefit from such calls due to the monthly charge for DSL or cable connections. Additionally, many startup companies started to offer analog tele­phone to Internet Protocol (IP) telephone converters, which used the standardized SIP (Session Initiation Protocol) protocol to transport phone calls over the Internet. Gateways ensured that such subscribers could be reached via an ordinary fixed line telephone number and could call any legacy analog phone in the world. Alternative long-distance carriers also made active use of the Internet to tunnel phone calls between countries and thus offer cheaper rates.
Starting in 2001, the General Packet Radio Service (GPRS) was introduced in public GSM networks for the first time. When the first GPRS-capable mobile phones quickly followed, mobile Internet access became practically feasible for a wider audience. Until then, mobile Internet access had only been possible via circuit-switched data calls. However, the data rate, call establishment times and the necessity of maintaining the channel even during times of inactivity were not suitable for most Internet applications. These problems, along with the small and monochrome displays in mobile phones and mobile software being in its infancy, meant that the first wireless Internet services (WAP 1.0) never became popular. Towards 2005, devices matured, high-resolution color displays made it into the mid-range mobile phone segment and WAP 2.0 mobile Web browsers and easy-to-use mobile e-mail clients in combination with GPRS as a packet-switched transport layer finally allowed mobile Internet access to cross the thresh­old between niche and mass market. Despite these advances, pricing levels and the struggle between open and closed Internet gardens, which will be discussed in more detail later on, slowed down progress considerably.
At this point it should be noted that throughout this book the terms ‘mobile access to the Internet’ and ‘mobile Internet access’ are used rather than ‘mobile Internet’. This is done on purpose since the latter term implies that there might be a fracture between a ‘fixed line’ and a ‘mobile’ Internet. While it is true that some services are specifically tailored for use on mobile devices and even benefit and make use of the user’s mobility, there is a clear trend for the same applications, services and content to be offered and useful on both small mobile devices and bigger nomadic or stationary devices. This will be discussed further in Chapter 6.
Another important milestone for wireless Internet access during this timeframe was 3G networks going online in many countries in 2004 and 2005. While GPRS came close to analog modem speeds, UMTS brought data rates of up to 384 kbit/s in practice, and the
4 Beyond 3G – Bringing Networks, Terminals and the Web Together
experience became similar to DSL. Again, network operator pricing held up mass adoption for several years.
1.4 From 2005 to today: Global Coverage, VoIP and Mobile Broadband
From 2005 to today, the percentage of people in industrialized countries accessing the Internet via broadband DSL or cable connections has continued to rise. Additionally, many network operators have started to roll out ADSL2+, and new modems enable download speeds beyond 15 Mbit/s for users living close to a central exchange. VDSL and fiber to the curb/fiber to the home deployments offer even higher data rates. Another trend that has accelerated since 2005 is Voice over IP (VoIP) via a telephone port in the DSL or cable modem router. This effectively circumvents the traditional analog tele­phone network and traditional network fixed line telephony operators see a steady decline in their customer base.
At the time of publication, the number of mobile phone users has reached 3 billion. This means that almost every second person on Earth now owns a mobile phone, a trend which only a few people foresaw only five years ago. In 2007, network operators registered 1000 new users per minute [3]. Most of this growth has been driven by the rollout of second-generation GSM/GPRS networks in emerging markets. Due to global competition between network vendors, network components reached a price that made it feasible to operate wireless networks in countries with very low revenue per user per month. Another important factor for this rapid growth was ultra-low-cost GSM mobile phones, which became available for less than $50. In only a few years, mobile networks have changed working patterns and access to information for small entrepreneurs like taxi drivers and tradesmen in emerging markets [4]. GSM networks are now available in most parts of the world. Detailed local and global maps of network deployments can be found in [5].
In industrialized countries, third-gener ation networks continued to evolve and 2006 saw the first upgrades of UMTS networks to High Speed Data Packet Access (HSDPA). In a first step, this allowed user data speeds between 1 and 3 Mbit/s. With advanced mobile terminals, speeds are likely to increase further. Today, such high data rates are mainly useful in combination with notebooks to give users broadband Internet almost anywhere. In the mid term, it is likely that HSDPA will also be very beneficial for mobile applications once podcasts , music downloads and video streaming on mobile devices become mass market applications.
While 3G networks have been available for some time, take-up was sluggish until around 2006/2007, when mobile network operators finally introduced attractive price plans. Prices fell below E40–E50 for wireless broadband Internet access and monthly transfer volumes of around 5 Gbytes. This is more than enough for everything but file sharing and substantial video streaming. Operators have also started to offer smaller packages in the range of E6–15 a month for occasional Internet access with notebooks. Packages in a similar price range are now also offered for unlimited Web browsing and e-mail on mobile phones. Pricing and availability today still vary in different countries. In 2006, mobile data revenue in the USA alone reached a $15.7 billion, of which 50–60% is non-SMS revenue [6]. In some countries, mobile data revenues now accounts for between 20 and 30% of the total operator revenue, as shown in Figure 1.1.
Evolution from 2G over 3G to 4G 5
35
30
25
20
15
10
5
0
T-MobileUSChina
Unicom
Figure 1.1 Percentage of data revenue of mobile operators in 2007 [6].
Vodafone
Ger
O2 UK3 Italy
%
While wireless data roaming is still in its infancy, wireless Internet access via prepaid SIM cards is already offered in many countries at similar prices to those for customers with a monthly bill. This is another important step, as it opens the door to anytime and anywhere Internet access for creative people such as students, who favor prepaid SIMs to monthly bills. In addition, it makes life much easier for travelers, who until recently had no access to the Internet while traveling, except for wireless hotspots at airports and hotels. An updated list of such offers is maintained by the Web community on the prepaid wireless Internet access Wiki [7].
1.5 The Future – the Need for Beyond 3G Systems
When looking into the future, the main question for network operators and vendors is when and why Beyond 3G wireless networks will be needed. Looking back only a couple of years, voice telephony was the first application that was mobilized. The Short Message Service (SMS) followed some years later as the first mass market mobile data application. By today’s standards comparably simple mobile phones were required for the service and little bandwidth. In a way, the SMS service was a forerunner of other data services like mobile e-mail, mobile Web browsing, mobile blogging, push-to-talk, mobile instant messaging and many others. Such applications became feasible with the introduction of packet-based wireless networks that could carry IP data packets and increasingly powerful mobile devices. Today, the capacity of current 3G and 3.5G netw orks is still sufficient for the bandwidth requirements of these applications and the number of users. There are a number of trends, however, which are already visible and will increase bandwidth requirements in the future:
Rising use – due to falling prices, more people will use mobile applications that require network access.
Multimedia content – while first attempts at mobilizing the Web resulted in mostly text-based Web pages, graphical content is now the norm rather than the exception.
6 Beyond 3G – Bringing Networks, Terminals and the Web Together
A picture may paint a thousand words, but it also increases the amount of data that has to be transferred for a Web page. Video and music downloads are also becoming more popular, which further increases in bandwidth requirements.
Mobile social networks – similar to the fixed-line Internet, a different breed of applications is changing the way people are using the Internet. In the past, users mainly consumed content. Blogs, podcasts, picture-sharing sites and video portals are now reshaping the Internet, as users no longer only consume content, but use the network to share their own ideas, pictures and videos with other people. Applications like, for example, Shozu [8] and Lifeblog [9] let users upload pictures, videos and Blog entries from mobile devices to the Web. In particular, picture, podcast and video transfers multiply the amount of data that users transmit and receive.
Voice over IP – the fixed line world is rapidly moving towards VoIP. It is likely that, five years from now, many of today’s fixed line circuit-switched voice networks will have migrated towards IP-based voice transmission. Likewise, on the network access side, many users will use VoIP as their primary fixed line voice service, for example over DSL or TV cable networks. The beginnings can already been observed today, as the circuit-switched voice market is under increasing pressure due to declining sub­scriber numbers. As a consequence, many operators are no longer investing in this technology. A similar trend can be observed in wireless networks. Here, however, the migration is much slower, especially due to the higher bandwidth requirements for transporting voice calls over a packet-switched bearer. This topic is discussed in more detail in Chapter 1.6.
Fixed-line Internet replacement – while the number of voice minutes is increasing, revenue is declining in both fixed line and the wireless networks due to falling prices. In many countries, wireless operators are thus trying to keep or increase the average revenue per user by offering Internet access for PCs, notebooks and mobile devices over their UMTS/HSDPA or CDMA networks. Thus, they have started to compete directly with DSL and cable operators. Again, this requires an order of magnitude of additional bandwidth on the air interface.
Competition from alternative wireless Internet providers – in some co untries, alter­native operators are already offering wireless broadband Internet access with Wi-Fi or WiMAX/802.16 networks. Such operators directly compete with traditional UMTS and CDMA carriers, who are also active in this market.
The broadband Internet is not a socket in the wall – this statement combines all previous arguments and was made by Anssi Vanjoki, Executive VP of Nokia’s Multimedia division [10], at a press conference. Today, many people already use Wi­Fi access points to create their personal broadband Internet bubble. Thus, broadband Internet is virtually all around them. In the future, people will not only use this bubble with desktop computers and notebooks, but also with smaller devices such as mobile phones with built-in Wi-Fi capabilities. Smaller devices will also change the way we perceive this Internet bubble. No longer is it necessary to sit down at a specific place, for example in front of a computer, in order to communicate (VoIP, e-mail, instant messaging), to get information or to publish information to the Web (pictures, Blog entries, videos, etc.). When the personal broadband bubble is left, mobile devices switch over to a cellular network. As we move into the future, the cellular network will extend into areas not covered today and available bandwidth will have to increase
Evolution from 2G over 3G to 4G 7
to cope with the rising number of users and their connected applications. Moving between the personal Internet bubble at home and the larger external cellular network will become seamless as devices and services evolve.
A number of wireless technologies are currently under development or in the early rollout phase that are designed to meet these future demands: 3GPP’s Long Term Evolution (LTE), HSPA+ and WiMAX. In addition, Wi-Fi is also likely to be an important network technology that is required to meet future capacity demands. All of these technologies will be further discussed in Chapter 2. The question that arises in this context is which of these technologies are 3G and which will be called 4G in the future?
The body responsible for categorizing wireless networks is the International Telecommunication Union (ITU). The ITU categorizes International Mobile Telecommunication (IMT) networks as follows:
IMT-2000 systems – this is what we know as 3G systems today, for example UMTS
and cdma2000. The list of all ITU-2000 systems is given in ITU-R M.1457-6 [11].
Enhanced IMT-2000 systems – the evolution of IMT-2000 systems, for example
HSPA, CDMA 1xEvDo and future evolutions of these systems.
IMT-Advanced systems – systems in this category are considered to be 4G systems.
At this time, there is still no clear definition of the characteristics of future IMT­Advanced (4G) systems. The ITU-R M.1645 recommendation [12] gives first hints but leaves the door wide open:
It is predicted that potential new radio interface(s) will need to support data rates of up to approximately 100 Mbit/s for high mobility such as mobile access and up to approximately 1 Gbit/s for low mobility such as nomadic/local wireless access, by around the year 2010 [...] These data rate figures and the relationship to the degree of mobility [...] should be seen as targets for research and investigation of the basic technologies necessary to implement the framework. Future system specifications and designs will be based on the results of the research and investigations.
When comparing current the WiMAX specifications to these potential requirements, it becomes clear that WiMAX does not qualify as a 4G IMT-Advanced standard, since data rates are much lower, even under ideal conditions.
3GPP’s successor to its 3G UMTS standard, known as LTE, will also have difficulties fulfilling these requirements. Even with a four-way Multiple Input Multiple Output (MIMO) transmission, data rates in a 20 MHz carrier would not exceed 326 Mbit/s. It should be noted at this point that this number is already a long stretch, since putting four antennas in a small device or on a rooftop will be far from simple in practice.
It is also interesting to compare these new systems with the evolution of current 3G systems. The evolution of UMTS is a good example. With HSDPA and HSUPA, user speeds now exceed the 2 Mbit/s that was initially foreseen for IMT-2000 systems. The evolution of those systems, however, has not yet come to an end. Recent new develop­ments in 3GPP Release 7 and 8 called HSPA+, which include MIMO technology and other enhancements, bring ev olved UMTS technology to the same capacity and
8 Beyond 3G – Bringing Networks, Terminals and the Web Together
bandwidth levels as currently specified for LTE on a 5 MHz carrier. HSPA+ is also clearly not a 4G IMT-Advanced system, since it enhances a current 3G IMT-2000 radio technology. Thus, HSPA+ is categorized as an ‘enhanced IMT-2000 system’.
To meet the likely requirements of IMT-Advanced, the WiMAX and LTE standards bodies have started initiatives to further enhance their technologies. On the WiMAX side, the 802.16m task group is working on standardizing an even faster radio interface. On the LTE side, a similar working program has become known as LTE+ or Enhanced LTE.
Current research indicates that the transmission speed requirements described in ITU-R M.1645 can only be achieved in a frequency band of 100 MHz or more. This is quite a challenge, both from a technical point of view and also due to a lack of available additional spectrum. Thus, it is somewhat doubtful whether these requirements will remain in place for the final definition of 4G IMT-Advanced.
In practice, several different network technologies will coexist and evolve in the future to meet the rising demands in terms of bandwidth and capacity. It is also likely that a combination of different radio systems, like for example LTE together with Wireless LAN, will be used to satisfy capacity demands.
From a user and service point of view, it does not matter if a network technology is considered 3.5G, 3.9G or 4G. Thus, this book uses the term ‘Beyond 3G systems’ (B3G), which includes all technologies which will be able to satisfy future capacity demands and which either evolve out of current systems or are a new development.
1.6 All Over IP
While on the radio network side it is difficult to foresee which mix of evolved 3G and 4G technologies will be used in the future, the future of fixed and mobile core networks is much easier to predict. One of the main characteristics of 3G networks is the support for circuit-switched and packet-switched services. The circuit-switched part of the core net­work and circuit-switched services of the radio network were specifically designed to carry voice and video calls. Service control rests with the Mobile Switching Center (MSC), the main component of a circuit-switched network. As subscribers can roam freely in a mobile network, a database is required to keep track of the current location of the subscriber in addition to the subscription information. This database is referred to as the Home Location Register (HLR). To establish a call, a mobile phone always contacts the MSC. The MSC then uses the destination’s telephone number to query the HLR for the location of the destination subscriber. The call is then routed to this MSC, which in turn informs the destination subscriber of the incoming call. This process is called signaling. For the speech path, a transparent circuit-switched channel is established between the two parties via the MSCs switching matrix. The signaling required for the call is transferred over an independent signaling network, as the circuit-switched channel only transports the speech signal.
In recent network designs, MSCs are split into an MSC Call Server component that handles the signaling and a media gateway that is responsible for forwarding the voice call as shown in Figure 1.2. Instead of fixed connections, media gateways use packet­switched ATM (Asynchronous Transfer Mode) or IP connections to forward the call. This removes the necessity to transport the voice data via circuit-switched connections in the core network.
p
Evolution from 2G over 3G to 4G 9
HLR
Location and Subscriber
Signaling connection
Database
Call Server
Media
Radio
Network
Radio base station
Figure 1.2 Circuit switching with dedicated network components.
Gateway
Exclusive channel for a connection or IP data flow with constant data rate
Call Server
Media
Gateway
Radio
Network
In the radio network ATM or A circuit switched connections is used for a call. Voice data and signaling for the call is not
orted over IP!
trans
While this approach is ideally suited to carry voice and video calls with a constant bandwidth and delay requirements, it performs poorly for a connection to the Internet. Here, all data is transported in data packets. Furthermore, data packets are not only exchanged between two endpoints while a connection is established, but usually between many. An example is a Web browsing session during which a user visits several Web sites, sometimes even simultaneously. While a Web page is transferred, it is desirable to use as much bandwidth as is currently available, rather than be limited to a circuit-switched channel that is designed to carry a digitized narrowband voice or video stream. An Internet connection is often also idle for a substantial duration. During this time, resources are best given to other users. This is also not possible with a circuit-switched connection, because it is an exclusive channel that offers a fixed amount of bandwidth between two parties while it is established.
For these reasons, 3G networks contain a separate core network to forward data packets rather than circuits. This is shown in Figure 1.3. The radio network serves both the circuit-switched and the packet-switched network and the kind of connection estab­lished to a user over the air depends on whether a circuit-switched connection or a packet­switched connection is required. Some systems such as UMTS even allow devices to simultaneously use packet and circuit connections so a phone call can be made while being connected to the Internet and transferring data.
Traditional fixed line networks use a similar split for simultaneous voice telephony and Internet access. Since DSL became popular, analog voice service and DSL use the same physical line to the customer’s home. A splitter is then used to separat e the analog telephone signal from the DSL service as they operate in different frequency bands. In the central exchange office, a similar splitter is used to connect the line of the subscriber to
10 Beyond 3G – Bringing Networks, Terminals and the Web Together
Call Server
Media
Radio
Network
ATM or IP
Figure 1.3 Typical circuit-switched and packet-switched dual architecture of 3G networks. The location and subscriber database is not shown.
Gateway
Radio Network
Packet Gateway
Call Server
Media
Gateway
Private
IP Network
Radio
Network
Internet
Gateway
Internet
the local circuit-switched exchange for voice calls and additionally to a DSL Access Multiplexer (DSLAM) for Internet connectivity. Telephone exchanges are then inter­connected via circuit-switched connections, while the DSLAM connects to a packet­switched backbone. In the meantime, however, there is a clear shift to transporting telephone calls over the Internet connection as well. Instead of connecting the analog phone to the splitter, the DSL access device is equipped with a jack for the phone. The DSL access device digitizes the voice signal and sends it as IP packets over the DSL connection. In many cases, an IP-based SIP server and RTP (Real Time Transport Protocol) replace the local circuit-switched telephone exchange. There are several advan­tages of this approach:
Only a single type of core network is needed, as the circuit-switched telephone exchanges and the circuit-switched network between them are no longer necessary.
Using an IP network for voice calls makes it a lot easier for companies other than the local telephone carrier to offer telephony services, as the controlling network element no longer needs to be at the local exchange.
Voice services can be combined with other services. Since there is more bandwidth available, users can, for example, exchange pictures with each other while being engaged in a voice call or add video at any point during the conversation.
While the trend to VoIP is already fully underway in fixed-line networks, wireless net­works have not yet caught up. Here, things are moving more slowly for a number of reasons. The main reason is that 3G mobile networks did not have the necessary bandwidth to support VoIP, which requires a higher data rate than circuit-switched
Evolution from 2G over 3G to 4G 11
voice calls. The gap has been somewhat reduced by the introduction of 3.5G networks. However, only B3G networks (evolved IMT-2000 and IMT-Advan ced) will have enough capacity and an optimized radio network to support VoIP on a large scale.
The challenges are significant, but none of the new B3G network architectures discussed in Chapter 2 have a circuit-switched core network. To be successful, it is essential for B3G wireless network operators to have a fully functioning VoIP solution in place in the future that is able to seamlesslytransfer the call to a circuit-switched wireless connection when the user roams out of network coverage. This is discussed in more detail in Chapter 4.
1.7 Summary
This chapter presented how fixed and wireless networks evolved in the past 15 years from circuit-switched voice-centric systems to packet-switched Internet access systems. Due to the additional complexity of wireless systems, enhancements are usuall y introduced in fixed-line systems first and only some years later in wireless systems as well. To date, fixed-line networks offer data rates to the customer premises of several megabits per second, in some cases already going beyond this. Wireless 3.5G networks are capable of data rates in the order of several megabits per second. In the future, more bandwidth and capacity will be achieved by evolving current wireless network technologies (evolved IMT-2000) and by designing new access networks (IMT-Advanced). This book therefore not only c oncentrates on 4G systems, but also discusses the evolution of 3G systems. Another important development is the use of packet-switched networks for transporting telephone calls, which is referred to as VoIP. This trend is already fully underway in fixed­line networks and will inevitably also happen in B3G networks, as systems such as WiMAX and LTE have been designed without a circuit-switched core network dedicated to voice calls.
References
1. Background on Fidonet (2008) http://www.fidonet.org.
2. Neuhetzki, T. (December 2005) German long distance tariffs in the 1990s, http://www.teltarif.de/arch/ 2005/kw52/s19950.html.
3. Sauter, M. (August 2006) 1000 new mobile phone users a minute, http://mobilesociety.typepad.com/ mobile_life/2006/08/1000_new_mobile.html.
4. Andersen, T. (19 February 2007) Mobile phone lifeline for world’s poor, http://news.bbc.co.uk/1/hi/ business/6339671.stm.
5. 2G and 3G coverage maps (2008) http://www.coveragemaps.com.
6. Sharma, C. (September 2007) Global wireless data market, http://www.chetansharma.com/ globalmarketupdate1H07.htm.
7. The prepaid wireless Internet access Wiki (2008) http://prepaid-wireless-internet-access.wetpaint.com.
8. Shozu (2008) http://www.shozu.com.
9. Lifeblog (2008) http://r2. nokia.com/nokia/0,71739,00.html.
10. Biography of Anssi Vanjoki Executive VP of Nokia Multimedia (2008) http://www.nokia.com/A4126347.
11. The International Telecommunication Union (2006) Detailed specifications of the radio interfaces of International Mobile Telecommunications-2000 (IMT-2000), ITU-R M.1457-6.
12. The International Telecommunication Union (2003) Framework and overall objectives of the future development of IMT-2000 and systems beyond IMT-2000, ITU-R M.1645.
2
Beyond 3G Network Architectures
2.1 Overview
As discussed in Chapter 1, the general trend in telecommunications is to move all applications to a common transmission protocol, the Internet Protocol. The tremendous advantage of this approach is that applications no longer require a specific network technology but can be used over different kinds of networks. This is important since, depending on the situation, an application might be used best over a cellular network while at other times it is more convenient and cheaper to use a wireless home or office networking technology such as Wi-Fi. The increasing number of multiradio devices supports this trend. Today and even more so in the future, a number of wireless technologies are deployed in parallel. This is necessary as the deployment of a new network requires a considerable amount of time and there are usually only a small number of devices supporting a new network technology at first. It is therefore important that different network technologies are deployed not only in parallel but also at the same location. As well as the introduction of new technologies, existing network technologies continue to evolve to offer improved performance while the new technology is not yet deployed or is just in the process of being rolled out. For these reasons, this chapter looks at a number of different Beyond 3G network technologies with an emphasis on those with the highest market share. In this context, the term ‘Beyond 3G networks’ is used for cellular networks that offer higher speeds than the original UMTS networks with their maximum data rate of 384 kbit/s per user.
In the cellular world, the Universal Mobile Telecommunication System (UMTS) with
its High-speed Packet Access (HSPA) evolution is currently the Beyond 3G system with the broadest deployment. This system, together with its future evolution, HSPAþ,is therefore discussed first.
Next, the chapter focuses on the successor technology of HSPA and HSPAþ, which is
commonly known as Long Term Evolution (LTE). In the standa rds, LTE is referred to as
Beyond3G–BringingNetworks,TerminalsandtheWebTogether:LTE, WiMAX, IMS, 4G Devices and the Mobile Web 2.0
MartinSauter © 2009JohnWiley&Sons,Ltd. ISBN: 978-0-470-75188-6
14 Beyond 3G – Bringing Networks, Terminals and the Web Together
the Evolved Packet System (EPS), which is divided into the Evolved Packet Core (EPC) and the Enhanced-UMTS Terrestrial Radio Access Network (E-UTRAN).
While LTE mainly addresses incumbent wireless operators, there is also great interest from new companies in building wireless networks for Internet access. Many of these companies are attracted by the Worldwide Interoperability for Microwave Access (WiMAX) standard, in particular with the 802.16e air interface. WiMAX is very similar to LTE, but designed from the ground up without the need for backwards-compatibili ty. Therefore, it is much more suitable for these companies’ needs. Since it is expected that both LTE and WiMAX will gain considerable market share, both technologies are discussed to show the similarities and also the differences between the two.
As will be shown throughout this book, 802.11 Wi-Fi networks will play an important role in overall wireless network architectures of the future. Consequently, this chapter also introduces Wi-Fi and the latest enhancements built around the original standard, such as an evolved air interface with speeds of up to 600 Mbit/s, security enhancements for home and enterprise use and quality of service extensions.
To give an initial idea about the performance of each system, some general observa­tions for each system in terms of bandwidth, speed and latency are discussed. Since these parameters are of great importance, and often grossly exaggerated by marketing depart­ments, Chapter 3 will then look at this topic in much more detail.
2.2 UMTS, HSPA and HSPAþ
2.2.1 Introduction
Initial drafts of UMTS standards documents appeared in working groups of the Third Generation Partnership Project (3GPP) at the end of 1999, but work on feasibility studies for the system began much earlier. A few UMTS networks were opened to the public in 2003, but it was not until the end of 2004, when adequate UMTS mobile phones became available and networks were rolled out to more than just a few cities, that even early adopters could afford and actually use UMTS. A time frame of five years from a first set of specifications to first deployments is not uncommon due to the complexity involved. This should also be considered when looking at emerging network technologies such as LTE and WiMAX, which are currently in this window between standardization and deployment.
2.2.2 Network Architecture
Figure 2.1 shows an overview of the network architecture of a UMTS network. The upper-left side of the figure shows the radio access part of the network, referred to in the 3GPP standards as the UMTS Terrestrial Radio Access Network (UTRAN).
2.2.2.1 The Base Stations
The UTRAN consists of two components. At the edge of the network, base stations, referred to in the standards as the NodeB, communicate with mobile devices over the air. In cities, a base station usually covers an area with a radius of about 1 km, sometimes
g
g
Beyond 3G Network Architectures 15
UE
UTRAN
RNC
NodeB
RNC
NodeB
Core Network
MSC GMSC
HLR
SCP
PSTN
GSM BSS
TRAUTRAU
BSCBSC
BTS
Figure 2.1 Common GSM/UMTS network. (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
PCUPCU
Data and signalin
SGSN
GGSN
Internet
Server
Signalin
less, depending on the population density and bandwidth requirements. To increase the amount of data and the number of simultaneous voice calls per base station, the coverage area is usually split into two or three sectors. Each sector has its own directional antenna and transceiver equipment. In the standards, a sector is sometimes also referred to as a cell. A NodeB with three sectors therefore consists of three individual cells. If a user walked around such a base station during an ongoing voice call or while data was exchanged, he would be consecutively served by each of the cells. During that time, the radio network would hand the connection over from one cell to the next once radio conditions deteriorated. From a technical point of view, there is thus little difference between a handover between cells of the same base station and between cells of different base stations. These and other mobility management scenarios will be discussed in more detail in Section 2.2.3. The radio link between mobile devices and the base station is also referred to as the ‘air interface’ and this term will also be used throughout this book. A device using a UMTS network is referred to in the standard as User Equipment (UE). In this book, however, the somewhat less technical terms ‘mobile’, ‘mobile device’ and ‘connected mobile device’ are used instead.
Today, base stations are connected to the network via one or more 2 Mbit/s links, referred to as E-1 connections in Europe and T-1 connections in the USA (with a slightly lower transmission speed). Each E-1 or T-1 link is carried over a pair of copper cables. An alternative to copper cables is a microwave connection, which can carry several
16 Beyond 3G – Bringing Networks, Terminals and the Web Together
logical E-1 links over a single microwave connection. This is preferred by many operators as they do not have to pay monthly line rental fees to the owner of the copper cable infrastructure. To make full use of the air interface capacity of a multisector base station, several E-1 links are required. The protocol used over these links is ATM (Asynchronous Transfer Mode), a robust transmission technology widely used in many fixed and wireless telecommunication networks around the world today. Figure 2.2 shows a typical base station cabinet located at street level. In practice, base stations are also frequently installed on flat rooftops close to the antennas, as there is often no space at ground level and as this significantly reduces the length and thus the cost for the cabling between the base station cabinet and the antennas.
As technology evolves, using E-1 links over copper cables becomes more difficult since the number of copper cables leading to a base station is limited and, more significantly, the line rental costs per month are high. Network operators have therefore begun using a number of alternative transmission technologies to connect base stations to the network:
High bandwidth microwave links – a single microwave link can be used to carry
several logical E-1 connections. The latest equipment is capable of speeds exceeding
50 Mbit/s [1].
Fiber links – especially in dense urban areas, many fixed line carriers are currently
deploying additional fiber cables for providing very high data rate Internet access to
businesses and homes. This infrastructure is also ideal for connecting base stations to
the rest of the infrastructure of a wireless network. In practice, however, only a fraction
of deployed base stations already have a fiber laid up to the cabinet.
ADSL/VDSL – a viable alternative to directly using fiber is to connect base stations via
a high-speed VDSL link to an optical transmission network. T-Mobile in Germany is
one operator that has chosen this solution [2]. In some cases, base stations still require
at least one E-1 link for synchronizing the base station with the rest of the network and
for carrying voice calls.
Ethernet – a transmission protocol becoming very popular today in radio access
networks is IP over Ethernet. This is reflected in new designs for UMTS/HSPA base
stations, which can be equipped with E-1 ATM-based interfaces or alternatively via IP
over Ethernet. The Ethernet interface is either based on the standard 100 Mbit/s
100Base-TX twisted pair copper cable interface commonly used with other IT equip-
ment such as PCs and notebooks or via an optical port. In the case of copper cabling,
additional equipment is usually required to transport the Ethernet frames over longer
distances, as 100Base-TX limits cable length to 100 m.
As the technology used for backhauling data from base stations has a significant impact on the bandwidth and cost of a network, this topic will be discussed in more detail in Section 3.17.
2.2.2.2 The Radio Network Controllers
The second component of the radio access network is the Radio Network Controller (RNC). It is responsible for the following management and control tasks:
Beyond 3G Network Architectures 17
The establishment of a radio connection, also referred to as bearer establishment.
The selection of bearer properties such as the maximum bandwidth, based on current available radio capacity, type of required bearer (voice or data), quality of service requirements and subscription options of the user.
Mobility management while a radio bearer is established, that is, handover control between different cells and different base stations of a network.
Overload control in the network and on the radio interface. In situations when more users want to communicate than there are resources available, the RNC can block new connection establishment requests to prevent other connections from breaking up. Another option is to reduce the bandwidth of established bearers. A new data con­nection might, for example, be blocked by the network if the load in a cell is already at the limit, while for a new voice call, the bandwidth of an ongoing data connection might be reduced to allow the voice call to be established. In practice, blocking the establishment of a radio bearer for data transmission is very rare, as most network operators monitor the use of their networks and remove bottlenecks, for example by installing additional transceivers in a base station, by increasing backhaul capacity between the base station and the RNC or by installing additional base stations to reduce the coverage area and thus the number of users per base station. Capacity management will be discussed in more detail in Chapter 3.
2.2.2.3 The Mobile Switching Center
Moving further to the right in Figure 2.1, it can be seen that the RNCs of the network are connected to gateway nodes between the radio access network and the core network. In UMTS, there are two independent core network entities. The upper right of the figure shows the MSC, which is the central unit of the circuit-switched core network. It handles voice and video calls and forwards SMS messages via the radio network to subscribers. As discussed in Chapter 1, circuit switching means that a dedicated connection is established for a call between two parties via the MSC that remains in place while the call is ongoing. Large mobile networks usually have several MSCs, each responsible for a different geographical area. All RNCs located in this area are then connected to the MSC. Each MSC in the network is responsible for the management of all users of the network in its region and for the establishment of circuit-switched channels for incoming and outgoing calls. When a mobile device requests the establishment of a voice call, the RNC forwards the request to the MSC. The MSC then checks if the user is allowed to make an outgoing call and instructs the RNC to establish a suitable radio bearer. At the same time, it informs the called party of the call establishm ent request or, if the called party is located in a different area or different network, establishes a circuit-switched connection to another MSC. If the subscriber is in the same network it might be possible to contact the MSC responsible for the called party directly. In many cases, however, the called party is not in the same network or not a mobile subscriber at all. In this case, a circuit-switched connection is established to a Gateway MSC (GMSC), shown in Figure 2.1 on the top right. Based on the telephone number of the called party, the GMSC then forwards the call to an external fixed or mobile telephone network. In practice, a MSC usually serves mobile subscribers and also acts as a GMSC.
18 Beyond 3G – Bringing Networks, Terminals and the Web Together
To allow the MSC to manage subscribers and to alert them about incoming calls, mobile devices need to register with the MSC when they are switched on. At the begin­ning of the registration process, the mobile device sends its International Mobile Subscriber Identity (IMSI), which is stored on a SIM card, to the MSC. If the IMSI is not known to the MSC’s Visitor Location Register (VLR) database from a previous registration request, the network’s main user database, the Home Location Register, is queried for the user’s subscription record and authentication information. The authenti­cation information is used to verify the validity of the request and to establish an encrypted connection for the exchange of signaling messages. The authentication infor­mation is also used later on during the establishment of a voice or video call to encrypt the speech path of the connection. Note that the exchange of these messages is not based on the IP protocol but on an out-of-band signaling protocol stack called Signaling System Number 7 (SS7). Out-of-band means that messages are exchanged in de dicated signaling connections, which are not used for transporting circuit-switched voice and video.
2.2.2.4 The SIM card
An important component of UMTS networks, even though it is very small, is the Subscriber Identity Module, the SIM card. It allows the network subscription to be separate from the mobile device. A user can thus buy the SIM card and the mobile device separately. It is therefore possible to use the SIM card with several devices or to use several SIM cards with a single device. This encourages competition between network operators, as users can change from one network to another quickly if prices are no longer compe­titive. When traveling abroad, it is also possible to buy and use a local prepaid SIM card to avoid prohibitive roaming charges. Separating network subscriptions from mobile devices has the additional benefit that mobile devices can not only be bought from a network operator but also from independent shops, for example electronic stores and mobile phone shops that sell subscriptions for several network operators. This stimulates competitive pricing for mobile devices, which would not happen if a device could only be bought from a single source. A further discussion of this topic can be found in [3].
2.2.2.5 The SMSC
A data service that became very popular long before the rise of current high-speed wireless Internet access technologies is the short message service, used to send text messages between users. As the service dates back to the mid 1990s, it is part of the circuit-switched core network. SMS messages are transported in a store and forward fashion. When a subscriber sends a message, it is sent via the signaling channel, the main purpose of which is to transport messages for call establishment and mobility manage­ment purposes, to the Short Message Service Center (SMSC). The SMSC stores the message and queries the Home Location Register database to find the MSC which is currently responsible for the destination subscriber. Afterwards, it forwards the message, again in an SS-7 signaling link, to the mobile switching center. When receiving the text message, the MSC locates the subscriber by sending a paging message. This is necessary, as in most cases the subscriber is not active when a text message arrives and therefore the
Beyond 3G Network Architectures 19
user’s current serving cell is not known to the MSC. On the air interface, the paging message is sent on a broadcast channel that is observed by all devices attached to the network. The mobile device can thus receive the paging message and send an answer to the network despite not having being in active communication with the network. The network then authenticates the subscriber, activates encryption and delivers the text message. In case the subscriber is not reachable, the delivery attempt fails and the SMSC stores the message until the subscriber is reachable again.
2.2.2.6 Service Control Points
Optional, but very important, components in circuit-switched core networks are inte­grated databases and control logic on Service Control Points (SCPs). An SCP is required, for example, to offer prepaid voice services that allow users to top-up an account with a voucher and then use the credit to make phone calls and send SMS messages. For each call or SMS, the MSC requests permission from the prepaid service logic on an SCP. The SCP then checks and modifies the balance on the user’s account and allows or denies the request. Mobile switching centers communi cate with SCPs via SS-7 connections. When a prepaid user roams to another country, foreign MSCs also need to communicate with the SCP in the home network of the user. As there are many MSC vendors, the interacti on model and protocol between MSCs and SCPs have been specified in the CAMEL (Customized Applications for Mobile Enhanced Logic) standard [4].
For providing the actual service (e.g. prepaid), only signaling connections between SCPs and MSCs are required. Some services, such as prepaid, however, also require an interface to allow a user to check his balance and to top up their account. In practice there are several possibilities. Most operators use some form of scratch card and an automated voice system for this purpose. Therefore, there are usually also voice circuits required between SCP-controlled interactive voice gateways and the MSCs. In addition, most prepaid services also let users top up or check their current balance via short codes (e.g. *100#), which do not require the establishment of a voice call. Instead, such short codes are sent to the SCP via an SS-7 signaling link.
2.2.2.7 Billing
In addition to the billing of prepaid users, which is performed in real time on SCPs, further equipment is required in the core network to collect billing information from the MSCs for subscribers who receive a monthly invoice. This is the task of billing servers, which are not shown in Figure 2.2. In essence, the billing server collects Call Detail Records (CDRs) from the MSCs and SMSCs in the network and assembles a monthly invoice for each user based on the selected tariff. Call detail records contain information such as the identity of the calling party, the identity of the called party, date and duration of the call and the identity of the cell from which the call was originated. Location information is required as calls placed from foreign networks while the user is roaming are charged differently from calls originated in the home network. Some network operators also use location information for zone-based billing, that is, they offer cheaper calls to users while they are at home or in the office. Another popular billing approach is
20 Beyond 3G – Bringing Networks, Terminals and the Web Together
Figure 2.2 A typical GSM or UMTS base station cabinet.
to offer cheaper rates at certain times. Most operators combine many different options into a single tariff and continuously change their billing options. This requires a flexible rule-based billing service.
2.2.2.8 The Packet-switched Core Network
The core network components discussed so far have been designed for circuit-switched communication. For communicating with services on the Internet, which is based on packet switching, a different approach is required. This is why a packet-switched core network was added to the circuit-switched core network infrastructure. As can be seen in Figure 2.2, the Radio Network Controller connects to both the circuit-switched core network and the packet-switched core network. UMTS devices are even capable of having circuit-switched and packet-switched connections established at the same time. A user can therefore establish a voice call while at the same time using his device as a modem for a PC, or for downloading content such as a podcast to the mobile device without interrupting the connection to the Internet while the voice call is ongoing. Another example of the benefits of being connected to both the packet-switched and circuit-switched networks is that an ongoing instant messaging session is not interrupted during a voice call.
Before a mobile device can exchange data with an external packet-switched network such as the Internet, it has to perform two tasks. First, the mobile device needs to attach to the packet-switched core network and perform an authentication procedure. This is
Beyond 3G Network Architectures 21
usually done after the device is switched on and once it has registered with the circuit­switched core network. In a second step, the mobile device can then immediately, or at any time later on, request an IP address from the packet-switched side of the network. This process is referred to as establishing a data call or as establishing a PDP (Packet Data Protocol) context. The expression ‘establishing a data call’ is interesting because it suggests that establishing a connection to the Internet or another external packet net­work is similar to setting up a voice call. From a signal ing point of view, the two actions are indeed similar. The connections that are established as a result of the two requests, however, are very different. While voice calls require a connection with a constant bandwidth and delay that remain in place while the call is ongoing, data calls only require a physical connection while packets are transmitted. During times in which no data is transferred, the channel for the connection is either modified or completely released. Nevertheless, the logical connection of the data call remains in place so the data transfer can be resumed at any time. Furthermore, the IP address remains in place even though no resources are assigned on the air interface. In UMTS, separating the attachment to the packet-switched core network from the establishment of a data connection makes sense when looked at from a historical and practical perspective. The majority of mobile devices today are mostly used for voice communication for which no Internet connection is required. Therefore, the mobile device can perform its main duty without establishing a data call. All other networks that will be discussed in this chapter, however, are only based on a packet-switched core network which is also used for voice calls. As a result, attaching to the network and requesting an IP address is part of the same procedure and the notion of a ‘data call’ is no longer part of the system design.
2.2.2.9 The Serving GPRS Support Node
The packet-switched UMTS core network was, like the circuit-switched core network, adapted from GSM with only a few modifications. This is the reason why the gateway node to the radio access network is still referred to as the Serving GPRS Support Node (SGSN). GPRS stands for General Packet Radio Service and is the original name of the packet-switched service introduced in GSM networks. Like the MSC, the SGSN is responsible for subscriber and mobility management. To enable users to move between the coverage areas of different RNCs, the SGSN keeps track of the location of users and changes the route of IP packets arriving from the core network when they change their locations. SGSNs are connected to RNCs via ATM links. Since RNCs often control several hundred cells and are physically distant from an SGSN, optical connections are used. In practice, a single RNC is connected to an SGSN via one or more OC-3 optical links with a speed of 155 Mbit/s each or an OC-12 optical connection with a speed of 622 Mbit/s [5]. Above the ATM layer, IP is used as a transport protocol.
The SGSN also has a signaling connection to the Home Location Register of the network that in addition to the data required for the circuit-switched network also contains subscription information for the packet-switched network. This includes, for example, if the user is allowed to use packet-switched services, their quality of service settings such as the maximum transmission speed that is granted by the network and which access points to the Internet they are allowed to use. The record also shows
22 Beyond 3G – Bringing Networks, Terminals and the Web Together
whether a user has a prepaid contract, in which case the SGSN has to contact a prepaid SCP before a connection request is granted.
2.2.2.10 The Gateway GPRS Support Node
At the edge of the wireless core network, Gateway GPRS Support Nodes (GGSNs) connect the wireless network to the Internet. Their prime purpose is to hide the mobility of the users from routers on the Internet. This is required as IP routers forward packets based on the destination IP address and a routing table. For each incoming packet, each router on the Internet consults its routing table and forwards the packet to the next router via the output port specified in the routing table. Eventually, the packets for wireless subscribers end up at the GGSN. Here, the routing mechanism is different. As RNCs and SGSNs can change at any time due to the mobility of the subscriber, a static database containing the next hop based on the IP address is not suitable. Instead, the GGSN has a database table which lists the IP address of the SGSN currently responsible for a sub­scriber with a certain IP address. The IP packet for the subscriber is then enclosed in an IP packet with the destination address of the SGSN. This principle is called tunneling because each IP packet to the subscriber is encapsulated in an IP packet to the responsible SGSN. On the SGSN the original IP packet is restored and once again tunneled to the RNC.
Figure 2.3 shows how tunneling an IP packet in the core network works in practice. The protocol stack shown uses the Ethernet protocol on layer 2 which is followed by IP on layer 3. The source and destination address of the IP packet belong to the SGSN and GGSN of the netwo rk. Afterwards, the GPRS Tunneling Protocol (GTP) encapsulates the original IP packet. Here, the source and destination address represent the subscriber and a host on the Internet such as a Web server.
In addition to tunneling, the GGSN is also responsible for assigning IP addresses to subscribers. During the connection establishment, the SGSN verifies the request of the subscriber with the information from the HLR and then requests the establishment of a
Figure 2.3 Encapsulated IP packet in a packet-switched core network.
Beyond 3G Network Architectures 23
tunnel and an assignment of an IP address from the GGSN. The GGSN then assigns an IP address from a pool and establishes the tunnel. In practice, there are two types of IP addresses. Many network operators use private IP addresses that are not valid outside the network. This is similar to the use of private IP addresses in home networks where the DSL or cable router assigns private IP addresses to all devices in the home network. To the outside world, the network is represented with only a single IP address and the DSL or cable router has to translate internal addresses in combination with TCP and UDP port numbers into the external IP address with new TCP and UDP port numbers. This process is referred to as Network Address Translation (NAT). If private IP addresses are used in wireless networks, the same process has to be performed by the GGSN. The advantage for the operator is that fewer public IP addresses are required, of which there is a shortage today. Other operators use public IP addresses for their subscribers by default. For subscribers this has the advantage that they are directly reachable from the Internet, which is required for applications such as hosting a Web server or for remote desktop applications. It should be noted, howeve r, that a public IP address also has disadvan­tages, especially for mobile devices, as unsolicited packets arriving from the Internet can drain batteries of mobile devices [6]. In practice, a subscriber can have several connection profiles, referred to as Access Point Names (APNs). The operator can therefore choose which subscribers and applications to use private or public IP addresses for.
2.2.2.11 Interworking with GSM
The lower left side of Figure 2.2 shows a GSM radio network, which is also connected to the circuit-switched and packet-switched core network. Depending on the network vendor, the GSM radio network can either be connected to the same MSCs and SGSNs or to separate ones. This is possible since the functionality of these components is very similar for GSM and UMTS. The interfaces to the 2G and 3G radio networks, however, are different. For the conn ections to the 3G radio network, ATM is used as the lower layer transport protocol while the connection between the SGSN and the 2G radio network is based on the Frame Relay protocol. In a newer version of the standard, an IP-based interface between SGSNs and the 2G radio network has been specified as well. At this point, however, it is not widely used. The MSC is connected to the GSM radio network via circuit-switched links.
2.2.3 Air Interface and Radio Network
2.2.3.1 The CDMA Principle
2G radio systems such as GSM are based on timeslots and channels on different carrier frequencies so a single base station can serve many users simultaneously. While such an approach is well suited for the transmission of circuit-switched voice calls, it only offers limited flexibility for packet-switched data transmission, as the GSM channel bandwidth is only 200 kHz and thus only a limited amount of bandwidth can be assigned to a single user at a time. For 3G systems such as UMTS, it was therefore decided to use a different transmission scheme for the air interface. An alternative transmission technology that overcomes the limitations of a narrow band and was already well understood at the end
24 Beyond 3G – Bringing Networks, Terminals and the Web Together
of the 1990s, when work on UMTS started, was Code Division Multiple Access (CDMA). Instead of a narrow channel bandwidth of 200 kHz as in GSM, the Wideband CDMA (W-CDMA) channel used in UMTS has a bandwidth of 5 MHz. Furthermore, instead of assigning timeslots, all users communicate with the base station at the same time but using a different code. These codes are also referred to as spreading codes, as a single bit is represented on the air interface by a codeword. Each binary unit of the codeword is referred to as a chip to clearly distinguish it from a bit. For a transmission speed of 384 kbit/s, a single bit is encoded in eight chips. The length of the spreading code thus equals 8. The codes used by different users for transmitting simultaneously are mathematically orthogonal to each other. The base station can separate several simulta­neous transmissions from mobile devices by applying the reverse algorithm to the one used by the mobile devices, as shown in Figure 2.4, and by knowing the codes used by each mobile device.
Figure 2.4 Simultaneous data streams of two users to a single base station. (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
To accommodate the different bandwidth requirements of different users, the system can use several different spreading code lengths simultaneously. For voice calls, which require a transmission rate of only 12.2 kbit/s, a spreading code length of 128 is used. This means that each bit is represented on the air interface by 128 chips and that up to 128 users can send their data streams simultaneously under ideal conditions. In practice, the number of voice calls per cell is smaller, due to interference from neighboring cells and transmission errors due to nonideal signal conditions. This requires a higher signal-to­noise ratio than could be achieved if there were 128 simultaneous transmissions. Furthermore, some of the codes are reserved for broadcasting system information channels to all devices in the cell, as will be discussed in more detail below. In practice, it is estimated in [7] that each UMTS cell can host up to 60 simultaneous voice calls, excluding subscribers using the cell for packet-switched Internet access.
Beyond 3G Network Architectures 25
To increase the data rate of UMTS beyond 384 kbit/s, HSPA and HSPAþ introduce,
among other enhancements, the simultaneous use of more than one spreading code per mobile device. This is discussed in more detail below.
2.2.3.2 UMTS Channel Structure
In wired Ethernet networks, which are commonly used in home and business environ­ments today, detecting the network and sending packets is stra ightforward for devices. As soon as the network cable is plugged in, the network card senses the transmission of packets over the cable and can start transmitting as well, as soon as it detects that no other device is sending a packet. In wireless networks, however, things are more compli­cated. First of all, there are usually several networks visible at the same time so the device needs to get information about which network belongs to which operator. Also, devices need to detect neighboring cells to be able to quickly react to changing signal levels when the user moves. As keeping the receiver switched on at all times is not very power­efficient, it is also necessary to have a mechanism in place that allows the device to power down the radio in situations where little or no data is transmitted and only wake up periodically to check for new data. Especially for voice calls, it is furthermore important to ensure a certain quality of service in order to prevent the network from breaking down in overload situations. For these reasons, the radio channel is split up into a number of individual channels. Access to these channels is controlled by the network and can be denied during overload situations.
In UMTS networks, a physical channel is represented by a certain spreading code. The following list gives an overview of the most important channels in the downlink direction (network to mobile device) and the uplink direction (mobile device to the network). All of these channels are transmitted simultaneously:
The Primary Common Control Physical Channel (P-CCPCH) – this channel carries
the logical broadcast control channel which is monitored by all mobile devices while
they do not have an active connection established to the network. Information dis-
tributed via this channel includes the identity of the cell, how the network can be
accessed, which spreading codes are used for other channels in the cell, which codes are
used by neighboring cells, timers for network access, and so on.
The Secondary Common Control Physical Channel (S-CCPCH) – this physical chan-
nel serves several purposes and thus carries a number of different logical channels.
First, it carries paging messages that are used by the network to search for mobile
devices for incoming calls, SMS messages or when data packets arrive after a long
period of inactivity. The channel is also used to deliver IP packets and control messages
to devices which only exchange small quantities of data at a particular time and thus do
not need to be put into a more active state.
Physical Random Access Channel (PRACH) – the random access channel only exists
in the uplink direction and is the only channel that a mobile device is allowed to
transmit without prior permission of the network. Its purpose is to allow the mobile to
request the establishment of a connection to set up a voice call, to establish a data call,
to react to a paging message or to send an SMS message. If a data call is already
26 Beyond 3G – Bringing Networks, Terminals and the Web Together
established and the mobile only sends or receives a small amount of data, the channel can also be used to send IP packets in combination with the S-CCPCH in the downlink direction. However, the data rate in both directions is very low and round trip delay times are high (in the order of 200 ms). The network therefore usually assigns different channels that are dedicated to packet-switched data transfers as soon as it detects renewed network ac tivity from a device.
Dedicated Physical Data and Control Channels (DPDCH, DPCCH) – once a mobile has contacted the network via the random access channel, the network usually decides to establish a full connection to the mobile. In case of a voice call, the network assigns a dedicated connection. On the air interface, the connection uses a dedicated spreading code that is assigned by the network. In the early days, UMT S networks also used dedicated connections for packet-switched data transmissions. In practice, however, the speed of such dedicated connections was limited to 384 kbit/s. Only a few users could get such a connection, also referred to as a bearer, in a cell simultaneously. Another downside of using a dedicated bearer for packet-switched connections is that the network has to frequently reassign spreading codes to different users. This is necessary, as most packet-switched data transmissions are bursty in nature and there­fore only require a high bandwidth bearer for a limited amount of time. In order to reuse the spreading code for oth er users, the network needs to change the spreading code length of a connection as soon as it is not fully used any more or even put the connection on a common control channel in order to have the resources available for other users. In practice, this creates a high signaling overhead and a mediocre user experience. It was thus decided that a new concept was needed that offers higher data rates and more flexibility for bursty data transmissions. The outcome of this process is known today as High-speed Packet Access, which many people also refer to as 3.5G. HSPA will be discussed in Section 2.2.4.
2.2.3.3 Radio Resource Control States
To save power and to only assign resources to mobile devices when necessary, a connec­tion to the network can be in one of the following Radio Resource Control (RRC) states:
Idle state – devices not actively communicating with the network are in this state. Here, they periodically listen to the paging channel for incoming voice or video calls and SMS messages.
Cell-FACH state – if a mobile device wants to contact the network, it moves to the Cell-Forward Access Channel (Cell-FACH) state. In this state the mobile sends its control messages via the random access channel and the network replies on the forward access channel, which is sent over the S-CCPCH described above. Power requirements also increase as the mobile device now needs to monitor the downlink for incoming control messages.
Cell-DCH state – once the network decides to establish a voice or data connection, the mobile is instructed to use a dedicated channel and is therefore moved to the Cell Dedicated Channel State (Cell-DCH). If the device is HSPA-capable, the network selects a shared channel instead, as will be discussed in more detail below. From a radio
Beyond 3G Network Architectures 27
resource control point of view, however, the mobile is still treated as being in Cell-
DCH state. Round trip delay times in Cell-DCH state range from 160 ms with a
dedicated bearer to 120 ms for an HSPA connection. In the case of a packet-switched
data connection, the network can decide to move a device back to Cell-FACH state if
its activity, that is, the amount of data transferred, decreases. Once activity increases
again, the network moves the connection back to Cell-DCH state. For even longer
periods of inactivity during a data session, the network can even decide to put the
mobile device back into idle state to reduce power consumption of the mobile device
and to reduce the management processing load for the network. Despite being in Idle
state, the mobile device can resume sending IP packets at any time via the random
access channel. However, there is a noticeable delay when moving to a more active
state. Depending on the radio network vendor, initial round trip delay times between
2 and 4 s can be observed [8].
Cell-PCH and URA-PCH states – even while in Cell-FACH state, the mobile device
requires a considerable amount of power to listen to the forward access channel despite
its minimal activity. Because of the noticeable delay when resuming data transmission
from the Idle state, two further states have been specified which are between Idle and
Cell-FACH state. In Cell-PCH and URA-PCH state, the mobile only needs to periodi-
cally listen to the paging channel while the logical connection between the radio network
and the device remains in place. If there is renewed activity, the connection can be
quickly resumed. The difference between the two states is that in Cell-PCH state the
mobile has to report cell changes to the network because the paging message is only sent
into one cell, whereas in URA-PCH state the mobile can roam between several cells that
belong to the same routing area without reporting a cell change to the network since the
paging message is sent into all cells belonging to the routing area. In practice, however,
only a few network operators make use of these two additional states.
2.2.3.4 Mobility Management in the Radio Network for Dedicated Connections
An important task in cellular networks is to handle the mobility of the user. In Cell­DCH state, the base station actively monitors the air interface connection to and from a mobile device and adjusts the transmission power of the base station and the transmission power of the mobile device 1500 times per second. In CDMA systems, such very quick power adjustments are necessary, as the transmissions of all mobile devices should be received by the base station w ith the same power level whether they are very close or very far away. In the direction from the network to a mobile device, as little transmission power as possible should be used as the transmission power of the base station is limited. In practice, data transfers to and from mobile devices close to a base station only require a small amount of transmission power while devices far away or inside buildings require significantly more transmission power. Adjusting the transmission power so frequently also means that, along with the user data, the mobile device has to continuously report to the network how well the user data is received. The network then processes this information and instructs the mobile device, at the same rate, whether it should increase, hold or decrease its current transmission power. For this purpose, a dedicated control channel is always established alongside a
28 Beyond 3G – Bringing Networks, Terminals and the Web Together
dedicated traffic channel. While the content of the dedicated traffic channel is given to higher layers of the protocol stack, which eventually extract a voice packet or an IP packet, t he information transported in the control channel remains in the radio soft­ware stack and is not visible to higher-layer applications.
When a user moves, they eventually leave the coverage area of a base station. While a DCH is established, the Radio Network Controller takes care that the connection to the mobile device is transferred to a more suitable cell. This decision is based on reception quality of the current cell and the neighboring cells, which is measured by the mobile device and sent to the network. In CDMA-based networks such as UMTS, there are two different kinds of handovers. The first type is referred to as a hard handover. When the network detects that there is a more suitable cell for a mobile device, it prepares the new cell for the subscriber and afterwards instructs the mobile to change to the new cell. The mobile device then interrupts the connection to the current cell and uses the handover parameters sent by the network, which include information on the frequency and spreading codes of the new cell, for establishing a connection. A hard handover is therefore a break-before-make handover, that is, the old connection is cut before the new one is established.
The second type of handover is a soft handover, which is the most commonly used type of handover in CDMA networks. Soft handovers make use of the fact that neighboring cells transmit on the same frequency as the current cell. It is thus possible to perform a make-before-break handover, that is, the mobile device commun icates with more than a single cell at a time. A mobile enters the soft handover state when the network sends control information to instruct it to listen to more than a single spreading code in the downlink direction. Each spreading code represents a transmission of a different c ell and the mobile combines the signals it receives from the different cells. In the uplink direction, the mobile device continues using only a single spreading code for a dedicated channel. All cells taking part in the soft handover are instructed by the RNC to decode the data sent with this code and forward it to the RNC. During soft handover, the RNC thus receives several copies of the user’s uplink data stream. This increases the likelihood that a packet is received correctly and does not have to be retransmitted. The cells which are involved in a soft handover for a dedicated channel are referred to as the active set. Up to six cells can be part of the active set. In practice, however, most networks limit the number of cells in the active set to two or three. Otherwise, the benefit of multiple transmissions that can be combined in both the network and the mobile station is outweighed by the additional capacity required in both the radio network and on the air interface. To improve the capacity of the network, around 30–40% of all dedicated conn ections in a network are in soft handover state, even if the users are not moving.
Figure 2.5 shows how hard and soft handovers are performed for a case in which not all cells are connected to the same radio network controller. In this example, the uplink and downlink data are collected and distributed by two RNCs. However, only one of the two RNCs, referred to as the Serving-RNC (S-RNC), communicates with the core network. The other RNC, referred to as the Drift-RNC (D-RNC), communicates with the S-RNC.
If the user moves further into the area covered by NodeB 2 and NodeB 3, the mobile will at some point lose contact with NodeB 1. At this point, the involvement of the RNC
a
Beyond 3G Network Architectures 29
SGSN
S-RNC
NodeB 1
Figure 2.5 Soft handover with several RNCs. (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
D-RNC
NodeB 2
Soft handover with three cells and two RNCs
This link is not used as all dat from and to the SGSN is sent via the Serving RNC (S-RNC)
NodeB 3
controlling NodeB 1 is no longer required. The current S-RNC then requests the SGSN to promote the current Drift-RNC as the new Serving RNC.
If the optional interface between the RNCs is not present, a soft handover as shown in Figure 2.5 is not possible. In this case, a hard handover is required when the mobile device moves out of the coverage area of the first NodeB and into the coverage area of the other two NodeB. Furthermore, the change of serving RNC has to be performed at the same time, which further complicates the handover procedure and potentially increases the period during which no data can be exchanged between the mobile device and the network.
When the user moves between the areas controlled by two different SGSNs, it is additionally necessary to change the serving SGSN. To make matters even more com­plicated, handovers can also take place while both a circuit-switche d and a packet­switched connection is acti ve. In such a case, not only the RNCs need to be changed, but also the SGSN and MSC. This requires close coordination between the packet­switched and circuit-switched core network. While in practice typical handover scenarios only involve cells controlled by the same RNC, all other types of handovers have to be implemented as well.
Since 3G/3.5G networks did not have the same coverage area as the already existing 2G GSM networks, and often still lack the same coverage area and indoor penetration today, it is very important to be able to handover active connections to GSM networks. This was and still is especially important for voice calls. Mobile devices, therefore, do not only search for neighboring 3G/3.5G cells but can also be instructed by the network to search for GSM cells. While a dedicated channel is established this is very difficult since in the standard transmission mode the mobile device sends and receives continuously. For cells at the edge of the coverage area, the RNC can activate a compressed transmission mode. The compressed mode opens up predefined transmission gaps that allow the mobile
30 Beyond 3G – Bringing Networks, Terminals and the Web Together
device’s receiver to retune to the frequencies of neighboring GSM cells, receive their signal, synchronize to them and return to the UMTS cell to resume communication and to report the result of the neighboring cell signal measurements to the network. Based on these measurements, the network can then instruct the mobile to perform an Inter-Radio Access Technology (Inter-RAT) hard handover to GSM. During an ongoing voice call, such a hard handover is usually not noticed by the user. A data connection handed over to a GSM/GPRS/EDGE network, however, gets noticeably slower.
2.2.3.5 Radio Network Mobility Management in Idle, Cell-FACH and Cell/URA-PCH State
When the user moves and the mobile device is in Idle or Cell/URA-PCH state, that is, no voice call is established and the physical connection of a data call has been removed after a phase of inactivity, the mobile device can decide on its own to move to a cell with a better signal. Cells are grouped into location areas and mobile devices changing to a different cell only need to contact the network to report their new position when selecting a cell in a new location area. This means that paging messages have to be broadcast in all cells belonging to a location area. In practice, a location area contains about 20–30 base stations, providing a satisfactory trade-off between the reduced number of location updates and the increased overhead of sending paging messages into all cells of a location area.
In Cell-FACH state, which is used if the mobile enters a phase of lower activity (i.e. it sends or receives only a few IP packets), the cell changes are also controlled by the mobile and not the network. Here, an efficient transition from one cell to another is deemed to be not important enough to justify the required processing capacity in the network and the higher power consumption in the mobile terminal.
2.2.3.6 Mobility Management in the Packet-switched Core Network
In the core network, the SGSN is responsible for keeping track of the location and reachability of a mobile device. In PMM (Packet Mobility Management) Detached state, the mobile is not registered in the network and consequently does not have an IP address. In PMM Connected state the mobile has a signaling connection to the SGSN (e.g. during a location update to report its new position) and also an IP connection. The SGSN, however, only knows the current RNC responsible for forwarding the packet and not the cell identity. This knowledge is not necessary, as the SGSN just forwards the data packet to the RNC and the RNC is responsible for forwarding it via the current cell to the user. In addition the SGSN does not know if the radio connection to the mobile is in Cell-FACH, Cell-DCH, Cell-PCH or URA-PCH state. This information is not necessary since it is the RNC’s task to decide in which state to keep the mobile based on the quality of service requirements of a connection and the current amount of transmitted data. Note that PMM Connected state is also entered for short periods of time during signaling exchanges such as a location update even if no IP connection is established.
If the RNC sets the radio connection to Idle state, it also informs the SGSN that the mobile is no longer directly reachable. The SGSN then modif ies its control state to PMM
Beyond 3G Network Architectures 31
Idle. For the SGS N this means that, if an IP packet arrives from the Internet later on, the mobile device has to be paged first. The IP packet is then buffered until a response is received and a signaling connection has been set up.
Figure 2.6 shows the different radio and core network states and how they are related to each other while a packet-switched data connection is established. When moving from left to right in the figure, it can be seen that, the deeper the node is in the network, the less information it has about the current mobility state and the radio network connection of the mobile device. Note that the state of a mobile device is also PMM Idle when the mobile has attached to the packet-switched core network after power on but has not yet established a data connection. The radio network mobility management states also apply when the mobile device communicates with the circuit-switched network. During a voice or video call via the MSC, the mobile device is in Cell-DCH state. The mobile is also set into Cell-DCH or Cell-FACH state during signaling message exchanges, for example during a location update with the MSC and SGSN.
UE
Idle
Cell-DCH
Cell-FACH
Cell/URA-PCH
Figure 2.6 Radio network and core network mobility management states for an active packet­switched connection.
RNC
PMM-Detached
PMM-Idle
PMM-Connected
SGSN
GGSN
IP Address assigned by GGSN
Internet
2.2.4 HSPA (HSDPA and HSUPA)
Soon after the first deployments of UMTS networks, it became apparent that the use of dedicated channels on the air interface for packet-switched data transmission was too inflexible in a number of ways. As per the specification, the highest data rate that could be achieved was 384 kbit/s with a spreading code length of 8. This limited the number of people who could use such a bearer simultaneously to eight in theory and to two or three in practice, as some spreading codes were required for the broadcast channels of the cell and for voice calls of other subscribers. Because of the bursty nature of many packet­switched applications, the bearer could seldom be fully used by a mobile device and therefore a lot of capacity remained unused. This was countered somewhat by first assigning longer code lengths to the user’s connection and then only upgrading the connection when it was detected that the bearer was fully used for some time (e.g. for a
32 Beyond 3G – Bringing Networks, Terminals and the Web Together
file download). Also, short spreading codes were quickly replaced by longer ones once it became apparent to the RNC that the capacity was no longer fully used. Despite these mechanisms, efficiency remained rather low. As a consequence, vendors started to specify a package of enhancements which are now referred to as High-Speed Downlink Packet Access (HSDPA) [9]. The specification of HSDPA started as early as 2001 but it took until late 2006 for the first networks and devices to support HSDP A in practice. Once standardization of HSDPA was a good way on, a number of enhancements were specified for the uplink direction as well. These are referred to as High-Speed Uplink Packet Access (HSUPA). Together, HSDPA and HSUPA are known today as HSPA.
The following paragraphs are structured as follows: first, an overview of the downlink enhancements is given. Afterwards, the performance improvements for the uplink direc­tion are discussed.
2.2.4.1 Shared Channels
To improve the utilization of the air interface, the concept of shared channels has been (re-) introduced. HSDPA introduces a High-Speed Downlink Physical Shared Channel (HS-DPSCH), which several devices can use quasi simultaneously as it is split into time­slots. The network can thus quickly react to the changing bandwidth requirements of a device by changing the number of timeslots assigned to a device. Thus, there is no longer a need to assign spreading codes or to modify the length of a spreading code to change the bandwidth assignment. HS-DPSCH timeslots are assigned to devices via the High-Speed Shared Control Channel (HS-SCCH). This channel has to be monitored by all mobile devices that the RNC has instructed to use a shared channel. The HS-DPSCH uses a fixed spreading code length of 16 and the HS-SCCH uses a spreading code length of 256.
2.2.4.2 Multiple Spreading Codes
To increase transmission speeds, mobile devices can listen to more than a single high­speed downlink shared channel. Category 6 HSDPA devices can listen to up to five high­speed downlink channels simultaneously, and category 8 devices to up to 10. Multiple high-speed downlink channels can also be used to transmit data to several devices simultaneously. For this purpose the network can configure up to four simultaneous shared control channels in the downlink direction and up to four mobile devices can therefore receive data on one or more shared channels simultaneously.
Figure 2.7 shows how user data is transferred through the radio network and which channels an HSDPA device has to decode simultaneously while it is in Cell-DCH state. At the RNC, the user’s data packets arrive from the core network in a Dedicated Traffic Channel (DTCH). From there, the RNC repackages the data into a single High-speed Dedicated Shared Channel (DSCH) and forwards it to the NodeB. In addition, the RNC sends control information to the device via a Dedicated Control Channel (DCCH). This channel is not shared between subscribers. It is also possible for the user to establish or receive a voice call during an ongoing HSDPA data transfer. In this case, the RNC additionally uses a dedicated traffic channel to forward the circuit-switched voice chan­nel to the user. At the base station (NodeB), data arriving via the HS-DSCH channel is buffered for a short while and then transmitted over one or several High-speed Physical
a
Beyond 3G Network Architectures 33
RNC
NodeB
UE
Figure 2.7 Simplified HSDPA channel overview in the downlink direction. (Reproduced from
Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
DTCH DCCH DTCH*
HS-DSCH
...
HS-PDSCH HS-SCCH DPDCH DPCCH
Shared Dedicated
DCH
Logical Channels
Transport Channels
Physical Channels
* optional, e.g. for simultaneous voice call
Downlink Shared Channels (HS-PDSCH). In practice, at least five simultaneous HS-PDSCHs are used and up to 10 simultaneous channels can be used for category 8 devices. In addition to those 5–10 shared channels, the device has to be able to monitor up to four high-speed downlink shared control channels to receive the timeslot assignments for the shared channels. Finally, the mobile has to decode control information arriving on the Dedicated Physical Control Channel (DPCCH) and potentially also the voice data stream arriving on the Dedicated Physical Data Channel (DPDCH).
In the uplink direction, an HSDPA-capable device also has to send a number of different data streams, each with a different spreading code. The first stream is the HSDPA control channel, which is used to acknowledge the correct reception of downlink data packets. In addition, a dedicated channel is required for IP packets transmitted in uplink direction. A further channel is necessary to transmit or to reply to radio resource commands from the RNC (e.g. a cell change command) and for the uplink direction of an ongoing voice call. Finally, the mobile needs to send control information (reception quality of the downlink dedicated channel) to the NodeB.
2.2.4.3 Higher Order Modulation
To further increase transmission speeds, HSDPA introduces a higher order modulation scheme to improve throughput for devices that receive the transmissions of the base station with a good signal quality. Basic UMTS uses Quadrature Phase Shift Keying (QPSK) modulation, which encodes two basic information elements (chips) per trans­mission step. HSDPA introduces 16QAM (Quadrature Amplitude Modulation) mod­ulation, which encodes four chips per transmission step. This doubles the achievable bandwidth under good reception conditions. Together with using 10 sim ultaneous
34 Beyond 3G – Bringing Networks, Terminals and the Web Together
downlink shared channels, the theoretical maximum downlink speed is 7.2 Mbit/s. In practice, transmission speeds range between 1.5 and 2.5 Mbit/s for devices that can bundle five shared channels (HSD PA category 6) and 4.5 Mbit/s for devices that can bundle 10 shared channels (HSDPA category 8).
2.2.4.4 Scheduling, Modulation and Coding, HARQ
The higher the transmission speed, the more important it is to detect transmission errors and react to them as quickly as possible. Otherwise, connection-orien ted higher-layer protocols such as TCP misinterpret air interface packet errors as congestion and slow down the transmission. To react to transmission errors more quickly, it was decided not to implement the HSDPA scheduler in the RNC, as was done before for dedicated channels, but to assign this task to the base station. HSDPA uses the Hybrid Automated Retransmission Request (HARQ) scheme for this purpose. With HARQ, data frames with a fixed length of 2 ms have to be immediately acknowledged by the mobile device. If a negative acknowledgement is sent, the packet can be repeated within 10 ms. The next data frame is only sent once the previous one has been positively confirmed. The mobile device must be able to handle up to eight simultaneous HARQ processes as the mobile device is allowed up to 5 ms to decode the packet before it has to send the ACK (Ackno wledgment) or the NACK (Negative Acknowledgment) to the network. During this time, two further data frames of other HARQ processes can arrive at the mobile station. In case data was not received correctly in one HARQ process, frames received via other HARQ processes have to be stored in the mobile device until all previous frames have been successfully received, so that the data can be forwarded in the correct order to higher protocol layer s.
Another reason to implement the scheduler in the base station is to be able to quickly react to changing signal conditions. Based on the feedback of the mobile device, the base station’s scheduler decides for each frame which modulation (QPSK or 16QAM) to use and how many error correction bits to insert. This process is referred to as Adaptive Modulation and Codi ng (AMC). Advanced algorithms in the base station use the knowledge about reception conditions of all mobile devices currently served on the high-speed shared channel. Devices with better signal conditions can be preferred by the scheduler, while devices that are in temporary deep fading situations receive fewer packets. This helps to reduce transmission errors and improves overall throughput of the cell as, on average, frames are transmitted with more efficient modulation and coding schemes. Studies in [10] and [11] have shown that an efficient scheduler can increase overall cell capacity by up to 30% compared with a simple scheduler that assigns time­slots in a round robin fashion.
2.2.4.5 Cell Updates and Handovers
As has been shown above, a mobile device in HSDPA Cell-DCH state receives its packet­switched data via shared channels. Nevertheless, additional dedicated channels are required alongside the shared channels to transport control information in both direc­tions. Furthermore, IP packets in the uplink direction are also transported in a dedicated channel. Finally, a parallel circuit-switched voice call is also transported in a
Beyond 3G Network Architectures 35
dedicated channel. These dedicated connections can be in soft handover state to improve reception conditions and to better cope with the user’s mobility. The shared channels, however, are only sent from one base station of a device’s active set. Based on the feedback received from the mobile device, the RNC can at any time promote another base station to forward the IP packets in the downlink. This procedure is referred to as a cell update. Since the procedure is controlled by the network, the interruption of the data traffic in the downlink direction is only very short.
2.2.4.6 HSUPA
In early 3G networks, uplink speeds were limited to 64–128 kbit/s. With the introduction of HSDPA, some vendors also included a radio network software update to allow dedicated uplink radio bearers with speeds of up to 384 kbit/s, if permitted by the conditions on the radio interface to a user. In practice, it can be observed that in many situations the mobile’s transmission power is sufficient to actually make use of such an uplink bearer. This is good for sending large files or e-mails with large file attachments, as well as for many Web 2.0 applications (cf. Chapter 6), where user-generated content is sent from a mobile device to a database in the network. As for dedicated downlink transmissions, however, assigning resources in the uplink in this way is not very flexible. The 3GPP standards body thus devised a number of improvements for the uplink direction, which are referred to as High-Speed Uplink Packet Access (HSUPA).
Unlike in the downlink, which was enhanced by using high-speed shared channels, the companies represented in 3GPP decided to keep the dedicated channel approach for uplink transmissions. There were several reasons for this decision. While in the downlink, the base station has an overview of how much data is in the buffers to be sent to all mobile devices, there is no knowledge in the base station about the buffer states of the mobile devices requesting to send data in the uplink. This makes assigning timeslots on a shared channel difficult as mobile devices continuously have to indicate if they have more data. The second reason to re-use the dedicated channel concept was that the soft handover concept is especially valuable for the uplink direction as the transmission power of mobile devices is limited. In the standards, an HSUPA dedicated uplink channel is referred to as an Enhanced-DCH or E-DCH.
A standard uplink dedicated channel is controlled by the RNC and spreading codes can only be changed to reach maximum data transfer rates of 64, 128 or 384 kbit/s. As the RNC is in control of the connection, the bearer parameters are only changed very slowly, for example only every few seconds once the RNC detects that the pipe is too big or too small for the current traffic load or once it detects that signal conditions have signifi­cantly changed. With the E-DCH concept, the control of the radio interface channel has been moved from the RNC to the base station, in a similar way to the solution for HSDPA. The base station controls overall scheduling of all E-DCH uplink transmissions by assigning bandwidth grants to all active E-DCH devices. Bandwidth grants are transmitted in the downlink direction via a new shared control channel, the Enhanced Absolute Grant Channel (E-AGCH). Access grants are translated into the maximum amount of transmission power each device is allowed to use. This way, all devices can still transmit their data to the base station at the same time while their maximum transmission speeds can be quickly adjusted as necessary. Staying with the dedicated channel concept
36 Beyond 3G – Bringing Networks, Terminals and the Web Together
and hence allowing all mobile devices to transmit at the same time has the additional benefit of shorter packet delay times compared with an approach where devices have to wait for their timeslot to transmit. Optionally, a second control channel, the Enhanced Relative Grant Channel (E-RGCH) can be used to quickly increase or decrease uplink transmission power step by step. This enables each base station involved in an E-DCH soft handover to decrease the interference caused by a mobile device, if this creates too much interference for communication with other devices.
The HARQ concept first introduced with HSDPA is also used with HSUPA for uplink transmissions. This means that the base station immediately acknowledges each data frame it receives. Faulty frames can thus be retransmitted very quickly. For this purpose, an additional dedicated downlink channel has been created, the Enhanced HARQ information channel (E-HICH). Each HSUPA device gets its own E-HICH while it is in Cell-DCH mode. If an E-DCH connection is in soft handover state, each of the contributing base stations sends its own acknowledgement for a frame to the mobile device. If only one acknowledgement is positive, the transfer is seen as successful and the HARQ process moves on to the next frame.
While it was decided to introduce a new modulation scheme for the downlink to further increase transmission speeds, it was shown with simulations that there would be no such effect for uplink transmissions. Uplink transmissions are usually power limited, which means that a higher order modulation cannot be used as the error rate would become too high. Instead, it was decided to introduce multicode transmissions in a single dedicated channel to allow the mobile to split its data into several code channels which are then transported simultaneously. This concept is similar to the concept of using several (shared) channels in the downlink to increase the data rates. The highest terminal category currently defined can use up to two spreading codes with a length of two and two spreading codes with a length of four. In theory, uplink data rates can reach up to 2 Mbit/s. In practice, data rates are lower in a similar way to those described above for HSDPA.
Figures 2.8 and 2.9 show the channels used in uplink and downlink direction for a terminal that supports both HSDPA and HSUPA for the most complicated case when a voice call is ongoing in parallel. In practice, encoding and decoding so many channels at the same time is very demanding and has only been made possible by the ever increasing processing power of mobile device chipsets, as is further discussed in Chapter 5. Furthermore, it can be observed that early HSPA network implementations fall back to the default dedicated channel approach for the packet-switched connection during a voice call.
2.2.5 HSPAþ and other Improvements: Competition for LTE
The quest to improve UMTS and make it even faster and more power-efficient and to allow more devices to use a cell simultaneously (e.g. for VoIP) has not ended with HSPA. For instance, there are a number of initiatives in Release 7 and Release 8 of the 3GPP standard to further improve the system. Improvements of the air interface are referred to as HSPAþ. In addition, the network architecture has also received an optional overhaul to be more efficient with a feature referred to as ‘one-tunnel’. The following section gives an overview of these improvements, many of which are likely to be introduced into networks over the next few years.
a
a
Beyond 3G Network Architectures 37
RNC
NodeB
UE-PHY
E-AGCH
E-HICH* E-RGCH*
HS-DSCH
HS-SCCH
HS-PDSCH
...
DCH**
DPDCH DPCCH
UE
HS-DSCH DCH**
*sent from different cells while in soft­handover state
**optional, e.g. for simultaneous voice call
Figure 2.8 Channels for a combined HSDPA and HSUPA transmission in the downlink direction. (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
RNC
E-DCH DCH**
Transport Channels
DCH**
DPCCH**
Physical Channels
Transport Channels
**optional, e.g. for simultaneous voice call
NodeB
UE-PHY
UE
E-DPDCH*
E-DCH
*potentially sent with several different spreading codes (multi-code)
E-DPCCH DPDCH**
DPCCH
(for HSDPA)
Figure 2.9 Channels for a combined HSDPA and HSUPA transmission in the uplink direction. (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
38 Beyond 3G – Bringing Networks, Terminals and the Web Together
2.2.5.1 Higher Order Modulation
One of the key parameters of a wireless system that is often cited in articles is the maximum transmission speed.So far, HSPA uses QPSK and 16QAM modulation to reach theoretical data rates of up to 14.4 Mbit/s in the downlink direction, which translates to 2–5 Mbit/s in practice under good radio conditions. To increase transmission rates further, 3GPP Release 7 introduces 64QAM modulation in the downlink, which transmits six chips per transmission step compared with four chips with 16QAM. In practice, it is expected that introducing 64QAM will result in a 30% throughputgain for users close to the center of the cell [12]. Most users, however, will not be able to use 64QAM modulation, as their signal­to-interference ratio will be too low. This will be discussed in more detail in Chapter 3. As the overall throughput in the cell increases, these users will nevertheless also benefit from this measure, as the cell will have more time to transmit its data, because data for users closer to the cell center can be sent faster. Furthermore, 64QAM modulation is also beneficial for micro cell deployments in public places such as shopping malls, where users are close to small cells and therefore create little interference.
In the uplink direction, no changes were made to the modulation scheme in 3GPP Release 7. Release 8, however, might include 16QAM modulation despite the earlier conclusion that there would be little benefit. More advanced receivers and a focus on microcell environments in public places may be responsible for changing this opinion.
2.2.5.2 MIMO
Another emerging technology to increase throughput under good signal conditions is Multiple Input Multiple Output, or MIMO for short. In essence, MIMO transmission uses two or more antennas at both the transmitter and the receiver side to transmit two independent data streams simultaneously over the same frequency band. This linearly increases data rates with the number of antennas. Two transmitter antennas and two receiver antennas (2 2) as currently specified for HSPAþ can thus double the data rate of the system under ideal signal conditions. Further technical background on MIMO can be found in Section 2.3. Release 7 of the standards foresees the use of MIMO in combination with 16QAM modulation in the downlink. If the network operator chooses to deploy a n extra set of base station antennas, the theoretical data rate of 14.4 Mbit/s is therefore increased to 28.8 Mbit/s.
Depending on the signal conditions, available antennas and device capabilities, the network can choose between a single data stream with 64QAM or two streams with 16QAM. Release 8 of the standar ds might combine MIMO with 64QAM, which would result in a peak data rate of 43.2 Mbit/s in the downlink. However, it should be once more noted at this point that such data rates can only be achieved by very few users of a cell. Details are given in Chapter 3. As uplink transmissions are usually power limited, MIMO has only been considered for the downlink direction.
2.2.5.3 Continuous Packet Connectivity
Continuous Packet Connectivity (CPC) is a package of features introduced in the 3GPP standards to improve handling of mobile subscribers while they have a packet connection
Beyond 3G Network Architectures 39
established, that is while they have an IP address assigned. Taken together, they aim at reducing the number of state changes to minimize delay and signaling overhead by introducing enhancements to keep a device on the high-speed channels (in HSPA Cell­DCH state) for as long as possible, even while no data transfer is ongoing. For this, it is necessary to reduce power consumption while mobiles listen to the shared channels and, at the same time, reduce the bandwidth requirements for radio layer signaling to increase the number of mobile devices per cell that can be held in HSPA Cell-DCH state.
CPC does not introduce new revolutionary features. Instead, already existing features are modified to achieve the desired results. To understand how these enhancements work, it is necessary to dig a bit deeper into the standards. 3GPP TR 25.903 [13] gives an overview of the proposed changes and the following descriptions refer to the chapters in the document which have been selected for implementation.
2.2.5.3.1 Feature 1: A new uplink control channel slot format (Section 4.1 of [11])
While a connection is established between the network and a mobile device, several channels are used simultaneously. This is necessary as there is not only user data sent over the connection but also control information to keep the link established, to control transmit power, and so on. Currently, the radio con trol channel in the uplink direction (the Uplink Dedicated Control Channel, UL DPCCH) is transmitted continuously, even during times of inactivity, in order not to lose synchronization. This way, the terminal can resume uplink transmissions without delay whenever required.
The control channel carries four parameters:
Transmit Power Control (TPC);
pilot (used for channel estimation of the receiver);
TFCI (Transport Format Combination Identifier);
FBI (Feedback Indicator).
The pilot bits are always the same and allow the receiver to get a channel estimate before decoding user data frames. While no user data frames are received, however, the pilot bits are of little importance. What remains important is the TPC. The idea behind the new slot format is to increase the number of bits to encode the TPC and decrease the number of pilot bits while the uplink channel is idle. This way, additional redundancy is added to the TPC field. As a consequence, the transmission power for the control channel can be lowered without risking corruption of the information contained in the TPC. Once user data transmission resumes, the standard slot format is used again and the transmission power used for the control channel is increased again.
2.2.5.3.2 Feature 2: CQI reporting reduction (Section 4.4 of [11]), uplink discontinuous transmission (Section 4.2 of [11]) in combination with downlink control information transmission enhancements
CQI reporting reduction:to make the best use of the current signal conditions in the downlink, the mobile has to report to the network how well its transmissions are received. The quality of the signal is reported to the network with the Channel Quality Index (CQI) alongside the user data in the uplink. To reduce the transmit power of the terminal while data is being transferred in the uplink but not in the downlink, this feature reduces the number of CQI reports.
40 Beyond 3G – Bringing Networks, Terminals and the Web Together
UL HS-DPCCH gating (gating = switch off): when no data is being transmitted in either the uplink or downlink, the uplink control channel (UL DPCCH) for HSDPA is switched off. Periodically, it is switched on for a short time to transmit bursts to the network in order to maintain synchronization. This improves battery life for applications such as Web browsing. This solution also lowers battery consumption for VoIP and reduces the noise level in the network (i.e. allowing more simultaneous VoIP users). Figure 2.10 shows the benefits of this approach.
Sporadically sent data packets
Default HSPA
HSPA + CPC
Control Channel is switched off
Figure 2.10 Control channel switch-off during times with little activity.
Control Channel
t
F-DPCH gating: terminals in HSDPA active mode always receive a Dedicated Physical Channel in the downlink, in addition to high-speed shared channels, which carries power control information and Layer 3 radio resource (RRC) messages, for example for handovers, channel modifications and so on. The Fractional-DPCH feature puts the RRC messages on the HSDPA shared channels and the mobile thus only has to decode the power control information from the DPCH. At all other times, that is when the terminal is not in HSDPA active mode, the DPCH is not used by the mobile (thus it is fractional). During these times, power control information is transmitted for other mobiles using the same spreading code. Consequently, several mobiles use the same spreading code for the dedicated physical channel but listen to it at different times. This means that fewer spreading codes are used by the system for this purpose, which in turn leaves more resources for the high-speed downlink channels or allows more users to be kept in HSPA Cell-DCH state simultaneously.
2.2.5.3.3 Feature 3: Discontinuous Reception (DRX) in the Downlink (Based on Section 4.5 of [11])
While a mobile is in HSPA mode, it has to monitor one or more high-speed shared control channels (HS-SCCH) to see when packets are delivered to it on the high-speed shared channels. This monitoring is continuous, that is the receiver can never be switched off. For situations when no data is transmitted, or the average data transfer rate is much lower than that which could be delivered over the high-speed shared channels, the base station can
Beyond 3G Network Architectures 41
instruct the mobile to only listen to selected slots of the shared control channel. The slots which the mobile does not have to observe are aligned as much as possible with the uplink control channel gating (switch-off) times. Therefore, there are times when the terminal can power down its receiver to conserve energy. Once more data arrives from the network than can be delivered with the selected DRX cycle, the DRX mode is switched off and the network can once again schedule data in the downlink continuously.
2.2.5.3.4 Feature 4: HS-SCCH-less Operation (Based on Sections 4.7 and 4.8 of [11])
This feature is not intended to improve battery performance but to increase the number of simultaneous real-time VoIP users in the network. VoIP service, for example via the IMS (cf. Chapter 4), requires relatively little bandwidth per user and thus the number of simultaneous users can be high. On the radio link, however, each connection has a certain signaling overhead. Therefore, more users mean more signaling overhead which decreases overall available bandwidth for user data. In the case of HSPA, the main signaling resources are the high-speed shared control channels (HS-SCCH). The more active users there are, the more they proportionally require of the available bandwidth.
HS-SCCH-less operation aims at reducing this overhead. For real-time users who require only limited bandwidth, the network can schedule data on high-speed downlink channels without prior announcements on a shared control channel. This is done as follows:the network instructs the mobile not only to listen to the HS-SCCHbut in addition to all packets being transmitted on one of the high-speed downlink shared channels. The terminal then attempts to blindly decode all packets received on that shared channel. To make blind decoding easier, packets which are not announced on a shared control channel can only have one of four transmission formats (number of data bits) and are always modulated using QPSK. These restrictions are not an issue for performance, since HS-SCCH-less operation is only intended for low-bandwidth real-time services.
The checksum of a packet is additionally us ed to identify for which device the packet is intended. This is done by using the terminal’s MAC address as an input parameter for the checksum algorithm in addition to the data bits. If the device can decode a packet correctly and if it can reconstruct the checksum, it is the intended recipient. If the checksum does not match then either the packet is intended for a different terminal or a transmission error has occurred. In both cases the packet is discarded.
In case of a transmission error, the packet is automatically retransmi tted since the mobile did not send an acknowledgement (HARQ ACK). Retransmissions are announced on the shared control channel, which requires additional resources but should not happen frequently as most packets should be delivered properly on the first attempt.
2.2.5.4 Enhanced Cell-FACH, Cell/URA PCH States
The CPC features described above aim to reduce power consumption and signaling overhead in HSPA Cell-DCH state. The CPC measures therefore increase the number of mobile devices that can be in Cell-DCH state simultaneously and allow a mobile device to remain in this state for a longer period of time even if there is little or no data being transferred. Eventually, however, there is so little data transferred that it no longer makes sense to keep the mobile in Cell-DCH state, that is it does not justify even the reduced signaling overhead and power consumption. In this case, the network puts the
42 Beyond 3G – Bringing Networks, Terminals and the Web Together
connection into Cell-FACH state as described above or even into Cell-PCH or URA­PCH state to reduce energy consumption even further. The downside of this is that a state change back into Cell-DCH state takes a long time and that little or no data can be transferred during the state change. In Release 7 and 8, the 3GPP standards were thus extended to also use the high-speed downlink shared channels for these states, as described in [14] and [15]. In practice this is done as follows:
Enhanced Cell-FACH – in the standard Cell-FACH state the mobile device listens to the secondary common control physical channel in the downlink as described above for incoming radio resource control messages from the RNC and for user data (IP packets). With the Enhanced Cell-FACH feature, the network can instruct a mobile device to observe a high-speed downlink control channel or the shared data channel directly for incoming radio resource control messages from the RNC and for user data. The advantage of this approach is that, in the downlink direction, information can be sent much faster. This reduces latency and speeds up the Cell-FACH to Cell-DCH state change procedure. Unlike in Cell-DCH state, no other uplink or downlink control channels are used. In the uplink, the mobile still uses the random access channel to respond to radio resource control messages from the RNC an d to send its own IP packets. This limits the use of adaptive modulation and coding since the mobile cannot send frequent measurement reports to the base station to indicate the downlink reception quality. Furthermore, it is also not possible to acknowledge proper receipt of frames. Instead, the RNC informs the base station when it receives measurement information in radio resource messages from the mobile.
Enhanced Cell/URA-PCH states – in these two states, the mobile device is in a deep sleep state and only observes the paging information channel to be alerted of an incoming paging message which is transmitted on the paging channel. To transfer data, the mobile device is then moved to Cell-FACH or Cell-DCH state. If the mobile device and the network support Enhanced Cell/URA-PCH states, the network can instruct the mobile device not to use the slow paging channel to receive paging information but to use a high-speed downlink shared channel instead. The high­speed downlink channel is then also used for subsequent RRC commands which are required to move the device back into a more active state. Like the measure above, this significantly decreases the wakeup time.
Figure 2.11 shows how this works in practice. While the message exchange to notify the mobile device of incoming data and to move it to another activity state remains the same, using the high-speed downlink shared channels for the purpose speeds up the procedure by several hundred milliseconds.
Which of the described enhancements will make it into networks in the future remains to be seen and will also depend on how quickly LTE and other competing network technologies are rolled out. While CPC and enhanced mobility management states increase the efficiency of the system, they also significantly increase the complexity of the air interface, as both old and new mobile devices have to be supported simulta­neously. This rising complexity is especially challenging for the development of devices and networks, as it creates ad ditional interaction scenari os which become more and more
Beyond 3G Network Architectures 43
UE
URA-PCH
paging
paging response
Cell-FACH
re-establish bearers
Cell-DCH
data exchanged
time
Figure 2.11 Message exchange to move a mobile device from URA-PCH state back to cell-DCH state when IP packets arrive from the network.
Network
IP packets arrive from the network but cannot be delivered as no bearer is established
High Speed Downlink Channels used for transmitting RRC messages to speed up the procedure
difficult to test and debug before. Already today, devices are tested with network equipment of several vendors and different software versions. Adding yet another layer of features will make this even more complex in the future.
2.2.5.5 Radio Network Enhancement: One-tunnel
Figure 2.12 shows the default path of user data between a mobile device and the Internet through the cellular network. In the current architecture, the packet is sent through the GGSN, the SGSN, the RNC and the base station. All user data packets are tunneled through the network as described above, since the user’s location can change at any time. The current architecture uses a tunnel between the GGSN and the SGSN and a second tunnel between the SGSN and the RNC. All data packets therefore have to pass through the SGSN, which terminates one tunnel, extracts the packets and puts them into another tunnel. This requires both time and processing power.
Since both the RNC an d the GGSN are IP routers, this process is not required in most cases. The one-tunnel approach, now standardized in 3GPP (see [16] and [17]), allows the SGSN to create a direct tunnel between the RNC and the GGSN. It thus removes itself from the transmission chain. Mobility management, however, remains on the SGSN which means, for example that it continues to be responsible for mobility management and tunnel modifications in case the mobile device is moved to an area served by another
44 Beyond 3G – Bringing Networks, Terminals and the Web Together
Default Network Setup One-Tunnel SGSN Bypass
GGSN
signaling (e.g. mobility management)
Figure 2.12 Current network architecture vs the one-tunnel enhancement.
SGSN
RNC
NodeB
user data
tunnels
GGSN
SGSN
NodeB
one-tunnel
for user
data
RNC
RNC. For the user, this approach has the advantage that the packet delay is reduced. From a network point of view, the advantage is that the SGSN requires fewer processing resources per active user, which helps to reduce equipment costs. This is especially important as the amount of data traversing the packet-switched core network is rising significantly.
A scenario where the one-tunnel option is not applicable is international roaming. Here, the SGSN has to be in the loop in order to count the traffic for inter-operator billing purposes. Another case where the one-tunnel option cannot be used is when the SGSN is asked by a prepaid system to monitor the traffic flow. This is only a small limitation, however, since in practice it is also possible to perform prepaid billing via the GGSN.
Proprietary enhancements even aim to terminate the user data tunnel at the NodeB, bypassing the RNC as well. However, this has not found the widespread support of companies in 3GPP and is not likely to be compatible with some HSPAþ extensions, like the enhanced mobile device states.
2.2.5.6 Competition for LTE in 5 MHz
With the enhancements described in this section, it is quite likely that enhanced HSPA networks become a viable alternative to LTE deployments in the short and medium term, as the spectral efficiency in a 5 MHz band of both systems is similar. As will be described in the next section, LTE scores over HSPA when more bandwidth is available, as it is not limited to 5 MHz channels. In practice, it is thus likely that some network operators will choose to improve their HSPA network and only later on move to LTE, while others will prefer to go straight to LTE.
Beyond 3G Network Architectures 45
2.3 LTE
2.3.1 Introduction
For several years, there has been an ongoing trend in fixed line networks to migrate all circuit-switched services to a packet-switched IP infrastructure. In practice, it can be observed that fixed line network operators are migrating their telephony services to a packet-switched architecture offering both telephony and Internet access either via DSL or a cable modem. This means, that circuit-switched technology is replaced by VoIP­based solutions, as will be described in more detail in Chapter 4. In wireless networks, this trend has not yet begun. This is mostly due to the fact that current 3G and 3.5G network architectures are still optimized for circuit-switched telephony both in the radio network and in the core network. In addition, today’s VoIP telephony implementations signifi­cantly increase the amount of data that has to be transferred over the air interface, which means fewer voice calls can be handled simultaneously. Besides these challenges, how­ever, migrating voice telephony to IP offers a number of significant benefits such as cheaper networks and integration with other IP-based applications as discussed in more detail in Chapters 5 and 6.
At the same time, the general trend of ever increasing transmission bandwidths is highlighting the limits of current 3G and 3.5G networks. It was therefore decided in 2005 by the 3GPP standardization body to start work on a next generation wireless network design that is only based on packet-switched data transmission. This research was performed in two study programs. The LTE program focused on the design of a new radio network and air interface architecture. Slightly afterwards, work started on the design of a new core network infrastructure with the Service Architecture Evolution (SAE) program. Later, they were combined into a single work program, the Evolved Packet System (EPS) program. By that time, however, the abbreviation ‘LTE’ was already dominant in literature and most documents still refer to LTE rather than EPS.
Besides being fully packet-based, the following design goals were set for the new network:
Reduced time for state changes – in HSPA networks today, the time it takes a mobile
device to connect to the network and start communication on a high-speed bearer is
relatively long. This has a negative impact on usability, as the user can feel this delay
when accessing a service on the Internet after a period of inactivity. It was thus decided
that with a new network design it should be possible to move from idle state to being
fully connected in less than 100 ms.
Reduced user plane latency – another downside of current cellular networks is the
much higher transmission delay compared with fixed line networks. While one-way
delay between a user’s computer at the edge of a DSL network to the Internet is around
15 ms today, HSPA networks have a delay of arou nd 50 ms. This is disadva ntageous
for applications such as telephony and real-time gaming. For LTE, it was decided that
air interface delay should be in the order of 5 ms to reach end-to-end delays equaling
fixed line networks.
Scalable bandwidth – HSPA networks are currently limited to a bandwidth of 5 MHz.
At some point, higher throughput can only be reasonably achieved by increasing the
46 Beyond 3G – Bringing Networks, Terminals and the Web Together
bandwidth of the carrier. For certain applications, a carrier of 5 MHz is too large and it was thus decided that the air interface should also be scalable in the other direction.
Throughput increase – for the new system, a maximum throughput under ideal conditions of 100 Mbit/s should be achieved.
The following secti ons now describe how these design goals are met in practice.
2.3.2 Network Architecture
2.3.2.1 Enhanced Base Stations
Figure 2.11 shows the main components of an LTE core and radio access network as described in [18]. Compared with UMTS, the radio network is less complex. It was decided that central RNCs should be removed and their functionality has been partly moved to the base stations and partly to the core network gateway. To differentiate UMTS base stations from LTE base stations, they are referred to as Enhanced NodeB, (eNodeB). As there is no central controlling element in the radio network any more, the base stations now perform air interface traffic management autonomously and ensure quality of service. This was already partly the case in UMTS with the introduction of HSPA, as discussed in the previous section. Control over bearers for circuit-switched voice telephony, however, rested with the RNC.
In addition, base stations are now also responsible for performing handovers for active mobiles. For this purpose, the eNodeB can now communicate directly with each other via the X2 interface. The interface is used to prepare a handover and can also be used to forward user data (IP packets) from the current base station to the new base station to minimize the amount of user data lost during the handover. As the X2 interface is optional, base stations can also communicate with each other via the access gateway to prepare a handover. In this case, however, user data is not forwarded during the hand­over. This means that some of the data already sent from the network to the current base station might be lost, as once a handover decision has been made, it has to be executed as quickly as possible before radio contact is lost. Unlike in UMTS, LTE radio networks only perform hard handovers, that is only one cell communicates with a mobile device at a time.
The interface that connects the eNodeB to the gateway nodes between the radio network and the core network is the S1 interface. It is fully based on the IP protocol and is therefore transport technology agnostic. This is a big difference to UMTS. Here, the interfaces between the NodeB, the RNCs and the SGSN were firmly based on the ATM protocol for the lower protocol layers. Between the RNC and the NodeB, IP was not used at all for packet routing. While allowing for easy time synchronization between the nodes, requiring the use of ATM for data transport on lower protocol layers makes the setup inflexible and complicated. In recent years, the situation has worsened as rising bandwidth demands cannot be satisfied any more with ATM connections over 2 Mbit/s E-1 connections. The UMTS standard was thus enhanced to also use IP as a transport protocol to the base station. LTE, however, is fully based on IP transport in the radio network from day one. Base stations are either equipped with 100 Mbit/s or 1 Gbit/s Ethernet ports, as known from the PC world, or with gigabit Ethernet fiber ports.
Beyond 3G Network Architectures 47
2.3.2.2 Core Network to Radio Access Network Interface
As shown in Figure 2.13, the gateway between the radio access network and the core network is split into two logical entities, the Serving Gateway (Serving-GW) and the Mobility Management Entity (MME). Together, they fulfill similar tasks as the SGSN (Serving GPRS Support Node) in UMTS networks. In practice, both logical components can be implemented on the same physical hardware or can be separated for independent scalability.
PDN-GW
HSS
SCP
S6
MME
eNodeB
Mobile Device
Figure 2.13 Basic LTE network architecture.
S5
S11
X2
SGi
Serving-GW
S1
Internet
eNodeB
The MME is the ‘control plane’ entity responsible for the following tasks:
Subscriber mobility and session management signaling. This includes tasks such as authentication, establishment of radio bearers, handover support between different eNodeB and to/from different radio networks (e.g. GSM, UMTS).
Location tracking for mobile devices in idle mode, that is while no radio bearer is established because they have not exchanged data packets with the network for a prolonged amount of time.
Selection of a gateway to the Internet when the mobile requests the establishment of a session, that is when it requests an IP address from the network.
48 Beyond 3G – Bringing Networks, Terminals and the Web Together
The Serving Gateway is responsible for the ‘user plane’, that is, for forwarding IP packets between mobile devices and the Internet. As already discussed in the section on UMTS, IP tunnels are used in the radio access network and core network to flexibly change the route of IP packets when the user i s handed over from one cell to another while moving. The GPRS Tunneling Protocol (GTP) is reused for this purpose and the mechanism is the same as shown for UMTS in Figure 2.3. The difference from UMTS is that the tunnel for a user in the radio network is terminated directly in the eNodeB itself and no longer on an intermediate component such as a radio network controller. This means that the BTS is directly connected via an IP interface to the Serving-Gateway and that different transport network technologies such as Ethernet over fiber or optical cable, DSL, microwave, and so on, can be used. In addition, the S1 interface design is much simpler than similar interfaces of previous radio networks, which relied heavily on services of complex lower layer protocols.
As the S1 interface is used for both user data (to the Serving-GW) and signaling data to the MME, the higher layer protocol architecture is split into two different protocol sets: the S1-C (control) interface is used for exchanging control messaging between a mobile device and the MME. As will be shown below, these messages are exchanged over special ‘non-IP’ channels over the air interface and then put into IP packets by the NodeB before they are forwarded to the MME. User data, however, is already transferred as IP packets over the air interface and these are forwarded via the S1-U (user) protocol to the Serving­Gateway. The S1-U protocol is an adaptation of the GTP from GPRS and UMTS (cf. Figure 2.3).
If the MME and the Serving Gateway are implemented separately, the S11 interface is used to communicate between the two entities. Communication between the two entities is required, for example for the creation of tunnels, when the user attaches to the network, or for the modification of a tunnel, when a user moves from one cell to another.
Unlike in previous wireless radio networks, where one access network gateway (SGSN) was responsible for a certain number of radio network controllers and each radio network controller in turn for a certa in number of base stations, the S1 interface supports a meshed architecture. This means that not only one but several MMEs and Serving-Gateways can communicate with each eNodeB and the number of MMEs and Serving-Gateways can be different. This r educes the number of inter-MME handovers when users are moving and allows the number of MMEs to evolve independently from the number of Serving-Gateways, as the MME’s capacity depends on the signaling load and the capacity of the Serving-Gateway depends on the user traffic load. These can evolve differently over time, which makes separation of these entities interesting. A meshed architecture of the S1 interface also adds redundancy to the network. If, for example one MME fails, a second one can take over automatically if it is configured to serve the same cells. The only impact of such an automatic failure recovery is that users served by the failed MME have to register to the network again. How the meshed capabilities of the S1 interface are used in practice depends on the policies of the network operator and on the architecture of the underlying transport network architecture.
Beyond 3G Network Architectures 49
2.3.2.3 Gateway to the Inter net
As in previous network architectures, a router at the edge of the wireless core network hides the mobility of the users from the Internet. In LTE, this router is referred to as the Packet Data Network (PDN)-Gateway and fulfills the same tasks as the GGSN in UMTS. In addition to hiding the mobility of the users, it also administers an IP address pool and assigns IP addresses to mobiles registering to the network. Depending on the number of users, a network has several PDN-Gateways. The number depends on the capabilities of the hardware, the number of users and the average amount of data traffic per user. As shown in Figure 2.13, the interface between the PDN-GW and the MME/Serving-GWs is referred to as S5. Like the interface between the SGSN and the GGSN in UMTS, it uses the GTP-U (user) protocol to tunnel user data from and to the Serving-Gateways and the GTP-S (signaling) protocol for the initial establishment of a user data tunnel and subsequent tunnel modifications when the user moves between cells that are managed by different Serving-GWs.
2.3.2.4 Interface to the User Database
Another essential interface in LTE core networks is the S6 interface between the MMEs and the database that stores subscription information. In UMTS, this database is referred to as the Home Location Register. In LTE, the HLR is reused and has been renamed the Home Subscriber Server (HSS). Essentially, the HSS is an enhanced HLR and contains subscription information for GSM, GPRS, UMTS, LTE and the IP Multimedia Subsystem (IMS), which is discussed in Chapter 4. Unlike in UMTS, however, the S6 interface does not use the SS-7-based MAP (Mobile Application Part) protocol, but the IP-based Diameter protocol. The HSS is a combined database and it is used simulta­neously by GSM, UMTS and LTE networks belonging to the same operator. It therefore continues to support the traditional MAP interface in addition to the S6 interface for LTE and also the interfaces required for the IMS as discussed in Chapter 4.
2.3.2.5 Moving Between Radio Technologies
In practice, most network operators deploying an LTE network already have a GSM and UMTS network in place. As the coverage area of a new LTE network is likely to be very limited at first, it is essential that subscribers can move back and forth between the different access network technologies without losing their connection and assigned IP address. Figure 2.14 shows how this is done in practice when a user roams out of the coverage area of an LTE network and into the coverage area of a UMTS network of the same network operator. When the user moves out of the LTE coverage area, the mobile device reports to the eNodeB that a UMTS (or GSM) cell has been found. This report is forwarded to the MME which contacts the responsible 3G (or 2G) SGSN and requests a handover procedure. The interface used for this purpose is referred to as S3 and is based on the protocol used for inter-SGSN relocation procedures. As a consequence, no software modifications are required on the 3G SGSN to support the procedure. Once the 3G radio network has been prepared for the handover, the MME sends a handover
50 Beyond 3G – Bringing Networks, Terminals and the Web Together
SGi
PDN-GW
S5
MME Serving-GW
S1
LTE eNodeB
Figure 2.14 LTE and UMTS interworking.
Internet
new tunnel (via S5 and S4)
S4
S3
3G SGSN
RNC
Handover
UMTS NodeB
command to the mobile device via the eNodeB. After the handover has been executed, the user data tunnel between the Serving-GW and the eNodeB is re-routed to the SGSN. The MME is then released from the subscriber management, as this task is taken over by the SGSN. The Serving-GW, however, remains in the user data path via the S4 interface and acts as a 3G GGSN from the point of view of the SGSN. From the SGSN’s point of view, the S4 interface is therefore considered to be the 3G Gn interface between the SGSN and the GGSN.
2.3.2.6 The Packet Call Becomes History
A big difference of LTE from GSM and UMTS is that mobile devices will always be assigned an IP address as soon as they register to the network. This has not been the case with GSM and UMTS because 2G, 3G and 3.5G devices are still mostly used for voice telephony and so it makes sense to attach to the network without requesting an IP address. In LTE networks, however, a device without an IP address is completely useless. Hence, the LTE network attach procedure already includes the assignment of an IP address. From the LAN/WLAN point of view this is nothing new. From a cellular industry point of view, however, this is revolutionary. The GPRS and UMTS procedure of ‘establishing a packet call’, a term coined with the old thinking of establishing a circuit-switched connection with a voice call in mind, will therefore become a thing of the past with LTE. Many people in the industry will have to change their picture of the mobile world to accommodate this.
Beyond 3G Network Architectures 51
2.3.3 Air Interface and Radio Network
While the general LTE network architecture is mainly a refinement of the 3G network architecture, the air interface and the radio network have been redesigned from scratch. In the 3GPP standards, a good place to start further research beyond what is covered below is TS 36.300 [19].
2.3.3.1 Downlink Data Transmission
For transmission of data over the air interface, it was decided to use a new transmission scheme in LTE which is completely different from the CDMA approach of UMTS. Instead of using only one carrier over the broad frequency band, it was decided to use a transmission scheme referred to as Orthogonal Frequency Division Multiple Access, or OFDMA for short. OFDMA transmits a data stream by using several narrow-band subcarriers simultane ously, for example 512, 1024, or even more, depending on the overall available bandwidth of the channel (e.g. 5, 10, 20 MHz). As many bits are transported in parallel, the transmission speed on each subcarrier can be much lower than the overall resulting data rate. This is important in a practical radio environment in order to minimize the effect of multipath fading created by slightly different arrival times of the signal from different directions. The second reason this approach was selected was because the effect of multipath fading and delay spread becomes independent of the amount of bandwidth used for the channel. This is because the bandwidth of each subcarrier remains the same and only the number of subcarriers is changed. With the previously used CDMA modulation, using a 20 MHz carrier would have been imprac­tical, as the time each bit was transmitted would have been so short that the interference due to the delay spread on different paths of the signal would have become dominant.
Figure 2.15 shows how the input bits are first grouped and assigned for transmission over different frequencies (subcarriers). In the example, 4 bits (representing a 16QAM modulation) are sent per transmission step per subcarrier. A transmission step is also referred to as a symbol. With 64QAM modulation, 6 bits are encoded in a single symbol, raising the data rate further. On the other hand, encoding more bits in a single symbol makes it harder for the receiver to decode the symbol if it was altered by interference. This is the reason why different modulation schemes are used depending on transmission conditions.
In theory, each subcarrier signal could be generated by a separate transmission chain hardware block. The output of these blocks would then have to be summed up and the resulting signal could then be sent over the air. Because of the high number of subcarriers used, this approach is not feasible. Instead, a mathematical approach is taken as follows. As each subcarrier is transmitted on a different frequency, a graph which shows the frequency on the x-axis and the amplitude of each subcarrier on the y-axis can be constructed. Then, a mathematical function called Inverse Fast Fourier Transformation (IFFT) is applied, which transforms the diagram from the frequency domain to the time domain. This diagram has the time on the x-axis and represents the same signal as would have been generated by the separate transmission chains for each subcarrier when summed up. The IFFT thus does exactly the same job as the separate transmission chains for each subcarrier would do, including summing up the individual results.
52 Beyond 3G – Bringing Networks, Terminals and the Web Together
t
OFDM(A)
1101110001
f
a
11
IFFT
01
b
c
11
d
00
01
e
A
A
Modulation & Amplification
t
1101110001
abcde
a
b
c
d
e
A
detect
detect
detect
detect
detect
abcde
f
t
f
FFT
A
f
Amplification &
De-modulation
t
Figure 2.15 Principles of OFDMA for downlink transmissions.
On the receiver side, the signal is first demodulated and amplified. The result is then treated by a fast Fourier transformation function which converts the time signal back into the frequency domain. This reconstructs the frequency/amplitude diagram created at the transmitter. At the center frequency of each subcarrier a detector function is then used to generate the bits originally used to create the subcarrier.
The explanation has so far covered the Orthogonal Frequency Division aspect of OFDMA transmissions. The Multiple Access (MA) part of the abbreviation refers to the fact that the data sent in the downlink is received by several users simultaneously. As will be discussed later, control messages inform mobile devices waiting for data which part of the transmission is addressed to them and which part they can ignore. This is, however, just a logical separation. On the physical layer, this only requires that modula­tion schemes ranging from QPSK over 16QAM to 64QAM can be quickly changed for different subcarriers in order to accommodate the different reception conditions of subscribers.
2.3.3.2 Uplink Data Transmission
For data transmission in the uplink direction, 3GPP has chosen a slightly different modulation scheme. OFDMA transmission suffers from a high Peak to Average Power Ratio (PAPR), which would have negative consequences for the design of an embedded mobile transmitter; that is, when transmitting data from the mobile terminal to the
Beyond 3G Network Architectures 53
network, a power amplifier is required to boost the outgoing signal to a level high enough to be picked up by the network. The power amplifier is one of the biggest consumers of energy in a device and should therefore be as power-efficient as possible to increase the battery life of the device. The efficiency of a power amplifier depends on two factors:
The amplifier must be able to amplify the highest peak value of the wave. Due to silicon constraints, the peak value determines the power consumption of the amplifier.
The peaks of the wave, however, do not transport any more information than the average power of the signal over time. The transmission speed therefore does not depend on the power output required for the peak values of the wave but rather on the average power level.
As both power consumption and transmission speed are of impor tance for designers of mobile devices, the power amplifier should consume as little energy as possible. Thus, the lower the difference between the PAPR, the longer is the operating time of a mobile device at a certain transmission speed compared with devices that use a modulation scheme with a higher PAPR.
A modulation scheme similar to basic OFDMA, but with a much better PAPR, is SC-FDMA (Single Carrier-Frequency Division Multiple Access). Due to its better PAPR, it was chosen by 3GPP for transmitting data in the uplink direction. Despite its name, SC-FDMA also transmits data over the air interface in many subcarriers, but adds an additional processing step as shown in Figure 2.16. Instead of putting 2, 4 or
SC-FDMA
1101110001
detect
1101110001
t
FFT
IFFT
01
c
10
d
A
abcde
c
d
A
abcde
01
10
f
IFFT
f
f
FFT
f
AtA
t
A
0
0
0
Modulation & Amplification
t
Amplification &
De-modulation
t
Figure 2.16 SC-FDMA modulation for uplink transmissions.
54 Beyond 3G – Bringing Networks, Terminals and the Web Together
6 bits together as in the OFDM example to form the signal for one subcarrier, the additional processing block in SC-FDMA spreads the information of each bit over all the subcarriers. This is done as follows: again, a number of bits (e.g. 4 representing a 16QAM modulation) are grouped together. In OFDM, these groups of bits would have been the input for the IDFT. In SC-FDMA, however, these bits are now piped into a Fast Fourier Transformation (FFT) function first. The output of the process is the basis for the creation of the subcarriers for the following IFFT. As not all subcarriers are used by the mobile station; many of them are set to zero in the diagram. These may or may not be used by other mobile stations.
On the receiver side the signal is demodulated, amplified and treated by the fast Fourier transformation function in the same way as in OFDMA. The resulting amplitude diagram, however, is not analyzed straight away to get the original data stream, but fed to the inverse fast Fourier transformation function to remove the effect of the additional signal processing originally done at the transmitter side. The result of the IFFT is again a time domain signal. The time domain signal is now fed to a single detector block which recreates the original bits. Therefore, instead of detecting the bits on many different subcarriers, only a single detector is used on a single carrier.
The differences be tween OFDM and SC-FDMA can be summarized as follows: OFDM takes groups of input bits (0s and 1s) to assemble the subcarriers which are then processed by the IDFT to get a time signal. SC-FDMA in contrast first runs an FFT over the groups of input bits to spread them over all subcarriers and then uses the result for the IDFT which creates the time signal. This is why SC-FDMA is sometimes also referred to as FFT spread OFDM.
2.3.3.3 Physical Parameters
For LTE, the following physical parameters have been selected:
Subcarrier spacing, 15 kHz.
OFDM symbol duration, 66.667 ms;
Standard cyclic prefix: 4.7ms. The cyclic prefix is transmitted before each OFDM
symbol to prevent inter-symbol interference due to different lengths of several trans-
mission paths. For difficult environments with highly diverse transmission paths a
longer cyclic prefix of 16.67 ms has been specified as well. The downside of using a
longer cycli c prefix, however, is a reduced user data speed since the symbol duration
remains the same.
The selected subcarrier spacing and symbol duration compensate for detrimental effects on the signal such as the Doppler effect (frequency shift) due to the mobility of sub­scribers. The parameters have been chosen to allow speeds of beyond 350 km/h.
To be flexible with bandwidth allocations in different countries around the world, a number of different channel bandwidths have been defined for LTE. These range from
1.25 MHz on the low end to 20 MHz on the high end. Table 2.1 shows the standardized transmission bandwidths, the number of subcarriers used for each and the FFT size (the number of spectral lines) used at the receiver side to convert the signal from the tim e to
Beyond 3G Network Architectures 55
Table 2.1 Defined bandwidths for LTE.
Bandwidth Number of subcarriers FFT size
1.25 MHz 76 128
2.5 MHz 151 256 5 MHz 301 512 10 MHz 601 1024 15 MHz 901 1536 20 MHz 1201 2048
the frequency domain. In practice, it is expected that operators using LTE will deploy networks in the frequency bands already available today for GSM and UMTS but use bandwidths of at least 10 MHz, since there is no speed advantage of using LTE in a 5 MHz band over HSPAþ. The smaller bandwidths of 1.25MHz and 2.5 MHz were specified for operators with little spectrum or for operators wishing to ‘re-farm’ some of their GSM spectrum in the 900 MHz band. In practice, however, it is questionable if this would bring a great ben efit since the achievable data rates in such a narrow band are lower than what can be achieved with HSPA today.
In addition to the use of already existing frequency bands, new bands are being made available for B3G wireless technologies. In Europe for example the 2.5 GHz band, also referred to as the IMT extension band, will be opened for LTE and possibly other wireless technologies. As will be shown in Chapter 3, however, there is still sufficient unused bandwidth available in existing bands which might make new bands unattractive for LTE in the near future.
2.3.3.4 From Slots to Frames
Data is mapped to subcarriers and symbols, which are arranged in the time and frequency domain in a resource grid as shown in Figure 2.17. The smallest aggregation unit is referred to as a slot or a resource block and contains 12 subcarriers and seven symbols on each subcarrier in case the default short cyclic prefix is used. The symbol time of 66.67 ms and the 4.7 ms cyclic prefix multiplied by 7 results in a slot length of 0.5 ms. In case the long cyclic prefix has to be used, the number of symbols per slot is reduced to six, again resulting in a slot length of 0.5 ms. The grouping of 12 subcarriers together results in a resource block bandwidth of 180 kHz. As the total carrier band width used in LTE is much larger (e.g. 10 MHz), many resource blocks are transmitted in parallel.
Two slots are then grouped into a subframe, which is also referred to as a Transmit Time Interval (TT I). In case of Time Division Duplex (TDD) operation (uplink and downlink in the same band), a subframe can be used for either uplink or downlink transmission. It is up to the network to decide which subframes are used for which direction. Most networks, however, are likely to use Frequency Division Duplex (FDD), which means that there is a separate band for uplink and downlink transmission. Here, all subframes of the band are dedicated to downlink or to uplink transmissions.
56 Beyond 3G – Bringing Networks, Terminals and the Web Together
frequency
1 frame (10 ms) = 10 sub-frames (1ms)= 20 slots (0.5ms)
next frame
180 kHz
t
=
1 symbol =
1 resource element
(15 kHz)
1 resource block= 1 slot (0.5 ms) =
12 sub-carriers * 7 symbols
Figure 2.17 The LTE resource grid.
1 sub-frame
Ten subframes are grouped together to form a single radio frame, which has a duration of 10 ms. Afterwards, the cycle repeat s with the next frame. This is important for mobile devices because broadcast information (e.g. uplink bandwidth assignments) is always transmitted at the beginning of a frame.
The smallest amount of resource elements (symbols) that can be allocated to a single mobile device at an instant in time is two resource blocks, which equals one subframe or one transmit time interval. To increase the data rate for the mobile device, the scheduler in the network can concatenate several resource blocks in both the time and the frequency direction. Since there are many resource blocks being transmitted in parallel, it is also possible to schedule several mobile devices simultaneously, each listening to different subcarriers.
2.3.3.5 Reference Symbols, Signals and Channels
Not all resource elements of a resource block are used for transmitting user data. Especially around the center frequency, some resource elements are used for other purposes, as described below.
To enable a mobile device to find the network after power on and when searching for neighboring cells, some resource elements are used for pilot or reference symbols in a predefined way. While data is transferred, pilot symbols are used by the mobile device for downlink channel quality measurements and, since the content of the resource element
Beyond 3G Network Architectures 57
is known, to estimate how to recreate the original signal that was distorted during transmission.
For the transmission of higher layer data, LTE re-uses the channel concept of UMTS as shown in Figures 2.8 and 2.9. Compared with UMTS, however, all devices use the shared channel on the physical layer. The LTE channel model is therefore much simpler than that of UMTS. LTE also re-uses the concept of logical channels (what is trans­mitted), transport channels (how is it transmitted) and physical channels (air interface) to separate data transmission over the air interface from the logical representation of data. Figure 2.18 shows the most important channels that are used in LTE and how they are mapped to each other.
Downlink Uplink
logical
channels
(what)
BCCH
DTCH
DCCHPCCH
CCCH DCCH DTCH
transport channels
(how)
physical channels
PCH
Figure 2.18 LTE uplink and downlink channels.
BCH
DL-SCH
PDSCHPBCH
UL-SCH
PUSCH
RACH
PRACH
2.3.3.6 Downlink: Broadcast Channel
While no data is transmitted and the mobile is in idle state, it listens to two logical channels. The logical Broadcast Control Channel (BCCH) is used by the network to transmit system information to mobile devices such as the network and cell , that is, which resource blocks and resource elements to find other channels, how the network can be accessed, and so on. The basic parameters sent on the BCCH are mapped to the BCH transport channel and the Physical Broadcast Channel (PBCH). The PBCH is then mapped to dedicated resource elements in the subchannels of the inner 1.25 MHz of the band. Which resource elements are used for the PBCH is calculated with a mathe­matical formula which generates a certain pattern and thus distributes the broadcast information between different subcarriers over time [20]. In addition, a number of additional resource elements are used for a Synchronization Channel (SCH), which is not shown in Figure 2.19. As the name implies, these resource elements help mobile devices to synchronize to the cell and to find the resource elements on which the broad­cast information can be found. In addition to basic cell parameters, the broadcast channel also carries further information which is necessary, but not essential from the
p
58 Beyond 3G – Bringing Networks, Terminals and the Web Together
first transmission path
Direct line of sight blocked
obstacle
BS
second transmission
ath
Figure 2.19 Principle of MIMO transmissions. (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
obstacle
obstacle
obstacle
MS
start. To save bandwidth and to be flexible in the future, this information is not carried on the PBCH but on the Physical Downlink Shared Channel (PDSCH) instead, which is also used for transferring user data (IP packets). A pointer on the PBCH informs mobiles where to find the broadcast information on the PDSCH.
2.3.3.7 Downlink: Paging Channel
The paging channel is used to contact mobile devices in an idle state when a new IP packet arrives in the core network from the Internet and needs to be delivered to the mobile device. In idle state, which is usually entered after a prolonged period of inactivity, only the tracking area (i.e the identity for a group of cells) where the mobile is located is known. The paging message is then sent into all cells of this group. When the mobile device receives the message, it establishes a connection with the network again, a bearer is set up and the packet is delivered. For services such as instant messaging, push e-mail and VoIP, paging for incoming IP packets is quite common. Such applications, if pro­grammed properly and no network address translation firewalls are used, have a logical connection with a server in the network but are dormant until either the user invokes a new action or the application is contacted by the network-based server, for example because of a new instant message coming in [21]. As can be seen in Figure 2.18, there is no physical channel dedicated for paging mess ages. Instead, paging messages are sent on the downlink shared channel and a pointer on the logical broadcast channel indicates where and when the paging messages can be found on the shared channel. The broadcast cycle for paging messages a mobile device needs to listen to is usually in the order of 1–2 s. This is a good balance between quick delivery of an incoming packet and power consumption of a mobile device while not being actively used.
Beyond 3G Network Architectures 59
2.3.3.8 Downlink and Uplink: Dedicated Traffic and Control Channels and Their
Mapping to the Shared Channel
From a logical point of view, user data and RRC control messages are transferred via dedicated traffic channels and the dedicated control channels. Each mobile device has its own dedicated pair of these channels. As can be seen in Figure 2.18, both channels are multiplexed to a single physical downlink or uplink shared channel (PDSCH, PUSCH) which are used for all devices. Higher softwar e layers are therefore independent of the physical implementation of the air interface.
2.3.3.9 Downlink: Physical Layer Control Channels
In addition to the previously mentioned channels, there are a number of additional physical layer control channels which are required to exchange physical layer feedback information. As these channels only carry lower layer control information and are originated by the base station and not the network behind them, they are not shown in Figure 2.18.
To inform mobile devices which resource blocks are assigned to them for transmitting in the uplink direction, the physical downlink shared channel is always accompanied by a Physical Downlink Control Channel (PDCCH). In addition, this channel informs mobile devices about the resource allocation of the PCH and the downlink shared channel.
Since the amount of data carried on the PDCCH varies, the number of OFDM symbols assigned to the physical control channel is broadcast via the Physical Control Format Indicator Channel (PCFICH). Finally, the Physical HARQ Indicator Channel (PHICH) carries acknowledgements for proper reception of uplink data blocks. The HARQ acknowledgment functionality used in LTE is similar to that used in UMTS. For details, see Section 2.2.4.
2.3.3.10 Uplink: Physical Layer Control Channels
In the uplink direction there are two physical layer control channels: The physical uplink control channel (PUCCH) is a per device channel and carries the following information:
HARQ acknowledgments for data frames received from the network. (cf. Section 2.2.4).
Scheduling requests from the mobile to inform the network that further uplink
transmit opportunities should be scheduled, as there is more data in the output buffer.
Channel Quality Indications (CQI) to the network, so the base station can determine
which modulation and coding to use for data in the downlink direction. CQI informa-
tion is also important for the scheduler in the base station, as it can decide to
temporarily halt data transmission to users in a temporary deep signal fading situation
where it is likely that data cannot be received correctly anyway.
The second control channel used in the uplink direction is the Physical Random Access Channel (PRACH). It is used when no bearer is established in the uplink direction to request new uplink transmission opportunities. It is also used when the mobile wants to establish a bearer for the first time, after a long timeout or in response to a paging from the network.
60 Beyond 3G – Bringing Networks, Terminals and the Web Together
2.3.3.11 Dynamic and Persistent Scheduling Grants
The packet scheduler in the base station decides which resource blocks of the physical downlink and uplink channels are used for which mobile device. This way, the base station controls both uplink and downlink transmissions for each mobile and is therefore able to determine how much bandwidth is available to a mobile device. Input parameters for the scheduler are, for example the current radio conditions as seen from each device, so the data transfer rate can be increased or decreased to mobiles temporarily experien­cing exceptionally good or bad radio conditions. Other input parameters are the quality of service parameters for a connection and the maximum bandwidth granted by the operator to a mobile device, based on the user’s subscription.
There are two types of capacity grants for uplink data transmissions: dynamic grants are announced once and are valid for one or more transmit time intervals. Afterwards, the network has to issue a new grant for additional transmit opportunities or for the mobile to receiver further data in the downlink direction. Dynamic grants are useful for data that arrives in a bursty fashion, like during Web browsing, for example, and sporadic downloading of content.
Applications such as voice and video calls require a constant bandwidth and as little variation as possible in the time difference between two adjacent packets (jitter). For such applications, the base station can also issue persistent grants which are given once and are then valid for all subsequent transmit time intervals. This way, no signaling resources are required to constantly re-assign air interface resources while a voice or video call is ongoing. This increases the overall efficiency of the cell and increases the amount of bandwidth that is available for user data. The issue arising with persistent grants is how the base station can know when to use this type of assignment. One way to achieve this is to base the decision on the connection’s quality of service req uirements which are signaled to the network during bearer establis hment. This works well for applications which require a constant bandwidth and are based on the IP Multimedia Subsystem. In addition to the Quality of Service (QoS) signaling initiated by the mobile device, the IMS has a connection to the transport network and can influence the bearer as well. This is discussed in more detail in Chapter 4. For Internet-based voice and other applications that are not using the IMS, however, using persistent grants is much more difficult. It is likely that such applications are used over a default bearer which has no guaranteed bandwidth and latency as it is used simultaneously for other applications on the same device such as Web browsing. In practice, it remains to be seen if schedulers will also take a look at the bandwidth usage of a mobile device over time and decide on this basis to use persistent or dynamic grants.
2.3.3.12 MIMO Transmission
So far, this chapter has focused on data transmission via a single spatial stream between a transmitter and receiver. Most wireless systems today operate in this mode and a second transmitter on the same frequency is seen as unwanted interference that degrades the channel. In practice, however, it can be observed that even a single signal is reflected and scattered by objects in the transmission path and that the other end receives several copies of the original signal from different angles at slightly different times. For simple wireless
Beyond 3G Network Architectures 61
transmission technologies, these copies are also unwanted interference. LTE, however, makes use of scattering and reflection on the transmission path by transmitting several independent data streams via individual antennas. The antennas are spaced at least half a wavelength apart, which in itself creates individual transmissions which behave differ­ently when they meet obstacles in the transmission path. On the receiver side, the different data streams are picked up by independent antenna and receiver chains. Transmitting severa l independent signals over the same frequency band is also referred to as Multiple Input Multiple Output, and Figure 2.19 shows a simplified graphical representation. In practice, this means that several LTE resource grids, as shown in Figure 2.17, are sent over the same frequency at the same time but via different antennas.
The LTE standard specifies two and four individual transmissions over the same band,
which requires two or four antennas at both the transmitter and receiver side respectively. Consequently, such transmissions are referred to as 2 2 MIMO and 4 4 MIMO. In practice, 2 2 MIMO is likely to be used at first, because of size constraints of mobile devices and due to the fact that antennas have to be spaced at least half a wavelength apart. Furthermore, most mobile devices support several frequency bands, each usually requiring its own set of antennas in case MIMO operation is supported in the band. More details on this topic will be discussed in Chapter 3 from a capacity point of view and in Chapter 5 from a mobile hardware point of view. On the network side, 2 2 MIMO transmissions can be achieved with a ‘single’ cross polar antenna that combines two antennas in a way that each antenna transmits a separate data stream with a different polarization (horizontal and vertical).
While Figure 2.19 depicts the general concept of MIMO transmission, it is inaccurate at the receiver side, as each antenna receives not only a single signal but the combination of all signals as they overlap in space. It is therefore necessary for each receiver chain to calculate a channel propagation that takes all transmissions into account in order to separate the different transmissions from each other. The pilot carriers mentioned above are used for this purpose. The characteristics required for these calculations are the gain, phase and multipath effects for each independent transmission path. A good mathema­tical introduction is given in [22].
As MIMO channels are separate from each other, 2 2 MIMO can increase the overall data rate by two and 4 4 MIMO by four. This is, however, only possible under ideal signal conditions. MIMO is thus only used for downlink transmissions since the base station transmitter is less power-constrained than the uplink transmitter. In less favor­able transmission conditions, the system automatically falls back to single stream trans­mission and also reduces modulation from 64QAM, to 16QAM or even QPSK. As has been shown in the previous section on HSPAþ, there is also a tradeoff between higher order modulation and MIMO use. Under less than ideal signal conditions, MIMO transmission is therefore only used with 16QAM modulation, which fails to double the data rate compared with a single stream transmission using 64QAM.
In the uplink direction, it is difficult for mobile devices to use MIMO due to their limited antenna size and output power. As a result, uplink MIMO is currently not part of the LTE standard. The uplink channel itself, however, is still suitable for uplink MIMO transmissions. To fully use the channel, some companies are thinking about implement­ing collaborative MIMO in the future, also known as multiuser MIMO [23]. Here, two mobile devices use the same uplink channel for their resource grid. At the base station
62 Beyond 3G – Bringing Networks, Terminals and the Web Together
side, the two data streams are separated by the MIMO receiver and treated as two transmissions from independent devices rather than two transmissions from a single device that have to be combined. While this will not result in higher transmission speeds per device, the overall uplink capacity of the cell is significantly increased.
2.3.3.13 LTE Throughput Calculations
Based on the radio layer parameters introduced in this section, the physical layer throughput of an LTE radio cell can be calculated as follows: the transmission time per symbol is 73.167 ms (66.667 ms for the symbol itself þ 4.7 ms for the cyclic prefix), the highest modulation order is 64QAM (6 bits per symbol) and there are 1201 subcarriers in a 20 MHz band:
Physical speed ¼ð1=0:000 073 167Þ
6
1201 ¼ 98:487:022 bit=s ði:e: about 100 Mbit=sÞ
When 2 2 MIMO is used, the physical layer speed doubles to about 200 Mbit/s and in case 4 4 MIMO is used for transmission, the theoretical data transmission speed is 400 Mbit/s, based on a 20 MHz channel.
These values are usually quoted in press releases. However, as already discussed for HSPA, these raw physical layer transmission speeds are not reached in practice for a variety of reasons:
64QAM modulation can only be used very close to the base station. For the majority of
users served by a cell, 16QAM (4 bits per symbol) or QPSK (2 bits per symbol) is more
realistic.
Error detection and correction bits (coding) are usually added to the data stream as
otherwise the bit error rate over the air interface would become too high. Under
average signal conditions it is common to see coding rates of 1/3. In practice, the
coding overhead is thus in the range of 25–30%.
Retransmissions – with a very conservative transmission strategy, the coding described
above is sufficient to correct most transmission errors. In practice, however, more
aggressive transmission strategies are used to make the best use of air interface
resources. This usually results in air interface packet retransmission rates in
the order of 20%.
There is a significant overhead from pilot channels and control channels such as the
broadcast channel and the dedicated signaling channels per user to acknowledge the
correct reception of data packets and to convey signal quality measurement results.
Many of those channels are transmit ted with a lower order modulation so even devices
in very unfavorable signal conditions can receive the information.
In many cases, less than 20 MHz of bandwidth is available for LTE.
When using MIMO, the modulation order has to be reduced under less than ideal
transmission conditions.
The overall capacity of the cell has to be shared by all users.
The interference caused by transmissions of neighboring cells on the same frequency
band has a further detrimental effect.
Beyond 3G Network Architectures 63
In practice, it is therefore likely that the throughput per cell is only about 30–50% of the theoretical values given above. For a cell with a 10 MHz carrier and 2 2 MIMO, an overall cell capacity on the IP layer of 30 Mbit/s may be achieved. A more detailed capacity analysis can be found in Chapter 3.
2.3.3.14 Radio Resource Control
As in UMTS and HSPA, the LTE network controls access to the air interface resources for both the uplink and the downlink. As there is no longer a central node in the radio network for the administration of resources, the base stations themselves are now responsible for the following tasks:
Broadcasting of system information.
Connection management – the mobile devices and the network use control channels such as the random a ccess channel, the paging channel and the dedicated control channels to exchange RRC messages. The first RRC messages exchanged when acces­sing the network for the first time, or after a long time of inactivity, are connection establishment messages. The eNodeB is then responsible for setting up a logical signaling bearer to the device via the shared uplink and downlink channel or denying the request where the system is overloaded. Connection managem ent also includes the establishment of dedicated bearers, again over the shared physical channel, based on the quality of service parameters of the user’s subscription.
Measurement control – as users move, the radio environment is very dynamic and devices therefore need to report signal strength measurements of the current and neighboring cells to the network.
Mobility procedures – based on signal measurements of the mobile device, the eNodeB can initiate a handover procedure to another cell or even to another radio network such as UMTS or GSM/GPRS where the LTE coverage area is left.
2.3.3.15 RRC Active State
To minimize the use of resources in the network and to conserve the battery power of mobile devices, there are several connection states. While data is exchanged between the network and a mobile device, the RRC connection is in the active state. This means the network can assign resources to the device on the shared channel at any time and data can be inst antly transmitted. The mobile remains in active state even if no data is transferred for some time, for example after the content of a Web page has been fully loaded. This ensures instant package transmission without any further resource control overhead, for example when the user clicks on a link.
While in full active state, the mobile has few opportunities to deactivate its receiver which has a negative impact on the battery capacity. After some time of inactivity, the network can thus decide to activate a Discontinuous Reception Mode (DRX) while the mobile is still in active state. This means that the mobile only has to listen to downlink bandwidth assignments and control commands periodically and can switch off its receiver at all other times. The DRX interval is flexible and can range from milliseconds to seconds.
64 Beyond 3G – Bringing Networks, Terminals and the Web Together
Even while in DRX mode, mobilityis still controlled by the network. This means that the mobile device has to continue sending signal measurement results to the network when a defined high or low signal threshold for the current cell or a neighboring cell is met. The eNodeB can then at any time initiate a handover procedure to another cell if required.
2.3.3.16 RRC Idle State
If no packets have been transmitted for a prolonged amount of time, the eNodeB can put the connection to a user in RRC Idle state. This means that, while the logical connection to the network and the IP address is retained, the radio connection is removed. The MME is informed of this state change as well, as IP packets arriving from the Internet can no longer be delivered to the radio network. As a consequence, on receipt of IP packets the MME needs to send a paging message to the mobile device, which leads to the re-establishment of a radio bearer. In case the mobile device needs to send an IP packet while in RRC idle state, for example because the user has clicked on a link on a Web page after a long time of inactivity, it also has to request the establishment of a new radio bearer before the packet can be transmitted.
Furthermore, the network no longer controls the mobility of a device in RRC idle state and the device can decide on its own to move from one cell to another. Several cells are grouped into a tracking area, which is similar to location and routing areas used in UMTS. The mobile only reports a cell change to the network if it selects a cell which belongs to a different tracking area. This means that the network, or more specifically the MME, has to send a paging message via all the cells that belong to the tracking area when a new packet for the device arrives from the Internet.
2.3.3.17 Treatment of Data Packets in the eNodeB
In addition to radio resource specific tasks, the eNodeB is also responsible for several tasks concerning the data packets themselves before they are transmitted over the air interface. To prevent data modification attacks, also referred to as man-in-the-middle attacks, an integrity checksum is calculated for each data packet before it is sent over the air interface. Input to the integ rity checksum algorithm is not only the content of the packet but also an integrity checking key which is calculated from a unique secret key that is shared between the eNodeB and each mobile device. If a message is fraudulently modified on the air interface, it is not possible to append a valid integrity checksum due to the missing key and the message is not accepted by the recipient. Integrity checking applies to IP packets, to radio resource control messages exchanged with the eNodeB and also to mobility and session management messages exchanged with the MME.
In addition to integrity checking, data packets are encrypted before being transmitted over the air interface. Again, the subscriber’s individual shared secret key, stored on the SIM card and the HSS, is used to calculate a ciphering key on both sides of the connection. Data intercepted on the air interface can thus not be decoded as the ciphering key is not known to an attacker. Ciphering applies to IP packets, to radio resource control messages and also to mobility and session management messages, the latter two not being based on IP.
Beyond 3G Network Architectures 65
A task only performed on user data IP packets before they are transmitted over the air interface is header compression. For LTE networks, this feature is very important, especially for real-time applications such as VoIP. As VoIP is very delay-sensitive, typically only 20 ms of speech data is accumulated in a single IP packet. With a data rate of around 12 kbit/s produced by sophisticated speech codecs, such as an Adaptive Multi-Rate (AMR) codec with a good voice quality, each IP packet carries around 32 bytes of data. In addition, there is an overhead of 40 bytes for an IPv4 header, the UDP header and the RTP header. W ith IPv6, the overhead is even larger due to the use of 128 bit IP addresses and additional header fields. This means that there is more overhead per packet than speech data. This greatly inflates the required data rate and therefore significantly reduces the potential number of simultaneous calls per base station. As voice calls are likely to be an important feature for LTE networks, it is necessary to compress IP packet headers before transmission. For LTE, the Robust Header Compression (ROHC) algorithm, originally specified in [24], was selected. Its advan­tages are:
A very good compression ratio. The 40 bytes overhead of various encapsulated
protocols are typically reduced to 6 bytes.
A built-in feedback mechanism detects compression process corruptions as a result of
air interface transmission errors. This allows an immediate restart of the compressor
logic instead of letting the error propagate into the compression process of subsequent
packets, as was the case with previously used header compression algorithms.
The ROHC algorithm not only compresses the IP header but analyzes the IP packet
and also compresses further encapsulated headers such as the UDP header and the
RTP (Real-time Transfer Protocol) header where the data packet contains audio
information.
In order not to focus only on VoIP packets, ROHC is able to detect different header
types in a packet and selects an appropriate overall header compression algorithm for
each packet. The different compression algorithms are referred to as profiles. For
VoIP packets, the RTP profile is used, whi ch compresses the IP header, the UDP
header and the RTP header of the packet. Further profiles are the UDP profile, which
compresses IP and UDP headers (e.g. of SIP signaling messages, cf. Chapter 4), and the
ESP (Encapsulated Security Payload) profile, which is used for compressing headers of
IPsec encrypted packets.
Integrity checking, ciphering and compression are all part of the Packet Data Convergence Protocol (PDCP), which sits below the IP layer and thus encapsulates IP packets. The packet size, howeve r, does not usually increase, as the additional PDCP header overhead is more than made up for by the header compression.
2.3.4 Basic Procedures
An important aspect of LTE, in addition to increasing the available bandwidth over the air interface, is to streamline signaling procedures to reduce delay for procedures such as setting up an initial connection and resuming data transfers from idle state. Figure 2.20
66 Beyond 3G – Bringing Networks, Terminals and the Web Together
Mobile Device
broadcast
to find network
Random Access Procedure
Attach request
Authentication Procedure
Bearer establishment
request
Bearer est. ack.
eNodeB
Attach request
Attach accept
MME
Update location
Insert subscr. data
Insert s. data ack.
Update location ack.
HSS
Create bearer request
Create ack.
Serving-GW
Create bearer request
PDN-GW
Create ack.
Figure 2.20 Attaching to the LTE network and requesting an IP address.
shows the message exchange of a device attaching to the network after it has been switched on until the point an IP address is assigned. Attaching to the network and getting an IP address is, as mentioned before, a single procedu re in LTE as all services are based on IP. It does not make sense, therefore, to attach to the network without requesting an IP address as is the case today in GSM and UMTS networks.
2.3.4.1 Network Search and Broadcasting System Information
The first step in attaching to the network after power on is to find all available networks and select an appropriate network to commun icate with. For this, the mobile performs an initial scan in all frequency bands it supports and tries to find downlink synchroniza­tion signals. As most LTE-capable devices also support 2G and 3G networks such as GSM and UMTS, the network search procedure also includes such networks. As most bands are not dedicated to a single network technology, the mobile must be able to correlate downlink signals to different radio systems. In case of LTE, the mobile searches for synchronization signals which are placed at regular intervals in the center subchan­nels (1.25 MHz) of an LTE carrier. Once these are found and properly decoded, the broadcast channel can be read and the mobile downloads the cell’s complete system information. In most cases, the mobile device starts its search on the last used channel before it was switched off. If the device has not moved since it has powered off, the network is found very quickly. If a network is found, the registration process continues. If not, the process is repeated until the network is found or all supported frequency bands
Beyond 3G Network Architectures 67
have been searched. Where the last used network before the device was powered down is not found, the mobile either selects a network on its own based on the preferences stored on the SIM card or presents the list of detected networks to the user who can then select the network of their choice (e.g. in case of roaming).
2.3.4.2 Initial Contact with the Network
After the broadcast information of a cell has been read and the decision has been made to use the network, the mobile devicecanattempttoestablishaninitial connection by sending a short message on the random access channel. The channel is referred to as a r andom access channel as the network cannot control access to this channel. There i s thus a chance that several devices attempt to send a message simultaneously which results in a network access collision. If this happens, the base station will not be able to receive any of the messages properly. To minimize this possibility, the message itself is only very short and only contains a 5 bit random number. Furthermore, the network offers many random access slots per second to randomize access requests over time. When the network picks up the random access request, it assigns a C-RNTI (Cell-Radio Network Temporary) identifier to the mobile and answers the message with a Random Access Response message. The message contains the 5 bit random number, so the mobile device can correlate the response to the initial message and the C-RNTI, which is used to identify the mobile device from now until the physical connection to the network is released (e.g. after a l onger time of inactivity). In addition, the message contains an initial uplink bandwidth grant, that is a set of resource blocks of the shared uplink channel that the mobile device can use in uplink direction. These resources are then used to send the RRC connection request message that encapsulates an initial attach request.
2.3.4.3 Authentication
When the eNodeB receives the connection request mess age, it forwards the contained attach request to the MME. The MME extracts the user identity from the message which is either the International Mobile Subscriber Identity or a temporary identity that was assigned to the mobile device during a previous connection with the netwo rk. In most cases, a temporary identity is sent which is changed once the user has been authenticated to reduce the number of IMSIs that have to be transmitted before encryption can be activated. If the IMSI of the subscriber is not known to the MME, the Authentication Center in the HSS is queried for authentication information. If a temporary identity is sent that is unknown to the MME, a request is sent to the mobile to send its IMSI instead. Afterwards, the network and mobile authenticate each other using secret private keys which are stored both on the SIM cards and in the authentication center, which is part of the HSS. Once the subscriber is properly authenticated, the eNodeB and the mobile device activate air interface encryption. At the same time, the MME continues the attach process by informing the HSS with an Update Location message that the subscriber is now properly authenticated. The HSS in turn sends the user’s subscription data, for
68 Beyond 3G – Bringing Networks, Terminals and the Web Together
example what types of connections and services the user is allowed to use, to the MME in an Insert Subscriber Data Message. The MM E confirms the reception of the message which in turn terminates the Update Location procedure with an acknowledgement message from the HSS, as shown in Figure 2.20.
2.3.4.4 Requesting an IP Address
In the next step, the MME then requests an IP address for the subscriber from the PDN-GW via the Serving-GW with a Create Bearer Request message. When the PDN GW receives the message it ta kes an IP address from its address pool, creates a subscriber tunnel endpoint and r eturns the IP address to the MME, again via the Serving-GW. Involving the Serving-GW in the process is required, as the user data tunnel is not established between the MME and the PDN-GW but between the Serving-GW and the PDN-GW. Once the MME receives the IP address, it for­wards it to the eNodeB in an Attach Accept message, which is the reply to the initial Attach Request message. The eNodeB in turn forwards the Attach Accept message including the IP address as part of a Radio Bearer Establishment Request Message to the mobile device, which answers with a Radio Bearer Establishment Response message containing an Attach Complete message. The Attach Complete Message is then forwarded to the MME and the mobile device can now commu­nicate with the Internet or the network operator’s internal IP network (e.g. to connect to the IMS).
Despite the many messages being sent back and forth between the different functions in the network, the number of messages exchanged between the mobile device and the network has been reduced compared with GSM and UMTS by performing several tasks with a single message. This should significantly speed up the overall process. Excluding network detection and reading the broadcast channel, the procedure is likely to take only a few hundred milliseconds.
2.3.5 Summary and Comparison with HSPA
At its introduction, LTE competes with already deployed HSPA networks. It is likely, that some network operators will decide to upgrade their HSPA networks and eventually upgrade the network for the use of higher-order modulation, MIMO and Ethernet-based backhaul. In a 5 MHz band, the performance of LTE and HSPAþ is similar, so other reasons are required for operators to add LTE to their already existing GSM and HSPA infrastructure. One reason for adding LTE to a cell site is to increase the available bandwidth for a certain region by off-loading traffic from HSPA to LTE, once LTE devices become more commonplace. As network v endors are offering base stations capable of supporting several radio technologies simultaneously, the move to LTE could come as part of replacing aging base stations. By 2012, for example, many UMTS base stations will have been in the field for almost 10 years and therefore will have to be replaced anyway. Since it is unlikely that HSPA will be directly replaced by LTE due to many devices still ‘only’ being HSPA-capable, multiradio technology base stations will become very interesting for mobile network operators. If LTE is used in the
Beyond 3G Network Architectures 69
same band as the other radio technologies, a single antenna can be used. Therefore, no additional antennas will be required for many base station sites. The existing antenna might be replaced with a new one, however, to enable MIMO transmissions for LTE. When more than a 5 MHz bandwidth is available, LTE can clearly show its advantages over 3G technologies, as LTE radio channels can be easily extended to 10, 15 or even 20 MHz. In addition, the simpler radio and core network with fewer components and new technologies for backhaul transmission from the base station to the rest of the network will lower network operation costs. This will be an interesting driver for network operators, as bandwidth demands keep rising while the revenue per user is flat or even declining. Finally, due to the simplified air interface signaling, LTE is much more suitable for always-on IP connectivity and applications such as instant messaging and push e-mail, which frequently communicate with the network, even while the mobil e device is not actively used.
2.3.6 LTE-Advanced
As discus sed in Section 1.5, LTE is unlikely to meet the transmission speed requirements of the ITU for 4G wireless systems. 3GPP has therefore started to investigate, how an evolution of LTE could meet these requirements. The following list shows some initial ideas for such a system, currently referred to as LTE-Advanced and LTE Plus, which were presented during a 3GPP worksh op in 2008 [25]:
LTE advanced shall be backwards-compatible to LTE (i.e. like HSPA is backwards­compatible to UMTS).
The primary focus should be on low-mobility users in order to reach ITU-Advanced data rates.
Channel bandwidths should be used beyond the 20 MHz currently standardized for LTE (e.g. 50 MHz, 100 MHz).
The number of antennas for MIMO should be increased beyond what is currently specified in LTE.
MIMO should be combined with beamforming.
There should be a further increase in Voice over IP capacity.
Cell edge data rates should be further improved.
Self-configuration of the network should be improved.
Details for these and other possible features for LTE-Advanced can be found in [26]. With LTE still in the specification phase and not yet being widely deployed, working on a further evolution presents a significant challenge to 3GPP members for a number of reasons. First, it is likely that significant work on the original LTE and SAE standards will be required beyond 2008. This leaves little room for significant simultaneous devel­opments. Also, there has not yet been much time to allow for adjustments of the standards based on the experience gained from developing and deploying LTE. Finally, some network operators might decide to halt their LTE plans and focus on improving their current HSPA networks to try to bridge the time until LTE-Advanced systems become avail able.
70 Beyond 3G – Bringing Networks, Terminals and the Web Together
2.4 802.16 WiMAX
2.4.1 Introduction
Another successor to current 3.5G wireless network technologies is WiMAX, a system based on the IEEE (Institute of Electrical and Electronics Engineers) 802.16 air interface standard. Major infrastructure vendors backing this technology are, for example, Intel, Motorola, Nortel, Alcatel-Lucent and Nokia Siemens Networks. As will be shown in this section, WiMAX shares many basic properties with LTE. From a timing perspective, WiMAX has a head start over LTE, as standards activities were started earlier.
While LTE is mainly attractive for incumbent 2G and 3G operators, WiMAX is very appealing to greenfield network operators, that is operators without an already existing network in place. There are several reasons for this:
In most countries, frequency bands for UMTS and LTE have already been sold many
years ago. WiMAX, however, can be operated in so far unused frequency bands that
are still in the process of being auctioned.
The WiMAX network architecture is fully based on IP, which simplifies the network
architecture design and deployment and cuts operating costs compared with current
ATM backhaul-based 3G networks.
New network operators are aiming to offer Internet access and thus compete with fixed
line high-speed Internet solutions such as DSL and TV cable. As a consequence, their
business model is quite different from that of incumbent wireless network operators,
who are still mainly focused on mobile voice services.
WiMAX network equipment is available today, while LTE equipment will only
become commercially available in the 2010–2012 timeframe.
An exception to the rule seems to be the US market, where Sprint, a major incumbent operator, together with Clearwire has decided to use WiMAX due to a lack of a convincing future perspective of its 3G technology.
2.4.2 Network Architecture
2.4.2.1 Small Networks for Stationary Clients
Many early WiMAX network operators have started to deploy small networks based on the early 802.16-2004 standard (previously referred to as 802.16d), which mainly targeted stationary devices with roof-mounted antennas or indoor WiMAX routers with large omnidirectional antennas built in. Such networks require little more than a few base stations and possibly a central server for storing subscription data for subscriber authen­tication purposes and to control access to the network. Since such networks are designed for stationary devices, no handovers of connections between base stations are required. Each cell can therefore act as a little network of its own.
2.4.2.2 Medium to Large Networks and Mobility
More recently, the air interface standard was enhanced to also support mobility of subscribers, including handovers from one base station to another. This version of the
Beyond 3G Network Architectures 71
standard is referred to as 802.16e, or 802.16-2005. Since handovers and mobile subscri­bers require more administration in the network, it became necessary to also specify the network behind the base stations. As the IEEE is only responsible for the air interface standardization, this part was taken over by the WiMAX Forum [27]. Standardizing the radio network and the core network is an important task, since only standardized functionalities and interfaces allow network operators to select compatible components from a wide variety of network vendors. This way, competition between vendors is fostered, which results in lower prices for network equipment. To ensure interoperability of components from different vendors, the WiMAX Forum is also in charge of a certification program for base stations, end user devices and other network equipment.
Figure 2.21 shows how one of the possible WiMAX network infrastructure setups looks like in practice. When compared with the LTE network infrastructure shown in Figure 2.14, there are remarkable similarities. As in LTE, WiMAX base stations commu­nicate with each other for handovers. Furthermore, there is no central element in the radio network, as was the case with UMTS. Instead, WiMAX networks only require gateways between the radio network and the core network, the Access Service Network Gateways (ASN-GW).The ASN-GWs are responsible for user management and mobility.
ASN
BS
R1
(802.16)
R8
R8
BS
BS
BS
BS
BS
R6
ASN
R6
ASN-GW
R8 R4
ASN-GW
R3
R3
Core Network
HA
AAA
IMS
DNS
DHCP
R5 (roaming)
Figure 2.21 WiMAX network architecture. (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
The air interface between mobile devices and the base stat ion is referred to as the R1 reference point or interface. In this section, the two words are used interchangeably. The protocol used over this interface is either 802.16-2004 (formerly 802.16d) for stationary wireless installations or 802.16-2005 (formerly 802.16e) for both stationary and mobile clients. Air interface details are discussed in Section 2.4.3.
In the radio network, base stations are connected to the ASN-GW via the R6 reference point. In the first version of the WiMAX standard, the interface is proprietary, which
72 Beyond 3G – Bringing Networks, Terminals and the Web Together
means that an ASN-GW and all base stations connected to it need to be from the same vendor. In the second version of the standard, the R6 reference point will also be standardized to allow mixed configurations. This further increases competition, which results in more competitive pricing. Like in LTE, the R6 interface be tween the base station and the access gateway is fully based on IP. Consequently, any transport technol­ogy that is capable of carrying IP packets can be used. Since multisector WiMAX base stations are capable of air interface data rates of 30 Mbit/s and beyond, as shown in Chapter 3, suitable transport technologies in the last mile to the base station are Ethernet-based microwave systems, VDSL and fiber connections. Because of the prohi­bitive cost of such connections and their slow sp eed compared with the capabilities of the air interface, 2 Mbit/s E-1-based co nnectivity is not likely to be used.
For smooth handovers of connections between base stations, the R8 reference point has been defined as shown in Figure 2.21. Like the R6 reference point, it is also fully based on the IP protocol. In practice, a base station is only co nnected to the network with a single physical interface. Data between different base stations therefore trave rses one or more routers before reaching another base station. In practice, the overhead created by this is likely to be small, as the amount of user data on the R6 interface is likely to be several orders of magnitude higher than the amount of data exchanged for a handover of a connection on the R8 interface.
2.4.2.3 The Access Service Network Gateway
In WiMAX networks, the gateway between the radio network and the core network is referred to as the Access Service Network Gateway. In principle, it is responsible for the same tasks as the Access Gateway (AGW) in LTE. These are:
subscriber management tasks such as authentication and subscription management;
mobility management to redirect the connection from one cell to another when the user
is moving;
to actively support the handover procedure in case the R8 reference poin t between two
base stations is missing.
2.4.2.4 Authentication and Ciphering
In the WiMAX standard, user devices are referred to as a Customer Premises Equipment (CPE), a term from the 802.16-2004 standard, which mainly addressed stationary equip­ment. The term, however, is still being used with the 802.16-2005 standard and mobile devices alongside the term Mobile Subscriber Station (MSS). When a device is powered on, its first task is to search for available networks and to get an IP address from the user’s home network, or , in the case of roaming, from another suitable network. Before a device is admitted to the network by the ASN-GW, an authentication procedure is required. Unlike 3GPP networks such as UMTS and LTE, WiMAX does not make use of a secret key that is stored on a SIM card and in the network. Instead, authentica­tion is performed with a public/private key pair in addition to an X.509 certificate.
In theory, the keys and the certificate could be stored on a SIM card. In practice, however, this is not the case today. Instead, the keys and certificate are stored in a safe
g
Beyond 3G Network Architectures 73
location in the device itself that cannot be directly accessed to prevent applications from reading the secret private key.
At the beginning of the authentication procedure, the device and the network exchange their public keys with each other, which are then used to derive temporary keys to encrypt further traffic. Data encrypted with a temporary public key can only be decrypted with the corresponding temporary private key which was derived from the secret private key. As private keys are never transmitted over the air interface, it ensures that an attacker cannot decipher or modif y the data. Another benefit of having public/private keys is that the private key of the subscriber is only stored in the client device but not in the network, as the network only requires the public key for encrypting the data. Consequently, no sensitive key information has to be stored on any equipment of the network operator.
In addition to the public key exchange, an additional mechanism is used to ensure that the public key sent by a device is tied to its MAC hardware address. This is done by sending a certificate, signed by a certificate authority, in addition to the public key, as shown in Figure 2.22. The certificate authority signs the certificate by encrypting the device’s public key and MAC hardware address with its private key. When the client device sends its public key together with the certificate, the network then decrypts the certificate with the public key of the certificate authority. Afterwards, it checks if the subscriber’s public key matches the one extracted from the certificate. Furthermore, it is verified that the MAC address, which is part of each data packet, also matches the one given in the certificate. If they match, the client device is authenticated. Tampering with the certificate is not possible since it can only be decrypted by everyone who knows the public key of the certificate authority. However, it cannot be changed and re-encrypted. The certificate also prevents a successful attack by duplicating MAC addresses, as the private key of the original device used in combination with a MAC address is securely stored in the device. Therefore, it cannot be duplicated together with the MAC address.
CA
1. Manufacturer gets a certificate
from CA that links the
MAC address and public key.
CPE
2. During authentication, client
sends certificate to network.
Figure 2.22 WiMAX authentication with the help of a Certificate Authority (CA). (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
3. Network has public key of the trusted CA and can thus verify the certificate.
Network
4. Client is authenticated and the
public key of the client contained in the certificate is used to encrypt the remainin
authentication process.
74 Beyond 3G – Bringing Networks, Terminals and the Web Together
It should be noted that certificates and public/private keys are also used for authenti­cation and encryption of secure Web sessions, such as for example for online shopping or Web banking. Here, a Secure HTTP (HTTPS) session is established instead of a standard HTTP session and the process is very similar to the one described above. During connection establishment, the Web server sends a certificate signed by a certificate authority to the Web browser. The certificate contains the URL of the Web site and the public key. The Web browser compares the URL the user has typed in with the one in the certificate. If both match, the Web browser can be certain that the connection was not redirected by an attacker. To an attacker, the public key in the certificate is worthless, as they do not have the private key to decrypt the information, which is encrypted by the client with the public key.
Both the WiMAX and the HTTPS authentication processes require certificates to be generated by a trusted certificate authority. Trust is established by storing the certificate authority’s public key locally. In the case of WiMAX, the certificate authority’s public key is stored in the device itself. In case of HTTPS, the certificate authority’s public key is stored in the Web browser. In practice, there are many different certificate authorities that can issue certificates. Verisign, for example is a company issuing both HTTPS and WiMAX certificates [28].
2.4.2.5 Client IP Address Assignments and R6 Tunnels
Once a device is authenticated and air interface ciphering has been activated, the ASN­GW is also in charge of assigning an IP address to a device or to request it from a Home Agent (HA) in the core network, as will be discussed below in Section 2.4.2.7.
As the network between the base stations and ASN-GWs is not necessarily owned by the WiMAX network operator, data traffic on the R6 reference point between gateways and base stations should be encrypted. For this purpose, an encrypted IPSec tunnel could be established between each base station and the ASN-GW. The user’s data is thus not only protected on the air interface, but also throughout the radio network up to the ASN-GW.
2.4.2.6 Micro Mobility Management
When a user moves from the coverage area of one base station to another, it is the base station’s task to handover the connection. Both the network and the client device can initiate a handover. In the radio network, this means that the current and new base station communicate with each other over the R8 interface or via the ASN-GW during the handover. Part of the handover process is also to inform the ASN-GW that the location of the subscriber has changed, as user data packets now have to be exchanged over a different IPSec tunnel. Figure 2.23 shows how this works in practice. In the example, base stations and ASN-GWs in the radio network use the 10.x.x.x IP subnet. One tunnel is established between the upper base station, which has been assigned
10.0.0.2 as an IP address, and the ASN-GW (10.0.0.1). Another tunnel is established to the base station in the lower part of the figure (10. 0.0.3). Before the handover, the user’s data, identified by the user’s IP address (195.36.219.196), is sent through the tunnel between 10.0.0.1 (ASN-GW) and the upper base station (10.0.0.2). Once the handover
Beyond 3G Network Architectures 75
10.0.0.2
BS
195.36.219.196
BS
Figure 2.23 Base station and user tunnel for micro mobility management. (Reproduced from
Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
BS tunnel
BS tunnel
10.0.0.3
R6 reference point
ASN-GW
10.0.0.1
Web
server
193.99.144.85
has been performed, the ASN-GW redirects the data flow to the tunnel between itself (10.0.0.1) and the lower base station (10.0.0.3).
2.4.2.7 Macro Mobility Management
Like in larger UMTS and LTE networks, it is required at some point to install several WiMAX ASN-GWs in the network to support a growing number of base stations and users. Once there are several radio network gateways in the network, it is possible that a user will change between cells controlled by different gateways. This means that packets arriving for a user from the Internet can no longer only be routed by default to a single ASN-GW. Instead, it is necessary to introduce core network mobility management as well. In UMTS and LTE, the GPRS Tunneling Protocol (GTP) takes care of this task between the single point of entry to the network (the GGSN in case of UMTS and the PDN in case of LTE) and the radio access network gateway. In WiMAX, it has been decided to use a different approach, as shown in Figure 2.24.
Instead of relying on a proprietary protocol such as GTP, it was decided to use Proxy Mobile IP (Proxy MIP), an already existing IP-based mobility management standard [29]. In principle, Proxy MIP works as follows: when a device requests access to the network, the ASN-GW requests an IP address for the device from the Mobile IP Home Agent. The HA has a pool of IP addresses it is responsible for and all packets arriving from the Internet destined to these IP addresses are always routed to the HA. From this pool, one IP address is assigned to the device and returned to the ASN-GW. The HA notes the IP address of the ASN-GW (64.236.23.28 in Figure 2.24) and begins forwarding all packets arriving from the Internet to the ASN-GW. The ASN-GW in turn forw ards packets it receives for this IP address to the base station, to which the subscriber is currently attached via a micro mobility management tunnel as described above.
g
76 Beyond 3G – Bringing Networks, Terminals and the Web Together
3. MIP tunnel between HA
and ASN-GW (Proxy-MIP)
MS
ASN-GW
ASN-GW
5. IP packet with destination address
195.36.219.196 forwarded through micro mobility
ement tunnels
mana
(195.36.219.196)
64.236.23.28
4. Care-Of IP address (COA) is the end point of the tunnel
Core Network
HA
IP pool
195.36.219.196
2. address for the client device taken from the address pool
1. packets are always delivered to the HA first
Web
server
193.99.144.85
Figure 2.24 ASN-GW mobility management using mobile IP. (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
For subscriber devices using IPv4, the ASN-GW terminates the MIP tunnel to make the process transparent for the subscriber device. This is required since it is not desirable to change the protocol stack of the device. Consequently, the ASN-GW becomes a proxy for the subscriber’s device, which is why the approach is referred to as Proxy MIP.
For devices using only IPv6 addresses, which is expected to become more common­place in the future, no proxy is required, as MIP is part of the protocol stack. This means that the mobile device communicates with the HA in the network on its own instead of leaving the task to the ASN-GW.
2.4.3 The 802.16d Air Interface and Radio Network
Like LTE,WiMAX uses Orthogonal Frequency Division Multiplexing (OFDM) to transmit data overthe air interface. The systems are therefore very similaron the physical layer and this section assumes the reader is familiar with the LTE air interface described in Section 2.3.3.
2.4.3.1 Fixed WiMAX
The first version of the 802.16 air interface standard, referred to as 802.16d or IEEE
802.16-2004 [30], is currently in use to connect devices such as notebooks and PCs to the Internet via WiMAX modems inst alled at home or at the office. It is not compatible with the current 802.16e or 802.16-2005 standards, which were developed later and introduce many enhancements required for mobility. The lifetime and use case scenarios of fixed WiMAX deployments are therefore limited, as devices and network equipment are unlikely to be upgradeabl e to the mobile standard. Also , fixed WiMAX networks do not usually use most of the standardized infrastructure described above, since their
Beyond 3G Network Architectures 77
network architecture is much simpler and also because no support for mobility is required. Nevertheless, this section takes a look at the fixed WiMAX air interface standard as well, because it is used to some degree in practice and forms the basis for mobile WiMAX, which is discussed later.
The big difference of IEEE 802.16-2004 compared with mobile WiMAX (IEEE
802.16-2005) and LTE is that it uses 256 OFDM subcarriers independent of the band­width used for the channel. Out of these, 193 are used for data transmissions. The remaining subcarriers are either unused at the edge of the band or provide pilot signals which are used by devices for channel estimation and filter approximation. Channel bandwidths defined are 1.25, 3, 3.5, 5.5, 7 and 10 MHz. The smaller bandwidths, however, are unlikely to be used in practice, as the resulting throughput is not sufficient to support high-speed Internet access even for a small number of users. Using the same number of subcarriers for all bandwidths means that the symbol transmit time varies depending on the bandwidth. For a bandwidth of 1.25 MHz, for example, the symbol transmit time is 128 ms, while for a 10 MHz deployment, the symbol time is only
22.408 ms. Therefore, the transmission characteristics on the physical layer depend on the bandwidth used for a channel.
Two profiles have been defined by the WiMAX Forum for fixed WiMAX. The wireless Metropolitan Area Network OFDM profile (wirelessMAN-OFDM) is used when a national regulator has officially assigned a frequency band for the use of WiMAX, for example as the outcome of a spectrum auction. Depending on the properties of the assigned band, fixed WiMAX devices are used in either TDD or FDD mode [31].
In TDD mode, the same band is used for uplink and downlink transmissions and the system continuously switches between transmission and reception. The advantage of this approach is that the system can be tuned to reflect the ratio between uplink and down­link traffic. Currently, more bandwidth is required in the downlink direction, which is why more time is allocated for downlink than for uplink transmissions. It should be noted at this point, however, that a 3:1 downlink/uplink ratio on the air interface does not exactly reflect the bandwidth ratio, because uplink transmissions are usually not as efficient due to the limited output power and antenna restrictions of a small device. Furthermore, TDD requires base stations to be tightly synchronized with each other to prevent uplink transmissions of devices in one cell to interfere with downlink transmis­sions of neighboring cells.
In FDD mode, different frequency bands are used for downlink and uplink trans mis­sions. For licensed frequency bands, this is often the preferred transmission mode. The advantage of FDD is that data can be transmitted in uplink and downlink in parallel. Furthermore, no transmission pause is necessary to give devices the necessary time to switch from transmission to reception mode. In addition, FDD transmission allows more sensitive receivers in mobile devices, which benefits overall data rates.
Figure 2.25 shows what an FDD downlink data transmission looks like in WiMAX with the IEEE 802.16-2004 standard for stationary devices. Downlink trans­missions are separated into individual frames with a fixed length between 2.5 and 20 ms. Each frame in turn holds a number of consecutive fields. The first field is the preamble, which has a known bit pattern that mobile devices can use to detect the beginning of a frame. The Frame Control Header (FCH) is next and contains information about the modulation and coding scheme of the first downlink burst that immediately follows.
78 Beyond 3G – Bringing Networks, Terminals and the Web Together
Frame 1 Frame 2 Frame 3 Frame 4 Frame 5
Preamble FCH
Broadcast
Management data for all devices
Figure 2.25 WiMAX 802.16-2004 downlink data transmission. (Reproduced from Communi- cation Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
DL-Burst 1
MAC PDU MAC PDU MAC PDU MAC PDU
A packet for client device x
DL-Burst 2 DL-Burst 3
A packet for client device y
t
BPSK modulation and a 1/2 coding rate are used for the FCH to ensure that all devices can receive the information correctly. The first downlink burst of a frame contains downlink broadcast data at the beginning to inform devices if and at what point in the frame data will be transmitted for them. Furthermore, the broadcast zone also contains information for devices when they are allowed to send data in uplink direction. The remainder of the first downlink burst of a frame then contains user data packets for one or more devices. A frame usually contains more than a single downlink burst and each burst can use a different mo dulation and coding scheme. The location of data for a device thus depends on the radio condition it experiences. Figure 2.25 also shows that data is only sent to one device at a time. This means that subscribers are only multiplexed in time but not in the frequency domain.
As the IEEE 802.16-2004 radio interface was designed for stationary use, devices do not report signal conditions as frequently as is required in systems supporting mobility. Instead, it is the device’s responsibility to judge radio conditions and to send a dedicated management message to the network to change the modulation and coding scheme for the uplink and downlink directions when required. Likewise, error detection and correc­tion on the MAC layer is also optional. If used for a connection, a basic Automatic Retransmission Request (ARQ) scheme splits packages into ARQ blocks and each side reports to the other which blocks have been received correctly. Blocks not received correctly are then retransmitted. This mechanism is similar to that used for the Transmission Control Protocol (TCP), which sits above the IP layer. The advantage of additionally checking and retransmitting packets in the MAC layer is that errors can be detected more quickly and that less data has to be retransmitted. This also helps to keep throughput high, as TCP automatically throttles a transmission once errors occur, as it interprets missing packets as congestion.
Beyond 3G Network Architectures 79
2.4.4 The 802.16e Air Interface and Radio Network
As fixed wireless applications only address a limited customer base, the IEEE soon decided after the fixed air interface standard was finalized to go one step further and to enhance the air interface with additional functionality for mobility and for power­constrained devices. To support mobility, the air interface management must react quickly to changing signal conditions and it must be able to han d over a connection between base stations when the user is moving. For battery-driven devices, the air inter­face had to be optimized to be as power-efficient as possible during times when no data is being transferred. In addition, the WiMAX air interface was enhanced to allow higher transmission speeds. The working group responsible for this task is referred to as 802.16e. After finalization of the standard, it is now known as IEEE 802.16-2005 [32]. The mobile WiMAX standard supports FDD and TDD, although initial deployments will only make use of the TDD option.
2.4.4.1 Orthogonal Frequency Division Multiple Access
On the physical layer, the main difference from the fixed WiMA X standard is that the subcarrier spacing is fixed so the number of subcarriers now varies with the bandwidth, as shown in Table 2.2. According to [33] , the maximum channel bandwidth supported by the first Intel WiMAX chips for notebooks is 10 MHz. In the future, however, larger
Table 2.2 Bandwidths and subcarriers for WiMAX.
Bandwidth Number of subcarriers FFT size
1.25 MHz 85 128 5 MHz 421 512 10 MHz 841 1024 20 MHz 1684 2048
bandwidths are going to be supported as well.
In addition to the parameters shown in Table 2.2, the following physical parameters
were selected:
subcarrier spacing, 10.94 kHz;
OFDM symbol duration, 91.4 ms;
cyclic prefix, 11.4 ms.
It is interesting to note that these values are similar but not identical to the values used in LTE. In LTE, the subcarrier spacing is 15 kHz and a shorter OFDM symbol duration of 66 667 ms is used (cf. Table 2.1).
Based on these radio layer parameters , the physical layer throughput of a WiMAX cell
can be calculated as follows: The transmission time per symbol is 102.8 ms (91.4 ms for the
80 Beyond 3G – Bringing Networks, Terminals and the Web Together
symbol itself þ 11.4 ms for the cyclic prefix), the highest modulation order is 64QAM (6 bits per symbol) and there are 1684 subcarriers in a 20 MHz band:
Physical speed ¼ð1=0:000 102 8Þ
61684 ¼ 98 287 937 bit=s ði:e about 100 Mbit=sÞ
This is almost exactly the same value as calculated for LTE earlier in this chapter and shows that, from this perspective, the two systems provide very similar performances. As was described in the section on LTE, it should be noted that in practice, the throughput of a cell is likely to be only 30–50% of this value. This is because of the overhead for coding, retransmissions of faulty packets, pilot s ignals and the overhead of the higher protocol layers, and also because of the less than ideal signal conditions for most users in the cell.
Figure 2.26 shows the structure of the downlink subframe. The major difference from the frame shown for the fixed WiMAX air interface is the fact that data transmissions to individual users can now also be multiplexed in both time and frequency, due to the much higher number of available subchannels. For this reason, this form of data transmission is not referred to as OFDM (Orthogonal Frequency Division Multiplexing) but instead as OFDMA (Orthogonal Frequency Division Multiple Access). At the beginning of a frame, the Downlink-MAP (DL-MAP) informs devices when and where data is scheduled for them in the frame. An optional Uplink-MAP (UL-MAP) c an also be present in the frame to assign uplink transmis­sion opportunities for this and the following frames.
time
FCH
User 2
DL-MAP
DL-MAP
User 1
frequency/subchannels
User 3
User 4
User 5
User 6 User 7
User n
Figure 2.26 A WiMAX OFDMA frame.
2.4.4.2 MIMO
WiMAX also supports MIMO transmissions in the downlink direction with multiple antennas (e.g. two input, two output=2 2) in the same way as already described for LTE. In the uplink direction, mobile devices only transmit a single data stream. Advanced base stations, however, can activate collaborative MIMO and instruct two devices to transmit at the same time. At the base station, the signals are recognized as coming from different devices due to their separat e multipath characteristics and are separated accordingly.
Beyond 3G Network Architectures 81
Depending on transmission conditions, one of the following two different MIMO
transmission modes can be used.
Matrix A: coverage gain. In a 2 2 antenna configuration (two transmitter antennas, two receiver antennas), a single data stream is transmitted in parallel by two separate antennas. A mathematical algorithm known as Space Time Block Codes (STBC) is used to encode the data streams of the two antennas to make them orthogonal to each other. This improves the signal-to-noise ratio at the receiver, which can be used to:
increase the cell radius;
provide better throughput for subscribers that are difficult to reach (e.g. in difficult indoor conditions or when moving at higher speeds);
transmit with higher-order modulation (e.g. 64QAM) while using fewer error correc­tion bits, which in turn increases transmission speeds to that subscriber.
Matrix B: Capacity Increase. This flavor of MIMO, also known as Spatial Multiplexing MIMO (SM-MIMO), sends a completely independent data stream over each antenna, as described in previous sections. Thus, the data rate can be doubled, given that the mobile device is close to the base station and has excellent reception conditions.
2.4.4.3 Adaptive Antenna Systems
Another feature that is already in the 802.16e standard document, but not used in early networks, is AAS (Adaptive Antenna Systems). By using several antennas and connect­ing them electrically, a beam can be formed towards a client device, thus increasing the signal-to-noise ratio experienced by the client device. To form a beam, the signal is sent over each antenna with a calculated phase shift and amplitude relative to the other antennas. There are no moving parts required for directing the beam in a certain direction, as the beam-forming effect is based on the phase and amplitude differences of the signals. Beamforming can be used in both the uplink and the downlink. For the uplink, beamforming improves the reception of the signal from a device and in the downlink beamforming lowers interference for other devices receiving a transmission from a neighboring cell on the same frequency.
2.4.4.4 Handover Procedures
In 802.16e, both the mobile station and the network can initiate a handover procedure. This is different from UMTS and LTE, in which a handover of a connection is always initiated by the network. In the IEEE specification, a handover is sometimes also referred to as a cell resel ection. This is somewhat unfortunate since in other systems cell reselec­tion is the process to change to another cell while no connection is established to the network.
To perform a handover to a new cell, a mobile has to search for neighboring cells when signal conditions deteriorate. During this process, the mobile will not be able to receive data from the current cell. For this reason, the mobile and base station have to agree when such searches can be performed and the base station then buffers all incoming data
82 Beyond 3G – Bringing Networks, Terminals and the Web Together
packets until the mobile device is back and receiving incoming data. When neighboring cells are detected by the mobile. it reports the reception conditions detected to the network via the serving base station. Both the network and the mobile can then initiate a handover procedure, if required. The mobile device on the one hand can initiate the handover if it feels that it would get better service from a neighboring cell. The network on the other hand can initiate a handover for the same reason or for load balancing purposes if it detects that a neighboring cell with less traffic can be received equally well by the device as the current cell.
In the simplest handover variant, the network or the mobile initiates a handover, which causes a short service interruption while the mobile connects to the new cell. If it is already synchronized, the outage will be shorter than if the mobile first has to associate to the cell and the new cell has to request the subscriber’s current parameters from the previous cell.
For real-time services such as VoIP, interruptions are undesired and two further options have been standardized to improve the handover behavior. The first optional procedure is referred to as Fast Base Station Switching (FBBS). As in the basic approach above, the base station requests the mobile to frequently scan for the availability of neighboring cells. These a re then reported to the serving base station. If signal conditions are strong enough, the serving base station contacts the neighboring base station via the backhaul connection and requests them to set up a context for this subscriber. If the base stations agree, the mobile device is informed that it can select from which base station it wants to receive its downlink data packets. The base stations are kept in a diversity list (the active set) in the mobile which can, by sending a short command, instruct the network to change the cell for downlink transmission. This way, the mobile can quickly react to changing signal conditions. In uplink direction, all base stations that are part of the active set receive the data stream from the mobile and forward correct ly received packets to the ASN-GW. This increases the probability that the network receives at least one copy of each packet but has the disadvantage of increasing bandwidth requirements on the backhaul links.
The Macro Diversity Handover (MDHO) is an even smoother form of handover. Here, all base stations in the active set transmit frames on the downlink to the subscriber. The mobile can then combine the received signal and thus increase the chance of successfully receiving a packet. This approach is quite similar to the UMTS soft hand­over, which, however was abandoned again with HSPA and LTE as it was seen as too costly in terms of capacity requirements on the air interface and the backhaul connection. Both FBBS and MDHO require that all cells that are part of the handover procedure use the same frequency, as the mobile only has a single transceiver and can therefore only receive and transmit on a single frequency.
2.4.4.5 Power Saving and Idle Mode
To minimize the power requirements of battery-driven devices, mobile WiMAX intro­duces a number of power-saving modes, referred to in the standard as power-saving classes. With power-saving class 1 the mobil e and network agree on a pattern in which the device periodically listens to the downlink and afterwards enters sleep mode for some time, during which it cannot be reached. Over time, the sleep periods are automatically
Beyond 3G Network Architectures 83
extended as it becomes less likely that data arrives for the device. Should data arrive while a mobile is in the sleep state, it is buffered in the base station and sent to the mobile as soon as it reactivates its transceiver. This automatically ends the power-save mode. Power save mode is also left when the mobile sends data on the uplink.
While this power-save mode is suitable for bursty data traffic with applications such as Web browsing, it is less suitable for VoIP transmissions, which also have long but predictable periods of inactivity between two packets. Here, it would not make sense to enter or exit power-save mode every time a packet has to be sent or received. Power-saving class 2 thus limits data transmissions to certain intervals. Outside these intervals, the device can turn off its transmitter. In practice, this limits the bandwidth available to a device, which is quite acceptable for VoIP applications that require little bandwidth anyway.
Finally, with power-saving class 3, the network and mobile can agree on a single sleep period after which the connection automatically becomes active again.
In practice, several service flows, each with a different IP address and for different applications, can be active per device. Each service flow can be in a different power­saving mode. The transmitter is then only switched off at times in which all service flows have entered a power-saving state.
2.4.4.6 Idle Mode
Even while in a power-save mode, a mobile device is required to wake up periodically and communicate with the network. In cases of long inactivity, this is undesired as it requires resources on the network side to keep a connection active and has a negative impact on the overall standby time of the device. The standard therefore also defines an idle mode state, in which the radio connection to the network is removed, while the device still keeps its IP address(es). Once in idle mode, the mobile can switch off its transceiver and only occasionally check the reception level of the current and neighboring cells and to observe incoming paging messages which could announce waiting packets on the network side. The paging interval is usually in the order of a few seconds. Another advantage of being in idle mode is that the mobile device can roam between different cells of the network that are in the same paging group without reporting the location change to the network. Only if the mobile roams to a cell in a different paging group does it have to inform the network of its new location, that is of its new paging group, so paging messages can be sent to the new paging group in the future. As a paging is sent via several cells, paging coordination is a task of the ASN-Gateway.
2.4.5 Basic Procedures
Figure 2.27 shows the basic procedures to establish a connection with the network after a mobile device has been powered on. In the first step, it will try to find the previously used network, whose parameters it might have saved in nonvolatile memory. Use of this information has the advantage that the mobile can go directly to a certain frequency and, if the user has not moved since the device was switched off, is very likely to receive a signal instantly. If the user has moved and no signal is found, a standard network search procedure is started. At first, downlink transmissions of a network are detected by searching for the preamble of each frame which has a known bit pattern. Once the
k
84 Beyond 3G – Bringing Networks, Terminals and the Web Together
Mobile Device Networ
1. Synchronization
2. Ranging
3. Capability Information Exchange
4. Authentication
5.Network Registration
6. Service Flow Creation
Connection established, user data can now be exchanged
Figure 2.27 Stages required to connect to a WiMAX network.
preamble is detected, the mobile device is synchronized to the frame structure of the downlink transmissions and can start to receive and decode cell information, which is sent after the frame’s preamble.
The cell information describes, among other things, where to find the contention-based ranging area, which is used in the second step to get into contact with the network. This area, referred to in other standards as the random access channel, is then used by the device to send a ranging request message with a low power level. The message includes the MAC hardware address of the device and the modulation and coding scheme the device suggests the network use to send an answer. If no answer is received, the message is repeated with a higher transmission power. Once the network has successfully received the ranging request, it answers with a ranging response message. The message contains information on how the device has to adjust its power lever for further communication and its synchronization. Synchronization corrections are required as the mobile station is not aware of the distance to the base station. The more distant it is from the base station, the earlier it has to send its transmissions to arrive in synch with packets of other subscribers that have different distances from the base station. The mobile device applies these values and sends another ranging request to the network. The network verifies that the values have been applied correctly and then returns a ranging response message, confirming the procedure and including Connection IDs (CIDs) that will identify packets to the mobile device in the downlink direction and bandwidth grants in the uplink direction.
In the third step, the mobile device sends a capability request message to the network which contains information about its capabilities such as the supported modulation and coding schemes. The network answers with a capability response message which contains its own capabilities. Capability exchange messages are not transmitted in the
Beyond 3G Network Architectures 85
contention-based area, but with the basic CID as part of the data area of a frame. This means that, as soon as the network has sent a final range response message, it starts to schedule uplink resourc es for the mobile device.
In step 4, the device and network authenticate each other as described in Section 2.4.2. Afterwards, the mobile registers to the network. Once registered, the final step in Figure 2.27 consists of establishing a service flow and dedicated CIDs for the user data. This procedure can be initiated by the mobile device or the network, where the service flow is pre-provisioned. Since service flows are agnostic to higher-layer protocols, no IP address is assigned at this point. This is a separate action, which has to be performed by the device once the service flow is active by sending a DHCP (Dynamic Host Configuration Protocol) request to the network in a similar way as is done today in fixed Ethernet and Wi-Fi networks.
2.4.6 Summary and Comparison with HSPA and LTE
When comparing the physical parameters of WiMAX, LTE and HSPAþ, it becomes apparent that in a 5 MHz band, performance of the three systems is very similar. Beyond a bandwidth of 5 MHz, LTE and WiMAX perform on a similar level, as both use OFDM modulation and very similar radio parameters. The major difference between the two is that WiMAX will first be deployed in TDD mode, while LTE will mainly be used in FDD mode for historical reasons. Another difference is that LTE uses a different uplink scheme, making it more power-efficie nt. How much difference this will make in practice remains to be seen. All other differences between the two systems are in the higher layers in the system. While the LTE air interface has inherited a strict channel structure from UMTS, the WiMAX air interface design is much simpler and adheres more to the simple Ethernet-style based MAC layers.
In practice, achievable transmission speed is just one of several important para­meters. Equally important is how well a system is able to handle potentially hundreds of always-on devices per cell, each communicating with the system several times a minute, as applications such as VoIP, encrypted VPN ( Virtual Private Network) tunnels and instant messengers constantly communicate with their servers in the network to keep their channels open through firewalls and NAT gateways. This requires that the system is not only streamlined for high bandwidths but to also able to handle a significant number of bandwidth requests per second for the keep alive messaging without sacrificing bandwidth and mobile battery power. As can be seen with HSPA, air interfaces are continuously enhanced to also take this issue into account. It therefore remains to be seen how first WiMAX and LTE networks fare in this regard and how their evolution accommodates such device behavior. As a consequence of this continuing evolution, it is impossible to describe one system as better than another in terms of performance.
2.4.6.1 Good Competition Between Network Technologies
Since LTE and HSPA on the one hand and WiMAX on the other are very similar in terms of through put and usage scenarios, many observers raise the question of whether we will
86 Beyond 3G – Bringing Networks, Terminals and the Web Together
see a similar destructive competition as in the days of the 2G GSM and CDMA networks. Here, users and operator s did not benefit greatly from this competition because networks and applications were both in the hands of the operators. This created many incompatibility issues for users. One example is text messaging. While in Europe, text messaging has flourished for a long time, it has only recently become popular in the USA. The main reason for this was that it was not possible for users of different networks to exchange text messages. Thus, the service did not take off until interoperability was finally introduced.
With HSPA, LTE and WiMAX, however, the application landscape is quite different. Here, the networks and applications are separated and do not depend on each other. Applications are based on the Internet Protocol and use w hatever network is available. Internet Protocol applications are not and should not be aware of the underlying network technology, which allows people to develop applications independently of the wireless network architecture. Some applications will still be developed by operators but the vast majority will come from Internet­based companies, as will be shown in more detail in Chapter 6. As a result of this split between the application and the network, the competition between different wireless technologies becomes very beneficial because:
It encourages faster network roll outs, as this is one of the few differentiators between
network operators.
It offers possibilities for new player s in the market.
It creates competition between device manufacturers.
New applications can be introduced much more easily and quickly, as they are no
longer forced into a tight network operator controlled framework.
2.4.7 802.16m: Complying with IMT-Advanced
Like LTE, WiMAX is also set to compete for a place in IMT-Advanced 4G. As the current specification is also not likely to qualify for 4G, several activities have been started to enhance the system. The 802.16m working group has been tasked to specify an air interface with a higher bitrate and the following enhancements are foreseen to improve system performance [34].
2.4.7.1 Use of Several Car riers
Like other standards bodies, the IEEE has recognized that increasing the bandwidth used for data transmission is one of the best ways to increa se overall data transfer rates. A multicarrier approach, in which two or even more carriers are used for transferring data, will be used by the future WiMAX air interface. The approach used by WiMAX is backwards-compatible, that is 802.16e and 802.16m mobile devices can be served by the same base station on the same carrier. An 802.16e device, however, does not see the channel bundling and continues to use only one carrier. To be backwards-compatible, high-speed zones are introduced in a frame, which are only available for 802.16m devices. If the carriers used for transmission are adjacent, guard bands that are normally in place to separate the carriers can be used for transferring data.
Beyond 3G Network Architectures 87
2.4.7.2 Self Organization and Inter Base Station Coordination
Interference from neighboring base stations and mobile devices is undesirable in wireless systems, as it reduces the overall system throughput. The new version of the standard introduces methods and procedures to request mobile devices to perform interference measurements at their location and send them back to the base station. The base station can then use the information gathered from different devices to adjust its power settings and potentially also to coordinate the frequency use with neighboring base stations.
2.4.7.3 New Frame Structure
In practice, it has been observed that the 802.16e frame structure with frame lengths of up to 20 ms is too inflexible. The downside of such long frames is slow network access and slow repetition of faulty data blocks, as devices only have one transmission opportunity per frame. The standard 802.16 m uses a new frame structure which consists of super­frames (20 ms) which are further divided into frames (5 ms) and again divided into eight subframes (0.617 ms). Within each frame of 5 ms, the transmission direction can be changed once. Since eight subframes fit into a frame, downlink/uplink time allocations of 6/2, 5/3, and so on can be achieved. By switching the transmission direction at least every 5 ms, HARQ retransmission delays are cut by three-quarters, the idle-to-active state transmission time is reduced from above 400 ms to less than 100 ms and the one-way access delay is reduced from almost 20 ms to less than 5 ms [34].
2.4.8 802.16j: Mobile Multihop Relay
In many scenarios, especially in rural areas, there are often only few or no possibilities at all to backhaul high-bandwidth connections via a fixed line copper or fiber links. Consequently, cells need to be connected wirelessly to the network either over high­bandwidth microwave links or via a concept in which the base stations themselves form a mesh-like network to forward traffic between base stations with no dedicated backhaul connection. WiMAX is the first standard to incorporate such a backhaul method and the
802.16j working group specifies how this should work in practice [35]. In addition to rural backhau ling, forwarding traffic between wireless network nodes is
also an interesting method to fill coverage holes and to improve in-building coverage. At first, it might seem illogical that sending a data packet over the air interface more than once actually increases the data rate. In practice, however, transmitting the packet over two or more links with a high signal-to-noise ratio is better than only transmitting it once but very slowly over a low-quality channel.
The 802.16j amendment to the standard, also referred to as Mobile Multih op Relay
(MMR), covers the following points to achieve these goals without increasing the number of base stations with expensive backhaul links:
2.4.8.1 Backwards Compatibility
MMR has been specified in a way that does not require mobile devices to be aware of relay nodes. This is important as introducing relaying would otherwise not be possible in already deployed networks.
88 Beyond 3G – Bringing Networks, Terminals and the Web Together
2.4.8.2 Multihop Capability
The standard is designed in a way that allows packets to traverse several hops to and from a base station that has a backhaul connection.
2.4.8.3 Relay Station Implementation Options
From the point of view of mobile devices, relays without a dedicated backhaul connec­tion look like a standard base station and have their own base station ID. The specifica­tion defines two kinds of Relay Stations (RS). A simple RS behaves almost like a simple repeater and leaves most of the work to a real base station, including even the handling of simple messages such as ranging requests. Such simple relays are also referred to as transparent relays as all links to mobile devices via relay stations are controlled by a base station. More complex relays, referred to as nontransparent relays, are able to locally manage the link to the subscriber and only forward user data packets to a base station and higher layer signaling information.
2.5 802.11 Wi-Fi
2.5.1 Introduction
At the end of the 1990s the first devices appeared on the market using a new wireless local area network technology that is commonly referred to today as Wireless LAN or Wi-Fi. Wi-Fi is specified by the IEEE in the 802.11 standard. It is very similar to the 802.3 fixed line Ethernet standard and reuses all protocol layers down to layer 2, as shown in Figure 2.28. The major difference between the two protocols is on layer 1, where the fixed line medium access has been replaced with several wireless variants. Furthermore, some additional management features were specified that address the specific needs of wireless transmissions that do not exist in fixed line networks, such as network announce­ments, automatic packet retransmission, authentication procedures and encryption. Over the years, several physical layer standards were added to increase transmission speeds and to introduce additional features. Devices are usually backwards-compatible and support all previous standards to enable newer and older devices to communicate with each other.
5–7
4
3
2
802.3
1
(Ethernet)
Figure 2.28 The 802.11 protocol stack. (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
Application specific
TCP/UDP
IP
802.2 Logical Link Control (LLC)
802.11b 802.11g
802.11a
802.11n
Beyond 3G Network Architectures 89
Initially, Wi-Fi was not very popular or widely known as network interface cards were
expensive and transmission speeds ranged between 1 and 2 Mbit/s. Things changed significantly with the introduction of 802.11b, which specified a physical layer for transmission speeds of up to 11 Mbit/s. Network interface cards became cheaper and devices appeared that could be connected to PCs and notebooks over the new high-speed USB (Universal Serial Bus) interface. Prices fell significantly and Intel decided to include Wi-Fi capabilities in their ‘Centrino’ notebook chipsets. At the same time, the growing popularity of high-speed DSL and TV cable Internet connectivity made wireless net­working more interesting to consumers, since the telephone or TV outlet was and still is often not close to where a PC or notebook is located. Wi-Fi was the ideal solution to this problem and Wi-Fi access points were soon integrated into DSL and cable modems. Likewise, Internet access in public places such as cafes, hotels, airports and so on became popular, again enabled by Wi-Fi and cheap high-speed Internet access at the other end of the wireless connection via DSL. Today, Wi-Fi has become ubiquitous in notebook s and many other mobile and portable devices such as game consoles, mobile phones, Internet tablets and Mobile Internet Devices (MID).
Over time, two additional physical layer specifications were added to further increase
transmission speeds. The 802.11g standard increased data transfer speeds to up to 54 Mbit/s on the air interface, and the recent 802.11n standard has the potential for up to 300 Mbit/s. It should be noted at this point that these speeds are only theoretical and not measured on the air interface. In practice, protocol overhead reduces the achievable speeds at the application layer to about half those values. This is discussed in more detail in the following sections. Standard 802.11a is another Wi-Fi air interface variant, but has never gained much popularity because it does not use the same standard frequency band as the other 802.11 variants.
The remainder of this chapter is structured as follows: as a first step, the Wi-Fi network
infrastructure model is discussed. This is followed by an introduction to the different physical layers and their properties. Like other wireless networking technologies, the network needs to be managed and organized and basic management procedures are discussed next. Wi-Fi security is a very important topic and, as initially encryption algorithms were found to be insecure, it is worth taking a look at this topic and discussing how security was improved over time. Due to the tremendous popularity of Wi-Fi and the growing use of the technology for real-time applications such as VoIP and video streaming, quality of service is becoming an important topic. As a consequence, this chapter then discusses the QoS extension of the Wi-Fi standard and how it can improve reliability for such applications.
2.5.2 Network Architecture
2.5.2.1 The Wireless Network in a Box
Unlike the network technologies described before, Wi-Fi is foremost a local area net­working technology and most Wi-Fi networks are deployed as a ‘network in a box’ as shown in Figure 2.29. In a typical home network, the Wi-Fi network bridges the final meters between the DSL modem and the devices using the fixed line Internet connection. The Wireless LAN Access Point (AP) is usually combined with the DSL modem and also
90 Beyond 3G – Bringing Networks, Terminals and the Web Together
Multi purpose WLAN Access Point
WLAN
Access Point
Ethernet
10 Mbit/s–1 GBit/s
Ethernet Switch
(Layer 2)
Wireline Ethernet devices
Figure 2.29 A DSL router with a wireless LAN interface. (Reproduced from Communication Systems for the Mobile Information Society, Martin Sauter, 2006, John Wiley and Sons.)
(internal)
DHCP
Server
IP Router
with NAT
(Layer 3)
Ethernet PPPoE
DSL
Modem
To DSL Splitter
serves as the DHCP server, which provides network configuration parameters such as the IP address to notebooks and other wireless devices when they connect to the network. Most multipurpose DSL or cable routers also have a built-in Ethernet switch with several ports to connect PCs and other devices with a twisted-pair Ethernet cable.
2.5.2.2 Network Address Translation
Multipurpose routers such as the WLAN access point can connect many wireless and wired clients to the network; they usually also include NAT functionality, that separates the local network from the DSL connection and translates IP addresses between the LAN and the WAN. This translation is required because the Internet service provider’s network usually only assigns a single IP address per DSL connection. By using NAT, local IP addresses and TCP or UDP port numbers are mapped to the external IP address and the same or different TCP or UDP port numbers. This allows all local devices to communicate with servers on the Internet simultaneously via separate connections. From an external network point of view, all devices use the same IP address. While this works well for many applications, there are some for which this translation creates a problem, as incoming packets are discarded if they do not belong to a mapping that was created by an outgoing packet first. This can be solved by configuring static mappings, which forward incoming packets for a server (e.g. a Web server) to a specific internal IP address.
Static mappings, however, are only useful for servers that always use the same TCP or UDP port numbers. SIP, however, which is the dominant protocol for VoIP applications, uses dynamic port numbers. In addition, SIP applications use an IP address they can query from the local protocol stack and include it in application layer messages. As the local network stack is not aware of the external IP address assigned by the DSL
Loading...