dissertation malkowski[1]

Performance Evaluation ofPacket Switched Services in

UMTS

Von der Fakultat fur Elektrotechnik und Informationstechnik derRheinisch-Westfalischen Technischen Hochschule Aachen zur Erlangung

des akademischen Grades eines Doktors der Ingenieurwissenschaftengenehmigte Dissertation

vorgelegt von

Diplom-IngenieurMatthias Malkowski

aus Essen

Berichter: Universitatsprofessor Dr.-Ing. Bernhard WalkeUniversitatsprofessor Dr.-Ing. Peter Vary

Tag der mundlichen Prufung: 16. Januar 2012

ABSTRACT

In order to deal with the increasing demands for high bit rate data ser-vices the Universal Mobile Telecommunications System (UMTS) has been

introduced as a 3rd Generation (3G) system for mobile communication.With the goal to achieve high spectrum efficiency and differentiated Qual-ity of Service (QoS) provisioning, UMTS traversed several evolutionarysteps. From Release 99 to Release 7 of the 3rd Generation PartnershipProject (3GPP) specifications numerous features have been added and havebeen extended.

Especially new transport channels like the High Speed Downlink SharedChannel (HS-DSCH) as part of the High Speed Downlink Packet Access(HSDPA) significantly improved the performance for Packet Switched (PS)services. At the time of the introduction of UMTS in 2002 the Release 99systems only reached a relatively low maximum data rate of less than400 kbit/s. With the Release 5 HS-DSCH, deployed in Germany since 2007,data rates of up to 3.6 Mbit/s are offered. In certain areas an even highertheoretical throughput of 7.2 Mbit/s was made available. The maximumtheoretical Release 5 throughput which is upper bounded to 14.4 Mbit/swas planned to be useable in Germany in 2010. Further enhancements asstudied within this thesis are introduced in later releases.

This thesis gives a comprehensive overview of UMTS with respect toits possibilities for the provision of PS services. The central question thisthesis gives answers to is in how far the individual release of UMTS improvethe ability to support mobile Internet services. Based on both link-level aswell as system-level simulation the performance of the available options, i.e.Release 99 without HSDPA and the various HSDPA releases, are quanti-tatively compared. By referring to analytical models or related empiricalwork the results are being validated. Furthermore, the acquired resultsare benchmarked with the performance of Mobile Worldwide Interoperabil-ity for Microwave Access (WiMAX), another state-of-the-art system. Withthese comparisons conclusions about the possibilities and limitations of thirdgeneration mobile networks are given.

Within the evaluations special focus is being put on the HSDPA as themost promising technology for the delivery of PS services. In this context

iv Abstract

in-depth analyses of the most important features, i.e. Adaptive Modulationand Coding (AMC), Hybrid ARQ (HARQ) and fast scheduling, are per-formed. For vendor specific algorithms, e.g. receiver algorithms, schedulingalgorithms or HARQ schemes, comparative analyses in terms of complexityand achievable performance are made. By presenting the release specificoptions and their performance, the system capacity and efficiency, the tech-nical limitations and drawbacks as well as possibilities for optimization andimprovement, an audience including operators, vendors, interested users andmainly researchers is addressed.

KURZFASSUNG

Um der wachsenden Nachfrage nach hochbitratigen Datendiensten nach-zukommen, wurde UMTS als ein Mobilfunksystem der dritten Gene-

ration eingefuhrt. Mit dem Ziel, hohe spektrale Effizienz und differenzierteDienstgute zu gewahrleisten, durchlief UMTS einige evolutionare Schritte.Seit der ersten Version des durch die 3GPP spezifierten Systems (Release 99)wurden zahlreiche Funktionen hinzugefugt und verbessert.

Insbesondere neue Transportkanale, wie zum Beispiel der zum HSDPAgehohrende HS-DSCH, verbesserten die Leistungsfahigkeit fur paketvermit-telte Dienste signifikant. Zum Zeitpunkt der Einfuhrung von UMTS im Jahr2002 konnte das Release 99 System nur eine relativ niedrige maximale Da-tenrate von unter 400 kbit/s erreichen. Mit dem in Release 5 hinzugekom-menen HS-DSCH, welcher in Deutschland seit 2007 zum Einsatz kommt,werden Datenraten bis zu 3.6 Mbit/s angeboten. In einigen Regionen istsogar ein hoherer theoretischer Durchsatz von 7.2 Mbit/s verfugbar. Dermaximal mogliche Durchsatz mit Release 5, welcher mit 14.4 Mbit/s seineobere Grenze hat, ist geplant ab 2010 in Deutschland nutzbar zu sein. Wei-tere Verbesserungen, welche in dieser Arbeit studiert werden, kommen inspateren Versionen des Standards hinzu.

Diese Arbeit liefert einen umfangreichen Uberblick uber UMTS in Bezugauf seine Moglichkeiten, paketvermittelte Dienste anzubieten. Die zentra-le Frage, zu der diese Arbeit Antworten gibt, ist, inwieweit die einzelnenUMTS Versionen die Fahigkeit steigern, mobilen Zugang zum Internet zubieten. Basierend auf Simulationen der Radioverbindung und des Kom-plettsystems wird die Leistungsfahigkeit der verfugbaren UMTS Optionenquantitativ verglichen. Unter Einbeziehung von analytischen Modellen undverwandten empirischen Arbeiten werden die Ergebnisse validiert. Des Wei-teren werden die erlangten Ergebnisse mit der Leistungsfahigkeit von MobileWiMAX als modernem Referenzsystem verglichen. Mit Hilfe dieser Verglei-che wird ein Bild der Moglichkeiten und Grenzen der Mobilfunknetze derdritten Generation gegeben.

Im Rahmen der Auswertungen wird speziell HSDPA als eine vielverspre-chende Technologie zur Erbringung paketvermittelter Dienste fokussiert.In diesem Kontext wird eine tiefgehende Analyse der wichtigsten Funk-

vi Kurzfassung

tionen wie adaptive Modulation und Kodierung, hybrides ARQ Protokollund schnelle Ressourcenzuteilung vorgenommen. Fur herstellerspezifischeAlgorithmen, zum Beispiel Empfangsalgorithmus, Ressourcenzuteilungsal-gorithmus und das verwendete hybride ARQ Schema, wird eine verglei-chende Analyse in Bezug auf die Komplexitat und die Leistungsfahigkeitdurchgefuhrt. Durch die Prasentation der spezifischen Optionen und derenLeistungsfahigkeit, der Kapazitat und Effizienz des Systems, der technischenLimitierungen und Nachteile als auch der Moglichkeiten fur Optimierungenund Verbesserungen werden Betreiber von Mobilfunknetzen, Hersteller, in-teressierte Nutzer dieser Technologie und vor allem Forscher als Leser gezieltangesprochen.

CONTENTS

Abstract iii

Kurzfassung v

Contents vii

1 Introduction 11.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.3 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.4 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 UMTS Architecture 72.1 Network Topology . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Radio Interface Protocols . . . . . . . . . . . . . . . . . . . . 9

2.2.1 Radio Resource Control . . . . . . . . . . . . . . . . . 122.2.2 Packet Data Convergence Protocol . . . . . . . . . . . 162.2.3 Broadcast/Multicast Control . . . . . . . . . . . . . . 202.2.4 Radio Link Control . . . . . . . . . . . . . . . . . . . . 21

2.2.4.1 Transparent Mode . . . . . . . . . . . . . . . 222.2.4.2 Unacknowledged Mode . . . . . . . . . . . . 232.2.4.3 Acknowledged Mode . . . . . . . . . . . . . . 26

2.2.5 Medium Access Control . . . . . . . . . . . . . . . . . 302.2.5.1 Logical Channels . . . . . . . . . . . . . . . . 32

2.2.5.1.1 Traffic Channels . . . . . . . . . . . 322.2.5.1.2 Control Channels . . . . . . . . . . 33

2.2.5.2 Transport Channels . . . . . . . . . . . . . . 342.2.5.2.1 Dedicated Transport Channels . . . 342.2.5.2.2 Common Transport Channels . . . . 35

2.2.5.3 Dedicated Channels and the MAC-d Entity . 372.2.5.4 High Speed Downlink Packet Access . . . . . 39

2.3 Physical Layer . . . . . . . . . . . . . . . . . . . . . . . . . . 43

viii Contents

2.3.1 Transport Channel Coding and Multiplexing . . . . . 442.3.1.1 Error Detection . . . . . . . . . . . . . . . . 482.3.1.2 Forward Error Correction . . . . . . . . . . . 492.3.1.3 Rate Matching . . . . . . . . . . . . . . . . . 512.3.1.4 Interleaving . . . . . . . . . . . . . . . . . . . 542.3.1.5 Constellation Rearrangement . . . . . . . . . 56

2.3.2 Physical Channels . . . . . . . . . . . . . . . . . . . . 562.3.2.1 Downlink Physical Channel Processing . . . 572.3.2.2 Uplink Physical Channel Processing . . . . . 582.3.2.3 Frame Structure . . . . . . . . . . . . . . . . 602.3.2.4 Modulation Mapping . . . . . . . . . . . . . 622.3.2.5 Multiple Access . . . . . . . . . . . . . . . . 642.3.2.6 Modulation . . . . . . . . . . . . . . . . . . . 66

3 Simulation Environment 693.1 Application Models . . . . . . . . . . . . . . . . . . . . . . . . 713.2 Transmission Control Protocol/Internet Protocol . . . . . . . 713.3 UMTS Protocol Stack . . . . . . . . . . . . . . . . . . . . . . 723.4 Radio Interference Simulation Engine . . . . . . . . . . . . . 743.5 Link-Level Simulation Module . . . . . . . . . . . . . . . . . . 753.6 Real Time Wireless Network Demonstrator . . . . . . . . . . 783.7 Graphical User Interface . . . . . . . . . . . . . . . . . . . . . 80

4 Link-Level Performance Evaluation 814.1 Modulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834.2 Coding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 864.3 Dedicated Channel . . . . . . . . . . . . . . . . . . . . . . . . 89

4.3.1 Downlink . . . . . . . . . . . . . . . . . . . . . . . . . 924.3.2 Uplink . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

4.4 High Speed Downlink Shared Channel . . . . . . . . . . . . . 964.5 Mobile WiMAX . . . . . . . . . . . . . . . . . . . . . . . . . . 100

4.5.1 Comparison of HSDPA and Mobile WiMAX PhysicalChannel Capacity . . . . . . . . . . . . . . . . . . . . 103

4.6 Throughput Comparison . . . . . . . . . . . . . . . . . . . . . 1044.6.1 AWGN Channel . . . . . . . . . . . . . . . . . . . . . 104

4.6.1.1 Physical Layer Throughput . . . . . . . . . . 1044.6.1.2 Maximum Throughput above RLC Layer . . 105

4.6.2 Throughput for Pedestrian Channel Model . . . . . . 108

Contents ix

4.6.3 Throughput for Vehicular Channel Model . . . . . . . 1094.7 Hybrid ARQ . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

4.7.1 Chase Combining and Incremental Redundancy . . . . 1144.7.2 Constellation Rearrangement . . . . . . . . . . . . . . 1154.7.3 Comparison . . . . . . . . . . . . . . . . . . . . . . . . 116

5 Performance on System Level 1195.1 Related Work of the Author . . . . . . . . . . . . . . . . . . . 1205.2 Fast Scheduling for HSDPA . . . . . . . . . . . . . . . . . . . 122

5.2.1 Scheduling Strategies . . . . . . . . . . . . . . . . . . . 1245.2.1.1 Maximum SINR Scheduling . . . . . . . . . . 1245.2.1.2 Proportional Fair Scheduling . . . . . . . . . 1245.2.1.3 Modified Largest Weighted Delay First Schedul-

ing . . . . . . . . . . . . . . . . . . . . . . . 1255.2.1.4 Exponential Rule Scheduling . . . . . . . . . 1265.2.1.5 Channel-Dependent Earliest Due Date Schedul-

ing . . . . . . . . . . . . . . . . . . . . . . . 1265.2.1.6 Expo-Linear Scheduling . . . . . . . . . . . . 127

5.2.2 Qualitative Comparison of Scheduling Metrics . . . . . 1275.3 HS-DSCH Performance for RT Services . . . . . . . . . . . . 129

5.3.1 MAC-d PDU Queueing Delay . . . . . . . . . . . . . . 1325.3.2 Inter-Scheduling Interval . . . . . . . . . . . . . . . . . 133

5.4 HS-DSCH Performance for mixed Services . . . . . . . . . . . 1355.4.1 MAC-hs PDU Throughput . . . . . . . . . . . . . . . 1365.4.2 Throughput of NRT Services . . . . . . . . . . . . . . 1375.4.3 Queueing Delay of RT Services . . . . . . . . . . . . . 1385.4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . 140

6 Conclusion and Outlook 141

A Additional MAC Entities 145A.1 System Broadcast by the MAC-b Entity . . . . . . . . . . . . 145A.2 Entities for Common, Shared and MBMS Channels . . . . . . 146A.3 High Speed Uplink Packet Access . . . . . . . . . . . . . . . . 150

A.3.1 MAC-e/es entity in the UE . . . . . . . . . . . . . . . 151A.3.2 MAC-e entity in the Node B . . . . . . . . . . . . . . 153A.3.3 MAC-es entity in the SRNC . . . . . . . . . . . . . . . 155

x Contents

List of Figures 157

List of Tables 161

List of Equations 163

List of Abbreviations 165

Bibliography 175

Non-Public References 189

Acknowledgment 193

CHAPTER 1

Introduction

Contents1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 Contributions . . . . . . . . . . . . . . . . . . . . . . 4

1.4 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . 5

S ince the first demonstrations of information transfer using radio trans-missions by Guglielmo Marconi in 1897, where he was able to transmit

Morse code signals over a distance of several kilometers, a huge amount of re-search and development created a large industry sector dealing with mobilecommunication. In the early 1980s first mobile communication systems,referred to as 1st Generation (1G) systems, entered the personal market.Because the analog systems suffered from several drawbacks in terms ofmobility, service provisioning and cost, the mass market was only reachedby the following generation, the 2nd Generation (2G) systems.

Launched in 1991 the digital Global System for Mobile Communication(GSM) became a European and, later, a worldwide success. Still focusingon Circuit Switched (CS) speech services only, it soon became clear thatalso non-voice services are demanded by the customers. Short MessageService (SMS) is an early example of a service originally not envisaged bythe system. Furthermore, the rapid development of the Internet caused theinformation society to strongly desire Packet Switched (PS) services also inthe wireless mobile world [96]. As a result 2G networks were extended byimplementing the General Packet Radio Service (GPRS) [4, 8], specified inRelease 97, on top of the legacy GSM as proposed in [54]. In the year 2000first GPRS networks were rolled out. At a later point in time, Release 99Enhanced Data Rates for GSM Evolution (EDGE) further improved theachievable throughput. Those two extensions of the 2G networks are oftenreferred to as 2.5G and 2.75G, respectively.

2 1. Introduction

With the goal to achieve a high data rate as known from fixed-line net-works (e.g. Integrated Services Digital Network (ISDN) and later DigitalSubscriber Line (DSL)), spectrum efficiency and differentiated QoS pro-visioning, the International Mobile Telecommunications-2000 (IMT-2000)initiative defined the requirements for the 3G networks. UMTS, speci-fied by 3GPP and also known as Wideband Code Division Multiple Access(WCDMA), became the most popular 3G system. After several initial trialsthe first commercial UMTS networks were launched around 2002. Since thenseveral releases of the 3GPP specification continuously enhanced the avail-able features of this system. The High Speed Downlink Packet Access (HS-DPA), for example, which has been rolled out in Germany in 2007 signifi-cantly increased the achievable throughput. Figure 1.1 illustrates the majorfeature growth over time [C12].

LTE(Release 8)HSPA+

(Release 7)HSUPA(Release 6)HSDPA

(Release 5)UMTS TDD

TD-CDMA

(Release 4)

UMTS FDD

W-CDMA

(Release 99)

OFDMA

Scaleable bandwidth

Spectrum efficiency

Higher cell edge bit rate

Latency between UE and

NodeB below 5 ms (1 ms TTI)

Compatible with earlier

releases and other systems

64QAM

TDD enhancements

10 MHz / 7.68 Mcps

MBMS evolution

MIMO antennas

Uplink Hybrid ARQ

Enhanced dedicated

uplink channels

Adaptive code rate

Centralized grant based

uplink scheduling in

NodeB

MBMS

Downlink Hybrid ARQ

High speed shared

downlink channels

Adaptive code rate and

modulation (16 QAM)

Downlink scheduling

in NodeB

Shorter TTI (2 ms)

Low chip rate

1.28 Mcps

Separation of

Transport and

Control in

CS domain

UTRAN instead of

GERAN

DS-CDMA

5 MHz bandwidth

3.84 Mcps

Soft Handover

Features

Time2000 2007/20082005/2006

Figure 1.1: UMTS releases and key features

1.1. Motivation 3

1.1 Motivation

In general this thesis is motivated by the questions which capabilities theevolutionary steps of UMTS have and how well they are suited for thedelivery of today’s and future PS services. Depending on the perspectivethis central questions split up into several subitems.

From customer perspective the applicability of UMTS to provide a mo-bile Internet connection which competes with the experiences known fromfixed-line broadband connections is of interest. In detail, important and in-teresting user performance metrics are the achievable individual throughputand the perceived QoS for both Non-Real Time (NRT) and Real Time (RT)services.

Additionally, from operator perspective the system capacity and theefficiency compared to other competing technologies and the theoreticallimits are of interest. A network and User Equipment (UE) vendor aspectof interest is the performance of not specified algorithms, e.g. schedulingand receiver algorithms.

Last but not least the identification of drawbacks and possibilities forimprovement of the developed UMTS are motivations from research point ofview. In addition to the improvement of UMTS itself the scientific findingsof this work facilitate the enhancement of competing and future mobilecommunication systems as well.

1.2 Objectives

This thesis intends to give a detailed overview about the various possibilitiesfor the provision of PS services in UMTS. In order to do so all relevantsystem aspects and protocol layers are described in detail with respect tothe release of the specification they have been introduced with. As the3GPP specification leaves diversification freedom for vendors and operatorsin certain areas, typical configuration options are presented and discussed.

With the help of bit-exact simulation models quantitative performanceevaluations of various UMTS configurations and releases with respect totheir applicability for PS services are made. To acquire in-depth perfor-mance results, simulations are performed on link and system level. Forvendor specific algorithms, e.g. receiver algorithms, scheduling algorithms

4 1. Introduction

or HARQ schemes, comparative analyses in terms of complexity and achiev-able performance are made. Furthermore, the evaluated efficiency of UMTSis assessed by comparing it to a competing state-of-the-art system, namely,Mobile WiMAX.

1.3 Contributions

The present thesis mainly contributes to the field of computer simulationaided performance evaluations of wireless networks. In detail, a simulationenvironment which allows bit-exact simulations on link level as well as highlydetailed system level evaluations has been developed by the author. Basedon this framework comprehensive simulation studies have been performedto examine the performance of UMTS with respect to both NRT and RT PSservices. Whenever possible the results have been validated by comparisonsto analytical models or related empirical work.

This thesis provides detailed and neutral performance and efficiencycomparisons of various configuration options of the different UMTS releasesas well as of Mobile WiMAX as a competing technology. Furthermore, thegathered results are assessed with respect to theoretical limits. With thiscomparison the reader of this thesis gets figures on what can be expectedfrom the competing technologies and the various releases of UMTS. Severalmethods for the provision of QoS in UMTS are presented and evaluated onboth qualitative and quantitative level. Based on this in-depth study thetrade-off between possible QoS and system capacity is shown.

Further contributions of this thesis are comparisons of vendor specificalgorithms, e.g. receiver algorithms, scheduling algorithms and HARQschemes. Several alternatives are presented and compared with respect totheir applicability, advantages and drawbacks. Finally, several drawbacksand limitations of the studied configurations are identified and discussedwith respect to future research (e.g. Radio Link Control (RLC) ProtocolData Unit (PDU) size granularity).

1.4. Outline 5

1.4 Outline

This thesis is structured as follows:

Chapter 2 gives an overview of the UMTS radio interface architecture.After a short introduction of the network topology and the involved nodes(Section 2.1) the protocol layers of the Access Stratum (AS) are presented.In detail, this includes a description of the UMTS Data Link Layer (DLL)and Radio Resource Control (RRC) (Section 2.2) as well as the PhysicalLayer (PHY) as specified by 3GPP (Section 2.3). The focus of this chapteris on features which have been implemented in the simulation environmentand which are required by the performance evaluations carried out withinthis thesis.

The following chapter gives an overview of the simulation environmentand simulation models developed and used within this thesis. All WirelessNetwork Simulator (WNS) modules required for the following performanceevaluations are introduced in Chapter 3.

Based on the simulation environment, comprehensive performance eval-uations are made. In Chapter 4 detailed link level results for typical UMTSconfigurations are presented. At the beginning of the chapter simulationsvalidating the implemented functionality are discussed. Where possible,these results are compared to analytical models. As a next step the phys-ical and transport channel performance of the different UMTS releases isevaluated. Both Dedicated Channel (DCH) and HS-DSCH results are ex-amined. In order to compare the results to another state-of-the-art system,performance results of Mobile WiMAX are presented as well. Starting withAdditive White Gaussian Noise (AWGN) channel based results the per-formance of the systems and different configurations is studied for fadingchannel environments. Several receiver algorithms are applied and examinedwithin these environments. Finally, different HARQ schemes are comparedwith respect to their performance and technical complexity.

While the previous chapter focused on the achievable single user per-formance, Chapter 5 concentrates on multiuser scenarios. After an intro-duction to related work published by the author besides this thesis, thescheduling principle and several scheduling algorithms are presented and

6 1. Introduction

qualitatively discussed. Based on system-level simulation, quantitative per-formance results in terms of QoS and system capacity are gathered. Specialinterest is put on scenarios including RT services as well as traffic mix sce-narios with RT and NRT services.

Finally, in Chapter 6 the work is concluded by summarizing the mainoutcome of this thesis. Furthermore, an outlook to potential future researchis given.

CHAPTER 2

UMTS Architecture

Contents2.1 Network Topology . . . . . . . . . . . . . . . . . . . . 7

2.2 Radio Interface Protocols . . . . . . . . . . . . . . . . 9

2.2.1 Radio Resource Control . . . . . . . . . . . . 12

2.2.2 Packet Data Convergence Protocol . . . . . . 16

2.2.3 Broadcast/Multicast Control . . . . . . . . . 20

2.2.4 Radio Link Control . . . . . . . . . . . . . . . 21

2.2.5 Medium Access Control . . . . . . . . . . . . 30

2.3 Physical Layer . . . . . . . . . . . . . . . . . . . . . . 43

2.3.1 Transport Channel Coding and Multiplexing . 44

2.3.2 Physical Channels . . . . . . . . . . . . . . . . 56

In this chapter the UMTS network architecture as specified by 3GPP isintroduced. First the general network topology of the UMTS Terrestrial

Radio Access Network (UTRAN) with all its nodes is presented. In the sec-ond part of this chapter the focus is on the UTRAN layer 2 and 3 protocolsas well as the involved entities. Finally, the physical layer functionality withrespect to the most important transport and physical channels is explained.

This chapter primarily concentrates on those parts of the UMTS systemwhich are relevant for the PS domain. Especially, aspects which are requiredby the performance evaluation in Chapter 4 and Chapter 5 are addressed.Furthermore, a detailed technical description of the features implementedin the simulation framework presented in Chapter 3 is given.

2.1 Network Topology

In Figure 2.1 a typical network deployment is introduced [6]. In UMTS theBase Transceiver Station (BTS) and Base Station Controller (BSC) of theGSM network have been replaced by the Node B and the Radio Network

8 2. UMTS Architecture

Node B

RNC

RNC

MSC

SGSN

Node B

Node B

Node B

Iub

Iub

Iub

Iub

Iur Iu

Iu

Uu

Figure 2.1: UTRAN network topology

Controller (RNC). The Node B is the entity located closely to the antennasof a site, i.e. the location where antennas are installed. In case sectorizedantennas are used one Node B is controlling all sectors at this site. It iseven possible that multiple cells at one location, e.g. macro and micro cells,are controlled by one Node B. The Node B primarily terminates the PHY ofthe radio interface. In UMTS the radio interface between UE and Node Bis called the Uu interface. Several Node Bs are controlled by one RNC towhich they are connected by the so called Iub interface.

The RNC is the network element which controls the radio resources ofits Node Bs. One RNC together with its Node Bs forms a Radio Network

2.2. Radio Interface Protocols 9

Subsystem (RNS). Different from the BSCs in GSM the RNCs in UMTS areinterconnected, namely through the Iur interface. The Iur interface easeshandover and allows a combined radio resource management, e.g. loadsharing between cells controlled by different RNCs. Both the Node Bs andthe RNCs belong to the Radio Access Network (RAN) of UMTS. EachRNC is connected to the Core Network (CN) by the Iu interface.

With the introduction of UMTS the network elements in the CN werekept the same. Similar to GSM the nodes to which the RAN is connectedto are the Mobile Switching Center (MSC) for CS services and the Serv-ing GPRS Support Node (SGSN) for PS services. If both the MSC andthe SGSN are located in the same node it is referred to as UMTS MobileSwitching Center (UMSC). Further information about these and additionalnodes in the CN of UMTS can be found in [138] and [89].

2.2 Radio Interface Protocols

As illustrated in Figure 2.2 the UMTS architecture can be divided into anAS and a Non-Access Stratum (NAS) [9]. This separation is done on afunctional basis. The AS contains the functionality which depends on theaccess technology, the NAS contains the functionality which is independentof it.

The AS provides radio bearer access to the NAS in an abstract andbearer independent way [12]. The type of information and the requiredQoS are parameters for the NAS to characterize a connection over the AS.According to Figure 2.2 there are three Service Access Points (SAPs), shownby ellipses, offered to the NAS. These SAPs are called Notification (Nt)SAP, General Control (GC) SAP and Dedicated Control (DC) SAP. TheNt SAP is used to broadcast data to identified users, the GC SAP’s purposeis to enable the CN to provide information and to give commands which donot relate to specific users, and the DC SAP’s task is the establishment andrelease of connections to specific UEs and to exchange information relatedto these connections.

The AS includes the network oriented layers of the ISO/OSI (Inter-national Organization for Standardization/Open Systems Interconnection)reference model [49]. These are the Physical Layer (PHY), the Data LinkLayer (DLL) and the Network Layer (NL). In the UMTS protocol stackthe data link layer is split into four sublayers [25]. As shown in Fig-


Nt DC GC Nt DC

Radio ProtocolIu

Prot.Iu Protocols

Iu

Access Stratum (AS)

Non-Access Stratum (NAS)

Core Network (CN)UTRANUE

Radio

(Uu)

GC

Prot.Radio

Figure 2.2: UMTS protocol architecture

ure 2.3 these sublayers are called Medium Access Control (MAC), Ra-dio Link Control (RLC), Packet Data Convergence Protocol (PDCP) andBroadcast/Multicast Control (BMC). The layers above the MAC layer aredivided into a control and a user plane. The PDCP layer and the BMClayer exist in the user plane only.

The task of the user plane is to provide services for the transport ofuser data while the control plane deals with signalling for connection setup,release and radio bearer management. The control plane tasks are mainlyfulfilled by the Radio Resource Control (RRC) protocol which belongs tothe network layer and exists in the control plane only.

Each block in Figure 2.3 symbolizes an instance of the respective proto-col. The instances of the physical layer communicate over physical channelswith each other. Among other aspects, the physical layer, described in Sec-tion 2.3, contains blocks for Forward Error Correction (FEC), Cyclic Re-dundancy Check (CRC) and multiplexing. These blocks are used to providetransport channels to the MAC layer.

There exist dedicated transport channels and common transport chan-nels provided by the physical layer to the MAC layer. Dedicated transportchannels belong to one UE and common transport channels are shared be-tween all UEs. The MAC layer maps these transport channels to so-calledlogical channels which are used by the RLC layer. A more detailed descrip-


UMTS Radio Interface(Uu Stratum) boundary

user plane

Channels

Transport

Logical

Channels

L2/RLC

L2/MAC

L2/PDCP

L3/RRC

L2/BMC

L1/PHY

control plane

RLC

MAC

PHY

RRC

BMC

PDCP

Control

Radio Bearers

Signalling

Radio Bearers

PS Domain CS Domain

Figure 2.3: Radio interface protocol architecture

tion of the functionality of the MAC layer can be found in Section 2.2.5.The RLC layer performs segmentation/reassembly of RLC Service Data

Units (SDUs) and may retransmit erroneously received data using Au-tomatic Repeat Request (ARQ) protocols. There are three transmissionmodes offered by RLC to upper layers which are the Acknowledged Mode(AM), the Unacknowledged Mode (UM) and the Transparent Mode (TM).Further information about the RLC layer is provided in Section 2.2.4.

Above the RLC layer the PDCP layer and the BMC layer are located.The use of both is service dependent. The PDCP, presented in Section 2.2.2,adapts packet data protocols, e.g. the Internet Protocol (IP), to the UMTSradio interface. An example for this adaptation is the Transmission Control


Protocol (TCP)/IP header compression. The BMC layer (see Section 2.2.3)is used for the Cell Broadcast Service (CBS). The service offered by the userplane DLL to the NL is implemented by the so-called Radio Bearers (RBs).The RB service may be offered by the PDCP for the PS domain or by theBMC and RLC entities for the CS domain. For speech services in the CSdomain, for example, the data of the Adaptive Multi-Rate (AMR) speechcodec [40] is directly sent to the TM RLC entities.

The RRC layer uses the RLC layer to transmit signalling data. Thechannels on which the RLC offers the data transfer service to the RRCare called Signalling Radio Bearers (SRBs). Furthermore, the RRC layerhas connections to all protocol layers mentioned so far. The purpose ofthese control SAPs is the configuration of the corresponding layers and theexchange of measurement information between these layers and the RRC.More detailed information about the RRC protocol is given in Section 2.2.1.

Above these protocol layers the border to the NAS with its three SAPs islocated (see Figure 2.2 and Figure 2.4). Higher layer signalling such as CallControl (CC), Mobility Management (MM) and Session Management (SM)is performed above this border by the Layer 3 (L3) CN protocols. Alsosecurity related procedures like authentication [42] and the Short MessageService (SMS) [7] are examples for functionality located in the NAS. As theNAS is out of the scope of this thesis the reader might wish to consider [11]and [13] for more information.

2.2.1 Radio Resource Control

The RRC layer specified in [34] is in charge of establishment, managementand release of the SRBs and RBs between UE and UTRAN as well asthe associated resources. Further important functions of the RRC are thebroadcast of system information and paging information in the downlinkas well as the transmission of UE generated random access messages andmeasurement reports in the uplink. Handover procedures and the initialcell selection [27] are tasks of the RRC as well.

Figure 2.4 shows the structure of the RRC layer at the UE side. In theUTRAN the structure is the same with the arrows, illustrating the up- anddownlink direction, pointing in the opposite direction. Services to the NASare offered by the General Control (GC), Notification (Nt) and DedicatedControl (DC) SAP. The GC SAP provides access to an information broad-


RFERFERFE

RLC

MAC

L1

UM−SAP AM−SAPTR−SAP

Nt−SAP DC−SAPGC−SAP

RLC−Ctrl

L1−Ctrl

MAC−Ctrl

Access Stratum

RRC SAPs

BCFE DCFE SCFE

Non−Access Stratum

TME

PNFE

RRC

Figure 2.4: UE side model of RRC

cast service which distributes information to all UEs in a certain geograph-ical area. The Nt SAP provides access to paging and notification broadcastservices. The paging service broadcasts its information in a certain geo-graphical area as well. In contrast to the GC service, this information isaddressed to a specific UE. The DC SAP provides access to services forthe establishment and release of an RRC connection to a specific UE. Thetransfer of messages using this connection is part of the DC SAP task aswell. When the RRC is instructed to establish a connection it sets up theSRBs. Furthermore, the DC SAP provides access to services for establish-ment and release of RBs which are to be used for data transfer in the user


plane. The release of the RRC connection causes all bearers in both thecontrol and user plane to be released as well. The RRC layer has controlconnections in the so-called management plane to all lower layers of the pro-tocol stack. These connections are used to configure the lower layers and toreceive measurement reports from them.

Within the RRC layer several functional entities can be distinguished.These are the Broadcast Control Functional Entity (BCFE), the Pagingand Notification Control Functional Entity (PNFE), the Dedicated ControlFunctional Entity (DCFE), the Shared Control Functional Entity (SCFE),the Transfer Mode Entity (TME) and the Routing Functional Entity (RFE).The BCFE, offering its services through the GC SAP, handles system infor-mation broadcast functions. To fulfill its task, the BCFE uses the BroadcastControl Channel (BCCH) as the logical channel for distributing the infor-mation in the downlink. Either the TM or UM RLC mode is used by theBCFE. There is at least one BCFE for each cell in the RNC.

Paging of UEs that do not have an RRC connection, i.e. they are in idlemode, is controlled by the PNFE. The PNFE is used to provide the RRCservices offered by the Nt SAP. Similar to the BCFE both the UM andTM mode may be used. The logical channel the PNFE makes use of is thePaging Control Channel (PCCH). At least one PNFE exists for each cellcontrolled by the regarded RNC.

The DCFE handles all functions and signalling specific to one UE. TheDCFE is used to perform the RRC tasks that are requested at the DCSAP. Depending on both, the message that is exchanged between the peerentities and the current UE service state, one of the three RLC modesapplies. In most of the cases the AM is used. The UM is applicable forthe transmission of messages like the RRC Connection Release or RRCConnection Reject while the TM carries messages like a Cell Update or RRCConnection Request. In the Serving Radio Network Controller (SRNC)there is one DCFE for each UE which has an RRC connection with thisRNC.

In Time Division Duplex (TDD) mode the DCFE is assisted by the SCFEwhich is located in the Controlling Radio Network Controller (CRNC). Thetask of the SCFE is to control the allocation of the Physical DownlinkShared Channel (PDSCH) and Physical Uplink Shared Channel (PUSCH)resources. In order to exchange this information the TM and UM modes ofthe RLC layer are used.


Located between the four entities described above and the RLC layer,the TME is responsible for the mapping between the different entities insidethe RRC layer and the SAPs provided by the RLC layer.

Routing of higher layer NAS messages to different Mobility Management(MM) and Connection Management (CM) entities at the UE side or differ-ent CN domains at the UTRAN side is handled by the RFE. The transfer ofthe NAS messages on RRC level is realized with the RRC Direct Transfermessages. Every NAS message is either carried by the Initial, Uplink orDownlink Direct Transfer message. Details about these and further PDUsexchanged by the RRC entities are described in the following paragraphs.

The control messages of the RRC protocol are exchanged by the RRCentities in the UE and the UTRAN using an Abstract Syntax NotationOne (ASN.1) based protocol [81, 82]. By applying the unaligned Packed En-coding Rules (PER) [83] a very compact encoding of the exchanged messagesis achieved. Figure 2.5 illustrates the tree-like data structure by showing asmall subset of the available nodes.

Message Class (Seq.)

Message Type (Choice) IE (Sequence, Optional)

Integrity

PDU (Sequence)

DL DCCH

DL DCCH

RB Setup

PDU (Sequence)

RB Release

PDU (Sequence)

RB Reconfiguration

IE (Sequence)

Frequency

Information

IE (Sequence)

Added Transport

Channel Information

IE (Array)

RB Information

Setup List

Figure 2.5: Abstract illustration of ASN.1 data structure

The ASN.1 data structure used by the RRC protocol can be groupedinto three main levels. On the highest level, the class definition level, the


message-class data types are located. Every message-class data type refersto an appropriate logical channel of the UMTS protocol stack and contains,apart from an optional integrity check message field, the message data type.The message data type contains all possible PDUs that may be transmittedover the corresponding logical channel. Hence, all possible configurationmessages which can be sent or received by the RRC protocol are assignedto one of these root message data types. A further differentiation betweenup- and downlink is done on the class definition level. The receiver of anRRC message has to identify the message-class type by the logical channelthe message is transmitted with.

The mid level contains the PDU definitions. On this level all PDUs arelisted without refering to the logical channel they are sent on. Every PDUconsists of Information Elements (IEs) which contain the actual information.The IEs, which can contain further IEs, describe the lowest level of theASN.1 data structure.

2.2.2 Packet Data Convergence Protocol

The PDCP sublayer [32] is the topmost sublayer of the user plane in theAS. As shown in Figure 2.6, the PDCP layer is located in the UMTSradio interface protocol stack above the RLC layer and uses the servicesprovided by the RLC protocol. To the higher layers the PDCP offers theso-called RBs for the transmission of user data. For the NL the PDCPacts as an adaptation layer for the transparent transmission of NL PDUs,e.g. IP packets. In this context the main task of the PDCP is to providefunctions for improving the channel efficiency by using header compressiontechniques.

The PDCP header compression is available only in the user plane forthe services of the PS domain. In this context significant improvementsfor PS services can be achieved [91]. Supported algorithms are the Inter-net Protocol Header Compression (IPHC) [56], which is an extension ofthe Van Jacobson compression [84], and the Robust Header Compression(ROHC) [52, 87]. The IPHC describes how to compress TCP, User Data-gram Protocol (UDP) and IP headers. The ROHC is a highly robustand efficient header compression method for Real-Time Transport Protocol(RTP)/UDP/IP, UDP/IP and Encapsulating Security Payload (ESP)/IPheaders over links with significant error rates and long round-trip times(e.g. cellular links). In contrast to IPHC the ROHC profiles available in


HC

Protocol 1

HC

Protocol 2

PDCP entity

HC

Protocol 1

HC

Protocol 2

PDCP entity

HC

Protocol 1

PDCP entity

PDCP sublayer

PDCP Control Radio Bearers

RLC UM RLC AM RLC TM

Figure 2.6: PDCP structure

UMTS do not support the compression of TCP/IP headers. Optionally,the PDCP may be configured to maintain sequence numbers for RBs whichsupport lossless Serving Radio Network Subsystem (SRNS) relocation orlossless Downlink (DL) RLC PDU size change.

The configuration of the PDCP layer is done by the RRC layer whichnegotiates the algorithms and their parameters. Each PDCP entity, corre-sponding to one RB, can be configured individually. Hence, a service specificparameter set can be established by the network. Important configurationparameters are the compression algorithms to be used by each PDCP entityand the mapping to RLC entities. One PDCP entity may use the servicesoffered by either one AM RLC entity or a combination of two UM or TMRLC entities. In the latter case one entity serves the DL, the other oneserves the Uplink (UL). A unidirectional configuration with one UM or TMentity is possible as well.

Both the IPHC and the ROHC may be configured for one PDCP entity.The compression protocols are distinguished by the Packet Identifier (PID)field of the PDCP protocol. Additionally, the IPHC uses the PID fieldto distinguish the different types of compressed headers (e.g. Full header,Compressed TCP and Compressed non TCP).

In general, both protocols make use of the fact that parts of the higherlayer headers change in a predictable way within one packet stream. Apacket stream which often could be interpreted as one session is generated


by grouping packets with similar characteristics by the so-called definingfields. Each packet stream is associated with a Context Identifier (CID)which is used at the receiver to map the packets to the correct compressioncontext.

The compression protocols are specific for the combination of higherlayer protocols, e.g. TCP/IP or RTP/UDP/IP, and distinguish betweencertain types of fields depending on the way they change within one packetstream. Static fields are not expected to change within one packet stream(e.g. protocol type and version [123], type of service or differentiated ser-vice [50, 112] as well as the source and destination IP addresses) and donot need to be transmitted. All fields which characterize a packet stream,i.e. the defining fields, belong to the group of static fields. Inferred fieldsare fields that can be regenerated from other fields at the receiver (e.g. IPpacket total length). Those header fields do not need to be transmittedas well. Fields which typically have a small difference between consecu-tive packets (e.g. TCP sequence number and acknowledgment number) arenot completely transmitted either. Instead the relative difference, the deltavalue, compared to the last uncompressed packet may be transmitted. Fi-nally, those fields which absolutely can not be predicted need to be alwayssent together with the delta compressed fields. An example for those fields isthe TCP checksum field. Figure 2.7 and Figure 2.8 illustrate how the fieldsof an Internet Protocol version 4 (IPv4) and a TCP header are categorized.

Version * IHL * Type of Service Total Length

Identification Flags Fragment Offset

Time to Live Protocol Header Checksum

Source Address *

Destination Address *

Options and Padding

0 3 4 7 8 15 16 18 19 31

No change in a packet stream

(* defining field)Inferred from other values

Delta encoded in

compressed headersNot present

Figure 2.7: Header compression for IPv4


Source Port * Destination Port *

Sequence Number

Window

Checksum

Options and Padding

0 3 4 109 15 16 31



Delta encoded in

compressed headers

Random field (must be included

in compressed headers)

Acknowledgment Number

Offset Reserved AU P R S F

Urgent Pointer

Figure 2.8: Header compression for TCP

As can be seen most of the header fields are either static fields or fieldswhich can be differentially encoded. Especially for delay sensitive low data-rate traffic with small packets, e.g. Voice over IP (VoIP), the PDCP headercompression techniques can significantly reduce the header overhead. Thecompression gain is even higher for Internet Protocol version 6 (IPv6) trafficas most of the header fields in IPv6 are static and the uncompressed IPv6header is, mainly because of the longer address fields [69], twice as big as astandard IPv4 header. Figure 2.9 shows such an IPv6 header.

Version * Traffic Class Flow Label *

Payload Length Next Header Hop Limit

Source Address *

Destination Address *

0 3 4 15 16 31



12 13 23 24

Figure 2.9: Header compression for IPv6


2.2.3 Broadcast/Multicast Control

Similar to the PDCP the BMC [33] is an entity existing only in the userplane. The task of the BMC is to implement the CBS where cell specificmessages can be broadcasted to all UEs within one cell. Therefore, thereexists one BMC entity per cell in the UTRAN.

Figure 2.10 illustrates the model of the BMC and its position in theUTRAN protocol architecture. The BMC is located above the RLC anduses the UM mode of the RLC for the DL broadcast of Cell Broadcast (CB)messages on the Common Traffic Channel (CTCH). The service of theBMC is offered to higher layers by the BMC SAP. The functions of theUTRAN BMC entity are the storage, the scheduling and the transmissionof messages for the CBS. The configuration of the radio resources and thereports about traffic volume measurements are exchanged using a controlSAP which is used by the RRC layer. The Cell Broadcast Center (CBC)is in charge of delivering the messages to the correct BMC entity which inturn broadcasts it to the corresponding cell. After receiving a CB messagethe BMC entity in the UE delivers the message to the higher layers.

BMC entity

RLC UM entity

BMC sublayer

RLC sublayer

RRC

BMC-SAP

UM-SAP

CTCH-SAP

user-plane

CBMC-SAP

Figure 2.10: Model of the Broadcast Multicast Control sublayer


2.2.4 Radio Link Control

The RLC protocol, specified in [31], provides segmentation and retransmis-sion services for both user and control data. In the control plane the RLCprovides SRBs as its service to the RRC. In the user plane the PDCP orthe BMC make use of the services offered by the RLC. If neither PDCPnor BMC is used, the RLC provides the RB service directly to the NL.Considering the functionality of the entities there is no difference betweenentities in the control plane and the user plane.

The RLC protocol provides three different data transfer services to thehigher layers. These are the TM, the UM and the AM data transfer service.

As illustrated in Figure 2.11 the TM and the UM data transfer is aunidirectional service, while the AM always exists in a bidirectional config-uration. Hence, the TM and UM entities are split into transmitting andreceiving entities. A bidirectional connection between two RLC entities inUM or TM consists of two RLC connections, one for the uplink and theother one for the downlink. The AM entity combines both the transmittingand receiving part in order to support the ARQ mechanism. No differenceexists between entities in the UE and the UTRAN from the functionalitypoint of view.

The configuration of the RLC layer is done by the RRC using the RLCcontrol SAP. The RRC is in charge of establishing, releasing and configur-ing the individual RLC entities. The entities are addressed by the logicalchannel the respective entity is responsible for. Furthermore, the RRC maysuspend and resume RLC entities. UM and AM entities have to be sus-pended during the security mode control procedure of the RRC when newciphering keys are going to be configured for those RLC entities in a de-

Transm.TM entity

Transm.UM entity

ReceivingUM entity

AM entityReceivingTM entity

Transm.TM entity

Transm.UM entity

ReceivingTM entity

ReceivingUM entity

AM entity

Uu Interface

Logical channels

UE UTRANTransmitting side Receiving side Transmitting side Receiving side

Figure 2.11: Overview of the RLC sublayer


terministic way. Details about the individual configuration parameters aredescribed in the following sections. In case of unrecoverable errors the RLCentities may send status information to the RRC.

2.2.4.1 Transparent Mode

The TM is a unidirectional service mostly used for the transmission of datastreams (e.g. audio or video), where the latency of the transmission is moreimportant than the error rate, or for broadcast purposes. When using thetransparent data transfer service the RLC protocol adds no header to thehigher layer payload. Hence, the RLC PDU just contains the payload (theRLC SDU). No requirement exists that the payload size has to be a multipleof 8 bits. As there is no Sequence Number (SN) included in the TM PDU,the TM has no chance to detect missing PDUs at the receiving entity.

A Limited segmentation and reassembly functionality is available in theTM. Segmentation of one RLC SDU can take place if all segments of thisSDU can be transmitted within one Transmission Time Interval (TTI). TheTTI is defined by the interval at which data is sent on a transport channel(see Section 2.2.5.2). When segmentation is configured by the RRC thereceiving entity reassembles all RLC PDUs received in a single TTI to oneSDU.

A timer based SDU discard mechanism is possible to be used as well.After an RLC SDU has been queued for a configured amount of time, thesender drops this SDU without any further notification to higher layers orthe peer entity.

Figure 2.12 shows the model of two interacting TM entities with thelogical channels which can be used by them.

The logical channel can be either BCCH, PCCH, Common ControlChannel (CCCH), Shared Channel Control Channel (SHCCH), DedicatedControl Channel (DCCH) for the control plane or Dedicated Traffic Channel(DTCH) for the user plane. The BCCH and PCCH are downlink only chan-nels. The CCCH and SHCCH may be used in both uplink and downlink.Nevertheless, the TM is only used in an uplink configuration.


Transmissionbuffer

Segmentation Receptionbuffer

Reassembly

DTCH/DCCHCCCH/SHCCH (UE)

BCCH/PCCH (UTRAN)

DTCH/DCCHCCCH/SHCCH (UTRAN)

BCCH/PCCH (UE)

TransmittingTM entity

ReceivingTM entity

TM-SAP TM-SAP

UE/UTRAN UTRAN/UEUu Interface

Figure 2.12: Model of Transparent Mode peer entities

2.2.4.2 Unacknowledged Mode

The UM is mainly used in certain RRC signalling procedures, where theacknowledgment and retransmission is part of the signalling procedure itself.VoIP is also a feasible service to be carried by the UM. The UM data transferservice does not guarantee the delivery to the peer entity. Figure 2.13 showsthe model of two interacting UM entities and the logical channels which maybe used by the UM.

In contrast to the TM a more flexible segmentation and reassemblymechanism exists for the UM. Segments of variable sized RLC SDUs mightbe transmitted with multiple RLC PDUs which can also be transmittedin different TTIs. The flexible segmentation mechanism is achieved by in-troducing Length Indicators (LIs) to the UM PDU header. With the helpof these LIs it is furthermore possible to transmit segments of consecutiveRLC SDUs in one PDU. This mechanism is called concatenation. If anRLC PDU can not be completely filled by segments of higher layer payloadthe UM adds padding octets to the PDU. This is achieved by a specialvalue of the LI. The LI can either be 7 or 15 bit long and it indicates thenumber of octets of a segment an RLC PDU contains. Therefore, the UMcan only be applied for octet aligned higher layer protocols. A LI of 7 bits


Reassembly

RemoveRLC header

Reception buffer

Deciphering

Transmissionbuffer

Segmentation &Concatenation

Add RLC header

Ciphering

TransmittingUM entity

ReceivingUM entity

DTCH/DCCHCTCH/CCCH/SHCCH (UTRAN)MTCH/MCCH/MSCH (UTRAN)

DTCH/DCCHCTCH/CCCH/SHCCH (UE)MTCH/MCCH/MSCH (UE)

UE/UTRAN Uu Interface UTRAN/UE

UM-SAP UM-SAP

Figure 2.13: Model of Unacknowledged Mode peer entities configured withoutduplicate avoidance and reordering

is used in case the UM PDU is smaller than 126 bytes while a LI of 15 bitsis needed for larger PDUs. A maximum PDU size of 624 bytes is allowedfor the UM mode.

By including a 7 bit SN in the UM PDU header the receiving entity isable to perform a SN check and is, therefore, capable of detecting missingPDUs. SDUs which are affected by missing UM PDUs are not deliveredto the higher layer. Error recovery in terms of Backward Error Correction(BEC) is not performed as the UM provides a unidirectional connectionwithout the possibility to use an ARQ mechanism.

To avoid the transmission of UM SDUs which have been queued for acertain amount of time and are, therefore, useless for the receiving side,e.g. in case of a VoIP service, the UM includes an SDU discard mechanism.When the configurable timer of such an SDU expires it is simply dropped.

An additional feature of the RLC UM in comparison to the TM is theciphering of the UM PDUs. When this feature is enabled by the RRCthe transmitting entity encrypts the user data and the receiving side isdecrypting it. By doing so the user data is protected from unauthorized


acquisition at the radio interface. In order to also secure data of the TMa ciphering mechanism in the MAC layer replaces this mechanism typicallyapplied in the RLC layer.

In Release 6 of the 3GPP specification the UM has been extended inorder to support the Multimedia Broadcast Multicast Service (MBMS).For the MBMS downlink data may be received by a UE from differentcells. Hence, the UE side of the UM is extended to also contain a dupli-cate avoidance and reordering mechanism. Out of sequence SDU delivery isan optional functionality as well. Figure 2.14 shows the configuration andthe logical channels of the UM in case it is used in an MBMS scenario.All MBMS channels, which are the MBMS point-to-multipoint ControlChannel (MCCH), the MBMS point-to-multipoint Traffic Channel (MTCH)and the MBMS point-to-multipoint Scheduling Channel (MSCH), are down-link only channels and always make use of the UM. The same applies to theCTCH. Both SHCCH and CCCH make use of the UM only in the downlinkdirection. For the DTCH and the DCCH the usage of the UM is optionalas all RLC modes can be used for the dedicated channels.

Reassembly

RemoveRLC header

Reception buffer

Duplicateavoidance and

reordering

Transmissionbuffer


Add RLC header

TransmittingUM entity

ReceivingUM entity

MTCH (UTRAN) MTCH (UE)

UTRAN Uu Interface UE

UM-SAP UM-SAP

Figure 2.14: Model of Unacknowledged Mode peer entities as used for MBMS


2.2.4.3 Acknowledged Mode

The AM is used for the error-protected transmission of both user data orsignalling data. Error protection is achieved by a BEC retransmission mech-anism. This ARQ functionality is needed for the provision of services whereerror-free delivery is mandatory. In order to provide a BEC an AM RLCentity always is deployed in a bidirectional configuration. Figure 2.15 showsthe model of the AM entity. As can be seen it is possible to use either one ortwo logical channels for the AM RLC connection. If two logical channels areused the AM entity is able to use one channel exclusively for data transferand the other one for the transfer of acknowledgments and other protocolinformation elements.

From the perspective of the provided functionality the AM inherits thecomplete UM functionality. Segmentation and reassembly as well as con-catenation and padding are provided in a very similar way. Also ciphering,as described in the previous section, is available. A timer based SDU dis-card mechanism as well as a discard mechanism based on the number ofretransmissions of associated RLC PDUs can optionally be enabled. If en-abled the AM supports explicit signalling of PDUs which are discarded atthe sending side. By transmitting a status PDU with Super Field (SUFI)Move Receiving Window (MRW) the reception window of the AM can bemoved even though not all PDUs have been received. The receiving entityin turn confirms this discard with the MRW acknowledgment SUFI. Upperlayers at the receiving entity can be informed about SDUs which are notdelivered due to an RLC discard.

Because multiple copies of an AM PDU may be sent by the retrans-mission based AM protocol a duplicate detection is always enabled for thereceiving part of an AM entity. The AM provides the mechanism to pre-serve the order of higher layer PDUs. If this in-sequence delivery is notconfigured an RLC SDU is directly delivered to the higher layer when thelast PDU containing segments of this SDU is received.

Using a sequence-number check function in the receiving entity the AMis capable of detecting PDU losses which are caused by lower layer trans-mission errors. Missing PDUs are requested for retransmission from thesending entity, thus providing an error-free transport service for the higherlayers. In order to do so each AM entity has two buffers, a TransmissionBuffer and a Retransmission Buffer. Every AM PDU is stored in the Re-transmission Buffer until its successful transmission is acknowledged by the



Add RLC header

Retransmissionbuffer &

management

Transmission buffer

Ciphering (only AMD PDU)

Deciphering

Demultiplexing & Routing

Reception buffer &retransmission management

Remove RLC header& extract piggybacked

information

Reassembly

RLC Control Unit

Piggybacked StatusInformation

ReceivedAcknowledgements

Acknowledgements

Set fields in PDU Header(e.g. set poll bits) &

piggybacked STATUS PDU

Multiplexing

AM-SAPUE/UTRAN

Receiving sideTransmitting side AM entity

DTCH/DCCH

DTCH/DCCH

DTCH/DCCH

DTCH/DCCH

(optional secondlogical channel)

(optional secondlogical channel)

Figure 2.15: Model of an Acknowledged Mode entity

peer entity. In case of a negative acknowledgment the Control Unit ordersa retransmission of the erroneous PDUs which are then copied from the Re-transmission Buffer back into the Transmission Buffer. By doing so a PDUcan be transmitted multiple times until it is received correctly. In case aPDU can not be delivered even after a number of retransmissions and ifno discard mechanism is configured, this unrecoverable error is signalled tohigher layers and the AM entity executes a reset procedure to enable thecontinuation of data transfer. A reset is also triggered if status report in-formation is inconsistent or if information within data or control PDUs isinvalid. The reset procedure consists of a special control PDU, the reset


PDU, that is sent to the peer-entity which in turn responds with a resetacknowledgment PDU.

If requested by higher layers the AM may indicate with a transmissionconfirmation the successful delivery of an SDU to the AM peer-entity. AMessage Unit Identifier (MUI) is used to reference the SDU in a later con-firmation.

In UMTS the RLC always applies a selective repeat ARQ mechanism. Incontrast to a stop-and-wait ARQ the sending RLC entity can transmit PDUsas long as their SN is within the transmission window without waiting for anacknowledgment. This avoids idle times during the transmission which arecaused by the sender if it would need to wait for a positive acknowledgmentfor every single PDU. To be able to limit the amount of unacknowledgedPDUs the receiving entity might implement flow-control mechanisms bychanging the size of the transmission window of the peer-entity. In UMTSthe maximum transmitting window size can be configured between 1 to4095 RLC PDUs. This limitation is in relation to the 12 bit SN used bythe AM data PDUs. The possibility of the selective repeat ARQ to signalboth positive and negative acknowledgments for received PDUs allows theRLC to only retransmit those PDUs which are not correctly received. Thismakes the RLC ARQ more efficient compared to the go-back-n ARQ whereall PDUs starting from the first missing PDU need to be retransmitted. Thedisadvantage of this is the necessary receiving buffer and a higher effort forthe retransmission management.

The UMTS standard does not specify a special sub-type of the selectiverepeat ARQ. Various mechanisms are possible to be used with the availableSUFIs of the status PDUs which exist for the RLC peer-to-peer communica-tion. Those mechanisms are differentiated by the way missing and correctlyreceived PDUs are reported to the peer entity. The basic mechanism of ac-knowledging PDUs is the use of the acknowledgment SUFI which indicatesthe next expected SN. All PDUs up to the PDU with the signalled SN areacknowledged by this field. Further SUFIs are available for the reporting ofmissing PDUs using lists, relative lists or bitmaps. A list contains pairs ofnumbers where the first element of the pair is the SN of a missing packetand the other one gives the number of consecutive missing packets. A rela-tive list uses 4 bit codewords that determine the distances between missingpackets. A bitmap uses a mapping of single bits to data packets to specifymissing and correctly received packets. Both relative list and bitmap re-ports contain a 12 bit First Sequence Number (FSN) field pointing to the


first missing PDU. A length field exists in every status report SUFI whichspecifies the number of value pairs, codewords or octets in the list, relativelist or bitmap, respectively. The RLC entity may dynamically select oneof the available methods for reporting and is therefore able to reduce theacknowledgment size and use the channel capacity as efficient as possible. Itis even possible that a status PDU containing several SUFIs is piggybackedin a normal AM data PDU. Special LI values are defined to indicate thepresence of such status PDUs. Every RLC entity must be able to processall kinds of status reports. The selection of the type of reporting is notspecified and is, therefore, implementation dependent.

Several methods to trigger the exchange of status reports are specifiedby the 3GPP. An AM entity can send a status report periodically or when-ever a missing PDU is detected. Furthermore, a so-called poll bit withinevery data PDU allows an entity to explicitly request an immediate statusreport from the peer-entity. This explicit poll request can be sent whenone of seven configurable criteria is fulfilled. Each criterion may be enabledor disabled by the RRC layer which is in charge of the RLC configuration.A poll can be triggered if the last PDU of the transmission buffer or theretransmission buffer is going to be sent. Furthermore, the sender may setthe poll bit statically for every n-th PDU as well as every m-th SDU. A win-dow based polling mechanism, which polls whenever a certain percentage ofthe transmission window is used, prevents the case of a closed transmissionwindow. A simple timer based mechanism allows to periodically request astatus report from the peer-entity. Whenever the transmitter sends a PDUwith the poll bit set it may start the so-called poll timer to assure that acorresponding status report is received. Upon expiration and if no statusreport for the PDUs up to that PDU is received a new poll is started.

To avoid deadlocks the specification requires that the RRC must eitherenable the combination of poll timer, last PDU in buffer and retransmissionbuffer or, alternatively, the timer based polling must be used. In order toavoid excessive signalling by too many status reports a poll prohibit timerand a status probibit timer can be used to prevent the transmission of morethan one poll request or status report within a certain amount of time.

The logical channels which can be used by the AM are the DTCH in theuser plane and the DCCH in the control plane.


2.2.5 Medium Access Control

In this section a detailed description of the UMTS MAC layer is givenregarding its services provided to the higher layers, its functions and theinterface to the lower layer. The most important tasks of the MAC layer,documented in the 3GPP specification [30], are the following ones. One ofthe main functions of the MAC layer is to map the logical channels, whichare offered as a service to the RLC layer, onto transport channels, whichcharacterize the interface between the MAC layer and PHY. This mappingas well as the establishment and release of channels is controlled by the RRClayer. Section 2.2.5.1 and 2.2.5.2 give an overview about all available logicaland transport channels, respectively. Because multiple logical channels maybe mapped onto the same transport channel the MAC layer is in charge ofmultiplexing the different data flows. Both traffic flows of one user or ofdifferent users can be multiplexed on one transport channel. In order todemultiplex received PDUs the MAC layer adds header fields to the MACSDUs which are used to identify different users on shared channels, differentservices received by multiple users and different logical channels belongingto one user. When transmitting SDUs of different logical channels or usersthe MAC layer is in charge of scheduling the data flows based on theirpriorities and further scheduling parameters.

On request by the RRC layer the mapping of logical to transport chan-nels can dynamically be modified. This MAC function is called transportchannel type switching. The RRC may base this decision on traffic volumemeasurements which are performed by the MAC layer. These measurementsinclude information about the Buffer Occupancys (BOs) of the RLC enti-ties which make use of the logical channels. As UTRAN is in charge of thetransport channel configuration the UE measurements as well as the UEconfiguration are exchanged by the RRC protocol.

Based on this amount of data to be transmitted and the available radioresources as well as the available transmission power and the instantaneouschannel conditions the MAC layer chooses appropriate Transport Formats(TFs) and Transport Format Combinations (TFCs) on a TTI basis. A TFspecifies the number and size of RLC PDUs which can be transmitted withinone TTI. For every transport channel there exists one Transport FormatSet (TFS) of possible TFs. For transport channels which share resourcesin the PHY the MAC layer takes care of selecting valid TFCs out of theTransport Format Combination Set (TFCS). This is performed in order to


BCCH MTCH MSCH MTCH MSCH MCCHPCCH BCCH CCCH CTCH SHCCHMAC Control DCCH DTCH DTCH

Associated

Downlink

Signalling

E-DCH

Associated

Uplink

Signalling

FACHBCH

Associated

Downlink

Signalling

Associated

Uplink

Signalling

HS-DSCH FACHRACHPCH FACH USCH DSCH DSCHUSCH

MAC-b MAC-mMAC-e/es MAC-hs MAC-c/sh/m

MAC-d

DCH DCH(TDD only) (TDD only) (TDD only) (TDD only)

(TDD only)

Figure 2.16: UE side MAC architecture

assure that a maximum bit rate given by the available physical resources isnot exceeded. The selected TFs are sent together with the data that is tobe transmitted to the PHY for further processing. For transport channelswhich are using link adaptation the MAC layer controls the transmissionand reception parameters as well as the Hybrid ARQ (HARQ) functionality.

Depending on the type of transport channel, different network elementsand MAC entities are involved. Figure 2.16 and Figure 2.17 give an overviewof these MAC entities and the logical and transport channels they are con-nected to. Common to all entities is the MAC control SAP which connectsall MAC entities to the RRC layer. This SAP is used by the RRC to config-ure the MAC entities, e.g. establishment and release of channels, and to getperiodical and event triggered measurement results and status informationback. An event triggered report is based on configurable thresholds of thedata volume stored in RLC while a periodical report is sent in configurablereporting intervals. Both modes of traffic volume reporting can be requestedby the RRC layer at the same time. In addition to the absolute BOs theMAC layer may also be configured to report the average BOs and the vari-ance of the BOs. In order to do so MAC receives with every data request oras a stand-alone signal the status of the entities’ transmission buffer wherethis request originated from.


MAC-eMAC-b

MAC-d

MAC-c/sh/mMAC-hs

MAC-es

BCH E-DCH

Associated

Downlink

Signalling

Associated

Uplink

Signalling

HS-DSCH

Associated

Downlink

Signalling

Associated

Uplink

Signalling

PCH FACH FACH RACH USCH USCH DSCH DSCH DCH DCH(TDD only) (TDD only) (TDD only)(TDD only)

MAC Control DTCHDTCHDCCHBCCH MAC Control

Iub Iur or local

Configuration with MAC-c/sh/m

Configuration without MAC-c/sh/m

BCCH CCCH CTCH SHCCHMCCHMSCH MTCH

MAC

ControlPCCB(TDD only)

Figure 2.17: UTRAN side MAC architecture

The structure and tasks of the MAC entities which are relevant for the per-formance as studied within this thesis are introduced in the Sections 2.2.5.3and 2.2.5.4. The remaining MAC entities which are implemented but donot directly affect the performance are described in Appendix A.

2.2.5.1 Logical Channels

The MAC layer offers the logical channels as its service to the RLC layer.Logical channels are characterized by the type of information transferred bythem. In general logical channels can be grouped into control channels andtraffic channels.

2.2.5.1.1 Traffic Channels

The purpose of the traffic channels is to transfer user plane data. Availabletraffic channels are the DTCH, CTCH and MTCH.

The DTCH is a point-to-point channel dedicated to one UE. A DTCHcan exist in uplink and downlink. Multiple DTCHs can be established forone single UE.

The CTCH is a point-to-multipoint channel existing in downlink only.User information can be transferred to all or a group of UEs. In each RNCthere exists one CTCH for every cell which is controlled by that RNC.


Similar to the CTCH the MTCH is a downlink point-to-multipoint chan-nel which is used for MBMS traffic. An MBMS service can be broadcastedby multiple cells. In this case the UE is configured to receive correspondingMTCHs from all those cells and the RLC is doing a selection combining ofthe received content.

2.2.5.1.2 Control Channels

The control channels transfer control plane information only. This signallinginformation is transmitted by the DCCH, CCCH, SHCCH, BCCH andPCCH. The MBMS service add the channels MCCH and MSCH to theset of logical control channels.

The DCCH is a bidirectional point-to-point channel between a UE andthe UTRAN. A DCCH is established by the RRC connection setup proce-dure and carries dedicated control information for one UE. Multiple DCCHswhich carry the data of SRBs exist for each UE.

The CCCH is a bidirectional channel that is shared between all UEs.There is no need for the UEs to have an RRC signalling connection to usethis channel (e.g. initial random access). The negotiation of the attributesof the SRBs between the RRC of the UTRAN and the UE is an examplefor the use of this channel. The UE is typically using the CCCH channelfor initial cell selection or cell reselection.

Similar to the CCCH the SHCCH is a bidirectional channel. The pur-pose of the SHCCH is the transmission control information for the Down-link Shared Channel (DSCH) and Uplink Shared Channel (USCH) betweenUTRAN and UE. The SHCCH exists in TDD only since its associatedshared transport channels are only used in this mode of operation.

The BCCH and the PCCH are downlink only channels. The BCCHbroadcasts system control information and the PCCH is used to indicateto a UE that it should establish a dedicated connection to the network.This paging is performed when the UE can not be reached by dedicatedsignalling. The network might not know the exact location of the UE oncell level.

The MCCH and the MSCH are control channels added for the MBMSservice in Release 6 of the 3GPP specification. Both are downlink point-to-multipoint channels. The MCCH is responsible for transmitting MBMScontrol information from UTRAN to the UEs. The MSCH is used to sendscheduling control information for the associated MTCHs to the UEs.


2.2.5.2 Transport Channels

The interface between the MAC layer and the PHY is characterized by theso-called transport channels. Transport channels are distinguished by theway the data is transferred over the radio interface. The units which areused to transfer data are called Transport Blocks (TBs). TBs are bit stringswith a length not necessarily being a multiple of eight bits. Multiple equallysized TBs which are transmitted within one TTI on one transport channelform a so-called Transport Block Set (TBS). A TTI is typically 10, 20, 40 oreven 80 ms long. With the introduction of the HS-DSCH and the EnhancedDedicated Channel (E-DCH) a 2 ms TTI has been added for those channels.

The characteristics of a TBS during a TTI on one transport channelare described by the TF. A TF consists of two parts, a dynamic part anda semi-static part. The dynamic part contains information about the TBsize, the TBS size and, in case of TDD, the TTI length. In FrequencyDivision Duplex (FDD) the TTI length is fixed for each transport channeland, therefore, is part of the semi-static part. Furthermore, physical layerparameters used for error detection and correction, e.g. CRC length andtype of FEC, as well as the rate matching parameters belong to the semi-static part. The dynamic part varies between TFs of one TFS. Parametersbelonging to the semi-static part are the same for all TFs within a TFS.The semi-static parameters can only be changed by a transport channelreconfiguration done by RRC.

The transport channels can be divided into dedicated and common trans-port channels. Details about both types are given in the following.

2.2.5.2.1 Dedicated Transport Channels

Dedicated transport channels provide a connection between the UTRANand one designated UE. The UE is identified by the characteristics of thephysical channel, i.e. frequency and code. In TDD mode the timeslot is anadditional characteristic. The MAC standard [30] specifies two dedicatedtransport channels which are the DCH and the E-DCH.

The DCH is dedicated to one UE. It exists in uplink and downlink.Both unidirectional and bidirectional configurations are possible. A UE canhave more than one DCH dedicated to it. For the DCHs features like fastpower control, soft handover, beam forming with smart antennas and theopportunity of fast rate changes, i.e. every 10 ms TTI, are applicable.


The E-DCH is an uplink only transport channel. Introduced in Release 6the E-DCH extends the DCH with features which originally have been in-troduced for the HS-DSCH. Primarily these features are link adaptationand HARQ. The E-DCH has a mandatory TTI of 10 ms. Furthermore, anoptional TTI of 2 ms has been added to the E-DCH feature scope. Hence,a fast rate change within a 2 ms interval can be achieved. Inner loop powercontrol is available for the E-DCH by always being associated to anotherfast power controlled channel.

2.2.5.2.2 Common Transport Channels

Common transport channels are shared between all UEs. Hence, in-bandidentification is required if not all UEs should be addressed. The identifica-tion of UEs is a task performed by the MAC layer. The following channelsare included in the Release 6 of the MAC specification:

The UTRAN broadcasts system information into an entire cell with theBroadcast Channel (BCH). There exists one BCH in each cell. As a UEcan not register to the network without the possibility to decode the BCHthis channel transmits with relatively high power in order to reach all UEswithin the intended coverage area. Since all UEs have to be able to decodethe BCH it is fixed to a low data rate of 12.3 kbit/s. The BCH always hasone fixed TF which is known by all UEs and a TTI of 20 ms.

The UTRAN broadcasts control information like paging and notificationinformation elements with the Paging Channel (PCH) into an entire cell.Since the initiation of the communication with a UE, e.g. because of aspeech call, is done by the PCH the PCH has to be received in the wholecell. The physical layer allows for efficient UE sleep mode procedures toreduce power consumption using PHY signals called page indicators. OnePCH is allowed in each cell. The TTI of the PCH is 10 ms for FDD and20 ms for TDD.

The Random Access Channel (RACH) is a contention based uplink chan-nel used for the transmission of small amounts of data, e.g. for initialaccess or location update. It can also be used to send small amounts ofpacket data like a Uniform Resource Identifier (URI) in a Hypertext Trans-fer Protocol (HTTP) request before the DCHs are established. In princi-ple the RACH uses the Slotted ALOHA protocol [137] to resolve collisionsbetween UEs. A maximum theoretical data rate of 16.8 kbit/s is sharedbetween all UEs in one cell. Several RACHs may be configured in one cell.


The RACH uses open loop power control and has a TTI of 10 ms or 20 ms.In TDD mode only the 10 ms TTI is used.

The Forward Access Channel (FACH) is a common downlink channelused for the transmission of small amounts of data [70]. In one cell therecan be more than one FACH. Because the FACH is used to answer requestson the RACH there has to be at least one FACH with such a low bit ratethat it can be received by all UEs in the regarded cell. Additional FACHsmay have higher data rates. Fast rate change on a 10 ms TTI basis is apossible feature of the FACHs. The FACH lacks the ability to use fast powercontrol, only slow power control is utilizable.

The DSCH and the USCH are channels shared by several UEs. Bothcarry dedicated control or traffic data and exist in TDD mode only. Morethan one DSCH and USCH may be configured in a TDD cell. Beam formingis an optional feature of both channels which is applicable in certain con-figurations. Without using beam forming the DSCH may be broadcastedin an entire cell. The DSCH is always associated with a DCH or a FACH.When associated with a DCH the DSCH is able to use inner loop powercontrol. Otherwise only slow power control is available. The USCH canalso be power controlled and is able to change its rate fast. As the USCH isan uplink channel features like uplink synchronisation and timing advanceare available.

The HS-DSCH is a downlink channel which is shared by serveral UEs.A fixed TTI of 2 ms for FDD and 10 ms for TDD is used by the HS-DSCH.Data to up to four UEs may be transmitted within each TTI. The targetedUEs and the amount of data are selected by the fast scheduling taking theindividual channel conditions into account. The addressing of these UEs isdone by one or more associated shared physical channels, the High SpeedShared Control Channels (HS-SCCHs), which also carry information neededfor the decoding of the HS-DSCH data. The HS-DSCH is using HARQ andis able to do link adaptation by varying the modulation, channel codingand transmission power. When addressing individual UEs the HS-DSCHmay make use of beam forming but also a broadcast to the entire cell isavailable. HARQ feedback and channel quality information is transmittedin the uplink by an associated signalling channel, the High Speed DedicatedPhysical Control Channel (HS-DPCCH). In contrast to DCH and E-DCHno macro diversity is applied to the HS-DSCH.


2.2.5.3 Dedicated Channels and the MAC-d Entity

The MAC-d (MAC-dedicated) entity is responsible for handling the DCHsallocated to a UE. There is one MAC-d entity in the UE and one MAC-dentity in the UTRAN for each UE. In the UTRAN the MAC-d entity islocated in the SRNC which is in charge of the corresponding UE. Figure 2.18shows the structure of the MAC-d entity in both the UE and the UTRAN.Compared to the other MAC entities the MAC-d entity is constructed ina mostly symmetrical way in terms of UE and UTRAN functionality. Thelogical channels which can be mapped on the DCH are the DCCH and theDTCH. Alternatively, these logical channels can be mapped to commontransport channels. In order to do so the MAC-d entity has a functioncalled Transport Channel Type Switching. On request by RRC the MAC-d entity may forward higher layer PDUs to other MAC entities which arein charge of the corresponding transport channels. In detail these are theMAC-c/sh/m for the RACH and FACH as well as the DSCH and USCH,the MAC-hs for the HS-DSCH and the MAC-e/es for the E-DCH.

In case one of these entities is not in the same network element theMAC-d entity applies Flow Control mechanisms to limit the buffering inthis network element. It is important to keep the amount of buffered datasmall so that the RLC ARQ mechanism can work in an efficient way. Whena DTCH or DCCH is mapped on the HS-DSCH the Flow Control is alwaysnecessary as the MAC-hs entity is located in the Node B. When a mappingto the FACH or DSCH is configured the Flow Control is needed becausethe CRNC containing the MAC-c/sh/m entity is not necessarily the SRNCwhere the MAC-d entity is located. In case the entities are in the samenetwork element (i.e. MAC-es and MAC-d in the SRNC as well as allentities in the UE) Flow Control is not needed. A detailed description ofthe Flow Control mechanisms on the Iur interface can be found in [35].

When more than one DTCH or DCCH is mapped on the same transportchannel the MAC-d entity is in charge of multiplexing/demultiplexing theindividual data flows. By introducing a 4 bit multiplexing header, the so-called Control/Traffic (C/T) field, up to 15 different logical channels carriedon the same transport channel can be distinguished. If there is a one-to-one relationship between a logical channel and a transport channel no C/Theader field is added to the MAC SDU.

This multiplexing also implies a scheduling to be done by the MAC-dentity. Based on the MAC Logical Channel Prioritys (MLPs) of the logical


C/T MUX

Transport Channel Type Switching

Deciphering

C/T MUX

UL: TFC selection

Ciphering

DCH DCH

DCCH DTCH DTCHMAC Control

MAC-d

from MAC-hs

to/fromMAC-c/sh/m

to MAC-e/es

(a) UE side

C/T MUX /Priority setting

(DL)

Flow Control

Transport Channel Type Switching

Deciphering

C/T MUX

DL scheduling /priority handling

Ciphering

MAC-d

DCCH DTCH DTCHMAC Control

DCH DCH

from/toMAC-c/sh/m

to MAC-hs

from MAC-es

(b) UTRAN side

Figure 2.18: MAC-d architecture

channels and the BOs of the RLC entities the MAC-d scheduler requestsa certain amount of RLC PDUs from each RLC entity. The amount ofdata which can be transmitted on a DCH within one TTI is given by theselected TF. Every DCH has a TFS out of which the MAC-d schedulerselects an appropriate TF every DCH specific TTI, i.e. every 10, 20, 40 or80 ms. Because multiple DCHs may be jointly encoded on the same CodedComposite Transport Channel (CCTrCH) in the PHY not all combinationsof TFs are possible. In order to not exceed limitations within the PHY inthis case, e.g. puncturing limits or transmission power, a restricted set ofTFCs is given by the TFCS. This set of allowed TFCs is configured by theRRC layer.

The data unit exchanged between the MAC-d entity and the physicallayer is called TBS. A TBS consists of TBs which correspond to the ex-changed MAC PDUs. These PDUs may contain MAC SDUs originatingfrom different logical channels. Even if this multiplexing is applied the TBswithin one TBS are of the same size. The size of the TBs and the number


of TBs transmitted within one TTI are given by the dynamic part of theTFs. In TDD mode even the TTI length belongs to the dynamic part ofthe TFs and, therefore, it may vary between the TFs of one TFS. Thesemi-static part of a TF mostly covers parameters relevant for the PHY likeerror protection scheme and CRC size. Within one TFS the semi-static isthe same for all TFs.

Additionally to the scheduling of PDUs transmitted on the DCHs theMAC-d entity in UTRAN has the task to set the priorities associated withthe PDUs which are transmitted on shared channels. The forwarding ofthis priority information to the MAC-c/sh/m or MAC-hs entity allows thoseentities to schedule individual UEs and traffic flows of a UE.

For logical channels connected to TM RLC entities the MAC-d entityperforms ciphering and deciphering. Only when these logical channels aremapped onto a DCH this enciphering is possible. The ciphering unit is theMAC SDU. The MAC header is not ciphered. The selection of the cipheringalgorithm and the key configuration are done by the RRC using the controlSAP of the MAC-d entity. Details about ciphering can be found in [42].

2.2.5.4 High Speed Downlink Packet Access

The MAC-hs entity, introduced in Release 5 of the UMTS specification,is responsible for handling the HSDPA functionality in the MAC layer.Figure 2.19 illustrates the architecture of this entity at both the UE andUTRAN side. The transport channel the MAC-hs entity is in charge of isthe HS-DSCH. Additionally to this transport channel further uplink anddownlink signalling channels are required by the MAC-hs entity. There is nological channel directly mapped to the MAC-hs entity. Instead the MAC-hsentity is connected to the MAC-d entities and, therefore, carries traffic ofthe DTCH and DCCH. These connections are achieved by so-called MAC-dflows. In the UE the MAC-hs entity is directly connected to the MAC-dentity, while in UTRAN two alternatives exist [29]. Either MAC-hs andMAC-d have a direct connection, like in the UE case, or the MAC-d trafficis routed via the MAC-c/sh/m entity (compare Figure 2.16 and Figure A.3).

In the UTRAN there exists one MAC-hs entity for every cell which sup-ports HS-DSCH transmission. These MAC-hs entities are located in theNode B. For both of the above configuration alternatives (with or withoutMAC-c/sh/m) a companion Flow Control function to either the Flow Con-trol function in the MAC-c/sh/m or MAC-d exists. By these Flow Control


Disassembly

Reordering

Reordering queue distribution

HARQ

Disassembly

Reordering

MAC-hs

to MAC-d

HS-DSCH

AssociatedUplink Signalling

AssociatedDownlink Signalling

MAC Control

(a) UE side

Priority Queuedistribution

PriorityQueue

PriorityQueue

Priority Queuedistribution

PriorityQueue

PriorityQueue

HARQ entity

TFRC selection

Scheduling /Priority handling

MAC Control

MAC-hs

HS-DSCH

AssociatedUplink Signalling

AssociatedDownlink Signalling

MAC-d Flows

(b) UTRAN side

Figure 2.19: MAC-hs details

functions the layer 2 signalling latency is limited and, therefore, the discard-ing and retransmission of RLC PDUs as a result of HS-DSCH congestionis reduced. In general, the Flow Control tries to keep the buffering in theNode B at an adequate level where sufficient data is available when a UE isto be scheduled. This is a requirement to use the radio resources in a mostefficient way. On the other hand the amount of buffered data is kept assmall as possible so that the Round Trip Time (RTT) is not unnecessarilyincreased and the amount of data which is lost during a buffer flush in caseof handover is minimized. Details about the Flow Control mechanism andprotocol can be found in [37] and [36].

In UTRAN the data received on the MAC-d flows for one UE is put intoup to 8 Priority Queues (PQs) which are used to buffer the MAC-d PDUsfor this user [66]. One UE may be associated with one or more MAC-d flows.Each MAC-d flow contains MAC-d PDUs for one or more PQs. Every singleMAC-d PDU is assigned a specific Scheduling Priority Indicator (SPI) sothat PDUs from different services can be handled with different priorities


by the priority handling function. The PQ distribution function allocatesthe arriving MAC-d PDU to the corresponding PQ whose priority is thesame as the SPI of the PDU. Hence, at any time all the PDUs buffered inone PQ have the same SPI.

The UTRAN MAC-hs entity has one scheduling and priority handlingfunction which takes care of the scheduling of different UEs and the dataflows within a single UE. The scheduling and priority handling functionmanages HS-DSCH resources between HARQ entities of different UEs anddata flows according to their priority. In order to do so it selects the HARQentities and corresponding PQs which are to be serviced. Assisted by sta-tus reports from the associated uplink signalling either new transmissionsor retransmissions can be triggered by the scheduling function. For bothtransmitted and retransmitted MAC-hs PDUs the scheduler determines asuitable Redundancy Version (RV) and indicates it to the PHY. Detailedinformation about these RVs and their role for supporting different HARQschemes will be presented in Section 2.3.1.3 and Section 2.3.1.5. New trans-missions may be initiated in favor of pending retransmissions to support thepriority handling. Furthermore, the scheduler triggers new transmissions incase a maximum number of retransmissions has been reached or the MAC-hs PDUs are out-of-date and to be discarded. In this case it is up to theRLC layer to retransmit missing data.

For every MAC-hs PDU to be transmitted the scheduler indicates thequeue identifier and Transmission Sequence Number (TSN) to the HARQentity. Both the queue identifier and the TSN are part of the MAC-hs PDUheader. The 3 bit queue identifier allows the reordering queue distributionfunction in the UE to map received PDUs to the correct reordering queues.Each reordering queue in the UE corresponds to a PQ in the MAC-hs en-titiy in UTRAN. The reordering queues store received MAC-hs PDUs andreorder them with a sliding window scheme according to the received TSNs.The TSN is a queue specific counter ranging from 0 to 63 which is increasedfor every transmitted MAC-hs PDU. If no PDUs with a lower TSN aremissing the MAC-hs PDUs with consecutive TSNs are delivered in the cor-rect order to the disassembly function. The disassembly function extractsthe MAC-d PDUs from the MAC-hs PDU and forwards them to the UEside MAC-d entity which in turn sends the contained SDUs to the higherlayers. In case a MAC-hs PDU is not yet correctly received all MAC-hsPDUs with a higher TSN are kept in the reordering buffer and a timer forthe missing PDU is started. This Reordering Release Timer (T1) is a timer


used to control the stall avoidance in the UE reordering buffer. On expirythe UE assumes the MAC-hs PDU to be finally lost and forwards PDUswith higher TSNs to the disassembly function anyhow. In the UTRAN theT1 timer with the same value exists for every transmitted but unacknowl-edged MAC-hs PDU. Upon expiry the Node B does not try to retransmitthe not yet acknowledged PDU. The value of T1 is configured by the RRClayer using the control SAP of the MAC-hs entity.

In the UTRAN MAC-hs entity there is one HARQ entity per UE. EachHARQ entity consists of up to 8 HARQ processes which implement par-allel stop-and-wait HARQ protocols. These parallel HARQ processes areneeded because of the round-trip time from the transmission of a PDUuntil its acknowledgment. As it takes 5 TTIs for the Node B to get theassociated Acknowledgement (ACK)/Negative Acknowledgement (NACK)response for a single HARQ process, at least 6 processes are required to con-tinuously transmit data to a single UE. For UEs having a higher minimuminter-TTI interval for HSDPA reception, i.e. every 2nd or 3rd TTI, a smallernumber of HARQ processes is sufficient. When the scheduler forwards thescheduling result for an initial transmission to the HARQ entity of the sched-uled UE, the HARQ entity is constructing the MAC-hs PDU containing thequeue identifier and the TSN. In addition a suitable HARQ process is se-lected to transmit the MAC-hs PDU. The HARQ process number is notpart of the MAC-hs PDU. Instead the associated downlink signalling car-ries this information. Additionally, the receiver is informed by toggling theNew Data Indicator (NDI) bit of the associated downlink signalling that aMAC-hs PDU belonging to an initial transmission is to be expected. Whena retransmission is triggered by the scheduler, the HARQ entitity does notneed to create a new MAC-hs PDU. The HARQ process just repeats thetransmission with an unchanged NDI bit and, potentially, a changed RVparameter. Status messages, e.g. ACK/NACK signalling, are received bythe HARQ process and delivered to the scheduler.

The Transport Format and Resource Combination (TFRC) selection inUTRAN is in charge of choosing an appropriate TF and the resources usedfor the transmission of the MAC-hs PDU on the HS-DSCH. The selectedresources are the used channelization codes and the modulation scheme.Based on the TF a number of MAC-d PDUs is put into the MAC-hs PDUand the generated TB is sent to the PHY. The TB size information togetherwith the selected resources are signalled to PHY which is in charge of en-coding and sending this associated downlink signalling using the HS-SCCH.

2.3. Physical Layer 43

In the UE the HARQ entity of the MAC-hs is in charge of handling theUE side functions related to the HARQ protocol. Based on the downlink sig-nalling the HARQ entity selects the used HARQ process for the succeedingHS-DSCH transmission. Depending on the decoding status of the receivedtransmission the HARQ process sends either a positive acknowledgment(ACK) or negative acknowledgment (NACK) back to the Node B via theassociated uplink signalling. If a MAC-hs transmission is correctly receivedthe MAC-hs PDU is forwarded to the correct reordering queue based on thequeue identifier within the PDU. In case a MAC-hs transmission can not bedecoded correctly, the content of the transmission is kept in the UE’s softbuffer until a further retransmission adds enough redundancy to achieve apositive decoding result. In case the NDI bit is toggled the UE flushes thesoft buffer and starts with the reception of an initial transmission. Whenthe soft buffer contained information from previous unsuccessful transmis-sion attempts this information is lost and the partially received MAC-hsPDU is discarded. As in UTRAN the HARQ entity contains up to 8 HARQprocesses. Per TTI only one HARQ process may receive data from theNode B.

2.3 Physical Layer

In this section the PHY of the UMTS radio interface is described withrespect to these aspects which are relevant for the performance evaluationpresented in Chapter 4. The general description of the PHY and an overviewof the 3GPP document structure specifying the PHY can be found in [16].The services offered by the PHY to the higher layers are summarized in [26].

One of the most important services provided to the higher layers are thetransport channels which have already been introduced in Section 2.2.5.2.The PHY maps these transport channels to physical channels. The physicalchannels differ depending on the duplexing mode. UMTS supports twoduplexing schemes. In the FDD mode the uplink and downlink have theirown dedicated frequency bands. Hence, FDD can only be operated withpaired frequency bands. In TDD the uplink and downlink are transmittedwithin the same frequency band and, therefore, unpaired frequency bandscan be used. Duplexing is achieved by using the time domain instead of the


frequency domain. Based on Radio Resource Management (RRM) decisionsor static configuration certain time slots are used for the uplink and theothers for the downlink. By allocating an uneven number of time slots foruplink and downlink the TDD mode is suited well for asymmetric services.However, this flexibility is limited as the switching point between uplink anddownlink is typically the same for neighbouring cells because of interferenceaspects.

The FDD mode is always used with two paired 5 MHz carriers. Thechip rate (see Section 2.3.2.5) is fixed to 3.84 Mcps1. In TDD three sub-modes are available which are characterized by their carrier bandwidth andchip rate. Similar to the FDD mode a 3.84 Mcps configuration deployedon a single 5 MHz carrier exists. Furthermore, a 7.68 Mcps mode using a10 MHz bandwidth and a 1.28 Mcps mode on a 1.6 MHz carrier are available.The 5 MHz modes are referred to as WCDMA while the 1.6 MHz mode isfrequently called Narrowband Code Division Multiple Access (NCDMA).Depending on operator and country the carrier spacing might be smallerthan the nominal bandwidth. 5 MHz carriers may, for example, be placedat distances ranging from 4.4 to 5 MHz.

The complete mapping of transport channels to physical channels forboth FDD [17] and TDD [21] is shown in Table 2.1. The physical channelswhich carry the data parts of the transport channels are written on thesame line in the table. Associated signalling channels are listed below thesephysical channels. The FACH and the PCH are the only transport channelsof different type which are both mapped on the same physical channel.Physical channels which are not directly associated to one of the transportchannels are listed at the bottom part of the table.

2.3.1 Transport Channel Coding and Multiplexing

The transport channels, used for data transmission by the MAC layer, tra-verse several processing steps in the PHY. Figure 2.20 illustrates these stepsfor the majority of transport channels. In detail these are the DCH, RACH,FACH, BCH and PCH as well as the TDD only DSCH and USCH. TheFDD downlink processing is shown on the left hand side of the figure, whilethe FDD downlink as well as both TDD transmission directions are depictedon the right hand side.

13.84 · 106 chips per second


Table 2.1: Mapping of transport channels to physical channels

Transport Physical ChannelChannel FDD TDD

BCH P-CCPCH P-CCPCHFACH S-CCPCH S-CCPCHPCH

RACH PRACH PRACHDSCH PDSCHUSCH PUSCHDCH DPDCH DPCH

DPCCHF-DPCH

HS-DSCH HS-PDSCH HS-PDSCHHS-SCCH HS-SCCH

HS-DPCCH HS-SICHE-DCH E-DPDCH E-PUCH

E-DPCCH E-UCCHE-AGCH E-AGCHE-RGCHE-HICH E-HICH

E-RUCCH

CPICHSCH SCH

AICHPICH PICHMICH MICH

PNBSCH (3.84 Mcps)PLCCH (1.28 Mcps)DwPCH (1.28 Mcps)UpPCH (1.28 Mcps)FPACH (1.28 Mcps)


Radio frame segmentation

Channel coding

TB concatenation /

Code block segmentation

CRC attachment

First interleaving

Rate matching

Physical channel segmentation

TrCH multiplexing

Second insertion of

DTX indication

CCTrCH

PhCH 1 PhCH Y

Second interleaving

Physical channel mapping

TrCH 1 TrCH X


Channel coding

TB concatenation /


CRC attachment

First interleaving

Rate matching

First insertion of

DTX indication

First insertion of

DTX indication

(a) FDD downlink


Channel coding

TB concatenation /


CRC attachment

Radio frame equalization

First interleaving

Rate matching

Physical channel segmentation

TrCH multiplexing

Bit scrambling (TDD)

CCTrCH

PhCH 1 PhCH Y

Second interleaving


TrCH 1 TrCH X


Channel coding

TB concatenation /


CRC attachment

Radio frame equalization

First interleaving

Rate matching

(b) FDD uplink and TDD

Figure 2.20: Transport channel multiplexing and coding

The detailed processing steps per transport channel, as specified in [18]for FDD and [22] for TDD, are described in the following sections. Er-ror detection by using a CRC is described in Section 2.3.1.1. All TBs andtheir corresponding CRC values are concatenated. Section 2.3.1.2 sum-marizes the available error correction schemes for the resulting concate-nated sequence. The channel coded sequence with its redundancy informa-tion is adapted to the number of available transmission bits by the RateMatching (RM) algorithm as outlined in Section 2.3.1.3. The downlink Dis-continuous Transmission (DTX) mechanism is treated by Section 2.3.1.3as well. As multiple transport channels can be multiplexed on the sameCCTrCH the RM algorithm jointly processes this set of channels.


For transport channels with a TTI spanning multiple radio frames, firstinterleaving across radio frames as described in Section 2.3.1.4 takes place.Radio frame segmentation is in charge of distributing the interleaved bitsto the individual radio frames belonging to the same TTI. Padding bits areinserted by the radio frame equalization in case the bit sequence can not beexactly split into equally sized parts.

Transport channel multiplexing is performed by concatenating the bitsequences of each transport channel on a radio frame basis. In the TDDmode bit scrambling is performed. In the FDD downlink additional DTXindicators are appended to the concatenated sequence to fill the remainingavailable bits of the physical channels. After the above processing steps theresulting bit sequence, also referred to as CCTrCH, is split into segmentsaccording to the number of physical channels to be used for the trans-mission. On each of these segments a second interleaving as described inSection 2.3.1.4 is made. Finally, the interleaved segments are given to thephysical channels for further processing.

Slightly different to the processing of the channels described above is thehandling of the HS-DSCH and E-DCH. Figure 2.21 illustrates the processingchains of these channels for both FDD and TDD mode.

The upper parts of the chains are similar to the general handling asillustrated before. A CRC of a fixed size of 24 bits (Section 2.3.1.1) andturbo coding as presented in Section 2.3.1.2 are used. The HARQ basedHS-DSCH and E-DCH allow Incremental Redundancy (IR) to be applied.IR is achieved by varying parameters of the RM algorithm, depicted in Sec-tion 2.3.1.3, and rearranging the bit to symbol mapping (constellation rear-rangement) as described in Section 2.3.1.5. The interleaving on radio framelevel is basically the same as the Second Interleaving for the non-HARQbased channels. For higher order modulation schemes parallel interleaversare used as illustrated in Section 2.3.1.4. A further difference to the normalprocessing of transport channels exists for the TDD case. For the TDD E-DCH and HS-DSCH the interleaving is done for all bits of one radio frame,while in case of FDD the interleaving is done per physical channel (dot-ted lines in Figure 2.21). Therefore, physical channel segmentation is onlyneeded for FDD.


HS-PDSCH 1 HS-PDSCH Y

CRC attachment

Bit scrambling (FDD)


Channel coding

PHY HARQ functionality /

Rate matching

Physical channel

segmentation (FDD)

HS-DSCH interleaving

Constellation rearrangement

for 16QAM and 64QAM



(a) HS-DSCH


Channel coding

Physical channel

segmentation (FDD)

PHY HARQ functionality /

Rate matching

E-DPDCH 1 (FDD)

E-PUCH 1 (TDD)



Constellation rearrangement

for 16QAM (TDD)

E-DCH interleaving

CRC attachment

E-DPDCH Y (FDD)

E-PUCH Y (TDD)

(b) E-DCH

Figure 2.21: Transport channel coding chain for HS-DSCH and E-DCH

2.3.1.1 Error Detection

Transmission errors of TBs are detected in the receiver by a CRC check.To do so every transmitted TB is appended by CRC parity bits which arecalculated by the bits of the TB and one of the cyclic generator polynomialsfrom Eqs. (2.1) to (2.4).

gCRC24(D) = D24 +D23 +D6 +D5 +D + 1 (2.1)

gCRC16(D) = D16 +D12 +D5 + 1 (2.2)

gCRC12(D) = D12 +D11 +D3 +D2 +D + 1 (2.3)

gCRC8(D) = D8 +D7 +D4 +D3 +D + 1 (2.4)


Depending on which polynomial is used the CRC parity has a lengthof 8, 12, 16 or 24 bits. The CRC length is either specific to the type oftransport channel or it can be configured by the RRC using the semi-staticpart of the TBS.

Some transport channels, e.g. HS-SCCH and E-DCH Absolute GrantChannel (E-AGCH), use a CRC masking technique to address UEs onshared transport channels. By masking the CRC value with the RadioNetwork Temporary Identifier (RNTI) of the addressed UE, i.e. performingan Exclusive Or (XOR) calculation, at the Node B and the configured RNTIat the UE, only the addressed UE is able to successfully perform the CRCcalculation.

2.3.1.2 Forward Error Correction

Because of the relatively high bit error rate of the wireless channel FECis required to assure a reliable transmission. In UMTS a ConvolutionalCode (CC) of rate 1/2 and 1/3 as well as a Turbo Code (TC) of rate 1/3 areavailable for FEC.

The generator polynomials for the convolutional code of rate 1/2 can befound in Eqs. (2.5) and (2.6). The polynomials for the 1/3 rate code arelisted in Eqs. (2.7), (2.8) and (2.9). Figure 2.22 illustrates the structureof both convolutional coders. The coders of both rates have a constraintlength of 9.

GCC 12 ,1

(x) = x8 +x6 +x5 +x4 +1 (2.5)

GCC 12 ,2

(x) = x8 +x7 +x6 +x5 +x3 +x +1 (2.6)

GCC 13 ,1

(x) = x8 +x6 +x5 +x3 +x2 +x +1 (2.7)

GCC 13 ,2

(x) = x8 +x7 +x5 +x4 +x +1 (2.8)

GCC 13 ,3

(x) = x8 +x7 +x6 +x3 +1 (2.9)

The turbo coder in UMTS uses a Parallel Concatenated ConvolutionalCode (PCCC). The repetition of the input bits, i.e. the systematic bits, and


Output 0G0= 557 (octal)

Input

D D D D D D D D




Input

D D D D D D D D


(a) Rate 1/2 convolutional coder

(b) Rate 1/3 convolutional coder

Figure 2.22: Convolutional coding with rate 1/2 and rate 1/3

the output of two parallel encoders with the transfer function in Eq. (2.10)lead to a code rate of 1/3. The input bits of the second encoder are shuffledby an interleaver. The block diagram of the complete turbo coder is shownin Figure 2.23. Further details can be found in [18].

GTC(x) =x3 + x+ 1

x3 + x2 + 1(2.10)

The input bits for the coding are the concatenated TBs including theirCRC values. If the number of bits in this bit sequence exceeds certainthresholds, i.e. 504 bits for convolutional coding and 5114 bits for turbocoding, code block segmentation takes place. In case of code block segmen-tation the input sequence is split into equally sized blocks which are encodedindependently.

The selected coding algorithm (turbo coding, convolutional coding orno coding) and the coding rate are part of the semi-static part of the TFSwhich is configured by RRC. After all input bits have been shifted into theencoder the trellis termination takes place. Trellis termination is neededto get the same redundancy information also for the last bits fed to theencoder. For the convolutional coder 8 zero bits are added to the inputsequence. These padding bits lead to a deterministic final state of the shift


xk

xk

zk

Turbo codeinternal interleaver

x'k

z'k

D

DDD

DD

Input

OutputInput

Output

x'k

1st constituent encoder

2nd constituent encoder

Figure 2.23: Turbo coding with rate 1/3

registers. All registers contain binary zeros after termination. For the turbocoder the same is achieved by internally disabling the shift register feedbackand shifting zeros into the 3 registers of both constituent encoders. Trellistermination leads to 16 or 24 additional output bits for the convolutionalcoder variants and 12 additional bits for the turbo coder for every codeblock. Finally, the encoded code blocks are concatenated again.

2.3.1.3 Rate Matching

RM is a mechanism to map the encoded bits of the turbo or convolutionalcoder to the bits available for transmission on the physical channels. Ifless bits are available on the physical channels, puncturing of the encodedbit sequence takes place. The RM of UMTS allows to remove an arbitrarynumber of bits from the encoded sequence by a predefined algorithm. Bythis puncturing mechanism variable coding rates between 1 (no coding) and1/2 or 1/3 (the mother code rates of the convolutional and turbo coder) canbe achieved.

The opposite case, i.e. the physical channels have more bits availablethan required by the encoded bit sequence, can be handled by the RM


algorithm in a similar way. Repetition of input bits allows to match to thenumber of available transmission bits for this case. The effective codingrate of the transmission is reduced compared to the mother code rate ifrepetition of bits is applied. The number of input bits to be repeated canbe controlled with a bit granularity as well.

In general the repetition of a large amount of bits is avoided with respectto the spectrum efficiency of the system. In UMTS several mechanismsare used to limit repetition of coded bits to a sufficient amount or avoidit completely. In the uplink the Spreading Factor (SF) to be used for atransmission is calculated before the RM takes place. For the lowest SF 4the number of required codes may be chosen as well. The required SFand the amount of channelization codes to achieve a certain Block ErrorRate (BLER) are derived by the Puncturing Limit (PL) and RM parameteras configured by higher layers. By selecting a high SF and a small number ofchannelization codes the transmission power and, therefore, UL interferenceis reduced.

In the downlink the usage of a variable SF is not possible as the allocationof channelization codes is fixed and may only be changed by RRC signalling.Instead of the UL mechanism the PHY in the Node B uses DTX to avoidtransmitting unnecessary information bits. If only a small amount of datais to be transmitted, DTX indicators are inserted into the bit sequence afterthe RM. The DTX indicators disable the transmitter for the correspondingbit positions and, therefore, reduce DL interference.

For the case that multiple transport channels are mapped to the samephysical channels, i.e. they are multiplexed on a so-called CCTrCH, theRM is in charge of distributing the available bits to the individual transportchannels. Every transport channel is configured with a rate-matching at-tribute which is included in the semi-static part of its TFS. Based on theseattributes the RM algorithm may prioritize transport channels with respectto the bits allocated to them. QoS differentiation in terms of service specificresidual Bit Error Rate (BER) and BLER targets can be achieved by thismethod. One example where this method is applied is the speech servicesof UMTS where important and less important bits of the AMR codec haveunequal residual BERs.

Different RM schemes are selected for convolutionally coded and turbocoded channels. As illustrated in Figure 2.24(a), transport channels whichmake use of convolutional coding use the basic RM algorithm on the com-plete bit sequence as fed to the RM block. Both puncturing and repetition


Radio frame segmentation (uplink)

Channel coding (downlink)

Bit separation

Rate matching

Bit collection

TrCH multiplexing (uplink)

1st insertion of DTX indication (downlink)

Rate

matching

algorithm

(a) Convolutionally coded channels andturbo coded channels with repetition

Radio frame segmentation (uplink)

Channel coding (downlink)

Bit separation

Rate matching

Bit collection

TrCH multiplexing (uplink)

1st insertion of DTX indication (downlink)

Rate

matching

algorithm

Rate

matching

algorithm

Systematic

bits

Parity 1

bits

Parity 2

bits

(b) Puncturing of turbo coded channels

Bit separation

Rate matching

Bit collection

RM_P1_2 RM_P2_2RM_S

RM_P1_1 RM_P2_1First rate

matching

Second rate

matching

Virtual IR

buffer

Systematic

bits

Parity 1

bits

Parity 2

bits

(c) HS-DSCH HARQ

Bit separation

Rate matching

Bit collection

RM_P1_2 RM_P2_2RM_S

Systematic

bits

Parity 1

bits

Parity 2

bits

(d) E-DCH HARQ

Figure 2.24: Rate Matching

of bits can be done with this algorithm. Turbo coded channels apply thesame algorithm when repetition of bits takes place. For puncturing theturbo coded bit stream is split into three sequences (bit separation) whichis depicted in Figure 2.24(b). The first one contains the systematic bitswhile the second and third sequences contain the non-systematic bits from


the upper and lower constituent encoder, respectively. The systematic bitsare not punctured for most of the transport channels (see below). The non-systematic bits are equally punctured by independently applying the RMalgorithm to both non-systematic bit sequences. After the RM is performedthe three sequences are combined again by the bit collection mechanism.

An exception to the above rule exists for the HARQ based channelsHS-DSCH and E-DCH. Figure 2.24(c) and Figure 2.24(d) show the RMblocks for these channels, respectively. Both channels are always turbocoded and independent RM takes place for the systematic bits as well asfor the two non-systematic bit sequences. Even in case of repetition theRM is performed for each sequence separately. The RV parameters s and rcontrol how the RM is to be done for every transmission and retransmission.The parameter s indicates if either systematic bits or non-systematic bitsare prioritized during RM. If non-systematic bits are prioritized, differentpuncturing patterns can be applied by varying the r value.

In order to support UE categories with limited HARQ soft buffer ca-pacities the HS-DSCH rate matching is split into two stages. The first onereduces the amount of bits to that amount which could be stored in the UE’smemory. If sufficient memory is available in the UE the first rate matchingstage is transparent. In the second stage the bit sequence is finally matchedto the available physical channel bits. By varying the RV parameters dif-ferent redundancy schemes can be applied in this stage. However, reducedsoft buffer capacities in the UE also reduce this flexibility as the ratherstatic first rate matching gets dominant. As will be shown in Section 4.7the gain of Incremental Redundancy (IR) over Chase Combining (CC) willbe reduced in this case.

2.3.1.4 Interleaving

Interleaving of bits is performed in both the transmitter and the receiver toreduce the impact of burst errors which typically occur during radio trans-missions. By reshuffling the bit sequence, burst errors on the radio interfacebecome distributed single bit errors at the decoder which significantly im-proves the reliability of the FEC. In UMTS two levels of bit interleavingexist. The first interleaving stage deals with the distribution of bits overmultiple radio frames. In case the TTI of the transport channel is 2 msor 10 ms the first interleaving is transparent. For larger TTIs a block in-


Interleaver

(R2 x 30)

Interleaver

(R2 x 30)

up,k (QPSK)

up,k up,k+1(16QAM)

up,k+2 up,k+3(16QAM)

p,v

k(QPSK)

vp,k vp,k+1(16QAM)

vp,k+2 vp,k+3(16QAM)

up,k up,k+1(64QAM) vp,k vp,k+1(64QAM)

up,k+2 up,k+3(64QAM) vp,k+2 vp,k+3(64QAM)

Interleaver

(R2 x 30) vp,k+4 vp,k+5(64QAM)up,k+4 up,k+5(64QAM)

up,k(BPSK)

up,k(4PAM)

up,k+1(4PAM)

vp,k

(BPSK)

vp,k(4PAM)

vp,k+1(4PAM)

Figure 2.25: Block interleaver

terleaver with inter-column permutation is used. The number of columnsof the interleaver is given by the number of radio frames per TTI, i.e 2columns for a 20 ms TTI, 4 columns for a 40 ms TTI and 8 columns for an80 ms TTI. This selection of columns assures that consecutive input bitsare always mapped to different radio frames. The number of rows of theinterleaver depends on the length of the bit sequence.

The second interleaving stage is responsible for shuffling the bits withina radio frame. This interleaving is always done after the physical channelsegmentation mapped the bits to the physical channels. Therefore, the in-terleaving takes place for every physical channel independently. Similar tothe first interleaving, a block interleaver with inter-column permutation isused as well. The number of columns of the interleaver has been chosen tobe 30 so that consecutive bits are always mapped onto different modula-tion symbols and different slots of the radio frame (15 slots and 2 bits perQuadrature Phase-Shift Keying (QPSK) symbol). Based on the availablebits per physical channel (given by the SF, the modulation and the burststructure) the number of rows of the interleaver is derived. In case of theFDD HS-DSCH the characteristics of the physical channels are fixed andthe number of rows is always 32.

If multiple information bits are modulated on the same modulationbranch, which is the case for 16QAM, 64QAM and the single branch 4PAM,


multiple parallel interleaver blocks are used to keep those information bitstogether. Figure 2.25 illustrates how the input bit sequence is mapped ontothe parallel interleavers. Not interleaving the bits of one modulation symbolassures that bits mapped onto preferred positions in the modulation con-stellation are maintained by the interleaver. Typically systematic bits ofthe turbo coder are mapped onto bit positions which have a higher demod-ulation probability. For retransmissions the bit to symbol mapping may bechanged by the constellation rearrangement.

2.3.1.5 Constellation Rearrangement

For the higher order modulation schemes 16-State Quadrature AmplitudeModulation (16QAM) and the 64-State Quadrature Amplitude Modulation(64QAM) the HARQ based HS-DSCH and the TDD E-DCH have a mech-anism to deal with the imperfectly Gray-coded symbol constellation. Theconstellation rearrangement can be used to modify the bit to symbol map-ping for HARQ retransmissions. Bits which are initially transmitted witha weaker symbol constellation may be retransmitted on a preferred bit po-sition. The available rearrangements for 16QAM and 64QAM are listed inTable 2.2.

Table 2.2: Constellation rearrangement for 16QAM and 64QAM

Constellation Output bit sequence forversion b 16QAM 64QAM

0 vk vk+1 vk+2 vk+3 vk vk+1 vk+2 vk+3 vk+4 vk+5

1 vk+2 vk+3 vk vk+1 vk+4 vk+5 vk+2 vk+3 vk vk+1

2 vk vk+1 vk+2 vk+3 vk+2 vk+3 vk+4 vk+5 vk vk+1

3 vk+2 vk+3 vk vk+1 vk vk+1 vk+2 vk+3 vk+4 vk+5

2.3.2 Physical Channels

The physical channels receive the input bit stream from the physical chan-nel mapping function of the transport channels. In general the task of thephysical channels is to prepare the input bit sequence for the radio transmis-sion. Main aspects which are considered are the multiple access scheme andduplexing mode in terms of frequency, code and time domain. Furthermore,


the modulation and the control of transmission power are considered. Theprocessing steps for the physical channels are specified in [19] and [23].

Each physical channel transmits its information in so-called slots. A slotis the resource allocation unit in the time domain of TDD. Furthermore,the granularity of the inner-loop power control is defined by these slots.Section 2.3.2.3 gives an overview of the slot- and overlying frame structures.

2.3.2.1 Downlink Physical Channel Processing

Figure 2.26 illustrates the processing steps in the FDD downlink. All phys-ical channels except the Synchronization Channel (SCH) are handled in thesame way. First the input bits are mapped to symbols by the modulationmapping. Groups of consecutive bits are mapped to one modulation sym-bol as described in Section 2.3.2.4. Both the I-branch and the Q-branchof the modulation symbols are spread by a channel specific channelizationcode CSF,n. The two resulting chip sequences are combined into a complexvalued sequence which is scrambled by a complex valued cell specific scram-bling code Sdl,i. Details about spreading as well as scrambling are given inSection 2.3.2.5. Before summing the chip sequences of all physical chan-nels each channel is multiplied by a channel specific gain factor Gi whichis set according to SF, channel specific quality and transmission power re-

j

Q

I

I+jQ

Sdl,1

CSF,n

ModulationMapper

(QPSK,16QAM or64QAM)

S/P

PhysicalChannel

G1

G2

Σ

Σ

GP

GS

P-SCH

S-SCH

toModulation

Figure 2.26: Physical channel processing in FDD downlink


quirements. The transmission power is, for example, controlled by the fastinner-loop Transmit Power Control (TPC) as specified in [20] and [24].

In contrast to all other physical channels the SCH, consisting of a pri-mary and a secondary sub-channel, is not scrambled. In order to find acell the primary SCH transmits a constant 256 chip sequence within everyslot. The secondary SCH transmits 256 chip long slot specific sequences.Based on the permutation of these sequences the UE is able to detect theslot number as well as the scrambling code group of a cell. By trying outthe 8 primary scrambling codes within a group, the UE deduces the primaryscrambling code of that cell. Both the primary and the secondary SCH aremultiplied by the gain factors GP and GS, respectively. Finally, all chipsequences are combined and modulated as depicted in Section 2.3.2.6.

2.3.2.2 Uplink Physical Channel Processing

In the FDD uplink the modulation is exactly the same as in the down-link. The processing before the modulation, however, differs as illustratedby the example in Figure 2.27. Each physical channel is either mapped tothe I-branch or Q-branch. Therefore, the modulation mapping only usesone-dimensional mappings (compare Section 2.3.2.4). The real valued sym-bol sequences originating from the modulation mapping of each physicalchannel are spread to a real valued chip sequence by the channelizationcodes. Because of the orthogonality of the I-branch and the Q-branch, twoindependent code trees are used by the spreading stage (see Section 2.3.2.5).

The spreaded sequences are independently weighted by gain factors. Ev-ery gain factor is set according to channel specific offsets based on the SFof the channelization code and the chosen TFC. Furthermore, inner loopPower Control (PC) and transmission specific quality requirements are ap-plied by the gain factors. After the gain control stage the signals from thechannels of the I-branch and Q-branch are summed and a complex additionis performed. Finally, the complex valued chip sequence is scrambled witha UE specific scrambling code (Section 2.3.2.5) and sent to the modulationstage as depicted in Section 2.3.2.6.

The mapping of physical channels to the I-branch and Q-branch is ac-cording to the following principle. The Dedicated Physical Control Channel(DPCCH) and the E-DCH Dedicated Physical Control Channel (E-DPCCH)are always transmitted on the Q-branch and the I-branch, respectively. Up


to 6 Dedicated Physical Data Channels (DPDCHs), which are alternatinglymapped to the I-branch and the Q-branch, can be used in parallel by oneUE. DPDCHs with an odd number are mapped to the I-branch and thosewith an even number are mapped to the Q-branch. The mapping of theHS-DPCCH depends on the number of configured DPDCHs. For 0, 1, 3or 5 DPDCHs the HS-DPCCH is mapped to the Q-branch. Otherwise itis mapped to the I-branch. Similar to the DPDCHs the E-DCH DedicatedPhysical Data Channels (E-DPDCHs) are mapped to the branches in analternating manner. Depending on if a DPDCH or a HS-DPCCH is usedthe first E-DPDCH is mapped on the I-branch or the Q-branch.

The Physical Random Access Channel (PRACH) processing in the UE issimilar to the one of the dedicated channels illustrated in Figure 2.27. ThePRACH message data part is mapped on the I-branch and the control parton the Q-branch. Both parts are spread by individual channelization codesand weighted by independent gain factors. Finally, the combined complexchip sequence is scrambled and modulated.

Sdpch,n

ModulationMapper

(BPSK)

S/P

toModulation

Σ

Σ

j

Q

I

I+jQ

βd

βed1

βed2

βhs

βec

βc

Cd1

Ced1

Ced2

Chs

Cec

Cc

DPDCH1

DPCCH

HS-DPCCH

E-DPDCH1

E-DPDCH2

E-DPCCH

ModulationMapper

(BPSK or 4PAM)

ModulationMapper

(BPSK or 4PAM)

ModulationMapper

(BPSK)

ModulationMapper

(BPSK)

ModulationMapper

(BPSK)

S/P

Figure 2.27: Processing of dedicated physical uplink channels in FDD


2.3.2.3 Frame Structure

As described in Section 2.3.1 the transport channels in UMTS transfer theirinformation on a TTI basis. For most transport channels a TTI spans oneor more 10 ms radio frames. In detail, a TTI may last 1, 2, 4 or 8 radioframes. A radio frame is subdivided into 5 subframes. The HSDPA relatedtransport channels and the High Speed Uplink Packet Access (HSUPA)when operated in the optional 2 ms configuration are an exception as theyare able to transmit on a subframe basis. In other words a TTI of 2 ms isavailable for these transport channels.

Radio frames and subframes are further divided into slots. 3 slots forma subframe and 15 slots compose a radio frame. The transmission of thephysical channels is based on these slots. Channels which apply the inner-loop power control may change their transmission power 1500 times persecond on a slot basis. Depending on the physical channel different slotformats are defined. Figure 2.28 illustrates the described frame structureand the physical channel dependent slot structures. At the top of the figurephysical channels related to the DCH transmission are depicted. Below thesethe E-DCH related traffic and associated signalling channels are shown.Finally, the physical channels in charge of the HS-DSCH transmission aredescribed.

Depending on the SF of a physical channel the number of bits per slotvaries. The control channels depicted in the figure use fixed SFs. For phys-ical channels which use a variable SF, i.e. the downlink Dedicated PhysicalChannel (DPCH) as well as the uplink DPDCH and E-DPDCH, the respec-tive formulas give the available number of bits per slot. The value range ofvariable k corresponds to the range of the SF (2k) allowed for these physi-cal channels. In case of the E-DPDCH and High Speed Physical DownlinkShared Channel (HS-PDSCH) the modulation scheme may be altered aswell. Here, the variable M denotes the number of bits per modulationsymbol.

The most complex slot formats exist for the physical channels associatedto the DCH (compare Table 2.1). Depending on the transmission directionthe mapping of control information and data to the physical channels differs.In the downlink both kinds of information are multiplexed in time. EachDPCH slot carries DPDCH fields, containing DCH data, and DPCCH fieldswhich transmit associated control information. The TPC field carries theinner-loop power control commands which indicate to the receiver of this


1 subframe (2 ms)

HARQ-ACK

10 bits

CQI

20 bits

Tslot = 2560 chips

1 radio frame (10 ms)

Subframe #0 Subframe #1 Subframe #2 Subframe #3 Subframe #4

HS-DPCCH

Data

Ndata bitsHS-PDSCH

Data

Ndata bits

Data

Ndata bits

Slot #0 Slot #1 Slot #2

HS-SCCHData

40 bits

Data

40 bits

Data

40 bits

Slot

#0

Slot

#14

Slot

#1

Slot

#2

Slot

#3

Slot

#4

Slot

#5

Slot

#6

Slot

#7

Slot

#8

Slot

#9

Slot

#10

Slot

#11

Slot

#12

Slot

#13

E-AGCH 20 bits

E-DPCCH 10 bits

E-DPDCHData

Ndata bits

bi,39bi,1bi,0

E-RGCH

E-HICHb0 b1 b39

10 bits

N data= 10*2kbits (k=0..6)UL DPDCH

UL DPCCH

Data

Ndata bits

Pilot

Npilot bitsTFCI

NTFCI bits

FBI

NFBI bits

TPC

NTPC bits

10*2kbits (k=0..7)DL DPCH

Pilot

Npilot bitsTFCI

NTFCI bits

TPC

NTPC bits

Data1

Ndata1 bits

Data2

Ndata2 bits

DPDCH DPDCHDPCCH DPCCH

TPC

NTPC bits(Tx OFF)

512 chips

(Tx OFF)F-DPCH

Frame

N data = M*10*2kbits (k=0..7)

N data = M*10*2kbits (k=4)

Figure 2.28: FDD frame and slot structure


information if it shall increase or reduce the transmission power in the nextslot. The Transport Format Combination Indicator (TFCI) field containsbits of the Reed-Muller coded TFCI which allows the receiver to decode theTFCS carried on the DPDCH. Finally, known pilot bits are transmitted forchannel estimation purposes. The exact number of bits within the DPDCHand DPCCH fields depends on the selected SF and the number of used slotswithin one radio frame in case compressed mode is applied. Furthermore,alternative slot formats are defined for the case that blind TF detection isconfigured and the TFCI is not transmitted.

Contrary to the downlink, control information and data are not timemultiplexed in the uplink. Instead orthogonal phases of the modulation areused to distinguish both. Additionally to the already described DPCCHfields the Feedback Information (FBI) field is used in case closed-loop trans-mit diversity is applied in the downlink.

In TDD the bursts transmitted within one slot significantly defer fromthe described FDD slot formats. For details please refer to [21].

2.3.2.4 Modulation Mapping

The modulation mapping is the first stage of the processing chain of a phys-ical channel. The input sequence is the bit sequence as mapped to thisphysical channel by the transport channel processing. In the downlink,depending on the chosen modulation scheme, either 2, 4 or 6 bits of this se-quence are jointly mapped to one modulation symbol. A modulation symbolis represented by two real valued components, the in-phase and quadraturecomponent. Figure 2.29 shows the bit to symbol mapping alternatives forthe FDD downlink.

If the input sequence contains DTX indicators the symbol mapping dif-fers from the figure. For the QPSK modulation the component correspond-ing to the DTX bit is set to 0. For one DTX bit this results in a BinaryPhase-Shift Keying (BPSK) modulation. If both bits of a QPSK symbolcontain DTX indicators the transmitter is completely switched off duringthis symbol. The higher order modulation schemes result in 0 valued com-ponents only in case all bits of a symbol contain DTX indicators. If at leastone bit of a symbol does not contain a DTX indicator, all DTX positionsare replaced by bits which are selected in order to improve the decodingprobability of the non-DTX bits.


Q

I

10 00

0111

1-1

(a) QPSK

Q

I

1011

0000 0010

0001 00111001

1010 1000

0100 01101110 1100

0101 01111111 1101

-0.4472-1.3416 0.4472 1.3416

(b) 16QAM

Q

I

100011

100010

100110

100111

100001

100000

100100

100101

101001

101000

101100

101101

101011

101010

101110

101111

000011

000010

000110

000111

000001

000000

000100

000101

001001

001000

001100

001101

001011

001010

001110

001111

010111

010110

010010

010011

010101

010100

010000

010001

011101

011100

011000

011001

011111

011110

011010

011011

110111

110110

110010

110011

110101

110100

110000

110001

111101

111100

111000

111001

111111

111110

111010

111011

0.6547 1.0911 1.52750.2182-0.6547-1.0911-1.5275 -0.2182

(c) 64QAM

Figure 2.29: Downlink modulation mapping

In the FDD uplink only one-dimensional modulation mappings are usedper physical channel. Figure 2.30 illustrates how 1 or 2 bits are mappedfor the BPSK and 4-State Pulse Amplitude Modulation (4PAM) case, re-spectively. As physical channels in the UE are transmitted on either the in-phase or the quadrature branch (compare Figure 2.27), QPSK-like and, forthe E-DPDCH case, 16QAM-like constellations are possible by multi-codetransmissions. The amplitudes of the in-phase and quadrature branchestypically differ because of different gain factors for the individual physicalchannels.

Q

I1 0

-1 1

(a) BPSK

Q

I0010 0111

-0.4472-1.3416 0.4472 1.3416

(b) 4PAM

Figure 2.30: Uplink modulation mapping

The TDD uplink and downlink is similar to the FDD downlink withrespect to the modulation constellations. The only differences are a modifiedbit-to-symbol mapping and a phase shift of 45◦. Details about the TDDmodulation mapping can be found in [23].


2.3.2.5 Multiple Access

For a mobile communication system the multiple access scheme is an im-portant feature. Multiple access allows to differentiate the connections ofmultiple UEs. In UMTS the main multiple access scheme is a Code DivisionMultiple Access (CDMA) based technique named Direct-Sequence Code Di-vision Multiple Access (DS-CDMA). DS-CDMA uses orthogonal sequencesto spread the transmitted information to the target bandwidth. By usingdifferent orthogonal sequences parallel physical channels can be realized.The sequences, also referred to as spreading codes or channelization codes,are taken from a so-called Orthogonal Variable Spreading Factor (OVSF)code tree as illustrated in Figure 2.31.

SF = 1 SF = 2 SF = 4

C 1,0= (1)

C 2,0= (1,1)

C 2,1= (1,-1)

C 4,0= (1,1,1,1)

C 4,1= (1,1,-1,-1)

C 4,2= (1,-1,1,-1)

C 4,3= (1,-1,-1,1)

SF = 8

Figure 2.31: Channelization code tree

The building principle with which the sequences of the two sub-codes ofa code CSF,n are derived is given by the recursive Eqs. (2.11) and (2.12).The sequence of the root code is defined to be C1,0 = (1).

C2∗SF,2∗n = [+CSF,n + CSF,n]∀n ∈ [0, SF − 1] (2.11)

C2∗SF,2∗n+1 = [+CSF,n − CSF,n]∀n ∈ [0, SF − 1] (2.12)

In one code tree all codes on one level are orthogonal to each other.Furthermore, codes on different levels are orthogonal as long as none of


these codes is part of the sub-tree of another one. As an example code C2,0

from the illustrated code tree is orthogonal to code C4,3 but not to C4,1.Every level of the code tree is characterized by a SF which gives the

length of the spreading sequences as well as the number of available codeson this level. The SF determines how many so-called chips are used totransmit one information symbol which is, depending on the chosen modu-lation, either 1, 2, 4 or 6 bits. As the chip rate is fixed for the FDD and thethree TDD modes the SF of a channelization code is inverse proportionalto the data rate a physical channel with this SF can carry. Table 2.3 sum-marizes the available SFs for the duplexing mode and the resulting symbolrate ranges.

Table 2.3: Symbol rates per physical channel depending on SF

Duplex mode Chip rate SF Symbol rate[Mcps] [ksps]

FDD UL 3.84 256-2 15-1920FDD DL 3.84 512-4 7.5-960

TDD 7.68 32-1 240-7680TDD 3.84 16-1 240-3840TDD 1.28 16-1 80-1280

Note that multi-code transmissions where one CCTrCH is mapped ontoseveral physical channels are available in all modes. Especially for the FDDthe effective symbol rate in case of such transmissions could be higher thandepicted in the table. Furthermore, the TDD symbol rates are theoreticalvalues for the case that all slots could be allocated to one transmissiondirection.

Additionally to the code domain TDD allows to allocate resources for in-dividual UEs on a slots basis in the time domain. Therefore, the TDD modesadditionally contain a Time Division Multiple Access (TDMA) component.Also the FDD HS-DSCH can be seen as a TDMA based channel as UEs arescheduled on a TTI basis. When operating UMTS with multiple carriersa Frequency Division Multiple Access (FDMA) component is added to allmodes as well. Combining all multiple access schemes, a physical channelis characterized by its frequency, code and time slot.


Every UE and every cell has its own code tree from which it uses or-thogonal channelization codes to separate parallel transmissions on multiplephysical channels. FDD UEs are an exception to this as the in-phase and thequadrature components of the modulated signal are spread independently.Therefore, these UEs make use of two code trees whose orthogonality isachieved by the phase shift of the modulated symbols (see Section 2.3.2.6).In an ideal case all physical channels transmitted by one UE or one cell donot interfere with each other because of the code orthogonality principle.Transmissions originating from different UEs or cells interfere as no orthogo-nality in between their code trees exists. This interference is called MultipleAccess Interference (MAI). In order to uncorrelate the MAI every UE andcell scrambles its complex valued spreaded signal with a pseudo-random se-quence. After this scrambling the interference from other UEs or cells canbe interpreted as uncorrelated noise. The scrambling sequences in UMTSare Gold sequences [63] which have a bounded and small cross-correlationwithin certain sets. In the downlink each cell is assigned one and only oneprimary scrambling sequence. Furthermore, it may additionally use a set ofassociated secondary scrambling sequences . In the uplink the UTRAN con-figures UE specific scrambling codes by higher layer signalling. Either longor short scrambling sequences may be configured for the dedicated physicalchannels.

2.3.2.6 Modulation

In UMTS the modulation technique as illustrated in Figure 2.32 is the samefor all duplexing modes.

Complex valued

chip sequenceSplit

real &

imaginary

parts Q

I

Pulse

shaping

Pulse

shaping

cos(ωt)

-sin(ωt)

Figure 2.32: Modulation


The input to the modulation is the complex valued chip sequence asoutput from the blocks illustrated in Figure 2.26 and Figure 2.27. Theinput sequence is split into its real and imaginary parts, i.e. its in-phase andquadrature branches. In order to bound the bandwidth of the modulatedsignal, both sequences are pulse-shaped by a root-raised cosine filter withthe impulse response as given by Eq. (2.13).

h(t) =sin(π tTC

(1− α))

+ 4α TTC

cos(π tTC

(1 + α))

π tTC

(1−

(4α t

TC

)2) (2.13)

The roll-off factor α = 0.22 and the chip duration TC = 1chiprate are fixed

constants for all FDD and TDD modes. By multiplying the pulse-shapedsignals with the carrier waveform the modulated analogue signal is created.Further details about the transmission and reception of the modulated signalcan be found in [14] and [15].

CHAPTER 3

Simulation Environment

Contents3.1 Application Models . . . . . . . . . . . . . . . . . . . 71

3.2 Transmission Control Protocol/Internet Protocol . . . 71

3.3 UMTS Protocol Stack . . . . . . . . . . . . . . . . . . 72

3.4 Radio Interference Simulation Engine . . . . . . . . . 74

3.5 Link-Level Simulation Module . . . . . . . . . . . . . 75

3.6 Real Time Wireless Network Demonstrator . . . . . . 78

3.7 Graphical User Interface . . . . . . . . . . . . . . . . 80

T he simulative performance evaluations within this thesis have been per-formed with the WNS framework which is under development by the

Communication Networks (ComNets) Research Group at RWTH AachenUniversity. A big part of this simulation environment became open source(OpenWNS) and is available for download at [1]. The WNS is a modularevent-driven simulation framework allowing the simulation of complex mo-bile communication systems (e.g. UMTS and WiMAX) on various levels ofdetailedness. The WNS consists of several modules each contributing partsof the protocol stacks and the simulation framework. The complete sourcecode for all simulation models developed and used within this thesis can bedownloaded from [98].

In this chapter those modules which have been developed, extended andused are presented. Besides giving an overview about the simulation envi-ronment which has been established during the work for this thesis, impor-tant simulation models required by the evaluation in Chapter 4 and Chap-ter 5 are introduced as well. Special focus within the following sectionsis put on simulation aspects exceeding the implementation of the UMTSprotocol stack as described in the previous chapter. Figure 3.1 illustratesa typical configuration of a protocol stack used for performance evaluationwithin this thesis.

70 3. Simulation Environment

RLC

MAC

PHY

Layer 2

Layer 1

Channel

RIS

EU

RIS

UDP

PDCP

IP

MPEG

TCP

RRC

AMR

TC

PIP

Ap

pl.

WWW

Layer 4

Layer 3

Layer 5-7

WNS

Figure 3.1: Wireless Network Simulator protocol stack

Traffic models of popular applications and their protocols are providedby the application module introduced in Section 3.1. The wide spread trans-port and network layer protocols of the Internet protocol suite are imple-mented by the Transmission Control Protocol/Internet Protocol (TCP/IP)module presented in Section 3.2. In Section 3.3 details about the implemen-tation of the UMTS protocol stack are given. The module which simulatesthe radio channel is described in Section 3.4.

Further modules, not shown in Figure 3.1, are a module for the provisionand calculation of link-level mappings (see Section 3.5), the Graphical UserInterface (GUI) module as depicted in Section 3.7 and a module allowingthe real time emulation of a (small) UMTS network (see Section 3.6).

3.1. Application Models 71

3.1 Application Models

Located on top of the protocol stack is the module which is in charge ofsimulating the traffic of various packet and circuit switched applications.The traffic models provided by the application module are implemented withthe Specification and Description Language (SDL). In order to use thesemodels in the C++ based simulation framework a code generator [131] aswell as class library called SDL Performance Evaluation Tool Class Library(SPEETCL) [130] are used.

Traffic models simulating Internet traffic as well as traffic typically foundin mobile communication systems, e.g. speech and Multimedia MessagingService (MMS), are available. The Internet traffic is composed by trafficmodels for World Wide Web (WWW) traffic [46], File Transfer Protocol(FTP) uploads and downloads [117, 118] as well as e-mail transmissionand reception using the Simple Mail Transfer Protocol (SMTP) and PostOffice Protocol version 3 (POP3) protocols. Traffic models focusing onmobile communication are represented by Wireless Application Protocol(WAP) and MMS models. Furthermore, speech services can be simulated byconfigurable speaker models generating AMR and VoIP traffic with variousCoder/Decoders (Codecs). Finally, video telephony and video streaming aswell as a simple Constant Bit Rate (CBR) traffic model are provided by theapplication module of WNS.

3.2 Transmission Control Protocol/Internet Protocol

The TCP/IP module of WNS provides an implementation of the most im-portant protocols of the Internet protocol suite [C18]. These protocols arethe UDP [119] and TCP [121] as well as basic IP functionality. Both IPv4[120] and IPv6 [55] may be used as the NL protocol. The TCP entity imple-ments the slow start, congestion avoidance, fast retransmit and fast recov-ery algorithms according to [44] as well as the retransmission mechanism asspecified in [116]. Optionally, selective acknowledgments as introduced by[109] may be enabled. Besides the basic TCP Reno as described by the citedRequest for Comments (RFCs), the TCP/IP module allows to make use ofmore advanced congestion control algorithms from the Linux kernel. Theseare, for example, TCP CUBIC [65], Scalable TCP [90] and Compound TCPas well as the TCP Veno [61] and Westwood+ [108] algorithms, which claimto improve performance especially in the presence of wireless links.


3.3 UMTS Protocol Stack

The module implementing the UMTS protocol stack originated from astand-alone simulator called UMTS Radio Interface Simulator (URIS) [C13].URIS includes a mostly bit accurate implementation of those UMTS pro-tocols which are relevant for the performance of the UMTS radio interface.Encryption, for example, is therefore not included. As the focus is on theRAN part of UMTS the nodes which are simulated are the UE, the Node Band the RNC. Figure 3.2 illustrates a setup with these nodes and how theprotocols provided by the various modules, introduced in Figure 3.1, aremapped onto these nodes.

Layer 5-7

"Layer 0"

Layer 1

Layer 1-3

Layer 3-4

UE

RISE

Client Application Models

UMTS UE Layer 1-3 (URIS)

Modulation/Coding

(Link Level Mapping)

MobilityRXTX

TCP/IP TCP/IP

Server Application Models

UMTS RNC Layer 1-3 (URIS)

RISE TX RX

Node B

UMTS Node B Layer 1-2 (URIS)

Modulation/Coding

(Link Level Mapping)

SRNC

Figure 3.2: Simulated nodes of the UMTS RAN

The client part of the simulated applications as listed in Section 3.1 ispart of the UE. The server side application part, which in reality would belocated in the Internet or the CN, is instantiated in the RNC node. Bothnodes include an instance of the TCP/IP stack as described in Section 3.2.The protocols of the UMTS RAN are mapped to the three nodes in ac-cordance to the UMTS specification which is presented in Chapter 2. Thenodes which directly take part in the radio communication, that is the UEand the Node B, both contain components provided by the module simu-lating the radio channel (see description of Radio Interference SimulationEngine (RISE) module in Section 3.4) and the module which is responsiblefor the link-level mapping (see Section 3.5).

3.3. UMTS Protocol Stack 73

The simulated units with which the UE and the Node B exchange dataare UMTS radio bursts. The physical channels exchange these bursts witheither fixed transmission power, e.g. in case of the Common Control Phys-ical Channel (CCPCH) and the Common Pilot Channel (CPICH), or withvarying transmission power based on the fast power control and selected SF,e.g. for DPCCH and DPDCH. Power offsets between the channels may alsobe configured. The Signal to Interference plus Noise Ratio (SINR) valuescalculated by the RISE and the link-level mappings are used to calculatethe BER and BLER. Several methods for combining the received bursts incase of Soft Handover (SHO) and HARQ retransmissions are available.

For the transmission of user data, the following transport and associatedsignalling channels are implemented. The Release 99 DCH is supported bythe MAC layer [C14] and the PHY [C2]. Several scheduling algorithms forlogical channels mapped onto DCHs are available in the MAC-d entity [C15].User traffic may also be routed from the MAC-d to the MAC-c/sh/m en-tity [C10]. The MAC-c/sh/m entity [C5] is able to schedule and transmituser data on FACH and RACH [C1] as well as on USCH and DSCH [C17].The HS-DSCH [C16] and the MAC-hs entity [C24] allow to simulate the HS-DPA. Available scheduling algorithms which are compared in Section 5.2 areRound Robin (RR), Maximum SINR (MaxSINR), Proportional Fair (PF),Expo-Linear (EL), Modified Largest Weighted Delay First (M-LWDF) andEarliest Deadline First (EDF). Finally, the MAC-e and MAC-es entities aswell as the corresponding E-DCH are implemented in the MAC layer [C12].

The RLC sublayer implements the AM, UM and TM [C7]. Selective re-peat ARQ with all specified options is implemented by the AM entity [C9].Header compression according to RFC 2507 [56] is applied by the PDCPsublayer. IPHC for both IPv4 [C23] and IPv6 [C11] is available. Further-more, broadcast and multicast traffic configurations are supported by thelayer 2 of the user plane [C6].

By the implementated protocols and entities which are summarizedabove the user plane protocols of the UMTS radio interface are completeand protocol-level simulations are possible. System-level simulations canbe performed by the following features implemented in the control plane.Control information, e.g. measurements, handover commands and radiobearer reconfigurations, are exchanged by the ASN.1 based RRC protocolin the control plane [C8]. Like traffic in the user plane the RRC PDUs areexchanged between UE and RNC by the RLC protocol. The IEs neededfor the ASN.1 based protocol are automatically generated from the RRC


specification [C20]. This generated ASN.1 message catalog allows bit-exactsystem-level simulations.

Several important RRM algorithms which use the RRC protocol areimplemented. The most important ones are the Connection AdmissionControl (CAC) with code tree management and RB up-/downgrade bytransport channel reconfiguration [C22] as well as the handover and SHOalgorithms [C19]. Furthermore, power control optimization algorithms forfemto cells [C4] as well as smart system handover algorithms using fuzzylogic techniques [C21] have been implemented. References to published per-formance evaluations of above algorithms can be found in Section 5.1.

3.4 Radio Interference Simulation Engine

The RISE module is responsible for the simulation of the radio channel. Forthe calculation of signal strength and interference various path loss, shad-owing and fading models are included. The RISE implements the genericpart needed for interference calculation. A small submodule implements theUMTS specific aspects required for interference calculation, e.g. orthogo-nality and spreading gain.

Every station in RISE is equipped with a transmitter, a receiver anda configurable antenna. Available antennas range from simple omnidirec-tional antennas to realistic three-dimensional antenna patterns. Figure 3.3illustrates different antenna patterns in a hexagonal scenario (only the cen-ter cell is shown). The relative signal strength as measured by the UEs isdepicted for a cell with a simple omnidirectional antenna, a site with threesectors using directed antennas and a three-sectored site using a realisticthree-dimensional antenna pattern.

Furthermore, station specific mobility models can be applied. Thesemobility models range from simple Brownian motion models, e.g. as usedin scenarios illustrated in Figure 3.3, to rather complex models which, forexample, simulate a realistic highway scenario including driver decisionsfor accelerating, breaking and lane changing. Figure 3.4 presents a GUIsnapshot of such a scenario on a 4 lane highway as used, for example, withinthe German research project CoCar [2].

3.5. Link-Level Simulation Module 75

(a) Omnidirectional (b) Sectorized (c) Realistic 3D pattern

Figure 3.3: Hexagonal scenario with different antenna configurations

3.5 Link-Level Simulation Module

The link-level module provides the SINR to BER and BLER mapping. Forthis purpose the physical layer coding chain and the radio channel are mod-elled and simulated by this module. The SINR values are calculated bythe RISE module. Based on the mapped BLER, the UMTS protocol stackis able to make an accurate stochastic decision on the CRC of the TBs.Depending on the probability of a correctly received TB it is either pro-cessed or dropped. Furthermore, the calculated residual bit errors may bepropagated to higher layers in case the RLC TM mode of UMTS is used.By doing so quantitative performance evaluations of services like speech orvideo telephony are possible [101]. For plain physical layer link-level studiesthe link-level module can be run stand-alone without the system-level andprotocol-level features provided by the UMTS stack.

The BER and BLER mappings can either be calculated on demandor in advance. Because of the high computational complexity the mostimportant and most frequently used mappings are, typically, precalculated.In order to calculate link-level mappings, the described module provides aset of building blocks for physical layer coding chains. Within this thesis thebuilding blocks needed for the bit-exact simulation of UMTS and WiMAX[74] have been developed.


Figure 3.4: Highway mobility model

Like all settings within WNS, the configuration of the link-level simula-tion chains is done with the programming language Python [3]. Figure 3.5gives an example for the configuration of a HS-DSCH chain. A screenshot ofthe resulting chain as shown by the GUI of a running simulation is presentedin Figure 3.6.

bs = 1262 + 24 # block size

t = 1 # transmissions

s = [1,0,1,0,1,0,1,0] # RV_s for QPSK

r = [0,0,1,1,2,2,3,3] # RV_r for QPSK

b = [0,0,1,1,1,2,3,0] # RV_b

mod = "QPSK" # modulation

cod = 3 # no. of codes

blocks = [

Source(bs),

TurboEncode(bs,[013, 015],4),

HARQTransmitter(t),

HSDPARateMatch(bs,mod,cod,s,r),

Interleave(32,30,mod,cod),

ConstellationRearrangement(mod,b),

Modulator(mod),

Spread(16,cod),

AWGN(),

MMSE(),

Despread(16,cod),

Demodulator(mod),

ConstellationRearrangement(mod,b),

Deinterleave(32,30,mod,cod),

HSDPARateMatch(bs,mod,cod,s,r),

HARQReceiver(t),

TurboDecode(bs,[013, 015],4),

Sink(bs) ]

Figure 3.5: Python configuration example of HS-DSCH chain

Every coding chain contains a source and a sink block. The source gen-erates a random bit sequence while the sink determines the number of bitand block errors. By performing these steps multiple times for a given SINRthe BER and BLER are derived. In between source and sink the blocks of

3.5. Link-Level Simulation Module 77

Figure 3.6: Screenshot of HS-DSCH link-level simulation

the transmitter coding chain, the channel, the receiver and the blocks ofthe receiver’s decoding chain are located. As visualized by the GUI in Fig-ure 3.6 the blocks of the chain exchange vectors containing different datatypes. The three types of vectors are binary valued bit vectors, floatingpoint vectors that are used to exchange soft bit information between severalreceiver blocks and complex valued vectors that carry modulated informa-tion. Every block used in the transmitter coding chain has a counterpartto be used in the chain of the receiver, e.g. modulator and demodulator.Channel coding is possible by a convolutional, a turbo and a Reed-Solomonencoder/decoder pair. Additional coding rates are achieved by blocks per-forming the RM algorithms of UMTS (see Section 2.3.1.3) or fixed ratepuncturing blocks as used in e.g. WiMAX. By including HARQ transmitterand receiver blocks, different HARQ types can be simulated. In combinationwith RM and constellation rearrangement (compare Section 2.3.1.5) blocks,incremental redundancy as used by the HS-DSCH is available. Robustnessagainst burst errors is achieved by a configurable block interleaver, used inboth UMTS and WiMAX, and a bit interleaver as specified for WiMAX.For CDMA based systems a block responsible for spreading and multicodetransmission exists. Orthogonal Frequency Division Multiplexing (OFDM)based systems are supported by an OFDM modulator with configurable FastFourier Transform (FFT) size and Cyclic Prefix (CP) length as well as a set


of subcarrier mapping blocks, e.g. Partial Usage of Subchannels (PUSC) asused in Mobile WiMAX. Further blocks deal with the padding of data andmultiplexing of control and data channels within single bursts.

The radio channel is simulated by either a block containing a simpleAWGN channel or more complex blocks which apply predefined and con-figurable fading profiles. Available predefined fading profiles are the Inter-national Telecommunication Union (ITU) pedestrian and vehicular fadingmodels. For fading channels several receiver techniques have been imple-mented within the link-level module. These receivers range from a sim-ple rake receiver to receivers with higher computational complexity using,e.g., Zero Forcing (ZF) and Minimum Mean Square Error (MMSE) channelequalization to eliminate Inter Symbol Interference (ISI) caused by multi-path propagation.

3.6 Real Time Wireless Network Demonstrator

The Real Time Wireless Network Demonstrator (RTWND) module allowsto turn the WNS into a real time network emulator. To do so the RTWNDmodule fulfills several tasks within WNS. In order to run in real time themodule interacts with the event scheduler of the simulation environment.A prerequisite of the real time capability is a sufficiently small number ofsimulated UEs and Node Bs to not exceed the processing capabilities of thehost system. Successful emulations have been made with one UE and oneNode B on an 800 MHz host system [C3].

A further task of the RTWND module is to inject real traffic insteadof traffic originating from traffic generators into the WNS. Therefore, itreplaces the application and TCP/IP modules. Higher layer IP traffic isreceived and sent by an Ethernet device implemented in the Linux kernelof the host system. Figure 3.7 illustrates a typical instance of the networkemulation system as used by the RTWND module.

By default the emulator acts as a bridge between two segments of anEthernet network and, therefore, is transparent to most of the networktraffic. Hence, traffic that is required by the network infrastructure (e.g.Address Resolution Protocol (ARP), Dynamic Host Configuration Protocol(DHCP), Network File System (NFS), Network Time Protocol (NTP), Net-work Information Service (NIS)) is not influencing the traffic on an emu-lated radio channel and vice versa. Only traffic of selected hosts is routed

3.6. Real Time Wireless Network Demonstrator 79

Bridging

Routing

br0

eth0 eth1

wns0

Emulator Box

Wireless Network Simulator

Mobile 1

Internet

Mobile 3Mobile 2

Application

Server

RNCUEUE

Figure 3.7: Real Time Wireless Network Demonstrator

through the simulator. Furthermore, it is possible to specify the UDP andTCP source and destination ports of connections for which the traffic shouldbe sent over the emulated radio channels. The mapping of IP traffic to thesimulated UEs and the decision about uplink and downlink is performedinside the simulator by the source and destination IP addresses. A detailedoverview about the open source utilities which have been used for this setupcan be found in [124] and [107].

The traversing packets sent to the WNS are influenced by the followingaspects. IP packets are delayed because of queueing, scheduling, trans-mission duration and retransmissions. Packets may get lost in case of fulltransmission queues, configurations without ARQ, maximum number of re-transmissions and other discard mechanisms. Finally, bit errors based on theresidual BER of the radio transmissions might be generated in the payloadrouted through WNS. In order to forward such packets RTWND recalcu-lates the checksum of the TCP or UDP headers used to tunnel the payloadto the emulator box. Quality evaluations of speech services, video streamingand video telephony can be performed using this configuration. In [102], forexample, the RTWND module which has been developed within the scopeof this thesis has been used to study the performance of video telephonyservices in UMTS.


3.7 Graphical User Interface

In order to study and assist the configuration of simulation scenarios inWNS a GUI has been developed. Furthermore, the visualization of scenar-ios and results, e.g. as used in research projects like [2], is aided by theGUI. Both system-level and link-level aspects can be displayed. Figure 3.3and Figure 3.4 show parts of a running system-level scenario. Lines betweenthe UEs and the Node Bs illustrate the physical layer communication re-lationship. Different colors represent different physical channels. Next tothose ”radio links” detailed information about transmission power, frequen-cies, channelization and scrambling codes can be shown. As depicted inthe figures a best-server map and a path loss map can be overlaid on theillustrated scenario. Finally, detailed station specific information, e.g. codetree usage, SHO measurements or buffer occupancies, can be illustrated byclicking on the corresponding Node B or UE.

In a further perspective (see Figure 3.6) information about link-levelchains is visualized. The blocks of each chain include information aboutthe most important configuration parameters. In between the blocks thetype and length of the exchanged vectors are shown. Depending on thetype of vector, i.e. bit vector, soft-bit vector or complex-valued vector,different illustrations of the exchanged data are drawn next to the chain.More detailed information about an ongoing link-level simulation, e.g. thecurrent state of the BER and BLER mapping as well as a visualization ofthe configured fading, is given in another window.

CHAPTER 4

Link-Level Performance Evaluation

Contents4.1 Modulation . . . . . . . . . . . . . . . . . . . . . . . . 83

4.2 Coding . . . . . . . . . . . . . . . . . . . . . . . . . . 86

4.3 Dedicated Channel . . . . . . . . . . . . . . . . . . . . 89

4.3.1 Downlink . . . . . . . . . . . . . . . . . . . . 92

4.3.2 Uplink . . . . . . . . . . . . . . . . . . . . . . 94

4.4 High Speed Downlink Shared Channel . . . . . . . . . 96

4.5 Mobile WiMAX . . . . . . . . . . . . . . . . . . . . . 100

4.5.1 Comparison of HSDPA and Mobile WiMAXPhysical Channel Capacity . . . . . . . . . . . 103

4.6 Throughput Comparison . . . . . . . . . . . . . . . . 104

4.6.1 AWGN Channel . . . . . . . . . . . . . . . . . 104

4.6.2 Throughput for Pedestrian Channel Model . . 108

4.6.3 Throughput for Vehicular Channel Model . . 109

4.7 Hybrid ARQ . . . . . . . . . . . . . . . . . . . . . . . 111

4.7.1 Chase Combining and Incremental Redundancy 114

4.7.2 Constellation Rearrangement . . . . . . . . . 115

4.7.3 Comparison . . . . . . . . . . . . . . . . . . . 116

In this chapter the performance of UMTS for PS services is evaluated onlink level. One main aspect this chapter targets is the quantification of

the achievable maximum throughput depending on certain variables, e.g.UMTS release, configuration options, radio conditions and receiver algo-rithms. Because only basic physical layer configurations can be evaluatedanalytically a link-level simulator has been used to get detailed results of thebit-accurate coding chains of the individual systems. As in most link-levelsimulations, only one radio link is modelled in detail. Neighbour cells andother UEs are considered by the interference they contribute to the SINR.

82 4. Link-Level Performance Evaluation

PS services in UMTS may be provided by several transport channels. Typ-ically, the DCH or the HS-DSCH is used to deliver packet based contentto the UEs. Depending on the release of the UMTS specification, the UE’scategory and operator specific preferences, several configuration options arepossible. Based on the module presented in Section 3.5 the most commonsetups are evaluated with respect to their performance. A comparison toanother state-of-the-art system is achieved by including similar results formobile WiMAX which has been added under the name IP-OFDMA andlater OFDMA TDD WMAN as a sixth mobile communication system tothe IMT-2000 family.

The configuration files used for the scenarios studied in this and Chap-ter 5 together with the respective simulation results are available for down-load from [99].

In the first two sections of Chapter 4 important elements of the physicallayer coding chains of all considered systems are evaluated and compared.In detail the modulation and channel coding are examined in Section 4.1 andSection 4.2, respectively. For the validation of the implemented functionalityand the simulation results, a comparison to analytical models and relatedempirical work is made.

The following sections analyze the complete physical layer coding chainsof the regarded systems for an AWGN channel. In Section 4.3 the DCH isstudied in both uplink and downlink direction. Several typical configura-tions but also non-common multi-code configurations are considered. Al-ternative TTI settings are simulated as well. Section 4.4 describes corre-sponding results for the HS-DSCH. Both Release 5 and Release 7 HS-DSCHconfigurations are evaluated. For validation purposes the obtained resultsare compared to empirical models. In order to benchmark the UMTS re-sults with another state-of-the-art system, comparable evaluations for Mo-bile WiMAX are performed and presented in Section 4.5. Furthermore, abasic comparison of the available physical channel data bits of HS-DSCHand Mobile WiMAX is made in this section.

Based on the previous results Section 4.6 draws a comparison in termsof achievable throughput with respect to theoretical limits. Additionallyto AWGN based evaluations, all systems and configurations are comparedon fading channels with various receiver techniques. A neutral and faircomparison is realized by using the same channel models and similar physicallayer configuration and simulation assumptions. The maximum theoretical

4.1. Modulation 83

throughput and the efficiency of the UMTS configurations on various levelsalso above the physical layer are compared as well.

Finally, Section 4.7 studies HARQ by means of comparing differentHARQ schemes. In detail the HARQ schemes Chase Combining and Incre-mental Redundancy as well as a scheme using Constellation Rearrangementonly are evaluated according to their performance and technical complexitywith respect to the UE categories.

4.1 Modulation

In order to transmit information over a radio channel the carrier needs tobe modulated. In mobile communication systems like UMTS and WiMAXthe phase of the (sub-)carrier frequency is modulated. In addition to thisPhase-Shift Keying (PSK) both systems make use of Quadrature AmplitudeModulation (QAM) which also varies the amplitude of the signal to transmitinformation.

The bit error and symbol error probability of both modulation schemesfor an AWGN channel can be calculated by using the Marcum Q-function

Q(x) =1

2erfc(

x√2

) =1√2π

∫ ∞x

e−t2

2 dt , x ≥ 0 (4.1)

where erfc is the complementary Gaussian error function.For the most simple PSK modulations, i.e. BPSK and QPSK as il-

lustrated in Figure 2.30(a) and Figure 2.29(a), respectively, the bit errorprobability can be expressed by

Pb,PSK = Q

(√2EbN0

)(4.2)

where Eb is the energy per bit and N0 is the noise power spectral den-sity [122]. It must be noted that for QPSK twice the transmission poweris needed to achieve the same bit error probability since two bits are trans-mitted by one modulation symbol. In Figure 4.1(a) this difference resultsin a shift of 3 dB between the BPSK and QPSK mapping curves.

The symbol error probability of QPSK, i.e. the probability that at least


one bit is received incorrectly, can be expressed by

Ps,QPSK = 1− (1− Pb)2 = 2 Q

(√EsN0

)−Q

(√EsN0

)2

(4.3)

where Es is the energy per transmitted symbol. For BPSK the bit errorprobability and the symbol error probability are the same.

The relation between SINR, as used as the unit of the x-axis in thefollowing diagrams, and Eb and Es is given by

SINR =EbN0· fbB

=EsN0· fsB

(4.4)

where fb and fs are the bit rate and symbol rate, respectively. The variableB is the channel bandwith which is 5 MHz in UMTS.

For rectangular QAM constellations as shown in Figure 2.29 the expres-sions for both the bit error probability and the symbol error probability areslightly more complex. The bit error probability depends on the exact map-ping of bits to modulation symbols and the bits that are transmitted. For aGray-coded mapping the average bit error probability can be calculated by

Pb,QAM = 1−

(1− 2

log2M

(1− 1√

M

)Q

(√3 log2M

M − 1

EbN0

))2

(4.5)

and the symbol error probability is expressed by

Ps,QAM = 1−

(1− 2

(1− 1√

M

)Q

(√3

M − 1

EsN0

))2

(4.6)

where M is the number of modulation symbols, i.e. 16 for 16QAM and64 for 64QAM. As can be seen in Figure 4.1 the analytical bit error andsymbol error results provided by the equations above perfectly match theresults of the link-level simulator introduced in Section 3.5.

For the two-dimensional modulation schemes 16QAM and 64QAM thebit error probability depends on the bit pattern that is to be transmitted.Each bit of this bit pattern has a different demodulation probability. Fig-ure 4.2 shows the SINR to BER mapping for the individual bits belongingto one modulation symbol. A random bit sequence is used as input for the

4.1. Modulation 85

20 10 0 10 20 30SINR [dB]

10-4

10-3

10-2

10-1

BER

BPSK QPSK 16QAM 64QAM

Bit Error Rate

Analytical

Simulated

(a) Bit error rate

20 10 0 10 20 30SINR [dB]

10-4

10-3

10-2

10-1

100

SER

BPSK QPSK 16QAM 64QAM

Symbol Error Rate

Analytical

Simulated

(b) Symbol error rate

Figure 4.1: Bit error and symbol error rate of modulation schemes

modulation. The dotted line marks the average BER of all bits as alreadyshown in Figure 4.1.

In UMTS the preferred bit positions i1 and q1 are typically used totransmit the systematic bits of the turbo coder (see Section 2.3.1.2) as theyare of higher importance to correctly decode a transmission at the receiver.Parallel interleavers as illustrated in Section 2.3.1.4 are used to maintainthese bit positions through the block interleaving. When HARQ is used theconstellation rearrangement introduced in Section 2.3.1.5 may be used tomodify the bit to symbol mapping by reshuffling and inverting individualbits of one modulation symbol. Bits which are transmitted on a weak bitposition may be retransmitted on a bit position with a stronger demodula-tion probability and vice versa. A detailed analysis of this feature is givenin Section 4.7.2.

It can be concluded that the modulation models implemented withinthe link level simulation chain are validated. Quantitative comparisons ofthe modulation schemes which are used for both UMTS and WiMAX havebeen carried out. In order to compare the previous and the following resultswith those found in other scientific literature, important definitions, e.g.SINR, as being used in this thesis have been introduced. It has been shownthat the bit to symbol mapping has an impact on the decoding probability


10 5 0 5 10 15 20 25SINR [dB]

10-4

10-3

10-2

10-1

BER

i1 ,q1

i2 ,q2

16QAM

Single bit BER

Average BER

(a) 16QAM

10 5 0 5 10 15 20 25SINR [dB]

10-4

10-3

10-2

10-1

BER

i1 ,q1

i2 ,q2

i3 ,q3

64QAM

Single bit BER

Average BER

(b) 64QAM

Figure 4.2: Simulated bit error probability of 16QAM and 64QAM

of the individual bits. This circumstance will be further investigated inlater sections. In the next section channel coding methods are added andcompared with respect to their performance.

4.2 Coding

Because of the relatively high bit error rate of the wireless channel FEC isrequired to assure a reliable transmission. In UMTS a convolutional coderof rate 1/2 and 1/3 as well as a turbo coder of rate 1/3 are available for FEC.Both have been introduced in Section 2.3.1.2.

The generator polynomials for the convolutional coder of rate 1/2 aregiven by Eqs. (2.5) and (2.6). The polynomials for the 1/3 rate coder arelisted by Eqs. (2.7) to (2.9). The coders of both rates have a constraintlength of 9. By using Eq. (4.7) an upper bound of the residual bit errorprobability of these convolutional codes can be calculated [58]:

Pb <1

k

∞∑d=dfree

Bd exp

{−EbN0

}d(4.7)

The free distance dfree, the minimal Hamming distance between differentcode words, is 12 and 18 for the 1/2 rate and 1/3 rate convolutional code,respectively. The number of non-zero information bits on all weight dfree

4.2. Coding 87

paths through the state diagram of the coder is Bdfree = 33 and Bdfree = 11.The factor k is the number of information bits per time unit. For theconvolutional codes in UMTS it is 1.

The turbo coder of UMTS contains two encoders with the transfer func-tion found in Eq. (2.10). The constraint length of the turbo coder is 4.In Figure 4.3 the residual bit error probabilities of both the convolutionalcoders and the turbo coder are shown for a BPSK transmission on an AWGNchannel. The bit error upper bound calculated by Eq. (4.7) is also shownas a reference. It can be seen that turbo decoding with 8 iteration showsa better performance compared to the convolutional code of rate 1/3. Fora residual BER of 10−5, for example, the turbo code has a coding gain2 dB higher than the corresponding convolutional code. A similar codinggain compared to the 1/2 rate convolutional code is achieved by the turbocode when punctured to the same rate. Because of the better properties PSservices typically make use of turbo coding although of its higher decodingcomplexity.

10 8 6 4 2 0SINR [dB]

10-5

10-4

10-3

10-2

10-1

BER

CC 1/2CC 1/3

Convolutional Code

Analytical

Simulated

(a) Convolutional code

10 8 6 4 2 0SINR [dB]

10-5

10-4

10-3

10-2

10-1

BER

TC 1/3 TC 1/2(punctured)

Turbo Code

(b) Turbo code

Figure 4.3: Bit Error Rates of UMTS Forward Error Correction

In Figure 4.4 a comparison of the BLER and the achievable throughputfor the different channel coding options is given. As before the assumedchannel is an AWGN channel and BPSK modulation is used. A code blocksize of 1912 bits has been chosen for all configurations. The coding gain ofthe turbo coder compared to the convolutional coder can be observed again.


For a BLER of 1% it is 2.1 dB for rate 1/3 and 1.2 dB for a coding rate of1/2.

As a comparison Figure 4.4 shows the coding performance of the con-volutional coder of the 802.16 OFDM PHY. In order to only compare thecoding properties other differences of 802.16 are not considered. Like inmany OFDM based systems a concatenation of a convolutional code and ablock code is used. In the OFDM PHY of 802.16 a (255,239) Reed-Solomoncode is concatenated with a convolutional code of rate 1/2 using the poly-nomials x6 + x5 + x4 + x3 + 1 and x6 + x4 + x3 + x + 1. Although theproperties of the convolutional code of 802.16 are worse compared to thehalf-rate code of UMTS, e.g. smaller constraint length, the concatenationwith the Reed-Solomon achieves a performance between the two convolu-tional codes of UMTS. Compared to the UMTS convolutional code of rate1/2 a coding gain of 0.9 dB exists at 1 % BLER. The turbo code puncturedto rate 1/2 is, however, still 0.8 dB better at the same BLER.

6 5 4 3 2 1 0 1SINR [dB]

10-3

10-2

10-1

100

BLE

R

TC 1/3

TC 1/2

RS +CC 1/2

CC 1/2

CC 1/2CC 1/3

Block Error Rate

UMTS

802.16

(a) Block Error Rate

7 6 5 4 3 2 1 0SINR [dB]

101

102

103

104

105

Thro

ughput

[kbit

/s]

TC 1/3 TC 1/2

RS +CC 1/2

CC 1/2

CC 1/2CC 1/3

Throughput

UMTS

802.16

Shannon bound

Shannon bound shifted by 3 dB

(b) Throughput

Figure 4.4: Block Error Rates and Throughput of Forward Error Correction

A good evaluation of the coding efficiency can be made by comparing theachievable throughput against the theoretical maximum based on Shannon’sformula. This throughput is calculated by

C = B · log2(1 +S

N) (4.8)

4.3. Dedicated Channel 89

where B is the channel bandwidth in Hertz, i.e. 5 MHz for this UMTSperformance evaluation, and S

N is the signal to noise power ratio. In UMTSthe interference from other stations can be interpreted as uncorrelated noise(compare Section 2.3.2.5). To express this contribution caused by the Mul-tiple Access Interference (MAI) the term SINR is used for this power ratio.Additionally to the theoretical maximum throughput, the throughput com-parison of Figure 4.4 includes the Shannon bound shifted by 3 dB as amore realistic limit. The uncoded data rate in the illustrated scenario hasbeen chosen to be 3.84 Mbit/s which corresponds to the chip-rate of UMTSwhen BPSK without spreading is applied. As can be seen the turbo coderof UMTS has the highest efficiency. Compared to all other coding schemesits throughput distribution is closest to the Shannon bound.

In summary it can be said that the implemented channel coding blockshave been validated by comparing the simulated residual bit error proba-bilities with corresponding analytical bounds. The turbo code of UMTSshowed to have a significantly higher performance compared to the convo-lutional codes of same rate. Comparing the convolutional codes of UMTSand WiMAX it turned out that the UMTS coder is slightly superior. Bycombining the convolutional code of WiMAX with a Reed-Solomon blockcode the performance exceeds the convolutional code of UMTS but stilldoes not reach the efficiency of the UMTS turbo code. After modulationand channel coding have been investigated standalone the following sectionscompare the systems by incorporating the complete bit accurate physicallayer coding chains.

4.3 Dedicated Channel

The DCH already introduced in the first releases of the UMTS specifica-tions is the standard channel used for the delivery of dedicated user trafficin today’s UMTS networks. Speech services, video telephony and a basicInternet connectivity are realized by DCHs. The data rate of a DCH is typ-ically fixed within a short time-interval. For CBR services like speech andvideo telephony this is the preferred choice. However, PS data services oftenhave a bursty traffic pattern. In order to support the transmission of suchservices the data rate of the DCH may be varied by changing the Spread-ing Factor (SF) of the physical channel on which the DCH is mapped. By


including the spreading gain in Eq. (4.2) and normalizing it to the energyper chip Ec the bit error probability can be calculated with

Pb,QPSK = Q

(√SF · EcN0

)and Pb,BPSK = Q

(√SF · 2EcN0

). (4.9)

The SF may range from 4 to 512 in the downlink and from 4 to 256 in theuplink. In the downlink a QPSK modulation is used and the DPDCH, whichcarries the user traffic of the DCH, and the DPCCH are time-multiplexed.In the uplink the DPDCH is transmitted using a BPSK modulation on thein-phase branch. The DPCCH is transmitted in parallel on the quadra-ture branch. Figure 4.5 shows the resulting bit error probabilities bothfor the downlink and the uplink. Because of the reduced spreading gainthe transmission power needs to be increased to achieve the same bit errorprobabilities in case the data rate is increased by SF reduction. Again thesimulation results are validated by the analytic results gained from Eq. (4.9).

40 30 20 10 0 10SINR [dB]

10-4

10-3

10-2

10-1

BER SF4

SF8

SF16

SF32

SF64

SF128

SF256

SF512

Downlink

QPSK, analytical

QPSK, simulated

(a) Downlink

40 30 20 10 0 10SINR [dB]

10-4

10-3

10-2

10-1

BER SF4

SF8

SF16

SF32

SF64

SF128

SF256

Uplink

BPSK, analytical

BPSK, simulated

(b) Uplink

Figure 4.5: Bit error probability depending on SF

In the downlink the mechanism of changing the SF is often referred to asRB rate switching or RB up- and downgrading. Because layer 3 signallingis required for this modification and the modification is initiated after a BObased triggering condition needs to be fulfilled for some time the RB up-and downgrading is a rather slow process. In between these reconfiguration


procedures the Node B continuously transmits with the configured SF. Ifmore than those bits that fit into the burst are to be transmitted the RMalgorithm applies puncturing. If less data is available for transmission thancould be transmitted the Node B uses DTX. No padding or repetition ofcoded bits is performed and the transmitter is switched off for the accordingbits. In the downlink a typical set of RB configurations is listed in Table 4.1.The SF is varied between 32 and 8 which leads to a maximum user datarate ranging from 64 kbit/s to 384 kbit/s. This data rate, achievable on RBlevel, results from the 5% higher data rate of the DCH assuming that noMAC multiplexing is performed and 2 bytes RLC AM header are includedin every TB. According [43] there exist two alternative configurations forSF 8. The more common one uses a TTI of 10 ms. The alternative one,shown in brackets in Table 4.1, has a TTI of 20 ms.

In the uplink the UE is in charge of changing the SF according to theamount of data that is to be transmitted. Both puncturing and repetition ofcoded bits may be applied to exactly fill the radio burst. DTX is never usedfor DCH transmission in the uplink. The same TFSs as in the downlink areavailable in the uplink. Depending on the configured data rate a minimumSF of 16, 8 or 4 is required (compare Table 4.1).

Table 4.1: Typical RB configurations

SF (downlink) 32 16 8SF (uplink) 16 - 64 8 - 64 4 - 32 (4 - 64)TTI 20 ms 20 ms 10 ms (20 ms)TF 0 0 · 336 bits 0 · 336 bits 0 · 336 bitsTF 1 1 · 336 bits 1 · 336 bits 1 · 336 bitsTF 2 2 · 336 bits 2 · 336 bits 2 · 336 bitsTF 3 3 · 336 bits 4 · 336 bits 4 · 336 bitsTF 4 4 · 336 bits 8 · 336 bits 8 · 336 bitsTF 5 12 · 336 bitsTF 6 (16 · 336 bits)TF 7 (20 · 336 bits)TF 8 (24 · 336 bits)Maximum DCH data rate 67.2 kbit/s 134.4 kbit/s 403.2 kbit/sAchievable RB throughput 64 kbit/s 128 kbit/s 384 kbit/s


The bit-exact simulation of the typical TFSs found in Table 4.1 is in-vestigated in the following. For an AWGN channel the BLER mappingsfor the downlink and the uplink are depicted in Figure 4.6 and Figure 4.7,respectively. Both the results for a 20 ms and a 10 ms configuration arepresented.

4.3.1 Downlink

25 20 15 10 5SINR [dB]

10-3

10-2

10-1

100

BLE

R

-17.44dB

1TBSF32

4TBsSF32

-14.43dB

1TBSF16

8TBsSF16

-11.42dB

1TBSF8

12TBsSF8

24TBsSF8

-8.38dB

1TB2*SF8

28TBs2*SF8

48TBs2*SF8

DCH Downlink (20 ms TTI) - AWGN channel - Rake receiver

64 kbps, SF 32

128 kbps, SF 16

384 kbps, SF 8

768 kbps, 2*SF 8

(a) 20 ms TTI

25 20 15 10 5SINR [dB]

10-3

10-2

10-1

100

BLE

R

-11.41dB

1TBSF8

8TBsSF8

12TBsSF8

-8.40dB

1TB2*SF8

12TBs2*SF8

24TBs2*SF8

DCH Downlink (10 ms TTI) - AWGN channel - Rake receiver

384 kbps, SF 8

768 kbps, 2*SF 8

(b) 10 ms TTI

Figure 4.6: Simulated DCH downlink BLER on an AWGN channel

In the downlink the mapping curves for each RB configuration are closelygrouped together in terms of required SINR. To achieve a certain BLERthe Node B can transmit to each UE with a TF independent signal poweras controlled by the inner loop PC. TF specific gain factors like used in theUE are not required (compare Section 2.3.2). When only a small numberof TBs are to be transmitted no repetition of bits which would introducean unnecessary coding gain is done. DTX is used instead to save radioresources and reduce downlink interference.

As depicted in Figure 4.6 a 3 dB distance between the TFSs of the64 kbit/s, 128 kbit/s and 384 kbit/s RB configurations can be identified.This separation directly results from the spreading gains of the different SFs.When a SF change by RB reconfiguration is performed the transmissionpower at the Node B must be changed accordingly.


Higher data rates than 384 kbit/s can be achieved by multicode trans-mission. A rarely used 768 kbit/s configuration exists in the downlink whichuses two codes of SF 8 [43]. The 768 kbit/s TFS for a 10 ms TTI is exactlythe one of a 20 ms TTI 384 kbit/s channel. To achieve 768 kbit/s with a20 ms TTI this TFS is extended by TFs which support transmissions up to48 TBs. As shown in Figure 4.6 the multicode transmission using two codesof SF 8 has similar properties as if the SF would have been reduced to 4. A3 dB distance to the 384 kbit/s configuration exists even though the sameSF is used.

A bigger spacing between the curves exists for SF 8 in case TFs cor-responding to data rates above 256 kbit/s are used. When two codes ofSF 8 are used data rates above 512 kbit/s require a higher SINR. As thenumber of turbo coded bits exceeds the available DPDCH bits given by Ta-ble 4.2, puncturing is performed by the RM algorithm. For TF 8, allowingthe highest data rate of 384 kbit/s, 7128 coded bits are punctured. Thesame coding rate is achieved in a 10 ms TTI configuration by puncturing3564 bits. 2904 bits still need to be punctured when 20 TBs are to betransmitted (TF 7). All other TFs of a single code transmission using SF 8do not require any puncturing. For the multicode configuration introducedabove the same puncturing ratios are used for data rates above 512 kbit/s.

Looking at the higher SFs puncturing is only needed for TF 4 whenSF 32 is configured. Here the worse ratio of DPDCH to DPCCH bits fromthe DPCH time multiplexing (compare Table 4.2) requires 36 bits to bepunctured. This small amount of puncturing only has a minor negativeimpact on the mapping by shifting it slightly to higher required SINR values.

Table 4.2: Downlink burst formats used in simulation

SF 32 16 8DPCCH bits/slot 20 (4+8+8) 32 (8+8+16) 32 (8+8+16)

(NTPC + NTFCI + NPilot)DPDCH bits/slot 140 (28+112) 288 (56+232) 608 (120+488)

(NData1 + NData2)DPDCH bits/10 ms TTI 2100 4320 9120DPDCH bits/20 ms TTI 4200 8640 18240DPDCH data rate [kbit/s] 210 432 912


The exact amount of required puncturing can be calculated as follows.By considering the 16 bit CRC value which is used for DCHs and the max-imum number of turbo coded bits (5114 as introduced in Section 2.3.1.2)the number of code blocks nc and their size sc can be calculated with

nc =

⌈nt · (st + 16)

5114

⌉and sc =

⌈nt · (st + 16)

nc

⌉(4.10)

where nt is the number of TBs and st is the TB size. Taking into accountthe coding rate 1/3 and the 12 bit trellis termination of the turbo coder, thenumber of bits before RM

br = nc · (3 · sc + 12) (4.11)

can be derived. If br is larger than the number of available DPDCH bitsper TTI, i.e. 30 or 15 times the available DPDCH bits per slot as given byTable 4.2, puncturing is done. If br is smaller, repetition or DTX insertionis performed. The difference of both values gives the number of bits thatare punctured or repeated.

For TFs where DTX is applied the same effective coding rate of 1/3,i.e. the native code rate of the turbo code, is used. Nevertheless, for lowBLERs the curves of these TFs divide. The reason is a gain contributedby larger code block sizes. For each SF the lowest SINR is required for theTFs having the largest code block size. In case of SF 32 and SF 16 this isthe TF with the highest number of TBs. When SF 8 is used with a 20 msTTI the TF allowing to transmit 12 TBs requires the lowest SINR. Eventhough the TF with 16 TBs uses DTX its coding probabilities are worth ascode block segmentation according to Eq. (4.10) takes place.

4.3.2 Uplink

In the uplink the UE switches the SF according to the instantaneous datarate. Figure 4.7 shows the BLERs for all uplink configurations of the TFsgiven in Table 4.1. When a 64 kbit/s TFS is configured the SF is switchedbetween 64 and 16. 1 to 4 TBs can be transmitted per TTI. A 128 kbit/sTFS adds a TF for SF 8, allowing to transmit 8 TBs, to the 64 kbit/s TFS.The TF for 3 TBs (dashed line in Figure 4.7) is not used in a 128 kbit/sconfiguration. The 20 ms TTI TFS for 384 kbit/s extends the 128 kbit/sTFS by 4 TFs of SF 4. In a 10 ms setup the SF ranges from 32 to 4 and 1to 12 TBs can be transmitted per TTI.


25 20 15 10 5SINR [dB]

10-3

10-2

10-1

100B

LER

1TBsSF64

-23.76dB

2TBsSF32

-20.73dB

3TBsSF16

4TBsSF16

-17.72dB

8TBsSF8

-14.72dB

12TBsSF4

16TBsSF4

-11.71dB

20TBsSF4

24TBsSF4

DCH Uplink (20 ms TTI) - AWGN channel - Rake receiver

64 kbps

128 kbps

384 kbps

(a) 20 ms TTI

25 20 15 10 5SINR [dB]

10-3

10-2

10-1

100

BLE

R

1TBsSF32

-20.73dB

2TBsSF16

-17.73dB

4TBsSF8

-14.72dB

8TBsSF4

-11.71dB

12TBsSF4

DCH Uplink (10 ms TTI) - AWGN channel - Rake receiver

384 kbps

(b) 10 ms TTI

Figure 4.7: Simulated DCH uplink BLER on an AWGN channel

Most of the TFs used in the uplink require a small amount of bit rep-etition. The TFs allowing to transmit 1, 2, 4, 8 and 16 TBs all share aneffective coding rate of 0.29. As depicted in Figure 4.7 3 dB spacing causedby the different SFs exists between these TFs. A higher amount of repeti-tion is only applied when 3 or 12 TBs are transmitted. A coding gain of1 dB resulting from the effective code rate of 0.22 is responsible for shiftingthe BLER curve. As derived by Table 4.3 puncturing is only needed for thehigh rate TFs of SF 4. When 20 TBs are transmitted the effective codingrate is 0.37. The TFs used to transmit with 384 kbit/s have a coding rateof 0.44.

Table 4.3: Uplink DPDCH configurations

SF 64 32 16 8 4DPDCH bits/slot 40 80 160 320 640DPDCH bits/10 ms TTI 600 1200 2400 4800 9600DPDCH bits/20 ms TTI 1200 2400 4800 9600 19200DPDCH data rate [kbit/s] 60 120 240 480 960


To achieve the same BLER for all data rates the UE changes the trans-mission power depending on the TF. TF specific gain factors are used toreduce the transmission power and, therefore, save battery power when onlya small amount of TBs are to be transmitted. The gain factors control therelative transmission power between DPDCH and DPCCH which uses afixed SF of 256 and is controlled by the inner loop PC. DTX is not used inthe uplink in order to reduce the peak-to-average power ratio and to avoidaudible interferences.

It can be summarized that the DCH allows to adapt its data rate bychanging the SF. Nevertheless, switching the RB configuration is a ratherslow process which, at least in the downlink, has some overhead with re-spect to the required radio resources. Therefore, the DCH configurationis typically kept constant for a relatively long time. Once configured for acertain data rate the fast power control of the DCH adapts the transmissionpower in order to assure that a reliable transmission with this data rate ispossible. This makes the DCHs suited well for the delivery of CBR serviceswith a relatively low data rate below 384 kbit/s.

4.4 High Speed Downlink Shared Channel

In contrast to the DCH which uses fast power control and TF specific gainfactors as methods to assure a certain SINR target the HSDPA primarilyapplies AMC for link adaptation. The transmission parameters which arevaried by the link adaptation according to the instantaneous channel condi-tions are the modulation scheme, the number of codes and the coding rate.If a higher order modulation scheme or a certain number of codes is notsupported by the UE’s category the Node B transmits with the maximumpossible AMC scheme and reduces the transmission power instead. Thepower offset is set so that the targeted BLER is still reached.

The data resources allocated to one mobile are always modulated withone single modulation scheme at one point in time. UMTS Release 5 sup-ports QPSK and 16QAM. In Release 7 of the UMTS specification 64QAMhas been added as a modulation scheme for the HS-DSCH. The HS-DSCHis able to use up to 15 codes, i.e. 15 HS-PDSCHs, of SF 16. The remainingcode of SF 16 can not be used because several codes of a higher SF are allo-cated for signalling and system management. The corresponding channels

4.4. High Speed Downlink Shared Channel 97

are, for example, the CPICH, the HS-SCCH and associated DPCHs.Channel coding is always done by the turbo code as described in Sec-

tion 2.3.1.2. While the turbo coder is the same as the one already beinganalyzed in Section 4.2 and Section 4.3, the RM of the HS-DSCH differsfrom that of the DCH. As described in Section 2.3.1.3 the RM for theHS-DSCH allows to apply different RM patterns and to prioritize eithersystematic or parity bits.

In HSDPA the Node B has a large flexibility to select an AMC scheme.In addition to the modulation schemes the number of codes can be variedbetween 1 and 15. Furthermore, a large set of effective coding rates is possi-ble due to the RM algorithm which maps an arbitrary number of bits fromthe turbo coder onto the available bits specified by the modulation schemeand number of codes. These coding rates vary between 0.17 (repetition ofbits) and 0.89 (puncturing of bits). In order to reduce the number of po-tential simulations this study limits the number of schemes to those whichare used for Channel Quality Indicator (CQI) reporting by the mobile [20].Table 4.4 lists these reference schemes as used by the UE to signal the max-imum possible AMC based on the instantaneous channel conditions. Thecomplete set of possible TB sizes can be found in [30].

For the HS-DSCH a set of one to three block interleavers of 32 rows and30 columns is used for each physical code. As described in Section 2.3.1.4the number of these interleavers depends on the chosen modulation scheme.A CRC field of 24 bits is added to every TB received from the MAC layer.

Figure 4.8 shows the SINR to BER and SINR to BLER mappings forthe reference AMC schemes as listed in Table 4.4. The simulated channel isa simple AWGN channel. The curves of the BER mapping are grouped bythe chosen modulation scheme which can be seen in particular for high BERvalues. These groups are further subdivided by the number of codes usedfor transmission. For lower BERs these groups branch off based on theircoding rate. In general higher order modulation schemes, higher numberof codes and higher coding rates require a higher SINR as well in order toreach the same BER.

The combinations of the variable AMC parameters have been selectedin such a way as to allow Release 5 UEs to report the channel quality with agranularity of 1 dB. This granularity can be observed at the BLER mappingswhich are roughly spaced by 1 dB. When transmitting the Node B may evenselect Modulation Coding Schemes (MCSs) which would result in mappingsbetween the illustrated ones. The reporting range for a targeted BLER of


Table 4.4: CQI table for category 10 and category 14 (Release 7)

CQI Modulation Codes TB size Code rate0 NA 0 0 01 QPSK 1 137 0.1682 QPSK 1 173 0.2053 QPSK 1 233 0.2684 QPSK 1 317 0.3555 QPSK 1 377 0.4176 QPSK 1 461 0.5057 QPSK 2 650 0.3518 QPSK 2 792 0.4259 QPSK 2 931 0.49710 QPSK 3 1262 0.44711 QPSK 3 1483 0.52312 QPSK 3 1742 0.61313 QPSK 4 2279 0.60014 QPSK 4 2583 0.67915 QPSK 5 3319 0.69616 16QAM 5 3565 0.37417 16QAM 5 4189 0.43918 16QAM 5 4664 0.48819 16QAM 5 5287 0.55320 16QAM 5 5887 0.61621 16QAM 5 6554 0.68522 16QAM 5 7168 0.74923 16QAM 7 9719 0.72524 16QAM 8 11418 0.74525 16QAM 10 14411 0.75226 16QAM / 64QAM 12 / 10 17237 / 15761 0.749 / 0.54827 16QAM / 64QAM 15 / 12 21754 / 21754 0.756 / 0.63028 16QAM / 64QAM 15 / 13 23370 / 26490 0.812 / 0.70829 16QAM / 64QAM 15 / 14 24222 / 32257 0.842 / 0.80130 16QAM / 64QAM 15 / 15 25558 / 38582 0.888 / 0.893

4.4. High Speed Downlink Shared Channel 99

20 15 10 5 0 5 10 15 20SINR [dB]

10-4

10-3

10-2

10-1

BER

CQI1 CQI30 CQI30

HS-DSCH - AWGN channel - Rake receiver

HS-DSCH release 5

HS-DSCH release 7

(a) Bit Error Rate

20 15 10 5 0 5 10 15 20SINR [dB]

10-3

10-2

10-1

100

BLE

R

CQI1 CQI30 CQI30

HS-DSCH - AWGN channel - Rake receiver

HS-DSCH release 5

HS-DSCH release 7

(b) Block Error Rate

Figure 4.8: Simulated HS-DSCH BER and BLER for an AWGN channel

10% is from -16 dB to 13 dB. For Release 7 UEs which support 64QAM theupper bound of reportable SINR values has been extended to approximately18 dB. As illustrated by the dashed lines a granularity of 2 dB is used forthe Release 7 AMC schemes which make use of 64QAM. Noticeable are themappings in the negative SINR scale which are possible due to the spreadinggain and very low coding rates (and throughputs).

As an analytical approximation of the simulated results Eq. (4.12) canbe used [53]. Eq. (4.12) has been adopted in several publications of systemlevel simulation results, e.g. [140]. Valid for an AWGN channel it allowsto translate a CQI value to the SINR which is required to achieve a certainBLER. A BLER of typically 10% is aimed for when the UE is reporting thechannel quality.

SINR =

√3− log10(CQI)

2· log10(BLER−0.7− 1) + 1.03 ·CQI− 17.3 (4.12)

In Figure 4.9 the mappings as calculated by Eq. (4.12) are shown. Com-pared to Figure 4.8 it can be observed that the analytical results closelymatch the simulated ones. Only small differences mostly related to codeblock segmentation effects for high CQI values may be noticed.

As a conclusion of the above comparison the implemented HS-DSCH


20 15 10 5 0 5 10 15 20SINR [dB]

10-3

10-2

10-1

100

BLE

R

HS-DSCH - AWGN channel - Analytical model

HS-DSCH release 5, analytical model

Figure 4.9: BLER mapping for an AWGN channel based on analytical model

coding chain can be considered to be validated. In contrast to the DCHthe transmission power control is not the main method used to adapt tochanging channel conditions. Instead the modulation, number of codes andchannel coding are adapted in order to transmit as much information aspossible by considering the instantaneous channel conditions and the UEcapabilities. As the channel assignment can be changed on a 2 ms basis theHS-DSCH is able to support highly variable PS traffic. Mostly because ofthe higher order modulation schemes the HS-DSCH reaches a significantlyhigher throughput than the DCH. While in a DCH configuration with opti-mal channel conditions up to 7 users of 403.2 kbit/s can be supported in onecell, the corresponding HS-DSCH data rate per user would be 1.83 Mbit/swith Release 5 and 2.76 Mbit/s with Release 7 if all 7 UEs report CQI 30 andget the same scheduling share. In the next section a comparison betweenthe HS-DSCH results and the performance of Mobile WiMAX is drawn.

4.5 Mobile WiMAX

In order to compare the link-level performance of the HS-DSCH with an-other state-of-the-art system the PHY coding chain of Mobile WiMAX hasbeen implemented and studied within this thesis as well. Mobile WiMAXhas become a member of the IMT-2000 family of mobile communicationsystems in the year 2007 and, therefore, it may be operated in the samefrequency bands as the former IMT-2000 systems. In detail the IMT-2000

4.5. Mobile WiMAX 101

extension bands at 2.496 to 2.69 GHz are of interest here. Based on the Or-thogonal Frequency Division Multiple Access (OFDMA) mode of 802.16e[74] a subset of features has been identified and standardized under the of-ficial name OFDMA TDD Wireless Metropolitan Area Network (WMAN).In early contributions to the approval process for the RadiocommunicationStandardization Sector of ITU (ITU-R) M.1457 recommendation [77] thisprofile was referred to as IP-OFDMA.

The major difference between Mobile WiMAX and the earlier membersof the IMT-2000 family is the OFDM modulation scheme, which will also beused for the 3GPP Long Term Evolution (LTE) systems. Because UMTSFDD typically uses a bandwidth of 5 MHz the same bandwidth for theMobile WiMAX system has been chosen in this study. For a 5 MHz config-uration Mobile WiMAX uses a FFT of size 512 and a sampling frequencyof 5.6 MHz as OFDM parameters. A fixed Cyclic Prefix (CP) duration of1/8 the OFDM symbol time is prepended to each symbol. In order to notexceed the bandwidth limitation of 5 MHz, guard carriers are introducedat the outer bins of the FFT. The number of guard carriers depends onthe transmission direction (uplink/downlink) and on the chosen subcarriermapping/permutation scheme. In the following the focus is on the downlinkPartial Usage of Subchannels (PUSC) mapping scheme which is mandatoryfor both 802.16e and Mobile WiMAX. At the left and right side of the FFT46 and 45 carriers, respectively, are left unused. Additionally to those 91guard carriers the DC subcarrier in the middle of the FFT does not transmitany information. From the remaining 420 subcarriers 60 subcarriers con-tain pilot symbols which are used for channel estimation. The remaining360 data subcarriers are used for data transmission to individual users.

Available modulation schemes for Mobile WiMAX are QPSK, 16QAMand 64QAM. Mobile WiMAX requires a convolutional code of rate 1/2 witha constraint length of 7 and a Convolutional Turbo Code (CTC) of rate 1/3with a constraint length of 4 as mandatory FEC codes. In Mobile WiMAXthere is, depending on the type of FEC, either a fixed puncturing patternor a symbol selection formula used to identify punctured and transmittedbits. Both FEC mechanisms lead to the available MCSs which can be foundin Table 4.5.

In Mobile WiMAX one block interleaver with 16 rows and the requirednumber of columns to interleave all bits from the puncturing unit is usedin case of convolutional coding. For CTC setups an interleaving mechanismbetween coding and puncturing takes place.


Table 4.5: Modulation and coding schemes of Mobile WiMAX

Modulation Code ratesQPSK 1/2 3/4

16QAM 1/2 3/464QAM 1/2 2/3 3/4 5/6

For the 2-dimensional modulation schemes 16QAM and 64QAM bothHSDPA as well as Mobile WiMAX include a mechanism to deal with theimperfectly Gray-coded symbol constellation. While in HSDPA the con-stellation rearrangement is used to modify the bit to symbol mapping forretransmissions caused by the HARQ, Mobile WiMAX uses bit interleavingto shuffle the bits of each symbol within an OFDM symbol in order to avoidlong runs of lowly reliable bits. In Mobile WiMAX a 16 bit CRC field isused in case HARQ is applied. A further optional CRC check exists in theMAC layer for every MAC PDU.

The BLER results for the 8 AMC schemes of Mobile WiMAX are shownin Figure 4.10. The difference of the gradients of the curves compared tothe HS-DSCH mappings in Figure 4.8 is caused by the convolutional coderof rate 1/2. The CTC results in gradients equal to those of the very similarturbo coder of UMTS. Nevertheless, the performance improvement of CTCover the convolutional coding is marginal for short code length [48].

10 5 0 5 10 15 20SINR [dB]

10-2

10-1

100

BLE

R

Mobile WiMAX - AWGN channel

Mobile WiMAX

Figure 4.10: Mobile WiMAX BLER mapping for an AWGN channel

4.5. Mobile WiMAX 103

4.5.1 Comparison of HSDPA and Mobile WiMAX Physical ChannelCapacity

Table 4.6 shows the achievable throughput for each modulation scheme un-der the assumption that all data subcarriers or channelization codes, avail-able for data transmission, are used. For the HSDPA the uncoded physicallayer throughput for 1 code and for the maximum number of codes, i.e. 15codes, are listed. Frequently, these maximum data rates are referred to whendetermining HSDPA capabilities. In a similar way the maximum throughputcontribution per OFDM symbol of Mobile WiMAX is shown. Additionally,the maximum theoretical throughput without coding for 44 OFDM sym-bols is depicted. Because Mobile WiMAX is specified for TDD mode alsothe uplink OFDM symbols are included as downlink symbols in order tomake a fair comparison. According to [139] the ratio of downlink/uplinkOFDM symbols may vary between 35/12 and 26/21. Furthermore, it isassumed that exactly 3 of the available 47 OFDM symbols are used for thepreamble (1 OFDM symbol) as well as the Frame Control Header (FCH)and all Medium Access Protocol (MAP) fields (2 OFDM symbols becauseof PUSC). The maximum throughput of the HS-DSCH using 64QAM isonly available with Release 7 of UMTS.

The comparison shows that both systems have a similar number of avail-able physical channel bits. The number of bits available for the HS-DSCHis slightly higher because less bits are used for management, e.g. no FCHor MAP fields. Furthermore, the pilot signals and the CP of the OFDMbased Mobile WiMAX require additional capacity compared to UMTS. InUMTS all management channels and all physical layer signals, e.g. CCPCHand CPICH, are contained in the remaining subtree of SF 16. In the nextsection more detailed results including the complete PHY coding chain andvarious receiver algorithms are presented.

Table 4.6: Available maximum data bits without coding [Mbit/s]

Modulation HS-PDSCH HS-PDSCH Mobile WiMAX (Mobile WiMAX)1 code 15 codes 1 symbol 44 symbols

QPSK 0.48 7.2 0.144 6.33616QAM 0.96 14.4 0.288 12.67264QAM 1.44 21.6 0.432 19.008


4.6 Throughput Comparison

In this section the BLER results from the previous sections are used tocalculate the achievable throughput by simply using the best MCS for agiven SINR. In order to really achieve the illustrated throughput an ac-curate channel measurement and CQI reporting are required. In the fol-lowing perfect channel knowledge is assumed. Therefore, the presented re-sults illustrate the upper bound of the achievable throughput. Because theAMC relies on delayed CQI reports the achieved throughput is typicallylower [53]. Especially in high mobility scenarios the throughput degrada-tion increases [93]. Nevertheless, proposals exist to improve the CQI accu-racy by modifications on either the network side [133] or the UE side [62].Simulation results including realistic CQI reporting are shown in Chapter 5.

4.6.1 AWGN Channel

4.6.1.1 Physical Layer Throughput

In Figure 4.11 the physical layer throughput, achievable on transport chan-nel level, is depicted. The left subfigure illustrates the HS-DSCH through-put for both Release 5 and Release 7. As a reference the Mobile WiMAXthroughput achieved with a fully used frame, i.e. uplink and downlinkperiods accumulated, is shown as well. Because variable block sizes in Mo-bile WiMAX affect the achievable throughput both the best case through-put (small blocks exactly allocating 2 OFDM symbols) and the worst casethroughput (one large convolutionally encoded block spanning all availableOFDM symbols) are shown.

It can be seen that the throughput over SINR on an AWGN channelis up to 3 dB away from the theoretical Shannon limit for a bandwidthof 5 MHz. Because the 5 MHz bandwidth of UMTS includes guard bandsand in some countries a carrier spacing smaller than 5 MHz is used thethroughput is often compared to a 3.84 MHz bandwidth which correspondsto the 3.84 Mcps of UMTS [71].

The SINR range in which the AMC of the HS-DSCH allows a throughputclose to the Shannon bound is around 30 dB. The 30 steps of the referenceMCSs have a higher distance to the bound when the modulation schemechanges. Especially when 64QAM in Release 7 is used the distance to thetheoretical limit for CQI 25 to CQI 27 is significantly higher compared tothe corresponding 16QAM MCSs. The maximum throughput of the HS-

4.6. Throughput Comparison 105

20 10 0 10 20 30SINR [dB]

102

103

104

Thro

ughput

[kbit

/s]

19291 kbps

12779 kbps

15840 kbps

HS-DSCH Throughput - AWGN channel - Rake receiver

HS-DSCH release 5

HS-DSCH release 7

Mobile WiMAX, small blocks

Mobile WiMAX, large blocks

Shannon bound


(a) HS-DSCH, Mobile WiMAX

20 10 0 10 20 30SINR [dB]

102

103

104

Thro

ughput

[kbit

/s]

67.2 kbps

134.4 kbps

403.2 kbps

806.4 kbps

403.2 kbps

806.4 kbps

67.2 kbps

134.4 kbps

403.2 kbps403.2 kbps

DCH Throughput - AWGN channel - Rake receiver

DCH Downlink, 20 ms TTI


DCH Uplink, 20 ms TTI


Shannon bound


(b) DCH

Figure 4.11: Physical layer throughput on an AWGN channel

DSCH is approximately 12.8 Mbit/s in Release 5 of the UMTS specificationand 19.3 Mbit/s with the 64QAM extension introduced in Release 7. Thecomparable Mobile WiMAX configuration reaches a peak throughput of15.8 Mbit/s which lies in between the throughputs of the two HS-DSCHresults. A more detailed comparison of HSDPA and Mobile WiMAX canbe found in [97].

The right subfigure of Figure 4.11 illustrates the DCH throughput ofthe configurations as introduced in Section 4.3. Both the throughput inuplink and downlink direction is presented. Within those ranges targetedby the inner loop power control, the curves have a distance to the Shannonbound comparable to the HS-DSCH results. The maximum DCH through-put achievable with one physical channel, i.e. one code, is 403.2 kbit/s usingthe standard TFSs. A multicode downlink configuration with two physicalchannels of SF 8 achieves a throughput of 806.4 kbit/s.

4.6.1.2 Maximum Throughput above RLC Layer

As a next step the maximum theoretical throughput of the RBs above theRLC layer is compared. This throughput is achieved under the assumptionthat the smallest possible PDU header of 16 bits is added by the RLC AMentities. As listed in Table 4.1 the DCH uses a TB size of 336 bits which leadsto an RLC SDU size of 320 bits. Therefore, a minimum layer 2 overhead


20 10 0 10 20 30SINR [dB]

102

103

104

Thro

ughput

[kbit

/s]

64 kbps

128 kbps

384 kbps

768 kbps

384 kbps

768 kbps

64 kbps

128 kbps

384 kbps384 kbps

18240 kbps

11200 kbps12160 kbps

11200 kbps

RLC Throughput - 336 bit MAC-d PDUs - AWGN channel - Rake receiver

HS-DSCH release 5

HS-DSCH release 7





Shannon bound


(a) 336 bit MAC-d PDUs

20 10 0 10 20 30SINR [dB]

102

103

104

Thro

ughput

[kbit

/s]

18560 kbps

12160 kbps

RLC Throughput - 656 bit MAC-d PDUs - AWGN channel - Rake receiver

HS-DSCH release 5

HS-DSCH release 7

Shannon bound


(b) 656 bit MAC-d PDUs

Figure 4.12: Maximum throughput on RLC level

of 4.8% for 320 bit RLC SDUs can be assumed if no MAC multiplexingis applied. For the HS-DSCH the alternative MAC-d PDU sizes of 336and 656 bits are foreseen [43] which result in RLC SDU size of 320 and640 bits, respectively. Even though a minimum theoretical RLC overheadof 2.4% could be achieved with 640 bit RLC SDUs, the overall layer 2overhead including MAC-hs headers and padding, caused by the MAC-hsPDU granularity, is between 4% and 5%, as can be seen by the followingHS-DSCH calculations.

Figure 4.12 shows the resulting throughputs for both 336 bit MAC-d PDUs (left subfigure) and 656 bit PDUs (right subfigure). Because ofthe 16 bit RLC header the maximum DCH throughput as shown beforeis reduced to 64 kbit/s, 128 kbit/s and 384 kbit/s above the RLC layer.When two channelization codes are used in the downlink the maximumRLC throughput reaches the nominal throughput of 768 kbit/s. Comparedto the 10 ms TTI configurations of the 384 kbit/s and 768 kbit/s RBs thealternative 20 ms configurations show more steps due to the larger TFS.

The throughput results of the HSDPA as measurable above the RLClayer differ from the previous HS-DSCH results as well. Especially for lowCQI values the RLC PDU granularity significantly influences the achievablethroughput. In order to transmit RLC PDUs of 336 bits and 656 bits theminimum required CQI is 5 and 8, respectively. In the 656 bit case less steps


in the throughput curve can be observed because of the PDU granularity.Because of an FDD HS-DSCH limitation of 70 PDUs per TTI the max-

imum throughput with 336 bit PDUs is 11.2 Mbit/s. Theoretically, athroughput of 12.2 Mbit/s with 16QAM and 18.2 Mbit/s using 64QAMwould be possible. When the MAC-d PDU size is increased to 656 bits thethroughput on RLC level reaches 12.2 Mbit/s and 18.6 Mbit/s for Release 5and Release 7, respectively.

Table 4.7 summarizes the achievable throughput for the 384 kbit/s DCHuplink and downlink as well as the HS-DSCH using QPSK, 16QAM and64QAM. The levels at which the throughputs are compared are the uncodedphysical channels, the transport channels (PHY throughput), the MAC-hsentity, the MAC-d entity and the RBs above the RLC AM entity. For thedual-code 768 kbit/s DCH downlink the 384 kbit/s throughput results needto be multiplied with 2. Comparing the efficiencies, i.e. the ratio of the RBthroughput to the throughput on physical channel level, it turns out thatthe Release 7 HS-DSCH has the highest efficiency of 0.86 while the DCHuplink only has an efficiency of 0.4. Nevertheless, it must be noted that amuch higher SINR is required by the HS-DSCH to achieve the correspondinghigh throughputs.

Table 4.7: Throughput comparison on various layers

DCH UL DCH DL HS-DSCH HS-DSCH HS-DSCHRelease 99 99 5 5 7UE category CAT 12 CAT 10 CAT 14Codes/SF 1/4 1/8 5/16 15/16 15/16Modulation BPSK QPSK QPSK 16QAM 64QAMMAC-d PDU size 336 336 336 656 336 656 336 656One code [kbit/s] 960 912 480 960 1440All codes [kbit/s] 960 912 2400 14400 21600PHY [kbit/s] 403.2 403.2 1659.5 12779 19291MAC-hs [kbit/s] - - 1649 12768.5 19280.5MAC-d [kbit/s] 403.2 403.2 1512 1640 11760 12464 11760 19024

(12768) (19152)RLC [kbit/s] 384 384 1440 1600 11200 12160 11200 18560

(12160) (18240)Efficiency 0.4 0.421 0.6 0.667 0.77 0.844 0.56 0.859

(0.844) (0.844)


4.6.2 Throughput for Pedestrian Channel Model

After the preceding comparisons which have been performed using a simpleAWGN channel, this and the following sections analyze the above configura-tions with Tapped Delay Line (TDL) based ITU channels [76]. In detail thechannel model ITU Pedestrian A (PA) at 3 km/h and the model ITU Vehic-ular A (VA) at 100 km/h have been used for the performance evaluation offading channels. As in the previous evaluations the co-channel interferenceof neighbour cells is considered as uncorrelated noise which contributes tothe overall interference level. Several receiver techniques have been appliedfor both channel models. For the WCDMA system a traditional Rake re-ceiver with Maximum Ratio Combining (MRC) of the individual paths hasbeen used. Furthermore, a ZF receiver, which tries to completely eliminateISI, and an MMSE receiver, which combats the ISI and takes the noise intoaccount, have been studied. For the reception in the OFDM based MobileWiMAX a simple equalization within the frequency domain is made. This isachieved by dividing the complex symbols of the subcarriers by the Fouriertransformation of the estimated channel impulse response.

In Figure 4.13 the throughput for a fading channel based on the PAmodel with low mobility, i.e. 3 km/h, is presented. The left subfigureillustrates the achievable throughput when a Rake receiver with 5 Rakefingers is used, the right subfigure shows the corresponding throughput ofan MMSE receiver. Similar to the AWGN results all curves are very closeto each other within the ranges where the curves are close to the Shannonbound. Nevertheless, the absolute distance to the theoretical bound is largerwith the fading channel.

It can be observed that the Rake receiver achieves a lower maximumthroughput compared to the AWGN results. Only 63% of the theoreticalRelease 5 and 34% of the Release 7 throughput is reached at maximum.Also in comparison to Mobile WiMAX the performance of the HS-DSCHwith Rake receiver is worse. As already discussed in [127] and [128] theRake receiver suffers in fading environments for higher order modulationslike 16QAM and 64QAM. In such scenarios receivers which try to eliminatethe ISI, e.g. ZF and MMSE, are superior. When using an MMSE receiverthe same maximum throughput as achieved for an AWGN channel can bereached (see right subfigure). Nevertheless, a higher SINR is required inthis case.

Different from the higher order modulation schemes, the receiver algo-


20 10 0 10 20 30 40SINR [dB]

101

102

103

104

Thro

ughput

[kbit

/s]

Throughput - ITU PA channel - Rake receiver

HS-DSCH release 5

HS-DSCH release 7

Mobile WiMAX





Shannon bound


(a) Rake receiver

20 10 0 10 20 30 40SINR [dB]

101

102

103

104

Thro

ughput

[kbit

/s]

Throughput - ITU PA channel - MMSE receiver (order 32)

HS-DSCH release 5

HS-DSCH release 7

Mobile WiMAX





Shannon bound


(b) MMSE receiver, order 32

Figure 4.13: Throughput for ITU PA channel, 3 km/h

rithms compared in Figure 4.13 do not differ under QPSK and BPSK modu-lation. Therefore, the throughputs of the DCH and QPSK based HS-DSCHconfigurations are comparable for Rake and MMSE receivers.

4.6.3 Throughput for Vehicular Channel Model

In Figure 4.14 the same setup as in Section 4.6.2 is shown for a fadingchannel with high velocity. Based on the VA channel profile at 100 km/hthe performance of an MMSE receiver of order 32 and order 128 as wellas the OFDM receiver of Mobile WiMAX is compared. As can be seen allthroughput results are worse compared to the PA channel results.

The receiver that suffers most in the vehicular scenario is the OFDMreceiver as implemented in this thesis. In comparison to the MMSE re-ceiver of UMTS its throughput is a little lower at SINR greater than 10 dB.Because of the high velocity the channel coefficients change very quickly.Since with the OFDM receiver the channel estimation and the equalizationare only performed one time for each OFDM symbol, the averaging errorof the estimated channel is increased because of the relatively long symboltime of the FFT of size 512 .

For SINR values below 10 dB the throughput of the UMTS MMSE re-ceiver is much closer to the Shannon bound than that of the OFDM WiMAXreceiver. In such operation areas, where only a low theoretical through-


20 10 0 10 20 30 40SINR [dB]

101

102

103

104

Thro

ughput

[kbit

/s]

Throughput - ITU VA channel - MMSE receiver (order 32)

HS-DSCH release 5

HS-DSCH release 7

Mobile WiMAX





Shannon bound


(a) MMSE receiver, order 32

20 10 0 10 20 30 40SINR [dB]

101

102

103

104

Thro

ughput

[kbit

/s]

Throughput - ITU VA channel - MMSE receiver (order 128)

HS-DSCH release 5

HS-DSCH release 7

Mobile WiMAX





Shannon bound


(b) MMSE receiver, order 128

Figure 4.14: Throughput for ITU VA channel, 100 km/h

put capacity is available, UMTS shows a more graceful degradation of thethroughput compared to WiMAX.

The HS-DSCH MMSE receiver in its current implementation can updatethe filter coefficients more frequently based on the reception of the CPICH.For such a receiver the order of the filter significantly influences the perfor-mance. For a fading environment as simulated in this scenario a higher filterorder improves the throughput for a given SINR significantly as shown forthe two receiver configurations. The MMSE receiver of order 32 has a lowermaximum throughput compared to Mobile WiMAX for both the Release 5and Release 7 HS-DSCH. When the filter length is increased to 128 bothHS-DSCH variants reach a higher throughput. The major reason why theMMSE receiver of order 128 is superior is because of its ability to bettercompensate ISI introduced by paths with a significantly larger delay.

Comparing the different receivers for the DCH it turns out that theMMSE receiver of order 32 has a small gain of about 1 dB in relation tothe Rake receiver. When enhancing the MMSE order to 128 no noticableadditional gain for the QPSK based DCH can be achieved.

Results show that the systems in the analyzed configuration have similarefficiencies with respect to the theoretical limit if state-of-the-art receiversare used. Depending on the scenario either HSDPA or Mobile WiMAX

4.7. Hybrid ARQ 111

achieve a higher throughput. The receivers used in this study show a greaterinfluence than the system itself. Especially the MMSE receivers showed tobe superior compared to basic Rake receivers or the channel equalization asused in the Mobile WiMAX configuration. For the higher order modulationschemes of the HS-DSCH the Rake receiver degredation in performance Re-garding the throughput in an AWGN environment both systems appear tobe close to an economically reasonable bound. The maximum throughputof Mobile WiMAX lies between the ones of HS-DSCH Release 5 and Re-lease 7. Compared to the DCH throughput of UMTS both Mobile WiMAXand HS-DSCH reach a significantly higher throughput. It is worth men-tioning that the maximum theoretical throughput is shown. In a real TDDbased Mobile WiMAX system only 35 out of the 47 OFDM symbols can beused for the downlink. Furthermore, at the time of writing many UMTSHSDPA mobiles support only 5 codes [28].

The performance of the link adaptation under real world conditions thatis based on signalling of the channel measurement results can be expectedto achieve a suboptimal link adaptation and, therefore, a lower throughputthan with perfect link adaptation as assumed in this section. Furthermore,HARQ mechanisms gain importance for improving performance under im-perfect channel knowledge at the transmitter. Last but not least the higherlayer functions, especially the scheduling algorithms, significantly influencethe performance of the overall system. In the following section the perfor-mance of HARQ mechanisms is analyzed. In Chapter 5 dynamic system-level simulations are presented that consider scheduling algorithms and bit-accurate signalling of the measurement reports from the UE to the Node B.

4.7 Hybrid ARQ

The previous sections dealt with the impact of the HS-DSCH AMC onthroughput. The focus is now extended to study its HARQ scheme. AMCand HARQ are the HS-DSCH features that correspond logically to the vari-able spreading factor and fast power control of the DCH as measures toadapt to the current channel needs. Both the variable spreading factor andthe fast power control belong to the most fundamental features of WCDMA[94]. The AMC is based on measurement reports, the CQI reports, whichbecause of their timing do not necessarily reflect the instantaneous channelconditions. Different from the fast power control used for the DCH, the


MCS of the HS-DSCH can less frequently be adapted to the SINR currentlyperceived at the UE receiver. Therefore, HARQ plays an important rolefor the HS-DSCH in order to improve the throughput efficiency within thelimitations present in UMTS. If the SINR of the channel is lower thancommunicated in the CQI report, HARQ allows to retransmit data withmodified redundancy information to increase the decoding probability ofa previous failed TB transmission. If the channel quality is higher thancommunicated in the latest CQI report, the MCS typically is chosen toopessimistic and radio resources are wasted. In order to reduce consequencesof these mismatches, the HS-DSCH is designed to target a BLER of 10%.

The HS-DSCH allows up to 8 transmissions of a given transport block.For each transmission several parameters can be changed to increase decod-ing probability. Two parameters affect the Rate Matching (RM) algorithm.The first one controls if systematic or non-systematic bits of the turbo codershould be prioritized. The second parameter is used to change the set ofbits which are punctured or repeated. For a TB transmission with a higheffective coding rate, up to three transmissions are needed until every bit ofthe coded bit sequence is transmitted at least once. Additionally to the ratematching the mapping of bits onto modulation symbols can be altered. Thisconstellation rearrangement is available for 16QAM and 64QAM (but notQPSK) where different decoding probabilities for the individual modulationsymbols exist (see Section 4.1). As shown in Table 2.2 up to four rear-rangement options are specified. The complete set of specified RedundancyVersion (RV) combinations is listed in Table 4.8.

Table 4.8: Allowed redundancy combinations

RV Prioritize systematic bits Redundancy bits ConstellationQPSK / QAM QPSK / QAM QAM

0 1 0 01 0 0 02 1 1 13 0 1 14 1 2 / 0 15 0 / 1 2 / 0 26 1 3 / 0 37 0 / 1 3 / 1 0

4.7. Hybrid ARQ 113

Two cases for the HARQ are to differ. The case where retransmissionscontain exactly the same bits as the initial transmission is called ChaseCombining (CC). The other case where RM parameters are modified forretransmitted TBs is called Incremental Redundancy (IR). Because of themodified RM parameters a different set of bits is transmitted there. IR isadvantageous with respect to the decoding probability, but a larger amountof soft memory is needed. As such memory increases the technical complex-ity of the UEs, already the initial technical 3GPP evaluations had a smallamount of soft bits as a design goal [38]. Because of this requirement notall UE categories support IR for high AMC schemes. Those categories usechase combining at higher data rates. Table 4.9 lists the UE capabilities upto HS-DSCH category 14 [28].

Table 4.9: HS-DSCH UE categories

Category Codes Modulation Min. interval TB size Total soft bits1 5 16QAM 3 7298 192002 5 16QAM 3 7298 288003 5 16QAM 2 7298 288004 5 16QAM 2 7298 384005 5 16QAM 1 7298 576006 5 16QAM 1 7298 672007 10 16QAM 1 14411 1152008 10 16QAM 1 14411 1344009 15 16QAM 1 20251 17280010 15 16QAM 1 27952 17280011 5 QPSK 2 3630 1440012 5 QPSK 1 3630 2880013 15 64QAM 1 35280 25920014 15 64QAM 1 42192 259200

The categories 1 to 12 have been introduced in Release 5 of the UMTSspecification. In Release 7 the 64QAM capable categories 13 and 14 havebeen added. Categories 10 and 14 are the most advanced ones in Table 4.9.All other categories use a reduced feature set only. In case the Node Breceives a CQI from a UE which maps to an AMC scheme that is not sup-ported by the mobile’s category the Node B selects the highest possible AMCscheme and may reduce the transmission power instead. Furthermore, the


AMC parameters themselves could be altered. For example a category 9UE would typically only receive either transmissions with up to 12 codes ortransmissions with a reduced coding rate because of its transport block sizelimitation. Therefore, the category 10 data rates for CQIs above 26 are notachievable by a category 9 UE. Additionally to the above differences theavailable soft memory and the intervals at which a mobile might be sched-uled vary within the different categories. As discussed in Section 2.3.1.3the size of the soft-bit buffer influences the HARQ gain. The lowest gain isachievable if only chase combining can be applied. Incremental redundancywith an unlimited soft memory leads to the optimum gain. In the followingsections these two extreme cases as well as one option to improve the chasecombining gain without the necessity of a large soft memory are analyzed.

4.7.1 Chase Combining and Incremental Redundancy

The decoding differences between chase combining and incremental redun-dancy are illustrated in Figure 4.15. The x-axis of the diagrams determinesthe SINR which is required to achieve a 10% BLER. The number of trans-missions is given by the y-axis. For every AMC scheme of Table 4.4 onecurve is shown. The flatter the curves are, i.e. the lower the required SINRfor an increasing number of transmissions, the higher the decoding gain is.As a reference the basic gain expected by the increase of mutual information[135] for transmission n

gbasic = 10 · log10(n) (4.13)

is illustrated by dotted lines at the left and right side of the simulated curves.Results for a UE of category 10 (solid lines) as well as a category 14 UE(dashed lines) are shown in the figure. In case of incremental redundancyall eight RVs are used in sequential order (compare Table 4.8). A detailedcomparison of different schemes for selecting the RVs can be found in [51].Link-level results as shown in the following are used as input parametersfor the development of empirical models. These models can be used insystem-level simulations. Such a model, which approximates the link-levelperformance of HARQ schemes by a low complexity formula, is presentedin [59], for example.

When chase combining is applied (see lines in Figure 4.15(a)) the achiev-able gain for the QPSK based modulation schemes exactly matches the the-oretical gain introduced in Eq. (4.13). For 16QAM and 64QAM the gain

4.7. Hybrid ARQ 115

20 10 0 10SINR [dB] for 10% BLER

1

2

3

4

5

6

7

8

Tra

nsm

issi

on

CQI1 CQI30 CQI30

HS-DSCH - CC - AWGN channel - Rake receiver

(a) Chase Combining


1

2

3

4

5

6

7

8

Tra

nsm

issi

on

CQI1 CQI30 CQI30

HS-DSCH - IR - AWGN channel - Rake receiver

(b) Incremental Redundancy

Figure 4.15: HARQ gain for Chase Combining and Incremental Redundancy

is smaller because of the higher order modulation schemes. Especially forlow coding rates, e.g. as used for the reference MCSs corresponding to CQI16 and CQI 26 for 16QAM and 64QAM, respectively, the theoretical basicgain is not reached. For the first retransmission the gain is reduced by up to0.45 dB. For the fourth transmission this loss reaches a maximum of 0.95 dBfor 16QAM and 1.3 dB for 64QAM.

Contrary to chase combining, incremental redundancy shows its decod-ing gain especially in the presence of high order modulation schemes andhigh coding rates. Figure 4.15(b) illustrates how the corresponding gainsreduce the required SINR in case of a retransmission. Already for the firstretransmission in case of CQI 30 a gain of 5.3 dB for 64QAM and 3.9 dB for16QAM compared to the gain calculated from Eq. (4.13) is reached. Similargains have been measured by [60] for a small subset of HSDPA MCSs.

4.7.2 Constellation Rearrangement

As a trade-off between the large soft memory required for incremental re-dundancy and the low gain of pure chase combining, a HARQ scheme usingonly constellation rearrangement can be used. For this HARQ scheme theparameter controlling the constellation is varied. The parameters affectingthe rate matching are kept constant and, therefore, the same informationbits are transfered in retransmitted TBs. Figure 4.16 illustrates the gains



1

2

3

4

5

6

7

8

Tra

nsm

issi

on

CQI1 CQI30 CQI30

HS-DSCH - CR - AWGN channel - Rake receiver

Figure 4.16: HARQ gain using Constellation Rearrangement only

achieved by this scheme. A gain of 1 dB is achieved for the first retransmis-sion if 16QAM is used as the modulation scheme. With 64QAM a highergain of up to 3 dB is reached.

4.7.3 Comparison

Figure 4.17 compares the gains for the schemes Chase Combining (CC),Constellation Rearrangement (CR) and Incremental Redundancy (IR) atthe first retransmission. In case of QPSK based transmissions, i.e. CQIvalues below 16, only IR brings an additional gain. The reachable gaindepends on the coding rate of the selected MCS. When bit repetition isapplied by the rate matching algorithm no big gain can be observed asall coded bits are transmitted already. For higher effective coding rates,e.g. in case of CQI 6, CQI 9 and the QPSK based schemes above CQI 11,incremental redundancy shows a significant gain (up to 1 dB for CQI 15).

In case of 16QAM and 64QAM CC has a negative gain, especially, forlow code rates. CR has a rather constant gain of around 1 dB for 16QAMand up to 3 dB for 64QAM. The highest gains are reached by incrementalredundancy. The code rate affects how big this gain is. For low code ratesas used for CQI 16, CQI 17 and the 64QAM based CQI 26 the gain of theCR scheme is a little higher than that of IR.

4.7. Hybrid ARQ 117

0 5 10 15 20 25 30CQI

1

0

1

2

3

4

5

6

Rela

tive g

ain

[dB

]

CC

CR

IR

QPSK 16QAM 64QAM

HS-DSCH - HARQ gain - AWGN channel - Rake receiver

Figure 4.17: Comparison of HARQ gains for first retransmission

Figure 4.18 shows the effective coding rate of the MCSs studied above.The effective coding rate for the HS-DSCH within this comparison is definedby

ECR =bTB

bPHY=

bTB

480 · bSymbol · ncodes − 24(4.14)

where bTB is the transport block size in bits, bSymbol is the number of bitsper modulation symbol and ncodes is the number of codes as introduced intable 4.4. The number of codes used for a certain MCS is annotated toeach column in the figure. Furthermore, the code rate is shown at which nobit repetition or puncturing is performed by the rate matching algorithm.Below this line repetition is needed, above puncturing is required. It can beseen that the effective coding rate is not strictly increasing with increasedCQI. Whenever more codes are added to a higher MCS, or a higher ordermodulation scheme is used, the coding rate must significantly be reducedin order to meet the SINR target. The reason for this adjustment is tocompensate for the changed spreading gain and modulation sensitivity.


0 5 10 15 20 25 30CQI

0.0

0.2

0.4

0.6

0.8

1.0

Eff

ect

ive c

odin

g r

ate

1

2

3

4

57

8 10

12

15

HS-DSCH coding rate - Category 10

No puncturing/repetition

(a) Category 10

0 5 10 15 20 25 30CQI

0.0

0.2

0.4

0.6

0.8

1.0

Eff

ect

ive c

odin

g r

ate

1

2

3

4

57

8 10

12

13

14

15

HS-DSCH coding rate - Category 14

No puncturing/repetition

(b) Category 14

Figure 4.18: Coding rates of reference AMC schemes

It can be concluded that the type of HARQ significantly impacts theperformance of the HS-DSCH in the presence of delayed and, therefore,inaccurate CQI reports. As the type of HARQ depends on the UE capabil-ities and the Node B algorithms, there is diversification freedom for bothUE and network vendors. In the comparison of HARQ schemes it has beenshown that CC has the lowest performance. For QPSK only the minimumexpected gain has been reached. In QAM based scenarios with low cod-ing rates an even lower gain was observed. IR, on the contrary, showed tohave the highest gain. Especially for high effective code rates and QAM anoticeable gain was achieved. CR demonstrated to be a good compromisebetween the simple CC and the soft memory demanding IR. A rather coderate independent gain which solely was based on the chosen QAM schemewas identified.

CHAPTER 5

Performance on System Level

Contents5.1 Related Work of the Author . . . . . . . . . . . . . . 120

5.2 Fast Scheduling for HSDPA . . . . . . . . . . . . . . . 122

5.2.1 Scheduling Strategies . . . . . . . . . . . . . . 124

5.2.2 Qualitative Comparison of Scheduling Metrics 127

5.3 HS-DSCH Performance for RT Services . . . . . . . . 129

5.3.1 MAC-d PDU Queueing Delay . . . . . . . . . 132

5.3.2 Inter-Scheduling Interval . . . . . . . . . . . . 133

5.4 HS-DSCH Performance for mixed Services . . . . . . 135

5.4.1 MAC-hs PDU Throughput . . . . . . . . . . . 136

5.4.2 Throughput of NRT Services . . . . . . . . . 137

5.4.3 Queueing Delay of RT Services . . . . . . . . 138

5.4.4 Conclusions . . . . . . . . . . . . . . . . . . . 140

Based on the link-level results of the previous chapter the performance ofUMTS is further analyzed from system-level perspective. In contrast to

link-level evaluations which only consider the radio link of a single UE theperformance for a group of users is studied in this chapter. Dynamic sim-ulations, as typically being used when evaluating the HSDPA performancefor multiple users [134], are utilized for this purpose. These simulationsare performed with the framework developed within the scope of this the-sis (compare Chapter 3) and include bit-accurate protocol implementationsas well as realistic and configurable algorithms. The goal of this chapteris to study the performance of UMTS in a realistic traffic scenario. Byexchanging the scheduling algorithms and varying the traffic mix the im-pact on perceived QoS parameters, e.g. throughput and queueing delay, isevaluated.

120 5. Performance on System Level

In Section 5.1 related evaluations which were performed with the help ofthe developed simulation framework and which were published by the au-thor are presented. After the previous chapter dealt with two of the threemost important HSDPA features, i.e. Adaptive Modulation and Coding(AMC) and Hybrid ARQ (HARQ), Section 5.2 introduces the concept offast scheduling [105]. First the general principles and basic simulation as-sumptions are presented. In a next step several proposed scheduling algo-rithms are qualitatively compared. The quantitative evaluation of a subsetof these algorithms is given in Sections 5.3 and 5.4. The first investigatedscenario is used to analyze the performance for Real Time (RT) services.In a second scenario a traffic mix of both RT and Non-Real Time (NRT)services is studied.

5.1 Related Work of the Author

On system level one of the key functionalities which affects the QoS per-ceived by the user is the scheduling of radio resources. Depending on thetransport channel, different scheduling concepts may be applied. For theDCH the scheduling options are limited as only traffic flows of one user arescheduled there. These traffic flows are scheduled strictly priority based.Only for flows within the same priority class underlying sub-scheduling algo-rithms are possible. The scheduling of dedicated channels has been studiedin [67] and [68].

For dedicated channels the resource partitioning between users is fixedfor a relatively long time. Every transport channel of each UE has an allo-cated TFS which is used by the MAC layer for scheduling purposes. Onlylayer 3 may change these allocations by assigning a new channelization codeand a new TFS. This transport channel reconfiguration requires RRC sig-nalling. Parallel downlink resources are occupied during a reconfigurationphase. In [106] an extended CAC scheme allowing the up- and downgradingof TFSs is proposed and evaluated. Within this scheme the users may havedifferent priorities and QoS profiles. By applying the proposed method aform of slow scheduling of radio resources between users is realized. Nev-ertheless, since this method of channelization code reallocation can be veryslow, e.g. in the range of 500 ms [94], it is obviously inefficient for burstyand low duty cycle data applications.

5.1. Related Work of the Author 121

Typically, PS traffic in mobile communication systems results from theTCP/IP traffic as used in the fixed-line Internet. The wireless link is usu-ally the bottle neck to which the congestion control algorithms of TCP tryto adapt the transmission rate to. From end-user perspective, delay andthroughput of the TCP/IP traffic are important QoS parameters. In [104]and [103] it has been shown by the author that the scheduling of dedi-cated channels in UMTS may negatively influence the traffic performanceof the TCP protocol. Depending on the scenario and scheduling algorithmthe bandwidth-delay product is not accurately estimated by the TCP and,therefore, the TCP throughput is reduced. Both TCP and IP add relativelylarge protocol headers to the user-data that is to be transmitted. Especiallywhen IPv6 is used or when the packet size of the service is small, e.g. in caseof VoIP, these headers substantially reduce the throughput experienced bythe user. In [91] it has been shown that the header compression algorithmsof the PDCP can significantly reduce the related header overhead.

The dedicated channels of UMTS can be used for both PS and CS ser-vices. Examples for CS services are ordinary voice calls [40] and VideoTelephony (VT) [41, 79]. In the latter case video is either encoded usingMPEG4 [75] or H.263 [78]. In [101] and [102] the performance of videotelephony as specified by the 3GPP is evaluated and proposals for improve-ments are given. By using the module presented in Section 3.6 the qualityof video transmitted over the emulated channel is evaluated using standard-ized metrics as specified in [80, 114, 115].

Another improvement for the UMTS radio interface with respect to itsefficiency is the use of MBMS. With MBMS content that is to be sent to agroup of users can be transmitted using point-to-multipoint channels. Theefficiency of such configurations has been studied in [95] and [100] by theauthor.


5.2 Fast Scheduling for HSDPA

The HSDPA is an extension to UMTS which has been introduced in Re-lease 5. In later releases it has been improved with the goal to achieve ahigher data rate. The HS-DSCH is the transport channel the HSDPA isbased on. The HSDPA concept has been designed to increase the downlinkpacket data throughput by means of fast physical layer retransmission andtransmission combining (HARQ) as well as fast scheduling and link adapta-tion (AMC), controlled by the Node B. AMC and HARQ have already beenevaluated in Section 4.4 and Section 4.7, respectively. In Section 4.6 themaximum theoretical throughput for one user depending on channel condi-tions was shown. Among the above features, the fast scheduling, taking intoaccount the current channel conditions of multiple users, contributes signif-icantly to the performance of HSDPA in terms of the so-called multiuserdiversity gain. In this section several scheduling algorithms are introducedand analyzed with respect to both the overall system performance and theQoS perceived by individual users.

The basic scheduling principle is illustrated in Figure 5.1 where thescheduler is located in the Node B rather than in the RNC, as typicalfor DCH scheduling. In this way, the delay introduced by the schedulingprocess is minimized. The radio channel condition is measured and reportedby means of the CQI by each mobile [20]. As the radio channel experiences

UE 2

Node B

UE 1

L1 feedback

Data

L1 feedbackData

Fast Node B scheduling based on:

Channel qualityUE capabilityQoS and priority

Figure 5.1: Elementary Node B scheduling principle

5.2. Fast Scheduling for HSDPA 123

small and large scale variations [129] the CQI reports allow the channelaware scheduling to favor UEs with temporarily good channel conditions.This exploitation of multiuser diversity can significantly increase the sys-tem capacity. On the other hand, the issues of fairness and guaranteedQoS among different users have to be considered by the scheduling algo-rithms. As no particular scheduling algorithm is specified for the HS-DSCHvarious alternative algorithms may be applied. Within this thesis severalalgorithms from literature (mostly not being intended to be used for theHS-DSCH, originally) have been analyzed with respect to their pros andcons in a realistically modeled HSDPA scenario.

The UMTS standard defines four QoS classes [5, 10], namely conversa-tional, streaming, interactive and background class. These classes mainlydiffer in the requirements on transmission delay and reliability in terms ofBLER. According to [39] HSDPA focuses on the last three classes. Whenclassifying services according to their QoS requirements the concept of RTand NRT services is useful. RT services impose strict delay requirementson the end-to-end communication. As a result the involved network nodessupporting RT traffic have to transfer the packets within a maximum tol-erable delay. Due to these severe delay constraints, the error correctionpossibilities for RT communication are very limited. On the other hand,NRT traffic is considered error sensitive, but has less demanding delay con-straints than RT traffic. These characteristics of NRT traffic allow for linkand also end-to-end level error recovery, enabling an error-free delivery ofuser data.

The QoS classes of UMTS can be grouped by these two categories. Con-versational and streaming traffic can be considered as RT services, whereasinteractive and background traffic belong to the NRT services. The maindifference between conversational and streaming traffic is the maximum tol-erable delay. Because of the bidirectional communication pattern the con-versational traffic is more delay sensitive than streaming traffic. Interactiveand background traffic do not have strict delay constraints and are onlydistinguished by the priority they are scheduled with. As this schedulingis strictly priority based the results are very deterministic and, therefore,not considered in the analyzed scenarios. The NRT services of all UEs areconfigured to be of the same priority. In the following we differ only be-tween NRT and RT traffic. NRT traffic requires user data to be transferrederror free, whereas delay requirements still allow end-to-end error recoveryas carried out by the TCP. In contrast, RT services have strict delay re-


quirements which exclude end-to-end retransmission protocols. Hence, anunreliable transport protocols like UDP is used.

The performance metrics for NRT services are mostly user and systemthroughput. For RT services the delay experienced by the MAC-d PDUsand the packet loss rate due to the discard timer [36] are the main evalua-tion metrics. In addition, the inter-scheduling interval and the Jain indexare measured to compare the fairness of different scheduling algorithms.The interval refers to the time period between two consecutive schedulingevents for an individual user. In the following sections the scheduling al-gorithms used within the performance evaluations are introduced and theirmain characteristics are qualitatively discussed.

5.2.1 Scheduling Strategies

5.2.1.1 Maximum SINR Scheduling

In every TTI the Maximum SINR (MaxSINR) scheduling algorithm servesthe user with best channel conditions and, therefore, with the highest instan-taneous supportable data rate. The serving principle obviously has benefitsin terms of cell throughput. Consequently, under idealized conditions it isthe system throughput optimal scheduler. Mathematically, it schedules userj that meets

j = arg maxi{Ri(t)} (5.1)

at time t where the time granularity of t is one TTI. Ri(t) is the instanta-neous data rate experienced by user i if being served by the packet scheduler.In general the term {·} represents the relative priority of user i. The maindisadvantage of the MaxSINR algorithm is its inherent unfairness. For in-stance, when a UE is far away from the Node B and its mobility is low itrarely is being scheduled [113].

5.2.1.2 Proportional Fair Scheduling

The Proportional Fair (PF) scheduling algorithm was initially proposed in[72] and further analyzed in [73] and [57]. According to [86] the PF schedulerserves the user j with best relative channel quality at time t that meets

j = arg maxi{Ri(t)λi(t)

}. (5.2)


Here Ri(t) is defined as above and λi(t) is the average data rate for user imeasured over the past. This rule ranks the users according to their instan-taneous channel quality relative to their own average channel conditions.Accordingly, users with a higher average throughput are not necessarilyprivileged. In this way not only the multiuser diversity can be exploited,but at the same time also the issue of fairness among the users is taken intoaccount.

Up to now the presented scheduling algorithms do not take into accountthe delay experienced by each individual user and, therefore, are not suitablefor scheduling RT services. In order to meet delay requirements, several QoSoriented scheduling algorithms have been proposed of which some relevantexamples are introduced below.

5.2.1.3 Modified Largest Weighted Delay First Scheduling

Modified Largest Weighted Delay First (M-LWDF) [45] is an algorithm tokeep the probability of delayed packets exceeding the discard bound belowthe maximum allowed SDU error ratio

Pr(Di > Ti) ≤ δi , (5.3)

where Di represents the Head Of Line (HOL) packet delay of user i, Ti thedelay bound and δi is the allowed percentage of discarded packets. TheM-LWDF scheduler selects at time t user j that meets

j = arg maxi{ai ·

Ri(t)

λi(t)·Di(t)} , (5.4)

where term ai is a constant used for QoS differentiation. Consequently,varying services can have a different δ so that the priority between userswith different demands in terms of error rate can be adjusted. Accordingto a suggestion in [132] an approximating practical rule for choosing ai is

ai = − log(δi)Ti

. The term Ri(t)λi(t)

is derived from the PF algorithm and Di(t)

is the HOL packet delay.By combining the PF metric and the HOL delay this algorithm not only

takes advantage of the multiuser diversity available in the shared channelthrough the PF algorithm, but also increases the weight of flows with HOLpackets that are close to their deadline violation. The value of the HOLdelay Di(t) has a significant impact on the total scheduling priority. SDUs


arriving in an empty queue initially have the HOL delay Di(t) = 0 and,therefore, have to wait for the increase of their priority. This intrinsic delayis experienced by each SDU, i.e. MAC-d PDU in HSDPA.

5.2.1.4 Exponential Rule Scheduling

The Exponential Rule (ER) scheduling relies on the PF algorithm for regularsituations and equalizes the weighted delays of the queues of all flows if theirdifferences are large. According to [126] this scheme schedules the user j attime t that meets

j = arg maxi{γiRi(t)e

aiDi(t)−aD

1+√

aD } (5.5)

where

aD =1

N

N∑i=1

aiDi(t) (5.6)

and γi > 0 (i = 1,..,N) are set in the form γi = ai/λi(t) to represent thetrade-off between the QoS requirements and the proportionally fair metric.

Similar as for the M-LWDF algorithm ai can be set to ai = −log(δi)Ti

.

5.2.1.5 Channel-Dependent Earliest Due Date Scheduling

The Channel-Dependent Earliest Due Date (CD-EDD) scheduling algorithm[92] extends the PF algorithm with an Earliest Due Date (EDD) compo-nent. The EDD functionality is realized by assigning each packet of user i adelivery deadline Ti. The user j selected by the CD-EDD scheduler at timet meets

j = arg maxi{ai ·

Ri(t)

λi(t)· Di(t)

Ti −Di(t)} , (5.7)

where ai, Ri(t), λi and Di(t) are defined as for M-LWDF. The term Ti −Di(t) is the time until the deadline is reached. As the delay of the HOLpacket gets closer to Ti the EDD term dominates Eq. (5.7). If the HOLpacket delay of user i is low the EDD term gives the flow a rather lowpriority.


5.2.1.6 Expo-Linear Scheduling

In order to avoid the intrinsic delay in M-LWDF and CD-EDD some otheralgorithms have been proposed. One example is the Expo-Linear (EL) al-gorithm proposed in [64]. It schedules user j at time t meeting

j = arg maxi{ai ·

Ri(t)

λi(t)· eaiDi(t)} , (5.8)

where ai, Ri(t), λi(t) and Di(t) are defined as for M-LWDF. EL introducesan exponential term to better equalize the weighted delays. If the HOLdelay of user i is low, the PF metric dominates the scheduling decision. Incase the HOL delay approaches the delay bound, the total priority increasesin an exponential manner.

EL can be seen as a modification to the ER scheduler. The differencesto the ER scheduler are that the mean value in the numerator and thecomplete denominator of the exponent are removed1. The removal of thedenominator has two advantages. Firstly, it allows the algorithm to increasethe total throughput when no user’s QoS is jeopardized. Secondly, it allowsfor a more direct influence by the service whose HOL packet delay is ap-proaching its deadline since this influence is not anymore affected by thecondition of the HOL packets of the other users.

5.2.2 Qualitative Comparison of Scheduling Metrics

Common to all algorithms introduced in the previous sections is their ex-ploitation of the multiuser diversity gain. The priority term in all schedulingformulas is linearly dependent on the instantaneous data rate Ri which issupported by the current channel conditions.

Qualitative differences between the algorithms can be observed with re-spect to the fairness between users and their tendency to maintain QoSconstraints. In Figure 5.2 the relative scheduling priorities of user i basedon the average user throughput λi and the HOL packet delay Di are illus-trated.

1According to [125] the reason for the removal of the mean value is that the mean valuein the numerator of the exponent is introduced only for better understanding of thecriterion’s functionality and the removal of this term does not change the functionalityof the algorithm.


0 5 10 15 20λi

0.5

0.0

0.5

1.0

1.5

Pri

ori

ty

MaxSINR

PF, M-LWDF, CD-EDD, ER, EL

Priority vs. average throughput

(a) Average throughput

0 5 10 15 20Di

0.5

1.0

1.5

2.0

2.5

Pri

ori

ty

CD-EDDER, EL

M-LWDF

MaxSINR, PF

Priority vs. HOL packet delay

(b) HOL packet delay

Figure 5.2: Qualitative comparison of scheduling metrics

As shown in Figure 5.2(a) most of the algorithms reduce the schedulingpriority of users which already observed a high average throughput in thepast. Only the MaxSINR algorithm does not take this information intoaccount and, therefore, does not schedule the users in a fair manner.

The dependency of the priority on the HOL packet delay is depicted inFigure 5.2(b). While the M-LWDF algorithm linearly increases the priorityof users with packets close to the deadline violation, the CD-EDD, EL andER algorithms increase the priority exponentially. As visualized by thehorizontal line, the other algorithms are not QoS aware at all.

Because of the benefits identified in the qualitative analysis in Sec-tion 5.2.1 the EL algorithm is used in the following as a candidate schedulingalgorithm which takes delay constraints into account. Simulation results ofthe other QoS aware algorithms can be found in [99].

5.3. HS-DSCH Performance for RT Services 129

5.3 HS-DSCH Performance for RT Services

The system used for the realistic performance evaluation of scheduling algo-rithms is modeled in this thesis with the relevant MAC-hs protocol, a phys-ical layer and a traffic generator. The TTI equals 2 ms and one HS-SCCH isconfigured in the simulation studies. Hence, only one of the configured UEscan get resources in each TTI. The mobility of the users is assumed to benormally distributed with a mean speed of 3 m/s and a variance equal to 1.All users are confined to a fixed size area and move in a Brownian motionmanner. The movement area is chosen in a way so that the UEs stay withinthe area of the evaluated cell. Furthermore, the configured serving cell, i.e.the cell the UEs receive the HS-DSCH from, is kept the same. Only SHOin terms of radio link addition for the dedicated resources of the UEs maytake place at the cell border. Further influences from the neighbouring cellsare limited by configuring a constant power allocation for the cells whichare not evaluated.

In the simulation studies a traffic generator is employed which createsMAC-d PDUs whose size equals to 336 bits. For the NRT services a full-queue model is applied which keeps the Buffer Occupancy (BO) of the cor-responding MAC-hs priority queues in the Node B at a constant level. Ac-cordingly, there is always sufficient data available for transmission to eachmobile. By doing so, effects originating from the Iub flow control betweenNode B and RNC [110] do not disturb the simulation of the radio interface.Further proposals for limiting flow control effects on HSDPA system perfor-mance are given in [111] and [47]. The RT services are configured to use aconstant data rate model which is realized by creating the same amount ofMAC-d PDUs periodically, e.g. every 5 TTIs.

In the following sections the basic behavioral differences of the PF andEL algorithms for RT services are shown. Relevant details of the simulationscenario are listed in Table 5.1. The simulation parameters have been se-lected as frequently being used in literature, see e.g. [136], as well as beingused in deployed networks [88].

The 9 UEs assumed in this scenario are separated in 3 groups withdifferent path loss profiles characterized by their mean perceived CQI values(see Figure 5.3 and Table 5.2). The CQI distribution for each profile isshown in Figure 5.4. All UEs have configured a constant-rate data sourcegenerating 3 MAC-d PDUs every 5 TTIs, i.e. 10 ms. Consequently, thesource data rate per UE is 100.8 kbit/s as 336 bit MAC-d PDUs are to be


Table 5.1: Scenario details and configuration parameters

Parameter ValueSimulation time 10000 sNumber of evaluated UEs 9Neighbour cells 1 ring, same frequencyCell antenna type OmnidirectionalPath loss Variable, groupedTraffic model Constant, 100.8 kbit/sMAC-d PDU arrival interval 5 TTIs (10 ms)Number of HS-PDSCH codes 5Number of HS-SCCH codes 1UE category 6Target BLER 10%Maximum number of retransmissions 4Transmit window size 12Receive window size 12Release timer 140 msMaximum delay 400 msCQI feedback cycle 2 msCQI repetition factor 1Filter length 50 TTIsδ (allowed fraction of discarded packets) 0.01Throughput measurement interval 50 TTIs

transmitted. By doing so a streaming service with a data rate of 100.8 kbit/sis simulated. The maximum tolerable delay is set to 400 ms which meansthat MAC-d PDUs experiencing a longer queueing delay will be discarded.

Table 5.2: Path loss in different UE groups

UE 1-3 UE 4-6 UE 7-9Path loss High Average LowMean CQI 11 15 19


UE 1-3

UE 7-9

UE 4-6

Figure 5.3: RT service scenario with various path loss profiles

0 5 10 15 20 25 30CQI

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

Frequency

UE 1-3

UE 4-6

UE 7-9

UE 1-3, highpath lossUE 4-6, averagepath loss

UE 7-9, lowpath loss

Figure 5.4: CQI distribution


5.3.1 MAC-d PDU Queueing Delay

As stated earlier, for RT services the packet delay must be limited to somespecified value, otherwise packets are discarded. Hence, packets from allusers, even those experiencing bad channel conditions, should be transmit-ted within the 400 ms delay bound. Comparing the Complementary Cumu-lative Distribution Function (CCDF) of PF and EL algorithm in Figure 5.5it turns out that EL achieves queueing delays that for the different pathloss profiles are much closer to each other than with PF. UE 1-6 suffer arather long MAC-d PDU queueing delay while UE 7-9 experience a rela-tively short delay under PF scheduling. The EL algorithm is much moreQoS aware than PF, since the variance of the delay across user groups ismuch lower.

0.0 0.1 0.2 0.3 0.4 0.5Delay [s]

0.0

0.2

0.4

0.6

0.8

1.0

CC

DF

Proportional Fair (PF) - MAC-d PDU delay

UE 1-3

UE 4-6

UE 7-9

(a) Proportional Fair (PF)

0.0 0.1 0.2 0.3 0.4 0.5Delay [s]

0.0

0.2

0.4

0.6

0.8

1.0

CC

DF

Expo-Linear (EL) - MAC-d PDU delay

UE 1-3

UE 4-6

UE 7-9

(b) Expo-Linear (EL)

Figure 5.5: MAC-d PDU queueing delay

The mean delay and packet loss probability per user group under PF andEL scheduling are given in Table 5.3. When comparing the mean queueingdelay of the PF and EL algorithms it can be seen that for PF delay is not aconcern. Moreover, there is a high packet loss under PF for UE 1-6 due tothe expiration of the discard timer. In contrast, the EL scheduler providesfor most of the UEs a better QoS in terms of delay. With this schedulerUE 4-9 have almost no packet loss during the whole simulation time. Thepacket loss of UE 1-3 is significantly reduced compared to the PF scheduler.


Table 5.3: Mean queueing delay and packet loss

Algorithm Mean queueing delay [ms] Discarded packetsUE 1-3 UE 4-6 UE 7-9 UE 1-3 UE 4-6 UE 7-9

PF 385.2 224.8 21.7 52.3% 12.0% 0.2%EL 275.1 140.4 44.2 5.4% 0.2% 0.0%

Despite this relatively low packet loss for the UEs at the cell border the ELscheduler allows to meet the delay requirements of the RT service.

5.3.2 Inter-Scheduling Interval

Besides delays caused by retransmitted packets the user queueing delaylargely depends on the inter-scheduling interval which is the time span be-tween successive resource allocations to a UE by the scheduler. As shownin Figure 5.6 the inter-scheduling interval gradient is quite different for thePF and EL algorithm. The PF scheduler selects users mainly based onthe channel quality. Therefore, UE 7-9 which experience the best channelconditions get the shortest inter-scheduling interval. In contrast the ELalgorithm tries to reduce delays of distant UEs at the expense of close-byusers as it considers the 400 ms deadline of the UEs. It can be observedthat under EL the UEs with the worst channel condition (UE 1-3) get theshortest inter-scheduling interval. In other words, they are scheduled mostfrequently. These UEs in principle suffer the longest MAC-d PDU queueingdelay due to the PF metric in Eq. (5.8). However, when the actual MAC-dPDU delay of a UE increases, its priority grows because the exponentialterm becomes dominant, then. Therefore, UEs with best channel qualityare scheduled under EL less frequent as visible from Figure 5.6. It can beconcluded that the PF scheduler is not suited to schedule delay sensitiveservices.


0.00 0.02 0.04 0.06 0.08 0.10Inter-scheduling interval [s]

0.0

0.2

0.4

0.6

0.8

1.0

CC

DF

Proportional Fair (PF) - Inter-scheduling interval

UE 1-3

UE 4-6

UE 7-9

(a) Proportional Fair (PF)

0.00 0.02 0.04 0.06 0.08 0.10Inter-scheduling interval [s]

0.0

0.2

0.4

0.6

0.8

1.0C

CD

F

Expo-Linear (EL) - Inter-scheduling interval

UE 1-3

UE 4-6

UE 7-9

(b) Expo-Linear (EL)

Figure 5.6: MAC-hs inter-scheduling interval

5.4. HS-DSCH Performance for mixed Services 135

5.4 HS-DSCH Performance for mixed Services

In this section the performance of the PF and EL algorithms with a mix-ture of RT and NRT users is compared. The differences in the simulationassumptions, compared to the previous scenario in Table 5.1 and Table 5.2,are listed in Table 5.4. Now the 9 users are categorized by group specifictraffic models. The first group (UE 1-3) is loaded by a full queue model tosimulate NRT services. The other groups are loaded with constant data rateservices. For UE 4-6 two MAC-d PDUs with a size of 336 bits are generatedevery 5 TTIs (10 ms). UE 7-9 are loaded by three instead of two PDUswhich are generated every 10 ms. Different from the previous scenario theaverage path loss is now the same for all UEs. As illustrated in Figure 5.7all UEs are allowed to move within the complete area of the center cell.

Table 5.4: Traffic parameters for mixed service scenario

UE 1-3 UE 4-6 UE 7-9Traffic model Full queue CBR, 67.2 kbit/s CBR, 100.8 kbit/sDelay [ms] 5000 1000 400

UE 1-3 (NRT)

UE 7-9 (RT2)

UE 4-6 (RT1)

Figure 5.7: Simulation scenario with mixed services


5.4.1 MAC-hs PDU Throughput

Since the number of HS-SCCH codes is set to 1 the MAC-hs PDU through-put can be considered as the aggregate cell throughput in this scenario.Figure 5.8 shows this throughput measured before and after the reorderingbuffer (see Section 2.2.5.4). The throughput before the reordering buffer iscalled HARQ throughput, the throughput of the reordered PDUs is referredto as buffer throughput. As the PDU throughput is measured as the quotientof the PDU size and the transmission time the buffer throughput is typicallylower because of the queueing delay introduced by the reordering buffer. Itcan be observed in general that the cell throughput of the PF scheduler ishigher than that of the EL scheduler. The corresponding mean values arelisted in Table 5.5. The MaxSINR scheduler which always schedules the UEwith the best channel conditions achieves the highest throughput.

0 500 1000 1500 2000 2500 3000 3500PDU throughput [kbit/s]

0.0

0.2

0.4

0.6

0.8

1.0

CC

DF

EL

PF

MaxSINR

PDU throughput

PF HARQ

PF Buffer

EL HARQ

EL Buffer

MaxSINR HARQ

MaxSINR Buffer

Figure 5.8: MAC-hs PDU throughput with mixed services

In order to compare the present results with those shown in Section 4.6it must be noted that category 6 UEs are studied here. In contrast tocategory 10 UEs, as shown in Table 4.7, only 5 of the 15 codes can be


Table 5.5: Mean cell throughput

Algorithm HARQ throughput Buffer throughput Jain’s fairness[kbit/s] [kbit/s] index

MaxSINR 1905 1534 0.502PF 859 768 0.971EL 695 550 0.848

used. The 4.8 Mbit/s of 5 16QAM codes results in a coded physical layerthroughput of 3.58 Mbit/s. This throughput is further reduced by MAC-hs,MAC-d and RLC headers as well as the overhead introduced by the MAC-hsPDU granularity (compare Section 4.6.1.2).

Additionally to the mean throughput, Table 5.5 compares the schedulingalgorithms with respect to Jain’s fairness index [85]. This index is definedby Raj Jain’s equation

J(x1, x2, . . . , xn) =(∑ni=1 xi)

2

n ·∑ni=1 x

2i

(5.9)

where xi is the average throughput of user i and n is the number of users.In case the available resources are evenly shared among the users the indexreaches its maximum value J = 1. In the worst case only one user isscheduled and J calculates to be 1

n . For the given comparison the meanthroughput per group is evaluated and, therefore, the minimum of J is 1

3 .As fairness between users is not considered at all by the MaxSINR met-

ric, the lowest index of around 0.5 can be observed with this algorithm.The PF algorithm applies a fairness metric in addition to the channel qual-ity based scheduling. Therefore, a high fairness index of 0.97 is achieved.Nevertheless, QoS constraints are not taken into account by the PF prioritycalculation. When considering the delay requirements of the users by usingthe EL algorithm the fairness index is reduced to 0.85.

5.4.2 Throughput of NRT Services

For the NRT UEs 1-3 the throughput is an important factor for the evalua-tion. By looking into Figure 5.9 it turns out that the PF scheduler provideshigher throughput for the NRT users compared to the EL scheduler. Themean UE throughput with the PF algorithm is 129 kbit/s while the mean


UE throughput with the EL scheduler is only 46 kbit/s. This results fromthe fact that the EL scheduler, besides throughput, also considers delay anddue dates of UEs.

0 50 100 150 200 250 300 350 400Throughput [kbit/s]

0.0

0.2

0.4

0.6

0.8

1.0

CC

DF

EL PF

UE throughput

PF HARQ

PF Buffer

EL HARQ

EL Buffer

Figure 5.9: MAC-hs UE throughput of NRT service

5.4.3 Queueing Delay of RT Services

On the contrary, as shown in Figure 5.10, the EL scheduler provides asmaller delay for the RT users whose services are delay sensitive. The meanMAC-d PDU queueing delay and the number of transmitted MAC-d PDUsper second on UTRAN side are listed in Table 5.6. It can be observed fromthese results that there are packet losses due to discard timer expirationwhen the PF scheduler is applied. Compared with PF the EL schedulerensures that there is no packet loss due to the delay bounds for the delaysensitive services. Furthermore, the most delay sensitive services (UE 7-9)get a lower MAC-d PDU queueing delay than the less delay sensitive UEs4-6, which indicates that they are treated preferentially during scheduling.The MaxSINR scheduler is not useful for RT services because it neither


takes delay bounds into account nor it generates a fair share of the availableresources.

10-3 10-2 10-1 100

MAC-D PDU delay [s]

0.0

0.2

0.4

0.6

0.8

1.0

CC

DF

PF UE 4-6 PF UE 7-9

EL UE 4-6EL UE 7-9

MAC-d PDU delay

PF UE 4-6

PF UE 7-9

EL UE 4-6

EL UE 7-9

Figure 5.10: MAC-d PDU queueing delay of RT users

Table 5.6: MAC-d PDU delivery statistics

Algorithm Buffer delay [ms] PDUs per second Discarded packetsUE 4-6 UE 7-9 UE 4-6 UE 7-9 UE 4-6 UE 7-9

MaxSINR 98.8 22.0 82 110 59.2% 63.4%PF 74.8 80.0 197 279 1.7% 7.0%EL 76.9 17.5 200 300 0% 0%


5.4.4 Conclusions

The simulation results show that the EL scheduler behaves similar to the PFscheduler when applied to NRT services. Nevertheless, the NRT throughputis lower for the EL scheduler as RT traffic needs to be prioritized. Further-more, it turned out that the PF scheduler is not suited for RT services.The delay requirements of RT users are not taken into account by the PFscheduler. Consequently, there is a severe packet loss when the PF algo-rithm is employed for RT services. In contrast to the PF scheduler the ELscheduler calculates the user priority not only based on the PF metric butalso based on the delay bound. Therefore, it is able to meet the differentQoS requirements of RT users.

Considering a mixture of RT and NRT services, there is a trade-off be-tween the throughput of NRT users and the delay requirement of RT users.The PF scheduler outperforms the EL scheduler with a higher aggregatethroughput. However, it can not guarantee the delay requirement of RTusers. The EL scheduler provides a relatively low cell throughput but itmeets the delay requirement of RT users. Hence, the EL scheduler is a bet-ter option for supporting a mixture of NRT and RT services. The highestaggregate throughput is achieved by the MaxSINR scheduler. Nevertheless,neither the QoS constraints are met nor fair service is given to the users bythis scheduler.

CHAPTER 6

Conclusion and Outlook

W ithin the scope of this thesis a detailed simulation framework has beendeveloped allowing the bit-exact simulation of UMTS on both link-

level and protocol-level. Based on this simulation environment extensivesimulation studies of various UMTS configurations have been carried out.Both the simulation framework with its models as well as the simulationconfiguration and the corresponding results have been published [98, 99].In general, configurations using the DCH and the HS-DSCH have beenanalyzed with respect to their ability to deliver PS based services to theUEs. As a reference to another state-of-the-art system simulation resultsfor Mobile WiMAX are presented as well.

It has been shown that all considered systems have similar physicallayer efficiencies with respect to theoretical limits. The compared receiveralgorithms introduce higher deviations on system and individual UE per-formance than the type of system itself. In the scenarios with fast fad-ing channels, for example, the HSDPA using a modern Minimum MeanSquare Error (MMSE) receiver performs better than the OFDM based Mo-bile WiMAX in the studied configuration. For the QPSK based DCH thereceiver algorithm showed only a minor impact. A simple Rake receiverperforms sufficiently well even in fading environments.

Larger deviations between the evaluated system configurations exist withrespect to Adaptive Modulation and Coding (AMC). The DCH does notmake use of AMC and, therefore, only a very small Spreading Factor (SF)specific working area in terms of required SINR exists. Fast power control isused to stay within this area. In comparison to Mobile WiMAX the HSDPAmay adapt its Modulation Coding Scheme (MCS) within a large dynamicrange and with a fine granularity of 1 dB step size. The fixed size RLCpayload, however, introduces padding overhead and shrinks the dynamicrange.

The DCH typically operates with a relatively low throughput, i.e. 384kbit/s peak throughput without multi-code transmission and 768 kbit/s

142 6. Conclusion and Outlook

in case multi-code transmission is supported by both UE and network.The Transport Format Set (TFS) is configured according to the maximumthroughput and only a relatively slow reconfiguration can take place. If lessresources are needed by one UE within a Transmission Time Interval (TTI)the other UEs do not directly benefit as the resources can not be reassignedfast enough. Only indirectly other UEs gain by the reduced interference.Therefore, the DCH is better suited for CS and guaranteed bit rate servicesas the ’bundling gain’ for PS services is rather limited.

The achievable peak throughput of the HSDPA depends on the UE’sHS-DSCH category. Category 12 UEs which are only able to use QPSKmay reach a peak throughput of 1.6 Mbit/s above RLC. In Release 5 amaximum throughput of 12.16 Mbit/s is achieved by HS-DSCH category 10.With the introduction of 64QAM in Release 7 a comparable throughput of18.56 Mbit/s can theoretically be reached with category 14. As shownby this thesis, Mobile WiMAX in the comparable reference configurationachieves a throughput in between that of Release 5 and Release 7 HSDPA.

For more realistic throughput figures multiuser scenarios must be takeninto account. As evidenced by dynamic system-level simulations the schedul-ing algorithm and the traffic mix have significant impact on the overall sys-tem performance and the individual perceived QoS. Results illustrate thetrade-off between the conflicting targets of high cell throughput and thedelay requirements for real-time services. If real-time requirements are in-troduced the multiuser diversity gain of the fast scheduling is significantlyreduced. Furthermore, fairness requirements between users reduce the over-all system throughput. As an example, the average cell throughput in ascenario with HS-DSCH category 6 UEs, i.e. at maximum 5 codes are used,is reduced from 1.5 Mbit/s to 0.77 Mbit/s when fairness between users isachieved by applying the proportional fair metric. If additionally the delayrequirements for a traffic mix of services with and without real time require-ments are considered the Expo-Linear (EL) scheduling algorithm furtherreduces the mean cell throughput to 0.55 Mbit/s.

As no inner-loop power control is used for the HSDPA, HARQ playsan important role for the efficiency of the system due to the variabilityof the radio channel. Especially in fading environments HARQ signifi-cantly contributes to the performance of the system. Within this thesisthe performance of HARQ schemes is compared with respect to their per-formance and technical complexities. For high code rates the IncrementalRedundancy (IR) scheme has the highest performance but also the high-

143

est complexity with respect to the amount of required soft memory. Up to4 dB gain compared to Chase Combining (CC) can be achieved for the firstretransmission. A more economic HARQ variant is achieved by using theConstellation Rearrangement (CR). Up to 1 dB gain for 16QAM and 3 dBgain for 64QAM can be achieved compared to plain CC. A disadvantageof the HARQ scheme in UMTS is that retransmissions must reuse the ini-tially selected MCS. Even if only a small amount of mutual information ismissing for the correct decoding, the same radio resources as for the initialtransmission have to be used.

Based on the presented performance results future research may focus onRelease 8 of the 3GPP specification. With Release 8 even higher through-puts can be reached by using dual cell HSDPA, i.e. two 5 MHz carriers areused in parallel, and Multiple Input Multiple Output (MIMO) with 64QAM.Drawbacks which have been identified within this thesis, e.g. fixed RLCPDU sizes, have been resolved by this release as well.

A further technology introduced with Release 8 is the so-called LongTerm Evolution (LTE). LTE introduces both a new radio access network,the Evolved UTRAN (E-UTRAN), as well as a new core network calledEvolved Packet System (EPS). In LTE the primary channel for the provisionof all kinds of services is the downlink shared channel which comprises thethree main features evaluated within this thesis. Similar to the HSDPA,AMC based on CQI reports and asynchronous downlink HARQ is applied.As no dedicated traffic channels are available, the fast scheduling, whoserate is increased to the 1 ms TTI, is the only method to provide QoS andfairness between users.

It is expected that the Release 8 HSDPA and LTE exist in parallel fora while. Therefore, a performance comparison of both could be a valuableresearch topic.

144 6. Conclusion and Outlook

APPENDIX A

Additional MAC Entities

ContentsA.1 System Broadcast by the MAC-b Entity . . . . . . . . 145

A.2 Entities for Common, Shared and MBMS Channels . 146

A.3 High Speed Uplink Packet Access . . . . . . . . . . . 150

A.3.1 MAC-e/es entity in the UE . . . . . . . . . . 151

A.3.2 MAC-e entity in the Node B . . . . . . . . . . 153

A.3.3 MAC-es entity in the SRNC . . . . . . . . . . 155

T he following sections are a supplement to Section 2.2.5. They introducethese MAC entities which are not covered so far as they do not directly

influence the performance as studied within this thesis. Nevertheless, forcompleteness and because most of the entities are important for the overallsystem functionality they are described in the following on the same levelof detail as the entities in charge of the DCH and the HS-DSCH (compareSection 2.2.5.3 and 2.2.5.4).

A.1 System Broadcast by the MAC-b Entity

The MAC-b (MAC-broadcast) entity is responsible for the broadcast ofsystem information. In the UTRAN the MAC-b entity stores and schedulesthe transmission of the System Information Blocks (SIBs) broadcasted onthe BCH of one cell. In order to do so there exists one MAC-b entity forevery cell. These MAC-b entities are located in the Node B. In the UEthere exist one MAC-b entity for the serving cell and multiple entities forthe neighbouring cells the UE is receiving system information from. Herethe MAC-b entity is in charge of receiving the BCH of the measured cells.As can be seen in Figure A.1 the only logical channel the MAC-b entity isresponsible for is the BCCH.

146 A. Additional MAC Entities

MAC-b

BCH

MAC control BCCH

Figure A.1: MAC-b entity at UE and UTRAN side

A.2 Entities for Common, Shared and MBMS Channels

The MAC-c/sh/m (MAC-control/shared/mbms) handles common, sharedand MBMS channels. In detail the transport channels controlled by thisentity are the PCH, the FACH and the RACH as well as the TDD onlyDSCH and USCH. There is one MAC-c/sh/m entity in each UE which isresponsible for the above channels of the serving cell. In the UTRAN oneentity for every cell exists in the RNC which is controlling the correspondingcells. Figure A.2 and Figure A.3 illustrate the MAC-c/sh/m entities andtheir functionalities for both cases.

Several logical channels are mapped onto transport channels by theMAC-c/sh/m entity. The PCCH has a static mapping to the PCH. Broad-cast information carried by the BCCH is mapped to the FACH. The bidi-rectional CCCH has a fixed mapping to the FACH in the downlink and tothe RACH in the uplink. The downlink only CTCH as well as the MBMSchannels MTCH, MCCH and MSCH are always mapped to the FACH. InTDD the control information of the SHCCH can be mapped to the FACHor DSCH for the downlink direction and to the RACH or USCH for theuplink direction.

Dedicated user traffic carried by the DTCH and DCCH can also be trans-mitted using shared channels. As both logical channels are only connectedto the MAC-d entity the MAC-c/sh/m entity has connections to all MAC-dentities for which a mapping to a shared channel exists. In FDD the trafficof multiple UEs can be transmitted using the FACH and RACH. In caseof TDD the DSCH and USCH are available as well. Because the MAC-d

A.2. Entities for Common, Shared and MBMS Channels 147

read MBMS Idadd/read

UE Id

TCTF MUX

Scheduling/Priority Handling

UL: TF selection

ASC selection

TFC selection

MCCH

MAC-c/sh/m

MSCH MTCH MTCH

to MAC-d

MAC ControlPCCH SHCCH CCCH CTCH BCCH(TDD only)

PCH DSCH DSCH(TDD only) (TDD only) (TDD only) (TDD only)

USCH USCH FACH FACH RACH

Figure A.2: UE side MAC-c/sh/m architecture

entity for one UE is not necessarily located in the same network elementin the UTRAN, the MAC-c/sh/m entity includes a flow control mechanismwhich limits the amount of required buffering between both entities. WhenMAC-d traffic is forwarded to the MAC-hs entity a similar mechanism existstowards this entity, located in the Node B, as well. In the UE such a flowcontrol mechanism is not needed.

Because several logical channels of different types, different MBMS ser-vices or different users can be mapped on the same transport channel theMAC-c/sh/m entity provides a mechanism to distinguish those channels.Whenever logical channels of a different type are multiplexed on the sametransport channel the transmitting MAC-c/sh/m entity adds a so-calledTarget Channel Type Field (TCTF) to the MAC PDUs. The TCTF is avariable sized header field allowing the receiver to map the MAC SDUs tothe correct logical channel type. The TCTF indicates the common logical


MAC-c/sh/m

to MAC-d

to MAC-hs

MAC Control

Flow ControlMAC-d /

MAC-c/sh/m

Flow ControlMAC-hs /

MAC-c/sh/m

TCTF MUX /UE Id MUX /

MBMS Id MUX

Scheduling / Buffering /Priority Handling / Demultiplexing

TFC selection

DL: codeallocation

TFC selection

PCH FACH FACH DSCH DSCH USCH USCH RACH(TDD only) (TDD only) (TDD only) (TDD only)

PCCH BCCH SHCCH CCCH CTCH MCCH MSCH MTCH(TDD only)

Figure A.3: UTRAN side MAC-c/sh/m architecture

channel type or if a dedicated logical channel is used. In case of dedicatedlogical channels or multiple MTCHs providing different MBMS services thisinformation is not sufficient to identify a logical channel unambiguously. Forthese scenarios multiplexing is achieved by the UE Id and MBMS Id headerfields which are added to the MAC PDUs by the transmitting MAC-c/sh/mentity. The UE Id field contains either the 16 bit Cell-RNTI (C-RNTI) orthe 32 bit UTRAN-RNTI (U-RNTI) which are allocated by the CRNC andSRNC, respectively. A UE Id Type field of 2 bits is used to distinguish bothvariants of identities. The U-RNTI is unique within the whole Public LandMobile Network (PLMN) and consists of two parts, the SRNC identifier andthe SRNC-RNTI (S-RNTI) which identifies a UE within the SRNC. TheC-RNTI is cell specific and can be reallocated when the UE accesses a newcell. The MBMS Id field has a length of 4 bit and allows to map up to 15MTCHs on the same FACH.

The next step after the addition of multiplexing headers consists of

A.2. Entities for Common, Shared and MBMS Channels 149

scheduling and priority handling. In UTRAN the MAC-c/sh/m schedulerallocates the resources of the FACH and DSCH to different UEs and todifferent data flows of single UEs according to their priority (MLP) and de-lay requirements as set by higher layers. Furthermore, fairness between thedata flows of the UEs must be assured. In the uplink the MAC-c/sh/m en-tity in the UE performs the same task for traffic flows of different prioritiesreceived from MAC-d which are mapped on the RACH. For the resourcesassociated to the USCH no scheduling is performed by the MAC layer. Al-location of resources to the UEs is centrally done by RRC in the CRNC.Based on this scheduling the CRNC is able to demultiplex uplink transmis-sions and forward them to the correct MAC-d entities. The flow control inthe UTRAN can not completely avoid scheduling related buffering. Hence,limited buffering is an additional functionality of the MAC-c/sh/m entityin UTRAN.

Depending on the amount of data that is to be transmitted the MAC-c/sh/m entity is in charge of selecting appropriate TFs for each transportchannel. In case multiple transport channels share physical resources, i.e.they are mapped onto the same CCTrCH, the combination of selected TFsmust not exceed the available capacity. It is a task of the MAC-c/sh/mentity to select appropriate TFCs which fulfill this criterion. These TFCsare configured by RRC and form the so-called TFCS. At the UE side theMAC-c/sh/m entity is in charge of selecting TFCs for the USCHs and TFsfor the RACH. In the downlink direction the MAC-c/sh/m entity in theUTRAN selects TFCs for the DSCHs and for the PCH and jointly encodedFACHs. If only one transport channel is mapped onto a CCTrCH that TFCselection is reduced to a simpler TF selection.

Further functionalities covered by the MAC-c/sh/m entity are the down-link code allocation of the DSCH and the Access Service Class (ASC) selec-tion of the RACH. In the UTRAN the MAC-c/sh/m entity allocates codesfor the DSCH to individual users on a TTI basis. This allocation is signalledto the physical layer. In the uplink the ASC associated with a RACH PDUis forwarded to the physical layer. This ASC is used in the physical layerto send this PDU on appropriate physical resources (i.e. access slot andsignature in FDD or time slot and channelization code in TDD). By havingdifferent ASCs the transmission probability on the RACH can be controlled.Emergency calls, for example, can be established by a random access of ahigher priority compared to other transmissions on this channel.

For MBMS the channels from multiple cells can be received simultane-


MAC Control

FACHFACH

MTCH MTCH MSCH

read MBMS Id

TCTF MUX

MAC-m

Figure A.4: Overview of MAC-m as used for MBMS

ously and selection combining of the received PDUs is performed. In casesuch a setup is configured there exist additional MAC-m entities in the UE.For every neighbouring cell from which MBMS services are received oneMAC-m entity as shown in Figure A.4 is established. The MAC-c/sh/mentity contains the MBMS functionality for the serving cell. The MAC-mentities contain a subset of the features of a MAC-c/sh/m entity. Only theTCTF demultiplexing and the demultiplexing using the MBMS Id are ap-plied to map received PDUs to the MSCH and one or more MTCHs. Theselection combining of the PDUs from different cells is done in the RLClayer.

A.3 High Speed Uplink Packet Access

The concept of the HSUPA is based on the E-DCH. With the introduc-tion of the E-DCH in Release 6, three new MAC entities have been addedto the 3GPP specification. In the UE the MAC-e/es entity (Figure A.5)is responsible for the handling of the E-DCH. The network counterpart,handling the E-DCH in the Node B, is the MAC-e entity, depicted in Fig-ure A.6. Because multiple Node Bs may receive the E-DCH transmissionsof a single UE combining functionality is implemented in the SRNC. Thecorresponding entity is called MAC-es and is illustrated in Figure A.7.

A.3. High Speed Uplink Packet Access 151

A.3.1 MAC-e/es entity in the UE

The MAC-e/es entity in the UE got its name from the fact that it includesthe counterpart functionality of both the MAC-e entity and MAC-es entityof the UTRAN. No clear split between MAC-e and MAC-es functionalitycan be depicted in this entity. The MAC-e/es entity is connected to theMAC-d entity. From there MAC-d PDUs are received. Equally sized MAC-d PDUs from one logical channel are multiplexed within one MAC-es PDU.For every logical channel there exists a 6 bit counter which is incrementedfor every created MAC-es PDU. The value of this counter is used to set theTSN which is the only header field of the MAC-es PDU. Within every TTIexactly one MAC-es PDU may be created per logical channel.

Multiple MAC-es PDUs originating from different logical channels aremultiplexed within one MAC-e PDU. To do so the MAC-e PDU has a headercontaining a list of value pairs being the Data Description Indicator (DDI)and the number of MAC-d PDUs corresponding to this DDI. The 6 bitDDI value identifies the logical channel the MAC-d PDUs originated fromand the size of the MAC-d PDUs. The exact mapping of DDI values isconfigured by the RRC using the MAC control SAP. The number of MAC-d PDUs is encoded by a 6 bit field as well. If enough space is available orif a Scheduling Information (SI) needs to be transmitted an SI is appendedat the end of the MAC-e PDU. The SI has a length of 18 bits and containsinformation about the UE’s buffer status, the logical channel of highestpriority where data is available for transmission and the power headroom interms of ratio between maximum transmission power and current DPCCHtransmission power. A special DDI value is used to indicate the presence ofthe SI in case it can not be unambiguously derived from the PDU length.

The E-DCH Transport Format Combination (E-TFC) selection is incharge of controling the MAC-e/es multiplexing functionality. Based onRRC configuration parameters the E-TFC selection handles the multiplex-ing according to the priorities of the MAC-d flows and the flow specificmultiplexing lists and HARQ profiles. The amount of data which can betransmitted within a TTI is deduced by a number of grant variables main-tained by the E-TFC selection entity. These grant values are initially setby RRC signalling and can be updated by Layer 1 (L1) signalling. L1 up-dates can either be absolute grant values received from the serving cell or, incase of FDD, relative grants received from both the serving cell and furthernon-serving cells. Cells controlled by the Node B of the serving cell may in-


E-TFC SelectionMultiplexing and

TSN setting

HARQ

E-DCHE-DPCCH (FDD)E-UCCH (TDD)

(E-TFC)

E-HICH(ACK/NACK)

to MAC-d MAC Control

MAC-e/es

E-AGCHE-RGCH(s) (FDD)

(Scheduling)

SchedulingAccess Control

(TDD)

E-RUCCH (TDD)

AssociatedDownlinkSignalling

AssociatedUplink

Signalling

Figure A.5: Architecture of MAC-e/es in the UE

crease, hold or decrease the grant relatively to the last serving grant. Thesecells form the serving E-DCH Radio Link Set (RLS) and always transmitthe same relative grant update commands. Cells which are controlled bydifferent Node Bs belong to the non-serving RLS and are not allowed toincrease the grant.

The selected E-TFC, represented by its E-DCH Transport Format Com-bination Indicator (E-TFCI), is provided to the HARQ entity. Togetherwith the MAC-e PDU created by the multiplexing entity the HARQ en-tity is able to handle the HARQ protocol of the E-DCH. Similarly to theHARQ entity of the MAC-hs the MAC-e/es’ HARQ entity contains sev-eral parallel HARQ processes in order to allow the continuous transmissionbased on the stop-and-wait HARQ protocol. The number of required HARQprocesses depends on the RTT based on the selected E-DCH TTI. For a10 ms TTI 4 HARQ processes are used by the HARQ entity and for a 2 msTTI the number of HARQ processes is 8. In contrast to the HS-DSCHthe FDD E-DCH does not have an explicit signalling of the HARQ processa transmission originated from. Instead a synchronous HARQ scheme isused where the HARQ process identification is derived from the Connec-tion Frame Number (CFN) and, in case of a 2 ms TTI, additionally fromthe subframe number. Also the ACK/NACK as well as the process spe-


cific grant signalling is mapped to the correct HARQ process based on therelative timing.

Every HARQ process stores its MAC-e payload and retransmits it as longas it is not positively acknowledged and the maximum number of transmis-sions is not reached. If a positive acknowledgment is received the E-TFCselection is informed and a new MAC-e PDU is created. The maximumnumber of transmissions and an E-DPDCH power offset are part of theMAC-d flow specific HARQ profile which is used to differentiate QoS ofindividual MAC-e PDU transmissions. For every transmission and retrans-mission the HARQ process informs PHY about the power offset and signalsthe chosen E-TFCI, the Retransmission Sequence Number (RSN) and theso-called ’happy bit’ using the associated uplink signalling. The RSN isused to indicate to the receiver that a new transmission has been started.Furthermore, the RV is derived from the RSN in case incremental redun-dancy is configured by the RRC layer. The happy bit is used to indicateto the Node B whether the UE could use more resources or not. In TDDmode the HARQ process identifier is an additional parameter signalled tothe Node B by the associated uplink signalling.

In case of TDD the Scheduling Access Control is an additional func-tionality of the MAC-e/es entity. The Scheduling Access Control entity isresponsible for routing associated signalling and SI via the E-DCH RandomAccess Uplink Control Channel (E-RUCCH) in case no E-DCH resourcesare assigned to the UE.

A.3.2 MAC-e entity in the Node B

In the UTRAN the E-DCH reception is done by the MAC-e entity withinthe Node B. One MAC-e entity is instantiated for every UE the Node Breceives E-DCH transmissions from. As macro diversity is applied to theFDD E-DCH there is no one-to-one mapping between MAC-e/es entitiesin the UE and MAC-e entities in the UTRAN. Every Node B which takespart in the RLS of one UE contains one MAC-e entity for this UE. In TDDmode macro diversity for the E-DCH is not used. All MAC-e entities shareone common E-DCH scheduling function which is in charge of schedulingthe resources of the individual UEs.

The E-DCH scheduling function receives the scheduling requests fromthe UEs by their MAC-e entities. Based on this information the cell re-sources are allocated to the UEs by transmitting the scheduling grants


E-DCHScheduling E-DCH Control

HARQ entity

Demultiplexing

MAC-e

MAC Control MAC-d Flows

E-DCH

AssociatedACK/NACKDownlinkSignalling

AssociatedE-TFCUplink

Signalling

E-AGCHE-RGCH(s) (FDD) E-DPCCH (FDD)

E-UCCH (TDD)

E-HICH

AssociatedSchedulingDownlinkSignalling

E-UCCH (TDD)

AssociatedUplink

Signalling

E-RUCCH (TDD)

Figure A.6: Node B MAC-e details

using the associated downlink signalling. For UEs for which the Node Bcontrols the serving cell both absolute and relative grant settings may beused to increase or decrease the serving grants. Interference from other cellsis measured and the Node B may instruct UEs for which it does not actas a serving Node B to reduce their grants by relative grant updates. InTDD only absolute grant settings are available. The detailed schedulingalgorithm itself is not specified. Instead, the algorithm and the underlyingRRM strategies are left open for network vendor specific solutions.

The HARQ entity in MAC-e receives the E-DCH transmission and routesit together with the associated RSN value to the correct HARQ processbased on its relative timing. Derived from the RSN, the CFN and thesubframe number, the HARQ process calculates the transmission numberand the RV. Based on this information it tries to decode the E-DCH payloadand, depending on the decoding result, sends an ACK or a NACK indicationback to the UE. If the decoding is successful the decoded MAC-e PDU issent to the demultiplexing entity.


The demultiplexing of the MAC-e PDU is done by using the DDI andnumber of PDU pairs of the MAC-e PDU header. The DDI mapping isconfigured by the SRNC and indicates the MAC-d flow and PDU size of theencapsulated MAC-es payload. Furthermore, the Iub bearer to be used forthe individual MAC-d flows is configured. By knowing the PDU sizes andthe number of corresponding PDUs the demultiplexing function is able toextract the MAC-es PDUs and to forward them to the correct RNCs.

A.3.3 MAC-es entity in the SRNC

For every UE which is using the E-DCH there exists one MAC-es entityin its SRNC. The MAC-es entity receives the MAC-es PDUs from thoseNode Bs of the UE’s E-DCH active set which could correctly decode theE-DCH transmission. The soft combining of the receptions for the cellscontrolled by one Node B is already done within this Node B. In the SRNCselection combining takes place. For every MAC-es PDU it is sufficient tobe correctly received by at least one Node B.

Based on the SRNC configuration and information provided by theNode B the reordering queue distribution function routes the MAC-es PDUs

Disassembly Disassembly Disassembly

Reordering QueueDistribution

Reordering QueueDistribution

Reordering /Combining



MAC-es

MAC Controlto MAC-d

MAC-d Flows from MAC-e Entities

Figure A.7: MAC-es details in SRNC


to the correct reordering buffer. There exists one reordering buffer per log-ical channel. The reordering function uses the TSN of the MAC-es PDUand Node B reception parameters, e.g. CFN and subframe number, to re-construct the correct PDU order. Duplicate MAC-es PDUs provided bymultiple Node Bs are eliminated by the reordering function. MAC-es PDUswith consecutive TSNs are delivered to the disassembly function.

The disassembly function extracts the MAC-d PDUs from the MAC-esPDU based on the DDI and number of PDUs values carried by the MAC-eheader and provided to the MAC-es by the MAC-e entity in the Node B.The disassembled MAC-d PDUs are finally delivered to the MAC-d entitythe MAC-es entity is connected to.

LIST OF FIGURES

1.1 UMTS releases and key features . . . . . . . . . . . . . . . . . 2

2.1 UTRAN network topology . . . . . . . . . . . . . . . . . . . . 82.2 UMTS protocol architecture . . . . . . . . . . . . . . . . . . . 102.3 Radio interface protocol architecture . . . . . . . . . . . . . . 112.4 UE side model of RRC . . . . . . . . . . . . . . . . . . . . . . 132.5 Abstract illustration of ASN.1 data structure . . . . . . . . . 152.6 PDCP structure . . . . . . . . . . . . . . . . . . . . . . . . . 172.7 Header compression for IPv4 . . . . . . . . . . . . . . . . . . 182.8 Header compression for TCP . . . . . . . . . . . . . . . . . . 192.9 Header compression for IPv6 . . . . . . . . . . . . . . . . . . 192.10 Model of the Broadcast Multicast Control sublayer . . . . . . 202.11 Overview of the RLC sublayer . . . . . . . . . . . . . . . . . . 212.12 Model of Transparent Mode peer entities . . . . . . . . . . . . 232.13 Model of Unacknowledged Mode peer entities configured with-

out duplicate avoidance and reordering . . . . . . . . . . . . . 242.14 Model of Unacknowledged Mode peer entities as used for

MBMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.15 Model of an Acknowledged Mode entity . . . . . . . . . . . . 272.16 UE side MAC architecture . . . . . . . . . . . . . . . . . . . . 312.17 UTRAN side MAC architecture . . . . . . . . . . . . . . . . . 322.18 MAC-d architecture . . . . . . . . . . . . . . . . . . . . . . . 38

(a) UE side . . . . . . . . . . . . . . . . . . . . . . . . . . . 38(b) UTRAN side . . . . . . . . . . . . . . . . . . . . . . . . 38

2.19 MAC-hs details . . . . . . . . . . . . . . . . . . . . . . . . . . 40(a) UE side . . . . . . . . . . . . . . . . . . . . . . . . . . . 40(b) UTRAN side . . . . . . . . . . . . . . . . . . . . . . . . 40

2.20 Transport channel multiplexing and coding . . . . . . . . . . 46(a) FDD downlink . . . . . . . . . . . . . . . . . . . . . . . 46(b) FDD uplink and TDD . . . . . . . . . . . . . . . . . . . 46

2.21 Transport channel coding chain for HS-DSCH and E-DCH . . 48(a) HS-DSCH . . . . . . . . . . . . . . . . . . . . . . . . . . 48

158 List of Figures

(b) E-DCH . . . . . . . . . . . . . . . . . . . . . . . . . . . 482.22 Convolutional coding with rate 1/2 and rate 1/3 . . . . . . . 502.23 Turbo coding with rate 1/3 . . . . . . . . . . . . . . . . . . . 512.24 Rate Matching . . . . . . . . . . . . . . . . . . . . . . . . . . 53

(a) Convolutionally coded channels and turbo coded chan-nels with repetition . . . . . . . . . . . . . . . . . . . . 53

(b) Puncturing of turbo coded channels . . . . . . . . . . . 53(c) HS-DSCH HARQ . . . . . . . . . . . . . . . . . . . . . 53(d) E-DCH HARQ . . . . . . . . . . . . . . . . . . . . . . . 53

2.25 Block interleaver . . . . . . . . . . . . . . . . . . . . . . . . . 552.26 Physical channel processing in FDD downlink . . . . . . . . . 572.27 Processing of dedicated physical uplink channels in FDD . . . 592.28 FDD frame and slot structure . . . . . . . . . . . . . . . . . . 612.29 Downlink modulation mapping . . . . . . . . . . . . . . . . . 63

(a) QPSK . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63(b) 16QAM . . . . . . . . . . . . . . . . . . . . . . . . . . . 63(c) 64QAM . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

2.30 Uplink modulation mapping . . . . . . . . . . . . . . . . . . . 63(a) BPSK . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63(b) 4PAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

2.31 Channelization code tree . . . . . . . . . . . . . . . . . . . . . 642.32 Modulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

3.1 Wireless Network Simulator protocol stack . . . . . . . . . . . 703.2 Simulated nodes of the UMTS RAN . . . . . . . . . . . . . . 723.3 Hexagonal scenario with different antenna configurations . . . 75

(a) Omnidirectional . . . . . . . . . . . . . . . . . . . . . . 75(b) Sectorized . . . . . . . . . . . . . . . . . . . . . . . . . . 75(c) Realistic 3D pattern . . . . . . . . . . . . . . . . . . . . 75

3.4 Highway mobility model . . . . . . . . . . . . . . . . . . . . . 763.5 Python configuration example of HS-DSCH chain . . . . . . . 763.6 Screenshot of HS-DSCH link-level simulation . . . . . . . . . 773.7 Real Time Wireless Network Demonstrator . . . . . . . . . . 79

4.1 Bit error and symbol error rate of modulation schemes . . . . 85(a) Bit error rate . . . . . . . . . . . . . . . . . . . . . . . . 85(b) Symbol error rate . . . . . . . . . . . . . . . . . . . . . 85

4.2 Simulated bit error probability of 16QAM and 64QAM . . . . 86

List of Figures 159

(a) 16QAM . . . . . . . . . . . . . . . . . . . . . . . . . . . 86(b) 64QAM . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

4.3 Bit Error Rates of UMTS Forward Error Correction . . . . . 87(a) Convolutional code . . . . . . . . . . . . . . . . . . . . . 87(b) Turbo code . . . . . . . . . . . . . . . . . . . . . . . . . 87

4.4 Block Error Rates and Throughput of Forward Error Correction 88(a) Block Error Rate . . . . . . . . . . . . . . . . . . . . . . 88(b) Throughput . . . . . . . . . . . . . . . . . . . . . . . . . 88

4.5 Bit error probability depending on SF . . . . . . . . . . . . . 90(a) Downlink . . . . . . . . . . . . . . . . . . . . . . . . . . 90(b) Uplink . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

4.6 Simulated DCH downlink BLER on an AWGN channel . . . 92(a) 20 ms TTI . . . . . . . . . . . . . . . . . . . . . . . . . 92(b) 10 ms TTI . . . . . . . . . . . . . . . . . . . . . . . . . 92

4.7 Simulated DCH uplink BLER on an AWGN channel . . . . . 95(a) 20 ms TTI . . . . . . . . . . . . . . . . . . . . . . . . . 95(b) 10 ms TTI . . . . . . . . . . . . . . . . . . . . . . . . . 95

4.8 Simulated HS-DSCH BER and BLER for an AWGN channel 99(a) Bit Error Rate . . . . . . . . . . . . . . . . . . . . . . . 99(b) Block Error Rate . . . . . . . . . . . . . . . . . . . . . . 99

4.9 BLER mapping for an AWGN channel based on analyticalmodel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

4.10 Mobile WiMAX BLER mapping for an AWGN channel . . . 1024.11 Physical layer throughput on an AWGN channel . . . . . . . 105

(a) HS-DSCH, Mobile WiMAX . . . . . . . . . . . . . . . . 105(b) DCH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

4.12 Maximum throughput on RLC level . . . . . . . . . . . . . . 106(a) 336 bit MAC-d PDUs . . . . . . . . . . . . . . . . . . . 106(b) 656 bit MAC-d PDUs . . . . . . . . . . . . . . . . . . . 106

4.13 Throughput for ITU PA channel, 3 km/h . . . . . . . . . . . 109(a) Rake receiver . . . . . . . . . . . . . . . . . . . . . . . . 109(b) MMSE receiver, order 32 . . . . . . . . . . . . . . . . . 109

4.14 Throughput for ITU VA channel, 100 km/h . . . . . . . . . . 110(a) MMSE receiver, order 32 . . . . . . . . . . . . . . . . . 110(b) MMSE receiver, order 128 . . . . . . . . . . . . . . . . . 110

4.15 HARQ gain for Chase Combining and Incremental Redundancy115(a) Chase Combining . . . . . . . . . . . . . . . . . . . . . 115(b) Incremental Redundancy . . . . . . . . . . . . . . . . . 115

160 List of Figures

4.16 HARQ gain using Constellation Rearrangement only . . . . . 1164.17 Comparison of HARQ gains for first retransmission . . . . . . 1174.18 Coding rates of reference AMC schemes . . . . . . . . . . . . 118

(a) Category 10 . . . . . . . . . . . . . . . . . . . . . . . . . 118(b) Category 14 . . . . . . . . . . . . . . . . . . . . . . . . . 118

5.1 Elementary Node B scheduling principle . . . . . . . . . . . . 1225.2 Qualitative comparison of scheduling metrics . . . . . . . . . 128

(a) Average throughput . . . . . . . . . . . . . . . . . . . . 128(b) HOL packet delay . . . . . . . . . . . . . . . . . . . . . 128

5.3 RT service scenario with various path loss profiles . . . . . . 1315.4 CQI distribution . . . . . . . . . . . . . . . . . . . . . . . . . 1315.5 MAC-d PDU queueing delay . . . . . . . . . . . . . . . . . . 132

(a) Proportional Fair (PF) . . . . . . . . . . . . . . . . . . 132(b) Expo-Linear (EL) . . . . . . . . . . . . . . . . . . . . . 132

5.6 MAC-hs inter-scheduling interval . . . . . . . . . . . . . . . . 134(a) Proportional Fair (PF) . . . . . . . . . . . . . . . . . . 134(b) Expo-Linear (EL) . . . . . . . . . . . . . . . . . . . . . 134

5.7 Simulation scenario with mixed services . . . . . . . . . . . . 1355.8 MAC-hs PDU throughput with mixed services . . . . . . . . 1365.9 MAC-hs UE throughput of NRT service . . . . . . . . . . . . 1385.10 MAC-d PDU queueing delay of RT users . . . . . . . . . . . . 139

A.1 MAC-b entity at UE and UTRAN side . . . . . . . . . . . . . 146A.2 UE side MAC-c/sh/m architecture . . . . . . . . . . . . . . . 147A.3 UTRAN side MAC-c/sh/m architecture . . . . . . . . . . . . 148A.4 Overview of MAC-m as used for MBMS . . . . . . . . . . . . 150A.5 Architecture of MAC-e/es in the UE . . . . . . . . . . . . . . 152A.6 Node B MAC-e details . . . . . . . . . . . . . . . . . . . . . . 154A.7 MAC-es details in SRNC . . . . . . . . . . . . . . . . . . . . . 155

LIST OF TABLES

2.1 Mapping of transport channels to physical channels . . . . . . 452.2 Constellation rearrangement for 16QAM and 64QAM . . . . 562.3 Symbol rates per physical channel depending on SF . . . . . 65

4.1 Typical RB configurations . . . . . . . . . . . . . . . . . . . . 914.2 Downlink burst formats used in simulation . . . . . . . . . . . 934.3 Uplink DPDCH configurations . . . . . . . . . . . . . . . . . 954.4 CQI table for category 10 and category 14 (Release 7) . . . . 984.5 Modulation and coding schemes of Mobile WiMAX . . . . . . 1024.6 Available maximum data bits without coding [Mbit/s] . . . . 1034.7 Throughput comparison on various layers . . . . . . . . . . . 1074.8 Allowed redundancy combinations . . . . . . . . . . . . . . . 1124.9 HS-DSCH UE categories . . . . . . . . . . . . . . . . . . . . . 113

5.1 Scenario details and configuration parameters . . . . . . . . . 1305.2 Path loss in different UE groups . . . . . . . . . . . . . . . . 1305.3 Mean queueing delay and packet loss . . . . . . . . . . . . . . 1335.4 Traffic parameters for mixed service scenario . . . . . . . . . 1355.5 Mean cell throughput . . . . . . . . . . . . . . . . . . . . . . 1375.6 MAC-d PDU delivery statistics . . . . . . . . . . . . . . . . . 139

162 List of Tables

LIST OF EQUATIONS

2.1 Generator polynomial for 24 bit CRC . . . . . . . . . . . . . 482.2 Generator polynomial for 16 bit CRC . . . . . . . . . . . . . 482.3 Generator polynomial for 12 bit CRC . . . . . . . . . . . . . 482.4 Generator polynomial for 8 bit CRC . . . . . . . . . . . . . . 482.5 1/2 rate convolutional code generator polynomial 1 . . . . . . 492.6 1/2 rate convolutional code generator polynomial 2 . . . . . . 492.7 1/3 rate convolutional code generator polynomial 1 . . . . . . 492.8 1/3 rate convolutional code generator polynomial 2 . . . . . . 492.9 1/3 rate convolutional code generator polynomial 3 . . . . . . 492.10 Transfer function of Turbo Code . . . . . . . . . . . . . . . . 502.11 Recursive OVSF generation formula 1 . . . . . . . . . . . . . 642.12 Recursive OVSF generation formula 2 . . . . . . . . . . . . . 642.13 Root-raised cosine filter . . . . . . . . . . . . . . . . . . . . . 67

4.1 Marcum Q-function . . . . . . . . . . . . . . . . . . . . . . . 834.2 Bit error probability for PSK modulations . . . . . . . . . . . 834.3 Symbol error probability for QPSK modulation . . . . . . . . 844.4 Relation between SINR, bit energy and symbol energy . . . . 844.5 Bit error probability for QAM modulation . . . . . . . . . . . 844.6 Symbol error probability for QAM modulation . . . . . . . . 844.7 Convolutional code upper bound bit error approximation . . 864.8 Shannon’s formula . . . . . . . . . . . . . . . . . . . . . . . . 884.9 Bit error probability depending on SF and chip energy . . . . 904.10 Code block number and size for Turbo Coded channels . . . . 944.11 Number of Turbo Coded bits including trellis termination . . 944.12 Approximation for SINR to CQI mapping . . . . . . . . . . . 994.13 Expected basic HARQ gain . . . . . . . . . . . . . . . . . . . 1144.14 Effective coding rate . . . . . . . . . . . . . . . . . . . . . . . 117

5.1 Maximum SINR scheduling . . . . . . . . . . . . . . . . . . . 1245.2 Proportional Fair scheduling . . . . . . . . . . . . . . . . . . . 1245.3 Probability of discarded packets for M-LWDF scheduling . . . 125

164 List of Equations

5.4 Modified Largest Weighted Delay First scheduling . . . . . . 1255.5 Exponential Rule scheduling . . . . . . . . . . . . . . . . . . . 1265.6 Exponential Rule scheduling (2) . . . . . . . . . . . . . . . . 1265.7 Channel-Dependent Earliest Due Date scheduling . . . . . . . 1265.8 Expo-Linear scheduling . . . . . . . . . . . . . . . . . . . . . 1275.9 Jain’s fairness index . . . . . . . . . . . . . . . . . . . . . . . 137

LIST OF ABBREVIATIONS

#16QAM 16-State Quadrature

AmplitudeModulation

1G 1st Generation

2G 2nd Generation

3G 3rd Generation

3GPP 3rd GenerationPartnership Project

4PAM 4-State PulseAmplitudeModulation

64QAM 64-State QuadratureAmplitudeModulation

AABMT Aachener Beitrage

zur Mobil- undTelekommunikation

ACK Acknowledgement

AICH Acquisition IndicatorChannel

AM Acknowledged Mode

AMC AdaptiveModulation andCoding

AMR AdaptiveMulti-Rate

ARP Address ResolutionProtocol

ARQ Automatic RepeatRequest

AS Access Stratum

ASC Access Service Class

ASN.1 Abstract SyntaxNotation One

AWGN Additive WhiteGaussian Noise

BBCCH Broadcast Control

Channel

BCFE Broadcast ControlFunctional Entity

BCH Broadcast Channel

BEC Backward ErrorCorrection

BER Bit Error Rate

BLER Block Error Rate

BMC Broadcast/MulticastControl

BO Buffer Occupancy

BPSK Binary Phase-ShiftKeying

BS Base Station

BSC Base StationController

BSS Base StationSubsystem

BTS Base TransceiverStation

166 List of Abbreviations

CC- Control-CAC Connection

Admission ControlCB Cell BroadcastCBC Cell Broadcast

CenterCBR Constant Bit RateCBS Cell Broadcast

ServiceCC Call ControlCC Chase CombiningCC Convolutional CodeCCCH Common Control

ChannelCCDF Complementary

CumulativeDistributionFunction

CCPCH Common ControlPhysical Channel

CCTrCH Coded CompositeTransport Channel

CD-EDD Channel-DependentEarliest Due Date

CDF CumulativeDistributionFunction

CDMA Code DivisionMultiple Access

CFN Connection FrameNumber

CID Context IdentifierCIR Carrier to

Interference RatioCM Connection

ManagementCN Core Network

CoCar Cooperative CarsCodec Coder/DecoderComNets Communication

Networks ResearchGroup at RWTHAachen University

CP Cyclic PrefixCPICH Common Pilot

ChannelCQI Channel Quality

IndicatorCR Constellation

RearrangementCRC Cyclic Redundancy

CheckCRNC Controlling Radio

Network ControllerC-RNTI Cell-RNTICS Circuit SwitchedC/T Control/TrafficCTC Convolutional Turbo

CodeCTCH Common Traffic

Channel

DDC Dedicated ControlDCCH Dedicated Control

ChannelDCFE Dedicated Control

Functional EntityDCH Dedicated ChannelDDI Data Description

IndicatorDHCP Dynamic Host

ConfigurationProtocol


DL Downlink

DLC Data Link Control

DLL Data Link Layer

DPCCH Dedicated PhysicalControl Channel

DPCH Dedicated PhysicalChannel

DPDCH Dedicated PhysicalData Channel

DRNC Drift Radio NetworkController

DRX DiscontinuousReception

DS-CDMA Direct-SequenceCode DivisionMultiple Access

DSCH Downlink SharedChannel

DSL Digital SubscriberLine

DTCH Dedicated TrafficChannel

DTX DiscontinuousTransmission

DwPCH Downlink PilotChannel

E

E-AGCH E-DCH AbsoluteGrant Channel

ECR Effective Code Rate

E-DCH Enhanced DedicatedChannel

EDD Earliest Due Date

EDF Earliest DeadlineFirst

EDGE Enhanced DataRates for GSMEvolution

E-DPCCH E-DCH DedicatedPhysical ControlChannel

E-DPDCH E-DCH DedicatedPhysical DataChannel

E-HICH E-DCH Hybrid ARQIndicator Channel

EL Expo-Linear

EPC Estimated PDUCounter

EPS Evolved PacketSystem

E-PUCH E-DCH PhysicalUplink Channel

ER Exponential Rule

E-RGCH E-DCH RelativeGrant Channel

E-RUCCH E-DCH RandomAccess UplinkControl Channel

ESP EncapsulatingSecurity Payload

E-TFC E-DCH TransportFormat Combination

E-TFCI E-DCH TransportFormat CombinationIndicator

ETSI EuropeanTelecommunicationStandards Institute

E-UCCH E-DCH UplinkControl Channel

E-UTRAN Evolved UTRAN


FFACH Forward Access

Channel

FBI FeedbackInformation

FCH Frame ControlHeader

FDD Frequency DivisionDuplex

FDMA Frequency DivisionMultiple Access

F-DPCH Fractional DedicatedPhysical Channel

FDT Formal DescriptionTechnique

FEC Forward ErrorCorrection

FER Frame Erasure Rate

FFT Fast FourierTransform

FPACH Fast Physical AccessChannel

FSN First SequenceNumber

FTP File TransferProtocol

GGC General Control

GoS Grade of Service

GPRS General PacketRadio Service

GSM Global System forMobileCommunication

GUI Graphical UserInterface

HHARQ Hybrid ARQ

HC Header Compression

HFN Hyper FrameNumber

HOL Head Of Line

HSCSD High Speed CircuitSwitched Data

HSDPA High SpeedDownlink PacketAccess

HS-DPCCH High SpeedDedicated PhysicalControl Channel

HS-DSCH High SpeedDownlink SharedChannel

HS-PDSCH High Speed PhysicalDownlink SharedChannel

HS-SCCH High Speed SharedControl Channel

HS-SICH High Speed SharedInformation Channel

HSUPA High Speed UplinkPacket Access

HTTP Hypertext TransferProtocol

IID Identifier

IE Information Element

IEEE Institute of Electricaland ElectronicsEngineers

IMEI International MobileEquipment Identity


IMSI International MobileSubscriber Identity

IMT-2000 International MobileTelecommunications-2000

IP Internet Protocol

IPHC Internet ProtocolHeader Compression

IPv4 Internet Protocolversion 4

IPv6 Internet Protocolversion 6

IR IncrementalRedundancy

ISDN Integrated ServicesDigital Network

ISI Inter SymbolInterference

ISO InternationalOrganization forStandardization

ITU InternationalTelecommunicationUnion

ITU-R RadiocommunicationStandardizationSector of ITU

ITU-T TelecommunicationStandardizationSector of ITU

Iu Interface betweenRNS and CN

Iub Interface betweenNode B and RNC

Iur Interface betweenRNCs

Kkbps kilobits per secondksps kilosymbols per

second

LL1 Layer 1L2 Layer 2L3 Layer 3LI Length IndicatorLoCH Logical ChannelLTE Long Term

Evolution

MMAC Medium Access

ControlMAC-b MAC entity for

BCHMAC-c/sh/m MAC entity for

common, shared andMBMS channels

MAC-d MAC entity forDCHs

MAC-e MAC entity forE-DCH in Node B

MAC-e/es MAC entity forE-DCH in UE

MAC-es MAC entity forE-DCH in SRNC

MAC-hs MAC entity forHS-DSCH

MAC-m MAC entity forMBMS channels inUE

MAI Multiple AccessInterference


MAP Medium AccessProtocol

MaxSINR Maximum SINR

MBMS MultimediaBroadcast MulticastService

Mbps Megabits per second

MCCH MBMSpoint-to-multipointControl Channel

Mcps Megachips persecond

MCS Modulation CodingScheme

MIB Master InformationBlock

MICH MBMS IndicatorChannel

MIMO Multiple InputMultiple Output

MLE Mobile Link Entity

MLP MAC LogicalChannel Priority

M-LWDF Modified LargestWeighted DelayFirst

MM MobilityManagement

MMS MultimediaMessaging Service

MMSE Minimum MeanSquare Error

MPEG Moving PictureExperts Group

MPIC Multi PathInterferenceCancellation

MRC Maximum RatioCombining

MRW Move ReceivingWindow

MS Mobile StationMSC Mobile Switching

CenterMSCH MBMS

point-to-multipointScheduling Channel

MT Mobile TerminalMTCH MBMS

point-to-multipointTraffic Channel

MUI Message UnitIdentifier

MUX Multiplexer

NNACK Negative

AcknowledgementNAS Non-Access StratumNBAP Node B Application

PartNCDMA Narrowband Code

Division MultipleAccess

NDI New Data IndicatorNFS Network File SystemNIS Network Information

ServiceNL Network LayerNRT Non-Real TimeNt NotificationNTP Network Time

ProtocolNW Network


O

OFDM OrthogonalFrequency DivisionMultiplexing

OFDMA OrthogonalFrequency DivisionMultiple Access

OMT Object ModelingTechnique

OOSE Object-OrientedSoftware Engineering

OSI Open SystemsInterconnection

OVSF Orthogonal VariableSpreading Factor

P

PA ITU Pedestrian A

PB ITU Pedestrian B

PC Power Control

PCCC ParallelConcatenatedConvolutional Code

PCCH Paging ControlChannel

P-CCPCH Primary CommonControl PhysicalChannel

PCH Paging Channel

P-CPICH Primary CommonPilot Channel

PDCP Packet DataConvergenceProtocol

PDN Public DataNetwork

PDSCH Physical DownlinkShared Channel

PDU Protocol Data UnitPER Packed Encoding

RulesPF Proportional FairPhCH Physical ChannelPHY Physical LayerPI Page IndicatorPICH Paging Indication

ChannelPID Packet IdentifierPL Puncturing LimitPLCCH Physical Layer

Common ControlChannel

PLMN Public Land MobileNetwork

PNBSCH Physical Node BSynchronisationChannel

PNFE Paging andNotification ControlFunctional Entity

POP3 Post Office Protocolversion 3

PQ Priority QueuePRACH Physical Random

Access ChannelPS Packet SwitchedP-SCH Primary

SynchronizationChannel

PSK Phase-Shift KeyingPUSC Partial Usage of

SubchannelsPUSCH Physical Uplink

Shared Channel


QQAM Quadrature

AmplitudeModulation

QoS Quality of ServiceQPSK Quadrature

Phase-Shift Keying

RRAB Radio Access BearerRACH Random Access

ChannelRAN Radio Access

NetworkRAT Radio Access

TechnologyRB Radio BearerRFC Request for

CommentRFE Routing Functional

EntityRISE Radio Interference

Simulation EngineRLC Radio Link ControlRLS Radio Link SetRM Rate MatchingRNC Radio Network

ControllerRNS Radio Network

SubsystemRNTI Radio Network

Temporary IdentifierROHC Robust Header

CompressionRR Round RobinRRC Radio Resource

Control

RRM Radio ResourceManagement

RS Reed-Solomon CodeRSN Retransmission

Sequence NumberRT Real TimeRTP Real-Time

Transport ProtocolRTT Round Trip TimeRTWND Real Time Wireless

NetworkDemonstrator

RV Redundancy VersionRWTH Rheinisch-

WestfalischeTechnischeHochschule Aachen

RX Receive

SSAP Service Access PointSAW Stop-and-WaitS-CCPCH Secondary Common

Control PhysicalChannel

SCFE Shared ControlFunctional Entity

SCH SynchronizationChannel

S-CPICH Secondary CommonPilot Channel

SDL Specification andDescriptionLanguage

SDU Service Data UnitSER Symbol Error RateSF Spreading Factor


SGSN Serving GPRSSupport Node

SHCCH Shared ChannelControl Channel

SHO Soft Handover

SI SchedulingInformation

SIB System InformationBlock

SINR Signal to Interferenceplus Noise Ratio

SIR Signal to InterferenceRatio

SM Session Management

SMS Short MessageService

SMTP Simple MailTransfer Protocol

SN Sequence Number

SNR Signal-to-NoiseRatio

SPEETCL SDL PerformanceEvaluation ToolClass Library

SPI Scheduling PriorityIndicator

SRB Signalling RadioBearer

SRNC Serving RadioNetwork Controller

SRNS Serving RadioNetwork Subsystem

S-RNTI SRNC-RNTI

S-SCH SecondarySynchronizationChannel

SUFI Super Field

TTB Transport BlockTBS Transport Block SetTC Turbo CodeTCP Transmission

Control ProtocolTCP/IP Transmission

ControlProtocol/InternetProtocol

TCTF Target ChannelType Field

TDD Time DivisionDuplex

TDL Tapped Delay LineTDMA Time Division

Multiple AccessTF Transport FormatTFC Transport Format

CombinationTFCI Transport Format

CombinationIndicator

TFCS Transport FormatCombination Set

TFI Transport FormatIndicator

TFRC Transport Formatand ResourceCombination

TFS Transport FormatSet

TL Transport LayerTM Transparent ModeTME Transfer Mode

EntityTPC Transmit Power

Control


TrCH Transport Channel

TS Time Slot

TSN TransmissionSequence Number

TTI Transmission TimeInterval

TX Transmit

U

U- User-

UDP User DatagramProtocol

UE User Equipment

UE-Id User EquipmentIdentity

UE-Id Type User EquipmentIdentification Type

UL Uplink

UM UnacknowledgedMode

UMSC UMTS MobileSwitching Center

UMTS Universal MobileTelecommunicationsSystem

UpPCH Uplink PilotChannel

URI Uniform ResourceIdentifier

URIS UMTS RadioInterface Simulator

U-RNTI UTRAN-RNTI

USCH Uplink SharedChannel

USIM UMTS SubscriberIdentity Module

UTRAN UMTS TerrestrialRadio AccessNetwork

Uu Radio interfacebetween UE andUTRAN

VVA ITU Vehicular AVB ITU Vehicular BVBR Variable Bit RateVoIP Voice over IPVSF Variable Spreading

FactorVT Video Telephony

WWAP Wireless Application

ProtocolWCDMA Wideband Code

Division MultipleAccess

WiMAX WorldwideInteroperability forMicrowave Access

WMAN WirelessMetropolitan AreaNetwork

WNS Wireless NetworkSimulator

WWW World Wide Web

XXOR Exclusive Or

ZZF Zero Forcing

BIBLIOGRAPHY

[1] OpenWNS - Open Wireless Network Simulator. http://www.

openwns.org/, 2007. 69

[2] Project Cooperative Cars (CoCar). http://www.aktiv-online.org/english/aktiv-cocar.html, 2007. 74, 80

[3] Python Programming Language - Official Website. http://www.

python.org/, 2007. 76

[4] 3GPP TS 22.060. General Packet Radio Service (GPRS); ServiceDescription; Stage 1. Technical Specification V7.0.0, 3rd GenerationPartnership Project, Technical Specification Group Services and Sys-tem Aspects, June 2007. 1

[5] 3GPP TS 22.105. Service Aspects: Service and Service Capabilities.Technical Specification V7.1.0, 3rd Generation Partnership Project,Technical Specification Group Services and System Aspects, Decem-ber 2006. 123

[6] 3GPP TS 23.002. Network Architecture. Technical SpecificationV7.3.1, 3rd Generation Partnership Project, Technical SpecificationGroup Services and System Aspects, October 2007. 7

[7] 3GPP TS 23.040. Technical Realization of the Short Message Service(SMS). Technical Specification V7.0.1, 3rd Generation PartnershipProject, Technical Specification Group Core Network and Terminals,March 2007. 12

[8] 3GPP TS 23.060. General Packet Radio Service (GPRS); ServiceDescription; Stage 2. Technical Specification V7.5.0, 3rd GenerationPartnership Project, Technical Specification Group Services and Sys-tem Aspects, October 2007. 1

[9] 3GPP TS 23.101. General UMTS Architecture. Technical Specifica-tion V7.0.0, 3rd Generation Partnership Project, Technical Specifica-tion Group Services and System Aspects, June 2007. 9

[10] 3GPP TS 23.107. Quality of Service (QoS) Concept and Archi-tecture. Technical Specification V7.1.0, 3rd Generation PartnershipProject, Technical Specification Group Services and System Aspects,October 2007. 123

http://www.openwns.org/

http://www.openwns.org/

http://www.aktiv-online.org/english/aktiv-cocar.html

http://www.aktiv-online.org/english/aktiv-cocar.html

http://www.python.org/

http://www.python.org/

176 Bibliography

[11] 3GPP TS 23.108. Mobile radio interface Layer 3 specification; Corenetwork protocols; Stage 2. Technical Specification V7.0.0, 3rd Gen-eration Partnership Project, Technical Specification Group Core Net-work and Terminals, June 2007. 12

[12] 3GPP TS 23.110. UMTS Access Stratum Services and Functions.Technical Specification V7.0.0, 3rd Generation Partnership Project,Technical Specification Group Services and System Aspects, June2007. 9

[13] 3GPP TS 24.008. Mobile radio interface Layer 3 specification; Corenetwork protocols; Stage 3. Technical Specification V7.9.0, 3rd Gen-eration Partnership Project, Technical Specification Group Core Net-work and Terminals, October 2007. 12

[14] 3GPP TS 25.101. User Equipment (UE) radio transmission and re-ception (FDD). Technical Specification V7.9.0, 3rd Generation Part-nership Project, Technical Specification Group Radio Access Network,October 2007. 67

[15] 3GPP TS 25.102. User Equipment (UE) radio transmission and re-ception (TDD). Technical Specification V7.8.0, 3rd Generation Part-nership Project, Technical Specification Group Radio Access Network,October 2007. 67

[16] 3GPP TS 25.201. Physical layer - general description. TechnicalSpecification V7.4.0, 3rd Generation Partnership Project, TechnicalSpecification Group Radio Access Network, October 2007. 43

[17] 3GPP TS 25.211. Physical channels and mapping of transport chan-nels onto physical channels (FDD). Technical Specification V7.2.0,3rd Generation Partnership Project, Technical Specification GroupRadio Access Network, May 2007. 44

[18] 3GPP TS 25.212. Multiplexing and channel coding (FDD). TechnicalSpecification V7.5.0, 3rd Generation Partnership Project, TechnicalSpecification Group Radio Access Network, May 2007. 46, 50

[19] 3GPP TS 25.213. Spreading and modulation (FDD). TechnicalSpecification V7.2.0, 3rd Generation Partnership Project, TechnicalSpecification Group Radio Access Network, May 2007. 57

[20] 3GPP TS 25.214. Physical layer procedures (FDD). Technical Spec-ification V7.5.0, 3rd Generation Partnership Project, Technical Spec-ification Group Radio Access Network, May 2007. 58, 97, 122

[21] 3GPP TS 25.221. Physical channels and mapping of transport chan-nels onto physical channels (TDD). Technical Specification V7.3.0,

Bibliography 177

3rd Generation Partnership Project, Technical Specification GroupRadio Access Network, May 2007. 44, 62

[22] 3GPP TS 25.222. Multiplexing and channel coding (TDD). TechnicalSpecification V7.3.0, 3rd Generation Partnership Project, TechnicalSpecification Group Radio Access Network, May 2007. 46

[23] 3GPP TS 25.223. Spreading and modulation (TDD). TechnicalSpecification V7.4.0, 3rd Generation Partnership Project, TechnicalSpecification Group Radio Access Network, June 2007. 57, 63

[24] 3GPP TS 25.224. Physical layer procedures (TDD). Technical Spec-ification V7.3.0, 3rd Generation Partnership Project, Technical Spec-ification Group Radio Access Network, May 2007. 58

[25] 3GPP TS 25.301. Radio Interface Protocol Architecture. TechnicalSpecification V7.3.0, 3rd Generation Partnership Project, TechnicalSpecification Group Radio Access Network, October 2007. 9

[26] 3GPP TS 25.302. Services Provided by the Physical Layer. TechnicalSpecification V7.5.0, 3rd Generation Partnership Project, TechnicalSpecification Group Radio Access Network, October 2007. 43

[27] 3GPP TS 25.304. User Equipment (UE) procedures in idle modeand procedures for cell reselection in connected mode. Technical Spec-ification V7.3.0, 3rd Generation Partnership Project, Technical Spec-ification Group Radio Access Network, October 2007. 12

[28] 3GPP TS 25.306. UE Radio Access Capabilities. Technical Specifi-cation V7.5.0, 3rd Generation Partnership Project, Technical Specifi-cation Group Radio Access Network, October 2007. 111, 113

[29] 3GPP TS 25.308. High Speed Downlink Packet Access (HSDPA);Overall Description; Stage 2. Technical Specification V7.4.0, 3rd Gen-eration Partnership Project, Technical Specification Group Radio Ac-cess Network, December 2007. 39

[30] 3GPP TS 25.321. Medium Access Control (MAC) protocol speci-fication. Technical Specification V7.6.0, 3rd Generation PartnershipProject, Technical Specification Group Radio Access Network, Octo-ber 2007. 30, 34, 97

[31] 3GPP TS 25.322. Radio Link Control (RLC) protocol specification.Technical Specification V7.4.0, 3rd Generation Partnership Project,Technical Specification Group Radio Access Network, October 2007.21

[32] 3GPP TS 25.323. Packet Data Convergence Protocol (PDCP) spec-ification. Technical Specification V7.6.0, 3rd Generation Partnership

178 Bibliography

Project, Technical Specification Group Radio Access Network, Octo-ber 2007. 16

[33] 3GPP TS 25.324. Broadcast/Multicast Control (BMC). TechnicalSpecification V7.1.0, 3rd Generation Partnership Project, TechnicalSpecification Group Radio Access Network, June 2006. 20

[34] 3GPP TS 25.331. Radio Resource Control (RRC) protocol speci-fication. Technical Specification V7.6.0, 3rd Generation PartnershipProject, Technical Specification Group Radio Access Network, Octo-ber 2007. 12

[35] 3GPP TS 25.425. UTRAN Iur interface user plane protocols forCommon Transport Channel data streams. Technical SpecificationV7.5.0, 3rd Generation Partnership Project, Technical SpecificationGroup Radio Access Network, October 2007. 37

[36] 3GPP TS 25.433. UTRAN Iub Interface Node B Application Part(NBAP) signalling. Technical Specification V7.6.0, 3rd GenerationPartnership Project, Technical Specification Group Radio Access Net-work, October 2007. 40, 124

[37] 3GPP TS 25.435. UTRAN Iub Interface User Plane Protocols forCommon Transport Channel Data Streams. Technical SpecificationV7.6.0, 3rd Generation Partnership Project, Technical SpecificationGroup Radio Access Network, October 2007. 40

[38] 3GPP TS 25.848. Physical Layer Aspects of UTRA High SpeedDownlink Packet Access. Technical Specification V4.0.0, 3rd Genera-tion Partnership Project, Technical Specification Group Radio AccessNetwork, April 2001. 113

[39] 3GPP TR 25.855. High Speed Downlink Packet Access; OverallUTRAN Description. Technical Report V5.0.0, 3rd Generation Part-nership Project, Technical Specification Group Radio Access Network,September 2001. 123

[40] 3GPP TS 26.073. Adaptive Multi-Rate (AMR) speech codec; C-source code. Technical Specification V7.0.0, 3rd Generation Partner-ship Project, Technical Specification Group Services and System As-pects, June 2007. 12, 121

[41] 3GPP TR 26.911. Codec(s) for Circuit-Switched (CS) MultimediaTelephony Service; Terminal Implementor’s Guide. Technical ReportV7.1.0, 3rd Generation Partnership Project, Technical SpecificationGroup Services and System Aspects, September 2006. 121

[42] 3GPP TS 33.102. 3G Security; Security Architecture. Technical

Bibliography 179

Specification V7.1.0, 3rd Generation Partnership Project, TechnicalSpecification Group Services and System Aspects, December 2006.12, 39

[43] 3GPP TS 34.108. Common Test Environments for User Equipment(UE) Conformance Testing. Technical Specification V7.2.0, 3rd Gen-eration Partnership Project, Technical Specification Group Radio Ac-cess Network, October 2007. 91, 93, 106

[44] M. Allman, V. Paxson and W. Stevens. TCP Congestion Control.Standards Track RFC2581, April 1999. http://www.ietf.org/rfc.

html. 71[45] M. Andrews and others. Providing Quality of Service over a Shared

Wireless Link. IEEE Communications Magazine, vol. 39, pp. 150–154,February 2001. 125

[46] Martin F. Arlitt and Carey L. Williamson. A Synthetic Work-load Model for Internet Mosaic Traffic. In Proceedings of the 1995Summer Computer Simulator Conference, pp. 24–26. Department ofComputer Science - University of Saskatchewan, 1995. 71

[47] Lajos Bajzik, Laszlo Korossy, Kari Veijalainen and CsabaVulkan. Cross-Layer Backpressure to Improve HSDPA Performance.In Proceedings of the IEEE 17th International Symposium on Per-sonal, Indoor and Mobile Radio Communications, PIMRC 2006, pp.1–5, Helsinki, Finland, September 2006. 129

[48] B. Baumgartner, M. Reinhardt, G. Richter and M. Bossert.Performance of Forward Error Correction for IEEE 802.16e. In 11thInternational OFDM Workshop, 2006. 102

[49] D. Bertsekas and R. Gallager. Data Networks, Second Edition.Prentice Hall, 1987. 9

[50] S. Blake, D. Black, M. Carlson, E. Davies, Z. Wang andW. Weiss. An Architecture for Differentiated Services. StandardsTrack RFC2475, Internet Engineering Task Force, December 1998.http://www.ietf.org/rfc.html. 18

[51] S. Bliudze, N. Billy and D. Krob. On optimal hybrid ARQ controlschemes for HSDPA with 16QAM. In Wireless And Mobile Comput-ing, Networking And Communications, 2005. (WiMob’2005), IEEEInternational Conference on, vol. 1, pp. 121–127, August 2005. 114

[52] C. Bormann, M. Degermark and H. Fukushima. RObust HeaderCompression (ROHC). Standards Track RFC3095, Internet Engineer-ing Task Force, July 2001. http://www.ietf.org/rfc.html. 16

http://www.ietf.org/rfc.html




180 Bibliography

[53] F. Brouwer, I. de Bruin, J.C. Silva, N. Souto, F. Cercasand A. Correia. Usage of Link-Level Performance Indicators forHSDPA Network-Level Simulations in E-UMTS. In Spread SpectrumTechniques and Applications, 2004 IEEE Eighth International Sym-posium on, pp. 844–848, Sydney, Australia, September 2004. 99, 104

[54] P. Decker and B. Walke. A General Packet Radio Service Proposedfor GSM. In GSM in a Future Competitive Environment, pp. 0–20,Helsinki, Finland, October 1993. 1

[55] S. Deering and R. Hinden. Internet Protocol, Version 6 (IPv6)Specification. Standards Track RFC2460, Internet Engineering TaskForce, December 1998. http://www.ietf.org/rfc.html. 71

[56] M. Degermark. Internet Protocol Header Compression. StandardsTrack RFC2507, Internet Engineering Task Force, February 1999.http://www.ietf.org/rfc.html. 16, 73

[57] R.C. Elliot and others. Scheduling Algorithms for the CDMA2000Packet Data Evolution. In Proceedings of IEEE Vehicular TechnologyConference, vol. 1, pp. 304–310, Vancouver, Canada, September 2002.124

[58] I. Forkel. Performance Evaluation of the UMTS Terrestrial RadioAccess Modes. Dissertation, Aachen University, Lehrstuhl fur Kom-munikationsnetze, November 2005, ISBN 3-86130-177-6. 86

[59] F. Frederiksen and T.E. Kolding. Performance and modeling ofWCDMA/HSDPA transmission/H-ARQ schemes. In Vehicular Tech-nology Conference, 2002. Proceedings. VTC 2002-Fall. 2002 IEEE56th, vol. 1, pp. 472–476, September 2002. 114

[60] P. Frenger, S. Parkvall and E. Dahlman. Performance Com-parison of HARQ with Chase Combining and Incremental Redundancyfor HSDPA. In Vehicular Technology Conference, 2001. VTC 2001-Fall. IEEE VTS 54th, vol. 3, pp. 1829–1833, Atlantic City, USA,October 2001. 115

[61] Cheng Peng Fu and Soung C. Liew. TCP Veno: TCP Enhance-ment for Transmission Over Wireless Access Networks. IEEE Journalon Selected Areas in Communications, vol. 21, pp. 216–228, 2003. 71

[62] Noriyuki Fukui. Study of Channel Quality Feedback in UMTS HS-DPA. In Proceedings of the 14th IEEE International Symposium onPersonal, Indoor and Mobile Radio Communications, PIMRC 2003,vol. 1, pp. 336–340, Beijing, China, September 2003. 104



Bibliography 181

[63] R. Gold. Optimal binary sequences for spread spectrum multiplex-ing. Information Theory, IEEE Transactions on, vol. 13, pp. 154–156,October 1967. 66

[64] A. Gougousis and M. Paterakis. Scheduling with QoS Supportfor Multirate Wireless Systems with Variable Channel Conditions. InProceedings of the 12th European Wireless Conference 2006. TechnicalUniversity of Crete, Greece, April 2006. 127

[65] Sangtae Ha, Injong Rhee and Lisong Xu. CUBIC: A New TCP-Friendly High-Speed TCP Variant. SIGOPS Operating Systems Re-view, vol. 42, no. 5, pp. 64–74, 2008, ISSN 0163-5980. 71

[66] B. Haberland, S. Bloch and V. Braun. 3G Evolution TowardsHigh Speed Downlink Packet Access. Alcatel Telecommunications Re-view, 4th Quarter 2003/1st Quarter 2004. 40

[67] S. Heier, C. Ellerbrock and M. Malkowski. UMTS Medium Ac-cess Control Quality of Service Scheduling. In Proceedings of PIMRC2002 - The 13th IEEE International Symposium on Personal, In-door and Mobile Radio Communications, Lisboa, Portugal, September2002. 120

[68] S. Heier and M. Malkowski. UMTS Radio Resource Managementby Transport Format Assignment and Selection. In Proceedings ofWPMC 2002 - The 5th International Symposium on Wireless Per-sonal Multimedia Communications, Honolulu, USA, October 2002.120

[69] R. Hinden and S. Deering. IP Version 6 Addressing Architec-ture. Standards Track RFC2373, Internet Engineering Task Force,July 1998. http://www.ietf.org/rfc.html. 19

[70] Harri Holma and Antti Toskala. WCDMA for UMTS, RadioAccess For Third Generation Mobile Communications. John Wiley &Sons, Ltd, 3rd Edition, 2004. 36

[71] Harri Holma and Antti Toskala. HSDPA/HSUPA for UMTS.John Wiley & Sons, Ltd, 2006. 104

[72] J.M. Holtzman. CDMA Forward Link Waterfilling Power Control.In Proceedings of IEEE Vehicular Technology Conference, vol. 3, pp.1663–1667, September 2000. 124

[73] J.M. Holtzman. Asymptotic Analysis of Proportional Fair Algo-rithm. In Proceedings of IEEE Personal Indoor Mobile Radio Com-munications (PIMRC), pp. F33–F37, September 2001. 124


182 Bibliography

[74] IEEE Std 802.16e-2005. IEEE Standard for Local and metropolitanarea networks, Part 16: Air Interface for Fixed and Mobile BroadbandWireless Access Systems. Technical Specification, 2005. 75, 101

[75] ISO/IEC 14496-2. Information technology - Coding of audio-visualobjects - Part 2: Visual, 2005. 121

[76] ITU-R Recommendation M.1225. Guidelines for evaluation ofradio transmission technologies for IMT-2000. Technical Recommen-dation, February 1997. 108

[77] ITU-R Recommendation M.1457. Detailed specifications of theradio interfaces of International Mobile Telecommunications-2000(IMT-2000). Technical Recommendation, April 2006. 101

[78] ITU-T Recommendation H.263. Video coding for low bit rate com-munication, January 2005. 121

[79] ITU-T Recommendation H.324. Terminal for low bit rate multi-media communication, September 2005. 121

[80] ITU-T Recommendation P.862. Perceptual evaluation of speechquality (PESQ), an objective method for end-to-end speech quality as-sessment of narrowband telephone networks and speech codecs, Febru-ary 2001. 121

[81] ITU-T Recommendation X.208. Specification of Abstract SyntaxNotation One (ASN.1), November 1988. 15

[82] ITU-T Recommendation X.690. Information technology - ASN.1encoding rules: Specification of Basic Encoding Rules (BER), Canon-ical Encoding Rules (CER) and Distinguished Encoding Rules (DER),July 2002. 15

[83] ITU-T Recommendation X.691. Information technology - ASN.1encoding rules: Specification of Packed Encoding Rules (PER), July2002. http://www.itu.int/. 15

[84] V. Jacobson. Compressing TCP/IP Headers for Low-Speed SerialLinks. Standards Track RFC1144, Internet Engineering Task Force,February 1990. http://www.ietf.org/rfc.html. 16

[85] Raj Jain, Dah-Ming Chiu and W. Hawe. A Quantitative MeasureOf Fairness And Discrimination For Resource Allocation In SharedComputer Systems. Computing Research Repository, 1998. 137

[86] A. Jalali and others. Data Throughput of CDMA-HDR a HighEfficiency-High Data Rate Personal Communication Wireless System.In Proceedings of IEEE Vehicular Technology Conference, vol. 3, pp.1854–1858, 2000. 124

http://www.itu.int/


Bibliography 183

[87] L-E. Jonsson, K. Sandlund, G. Pelletier and P. Kremer. RO-bust Header Compression (ROHC): Corrections and Clarifications toRFC 3095. Standards Track RFC4815, Internet Engineering TaskForce, February 2007. http://www.ietf.org/rfc.html. 16

[88] M. Jurvansuu, J. Prokkola, M. Hanski and P. Perala. HSDPAPerformance in Live Networks. In Communications, 2007. ICC ’07.IEEE International Conference on, pp. 467–471, Glasgow, Scotland,June 2007. 129

[89] Heikki Kaaranen and others. UMTS Networks - Architecture,Mobility and Services. John Wiley & Sons, Ltd, 2nd Edition, 2005. 9

[90] Tom Kelly. Scalable TCP: Improving Performance in HighspeedWide Area Networks. SIGCOMM Computer Communication Review,vol. 33, no. 2, pp. 83–91, 2003, ISSN 0146-4833. 71

[91] A. Kemper, M. Malkowski and I. Karacan. IPv6 PerformanceEvaluation on Dedicated Channels in UMTS. In Proceedings of Eu-ropean Wireless 2004, pp. 279–283, Barcelona, Spain, February 2004.ISBN 84-7653-846-4. 16, 121

[92] Ahmed K.F. Khattab and Khaled M.F. Elsayed. Channel-Quality Dependent Earliest Deadline Due Fair Scheduling Schemes forWireless Multimedia Networks. In Proceedings of the Seventh ACMModeling and Simulation of Wireless and Mobile Systems (MSWiM),Venice, Italy, October 2004. 126

[93] Troels E. Kolding. Link and System Performance Aspects of Pro-portional Fair Scheduling in WCDMA/HSDPA. In Vehicular Tech-nology Conference, 2003. VTC 2003-Fall. 2003 IEEE 58th, vol. 3, pp.1717–1722, Orlando, USA, October 2003. 104

[94] Troels E. Kolding and others. High Speed Downlink Packet Ac-cess: WCDMA Evolution. IEEE Vehicular Technology Society News,pp. 4–10, February 2003. 111, 120

[95] P. Leaves, D. Grandblaise, M. Malkowski, C. Hamacher andL. Vignali. OverDRiVE’s Approach for Spectrum Efficient RadioResource Management. In Proceedings of ITG-Workshop on Multira-dio Multimedia Communications (MMC 2003) Communication Tech-nology for Vehicles, pp. 9–26, Dortmund, Germany, February 2003.121

[96] Petri Mahonen, Rudolf Mathar and Peter Vary. Kommu-nikationstechnik fur das mobile Internet von morgen. RWTH-Themen,vol. 1, May 2007. 1


184 Bibliography

[97] M. Malkowski. Link-Level Comparison of IP-OFDMA (MobileWiMAX) and UMTS HSDPA. In Proceedings of the 18th AnnualIEEE International Symposium on Personal, Indoor and Mobile RadioCommunications (PIMRC 2007), pp. 5, Athens, Greece, September2007. ISBN 1-4244-1144-0. 105

[98] M. Malkowski. Performance Evaluation of Packet Switched Ser-vices in UMTS - Simulation Framework. http://www.comnets.

rwth-aachen.de/downloads/malkowski_src.tgz, 2011. 69, 141[99] M. Malkowski. Performance Evaluation of Packet Switched Services

in UMTS - Simulation Results. http://www.comnets.rwth-aachen.de/downloads/malkowski_sim.tgz, 2011. 82, 128, 141

[100] M. Malkowski and J. Chua. UTRAN Enhancements for Multicast.In Proceedings of HyWiN 2003, pp. 1–11, Turin, Italy, December 2003.121

[101] M. Malkowski and D. Claßen. Performance of Video TelephonyServices in UMTS using Live Measurements and Network Emulation.In Proceedings of IST Mobile Summit 06, pp. 5, Myconos, Greece,June 2006. 75, 121

[102] M. Malkowski and D. Claßen. Performance of Video TelephonyServices in UMTS using Live Measurements and Network Emulation.Wireless Personal Communications, Springer Journal on, vol. 46,no. 1, pp. 19–32, July 2008, ISSN 0929-6212. 79, 121

[103] M. Malkowski and S. Heier. Interaction between UMTS MACScheduling and TCP Flow Control Mechanisms. In Proceedings of2003 International Conference on Communication Technology, pp.1373–1376, Beijing, China, April 2003. ISBN 7-563-50686-1. 121

[104] M. Malkowski and S. Heier. Quality of Service Scheduling forUMTS. In Proceedings of European Personal Mobile CommunicationsConference 2003, pp. 529–533, Glasgow, United Kingdom, April 2003.ISBN 0 85296 753 5. 121

[105] M. Malkowski, A. Kemper and X. Wang. Performance ofScheduling Algorithms for HSDPA. In Proceedings of ChinaCom 2007,pp. 5, Shanghai, China, August 2007. ISBN 1-4244-1009-6. 120

[106] M. Malkowski, M. Schnick, M. Schinnenburg andM. Schmocker. Connection Admission Control in UMTS withrespect to Network Capacity and Quality of Service. In Proceedingsof the 11th European Wireless Conference 2005, vol. 2, pp. 491–497,Nicosia, Cyprus, April 2005. ISBN 3-8007-2886-9. 120

http://www.comnets.rwth-aachen.de/downloads/malkowski_src.tgz

http://www.comnets.rwth-aachen.de/downloads/malkowski_src.tgz

http://www.comnets.rwth-aachen.de/downloads/malkowski_sim.tgz

http://www.comnets.rwth-aachen.de/downloads/malkowski_sim.tgz

Bibliography 185

[107] M. G. Marsh. Policy Routing Using Linux. Sams, 2001. 79[108] Saverio Mascolo, Claudio Casetti, Mario Gerla, M. Y.

Sanadidi and Ren Wang. TCP Westwood: Bandwidth Estimationfor Enhanced Transport over Wireless Links. In MobiCom ’01: Pro-ceedings of the 7th annual international conference on Mobile comput-ing and networking, pp. 287–297, New York, USA, 2001. ACM, ISBN1-58113-422-3. 71

[109] M. Mathis, J. Mahdavi, S. Floyd and A. Romanow. TCP Selec-tive Acknowledgement Options. Standards Track RFC2018, October1996. http://www.ietf.org/rfc.html. 71

[110] Marc C. Necker and Andreas Weber. Impact of Iub Flow Con-trol on HSDPA System Performance. In Personal, Indoor and MobileRadio Communications, 2005. PIMRC 2005. IEEE 16th InternationalSymposium on, vol. 3, pp. 1703–1707, September 2005. 129

[111] Marc C. Necker and Andreas Weber. Parameter Selection forHSDPA Iub Flow Control. In Wireless Communication Systems, 2005.2nd International Symposium on, pp. 233–237, Siena, September 2005.129

[112] K. Nichols and S. Blake. Definition of the Differentiated ServicesField (DS Field) in the IPv4 and IPv6 Headers. Standards TrackRFC2474, Internet Engineering Task Force, December 1998. http:

//www.ietf.org/rfc.html. 18[113] Yoshiaki Ofuji, Akihito Morimoto, Sadayuki Abeta and

Mamoru Sawahashi. Comparison of Packet Scheduling Algorithmsfocusing on User Throughput in High Speed Downlink Packet Ac-cess. In Proceedings of the IEEE 13th International Symposium onPersonal, Indoor and Mobile Radio Communications, PIMRC 2002,vol. 3, pp. 1462–1466, Lisbon, Portugal, September 2002. 124

[114] OPTICOM GmbH. User Manual PESQ OEM. Manual Version 1.6,2003. 121

[115] OPTICOM GmbH. PRELIMINARY User Manual PEVQ OEM.Manual Version 0.9.2, 2005. 121

[116] V. Paxson and M. Allman. Computing TCP’s RetransmissionTimer. Standards Track RFC2988, November 2000. http://www.

ietf.org/rfc.html. 71[117] Vern Paxson. Empirically-Derived Analytic Models of Wide-Area

TCP Connections. In IEEE/ACM Transactions on Networking, pp.316–336. Lawrence Berkeley Laboratory and EECS Division - Uni-






186 Bibliography

versity of California, August 1994. ftp://ftp.ee.lbl.gov/papers/WAN-TCP-models.ps.Z. 71

[118] Vern Paxson and Sally Floyd. Wide-Area Traffic: The Failure ofPoisson Modeling. In IEEE/ACM Transactions on Networking, pp.226–244. Lawrence Berkeley Laboratory and EECS Division - Uni-versity of California, June 1994. ftp://ftp.ee.lbl.gov/papers/

WAN-poisson.ps.Z. 71[119] J. Postel. User Datagram Protocol. Standards Track RFC768, Au-

gust 1980. http://www.ietf.org/rfc.html. 71[120] J. Postel. Internet Protocol, Darpa Internet Program Protocol Spec-

ification. Standards Track RFC791, University of South California,September 1981. http://www.ietf.org/rfc.html. 71

[121] J. Postel. Transmission Control Protocol, Darpa Internet ProgramProtocol Specification. Standards Track RFC793, University of SouthCalifornia, September 1981. http://www.ietf.org/rfc.html. 71

[122] John G. Proakis. Digital Communications. McGraw-Hill, NewYork, 4th Edition, 1995. 83

[123] J. Reynolds and J. Postel. Assigned Numbers. Standards TrackRFC1700, Internet Engineering Task Force, October 1994. http:

//www.ietf.org/rfc.html. 18[124] B. De Schuymer and N. Fedchik. ebtables/iptables interaction on a

Linux-based bridge. http://ebtables.sourceforge.net, November2003. 79

[125] S. Shakkottai and A. L. Stolyar. Scheduling for Multiple FlowsSharing a Time-Varying Channel: The Exponential Rule. technicalreport, Bell Labs, December 2000. 127

[126] S. Shakkottai and A. L. Stolyar. Scheduling Algorithms for aMixture of Real Time and Non-Real Time Data in HDR. In Pro-ceedings of the 17th International Teletraffic Congress, pp. 793–804,Salvador da Bahia, Brazil, September 2001. 126

[127] Simon Shin, Chan koo Kang, Joung-Cheol Kim and Se-HyunOh. The Performance Comparison between WiBro and HSDPA. InProceedings of the Second International Symposium of Wireless Com-munication Systems, ISWCS 2005, pp. 346–350, Siena, Italy, Septem-ber 2005. 108

[128] J.C. Silva, N. Souto, A. Correia, F. Cercas and A. Rodrigues.Multipath interference canceller for high speed downlink packet accessin enhanced UMTS networks. In Spread Spectrum Techniques and Ap-

ftp://ftp.ee.lbl.gov/papers/WAN-TCP-models.ps.Z

ftp://ftp.ee.lbl.gov/papers/WAN-TCP-models.ps.Z

ftp://ftp.ee.lbl.gov/papers/WAN-poisson.ps.Z

ftp://ftp.ee.lbl.gov/papers/WAN-poisson.ps.Z






http://ebtables.sourceforge.net

Bibliography 187

plications, 2004 IEEE Eighth International Symposium on, pp. 609–612, September 2004. 108

[129] B. Sklar. Rayleigh Fading Channels in Mobile Digital Telecommu-nication Systems. IEEE Communications Magazine, pp. 90–100, July1997. 123

[130] M. Steppler. SDL Performance Evaluation Tool Class Li-brary (SPEETCL). http://www.comnets.rwth-aachen.de/intern/speetcl/release/doc/html/home.html, 2002. 71

[131] M. Steppler. SDL-PR to C++ code generator using the Com-Nets Class Library. http://www.comnets.rwth-aachen.de/intern/sdl2cncl/sdl2cncl_toc.html, 2002. 71

[132] A. L. Stolyar and K. Ramanan. Largest Weighted Delay FirstScheduling: large Deviations and Optimality. Annals of Applied Prob-ability, 2001. 125

[133] Pablo Tapia, Dan Wellington, Jun Liu and Yasmin Karimli.Practical Considerations of HSDPA Performance. In Proceedings ofthe 66th IEEE Vehicular Technology Conference, VTC Fall 2007, pp.111–115, Baltimore, USA, October 2007. 104

[134] U. Turke, M. Koonert, R. Schelb and C. Gorg. HSDPA Per-formance Analysis in UMTS Radio Network Planning Simulations.In Proceedings of Vehicular Technology Conference, 2004. VTC 2004-Spring. IEEE 59th, vol. 5, pp. 2555–2559, Milan, Italy, May 2004.119

[135] Peter Vary. Nachrichtensysteme I - Unterlagen zur Vorlesung.2000. 114

[136] J. Voigt, J. Deissner, J. Hubner, D. Hunold and S. Mobius.Optimizing HSDPA Performance in the UMTS Network PlanningProcess. In Vehicular Technology Conference, 2005. VTC 2005-Spring. 2005 IEEE 61st, vol. 4, pp. 2384–2388, Stockholm, Sweden,May 2005. 129

[137] B. Walke. Mobile Radio Networks - Networking, Protocols and Traf-fic Performance, Second Edition. John Wiley & Sons, Chichester, NewYork, Weinheim, Brisbane, Singapore, Toronto, 2002. 35

[138] B. Walke, P. Seidenberg and M. P. Althoff. UMTS - TheFundamentals. John Wiley & Sons, Ltd, Chichester, UK, December2003. 9

[139] WiMAX Forum. WiMAX Forum Mobile System Profile. TechnicalSpecification Release 1.0, WiMAX Forum, 2006. 103

http://www.comnets.rwth-aachen.de/intern/speetcl/release/doc/html/home.html

http://www.comnets.rwth-aachen.de/intern/speetcl/release/doc/html/home.html

http://www.comnets.rwth-aachen.de/intern/sdl2cncl/sdl2cncl_toc.html

http://www.comnets.rwth-aachen.de/intern/sdl2cncl/sdl2cncl_toc.html

188 Bibliography

[140] M. Wrulich, W. Weiler and M. Rupp. HSDPA Performance ina Mixed Traffic Network. In 67th IEEE Vehicular Technology Confer-ence, 2008. VTC Spring 2008, pp. 2056–2060, Singapore, May 2008.99

NON-PUBLIC REFERENCES

Oppositely to the bibliography, this list of sources references unpub-lished documents. These are mainly references to diploma and student

project theses with significant contribution to the simulation environmentused within this dissertation. The references with dates after 2002 have beensupervised by the author at Communication Networks (ComNets) ResearchGroup at RWTH Aachen University during the work on this dissertation.

[C1] Yashesh Buch. Modelling and Simulation of the Random AccessChannel in UMTS. Diploma Thesis, ComNets, February 2003. 73

[C2] Gianluca Caminiti. Performance Evaluation of different UMTSPhysical Layer Configurations. Diploma Thesis, ComNets, July 2002.73

[C3] Daniel Claßen. Experimental Performance Evaluation of VideoTelephony Services in UMTS including Analysis of potential Improve-ments at various Protocol Layers. Diploma Thesis, ComNets, January2006. 78

[C4] Michael Dittrich. Performance Evaluation of Deployment Sce-narios for UMTS Femto Base Station. Diploma Thesis, ComNets,January 2008. 74

[C5] Christian Ellerbrock. Simulative Performance Evaluation ofUMTS Medium Access Control Protocol for Common Transport Chan-nels. Diploma Thesis, ComNets, December 2001. 73

[C6] Alex Fernandez Yeste. Simulative Performance Evaluation ofMulticast Services in UMTS. Diploma Thesis, ComNets, June 2003.73

[C7] Sebastian Grabner. Simulative Performance Evaluation of theUMTS Radio Link Control Protocol. Diploma Thesis, ComNets,March 2001. 73

[C8] Kerrin Gunter. Radio Resource Control Protocol for the UMTS-Simulator URIS. Diploma Thesis, ComNets, December 2001. 73

[C9] Dirk Heinrichs. Simulative Performance Evaluation of ARQ-Protocols for UMTS. Diploma Thesis, ComNets, November 2001. 73

190 Non-Public References

[C10] Georg Himmrich. Interconnection between MAC-d and MAC-c En-tity for multiple Mobiles in URIS. Student Project Thesis, ComNets,January 2004. 73

[C11] Ibrahim Karacan. Simulative Performance Evaluation of IPv6 inUMTS. Master Thesis, ComNets, April 2003. 73

[C12] Burge Kurt. Implementation and Evaluation of the UMTS HSUPA(High Speed Uplink Packet Access). Master Thesis, ComNets, Septem-ber 2007. 2, 73

[C13] Alexander Latsch. Design and Implementation of a Simulator forUMTS. Diploma Thesis, ComNets, June 1999. 72

[C14] Matthias Malkowski. Implementation Framework of the UMTSMedium Access Control Layer. Student Project Thesis, ComNets, July2001. 73

[C15] Matthias Malkowski. Simulative Performance Evaluation ofUMTS Medium Access Control Protocol for Dedicated TransportChannels. Diploma Thesis, ComNets, January 2002. 73

[C16] Enric Navarro. Implementation and Evaluation of HSDPA inURIS. Diploma Thesis, ComNets, May 2006. 73

[C17] Francesco Pellicano. Simulative Performance Evaluation of theDownlink Shared Channel in UMTS. Diploma Thesis, ComNets, July2003. 73

[C18] Jan-Oliver Rock. Simulative Performance Evaluation of IP basedServices over UMTS. Diploma Thesis, ComNets, May 2001. 71

[C19] Michael Salzmann. Methods for Performance Optimization of aUMTS Network at high Density Traffic Locations. Diploma Thesis,ComNets, March 2004. 74

[C20] Klaus Sambale. Development of a Code Generator for the ASN.1based RRC Protocol of UMTS. Student Project Thesis, ComNets,August 2004. 74

[C21] Alicia Sanchez Mora Moreno. Smart Handover Decisionsin Heterogeneous Networks using Location based Link Information.Diploma Thesis, ComNets, November 2006. 74

[C22] Michael Schnick. Simulative Performance Evaluation of Connec-tion Admission Control Algorithms in UMTS. Diploma Thesis, Com-Nets, November 2004. 74

[C23] Helmut Seidler. Performance Evaluation of TCP/IP over UMTS.Diploma Thesis, ComNets, January 2001. 73

Non-Public References 191

[C24] Xiaohua Wang. Implementation and Evaluation of UMTS SharedChannels with Fast Scheduling. Master Thesis, ComNets, August2006. 73

192 Non-Public References

ACKNOWLEDGMENT

The present work has been developed during my activities as research assis-tant at the Communication Networks (ComNets) Research Group at RWTHAachen University.

I am deeply grateful to my advisor Prof. Dr.-Ing. Bernhard Walke forthe detailed guidance, fruitful discussions, critical review of this thesis andthe experienced freedom in research during the elaboration of this work.Special thanks also go to Prof. Dr.-Ing. Peter Vary from the Institute ofCommunication Systems and Data Processing for his kind support as sec-ond examiner of this thesis.

Best wishes go to all the great colleagues I was able to work with duringmy time as a research assistant at ComNets. Many thanks for the creativeand inspiring working atmosphere.

Sincere thanks go to my diploma students and student project workerswho contributed to my research during their time at ComNets: YasheshBuch, Daniel Claßen, Michael Dittrich, Georg Himmrich, Ibrahim Karacan,Burge Kurt, Alicia Sanchez Mora Moreno, Enric Navarro, Francesco Pelli-cano, Michael Salzmann, Klaus Sambale, Michael Schnick, Xiaohua Wang,Alex Fernandez Yeste and Aydar Zimaliev.

Finally, I would like to thank my family and friends for their supportand understanding during the elaboration of this thesis.

Nuremberg, January 2012 Matthias Malkowski

dissertation malkowski[1]

Documents