jarirasinen evaluation of multi-repeater · pdf filejarirasinen evaluation of multi-repeater...

78
JARI RASINEN EVALUATION OF MULTI-REPEATER PERFORMANCE IN WCDMA NETWORKS Master of Science Thesis Examiners: Prof. Jukka Lempiäinen M.Sc. Panu Lähdekorpi Examiners and topic approved in the Faculty of Computing and Electrical Engineering Council meeting on 7th of April 2010

Upload: buithuy

Post on 07-Feb-2018

218 views

Category:

Documents


0 download

TRANSCRIPT

JARI RASINENEVALUATION OF MULTI-REPEATER PERFORMANCE INWCDMA NETWORKSMaster of Science Thesis

Examiners: Prof. Jukka LempiäinenM.Sc. Panu Lähdekorpi

Examiners and topic approved in theFaculty of Computing and ElectricalEngineering Council meeting on 7th ofApril 2010

II

ABSTRACT

TAMPERE UNIVERSITY OF TECHNOLOGYMaster’s Degree Programme in Signal Processing and Communications EngineeringRasinen, Jari: Evaluation of Multi-Repeater Performance in WCDMANetworksMaster of Science Thesis, 67 pagesMay 2010Major: Communications EngineeringExaminers: Professor Jukka Lempiäinen, M.Sc. Panu LähdekorpiKeywords: Radio network planning, HSDPA, WCDMA, multi-repeater, cellular network

Constantly increasing demand for mobile data services has caused pressure for ope-rators to enhance the coverage of third generation networks from the densely popu-lated city centers also to the suburban and rural regions. This is where radio networkplanning aims to design network implementations with maximum performance andminimum costs. Choosing the right equipment for the network infrastructure playsa key role in this optimization task.

Third Generation Partnership Project standardized UMTS uses interference limitedcode division multiple access scheme in the radio interface. In the downlink, the th-roughput has been further improved with techniques that are called together ’highspeed downlink packet access’, in which the interference aspect plays even biggerrole. Repeaters have potential to provide a cost-efficient way to improve the networkperformance. In this thesis, performance of multiple air-to-air repeaters parallel in asingle cell was studied. Firstly, network coverage was studied with varying number ofrepeaters in rural macrocellular environment. Secondly, the received signal qualitywas studied as the ratio of desired signal and interference and noise in suburban en-vironment with different number of repeaters, and with different repeater distancesto the base stations. In addition, an alternative repeater configuration was studiedin both the scenarios to see the effect of repeater antenna locations and directionson the system performance.

The network coverage was noticed to improve in rural macrocell with multiple repea-ters, but the number of repeaters was limited. In suburban macrocell, the studiedmulti-repeater configurations provided minor improvements in the received signalquality. Both, the number of repeaters, and the distance to the serving base stationwere shown to have effect on the multi-repeater performance. Additionally, it wasnoticed that the alternative repeater configuration provided poor coverage improve-ment in the rural scenario, but in suburban scenario it provided the best results.

III

TIIVISTELMÄ

TAMPEREEN TEKNILLINEN YLIOPISTOSignaalinkäsittelyn ja tietoliikennetekniikan koulutusohjelmaRasinen, Jari: Usean toistimen suorituskyvyn arviointi WCDMA-verkossaDiplomityö, 67 sivuaToukokuu 2010Pääaine: TiedonsiirtotekniikkaTarkastajat: Professori Jukka Lempiäinen, DI Panu LähdekorpiAvainsanat: radioverkkosuunnittelu, toistin, solukkoverkko, HSDPA, WCDMA

Mobiilien datapalvelujen alati kasvanut kysyntä on lisännyt operaattoreiden pai-netta parantaa kolmannen sukupolven matkaviestinverkkojen kuuluvuutta tiheästiasutettujen kaupunkien keskusta-alueilta lähiöihin ja harvaan asuteteuille seuduil-le. Radioverkkosuunnittelun lähtökohtana on suunnitella verkkoja, jotka tarjoavatmaksimaalisen suorituskyvyn mahdollisimman pienin kustannuksin. Tämän vuoksierityisesti laitteiston valinta verkkoinfrastruktuurissa on tärkeässä osassa radiover-kon kustannusten optimoinnissa.

Euroopassa kolmannen sukupolven matkaviestinverkoksi on valittu UMTS, jonkakäyttämä koodijakoiseen monipääsytekniikkaan pohjautuva radiorajapinta on häi-riörajoitettu. Tätä tekniikkaa on edelleen parannettu alalinkin siirtonopeuksien osal-ta usein parannuksin, joita yhdessä kutsutaan nimellä ’high speed downlink packetaccess’. Suunniteltaessa nopeampia siirtonopeuksia tukevia verkkoja, häiriösuun-nittelu on entistä keskeisemmässä osassa. Oikein hyödynnettynä toistinten avullaverkon suorituskykyä on mahdollista parantaa kustannustehokkaasti. Tämän työntarkoituksena oli tutkia monen rinnakkaisen toistimen vaikutusta verkon peittoonja alalinkin signaalin laatuun. Verkon peittoa on ensiksi tutkittu harvaan asutullamakrosolualueella usealla eri määrällä toistimia asennettuna jokaisen solun palve-lualueelle. Toiseksi, radiorajapinnan kautta kulkevan signaalin laatua tutkittiin lä-hiöaluetopologiassa vaihdellen toistimien määriä soluissa ja toistimien etäisyyksiätukiasemiin nähden. Lisäksi vaihtoehtoisella tointinkonfiguraatiolla tutkittiin toisti-men antennien sijoittelun ja suuntauksen vaikutusta verkon suorituskykyyn.

Työn tuloksena huomattiin usean rinnakkaisen toistimen parantavan verkon peit-toa harvaanasutulla alueella, mutta käytettävien toistimien lukumäärällä on selväyläraja. Lähiöalueen makrosolussa usean toistimen käyttö ei tuonut huomattavaa pa-rannusta käytettyihin mittareihin, vaikkakin sekä toistimien lukumäärällä että etäi-syydellä palvelevaan tukiasemaan oli selvä vaikutus monen toistimen tarjoamaansuorituskykyyn. Toistimen antennien sijoittelulla ja suuntauksella huomattiin ole-van vaikutusta sekä verkon peittoon että vastaanotetun signaalin laatuun.

IV

PREFACE

This Master of Science Thesis was written between autumn 2009 and spring 2010 inmy spare time to complete my M.Sc. degree at Tampere University of Technology(TUT). The thesis work was done for the Radio Network Group that operates underthe Department of Communications Engineering.

I would like to thank professor Jukka Lempiäinen and M.Sc. Panu Lähdekorpifor providing me the topic for this thesis. Especially Panu’s effort guiding and help-ing me throughout the work deserves an extra acknowledgement - his knowledge ofthe simulator tool was a necessity to me to be able to carry out the research workoutside the university environment. I would also like to express my gratitude to mycurrent employer, Bitville Oy, and especially the managing director Antti Keuru-lainen, the chief operating officer Janne Viskari, and project manager Jonni Purho,for the encouraging atmosphere to finish my M.Sc. degree aside the work.

A very special thanks is dedicated to my parents Risto and Raili, and to my long-term girlfriend Corina Maiwald, who have been supporting me through my wholestudying time period at TUT.

Tampere, April 14, 2010

Jari [email protected]

Näyttämönkatu 6 C 3433720 Tampere, Finland

Tel. +358 45 671 0428

V

CONTENTS

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12. Introduction to Cellular Networks . . . . . . . . . . . . . . . . . . . . . . . 32.1 Evolution of Cellular Networks . . . . . . . . . . . . . . . . . . . . . . 32.2 Multiple Access and Duplexing . . . . . . . . . . . . . . . . . . . . . . 52.2.1 Time Division Multiple Access . . . . . . . . . . . . . . . . . . . . 52.2.2 Frequency Division Multiple Access . . . . . . . . . . . . . . . . . 62.2.3 Code Division Multiple Access . . . . . . . . . . . . . . . . . . . . 62.2.4 Duplexing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.3 Cellular Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.4 Radio Propagation in Cellular Systems . . . . . . . . . . . . . . . . . . 92.4.1 Propagation Environments . . . . . . . . . . . . . . . . . . . . . . 92.4.2 Propagation Models . . . . . . . . . . . . . . . . . . . . . . . . . . 92.4.3 Multipath Propagation . . . . . . . . . . . . . . . . . . . . . . . . 112.4.4 Slow and Fast Fading . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.5 WCDMA for UMTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.5.1 UMTS Reference Model . . . . . . . . . . . . . . . . . . . . . . . . 132.5.2 WCDMA Air Interface . . . . . . . . . . . . . . . . . . . . . . . . 142.5.3 Power Control in WCDMA . . . . . . . . . . . . . . . . . . . . . . 162.5.4 Handovers in WCDMA . . . . . . . . . . . . . . . . . . . . . . . . 16

2.6 HSDPA for UMTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172.7 WCDMA Frequency Variants . . . . . . . . . . . . . . . . . . . . . . . 19

3. Radio Network Planning for UMTS . . . . . . . . . . . . . . . . . . . . . . 203.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.2 Planning Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203.2.1 Dimensioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.2.2 Detailed Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.2.3 Optimization and Monitoring . . . . . . . . . . . . . . . . . . . . . 23

3.3 Planning Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.4 Planning for HSDPA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.5 Performance Indicators for HSDPA Air Interface . . . . . . . . . . . . 25

4. Repeaters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294.1 Introduction to Cellular Repeaters . . . . . . . . . . . . . . . . . . . . 294.2 Repeater Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304.2.2 Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314.2.3 Repeater Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

4.3 The Effects of Repeaters on Transmission Path . . . . . . . . . . . . . 32

VI

4.3.1 Thermal Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324.3.2 Transmission Delay . . . . . . . . . . . . . . . . . . . . . . . . . . 344.3.3 Repeaters in WCDMA network . . . . . . . . . . . . . . . . . . . 35

5. Numerical Analysis for System Level Performance Evaluation . . . . . . . . 365.1 Overview for Analysis Method . . . . . . . . . . . . . . . . . . . . . . 365.2 Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395.2.1 Rural Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415.2.2 Suburban Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . 43

5.3 Noise Rise in the Uplink . . . . . . . . . . . . . . . . . . . . . . . . . . 455.4 Theoretical Behavior of Multiple Repeaters . . . . . . . . . . . . . . . 47

6. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 496.1 Rural Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 496.2 Suburban Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 536.3 Error Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

7. Conclusions and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 62References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

VII

LIST OF ABBREVIATIONS AND SYMBOLS

ABBREVIATIONS

16-QAM 16 quadrature amplitude modulation

1G The first generation cellular system

2G The second generation cellular system

BER Bit error ratio

BPL Building penetration loss

BS Base station

BTS Base transceiver station

CAPEX Capital expenses

CDF Cumulative distribution function

CDMA Code division multiple access

CN Core network

CPICH Common pilot channel

CQI Channel quality indicator

DS-SS Direct sequence spread spectrum

EDT Electrical downtilt

EIRP Effective isotropically radiated power

FDD Frequency division duplexing

FDMA Frequency division multiple access

GSM Global System for Mobile Communications

HS-DSCH High speed downlink shared channel

HSPA High Speed Packet Access

KPI Key performance indicator

ME Mobile equipment

VIII

MHA Mast head amplifier

MS Mobile station

NMT Nordic Mobile Telephone

OPEX Operating expenses

OVSF Orthogonal variable spreading factor

PC Power control

QoS Quality of service

RAN Radio access network

RNC Radio network controller

RSCP Received signal code power

SfHO Softer handover

SHO Soft handover

TACS Total Access Communications System

TDD Time division duplexing

TDMA Time division multiple access

UE User equipment

UMTS Universal Mobile Telecommunications System

USIM UMTS subscriber identity module

UTRAN UMTS terrestrial radio access network

IX

SYMBOLS

α Orthogonality factor

ηUL Uplink load factor

γ Propagation environment dependent parameter for orthogonality model

λ Wavelength

a1 Propagation environment dependent parameter for orthogonality model

a2 Propagation environment dependent parameter for orthogonality model

C Area correction factor for COST-231-Hata model

c Speed of light

Eb/N0 Energy-per-bit to spectral noise density ratio

EFB Effective base station noise factor

f Carrier frequency

FB Base station noise figure

FR Repeater noise figure

G Geometry factor

GA Base station antenna gain

GD Repeater donor antenna gain

GR Repeater gain

GS Repeater service antenna gain

GT Total repeater path gain

hBS Base station antenna height

hMS Mobile station antenna height

I Received interference power

i Other-to-own-cell interference ratio

Ioth Received power from the other cells

X

Iown Received power from own cell

Iown Received power from the own cell

IM Interference margin

k Boltzman’s constant

LD Direct link loss

LP Path loss between repeater and BS

LR Repeater link loss

LS Path loss between MS and repeater

No Noise power at the base station output

NiB Noise power at the base station input

NTH Thermal noise

Pi Total transmit power of the serving cell

pi Transmit power of HS-DSCH

PN Noise power

Pn Transmit power of other cell

PHS−SCCH High speed shared channel power

PHSDPA Received HSDPA power

prx Received power

Ptot Total received power including noise and interference

r Distance between MS and BS

Rc Chip rate

Rbit User bit rate

S Received HSDPA power

So Signal power at base station output

SiB Signal power at base station input

XI

T Noise temperature of a component

TaB Base station antenna noise temperature

TaR Repeater service antenna noise temperature

TeB Inherent noise temperature of the base station

TeR Inherent noise temperature of the repeater

W Bandwidth

W Repeater weight factor

1

1. INTRODUCTION

Evolution in cellular communication has been rapid within the past decades, movingstrongly towards packet switched data optimized networking. For example, recentlythe usage of mobile data has experienced a major boost. This movement has puta great demand on densely installed base station sites in order to provide highcapacity and high mobility to the end-users. Currently, high speed packet accesstechnologies (HSDPA/HSUPA, in downlink and uplink, respectively) cover urbancity center regions, moving towards the suburbs, and with new lower frequencyband refarming also to the rural regions.

Cost-effective capacity and coverage improvement strategies are needed to main-tain the network profitable. This leads to a situation where deploying more basestation sites to sparsely populated rural regions is not economically feasible for theoperators, and other solutions should be investigated. One solution could be low-costradio relaying nodes, repeaters.

The purpose of this thesis work is to study the system level performance of mul-tiple parallel installed repeaters within the same cell. The work is divided into twomacrocellular scenarios: rural and suburban. The simulations were carried out withstatic system level simulator as a framework. The studied key aspects were the pilotchannel received signal code power (RSCP) coverage and signal quality in termsof signal-to-interference-and-noise ratio (SINR). In each scenario, the results werecompared to simulations without repeaters. Also, average RSCP and SINR valueswithin the studied area were compared in respect to the number of repeaters. It isshown that in rural areas with sparsely populated base station sites the upper limitfor coverage improvement is found already with a few repeaters, if the uplink qual-ity is wanted to maintain in an adequate level in terms of noise rise. Similarly, it isshown for the suburban scenario that adding more than one repeater in the cell doesnot necessarily provide additional gain in SINR, but with well planned configurationmultiple repeaters can be utilized if needed. Moreover, with an alternative repeaterconfiguration the effect of repeater donor antenna location and service antenna di-rection on the system performance was studied in the both scenarios. It was noticedthat the alternative configuration provided poor coverage improvement in the ruralscenario in respect to the other repeater configurations. However, in the suburbanscenario it provided good SINR improvement when the repeaters were located at

1. Introduction 2

longer distances from the serving base stations. Although the research was carriedout with Universal Mobile Telecommunications System (UMTS) specific parametersand metrics, the results should be applicable also for other interference limited codedivision multiple access systems.

The content of this thesis is divided into seven chapters. The Chapter 2 firstprovides a glance to the evolution of cellular systems, and step-by-step leads tothe theory behind the radio access technology that is used in high speed downlinkpacket data connections. Chapter 3 goes through the radio network planning processfor UMTS network, and introduces the performance metrics that are used in thesimulations that were carried for this thesis work. In Chapter 4, definition for therepeater is given, and the properties of different repeater equipment components areinvestigated. The work that was done for the this thesis itself begins in Chapter5 by introducing the simulator environment and the created scenarios. In the lastchapters, the results for the simulations are given with the error analysis, and finallythe conclusions are made based on the simulation results.

3

2. INTRODUCTION TO CELLULAR

NETWORKS

Some basic concepts in all the cellular systems are the same, and same technicalapproaches are used in many different systems. Thus, understanding the basic prin-ciples of cellular communications and radio propagation related issues is beneficialregarding the work that was done in this thesis. Moreover, this thesis concentrates onUMTS system and the high speed downlink packet access radio link aspects whichare introduced later in this chapter. Hence, also some wideband CDMA relatedconcepts are introduced more in detail as an introduction to the UMTS.

2.1 Evolution of Cellular Networks

The first steps for cellular networking took place in the 1940s when AT&T startedoperating the first commercial car-borne telephony service. Within the same decadeAT&T also introduced the cellular concept of reusing radio frequencies, which cre-ated the basis for cellular communications. In spite of the limitations of the serviceand bulky, power-hungry equipment, cellular systems were deployed in many coun-tries during the 1950s and 1960s, though with modest amount of users that wascounted in thousands. [1]

Mobile cellular communication got a real boost after several decades in the 1980s,when the first mobile cellular systems started emerge driven by the telecommunica-tions industry. In that time there was no dominant standard, but instead numeroussystems were introduced around the world. The most widespread systems werein 1981 introduced Scandinavian Nordic Mobile Telephone (NMT) and AdvancedMobile Phone Service (AMPS) in America. Also Total Access CommunicationsSystem (TACS) that was deployed in the UK is one of the major first generation(1G) systems. Common to all of these systems, for the first time the coverage ar-eas were geographically divided into smaller sectors, called cells. The cell rangesvaried between a couple of kilometers up to tens of kilometers. Besides the newcellular network structure, 1G systems started to provide additional services, suchas billing and roaming. As an open standard NMT gained some popularity alsooutside Europe, like in the Middle East, Baltic region and Russia, and Asia. [1, 2, 3]

Problems with the first generation analog mobile systems were, for example,cross-talk between users, bad voice quality and, still, bulky equipment. In the first

2. Introduction to Cellular Networks 4

implementations the total lack or poorly implemented encryption made it possibleto eavesdrop the phone calls with right scanning equipment. Although the 1Gsystems kept evolving and improving, the higher subscriber amounts led to capacityproblems in the networks. In addition, none of the systems were compatible betweeneach other. In the 1980s the telecommunications industry started moving towardsdevelopment of digital second generation (2G ) systems. [1, 2]

In Europe, a common mobile standard was found to be Global Systems for Mo-bile Communications (GSM). The purpose of GSM was to provide a single standardin Europe supporting seamless speech services throughout Europe in terms of in-ternational roaming [4]. GSM was first developed in the 1980s; from 1982 to 1985discussions were held in the GSM group to decide between building an analog or adigital system [3]. The system was built on narrowband TDMA solution with Gaus-sian Minimum Shift Keying modulation (GMSK). Digitalization of both equipmentand the system provided better voice call quality, more capacity in terms of speechcoding, and new services. Especially SMS (short messaging system) ended up be-coming a real killer application. The deployment of new digital GSM system startedcommercially in the early 1990s. Other main 2G standards are digital AMPS (D-AMPS), code division multiple access (CDMA) based IS-95, and personal digitalcellular (PDC). Similar to GSM, in D-AMPS multiple access is handled in timedomain. Using the existing AMPS channels it allowed a smooth transition fromanalog AMPS to the new digital version. The capacity was increased in respectto AMPS by dividing each 30 kHz channel pair into three time slots and digitallycompressing the voice data, leading to tripled call capacity in a single cell. IS-95took a completely different approach choosing CDMA as the access technique beingone of the first systems utilizing CDMA in commercial networks. [3, 1]

In the 1990s, the growth of the packet switched data traffic in fixed networks wasalso noted in the design requirements of the third generation (3G) radio networkstandards. The requirements for 3G systems are defined by ITU in InternationalMobile Telecommunications-2000 (IMT-2000), which covers today all the 3G stan-dards, such as UMTS, cdma2000, and WiMAX.

In the present situation the development of 3G systems is centralized under twostandardization organizations: Third Generation Partnership Project (3GPP) andThird Generation Partnership Project 2 (3GPP2). 3GPP is responsible for thestandardization Universal Mobile Telecommunications System (UMTS), that utilizesWideband CDMA (WCDMA), but also continues to development of the ETSI basedGSM standards as one of the design targets of UMTS was intersystem compatibilityto legacy GSM. 3GPP2 in its turn continued the development of IS-95 compatiblecdma2000, since the deployment of IS-95 was already comprehensive, and gatheredalso the Asian standardization bodies in the same forum. The first specifications for

2. Introduction to Cellular Networks 5

WCDMA based UMTS was released in late 1999 (these specifications are generallycalled Release 99), and around the same time also cdma2000 was introduced. [2, 5, 6]

Today the evolution in 3G networks is happening in radio access and core net-works. In radio access, higher order modulation and adaptive coding schemes areused for providing higher throughput for data service subscribers in the network.In 3GPP this evolution is called high speed downlink packet access (HSDPA) andhigh speed uplink access (HSUPA) - together called as high speed packet access(HSPA). In 3GPP2 the corresponding evolution is called cdma2000 EV-DO. In thecore network side the evolution is on its way to All-IP based packet switched com-munication. In the same time with 3G evolution, the next major step is to introducecommercially the long-term evolution radio access (LTE) with evolved packet core(EPC), which should provide the common framework for a truly world wide stan-dard, supported by both, 3GPP and 3GPP2, organizations. Yet, it is still good toremember, that the most widely spread system currently is GSM [7], and the mar-ket growth is only now moving from 2G systems to 3G systems in many developingmarkets, such as Africa.

2.2 Multiple Access and Duplexing

There are different approaches to allocate resources to multiple simultaneous userswithin the same medium, and in this context the term ’multiple access’ is used.The orthogonality between the signals from different users is especially importantin radio communications in order to avoid interference and crosstalk phenomena.The chosen multiple access scheme in the radio system affects also to radio resourceand mobility management due to the limitations and benefits within each multipleaccess scheme. In this section the three most common multiple access schemes aredescribed together with the two basic duplexing schemes.

2.2.1 Time Division Multiple Access

Time division multiple access (TDMA) shares the transmission medium in timedomain between the subscribers. The basic structure of TDMA consists of frameswhich are transmitted as bursts in the medium. A frame is divided in turn intosmaller pieces called time slots that carry the payload traffic from multiple users.The principle of TDMA is illustrated in Figure 2.1.

The basic requirement of TDMA utilization is frame level synchronization, whichis done with reference bursts. Also additional preamble bits are used in the beginningof each traffic burst to allow the receiver to acquire the correct carrier phase. Typicalto TDMA is also a requirement for guard bands in time domain between the timeslots due to the delay variations from different users within the network. In practice

2. Introduction to Cellular Networks 6

this means that the time slots occupy the medium longer than it is required fortransmitting the payload, which is inefficient, but required for proper functioning ofthe system. [8]

Time

Frequency

Amplitude

Figure 2.1: Time division multiple access concept.

2.2.2 Frequency Division Multiple Access

In frequency division multiple access (FDMA) user separation is done in frequencydomain (Figure 2.2). For example, in GSM the frequency separation of adjacentchannels is 200 kHz, but additional guard band is used in order to avoid interfer-ence. Similar to TDMA, FDMA is also able to provide a fully orthogonal trans-mission medium between the users, if proper guard bands are applied, and thereare not strong interfering signals blocking the channels. The drawback of FDMA isthe linearity requirement for the transmission medium, which may lead to complexequalization schemes in the receiver design. [8]

Time

Frequency

Amplitude

Figure 2.2: Frequency division multiple access concept.

2.2.3 Code Division Multiple Access

Code division multiple access (CDMA) separates users in a network with orthogonalcodes that have as low cross-correlation between each other as possible. Because of

2. Introduction to Cellular Networks 7

division in code domain each user can be deployed in the same frequency band, andfrequency re-use is not required anymore in operator’s network. CDMA concept isillustrated in Figure 2.3

Code

Time

Frequency

Figure 2.3: Code division multiple access concept.

Using CDMA is beneficial on channels which suffer from time-varying multipathdistortion or jamming of a narrowband interfering signal. The problems with CDMAare the near-far problem and partial correlation problem. The near-far problemoccurs when two transmitters from different distances are transmitting to the samereceiver, for example two mobiles (one in the cell edge and one near the base station)that are transmitting to the base station. The signal is received weaker in the receiverfrom the transmitter located further away from the receiver, and in the worst caseit will not be detected. To cope with the near-far problem, power controlling isused in CDMA systems, so the transmitter further away from the receiver usesmore power to the transmission. In TDMA and FDMA the near-far problem doesnot exist in that extend as in CDMA because either one transmitter is active ata time (TDMA), or the signals may be filtered by bandpass filters with very largestop-band attenuations. As the amount of users increases in the network, also theinterference level rises, which leads to degradation in capacity and coverage - thisis called soft capacity and cell breathing. In the system, capacity between users isfixed to the time and frequency slot allocations, which are limited resources, but inCDMA the limitation is ’softer’ and depends on the number of available codes, andon the interference originating from the other subscriber in the cell.[8]

2.2.4 Duplexing

Duplexing is the method to separate the traffic in uplink and downlink directions,respectively. The most common duplexing schemes are frequency division duplex-ing (FDD) and time division duplexing (TDD). In FDD the uplink and downlinkchannels are allocated in their own frequency bands, which provides the maximumcapacity for the both directions in the sense of utilizing the whole allocated spectrum

2. Introduction to Cellular Networks 8

per direction. In TDD the uplink and downlink traffic shares the same frequencyband in time domain. Due to these differences it is logical to say that FDD suitsbetter data oriented packet switched traffic, and TDD for circuit switched trafficwith constant bit rate (such as voice calls).

Closely related to duplexing schemes are the concepts of paired and unpairedspectrum allocations. As TDD requires only one frequency band, it is allocatedin unpaired bands, whereas FDD always requires two paired bands for the uplinkand downlink. When an operator starts planning a new network, it has to considerin the planning strategy whether it will use TDD or FDD schemes in the system.TDD may be more spectral efficient, and the system could be allocated to separatedunpaired frequency bands, but in the cost of the system capacity. On the other hand,licensing paired bands for an FDD system may be significantly more expensive forthe operator point of view.

2.3 Cellular Concept

The principal idea in cellular communications is to divide the coverage areas insmall areas called cells. This reduces the required transmission powers, and enablesthe use of small battery powered radio transmitters in personal use. In [9] threeaims for cellular systems are listed: coverage and mobility, capacity, and quality. Inother words, with cellular systems coverage is supposed to be offered in areas whereit is demanded by the users, with provision of mobility. Capacity demand relatesto the number of dropped calls, but nowadays also to the throughput experiencedby the end-user, since the mobile communications is more and more packet dataoriented. Quality is measured for a mature network by the means of speech quality,throughput, jitter, and bit error ratio (BER).

The cellular network basic infrastructure consists of base transceiver stations(BTS), or in generally base stations (BS), mobile stations (MS), and most likely alsonetwork elements that multiplex the traffic from multiple base stations to transportthe traffic in the backbone network. Mobile stations are connected to the basestations via radio links, and the base stations in turn either connect to other basestations or other collective nodes. In this thesis a cell site term is used for thelocation comprising the base station site equipment (transceiver, antennas, cablesand mast). Cell relates to the coverage area of a base station antenna. Also the termbase station is used throughout this thesis to refer to the base station equipmentand the antennas instead any system specific terms, or more specific BTS term.

One of the aspects that is commonly applied to increase the cellular systemcapacity is sectorizaton, in which a BS site is split in narrower coverage areas bydirectional antennas. Although sectorization requires more base station equipment,it is widely used in modern cellular systems, since it has been proven to reduce the

2. Introduction to Cellular Networks 9

overall interference in the network, and thus increase capacity. Sectorization hasbeen seen particularly useful in areas with high traffic density.

When the user locates at the cell edge, and is on the way to change to the cover-age area of the neighboring cell, it is desired to maintain the connection during therelocation to the new cell. The procedure that is used for serving cell change (toprovide mobility) is called a handover. Basically, a handover is triggered when cer-tain threshold values are reached in the received signal level from the base stationsparticipating in the handover. Handover is called hard handover, when the connec-tion is closed to the previous serving base station before connecting to the new one.In this case, some buffering may be needed in the network elements to maintainthe handover seamless, and also the handover should occur in such a short timeframe that it can be hardly noticed by the user. In soft, or softer, handover connec-tions are remained in the cell edge region to two or more base stations, or sectors,respectively. Handovers may also be divided into inter-system or inter-frequencyhandovers, which are in practice hard handovers.

2.4 Radio Propagation in Cellular Systems

2.4.1 Propagation Environments

Radio propagation environment is typically classified in three different categories:urban, suburban and rural – and additionally to special classes called microcellularand indoor. Classification between the area types is based on the obstacles (build-ing, trees, etc.) surrounding the base station and mobile station antennas. Theseparation is also done according to the base station antenna height. If the base sta-tion antenna height in the area is above the average rooftop level, the environmentis considered to be macrocellular. In microcellular cases the average base stationantenna height is below the rooftop level. [10]

Only urban areas with high buildings are considered to have microcellular prop-agation environment, since the surrounding buildings prevent the radio propagationin larger area from the base station antennas that are located below rooftop level.

2.4.2 Propagation Models

A propagation model is needed to estimate the path loss between the mobile sta-tion and base station. This need occurs for example in the network dimensioningphase. Propagation models are divided in empirical, semi-empirical and determin-istic models. Empirical models are based on measurement campaigns: the mea-surement statistics are turned into mathematical models. Semi-empirical modelsrely on physical phenomenas, such as diffraction, refraction and reflection, and com-

2. Introduction to Cellular Networks 10

bine these with field measurements. Deterministic models, such as ray tracing andray launching, have a basis on the electromagnetic theory, therefore providing moreaccuracy in path loss calculations in cost of computational power requirement.

For simple line-of-sight (LOS) conditions (2.1) applies for free space loss whichis expressed as a ratio of the radiated power Pt to the power Pr received by theantenna

PtPr

=(4πd)2

λ2=

(4πfd)2

c2, (2.1)

where λ is wavelength, d distance between the transmitter and receiver, f the fre-quency and c speed of light.

In macrocellular environments the Okumura-Hata (or simply Hata) model is themost widely used empirical propagation model. The model is based on the fieldmeasurements made by Okumura in Tokyo in 1968 [11]. Okumura’s measurementswere later fitted in a mathematical formula by Hata [12]:

Lurban = 69.55 + 26.16 log10(f)− 13.82 log10(hBS)− a(hMS)

+ [44.9− 6.55 log10(hBS)] log10(d),(2.2)

where a(hMS) is a mobile antenna correction factor, that depends on the propagationenvironment (defined for small city and large city), hBS and hMS. Input parametersfor the Okumura-Hata model are presented in Table 2.1.The model applies undercertain conditions for urban areas. With additional terms the model can be tunedto fit better in other propagation conditions. For example, the equation

Lopen = Lurban − 2

(log10

(f

28

))2

− 5.4 (2.3)

presents the Okumura-Hata-model for open rural areas. [13]

Table 2.1: Okumura-Hata model parameters.

Parameter Description

f Carrier frequency, 500 < f < 1500 MHzhBS Base station antenna height in meters, 30 < hBS < 200 mhMS Mobile station antenna height in meters, 1 < hMS < 10 md Distance between mobile and base station in kilometers,

1 < d < 20 km

Later, the European Union initiated COST forum for co-operative scientific re-search covered mobile radio evolution related topics in COST-231 program. Asone result Hata model was improved to cover frequency range up to 2000 MHz,

2. Introduction to Cellular Networks 11

and thereby to be applicable also for propagation predictions for 3G systems. Theformulation of COST-231-Hata model is presented as follows:

Lurban = 46.3 + 33.9 log10(fc)− 13.82 log10(hBS)− a(hMS)

+ [44.9− 6.55 log10(hBS)] log10(d) + C,(2.4)

where C is an area correction factor. The other parameters that were earlier definedfor the Hata model apply also for the COST-231-Hata model, but the frequencyrange must follow COST-231-Hata limitations. [14, 13]

2.4.3 Multipath Propagation

Multipath propagation occurs, when the same transmitted signal arrives to thereceiver in different time instances. Thee mechanisms play a role in multipathpropagation: reflection, scattering and diffraction, which are illustrated in Figure2.4. Reflection occurs when a radio wave encounters a surface that is large relativeto the wavelength of the signal. An example of reflection could be a ground-reflectedsignal from a base station to a mobile. A reflection causes a 180 degrees phase shiftto the electromagnetic signal, which may cause signal cancellation to the reflectedsignal arrived to receiver antenna with the LOS component. On the other hand,the path length of the reflected signal may cause noticeable delay relative to theunreflected signal. If the path delay equals to half a wavelength, the two signals areback in phase, and thus are received constructively in the receiver antenna. [15]

obstacle

building

building building

building

diffraction

scattering

reflection

Figure 2.4: Multipath propagation concept.

When a radio wave encounters an obstacle that is in order of the signal wavelengthor less, it scatters into several weaker components. In mobile communications, there

2. Introduction to Cellular Networks 12

are several obstacles in this size scale, such as lamp posts and traffic signs. Scatteringis very hard to predict, because the amount and places of obstacles may vary a lotin time in every propagation environment. [15]

Diffraction occurs when the signal is passing a sharp edge that is large comparedto the signal wavelength. Diffracted signal changes its direction but not its phase.Due to diffraction signal reception is possible even without a LOS component of thesignal. [15]

Multipath propagation can be presented with impulse response, and the key figuredescribing multipath properties of a propagation environment is delay spread. Delayspread is the time difference of first arrived line-of-sight component and last arrived’echo’ of the signal. The multipath propagation effect is especially strong in urbanenvironment with a lot of dynamics and high concentration of small obstacles.

In CDMA systems, a RAKE receiver is used to mitigate multipath propagationeffects. It tries to recover the signals from multiple paths and combine them withsuitable delays. A RAKE receiver is illustrated in Figure 2.5. First a transmitteddata is spread with a spreading code and modulated for radio transmission. In thepropagation channel multiple copies of the signal are generated with different timedelays and attenuation factors. The receiver sees the multiple copies of the signalas a sum of multipath components, and demodulates the multipath propagatedsignal. In RAKE receiver the demodulated chip stream is fed to multiple correlatorswith different delays. Each correlated component is then weighted with factors thatare estimated from the propagation channel, and finally the weighted signals arecombined as one. [15]

a'1

a'2

a'3

c(t-τ1)

c(t-τ2)

c(t-τ3)

Σ

a1

a2

a3

ΣDemodulator

Multipath channel RAKE receiver

Modulated signal

τ1

τ2

τ3

Figure 2.5: Rake receiver. [15]

2.4.4 Slow and Fast Fading

In addition to distance dependent path loss, the received signal level experiences alsofluctuations that are are called slow and fast fading. When the receiver is placed in acoverage area that either omits the LOS component or other dominant component,the receiver is considered to be shadowed. Usually these shadowing obstacles are

2. Introduction to Cellular Networks 13

big trees, buildings, or in rural environment the hills. Slow fading is shown to followlog-normal distribution, and thus it is also called log-normal fading.

Fast fading is a consequence of the multipath propagation phenomena. As de-scribed earlier, the multipath propagated signal components can be received eitherconstructively or de-constructively depending on the relative phase difference be-tween the components. This causes very rapid fluctuations in the received signallevel. Slow and fast fading components of received signal are illustrated in Figure2.6 in respect to receiver-transmitter distance.

fast fading

slow fading

Amplitude

Distance

Figure 2.6: Slow and fast fading of propagated signal in respect to receiver-transmitterdistance.

2.5 WCDMA for UMTS

2.5.1 UMTS Reference Model

At high level of abstraction UMTS architecture consists of the user equipment (UE),UMTS Terrestrial Radio Access Network (UTRAN) and core network (CN). Eachof the elements are hierarchically connected to higher or lower level elements viainterfaces that are defined by 3GPP (see Figure 2.7). Only the UE and RAN ele-ments are described in this subsection as the thesis covers only radio access aspectsof UMTS. For further reading of the topic [16] is suggested.

The user equipment consists of the mobile equipment (ME) and UMTS subscriberidentity module (USIM). ME is the mobile device that the subscriber uses for radiocommunication over the Uu interface. In the mobile equipment the subscriber hasthe USIM, which is a small smart-card containing the identity information andperforming authentication and encryption algorithms. [16]

2. Introduction to Cellular Networks 14

In UMTS the radio access network, also called UTRAN, consists of Node Bs andradio network controllers (RNC). The Node B is a 3GPP specific name for a 3Gbase transceiver station (or base station, which name is further used in this thesis),and it converts the data flow between Iub and Uu interfaces, and also participatesin radio resource management [16]. The RNC is the service access point to all theservices that UTRAN provides to CN, and it also owns and manages all the radioresources of the Node Bs connected to it [16]. In the following chapters more generalterms (mobile station and base station) are used instead of the 3GPP terms.

UE

ME

USIM

UTRAN

NB

RNC

CN

NB

RNC

Uu

Cu

Iub

Iur

Figure 2.7: UMTS reference model.

2.5.2 WCDMA Air Interface

Direct sequence spread spectrum (DS-SS) is the most commonly used spreadingscheme in spread spectrum communications. This is also used in WCDMA (andin IS-95, respectively). The key features of the WCDMA speficiations are listed inTable 2.2

Table 2.2: Key features of WCDMA radio interface. [10]

Multiple access DS-CDMAChip rate 3.84 Mchip/sDuplexing scheme FDD and TDDChannel bandwidth 5 MHzFrame structure 10 ms frame with 15 time slots

In DS-SS, binary modulated data is modulated for a second time with a widebandsignal, called as spreading code. The spreading codes are noise-like pseudonoisesequences, that have a symbol rate (also called as chip rate) much higher thanthe user bit rate. In WCDMA two codes are applied in the spreading operation tomaximize the system capacity. Firstly, orthogonal variable spreading factor (OVSF)codes are used as channelization codes to separate different users from each otherwithin a cell and to spread the narrowband signal into wider bandwidth. OVSF

2. Introduction to Cellular Networks 15

codes provide maximum capacity, measured by the number of active users, with theproperty that two codes from the same family are perfectly orthogonal. Secondly,spread signal is multiplied with a cell-specific scrambling code, which has statisticalproperties of a random sequence. The concept of signal spreading and despreading inWCDMA is presented in Figure 2.8. Since the narrowband signal is multiplied withtwo different codes, the WCDMA signal is relatively flat in the frequency domainwithout any dominant peaks. [4, 10]

Radio channel

noise and

interference

scrambling

codechannellisation

code

scrambling

codechannellisation

code

narrowband signal

wideband signal

narrowband signal

spreading de-spreading

Figure 2.8: Spreading operation in WCDMA.

Spreading operation brings along several benefits. First of all, tolerance to nar-rowband interference is achieved. When the spread wideband signal is interferedwith a narrowband signal, and the wideband signal is despread again in the re-ceiver, the received signal is formed now from a sum of the narrowband receivedsignal and spread interfering signals (Figure 2.9). The received narrowband signalis then filtered with a bandpass filter, and in the end, only a minimal part of origi-nal narrowband interfering signal energy is left. The ratio between the transmittedmodulation bandwidth and the information signal bandwidth is called spreadingfactor:

SF =Rc

Rbit

, (2.5)

where Rc is the chip rate and Rbit the user data rate, respectively. [4]

wideband signal

narrowband interference despread narrowband signal

spread interference

bandpass filter

despreading

Figure 2.9: Despreading a wideband signal in a presence of interference. [4]

When the spreading factor is transformed to logarithmic scale the result is calledprocessing gain

PG = 10 log10(SF ) = 10 log10(Rc

Rbit

). (2.6)

2. Introduction to Cellular Networks 16

In radio network planning the processing gain is used as an additional gain in linkbudget calculations, as the spreading is seen to bring additional gain to receiversensitivity.

Since the system chip rate is fixed to 3.84 Mcps and the spreading factor rangesbetween 4 and 512, the net user bit rates supported by one code channel vary from1 to 936 kbps. In theory, the maximum system downlink bit rate of 2.3 Mbps isachieved by allocating three codes for a user. However, in Release 99 multicodes arenot used in practice. In uplink the corresponding bit rates are half to the downlinkdue to the lower order modulation scheme. [10]

The WCDMA system also has a benefit when operating in multipath environ-ments. As discussed in Section 2.4.3, a RAKE receiver can be used in CDMAsystems to mitigate the multipath effects. The RAKE receiver can cope with themultipath components with a resolution of the system chip rate - the shorter the chipduration is, the better the RAKE receiver can exploit the multipath components. InWCDMA 3.84 MHz system bandwidth allows approximately 0.26 µs RAKE receiverresolution. [4]

2.5.3 Power Control in WCDMA

As discussed in Section 2.2, to cope with the near-far effect sophisticated powercontrol scheme has to be used in CDMA based systems. Especially in the uplinkdirection only one mobile station may in the worst case block the whole cell, or atleast degrade the signal level received in the mobiles close to the cell edge. The ideaof power control (PC) is to maintain the transmitting power level as low as possible,yet still maintaining an adequate signal level in the reception.

In WCDMA, the power control functions are divided into open-loop PC, inner-loop PC and outer-loop PC. Open-loop PC sets the initial transmission powers whenthe mobile is accessing the network. When the mobile is attached to the networkand is active, inner-loop PC is utilized in 1500 Hz basis to dynamically adjust thetransmission powers maintaining the signal quality at required level. The outer-loop PC estimates the received signal quality and adjusts the target SIR (signal-to-interference ratio) for the inner-loop PC. [4]

2.5.4 Handovers in WCDMA

Both hard and soft handover schemes are supported in WCDMA. Hard handoversare used, when the carrier frequency changes between the cells (for example from900 MHz to 2100 MHz), or when the system changes (between GSM and WCDMA).

In normal WCDMA operation soft handover (SHO) is the most common handoverscheme. In soft handover the mobile is connected simultaneously to two or more

2. Introduction to Cellular Networks 17

Node-Bs (see Figure 2.10(a)), which in turn may be managed by one common orseveral different RNCs. Besides SHO, it is also possible to maintain simultaneousconnections to different sectors within the same serving Node B. This scheme iscalled softer handover (SfHO) (see Figure 2.10(b)). Moreover, in SfHO the handoveris managed by the one RNC that manages the serving Node B.

(a) Soft handover between two cells.

(b) Softer handover between two sectors.

Figure 2.10

S(f)HO window margins are tunable RRM parameters, and are part of the opti-mization and maintenance process of the network. SHO and SfHO schemes providegain order of couple of decibels in reception diversity sense. However, simultaneousconnections to multiple base stations are also decreasing the overall capacity in thesystem level, and thus the soft handover regions should be optimized (as a part ofradio network planning process).

2. Introduction to Cellular Networks 18

2.6 HSDPA for UMTS

At the time when Release 99 WCDMA was introduced, 384 kbps was sufficientexpectation for any data connection, and it was considered as ’broadband’. In3GPP Release 5 the technical specifications for HSDPA were introduced. Basically,HSDPA is a technical update to WCDMA radio access architecture in providing newmodulation and coding schemes, and also changes in radio resource management(power control and scheduling). One major difference is also that HSDPA uses ashared data channel for packet switched communication. In Release 99 WCDMAeach subscriber have a dedicated channel allocated for voice or data traffic, but inHSDPA they share a channel called high speed downlink shared channel (HS-DSCH)at a time for the time instance that is allocated by the scheduler in the Node B.Also, HSDPA does not support soft handovers in downlink (but in uplink theyare optionally supported as uplink uses the same channels as the basic Release 99WCDMA). The key differences between Release 99 WCDMA and Release 5 HSDPAare listed in Table 2.3.

Table 2.3: Differences between Release 99 WCDMA and Release 5 HSDPA.

Release 99 WCDMA Release 5 HSDPA

Soft handover Yes NoFixed spreading factor No YesAdaptive Modulation and Coding No YesFast Power Control Yes No

Instead of fast power control, HSDPA responds to propagation channel variationswith adaptive coding and modulation (AMC). The SINR requirement (and thusachieved instantaneous throughput) is adjusted either tightening or relaxing thecoding and modulation schemes instead of adjusting instantaneous power levels.The outer loop power control is still used to keep the transmitted signal power atadequate level. In practice AMC is done based on the reports, called channel qualityindicators (CQI) which the mobile sends to the network. The link adaptation is donein 2 ms time intervals. [16]

The maximum cell (and end-user) throughput depends mainly on the number ofcodes that are allocated for the connection, and the used modulation and codingscheme. When 3/4 coding and the highest order modulation, 16 quadrature ampli-tude modulation (16-QAM), are used, up to 10.7 Mbps cell throughput is achievedwith 15 parallel multicodes [16]. From the subscriber point of view, the achievedthroughput depends on the UE capabilities, that the 3GPP has divided in differentcategories listing the supported coding and modulation schemes for each class. Forexample, 16-QAM modulation scheme is only an optional feature, which was not

2. Introduction to Cellular Networks 19

supported in the first UEs, and thus with them the maximum throughput wouldnot be achieved.

2.7 WCDMA Frequency Variants

The typical carrier frequency for WCDMA UMTS has been 2100 MHz, but 3GPPhas produced specifications also for other bands that are used in different regions.Lately, WCDMA deployment in Europe to 900 MHz band has started in order toprovide wider coverage for 3G services, especially to provide better indoor coveragefor mobile data connections without need to build a dense network. The WCDMAfrequency variants are release independent, meaning that, even if the frequencyvariant is added with later 3GPP release schedule, the products for that band canuse earlier 3GPP release as design basis. The frequency variants and their usageareas are listed in Table 2.4. [16]

Table 2.4: WCDMA frequency bands in 3GPP. [16]

Operatingband

3GPPname

Total spec-trum

Uplink[MHz]

Downlink[MHz]

Region

Band I 2100 2x60 MHz 1920 - 1980 2110 - 2160 Mainstream WCDMABand II 1900 2x60 MHz 1850 - 1910 1930 - 1990 PCS band in AmericaBand III 1800 2x75 MHz 1710 - 1785 1805 - 1880 Europe, Asia and

BrazilBand IV 1700/

21002x45 MHz 1710 - 1755 2110 - 2155 New 3G band in Amer-

ica and AsiaBand V 850 2x25 MHz 824 - 840 869 - 894 USA, America, and

AsiaBand VI 800 2x10 MHz 830 - 840 875 - 885 JapanBand VII 2600 2x70 MHz 2500 - 2570 2620 - 2690 New 3G bandBand VIII 900 2x35 MHz 880 - 915 925 - 960 Europe and AsiaBand IX 1700 2x35 MHz 1750 - 1785 1845 - 1880 JapanBand X 1700/

21002x60 MHz 1710 - 1770 2110 - 2170 Extended band IV

20

3. RADIO NETWORK PLANNING FOR UMTS

Radio network planning is an on-going process that aims to cost efficient networkroll-out and operation from the operator point of view – at the same time aiming toperformance optimization from the end-user point of view. This chapter introducesthe basic concept of the planning process in WCDMA networks. In addition, theadditional aspects, that are needed in planning HSDPA optimized network, areintroduced together with the performance metrics that were used in the studiesdone for this thesis.

3.1 Background

Radio network planning is a process that supports the operator by producing net-work plans and maintaining the network operation in optimal level. Operating andbuilding a network is always a trade-off between the expenses, capital expenses(CAPEX) and operating expenses (OPEX). In 3G networks, the trade-off can beseen also between the coverage and capacity. It could be said the main target forthe network planner from the operator point of view is to design a network withminimal costs and with maximal performance.

The network planner can affect capital expenses when defining the network con-figuration, and deciding the number of required base station sites. Less sites leads tosmaller investments. Similarly, less sites require also less operating costs (for exam-ple electricity or maintenance). On the other hand, even though the number of sitesshould be minimized, the system capacity and coverage should remain in adequatelevel - that is, the number of dropped calls should be minimal, and mobility aspectshould be remembered (no coverage holes, especially in regions where the use of thenetwork is active).

3.2 Planning Process

As a process, radio network planning can be seen consisting of phases starting fromnetwork dimensioning, and continuing to detailed planning and optimization. Theplanning process is illustrated in Figure 3.1. The planning process is started byroughly estimating the capacity and coverage need for the network, that is dimen-sioning. When the rough idea of the network is available, more detailed plans for thenetwork topology, equipment configuration and parameters can be planned. After

3. Radio Network Planning for UMTS 21

this, the operator has a plan to start the roll-out phase. The capacity and config-uration requirements for the network may change as the time passes, for exampledue to increase of subscribers in the network, or topology changes in the propaga-tion environment. Thus, on-going network measurements have to be conducted, andperformance data collected, to maintain the optimal performance for the network.

DIMENSIONING

- traffic

- area

- coverage threshold

DETAILED PLANNING

- traffic

- site configuration

- coverage thresholds and

capacity requirements

OPTIMIZATION AND

MONITORING

- traffic

- coverage verification

- capacity availability

Figure 3.1: The planning process for cellular radio networks. [10]

3.2.1 Dimensioning

In the dimensioning phase all the input data for detailed planning is gathered pro-viding an initial rough estimate for the amount of network elements and capacity.In radio network dimensioning the target is to estimate the site density and the sitelocations in the target area. Also services provided to the subscribers must be takeninto account, as they relate to the required capacity, and to quality of service (QoS)requirements. [4]

When the network planning is in dimensioning phase, fixed load for all the cellsis assumed. This may lead to too high density of cells, if average load is assumedtoo high. Also an assumption of similar propagation to all the cells, and evenlydistributed traffic leads to inaccurate assumption that all the cells are the same.Thus in the detailed planning phase the coverage predictions may be quite differentto the ones in dimensioning. [10]

Link budgets are used for defining the cell ranges in the radio network. Linkbudget defines the maximum allowed path loss for the radio link by taking intoaccount transmitter power, receiver sensitivity, RF equipment gains and losses, anddifferent system dependent gains and margins. The output path loss value canbe translated into cell coverage with a propagation model or measurements. Linkbudgets are calculated for all the provided services, and the maximum cell rangeis defined by the service with the tightest path loss requirement. An example of aWCDMA-specific link budget is presented in Table 3.1. In the example, it is assumedthat the traffic is asymmetric packet data in downlink and uplink directions. Thelink budget is slightly unbalanced, which should be noted in detailed planning phase.It is good to notice that in WCDMA the link budget calculations depend on the cell

3. Radio Network Planning for UMTS 22

traffic activity and the user bit rates (that is, the offered service: speech or data).The bit rate affects the processing gain, and load affects the coverage (taken intoaccount as interference margin). [10, 4]

Table 3.1: Example of WCDMA link budget. Extracted from [10].

Uplink Downlink

Load 30 % 50 %Bit rate 64 kbps 144 kbps

Receiving end

Thermal noise density -174 dBm/Hz -174 dBm/HzReceiver noise figure 3 dB 6 dBReceiver noise density -170 dBm -168 dBmReceiver noise power -105 dBm -102 dBmInterference margin 1.55 dB 3.01 dBTotal noise -103.45 dBm -98.99 dBmProcessing gain 17.78 dB 14.26 dBRequired Eb/N0 5 dB 4 dBReceiver Sensitivity -116.23 dB -109.25 dB

Receiver antenna gain 18 dBi 0 dBiCable loss 4 dB 0 dBLNA gain 0 dB 0 dBAntenna diversity gain 0 dB 0 dBSoft handover gain 3 dB 3 dBPower control headroom 0 dB 0 dBRequired signal power -133.23 dBm -112.25 dBm

Transmitting end

Transmitter power 125 mW 1000 mW20.97 dBm 30 dBm

Transmit antenna gain 0 dBi 18 dBiCable/body losses 2 dB 4 dBTransmitter EIRP 18.97 dBm 44 dBm

Allowed propagation loss 152.20 dB 156.25 dB

3.2.2 Detailed Planning

Detailed planning consists of configuration planning, topology planning and parame-ter planning. When the dimensioning phase is ready, and estimates for site locationsand antenna mast heights are known, the site configuration can be chosen more indetail. For the detailed link budget calculations computer simulations may provideadditional information about expected interference levels in the network and re-quired signal quality for given service. Also through detailed simulations the precise

3. Radio Network Planning for UMTS 23

site locations and the cell coverages can be defined.One part of the detailed planning is to balance the link budget, if the maximum

path loss values differ between uplink and downlink, since imbalance would lead tolimitations either in uplink or downlink coverage. Like in the example link budget inTable 3.1, the uplink limitation could be compensated with a low noise mast headamplifier (MHA). Thus, configuration planning plays a key role in this balancingtask.

Parameter planning gives the initial parameters for the network operation. Theseparameters can be definitions of soft handover window thresholds (affects to theoverall capacity of the network), or admission control related parameters.

Detailed planning phase may include a task that is called model tuning. Propa-gation (and coverage) predictions are usually done with semi-empirical models, thatare described in Section 2.4. The models are often too general and fully valid onlyfor the regions, where the original measurement campaign was done. The modeltuning is done to minimize the error of the propagation model, and thus improvethe accuracy of coverage predictions.

3.2.3 Optimization and Monitoring

Optimization and monitoring begins, when the network roll-out is ready and thesites are in use. This phase reflects well the nature of the iterative work neededin radio network planning. During the network evolution the traffic demand anddistribution will most likely change, which has to be considered in the capacityand coverage planning. Also the propagation environment may change due to theweather conditions in each time of the year, or due to the changes in the terrain(for example, more buildings are build in suburban region, which causes shadowingin some parts of the coverage area). Thus, the network configuration and topologyplan may change during the network evolution.

In network optimization, on-going measurements can help finding the optimumsolution for the current network configuration. Aspects that can be considered inthis phase are, for example, sectorization, antenna configurations (height, direction,beamwidth and tilting) or using additional amplifiers [4]. Also network monitoringtools provide useful data of the key performance indicators (KPI) from the network.With optimized operating parameters for radio resource management the networkQoS can be maintained at desired level.

3. Radio Network Planning for UMTS 24

3.3 Planning Tools

The planning tools mostly consist of network simulator and planning software, andmeasurement equipment. Both the measurement equipment and the simulationsoftware are versatilely used throughout the planning process, but the requirementsfor the tools in each phase may differ. For example, in the dimensioning phase theoutput information need from the tool can be more relaxed, as the purpose is toproduce only the initial requirements for network performance and configuration.

The simulators can be divided into static and dynamic simulators according tohow they handle the traffic distribution over time. In the static simulations thesubscribers are distributed to the surveyed area, and the network performance isstudied on a fixed time instance. The set of results, that a static simulator giveson this time instance, is called a snapshot. Static simulators are often used in thedimensioning phase because they provide fast and rough estimates for the coveragearea with different planned services, and the effects of network configuration changescan be rapidly evaluated.

In the dynamic simulations the subscriber locations and the propagation environ-ment vary over the time, and thus more computational power is required. Dynamicsimulations can be useful when more detailed information from the system per-formance is required. To make static simulations statistically more reliable, Monte-Carlo analysis approach can be used. The idea of Monte-Carlo simulations is to takeseveral snapshots of the network in different time instances, and estimate an averagevalue for the results. The benefit of Monte-Carlo simulations is that less computa-tional power is required in respect to dynamic simulations, especially if already afew snapshots can provide the wanted information of the network behavior.

Besides the simulations, also measurement equipment are essential tools in ra-dio network planning. They can be used for allocating the problematic locationsfrom radio propagation point of view, or used for model tuning purposes. Utilizingmeasurement equipment can be beneficial in each phase of the planning process pro-viding realistic data to support, for example, the simulation results. Typical toolsused in radio network planning are, for example, spectrum analyzer, test transmit-ter and receiver, UMTS terminal, a computer collecting, storing and visualizing themeasurement data, and a GPS receiver (for recording the measurement route).

3. Radio Network Planning for UMTS 25

3.4 Planning for HSDPA

Since HSDPA is based on a shared data channel between the users, opposed to adedicated one in Release 99 WCDMA, average throughput over the cell is oftentaken as a planning criteria in HSDPA dimensioning. Already in the dimensioningphase it should be considered whether HSDPA will be shared with Release 99, orif the whole transmission power in the cell will be dedicated for HSDPA traffic. Ifthe power is shared with Release 99, there are two strategies for dimensioning thenetwork. In the first case remaining power from the Release 99 can be allocated forHSDPA connections, and throughput can be estimated according to the availablepower. In the second approach the desired throughput could be defined first andthen allocate the HSDPA power accordingly to achieve the throughput. In all, themain input parameters for HSDPA dimensioning are Release 99 traffic for traditionallink budgets and desired HSDPA throughput. [4, 3]

When planning the HSDPA radio interface, capacity, coverage and throughput areeven more firmly tied together in HSDPA than they were in Release 99. In HSDPAplanning, SINR takes the key role for estimating the throughput and coverage.SINR, and other related performance metrics for HSDPA are described more indetail in Section 3.5. When the HSDPA throughput requirement is set, SINR valuecan be estimated for example with link level simulations, and placed as a planningmargin in the link budget instead of Eb/N0 which depended on required Release 99service bit rate. Thus SINR value is also one of the factors that affect on HSDPAcoverage. Other differences to Release 99 that reflect to the coverage estimation viathe link budget are the lack of support for soft handover and fast power control, andalso the fixed spreading factor for shared HSDPA downlink channel. Due to missingsoft handover support, soft handover gain cannot be obtained anymore in HSDPA.On the other hand, the power control overhead margin is not needed to be taken intoaccount anymore. A fixed spreading factor of 16 can be seen as a 12 dB processinggain for all the supported bit rates. In uplink, additional margins ranging around afew decibels have to be considered due to suboptimal power control for users thatare applied soft handover scheme in uplink, and increased uplink signaling [4].

3.5 Performance Indicators for HSDPA Air Interface

Mostly the same performance indicators apply for the HSDPA as for basic WCDMA.However, from system throughput point of view the interference and signal qualityrelated SINR value plays a major role in HSDPA planning and performance moni-toring. In this section the basic performance indicators, that are used for analysisin this thesis, are described based on the definitions in [17] and [4].

3. Radio Network Planning for UMTS 26

Common Pilot Channel Received Signal Code Power, CPICHRSCP

Common pilot channel (CPICH) received signal code power (RSCP) is one of themeasurement indicators defined by 3GPP. It is defined as received power on one codemeasured on the Primary CPICH. It is measured by the mobile, and the value is usedfor estimating path losses, and indirectly in power control and handover algorithms.The reference point for the RSCP is the antenna connector of the mobile. [18]

Energy per Bit to Spectral Noise Density Ratio, Eb/N0

This value is rather used for estimating the required power level in Release 99WCDMA power control to achieve certain block error rate (BLER) with a givenservice (that is, certain user bit rate). Although in HSDPA SINR values are usedin downlink coverage and capacity estimations, Eb/N0 is still needed for uplink di-mensioning and detailed planning.

In uplink Eb/N0 can be defined as it is presented in (3.1). It is assumed that thesignal is received at a constant power prx and that the received interference powerequals to I. With user data rate Rbit and signal bandwidth W (equalling to the chiprate Rchip) the received interference is distributed over the whole signal bandwidthand the bit energy corresponds to prx/R. Thus, Eb/N0 can be given as:

EbN0

=W

Rbit

prxI. (3.1)

The required Eb/N0 value depends, in addition, on the radio propagation envi-ronment. For example multipath propagation and the mobile speed have an effecton the value. The required values for certain target data rate can be obtained fromlink level simulations. [4, 19]

Orthogonality factor, α

Orthogonality factor (α) is a theoretical figure that describes the orthogonality ofthe codes in the propagation channel conditions. It depends on the instantaneousmultipath conditions, and the value ranges between 0 and 1. In case there is aclear dominant component that is interfered with weak multipath components, theorthogonality factor value is closer to its maximum.

In analysis, fixed estimate of average α = 0.5 is often used. However, the averagevalue depends on the propagation environment. In rural areas propagation envi-ronment is expected to have more multipaths due to longer propagation distancescompared to urban and suburban environments. Thus the used value for rural areasshould be lower than for urban areas. However, if the amount of obstacles is low in

3. Radio Network Planning for UMTS 27

the rural area, the orthogonality factor can be higher due to less occurred scattering.[4]

Other-to-own-cell interference ratio, i

Other-to-own-cell interference ratio is defined as:

i =IothIown

, (3.2)

where Iown is the received interference coming from the own cell and Ioth the receivedpower from the other cells. The ratio describes how well the cells are isolated fromeach other, and it relates to the spectral efficiency of CDMA system (since wellisolated cells can deploy more users and offer better throughput). Typical averagevalues for i range between 0.1-2, where low values are often obtained for well isolatedmicrocells, and higher values (over 1) poorly designed macrocells.

The i behaves a bit differently in uplink and downlink. In the uplink i is thesame for all the connections, since the value is estimated for the receiver in the basestation. In the downlink, the value is estimated for each mobile, and thus it dependson the mobile location. However, the average value over over all the mobiles is thesame as for all the base station in the network, assuming the traffic distribution ishomogenous. [4]

Geometry factor, G

Geometry factor is almost similar to inverse of the previously defined i being definedas the ratio of the received power from the serving cell (Iown) to the sum of powerfrom the other cells (Ioth) and the noise power (PN):

G =Iown

Ioth + PN. (3.3)

The geometry factor reflects to the distance between the mobile and the basestation, and the value ranges between -3 dB and 20 dB, where the lowest values areobtained at the cell edge. Geometry factor is used as an input value in downlinklink level simulations. [4]

3. Radio Network Planning for UMTS 28

Signal-to-interference-and-noise ratio, SINR

Signal-to-interference-and-noise ratio (SINR) for HSDPA ties together the downlinkrelated performance indicators:

SINR = SF · PHSDPA − PHS−SCCH

Ptot ·(1− α +

1

G

) , (3.4)

where SF is the spreading factor (16 in HSDPA), PHSDPA is the received HSDPApower, PHS−SCCH is the received power of high speed shared control channel Ptot thetotal received power including interference and noise, α the orthogonality factor, andG the geometry factor. The HSDPA SINR value is an essential metric in the linkbudget analysis, and it can be used for evaluating the achievable cell throughput, orfrom other perspective a target SINR value can be set to plan the network so thatthe desired throughput value can be achieved. [4]

29

4. REPEATERS

Repeater is the key element that was studied in this thesis. Hence, this chapter pro-vides the motivation why repeaters are used in cellular CDMA networks, and howthey can be used. The concept of a repeater unit is also briefly introduced. Further-more, the repeater related properties affecting the cellular system level performance,that were assumed in the studies, are presented.

4.1 Introduction to Cellular Repeaters

This thesis studies analog, air-to-air repeaters that communicate via radio links. Ananalog repeater is completely transparent to the base station only amplifying therelayed signal uplink and downlink frequency bands. It does not digitally regeneratethe signal or separate the users. The structure of such a repeater is very simplesince no intelligent hardware of software are needed. Repeater installations can bemade after network roll-out without hardware or software changes in the network,which offers a flexible way to optimize the network. Thus, analog repeaters are verycost-efficient to temporarily increase cell coverage or capacity, or to fill a coveragegap in shadowed regions.

A repeater may provide enhanced coverage when it is placed near to the cell edge,or deploying them in areas that are shadowed from the base station coverage. Arepeater can also enhance cell capacity by decreasing the required transmit powerin base station and mobile ends, and thus directly decreasing the total interferencelevel in CDMA networks. The repeater coverage area may be an outdoor or indoorlocation [4, 20] . The capacity improvement property of repeaters has been studiedin [21, 22], showing that in hotspot regions repeaters can provide noticeable capacityimprovement in the network.

Multiple repeaters can be installed to the same donor cell, serving multiple cov-erage gaps or for coverage extension [20]. Coverage extension can be achieved by,for example, daisy-chaining multiple repeaters (Figure 4.1(a)), which increases therepeater coverage beyond single repeater capability, but the inserted delays set theupper limit for the amount of chained repeaters [4]. In addition, they also amplifythe noise decreasing the receiver sensitivity levels. Daisy chained repeaters couldbe utilized to cover shadowed parts of a highway or a tunnel in a rural macrocelltopology. Another approach is to feed the donor cell with multiple parallel repeaters

4. Repeaters 30

(a) Multiple repeaters in series (daisy chained).

(b) Multiple repeaters located parallel in a cell.

Figure 4.1

(Figure 4.1(b)), for example covering multiple hot spot areas, or providing extendedcoverage in sparsely populated rural areas.

Not much research has been done about multi-repeater utilization in cellularnetworks. In this thesis this topic is covered by investigating the effects of parallelmulti-repeater deployment in macrocellular environments. Also the challenges ofmulti-repeater deployment are covered in the following chapters.

4.2 Repeater Equipment

4.2.1 Overview

Repeaters are amplifying relay stations for the signal. They are located in betweenthe target subscribers in a cell and the corresponding parent cell. An example of arepeater configuration is presented in Figure 4.2. A repeater consists of the repeaterunit and two antennas: one forms a link to the donor cell (also known as mothercell) and the other for service area coverage. The antenna directed from the repeaterlocation to the base station is called donor antenna, and the antenna pointing tothe repeater service area is called service antenna. A repeater site also requires apower supply as the repeater unit requires power feed. The repeater antennas maybe located on a mast or building rooftops.

4. Repeaters 31

Figure 4.2: An example of repeater configuration.

4.2.2 Antennas

A repeater configuration utilizes usually two antennas, donor antenna and repeaterservice antenna, although it would be possible to transmit the signal between therepeater and base station via wired optical or copper link (in cost of increasedtransmission delay in comparison to wireless transmission). Normally, donor linkantennas have high horizontal directivity, for example 33 degrees, in order to de-crease the inter-cell interference and multipath channel effects. For the service areadirective antennas with wider beamwidth can be used.

A challenge that must be considered when setting up a repeater antenna config-uration, is repeater self-oscillation. Self-oscillation occurs when donor and serviceantennas leak the signal between each other, and the repeater amplifies its ownsignal (Figure 4.3). The antenna isolation should be at least 15 dB higher thanthe used repeater gain [23]. It is also suggested to use repeater antennas with highfront-to-back ratio to avoid the leakage.

Repeater

GR

service

antennadonor

antenna

Repeater

GR

service

antennadonor

antenna

Good isolation Poor isolation

Figure 4.3: Example of good and poor antenna isolation in repeater site.

4. Repeaters 32

4.2.3 Repeater Units

A WCDMA repeater that is under the scope in this thesis, basically consists of alinear amplifier and bandpass saw-tooth filters for uplink and downlink frequencybands. Repeater noise figure value is typically 3-5 decibels. The repeater gain canbe adjusted separately for UL and DL directions ranging typically between 45-90decibels. A repeater may also contain an automatic gain control (AGC) controllingthe repeater gain to avoid the self-oscillation phenomena. In addition, a repeaterincludes connector for donor and service antennas, and a possible communicationlink for adjusting the repeater settings. [23]

In this thesis, for simplicity for the simulations, the filters and amplifiers areassumed to be ideal with ideal frequency responses and linearly operating amplifier.Also transmission delay is neglected since the repeaters are placed within a cellrange, and the transmission delay effect remains negligible.

4.3 The Effects of Repeaters on Transmission Path

4.3.1 Thermal Noise

Repeater is not an ideal amplifier, and thus it adds noise to the signal passingthe repeater unit. The repeater system block diagram with noise addition is illus-trated in Figure 4.4. The cell structure with an embedded repeater also changes thestructure of the transmission path. The total path loss is divided between mobile-to-repeater link (service path loss LS) and donor links (repeater path loss LP ). Therepeater path from the noise generation perspective is illustrated in Figure 4.4 andthe parameters are described in Tables 4.1 and 4.2. [24]

Repeater

GR

TeR

TaR

Base station

GBS

TeB

TaB

NO

NOR

GAGD

NiR

SiR

Lp

GT

Figure 4.4: Repeater system block diagram. [24]

4. Repeaters 33

Table 4.1: Repeater transmission path parameters.

Symbol Definition

FB Base station noise figureFR Repeater noise figureGA Base station antenna gainGD Repeater donor antenna gainGR Repeater gainGS Repeater service antenna gainLP Path loss between repeater and BSLS Path loss between MS and repeater

Table 4.2: Repeater system noise parameter definitions.

Symbol Definition

N iB Noise power at the base station inputNo Noise power at the base station outputSiB Signal power at base station inputSo Signal power at base station outputTaB Base station antenna noise temperatureTaR Repeater service antenna noise temperatureTeB Inherent noise temperature of the base stationTeR Inherent noise temperature of the repeater

The thermal noise density of a component is generally expressed in form

NTH = kT, (4.1)

where k is the Boltzman’s constant and T is the noise temperature of a component.Using Figure 4.4 and (4.1), the thermal noise at the base station can be expressedin form

No = k(TaR + TeR)GT + k(TaB + TeB), (4.2)

where GT is total repeater path gain combining all the gains and losses on thetransmission path:

GT = GRGDLPGA. (4.3)

As the noise factor for a system is generally defined as the ratio of input signal-to-noise ratio to output signal-to-noise ratio, the effective base station noise factorfor the total noise contribution in unloaded network is

4. Repeaters 34

EFB =

SiB

NiBW

So

NoW

=

SiB

kTaB

So

k(TaR + TeR)GT + k(TaB + TeB)

, (4.4)

where W is the signal bandwidth. Assuming SiB = So, TaB = TaR and TeR = TeB,and by using the relation between thermal noise and noise factor, (4.4) can besimplified in form

EFB = FB +GTFR. (4.5)

In parallel multi-repeater architecture it can be logically deduced from Figure 4.4that the repeater generated noise sum up at the BS. Considering the summed noisefrom multiple repeaters, (4.1) can be shown to get form

EFB = FB +n∑i=1

GT,iFR,i, (4.6)

where n is the number of repeaters connected to the base station. In (4.5) it canbe seen that the effective base station noise factor depends on the distance of therepeater from the base station (due to path loss dependency of GT ). Also repeatergain GR clearly affects on the noise level in the base station. In further chapters,EFB may also refer to as the effective base station noise figure, meaning the valuein a logarithmic scale.

4.3.2 Transmission Delay

According to [23], repeater introduces an additional propagation delay of 5-6 µs

to the relayed signal due to commonly used narrowband filtering technology. Thiscauses extra requirements for the receiver’s search window.

A repeater can be seen to cause an additional multipath propagation effect in thereceiver point of view. In this case, to capture a certain amount of total receivedsignal energy, more RAKE fingers may be required in the receiver end. However, itcan be expected that a repeater embedded in a cell covers a smaller area comparedto the base station, and thus the number of resolvable multi-paths could be smaller.On the other hand, according to [23] maximum time delay between two paths shouldbe 20 µs or less in order the RAKE receiver could constructively combine the multi-path components. This sets a practical limit for three or less repeaters embeddeddaisy-chained in the same cell.

4. Repeaters 35

4.3.3 Repeaters in WCDMA network

Installing a repeater increases the network coverage, especially when a repeater isinstalled in cell edge (or shadowed area). The capacity improving effect of a repeateris much more complicated. It is notable that a repeater is a passive componentin a network and does not bring any additional capacity to the theoretical systemcapacity – quite in opposite. However, in optimization sense it is possible to increasethe capacity that is obtained in practice in a cell. This is achieved by limiting other-cell interference (improving cell isolation), and thus increasing the proportion ofown-cell power in respect to interfering other cell power. Balancing the uplink anddownlink capacity needs makes repeater deployment even trickier, since in uplinkrepeater always causes additional noise rise in the base station. Thus, the transmitpower in mobile stations may be increased causing a higher interference level in thewhole cell, and in that sense eating more cell capacity.

Aspects on setting a repeater operating point are discussed in [24] and in [20].There is always a trade-off between increasing the noise level either in uplink ordownlink, and therefore this has to be considered when repeaters are included inthe radio network plan. Adequately low repeater gain leads to low rise in the BSnoise floor, but in turn it also increases the effective repeater noise figure limiting thedownlink coverage that could be achieved with repeaters. If the downlink coverageand capacity of the mother cell can be sacrificed, with higher repeater gains it ispossible to extend the repeater coverage. This can be the case, if, for example,capacity can be allocated from mother cell to a dedicated repeater service area,such as an indoor location.

Deploying repeaters in existing cells may be beneficial in the optimization phaseof the network, when the base station site infrastructure already exists, and thecell density is not wanted to be increased. For example, in [25] major through-put enhancement up to 50% was observed in dynamic simulations, when repeaterswere deployed in dense urban areas to enhance the performance of macrocell sitedeployment. Also in [26] repeaters have been acknowledged to increase the downlinkcapacity through other-cell interference reduction. However, research about multi-repeater performance seems not to exist in the extend that the results would bepublicly available.

36

5. NUMERICAL ANALYSIS FOR SYSTEM

LEVEL PERFORMANCE EVALUATION

In this chapter an overview is given for the numerical analysis method. The usedanalysis tool is introduced, and the parameters and site configurations for the an-alyzed scenarios are given. The results are followed up in Chapter 6 together withthe error analysis.

5.1 Overview for Analysis Method

The numerical calculations for the studies were performed using a MATLABTMbasedstatic system level network simulator [27] as a framework. The original simulatordid not have a support for cellular repeaters, and the support was later implementedto support repeater research work [28]. In the implementation it is assumed that arepeater unit is an ideal linear, bidirectional, amplifier for the input signals, and thatthe repeater donor and service antenna are fully isolated. Modeling of antennas isalso included in the repeater implementation, and the simulator takes the antennapatterns as input. For interference analysis the interfering signals from adjacentcells are taken into account in the donor antenna reception.

In the analysis, two factors are studied: the link quality and the received signalstrength. This is done studying downlink SINR and CPICH RSCP values, respec-tively. For a HSDPA connection the key link quality indicators that give informationon the signal quality are also geometry factor and orthogonality factor. However,these values are already included in the SINR value (see (3.4)). Thus, surveyingthe orthogonality and geometry factors separately is not required (but the simula-tor does calculate those for SINR evaluation). The analysis is concentrated on thedownlink because HSDPA is a downlink specific technique (but it also has to be re-membered that careful uplink planning is required for error free feedback signaling).On the other hand, as a design target a modest noise rise in the uplink was chosen,so that the uplink capacity and coverage should not be deteriorated by the repeaterdeployment. The analysis was conducted over the map area.

The simulator evaluates on each map pixel, which is the serving cell (and whichareas are served by a repeater), and the RSCP and SINR values are calculatedaccordingly. In the analytical calculations, the RSCP value is evaluated simply asa difference of CPICH transmit power and the path loss between the serving BS

5. Numerical Analysis for System Level Performance Evaluation 37

and a map pixel. If the pixel locates in repeater serving area, repeater gain is takeninto account in the total path loss. For SINR analysis on a map pixel m (that is, alocation for MS) and serving cell i, a similar re-formulation of (3.4), that is presentedin [29], was used in the simulations:

SINR = SF · S

Iown + Ioth +N, (5.1)

where S, Iown and Ioth are received powers defined as

S =piLmi

(1 +WRi), (5.2)

Iown =PiLmi

((1− αS,i) + (1− αr,i)WRi), (5.3)

Ioth =∑n6=i

PnLmn

(1 +WRn), (5.4)

andN is the noise level at the MS including the effect of the repeater in the downlink.The parameters used in (5.1)-(5.5) are further described in Table 5.1. Furthermore,WR is the repeater weight factor

WR =LDLR

, (5.5)

where LD is the link loss of the direct link between BS and MS, and LR is link lossof repeater link.

Table 5.1: Parameters for SINR analysis.

Parameter Description

SF Spreading factorS Received HS-DSCH powerIown Received interference from the serving cellIoth Received interference from other cellsN Noise level at receiving MSpi Transmit power of HS-DSCHPi Total transmit power of the serving cellPn Total transmit power of other cellsαS,i Orthogonality factor of the direct linkαr,i Orthogonality factor of repeater link

In the path loss calculations, the shadowing was neglected in order to highlightthe effect of repeater configuration changes in the results. Thus, it was possibleto visualize clearly the repeater service area RSCP and SINR values in respect to

5. Numerical Analysis for System Level Performance Evaluation 38

network configuration in which repeaters are not installed in the cells. Moreover,shadowing effect would not expect to have effected on the average performance overthe network. Also, if shadowing would have been taken into account, the samecalculations would have needed to be done for several snapshots. In turn, the mapdata from several snapshots would have needed to be averaged, but in practicehandling, for example, data with nine installed repeaters in a cell would have beenvery inefficient, and most likely even impossible with the hardware platform thatwas used in the simulations.

In order to provide more accurate SINR values, the orthogonality factor, that wasdescribed in Section 3.5, is treated more sophisticated than merely setting a constantaverage value over the whole map area. According to [30], the orthogonality factorcan be represented as a function of distance between the mobile and base station:

α(r) = 1−(a1 − a2 · e(−r/γ)

), (5.6)

where r is the distance between MS and BS, and a1, a2 and γ are propagationenvironment dependent parameters. The parameter values that were used in thesuburban scenario are presented in Table 5.6.

When defining the scenarios, the approach that was chosen, was to set the noiserise limit in uplink to one decibel in order to avoid losing too much uplink per-formance while increasing downlink performance. As it was seen in Section 4.3.1,repeater introduces additional noise in uplink, which in turn either decreases thenumber of served subscribers, or reduces the coverage area of the cell. The relationbetween the cell capacity in terms of served users and the noise rise in the uplink isinvestigated in Section 5.3.

To meet the uplink noise rise limit requirement, the simulator tunes the repeatergains automatically according to an algorithm (see Figure 5.1). The algorithm firstforms a target value for GT on each repeater link based on (4.6), with the assump-tion that GT is the same for every repeater link. In practice this was done by firstdefining one target GT value that was then divided to all the repeaters that wereplaced in the cell. When the target GT is known, GR values for each repeater arecalculated applying (4.3). Since GT target is set the same for all the repeater linksin the cell, a repeater that is located directly in the BS antenna mainlobe obtainssignificantly smaller repeater gain, than the ones in the sides with larger antennalosses. Otherwise the repeater gains were distributed approximately symmetricallybetween the repeaters (due to the symmetries in the repeater locations). The max-imum and minimum values of used repeater gains are listed in Section 5.2 in Tables5.5 and 5.8.

5. Numerical Analysis for System Level Performance Evaluation 39

Uplink noise rise limit

Calculate target GT,target

G T,target

Calculate target G

for each repeaterR

Repeater gains

Repeater-BS link path losses

Antenna losses

Antenna gains

Figure 5.1: Repeater gain tuning algorithm.

5.2 Scenarios

The studied scenarios are divided to rural ones and suburban ones in this thesis. Thedivision is done with the propagation environment parameters and the base stationsite-to-site (and thus also repeater-to-site) distances. Also the system parametersare different between the rural and suburban scenarios. In suburban environmentnominal UMTS with 2100 megahertz frequency band is used, whereas in rural sce-narios 900 megahertz frequency variant was used due to much longer BS site-to-sitedistances.

In all the scenarios sixty-nine three-sectored base station sites were homoge-neously distributed over the map area to provide enough interfering signals prop-agating to the studied area from the adjacent cells. This provides more realisticresults when the signal quality is studied. The studied area concentrated over ahexagonal area, and the corners of the hexagon located at the first tier base stationsites. A simplified scheme of the base station layout is illustrated in Figure 5.2. Forsimplification reasons repeaters are shown in the figure only for one base stationsite, but in the simulations all the cells had the same number of repeaters.

Repeater locations were chosen so that they would be uniformly distributed overthe cell area with fixed BS-to-repeater distances (in respect to other repeaters withinthe cell). Topologies were created with 1, 2, 3, 5, 7 and 9 repeaters per cell (2-, 3-and 9-repeater cases are illustrated in Figure 5.3). In Table 5.2 the misalignmentangles are listed between the repeater donor antennas and the base station antenna.Repeaters are numbered in the Table 5.2 in clockwise order. The repeater service

5. Numerical Analysis for System Level Performance Evaluation 40

D

BS

BS

BS

BS

BSBS

BS

repeater

repeater

repeater

Figure 5.2: Schematic picture about the base station locations and the studied area. Hexag-onal polygon represents the studied area, and D site-to-site distance.

antenna directions in respect to the BS antennas are the same as the directionof repeater location from the BS antenna. Repeater donor antennas were alwaysdirected towards the BS antenna. The 2-repeater case differs from the other cases,since the repeater service and donor antenna locations for the both repeaters are thesame as in the 1-repeater case. See Figure 5.3(a) for further details. This was doneto bring out some possible effects of different repeater donor antenna locations andservice antenna directions. Nominal repeater distance from the BS site was set to afourth of the site-to-site distance D. More detailed information for the scenarios isprovided in following subsections.

Table 5.2: Repeater donor antenna direction for each cell (in respect to the mother cell BSantenna).

Repeater1 2 3 4 5 6 7 8 9

1 repeater 0 ◦2 repeaters 0 ◦ 0 ◦3 repeaters -30 ◦ 0 ◦ 30 ◦5 repeaters -40 ◦ -20 ◦ 0 ◦ 20 ◦ 40 ◦7 repeaters -45 ◦ -30 ◦ -15 ◦ 0 ◦ 15 ◦ 30 ◦ 45 ◦9 repeaters -50 ◦ -37.5 ◦ -25 ◦ -12.5 ◦ 0 ◦ 12.5 ◦ 25 ◦ 37.5 ◦ 50 ◦

5. Numerical Analysis for System Level Performance Evaluation 41

BS

60o

(a)

BS

30o

(b)

BS

50o

(c)

Figure 5.3: Repeater locations for two repeaters (a), three repeaters (b) and nine repeaters(c) deployed in each cell.

5.2.1 Rural Scenario

The rural scenario was analyzed with one to nine repeaters as defined in Section 5.2(for the donor antenna directions, see Table 5.2). Since the BS sites were distributedover a large area with long site-to-site distances, this was considered to be feasiblealso in practice. The propagation environment in this case corresponds open ruralterrain, and no additional building penetration loss (BPL ) is taken into account inradio link propagation calculations. Path loss calculations were done with Okumura-Hata model for open area shown in (2.3). Due to long site-to-site distances, theUMTS 900 system was assumed to be in use. Using UMTS 900 frequency band isa commercial trend for the operators at the moment for deploying 3G coverage insparsely populated rural regions. Details for the scenario are listed in Table 5.3,and the configuration parameters in Table 5.4. The antenna patterns for the basestation antenna and repeater donor antenna are illustrated in Figure 5.4(a) and5.4(b), respectively. The antennas correspond to the ones that are available fromthe manufacturer [31]. For the repeater service antenna the same model is used asfor BS antenna, and the both antennas utilize fixed electrical downshatilt (EDT).Moreover, the minimum and maximum values for the used repeater gains, GR,min

and GR,max, respectively, are listed in Table 5.5.For the rural scenarios the main target was to see how the different number of

repeaters affect the pilot coverage in the studied area, while keeping the uplink noise

5. Numerical Analysis for System Level Performance Evaluation 42

Table 5.3: Details for rural scenarios.

Topology

Terrain Flat / openSite-to-site distance 9.6 kmRepeater-to-site distance 2.4 km (0.25 D)Number of repeaters per cell Varies (1, 2, 3, 5, 7 or 9)

System UMTS 900

Table 5.4: Equipment configurations used in the rural simulations.

Parameter Value

Base station Noise figure 4 dBAntenna height 32 mAntenna gain 14.9 dBiAntenna downtilt 5 ◦ EDTAntenna beamwidth (horizontal) 65 ◦CPICH transmit power 30 dBm

Mobile Noise figure 8 dBAntenna gain 0 dBiAntenna height 1.5 m

Repeater Noise figure 3 dBDonor antenna height 32 mDonor antenna gain 20.57 dBiDonor antenna beamwidth (horizontal) 33 ◦Donor antenna downtilt noneService antenna height 32 mService antenna gain 14.9 dBiService antenna beamwidth (horizontal) 65 ◦Service antenna downtilt 5 ◦ EDTRepeater gain See Table 5.5

Table 5.5: Repeater gains for rural scenario.

Number of repeaters 1 2 3 5 7 9

GR min. [dB] 62 59 57 55 54 53GR max. [dB] - - 60 60 59 59

5. Numerical Analysis for System Level Performance Evaluation 43

−35−30−25−20−15−10 −5 0

30

210

60

240

90270

120

300

150

330

180

0

Radiation patterns for 65o antenna − 900 MHz

(a)

−35−30−25−20−15−10 −5 0

30

210

60

240

90270

120

300

150

330

180

0

Radiation patterns for 33o antenna − 900 MHz

(b)

Figure 5.4: Antenna patterns for 65 ◦ directional antenna (a) and 33 ◦ directional antenna(b) that are used in rural scenarios. The blue curve illustrates the horizontal pattern, andthe red curve vertical pattern. The 65 ◦ antenna pattern includes 5 ◦ EDT.

rise limited to one decibel. It was also targeted to show how many repeaters couldbe deployed in a rural cell, and still be able to improve the coverage. It was expectedthat after deploying a high enough number of repeaters, the benefit of the repeaterswould start to saturate and decline, since the repeater gain would need to be reducedat the same time to keep the uplink noise rise at a feasible level.

5.2.2 Suburban Scenarios

The suburban scenarios were analyzed with one to three repeaters within each cell,because the BS sites were much more densely located compared to the rural scenario.The site-to-site distance was chosen now to be 1.2 kilometers, and similarly to therural scenario, the nominal repeater-to-site distance was a fourth of the site-to-sitedistance. In addition, the distance was varied with 100 meter steps in order to seethe effect of repeater-to-BS distance when the distance is limited due to more denseBS site distribution. The variation of 100 meters was considered to be feasibleto bring out the effects with a minimum number of additional simulation cases.Since the BS sites were placed more densely, there was not seen to be room todistribute more repeaters in the mother cell without introducing additional own-cellinterference (and thus decreasing SINR). For the suburban propagation environmentnominal UMTS 2100 frequency variant was used with pre-implemented COST-231-Hata propagation model. The suburban scenario parameters are given in Tables 5.6and 5.7. The minimum and the maximum values for the repeater gains (GR,min andGR,max, respectively) are given in Table 5.8. Due to the different frequency bandin comparison to the rural scenario, specific antennas for 2100 MHz frequency bandwere used in the simulations (see Figure 5.5). For the antenna patterns see Figures5.5(a) and 5.5(b).

5. Numerical Analysis for System Level Performance Evaluation 44

Table 5.6: Details for suburban scenarios.

Topology

Terrain Flat (BPL included)Site-to-site distance 1.2 kmRepeater-to-site distance 200 - 500 m, 100 m stepsNumber of repeaters per cell Varies (1, 2, or 3)

System UMTS 2100

Orthogonality factor propagationenvironment parameters

a1 0.596a2 0.528γ 316.2

Table 5.7: Equipment configurations used in the suburban simulations.

Parameter Value

Base station Noise figure 4 dBAntenna height 32 mAntenna gain 18 dBiAntenna downtilt 5 ◦ EDTAntenna beamwidth (horizontal) 65 ◦CPICH transmit power 30 dBmHS-DSCH transmit power (pi) 42 dBm (16 W)Total transmit power of the serving cell (Pi) 43 dBm (20 W)Total trasnmit power of other cells (Pn) 40.8 dBm (12 W)

Mobile Noise figure 8 dBAntenna gain 0 dBiAntenna height 1.5 m

Repeater Noise figure 3 dBDonor antenna height 32 mDonor antenna gain 20.99 dBiDonor antenna beamwidth (horizontal) 33 ◦Donor antenna downtilt noneService antenna height 32 mService antenna gain 18 dBiService antenna beamwidth (horizontal) 65 ◦Service antenna downtilt 5 ◦ EDTRepeater gains See Table 5.8

5. Numerical Analysis for System Level Performance Evaluation 45

−40 −30 −20 −10 0

30

210

60

240

90270

120

300

150

330

180

0

Radiation patterns for 65o antenna − 2100 MHz

(a)

−40 −30 −20 −10 0

30

210

60

240

90270

120

300

150

330

180

0

Radiation patterns for 33o antenna − 2100 MHz

(b)

Figure 5.5: Antenna patterns for 65 ◦ directional antenna (a) and 33 ◦ directional antenna(b) that are used in suburban scenarios. The blue curve illustrates the horizontal pattern,and the red curve vertical pattern. The 65 ◦ antenna pattern includes 5 ◦ EDT.

Table 5.8: Repeater gains for suburban scenario.

Number of repeaters 1 2 3

Repeater distance GR [dB]200 m min. 52 49 48

max. - - 50300 m min. 56 53 51

max. - - 54400 m min. 58 55 54

max. - - 56500 m min. 60 57 55

max. - - 58

5.3 Noise Rise in the Uplink

The noise level at the receiver in the downlink and uplink directions depends on thecell loading. It can be defined in the uplink as [19]:

ηUL =EbN0

Rbit

Rc

·N · v · (1 + i), (5.7)

where N is the number of users in the cell, i other-to-own-cell interference ratioand v activity factor. The activity factor is used for indicating the probability thatthe mobile is transmitting traffic. Since in this thesis Release 99 uplink connectionis assumed (with the required additions for HSDPA feedback channel), the uplinkload can be estimated with (5.7). In HSDPA the uplink is used only for signalingpurpose, and for the user data traffic Release 99 channels are used. The noise rise,that should be taken into account in the uplink link budget as an additional margin,

5. Numerical Analysis for System Level Performance Evaluation 46

is given followingly [17]:

IM = 10 · log10(1− ηUL). (5.8)

Usually the network is designed either to tolerate load of 50% or 75% [4], whichcorrespond 3 dB and 6 dB noise rise in uplink, respectively.

In Figure 5.6 is presented the relation of noise rise in uplink with different userdata rates and the number of users in the cell. In the noise rise calculations, param-eters given in Table 5.9 are assumed. The values for Eb/N0 and i are obtained from[19] and [4]. Activity factor estimation is based on the assumption that the usedservice is downlink traffic oriented, for example web browsing, and therefore uplinkactivity remains low.

Table 5.9: Parameters used in the noise rise calculations.

Parameter Value Description

Eb/N0 1.5 dB For 128 kbps PS data and mobilespeed of 3 km/h.

1.0 dB For 384 kbps PS data and mobilespeed of 3 km/h.

Activity factor v 0.2 Applies if downlink traffic orientedapplication is used, for example webbrowsing.

Other to own cell interference i 0.65 Applies for macrocells.

0 20 40 60 80 100 1200

1

2

3

4

5

6

7

8

9

10

Number of users

No

ise

ris

e [

dB

]

Interference margin vs. number of users in the cell

128 kbps

384 kbps

Figure 5.6: Noise rise in the base station receiver as a function of users in the cell.

In the Figure 5.6 it can be seen that, for example, with cell load of 75% the cellcan support approximately 5-10 users less (depending on the uplink data rate) if the

5. Numerical Analysis for System Level Performance Evaluation 47

uplink noise level rises by 1 dB. Although the values for v and i do not necessarilycorrespond to ones that were used in the simulation scenarios in this thesis, the idearemains the same: 1 dB noise rise causes tolerable capacity loss in terms of numberof users. In addition, deployment of repeaters could also expect to decrease theiUL by mitigating the other-cell interference in respect to the own-cell power in therepeater service areas, and thus decrease the loading.

5.4 Theoretical Behavior of Multiple Repeaters

To investigate how multiple repeaters could expect to affect the uplink performanceof a cell, theoretical calculations were performed. The effective base station noisefigures were calculated with (4.6). In the calculations, both scenarios, rural andsuburban, were assumed with the similar parameters that were introduced earlier(see Table 5.10). It was assumed that the repeater distances from the base stationremained constant for each repeater. The antenna gains used in the calculationsinclude the base station antenna loss in horizontal plane, if the repeater donorantenna misaligned from the bore-sight of the base station antenna main lobe. Thebase station antenna losses in horizontal plane are presented in Figures 5.7(a) and5.7(c). In Figures 5.7(a) and 5.7(c) it can be seen how the antenna losses increasethe misalignment from the bore-sight of the BS antenna increases.

Table 5.10: Equipment configurations used in the suburban simulations.

Rural Suburban

Carrier frequency 900 MHz 2100 MHzBS noise figure 4 dB 4 dBBS antenna gain 14.9 dBi 18 dBiRepeater noise figure 3 dB 3 dBDonor antenna gain 20.57 dBi 20.99 dBiRepeater distance 2400 m 400 m

The results for the calculations are shown in Figures 5.7(b) and 5.7(d). In thecalculations, different to the simulations was that the repeater gain was set constantfor each repeater, since the purpose was to show how the effective base station noisefigure behaves as a function of repeater gain. In the 2-repeater case, clearly lowerrepeater gain values can be used than it might have expected. For example, in therural scenario it seems that two repeater donor antennas locating at the bore-sightof the BS antenna would cause approximately the same noise rise as seven repeaterswith similar repeater gains. This is due to the repeater donor antenna locationswhich correspond to 1-repeater case, as mentioned in Section 5.2. Thus, the antennalosses are minimal, and the summed noise amplification from two repeaters is high.

5. Numerical Analysis for System Level Performance Evaluation 48

In the suburban case, according to Figure 5.7(d), much lower repeater gains areexpected to be able to be used than in the rural case. This is mostly due to muchshorter repeater distances.

0 10 20 30 40 500

1

2

3

4

5

6

7

An

ten

na

Lo

ss [

dB

]

Horizontal Angle [degrees]

Base Station Antenna Lossvs.

Misalignment Angle from Antenna Bore−sight (Rural)

(a) Base station antenna loss in horizontalplane, rural scenario.

45 50 55 60 654

4.2

4.4

4.6

4.8

5

5.2

5.4

5.6

5.8

6

Eff

ective

Ba

se

Sta

tio

n N

ois

e F

igu

re [

dB

]

Repeater Gain GR

[dB]

Effective Base Station Noise Figurevs.

Repeater Gain (Rural)

1 repeater

2 repeaters

3 repeaters

5 repeaters

7 repeaters

9 repeaters

(b) Effective base station noise figure in respectto repeater gain, rural scenario.

0 10 20 30 40 500

1

2

3

4

5

6

7

An

ten

na

Lo

ss [

dB

]

Horizontal Angle [degrees]

Base Station Antenna Lossvs.

Misalignment Angle from Antenna Bore−sight (Suburban)

(c) Base station antenna loss in horizontalplane, suburban scenario.

45 50 554

4.2

4.4

4.6

4.8

5

5.2

5.4

5.6

5.8

6

Eff

ective

Ba

se

Sta

tio

n N

ois

e F

igu

re [

dB

]

Repeater Gain GR

[dB]

Effective Base Station Noise Figurevs.

Repeater Gain (Suburban)

1 repeater

2 repeaters

3 repeaters

(d) Effective base station noise figure in respectto repeater gain, suburban scenario.

Figure 5.7

49

6. RESULTS

This chapter provides the results for the scenarios that were defined in Chapter5. The main results are given in two formats: as statistical plots and networkmaps. Statistical plots illustrate the averages of simulated (RSCP or SINR) valuesor the improvement provided by the repeaters for a given metric. The statisticsare presented as a function of repeater-to-site distance or number of repeaters. Thenetwork maps either present a simulated value, in each map pixel, or a difference indecibels between situations where repeaters are not in use and a given number ofrepeaters are in use in each cell. The purpose of the maps is to clarify the results thatare given as statistical plots. In the end, an error analysis is given for the researchwork. The error analysis covers the whole work from defining the topologies tosetting the simulation parameters, and the credibility of the results.

6.1 Rural Scenario

This section provides the results for the rural scenario as a main purpose to presentthe change in the cell coverage (that is, received pilot channel power) due to differentamount of repeaters installed within the cell. All the results in this section dealwith the RSCP values. One of the observed results was the coverage improvement,although it is not trivial to say what is a good criteria for coverage improvement. Inthis case, the coverage improvement was defined as a three-decibel rise in the RSCPvalue, which implies to doubled power in the linear scale.

At first, Figures 6.1(a) - 6.1(c) show the RSCP coverage when no repeaters, one,and three repeaters are installed in each cell. It can be seen that the rural scenariois coverage limited, since the RSCP value in regions without a clear cell dominancefalls down to around -100 dBm. Moreover, adding repeaters in the network increasesthe coverage, but the improvement areas for each repeater are smaller when morerepeaters are installed.

In Figure 6.2(a), the change in the average RSCP value (in respect to site config-uration without repeaters) is given as a function of installed repeaters in each cell.Here on, the relative difference in RSCP value in respect to case without repeaters isreferred to as delta RSCP. The delta RSCP values were calculated over the studiedarea that was described earlier in Chapter 5. It seems that the number of repeatersdoes not reflect much to the delta RSCP, but still small differences between different

6. Results 50

(a) (b)

(c)

Figure 6.1: RSCP maps in rural scenario with no repeaters (a), 1 repeater (b) and 3repeaters (c) deployed in each cell.

1 2 3 4 5 6 7 8 90.5

0.6

0.7

0.8

0.9

1

1.1

1.2

1.3

1.4

1.5

Number of repeaters

∆R

SC

P (

aver

age)

[dB

]

Change in average RSCP vs. number of repeaters

(a)

1 2 3 4 5 6 7 8 920

25

30

35

40

45

Number of repeaters

Impro

ved

cover

age

area

[%

]

Improved RSCP coverage area vs. number of repeaters − 3 dB threshold

(b)

Figure 6.2: Improvement in average RSCP (a) and RSCP coverage (b) in respect to thenumber of repeaters.

number of installed repeaters can be noticed. However, when investigating the cov-erage improvement, that is presented in Figure 6.2(b), much clearer difference can

6. Results 51

be noticed if more than one repeater is installed in each cell area: one repeater pro-vides slightly under 30% improvement, whereas five repeaters provide approximately42.5% improvement. Installing more than five repeaters in the cell seems to startdecreasing the coverage improvement. Based on Figures 6.2(a)-6.2(b), an indicationfor optimum repeater configuration for coverage improvement can be deduced.

Noticeable is that the case in which two repeaters are installed in a cell clearly dif-fers from the expected value. However, this was already proposed in the theoreticalplot of the behavior of EFB in Figure 5.7(b), which implied that the two-repeatercase would behave somewhat differently in respect to the expected trend. In addi-tion, Figure 6.3 illustrates the distribution of RSCP values over the studied area.It supports the results in Figures 6.2(a) and 6.2(b): at the areas with low recep-tion level RSCP, two repeaters provide the lowest improvement in the RSCP values,whereas at areas with strong signal level each step of adding a repeater seems toprovide additional gain in received power. Also, if more than three repeater are in-stalled in a cell, no significant additional improvement in RSCP performance can beachieved. The behavior of the RSCP cumulative distribution function (CDF) can beexplained with the delta RSCP maps in Figure 6.4. The maps in Figure 6.4 show thedifference in the RSCP values with repeaters installed to the cells in respect to casewhen no repeaters are installed. In one repeater case the coverage improvement areaof a single repeater is larger than of a single repeater in any multi-repeater case.On the other hand, in multi-repeater cases, the regions with higher RSCP levelsclearly increase, but also the repeater service areas clearly become smaller. In 3-and 5-repeater cases it can be also noticed that the repeater service area is smallerfor the repeater located in the bore-sight of the BS antenna than of the ones thatare located out of the bore-sight.

−100 −90 −80 −70 −60 −500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

CPICH RSCP [dBm]

CPICH RSCP Cumulative Distribution Function (over studied area)

No repeaters

1 repeater

2 repeaters

3 repeaters

5 repeaters

9 repeaters

Figure 6.3: Cumulative distribution function for CPICH RSCP with different number ofrepeaters.

6. Results 52

(a) (b)

(c) (d)

Figure 6.4: Change in RSCP in respect to no repeaters deployed in the cells. 1 repeater(a), 3 repeater (b), 5 repeaters (c) and 9 repeaters (d) deployed in each cell.

When combining all the information provided by the maps and the statisticalplots, an optimal number of repeaters can be pointed out to be three to five repeaterin a cell, in the sense of coverage optimization. It was noticed that observing onlythe average RSCP values over the studied area is not enough, since the differencesbetween different repeater configurations were rather small when the RSCP valuewas averaged over the studied area. Complementing this with illustrative mapsand evaluating coverage improvement on pixel basis added valuable information forevaluating the optimum solution. in summary, the improved coverage area andimprovement in average RSCP with different repeater configurations are presentedin Table 6.1. The highest value for the improved coverage area (42.4%), and also forthe improved average RSCP (1.23 dB), were achieved with five repeaters per cell.

6. Results 53

Table 6.1: The conclusive results of rural scenario.

1repe

ater

2repe

aters

3repe

aters

5repe

aters

7repe

aters

9repe

aters

Impr. coverage area [%] 28.9 23.7 40.8 42.4 39.8 37.2Impr. in average RSCP [dB] 1.19 1.16 1.21 1.23 1.19 1.19

6.2 Suburban Scenario

In the suburban scenario, the interest was focused on the SINR to estimate the pos-sible capacity changes in HSDPA as a result of multi-repeater deployment. Resultedplots are based on the average SINR value over the studied area, and the cell areathat experienced SINR improvement. The threshold value for SINR improvement isvaried between zero to three decibels (where zero decibels is in practice just a smallenough number to avoid inaccuracy problems to denote the result when subtractingtwo numbers in MATLABTM). The plots are presented both as a function of re-peater distance, and as a function of number of repeaters. The results are supportedwith maps illustrating the SINR values, and the SINR improvement areas.

In Figure 6.5 is illustrated the SINR distribution when repeaters are not installedin the network. Figures 6.6(a) - 6.6(d) further illustrate the instances when oneand three repeaters are installed in the distances of 200 meters and 500 metersfrom the BS sites. Figures 6.6(a) and 6.6(c) show quite well how the shape of theSINR dominant area for each cell is modestly changed when the repeater distance isshort. Furthermore, with the maximum repeater distance the change is very notice-able. With the maximum repeater distance the repeater coverage clearly reaches theneighboring cell areas and cause other-cell interference (shown in Figures 6.6(b) and6.6(b)). It can be also seen in Figure 6.6, that in the suburban scenario, SINR valuesat the cell edges are approximately eight to ten decibels, which is according to [17]sufficient level to provide average single-user throughput of 1 Mbps with five codesallocated to a user. To double the average throughput (with five codes), the averageSINR values should be improved by five decibels. One decibel improvement wouldprovide 100-200 kbps improvement in average throughput. If ten or fifteen codeswere allocated to a user, the corresponding improvements in throughput would beeven higher.

Figure 6.7 illustrates the improvement of average SINR values over the studiedarea. Some trend on the average SINR behavior with different repeater configura-tions can be deduced, but again similar to average RSCP for the rural case presentedin Section 6.1, the differences between different cases are rather small varying within

6. Results 54

Figure 6.5: SINR map, no repeaters installed.

(a) (b)

(c) (d)

Figure 6.6: Change in SINR in respect to no repeaters deployed in the cells. 1 repeaterwith 200 m repeater distance (a), 1 repeater with 500 m repeater distance (b), 3 repeaterswith 200 m repeater distance (c) and 3 repeaters with 500 m repeater distance (d) deployedin each cell.

6. Results 55

200 250 300 350 400 450 5000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Repeater−to−BS distance [m]

∆ S

INR

[d

B]

Average SINR improvement over studied area vs. repeater−to−BS distance

1 repeater

2 repeaters

3 repeaters

Figure 6.7: Improvement in average SINR (in dB) over the studied area as a function ofrepeater distance.

half decibels. However, local improvement in repeater service areas is much higher,as it can be seen later in Figure 6.12.

The improved SINR area is illustrated in Figures 6.8(a) - 6.8(d) with differentSINR improvement threshold values (zero to three decibels). A clear trend for thearea that experiences SINR improvement is visible in Figures 6.8(b) and 6.8(c):the improved area starts to saturate around 300-400 meter repeater distance. Thesame applies partly in Figure 6.8(d), except for the 2-repeater case. Instead, 2-repeater case still shows rising trend in improved SINR area even with 500 m repeaterdistance. Also in the Figures 6.8(a) - 6.8(c) it can be seen that the 2-repeater caseprovides the best SINR area improvement with long repeater distances. In Figure6.8(a) it can be also seen that when repeaters are installed near the BS antenna,larger area experiences gain in SINR if the SINR improvement is evaluated with alow threshold. When the threshold is raised, short repeater distances do not providemuch improvement in SINR (Figure 6.8(b)). On the other hand, when repeaters arebrought near cell edges, higher repeater gains are used, but SINR improvement startsto saturate due to other-cell interference (Figures 6.8(b) and 6.8(c)). Additionally,in Figures 6.8(b) - 6.8(d) it can be seen that 1-repeater case provides the best resultsin improved SINR area with repeater distances of 300-400 meters (evaluated withover 1 dB improvement threshold).

Similar figures were also plotted for the SINR behavior as a function of installednumber of repeaters. Figures 6.9 and 6.10 give some idea for the optimum repeaterdistance for each repeater configuration. In Figure 6.9, average SINR improvement

6. Results 56

200 250 300 350 400 450 50025

30

35

40

45

50

55

60

65

70

75

Repeater−to−BS distance [m]

Imp

rov

ed S

INR

are

a [%

]

Improved SINR area vs. repeater−to−BS distance − threshold 0 dB

1 repeater

2 repeaters

3 repeaters

(a)

200 250 300 350 400 450 50015

20

25

30

35

40

Repeater−to−BS distance [m]

Imp

rov

ed S

INR

are

a [%

]

Improved SINR area vs. repeater−to−BS distance − threshold 1 dB

1 repeater

2 repeaters

3 repeaters

(b)

200 250 300 350 400 450 5000

5

10

15

20

25

30

35

Repeater−to−BS distance [m]

Imp

rov

ed S

INR

are

a [%

]

Improved SINR area vs. repeater−to−BS distance − threshold 2 dB

1 repeater

2 repeaters

3 repeaters

(c)

200 250 300 350 400 450 5000

5

10

15

20

25

Repeater−to−BS distance [m]

Imp

rov

ed S

INR

are

a [%

]

Improved SINR area vs. repeater−to−BS distance − threshold 3 dB

1 repeater

2 repeaters

3 repeaters

(d)

Figure 6.8: Improved SINR area with different thresholds for SINR: threshold of 0 dB (a),1 dB (b), 2 dB (c) and 3 dB (d).

over the studied area is presented as a function of installed repeaters. It shows thesimilar results as Figures 6.8(a) - 6.8(d): with longer repeater distances two repeatersprovide the highest improvement, whereas with 200-meter repeater distance tworepeaters provide the lowest improvement. It can also be seen that with 200- and300-meter distances one and three repeaters provide about the same average SINRimprovements. In Figures 6.10(a) and 6.10(b) improved SINR area is shown inrespect to the number of installed repeaters. With lower threshold (Figure 6.10(a)),200 meter repeater distance can be seen to provide clearly the worst results inimproved SINR area, and the best result is obtained with two repeaters at distanceof 400 meters. Moreover, repeater distances of 300-400 meters seem to provide thebest values with all the regardless of the number of repeaters. On the other hand,in Figure 6.10(b) it is shown that 400-500 meters for repeater distance providenoticeable better results in respect to shorter repeater distances. In Figures 6.8and 6.10, it can also be noticed that choosing the threshold value to evaluate the

6. Results 57

improvement in the SINR performance area has to be done carefully, or the resultshave to be analyzed with different threshold values, and using SINR maps as help.Furthermore, CDF for the SINR values were plotted (Figure 6.11) to get some ideahow the repeater configuration affects on the SINR distribution over the studiedarea. Based on Figure 6.11, no significant changes can be observed between differentconfigurations, and the minor differences correlate quite well with the results thatwere seen in Figure 6.7.

1 2 30

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Number of repeaters

∆ S

INR

[dB

]Change in average SINR over studied area vs. number of repeaters

Distance 200 m

Distance 300 m

Distance 400 m

Distance 500 m

Figure 6.9: Change in average SINR in the studied area as a function of installed repeaters.

1 2 310

15

20

25

30

35

40

45

50

Number of repeaters

Impro

ved

are

a [%

]

Improved SINR area vs. number of repeaters − threshold 1 dB

Distance 200 m

Distance 300 m

Distance 400 m

Distance 500 m

(a)

1 2 3−5

0

5

10

15

20

25

30

35

Number of repeaters

Impro

ved

are

a [%

]

Improved SINR area vs. number of repeaters − threshold 3 dB

Distance 200 m

Distance 300 m

Distance 400 m

Distance 500 m

(b)

Figure 6.10: Area that experienced SINR improvement in respect to the case where norepeaters were installed in the cells: threshold of 1 dB (a) and 3 dB (b) for the improvement.

When observing the SINR difference maps (Figures 6.12(a) - 6.12(d)) for the samecases that were illustrated in Figure 6.6, it is shown that the SINR improvementarea is much larger when repeaters are installed close to the sites (as it was seen inFigures 6.8 and 6.10), but with much smaller peak improvement in the SINR values.The improvement is rather spread over a larger area, whereas with longer repeater

6. Results 58

6 8 10 12 14 16 18 200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

SINR [dB]

CD

F

DL SINR Cumulative Distribution Function (over studied area)

No repeaters

1 repeater

2 repeaters

3 repeaters

(a)

5 10 15 200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

SINR [dB]

CD

F

DL SINR Cumulative Distribution Function (over studied area)

No repeaters

1 repeater

2 repeaters

3 repeaters

(b)

Figure 6.11: Cumulative distribution functions for downlink SINR values with repeaterdistance of 300 meters (a) and 400 meters (b).

(a) (b)

(c) (d)

Figure 6.12: Change in SINR in respect to no repeaters deployed in the cells. 1 repeaterwith 200 m repeater distance (a), 1 repeater with 500 m repeater distance (b), 3 repeaterswith 200 m repeater distance (c) and 3 repeaters with 500 m repeater distance (d) deployedin each cell.

6. Results 59

distances the improvement affects a smaller area with higher improvement. In both,1- and 3-repeater, cases the average improvement over the studied area does notdiffer too much (as it was seen in Figures 6.7 and 6.9), but the concentrated areaand improvement on the corresponding repeater service area behaves differentlywhen the repeater distance varies. This can be explained with the repeater gainvalues, that are lower for the repeaters installed near the BS sites. Also, near theBS signal from the mother cell is stronger. Improved SINR area reduces in size withlong repeater distances due to interference coming from the neighbor cells (repeatersbegin to overlap with the neighbor cells and neighboring repeaters from the the othercells, see Figure 6.12).

Finally, the key values from the results are collected in the Table 6.2. It presentsthe values for the improved SINR area (when threshold value for improvement isset to 3 decibels), SINR improvement over the studied area, and also improvementin average SINR in repeater serving area. The highest values for average SINRimprovement in the repeater service areas are obtained with 1- and 3-repeater caseswhen the repeater distance is 400 meters. They also provide the highest valuesfor the improved SINR area (20.7% and 16.7%, respectively), if the improvementthreshold is considered to be 3 decibels. However, it should be noted, that, with a3-decibel threshold, the highest improvement in SINR area was achieved with tworepeaters per cell at distance of 500 meters from BS (see Figure 6.10(b)). Lookingat the improvement over the whole studied area, the highest value is provided by2-repeater case, and the second highest value by 1-repeater case (in both cases therepeater distance was 400 meters).

Table 6.2: The results for suburban scenario (with 3 dB threshold for the improved area).

1 repeater 2 repeaters 3 repeaters

300m

400m

300m

400m

300m

400m

Improved SINR area [%] 9.2 20.7 0.3 12.4 9.3 16.7Improvement in average SINR [dB] 0.63 0.71 0.59 0.74 0.66 0.53(whole studied area)Improvement in average SINR [dB] 2.31 2.51 2.16 2.33 2.24 2.37(repeater serving area, no threshold)

6. Results 60

6.3 Error Analysis

The error sources for this thesis work can be divided roughly into three sources: theradio interface planning approach, simulator tool inaccuracies, and ideal approxima-tions for several simulated parameters and components. When evaluating systemlevel performance and estimating RSCP and SINR values for a given map pixel,there are several possible error sources that should be acknowledged. First of all,estimating path loss values with a semi-empirical model, as it was done in thesesimulations, gives only more or less sufficient estimate for the path losses. There areseveral different semi-empirical models for the studied scenarios (rural and subur-ban), and each model could provide slightly different results for path losses. Createdtopologies do not either correlate to any specific realistic terrain in detailed level,and thus a common semi-empirical model provides sufficient approximation for pathloss values in the simulations. However, in this thesis, the results were taken as rel-ative values, which reduced the error by neglecting the absolute estimated values.Moreover, Okumura-Hata and COST-231-Hata are nowadays the most commonlyused models, and thus offer the best comparability to other UMTS radio interfacerelated macrocell studies. Also estimating the orthogonality factor with the givenmodel causes some error in the final SINR values, but the error is smaller comparedto the case when using the default fixed average value. More significant error sourcefor the calculated SINR values can be the model, that was used in the analysis (seeSection 5.1). The current model takes only the best server link into account, whichleads to inaccuracies if the repeater serving areas are overlapping within the samemother cell. Hence, in the created topologies such configurations were avoided, andthereby also the SINR analysis model applies with minimal error.

An issue that was related to the simulation tool and has influence on the statis-tical reliability of the results, is the used resolution. For the performed simulations,resulted maps consisted of approximately 90 000 pixels, of which the studied areacovered about fifteen percent. This causes some averaging in the performance met-rics that were calculated in the simulations. On the other hand, it has to be notedthat the current simulation implementation is rather inefficient when it comes to thememory usage of the system. The problem was faced especially when the number ofrepeaters in the network was increased up to five repeaters per cell or more, becausethe memory usage increases significantly. Therefore, the chosen map resolutionswere trade-offs between statistical reliability and simulator tool performance.

Assuming ideal operation of certain elements causes some error to the results.For example, repeaters in this thesis are assumed to operate fully linearly with idealbandpass filtering for the relayed channels. Also, a full antenna isolation for therepeater antennas would not occur in practice. Other error sources, related to ideal-

6. Results 61

ization, are simplified interference modeling when calculating other-cell and own-cellinterference, and several approximations in the multi-repeater model for calculat-ing the effective base station noise figure. However, for example in dimensioningpurposes the error from these idealizations should be rather small in system levelsimulations, especially considering the performance of the whole network. With therepeater gains, that were used in the simulations, the antenna isolation can be as-sumed to remain good, and full isolation can be assumed as an approximation. Inthe noise figure calculations, assuming the noise temperatures at different antennainputs the same is a way to simplify the model to be feasible to formulate for thesimulator. In practice, the antenna noise temperature would vary in each antenna,but if the differences in noise temperatures are expected to be small enough, theapproximation is valid.

It could also be discussed whether the scenarios are realistic enough to providesufficient data for repeater performance evaluation. The main idea in the simulationswas to roughly estimate feasible scenarios and the radio propagation parameters tocorrespond the propagation environments with values that were available in theliterature. Terrain flatness and homogenous distribution of sites and repeaters israther ideal, and hardly ever occurs in real life radio network planning - alreadydue to cell breathing, terrain variations and shadowed regions, and inaccuracies inconfiguration installations. In addition, from the research point of view, the re-peater antenna directions and locations should have been chosen differently for the2-repeater case, since the configuration symmetry clearly differs from the idea of theother configurations. This in turn makes it more difficult, if not impossible, to com-pare the 2-repeater cases in both rural and suburban scenarios to the other repeaterconfigurations. However, in this case the 2-repeater case provides information on theeffect of different antenna configuration approach for repeater site installation. Mostlikely, results for 2-repeater case with symmetrical repeater installation, similar toother multi-repeater cases, could be interpolated between 1- and 3-repeater cases.

62

7. CONCLUSIONS AND DISCUSSION

In this thesis, multi-repeater performance was studied in rural and suburban macro-cellular scenarios. In the rural scenario, the main interest was to study the coverageimprovement provided by multiple repeaters installed in a cell. In the suburban sce-nario, SINR was chosen as the studied performance metric. The simulation resultsshow the difference between the different multi-repeater configurations that werestudied. Applying a different number of repeaters in the mother cell affects bothon coverage and capacity of the base station. In the macrocellular rural scenario,where BS site-to-site distances are long, multiple repeaters can provide additionalcoverage. In suburban cells, the repeater distance has a clear effect on the the-oretical HSDPA performance measured by SINR, but the difference between theimprovement provided by one to three repeaters in a cell is rather small.

The results for the rural scenario propose that the optimum number of repeaterswould be approximately five repeaters in a cell, based on the pilot coverage, provid-ing up to 42% improvement in the coverage area and 1.23 decibels improvement inthe average RSCP. When more repeaters are installed in a cell, the area that experi-ences more than a three-decibel improvement in RSCP begins to decrease. However,3-, 7- and even 9-repeater cases still provide better results in comparison to one in-stalled repeater in a cell (with coverage improvements of 40.8%, 39.8% and 37.2%,respectively). For 1-repeater case less than 30% coverage improvement was shown.On the other hand, the average RSCP value does not seem to experience noticeablechanges between different repeater configurations. When less repeaters are installedin the cells, higher repeater gains can be utilized, but the improved coverage areabecomes more limited. Installing more repeaters by distributing them evenly in thecell region decreases the usable repeater gains – if the uplink performance is notwanted to suffer from significant noise rise. Decreased repeater gain values reflect tothe average RSCP value. At the same time, higher number of repeaters are able tocover larger area. However, eventually a dense repeater employment causes overlap-ping in the repeater serving areas, which reduces the overall coverage improvement.Moreover, in the 2-repeater case, it was seen that the used configuration was notan effective way to improve coverage. In contrary, this configuration reduces thecoverage improvement by approximately five percentage units in comparison to 1-repeater case. This is due to lower usable repeater gain values. In the 2-repeater

7. Conclusions and Discussion 63

configuration, all donor antennas were located at the bore-sight of the BS antennaand the serving antennas were spreading the relayed signal more towards the sidesof mother cell.

Evaluation of multi-repeater performance in the suburban scenario gave inter-esting results, especially considering the antenna configurations. According to theresults, the used 2-repeater configuration seems to perform quite well with long re-peater distances. The highest improvement in average SINR (over the whole studiedarea), 0.74 decibels, was achieved with two repeaters when the repeater distance was400 meters from the serving BS. Thus, it seems that the repeater antenna directionscould have effect on the SINR performance. In the studied suburban scenario, in2-repeater case, the repeater antenna directions were similar to the ones that wereused in the rural scenario. This seems to be beneficial in suburban macrocell envi-ronment, in comparison to cases where the repeater serving antennas are pointingstraight out from the cell area to the cell edge. A practical reason to this couldbe, that in the 2-repeater case, the other-cell interference coming from the othercell-border repeaters is smaller, in comparison, for example, to the 3-repeater case.However, the best values for improved SINR area were achieved with 1- and 3-repeater cases (repeater distance being 400 meters): 20.7% and 16.7%, respectively.For comparison, with two repeaters the value for improved SINR area was 12.4%with the same 400-meter repeater distance.

When excluding the 2-repeater case, three repeaters at distance of 300 - 400 me-ters seems to be the next best choice. On the other hand, the differences betweenall the cases are rather small. At 300-meter distance, three repeaters offer the bestimprovement in the lower end of SINR values. It could be said that with shorterrepeater distances the SINR improvement spreads in larger area over the cells, butlower repeater gains keep the improvement modest. Longer repeater distance allowsusing higher repeater gains, and the improvement in SINR is much higher in therepeater service areas. Furthermore, when the repeaters locate at the cell edge,the received signal power from the mother cell is smaller, which reflects to a higherimprovement in SINR in repeater serving areas. Due to other- and own-cell in-terference, however, the improved area becomes limited. Hence, the average SINRover the studied area does not vary much between different configurations. Froman other point of view, this could also imply, that installation of multiple repeatersin suburban cells with feasible repeater-to-site distances, and while choosing therepeater gains carefully, would not introduce significant capacity or coverage lossin the uplink. Thereby, utilizing multiple repeaters in severely shadowed regionswithin the same mother cell would be considerable alternative to denser BS sitedeployment, in cost-optimization point of view. Similarly, there is also potentialto utilize multiple repeaters in outdoor-to-indoor situations to provide performance

7. Conclusions and Discussion 64

improvement inside buildings. For example, reflecting the results of this thesis tooutdoor-to-indoor performance optimization, three repeaters could be used to pro-vide performance improvement inside buildings in a suburban macrocell. That is,three buildings in a cell could be covered with outdoor-to-indoor repeaters.

The simulations gave encouraging examples of multi-repeater performance bothin rural and suburban macrocells, when uplink noise rise limit was pre-defined. Eventhough a clear optimum solution was not found for the suburban scenario, the resultcan be also said to be that installing multiple repeaters in a suburban cell improvesslightly the system level performance, and at least two or three repeaters could beinstalled in a cell to improve the average performance. Moreover, the repeaters couldprovide a clear SINR improvement, when they are installed near the cell edge. Atthe same time, attention should be paid to setting the repeater gains to avoid noiserise in the uplink.

This work provided some insight to multi-repeater performance in two differentscenarios, but there is still plenty to do as further work. The results of this thesisshould be verified with real-life measurements in some extend, so that the resultscould be confirmed to be suitable to be applied in practice. In this thesis, allthe scenarios had idealized terrains with homogenous repeater and base stationsite distributions. Additionally, the simulations could be continued with differentkind of scenarios and multi-repeater configurations. One interesting application,that could be added as a simulation scenario, would be outdoor-to-indoor coverageand capacity simulations, which, however, requires additional implementations tothe MATLABTMtool. It would be interesting to see the upper limit for multiplerepeaters in a cell, and this the maximum number of buildings that could be servedin suburban, or urban, cells. The simulations could be also extended to cover systemlevel issues, such as RRM and scheduling, but this is not possible with the currentsimulator, and implementing this would require a tremendous amount of work.

65

BIBLIOGRAPHY

[1] E. Dahlman, S. Parkvall, J. Sköld, and P. Beming, 3G evolution: HSPA andLTE for mobile broadband. Elsevier Ltd., 2007.

[2] J. Korhonen, Introduction to 3G Mobile Communications. Artech House, Inc.,2001.

[3] A. R. Mishra, Ed., Advanced Cellular Network Planning and Optimisation:2G/2.5G/3G...Evolution to 4G. John Wiley & Sons, 2007.

[4] J. Laiho, A. Wacker, and T. Novosad, Radio Network Planning and Optimisa-tion for UMTS, 2nd ed. John Wiley & Sons, 2006.

[5] The Third Generation Partnership Project. Accessed on 7.3.2010. [Online].Available: http://www.3gpp.org/

[6] The Third Generation Partnership Project 2. Accessed on 7.3.2010. [Online].Available: http://www.3gpp2.org/

[7] GSM World. Accessed on 24.2.2010. [Online]. Available:http://www.gsmworld.com/

[8] E. Lee and D. Messerscmitt, Digital communication, 1st ed. Kluwer AcademicPublishers, 1988.

[9] S. R. Saunders and S. R. Simon, Antennas and Propagation for Wireless Com-munication Systems. New York, NY, USA: John Wiley & Sons, Inc., 1999.

[10] J. Lempiäinen and M. Manninen, Radio Interface System Planning forGSM/GPRS/UMTS. Kluwer Academic Publishers, 2001.

[11] Y. Okumura, E. Ohmori, T. Kawano, and K. Fukuda, “Field Strength and itsVariability in VHF and UHF Land Mobile Service,” Rev. Elec. Comm. Lab,vol. 16, p. 825, September-October 1968.

[12] M. Hata, “Empirical formula for propagation loss in land mobile radio services,”Vehicular Technology, IEEE Transactions on, vol. 29, no. 3, pp. 317 – 325,August 1980.

[13] J. S. Seybold, Introduction to RF Propagation. Wiley-IEEE, 2005.

[14] E. Damasso and L. M. Correia, Eds., Digital Mobile Radio towards FutureGeneration Systems, COST 231 Final Report, 1999.

BIBLIOGRAPHY 66

[15] W. Stallings, Wireless Communications and Networks. Prentice Hall Profes-sional Technical Reference, 2001.

[16] H. Holma and A. Toskala, WCDMA for UMTS: HSPA Evolution and LTE,4th ed. John Wiley & Sons, 2007.

[17] H. Holma and A. Toskala, HSDPA/HSUPA for UMTS: High Speed Radio Accessfor Mobile Communications. John Wiley & Sons, 2006.

[18] 3GPP, “Physical layer; Measurements (FDD),” 3rd Generation PartnershipProject (3GPP), TS 25.215 v.9.1.0, Dec. 2009. [Online]. Available:http://www.3gpp.org/ftp/Specs/html-info/25215.htm

[19] H. Holma and A. Toskala, WCDMA for UMTS - Radio Access for Third Gen-eration Mobile Communications, 3rd ed. John Wiley & Sons, 2004.

[20] J. Shapira and S. Y. Miller, CDMA Radio with Repeaters. Springer PublishingCompany, Incorporated, 2007.

[21] M. Rahman and P. Ernström, “Repeaters for hotspot capacity in DS-CDMAnetworks,” Vehicular Technology, IEEE Transactions on, vol. 53, no. 3, pp.626–633, May 2004.

[22] P. Lähdekorpi, J. Niemelä, J. Borkowski, and J. Lempiäinen, “WCDMA Net-work Performance in Variable Repeater Hotspot Traffic Cases,” in 6th IEEInternational Conference on 3G and Beyond, 2005.

[23] 3GPP, “Universal Terrestrial Radio Access (UTRA) repeater planning guide-lines and system analysis,” 3rd Generation Partnership Project (3GPP), TR25.956, Mar. 2008. [Online]. Available: http://www.3gpp.org/ftp/Specs/html-info/25956.htm

[24] Qualcomm, “Using Power Control to Improve Repeater Operation,” White pa-per, 2003.

[25] P. Rathinavelu, G. Schapeler, and A. Weber, “HSDPA Performance Enhance-ment Using Repeaters and Remote RF Heads,” in Personal, Indoor and MobileRadio Communications, 2006 IEEE 17th International Symposium on, Sept.2006, pp. 1–5.

[26] J. Niemelä, J. Borkowski, and J. Lempiäinen, “Assessment of repeaters forWCDMA UL and DL performance in capacity-limited environment,” in Proc.of14th IST Mobile Summit, June 2005, 2005.

BIBLIOGRAPHY 67

[27] NPSW - Network Planning Strategies for Wideband CDMA, v.5.0.0, Applica-tion note.

[28] P. Lähdekorpi, “Effects of Repeaters on UMTS network Performance,” Master’sthesis, Tampere University of Technology, 2006.

[29] P. Lähdekorpi, J. Itkonen, and J. Lempiäinen, “Comparison of RF-RepeaterEfficiency in Macrocellular Network Topologies,” in 16th European WirelessConference, Italy, April 2010.

[30] N. B. Mehta, A. F. Molisch, and L. J. Greenstein, “Macrocell-Wide Behaviorof the Orthogonality Factor in WCDMA Downlinks,” IEEE Transactions onWireless Communications, vol. 5, no. 12, pp. 3394–3399, 2006.

[31] Kathrein Scala Division. Accessed on 22.2.2010. [Online]. Available:http://www.kathrein-scala.com/