Uplink Interference Management of High Bit Rate Users in Evolved WCDMA

(1)

Institutionen för systemteknik

Department of Electrical Engineering

Examensarbete

Uplink Interference Management of High Bit Rate

Users in Evolved WCDMA

Examensarbete utfört i Kommunikationssystem vid Tekniska högskolan i Linköping

av

Samuel Axelsson

LITH-ISY-EX-3706-2005 Linköping 2005

Department of Electrical Engineering Linköpings tekniska högskola

Linköpings universitet Linköpings universitet

(2)

(3)

Uplink Interference Management of High Bit Rate

Users in Evolved WCDMA

Examensarbete utfört i Kommunikationssystem

vid Tekniska högskolan i Linköping

av

Samuel Axelsson

LITH-ISY-EX-3706-2005

Handledare: Erik Geĳer Lundin

isy, Linköpings universitet

Gunnar Bark

Ericsson Research, Linköping Examinator: Fredrik Gunnarsson

isy, Linköpings universitet

(4)

(5)

Avdelning, Institution

Division, Department

Division of Automatic Control Department of Electrical Engineering Linköpings universitet S-581 83 Linköping, Sweden Datum Date 2005-06-03 Språk Language Svenska/Swedish Engelska/English ⊠ Rapporttyp Report category Licentiatavhandling Examensarbete C-uppsats D-uppsats Övrig rapport ⊠

URL för elektronisk version

http://www.control.isy.liu.se

ISBN

—

ISRN

LITH-ISY-EX-3706-2005

Serietitel och serienummer

Title of series, numbering

ISSN

—

Titel

Title Interferenshantering med förbättrad WCDMA-upplänk för användare med högdatatakt Uplink Interference Management of High Bit Rate Users in Evolved WCDMA

Författare

Author Samuel Axelsson

Sammanfattning

Abstract

The WCDMA air interface, used in the third generation mobile communication systems, is currently being evolved to improve the uplink, i.e. the radio links car-rying traffic from the mobile user to the fixed network. An enhanced uplink concept is being developed to meet the expected needs from future applications like mul-timedia and video-streaming. This thesis studies interference management when high bit rates are introduced in the enhanced uplink. The study is performed through theoretical assessments and simulations using WCDMA system simula-tors.

An optimization scheme using a basic system throughput based scheduling is derived to attain a theoretical assessment of bit rate limits. The throughput optimization is achieved at the expense of user-experienced fairness. Users located on cell coverage area overlap show to be most complicated to manage.

The need for interference management is primary when the network deploy-ment consists of small cells while coverage requiredeploy-ments are most essential when the cell size increases. By exploiting the benefits of directional antennas the an-tenna tilt can be tuned to increase performance resulting in increased bit rates, increased system throughput and increased resource efficiency. The improvements are attained without trade-offs and the different components of the study concur unanimously.

Nyckelord

(6)

(7)

Abstract

The WCDMA air interface, used in the third generation mobile communication systems, is currently being evolved to improve the uplink, i.e. the radio links carrying traﬃc from the mobile user to the ﬁxed network. An enhanced uplink concept is being developed to meet the expected needs from future applications like multimedia and video-streaming. This thesis studies interference management when high bit rates are introduced in the enhanced uplink. The study is performed through theoretical assessments and simulations using WCDMA system simula-tors.

An optimization scheme using a basic system throughput based scheduling is derived to attain a theoretical assessment of bit rate limits. The throughput optimization is achieved at the expense of user-experienced fairness. Users located on cell coverage area overlap show to be most complicated to manage.

The need for interference management is primary when the network deploy-ment consists of small cells while coverage requiredeploy-ments are most essential when the cell size increases. By exploiting the benefits of directional antennas the an-tenna tilt can be tuned to increase performance resulting in increased bit rates, increased system throughput and increased resource efficiency. The improvements are attained without trade-offs and the different components of the study concur unanimously.

(8)

(9)

Acknowledgements

First of all I would like to thank the LinLab research group at Ericsson for giving me the opportunity to conduct my master thesis there. I especially want to thank my examiner Fredrik Gunnarsson for always taking the time to listen and help me with questions and guidance. I would also like to thank my supervisors Gun-nar Bark and Erik Geĳer Lundin for all support and Eva Englund and Ke Wang Helmersson for helping and encouraging me.

I would ﬁnally like to thank family and friends for all the support and reassurance given during this spring, especially Anna for picking up the pieces after a hard day’s work.

Linköping, June 2005

Samuel Axelsson

(10)

(11)

Abbreviations

3G 3rd Generation mobile communication system 3GPP 3rd Generation Partnership Project

ACK Acknowledgement

ARQ Autonomic Repeat Request

BLER Block Error Rate

BPSK Binary Phase Shift Keying

CDF Cumulative Distribution Function, see Appendix C CDMA Code Division Multiple Access

CIR Carrier to Interference Ratio CTIR Carrier to Total Interference Ratio

dB Decibel, see Appendix A

dBd Decibel with a dipole antenna reference, see Appendix A dBi Decibel with an isotropic antenna reference, see Appendix A dBm Decibel with respect to 1 mW, see Appendix A

dBW Decibel with respect to 1 W, see Appendix A

DCH Dedicated Channel

DS-CDMA Direct Sequence CDMA E-DCH Enhanced Dedicated Channel

EUL Enhanced Uplink

FTP File Transfer Protocol

GPRS General Packet Radio Service

GSM Global System for Mobile communications

HARQ Hybrid Autonomic Repeat Request

HSDPA High-Speed Downlink Packet Access

IS-95 Interim Standard 95

kbps Kilobit per second

Mb Megabit

Mbps Megabit per second

MB Megabyte

Mcps Megachips per second

MHz Megahertz

(12)

x

NACK Negative Acknowledgement

QoS Quality of Service

RAN Radio Access Network

RBS Radio Base Station

RNC Radio Network Controller

SHO Soft Handover

TCP Transport Control Protocol TDMA Time Division Multiple Access TTI Transmission Time Interval

UMTS Universal Mobile Telecommunications Services WCDMA Wideband Code Division Multiple Access

(13)

Introduction

As telecommunications are becoming more and more important, so are the re-quirements on available services. Mobile communications of today offer wireless communications beyond accustomed speech services. This chapter presents an in-troduction to the interference management problem which is the focus of the thesis. The research approach and other work related to the problem are also presented in this chapter.

1.1 Background

The evolvement of mobile telecommunication systems has come a long way since the ﬁrst generation mobile communication system (1G). Even though 1G showed poor stability, coverage and sound quality the interest in mobile communications was evident. The capacity of 1G was limited by the analogue technology em-ployed by the systems and when 2G was introduced using digital technology the improvements were apparent. Coverage, stability and security capacities were all increased, at the same time more users could be served and data services became available. Examples of 2G systems are the Global System for Mobile communi-cations (GSM) and Interim Standard 95 (IS-95) adopted in Europe and in the United States respectively. 2G systems of today are pushed to their limits using techniques like General Packet Radio Service (GPRS), oﬀering higher data rates and thus supporting transmission of low resolution photos and limited multimedia applications.

In order to satisfy the expected needs from future applications like multimedia and video-streaming the third generation mobile communication system (3G) will replace its predecessors. The 3G system used in Europe is called Universal Mobile Telecommunications Services (UMTS) and a similar system called CDMA2000 is used in the United States. The air interface used in UMTS is Wideband Code-Division Multiple Access (WCDMA) [1]. The ﬁrst fully commercialized WCDMA service was operational in 2001 and since then an ongoing evolvement has taken place to increase resource utilization. WCDMA Release 5 [3] introduced the high-speed downlink packet access (HSDPA) to improve downlink capacity, i.e. the

(18)

2 Introduction

radio links carrying the network-to-mobile traffic. The following release, WCDMA Release 6, offers the natural complement the enhanced uplink (EUL), i.e. the radio links carrying the mobile-to-network traffic.

The EUL standard is being developed within the telecommunications standard-ization body 3rd Generation Partnership Project (3GPP). The aim is to reduce delays, to improve capacity and feasible data rates up to 4 Mbps in the uplink. Data rates up to 4 Mbps bring diﬀerent aspects to the cell1_{planning compared to}

lower data rates allowed in previous releases. Users in a WCDMA network share available resources and granting a user a high bit rate also means allowing it to utilize most of a cells resources. The problem becomes even more complicated due to tight interference coupling between the cells in the network. Utilizing resources in one cell can also result in consumption of radio resources in neighboring cells if user location is unfortunate, e.g. on the boundary between two cells.

1.2 Problem Statement

The master thesis assignment is to study the consequences of introducing high uplink bit rates in WCDMA networks. Under what circumstances does the net-work deployment introduce significant bit rate limitation and how can interference issues be managed to attain an efficient system? What bit rates can actually be achieved in specific network scenarios without compromising network stability and coverage?

The aim is to show that cell planning options like cell radius and antenna tilt can be exploited to increase the system eﬃciency and achieve the EUL concept goals.

1.3 Research Approach

In order to answer the stated problem the following research approach is taken. • Theoretical assessments of bit rate limitations of a system throughput based

scheduling for a simple network with the aim to bring intuition to scenarios where certain network deployments can be unfavorable.

• Path gain map generation of speciﬁc network deployments with diﬀerent cell radius and antenna tilt. More realistic scenarios based on the derived net-work deployments from the theoretical assessments can thereby be studied. • The derived deployments are studied in dynamic simulations with a more advanced scheduling than the basic throughput based scheduling. Dynamic simulations with more realistic and advanced scheduling captures the dy-namics of a WCDMA network.

The three diﬀerent phases capture diﬀerent aspects of the problem as they grow more complex. The theoretical assessments describes the underlying principle and additional layers of realism are added when conducting the simulations.

1

(19)

1.4 Related Work 3

1.4 Related Work

Since the EUL is a fairly new concept there are not many studies yet published. Most studies on WCDMA interference and capacity in the literature assume traﬃc generated by low bit rate mobile users, uniformly distributed within a cell. This is not the case for high bit rate users, instead a scenario, where only a few potential high bit rate users exist in each cell, is more relevant.

1.4.1 WCDMA Capacity

Dehghan et al. [7] and Owen et al. [16] study capacity and planning issues in WCDMA. The article in [7] discusses the importance of more complex simulations compared to simplified theoretical approaches when optimizing a WCDMA net-work effectively. Theoretical approaches can be used to dimension and plan the network but simulations are needed to optimize performance. This motivates the research approach described in Section 1.3 containing both theoretical aspects de-scribing the general principles and simulations verifying the derived theories. The article in [16] focuses on capacity limited by intercell and intracell interference, i.e. interference caused by users in the own cell and interference caused by users outside the own cell respectively. A WCDMA network has a tight interference coupling between cells and the F-factor is derived to model this. The F-factor is the ratio between own cell interference and total cell interference. This ratio is crucial when high bit rates are adopted since a single user on the boundary between two cells can cause significant interference in neighboring cells.

Kim et al. [11] and Lei et al. [13] study the interference-based capacity in a CDMA cellular system. The second article focuses on the dependency between location and other-cell interference. It is concluded that prioritizing users close to the center of the cell over users close to the boarder of the cell will result in higher system capacity since users on the boundary generate more noise in neighboring cells. The expected conclusion results in a trade-oﬀ between capacity and fairness. The article does not consider high bit rates speciﬁcally but this will be even more explicit when applied to high bit rates. In order to achieve a theoretical assessment of the bit rate limit the article proposes a throughput based scheduling.

1.4.2 Resource Efficiency

Oh et al. [14] and Zhang et al. [19] study the optimal resource allocation for uplink data services. Only a single cell is studied and the intercell interference aspect is therefore missing. The scheduling policy for users on the boundary between cells will be crucial when using high bit rates. In the single cell case the articles con-clude that in order to achieve maximum system throughput, the optimal resource allocation distributes maximum transmission power to the user with the most fa-vorable path gain and then the user with second best path gain and so on until the resource limit is reached, i.e. maximum noise rise is reached or the case when all users in the cell transmit with maximum power. This single cell optimal allocation is not applicable in a multicell system due to intercell interference.

(20)

4 Introduction

1.4.3 Load Estimation

Geĳer et al. [8, 9] study uplink load estimation in WCDMA, which holds great challenges and is equally important to be able to utilize system resources eﬃciently. In this thesis theoretical limits are in focus rather than exact estimation which is why the true load will be assumed known both in the theoretical assessments and in the simulations even though not being available in reality.

The notation adopted in this thesis is similar to the one used in [8, 9]. The notation might seem cumbersome at ﬁrst but it holds several advantages.

1.5 Thesis Outline

The remaining part of the thesis is arranged as follows. Chapter 2 gives a brief introduction to WCDMA and especially to the enhanced uplink concept. Funda-mental techniques used in a CDMA system are described. The following chapter, Chapter 3, contains theoretical assessments with the aim to bring intuition to sce-narios where the network deployment can be unfavorable with respect to high bit rate users. Chapter 4 describes both the simulation models used for the path gain map generation and the dynamic simulations since they are similar in many ways. Simulation results from the static and the dynamic simulations are presented in Chapter 5. Finally conclusions and future work are presented in Chapter 6.

(21)

Chapter 2

Third Generation Mobile

Communication System

UMTS is the leading 3G technology today and the UMTS network use Wideband Code Division Multiple Access (WCDMA) as its air interface. WCDMA differs fundamentally from the Time Division Multiple Access (TDMA) air interface in 2G networks like GSM where a user is granted transmission during a specific time interval and using a certain frequency spectrum. The WCDMA radio interface instead allows the users to utilize the entire available frequency bandwidth at the same time. The WCDMA system utilizes the available bandwidth more efficiently compared to 2G systems like GSM and thereby increasing capacity, making higher bit rates available and allowing a higher number of users to be served. This chapter presents an overview of the WCDMA air interface followed by an introduction to the enhanced uplink concept. A more detailed description of WCDMA can be found in [4, 10, 12, 15, 17].

2.1 Wideband Code Division Multiple Access

The users in a WCDMA system are neither separated in time nor in frequency, instead they are separated by applying a scheme called Direct Sequence Code Division Multiple Access (DS-CDMA). On the transmitting side the user data is multiplied with a spreading code that has n times higher chip rate1_{than the rate}

of the user data, resulting in a spread signal in the frequency spectrum, hence the name spreading code. The ratio between the spreading code chip rate and the data rate is called the spreading factor, earlier denoted as n. A chip rate of 3.84 Mcps results in a carrier bandwidth of approximately 5 MHz which is wider than the bandwidth of about 1 MHz used in CDMA, hence the additional wideband in WCDMA.

1

Chip rate is the bit rate of the spreading code. 5

(22)

6 Third Generation Mobile Communication System

Example 2.1: Maximum data rate per spreading code

A spreading factor of n = 8, will result in a maximum data rate of 480 kbps assuming binary phase shift keying (BPSK) modulation. In a BPSK system the phase of the carrier signal is switched between two phase values corresponding to the binary symbols 1 and 0 respectively.

3.84 Mcps

8 = 480 kbps

Multiple simultaneous codes must be used to attain bit rates greater than this, assuming the spreading factor is not lowered instead. The lower the spreading factor, the fewer available spreading codes since the spreading codes are orthogonal with respect to each other.

On the receiver side the spread signal is again multiplied with the same unique spreading code and the original data sequence can be retrieved, this procedure is therefore called despreading. The spreading and despreading procedure is illus-trated in Figure 2.1. +1 +1 +1 +1 +1 -1 -1 -1 -1 -1 Spreading Despreading Data Spreading Code Data ×Code Code Data Chip Symbol

Figure 2.1. Spreading and despreading procedure. The spreading factor, n, is set to 8. Courtesy of Holma et al. [10].

If the spread signal is despread using the same spreading code as the original, the data sequence is retrieved. If any other spreading code orthogonal to the original is used to despread the result will be perceived as noise or interference. The procedure is illustrated in Figure 2.2.

Since all other users will be perceived as interference, see Figure 2.3, care must be taken so that the interference caused by other users does not exceed the own signal energy. If that is the case the own signal will drown in interference and disable the data retraction in the despreading procedure.

A fast power control, permitting users to transmit with a certain transmission power, is employed to control the interference caused by each user. The power

(23)

2.1 Wideband Code Division Multiple Access 7 +1 +1 +1 +1 +1 -1 -1 -1 -1 -1 Own Data Code Own Signal Code Despreaded Data Integrated Data Other Signal +1 -1 Integrated Data +1 -1 +1 -1

Figure 2.2. Despreading with the original spreading code retrieves the data while despreading using another spreading code than the original, or the original spreading code used on another signal, results in noise. Courtesy of Holma et al. [10].

Spreading Despreading Power Frequency Frequency Frequency Frequency Power Power Power

Figure 2.3. Multiple users are able to transmit using the same frequency band. The own data can be retrieved as long as interference produced by other users do not exceed the own signal energy. The shaded bars in the figure represent different user signals. Courtesy of Holma et al. [10].

(24)

control is essential to handle the near-far problem and also compensates for dif-ferences and variations in the radio channel. The near-far problem appears when two users are transmitting with equal power but one user experiences considerably lower path gain, i.e. higher signal attenuation. This user will then drown in in-terference caused by the user with high path gain. The power control can adjust transmission power so this problem is reduced.

2.2 Network Architecture

The UMTS network architecture consists of two major elements, the core network and the WCDMA Radio Access Network (WRAN). The core network routes traﬃc to external networks like the Internet while the WCDMA Radio Access Network handles the radio traﬃc. The WRAN consists of the mobile users which connect to the Radio Base Stations (RBS) which are controlled by the Radio Network Controllers (RNC), see Figure 2.4. The RBS handles all radio signalling to the mobile users while the RNC with its more central role handles communication with the core network. Resource allocation like user-service requests is shared by the RNC and the RBS.

RNC

RBS

User RBS

Core Network

Figure 2.4. The UMTS network architecture. The base stations use three sector an-tennas resulting in three cells per base station.

(25)

2.3 Enhanced Uplink 9

2.2.1 Soft and Softer Handover

WCDMA supports soft and softer handover. A user in soft handover is connected to multiple cells belonging to diﬀerent base stations while a user in softer handover is connected to multiple cells belonging to the same base station, see Figure 2.5. This feature is crucial since a user in handover experiences roughly the same path gain to multiple cells and is thereby able to cause considerable interference in multiple cells. This can occur when a user is located in an overlapping cell coverage area from multiple cells. The soft and softer handover features enable all cells to which the user is connected, i.e. the active set, to power control the user and thus reducing interference issues. A user moving away from one cell into another cell would increase transmission power to reach the original cell and inﬂict unnecessary interference if soft and softer handover were not used. The soft and softer handover features also imply that uplink data transmissions can be received at multiple cells2_.

RBS

User in soft handover

RBS

User in softer handover

RNC

Figure 2.5. Soft and softer handover.

2.3 Enhanced Uplink

As mentioned in Chapter 1 the evolvement of the WCDMA standard is an ongo-ing process illustrated in Figure 2.6. WCDMA Release 5 introduced the enhanced downlink, HSDPA, and the main goal with the following release, WCDMA Re-lease 6, is to complement the improved downlink with the enhanced uplink. A main requirement of the enhanced uplink is that it must be able to coexist with existing WCDMA releases and should be possible to introduce in the already de-ployed WCDMA networks. A new transport channel is therefore introduced, the

2

(26)

Enhanced

Uplink _enhancementsAdditional Enhanced Downlink

(HSDPA)

Rel 4 Rel 5 Rel 6

WCDMA WCDMAEvolvedEvolved WCDMA

WCDMA R99

Figure 2.6. The evolution of WCDMA.

Enhanced Dedicated Channel (E-DCH), alongside the existing Dedicated Channel

(DCH). E-DCH is intended for best eﬀort services rather than real time services like speech. The main EUL goals are similar to the goals set for the HSDPA in WCDMA Release 5:

• Reduced delays • Increased data rates • Increased capacity

• Increased high data rate availability

The main objective is not the increase of peak data rates but rather the avail-ability of high data rates, i.e. increased high data rate coverage. Reduced delays, increased data rates and increased capacity will add to the experienced quality of service (QoS) and support more demanding applications. The general feature in the EUL concept is to move functionality closer to the user, i.e. move deci-sion making functions from the RNC to the RBS, resulting in less signaling and reduced delays. Less information is however available because of this decentral-ization, making the design of eﬃcient system functions more complicated. Other more speciﬁc features the EUL concept is based on are:

• Short TTI

• Hybrid ARQ with soft combining • Fast scheduling

The general principles behind these features are described below.

2.3.1 Short Transmission Time Interval

The available radio resources are allocated between the users accessing the sys-tem at certain intervals, the length of these intervals are called transmission time

interval (TTI). Each user is signaled a maximum bit rate during each TTI. The

maximum rate results in a transmission power controlled by the fast power control described in Section 2.1. A short TTI reduces round trip time and enables better channel adaptation to improve the transmission performance. Channel adaptation is needed in response to changing channel conditions due to user mobility.

(27)

2.3 Enhanced Uplink 11

2.3.2 Hybrid ARQ with Soft Combining

Automatic Repeat Request, ARQ, is an error control method that repeats the transmission of an erroneous data block. If a data block is received correctly the cell replies by transmitting an acknowledgement (ACK) to the mobile user and the next block is transmitted. If the data block is erroneous the cell responds with a negative acknowledgement (NACK) and the data block is retransmitted. A user in soft handover only requires an ACK from one of its multiple connected cells.

Instead of discarding the erroneous block completely the Hybrid ARQ (HARQ) scheme combines the previous transmission attempts with the current retransmis-sion, this procedure is called soft combining. Even though the previous trans-mission attempt could not be decoded correctly there is still usable information present in the received signal. Short TTI lengths enable more retransmission at-tempts for a given time period and thus larger gain when using HARQ.

Example 2.2: The general principle of the HARQ feature

Assume a packet being sent with a certain bit rate resulting in a certain probability of being received correctly. If the bit rate is increased with a factor x, this allows the packet to be retransmitted x times during the same time as the original bit rate required. The probability of the packet being received correctly is lower for each of the x transmission attempts but since the channel quality varies between the attempts, a HARQ scheme with soft combining can terminate the transmission in advance if transmission quality is good. The variations in channel quality can therefore be utilized better resulting in less than x needed transmission attempts on average and thereby a more eﬃcient system.

2.3.3 Fast Scheduling

Scheduling is the mechanism that grants users transmission permission. For each TTI, the scheduler assigns a maximum bit rate that the user may use. The user’s ﬁnal choice of rate depends on available transmission power and, if in handover, rate limitations imposed from the other cells in the active set. A well designed scheduling algorithm ensures eﬃcient use of the interference resource. The schedul-ing function is moved from the RNC to the RBS in the EUL concept. Movschedul-ing functionality closer to the user results in reduced delays and faster radio channel adaptation. The algorithm will however be more complicated to design since less information is available in the RBS.

(28)

(29)

Chapter 3

Theoretical Assessments

Theoretical approaches might not capture the complex characteristics or the dy-namics of a WCDMA network but they bring intuition to potential problems and increase the ability to interpret simulation results. This chapter covers theoretical assessments with the aim to create understanding for potential interference issues in a WCDMA network.

3.1 Definitions

The quality of service (QoS) experienced by users in a network is the main per-formance measurement. No matter how central the QoS is, it is at the same time very ambiguous. Dropped calls, delays or poor sound quality are difficult to weight against each other since it depends on the user experiencing it. In the theoreti-cal assessments performance is measured in system throughput and no concern is taken to neither fairness nor delay. This might seem crude but it is justified by the fact that focus is on high bit rate users. The number of users will be moderate in order to attain high bit rates. Few users in the system also implies large scattering of the users, resulting in less competition and justification of the throughput based performance measurement.

Notation and a number of deﬁnitions used in the thesis will be declared in this section. In the theoretical assessments a system with B base stations and M mobile users is assumed. Each base station has one cell and base station and cell will therefore refer to the same object. All B cells are serving the same central node, i.e. the same radio network controller (RNC), of which there exists only one in the theoretical assessments. Since only one RNC is assumed the set of cells controlled by this RNC and the complete set of cells are the same.

Perfect power control is assumed in the theoretical assessments, i.e. the sig-nalled CIR is attained instantly, and the static situation at a certain instance in time is studied. The time index, t, is therefore not explicitly written in the following sections of this chapter.

A signal propagating through any medium will be attenuated, this is further explained in Section 4.1. In the theoretical assessments the signal attenuation is

(30)

14 Theoretical Assessments

captured in the path gain matrix deﬁned as (3.1). It is the ratio between the power of the received signal and the the power of the transmitted signal and thus always less than one.

Deﬁnition 3.1 Path gain matrix

G,    g1,1 . . . g1,B .. . . .. ... gM,1 . . . gM,B    M ×B <1 (3.1)

where gi,j is the uplink path gain experienced between user i and cell j.

There is always background interference due to thermal noise. The total in-terference power in a cell j, Itot

j , is the sum of the received signal power added to

the background noise power. Interference power and background noise power will also be referred to as interference and background noise respectively.

Deﬁnition 3.2 Total interference

Ijtot, M X i=1 Ci,j+ Nj = M X i=1 pigi,j+ Nj (3.2)

where pi is user i’s transmission power and Ci,j= pigi,j is the received power in

cell j from user i. Nj is the background noise power in cell j.

A diﬀerent way of expressing the total interference is to describe it as rise over thermal noise or noise rise. It is basically the ratio between total interference, Itot_,

and background noise, N.

Deﬁnition 3.3 Noise rise

Λj , Itot j Nj = PM i=1pigi,j+ Nj Nj = 1 + PM i=1pigi,j Nj ≥ 1 (3.3)

The mobile user equipment has a limited transmission power of about 21 dBm1

due to battery capacity and antenna features. The basic WCDMA principles described in Section 2.1 showed that in order to retrieve the data from a spread signal the interference must not be too high. In order to ensure system stability and coverage the maximum allowed interference is to be limited.

The quality of a communication link is a function of the received signal power and the ambient noise power. The carrier to interference ratio (CIR) is the ratio between own signal energy and experienced interference, i.e. total interference excluding the own signal power.

Deﬁnition 3.4 Carrier to Interference Ratio (CIR) from user i to cell j

γi,j,

Ci,j

Itot

j − (1 − αi,j)Ci,j

(3.4)

The orthogonality factor, αi,j, is described in Section 3.1.1.

1

(31)

3.1 Deﬁnitions 15

3.1.1 Multipath Propagation

Due to propagation mechanisms like reflection, diffraction and scattering, a trans-mitted signal can reach the receiver through multiple simultaneous propagation paths. A transmitted signal can therefore interfere with itself because of this mul-tipath propagation. The different components reach the receiver with different phases and different amplitudes. The orthogonality factor describes the receiver’s ability to handle the own signal interference and varies between users and in time depending on user mobility. Perfect own signal interference handling, i.e. no own signal interference, corresponds to an α = 0. The own signal interference will limit the attainable bit rates considerably as shown in Example 3.1.

Example 3.1: Limitation due to own signal interference

The following holds for the CIR deﬁned in (3.4), γi,j = Ci,j Itot j − (1 − αi,j)Ci,j < 1 αi,j .

This results in a signiﬁcant constraint on available bit rates as shown in Figure 3.1. In this example maximum noise rise is set to 7 dB and the potential bit rate of a single user in a cell is plotted versus the orthogonality factor. Intercell interference, i.e. interference caused by users in neighboring cells, is not taken into account resulting in a theoretical maximum. Equation (3.6) is used to map CIR values to bit rates. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 1 2 3 4 5 6 Orthogonal factor, α Bit rate [Mbps] 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−5 0 5 10 CIR, γ [dB] Bit rate CIR, γ

Figure 3.1. Theoretical bit rate limitation due to own signal interference. Maximum noise rise is set to 7 dB.

(32)

3.1.2 Shannon’s Theorem

When Shannon published his Communication Theory in the late 40s the perspec-tive on wireless communication changed. Focus shifted from the basic physical principles like electromagnetic propagation to a systems approach. Among other things he formulated Shannon’s theorem [18]. Shannon’s theorem deﬁnes a theo-retical bit rate limit based on available bandwidth and CIR.

Theorem 3.1 Shannon’s Theorem, The maximum attainable bit rate

R= W log2(1 + γ) [Mbps], (3.5)

where R is the maximum bit rate in Mbps, W is the bandwidth in MHz and γ the experienced CIR.

This is a theoretical restriction and this bandwidth eﬃciency can never be achieved in a real system. The theorem will however be used as a model to map CIR-values to corresponding bit rates using (3.6).

R= C ln (1 + γ) [Mbps] (3.6)

where C is a constant set to normalize the maximum available bit rate to 4 Mbps since that is the goal set for the enhanced uplink. Note that log2(x) =

ln(x) ln(2) and

that the actual value of C is set to is not crucial in this analysis but rather the model in general. −100 −8 −6 −4 −2 0 2 4 6 0.5 1 1.5 2 2.5 3 3.5 4 4.5 CIR [dB] Bit rate [Mbps]

Figure 3.2. Model of end-user bit rate as a function of experienced CIR.

Similar to the carrier to interference ratio is the carrier to total interference ratio (CTIR). The diﬀerence being that the own signal signal power is related to the total interference in the cell instead of the interference experienced with respect to the own signal.

(33)

3.2 Soft and Softer Handover 17

Deﬁnition 3.5 Carrier to Total Interference (CTIR) from user i to cell j

βi,j, Ci,j Itot j = pigi,j PM l=1plgl,j+ Nj <1 (3.7)

Resulting in a nonlinear relationship between the CIR and the CTIR according to,

γi,j=

βi,j

1 − (1 − αi,j)βi,j

. (3.8)

Example 3.2: Pole Capacity

Equation (3.2) can in the single cell case be expressed using (3.3) and (3.7), Itot= M X i=1 pigi+ N 1 = M X i=1 βi+ 1 Λ Λ = 1 1 −PM i=1βi , where PM

i=1βi < 1. When PM_i=1βi becomes close to 1 the corresponding noise

rise approaches inﬁnity and the system reaches its pole capacity. Even if inﬁnite transmission power, pi, would be available the system capacity cannot exceed this

theoretical limit. The uplink is therefore regarded as interference power limited.

3.2 Soft and Softer Handover

Best performance is obtained when all received signals from a user in softer han-dover are maximum ratio combined, this is a scheme resulting in the following CIR,

γi=

X

j∈B

γi,j, (3.9)

where B is the set of cells in the active set. This is possible when a user is connected to multiple cells in the same base station. Since each base station is assumed to support only one cell, the softer handover feature is not adopted in the theoretical assessments.

In the case of soft handover the best performance is obtained by using selection

combining. This is a scheme where the received signal with highest CIR is the CIR

experienced by the user,

γi= max

(34)

In the theoretical assessment the users are assumed to be connected to all cells in the system while there in reality is a limitation to the maximum number of connected cells. This yields,

βi= max j∈Bβi,j= maxj∈B pigi,j Itot j . (3.11)

Ultimately maximization of the system throughput is the main objective of the theoretical assessments, which will result in maximum allowed total interference, Ijtot,max, in each cell2. This can be used to approximate the soft handover feature

by choosing the cell j that power controls user i, ji, as follows,

ji= arg max j∈B

pigi,j

Ijtot,max

, (3.12)

where arg maxxf(x) is the argument x that maximizes f (x). When the maximum

noise rise in each cell is reached, i.e. when Itot j = I

tot,max

j ∀j, the following holds,

βi= max j∈B pigi,j Itot j = max j∈B pigi,j Ijtot,max = pigi,ji Ijtot,maxi

The soft handover feature is therefore adopted by setting ji as in (3.12).

3.3 System Throughput Optimization

A throughput based scheduling is assumed in the theoretical assessments. The scheduling is done by solving an optimization problem and this section is devoted to deﬁning this optimization problem. The deﬁnitions stated in Section 3.1 are used to derive a theoretical assessment of the bit rate limit with the system throughput maximization as main objective. By setting setting ji as in (3.12) and solving for

pi in (3.7) results in,

pi=

βiIjtoti gi,ji

, (3.13)

where αi, βi and γi refers to αi,ji, βi,ji and γi,ji respectively in the following sections of this chapter. Inserting this expression in (3.3) gives,

Λj = 1 + PM i=1pigi,j Nj = 1 + M X i=1 βi gi,j gi,ji Itot ji Nj . (3.14)

Equation (3.14) expresses the noise rise in cell j due to background noise and transmitting users in the system3_{. We see that an interesting ratio is that between}

gi,j and gi,ji. This motivates the following deﬁnition. 2

Assuming that there are enough users in each cell to utilize the entire available noise rise resource.

3

(35)

3.3 System Throughput Optimization 19

Deﬁnition 3.6 The relative path gain between user i and cell j is,

zi,j,

gi,j

gi,ji

. (3.15)

In order to achieve a compact notation of the relative path gain the normalized

path gain matrix is deﬁned as (3.16). The relative path gain for each user i with

respect to each cell j is found in the i:th row and j:th column of the matrix.

Deﬁnition 3.7 Normalized path gain matrix

Z,    z1,1 . . . z1,B .. . . .. ... zM,1 . . . zM,B    M ×B =     g1,1 g1,j1 . . . g1,B g1,j1 .. . . .. ... gM,1 g_M,jM . . . gM,B g_M,jM     M ×B ≤ 1 (3.16)

3.3.1 Equal Background Noise

The background noise in a cell depends on the thermal temperature. The thermal properties of the cells can be assumed equal and therefore also the background noise. If equal background noise in all cells is assumed, i.e. Nj = N ∀j, (3.14)

can be expressed as,

Λj = 1 + M X i=1 βi gi,j gi,ji Λji. (3.17)

This expression can be used to deﬁne a system of equations describing the noise rise in all cells in the system.

   Λ1 .. . ΛB    B×1 =    1 .. . 1    B×1 + ZT B×Mdiag Λj1 · · · ΛjM M ×M    β1 .. . βM    M ×1 (3.18) where diag Λj1 · · · ΛjM

M ×M is a diagonal matrix with the noise rise of the

cell to which user i is connected in each diagonal element. The product between the variables Λj and βi makes this a nonlinear system of equations. Assuming

that the maximum noise rise will be achieved yields the following expression.    Λmax 1 .. . Λmax B   ≥    1 .. . 1    B×1 + ZTdiag Λmax j1 · · · Λ max jM    β1 .. . βM   , (3.19)

where the equality in (3.18) has been replaced with an inequality for optimization simplicity. Equation (3.19) has the users’ CTIR as only free variable. Due to the approximation in the derivation of (3.19) the system of equations is therefore linearized.

(36)

3.3.2 Optimization Problem

By maximizing the sum of the users’ bit rates, PM

i=1Ri, the throughput of the

system is maximized. Finding the maximum throughput can therefore be written as a nonlinear optimization problem, where the constraints are linear but the objective function is nonlinear.

max M X i=1 Ri= M X i=1 Cln 1 + βi 1 − (1 − αi) βi (3.20a) subject to    Λmax 1 − 1 .. . Λmax B − 1   ≥ Z T_{diag Λ}max j1 · · · Λ max jM    β1 .. . βM    (3.20b) 0 ≤ βi≤ min ( pmax i gi,ji I_jtot,max_i ,1 ) ∀i (3.20c)

where the βis are seen as variables. Yet another constraint is incorporated on βi

since a maximal transmission power for each user i, pmax

i , is introduced.

3.3.3 Linearity and Convexity

Due to the nature of the nonlinearity of the objective function in (3.20) the op-timization problem is neither linear nor convex. A convex opop-timization problem implies that a local maximum is also a global maximum. Nonlinear and nonconvex optimization problems are generally very hard to solve. Further approximations will be done in the following sections to simplify the optimization problem.

3.3.4 Equal Maximum Noise Rise

By assuming equal conditions in all cells justiﬁes choosing equal maximum noise rise in all cells, i.e. Λmax

j = Λ

max _{∀j. Equation (3.20b) can then be simpliﬁed}

to,    Λmax_{− 1} .. . Λmax_{− 1}    B×1 ≥ ZT_Λmax IM ×M    β1 .. . βM   ,

where I is the identity matrix and Λmax_{is the scalar representing the maximum}

noise rise. This can be rewritten as, Λmax_{− 1} Λmax    1 .. . 1    B×1 ≥ ZT    β1 .. . βM   

where the scalar multiplying the vector on the left hand side can be identiﬁed as the relative load, L, deﬁned in (3.21).

(37)

Deﬁnition 3.8 Relative load

Lj, 1 −

1 Λj

(3.21) Equal maximum noise rise also implies the cell having maximum relative load,

Λmax₌ 1

1 − Lmax =

Imax

N . (3.22)

3.3.5 Nonlinear Optimization Problem

With the assumption from Section 3.3.4 and the deﬁnition of relative load, the optimization constraint in (3.20b) can be written as,

Lmax1B≥ ZTβ

where 1B = 1 . . . 1 T

B×1and β = β1 . . . βM

T_{. This simpliﬁes (3.20) to,}

max M X i=1 Cln 1 + βi 1 − (1 − αi) βi (3.23a) subject to Lmax 1B≥ ZTβ (3.23b) 0 ≤ βi≤ min pmax i gi,ji Itot,max ,1 ∀i (3.23c)

in the case of equal maximum noise rise and equal background noise in all cells.

3.3.6 Quadratic Optimization Problem

Due to the complex nature of the optimization problem the objective function is approximated with a quadratic function. There is extensive theory in [6], covering quadratic optimization problems or quadratic programming problems but that is beyond the scope of this thesis. Results from that theory is however used to approximate the optimization problem. The derivation of the approximated objective function is found in Appendix B and the resulting quadratic optimization problem is, max M X i=1 C βi+ 1 − 2αi 2 β 2 i (3.24a) subject to Lmax 1B≥ ZTβ (3.24b) 0 ≤ βi≤ min pmax i gi,ji Itot,max ,1 ∀i. (3.24c)

Equation (3.24) is used in the following examples to study basic scenarios with enlightening interference issues.

(38)

Example 3.3: Two Cell System with Four Users

Let a system have four users i, i ∈ {1, 2, 3, 4}, who are all able to connect to one of two cells j, j ∈ {1, 2}. Assume that users in cell 1 can not be heard in cell 2 but users in cell 2 are heard in both cells. The system then has the following path gain matrix, G=     −119 −∞ −122 −∞ −130 −119 −135 −119     [dB].

User 1 and 2 will connect to cell 1 and user 3 and 4 will connect to cell 2. The resulting Z-matrix is,

Z=     1 0 1 0 0.08 1 0.03 1     . Equation (3.14) is thus Λ1= 1 + β1Λ1+ β2Λ1+ β3 g3,1 g3,2 Λ2+ β4 g4,1 g4,2 Λ2 Λ2= 1 + β3Λ2+ β4Λ2 ,

in the separate cells. Note that equal background noise is assumed in all cells and that β1and β2are not present in the second equation since the corresponding users

are not heard in that cell. Actual noise rise is replaced with a maximum noise rise, Λmax _{= 7 [dB], in both cells. Together with the above system of equations this}

yields the inequality below. Λmax_{− 1} Λmax 1 1 =L max Lmax ≥ 1 1 g3,1 g3,2 g4,1 g4,2 g1,2 g1,1 g2,2 g2,1 1 1 !     β1 β2 β3 β4     = ZTβ

Assume a maximal transmission power, pmax _{= 21 [dBm], for all users. Solving}

the optimization problem in (3.24) with the above parameters yields,

β=     0.661 0.112 0.140 0.661     .

This solution results in maximum noise rise in both cells and the transmission power for each user is,

p=     21.0 16.3 14.3 21.0     [dBm].

(39)

Note that this is a special case where the optimization is done with relative ease due to the nature of the chosen path gain matrix and maximal transmission power. The users’ corresponding CIR values are

γ=     1.947 0.126 0.163 1.947     .

The CIR for each user is mapped to a bit rate using (3.6). This results in the following bit rate for each user.

R=     2.70 0.30 0.38 2.70     [Mbit/s]

The resulting average bit rate per user is thus 1.52 Mbps. This result takes no fairness into consideration which is clearly seen in the bit rate distribution. The user with lowest relative path gain in each cell is allowed to transmit with maximum transmission power while the rest of the users can utilize what is left of the noise rise resource.

Example 3.4: Minimum bit rate

Assume the same scenario as the previous example but also introduce a minimum bit rate, Rmin = 1.0 [Mbps]. This yields the following result.

p=     19.16 20.98 17.98 19.52     [dBm] γ =     0.76 0.49 0.49 0.89     R=     1.42 1.00 1.00 1.59     [Mbit/s]

The resulting average bit rate per user is thus 1.25 Mbps. This is obviously more fair but also a trade-oﬀ since system throughput decreases.

Example 3.5: Maximum bit rate

Yet again assume the same scenario as the previous example but instead of a minimum bit rate introduce a maximum bit rate, Rmax= 2.0 [Mbps]. This yields

the following result. p=     20.21 19.15 16.78 20.21     [dBm] γ =     1.23 0.28 0.33 1.23     R=     2.00 0.61 0.72 2.00     [Mbit/s]

(40)

The resulting average bit rate per user is thus 1.33 Mbps. Again this is more fair than Example 3.3 and system throughput is again decreased due to the trade-oﬀ between fairness and throughput capacity.

Example 3.3 through Example 3.5 illustrate that the optimal result throughput-wise is achieved by an unfair bit rate distribution and a more fair system is attained at the expense of system throughput.

Example 3.6: Two cell with one user in each cell

Let a system have two cells j, j ∈ {1, 2}, with one user i, i ∈ {1, 2}, in each cell. Assume that user 1 in cell 1 can not be heard in cell 2 but user 2 in cell 2 is heard in both cells. Both users have an equal and relatively good path gain, i.e. around 110 dB, to their serving cells. This implies that they are able to utilize all of the available noise rise resource in their own cell. The eﬀect user 2 has on user 1, governed by the relative path gain, is studied. Assume the following path gain matrix,

G= g0 −∞

g2,1 g0

[dB], which results in the following system of equations,

Lmax≥ β1+ g2,1 g0 β2 Lmax≥ β2 . Since g2,1

g0 <1 the best resource distribution throughput-wise is to let user 2 use as much as possible of the available resources and let user 1 use what is left. Figure 3.3 shows the result when the g2,1parameter is varied. Note that the x-axis is plotted

in logarithmic scale and thus representing the ratio between g2,1 and g0in [dB].

Example 3.6 illustrates that a user, located on the overlapping cell coverage area, can utilize a signiﬁcant amount of neighboring cells’ resources. The example shows that a user with a relative path gain greater than -7 dB has the potential to more than halve the attainable CIR for a user in a neighboring cell. A relative path gain of -7 dB will therefore be used as threshold to consider a user hazardous. In a real system it would not go this far since the soft handover feature would come into force and decrease the induced interference. That would however result in decreased performance both for the user and for the system and thus still seen as an interference issue, justifying the -7 dB relative path gain threshold.

(41)

3.3 System Throughput Optimization 25 −250 −20 −15 −10 −5 0 1 2 3 4 5

Relative path gain [dB]

Bit rate [Mbps] −25 −20 −15 −10 −5 0 −6 −4 −2 0 2 4 6 8

Relative path gain [dB]

CIR [dB]

User 1 User 2

Figure 3.3. Example how a user when close to the cell boundary affects another user in a neighboring cell.

(42)

(43)

Chapter 4

Simulation Models

Basic link budgets and theoretical equations may be used to dimension a network but due to the interactive and tight interference coupling of WCDMA networks a more complex approach based on simulations is required to optimize efficiency. The models used to simulate the WCDMA network will be presented in this chapter.

4.1 Radio Channel Model

The received energy of a transmitted electromagnetic wave will always be less than the energy with which it was transmitted. This is called signal attenuation and is due to physical propagation mechanisms like reﬂection, diﬀraction, scattering and especially the multidirectional radiation of any transmitting antenna. To attain a realistic model of the signal attenuation it will consist of four elements described in this section. The total path gain can be expressed as a product, Equation (4.1), of these four elements.

g= gpgsgfga<1 (4.1)

Here, the path gain is expressed in linear scale, but it is often referred to in logarithmic scale and will then be the sum of the separate elements in dB.

4.1.1 Distance Attenuation

A widely used distance attenuation model for coverage calculation and link budgets is the Okumura-Hata model [12]. Cell radiuses in the interval of 300 meters to 1500 meters, where the Okumura-Hata model has good accuracy, is studied. The Okumura-Hata model describes the distance attenuation as,

gp= 46.3 + 33.9 · log10(f ) − 13.82 · log10(hB) − 3.2 · [log10(11.75 · hU)]2

− 4.97 + (44.9 − 6.55 · log10(hB)) · log10(d), (4.2)

where gp is the distance attenuation (dB), f is the frequency (MHz), hB and hU

are the heights (m) of the base station and user antenna respectively. Finally d 27

(44)

28 Simulation Models

(km) is the distance between the cell and the user. Note that gpstrictly decreases

with distance.

4.1.2 Shadow Fading

Shadow fading occurs when large obstacles like buildings and hills block the line of sight. The shadow fading inﬂuence on the average signal level variations is commonly modeled with a log-normal distribution.

gs= N (µ, σ) , (4.3)

where gs, µ and σ are given in dB. The parameters µ and σ are commonly set to 0

dB and 8 dB respectively. When a user is located in such a way that it is subject to shadow fading it is likely that nearby locations also will be subject to shadowing, shadow fading is therefore spatially correlated. A decorrelation distance is used to represent the distance where the correlation has decreased with a factor 1

e.

4.1.3 Multipath Fading

Multipath fading or fast fading is due to the multiple signal paths taken by the propagating signal, hence the name multipath fading. The multiple components reach the receiver with different amplitude and phase resulting in constructive and destructive interference. The multipath fading can shift vary fast since a slight change in phase between two propagating components can have significant effect on the received signal, which is why it is also called fast fading.

4.1.4 Antenna Gain

The antenna gain models the gain oﬀered by directional antennas. The antenna characteristics is described in more detail in Section 4.2.

4.2 Antenna Model

An isotropic antenna radiates with equal intensity in all directions. Such an an-tenna can never be implemented in reality and even more importantly, it is not wanted. An isotropic antenna would be very ineﬃcient since the objective for the antenna is to cover a surface area. Antennas are always directional to some extent, meaning that it radiates more energy in some directions than others, depending on how it is designed. The antenna gain is described with reference to an ideal isotropic antenna (dBi) or a dipolar antenna (dBd)1_{. The antenna model used is}

a three-sector antenna illustrated by the antenna diagram in Figure 4.1. A full description of the antenna model can be found in [2].

Each cell covers 120 degrees in the horizontal plane resulting in full coverage when using three antennas. The vertical diagram in Figure 4.1(b) shows a 6-degree electrical tilt and to this a mechanical tilt is added or subtracted resulting in a

1

(45)

4.3 Network Deployment 29 −10 dB −3 dB 0 dB 30 210 60 240 90 270 120 300 150 330 180 0

(a) Horizontal antenna diagram

−10 dB −3 dB 0 dB 30 210 60 240 90 270 120 300 150 330 180 0

(b) Vertical antenna diagram

Figure 4.1. Antenna diagram. The gain in the main direction of the antenna is 15.85 dBd and the antenna diagrams show the gain (dB) in each direction with reference to the gain in the main direction.

total downward tilt which is the antenna tilt angle referred to throughout this thesis. Note that the electrical tilt is kept ﬁx and only the mechanical tilt varied.

4.3 Network Deployment

The network deployment refers to site2 _{placement, number of cells per site, cell}

radius, antenna tilt etc. The network deployment used in the simulations consists of seven sites using three sector antennas, resulting in 21 cells. Wrap-around is used to avoid border eﬀects, i.e. the seven sites are repeated creating an inﬁnite simulation surface without borders. The network deployment is illustrated in Figure 4.2.

4.4 Static Simulation Models

As a ﬁrst step of the simulations, static path gain maps are studied. The generated path gain maps are based on statistics from realistic scenarios. Each cell consists of a number of bins where the bin size is set to a tenth of the cell radius. Note that the number of bins will be ﬁxed regardless of cell radius and thus result in the bin size being proportional to the cell radius. It is reasonable since a small cell results in a small bin size and thus good accuracy. When studying large cells, the bin size increases and the need for detailed accuracy decreases.

2

(46)

Figure 4.2. Simulated network deployment. The cell radius is defined as the radius of the circle circumscribing the cell, i.e. the distance from the center of the cell to one of its corners, resulting in a site-to-site distance three times the cell radius.

4.4.1 Shadow Fading

When generating the path gain maps the shadow fading model is slightly diﬀerent from the model described in Section 4.1.2. Instead of being a single Gaussian distribution it is a sum of two Gaussian distributions. One distribution models the shadow fading over the cell while the other models the shadow fading in each bin. The cell shadow fading model has µcell= 0 dB, σcell= 5 dB and decorrelation

distance equal to one ﬁfth of the cell radius. The bin shadow fading model has µbin= 0 dB and σbin= 3 dB resulting in a total shadow fading model with µtotal=

0 dB and σtotal = 8 dB corresponding to the model described in Section 4.1.2.

Decorrelation distance in the bin shadow fading model is meaningless since a user does not have a certain location within a bin. A parameter, 0 ≤ ρ ≤ 1, is used in the bin shadow fading model to represent fading correlation to diﬀerent cells, where zero represents uncorrelated and one completely correlated. A user in a topographical cavity for example will probably experience a low path gain to all cells. ρ is set to 0.5 when generating the path gain maps.

4.4.2 Multipath Fading

Multipath fading is not modeled when studying the static path gain maps since it holds no interest when studying a static situation. Instead a margin is added to cover the multipath fading variations.

(47)

4.5 Dynamic Simulation Models 31

4.5 Dynamic Simulation Models

As a second step of the simulations, a study of the dynamic aspects of the WCDMA network is conducted. Simulations should reflect the world as realistically as possi-ble, limited of course by available computational capacity. Behavior over time and the introduction of a traffic model will offer a more realistic aspect to the analysis. Dynamic simulations are more demanding than the path gain map generation and thus restricted to the most essential scenarios. Interference issues are most evident when mobile users are not power limited, i.e. when cell radius is moderate. The dynamic simulations are therefore restricted to scenarios where the cell radius is set to 500 meters.

4.5.1 Enhanced Uplink

Only the uplink will be simulated to reduce computational load. The new features in the enhanced uplink concept, some described in Section 2.3, are adopted in the dynamic simulations.

4.5.2 Traffic Model

The traffic model governs the users’ mobility, intensity, service requests etc. Only E-DCH traffic is modeled, i.e. all users in the system are using the new introduced E-DCH channel. Speech users using the DCH channel are not modeled since the focus of the thesis is to study the effects of different network deployments rather than improvements in the EUL concept compared to previous WCDMA releases. A file transfer protocol (FTP) model is used to model the requested data service. Each user is modeled to transfer a single data packet each. Packet size is set to 1 megabyte (MB), i.e. 8 Mb, and a maximum bit rate is set to 4 Mbps. The packet size is set relatively high3_{to compensate for the maximum bit rate in}

order to be able to study the eﬀect of high bit rates.

The users are initially distributed uniformly over the simulation area and move according to a Gaussian walk, with a constant average speed and average acceler-ation.

4.5.3 Soft and Softer Handover

The soft and softer handover feature is adopted in the dynamic simulations. A user is allowed to maintain connection to up to three diﬀerent cells at the same time. Furthermore, a user is let in to soft handover when the relative path gain is greater than -2 dB and let go of when the relative path gain falls below -4 dB.

4.5.4 Logging

In order to attain reliable statistics, the simulation time is set to 200 seconds. The initial 20 seconds are discarded to let the system stabilize. User speciﬁc logging

3

(48)

like bit rate and path gain will be logged as an average for each user and only users entering the system after the 20 second threshold will be included in the statistics. System throughput is averaged over both time and cells while noise rise is logged in a histogram for each TTI. The simulations are performed three times using diﬀerent seeds to attain suﬃcient statistics.

4.5.5 Hybrid ARQ with Soft Combining

Hybrid ARQ with soft combining will be adopted using 8 parallel HARQ queues and a block error rate (BLER) set to 10 %. 8 parallel processes imply that 7 other queues will be handled until the ﬁrst process has received an ACK or NACK and can be retransmitted if needed. The number of parallel queues is set to correspond to the round trip time. The BLER is the probability of an erroneous block being received. A BLER set to 10 % corresponds to 90 % of the transmitted blocks being received correctly on the ﬁrst transmission attempt.

4.5.6 Scheduling

A maximum number of users will be admitted by the RNC and users not let in will wait until admitted. Active users will be given an initial transmission grant based on available noise rise resources. The resources will be shared equally among the active users by the scheduling algorithm. During each TTI the initial transmis-sion grant for each user is tuned to maximize eﬃciency according to the number of transmitting users. Noise rise measurements needed to tune transmission grants will be estimated with true noise rise to attain an optimistic assessment of perfor-mance.

A user in soft handover will receive transmission grants from all cells the active set and the minimum grant will be applied.

4.5.7 G-RAKE Receiver Model

In order to reduce effects from the multipath fading a RAKE-receiver [15, 17] has been used in previous WCDMA releases. A RAKE receiver has multiple antennas called fingers receiving the different signal components reaching the receiver. By using the correlation between the different fingers the own signal interference can be reduced. Figure 3.1 shows that an α ≈ 0.6, which is commonly used to model the own signal interference in a typical urban scenario4_{, corresponds to a}

maxi-mum bit rate of approximately 2 Mbps. A regular RAKE-receiver will obviously not be suﬃcient to reach the goal of up to 4 Mbps in the EUL concept. Instead the

generalized RAKE-receiver (G-RAKE) will be employed. The G-RAKE receiver

utilizes the correlation between the receiving ﬁngers of the antenna more eﬃciently resulting in less own signal interference. The G-RAKE receiver is modeled by set-ting the orthogonality factor α to zero resulset-ting in perfect own signal interference management.

4

The typical urban scenario is designed by 3GPP to be able to compare different simulation results.

(49)

Chapter 5

Simulation Results

Results from simulations will be presented and analyzed in this chapter. The sim-ulations are conducted using 3G simulators in Matlab developed by Ericsson Re-search. Section 5.1 presents results from the statistically generated path gain maps, followed by Section 5.2 where user traffic is introduced in dynamic simulations. Fi-nally comparisons between the theoretical assessments and the simulation results will be made in Section 5.3.

5.1 Path Gain Map Generation

The network deployment, i.e. how site location, antenna etc. is chosen, will affect the characteristics of the network and thereby the efficiency. In order to observe effects of cell radius and antenna tilt settings, statistical path gain maps representing static snapshots in time are generated and studied.

5.1.1 Path Gain

A user will theoretically experience a path gain to all cells in the system but only cells in a certain vicinity of the user will have any practical importance since path gain drops rapidly with distance. The cell to which a user has best path gain will most likely be the serving cell with the exception of users in handover which are connected to multiple cells. Figure 5.1 illustrates the best path gain experienced in each bin for diﬀerent antenna tilts in a scenario where cell radius is set to 500 meters.

The antenna tilt aﬀects the experienced path gain throughout the network, by tilting the vertical antenna pattern1_{the cell overlap can be controlled. As the tilt}

is increased the cell coverage overlap is decreased but so is the average path gain experienced in the network as shown by the histograms in Figure 5.1. At a certain point, when the antenna tilt is too extreme, coverage is lost. This eﬀect can be

1

Described in Figure 4.1(b).

Uplink Interference Management of High Bit Rate Users in Evolved WCDMA

Institutionen för systemteknik

Department of Electrical Engineering

Examensarbete

Uplink Interference Management of High Bit Rate

Users in Evolved WCDMA

Uplink Interference Management of High Bit Rate

Users in Evolved WCDMA

Examensarbete utfört i Kommunikationssystem

vid Tekniska högskolan i Linköping

av

Abstract

Acknowledgements

Abbreviations

Contents

Chapter 1

Introduction

1.1

Background

1.2

Problem Statement

1.3

Research Approach

1.4

Related Work

1.4.1

WCDMA Capacity

1.4.2

Resource Efficiency

1.4.3

Load Estimation

1.5

Thesis Outline

Chapter 2

Third Generation Mobile

Communication System

2.1

Wideband Code Division Multiple Access

2.2

Network Architecture

Core Network

2.2.1

Soft and Softer Handover

2.3

Enhanced Uplink

2.3.1

Short Transmission Time Interval

2.3.2

Hybrid ARQ with Soft Combining

2.3.3

Fast Scheduling

Chapter 3

Theoretical Assessments

3.1

Definitions

3.1.1

Multipath Propagation

3.1.2

Shannon’s Theorem

3.2

Soft and Softer Handover

3.3

System Throughput Optimization

3.3.1

Equal Background Noise

3.3.2

Optimization Problem

3.3.3

Linearity and Convexity

3.3.4

Equal Maximum Noise Rise

3.3.5

Nonlinear Optimization Problem

3.3.6

Quadratic Optimization Problem

Chapter 4

Simulation Models

4.1

Radio Channel Model

4.1.1