Coexistence of Real Time and Best Effort Services in Enhanced Uplink WCDMA

Institutionen för systemteknik

Department of Electrical Engineering

Examensarbete

Coexistence of Real Time and Best Effort Services

in Enhanced Uplink WCDMA

Examensarbete utfört i Kommunikationssystem vid Tekniska högskolan i Linköping

av Erik Axell LITH-ISY-EX-3615-2005

Linköping 2005

Department of Electrical Engineering Linköpings tekniska högskola

Linköpings universitet Linköpings universitet

Coexistence of Real Time and Best Effort Services

in Enhanced Uplink WCDMA

Examensarbete utfört i Kommunikationssystem

vid Tekniska högskolan i Linköping

av

Handledare: Erik Geijer Lundin

isy, Linköpigs universitet

Eva Englund

Ericsson Research, Linköping

Examinator: Fredrik Gunnarsson

isy, Linköpigs universitet

Avdelning, Institution

Division, Department

Division of Automatic Control Department of Electrical Engineering Linköpings universitet S-581 83 Linköping, Sweden Datum Date 2005-01-28 Språk Language ¤ Svenska/Swedish ¤ Engelska/English ¤ £ Rapporttyp Report category ¤ Licentiatavhandling ¤ Examensarbete ¤ C-uppsats ¤ D-uppsats ¤ Övrig rapport ¤ £

URL för elektronisk version

http://www.ep.liu.se/exjobb/isy/2005/3615

ISBN

—

ISRN

LITH-ISY-EX-3615-2005

Serietitel och serienummer

Title of series, numbering

ISSN

—

Titel

Title

Samexistens av Realtidstjänster och Förbättrade Datatjänster i WCDMA Upplänk Coexistence of Real Time and Best Effort Services in Enhanced Uplink WCDMA

Författare

Author

Erik Axell

Sammanfattning

Abstract

The increasing use of data services and the importance of IP based services in third generation mobile communication system (3G), requires the transmission from the cell phone to the base station, i.e. uplink, to manage high speed data rates. In the air interface for 3G in Europe, WCDMA, a concept for enhancing the transmission from the cell phone to the base station, called Enhanced Uplink, is being standardized. The overall goal is to provide high speed data access for the uplink. One of the requirements is that the enhanced uplink channels must be able to coexist with already existing WCDMA releases. For example, the enhanced uplink must not impact seriously on real time services, such as speech, carried on current WCDMA channels.

The purpose of this work is to study how the quality, coverage and capacity of real time services carried on previous WCDMA releases is affected when introduc-ing the Enhanced Uplink in a WCDMA network. The main focus of the study is thus to demonstrate the trade-off between voice and best effort performances.

Theoretical assessments and simulations show that the Enhanced Uplink has many advantages over previous WCDMA releases. For example the Enhanced Uplink yields a larger system throughput for all voice loads. The noise rise, i.e. the ratio of total received power to the background noise power is being considered as the resource. It is shown that user traffic carried on the Enhanced Uplink is able to operate under a higher noise rise level as well as to get a higher throughput per noise rise. The resource is hence more efficiently utilized.

Nyckelord

Abstract

The increasing use of data services and the importance of IP based services in third generation mobile communication system (3G), requires the transmission from the cell phone to the base station, i.e. uplink, to manage high speed data rates. In the air interface for 3G in Europe, WCDMA, a concept for enhancing the transmission from the cell phone to the base station, called Enhanced Uplink, is being standardized. The overall goal is to provide high speed data access for the uplink. One of the requirements is that the enhanced uplink channels must be able to coexist with already existing WCDMA releases. For example, the enhanced uplink must not impact seriously on real time services, such as speech, carried on current WCDMA channels.

The purpose of this work is to study how the quality, coverage and capacity of real time services carried on previous WCDMA releases is affected when intro-ducing the Enhanced Uplink in a WCDMA network. The main focus of the study is thus to demonstrate the trade-off between voice and best effort performances.

Theoretical assessments and simulations show that the Enhanced Uplink has many advantages over previous WCDMA releases. For example the Enhanced Uplink yields a larger system throughput for all voice loads. The noise rise, i.e. the ratio of total received power to the background noise power is being considered as the resource. It is shown that user traffic carried on the Enhanced Uplink is able to operate under a higher noise rise level as well as to get a higher throughput per noise rise. The resource is hence more efficiently utilized.

Acknowledgements

I have had the opportunity to perform my master thesis work at Ericsson Research in Linköping. It has been a great time and I have met a lot of inspiring and competent people working at the front line of telecommunication technology. I would like to thank you all for great commitment and interest in my work and, most important, for making me feel welcome.

Special thanks to my supervisor Eva Englund for guiding me through the work and taking time to answering my questions. Thanks also to Ke Wang Helmersson for all the help with the simulations.

I would also like to thank my supervisor at the University, Erik Geijer Lundin and my examiner Fredrik Gunnarsson for support, valuable comments on the work and for all the help with LA_TEX.

Finally, I would like to thank Andreas Bergström, parallel master thesis stu-dent and opponent, for good company and for helpful discussions and comments. Erik Axell

Linköping, January 2005

Abbreviations and Acronyms

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

ACK Acknowledgement

BLER Block Error Rate

CDMA Code Division Multiple Access

DCH Dedicated Channel (transport channel)

DS-CDMA Direct Sequence Code Division Multiple Access E-DCH Enhanced Dedicated Channel (transport channel)

EUL Enhanced Uplink

FDD Frequency Division Duplex

GSM Global System for Mobile communication HARQ Hybrid Automatic Repeat Request HSDPA High Speed Downlink Packet Access

IP Internet Protocol

ISI Inter Symbol Interference

ITU International Telecommunication Union kbps Kilobits per second

Mbps Megabits per second

MMS Multimedia Messaging Service NACK Negative Acknowledgement Node B Base station

PC Power Control

QoS Quality of Service

RAN Radio Access Network

RLC Radio Link Control

RNC Radio Network Controller

RTT Round Trip Time

Rx Receiver

SIR Signal to Interference Ratio SNR Signal to Noise Ratio

x

TCP Transmit Control Protocol TDD Time Division Duplex TTI Transmission Time Interval

Tx Transmitter

UE User Equipment

UMTS Universal Mobile Telecommunication Services WCDMA Wideband Code Division Multiple Access

Contents

2 Third Generation Mobile Communication System 5 2.1 UMTS Network Architecture . . . 5

2.2 Introduction to WCDMA . . . 7 2.3 WCDMA Evolvement . . . 8 2.3.1 Enhanced Uplink . . . 8 3 Theoretical Assessments 11 3.1 Pole Capacity . . . 11 3.2 Resource Efficiency . . . 13

3.3 Maximum Number of Users . . . 17

3.4 Throughput per Noise Rise . . . 20

3.5 Coverage . . . 21 4 Simulation Model 25 4.1 Propagation Model . . . 25 4.1.1 Shadow Fading . . . 25 4.1.2 Multipath Fading . . . 26 4.2 Simulation Scenarios . . . 26 4.2.1 Traffic Model . . . 26 4.2.2 Cell Deployment . . . 26 4.2.3 User Placement . . . 26 4.3 System Model . . . 27 4.3.1 Fast HARQ . . . 27 4.3.2 Power Control . . . 28 4.3.3 Admission Control . . . 28

4.3.4 Fast Rate Control . . . 29

4.4 Simulation Logging . . . 29 xi

5 Simulations and Results 31

5.1 Performance Measures . . . 31

5.2 Evaluation Criteria . . . 32

5.3 Voice Only . . . 33

5.4 Simultaneous Voice and Best Effort Data . . . 35

5.4.1 At the Capacity Limit . . . 38

6 Conclusions 45

Chapter 1

Introduction

The introduction of the third generation mobile communication system (3G) opened up new doors in wireless communication. The possibilities of transmitting all kinds of data between cellular phones has increased enormously during the last decade. The air interface standard for 3G in Europe, called WCDMA, makes it possible to use the cellular phone for web surfing, e-mailing, interactive gaming, video streaming and receiving several other data services like multimedia, video-clips and pictures.

However, new features create new demands on the communication system. In the WCDMA specifications a concept called high speed downlink packet access (HSDPA) has been evolved. This concept makes it possible to transmit high speed data from the base station to the cell phone, i.e. downlink.

The increasing use of data services and the importance of IP based services also requires the transmission from the cell phone to the base station, i.e. uplink, to manage high speed data rates. The standardization body for WCDMA is the 3rd generation partnership (3GPP). Within 3GPP a concept for enhancing the transmission from the cell phone to the base station, called Enhanced Uplink (EUL), is being standardized. The overall goal is to provide high speed data access also for the uplink. One of the requirements that has been agreed upon within 3GPP is that the enhanced uplink channels must be able to coexist with already existing WCDMA releases. For example, the enhanced uplink must not impact seriously on real time services, such as speech, carried on current WCDMA channels.

1.1

Problem Statement

The master thesis assignment is to study how the quality, coverage and capacity of real time services carried on previous WCDMA releases is affected when intro-ducing the Enhanced Uplink in a WCDMA network. Coverage is the geographical area within which a user can connect to the mobile network with acceptable qual-ity. Capacity is the number of users or the total amount of data bits per time instant that can be supported by the system. Specifically the impact on quality

2 Introduction

and capacity of speech carried on previous WCDMA releases when introducing the Enhanced Uplink in the same network will be studied.

The aim is to demonstrate the trade-off between real time and best effort performances and to investigate possible solutions to mitigate the impact on real time services.

1.2

Related Work

As far as we know, no previous work has been performed on evaluating simultane-ous voice and Enhanced Uplink users. However, a large amount of work has been done on performance measures and evaluations of CDMA systems. The following section will give an outline of such previous related work.

Quality is closely connected to both capacity and coverage measures. Capacity is most often measured as a maximum throughput while still keeping the quality requirements fulfilled, and coverage as a geographical area within which the quality requirements fulfilled. Therefore no specific section for the quality measure is given.

1.2.1

Capacity

Several publications deal with the problem of estimating the system capacity, e.g. [8] and [19] consider an integrated voice/data WCDMA system. Both define capacity as the maximum amount of users for which the probability that they are satisfied is greater than some certain percentage. In the latter a user is defined satisfied when the probability that the bit error rate is below some threshold, is greater than some value. Since the bit error rate is directly related to the received SIR, the satisfaction measure can also be formulated as the probability that the received SIR is greater than some threshold. The capacity region, in terms of number of users, is obtained for an integrated speech and long constraint delay data system.

This measure is also exemplified in [8], however a more general method is introduced. Furthermore the capacity is evaluated and optimized. It is shown that the total system capacity is maximized when the per-service capacities for all bearer services are equal. It is also shown that there is a linear relationship between speech capacity and interactive data capacity.

Another common measure for data capacity is the system throughput, i.e. the maximum possible transmission data rate (often in kbps) as a function of received interference power or noise rise. This measure is used in [5] to evaluate the uplink capacity gain in a WCDMA network due to faster scheduling. A capacity gain of approximately 10% is shown, simply by reducing the packet scheduling interval from 500 ms to 100 ms.

The quality in terms of bit error rates and blocking probabilities for voice calls in an integrated voice/data DS-CDMA network is discussed in [9]. The bit error rate is derived as a function of the number of voice and data calls, and the relationship is used to determine the capacity of the network. The tradeoff between delay and capacity for data users is also illustrated.

1.2 Related Work 3

An optimization approach to support multimedia services in CDMA is pre-sented in [15]. The total transmitted power is minimized and the sum of rates is maximized while all QoS constraints are still fulfilled. Each user specifies a QoS in form of maximum bit error rates that are mapped into equivalent SIR requirements. The solutions are plotted as number of voice users against number of data users, which yields a capacity region. The capacity region is defined by a straight line between approximately 32 speech users and 14 data users with bit rate 20 kbps.

Speech capacity is often measured in Erlangs, i.e. the traffic density computed as the ratio between call arrival rate and call departure rate. In [16], an integrated voice/data CDMA system is considered, and the voice and data capacity in Erlangs is calculated both analytically and numerically. In [16] perfect power control as well as imperfect power control is investigated with both Gaussian and Lognormal approximation of the traffic load. A similar assessment on a DS-CDMA system is performed in [11]. The two latter works only consider the single cell case.

In [13], the so called F-factor is introduced, defined as F = Ioc

Ioc+Isc where Ioc

is the interference power caused by users in the own cell (intra cell) and Isc is the

interference caused by users in the surrounding cells (inter cell). The difficulties and possibilities of deriving the F-factor is discussed and the capacity in terms of number of users is estimated.

Both [4] and [12] describes pole capacity, but with a slightly different approach. In [12] the results are also verified in simulations, while [4] gives a more general theoretical account. Pole capacity will be discussed in more detail in Section 3.1.

1.2.2

Coverage

A method of calculating the inter and intra cell interference in a UMTS system is discussed in [18]. The interference is calculated by solving a system of fixed-point equations, using an iteration algorithm. Furthermore, the result is used to reckon the cell coverage area as the distance within which the outage probability is less than a certain percentage.

An analytical approach to determine coverage probabilities are given in [17]. An algorithm is used where the interference at all cells, the transmission powers of all mobiles and the probability function of received powers at the base stations are calculated. From this the coverage probability for a given mobile is reckoned. Another analytical approach is described in [14]. The uplink coverage is in-vestigated in a UMTS system under non-homogenous and moving traffic load. In particular the coverage is investigated for different call assignment policies and for a hot spot moving among the cells. Inter cell interference is also taken into account. Some numerical results to verify the analysis are also given.

4 Introduction

1.3

Research Approach

To achieve the thesis goals the study have been made in the following steps:

• Literature study on mixing real time and best effort services in uplink in CDMA-systems, particularly on various ways of evaluating the system qual-ity, coverage and capacity. An outline of this was presented in the previous section. An introduction to the third generation mobile communication sys-tem and a description of WCDMA evolution to the Enhanced Uplink is given in Chapter 2.

• Theoretical assessments, also including model description and methods of evaluating and comparing the performance of speech versus data. The the-oretical assessments and results are described in more detail in Chapter 3.

• Simulation scenario design. The most important simulation models and as-sumptions as well as the simulation scenarios are described more in Chapter 4.

• Evaluation methods and criterions selection. The performance measures and evaluation criteria used to evaluate the simulations are presented in sections 5.1 and 5.2 respectively.

• Simulations and evaluation of simulation results. Comparison with the the-oretical assessments. The simulations and results are presented in sections 5.3 and 5.4.

• Documentation and conclusions from the results. The conclusions of the study are presented in Chapter 6.

Chapter 2

Third Generation Mobile

Communication System

The standard for third generation mobile communication system in Europe is referred to as Universal Mobile Telecommunication Services (UMTS), adopted by the International Telecommunication Union (ITU). The air interface used in UMTS is Wideband Code Division Multiple Access (WCDMA). The following sections will give an overview of WCDMA for UMTS and it’s evolution to the Enhanced Uplink concept. More about UMTS can also be found in [6], [1] and [10] and the basic principles of wireless communications are described in [3].

2.1

UMTS Network Architecture

The UMTS network consists basically of a core network, Radio Network Con-trollers (RNC), base stations (Node B) and user equipments (UE). Figure 2.1 shows a schematic picture of the UMTS network and it’s elements.

The core network is the connection to an external network such as the internet or the ordinary fixed telephony system. The UE can be a mobile phone or a computer card. The RNC and Node B constitute the Radio Access Network (RAN), or the connection between the UE and the core network.

Each RNC controls a number of Node Bs. The actual radio signal is transmitted and received by a Node B. Each Node B supports one or a number of cells covering a geographical area. If the antenna for example is of a three sector type, the Node B consists of three cells. Each UE within a cell area is connected to a certain Node B. The cells normally intersect near the cell borders, and UEs positioned in this area are connected to more than one Node B.

Handover is executed when a user moves between cells. Soft handover is when the UE is connected to several Node Bs, and softer handover is when the UE is connected to several cells within the same Node B, see Figure 2.2. Soft and softer handover enables the UE to maintain the continuity and quality of the connection while moving from one cell to another.

6 Third Generation Mobile Communication System

RNC

RNC RNC

Core network

Figure 2.1. UMTS architecture.

Node B Node B

RNC

Soft handover Softer handover

2.2 Introduction to WCDMA 7

2.2

Introduction to WCDMA

In previous generations of mobile communication systems, users are separated by transmitting in different time slots and/or using different frequencies. In WCDMA users are separated through Code Division Multiple Access (CDMA). In this scheme each user is assigned on a unique code which makes it possible for several users to transmit on the same frequency at the same time.

WCDMA uses Direct Sequence CDMA. The original signal is spread by a multiplication with a spreading code, consisting of a sequence of 1 and -1 bits, also called chips. The spreading codes are chosen from the code tree in Figure 2.3. The different levels of the code tree corresponds to different code lengths. Once a user has been dedicated a code, no code in the subtree of that code can be used. This preserves orthogonality between codes, even for codes with different lengths.

c (c, c) (c, -c) c_1,1= (1) c2,1= (1, 1) c_2,2= (1, -1) c_4,1= (1, 1, 1, 1) c_4,2= (1, 1, -1, -1) c4,3= (1, -1, 1, -1) c_4,4= (1, -1, -1, 1)

…

Figure 2.3. Channelization code tree.

Figure 2.4 shows the principal of spreading and despreading. In this example, every data symbol is multiplied by a spreading code sequence of 8 chips. We say in this case that we have used a spreading factor of 8, i.e. the ratio of the chip rate to the data rate.

The multiplication with a spreading code with chip rate larger than the data rate results in an ostensibly random signal. Multiplying the spread signal with the spreading code again, i.e. despreading, restores the signal to it’s original, see Figure 2.4. A multiplication of the spread signal with the wrong spreading code would result in a signal looking like noise.

The signal bandwidth is proportional to the bit rate. Since the signal is spread by the spreading factor, the bandwidth will also widen with the spreading factor. WCDMA uses a chip rate of 3.84 Mcps, which yields a bandwidth of approximately 5 MHz. Compared to other CDMA technologies, this bandwidth is wider, and hence the name Wideband CDMA.

8 Third Generation Mobile Communication System Data Spreading code Spread signal = data x code 1 -1 Spreading code 1 -1 Spreading Despreading Symbol Chip Data = spread signal x code

1 -1 1 -1 1 -1

Figure 2.4. Spreading and despreading in DS-CDMA.

Figure 2.5 shows the bandwidth widening as a consequence of spreading, and how the receiver easily can find and decode the correct signal.

WCDMA supports two ways of separating downlink and uplink. The separa-tion can be done either by Frequency Division Duplex (FDD) or Time Division Duplex (TDD). The FDD mode is the one that operators are now deploying in WCDMA, and TDD mode will thus not be considered in the sequel.

2.3

WCDMA Evolvement

The WCDMA standard is continuously evolving. Figure 2.6 shows the WCDMA evolvement in principle. In the standardized air interface WCDMA release 5, a concept for high bit rate downlink transmission was introduced, called High Speed

Downlink Packet Access (HSDPA).

Furthermore, the increasing use of data services and the importance of IP based services also requires the uplink transmission to manage high speed data rates. Within the 3rd generation partnership (3GPP) a concept for enhancing the transmission from the cell phone to the base station, called Enhanced Uplink, is being developed. The basic functions of Enhanced Uplink are described more in the following section. The basics of HSDPA can be found in [7].

2.3.1

Enhanced Uplink

Many solutions from HSDPA are also used in the Enhanced Uplink, but there are also differences in downlink and uplink. The Enhanced Uplink concept is described in more detail in [2]. The overall goal is to improve coverage and throughput as well as to reduce the delay of the uplink. Among the requirements that have been agreed on within 3GPP is that the enhanced uplink channels must be able

2.3 WCDMA Evolvement 9 frequency power frequency power frequency power frequency power Spreading Despreading a b c d

Figure 2.5. The spread signal in (b) occupies a larger bandwidth than the original signal in (a). Many user transmits at the same time in (c). After despreading in (d) the correct signal can easily be found.

Enhanced

Uplink _enhancementsAdditional Enhanced Downlink

(HSDPA)

Rel 4 Rel 5 Rel 6

WCDMA

WCDMAEvolvedEvolved

WCDMA

WCDMA

R99

Enhanced

Uplink _enhancementsAdditional Enhanced Downlink

(HSDPA)

Rel 4 Rel 5 Rel 6

WCDMA

WCDMAEvolvedEvolved

WCDMA

WCDMA

R99

10 Third Generation Mobile Communication System

to coexist with already existing WCDMA releases. For example, the enhanced uplink must not impact seriously on real time services, such as speech, carried on current WCDMA channels.

Basic Principles

The enhanced uplink introduces a new transport channel, the Enhanced Dedicated Channel (E-DCH). A dedicated channel (DCH) is assigned to only one UE at a time. A DCH is power controlled, meaning that the transmitter power is increased if the channel is bad and, important in the uplink, the power is limited if too much interference is caused by the user. The E-DCH may transmit on a 2 ms basis, which is five times as often as transmission on the previous DCH. The shorter

Transmission Time Interval (TTI) reduces uplink delays and makes it possible

to retransmit faster and to adapt faster to the system interference. E-DCH also supports 10 ms TTI in order to coexist with previous WCDMA releases. The E-DCH also supports the following new features, which improves system capacity.

- Fast hybrid automatic repeat request (HARQ). The enhanced uplink

sup-ports Node B controlled retransmissions, unlike previous WCDMA releases where retransmissions are controlled from the RNC. By rapidly request re-transmission of erroneous data, the delays are reduced essentially and the capacity is increased. The enhanced uplink also uses soft combining in the Node B, which means that data blocks that can not be correctly decoded are saved and combined with later retransmissions of the same blocks to find the correct data. With soft combining the number of retransmissions are reduced. Fast HARQ with soft combining leads to higher capacity and robustness against link adaption errors.

- Fast rate control. Since the uplink is interference limited a fast adaption

to the interference conditions is necessary. In the Node B the uplink data rate for each user is controlled. Moving the rate control to the Node B reduces delays which leads to a rapid adaption and a tight control of uplink interference. The fast rate control also allows admission control in the RNC to be more relaxed. A larger number of bursty high bit rate users can be allowed. This in all yields a higher uplink capacity.

The E-DCH also supports soft handover, just like previous WCDMA releases. During handover between cells the user is connected to both (in some cases even three) cells at the same time. This allows power control from multiple cells and is required to limit the inter cell interference, which is an important issue in the uplink.

In previous WCDMA releases both voice and best effort users are carried on the DCH. However, in the sequel of this work DCH refers to previous WCDMA release channels carrying best effort users, i.e. without the Node B controlled retransmissions and rate control, and with 10 ms TTI. Voice users are simply named voice or speech. In the same way E-DCH refers to best effort users carried on the enhanced uplink supporting 2 ms TTI and the features mentioned above.

Chapter 3

Theoretical Assessments

As mentioned before there are several ways of measuring system performance. We consider pole capacity and resource efficiency and assess the maximum number of users by combining these two quantities. The last section of this chapter will take coverage in consideration. Since the quality measure is incorporated in the other measures, no specific analysis of quality will be performed.

When talking about capacity one can consider hard capacity or soft capacity, where soft capacity depends on factors like where the users are positioned. It can be shown that soft capacity is a more optimistic measure than hard capacity. We only consider hard capacity measures.

We aim on showing that the enhanced uplink has many advantages over pre-vious WCDMA releases in many aspects, mainly in a better resource utilization and hence a larger amount of users and higher throughput.

3.1

Pole Capacity

In this section we assume perfect power control and unlimited UE transmitter power. Perfect power control means that the UE power is controlled such that the received SIR is always equal to the target SIR. With these assumptions the noise rise level will for a certain traffic load approach infinity. The maximum number of users where the noise rise level approaches infinity is called pole capacity. An analytical expression as well as a numerical approximation of the pole capacity is derived.

Assuming perfect power control yields that all voice users cause the same re-ceived power and, if we also assume all data users have the same bit rate, all data users cause the same received power. The SIR targets, γv and γd, for voice and

data respectively can be expressed

γv=

Pv

Itot− (1 − ω)Pv, γd=

Pd

Itot− (1 − ω)Pd, (3.1)

where ω is the own signal interference, Pv, Pd are the received powers from a

voice and data user respectively and Itot is the total received interference. The

12 Theoretical Assessments

own signal interference is caused by time dispersion due to multipath propagation. Some parts of the signal are so delayed that they interfere with the next symbol. This phenomena is known as inter symbol interference (ISI), and can be surpressed with more advanced receiver types. However with a standard RAKE receiver, the own signal interference can be significant. A RAKE receiver is basically a collection of multiple correlation receivers, called fingers, who combine all of the received signals. Here the own signal interference is assumed constant and equal for all users. This is a simplification, in reality it varies a lot from user to user and in time.

The total interference,

Itot= N0+

N

X

n=1

Pi+ Isc,

where N0 is the background noise power, Pi is the received power from user i,

N is the total number of users in the cell and Isc is the interference power from

surrounding cells. We denote the received powers from all voice users and all data users in the own cell Iv and Id respectively, such that Iv+ Id =

P_N

n=1Pi. The

interference caused by users in the own cell is hence Ioc = Iv+ Id.

Approximat-ing the interference from surroundApproximat-ing cells, Isc, by a factor ξ times the own cell

interference yields the following expression for the total interference.

Itot= N0+ Ioc+ Isc= N0+ (1 + ξ)Ioc= N0+ (1 + ξ)(Iv+ Id)

Solving (3.1) for Pv and Pd yields

Pv= ₁ Itot

γv + (1 − ω)

, Pd= ₁ Itot

γd + (1 − ω)

. (3.2)

We define the system noise rise, which is considered as the limiting system resource since the uplink is mainly interference limited.

Definition 3.1.1 (Noise rise) Noise rise, η, is defined as the ratio of the total received power to the background noise power, i.e.:

η = Itot

N0

.

¤ Definition 3.1.1 also yields

ηN0= Itot= N0+ (1 + ξ)(Iv+ Id), (3.3)

where ξ is the expansion factor. According to the definition above

ξ = Isc

Ioc.

The factor 1 + ξ is equal to the inverse of the F-factor [13], previously defined in section 1.2.1, i.e. 1 + ξ = Ioc+Isc

Ioc =

1

F.

We define a service mix as a vector whose elements are the fractional contri-butions of each bearer service to the expected total number of users [8].

3.2 Resource Efficiency 13

Definition 3.1.2 (Service mix)

α = (αv, αd) = ( uv

utot,

ud

utot)

where uv, ud is the expected number of voice and data users respectively and

utot= uv+ udis the expected total number of users.

¤ The assumption that all voice users cause the same received power, Pv, and all

data users cause the same received power, Pd, together with the service mix, α,

transforms (3.3) to

ηN0= N0+ (1 + ξ)(Pvuv+ Pdud) = N0+ (1 + ξ)utot(Pvαv+ Pdαd). (3.4)

Combining (3.2), (3.4) and Definition (3.1.1) and solving this for η yields

η = 1 1 − (1 + ξ)utot( 1 αv γv+1−ω+ αd 1 γd+1−ω) (3.5) Since η > 0 per definition we conclude from the above equation that

(1 + ξ)utot( ₁ αv γv + 1 − ω

+ ₁ αd

γd + 1 − ω

) ≤ 1, (3.6)

for any stable system. Letting (3.6) approach equality implies η → ∞, and from this we get the pole capacity.

utot= 1 (1 + ξ)( αv 1 γv+1−ω+ αd 1 γd+1−ω) . (3.7)

This expression for pole capacity will be exemplified later in Example 3.2, Section 3.3.

3.2

Resource Efficiency

Among the expectations of the enhanced uplink is a better resource utilization. Thanks to fast rate control, we have a more efficient use of the available resource. In this section we give a description of the utilization of the resource and also give some propositions and explanations to it’s behaviour.

Some useful variables are defined to describe the model. At this stage we consider the multiple service case, i.e. there are several different services simul-taneously in the system. Each service has some specific quality of service (QoS) requirements, such as maximum delay, maximum block error rate (BLER), mini-mum bit rate and so on. For a user to be satisfied, all of the requirements for the current service must be fulfilled. This leads to the definition of our first quantity,

14 Theoretical Assessments

Definition 3.2.1 (Maximum noise rise) The maximum noise rise allowed in the cell, i.e. the maximum noise rise at the receiving base station such that all of

the QoS requirements are fulfilled, is denoted ηmax.

¤ Note that the noise rise, η, is a stochastic process that varies over time and from one cell to another. Hence it is valuable to consider the expectation value of the noise rise.

Definition 3.2.2 (Maximum expected noise rise) The maximum expected noise rise, η, is defined as

η = max{E[η] : P r(η ≥ ηmax) ≤ P },

where η is the system noise rise.

¤ Furthermore, we define the resource efficiency, ρ, as the ratio of the maximum ex-pected resource utilization, i.e. the maximum exex-pected noise rise, to the available resource, i.e. the maximum allowed noise rise.

Definition 3.2.3 (Resource Efficiency)

ρ = η

ηmax

¤ This might seem as a bad measure of how good the resource is being utilized. It should be no problem to get a high noise rise without getting a good throughput. However it is a good measure of how much of the resource is being utilized. As-suming that the amount of utilized resource is correlated to a good throughput justifies this measure. That this is in fact the case is shown in Section 3.4 and later in simulations.

Now the single service case is considered. The single service case means there is only one service type in the system. We consider single service systems with different services n, for example only voice users or only best effort users. By analogy with Definition 3.2.2 the maximum expected noise rise for a single service

n is defined.

Definition 3.2.4 (Maximum expected noise rise per service) The maximum expected noise rise for a single service n, is defined as:

ηn= max{E[ηn] : P r(ηn ≥ ηmax) ≤ P }

3.2 Resource Efficiency 15

For the single service case the maximum expected noise rise is of course the same as the maximum expected noise rise for the single service.

The resource efficiency for a single service is defined analogous to the previous definition.

Definition 3.2.5 (Resource efficiency per service) The resource efficiency for a single service n is defined as:

ρn= ηn

ηmax.

¤ Again this of course looks exactly the same as Definition 3.2.3, since the single service case is only a special case of the multiple service case.

Now the interesting problem is to find out if and, if so, how the resource efficiency depends on the service mix. To do some valuable evaluation of the resource efficiency we have to make some assumptions about the system behaviour. The sequel of this section will describe some possible assumptions and how this will affect the resource efficiency. We will only consider the two service case with simultaneous voice and best effort data users.

Assumption 3.2.1 Resource efficiency, ρ, is independent of the service mix, α.

Assumption 3.2.2 Resource efficiency, ρ, depends linearly on the single service resource efficiencies according to:

ρ = βvρv+ βdρd,

where βv and βd are parameters dependent on the service mix, α.

For the latter assumption, the question is how βv and βd depend on the service

mix. All reasonable parameters βv, βd must apply to the following relation:

βv= 0 ⇐⇒ βd= 1 and βv= 1 ⇐⇒ βd= 0. (3.8)

A reasonable assumption would be that βvand βdsomehow relates to the

inter-ference contribution per service for a given α, or rather the fraction of interinter-ference caused by each service , i.e.

βv= Iv

Iv+ Id, βd=

Id

Iv+ Id, (3.9)

where Iv, Id is the interference caused by voice users and data users, respectively.

This definition yields βv+ βd = 1, and hence (3.8) is fulfilled. We will now try

to find an expression for the coefficients βv and βd. To start with we consider

only the voice coefficient, βv. The interference caused by voice users can also be

expressed as:

16 Theoretical Assessments

where Pv is the received power from one voice user. Analogous we will get the

same expression for Id. Equations (3.9) and (3.10) yield the following expression

for βv.

βv= Pvαv

Pvαv+ Pdαd

Using expressions (3.2) for the received powers yields

βv = Itot 1 γv+(1−ω)αv Itot 1 γv+(1−ω)αv+ Itot 1 γd+(1−ω)αd = 1 1 + γv1 +(1−ω) 1 γd+(1−ω)· αd αv

In the same way for data users we get

βd= 1 1 + 1 γd+(1−ω) 1 γv+(1−ω)· αv αd Assume ½ 1 − ω ¿ 1 γv 1 − ω ¿ 1 γd

. This is reasonable for small SIR, but not for large SIR, i.e. for high bit rates. In simulations we use a target bit rate of 320 kbps, which is not very high. Hence this is a reasonable simplification. Neglecting the term (1 − ω), yields βv= 1 1 + γdαd γvαv , βd= 1 1 + γvαv γdαd . (3.11)

For small SNR the ratio of SIR targets is approximately the same as the ratio of bit rates, i.e.

γv

γd ≈

Rv

Rd,

and hence (3.11) becomes

βv= 1 1 +Rdαd Rvαv , βd= 1 1 + Rvαv Rdαd .

Inserting this in Assumption (3.2.2) yields the following expression for the resource efficiency. ρ = ρv 1 + Rdαd Rvαv + ρd 1 + Rvαv Rdαd (3.12) Consequently, these assumptions gives us an expression for the resource efficiency dependent only on the service mix and the data rates. Also note that the depen-dency is on the ratio of bit rates and the ratio of service mix elements, rather than absolute values. This also leads to three interesting cases:

Rd> Rv. In this case the denominator of the first term in (3.12) will be large

and hence ρv will have a small impact on ρ.

Rd< Rv. By analogy with the previous case, ρv will now be dominating and ρd

3.3 Maximum Number of Users 17

Rd= Rv= R. If the bit rates are equal the resource efficiency will get a linear

dependency of ρd and ρv according to:

ρ = ρv 1 + Rαd Rαv + ρd 1 +Rαv Rαd = ρvαv+ ρdαd αv+ αd .

Since αd+ αv= 1 and hence αd= (1 − αv) this transforms into

ρ(αv) = ρvαv+ ρd(1 − αv),

which represents a straight line between ρd and ρv.

The conclusion is that the more alike the bit rates are, the more the resource efficiency looks like a straight line and the more the bit rates differ, the more the resource efficiency for the service with larger bit rate dominates. This is shown in the following example, with the intention to use as realistic values as possible.

Example 3.1: Resource Efficiency

This example shows what the resource efficiency from (3.12) looks like, i.e. the resource efficiency when using Assumption 3.2.2. We use as realistic parame-ters as possible, i.e. the same parameparame-ters as used in simulations. The voice bit rate is chosen to Rv = 12.2 kbps and the best effort data bit rate is chosen to

Rd = 64 kbps in Figure 3.1 and Rd = 320 kbps in Figure 3.2. We use ρv = 0.8

and ρd= 0.5 and 0.3 to illustrate the difference between E-DCH and DCH. These

values are chosen to be approximately the same as the simulation results shown in Table 5.3, section 5.4. In these examples the total resource efficiency is plotted versus the fraction of voice users.

It’s quite obvious that the total resource efficiency is dominated by the resource efficiency for best effort data when using 320 kbps bit rate. This is the case also when using 64 kbps bit rate, but not as significant. This is because the ratio Rd

Rv

is five times larger for 320 kbps than for 64 kbps.

3.3

Maximum Number of Users

We now derive an expression for the maximum number of users in the system by combining theory from the two previous sections. Using (3.5), but with η replaced by the maximum expected noise rise from Definition 3.2.4 yields

η = 1 1 − (1 + ξ)utot( 1 αv γv+1−ω+ αd 1 γd+1−ω) . (3.13)

18 Theoretical Assessments 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Fraction of voice users, α_v

Resource Efficiency,

ρ

ρ_d = 0.5

ρ_d = 0.3

Figure 3.1. Result from Example 3.1. Resource efficiency, ρ for ρv= 0.8, ρd= 0.5, 0.3, Rv= 12.2 kbps and Rd= 64 kbps. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Fraction of voice users, α_v

Resource Efficiency,

ρ

ρ_d = 0.5

ρ_d = 0.3

Figure 3.2. Result from Example 3.1. Resource efficiency, ρ for ρv= 0.8, ρd= 0.5, 0.3, Rv= 12.2 kbps and Rd= 320 kbps.

3.3 Maximum Number of Users 19

From Definition 3.2.3 we get

η = ρ · ηmax. (3.14)

Inserting this in (3.13) and solving it for the expected total number of users, utot,

yields utot= 1 − 1 ρ·ηmax (1 + ξ)( αv 1 γv+1−ω+ αd 1 γd+1−ω) . (3.15)

Since we are calculating with stochastic processes, or rather expectation values, this is not strictly mathematical correct. We can really not just insert the maxi-mum expected noise rise to reckon the expected number of users. However, this is an initial approximation.

We also note that when letting ρ · ηmax→ ∞, this is the exact same expression

as for the pole capacity, just as expected.

Inserting our assumed expression (3.12) for resource efficiency, ρ, in (3.15) yields utot= 1 − Rvαv+Rdαd (ρvRvαv+ρdRdαd)·ηmax (1 + ξ)( αv 1 γv+1−ω+ αd 1 γd+1−ω) . (3.16)

Example 3.2: Maximum number of users and pole capacity

The look of the capacity measure in (3.16) will now be shown in an example. Again we try to use as realistic parameters as possible for E-DCH, i.e. ρv = 0.85

and ρd = 0.5. The voice bit rate is set to 12.2 kbps and the data bit rate is set

to 320 kbps. For this example we also set ηmax = 7 dB, ξ = 0.7, γv = −21 dB

and γd = −8 dB. The own signal interference, ω = 0.68, is reckoned from the

simulator for a 3GPP Typical Urban model. The result is shown in Figure 3.3, where the total number of users is plotted against the fraction of voice users, αv.

As a comparison, the pole capacity from (3.7) is also plotted.

With these assumptions we can also reckon the maximum number of users in the single service cases. For the case with only voice users we see that uv ≈ 56.8,

and with only best effort users ud,EU L ≈ 2.34. The same quantities for the pole

capacity are uv,pole ≈ 74.2 and ud,pole≈ 3.90.

In the same way we reckon the maximum number of users for DCH with the same parameters, but ρd = 0.3. This yields ud,Rel5 ≈ 1.30. Hence, in the single

service best effort data case we get a capacity for E-DCH approximately ud,EU L

ud,Rel5 =

2.34

1.30 = 1.8 times the capacity for DCH. An even larger capacity gain is shown in

20 Theoretical Assessments 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 10 20 30 40 50 60 70 80

Fraction of voice users

Total number of users

Maximum Number of Users Pole Capacity

Figure 3.3. Total number of users, utot, calculated from (3.16) with parameters set to

ρv = 0.85, ρd = 0.5, Rv= 12.2 kbps, Rd = 320 kbps, ηmax = 7 dB, ξ = 0.7, ω = 0.68,

γv= −21 dB and γd= −8 dB.

3.4

Throughput per Noise Rise

This section will estimate the throughput per noise rise, i.e. the ratio of throughput to the noise rise. This is also a way of validating the resource efficiency measure. If we can show that the E-DCH gets a higher or equivalent throughput per noise rise as the DCH, this means that the resource efficiency is a legitimate measure to compare the system performances.

Thanks to the fast HARQ processes it is possible to get a higher bit rate for a given SIR. This leads to a lower noise rise caused by a user with a given bit rate, or equivalent a higher throughput for a given noise rise.

The throughput is calculated as the sum of bit rates for all active users. Since the resource efficiency is only valid at the capacity limit, we only consider the throughput per noise rise for a maximum number of users. Assuming that all users of service n have the same bit rate Rnyields the throughput

P_N

n=1λnRnun, where λn is the activity factor for service n and un is the maximum expected number of service n users. Since we consider the capacity limit, the maximum noise rise, ηmax, from Definition 3.2.1 is used as noise rise measure. This yields the following expression for throughput per noise rise

² =

PN

n=1λnRnun

3.5 Coverage 21

Which in the two service case transforms into

² = utot(λvRvαv+ λdRdαd)

ηmax . (3.17)

Using the expression (3.15) for the maximum expected total number of users,

utot yields ² = 1 − 1 ρ·ηmax (1 + ξ)( αv 1 γv+1−ω+ αd 1 γd+1−ω) ·(λvRvαv+ λdRdαd) ηmax . (3.18)

By a short glance at this expression it is quite obvious that a higher resource efficiency, ρ, also yields a higher throughput per noise rise. Hence with these assumptions, a better resource efficiency also yields a better throughput, and the resource efficiency is for sure a valid measurement.

3.5

Coverage

One of the goals of the Enhanced Uplink is to improve the coverage for high bit rates. The purpose of this section is to theoretically assess the maximum possible bit rate for a given cell radius. The cell radius is given by the maximum allowed noise rise, ηmax, in each cell. Again using (3.5) but with utotαv= uv and

utotαd= ud and η replaced by the maximum noise rise ηmaxyields

ηmax= 1 1 − (1 + ξ)( uv 1 γv+1−ω + ud 1 γd+1−ω) .

At first we assume there are only voice users in the system and calculate the maximum possible SIR target for one new best effort user entering the system, i.e. we assume ud= 1 and solve the equation for γd.

(1 + ξ)( ₁ uv γv + 1 − ω + ₁ 1 γd+ 1 − ω ) = 1 − 1 ηmax ⇒ 1 1 γd+ 1 − ω =1 − 1 ηmax 1 + ξ − uv 1 γv + 1 − ω | {z } K ⇒ 1 γd + 1 − ω = 1 K ⇒ γd= ₁ 1 K− 1 + ω , K = 1 − 1 ηmax 1 + ξ − uv 1 γv + 1 − ω . (3.19)

We now have an expression for the maximum SIR target for a new best effort user entering the system. For this SIR, we can calculate a bit rate. Thanks to the

22 Theoretical Assessments

fast HARQ processes in the enhanced uplink, we can admit multiple transmitting attempts and hence get a capacity gain.

Example 3.3: SIR target for a given coverage

This example will calculate the SIR target reckoned from (3.19). We use the same parameter values as for Example 3.2, i.e. ηmax= 7 dB, ξ = 0.7 and γv= −21 dB.

The own signal interference is again ω = 0.68 for 3GPP Typical Urban, and

ω = 0.17 for the ITU standardized traffic model Pedestrian A, also reckoned from

the simulator. We also know from Example 3.2 that the maximum number of voice users is uv ≈ 56.8. If uv exceeds this value we would get γd < 0, which is

not possible. Figure 3.4 shows the SIR target plotted against the number of voice users. 0 10 20 30 40 50 60 −25 −20 −15 −10 −5 0 5

Number of voice users

SIR target for a data user

Pedestrian A 3GPP Typical Urban

Figure 3.4. SIR target, γd, calculated from (3.19) with parameters set to ηmax= 7 dB,

ξ = 0.7, γv= −21 dB, ω = 0.68 for 3GPP Typical Urban and ω = 0.17 for Pedestrian

A.

To calculate the bit rates from (3.19), we need a relationship between bit rate and SIR target. Figure 3.5 shows a plot of the SIR target versus bit rate for E-DCH using a 3GPP Typical Urban traffic model with perfect own signal interference cancellation. This plot is performed by tuning the SIR targets in the simulator for different bit rates. To derive a analytical expression for the relationship between bit rate and SIR, we use the Shannon capacity formula, i.e.

3.5 Coverage 23

R = C log(1 + γ), (3.20)

where C is a constant. The constant, C, is adjusted such that the simulated and the analytical curves approaches each other. The constant is chosen to be

C = 2.5 · 103_{. This curve is also shown in Figure 3.5.}

0 500 1000 1500 2000 2500 3000 3500 4000 4500 −12 −10 −8 −6 −4 −2 0 2 4 6 Bit rate [kbps] SIR [dB] Simulated Analytical

Figure 3.5. SIR target versus bit rate, reckoned from simulator for 3GPP Typical Urban and analytical using Shannon capacity formula.

Example 3.4: Bit rate for a given coverage

This example estimates the maximum possible bit rate for a new best effort user entering the system, with a given coverage. Equation (3.20) is used to calculate the bit rate for a given SIR target. The SIR target is reckoned from (3.19) with the same parameters as for Example 3.3, using the 3GPP Typical Urban traffic model. Figure 3.6 shows the maximum bit rate versus number of voice users. We see that the offered bit rate seems to be linearly dependent on the number of voice users in the valid region.

One is often interested in the maximum bit rate a best effort user can be

of-fered. For the 3GPP Typical Urban model the maximum SIR is γd≈ −2.56 dB,

which yields approximately 1.1 Mbps maximum bit rate using (3.20).

Interesting is also the absolute maximum bit rate a best effort user can be of-fered. Again this is when uv = 0. Also assuming that there is no power limitation, i.e. η → ∞, and considering only a single cell, i.e. the other to own interference

24 Theoretical Assessments

ξ = 0. These assumptions turns (3.19) into

γd=

1

ω.

For the 3GPP Typical Urban model the mean value ω = 0.68 ⇒ γd = _0.681 =

1.68 dB. The bit rate for this SIR, reckoned from (3.20), is approximately 2.26 M bps. Again this is not strictly mathematical correct since we are dealing with mean val-ues. What is known in this case is that E[γd] = E[_ω1] ≥ _E[ω]1 . This inequality

means that the simplification results in an underestimation of the possible bit rate. As discussed earlier the own signal interference, ω, in reality varies a lot. Here the mean value for the 3GPP Typical Urban traffic model is considered. In reality this of course means also that this maximum bit rate varies a lot.

0 10 20 30 40 50 60 0 200 400 600 800 1000 1200

Number of voice users

Maximum bit rate for a data user [kbps]

Figure 3.6. Maximum bit rate for one new best effort user entering the system versus number of voice users. Result from Example 3.4.

Chapter 4

Simulation Model

To evaluate the system performance and validate the theoretical assessments, sim-ulations are performed. This chapter describes the simulation models and assump-tions in more detail.

4.1

Propagation Model

The propagation model characterizes the channel qualities by the attenuation of transmitted signals. The attenuation is the inverse of the path gain. The path gain, G, is thus the ratio of the received power to the transmitted power and it consists of four different parts,

G = GaGdGsGm< 1.

Here, Ga is the antenna gain, Gd is the distance attenuation, Gs is the shadow

fading and Gm is the multipath fading. The antenna gains and distance

atten-uations are given as lookup tables. The lookup table for distance attenuation is calculated using to the Okumura-Hata model. The fading models are described more in the following sections.

4.1.1

Shadow Fading

A mobile moving through an environment will be shadowed by major obstacles such as hills or buildings, which will cause fluctuations in the received signal. This phenomenon is called shadow fading. Shadow fading is mostly quite slow, and hence also called slow fading.

The shadow fading in logarithmic scale is modelled in the simulator by a normal distributed variable, with mean µ = 0 and standard deviation σ = 8. The shadow fading is also assumed to be correlated with decorrelation distance 100 meters.

26 Simulation Model

4.1.2

Multipath Fading

Radio waves travel different paths from transmitter to receiver, and some have been reflected several times during their way. Each reflection causes a phase shift, and the difference in length of path way also leads to a phase difference at the receiver. This is known as multipath fading. It is also called fast fading, due to rapid variations of the received signal power.

The multipath fading is modelled in the simulator with a standardized model called 3GPP Typical Urban.

4.2

Simulation Scenarios

The simulation scenarios should be chosen such that the simulations reflect a realistic system. On the other hand, a perfect realistic model is impossible to implement. Also, simplifications has to be done not to render too complicated calculations. The simulation scenarios are chosen to be a compromise between these demands.

4.2.1

Traffic Model

The simulated traffic consists of both speech and best effort data. Both user types are based on Poisson processes for arrival. The speech model includes a voice activity process, where the user is active 60% of the time. The speech calls are modelled with an exponential talk time with an average of 90 seconds.

The best effort traffic consists of a combined MMS and e-mail traffic model, where MMS occurs with 60% probability and e-mail with 40% probability. The packets are of random size with mean 12.7 kB for MMS and 60 kB for e-mail. The simulator also models the TCP flow control for packet based data.

4.2.2

Cell Deployment

The simulator models a two-dimensional environment and maintains positions for base stations and users. The simulation environment consists of seven sites, each with a three sector antenna. This yields 21 cells, forming a uniform hexagonal pattern as shown in Figure 4.1. To avoid border effects the plan is repeated through a wrap-around technique, so that the environment forms in fact an infinite hexagonal grid. The cell radius is set to 500 meters.

4.2.3

User Placement

Users are initially placed randomly throughout the simulated area according to a uniform distribution. They move with a Rayleigh distributed absolute velocity.

4.3 System Model 27

Figure 4.1. Cell deployment.

4.3

System Model

This section briefly describes the most significant assumptions and operations used to model the system.

4.3.1

Fast HARQ

Retransmissions are modelled with a stop-and-wait protocol. This means that no retransmission attempt is done in the same process before a negative acknowl-edgement is received. For E-DCH with 2 ms TTI, the simulator uses five parallel queues. The number of parallel queues are set such that the time to go through all the queues are approximately the same as the HARQ round trip time, so when the first queue is handled again, the Acknowledgement (ACK) or Negative Acknowl-edgement (NACK) has been received by the UE. Also, the time to go through the queues and the round trip time being as close together as possible. The principal operation of the HARQ processes is shown in Figure 4.2.

The outer loop power control model allows four transmission attempts, mean-ing that the probability that the transmission is still incorrect after the fourth transmission is 1%. If the number of retransmissions exceeds five, a Radio Link Control (RLC) retransmission is triggered, which will perform a retransmission request from the RNC.

When errors occur, all blocks transmitted in the current TTI will be retrans-mitted. The successive attempts are soft combined, using a model for chase com-bining, where each retransmission is an exact copy of the original transmission. Chase combining is modelled by adding the SIR for each transmission attempt. The transport format and resource combination is not changed if retransmission is required.

28 Simulation Model processing processing processing processing processing processing Node B processing processing processing

NACK ACK NACK ACK

processing

ACK ACK

UE 1 2 3 4 5 1 2 3 4 5

Known (fixed) timing relation 2 ms TTI 1 2 3 4 5 1 2 3 4 5

Figure 4.2. HARQ processes.

4.3.2

Power Control

Power control consists of an inner and an outer loop. The inner loop compares the received SIR with a target SIR. The transmission power is increased by 1 dB if the received SIR is below the target, and decreased by 1 dB if the received SIR is above the target.

The outer loop controls the target SIR. The target SIR is increased by a fixed step for each erroneously received block, and decreased by a smaller amount for every correctly received block. The ratio between the increase and decrease amounts is calculated based on the block error rate (BLER) target, such that the actual BLER, in a stationary situation, will converge to the BLER target.

4.3.3

Admission Control

Admission control makes sure that the system is not overloaded, by not admitting users that will cause the overload. When a user requests a channel, the noise rise that will be caused by the user is estimated. The estimate is added to the current system noise rise. If this noise rise estimation exceeds the maximum allowed noise rise of 7 dB, the user will be blocked and not admitted to the system. For voice users this means that they will not get a channel. For best effort users it means that they will actually be admitted to the system, but not allowed to start the transmission. Hence a best effort user will actually experience a very low bit rate because of the waiting time rather than being blocked.

The blocking of voice users will only be used in the voice only simulations with no best effort data users in the system. To model the prioritization of voice users, they are always admitted when letting simultaneous best effort users into the system. If done correctly, the admission control should look only at the non best effort load when admitting voice users. This might seem a bit simplified, however it will be shown that the capacity is in fact limited by the best effort users. But perhaps capacity could be increased further by rejecting voice calls.

4.4 Simulation Logging 29

4.3.4

Fast Rate Control

Node B rate control with a busy indicator is modelled using initial working as-sumptions for rate control not conformant with current agreement in 3GPP in all details. The node B controls the maximum bit rate, Rmax, i.e. the maximum

number of blocks that a EDCH user is allowed to transmit in each TTI. The busy indicator is set when the noise rise exceeds a certain value and unset when it falls below the same value.

If the busy flag is set, a new user with data in the transmit buffer sends a rate request to the Node B. If the busy flag is set the user will not be admitted, but users that already have a radio link will continue the transmission with bit rate ≤ Rmax. However, if the busy flag is set when the rate request is received

by the Node B, the maximum bit rate will be decreased and hence the users that are already transmitting are forced to lower their bit rates. Now the new user can be admitted and allowed to transmit with bit rate lower than the new Rmax.

Example 4.1 and Figure 4.3 shows how the rate control with busy indicator works. Example 4.1: Rate control with busy indicator

This example describes how the rate control with busy indicator works. The description refers to Figure 4.3.

1. U E2 starts transmitting and the busy indicator is set.

2. U E2 stops transmitting and the busy indicator is unset. U E1 starts

trans-mitting and the busy indicator is set again.

3. U E2wants to start transmitting but can not be admitted immediately since

the busy indicator is set. U E2sends a rate request.

4. Rmaxis lowered and hence also U E1bit rate is lowered. The busy indicator

is temporarily unset and U E2is admitted and starts transmitting.

5. U E1 stops transmitting and the busy indicator is again unset. Rmax is

increased and hence also U E2bit rate is increased.

4.4

Simulation Logging

The simulated time is set to 200 seconds, and the logging starts after 20 seconds when the traffic is assumed stable. Information on the total system is being logged every 2 ms. However, information about each user, such as number of transmitted blocks, is logged only for the total simulated time and not each time instant, due to memory limitations. Each scenario is simulated three times with different seeds to get more accurate values.

30 Simulation Model R_max Busy Indicator UE₁ REQ UE₂ SET UNSET 1 2 3 4 5

Chapter 5

Simulations and Results

The following chapter presents the simulations and results. The first section defines some performance measures, the second section describes the evaluation methods and the last sections presents the simulation results.

5.1

Performance Measures

Since the purpose of the master thesis is to demonstrate the trade-off between real time and best effort performances, a great issue is to decide what to evaluate and how to evaluate the performances. This section presents the measures that are calculated and studied from simulations. The next section describes how to use these measures. Some of these measures are already known from the theoretical assessments.

Throughput

Throughput is the data rate in the whole cell. It is calculated by dividing the total number of transmitted bits by the elapsed time and number of cells. As received bits we only consider blocks that have been delivered.

throughput = transmitted bits

elapsed time · number of cells

Normalized User Bit Rate

Normalized user bit rate is the data rate experienced by the user. It is calculated by dividing the size (bits) of the message the user has transmitted by the time elapsed from the message was generated until it was completely received.

normalized user bit rate = message size

transmission time

32 Simulations and Results

Normalized User Delay

Normalized user delay is the time it takes for the user to get a certain number of bits. It is calculated by dividing the time elapsed from the message was generated until it was completely received by the size (bits) of the message the user has transmitted. Normalized user delay is the inverse of normalized user bit rate. When calculating normalized user bit rate and normalized user delay, only users who have started and finished the transmission within the simulated time are considered.

normalized user delay = transmission time

message size =

1

normalized user bit rate

Block Error Rate

Block error rate (BLER) describes the fraction of data block errors transmitted by the user. BLER is calculated by dividing the number of blocks that are not transmitted correctly by the total number of transmitted blocks. Block error rates for voice users are calculated from users who have been given a channel and started the transmission within the simulated time.

BLER = N umber of f ailed blocks

T otal number of blocks

Resource Efficiency

Resource efficiency describes the utilization of the available resource. This is the same quantity as treated theoretically in Section 3.2. It is measured as the ratio of the noise rise mean value to the maximum noise rise limit at the capacity limit.

ρ = η

ηmax

Throughput per Noise Rise

Throughput per noise rise is the ratio of throughput to the noise rise mean value. This, again is the same quantity as treated theoretically in Section 3.4 earlier. This is a way of measuring the efficiency of the caused noise rise as well as a validation of the resource efficiency measure.

throughput per noise rise =throughput

η

5.2

Evaluation Criteria

To get the interesting information out of the simulations and the above measures, one needs some criteria to evaluate the system performance. The criteria that are applied in the sequel to evaluate the simulation results are presented below.