Voice-over-IP over Enhanced Uplink

(1)

Institutionen för systemteknik

Department of Electrical Engineering

Examensarbete

Voice-over-IP Capacity over Enhanced Uplink

Examensarbete utfört i Reglerteknik vid Tekniska högskolan i Linköping

av

Nils Brännström

LITH-ISY-EX-07/3941-SE

Linköping 2007

Department of Electrical Engineering Linköpings tekniska högskola

Linköpings universitet Linköpings universitet

(2)

(3)

Voice-over-IP Capacity over Enhanced Uplink

Examensarbete utfört i Reglerteknik

vid Tekniska högskolan i Linköping

av

Nils Brännström

LITH-ISY-EX-07/3941-SE

Handledare: Daniel Axehill

isy, Linköpings universitet

Erik Geĳer Lundin

Ericsson

Examinator: Fredrik Gunnarson

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 2007-03-02 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 http://www.ep.liu.se/2007/3941 ISBN — ISRN LITH-ISY-EX-07/3941-SE

Serietitel och serienummer

Title of series, numbering

ISSN

—

Titel

Title

Kapacitet för IP-telefoni i den förbättrade WCDMA-upplänken Voice-over-IP Capacity over Enhanced Uplink

Författare

Author

Nils Brännström

Sammanfattning

Abstract

The traditional voice service in mobile networks is an important service that mobile users expect high quality from. With the convergence of mobile networks towards an all-IP network, an IP-based speech service becomes important which is referred to as Voice-over-IP (VoIP). The traditional voice service is highly optimized and a VoIP service must therefore fulfil strict quality requirements to provide the same speech service quality. The air interface technology, WCDMA, which is used in third generation communication systems in Europe is constantly developed. An improved concept for the mobile-to-network trans-mission, called the Enhanced Uplink (EUL) provides for higher uplink capacity for packet data services. It also includes features that may provide a sufficient VoIP service quality in mobile networks, when considering the uplink transmission. The purpose of this thesis is to evaluate the VoIP capacity over EUL and iden-tify crucial aspects of radio resource management in order to increase the capacity. This is done through dynamic system simulations, using a realistic VoIP traf-fic model. The VoIP capacity is also estimated by a derived theoretical framework. It is shown by simulation results and theoretical estimations, that power control is a vital mechanism in order to increase the capacity. Simulation results indicate that a VoIP over EUL capacity of 65% of the traditional voice service capacity may be reached. The results also indicate that to improve the capacity for larger cells, the allowed VoIP packet delay must be increased.

Nyckelord

(6)

(7)

Abstract

The traditional voice service in mobile networks is an important service that mobile users expect high quality from. With the convergence of mobile networks towards an all-IP network, an IP-based speech service becomes important which is referred to as Voice-over-IP (VoIP). The traditional voice service is highly optimized and a VoIP service must therefore fulﬁl strict quality requirements to provide the same speech service quality. The air interface technology, WCDMA, which is used in third generation communication systems in Europe is constantly developed. An improved concept for the mobile-to-network transmission, called the Enhanced Uplink (EUL) provides for higher uplink capacity for packet data services. It also includes features that may provide a suﬃcient VoIP service quality in mobile net-works, when considering the uplink transmission.

The purpose of this thesis is to evaluate the VoIP capacity over EUL and identify crucial aspects of radio resource management in order to increase the ca-pacity. This is done through dynamic system simulations, using a realistic VoIP traﬃc model. The VoIP capacity is also estimated by a derived theoretical frame-work.

It is shown by simulation results and theoretical estimations, that power control is a vital mechanism in order to increase the capacity. Simulation results indicate that a VoIP over EUL capacity of 65% of the traditional voice service capacity may be reached. The results also indicate that to improve the capacity for larger cells, the allowed VoIP packet delay must be increased.

(8)

(9)

Acknowledgements

First of all I would like to express my gratefulness and a big thank you to all the people at the Radio Resource Management department at Ericsson, where I’ve carried out this thesis project. You’ve all made me feel very welcome from day one and supportive with your open minds and open doors where a lot of great competence and experience reside.

I would especially like to thank Erik Geĳer Lundin, who has been my supervi-sor at Ericsson and been very supportive and a light of guidance at all times with everything from the simulator to LA_{TEX. I would also like to thank Raimundas}

Gaigalas at the department, for your support with the VoIP traﬃc model and for your honest opinions about my work. Last but not least, thank you my dear master thesis colleague at Ericsson, Victor Laborda, for great company and many deep discussions about everything over countless cups of coﬀee.

I would also like to thank my supervisor at Linköpings universitet, Daniel Ax-ehill, and my examiner, Fredrik Gunnarsson, for your feed-back and ideas on my work.

To my invaluable family and dear friends I would ﬁnally like to say that I’m very grateful for your support and for being who you are. I treasure every minute with you and the smiles and laughters we share together. You mean the world to me.

Nils Brännström

(10)

(11)

Abbreviations

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

ACK Acknowledged AMR Adaptive Multi-Rate BER Bit Error Rate BLER Block Error Rate

CIR Carrier-to-Interference Ratio CS Circuit-Switched

CQI Channel Quality Information

DL DownLink

DPCCH Dedicated Physical Control Channel DPDCH Dedicated Physical Data Channel EUL Enhanced UpLink

GSM Global System for Mobile communications HARQ Hybrid Automatic Repeat Request HSDPA High Speed Downlink Packet Access IMS IP Multimedia Subsystem

kbps Kilobit per second Mbps Megabit per second Mcps Megachips per second MHz Megahertz

NACK Not Acknowledged

OLPC Outer Loop Power Control PS Packet-Switched

QoS Quality of Service RLC Radio Link Control RNC Radio Network Controller ROHC RObust Header Compression RTT Round-Trip Time

SF Spreading Factor SID SIlence Descriptor

(12)

x

SIR Signal-to-Interference Ratio TA Transmission Attempt

TFC Transport Format Combination TTE Transmission Target Error TTI Transmission Time Interval UE User Equipment

UL UpLink

UMTS Universal Mobile Telecommunications System UTRAN UMTS Terrestrial Radio Access Network VoIP Voice-over-IP

(13)

3.3.3 EUL conﬁguration . . . 33 3.3.4 Hybrid ARQ . . . 34 3.3.5 Power Control . . . 34 3.3.6 HS-DPCCH model . . . 35 3.3.7 Admission Control . . . 35 3.3.8 System Logging . . . 36 3.4 Capacity Requirements . . . 36 4 Theoretical Assessments 39 4.1 Deﬁnitions . . . 39 4.1.1 Propagation . . . 40 4.1.2 Uplink Load . . . 40

4.2 VoIP over EUL Adaptation . . . 42

4.2.1 Adaptation to VoIP Requirements . . . 42

4.2.2 Channel Activities . . . 45

4.3 Theoretical VoIP Capacity . . . 47

4.3.1 F-factor . . . 47

4.3.2 Orthogonality Factor . . . 48

4.3.3 Initial SIR targets . . . 48

4.3.4 Coverage . . . 51

4.3.5 Uplink DPCCH Gating . . . 53

5 Simulation Results 57 5.1 Dynamic Simulation Results . . . 57

5.1.1 Default Settings . . . 57

5.1.2 Adjusted initial SIR target . . . 58

5.1.3 Adjusted OLPC Target . . . 59

5.1.4 Increased Cell Radius . . . 61

5.2 Results Comparison . . . 63

6 Conclusions 65

(15)

A Simulation Parameters 71

(16)

(17)

Chapter 1

Introduction

Today’s state-of-the-art telecommunication technology, can provide mobile users with a range of services. The ordinary voice service is still the most important service that mobile users rely on and expect high quality from. With the advent of mobile broadband, an alternative to the traditional voice service may be pro-vided, which is called Voice-over-IP. This introductory chapter will describe the background and the interest in Voice-over-IP over Enhanced Uplink in order to formulate the problem statement of this thesis project.

1.1 Background

Communication systems are made out of a large number of network nodes that provide connections between users or access to different communication networks. Traditionally, voice communication is done in the Circuit-Switched (CS) domain, where dedicated links between users are created throughout the network. A CS speech service sets up a dedicated link between two users and the link is reserved as long as the voice session lasts, whether or not the users speak to each other. These direct and reserved links are not very efficient in terms of network performance, especially for a voice service or other types of bursty data traffic. Packet-Switched (PS) communication however, splits the data to be transmitted into packets and these are then routed individually between the network nodes. This increases link efficiency, when links between nodes may be shared by packets from other users and idle users do not preserve any communication links.

The Internet is a network that transmits data in the PS domain, using the Internet Protocol (IP) to route traffic to its specific destination. This protocol may also be used to route voice data, which is referred to as Voice-over-IP (VoIP). VoIP has become a strong competitor to the traditional CS voice service in the fixed networks, with PC applications such as Skype and broadband providers who offer VoIP services together with the Internet connection.

With the success of VoIP in the ﬁxed networks, the interest in providing VoIP 1

(18)

2 Introduction

in the mobile networks is large for many reasons. The development of mobile net-works is heading in a direction towards an all-IP architecture, where a PS voice ser-vice (VoIP) becomes important. The traditional voice serser-vice is highly optimized in the CS domain which users have been accustomed to, thus a VoIP service needs to fulfil the same quality requirements and capacity. The main benefits for a single packet-switched network are that operators can reduce their operational costs and increase revenues when providing new flexible IP-services. A VoIP service also enables end-user flexibility, when users may add other multimedia services to an ongoing voice session [17].

In Europe the standard for 3G systems is called Universal Mobile Telecom-munications System (UMTS). The radio technology in UMTS is Wideband Code

Division Multiple Access (WCDMA), which is speciﬁed by the 3rd Generation

Partnership Project (3GPP). The WCDMA technology is continuously develop-ing to support high capacity mobile broadband. The ﬁrst big step was taken by introducing High Speed Downlink Packet Access (HSDPA) in the standardized WCDMA 3GPP Release 5, providing higher bitrates in the downlink transmission from the network to the mobile.

Later, the complement in the uplink, mobile-to-network transmission, was de-veloped. This concept is called the Enhanced Uplink (EUL), which is speciﬁed in Release 6. EUL supports higher uplink bitrates and features that look encouraging for providing a suﬃcient VoIP service in WCDMA networks in terms of quality and capacity.

1.2 Problem Statement

With the expected interest in this service there is a need for evaluating the VoIP capacity over EUL. With the quality requirements given, it is also vital to inves-tigate and characterize the use of radio resources for a VoIP service over EUL.

Given the performance requirements, the aim of this thesis project is to evaluate the VoIP capacity over EUL in terms of maximum number of VoIP users that may be supported by the system simultaneously. Moreover, the goal is to characterize crucial aspects of radio resource management, aiming at optimizing the VoIP over EUL capacity.

1.3 Approach

The research approach in this thesis is divided into two parts. A theoretical esti-mation of the capacity will be derived and simulations will be set up and run in a dynamic radio network simulator. The results will be compared to the capacity of a CS voice service and expressed as a fraction of CS speech capacity in the main results.

(19)

1.4 Related Work 3

The theoretical framework will not only generate a more profound understand-ing of VoIP over EUL, but also give a hint of what will be expected from the dynamic simulation results. More importantly, it will generate an idea of the pos-sible capacity gains for diﬀerent settings, although a theoretical estimation has clear limitations.

Dynamic simulations of VoIP over EUL will be run, using a radio network sim-ulator developed by Ericsson with a realistic traffic model of VoIP. The simulations will be run for different conditions and parameter settings, aiming at optimizing the capacity and characterizing the impacts of different conditions.

1.4 Related Work

There has been quite a lot of research done in the past, considering a VoIP ser-vice in WCDMA systems. The majority of the reports evaluate the downlink VoIP scenario using HSDPA and only a few consider VoIP over EUL. In general, previous studies of VoIP over EUL are done with various assumptions and param-eter settings that not always reﬂect the realistic scenario. The focus is mainly on

technology potential. In comparison, these studies do not consider the impact on

control signalling done in the uplink to support HSDPA. It is also common that an essential part of the power control mechanism (the outer loop) is not activated. Moreover the VoIP traﬃc models do not consider the data that is transmitted when a user is silent, due to simplicity reasons.

A good introductory report to VoIP over WCDMA can be found in [17]. The interest and benefits of a VoIP service over WCDMA are discussed and the key features of EUL and HSDPA that may be used to reach CS speech capacity. It is also shown, under certain simulation conditions, that VoIP downlink capacity using HSDPA exceeds CS capacity. Different scheduling algorithms, that decides when and which users that are served in the downlink, are compared and evalu-ated. It also gives a good introduction to the characteristics of VoIP traffic.

A study of VoIP over EUL can be found in [18]. This study consists of a an-alytical part as well as simulation results. The anan-alytical framework is used to predict the user created interference from VoIP users and this is compared to the interference created for a CS speech service. The analytical framework is in some aspects similar to the framework in this report, but in general simpliﬁed. The sim-ulation conditions are very diﬀerent from this study and not directly comparable. It is however shown that VoIP over EUL may provide the same capacity as CS speech whereas for some conditions the capacity gain is around 9% compared to CS speech capacity.

Another study in [14], includes studies of VoIP over HSDPA and EUL. It takes into account the improvements and features speciﬁed in the 3GPP Release 7. Again, the conditions are not comparable when the assumptions are based

(20)

4 Introduction

on [18]. Instead it gives a hint of the relative gains with further improvements and the evolved WCDMA technology. 3GPP Release 7 includes an eﬀective way of reducing the uplink control signalling overhead. It is shown that these improve-ments may generate a capacity gain of around 50 % compared to VoIP over EUL in previous releases. The downlink capacity is shown to exceed CS capacity with a VoIP optimized scheduler.

1.5 Chapter Outline

Chapter 2 gives a brief overview of cellular systems in general, followed by a more detailed description of UMTS, WCDMA and EUL. Chapter 3 introduces VoIP followed by a more detailed description of the system model and the evalu-ation requirements. The theoretical framework and estimevalu-ation results are found in Chapter 4. The results from the dynamic simulations are found in Chapter 5, followed by overall conclusions in Chapter 6.

(21)

Chapter 2

Cellular Radio Systems

Following the success of the second-generation (2G) mobile communication sys-tems, for example GSM, new demands soon arose beyond ordinary services as voice calls and text messages. With the growing number of wireless users and new ser-vice and capacity demands, a new third generation (3G) technology was developed. The new technology aims at providing higher capacity allowing for new services as video telephony, streaming video and web browsing. In Europe the standard for 3G systems is called Universal Mobile Telecommunications System (UMTS) and is adopted by the standardization organization European Telecommunications Stan-dards Institute (ETSI). To transfer information over the air, UMTS systems uti-lize Wideband Code Division Multiple Access (WCDMA) as its air interface. This chapter will introduce the concepts of cellular systems and the radio technology WCDMA. It will also briefly describe the evolved concept of uplink transmission in 3G-systems, referred to as the Enhanced Uplink. More information about wireless communication and WCDMA can be found in [5], [13] and [16].

2.1 Propagation

Radio communication systems are characterized by the transmission of informa-tion over the air. Using the air as the transmission channel, implies that the channel is highly varying. The transmitter or the receiver will often have a certain mobility at the same time as objects in the environment are non-stationary. This results in varying and unpredictable radio conditions.

Wireless signals are transmitted by means of electromagnetic waves. When electromagnetic waves propagate in air they will be exposed to reflection, diffrac-tion and scattering, resulting in an attenuadiffrac-tion of the the signal. To model these phenomena by statistical means, one can divide the total path gain into a product of elements that will be described below. The total path gain g, which is the ratio between the received power, Prx, and the transmitted power, Ptxcan be expressed

as,

(22)

6 Cellular Radio Systems

Prx

Ptx

= g = gpgsgfga<1 (2.1)

where gp is the distance path gain, gs is the shadow fading, gf is the fast fading

and ga is the antenna gain. The total path gain is here expressed in linear scale.

2.1.1 Distance Path Gain

The distance path gain, which describes the attenuation due to the distance be-tween the transmitter and the receiver, is often modelled as,

gp=

C

Rα (2.2)

where R is the distance in meters, C and α are constants. A typical value for free-space propagation is α = 2. Larger values, up to α = 5 are used to model more urban environments [5]. Figure 2.1 shows the distance path gain as a function of distance in decibels. Here C = 0.0013 and α = 3.52, which are the values used throughout this report. The two constants are derived from the Okumura-Hata propagation model and the values depend on the base station height and the mobile terminal height above ground level, the area type, e.g., urban or rural environments and the carrier frequency.

0 500 1000 1500 2000 −120 −110 −100 −90 −80 −70 −60 −50 −40 −30 −20 Distance [m]

Distance path gain [dB]

(23)

2.2 Multiple Access 7

2.1.2 Shadowing Fading

Shadowing fading intends to describe the attenuation due to large objects blocking the line-of-sight between the transmitter and the receiver, e.g., hills and buildings. These variations are often modelled with a log-normal distribution with a mean value of µdB= 0 and a standard deviation of σdB in the range 5-12 dB [19].

Varia-tions due to shadowing are slow and are also spatially correlated. A decorrelation distance is used to describe the distance at which the correlation has decreased with a factor 1

ǫ. Typical values are in the range of 50-100 meters.

2.1.3 Fast Fading

Fast fading is also called multipath fading and describes the fact that the radio signal may propagate via multiple paths of different lengths. This will cause the multiple components to reach the receiver with varying phases. This in turn may generate constructive or destructive interference, resulting in rapid variations in received signal power. Multipath fading also means that the signal energy will arrive dispersed in time at the receiver, characterized by a multipath delay profile. It is called fast fading because a slight movement of the receiver, in order of a few wavelengths, can cause a big difference in phase for two signal components that travelled different paths. These fast variations in the received signal energy is commonly modelled by a Rayleigh distribution [5].

2.1.4 Antenna Gain

The antenna gain is the gain introduced by a directional antenna in relation to an an isotropic antenna. An ideal isotropic antenna radiates equal power in all directions and a directional antenna may concentrate its equal total power in a certain direction. The antenna gain is commonly expressed with reference in decibels to an isotropic antenna where the unit is named dBi.

2.2 Multiple Access

Wireless communication means that signals are transmitted in the air through modulation of a certain carrier frequency that is shared by many users. There are different schemes that allow for this resource to be shared, so called multiple access schemes. This section will briefly describe the different techniques with the focus on Code Division Multiple Access (CDMA), which is used in 3G systems.

2.2.1 Frequency Division Multiple Access

Frequency Division Multiple Access (FDMA) is a scheme where, as the name im-plies, the available frequency spectrum is divided into narrow frequency bands (channels) with a certain gap or guard band in between. Users are then assigned a

(24)

communication channel per user and transmit on respective frequency throughout time. The number of channels, i.e., users that are able to communicate simulta-neously, are obviously limited and therefore the system capacity is limited. This is regarded as a hard capacity system.

2.2.2 Time Division Multiple Access

Time Division Multiple Access (TDMA) divides a frequency channel into a number of short time slots in which users are allowed to transmit within their given slot. A user may be assigned one or many time slots and several frequency channels can be used in the system to further increase the capacity. In TDMA, the introduction of several frequency channels must be carefully planned geographically so that the transmissions on the same frequency do not interfere with each other. If the distance between the locations where the same frequency channel is used is large enough, the same frequency may be used in the system. This is also known as

frequency reuse. TDMA increases system capacity compared to FDMA, but it is

still a hard capacity system.

2.2.3 Code Division Multiple Access

Code Division Multiple Access (CDMA) is a multiple access scheme where users and their channels are separated by codes. This allows users to transmit on the same frequency at the very same time instant. This scheme allows more eﬃcient use of the frequency resource. Transmitting on the same frequency at the same time results in that users create interference that other users experience as noise. As opposed to TDMA and FDMA systems, a CDMA system is a soft capacity system when the system is limited by noise rather than lack of time or frequency resources.

Code division make use of codes that are called spreading codes. The codes spread the user signals by multiplying the user data with a number of bits, called chips. A code consists of a sequence of n number of chips, which is known both by the transmitter and the receiver. The spread data is now at a rate of nR, where R is the original bitrate of the user signal. In the frequency domain, the bandwidth of the original signal is eﬀectively spread out n times, hence CDMA systems are referred to as spread spectrum systems. The number n is also called the spreadingfactor.

The spreading codes are arranged so that they are perfectly orthogonal to each other and are designed using a code tree as illustrated in Figure 2.2. These codes are called Orthogonal Variables Spreading Factors (OVSF). To ensure complete orthogonality, once a code is in use none of the codes from the subtree under that code can be used.

At the receiver side the known code is multiplied and integrated with the spread data and if they are perfectly synchronized the original data bits will be retrieved

(25)

2.2 Multiple Access 9 SF = 1 SF = 2 SF = 4 c1 = {1} c2,1 = {1 1} c2,2 = {1 -1} c4,1 = {1 1 1 1} c4,2 = {1 1 -1 -1} c4,3 = {1 -1 1 -1} c4,4 = {1 -1 -1 1} c c -c c c SF = 64 SF = 32 SF = 16 SF = 8 SF = 128 SF = 256 SF = 1 SF = 2 SF = 4 c1 = {1} c2,1 = {1 1} c2,2 = {1 -1} c4,1 = {1 1 1 1} c4,2 = {1 1 -1 -1} c4,3 = {1 -1 1 -1} c4,4 = {1 -1 -1 1} c c -c c c SF = 64 SF = 32 SF = 16 SF = 8 SF = 128 SF = 256

Figure 2.2. Code tree that generates orthogonal code of varying length, hence varying spreading factors (SF).

as shown in Figure 2.3. Attempting to despread the data with another orthogonal spreading code will be interpreted as noise or interference. We can also see that the integration process increases the amplitude by the spreading factor and this is also known as the processing gain which makes the system robust against interference and that signals may be detected and despreaded even if the signal quality is low. This is also a necessity when all present channels will be spread out and transmitted on the same carrier frequency.

Figure 2.3. Principle of spreading and despreading data. Here the SF or the code length is equal to 4.

CDMA enables variable bitrates for users and is spectral eﬃcient in the sense that many users can be served simultaneously using a limited frequency bandwidth.

(26)

An important factor is the power level in the system or the presence of other users’ signals which is experienced as noise. A simpliﬁed illustration is shown in Figure 2.4, where the power spectral density is plotted. The overall power spectral density is decreased when spreading the signal over the carrier bandwidth. As long as the power level of other present signals do not add up to become larger than the own signal, it can be retrieved from despreading.

f

P

f

P

1

P

f

3 2

f

P

f

4 5

f

P

f

P

f

P

1

P

f

3 2

f

P

f

4 5

Figure 2.4. Spreading and despreading in the frequency domain. 1. Spreading of the original signal 2. Transmission 3. Shaded field represent other received wideband and narrowband interference 4. Despreading 5. Filtering.

2.3 Cellular Systems

2.3.1 Cellular Network Structure

A cellular network aims to provide network access for many mobile users within a geographical area. This area is further divided into several subareas each served by a particular base station, which is the node that transmits and receives all radio signals from the mobile users in the subarea. The subarea that a base station serves is commonly known as a cell. In this case the base station makes use of an omnidirectional antenna that radiates in all directions. A base station might also use directional antennas which further divides the subarea that the base station serves, into smaller cells or sectors. The location of the base station is called a site. Figure 2.5(a) illustrates the principle with a directional antenna serving several sectors or cells. In Figure 2.5(b), 3 cells are illustrated where each cell is served by its base station.

2.3.2 Handover

Users that are connected to a certain base station will sometimes move into a physical area that is served by another base station. In this case the user will

(27)

2.3 Cellular Systems 11

(a) 1 site with directional antenna serving 3 sectors (cells)

(b) 3 sites, each with omnidirec-tional antennas, serving 1 cell each

Figure 2.5. Illustration of different base station and cell structures

be handed over and establish connection to another base station. In TDMA and FDMA systems, this is refereed to as hard handover since the connection must be dropped before a new radio link can be setup towards the new serving base station. In CDMA systems however, users are transmitting on the same frequency simultaneously in the system, which makes soft handover possible. During soft handover the user is connected to both base stations at the same time and a new connection can be established before the ﬁrst one is dropped. This reduces the risk of, e.g., dropped calls and enables diversity in terms of utilizing the air interface eﬃciently. When a user is handed over to another sector (cell) that is served by the same base station, it is known as softer handover.

2.3.3 Uplink and Downlink

In a cellular system, the transmission of radio signals from the user to the base station is called uplink transmission. The other communication path from the base station to the users, is called the downlink. To be able to communicate in both directions a TDMA system for example, separates uplink and downlink transmission by means of diﬀerent frequency channels. This is known as Frequency Division Duplex (FDD), which is also commonly used for CDMA systems. The uplink and downlink traﬃc is hence separated using two carrier frequencies. All uplink transmission is thus done using the same carrier frequency in the system and another carrier frequency is used for all downlink transmission.

2.3.4 Power Control

Power control is an important mechanism in all cellular systems. For any trans-mission on a certain frequency, the transmitted power level needs to be sufficient to provide enough signal quality at the receiver. For different kinds of services there will be different types of requirements on the received signal quality. The received quality depends on the transmission power and the total path gain between the user and the base station. The radio environment is rapidly varying due to fading

(28)

characteristics and thus the transmission power of the user equipment needs to be regulated. To describe the signal quality received at a certain frequency, the term

Signal-to-Interference(SIR) is often used. SIR describes the ratio of the wanted

signal power to the total interfering power. The interfering power may originate from other unwanted signals or from background noise.

In a CDMA system, there will be many users transmitting on the same carrier frequency at the same time in the uplink transmission to the base station. These terminals may of course be transmitting very near the base station or far away from it. Hence a terminal transmitting with too high power or at a close distance would simply make other transmitted user signals further away undetectable. On the other hand a terminal transmitting with too little power or far away might never be distinguishable for the base station. This is often referred to as the

near-far problem and implies a need for a fast and eﬀective power control in CDMA

systems, when channel conditions are changing rapidly due to fading.

The overall goal is to keep all received signal powers in the base station at the

same minimum level, enough to provide the required SIR. This minimizes user

created interference and saves battery consumption of the terminals.

2.4 UMTS

This section will describe the UMTS network in more details and the evolved concepts of uplink transmission in WCDMA, namely EUL.

2.4.1 WCDMA

The principles of WCDMA is characterized by the CDMA scheme described in previous sections. WCDMA systems use a chip rate of 3.84 Mcps, which corre-sponds to a bandwidth of 5 MHz compared to other CDMA systems, commonly using a carrier bandwidth of 1.25 MHz. It is speciﬁed by the 3rd Generation Part-nership Project (3GPP) and is being continuously developed through joint project of international standardization bodies. The aim is to support higher bitrates, high spectrum eﬃciency and improved coverage, to meet the new demands for evolving high capacity multimedia services and internet access [13].

2.4.2 UMTS Network Architecture

UMTS networks are at a high-level system view made out of three subsystems, namely the core network, the UMTS Terrestrial Radio Access Network (UTRAN) and the User Equipment (UE). The core network routes traﬃc to external net-works, e.g., the public switched telephone network (PSTN) or the Internet. An overview of the architecture can be seen in Figure 2.6.

(29)

2.4 UMTS 13

UTRAN is the link between the UEs and the core network through the interface called Iu towards the core network and the Uu (radio) interface towards the UE. Within the UTRAN we ﬁnd a number of Node Bs (base stations) where each Node B in turn handles a number of cells (sectors) that are geographical areas. The Node Bs are the nodes that transmit and receive all radio signals. They are in turn controlled by a Radio Network Controller (RNC). One RNC may have several Node Bs connected to it and these two types of nodes are interconnected by the Iur and Iub interfaces.

Core Network Iu Iur Iub Uu Node B UE RNC Node B Node B RNC UTRAN Core Network Iu Iur Iub Uu Node B Node B UE RNC Node B Node B Node B Node B RNC UTRAN

Figure 2.6. UMTS network architecture.

2.4.3 UMTS QoS Classes

UMTS states four diﬀerent Quality of Service (QoS) classes that services might be divided into, mainly dependent on how sensitive the service is to delay.

• Conversational class, strict delay requirements, e.g., voice telephony

• Streaming class, moderate delay, e.g., streaming a video clip

• Interactive class, often characterized by response patterns, e.g., web browsing

• Background class, no requirements on delay, e.g., ﬁle download

When considering a service like VoIP, the aim is to provide a packet-switched speech service with the same QoS characteristics as a traditional speech service. Hence VoIP is assumed to be of conversational class with strict delay requirements to satisfy the human perception of a real-time conversation.

(30)

2.4.4 Admission Control

The resources of the UTRAN and especially the air interface, are important to manage in an eﬃcient way. This is done through Radio Resource Management (RRM), which is a set of algorithms that aim to maximize system capacity, cover-age and provide QoS for diﬀerent services during varying conditions [6]. Admission

control is one important RRM algorithm when using WCDMA as the air interface.

The principles of admission control is to make sure that the system does not get overloaded and that QoS can be provided after admitting new users into the system. In the downlink the admission control checks whether there are enough codes available for the requested radio links. There is also a downlink load check that estimates the total power used for the present radio links and checks if there is power available to serve the requested links. The limiting resource in the downlink is essentially the number of codes available and the power of the base station.

Considering the uplink, the limiting resource is the interference headroom and an uplink load check is performed before new requests are admitted. This is done through estimating the load impact from the number of present radio links. This is then compared to a speciﬁed uplink load limit in order to reject or admit the requests.

2.4.5 Uplink Power Control

Power control is another important RRM algorithm that is used both in the down-link and the updown-link. In the updown-link, the shared resource is the interference head-room. Using WCDMA, power control becomes crucial for a stabile system when the interference headroom must be managed. This section will further describe how the uplink power control is done in order to provide QoS and maintain the users’ signal qualities. A schematic illustration of the uplink power control can be seen in Figure 2.7.

The control is divided into the inner loop and outer loop. The inner loop is located between Node B and the UE and makes rapid estimation of the received Signal to Interference Ratio (SIR) and compares it with the present SIR target. This is done very frequently at a frequency of 1.5 kHz in order to ﬁght rapidly varying radio conditions known as fast fading. Depending on if the measured SIR is higher or lower than the target, Node B sends a Transmit Power Control (TPC) command to the UE to either raise or lower its transmission power.

The outer loop resides in the RNC, which provides a SIR target to the inner loop. The SIR target will be changing over time but not as frequently as the inner loop operates. The outer loop may be executed every Transmission Time Interval (TTI) which is the time frame in which a user may transmit over the air interface, e.g., every 10 ms. The Outer Loop Power Control (OLPC) algorithm is implementation-speciﬁc and aims at keeping the SIR-target at a level just enough

(31)

2.4 UMTS 15

to provide the required QoS. Regulating the inner loop at a too high SIR-target would create unnecessary interference to other users.

The operation is based on the measured quality of the uplink transmission. This quality is the compared to a specific quality target that the service is expected to meet. The basis for quality is often expressed in Block Error Rate (BLER). BLER is the ratio of erroneous received data blocks (data which is jointly encoded) to the total number of received data blocks. If the quality is better than wanted, the SIR target may consequently be decreased by a certain amount that depends on the BLER that is allowed. If a higher BLER is experienced than the allowed, the SIR target is raised to a higher level set by a certain step-up size. This is a type of jump-algorithm and is described in more details in [8]. When a new radio link is setup, there is an initial SIR-target configured and given to the inner loop before any data is sent. After the first transmission is done the SIR target will be changing in time and start to converge in order to provide the acceptable BLER. In the ideal situation where the radio environment is constant, the outer loop would converge and provide the acceptable BLER in a stationary steady-state. As the radio environment will be changing over time the SIR target that provides the wanted quality will be changing as well. Using a larger step-up size would allow for faster adjustment to large variations in the radio environment with the cost of larger oscillations around the critical SIR level. A smaller step-up size would keep the SIR target closer to the critical level over time but it would take longer time to adjust to variations in radio conditions. There is also an anti-windup mechanism to prevent that the difference between the SIR target and the effective SIR becomes too big. If the difference is more than the allowed, the outer loop is suspended. New SIR-target Quality target Send TPC Receive TPC adjust PUL Measure received SIR

Measure quality (BLER)

RNC

Node-B

UE

-+

=0?

-

=0?

+

Outer loop

Inner loop

New SIR-target Quality target Send TPC Receive TPC adjust PUL Measure received SIR

Measure quality (BLER)

RNC

Node-B

UE

-+

=0?

-

=0?

+

Outer loop

Inner loop

(32)

2.5 Enhanced Uplink

WCDMA is constantly being developed. In the early releases of WCDMA systems, various multimedia services were provided and a lot of effort has been made to further improve these services in terms of performance. Higher bitrates, reduced delays and improved coverage for packet data services, e.g., web browsing and other IP-based applications, improve the end-user experience. This is also benefi-cial for operators when offering a higher overall capacity and enriched services [11]. The first big step was taken by introducing High Speed Downlink Packet Access (HSDPA) in the WCDMA 3GPP Release 5, providing higher bitrates in the down-link. In Release 6 an improved concept of the uplink data transmission from the UE to the base station was introduced and is being developed. This concept is known as the Enhanced Uplink (EUL) and may theoretically support bitrates up to 5 Mbps in comparison to earlier releases providing 384 kbps.

2.5.1 EUL Channels

EUL introduces a new uplink transport channel, the Enhanced Dedicated Channel (E-DCH) which is dedicated to only one UE. This channel can co-exist with the existing DCH from earlier releases with its Dedicated Physical Control Channel (DPCCH). EUL introduces two uplink physical channels, the E-DPCCH and the E-DPDCH. The E-DPDCH carries the actual raw data whilst the E-DPCCH car-ries uplink control signaling.

The EUL physical channels are power controlled together with the DPCCH. The inner loop power control as described in Section 2.4.5 regulates the trans-mission power of the DPCCH. When data is transmitted the power level of the EUL physical channels are set according to two power offset values relative to the DPCCH power. The power offset to the physcial data channel, E-DPDCH, is of great importance and varies for different bitrates. The power level of E-DPDCH should be set for best overall performance taking into account the service require-ments, e.g., delay and the transmission channel. More power to the data channel reduces delay with the cost of increased interference and less power to the other physical channels when the total power of the UE is limited.

It is also important to note that the E-DPCCH is only transmitted when data is transmitted, whilst the DPCCH is always transmitted which carries control in-formation. Another uplink physical channel needed to support HSDPA (if present) is the HS-DPCCH which carries acknowledgment information (ACK/NACK) and Channel Quality Information (CQI) for the downlink traﬃc. The HS-DPCCH is also power controlled together with the DPCCH.

2.5.2 EUL Features

In order to improve the uplink performance a couple of new features are provided. The E-DCH may be conﬁgured to transmit with a short TTI, as short as 2 ms,

(33)

2.5 Enhanced Uplink 17

but it also supports to transmit on a 10 ms basis. A short TTI further reduces the delay and makes faster adaptation possible for mechanisms such as as power control. Fast scheduling is also an important factor, which controls when a user might transmit and at what maximum data rate. This can be done for every TTI and makes fast rate control possible.

A key feature is the fast hybrid automatic repeat request (HARQ) which reside in the Node B. This allows for a fast request for retransmissions if the decoding of the received data fails. This feature also allows for soft combining of the original transmitted data. Erroneously decoded data is not discarded and the retransmit-ted data may then be combined with the original transmission in order to make a new decoding attempt. One eﬀective combining scheme is Incremental

Redun-dancy(IR) where diﬀerent transmissions are coded diﬀerently, rather than simply

repeating the same coded bits. HARQ makes the system robust against sudden interference peaks and may improve the radio link eﬃciency [11].

The operation of the OLPC is based on the Transmission Target Error (TTE) rate after a number of targeted Transmission Attempts (TAs). These two param-eters are conﬁgured depending on the quality that is wanted for the service. The OLPC hence raises the SIR target when more TAs have been used than the tar-geted number of TAs. If decoding was successful using the tartar-geted number of TAs or less, the SIR target is decreased in a similar manner as described in Section 2.4.5. The gain of soft combining can be exploited when targeting multiple TAs which can be done through OLPC operation. When targeting multiple TAs, the trans-mission power will be adjusted to meet the target and the decoding may be suc-cessful after soft combining of multiple received transmissions at a lower SIR. This of course introduces delay, but on average not all transmission attempts will be used when channel conditions vary between the attempts. The HARQ scheme may be able to decode the data in advance, using fewer TAs than the targeted number. The typical trade-oﬀ here is between delay and coverage. Coverage may be improved by utilizing soft combining of multiple transmissions with the cost of increasing delay.

The concept of HARQ is illustrated in Figure 2.8. Data will be retransmit-ted until the UE receives an ACK (acknowledged) upon succesful decoding or the maximum number of transmission attempts is reached. If the maximum number of transmission attempts is reached, a retransmission from the Radio Link Control (RLC) layer will be done if it works in Acknowledged Mode (AM). In Unacknowl-edged Mode (UM) the data will simply be lost. Multiple HARQ processes are used in order to allow continuous transmission when there is a processing delay in the receiver before an ACK or NACK (not acknowledged) is returned. If the decoding failes for one of the HARQ processes the process will be occupied and a retransmission will be made in the next cycle. Thus retransmissions are done every NHARQ· T T I ms, where NHARQis the number of HARQ processes. If there

(34)

18 Cellular Radio Systems Receiver processing NAK Receiver processing ACK Receiver processing ACK Receiver processing ACK Receiver processing ACK Receiver processing NAK 0 1 Receiver processing 2 3 4 1 NAK Receiver processing NAK 2 3 4 1 2 TTI Receiver processing NAK Receiver processing ACK Receiver processing ACK Receiver processing ACK Receiver processing ACK Receiver processing NAK 0 1 Receiver processing 2 3 4 1 NAK Receiver processing NAK 2 3 4 1 2 TTI

Figure 2.8.Operation of multiple HARQ processes. This example uses 4 parallel HARQ processes.

(35)

Chapter 3

System Model and

Requirements

In order to evaluate the VoIP over EUL capacity, system simulations were done in this study, using a HSDPA and EUL capable radio network simulator. This MATLAB-based simulator developed by Ericsson, models the dynamic function-alities of the UTRAN, e.g., power control and soft handover down to the roaming of individual users and their individual radio link conditions. This chapter will in-troduce Voice-over-IP and its characteristics and the system model in more details, followed by the capacity requirements.

3.1 Voice-over-IP

Voice-over-IP is becoming a main-stream term with the success of various Inter-net applications, such as Skype, where users can make voice calls or even video calls to other users using the same software (PC-to-PC calls). This service in the fixed network is mostly free of charge but applications may also support VoIP calls to the PSTN. Internet providers offering the service of having home phones connected with a broadband connection, has further made VoIP a big competitor to the ordinary CS voice service with cheap rate plans and flexibility.

Oﬀering a VoIP service in the mobile networks has long been discussed within 3GPP and the challenges it will put on the network in order to reach the same reliability and QoS as CS speech. Voice calls is still the most important service in mobile networks, despite the advent of various IP services.

3.1.1 Architecture

The convergence of fixed and mobile networks relies on the architecture of the IP Multimedia Subsystem (IMS) which is specified by 3GPP [6]. IMS is a base for in-terconnecting fixed and wireless networks and enables new services to be provided

(36)

20 System Model and Requirements

such as VoIP or Push-to-Talk over Cellular (PoC), where instant communication is enabled by a single push of a button. As for session control signalling 3GPP has adopted the Session Initiation Protocol (SIP) which is speciﬁed by the Internet Engineering Task Force (IETF). SIP signalling is done whenever users setup a new session or add a new service. SIP signalling is not considered in the simulations when a VoIP session begins, however there is a ﬁxed channel setup time used in the simulator during which no VoIP packets are transmitted.

3.1.2 Adaptive Multi-Rate Speech Codec

3GPP has adopted the Adaptive Multi-Rate (AMR) speech codec as the manda-tory codec for WCDMA systems but AMR was already standardized by ETSI for the GSM systems. The AMR operates on 20 ms voice frames, at a sampling frequency of 8 kHz. The codec supports 8 diﬀerent modes with diﬀerent output bitrates in the range 4.75-12.2 kbps. The number of samples per voice frame is 8 · 103_·_{20 · 10}−3= 160 samples, but the current AMR mode decides the number

of bits per frame, hence output bitrate. It is also adaptive since it has a feature that allows it to switch mode, hence output bitrate every 20 ms upon command.

The codec also supports voice activity detection and generation of comfort noise parameters. Voice activity detection can determine if the speech frame ac-tually contains speech or if the user is silent. This is important since during silent periods it is not bandwidth eﬀective to encode the background noise with the same number of bits as for speech frames. If no bits are transmitted at all, it would be very uncomfortable to hear nothing whilst speaking and have the perception of a dropped call. Instead comfort noise parameters are generated and the frames carrying these are called Silence Descriptor (SID) frames which are sent every 160 ms and consist of a small payload of only 39 bits. A more detailed description of the AMR codec can be found in [1].

In this study it is assumed that the AMR codec operates in the 12.2 kbps mode throughout the sessions, and a model of SID frames is included and transmitted as described during silent periods. Hence the core payload from the AMR codec is 12.2 · 103_·_{20 · 10}−3 = 244 bits per AMR frame during active periods and 39 bits

during silent periods.

3.1.3 Protocol Overhead

The AMR frames will further be encapsulated into packets which introduces a certain amount of overhead consisting of headers from the RTP/UDP/IP proto-col stack. The fundamental protoproto-col aspects for VoIP are describes in detail in [12]. First one or many speech frames will be encapsulated into a Real-time Trans-port Protocol (RTP) packet. The number of AMR frames per RTP, nAM R, is a

parameter which can be tuned with a clear trade-oﬀ between header overhead and delay. With few frames per RTP packet the delay is kept low but the overhead

(37)

3.1 Voice-over-IP 21

from all protocols is significant and vice versa. There are two modes for forming the RTP payload, octet-aligned mode and bandwidth-efficient mode [15]. In the latter mode, each AMR frame will be accompanied by a table of contents field of 6 bits and a payload header of 4 bits describing the codec mode request. The total RTP payload is then padded to an even octet. The RTP protocol then adds a header of 96 bits to this payload.

Each RTP packet will then be placed in a Universal Datagram Protocol (UDP) packet. UDP is commonly used for VoIP instead of the alternative TCP. UDP is not as reliable as TCP but reduces the overhead compared to TCP. The UDP introduces a header of 64 bits. Finally the Internet Protocol (IP) introduces a header of 160 bits to the packet when IP version 4 is assumed.

As can be noted, the protocol overhead is signiﬁcant. Assuming the conditions above and nAM R=1, the total header is 320 bits and the RTP payload is 256 bits

when using the bandwidth-eﬀective mode. That gives an overhead factor of 1.25 which is not very eﬀective from a bandwidth perspective. In this study nAM Ris

set to 1 and the bandwidth-eﬀective mode is used, in order to prioritize minimum delay.

3.1.4 Robust Header Compression

As noted in the previous section, VoIP packets consist of a very small payload making the impact of protocol overhead a big factor. It is obviously not effective to have a overhead factor of 1.25 especially when VoIP means frequent transmis-sions (every 20 ms) of small payloads that are highly delay-sensitive. To reduce this effect there is a scheme called RObust Header Compression (ROHC) speci-fied by IETF [7]. ROHC exploits the fact that most fields in the RTP/UDP/IP headers remain static or introduce constant change throughout a session. This allows for the header to be compressed before transmission and decompressed at the receiver side. ROHC may compress the header to as little as 24 bits (3 bytes) but initially full headers must be transmitted before the scheme enters compressed mode. ROHC also requires synchronization with the decompressor and full head-ers might need to be transmitted again if synchronization is lost or after a number of consecutive packets are lost.

ROHC is though very eﬀective as can be seen in Figure 3.1 where the overhead ratio for a VoIP packet is illustrated. In this study, simulations are run with and without ROHC. ROHC is modelled by a constant 3 byte header throughout the session, hence the model is slightly optimistic. Figure 3.2 illustrates the average bitrate from the IP-layer dependent on nAM R for both cases with full header

and applying ROHC. It is clear that with increasing nAM R the overhead impact

decreases rapidly for the full header case, whilst ROHC is eﬀective even for small values of nAM R.

(38)

22 System Model and Requirements VoIP Packet Data UDP RTP IP Overhead Payload Data RoHC VoIP Packet Data UDP RTP IP Overhead Payload Data RoHC

Figure 3.1.Illustration of the overhead ratio for a VoIP packet, and the effect of ROHC.

1 2 3 4 5 6 7 8 9 10 12 14 16 18 20 22 24 26 28 30 n AMR Bitrate [kbps] Full Header ROHC

(39)

3.1 Voice-over-IP 23

3.1.5 Traffic Model

The flexibility of the radio network simulator makes it possible to specify and characterize various types of traffic in the network. In this study the only present traffic is uplink VoIP traffic characterized by means of parameters and principles that will be further explained here.

The average number of users in the system is specified by the offered load (users per cell) and is held constant on an average during the simulation. In order to maintain the offered load, new sessions are created by a Poisson process according to a session arrival intensity, λ which is specified through,

N = λT (3.1)

where N is the oﬀered load, and T is the mean session time.

Every session lasts for a certain time which is exponentially distributed with a mean session time of 90 seconds. The minimum session time is set to 20 seconds. During the ﬁrst tslow seconds of the simulation the sessions are created uniformly

so that the load after this time has elapsed, is equal to the desired load N . This is done in order to to slowly ramp up the traﬃc and avoid transient behavior. After the user’s session time has elapsed and the user has no undelivered bits to the system, the session is terminated.

The VoIP sessions are also characterized by a voice activity factor, av, which

is set to 50%. Thus the users speak and transmit VoIP packets on average half of the session time and are on average silent during the other half, during which SID frames are transmitted. The mean activity time is set to 2 seconds. During active periods the inter-packet time is ﬁxed. The number of AMR frames per packet, nAM R, is set to 1 hence packets arrive at 20 ms intervals. When new sessions

are created, the time until the ﬁrst packet arrives to the transmission buﬀer is randomly generated over the inter-packet time interval. This is done to ensure that the new users do not transmit packets at the very same time instant during their sessions.

The VoIP packets are stamped with the time of arrival to the transmission buffer. When bits are successfully delivered to the system (Node B) the remaining number of bits for the oldest packet is decreased until the delivered bits are assigned to an undelivered packet. The time of arrival to the transmission buffer is then deducted from the time of delivery and the delay for each packet can be logged per user. This method has clear limitations if packets are retransmitted. If a packet needs a retransmission, a new packet may be successfully transmitted and delivered whilst the previous packet is waiting to be retransmitted when using parallel HARQ processes. This implies that the calculated delays for the packets may not be correctly mapped, but the average delay for all packets is the same. This problem originates from the architecture of the simulator which is based on delivered bits, making the identification of complete packets cumbersome. This is

(40)

not expected to have an impact on the results in this study, when the incorrect mapping of delays don’t generate an absolute delay which exceeds the delay limit that is considered in each simulation case.

3.2 Propagation and Receiver Model

3.2.1 Fading

Section 2.1 described the characteristics of fading and common statistical mod-els. The distance attenuation in the simulator is derived from the Okumura-Hata propagation model, and the values given in Section 2.1.1 are used in the simula-tions. Shadow fading is derived in the simulator as described with a log-normal distribution with σdB= 8 and a decorrelation distance of 100 meters.

The fast fading in an urban environment is modelled through the standardized model named 3GPP Typical Urban. This model consist of a precomputed map of fast fading values. The antenna gain is derived from implemented antenna diagrams in the simulator. The simulations are run using a 2D-antenna diagram.

3.2.2 RAKE Receivers

Fast fading results from multipath propagation and at the receiver side this means that signal energy arrives dispersed in time with a certain delay between the signal components. A RAKE receiver consist of multiple correlation receivers called

fingers that may be allocated to those delay points where signiﬁcant signal energy

is received [13]. Each ﬁnger then correlates the received signal with the expected code in order to reduce the multipath eﬀect and reduce the own signal interference. Simulations are run with an implemented model of RAKE receivers.

3.3 System Model

This section will describe the most important settings used that deﬁne the system and its dynamic operations.

3.3.1 Network Layout

The layout of the network can be seen in Figure 3.3. The network consists of 7 sites, where each site (base station) uses a three sector antenna (directional antenna), resulting in 3 sectors or cells per site. In total, this layout consists of 21 cells in a hexagonal pattern. As the interference from surrounding cells is modelled in the simulator, there would be cells at the border of the layout where the interference situation is quite different from the situation that a cell within the pattern experiences. To avoid these border effects the simulator uses a wrapping-technique where imaginary copies of the system are repeated around the layout so that cells on the border will be affected equally as a cell inside the pattern. A 3

(41)

3.3 System Model 25

sectors per site layout generates a site-to-site distance of 3 times the cell radius. The cell radiuses used in the simulations are denoted R and 3R meters, where R [m] is a typical urban cell radius.

Figure 3.3. Network layout used in the simulations.

3.3.2 Mobility and User Placement

New users entering the system are initially placed randomly in the network accord-ing to an uniform distribution. After initial placement the users move speciﬁed by a constant absolute velocity with angular variations during the session. The velocity is set to 3 km/h in the simulations, which states a rather low mobility.

3.3.3 EUL configuration

Since the only present traffic in the system is VoIP over EUL, all users are us-ing the E-DCH transport channel. The UEs are of category 3 and uses a 10 ms TTI [3]. Non-scheduled mode is simulated, which means that the UEs are allowed to transmit E-DCH data at any time, up to a configured number of bits. It is the serving RNC that configures a non-scheduled transmission flow and transmission can be done without receiving any scheduling command from Node B. This re-duces signalling overhead and possible delay from scheduling.

The maximum number of bits a user may transmit per TTI in non-scheduled mode, is referred to as a non-scheduled grant. The simulations are run with a constant non-scheduled grant that allows for 2 RLC Protocol Data Units (PDUs) to be transmitted in one TTI, where one RLC payload holds 320 bits. This means that the non-scheduled grant allows for transmitting one VoIP packet per TTI with uncompressed protocol header. When ROHC is applied, a VoIP packet ﬁts into 1 RLC payload and may be transmitted during a TTI, allowing for a de-creased eﬀective bitrate from 64 kbps (full header) to 32 kbps. If the real ROHC

(42)

scheme would be modelled, there is as mentioned occasions where full headers are transmitted. This implies that the non-scheduled grant should be set to at least al-low for VoIP packets with full headers to be transmitted even when ROHC is used. When data is transmitted the power levels of the EUL physical channels are set in respect to the DPCCH power. This is done through two relative power oﬀsets to the DPCCH as described in section 2.5.1. The approach in this study has been to keep the EUL physical channels at their default power levels independent of other parameter settings used for diﬀerent simulation cases.

3.3.4 Hybrid ARQ

As described in section 2.5.2, parallel HARQ processes are used which allows for continuous transmission. 4 HARQ processes are specified for a 10 ms TTI and the maximum number of transmission attempts is set to 5. The soft combining scheme that is modelled is incremental redundancy, where retransmitted data is coded differently from the original transmission and each retransmission adds new redundant bits. The RLC works in unacknowledged mode which means that no RLC retransmission is done when the maximum number of transmission attempts is reached and the packet is lost. It is worth to note that a retransmission is done with the same number of bits as the first transmission, hence the transmission buffer is only decreased when data is assigned to an available HARQ process and transmitted there. Processes are always occupied until an ACK is received from Node B.

The fix minimum delay of a packet is always defined by the TTI which is the time window during which data is transmitted. Additional delay is generated from retransmissions which is specified by the HARQ Round-Trip Time (RTT), which becomes RT T = NHARQ· T T Ims. Thus every retransmission introduces an extra

uplink delay of 4 · 10 = 40 ms.

3.3.5 Power Control

The power control as described in section 2.4.5 regulates the transmission power of the DPCCH channel by the inner loop. The outer loop is triggered on a TTI basis and takes into account the TA target and the TTE. If more TAs in a HARQ process are used than the target, the SIR target is increased by a certain step-up size. If decoding is successful using the targeted number TAs or less, the SIR target is lowered by an amount dependent on the TTE and the step-up size. The outer loop only acts upon transmitted data and the SIR target remains unchanged when no data is transmitted on the E-DPDCH.

The TTE in this study is set to 1% and the initial SIR target as well as the TA target are varied as stated. The default initial SIR target is here denoted x [dB], which is the default value for traﬃc over EUL.

(43)

3.4 Capacity Requirements 27

3.3.6 HS-DPCCH model

The HS-DPCCH is an uplink physical channel that carries ACK/NACK and CQI reports back to Node B. In this study the assumption is that VoIP traffic is present in the downlink. In order to model this impact, an average power offset is used in the simulations. The HS-DPCCH is power controlled with the DPCCH and when an ACK/NACK or a CQI report is transmitted, the power level of the channel is set relative the DPCCH. This derived average power offset is based on how frequent CQI reports are transmitted and the downlink voice activity. The same power offset is used in the theoretical assessment as in the simulations.

3.3.7 Admission Control

In this study admission control is not performed and there will be no rejections from these mechanisms. Thus new users are always admitted regardless of the current uplink load situation or due to possible downlink resource limitations. In practice this is of course not a realistic situation, and it should be kept in mind when evaluating the capacity. In this study it can be justiﬁed from the fact that it makes it possible to examine the characteristics that set and limit the maximum possible capacity for diﬀerent parameter settings. The capacity is determined from clear requirements and at a point where the system is still stable though heavily loaded. A slight increase in the load will in that situation increase the interference level and become too high. The result will be that users are regulated to transmit with maximum power by the inner loop and the users’ SIR targets will rise quickly and the system collapses.

3.3.8 System Logging

All simulations are run with a simulation time of 200 seconds. Each simulation case is also run with three different seeds, that set the random functions in different states in MATLAB. This is done to increase the accuracy of the values. The logging starts after the initial 20 seconds of the simulation in order to let the traffic stabilize. Users who start a session less than 10 seconds before the simulation ends are not considered in the statistics for the same reason.

3.4 Capacity Requirements

In order to determine the capacity for a VoIP service there need to be clear criteria to evaluate results from simulations and for a theoretical estimation. To measure the capacity a couple of deﬁnitions are made in this section.

The main factor to ensure QoS is the strict delay that is allowed for each transmitted VoIP packet to be successfully received and decoded in Node B. Ev-ery delivered packet per user will be logged with their respective delay in the simulations. This allows for post processing of the statistics where the ratio of de-layed packets can be calculated in respect to a certain delay threshold. The delay

Voice-over-IP over Enhanced Uplink

Institutionen för systemteknik

Department of Electrical Engineering

Examensarbete

Voice-over-IP Capacity over Enhanced Uplink

Voice-over-IP Capacity over Enhanced Uplink

Examensarbete utfört i Reglerteknik

vid Tekniska högskolan i Linköping

av

Abstract

Acknowledgements

Abbreviations

Contents

Chapter 1

Introduction

1.1

Background

1.2

Problem Statement

1.3

Approach

1.4

Related Work

1.5

Chapter Outline

Chapter 2

Cellular Radio Systems

2.1

Propagation

2.1.1

Distance Path Gain

2.1.2

Shadowing Fading

2.1.3

Fast Fading

2.1.4

Antenna Gain

2.2

Multiple Access

2.2.1

Frequency Division Multiple Access

2.2.2

Time Division Multiple Access

2.2.3

Code Division Multiple Access

f

P

f

P

P

f

f

P

P

f

f

P

f

P

f

P

P

f

f

P

P

f

2.3

Cellular Systems

2.3.1

Cellular Network Structure

2.3.2

Handover

2.3.3

Uplink and Downlink

2.3.4

Power Control

2.4

UMTS

2.4.1