Nhiều giao thức truy cập đối với truyền thông di động P1

Multiple Access Protocols for Mobile Communications: GPRS, UMTS and BeyondAlex Brand, Hamid Aghvami Copyright  2002 John Wiley & Sons Ltd ISBNs: 0-471-49877-7 Hardback; 0-470-84622-4 El

Multiple Access Protocols for Mobile Communications: GPRS, UMTS and Beyond

Alex Brand, Hamid Aghvami Copyright  2002 John Wiley & Sons Ltd ISBNs: 0-471-49877-7 (Hardback); 0-470-84622-4 (Electronic)

1

INTRODUCTION

This book focuses on issues related to multiple access for cellular mobile communications,with a specific interest in access arbitration through multiple access protocols situated atthe lower sub-layer of the second OSI layer, namely the medium access control (MAC)layer

In this chapter, first an introduction to cellular mobile communication systems isprovided This introduction will be further expanded upon in Chapter 2, particularly withrespect to the features which distinguish the different generations of mobile communica-tion systems, from analogue first generation (1G) systems to possible fourth generation(4G) scenarios Next, it is discussed what impact the emergence of the Internet mayhave on cellular communication systems The importance of multiple access protocols isalso examined, particularly in the context of packet-based systems, and packet reserva-tion multiple access (PRMA) is considered as a case study Finally, together with somebackground information, an overview of our own research efforts related to PRMA-basedprotocols is provided These efforts are mainly concerned with how to combine PRMAwith code-division multiple access (CDMA), when such a combination is beneficial, andmore generally with different approaches to access control at the MAC layer and theirrespective benefits They are documented in detail in later chapters

1.1.1 The Cellular Concept

The first land mobile communication systems were based on wide area transmission [1].Each base station had to provide coverage for large autonomous geographical zones Calls

of customers leaving a zone had to be dropped and re-established in a new zone [2] Suchsystems suffered from low-capacity and high-transmit power requirements for mobiletransceivers, shortcomings that would not have allowed us to witness the tremendousgrowth in mobile communications in the past few years, with penetrations now exceeding

70% in many countries Only the introduction of cellular mobile communication systems

in the late 1970s made this development possible, by enabling frequencies, used in onecell, to be reused under certain conditions in other cells to increase capacity Nowadays,

mobile communication systems are almost by implication cellular communication systems

as well We use either of these two terms interchangeably, sometimes also the full term,

namely cellular mobile communication systems.

Cellular mobile communication systems are designed to provide moving users (frompedestrians to travellers in high-speed trains) with a means of communication In contrast

to (basic) cordless telephones, cellular telephones (also referred to as mobile phones,

mobile stations, mobile terminals or sometimes simply handsets) are not attached to aparticular base station, but may make use of any one of the base stations provided bythe company that operates the corresponding network Each of these base stations covers

a particular area of the landscape, called a cell The ensemble of base stations should

cover the landscape in such a way that the user can travel around and carry on a phonecall without interruption, possibly making use of more than one base station, as shown in

Figure 1.1 The procedure of changing a base station at cell boundaries is called handover

or handoff We prefer the first term, since it implies (unlike the second term) that an effort

is made to sustain a call across cell boundaries Obviously, these systems can also servestationary users, and do so increasingly, as fixed telephones are more and more substituted

by wireless phones

Communication from the mobile station (MS) to the base station (BS) takes place on

the uplink channel or reverse link and from BS to MS on the downlink channel or forward

link (Figure 1.1) To enable communication, some resources need to be allocated to the

base station (these may be frequency bands, time-slots, sets of codes, or any combination

of the three), which in turn may assign a portion of them to individual calls to supportcommunication on both uplink and downlink channels The amount of resources allocated

to users will depend on the current resource availability and the particular requirements

of each requesting user As the base station must be able to assign individual portions of

its resources to support multiple communications, basic multiple access techniques (such

as frequency-, time-, or code-division multiple access, with FDMA, TDMA, and CDMA

as their respective acronyms) are required together with multiple access protocols, which

govern access to these resources The basic multiple access schemes are briefly describedfurther below in this section, and the importance of the multiple access protocols isexamined in Section 1.3

1.1.2 Propagation Phenomena in Cellular Communications

The design of cellular communication systems is particularly challenging because of theadverse propagation conditions experienced on the radio channel Without discussing thecomplex underlying physical mechanisms, for which the reader may consult a mobilecommunications handbook such as that in Reference [3] or a book dedicated to radio

Figure 1.1 (a) Basic principle of cellular communications (b) Uplink and downlink channel

1.1 AN INTRODUCTION TO CELLULAR COMMUNICATION SYSTEMS 3

propagation such as that in Reference [4], three main propagation effects are usuallydistinguished These are the pathloss, slow fading or shadowing, and fast fading or multi-

path fading The pathloss describes the average signal attenuation as a function of the

distance between transmitter and receiver, which includes the free-space attenuation as onecomponent, but also other factors come into play in cellular communications, resulting

in an environment-dependent pathloss behaviour Shadowing or slow fading describes

slow signal fluctuations, which are typically caused by large structures, such as big

build-ings, obstructing the propagation paths Fast or multipath fading is caused by the fact

that signals propagate from transmitter to receiver through multiple paths, which can add

at the receiver constructively or destructively depending on the relative signal phases.The received signal is said to be in a deep fade when the paths add destructively in amanner that the received signal level is close to zero Fades occur roughly once everyhalf wavelength [3] Given that we are dealing with wavelengths of 30 cm and less incellular communication systems, it is clear that multipath fading can result in relativelyfast signal fluctuations; exactly how fast depends on the speed of the mobile station and

on the dynamics of the surrounding environment

When designing cellular communication systems and particularly when planning thedeployment of such systems (e.g choosing suitable base station locations), one willhave to account for these propagation phenomena appropriately One way to do this

is to use deterministic propagation tools such as ray tracing tools, which will late experienced signal levels for every specific location of the planned system coveragearea, taking into account every structure which could affect signal levels Another way

calcu-is to resort to statcalcu-istical models, which have to be establcalcu-ished by analysing tion measurements performed in suitably chosen environments, e.g classified as denseurban, typical urban, suburban and rural propagation environments, to name just a few.For the purposes of some of our investigations, we will deal with distance-independentpathloss coefficients and a so-called lognormal shadowing model, as outlined in detail inChapter 5

propaga-1.1.3 Basic Multiple Access Schemes

For reasons discussed in detail in Chapter 3, we make a distinction between basic multiple

access schemes, such as FDMA, TDMA, and CDMA, associated with the physical layer

(PHY) on the air interface of a mobile communications system, and multiple access

protocols, situated at the medium access control (MAC) layer above the PHY Roughly

speaking, the basic schemes provide the capability of dividing the total resources available

to a base station into individual portions, which can be assigned to different users, andthe protocols govern access to these resource portions, e.g provide access arbitration.Analogue first generation cellular communication systems made use of FDMA as a basicmultiple access scheme In digital 2G systems, TDMA is predominant, but a CDMA-basedsystem exists as well CDMA is the most commonly used form of multiple access for thirdgeneration systems, in some cases complemented by a hybrid CDMA/TDMA scheme

1.1.3.1 Frequency-Division Multiple Access

In FDMA, each communication is carried over one or two (depending on the duplexscheme, see below) narrowband frequency channels The channel bandwidth and themodulation scheme determine the gross bit-rate that can be sustained Because of non-ideal

3 2

Figure 1.3 TDMA frames, time-slots, and bursts

filters, guard bands must be introduced between these channels to avoid so-called adjacentchannel interference This is illustrated in Figure 1.2

1.1.3.2 Time-Division Multiple Access

In TDMA, rather than assigning each user a channel with its own frequency, users share

a channel of a wider bandwidth, which we shall call a (frequency) carrier, in the time

domain This is achieved by introducing a framing structure with each TDMA frame

subdivided into N time-slots, if N user channels are to be supported User i is then allowed to access the carrier only during time-slot i, by transmitting a so-called burst

which fits into this time-slot, as shown in Figure 1.3 In order to sustain a continuous

gross source bit-rate of R s bit/s, the transmission speed during the burst transmission must

be at least N R s bit/s

Provided that enough spectrum is available, multiple carriers may be assigned to eachcell Therefore, such TDMA systems feature typically also an FDMA element, and are inreality hybrid TDMA/FDMA systems

1.1.3.3 Code-Division Multiple Access

In CDMA, narrowband signals are transformed through spectrum spreading into signalswith a wider bandwidth, the carrier bandwidth Like in TDMA, multiple users sharethe carrier bandwidth, but like in FDMA, they transmit continuously during the call or

session The multiple access capability derives from the use of different spreading codes

for individual users Because of the spreading of the spectrum, CDMA systems are alsoreferred to as spread spectrum multiple access (SSMA) systems

Two basic CDMA techniques suitable for mobile communications are distinguished,namely frequency hopping (FH) and direct-sequence (DS) CDMA techniques ‘Proper’FH/CDMA systems have not been specified for mobile communications so far and are notdiscussed any further, but so-called slow frequency hopping (SFH) can also be applied

in TDMA systems The second generation Global System for Mobile Communications

1.1 AN INTRODUCTION TO CELLULAR COMMUNICATION SYSTEMS 5

(GSM) for instance features an SFH option, the benefits of which are discussed extensively

in Chapter 4

For one 2G system called cdmaOne and most 3G systems, e.g the Universal MobileTelecommunications System (UMTS), DS/CDMA was chosen as a basic multiple accessscheme In DS/CDMA, a bit-stream is multiplied by a direct sequence or spreading code

composed of individual chips They have a much shorter duration than the bits of the

user bit-stream, and this is why the original signal’s spectrum is spread The bandwidth

expansion factor, often simply referred to as spreading factor and in this book denoted

by the symbol X, is equal to the duration of a bit, T b, divided by the duration of a chip,

T c , i.e X = T b /T c

The same codes used at the transmitting side to spread the signals are used at thereceiving side to de-spread them again If codes assigned to different signals or userchannels are mutually orthogonal, then these signals can be perfectly separated at thereceiving side In practise, due to multipath propagation, fully orthogonal separation at thereceiving side may not be achieved even when the codes are orthogonal at the transmittingside Provided that appropriate measures are taken, this is not really a problem, but it has aninteresting consequence, which is highly relevant for some of the topics discussed in thisbook Non-orthogonality creates mutual interference between all users, so-called multipleaccess interference (MAI) The resource assigned to an individual user in a CDMA system

is therefore not so much a code, but rather a certain power level This is illustrated inFigure 1.4, which shows sharing of resources in the time-domain, the frequency-domain,and in terms of power levels, for FDMA, TDMA and CDMA respectively

1.1.3.4 Frequency-Division Duplex and Time-Division Duplex

To sustain a bi-directional communication between a mobile terminal and a base station,transmission resources must be provided both in the uplink and downlink directions.This can either happen through frequency-division duplex (FDD), whereby uplink anddownlink channels are assigned on separate frequencies, or through time-division duplex

Power Time

Frequency

Frequency

Figure 1.4 Sharing of time-, frequency- and power resources between three users in FDMA, TDMA and CDMA respectively

(TDD), where uplink and downlink transmissions occur on the same frequency, but nate in time Both methods can be applied in conjunction with any of the above-describedmultiple access schemes 1G and 2G systems apply FDD In UMTS, a 3G system, bothFDD and TDD modes are supported, not least because symmetric uplink and downlinkspectrum is normally required for FDD-only systems, but 3G spectrum consists of both

alter-so-called paired bands (i.e symmetric spectrum) and unpaired bands.

A description of the key features of 1G, 2G and 3G systems is provided in Chapter 2,which also considers possible 4G scenarios Advantages and disadvantages of the differentbasic multiple access schemes are examined in more detail in Section 3.2 Approaches

to the modelling of the physical layer performance for some of our investigations arediscussed in Chapter 5 Chapter 4 on multiple access in GSM and GPRS deals also toquite a considerable extent with physical layer issues

1.1.4 Cell Clusters, Reuse Factor and Reuse Efficiency

As pointed out earlier, resources used in one cell may be reused in other cells, butthis must be done in such a way that ongoing communications experience sufficientquality Assume for now that we are dealing with frequency resources, and that the main

factor affecting the quality is the so-called co-channel interference, that is, interference

generated by communications in other cells transmitting on the same frequency as adesired communication link in a test cell The required communication quality, togetherwith other factors, e.g related to propagation conditions, such as the pathloss coefficient,will determine the minimum distance that must be respected between two co-channel

cells, the so-called reuse distance This leads to the concept of cell clusters (or cellular

reuse patterns), namely a set of neighbouring cells within the reuse distance, any two

of which are not allowed to use the same frequency The frequencies are instead reused

in a cell occupying the same relative position in a neighbouring cluster, as illustrated inFigure 1.5 In other words, every cell in a cluster obtains a share of the total bandwidthavailable to an operator, and the same bandwidth is reused in other clusters The number

of cells within each cluster is called the (frequency) reuse factor or cluster size N f [5],

which is seven in the example depicted The reuse efficiency is the inverse of the reuse factor, hence 1/N f

With such a cellular approach, it is in theory possible to increase capacity without limitthrough cell splitting (i.e deploying multiple small cells in an area previously served by

a single big one), but there are certain practical constraints

1.1.5 Types of Interference and Noise Affecting Communications

The permitted co-channel interference, which depends on various physical layer aspectssuch as the modulation scheme employed, is a key parameter determining the minimumfrequency reuse factor Communications taking place on adjacent channels can also createnotable mutual interference because of non-ideal filters both at the transmit side (resulting

in some power being also radiated outside the allocated channel) and at the receive side(due to receive filters not fully rejecting out-of-band signals) This is referred to as adjacentchannel interference (ACI) It is strongest between directly adjacent or neighbouringchannels, and decreases as channels further apart are being considered owing to the filter

1.1 AN INTRODUCTION TO CELLULAR COMMUNICATION SYSTEMS 7

1 1

6

2

2 3

3

3

4

4 5 5

5

7 6

6 7

7 4

D Reuse distance

Cell radius R

Figure 1.5 Cell clusters assuming hexagonal cells with a frequency reuse factor of seven

attenuation In general, ACI is much less of a problem than co-channel interference All thesame, when several frequency channels are assigned to a cell, if they are neighbouringchannels, guard bands should separate them to avoid excessive ACI, otherwise non-neighbouring channels should be assigned Unfortunately, due to limited 3G spectrumand the fact that wideband channels are used, there are 3G scenarios for such systemswhere neither sufficient guard band is available nor non-neighbouring channels can bechosen, hence ACI becomes an issue, as discussed in Section 2.3

Compared to TDMA and FDMA systems, CDMA systems exhibit certain peculiarities.Firstly, the reuse factor can be set to one in CDMA systems (and in fact often is)

This is also referred to as universal frequency reuse In this case, mutual interference is

generated between all cells, hence rather than referring to this as co-channel interference,

the term intercell interference is used Secondly, while user channels within a cell are

separated from each other in an orthogonal manner both in TDMA systems (perfectseparation between time-slots can be achieved through guard periods) and in FDMAsystems (assuming sufficient guard bands to avoid ACI), this is not necessarily the case

in CDMA systems Since spreading codes do not always provide orthogonal separation,

interference within a cell, so-called intracell interference, can become an issue as well.

Therefore, intracell and intercell interference can both be non-negligible components ofthe total multiple access interference experienced by a communication link in a CDMAsystem

On top of interference generated by other users in the system, additional noise sourcesmay affect the quality of a communication, for instance thermal noise In the following,the term ‘interference’ refers to noise generated by other cellular users, and ‘noise’ tothermal noise as well as noise generated by sources outside the considered system Thecommunication quality in terms of bit error rate (BER) or frame erasure rate (FER) cantherefore be expressed as a function of the signal-to-noise ratio (SNR), or the signal-to-interference ratio (SIR), depending on which type of signal disturbance is dominant.Typically, at the beginning of a system build out, when there are few cells, the system is

coverage-limited, and the SNR is mostly relevant As cells are added to fill in coverage

holes and reduce cell radii, and the user traffic increases, the system becomes

capacity-limited and the SIR becomes more critical In situations where neither interference nor

noise can be ignored, the signal-to-interference-plus-noise ratio (SINR) may be considered

as a channel measure However, if the nature of the interference is significantly differentfrom that of the noise and affects the signal in a different manner, then it may not bepossible to lump the two together and describe the performance as a function of the SINR.Instead, one would have to use, for example, SNR curves parameterised to interferencelevels or SIR curves parameterised to noise levels

The signal quality can also be expressed as a function of the ratio of energy per bit

E b either to the noise power per Hertz, N0, or the interference power per Hertz, I0.Finally, instead of using signal and interference levels at the ‘base-band’, the so-calledcarrier-to-interference ratio (CIR) at the radio frequency level is often used According toReference [6],

CIR= E b · R b

I0· B c

with R b the bit-rate in bits per second, and B cthe radio channel bandwidth in Hertz

Cellular Communications

In the ‘wired world’, we are witnessing how traffic of all types is increasingly beingcarried on packet-switched networks using the connectionless internet protocol (IP) — orrather, the IP protocol suite, which features various other protocols on top of IP, e.g.transport protocols such as the transport control protocol (TCP) and the user datagramprotocol (UDP) This development is mainly due to the tremendous success the Internethas enjoyed in recent years (incidentally, not unlike cellular communications) Initiallyconstrained to non-real-time applications such as Telnet, file transfer, email and Webbrowsing, this move towards IP now embraces audio and video streaming with morestringent delay constraints, and even ‘proper’ real-time traffic such as Voice over IP (VoIP).Strictly speaking, it is typically voice over RTP (the real-time protocol, an application-level protocol), UDP and IP It is now widely anticipated that the same will eventuallyhappen in the wireless world as well, which has some serious technical implications oncellular communication systems In the following, we deal with the general implications;the impact specifically on multiple access protocols is discussed in the next section.Already in the late eighties (e.g in Reference [7]), Goodman, whom we will refer to

at various other occasions in this text, suggested that both the fixed architecture and theair interface of 3G systems should be based entirely on packet-switching for all types

of services Not only would the available resources be exploited more efficiently, butalso certain functions could be decentralised and distributed over many processors, whichwould improve the scalability of such systems He also proposed a packet-based multipleaccess protocol called packet reservation multiple access (PRMA) suitable for the wirelesslinks between mobile and base stations Although his vision of an all-packet systemwas probably not driven by the Internet at that time (he suggested that ‘3G systems, in

harmony with broadband integrated services digital networks, would use shared resources

1.2 THE EMERGENCE OF THE INTERNET 9

to convey many information types’), it is in some respects in tune with Internet architectureprinciples

Goodman’s vision has not really caught on during initial 3G standardisation efforts.True, unlike 2G systems, first releases of 3G systems have incorporated packet-datasupport right from the start, however, without proper support of real-time packet dataservices such as packet voice For instance, the first UMTS release might well provideimproved support for packet data over the air interface compared to the GSM GeneralPacket Radio Service (GPRS) However, the packet-based infrastructure in the fixednetwork, which has evolved from the GPRS infrastructure, was not designed for voice.Instead, it was intended that voice would always be delivered over the circuit-switchedinfrastructure

But why this reluctance towards packet-voice and an all-packet system? There was asignificant amount of scepticism in the industry regarding the feasibility of an all-packetsystem which could deliver the high-quality standards required for voice communications.Also, decentralisation, explicitly advocated by Goodman, to some extent inherent in themove to a packet-only system, and certainly consistent with the Internet architecture,means loss of operator control This is something that operators do not like too much, asthey tend to control the types of services delivered through their networks The Internet,

by contrast, is built according to the ‘end-to-end principle’, where the infrastructure between the end nodes is not concerned with the services to be delivered

in-In the case of the cellular communications industry, apart from commercial tions, there are sound technical reasons for this desire to control matters Take the issue

considera-of handovers as an example Goodman [7] proposed to decentralise them completely,effectively placing them into full control of the mobile terminals, in order to cope effi-ciently with the large volume of handover-related signalling traffic as a result of smallerand smaller cell sizes due to ever increasing traffic density Cellular operators, however,like to control which terminal is served by which cell and thus prefer network-controlledhandovers This is firstly because normally only the network has a complete view of

the communication quality on both up- and downlink (the latter through measurement

reports sent by terminals), which may be different Secondly, the quality of individualcommunications has to be traded off against system capacity and the quality of othercommunications, requiring careful admission control and load balancing by the network

In general, centralised algorithms exploiting the global view of a matter perform betterthan decentralised ones with only local information available Obviously, they are alsomore complex, and therefore sometimes inappropriate, but when it comes to efficient use

of scarce and precious air-interface resources, it is often worth the effort

In spite of the initial scepticism by the cellular industry towards packet-based systems,the power of the Internet is proving too strong, and things are moving on Subsequentreleases of 3G systems will be capable of supporting voice over the packet-switchedinfrastructure as well This does not necessarily imply a complete decentralisation of thearchitecture of cellular systems, at least not in the beginning Operators driving thesedevelopments are predominantly interested in the new services they hope to deliverover their networks, as discussed in somewhat more detail in Section 2.4 and again inChapter 11, and most of them are not (yet?) prepared to give up control In terms of ourmain topics of interest for this book, such developments will primarily impact multipleaccess protocols, less the basic multiple access schemes However, as decentralisationgoes further (assuming that it will eventually), autonomous ‘plug-and-play’ base stations

become desirable, something which could affect the choice of preferred basic multipleaccess schemes as well For instance, in CDMA systems, to improve the transmissionquality, terminals are often connected to the network via multiple base stations Thisimplies some co-ordination between these base stations Furthermore, entities are neededthat can process the signals of multiple base stations Dealing with such matters throughpacket-based systems is by no means impossible, but it is somewhat of an obstacle to fulldecentralisation.

Cellular Communications

In 2G systems, which were designed to carry voice and some low-bit-rate data services

by setting up ‘circuits’ or dedicated channels for the duration of a call, access arbitration

is only required at the time of setting up a call to request such dedicated channels Withthe advent of advanced 2G systems, e.g GSM complemented by GPRS, and first releases

of ‘true’ 3G systems such as UMTS, non-real-time data carried on common or sharedchannels becomes increasingly important, which calls for more sophisticated multipleaccess protocols As just discussed, these systems will further evolve to support real-time

IP traffic What does this mean in terms of choosing appropriate multiple access protocols?

We are continuing to use Reference [7] as a case study, since packet reservation multipleaccess, the multiple access protocol proposed by Goodman for the air-interface uplinkchannel, was a subject of extensive research efforts by the authors, as documented inthis book (see Section 1.4) Consider a traffic source which alternates between ‘off’ or

‘silent’ (no packets are generated) and ‘on’ or ‘active’ (packets are generated at a ratematching the channel transmission rate, e.g one packet per TDMA frame fitting into one

time-slot) A typical example would be a voice source subject to voice activity detection.

By reserving resources on the air interface only during ‘on’ phases, when packets need

to be transmitted, rather than hanging on to them for the entire duration of a call (aswould be the case in a ‘circuit-switched model’), PRMA attempts to make efficient use of

air-interface resources Compared to a conventional TDMA air interface, where N slots can sustain N calls, with PRMA M calls can be multiplexed onto N now shared time-slots, with M > N ; how many exactly depends obviously on the so-called activity

time-factor, i.e the fraction of time the traffic source is in ‘on’ state In other words, PRMA

allows for a certain degree of statistical multiplexing over the air One could therefore

conclude that in conjunction with a packet-switched infrastructure, a ‘packet-switchedair interface’ such as PRMA would also make sense However, while the split betweenpacket-switching and circuit-switching is fairly evident in the fixed network infrastructure,when dealing with the air interface, the situation is a little bit more complicated.Essentially, on the air interface, we can distinguish between dedicated channels on

one hand and common or shared channels on the other Typically, dedicated channels,

which are set apart for the sole use of one communication link between a mobile terminal

and a base station, are associated with the ‘circuit-switch model’ Conversely, shared

channels (shared between a limited number of users) or common channels (common to

the whole cell population), for which appropriate multiple access protocols are crucial,are often associated with the ‘packet-switch model’ This is indeed often appropriate,

particularly for packet-based services which exhibit very bursty traffic characteristics,

i.e traffic sources which alternate between short activity periods (e.g at high bit-rates)

1.3 THE IMPORTANCE OF MULTIPLE ACCESS PROTOCOLS 11

and long inactivity periods However, depending on the type of packet-traffic, the basicmultiple access scheme in use, and the frequency planning applied, a statistical multi-plexing gain may also be obtained when using dedicated channels over the air interface

In fact, this may even be the better choice in certain circumstances Roughly speaking,apart from traffic characteristics, this depends on whether the system is blocking-limited

or interference-limited

A blocking-limited system is one in which a fixed number of channels are available

per cell The transmission quality is largely independent of the resource utilisation, that

is the fraction of available channels which are assigned to ongoing calls New calls orsessions are admitted if a channel (e.g a time-slot) is available, and blocked if this is notthe case In such a scenario, using a protocol such as PRMA, which carries all traffic onshared or common channels, provides indeed a performance advantage for all types ofpacket traffic owing to statistical multiplexing

In an interference-limited system, by contrast, the transmission quality depends on the

system load: the more calls are admitted to the system, the worse it gets Obviously, theaim is to admit calls only if the required quality levels can be met, but exceeding theappropriate load level somewhat results only in a gradual degradation of quality, and may

occasionally be tolerated This is referred to as soft-capacity feature and the load level at which the required quality level can just be met is sometimes called soft-blocking level.

(Incidentally, by applying PRMA on a blocking-limited system, we can get soft capacity

as well.)

As we identified earlier, the key resources assigned to users in CDMA systems arepower levels, which makes these systems naturally interference limited Also a TDMAsystem can be operated in an interference-limited fashion, if it features the option ofslow frequency hopping In such interference-limited systems, statistical multiplexingoccurs naturally through interference multiplexing, even when dedicated channels areused Voice activity detection, for instance, leads to an interference reduction duringvoice silent phases also on dedicated channels, meaning that the total power budget isshared flexibly and dynamically between users Whether dedicated or shared channelsare the preferred option then depends on the statistical behaviour of a communicationsource, the available code resources (in the case of CDMA), the overhead that is required

to maintain a dedicated link while the source is silent, and the efficiency of the multipleaccess protocol in assigning and releasing, e.g shared channels If the used multipleaccess protocol is well designed, the balance may be tipped towards shared channels,particularly for bursty sources However, for packet-voice in a CDMA system, dedicatedchannels may often be the best choice, which means that from a statistical multiplexingperspective, the performances of ‘packet-switched voice’ and ‘circuit-switched voice’ aresimilar (assuming that the same type of voice activity detection is applied in both cases).This puts ‘voice over IP’ at a disadvantage compared to optimised circuit-switched voice,because of the additional overheads associated with IP protocol headers

We are dealing extensively with such topics throughout this book In Chapter 3, weprovide a review of basic multiple access schemes and multiple access protocols InChapter 4, where the GSM air interface and its GPRS additions are described, key topicsinclude resource utilisation and blocking-limited versus interference-limited system oper-ation In Chapter 10, the different options available on the UMTS air interface for thesupport of packet traffic are described In Chapter 11, the issue of interference-limitedoperation versus blocking-limited operation specifically for supporting VoIP in enhanced