Perspectives for present and future CDMA-based ... - IEEE Xplore

MULTIPLE ACCESS TECHNOLOGIES FORB3G WIRELESS COMMUNICATIONSPerspectives for Present and FutureCDMA-Based Communications SystemsRomano Fantacci, Francesco Chiti, Dania Marabissi, Giada Mennuti, Simone Morosi, and Daniele Tarchi,Università di FirenzeABSTRACTIn this article the main boosters of futureCDMA communication systems are presentedand described in order to highlight ways to copewith the more challenging requirements futuresystems and services claim. Particular attentionis devoted to the inherent weak points of CDMAsystems and detection techniques that can beused to overcome their impairments. Moreover,high-data-rate transmissions in wireless channelsrequire proper link adaptation techniques; theseare thoroughly described as well. Finally, the lastpart of this article focuses on the design of suitableprotocol strategies for new heterogeneousmultimedia packet services characterized bystrict QoS requirements.INTRODUCTIONIn recent years the increased demand foradvanced wireless communication services hasled to the definition of more and more powerfulsystems. Third-generation (3G) mobile serviceshave just begun, and in the 3G Partnership Project(3GPP) 3G enhancement activity has alreadystarted. Furthermore, various wireless researchorganizations such as the Wireless WorldResearch Forum (WWRF) and InternationalTelecommunication Union (ITU) are holdingactive discussions about systems beyond 3G tobe deployed around 2010. For next-generationmobile wireless systems, high data rates and customizablesystems enabled by flexible technologiesare envisaged. In order to satisfy the highdata rate requirement and efficiently supportmultimedia services, modulation and multipleaccess schemes must be spectrally efficient andflexible. Code-division multiple access (CDMA)and orthogonal frequency-division multiplexing(OFDM) are two of the most promising candidates,as well as hybrid schemes based on bothtechniques. In particular, CDMA has emergedas the predominant radio access technology toprovide multimedia services in 3G systems. However,high-data-rate transmissions with quality ofservice (QoS) requirements are strongly influencedby the propagation conditions of themobile wireless channel and multiple accessinterference due to simultaneous users. As aconsequence, to achieve efficient utilization ofthe limited radio spectrum and allow contemporaryconnections with different QoS requirements,remarkable improvements in both thewireless physical (advanced detection schemes,multiple antennas, coding, etc.) and link layer(access control, bandwidth allocation, etc.) arerequired.This article deals with some of these promisingapproaches and schemes of special interestin future beyond 3G wireless networks wherebursty high-speed traffic is foreseen, andadvanced detection schemes and high bandwidthefficiency are pursued. The outline of this articleis as follows. We deal with physical layer issuesbyproviding a brief survey of the principal detectionschemes. We also include discussions of theadvantages afforded by efficient detectionschemes with antenna array systems. We dealwith link layer issues: following a brief introductionto the principles of adaptive modulationand coding (AMC), a suitable scheme for efficientAMC implementation in 3G and B3G systemsis described. In addition, to increase servicereliability, a novel scheme that combines turbocodes’ advantages with those of efficient automaticrepeat request (ARQ) schemes is discussed.We then deal with radio resourcemanagement aspects including admission control,power control, and scheduling schemes.Finally, conclusions are drawn.PHYSICAL LAYER ISSUESThis section deals with physical layer issues. Inparticular, some advanced receiving approachesare discussed herein. In addition, a brief discussionof multicarrier CDMA (MC-CDMA) is provided.ADVANCED DETECTION TECHNIQUESAs is well known, the benefits of the directsequence (DS)-CDMA technique can be summarizedas a potential increase of system capaci-IEEE Communications Magazine • February 2005 0163-6804/05/$20.00 © 2005 IEEE95

NoReliable signalscancellationRake (MRC)Informative bitdecisionUnreliable usersBasebandreceived signal r(t)Rake (MRC)> Threshold?ear detectors are the decorrelating and minimummean square error (MMSE) detectors [1].The former performs the inverse of a cross-correlationmatrix with matched filter outputs inorder to decouple the data. The latter implementslinear transformation that minimizes themean squared error between the actual data andthe soft output of the conventional detector.Interference cancellation receivers explicitlyestimate the overall interference due to multipleaccess and multipath for each received signal,then subtract it from the received signal beforemaking a data decision [1]. IC receivers havelower implementation complexity than lineardetectors, so they are widely considered forfuture applications.As described in the literature, IC can be performedin parallel (i.e., simultaneously) for allusers, resulting in a parallel IC (PIC) receiver,or in a serial way, resulting in a successive IC(SIC) receiver [1].Of special interest for multirate applicationsare receivers based on the groupwise approachin which user signals are divided into groups(e.g., according to their spreading factor). Symboldetection and interference estimation areperformed for signals within the same group.Hence, the associated interference effects aresubtracted from signals in other groups. A specialcase of groupwise IC receiver is the selectivePIC (SPIC) receiver proposed and analyzed in[2, 3]. In this case user signals are divided intoreliable and unreliable signal groups based oncomparisons of the decision variables with aproper threshold value. Reliable signals aredirectly detected and (after signal reconstruction)canceled from the whole received signal.Unreliable signals are detected after cancellationof the MAI effects due to reliable signals. In Fig.1 a block scheme of an SPIC receiver is shown.It is shown in [3] that the SPIC receiverachieve almost the same performance as the PICreceiver with the advantage of lower implementationcomplexity.Recently, advanced receiving schemes havebeen developed by taking into account that practicalCDMA communication systems rely on theuse of error control coding and interleavingschemes. Optimal joint decoding/detection methodsallow excellent performance to be achieved,but unfortunately require prohibitive computationalcomplexity for actual applications. Tolower this drawback, suboptimal solutions inwhich symbol detection and channel decodingare performed and optimized separately havebeen proposed [1]. In particular, the successfulproposal of turbo codes [4] has led to the definitionof powerful MUD receivers based on theiterative (turbo) processing technique. In iterativeMUD, extrinsic information is determinedin each detection and decoding stage, and usedas a priori information for the next iteration.This procedure is adopted at each iteration, as inturbo codes; this detection philosophy is definedas turbo MUD, and the advantages of its introductionare remarkable for heavily loaded systemsas well. Even if turbo MUD receivers stillhave the drawback of high implementation complexity,they represent a promising approach toenhance performance in future high-speed mulnFigure 1. A block scheme of the SPIC receiver.YesInformative bitdecisionReliableusersty, the ability to support universal frequencyreuse, graceful degradation under loaded conditions,soft handoff capability, easy adaptation tovariable-rate services, low power flux densityemission, and finally, the possibility of using classicaldiversity techniques to cope with multipathfading effects. On the other hand, this techniqueis basically interference limited (i.e., performanceis limited by multiple access interference,MAI). Hence, suitable detection schemes haveto be considered in order to lower this drawback.The traditional DS-CDMA detection techniqueis based on the use of a correlation receiverin which signals of additional users areconsidered pure interference. Hence, if a correlatorreceiver is used, the MAI limits the numberof active users in relation to a specified biterror rate (BER) value. Another shortcoming isits vulnerability to the near-far effect (i.e.,impairments due to the strongest signals onweak ones), which mandates tight implementationof the power control scheme to prevent performancefrom excessive degradations. Duringrecent years more efficient detection schemes,multiuser detection (MUD) receivers, have beenproposed and investigated as a potential methodto improve CDMA system performance [1]. Themain feature of MUD receivers is joint detectionof all received signals to eliminate the near-farproblem and increase system capacity.The optimal MUD receiver, formed by abank of matched filters followed by the Viterbialgorithm (i.e., maximum likelihood sequenceestimator, MLSE), has implementation complexity(exponential in the number of users) prohibitivefor wide adoption in practicalapplications. Hence, research efforts havefocused on suboptimal MUD approaches, classifiedinto linear and interference cancellation(IC) MUD receivers [1]. In linear MUD receivingschemes, a linear transformation is appliedto the matched filter outputs to reduce the MAIeffects on each user. The two most popular lin-96IEEE Communications Magazine • February 2005

In recent years,Multi-Carriersystems, based onOFDM and CDMAschemes, have beenthe subject ofextensive researchto efficientlyevaluate thefeasibility of suchnew air interfacecandidate.requirements is an OFDM-based system. TheOFDM technique is robust against severe multipathfading and can be efficiently implementedby means of fast Fourter transform(FFT)/inverse FFT (IFFT). In recent years, multicarrier(MC) systems based on OFDM andCDMA have been the subject of extensiveresearch to efficiently evaluate the feasibility ofa such new air interface candidate.A comprehensive overview of different solutionscombining OFDM and CDMA to overcometheir inherent limits (e.g., lack of efficientmultiple access in OFDM, the near-far effectand cumbersome structure to combat fading inCDMA) was accurately carried out in [7].Advantages provided by these solutions are efficientexploitation of energy scattered in the frequencydomain, robustness toward fading, andthe high data rates achievable.LINK LAYER TOPICSThis section deals with two main link layer topics:adaptive modulation and coding techniquesand hybrid ARQ protocols.ADAPTIVE MODULATION ANDCODING TECHNIQUESAdaptive modulation and coding (AMC) belongsto the class of link adaptation (LA) algorithms.The basic idea behind LA techniques is to adapttransmission parameters to take advantage ofchannel conditions. The fundamental parametersto be adapted include modulation and codinglevels, but other quantities can be adjusted, suchas power levels (as in power control), spreadingfactors, and signaling bandwidth. LA is nowwidely recognized as a key solution to increasethe spectral efficiency of wireless systems. Animportant indication of the popularity of suchtechniques is the current proposals for 3G wirelesspacket data services, such as UniversalMobile Telecommunications System (UMTS), aswell as 2.5G, such as General Packet Radio Service(GPRS), to provide a higher data rate.One of the main challenges for LA is knowledgeof the channel parameters, because ofdelay in feedback response and, on the otherhand, channel parameters changing due to thefading nature of the wireless medium.AMC is one method used in order to performLA in wireless networks. AMC techniques workby defining some schemes, each characterized bydifferent modulation and coding types (modulationand coding scheme, MCS); in a systemwhere LA is performed using AMC, the bestMCS in terms of throughput maximization isselected, considering that in an environmentwith low SNR it is better to use low-order modulationand coding, while in a high SNR environmentit is better to use high-order modulationand coding.With AMC, the power of the transmitted signalis held constant, but the modulation andcoding formats are changed to match the currentreceived signal quality. In a system with AMC,users close to the base station are typicallyassigned higher-order modulations and highcode rates (e.g., 64-quadrature amplitude modulation[QAM] with rate-3/4 convolutional coding),but the modulation order and/or code ratewill decrease as distance from the base stationincreases.The choice of the best scheme is madeaccording to the measure of some parameterstaking into account the channel state informationin order to adapt the modulation and codingscheme. The information to be monitoredcan include a great variety of parameters. Forexample, a CDMA-based system is more vulnerableto power unbalance, so in its case it is betterto monitor the received power fluctuation ormultiple access interference.The channel state information (CSI) can bemonitored in two main ways: direct and indirectmode. In direct mode the channel state is measuredusing some pilot symbols and verifying atthe receiving end if the received power is withina specific range, while in the indirect methodspecific information such as the acknowledgment(ACK)/negative ACK (NACK) packets relatedto the reliability of the received information areused. Increasing interest in AMC techniques isshown by the attention devoted to them in thedefinition of some new telecommunications systemssuch as Enhanced Data Rate for GSMEvolution (EDGE), while for 3G system anextension of UMTS, high-speed downlink packetaccess (HSDPA), is foreseen based on the use ofsuitable AMC techniques to improve performancein terms of throughput.HYBRID ARQ PROTOCOLSThe traditional approach to establish a reliablelink-to-link connection is based on adding ARQprotocols to the logical link control (LLC); thebasic schemes are Stop and Wait (SW), Go BackN (GBN), and Selective Repeat (SR). The aboveschemes show increasing implementation complexitycounterbalanced by better throughput(i.e., lower average number of retransmissionsneeded to successfully deliver a packet).The wireless communication paradigm causesbursts of errors to occur during which packetscannot be successfully transmitted on the link,and proper error detection and correction techniquesare needed without drastically reducingspectral efficiency.In this case, to lower residual error rate andhence allow reliable services, powerful channelcoding schemes have been proposed [8]; amongthem, the best candidates seem to be turbocodes [4]. The main drawback of this approach,commonly called forward error correction(FEC), is represented by added redundancy thatdrastically reduces system throughput. HybridARQ (H-ARQ) techniques keep the advantagesoffered by previous approaches together withmitigation of the typical drawbacks, thus settingup highly mobile and reliable end-to-end connections.Basically there are three H-ARQ schemes:type I H-ARQ, in which each packet can beindependently decoded, thus discarding erroneouspackets; type II H-ARQ, in which jointpacket decoding is carried out by keeping trackof each message’s decoding results in a memorybuffer; and type III H-ARQ, in which the MCSsare varied according to the transmission out-98IEEE Communications Magazine • February 2005

come. One of the most promising type II H-ARQ schemes is based on uncorrelated datacombination, often called packet combining. Thispermits the Euclidean distance between thecodewords and the received sequence to be maximized,as shown in [9], thereby minimizing theerroneous detection probability. Packet combiningcan be based on either hard decision outputsor on soft channel outputs. In particular, two differenttype II H-ARQ schemes based on softcombining are possible: input and output combinedtype II ARQ. In the former the additionalinformation to be combined with the actualpacket may be a received frame replica, while inthe latter it is extrinsic information provided bythe decoding of the previous frame. The choicedepends mostly on the reliability of the decodingprocess that can increase or decrease soft output’sreliability. As an example, Fig. 3 sketchesthe block diagram of a receiver scheme in whichoutput combined type II ARQ is adopted.Demodulator Decoder CRC checkPreviousINSICLogicn Figure 3. A receiver scheme with output combined type II ARQ.New callSOCPreviousOUTACK/NACK + CSIRADIO RESOURCEMANAGEMENT ASPECTSFuture wireless systems ought to supply userswith a wide range of different multimedia applications,combining real-time with data servicesand supporting different QoS requirements. As aconsequence, flexible radio resource management(RRM) policies, suitable to varying trafficconditions become crucial. RRM functionsinclude all procedures implemented to dynamicallymanage radio resources in order to meetthe instantaneous demands of users movingaround in cells, providing data delivery withappropriate end-to-end QoS guarantees. Theywill impact overall system efficiency, so they willdefinitely play an important role in future wirelessnetworks [10]. We briefly discuss belowthree main topics in the definition of a suitableRRM scheme: admission control, power control,and scheduling.PowerreconfigurationSINR estimationAdmission controlSINR < SINR targetNoAcceptn Figure 4. An admission control scheme.ADMISSION CONTROLIn CDMA systems, where bandwidth is ashared resource, special attention must be paidto MAI arising among active connections. Inparticular, significant QoS degradations mayarise if this parameter is not kept under controlby limiting the number of active communications.This goal is achieved by the use ofproper admission control (AC) algorithms. Thetypical AC scheme preserves ongoing connections’QoS by rejecting new connections wheneverit is foreseen that they could causeunacceptable degradation of active communications(i.e., outage problems).The simplest AC allows admission of anincoming request if and only if the total numberof users does not exceed a specified limit definedduring the network planning phase.Conversely, dynamic AC algorithms are basedon measurements or estimations of the actualsystem load or interference level, so a variablenumber of connections may be handled in relationto actual conditions. An alternativeapproach is that of considering the AC schemein conjunction with power control or adaptiveratetransmission. As an example of such anapproach, Fig. 4 illustrates its basic functionalblocks. Typical gain values in terms of additionalaccepted data connections under very high calltraffic load conditions with respect to staticapproaches are greater than 60 percent.POWER CONTROLOne of the main widely investigated aspects ofdefined RRM wireless networks is, undoubtedly,power control. In particular, power control ismainly considered in order to improve systemperformance in a shared scenario. Minimizingpower consumption is the main objective of anypower control algorithm. This leads to maintainingthe necessary QoS by compensating for terminalmovements, fluctuating interference, andchannel degradation due to shadowing or multipathfading. But the benefits of a properapproach can be various; on one hand, lowerpower means lower interference and thereforeincreased network capacity; on the other hand, itreduces the problem of battery life in mobile terminals.Special and more efficient approachescan be applied for multimedia applicationswhose delay requirements are not too stringent,which is the case for packet communications.Indeed, for this kind of application the QoSparameters are mainly related to the BER, whilea certain delay in data transmission can beYesBlockIEEE Communications Magazine • February 2005 99

Quite apart fromwhich solution willovercome, it seemsreasonablethat futurecommunicationssystems will becharacterizedby highlydynamic resourcemanagement andextreme flexibility indeployment andimplementation.accepted and, in some cases, not even perceived.This peculiarity allows further refinement of PCalgorithms suitable for multimedia networks:when observing high interference in the channel,instead of using high power to overcome thisnoise and transmit a packet successfully, it canbe more advisable to postpone transmission,buffer the delay-tolerant traffic, and wait for theinterference to subside before transmittingagain.SCHEDULINGDue to the high number of different applicationsenvisaged for next-generation wireless networks,priority-based management of different queuesmust be supported in order to meet user requirementsin terms of QoS and cope with trafficburstiness. Moreover, the management policy ofevery single buffer must consider the nature oftraffic in terms of delay constraints (e.g., realtimeor non-real-time applications). Then thisapproach requires careful monitoring of bothchannel conditions and buffer occupation inorder to efficiently schedule the different typesof traffic. A scheduling technique is efficient inany CDMA-based scenario where flexible use ofthe spectrum could be achieved and differentQoS constraints must be satisfied.CONCLUSIONSFor future next-generation mobile telecommunications,the goals are high data rates and customizableapplications enabled by flexibletechnologies, assuring for the user a ubiquitoussystem that is able to always guarantee a connection.To realize this aim two possible scenariosare envisioned. One expects the definition of anew standard that substitutes the previous 3Gsystems and eliminates those problems thatmake 3G networks unsuitable for high-data-ratemultimedia applications. Alternatively, a wirelessinfrastructure based on the integration andinteroperability of different wireless systems canbe supposed. Quite apart from which solutionwill overcome, it seems reasonable that futurecommunications systems will be characterized byhighly dynamic resource management andextreme flexibility in deployment and implementation.Hence, CDMA represents one of themost promising technologies for future beyond3G systems. However, to provide enhanced serviceswith high data rates some key technologiesare needed. This article has discussed some ofthem, such as advanced receivers, multipleantennas, adaptive link layers, and resourcemanagement techniques.REFERENCES[1] D. Koulakiotis and A. H. Aghvami, “Data Detection Techniquesfor DS/CDMA Mobile Systems: A Review,” IEEE Pers.Commun., vol. 7, no. 3, June 2000, pp. 24–34.[2] R. Fantacci, “Proposal of an Interference CancellationReceiver with Low Complexity for DS/CDMA MobileCommunication Systems,” IEEE Trans. Vehic. Tech., vol.48, no. 4, July 1999, pp. 1039–46.[3] R. Fantacci, D. Marabissi, and S. Morosi, “PerformanceAnalysis and Optimization of a Space-time Selective PICfor CDMA System,” IEEE Trans. Wireless Commun., vol.3, no. 2, Mar. 2004, pp. 359–66.[4] C. Berrou and A. Glavieux, “Near Optimum Error CorrectingCoding and Decoding: Turbo-codes,” IEEETrans. Commun., vol. 44, Oct.1996, pp. 1261–71.[5] X. Wang and H. V. Poor, Wireless Communication Systems,Prentice Hall, 2004.[6] R. D. Murch and K. B. Letaief, “Antenna Systems forBroadband Wireless Access,” IEEE Commun. Mag., vol.40, no. 4, Apr. 2002, pp. 73–86.[7] L. Hanzo et al., OFDM and MC-CDMA for BroadbandMultiuser Communications, Wiley, 2003.[8] D. J. Costello and S. Lin, Error and Control Coding: Fundamentalsand Applications, Prentice Hall, 1983.[9] D. Chase, “Code Combining, a Maximum LikelihoodApproach for Combining an Arbitrary Number of NoisyPackets,” IEEE Trans. Commun., vol. 33, May 1985, pp.385–93.[10] J. S. Blogh and L. Hanzo, Third-Generation Systemsand Intelligent Wireless Networking, Wiley,2002.BIOGRAPHIESROMANO FANTACCI [M’87, SM’91, F’05] is a full professor oftelecommunication networks and digital communicationsat the University of Florence, Italy. He received the IEE IEREBenefactor Premium in 1990 and an IEEE ComSoc Awardfor Distinguished Contributions to Satellite Communicationsin 2002, and is currently serving as Associate Editorfor IEEE Transactions on Communications and IEEE Transactionson Wireless Communications.FRANCESCO CHITI [StM’01] received a degree in telecommunicationsengineering and a Ph.D. degree in informatics andtelecommunications engineering from the University of Florencein 2000 and 2004. His current research topics aredevoted to MAC, LLC, and network layer protocols for bothpublic and private wireless communications systems as wellas ad hoc and sensor networks. He has taken part in severalEuropean research projects, such as as IP GoodFood,NoEs Nexway and Newcom, and the COST 289 action.DANIA MARABISSI [StM’00] received a Ph.D. degree in informationand telecommunications engineering from the Universityof Firenze in 2004. She joined the Electronic andTelecommunications Department at the University of Florencein 2000. She currently conducts research on physicallayer design for spread-spectrum wireless systems. In particular,her interests include multiple access techniques,multi-user detection, and channel estimation in 3G andB3G mobile communications.GIADA MENNUTI [StM’03] received her degree in telecommunicationsengineering from the University of Firenze in2002 and is currently working toward her Ph.D. at thesame university. Her research interests mainly include MAC,LLC, and network layers, with special attention on resourcemanagement, in wireless communications in 3G and B3Gsystems. She is involved in two Network of Excellence(Newcom, Satnex) projects and the COST 289 actions.SIMONE MOROSI [StM’98, A’01, M’03] received a Ph. D.degree in information and telecommunication engineeringin 2000 from the University of Firenze. Since 1999 he hasbeen a researcher of the Italian Interuniversity Consortiumfor Telecommunications (CNIT). Since 2000 he has beenwith the Department of Electronics and Telecommunicationsof the University of Firenze. His present researchinterests involve CDMA communications, multi-user detectiontechniques, and B3G mobile communications.DANIELE TARCHI [StM’98] received a Ph.D. degree in informaticsand telecommunications engineering in 2004 from theUniversity of Florence. He is involved in several national projects(Rescue and Pattern) as well as European IP (Good-Food) and Network of Excellence (Nexway, Newcom, Satnex)projects. He is involved in COST 289 European Union actions.He has served as a Technical Program Committee memberfor IEEE VTC 2004-Spring and IEEE ICC 2005.100 IEEE Communications Magazine • February 2005