An Adaptive Channel Estimation Technique in MIMOOFDM Systems.docx

1、An Adaptive Channel Estimation Technique in MIMO OFDM SystemsAn Adaptive Channel Estimation Technique in MIMO OFDM SystemsJOURNALOFELECTRONICSCIENCEANDTECHNOLOGYOFCHINA,VOL.6,NO.3,SEPTEMBERQ08313AnAdaptiveChannelEstimationTechniqueinMIMOOFDMSystemsPei-ShengPanandBao-YuZhengAbstrac手_Inthispapeanadapt

2、echannelestimationforMooFDMisproposed.AsetofpilottonesfirstareplacedineachOFDMblockthenthechannelfrequencyresponseofthesepilottonesareadaptivelyestimatedbyrecursiveleastsquares(RLS)directlyinfrequencydomainnotintimedomain.Thenaftertheestimationofthechannelfrequencyresponseofpilottones,toobtainthecha

3、nnelfrequencyresponseofdatatones,anewinterpolationmethodbasedonDFTdifferentfromtraditionallinearinterpolationmethodaccordingtoadjaeentpilottonesisproposed.Simulationresultsshowgoodperformanceofthetechnique./ndexTerms-Adaptive,channelestimation,frequencydomain,M1MOOFDM,recursiveleastsquares?1.Introdu

4、ctionOrthogonalfrequencydivisionmultiplexing(OFDM)technologyturnsafrequency-selectivechannelintoaparallelcollectionoffrequencyflatsub-channelstomitigatetheeffectsofintersymbolinterference(ISI).Ontheotherhand,multi-inputmulti-output(MIMO)systemscangainmuchmorespectralefflciency(inbitspersecondperhert

5、byadoptingmulti-antennaonbothendsofthewirelesslinkthansingle-inputsingle-output(SISO).Hence.thecombinationOfM0systemswithOFDMtechnologyprovidesapromisingcandidatefornextgenerationfixedandmobilewirelesssystemsHoweveLaccuratechannelstateinformationsoisrequiredforspace-timedecodenThechannelstateinforma

6、tioncanbeobtainedthroughtwotypesofmethods.Oneiscalledblindchannelestimation,whichexploresthestatisticalinformationofthechannelandcertainpropertiesofthetransmittedsignals.Theotheriscalledpilot-basedchannelestimation,whichisbasedonthepilotdatasentattheManuscriptreceivedOctobor16,2007;revisedNovember16

7、,2007.ThisworkwassupposedbytheNafionaINatureScienceFoundationofChinaunderGrantNo.60372107.P.S.PanandB.-Y.ZhengarewithSchoolofCommunication&InformationEngineering,UniversityofNanjhagUniversityofPostsandTelecommunications,N眦jhag,210003,China(e-mail:panps,zby皿J).transmitterandknownaprioratthereceiv

8、enThoughtheformerhasitsadvantageofnooverheadcost,itneedsalongdatarecordandcomplexalgorithm.Asforthelattersomepilot-basedchannelestimationschemeshavebeenproposed4161.Buttheseschemeshavehighcomputationalcomplexityandmemoryrequirementsandpriorknowledgeofthesecondorderstatisticsofchanne1.Usingpilottones

9、toobtaintheCS1wasdiscussedin【7】,whereanoptimalplacementofthepilottoneswasproposedforMIMOOFDMsystems.In【81,anadaptivecharmelestimationschemewasproposedforMOOFDMsystems.Thisschemeadaptivelyestimatethechannelimpulseresponseintimedomainbasedontheleast-mean-square(LMS)algorithm.Inthispaper,weproposeanada

10、ptivecharmelestimationtechniquewhichadaptivelyestimatethechannelfrequencyresponsedirectlyinfrequencydomainbasedontherecursiveleastsquare(RLS)algorithmoffrequencydomain.Wefirstgiveorthogonalstructureofpilottones.ThenthechannelfrequencyresponseofthesepilottonesisadaptivelyestimatedbyRLSalgorithminfreq

11、uencydomain.Inordertoobtainthechannelfrequencyresponseofdatatones.aninterpolationmethoddifferentfromlinearinterpolationmethodaccordingtoaacentpilottonesisproposed.Here,theestimationofchannelfrequencyresponseisobtainedthroughthechannelimpulseresponseintimedomain,whilethechannelimpulseresponseintimedo

12、mainisobtainedthroughthechannelfrequencyresp0nseofpilottones.Simulationresultshaveshowngoodperformanceforthetechnique.2.SystemModelAsimplifiedblockdiagramofMIMOOFDMsystemswithtwotransmitantennasandtworeceiverantennasisshowninFig.1.Attime,adatablockbn,尼】,k=0,l,一listransformedintotwodifferentsignals:I

13、n,k】:k=0,l,Nl;l,2atthespacetimeencoder,whereMandfisthenttmberoftonesoftheOFDMsystems,toneindex.andantennaindexrespectivelyEachofthesesignalsformsanOFDMblock.Then,thetwotransmittingantennassimultaneouslytransmit0FDMblocksafteranN-lengthIFFToperation.314JOURNALOFELECTRONICSCIENCEANDTECHNOLOGYOFCHINA,V

14、OL.6,NO.3,SEPTEMBER2008Fig.1.MIMOOFDMsystem:(a)transmitterand0o)receiverAtthereceiver,thediscreteFouriertransform(DFT)ofthereceivedsignalateachreceiverantennaisthesuperpositionoftwodistortedtransmittedsignals.TheoutputoftheFFTcanbeexpressedas即,】=In,】五即,】+NIn,纠(1)whereH即,】denotesthechannelfrequencyre

15、sponseatthekthtoneofthenthOFDMblock,correspondingtotheithtransmitandthejthreceiveantenna,N,即,k】denotestheadditivecomplexGaussiannoiseonthejthreceivingantennaandisassumedtobezeromeanwithvariance.Wecanexpress(1)inmatrixformas:xH-lj+N(2)whereViii/=即,0】,1】,即,N一111N=Njn,othereceiveantenna,FdenotestheNxND

16、FTmatrix.ItiseasilyseenthatLtonescanbeusedfortrainingtorecoverthechanne1.LetK=白,bemesetofLtonesusedaspilottones.Wehave旦=(4)where望=】,L?,即,】F=略时FisaVandermondematrix,soitisnon-singular,andtherefore,hcanbefoundbyinvertingit.Oncehisfound,wecanobtainthechannelfrequencyresponserequiredforspacetimedecoder.

17、ThatmeansweonlyneedLpilottonesineachOFDMblocktorecoverythewholechannelequencyresponse.LpilottonesareplacedequallyspacedwithN/Linequencydomain.Pilottonesaredividedinto2groupsfortwotransmitantennasrespectivelytomakethemorthogonal:=,=小?,+.Forthe胁transmitantenna,sendingpilottoneson,sendingzerosignalson(

18、f)andsendingdataonothertonesareshowninFig.2.Time(a)PilotoZerosignaloDataTime(b)Fig.2.Pilotarrangement:(a)antenna1andCo)antenna2.Onpilottonekoftheithtransmitantenna,onlythetoneitselfcansendsignals,then(1)canbesimplifiedasfollowsaccordinglyLn,k:In,kx,In,k+In,k(5)HereAlamoutitllspacetimeencoderisadopte

19、d,sowhenmwehave,Y()=X()H()+N()(6)where,._II口h.Tg叮2.TPANet口:AnAdaptiveChannelEstimationTechniqueinMIMOOFDMSystems315Y()=Yj-,Yjn+1,xcH():H1In,H2+l,N()=NjIn,+1,.ThecorrespondingadaptiveestimationofthechannelfrequencyofpilottonesisshowninFig.3.Fig.3.Adaptivechannelestimationofpilottones()CaDbeupdatedbyu

20、singRLSalgorithmtinfrequencydomainasfollows:E十2():Y()一X()()(7)():()+G()E()(8)GP(:川(一G也(x(P川()(1o)where()isfiltergainmatrix.Initializing,()=o,P0()=81.(11)Aftertheestimationofthechannelfrequencyresponseofpilottones,wecaDobtainthechannelfrequencyresponseofdatatones.By(4)thechannelimpulseresponseintimed

21、omainwithmaximum1engthofchannelhcaDbefound,therbyf3)thechannelfrequencyresponseneededbyspacetimedecodercaDbeobtained.Wenoticethefactthatthelengthofwirelesschannelist-miteaDdusuallyfarsmallerthanthenumberoftonesinaOFDMblock,soweaddawindowtothechannelimpulseresponseintimedomain,thenbyDFToperationtoacq

22、uirethechannelfrequencyresponse.4.SimulationInoursimulationsystem,twotransmitantennasusinga2x2Alamoutispacetimecodeareutilizedfordiversity.TheOFDMblocksizeischosenasN=512addacyctcprefix(CP)withlengthof16isinsertedattheheadofeachOFDMblock.Thefoursub.channelsareassumedtobeindependenttoeachotheraddtoha

23、vemaximum1engthofchannelL=16individually,whicharesimulatedbytheJakesmodel“.Dopplerrespectivelysimulatedasshiftaddcarrierfrequencyarea148HzaDdc3.2GHz.QPSKisapplied.Fig.4showsthattheBERperformanceofthereceiverwithadaptiveRLSchannelestimationisclosetOthatofthereceiverknowingthechanne1.InFig.5.weshowthe

24、convergencebehaviorofourRLSalgorithmsat15dBinthepilottone1.TheMSEresultswereObtainedbvaveragingover1OOsimulationrtms.Bycomparison.theconvergencebehaviorofLMSalgorithmsproposedin【8isalsoshownwiththeadaptiveconstant1.Forconveniencetocomparethechannelfrequencyresponseatpilottone1ObtainedbyourRLSalgorit

25、hmswithLMSalgorithms,thechannelimpulseresponseintimedomainobtainedbyLMSalgorithmsineveryiteration(oneOFDMsymbo1duration)are仃aDSformedintothefrequencydomain.Sonot1osmgjustness,weshowtheconvergencebehaviorofbothRLSaddLMS.ThereasonthattheproposedILSalgorithmconvergesmuchfasterthanLMSliesinitsprocessing

26、doneinfrequencydomainnotintimedomain.024681Ol214161820SNR(dB)Fig.4.Biterrorratev.s.SNRwimDopplershiftis148Hz.050100150200250300350400450500OFDMsymbolFig.5.ExperimentallearningcurvesofRLSandLMS.5000OFDMsymbolsaredealttodemonstratethetrackingforthepilottone1at15dBinFig.6aDdFig.7.Therealcomponentofthef

27、iequencyresponseinthepilottone1aDdtheimaginarycomponentofthefrequencyresponseinpilot1are打ackedinFig.6aDdFig.7respectivelybyRLSalgorithm.Theresultshowsthegoodtrackingforthepilottone.ThisgoodtrackingisduetofasterconvergenceofRLSdoneinfrequencydomainthanLr【250505050505一pJ苫316JOURNALOFELECTRONICSCNCEAND

28、TECHN0【DGYOFCHrNA.VOL.6,NO.3,SEPTEMBER2008LMSdoneintimedomain.ItisclearthatthecomputationalcomplexityorderoftheRLSalgorithmiso(ZM)andthatoftheLMSalgorithmisOLMr,),whereMrdenotesthenumberofreceiveantennas.OFDMsymbolOFDMsymbolFig.6.Comparisonoftherealcomponentofthefrequencyresponseofpilottone1betweent

29、heestimationandrealchannelat15dB.OFDMsymbolOFDMsymbolFig.7.Comparisonoftheimaginarycomponentofthefrequencyresponseofpilottone1betweentheestimationandrealchannelat15dB.5.ConclusionsWeproposeanefficientadaptivechannelestimationtechniqueforMMOOFDMsystems.TwosetsofLpilot.tonesaretransmittedattwoantennas

30、simultaneouslyandthechannelcanbeestimatedemciently.Theadaptivechannelestimationtechniqueisbased0nISalgorithmanddonedirectlyinfrequencydomain.ThemainadvantageisrootedinitsabilitytotrackthechannelfrequencyresponseofpilottonesquicklyandtheacquirementofchannelfrequencyresponseindatatonesbasedonDFTnotlin

31、earinterpolation.References【1】GJ.Foschini,”Layeredspacetimearchitectureforwirelesscommunicationsinafadingenvironmentwhenusingmultielementantennas,BellLabs.Tech.,vo1.1,no.2,pp.41.59,1996.【2】GJ.FoschiniandM.J.Gans,”Onlimitsofwirelesscommunicationsinafadingenvironmentwhenusingmulti.elementantennas,”Fle

32、ssPersonalCommun.,vo1.6,no.3,pp.311-335,Mar.1998.【3】H.Sampath,S.Talwar,J.Tellado,VErceg,andA.Panlraj,Afourth-generationMM0.OFDMbroadbandwirelesssystem:design.performanceandfieldtrialresults,”衄CommunicationsMagazine,vo1.40,no.9,PP.143149,Sep.2002.【4YLi,”Channelestimationfor0FDMsystemswimtransmiaerdiversityinmobilewirelesschannel,IEEEJSAC.vo1.17,no.3,PP.461471,1999.【5】YLi,H.W.mters,andN.R.Sollenberger,”MIM0一OFDMforwirelesscommunications:signaldetectionwithenhancedchannelestimation,”衄Trans.Comm.,vo1.50,no.9,PP.14711477,2002.【6】YLi,”Sim