Joint Co-channel Interference Cancellation and Channel Short

The mitigation of co-channel interference (CCI) and inter-symbol interference (ISI) has been a major area of focus of researchers in wireless communications. An approach of interest separates CCI cancellation and ISI equalization in a slow frequency-select

JointCo-channelInterferenceCancellationandChannel

ShorteningwithSpace-TimeProcessing

RoopshaSamanta

LiteratureSurvey

MultidimensionalDigitalSignalProcessing

WirelessSystemsInnovationsLaboratory

TheUniversityofTexasatAustin,TX78712

roopsha@ece.utexas.edu

Spring2003

Abstract

Themitigationofco-channelinterference(CCI)andinter-symbolinterference(ISI)hasbeenamajorareaoffocusofresearchersinwirelesscommunications.AnapproachofinterestseparatesCCIcancellationandISIequalizationinaslowfrequency-selectiveRayleighfadingchannelintotwostages.Aspace-time lterreducestheCCI,followedbyaViterbiequalizerforISIequalization.ThecomplexityoftheViterbiequalizercanbereducedbyusinga lterforchannelshortening.Iaminvestigatingvariouswaystodesignaspace-time nite-impulseresponse(FIR) lterwhichwouldperformjointCCIcancellationandchannelshortening,whilemaximizingasignal-to-interference-plus-noise-likeobjectivefunction.

The mitigation of co-channel interference (CCI) and inter-symbol interference (ISI) has been a major area of focus of researchers in wireless communications. An approach of interest separates CCI cancellation and ISI equalization in a slow frequency-select

1Introduction

Co-channelinterference(CCI)isaprobleminherenttoallcellularwirelesscommunicationsys-tems.Itarisesduetothebasicfrequencyreuseschemeofcellularsystems.Thismeansthatanumberofcellsinagivencoverageareausethesamesetoffrequencies.Theinterferencebetweenthesignalsfromthesecellsistermedasco-channelinterference.Inter-symbolinterference(ISI)isacharacteristicproblemoffrequency-selectivefadingchannels.Themultipleversionsofthetransmittedwaveform,whichtravelovermultiplere ectivepaths,arriveatslightlydi erenttimesfromtheothers.Thisleadstoaspreadingofthesymbolintime(timedispersion).Someofthesecomponentsarriveduringthenextsymbolduration,causingISI(commonlytermedaschannel-inducedISItodi erentiateitfromtheISIcausedby lters).[1]and[2]arecomprehensivetutorialsonthecharacterizationandmitigationoffadinganditse ectsinmobileRayleighfadingchannels.

ThemitigationofbothCCIandISIhasbeenamajorconcernofresearchersinwirelesscommuni-cations.Thegoalhasbeentodemodulatethedatasymbolswhilemaximizingthesignal-to-interference-plus-noiseratio(SINR),usingequalizersofacceptablecomplexity.

Theareademandsatrade-o betweentheSINRandthecomplexityofthereceiverandtheoptimumsolutionstillremainsanopenproblem.

2Background

VariousapproachestodealwithCCIandISIequalizationhavebeensuggestedovertheyears.Amulti-user-maximum-likelihood-sequence-estimator(MLSE)[3]istheoptimalsolutionformaximizingthesignal-to-interferenceratio(SIR).Butitrequiresthechannelinformationoftheinterfereralongwiththatofthedesireduserandthiscannotbeeasilyestimated.

Whiletherehavebeensubsequentsub-optimalsolutionstotheproblem,oneapproachofinteresthasbeenpresentedbytheauthorsin[4].ThispaperisdiscussedinSection3.Itisbasicallyasingleusertwo-stageapproachforseparateCCIreductionandISIequalizationinaslowfrequency-selectiveRayleighfadingchannel.Aspace-time lterisusedforCCIcancellationinthe rststage.AViterbi

The mitigation of co-channel interference (CCI) and inter-symbol interference (ISI) has been a major area of focus of researchers in wireless communications. An approach of interest separates CCI cancellation and ISI equalization in a slow frequency-select

equalizerperformsISIequalizationinthesecondstage.Thecomplexityofthisequalizerisafunctionofthenumberofstatesinit,whichinturn,isanexponentialfunctionofthelengthofthechannelimpulseresponse.In[4],thecomplexityoftheequalizerisnotaddressed.

Ahelpfulreferenceforequalizerdesignforcellularmobileradiochannelsis[5].Here,theauthorshavepresentedacomprehensivetheoryforoptimumdiversitycombiningandequalizationindigitaldatatransmissionoverfrequency-selectiveRayleighfadingchannels.Variouscombiner-equalizers,whichminimizethemean-squarederror,aredetermined.

Auni edanalysisofoptimumspace-timeequalizerswithlinear ltersoneachantennabranch,followedbyadecisionfeedbackequalizer(DFE)orMLSEisprovidedin[6].

Channelshorteningfordiscretemulti-tone(DMT)transceiversisamuch-researchedareaand

[7]exploresvariousmethodsfordeterminingthecoe cientsofatime-domainchannelshortening lter.VariousrelevantreferencesforchannelshorteningareexploredinSection4.

ChannelmemorytruncationforreducingthecomplexityoftheViterbidecoderwassuggestedasearlyas1973in[8],[9]and1978in[10].Thedi erentapproachestothisproblemarediscussedandcomparedinSection5.

3CCI/ISIReductionwithSpace-TimeProcessing

ThispaperpresentsahybridapproachforseparateCCIreductionandISIequalizationinaslowRayleighfadingchannel.Thesignalisreceivedbyanarrayofantennas,followedbyaspace-time lterasshowninFigure1.The lterisdesignedtomaximizethesignal-to-interference-plus-noiseratio(SINR)byjointlyoptimizingthe ltersweightvectorandthemodi edchannelvector(e ectivechannelimpulseresponse).ThischannelvectorisusedbyaViterbiequalizer,thatformsthenextstage,todemodulatethedatasymbolswithISIequalization.

Thespace-time lterhasthee ectoftemporallycoloringthedisturbance.TheViterbialgorithmperformsoptimallyonlywhentheinputnoiseiswhite.So,atemporalwhitening lterisincludedinthedesignoftheViterbiequalizertotakecareofnon-GaussianresidualCCIandnoise.

The mitigation of co-channel interference (CCI) and inter-symbol interference (ISI) has been a major area of focus of researchers in wireless communications. An approach of interest separates CCI cancellation and ISI equalization in a slow frequency-select

TheauthorsdonotrelyonthechannelinformationofCCI.Atrainingsequenceisusedbythejointoptimizertoprovideaneigenvectorsolutionforthespace-time lterandtheViterbiequalizer.AsdiscussedinSection2,thenumberofreceiveroperationsperdatasymbolisanexponentialfunctionofthelengthofthechannelimpulseresponse,resultinginunacceptablylargereceivercomplexityforalengthgreaterthan4.Thecomplexitycanbereducedbyusingachannelshorteningequalizer.Thisissueisnotaddressedinthe

paper.

Figure1:Theblockdiagramofthehybridapproach.[4]

4ChannelShortening

Thedesignoftime-domainequalizers(TEQ)forchannelshorteningfordiscretemulti-tonetransceivershasreceivedextensiveattentionintheliterature.Akeyreferenceintheareais[7].Inthispaper,theauthorshavedevelopedvariouscomputationallye cientalgorithmsforreducingthee ectivelengthofthechannelforDMTtransceiversusingashorttime-domainFIR lter.

Inthe rstmethod,thealgorithmutilizeseigenvectorsandeigenvaluestogeneratethecoef- cientsofanoptimalshortening lter,giventheoriginalimpulseresponse,thedesiredlengthofthee ectivechannel(ν)andthe lterlength.Theobjectivefunction,whichismaximizedbythisalgo-

The mitigation of co-channel interference (CCI) and inter-symbol interference (ISI) has been a major area of focus of researchers in wireless communications. An approach of interest separates CCI cancellation and ISI equalization in a slow frequency-select

rithm,istheshortening-signal-to-noise-ratio(SSNR),whichistheratiooftheenergyofthelargestνconsecutivesamplestotheenergyintheremainingsamples.Theenergyintheseremainingsamplesisminimizedwhiletheenergyinsidethewindowofinterestisconstrainedtobeunity.

Thismethodcanbeincorporatedintotheapproachin[4]tobuildanobjectivefunctionthatincludesboththeSINRandtheSSNR.Thesolutiontothisjointoptimizationproblemwouldgivethecoe cientsofaspace-time lterthatcanjointlytackleCCIreductionandchannelshortening.

ThedesignofMIMOequalizersforchannelshorteninghasbeeninvestigatedinrecentliterature

[11]and[12].In[11],theauthorderives nitelengthMIMOchannelshorteningequalizersusingaminimum-mean-squared-error(MMSE)criterion(asopposedtothemaximumSSNRapproachof[7]).Thedesignedequalizersalsoperformnoisewhiteningandmulti-channelmatched ltering.[12]presentsanotherapproachtothesameproblem.Theauthorsbuildupanobjectivefunctionthatincludesatrade-o parameterbetweenmaximizingtheSINRandtheSSNR.Aneigenvectorsolutionisobtainedfortheoptimumequalizercoe cientsandthee ectiveMIMOchannel.

Thesolutionin[12]requiresknowledgeoftheautocorrelationsequenceofthenoisevector.Inaddition,thereisnoseparatetreatmentofCCIandISI.

5ChannelShorteningforMLSE

Thespeci cproblemofchannelmemorytruncationforMLSE,implementedbytheViterbialgorithm,wasaddressedinaseriesonpaperswaybackinthe1970’s.Inallthesepapers,thepre-equalizerwasdesignedtominimizethenoisevarianceseenbytheViterbiequalizer.Theydi eredinthechoiceofthedesiredimpulseresponse(DIR)ofthechannelthatwasusedforMLSEbytheViterbialgorithm.

In[8],theDIRischosentobeatruncatedversionoftheoriginalchannelimpulseresponse.In

[9],FalconerandMageeusedaunitenergyconstraintforthe xedlengthDIRtominimizethenoisevariance.Inotherwords,theyusedtheMMSEmethodforchannelshortening.

Thepre-equalizercolorsthenoiseandnoattemptismadetorestrictthecorrelationofthenoise

The mitigation of co-channel interference (CCI) and inter-symbol interference (ISI) has been a major area of focus of researchers in wireless communications. An approach of interest separates CCI cancellation and ISI equalization in a slow frequency-select

ineitherofthesepapers.In[10],theauthor,Beare,de nesane ectivesignal-to-noiseratio(SNR)thatincludesthee ectsofnoisecorrelationontheViterbialgorithm.Hethenproceedstoshowthatthise ectiveSNRismaximizedbychoosingaDIRwhosepowerspectrumcloselymatchesthatoftheoriginalresponseinthemeansquareerror(MSE)sense.

Inthispaper,acomparisonofthethreechoicesoftheDIRhasbeenmadeonthebasisofthee ectiveSNR.ThisisshowninFigure3.TheFalconerandMageechannelmodelisused(Figure

2).

Figure2:FalconerandMageediscretetimechannelA.

[9]

Figure3:E ectiveSNRforvariousDIRschemes.[10]

The mitigation of co-channel interference (CCI) and inter-symbol interference (ISI) has been a major area of focus of researchers in wireless communications. An approach of interest separates CCI cancellation and ISI equalization in a slow frequency-select

ItisobservedthatforDIRsofuptolength4,theBeareschemeachievesthebestresults.Inpractice,aDIRoflength2or3isusuallyallthatisfeasiblebecauseoftheincreaseinthecomplexityoftheViterbidetector.

6RelevantSimulationResults

Theapproachthatisclosesttomyproposedideaistheoneundertakenin[12].ThesimulationresultsarepresentedinFigures4and5.Twodi erentmetricshavebeenusedtocomparethemethodwiththeMMSEmethodof[11],namely,theenergycompactionratio,whichisanindicationofthechannelshorteningachievedandtheoverallSNR.αisthetrade-o parameterbetweentheSINRandtheSSNR,mentionedinSection4.

Iwillbecomparingtheperformanceofmymethodwiththesimulationresultsofthis

paper.Figure4:Energycompactionasafunctionofequal-

izerlength.

[12]Figure5:OverallSNRasafunctionofequalizerlength.[12]

The mitigation of co-channel interference (CCI) and inter-symbol interference (ISI) has been a major area of focus of researchers in wireless communications. An approach of interest separates CCI cancellation and ISI equalization in a slow frequency-select

7Conclusion

CCIreductionandISIequalizationforcellularmobileradiochannels,characterizedbyslowRayleighfadingchannels,hasbeenanactiveareaofresearch.In[2],theauthorsintroducedanoveltwo-stageapproachtothisproblemwhereintheydesignedaspace-time lterforCCIcancellation,whichwasfollowedbyaViterbiequalizerforISImitigation.Shorteningthee ectivechannelimpulseresponsecanreducethecomplexityoftheViterbiequalizer.Givena xednumberofstagesintheViterbiequalizer,space-time lterscanbedesignedtojointlytackleCCIreductionandtheappropriatechannelshorteningwhilemaximizingtheSINR.Themultipleantennasinthereceiverbringthisproblemintotherealmofmulti-dimensionaldigitalsignalprocessing.Iplantoinvestigatethisjointoptimizationproblemandcomeupwithvariouswaystodesignthe lter.Iwillbeexploringtheissueinthetime-domainforthesingle-input-multiple-output(SIMO)channelwherethespace-time lterwillbemulti-input-single-output.ThesimulationswouldbeperformedusingtheMATLABCommunicationsandSignalProcessingtoolboxes.

Iftimepermits,Iwilltrytoextendthesametothefrequencydomaintodealwithmultiple-input-multiple-output(MIMO)orthogonalfrequencydivisionmultiplexing(OFDM)systems.

The mitigation of co-channel interference (CCI) and inter-symbol interference (ISI) has been a major area of focus of researchers in wireless communications. An approach of interest separates CCI cancellation and ISI equalization in a slow frequency-select

References

[1]B.Sklar,“Rayleighfadingchannelsinmobiledigitalcommunicationsystems.I.characterization,”IEEE

Commun.Mag.,vol.35,pp.90–100,July1997.

[2]——,“Rayleighfadingchannelsinmobiledigitalcommunicationsystems.II.mitigation,”IEEECommun.

Mag.,vol.35,pp.102–109,July1997.

[3]S.Wales,“TechniqueforcochannelinterferencesuppressioninTDMAmobileradiosystems,”Proc.Inst.

Elect.Eng.–Commun,vol.142,pp.106–114,Apr.1995.

[4]J.W.Liang,J.T.Chen,andA.J.Paulraj,“Atwo-stagehybridapproachforCCI/ISIreductionwith

space-timeprocessing,”IEEECommun.Lett.,vol.1,pp.163–165,Nov.1997.

[5]P.BalabanandJ.Salz,“Optimumdiversitycombiningandequalizationindigitaldatatransmissionwith

applicationstocellularmobileradio—parti:Theoreticalconsiderations,”http://www.77cn.com.cnmun.,vol.40,pp.885–894,May1992.

[6]G.Arslan,B.L.Evans,andS.Kiaei,“Equalizationfordiscretemultitonetransceiverstomaximizebitrate,”

IEEETrans.SignalProcessing,vol.49,pp.3123–3135,Dec.2001.

[7]P.J.W.Melsa,R.C.Younce,andC.E.Rohrs,“Impulseresponseshorteningfordiscretemultitonetran-

ceivers,”http://www.77cn.com.cnmun.,vol.44,pp.1662–1672,Dec.1996.

[8]S.U.H.QureshiandE.E.Newhall,“Anadaptivereceiverfordatatransmissionovertimedispersive

channels,”http://www.77cn.com.cn.Theory,vol.19,pp.448–457,July1973.

[9]D.D.FalconerandF.R.Magee,“Adaptivechannelmemorytruncationformaximumlikelihoodsequence

estimation,”BellSyst.Tech.J.,vol.52,pp.1541–1562,Nov.1973.

[10]C.T.Beare,“Thechoiceofthedesiredimpulseresponseincombinedlinearviterbialgorithmequalizers,”

http://www.77cn.com.cnmun.,vol.26,pp.1301–1307,Aug.1978.

[11]N.Al-Dhahir,“Firchannel-shorteningequalizersforMIMOISIchannels,”http://www.77cn.com.cnmun.,vol.49,

pp.213–218,Feb.2001.

[12]http://www.77cn.com.cnacenkoandP.P.Vaidyanathan,“Eigen lterdesignofmimoequalizersforchannelshortening,”inProc.

IEEEICASSP,vol.3,May2002,pp.2361–2364.