AdaptiveCodedModulationforOFDMSystemOFDM系统的自适应编码调制

JournalofEngineering

Volume18February2012

Number2

AdaptiveCodedModulationforOFDMSystem

Assist.Prof.Dr.SerkoutN.Abdullah

ZainabMageedAbid

AdaptiveCodedModulationforOFDMSystem

Assist.Prof.Dr.SerkoutN.AbdullahZainabMageedAbid

sar_abd2000@

zainab.mageed@

Abstract

ThispaperstudiestheadaptivecodedmodulationforcodedOFDMsystemusingpuncturedconvolutionalcode,channelestimation,equalizationandSNRestimation.Thechannelestimationbasedonblocktypepilotarrangementisperformedbysendingpilotsateverysubcarrierandusingthisestimationforaspecificnumberoffollowingsymbols.Signaltonoiseratioisestimatedatreceiverandthentransmittedtothetransmitterthroughfeedbackchannel,thetransmitteraccordingtotheestimatedSNRselectappropriatemodulationschemeandcodingratewhichmaintainconstantbiterrorratelowerthantherequestedBER.Simulationresultsshowthatbetterperformanceisconfirmedfortargetbiterrorrate(BER)of(10-3)ascomparedtoconventionalmodulationschemes,theconvolutionalcodedmodulationoffersaSNRgainsof5dBcomparedtouncodedstateatBERof10-3.TheproposedadaptiveOFDMschememaintainsfixedBERunderchangingchannelconditions.

الخلاصة

فيهذهالمقالةتمدراسةمنظومةالتقسيمالمتعددالمتعامدللترددمتعددةالسرعةومتكيفةومرمزةباستخدامنوعالترميزPuncturedConvolutionalوتخمينوتعديلالقناةوتخميننسبةالإشارةإلىالضوضاء.تمدراسةطريقةلتخمينجودةالقناةتسمىBlockتعتمدعلىإرسالمعلوماتمعروفةلدىالمستلمعلىكلالتردداتوهيملائمةللقناةالتيتتغيرحالتهاببطءوطريقةلتخميننسبةالإشارةإلىالضوضاء,يتمقياسمستوىالإشارةإلىالضوضاءفيجهةالاستلاموإعادةإرسالهاإلىالمرسلمنخلالقناةالتغذيةالعكسيةالمثاليةاعتماداعلىمستوىالإشارةإلىالضوضاءيقومالمرسلباختيارالأسلوبالمناسبللاتصالالذييتضمنوحدةتضمينمرمزةتحققنسبةخطامحددةمسبقاعلىأنلايتجاوزعتبةنسبةالخطأالمطلوبة.لقدأظهرتالنتائجالعمليةعندنسبةخطا10-3أداءجيدلهذهالمنظومةمقارنةبمنظومةالتضمينالمرمزالغيرمتكيفوكذلكالتضمينالمرمزأعطىتحسنلSNRقدره5dBمقارنةبالتضمينالغيرمرمز.هذهالمنظومةتحققنسبةخطاثابتةتحتظروفالقناةالمتغيرة.

Keywords:OFDM,adaptive,SNRestimation,convolutionalcode,channelestimation

INTRODUCTION

Theideaofadaptivemodulationandcoding(AMC)istodynamicallychangethemodulationandcodingschemeinsubsequentframeswiththeobjectiveofadaptingtheoverallthroughputorpowertothechannelcondition.Infact,whenemployingorthogonalfrequencydivisionmultiplexing(OFDM)overaspectrallyshapedchanneltheoccurrenceofbiterrorsisnormallyconcentratedinasetofseverelyfadedsubcarriers,whichshouldbeexcludedfromdatatransmission.Ontheotherhand,thefrequencydomainfading,whileimpairingthesignal-to-noiseratioofsomesub-carriers,mayimproveothersabovetheaveragesignal-to-noiseratio.Hence,thepotentiallossofthroughputduetotheexclusionoffadedsubcarrierscanbemitigatedbyusinghigherordermodulationmodesonthesub-carriersexhibitinghighersignal-to-noiseratio.Inaddition,othersystemparameters,suchasthecodingrateoferrorcorrectioncodingschemes,canbeadaptedatthetransmitteraccordingtothechannelfrequencyresponse[BenvenutoandTosato,2004].

ADAPTIVECODINGANDMODULATION

Themainconceptofadaptivecodingandmodulationistomaintainaconstantperformancebyvaryingtransmittedpowerlevel,modulationscheme,codingrateoranycombinationoftheseschemes.ThisallowsustovarythedataratewithoutsacrificingBERperformance.Sinceinlandmobilecommunicationsystems,thelocalmeanvalueofthereceivedsignallevelvariesduetothefadingchannel,Adaptivecodingandmodulationisaneffectivewaytoachievehighdataratesandithasprovedtobeabandwidthefficienttechnologytotransmitmultimediainformationovermobilewirelesschannels.Itcanbedescribedasfollows:

Modulationmode=(M1 cq<I1

M2 I1<cq<I2

.

.

.

Mn In-1<cq)(1)

WhereM1,M2...Mn,arendifferentmodulationmodesvaryingfromlowermulti-levelmodulationtohighermultilevelmodulationwithincreasingorder.cqistheestimatedchannelqualityexpressedintermsofthesignal-to-noiseratio(SNR)ofthemobilewirelesschannel,I1,I2,…..,In-1aretheswitchingthresholdsbetweendifferentmodulationmodes.

Theselectionofmodulationmodeforthenexttransmissionheavilydependsonthecurrentchannelqualityestimation.Ifthechannelqualitycanbemeasuredaccurately,idealswitchingbetweendifferentmodesisavailable.Thesystemcouldhavethehighestperformanceundersuchcircumstances,iftheswitchingthresholdsareselectedcarefully.Inotherwords,inthescenarioofnochannelqualityestimationerror,theeffectivenessoftheadaptivemodulationsystemwillbedecidedmainlybytheselectionoftheswitchingthresholds[LongandLo,2003].

Modulationschemeandcodingratearethemostcommonparametersusedinadaptivemodulation.Basicguidelinesforefficientusageoftheseparametersareasfollows[SampeiandHarada,2007]:

ModulationSchemeControl:Modulationschemecontrolistheonlytechniquetoenhancetheupperlimitofspectralefficiencyintermsofbit/s/Hzinsingle-inputandsingle-outputsystems.However,ifonlythemodulationschemeisusedasacontrollablemodulationparameterMP,itsdynamicrangeisnotsowide.Table(1)summarizestheoreticallyobtainedrequiredenergypersymboltonoisespectraldensityEs/NotosatisfyarequiredBER(BERreq)foreachmodulationschemeunderadditivewhiteGaussiannoise(AWGN)conditions,whereGraymappingisassumedandnochannelcodingisemployed.Whenallthemodulationschemesinthistableareselectable,thesystem’sdynamicrangeisabout16dB.Ontheotherhand,whenachannelisinaflatRayleighfadingcondition,thesystem’sdynamicrangeismorethan20dB,whichiswiderthanthatcoveredbyanadaptivelycontrolledmodulationscheme.OnesolutiontothisproblemistointroduceanontransmissionmodeasoneoftheselectableMPsandtochoosethismodewhenthereceivedsignallevelistoolowtoguaranteetransmissionofevenabinaryphase-shiftkeying(BPSK).

Inthiscase,however,theratioofthenontransmissiontimeperiod(outageprobability)increasesastheaveragereceivedsignalleveldecreases.

CodingRateControl:Whencodingratecontrolisintroducedtoadaptivemodulationschemes,becausetherequiredEs/NoforaspecificBERcanbelowered,onecanhavemorechancetousehigheruserratescomparedtononcodingratecontrolledsystemsunderthesamereceivedsignallevel;thus,codingratecontrolenhancesaveragesystemthroughput.Moreover,thecodingratecontrolhasanotheradvantage;onecanreducetherequiredsignal-to-noisepowerratio(SNR)orSINRgapbetweenadjacentMPmodes,becauseanarbitrarynumberofcodingratescanbepreparedbyacombinationofpuncturedcodesandregularbitpuncturing.

TheruleswhichfollowtoChooseMPSetare:

ListallthepossiblecombinationsofmodulationSchemesandcodingrates,andsortthesecombinationsofMPsintheorderoftherequiredSNRorSINR.

IfthereareseveralMPsetsthatcanachievethesamespectralefficiency,asetwiththelowestrequiredSNRischosen.

IfthereareseveralMPsetslocatedverycloselyintherequiredSNR,someofthemcanberemoved.

InFigure(1),thecurvesfromlefttorightrepresenttheBERofQPSK,16QAM,64QAMand256-QAMinAWGNchannel,respectively.Inordertodecidetheproperswitchinglevelsfromthisplot,operatingpoint,ordesiredBERmustbedecided.Inthisstudy,BERof(10-3)isusedasoperatingpoint.ThismeansthatthesystemwilltryandkeepaBERlowerthan(10-3)withthemostspectrallyefficientmodulationschemewheneverpossible.Atthispointspectralefficiencyshouldbedefinedasthenumberofinformationbitsencodedonamodulatedtransmissionsymbol.Forexample,QPSKhasaspectralefficiencyof2bitspersymbol,16QAMhas4bitspersymbol,64QAMhas6bitspersymboland256QAMhas8bitspersymbol.

THEADAPTIVEOFDMSYSTEM

TheproposedadaptiveOFDMsystemusedinthetestisshowninFigure(2).Thesystemconsistsofatransmitter,areceiverandaRayleighcommunicationchannel.Thetransmittercodestheinputdatabytheconvolutionalcoderthatisefficientinthemultipathfadingchannel.Theconvolutionalcoderusesthecoderate(R=1/2)andtheconstraintlength(k=7).Theencodeddataarepuncturedtogeneratehighcoderatesfromamothercoderateof1/2,thecodedserialbitsequencesareconvertedtotheparallelbitsequencesandthenmodulated.TheOFDMtimesignalisgeneratedbyaninverseFFTandistransmittedovertheRayleighfadingchannelafterthecyclicextensionhasbeeninserted.Dopplerfrequencyisassumedtobe5Hz(slowflatfading).Inthereceiverside,thereceivedsignalisserialtoparallelconvertedandpassedtoaFFToperator,whichconvertsthesignalhacktothefrequencydomain.Thisfrequencydomainsignaliscoherentlydemodulated.ThenthebinarydataisdecodedbytheViterbiharddecodingalgorithm.ThesimulationparametersarelistedinTable(2)

Thesystemisoperatingatasamplingrateof20MHz.Ituses64-pointFFT.TheOFDMsymboldurationworth's66samplewhere64isfordatawhile2iscyclicprefix.Thiscorrespondstoefficiencyof(0.96).Usingdifferentmodulationschemescombinedwithpuncturingoftheconvolutionalencoder,5differentdataratearedefined.Datarateisafunctionofthemodulation(QPSK,16-QAM,64-QAMand256-QAM)andthecoderate.Thedatarateiscalculatedusing,

Datarate=(bitscarrier*Ncarriers*CR)/TOFDM (2)

Wherebitscarrieristhenumberofbitspercarrier,i.e.2forQPSK,Ncarriersisthenumberofsubcarrierswithinformation,CRisthecoderateandTOFDMistheOFDMsymbolduration.

TheframelengthisvariableconsistsoffixednumberofpilotsymbolsNpandvariablenumberofdatasymbolsNd.

SIGNALTONOISERATIOESTIMATION

TheadaptivemodulationschemeneedstheaccurateinformationofthemultipathchannelandestimatestheSNRvaluebymeasuringthequantityofnoiseamongthereceivedsignal.Therehavebeentwomethodsofmeasuringthequantityofnoiseamongthereceivedsignalthatpassedthroughthechannel.Thefirstmethodusesthepreviouslyknownpilotsymbolasthereferencesignalthatisspeedyandstablemeasuringmethod.ThesecondmethodusesthedemodulatedsignalasthereferencesignalandhastheBERvaluesufficientlylowfortheaccuratemeasurementoftheSNR.Therefore,thepilotsymbolisnotnecessary.SNRestimationalgorithmisshowninFig.(3).

TheSNRvalueiscomputedbycomparisonbetweenthereceivedsignalpowerandthenoisepowerthat<|x|²>isthereceivedsignalpowerand<|y|²>isthenoisepower[Chu,Parkandetc,2007].

CHANNELESTIMATION

ChannelestimationcanbeachievedbytransmittingpilotOFDMsymbolasapreamble.ThechannelestimationcanbeperformedbyeitherinsertingpilottonesintoallofthesubcarriersofOFDMsymbolswithaspecificperiodorinsertingpilottonesintoeachOFDMsymbol.

Thefirstone,blocktypepilotchannelestimation,hasbeendevelopedundertheassumptionofslowfadingchannel.Thisassumesthatthechanneltransferfunctionisnotchangingveryrapidly.Theestimationofthechannelforthisblock-typepilotarrangementcanbebasedonLeastSquare(LS)orMinimumMean-Square(MMSE).Thelater,thecomb-typepilotchannelestimationhasbeenintroducedtosatisfytheneedforequalizingwhenthechannelchangeseveninoneOFDMblock.Thecomb-typepilotchannelestimationconsistsofalgorithmstoestimatethechannelatpilotfrequenciesandtointerpolatethechannel.Theinterpolationofthechannelforcomb-typebasedchannelestimationdependsonlinearinterpolation.

Assumingisthetransmittedpilotdata,thereceivedsignalafterFFTis:

(3)

Wherew(k)isthenoisecomponents,andsince,thepilotdataisknownatthereceiver,thenthesimplestwaytoestimatethechannelisbydividingthereceivedsignalbytheknownpilot:

Ĥ(k)=Y(k)/P(k) (4)

WhereĤ(k)istheestimateofthechannel,andwithoutnoise,thisgivesthecorrectestimation.Whennoiseispresent,therecouldbeanerror[Omran,2007].

RESULTS

Figure(4)showstheestimator’sperformanceinallmodulationschemes,theSNRestimatorworkswellwithQPSKmodulation.AsshowninFigure(4),forM-QAMmodulationatlowSNR,theestimateisnotgood.ThisisbecauseatlowSNR,thereareagreaternumberofsymbolerrorsthatoccur.Thosesymbolsaretheinputtotheestimator.Themorereliablethesymbolinformationis,thebettertheSNRestimateisforQAMschemes.ThisiswhytheestimatesarebetterathighSNR.

Thelevelsintable(3)areconcludedinthefollowingway:AtanoperatingBERof(10-3),thereisnomodulationschemethatgivesthedesiredperformanceatanSNRbelow6.8dB.Therefore,1/2QPSKischosenasitisthemostrobust.Between6.8and12.5dB,thereisonlyoneschemethatgivesperformancebelow(10-3),andthatis1/2QPSK.Between12.5and15.5dB,1/216-QAMgivesthedesiredBERatabetterspectralefficiencythan1/2QPSK.Between15.5and23dB,3/416-QAMgivesthedesiredBERatabetterspectralefficiencythan1/216-QAM.Between23and30dB,3/464-QAMgivesthedesiredBERatabetterspectralefficiencythan3/416-QAM.AndatSNRhigherthan30dB,256-QAMgivesthebestspectralefficiencywhileprovidingthedesiredBERperformance.

InFigure(7),theperformanceofadaptivecodedmodulationbeginsbyoverlappingtheQPSKcurve.Itisanalogoustothespectralefficiencycurve,asQPSKistheprimarytransmissionmodeusedinlowSNR.However,astheSNRisincreasedto13dB,weseeaninterestingresult.TheperformanceofadaptivecodedmodulationbeginstoimprovebeyondwhatQPSKcanprovide.

Considerachannelthathasadeepfade.Optionsherearetouseoneoffivemodulationmodes,whichdifferinspectralefficiencyandrobustness.Ifthefadingconsideredtobeextremelydeep,perhapshalfofallbitswillbeinerror.Here,itisadvantageoustosendfewerbitsbecausethetotalnumberoferrorswillbedecreased,whichinfluencesbiterrorratesmuchmorethantotalnumberofbitssent.Whenthechannelisnotinafade,thenmanybitsarewantedtobesent.Inthissituation,TheBERisloweredbyincreasingthenumberofbitssentbecauseerrorsbecomelessfrequent.ItisthecombinationofthesetwoprinciplesthatallowstheBERperformanceofadaptivesystemstobemorerobustthanstaticsystemswhilesimultaneouslyprovidingbetterspectralefficiencyatmostrangesofSNR.

InFigure(8),aplotofthespectralefficiencyofadaptivecodedmodulationversusSNRindB.Here,spectralefficiencywillbedefinedtobethenumberofbitssentpermodulationsymbol.

NotethatatlowSNR,thesystemachieves2bitspersymbol,asQPSKisprimarilyused.However,astheSNRincreases,thethroughputalsoimprovesteadily,whichindicatesthatmorespectrallyefficienttransmissionmodeisbeginningtouse.

Thecurvebeginstoleveloutatcloseto30dB,as256QAMbecomesthetransmissionmodeusedmostoftenandQPSKisrarelyused.AsSNRimproves,thesystemismoreabletochoosemoreefficienttransmissionmode.

CONCLUSION

Themainconclusionsdrawnfromthisstudyare:

TheACMschemeenhancestheperformanceoftheOFDMwirelesscommunicationsystemsinceitcombinestwoadaptiveschemesbasedonmodulationandcoding.TheresultsshowthattheACMschemeadjustseffectivelytothechannelenvironmentbecauseitallocates(1/2QPSK)tothedecreasingSNRvalueand(1/216-QAM,3/416-QAM,3/464-QAMand3/4256-QAM)totheincreasingSNRvalue.

Whenchannelhasdeepfadeoneoffivemodulationmodeswhichdifferinspectralefficiencyandrobustnessisused,iffadingisextremelydeephalfofallbitswillbeinerror,itisadvantageoustosendfewerbitsbecausethetotalnumberoferrorswillbedecreased.Whenchannelisnotinafademanybitsaresend,InthissituationBERisloweredbyincreasingthenumberofbitssentbecauseerrorsaresmall.ItisthecombinationofthesetwoprinciplesthatallowstheBERperformanceofadaptivesystemtobemorerobustthanstaticsystem.

TheACMsystemprovidesbetterspectralefficiencyatmostrangesofSNR,atlowSNR,thesystemachieves2bitspersymbol,asQPSKisprimarilyused.However,astheSNRincreases,thethroughputalsoimprovesteadily,whichindicatesthatmorespectrallyefficienttransmissionmodeisused.

Theconceptofadaptivemodulationoptimizesthebandwidthefficiencyforwirelesscommunicationswithoutexcessivecomplexity.

REFERENCES

ButhainaMosaOmran,“IMPROVEMENTTECHNIQUESFORSATELLITEDIGITALVIDEOBROADCASTINGUSINGOFDM”,P.HD.Thesis,inelectronicandcommunicationengineering,UniversityofBaghdad,2007.

F.Y.LongandK.T.Lo,“Measurementsforperformancedegradationofadaptivemodulationschemecausedbychannelqualityestimationerror”IEE,ProceedingCommunicationVol150,No.2,April2003.

FernandoH.Gregorio,“802.11a-OFDMPHYCodingandInterleaving”,HelsinkiUniversityofTechnology.

HyungSukChu,ByungSuParkandChongKooAn,"WirelessImageTransmissionbasedonAdaptiveOFDMSystem",IEEE,2007.

NevioBenvenutoandFilippoTosato,“OntheselectionOfAdaptiveModulationandCodingModesOverOFDM”,IEEECommunicationsSociety,2004.

SeiichiSampeiandHiroshiHarada,“SystemDesignIssuesandPerformanceEvaluationsforAdaptiveModulationinNewWirelessAccessSystems”,ProceedingsoftheIEEE,Vol.95,No.12,December2007.

SinemColeri,MustafaErgen,AnujPuri,andAhmadBahai,“ChannelEstimationTechniquesBasedonPilotArrangementinOFDMSystems”,IEEETransactionsOnBroadcasting,VOL.48,NO.3,September2002.

LISTOFABBREVIATIONS:

ACM :AdaptiveCodedModulation

AWGN :AdditiveWightGaussianNoise

BER :BitErrorRate

LS :LeastSquare

MMSE :MinimumMeanSquareError

OFDM :OrthogonalFrequencyDivisionMultiplexing

QAM :QuadratureAmplitudeModulation

QPSK :QuadraturePhaseShiftKeying

SNR :SignaltoNoiseRatio

Table1TheoreticalRequiredEs/NotoSatisfyaBERrequnderAWGNConditions

Table2simulationparameters

Samplingrate

20MHz

NumberofFFTpoints

64

Numberofcarriers