Behavioral-Modeling-and-Predistortion

52December2009DigitalObjectIdentifier10.1109/MMM.2009.9345161527-3342/09/$26.00©2009IEEEDigitalObjectIdentifier10.1109/MMM.2009.934516Wirelesscommunicationsystemsarecontinuous-lyevolvingtosupportmoreusersandprovidehigherdatarateswithinthelimitedandovercrowdedradio-frequen-cy(RF)spectrum.Toensurespectralefficien-cy,compactconstellationsandmultipleaccesstechniquesbasedoncodedivisionmultipleaccess(CDMA)andorthogonalfrequencydivisionmultiplexing(OFDM)areused.Yet,thisspectralefficiencyisachievedattheex-penseofsignificantlytougherrequirementsontheperformanceoftheRFfrontend.In-deed,theresultingnonconstant-envelope,amplitude-modulatedsignalsunavoidablystimulatethetransmitters’nonlinearities.Thisnonlinearbehaviorismainlygeneratedbythepoweramplifier(PA)operatingclosetosaturationforpowerefficiencyconsider-ations.OperatingthePAinlinearmodewithsuchsignals,whichhavehighpeak-to-aver-agepowerratio(PAPR),willresultinsignifi-cantlylowpowerefficiencies.Accordingly,thedesignofthepoweramplificationstageisoftenperceivedasbeingdrivenbyatradeoffbetweenlinearityandpowerefficiency.Currentstate-of-the-artpoweramplifi-cationsystemsoftenuseaDohertyPAforFadhelM.Ghannouchi([e-mail])isaDistinguishedMicrowaveLectureroftheMTTSocietyandaFellowoftheIEEE.OualidHammi([e-mail])iswithGreenRadioTechnologiesInc.,Calgary,AB,CanadaandisaMemberoftheIEEE.TheyarebothwithiRadioLaboratory,ElectricalEngineeringDepartment,SchulichSchoolofEngineering,UniversityofCalgary,Calgary,AB,Canada.FadhelM.GhannouchiandOualidHammi©COMSTOCKFOCUSEDISSUEFEATUREAuthorizedlicenseduselimitedto:Alcatel-Lucent.DownloadedonApril15,2010at07:32:53UTCfromIEEEXplore.Restrictionsapply.December200953highefficiencyatoutputpowerbackoff,andadigitalpredistortertorestoretherequiredlinearityperfor-mance.Digitalpredistortioniscurrentlythepreferredlinearizationtechniqueandiswidelyusedforapplica-tionswith,typically,upto20MHzbandwidth.DigitalpredistortionconsistsofapplyingacomplementarynonlinearityupstreamofthePAsothatthecascadeofthedigitalpredistorterandthePAbehavesasalinearamplificationsystem.Inthiscontext,behavioralmod-elingiscrucialtopredictthenonlinearityofthePAinparticularandthetransmitteringeneral.Moreover,behavioralmodelingisevenmoreimportantsincepredistortioncanalsobeconsideredasabehavioralmodelingproblem.ThesynthesisofthepredistortionfunctionisequivalenttothebehavioralmodelingofthePA’sreversefunctionobtainedbyswappingthePA’sinputandoutputsignalswithappropriatesmall-signalgainnormalization.ThekeyadvantageofbehavioralmodelingresidesinthefactthatitdoesnotrequiredeepknowledgeoftheRFcircuitphysicsandfunctionality.AsillustratedinFigure1,behavioralmodelingsimplifiesthemodel-ingoftheRFcircuittotheidentificationofamathemat-icalformulationthatrelatestheinputandtheoutputofthedeviceundertest(DUT)thatcanbeconsideredasablackbox.Accordingly,behavioralmodelingappearsasatime-andresource-efficientprocessfortransmit-terperformanceevaluationanddigitalpredistorterdesign.Theperformanceofbehavioralmodelingisinfluencedbytwokeyaspects,theobservationandtheformulation.Theobservationreferstotheaccurateacquisition•ofthesignalsattheinputandoutputoftheDUTwhileexcitingtheappropriatebehavior.Theformulationcorrespondstothechoiceofa•suitablemathematicalrelationthatdescribesallthesignificantinteractionbetweentheDUT’sinputandoutputsignals.Accordingly,alimitedaprioriknowledgeoftheDUTisrequiredtoobservetherightbehaviorandchoosetheadequateformulation.Thisarticlecomplementspreviousworkover-viewingandcomparingbehavioralmodels[1],[2].ItaddressesthemajorissuesrelatedtothestateoftheartofbehavioralmodelinganddigitalpredistortionofPAsfor3G1wirelesscommunicationinfrastruc-tures.Inthisarticle,digitalpredistortionispresentedfromabehavioralmodelingperspectivewithanemphasisonthesimilaritybetweenthebehavioralmodelingandthedigitalpredistortionprocesses.Thus,thisarticlefocusesondistortioncancellationbasedondigitalpredistortiontechniquesanddoesnotcoverotherlinearizationtechniques.First,vari-ouskeyaspectsthatneedtobeconsideredforaccu-rateobservationoftheDUTbehaviorarediscussed.Then,athoroughreviewofstate-of-the-artbehavioralmodelsispresented.Modelperformanceassessmentisdiscussedandacomparisonbetweenstate-of-the-artmodelsperformanceandcomplexityiscarriedout.Finally,asoftwaresolutionforclosed-loopdigi-talpredistortionisbrieflyintroduced.DigitalPredistortionfromaBehavioralModelingPerspectiveTheblockdiagramofadigital-predistortion-basedlin-earamplificationsystemispresentedinFigure2.Thesignalattheinputofthedigitalpredistorter(xin_DPD(n))canbederivedfromthatattheoutputofthePAstageaccordingtoxin_DPD1n25xout1n2G,(1)wherexout(n)andGaretheoutputwaveformandthegainofthelinearizedPA,respectively.Accordingly,whilethebehavioralmodelingoftheDUTconsistsofidentifyingthefunctionfDUTthatsat-isfiesfDUT(xin(n))5xout(n),thesynthesisofthedigitalpredistortionfunctionisequivalenttotheestimationofthefunctionfDPDsuchthatfDPD1xout1n2@G25xin(n).Twoapproachescanbeconsideredtosynthesizethepredis-tortionfunction.Thefirstconsistsofsynthesizingandth