40射频功率放大器的数字自适应基带预失真技术

上海交通大学硕士学位论文射频功率放大器的数字自适应基带预失真技术姓名：虞强申请学位级别：硕士专业：电磁场与微波技术指导教师：沈海根20070101-67-RFRadioFrequencyPAPowerAmplifierWCDMAWidebandCodeDivisionMultipleAddressMPSKMultiplePhaseShiftKeyOFDMOrthogonalFrequencyDivisionMultiplexingIFIntermedialFrequencyAMAmplitudeModulationPMPhaseModulationIPInterceptPointIMDIntermodulationDistortionACPRAdjacentChannelPowerRatioLOLocalOscillatorDACDigitaltoAnalogConversionADCAnalogtoDigitalConversionLPFLowPassbandFilterLTILinearTimeInvariationNLNonLinearLRLinearRegressionLSLeastSquareSVDSingularValueDecompositionFPGAFieldProgrammableGateArraysASICApplicationSpecificIntegratedCircuitsW-HWienerHammersteinMSEMeanSquaredError-2-WCDMAWIMAXWienerHammerstein-3-WienerMatlabHammersteinWienerHammerstein-4-ADAPTIVEDIGITALPREDISTORTIONTECHNOLOGYFORRFPOWERAMPLIFIERSABSTRACTPoweramplifiersareessentialcomponentsincommunicationsystemsandareinherentlynonlinear.Thenonlinearitycreatesspectralgrowthbeyondthesignalbandwidth,whichinterfereswithadjacentchannels.Italsocausesdistortionswithinthesignalbandwidth,whichdecreasesthebiterrorrateatthereceiver.Toimprovethepoweramplifierefficiencywithoutcompromisingitslinearity,poweramplifierlinearizationisessential.Amongalllinearizationtechniques,digitalpredistortionisoneofthemostcosteffective.Withthedevelopmentofthedigitaltechnology,digitalpredistortiontechnologybecometothemostpotentialtechnologyforRFPAlinearization.Digitalpredistortionimplementationsinthecurrentliteraturemostlyfocusonthepoweramplifierthathasamemorylessnonlinearity.Asthesignalbandwidthgetswider,suchasinWCDMA,poweramplifiersbegintoexhibitmemoryeffects.Thisisespeciallytrueforthosehighpoweramplifiersusedinwirelessbasestations.Asaresult,thecurrentoutputofthepoweramplifierdependsnotonlyonthecurrentinput,butalsoonpastinputvalues.Inotherwords,thepoweramplifierbecomesanonlinearsystemwithmemory.Forsuchapoweramplifier,memorylesspredistortioncanachieveonlyverylimitedlinearizationperformance.Therefore,digitalpredistortersalsoneedtohavememorystructures.Thisdissertationinvestigatesrobustpredistortermodelsthatarecapableof-5-linearizingpoweramplifierswithmemoryeffects.Italsoinvestigatessystemimplementationissuesrelatedtothesewidebanddigitalpredistortionsystems.Amongallofthememorynonlinearmodels,WienerandHammersteinmodelsarethemost“specialized”withtheleastnumberofcoefficientsandeasiestforapplication,butarehardtoidentifyitsparameters.BasedonthecurrentresearchandconsideredthePA’scharacteristics,thehigh-ordersofpastsignalhaslittleeffecttotheoutputsignalwhencomparedtoit’sbasesignal,quasilinearregressionalgorithmwasproposedtoidentifythememoryWienermodelpredistortionsystem.Thenewmethodcouldimproveaccuracyoftheestimationanddecreasethecomputationwhileachievinginsufferingfromthelocalminimumvaluesofotheriterationalgorithm.Andalso,linearregressionalgorithmwasproposedtoidentifythememoryHammersteinpredistortionsystem.Simulationresultsshowthatitcandecreasecomputationrapidly.Itwillimprovesystemaccuracyifusetheestimatedvaluesasinitialvaluesofiterationmethod.Andmore,thisthesisanalyzedthepotentialissuesinpredistortiontechnologyimplement.KEYWORDSNonlinearDistortion,MemoryEffect,ACPR,IMD,WienerModel,HammersteinModel-8-1.11-1[1]Figure1-1Typicalwirelesstransmittersection,showingcompleteupconversionpathfrombaseband1-1WCDMAWimaxMPSKOFDMPAR1090-9-1.2PMCPM7815ItersilISL5239CDMA1-2PredistorterDSPLPFLPFLOPADigitalbasebandinputADCDACFigure1-2DigitalAdaptiveBasebandPredistortionBlock1-2-10-[2]AM/AMAM/PM,WCDMAWIMAX[3][4]WienerHammersteinVolterraVolterraWienerHammersteinDSPWienerGauss-NewtonGaussVolterraVolterraGauss-NewtonGauss-11-Hammerstein/1.3WienerMatlabHammersteinWienerHammerstein7-12-2.1AM/AM2-1)()(1tVKtVinout=2.11K0Vin(t)Vout(t)Slope=K1Figure2-1Idealamplifierwithalineartransfercharacteristic2-1Taylor...)()()()(332210+++=tsatsatsatsiii2.22122322...PPPPTHDn+++=2.3-13-)(0ts)(tsi...,,321aaa21P22322nPPP2-2Figure2-2AM-AMtransfercharacteristicVin(t)0Vout(t)2-2AM-AM)cos()(wtAtsi=2.4AM-PM...)3cos(41)2cos(21)cos()43(21...))cos(())cos(()cos()(33223312233221+⋅+⋅+⋅+⋅+⋅=+⋅+⋅+⋅=wtAawtAawtAaAaAawtAawtAawtAatso2.52331233222)43(...)41()21(AaAaAaAaTHD⋅+⋅+⋅+⋅=2.6ωω3,22-3-14-AmplitudeFrequencyFrequencyff2f3fInputOutputFigure2-3TransferCharacteristiconfrequencydomainof1-toneinputsignalAmplitude2-3))cos()(cos()(21twtwAtsi+=2.7AM-PM)]3cos(41)2cos(43)2cos(43)2cos(43)2cos(43)3cos(41)3cos(41[)]cos()cos()cos(21)cos(211[)]cos()49()cos()49[(...))cos()(cos(())cos()(cos(())cos()(cos()(22121122121332121212223113132132212211twtwtwtwtwtwtwtwtwtwtwaAtwtwtwtwtwtwaAtwaatwaaAtwtwAatwtwAatwtwAatso+++++−++−++⋅⋅+++−++++⋅⋅++++⋅=++⋅++⋅++⋅=2.8-15-Figure2-4TransferCharacteristiconfrequencydomainof2-toneinputsignal2-42-4[5])cos(),cos(21222122twtwaAtwtwaA+⋅−⋅2.9)2cos(43),2cos(43,)2cos(43),2cos(4312122121twtwtwtwtwtwtwtw+−+−2.102.2AM/PM2-5AM-PMTaylorAM-PM-16-InputPowe(dBm)OutputPhaseFigure2-5.AM-PMTransferCharacteristic2-5AM-PM...)()()()(33221+++=ΦtVatVatVatiii2.112.7AM-AM))(cos())(cos()(22110ttwKttwKtsΦ++Φ+=2.122-6AM-PM[6]123405-101-22time,nsecVin,V123405-101-22time,nsecVout,VFigure2-6Outp