面对3G的中国PAS市场发展策略研究

南京邮电学院硕士学位论文面对3G的中国PAS市场发展策略研究姓名：陈赛翡申请学位级别：硕士专业：企业管理指导教师：范鹏飞20040401面对3G的中国PAS市场发展策略研究作者：陈赛翡学位授予单位：南京邮电学院相似文献(10条)1.外文期刊TaijunLiu.SlimBoumaiza.FadhelM.GhannouchiDynamicBehavioralModelingof3GPowerAmplifiersUsingReal-ValuedTime-DelayNeuralNetworksInthispaper,weproposeanovelreal-valuedtime-delayneuralnetwork(RVTDNN)suitablefordynamicmodelingofthebasebandnonlinearbehaviorsofthird-generation(3G)base-stationpoweramplifiers(PA).Parameters(weightsandbiases)oftheproposedmodelareidentifiedusingtheback-propagationalgorithm,whichisappliedtotheinputandoutputwaveformsofthePArecordedunderrealoperationconditions.Time-andfrequency-domainsimulationofa90-WLDMOSPAoutputusingthisnovelneural-networkmodelexhibitagoodagreementbetweentheRVTDNNbehavioralmodel'spredictedresultsandmeasuredonesalongwithagoodgenerality.Moreover,dynamicAM/AMandAM/PMcharacteristicsobtainedusingtheproposedmodeldemonstratedthattheRVTDNNcantrackandaccountforthememoryeffectsofthePAswell.Thesecharacteristicsalsopointoutthatthesmall-signalresponseoftheLDMOSPAismoreaffectedbythememoryeffectsthanthePAslarge-signalresponsewhenitisdrivenby3Gsignals.ThisRVTDNNmodelrequiresasignificantlyreducedcomplexityandshorterprocessingtimeintheanalysisandtrainingprocedures,whendrivenwithcomplexmodulatedandhighlyvaryingenvelopesignalssuchas3Gsignals,thanpreviouslypublishedneural-network-basedPAmodels.2.外文期刊PeterClarkeRFdesignteamtoutsCMOSspinfor3GPAsIsitpossibletomakealow-costCMOSpowerampfor3Gcellphonesthatcanoperatewiththepowerefficiencyandrobustnessofagallium-arsenidePA?AccoSemiconductor(SaintGermainenLaye,France)believesithasfoundaway.Butobserversquestionedwhetherthecompany'sMASMOStechnologycandeliveraspromised.AndoneanalystsaidtheissueofCMOSPAsfor3Gcouldbemootasmorefront-endmodulesappearthathousethepowerampalongwithfilteringandaCMOScontroller.CMOShasbeenusedforGSMcellphonepowerampsandforthePArequirementsofBluetoothandwirelessLANs.Butforthemore-complexmodulationschemesandhigherfrequenciesofevolved3GandWiMax,GaAsorsimilartechnologyhasbeennecessary.3.外文期刊AndrevanBezooijenPAs:Migratingfrom2Gto3GConsideringthepressuretomaintainbackwardcompatibilitywithinthecellularindustry,seamlessmigrationfrom2Gto3Gisnotaluxury,it'samandate.Itiscommonknowledgethatdifferentregionsoftheworldemployseveraldifferenttypesofcellularsystems.Inmanyplaces,morethanonesystemisusedwithintheregionsimultaneously.Inthenearfuture,mostcountrieswillbeginagradualmigrationfromexistingsecond-generation(2G)systemstothenext-level,third-generation(3G)systems.Tomakethingsevenmorecomplicated,bothsystemsmustcoexistforatleastafewyears.Forthisreason,andbecausenosinglestandardwillachieveworldwidecoverageinthenearfuture,handsetsthatcanconnecttomultiplesystemswillberequired.Suchhandsetswillprovideaccesstothefeaturesofthe3Gnetworkswhereavailable,whilestillprovidingbackwardcompatibilitywith2Gnetworks.Thisimpliesthattherewillbeademandformultimodeandmultibandhandsets,whichextrapolatesintoademandformultimode,multibandpoweramplifiers(PA)aswell.4.外文期刊Luongvinh.D..Kwon.Y.AFullyRecurrentNeuralNetwork-BasedModelforPredictingSpectralRegrowthof3GHandsetPowerAmplifiersWithMemoryEffectsEfficientandaccuratebehavioralmodelsofpoweramplifiers(PAs)withmemoryeffectsareimportantforpredictingthedistortionsgeneratedbyPAsin3Ghandsets.Conventionalrecurrentneuralnetwork(RNN)hasbeenappliedforRFPAs,butitscapabilitytomodelPAswithmemoryeffectshasnotbeeninvestigated.Inthisletter,weproposeanewfullyRNNwithGammatapped-delaylinessuitableformodelingthedynamicbehaviorof3GPAswithmemoryeffects.Afterbeingtrainedwithwidebandcodedivisionmultipleaccess(W-CDMA)(3GPPUplink)signals,theproposedmodelisvalidatedwithnotonlyW-CDMAbutalsohigh-speeddownlinkpacketaccess(3GPPUplink)signalswithhigherpeak-to-averageratios(PARs),whichdemonstratesthegeneralityofthemodel.ThecomparisonswithpreviousRNNmodelsshowthattheproposedmodeloffersimprovedperformanceinpredictingspectralregrowthbyreducingerrorsby1.7-4dB5.外文期刊MagnusIsaksson.DavidWisell.DanielRonnowAComparativeAnalysisofBehavioralModelsforRFPowerAmplifiersAcomparativestudyofnonlinearbehavioralmodelswithmemoryforradio-frequencypoweramplifier(PAs)ispresented.Themodelsarestaticpolynomial,parallelHammerstein(PH),Volterra,andradialbasis-functionneuralnetwork(RBFNN).TwoPAswereinvestigated:onewasdesignedforthethird-generation(3G)mobiletelecommunicationsystemsandonewasdesignedforthesecond-generation(2G).TheRBFNNreducedthetotalmodelerrorslightlymorethanthePH,buttheerroroutofbandwassignificantlylowerforthePH.TheVolterrawasfoundtogivealowermodelerrorthandidaPHofthesamenonlinearorderandmemorydepth.ThePHcouldgivealowermodelerrorthanthebestVolterra,sincetheformercouldbeidentifiedwithahighernonlinearorderandmemorydepth.Thequalitativeconclusionsarethesameforthe2Gand3GPAs,butthemodelerrorsaresmallerforthelatter.Forthe3GPA,astaticpolynomialgavealowmodelerroraslowasthebestPHandlowerthantheRBFNNforthehardestcrossvalidation.Themodelswithmemory,PH,andRBFNN,showedbettercross-validationperformance,intermsoflowermodelerrors,thanastaticpolynomialforthehardestcro