电子科学与技术外文翻译

科学与技术外文翻译Designofintegrated1.6GHz,2WtunedRFpoweramplifierAbstract：ThispaperdescribesthedesignofanintegratedtunedpoweramplifierspecifiedtooperateatInmarsatsatelliteuplinkfrequenciesfrom1626.5to1660.5basictopologyoftheamplifierliesontheparalleltunedinverseclassEamplifierthatismodifiedbyplacingtheDC-blockingcapacitorintoanewpositionandbyadjustingthesizeofthecapacitortoimprovestabilitybelowthedesiredband.Further,thenewpositioningreduceslossesbetweendrainandload.Thehighcurrentsflowinginthecircuitmadeitnecessarytousewideinductorwidthandhigh-Qfingercapacitorsintheon-chipresonator.TheamplifierwasimplementedasaGalliumArsenide(GaAs)integratedcircuit(IC)thatdelivered2Wofoutputpowerwhilethedrainefficiencywasca.56%.Measurementsincludedsourceandloadpullstofurtherimprovetheperformanceoftheamplifierandtoinvestigatethestabilityatsmallinputdrivelevels.Keywords:InverseclassE•Power-oscillation•Biasnetwork1IntroductionTheusabilityoftraditionallinearamplifiersintoday’shighpowermunicationssystemsislimitedduetotheirlowefficiency.ThisfacthasdriventheinterestofresearchtowardsmoreefficientamplifierssuchasclassE[1–3]andinverseclassE[4].Also,thedemandofhigheroutputpowermeanshigherpeakcurrentsandvoltagesinthedrainorcollectorcircuits.Thiscreateshighrequirementsforbothmaximumbreakdownvaluesofthetransistorandtothepassivecircuitryofthemonolithicmicrowaveintegratedcircuit(MMIC).TheeffectoflimitedconductivityandlimitedcapabilitytocopewithheatcanbeminimizedthroughcarefuldesignofMMIC.Further,emergingtransistortechnologiesseemtowithstandlargercurrentdensitiesandpeakvoltages[5],andtherefore,thechoiceoftechnologyisincreasinglyimportantwhendesigninghighpowerdevices.Theaimofthispaperistoshowexperiencesrelatedtothedesignofswitchinghighpowerradiofrequency(RF)amplifiers(PA)withintegratedoutputpulseshaping.InthesecondchaptertheintroductiontoclassEandinverseclassEoperationisrevisitedandthedifferencesbetweenthetwotopologiesarethirdchapterdescribesthedesignoftheinputandoutputcircuitry,stabilizingcircuitsandprovidessometipstominimizetimingdifferencesattheinputofamulti-fingertransistor.Thefourthchaptershowsthefinalschematicandaphotooftheimplementedchip.Themeasuredperformanceisreportedinchapterfivebyusingbothbasicsingletonemeasurementequipmentandamodernloadpullsystemusingmulti-purposetuners(MPT).Thelastsectionprovidesasummaryofthearticleanddiscussionoftheissuesrelatedtostabilizingcircuits.2ClassEandinverseclassEamplifiersClassEandinverseclassEareregardedasswitchingamplifiers.Ideally,inbothofthemthetransistorisdriveneitheronoroffandthisswitchingoperationproducesaseriesofvoltageandcurrentpulsestotheoutput.Thesepulsesarephaseshiftedandthereforedonotoverlapwitheachother.Idealnon-overlapcausesthetransistortooperatewithdrainefficiencyof100%.ClassicalclassEdrainwaveforms,normalizedtoDCvaluesofsupplycurrentandvoltage,areshowninFig.1.Thesolidlineisnormalizeddraincurrentwaveformandthedashedlineisnormalizeddrainvoltage.TherequirementforoptimaloperationinclassEiszerovoltageswitching(ZVS),wherethedrainvoltageanditsderivativegoestozerojustbeforethetransistorstartstoconduct.IninverseclassEthewaveformshaveswappedplacessothatthesolidlinewaveforminFig.1isthedrainvoltageandthedashedlineisthedraincurrent.Theoptimaloperationisalsochangedtozero-currentswitching(ZCS),wherethecurrentanditsderivativegoessmoothlytozerobeforethetransistorentersnonconductingphase.AdvantagesofinverseclassEoverclassicalrealizationarethatthedrainpeakvoltagesarelowerthaninclassicalclassEandtheinductancevaluesintheoutputcircuitryaresmaller,whichcansaveareainaMMICchipimplementationandcanusuallygivesmallerelectricalseriesresistance(ESR)[4].Also,thepossibilitytoacmondateseriesinductanceasapartofresonatingcircuitryisuseful,sincetheparasiticreactancescancauseundampedresonancestodrainwaveforms[6,7].TheseadvantageswerethereasonsforchoosinginverseclassEtopologyasastartingpointforourinvestigation.However,thetunedimplementationisnottraditionalinverseclassE,althoughithassimilarpulsedoperation.3Designoftunedpoweramplifier3.1GaAsICprocessTheICprocessusedisaTriquintSemiconductor’spseudomorphichighelectronmobilitytransistor(pHEMT)processnamedTQPED.Theprocessutilizesbothenhancementanddepletionmodefieldeffecttransistors(FETs)with0.5lmlengthopticallithographygates,butinourcaseweusedonlydepletionmodetransistors.Theavailabledepletionmodetransistorshaveatransitionfrequency(Ft)of27GHz,drain-gatebreakdownvoltageof15Vandnominalpinch-offpointof-0.8V.TransistorsmodelsusedareTOM3FETmodels.Thereareseveralotherfeaturesintheprocess:nichrome(NiCr)resistorsforprecisionandbulkforhighvalueresistors,highvalueMetal–Insulator–Metal(MIM)capacitors,1localand2thickglobalmetallayers[8].3.2DesignoftheresonatorThedifferencebetweentheoriginalinverseclassEinFig.2andthefinaltunedtopologyusedinourdesign,showninFig.3,isthelocationofblockingcapacitorCs.TheoriginalplacinginFig.2providestheDC-blockingtotwodirections:totheoutput(load)and,moreimportant,itblocksthedirectDC-currentpaththroughLptoourcasetheblockingcapacitorisunderneaththeresonatingcircuitasshowninFig.3,wheretheCsobstructstheflowofDC-currentthroughLptoground,butnottotheoutput(load).Thereisadirectwayforfundamentalcurrenttoflowtotheoutput,withoutpassinganyblockingcapacitor.TheDCblockingcapacitorcannowbemadesignificantlysmaller.Inourcasethereductionwasfrom100pFtolessthan50pF,whichmeanssavingsinchipareaandasasecondaryeffect,theabilitytotuneastabilizingtraptowantedfrequency(moreinchapter3.2)whilemaintaininggoodamplifierperformance.ThedesignoftheDC-blockisnowalsoslightlyeasier,sincepeakcurrentflowingintotheblockingbranchissmaller.Furthermore,theESRbetweendrainandloadissmaller.ThefundamentalcurrentamplitudesinponentsCsandLwere1.2and2A,respectively.ThetotalpeakcurrentsintheparallelresonatorstructurecanbeseeninFig.3.ThetraditionalinverseclassEdimensioning[4]for1.6GHzandPout=3Wresultsinlargechiparea,asduetohighQ=10thecapacitorCtotislarge(63.5pF)and—duetohighpeakcurrents(ca.6A)—theinductorgetsphysicallyhuge.Togetreasonableon-chipponentvaluesthedesignwasgraduallydeviatedfromthedesignprocedurein[4]byshiftingittowardslowerloadresistanceandQvalue,andincreasingtheresonancefrequency.Thisendedupinadimensioningthatprovidesclean,nonoverlappingcurrentandvoltagepulses,reasonablesizepassives,butwhichiseventuallyclosertoclassC–Efundamentalload[9]thantooriginalinverseclassE.Thefinalponentvaluesofthesimulationwithdiscreteponentmodelsandanoff-chiplow-passimpedancematchingnetworkto50Ωresultedinthefollowingdimensioning:resistiveload4Ω,Ctot=30pF,Lp=0:22nH,andLsosmallitcouldbeomittedfromthefinalcouldbereduceddownto50pFwithoutaffectingtheoverallperformance,anditcanbeusedtotuneastabilizingbelowthe-carriernotch,asshownlaterinFig.8.Theoverallsimulationresultswithalargeswitchingtransistor(12paralleltransistorswith18×50µm/0.5µmfingers)estimated5.6Woutputpowerwith72%drainefficiency.Thechallengewasnowtomaintainasgoodoutputpowerandefficiencywhilereplacingtheidealcircuitponentswithprocessdesignkit(PDK)ponentsandwhileaddingsomestabilizingcircuitstothedesignproblemcamewiththephysicaldesignoftheinductor.DespitetheloweredQvaluethecurrentamplitudewasstillsohigh(4.2Apeak)thatca.200lmwidemetallinewasneededfortheinductor,andtokeepthecenterofthe3/4-turninductoropenitcouldnotbemadephysicallysmallerthan0.4nH.Hence,thecapacitanceCtotandQvaluewerefurtherreducedabit,andtoreduceresistivelossesthecapacitanceCtotwassplitinto12parallelhigh-Qcapacitors.Thedrawnlayoutoftheresonatorstructurewasimportedinto2.5Dfieldsimulator,andS-parametersweresimulatedandparedwiththoseofthediscretesimulationprototype.TheunloadedphaseandmagnitudeoftheimpedancedataforparisonsfromS-parametersimulationsareshowninFigs.4and5.Thephaseandmagnitudedataofadistributedresonatorismarkedwithadashedlineinbothfigures.Thephasesandmagnitudesoftheresonatorsfollowalmostthesamethepleteamplifierwassimulated,thedrainefficiencywasabout70%andoutputpowerwasabout3.4W.Thereductioninoutputpowermaybeexplainedbyparasiticresistancesandbytheadditionofstabilizingcircuits.ThedrainefficiencyissurprisinglygooddespitethesomewhatloweredQandempiricaloutputcircuitsimulatedandimplementeddistributedresonatorisshowninFig.6.3.3StabilizingtheamplifierTheamplifiershowedatendencyofinstabilityduringlarge-signalS-parameter(LSSP)simulations.Intheend,stabilityhadtobeevaluatedthroughLSSP-basedstabilitycirclessinceunconditionalstability(K>1)couldnotbeachievedwithoutheavylosses.Stabilitycirclesweredrawnthroughoutafrequencyrangeof0.5–5GHz.Afterseveralsimulations,avarietyofstabilizingcircuitshadtobeusedtopensateringingthediscretecapacitorCswastunedto50pFtogenerateatrapintheoutputresonatoratabout1.2GHzfrequency.ThishelpedinachievingstabilityatfrequenciesbelowthefrequencybandasshownbyRollett’sK-factorinFig.7.The50pFvaluewaschosenforbothsmalldegradationinoutputpowerandforgoodstabilityperformance.TheeffectoftuningofthecapacitorCsisshowninFig.8,wherethecapacitoristunedfrom30to70pF.Further,5ΩofseriesresistancewasaddedtothreegatelinesasshowninFig.9(b)tokeeptheamplifierstablewithoutputstandingwaveratio(SWR)rangeof4.6:1.Also,awidebandRC-sinkcircuitwasincludedintheinputoftheamplifiertoreducethegaininhigherstableoutputSWRrangeincreasedwiththeRCfilterto22.6:1.Accordingtothesimulationstheseriesresistancescausedabout0.46dBgainlossandtheRCfilteragainanadditional0.67dB.Iftheamplifierhadtobeunconditionallystable(K>1),inthefrequencyrangeof0.1GHzto8.0GHz,theincreaseofseriesresistancesto9Ωwouldcauseanadditional0.40dBgainlossandmoreattenuationtothedrivesignal.Thetotaldecreaseofgainduetostabilizationwouldthenbe1.53dB,frommaximumgainof11.36–9.83dB.Intheimplementedform,themaximumsimulatedgainis10.23dB.3.4InputsignaltiminginaphysicallylargetransistorDuringsimulationstherewasanoticeablephaseshiftbetweenextremefingersofthewidetransistorconsistingof12918transistorswithawidthof50lmeach.Thisphaseshiftcausedpartialoverlapbetweenoutputpulsesanddecreasedthedrainefficiency.Atthattimetheinputnetworkwasmadeofaladder-likestructureshownasanexampleinFig.9(a).ThedistanceofthelinebetweenFETAandFETBiscloseto1mm,whichasapurelinedelaywouldresultinabout20psofdelay.Butthedelaydifferencewasmorethan80psandalsothepulsewidthoftheinputsignalwaslargerthanpredicted.Sincethetransistorsdonotswitchsimultaneously,theystarttoloadeachotherandconsumemorepower.Thereasonforincreaseddelayandwidenedpulsewidthisthesignaldependentgatecapacitancethatcausesconsiderableamountofsecondharmonicdistortionintheunterminatedladder-likeinputnetworkinFig.9(a).wherethesignalpathsarealmostequalinimprovedthetimingbehaviourandtheinputwaveformphasinginthesimulationswasnearlythesame.Onlythepulsewidthwasstillsomewhatlarge.Theequalinputroutingincreasedthedrainefficiencyoftheamplifierfrom55to68%.Theeffectsofgatecapacitancetogetherwithadditionalsolutionstotimingproblemshavebeenpublishedin[10].4FinalcircuitTheamplifierdiesized1.96mm×3.62mm(W×L)wasglueddirectlytoa6mmthickaluminiumheatsink.Thegold-platedprintedcircuitboard(PCB)containingoutputmatchingnetworkandsomeofthegatebiasingnetworkwasmountedontotheheatsink.NextthechipwaswirebondedandSMAconnectorswereaddedtothecircuit.ThefinalcircuitschematicisshowninFig.10,wherethedashedlineisusedtoseparatetheon-andoff-chipponents.LowerleftsideinthefigureisanLC-matchingnetworkwhichisfollowedbyanRC-sinkcircuit.TheRC-circuitincreasesthestabilityoftheamplifierbyprovidingawidebandloadingathigherfrequencies.Inthegatebiascircuit,upwardsfromthematchingcircuitistheparallelRLC-circuitthatisahighimpedanceattheoperatingfrequencywhiletheoff-chipparallelRC-circuitprovidesadditionalbiasresistanceinthelowfrequencies,thusincreasingthestabilityoftheamplifier.Ontherightsideofthetransistor,thetunedoutputresonatortogetherwiththeoff-chipmatchingcircuitisshown.Drainsupplyvoltageisdirectedthroughalongtransmissionlinethatisahighimpedanceattheoperationfrequency.TheparallelgateRCbiascircuitandoutputmatchingcircuitwereimplementedonthePCB,whichsimplifiedtheimplementedchipshowninFig.11.Upperleftbox(a)istheLCinputmatching,lowerleftbox(b)istheRLCbiasnetworkandontherightofthebiasisthebox(c)containingtheRC-sinkcircuit.Theequallengthinputlinesareshowninthebox(d),wheretheaddedseriesresistors(5Ωeach)showaswidesectionsinbetweentheequallengthlines.Thetransistorsetisshowninbox(e)anditconsistsof12transistorseachofwhichcontain18fingerswithawidthof50lmeach.Thesaturationcurrentofthetransistorisabout4A.Ontherightfromthetransistorsetthereistheoutputpathtogetherwiththeparallelresonatorinthebox(f).Bondingpadsarebelow(Gatebias),ontheleft(RFin)andup(RFoutanddrainbias).Twoorthreebondwiresareusedtominimizeseriesresistanceandinductanceandalsotomaximizecurrentcapabilityofthewires.4.1TheimplementedamplifierTheimplementedamplifierisshowninFig.12togetherwithapictureofthechiplayout.Theresonatorstructureontherightsideofthelayoutisclearlyvisibleintheactualchip.ThePCBhadtobedrilledopenandthealuminiumbaseplatemachinedforlevellingthechipalongthePCBsurface.Thiswaythebondwiresarekeptasshortaspossible.ThePCBcontainstheimpedancetransformingnetworkrequiredbytheoutputoftheamplifier.Further,thesupplyisprovidedthroughlonglinethathasrelativelyhighimpedanceatthefundamentalandhaslowresistanceatDC.ApartofthegatebiasingnetworkisalsolocatedinthePCB.ThetotalsizeofthePCBis17.1mm×37.6mm(Width9Length).5Measuredperformance5.1MeasurementsetupsAsingletonemeasurementwasusedtomeasuretheamplifieroutputpowerandefficiency.AnIFR2025signalgeneratorandabufferamplifierfromMini-Circuitsprovidedthedrivesignallevelof25dBm.TheoutputwasmeasuredwithaRohde&SchwarzZVA8vectorspectrumanalyser(VSA).TheloadpullmeasurementswereperformedwithFocusMicrowavesMPT1820tunersthatwereappliedbothtotheinputandoutputoftheamplifier.AsasourcewasRohde&SchwarzSMU200Awithabufferamplifier.Theinputpowerlevelswerefrom15to25dBm.TheRFinputandoutputpowersweremeasuredwithAnritsuML2438Apowermeterwithdualinput.TheharmoniccontentofthespectrumandoscillationspikesweremeasuredwithRohde&SchwarzFSQ40VSA.5.2TuningoftheamplifierInthefirstmeasurementstheamplifierdidnotmeetthesimulatedresponse.Measurementsgaveonly0.96Wofoutputpowerat1575MHzwhenthesimulatedfigureswere3.4Wofoutputpoweranddrainefficiencyof70%,allatsupplyvoltageof5.5V.Oursuspiciondirectedtowardsascribelinethatpassedveryclosetotheoutputresonatorstructureandpossiblycouldcoupletheoutputtotheinputoftheamplifier.ThescribelinewascutwithanUV-laserbutthishadnoeffecttothefrequencymeasuredDCcurrentoftheamplifierwasconsiderablyhigherthansimulated,suggestingthattheloadimpedanceoftheswitchingstagewastoolow.Theimpedanceseenatthedrainwasincreasedbyreplacingapairof2.7pFhigh-Qceramiccapacitors(AmplifierA,inTable1)intheexternaloutputmatchingnetworkwithone2.9pFcapacitor(AmplifierB,inTable1).Thismodificationincreasedtheoutputpowerto2Wandthedrainefficiencyto56%atthefrequencyof1625MHz.Theoutputpowerandefficiencyinthefrequencyrangeof1.5–1.7GHzisshowninFig.13.Theoutputpowerismaintainedwithin0.53dBinthedesiredfrequencyrange(1626.5–1660.5MHz)asshowninFig.13.Withinthatsamefrequencybandthedrainefficiencystaysabove53%whilethehighestefficiency,56%,isachievedat1626MHz.Byadjustingthesupplyvoltagetheefficiencyoftheamplifiercanbeincreasedevenmore,whichcanbeseenfromFig.14.Drainefficiencyincreasessteadilywhensupplyislowered.Atasupplyvoltageof2.5Vandfrequencyof1625MHz,theamplifierhasadrainefficiencyof65%.Thisimpliesthattheamplifiercanmaintainanefficientoperationalsowhenusedinanenvelopeeliminationandrestoration(EER)system.ThehighpeaksinatlowestsupplyvoltagesinFig.14arecausedbydrivesignalfeedthroughthatsumsintotheoutputsignal.5.3LoadpullmeasurementsMeasurementswereperformedwithloadpullsystemat1.6GHzspotfrequencytoseveralmodifiedamplifiers.ThedifferencesbetweentheamplifiersareshowninTable1.InthenextchapterswewillmainlyconcentrateonamplifiersCandDforreasonsthatwillbeapparentlateron.LetusnowdiscusstheamplifierCwhichisverysimilartoamplifierBmeasuredearlier.TunersoftheloadpullsystemwereconnectedtotheoutputsandinputsoftheamplifierC.Theloadtuningoffundamental,secondharmonicandthirdharmonicresultedinabout2.4Wofoutputpower(33.8dBm)whilemaintainingabout57.4%drainefficiencyatthispeakpowerspot.ThefundamentalloadimpedanceintermsofpowerwasatslightlyhigherimpedancethantheoptimumdrainefficiencypointasshowninFig.15,wherethe-1dBoutputpowerpoints(triangles)and-5%unitefficiencypoints(circles)areshown.Thepeakefficiencypointinthefigureis(a)(58.6%)andpeakpoweris(b)(33.9dBm).Theoptimumefficiencyareaisratherlarge.Bothoftheloadharmonicswereevenmorerelaxed,anddifferencesforexampleinoutputpowerhadtobemeasuredintenthsofdecibelsratherthanindecibels.Further,theefficiencydifferencesweremeasuredinoneortwopercentageunitsinsteadoffivetoten.Asanexample,thethirdharmonicoptimumoutputpointswithin0.2dBfrommaximum(triangles)andefficiencypointswithin2%unitsfrommaximum(circles)areshowninFig.16.Asitcanbeseen,thethirdharmonicimpedanceisnotascriticalasthefundamentaltone.Theoptimumefficiencyismarkedwith(a)andmaximumoutputpowerismarkedwith(b).Itshouldbenotedthattheadjustmentofthethirdharmonicdidnotincreaseoutputpowerontheabsolutescalenorthedrainefficiency.Thepeakoutputpowervalueremainedwithin0.2dBofthepeakvalueofthefundamentalloadpullandthedrainefficiencyrosefrom58.6onlyto59.8%shownattheSmithchartpoint(a)inFig.16.Theinsensitivityoftheamplifiertoharmonictuningiscausedbythelongdrainbiaslinethatislowimpedanceatthesecondharmonicandthelowpassmatchingnetworkattheoutputthatattenuatesthethirdharmonic.5.4SourcepullmeasurementsThesourcepullofthefundamentalimpedancedidincreasetheoutputpowerandefficiencyofamplifierCslightly.Theoptimumdrainefficiency(circles)andoutputpowerpoints(triangles)(a)and(b),respectively,areshownFig.17.Thepowerpointsarewithinthelimitsof0.2dBandefficiencywithinadifferenceof2%units.Theamplifierefficiencyrosewiththefundamentalsourcetuningto62.1%(pointa)andtheoutputtoabout2.5W(34.1dBm,pointb).Harmonicsourcepullmeasurementsshowedthattheharmonicimpedanceswerenotascriticalasthefundamental.ThisisduetothelowpassinputmatchingandthewidebandRC-sinkcircuit.TheRC-sinkcircuitlowersthecalculatedmagnitudeoftheimpedanceespeciallyathighfrequencies,asshowninFig.18(b).Thishasaneffecttobothsecondandthirdharmonicimpedances.ThemagnitudeoftheimpedancewithouttheRC-sinkisshownasareferenceinFig.18(a).Inbothcasestheinputmatchingcircuitswerenotincludedinthecalculations.5.5StabilityoftheamplifierAtfirsttheamplifierAdidshowsomeunstablebehaviourduetosupplyvoltagemodulationcausedbyinsufficientbiasdecouplingatthedrain.Theinstabilityappearedatlowinputpowerlevelsasnoisesidebandsthatliedonbothsidesoffundamentalfrequency.Wheninputpowerwasloweredfurtheron,theamplifierdidbreakintofullscaleoscillation.Asacure,thesupplyimpedancewasloweredbyalargenumberofdecouplingcapacitors(4×470pF)addedtothedrain.InanEERapplication,thesupplymodulatorwillprovidelowenoughimpedanceatthedrain.WhentheloadpullwasdonetotheamplifiersA,CandD,aspuriousoscillationdetectionwasappliedatalevelof-50dBc.Withthissetupwewereabletoparethesensitivityofdifferentamplifierstooscillations.WefoundoutthattheRC-sinkcircuitusedinamplifiersAandCindeedimprovedstability,especiallyinthelowinputpowerlevels.DatausedforparisonwasmeasuredfromamplifierD,wheretheRC-sinkwascutusinganultravioletlaser.Theoscillationpointsofthefundamentalimpedanceloadpullwithalow15dBminputpowerareshowninFig.19.Theoscillationsidebandsdetectedare1701and1489MHz.IfweparethisresulttoamplifierA,theamountoffoundoscillationpointsisconsiderablysmallerandthelocationofthemisrathertightlyspacedinthelowimpedancearea,asshowninFig.20.Theoscillationfrequencyinthiscaseis1568MHz.TheamplifierAandamplifierDhadadifferentfrequencyforthemodulatingspuriousponents:Withoutthedampingcircuitthemodulatingspuriouswasca.±100MHz,whilewiththedamperthemodulationappearedatabout±30MHz.TheprobablereasonforthisliesindifferentdrainbypassingastheamplifierDhadlesssupplycapacitance(4×470pFless).Theeffectofthiswasstudiedbysimulatingfrominputtoloadtransmission(S21)inthedrainbiasandmatchingnetwork.ThecircuitfromamplifierDshowsresonanceataround93MHzasshowninFig.21(a),whileintheamplifierAthedrainresonanceisat27MHzfrequencyasshowninFig.21(b).Asareference,ameasuredspectrumofamplifierA’soutputisshowninFig.22,wherethemarkersoneandfourareat±33MHzdistancefromthefundamentalfrequency.Theinputpowerinthiscaseis15dBmandfundamentalloadimpedanceisatoneofthefoundoscillationimpedances(Г=0.370,Ø=165.3).Theincreaseofchipdecouplingcapacitors(4×470pF)inthecaseofamplifierAdidlowertheresonancefromaround100MHzinto30MHzregion.Insidethe±30MHzbandtherearealsootheroscillationtoneswhichresembleaquasi-periodicsolutionandachaosspectrumofquasi-periodicandchaosspectrumsareshownforexamplein[11].Aninterestingpointoftheresonanceseenwasthatthelargeelectrolyticcapacitorsappliedtothebiassupplydidnotattenuatetheabout30MHzresonance,duetoitshighinductance.5.6AdditionalfindingsSwitchingamplifiersaredependentonsufficientamountofgatedrive,andvariationsinthedrivesignalaffectquicklytheperformanceoftheamplifier.Inourcasesmallvariationsintransistorpinch-offvoltageresultedin2–3dBdifferencesingainbetweentheamplifierswhenbiasvoltagewaskeptconstant.Suchcleardifferencescouldbeseeninvectornetworkanalysermeasurementswitha0dBminputdrive.Moreconstantresultscouldbederivedbyadjustingthesmalldrainbiascurrentstoequalwhennodrivesignalwasapplied.Nowthemeasurementsofgainwerematchedwithin0.8dB.6SummaryAtunedRFpoweramplifierhasbeendesignedforoperationinafrequencybandof1.6–1.7GHz.Theamplifierwasdesignedempiricallytohavenon-overlappingdrainvoltageandcurrentpulses,andtherequiredresonantcircuitwasimplementedon-chip.AnewpositionoftheDC-blockingcapacitorgeneretedaresonatortrapthatstabilizestheamplifierbelowthedesiredband,andagateRC-sinkcircuitwasusedtostabilizetheamplifierathigherfrequencies.Further,equallengthinputlineswereimplementedtoequalizethetimingofgatesignals.ThestabilityoftheamplifierwasevaluatedthroughsimulationsoflargesignalS-parametersandstabilitycircles.AdditionalresistancestogetherwithanRC-sinkcircuithadtobeappliedtokeepthestableoutputSWRrangeatmorethan22.6:1.TheamplifierwasimplementedontoaGaAssubstratewithdepletionmodehighelectronmobilitytransistors(FETs).Theimplementedamplifierdelivers2Wofoutputpowerwhilemaintaining56%drainefficiency.Thefrequencyresponseiswithin0.53dBatafrequencybandof1626.5–1660.5MHzwhilethedrainefficiencystaysabove53%inthisdesiredband.Theamplifierwasalsomeasuredwithaloadpullsystemataspotfrequencyof1.6GHz.Withthehelpoftunerstheamplifierachievedabout2.5Wofoutputpowerand62%efficiency.Loadpullalsorevealedtheamplifier’ssensitivitytooscillationsatsmalldrivelevels.Itwasfoundoutthattheoutputmatchingnetworkhasalow-frequencyresonancethatmightcontributetotheunstableimplementedgatesinkwasverifiedtostabilizethecircuit.ItmightbeabeneficialideatodesignaninputLC-trapcircuittunedtothesecondharmonic,sincethereissecondharmoniccontentattheinputwhichwidenstheinputwaveforminthetimedomain,creatingproblemsintermsofefficiencyandpower.Further,increasingthethirdharmoniccontentcouldmaketheinputwaveformmoresquare-likewhichisadesiredfeatureinswitchingamplifiers.Acknowledgements：ThisworkhasbeensupportedbyTheAcademyofFinland,InfotechOuluGraduateSchool,TriQuintSemiconductorInc.,NokiaFoundation,TaunoTonningFoundation,UllaTuominenFoundationandThefoundationofRiittaandJormaJ.Takanen.MyspecialthankstothepersonneloftheDepartmentofElectronicsandTelemunicationsintheNorwegianUniversityofScienceandTechnology(NTNU)andtothepersonneloftheMicro-andNanotechnologyCentreintheUniversityofOulu.References1.Cripps,S.C.(2006).RFpoweramplifiersforwirelessmunications(2edn).685CantonStreet,Norwood,MA02062:ArtechHouseInc.2.Raab,F.(1977).IdealizedoperationoftheclassetunedpowerandSystems,IEEETransactionson,24(12),725–735.3.Sokal,N.O.,&Sokal,A.D.(1975).Classe–anewclassofhighefficiencytunedsingle-endedswitchingpowerJournalofSolid-StateCircuits,10(3),168–176.4.Mury,T.,&Fusco,V.(2005).Series-l/parallel-tunedclass-epoweramplifieranalysis.InProc.Europeanmicrowaveconference(Vol.1,p.4).doi:10.1109/EUMC.2005..5.Tayrani,R.(2007).Aspectrallypure5.0w,highpae,(6–12ghz)ganmonolithicclassepoweramplifierforadvancedt/rProc.IEEEradiofrequencyintegratedcircuits(RFIC)symposium(pp.581–584).doi:10.1109/RFIC.2007.380951.6.Hietakangas,S.,Rautio,T.,&Rahkonen,T.(2006).1ghzclasserfpoweramplifierforapolartransmitter.InProc.24thnorchipconference(pp.5–9).doi:10.1109/NORCHP.2006.329232.7.Hietakangas,S.,Rautio,T.,&Rahkonen,T.(2008).OneghzclasserfpoweramplifierforapolarIntegratedCircuitsandSignalProcessing,AnInternationalJournal,54(2),85–94.doi:10.1007/s10470-007-9109-x.8.TriquintSemiconductorInc.(2007).TQPED0.5umE/DpHEMTfoundryservice(2.2edn).foundryTQPEDv2_2.pdf.9.Kazimierczuk,M.,&Tabisz,W.(1989).Classc-ehigh-efficiencytunedpoweramplifier.CircuitsandSystems,IEEETransactionson,36(3),421–428.doi:10.1109/31.17589.10.Hietakangas,S.,Typpo,J.,&Rahkonen,T.(2008).Effectsofinputroutinginswitchedrfamplifiers.InProc.workshoponintegratednonlinearmicrowaveandmillimetre-wavecircuitsINMMIC2008(pp.35–38).doi:10.1109/INMMIC.2008..11.Suarez,A.,Jeon,S.,&Rutledge,D.(2006).Stabilityanalysisandstabilizationofpoweramplifiers.IEEEMicrowaveMagazine,7(5),51–65.doi:10.1109/MW-M.2006.247915.SimoHietakangaswasborninAlaha¨rma¨,Finland,in1980.HereceivedtheM.Sc.degreeinElectricalEngineeringfromtheUniversityofOulu,Oulu,Finland,in2005,andiscurrentlyworkingtowardthePh.D.degreeattheUniversityofOulu.HistechnicalinterestslieinthefieldofanalysisandmodelingofswitchingRFpoweramplifiers.JukkaTyppo¨wasborninOulu,Finland,in1963.HereceivedhisM.Sc.degreeinUniversityofOulu,in1992,andhisPh.D.degreeinNorwegianUniversityofScienceandTechnology,in2003.Currently,heworksasaresearchfellowatNorwegianUniversityofScienceandTechnology,DepartmentofElectronicsandTelemunications.HiscurrentresearchtopicisintegratedRFpoweramplifiers.TimoRahkonenwasborninJyva¨skyla¨,Finland,in1962.HereceivedtheDiplomaEngineer,Licentiate,andDoctorofTechnologydegreesfromtheUniversityofOulu,Oulu,Finland,in1986,1991,and1994,respectively.HeiscurrentlyaProfessorofcircuittheoryandcircuitdesignwiththeUniversityofOulu,whereheconductsresearchonlinearizationanderror-correctiontechniquesforRFpoweramplifiersandA/DandD/Aconverters.1.6GHZ，2W集成式调谐射频式接收功率放大器的设计摘要：这篇论文是对一种指定运行在上行频率为1626.5到1660.5MHZ范围内的国际海事卫星集成调谐功率放大器的描述。这个放大器基本的拓扑结构在于并行优化逆E类放大器，修改这类放大器是通过在一个新的位置安装直流隔离电容并且通过调整电容的尺寸来提高低于预期波段频率的稳定性。另外，新的安放位置减少了排水和负载的损失。在电路中很高的电流使得使用宽电感宽度和在芯片上的谐振器的高Q值的手指电容很有必要。该放大器被实现为砷化镓（GaAs电路（IC）的交付2W的输出功率，而漏极效率为约56％。测量包括信号源和负载以达到进一步改善该放大器的性能，并在小输入驱动电平进行时调查稳定性。关键词：逆E级，功率放大器，自激振荡，偏置网络1.介绍在今天的高功率通信系统传统的线性放大器的可用性是由于其低效率而被限制。这个事实推动了向更高效的放大器的研究兴趣，比如E级[1-3]和逆E级[4]。此外，较高的输出功率的需求意味着更高的峰值电流和电压在漏极或集电极电路。这将创建要求高的晶体管的两个最大击穿值和单片微波集成电路（MMIC）的无源电路。有限的导电性和有限的电容大小，以应付热的效果可以通过仔细设计MMIC而达到最小化。另外，晶体管的新兴技术似乎能承受更大的电流密度和峰值电压[5]，因此，在设计高功率器件时，技术的选择变得越来越重要。本文的目的是要显示与开关高功率射频（RFPA）具有集成输出脉冲整形的设计经验。在第二章中介绍的E级和逆E级操作是重新审视两种拓扑结构之间的差异.第三章介绍了输入和输出电路，稳定的电路的设计，并提供一些提示关于在一个多指晶体管的输入端减少时差的方式。第四章给出了最终的原理和实现芯片的照片。在第五章中，实测性能是通过使用两个基本的单音测量设备和现代化的负载拉移系统采用多功能调谐器（MPT）进行的报告。最后一个部分提供的有关稳定电路的问题的文章和讨论的摘要。2.E级与反E级放大器E级和逆E级被视为开关放大器。理想情况下，在两者的晶体管被驱动打开或关闭，该开关动作产生一系列电压和电流脉冲的输出。这些脉冲相移，因此不会彼此重叠。理想的非重叠将导致晶体管具有100％的漏极效率操作。经典E类DC1极电流的波形，虚线是归一化的漏极电压。在E类为最佳操作的要求是零电压开ZVSE类的波形已经交换了位置，以便在图1中的实线波形是漏极电压，虚线是漏电流。最佳的操作也被改变到零电流开关（ZCSE级相对于经典实现的优点是漏极峰值电压比传统的E级较低，在输出电路中的电感值较小，它可以在一个MMIC芯片实现，节省面积，并且通常可以得到更小的电器串联电阻（ESR[4]。此外，为了接收串联