MOSFET雪崩能量计算方法

AvalancheCharacteristicsandRatingsofPowerMOSFETGiovanniPriviteraProduct&ApplicationengineeringPowerMOSFETDivisionSTMicroelectronicsCataniaItaly1.1IntroductionBackinthemid80's,powerMOSFETmanufacturersstartedtoclaimanewoutstandingfeature:theAvalancheRuggedness.Suddenly,newfamiliesofdevicesevolved,allwiththis“new”feature.Theimplementationwasquitesimple:theverticalMOSFETstructurehasanintegralbodydraindiode,whichcannotbeeliminated.So,bychangingsomeprocessandlayoutparameters,itispossibletoguaranteetheuseoftheclampingcapabilityofthisdiodeinwithstandingaccidentalvoltage/powersurgesbeyondthenominaldrainsourcevoltage,Ofcoursetheconfusionaboutthemeaningofruggedness,andhowtoratethisinthedatasheetwassohuge,coupledwiththepoortheoreticalknowledgeofit.Despitethis,allPowerMOSFETmanufacturersstartedtoproduceavalancherateddevices,proposedatasheetratings(althoughimperfect),toprotectthemselvesandtheendusers,fromthisincompleteknowledge.Today,knowledgeondevice’sbehaviorduringavalancheconditionsisenhanced.Alotofapplicationnotesandpaperswereissuedwithdifferentapproachestoexplainratingsandavalanchebehavior.ThescopeofthisnoteistobrieflyreviewtheMOSFETphysicsonavalanchetosupplydesignerswithtoolsandhintstodealwithavalancheissues.1.2MOSFETfundamentalsThebasicandsimplifiedverticalstructureofaMOSFETisdrawninFigure1.TheactualMOSFETisaninfiniteparallelofthese'microscopic'structuresthatworktogethersharingthesameDRAIN,withalltheGATEsconnectedtogetherbyadeposedpolysiliconmeshandalltheSOURCEslinkedbythetopmetal.NN--PP++NN++SourceGateDrainDCRpSGNN--PP++NN++SourceGateDrainDCRpSGFigure1MOSFETverticalstructureandparasiticelementsFigure1referstothewellknownSTpatentedhighvoltageMOSFETstructure,MESHOVERLAYTMexceptsomeprocessoptimizationoftheshapeofthebody-drainjunctionandotherimportantimprovementsintheMESHoverlaydesign.Theconceptofthisverticalstructurecouldbeconsideredvalidalsoforvariousoldercellularorothertechnologies.Duringonstate,whilethegatesourcevoltageisabovethethreshold,theconductioncurrentislocalizedinthedrainandintheregionbelowthegate(channel).DuringoffstatetheVoltagedropacrossdrainandsourceissustainedbythePNjunctionatreversebias,andaverysmallcurrent(leakage)flowsthroughthejunction.Ifthevoltageincreasestoomuchandtheelectricalfieldreachesthecriticalvalue,thejunctiongoestobreakdown,andcurrentstartstoflowthroughthebodyregion.So,ifanover-voltageisappliedtothejunction,acurrentflowsthroughitwhiletheMOSFETlimitstheactualdrain–sourcebreakdownvoltage.Thebreakdownmechanismitself,isnotdestructiveforap-njunction.However,overheatingcausedbythelargebreakdowncurrentandhighbreakdownvoltagedamagesthePNjunctionunlesssufficientheatsinkingisprovided.PAvalanchecharacteristicsandratingsofPowerMOSFET2LookingatthestructureoftheMOSFET,onecanseethatthePNjunctionisnotasimplenorperfectdiode.ThediodeoftheMOSFETisthecollectorbasejunctionofaBJT(BipolarJunctionTransistor,alsocalledtheparasitictransistor)madebytheN+regionofsource,P/P+regionofthebodyandN+regionofthedrain,withthebaseshortedtotheemitterbythefrontmetal.ThecapabilityofaMOSFETtowithstandavalancheconditiontakesintoaccountthesetwoconcerns.Infact,twokindsoffailurearise:onerelatedtocurrent,andtheothertopowerdissipation.Intheformer,failureiscausedbythelatchingoftheparasiticbipolarduetothecurrentthatflowsthroughitsbaseresistance,multipliedbythegain.Thesecondisreachedwhenthetemperatureofthejunctionrisestoacriticalvaluethatprovokestheformationofhotspotswithaveragetemperaturesabout650°Candpeakofapproximately1000°Ccausedbyregenerativethermalrunaway,causingtheextremelyrapiddestructionofthedevice.1.2.1FailureModeDescriptionsAspreviouslydiscussed,theintegraldiodeofaMOSFETisthecollectorbasejunctionoftheparasitictransistor.Ifthecurrentflowslaterallythroughregionp,theincreaseinthevoltagedropacrosstheemitterbaseresistancecausestheBJTtoturn-on.Theinitialavalanchecurrentisconcentratedmainlyinthediodelocalizedinthedeepzoneofp+;assoonasthecurrentgrows,itbeginstointerestalsothep,lighterdoped,regions.Since,bydesign,thevalueoflateralresistanceRpishigherthantheoneoftheverticalresistanceoftheheavydopedp+region,andthecurrentisconcentratedintheregionp+,sotheBJTshouldnotturnon.Assoonasthecurrentbeginstointerestthepregion,causingasufficientdropofvoltagetoequaltheVBEoftheBJT,thecurrentofthebase,Ib,inconjunctionwiththeβofthetransistorwillcausetheBJTturn-on.VBEhasanegativetemperaturecoefficientconsequentlyleadingtothermalrunawayandfinally,thedestructionofthedeviceduetothesecondarybreakdownoftheparasiticBJT.TheadoptionofastronglydopedP+region,determiningthereductionofthegainofthetransistorandthebaseresistancehasbeenthefirststepfortheimprovementoftheMOSFET,followedbyothermoresubtleoptimizations.ThepowerthatisdissipatedintheMOSFETcausesanincreaseinjunctiontemperature.Ifthetemperatureincreasestoacriticalvaluesetbythepropertyofthesilicon[2],thefailure,withoutthecontributionoftheparasiticbipolar,occursbecauseofthecreationofthermallygeneratedcarriersintheepitaxyal/bulkregionandsothecreationofhotspots.Thecriticaltemperaturetohavethisphenomenonisbeyondthemaximumjunctiontemperatureofthedevicesandisrelatedtotheintrinsictemperatureofdopedsilicon,towhichtheconcentrationofthebulkequalstheoneofthethermalgeneratedcarriers.Thetemperatureincreaseduringavalanchephenomena,duetothermalcapacitanceofthesilicon,isnotinstantaneous.Hence,thiskindoffailureshouldbedistinguishedfromthatcausedbycurrentasthedeviceholdsthebreakdownvoltageforafinitetimebeforedestruction.1.3TestingtheAvalancheRuggednessTheAvalanchecapabilityofthedeviceisclassicallyevaluatedbyacircuitthatperformsanUnclampedInductiveSwitching(UIS)liketheonedescribedinFigure2.Figure2UISreferencediagramTheoperationisthefollowing;atzerotimethedeviceswitcheson,closingthecircuit.Duetothepresenceofaninductance,(consideringsomeresistanceduetothelayoutandtheONresistanceoftheMOSFET)thecurrentincreasesfollowinganRAvalanchecharacteristicsandratingsofPowerMOSFET3exponentiallaw,asafunctionoftheL/Rcharacteristicsofthecircuit.Figure3TypicalUISwaveformsAssoonasthedeviceisswitchedoff,asthemagneticfieldintheinductancecannotinstantaneouslygotozero,thedi/dtcausesanovervoltageonthedrainofthedevice.Naturally,thedeviceispracticallyanopencircuituptoitsownblockingvoltage,thereforetheextravoltageislimitedbytheBVDSSoftheDUT.Duringtheavalanche,thecurrentflowsthroughtheDUT,dissipatingtheaccumulatedenergythatwasstoredinthecoilduringthecharging.Table1explainsseveralrelationsforthetav,EavandthePavgwithseveralcircuitconfigurations.Figure4ConstantcurrentavalanchefixtureAdeviceiscommonlydefinedrugged,oravalanche-rated,ifatsomestatedconditionsofcoilandconductedcurrentitsurvivesthistest.Inthepast,othercircuitsweresuggestedtotestthisdevicecapabilityliketheoneinFigure4.Thecurrentismaintainedconstantforasettimeeliminatingthedependencyfortheenergytobewithstoodonthecoil,butactuallytheonlyrecognizedmethod(JEDECstandardNo.24-5,MIL-STD750Dmethod3040.2)isthecircuitdescribedinFigure2.+-VDDZENERDUTL+-VDDZENERDUTLFigure5disconnectedsupplyUISfixtureAcircuitcommonlyusedtotesttheAvalancheruggednessoftheMOSFETisshowninfigure5.IthasaspecialfeatureofapowerswitchinseriestotheVDDthatconnectsthevoltagesourcetothecircuitonlyduringthecoilcharging,disconnectingitafewmicrosecondsbeforetheswitch-offandtheavalancheoperation.ThistechniqueallowstoincreasetheVddbeyonditsmaximumratedVDS,speedsupthechargeofthecoilduringturnon,andconsequentlydecreasestheturnonstatetime.Also,theenergydissipatedisdifferentasonecanreadinthetable1.1.4DatasheetRatingsWhenthedeviceisclassifiedas“AvalancheRated”,thedatasheetprovidestheend-usersomeusefulparameters,whichdefinetheratingsofthedeviceduringavalanche:-Iar,definedasthemaximumcurrentthatcanflowthroughthedeviceduringtheavalancheoperationswithoutanyBJTlatchingphenomenon.ThisMaximumlimitmustbeconsideredasanabsolutemaximumrating.EvenifthecriticalcurrenttobringthedevicetofailureishigherthantheIAR,theproducerguaranteestheoperationofthedevicebelowthislimit.Besides,itisusuallytestedforseveralmicroseconds.AllSTMicroelectronics’HighVoltagePowerMOSFETaretestedaccordingtotheIar.Alltheavalancheoperations(singleeventorrepetitive)belowthiscurrentvaluecanbeconsideredsafeunlesspowerdissipationissues.ItisimportanttounderlinethatforMOSFETsconnectedinAvalanchecharacteristicsandratingsofPowerMOSFET4parallel,thecurrentthatisswitchedduringtheavalanchephenomenonisnotshared,differentlyfromtheoperationsinconductionstate.Infact,atturnoff,onlythedevicewiththelowerbreakdownand/orwiththefasterswitchwillgointoavalanche,withstandingthetotalcurrentthatduringtheonstateissharedwiththeotherMOSFETsinparallel.IfsuchcurrentismorethantheIar,thedevicecanfail.Eveniftheenergyassociatedtothateventisverylow,failureisduetotheactivationoftheMOSFET’sparasiticbipolar.-EAS(EnergyduringAvalancheforSinglepulse)isdefinedasthemaximumenergythatcanbedissipatedbythedeviceduringasinglepulseavalancheoperation,(atthecircuitconditionsdescribedabove),attheIarandatthestartingjunctiontemperatureof25°C,tobringthejunctiontemperatureuptothemaximumonestatedintheabsolutemaximumratings.Ofcourse,thisvaluedecreasesasthestartingjunctiontemperatureincreases.Usuallyinsomedatasheetanenergyderatingcurvecalled“AvalancheEnergyvsstartingTj”,isprovided[figure6].AllthesingleeventavalancheoperationsbelowthisEnergyvalueareconsideredsafeforthedeviceifthejunctionstartingTemperatureis25°C,switchingadraincurrentlessorequaltoIar.IfTj>25°C,theendusercanrefertothecurve“AvalancheEnergyvsstartingTj”,toapplytherightderating.Figure6EASvstemperatureplotofSTP9NK80ZExample.Let’ssupposetohaveanSTP9NK80ZworkingasmainswitchofaDC/DCconverter.ThedatasheetratingsofthisdevicesareEas=350mJIar=7.5A,Tjmax=150°CFromtheboardanalysis,wefindthatthedeviceforsuchpowersupplyconditioncanexperienceasinglepulseavalancheoperation.Themeasurementshaveshownthattheaveragejunctiontemperatureis100°C,thepeakdraincurrentswitchedduringtheavalancheis4Aandtheenergythatisdissipatedduringthatsingleavalancheoperationis0.24mJ.Tounderstandifthedeviceisworkingwithintheratings,wehavetochecktheswitchedcurrent,comparingittotheIar;becausetheIdpeakvalueis4AandbelowtheIar,thisfirstratingissatisfied.Now,tounderstandifthejunctiontemperatureisbelowtheTjmax,wesupposethatbeforetheavalanchethejunctiontemperatureistheaverageone,100°C.Lookingattheplot“AvalancheEnergyvsstartingTj”,theenergytobringthejunctiontemperaturetotheallowedmaximumrating,startingfrom100°C,isapproximately50mJ.Sincetheenergymeasuredisbelowthatvalue,themaximumjunctiontemperaturereachedduringtheavalanchewillbelessthantheTjmax.Consideringthatboththebonds,currentbelowIarandjunctiontemperaturebelowmaximumtemperaturearewellsatisfied,onecansafelystatethatthedevice(underthatsingleavalancheevent)isworkingwithintheratings.ItisinterestingtounderstandhowthisEASvalueissetbySTMicroelectronics,alsobecauseeachMOSFETmanufacturerstateshis/herownapproachandfindings.Inordertoprovideaclearexplanation,theEASvalueisnotsimpletostate,asitisverydifficulttolookatthejunctiontemperatureduringtheavalancheoperation.Somemanufacturerssetthisvaluebythethermalimpedancestatedinthedatasheet.Thiscouldbeaninterestingapproach,butsomeconcernswouldhavetobetakenintoconsideration.Thedatasheetthermalimpedanceistheresponseofthesystemtoarectangularpowerpulse,maintainingthepackagecaseat25°C.owerjcjcPtZthT)(=∆Id=IarEAS@Iar,Tj=25°CStartingTjStartingTj=100°CAvalanchecharacteristicsandratingsofPowerMOSFET5Theapplicationoftheabovementionedformulawithoutanymodificationcannotgivepreciseinformation,infactconsideringtwopowerpulseshavingthesamepeak,onerectangular,theothertriangular,weobviouslyfindthatthepeaktemperatureisquitedifferent(figure7).TousetheZth,thetriangularpulsecanbeapproximatedasrectangularwithscaledamplitudeandwidth.Figure7ThermalresponseoftriangularpulseOnemoreimportantconcernabouttheuseofthermalimpedance,isthatusuallyitisexperimentallyandtheoreticallycalculatedconsideringtheon-stateofthedeviceandsoapowerdistributionwithinthedevicedifferentfromtheoneduringtheavalanche-state.STMicroelectronics’approachtoEASstatementstartsfromatheoreticalthermalmodelofthediewithsomeexperimentalverifications.TocalculatethemaximumtemperatureinthejunctionduringAvalanche,anequationliketheonebelowcanbeused[5]:startingavjTtAKPT+=0eq.1Wheretavistheavalanchetime(s)AdieArea(m2)P0peakpower(W)Ksilicon'sthermalconstant(Wm-2s½K-1)Itcomesfromthearrangementofthesolutionofthegeneralequationofheattransmission(Fourierequation),forthespecialcaseofaninfinitemedia(bi-dimensionalcase)subordinatedtoashortrectangularpowersourcepulseuniformlydistributedoveranarea[1].Theeq1,appliedtoatriangularpowerpulse,isasimpleandgoodapproximationinordertocalculatethetemperatureincreasewithinthedeviceduringtheavalancheoperationifthepulsehasashortduration.Startingfromthismodel,theexperimentalverificationismadelookingtotheVDSshapeduringtheAvalanche.Infactthebreakdownvoltageofthedeviceisnotconstantwhenvaryingthetemperatureandthecurrent[2].0.111010001002003004005006007008009001000Vds(Volt)Id(Ampere)Figure8BVDSSvsIDThevariationoftheblockingvoltagewiththecurrentisnearlylinearandthisresistanceisspecificofthedie-sizeofagivensetofdesignrules,epitaxiallayerandpackage.Figure8showsthemeasuredcharacteristicofBVDSSvstheIdfora700Vdevice.Thevoltagevariationwiththetemperatureispositivetoo,asshowninthedatasheetplot,andthevariationdependsonthestructuralcharacteristicsofthesilicondie.Bytakingtheseintoaccount,thejunctiontemperaturefromtheVDSshapecanbeextrapolated.Figure9VariationofVDS/IDshapeswiththeCoilduringtheUISFig9showstherealVDSshapeduringavalanche,wheretheincreaseofVDSobservedisduetotheriseofthesilicontimeTj(t)P(t)TriangularpulseanditsresponseRectangularpulseanditsresponseAvalanchecharacteristicsandratingsofPowerMOSFET6temperature.Figure9alsoshowsthewaveformsforthesamedevicesubmittedtodifferentCoils(i.e.differentenergy)atthesamepeakcurrent.Itisevidentthatthedifferentvdsshapesareduetodifferentmaximumtemperaturesreachedbythejunction.ThelinearrelationshipoftheVdswiththedietemperaturebecomesclearwhencomparingtheVdsquasi-parabolicshapetothetemperatureprofileforatriangularpowerpulseinfigure7.-EAR(EnergyduringAvalancheforRepetitivepulse)isoftenstatedinthedatasheet.ThisvalueisdefinedastheenergyatIar,duringrepetitiveoperation,andthevalueisclassicallyfoundconsideringa10KHzpowerpulsetrainwithadutycycleof50%andthenominalpowerratingofthedevice:Ear=power/fForinstance,iftheavailablepoweroftheMOSFET(Pdinthedatasheet)is150W,theEARshouldbeequalto15mJ.Itiseasytosaythatthisvalueisredundantandnotveryusefulforthedesign.ItgivesverypoorinformationaboutthemaximumenergytobedissipatedtomaintainthejunctiontemperaturebelowtheTjmax.AquickwaytoevaluatethetemperatureduringavalancheistotakeintoaccounttheextrapowerdissipationintroducedbytheRepetitiveavalanche,tocalculatetheheatsinksizetodissipateitbecauseofatriangularpulse:AVtVIE⋅=21Ifthefrequencyisf,tavthetimeinavalanche,VandI,thepeakvoltageandcurrent,theadditionalaveragepowertobedissipatedis:ftVIfEPAVAV⋅⋅=⋅=21Inadditiontotheconduction(Pcond)andswitching(Psw)losses:PavPswPcondPtot++=Tomakeanexample,wecanstartfromthepreviousSTW9NK80Z.WeconsiderthattheAvalancheinthiscaseisnotasingleevent,butarepetitiveonewithafrequencyf=50kHz..Forthesingleeventexample,sincethecurrentisbelowIarandthejunctiontemperatureisbelow150°C,thedeviceworksinsafeoperatingmode.Now,wehavetocheckonlythatthedevicesubmittedtorepetitiveavalanchemaintainsthejunctiontemperaturebelow150°C.Becauseforeachavalanchepulsewedissipate0.24mJ,theaveragepowerforavalancheisWkHzmJfEPAV125024.0=⋅=⋅=Ifthedevicehas:Rthja=Rthjc+Rthcs+Rthsa=10°C/WIftheaverageswitchingandconductionlossescomponentequalsto2W,theaveragetemperatureis:WWWPavPswPcondPtot14122=+=++=Tj=PtotRthja+Ta=140°C+TaSincethisistheaveragetemperatureandthepeaktemperatureduringtheavalancheshouldbehigher,it’sclearthattheavalanchephenomenonisgeneratingpowersohighthatthesystemthermalbehaviorcannotdissipateinordertomaintainthejunctiontemperaturebelowtheTjmax.Theonlysolutionistoreducethethermalresistanceofthesystem,changingtheheat-sinkorre-designtheapplicationtoavoidtheAvalanchefailure.Inordertoobtainamoreaccuratecomputation,theonlywaytoevaluatethemaximumtemperatureduringrepetitiveavalancheistocalculateitbytheZthofthesystem.DespitealltheconcernsalreadydiscussedaboutadoptingthepublishedZthfortheavalanche,experienceletsussafelystatethatthecomputationresultsforrepetitiveavalanchearesufficientlyguardbandedforrealworldapplications.Severalmethodscanbeusedtofindthesteadystatemaximumtemperatureduringsteadystateorafterafinitenumberofavalancheoccurrences.Themostfrequentlyandsufficientlyconservativeequationusedinordertofindthemaximumtemperatureforperiodicrectangularpowerpulsesatsteadystateis:()Τ−+Τ=∆pjcpjcpjctZthtRthtPT1max0eq.2WherePoisthepeakpower,Ttheperiodofthepulsestrainandtpisthepowerpulsewidth.Consideringtheapplicationontherepetitiveavalanche(butitcanbeappliedtoseveraloperativemodes)thecalculationofthepeaktemperaturemustalsotakeintoaccounttheotherlosses,duetoconductionandswitching.Figure10ashowsthepowerprofileofatypicalswitching.AvalanchecharacteristicsandratingsofPowerMOSFET7aPavg-Pavg0.7*P1-0.7*P10.89*P2-0.89*P20.7*P3-0.7*P3bFigure10exampleofmodelingtriangularpulsesbyrectangularonesandapplicationofsuperimpositionprincipleItiswellknownthatatriangularpowerpulsecanbemodeledasarectangularpulseofamplitude0.7ofthepeakpowerandwidth0.71oftherealone.Moreover,thataramppowershapecanbemodeledasasteppulseofwidth0.56andamplitude0.89oftherealones(Table2).Figure11showsthesimulatedtemperatureresponseofanarbitrarythermalimpedancetothetriangularpowerpulse(blueline)comparedtotherectangularapproximation(redline).timeTemperaturepowerFigure11thermalresponsecomparisonSotheaveragepowercanbecalculatedfromeachpeakpower:3322115.05.05.0PtPtPtPAVE⋅Τ+⋅Τ+⋅Τ=Bysuperimpositionprincipleandthemanipulationofequation2,wecanfindthepeaktemperatureduetoeverysinglepowerpulseoffigure10.Forthefirstpulse:()()()1*111*17.0maxtZthPttZthRthPTjcDjcjcAVEjc⋅⋅++−=∆Forthesecond:()()()()()[]2121112*22127.089.0maxttZthtttZthPtZthPttZthRthPTDjcDjcjcjcjcAVEjc+−+−⋅⋅++⋅⋅++−=∆Andforthethirdone:()()()()()[]()()[]3*33*3*223*3213*321113*3*3321389.07.07.0maxDjcDjcDDjcDDjcjcDjcjcAVEjcttZthtttZthPttttZthtttttZthPtZthPttttZthRthPT+−++⋅⋅+++++−+++−⋅⋅++⋅⋅++++−=∆eq.3Figure12steadystateswitchingofSTP11NM60FPLet’smakearealexample.Figure12showstheswitchingwaveformsofSTP11NM60FP,anSTMicroelectronicMDMesh,workingasmainswitchinanACadapter.ThewaveformsshowthatthedraintosourcevoltageisgoingbeyondtheVdsAbsolutemaximumrating,repetitivelywithat1t2t3TP1P3P2t*1t*2t*3td3td2td1AvalanchecharacteristicsandratingsofPowerMOSFET8periodof12us.Figure13showsadetailedwaveformoftheavalanchephenomenon.Tocheckifthedeviceisworkingwithinthespecificationwehavetocheckthedatasheetratings.Thedatawehave:T=12usTcase=70°Cton=40nsPpeakon=160V*1.6A=256Wtoff=60nsPpeakoff=280V*2.4A=672Wtcond=2.4usPpeakcond=(4A)²*0.45*2.4=17.3Wtav=86nsPpeakav=680V*3.6A=2448WWhereforthepeakpowercalculationovertheconductionstate,weassumedtheRdsonattheworstconditioni.e.Tj=150°C:Rdson(150°C)=Rdson(25°C)2.4,Thefactor2.4isthederatingfactoronecanfindinthedatasheetlookingattheplot“normalizedRdsonvstemperature”.abFigure13detailsofon/offswitchingofSTP11NM60FPLookingatthedatasheetwefindthattheIaris5.5A,becausethepeakcurrentduringtheavalancheis3.6A,belowthedatasheetratings,theIarconstraintissatisfied.Pleasenotethatthepeakcurrentbeforetheturnoff,thatinthiscaseis4.8A,shouldbetakenintoaccountasveryworstcaseforavalancheincaseofahigherdv/dt,butalsointhislastcase,thespecificationissatisfied.Let’scalculatetheAveragepower:PEAKoffoffPEAKononPEAKcondcondPEAKavavAVEPtPtPtPtP⋅Τ+⋅Τ+⋅Τ+⋅Τ=5.05.05.05.0PAVE=8.77+1.73+0.43+1.68=12.61WFortheconductionlossesweconsideredtheworstcaseofatriangularpowershapeeveniftherightone