Power-Cycling-Lifetime-of-Advanced-Power-Modules-f

PowerCyclingLifetimeofAdvancedPowerModulesforDifferentTemperatureSwingsU.Scheuermann,U.HechtSEMIKRONElektronikGmbH,Sigmundstraße200,90431Nuremberg,GermanyTel.:++49-911-6559-381Fax:++49-911-6559-414E-mail:[e-mail]:Exactmodulelifetimepredictionrequiresastatisticalapproachforthequalificationofmodulesinend-of-lifetests.ThisdemandcanbemetbyaWeibullanalysisofagroupofsamples.Theresultsforactivepowercyclingtests,presentedhereforadvancedpowermoduleswithoutbaseplate,provethequalityofthedesignandprocesscapabilitybyanarrowdistributionofend-of-lifefailures.Thetestresultsforjunctiontemperatureswingsof80Kand110Kcanbeusedasabasistoinvestigatethesuperpositionofthebothconditions.Thevalidityofthelinearsuperpositionofaccumulatedstressisofgreatimportanceforthetransferoflifetimetestresultstoapplication-conformcomplexloadprofiles.I.INTRODUCTIONWiththeuseofIGBTpowermodulesinapplicationslikeautomotiveortraction,thereisanincreasingdemandforexactmodulelifetimeprediction.Anestimationofthepowermodulelifetimerequiresarelationshipbetweenthenumberofloadcyclesversusthetemperatureconditions(i.e.maximumheatsinktemperatureandmaximumjunctiontemperature)inthesystemaswellastheknowledgeofthedominantloadcycleconditioninthemissionprofileoftheapplication[1],[2].Ifthemissionprofileiscomplex,theassumptionmadeforthesuperpositionofdifferentconditionsbecomesimportant.Wewillpresentfirstexperimentalresultsonthesuperpositionofdifferentcycleconditions.II.POWERCYCLINGTESTThepowercyclingtest(PC)isanimportanttooltodeterminethemodulelifetimeofagivenmodulearchitecture.Duringtheheatingphaseofacycle,thepowerchipsareactivelyheatedbyaconstantloadcurrent(Fig.1).Thedissipatedenergyduringtheheatingphasewillheatupthewholesystemandestablishatemperaturegradientinthepowermodule.Ashortpulsetimeintherangeof1swillonlystressthechipitselfanditsconnections,especiallythebondwirecontact.Alongerpulsetimeallowsfatigueprocessesduetocreeporreconstructiontoevolve;itismorerealisticandshouldbeselectedforhightemperatureswings,wherethelongertesttimecanbeaccepted.Theloadcurrentshouldbeadjustedintherangeofthenominalcurrentatthegivencoolingconditions.Afterthemaximumjunctiontemperatureisreached,theloadcurrentisturnedoffandthechipcoolsdownfastduetoitssmallthermalmassuntilitreachestheheatsinktemperatureandthencontinuestocoolwithacharacteristicdeterminedbytheheatsinktimeconstant.Thedurationofthecoolingphasedependsontheefficiencyoftheemployedcoolingsystem.Forhighefficiencywater-cooledheatsinks,thecoolantflowcanbesuppressedduringtheheatingphaseandcanbeactivatedonlyinthecoolingphaseinordertodecreasethecycletime.Ifasmallsensecurrentisappliedduringthecoolingphase,themaximumjunctiontemperaturecanbemonitoredon-lineduringthetest.Thecontrolparametersshouldbecarefullyselected.Ifthejunctiontemperatureitselfisusedtocontroltheon-andoff-phase,thenanincreaseofthethermalresistancecausesareductionofthedissipatedenergy,whichisnotrealisticforapplicationconditions.AmoreFig.1:Powercyclingtestconditions4060801001200102030405060time[s]temperature[°C]coolantflowcoolantflowTheatsink,maxTheatsink,minTjunction,maxfavorableparameteristheheatsinktemperature;itwillexcludetheinfluenceofchangesinthecoolingefficiency(forexampleachangeincoolingliquidtemperature)butwillincludetheeffectscausedbychangesinthethermalresistanceofthemoduleortheheatsinkinterface.Anotheradvantageisthatthiscontrolparameterallowsthestandardizationoftestconditionsforastatisticalanalysis,becausethedatasheetthermalresistancecanbeusedtocalculatethenominaljunctiontemperatureandthereforesuppliesaparameterselectionindependentofthethermalresistancevariationofthespecificsamplesinatestgroup[3].Toavoidconfusionofterms,thetemperaturedifferencesinFig.1shouldbeclearlydistinguished.ThevalueTj=Tj,max-Tj,mincharacterizesthepowercyclingtemperatureswing.ItisadifferenceoftemperaturesatthesamelocationatdifferenttimesandshouldnotbeconfusedwiththetemperaturedifferencebetweentwolocationsatthesameinstantasinthedefinitionofthethermalresistanceRth,jh=T/P=(Tj-Th)/P.III.THELESIT-RESULTSAnimportantstudyofthepowercyclinglifetimeofstandardbaseplatemoduleswasconductedduringtheLESIT-project.Usingmodulesfromdifferentsuppliers,acomprehensivedatabasewascollected[4].Themostimportantresultwastheperception,thattheparameterTjaloneisnotsufficienttodescribethecharacteristicofthelifetimeofmodulesinpowercyclingtests.Asecondparameter,theabsolutemediumjunctiontemperatureTm=Tj,min+Tj/2,hasaconsiderableinfluenceonthetestresult.ThenumberofcyclestofailureNfcanbeexpressedinacompactform(1)withtheBoltzmann-constantkB,whichallowsagoodapproximationofalltestresults(Fig.2)byaleast-square-fitofasetofthreeparameters:theconstantparameterA,theexponentandtheactivationenergyEa.Ashortcyclingtimeof10swasusedforalltests.IV.POWERCYCLINGOFADVANCEDMODULESAnextendedseriesofpowercyclingtestswasperformedduringthequalificationoftheSKiM®3module[5].Thissixpackmodulecanbeconsideredasarepresentativeofafamilyofmodules,basedonthepressurecontacttechnology(Fig.3).Theadvantagesofthepressurecontactconcept–t