Defect Diagnosis using a Current Ratio based Quies

1DefectDiagnosisusingaCurrentRatiobasedQuiescentSignalAnalysisModelforCommercialPowerGridsChintanPatel,ErnestoStaroswiecki,SmitaPawar,DhruvaAcharyyaandJimPlusquellicDepartmentofCSEE,UniversityofMaryland,BaltimoreCountyforJETTAfromaspecialissueguesteditedbyDr.RaÞcMakkiContactAuthorJimPlusquellicUMBC,CSEE,ECS2121000HilltopCircle,Baltimore,MD-21250Email:[e-mail]:410-455-1349Fax:[phone]DefectDiagnosisusingaCurrentRatiobasedQuiescentSignalAnalysisModelforCommercialPowerGridsAuthors:ChintanPatel,[e-mail],[e-mail],[e-mail],[e-mail],[e-mail]:DepartmentofCSEE,UniversityofMarylandBaltimoreCounty1000HilltopCircle,Baltimore,MD-21250.AbstractQuiescentSignalAnalysis(QSA)isanovelelectrical-test-baseddiagnostictechniquethatusesIDDQmeasure-mentsmadeatmultiplechipsupplypadsasameansoflocatingshortingdefectsinthelayout.Theuseofmultiplesupplypadsreducestheadverseeffectsofleakagecurrentbyscalingthetotalleakagecurrentovermultiplemeasure-ments.Inpreviouswork,aresistancemodelforQSAwasdevelopedanddemonstratedonasmallcircuit.Inthispaper,theweaknessesoftheoriginalQSAmodelareidentiÞed,inthecontextofaproductionpowergrid(PPG)andprobecardmodel,andanewmodelisdescribed.ThenewQSAalgorithmisdevelopedfromtheanalysisofIDDQcon-tourplots.AÒfamilyÓofhyperbolacurvesisshowntobeagoodÞttothecontourcurves.Theparameterstothehyperbolaequationsarederivedwiththehelpofinsertedcalibrationtransistors.SimulationexperimentsareusedtodemonstratethepredictionaccuracyofthemethodonaPPG.Keywords:IDDQ,IDDT,quiescentsignalanalysis,test,powergridsAbstractQuiescentSignalAnalysis(QSA)isanovelelectrical-test-baseddiagnostictechniquethatusesIDDQmeasurementsmadeatmultiplechipsupplypadsasameansoflocatingshortingdefectsinthelayout.Theuseofmultiplesupplypadsreducestheadverseeffectsofleakagecurrentbyscalingthetotalleakagecurrentovermultiplemeasurements.Inpreviouswork,aresistancemodelforQSAwasdevelopedanddemonstratedonasmallcircuit.Inthispaper,theweaknessesoftheoriginalQSAmodelareidentiÞed,inthecontextofaproductionpowergrid(PPG)andprobecardmodel,andanewmodelisdescribed.ThenewQSAalgorithmisdevelopedfromtheanalysisofIDDQcontourplots.AÒfamilyÓofhyperbolacurvesisshowntobeagoodÞttothecontourcurves.Theparameterstothehyperbolaequa-tionsarederivedwiththehelpofinsertedcalibrationtransistors.SimulationexperimentsareusedtodemonstratethepredictionaccuracyofthemethodonaPPG.1.0IntroductionIDDQhasbeenamain-streamsupplementaltestingmethodfordefectdetectionformorethanadecadewithmanycompanies.Withtheadventofthedeepsubmicrontechnologies,theuseofsingle-thresholdIDDQtechniqueresultsinunacceptableyieldloss.Settinganabsolutepass/failthresholdforIDDQtestinghasbecomeincreasinglydifÞcultduetotheincreasingsubthresholdleakagecurrents[1].Currentsignatures[2],delta-IDDQ[3]andratio-IDDQ[4]havebeenproposedasameansforcalibratingforthesehighsubthresholdleakages.Thesetechniquesrelyonaself-relativeordifferentialanalysis,inwhichtheaverageIDDQofeachdeviceisfactoredintothepass/failthreshold.However,theseproposedformsofcalibrationareexpectedtobecomelesseffectiveoversuccessivetechnologygenerations.Analternativecalibrationstrategythatmayhavebetterscalingpropertiesistodistributethetotalleakagecurrentacrossasetofmeasurements.ThisisaccomplishedbyintroducingprobinghardwarethatallowsthemeasurementofIDDQateachofthesupplyports.Themethodproposedinthiswork,calledQuiescentSignalAnalysis(QSA),isdesignedtoexploitthistypeofleakagecalibrationfordefectdetectionandasameansofprovidinginformationaboutDefectDiagnosisusingaCurrentRatiobasedQuiescentSignalAnalysisModelforCommercialPowerGridsChintanPatel,ErnestoStaroswiecki,SmitaPawar,DhruvaAcharyyaandJimPlusquellicDepartmentofCSEE,UniversityofMaryland,BaltimoreCountyThisworkissupportedbyaFacultyPartnershipAwardfromIBMÕsAustinCenterforAdvancedStudies(ACAS)ProgramandbyanNSFgrant,awardnumber0098300.4thedefectÕslocationinthelayout[5][6].ThislatterdiagnosticattributeofQSAmayprovideanalternativetoimage-basedphysicalfailureanalysisproceduresthatarechallengedbytheincreasingnumberofmetallayersandßipchiptechnology.Aresistance-baseddiagnosticmodelforQSAwasdevelopedinpreviousworksandsimulationexperimentswereusedtodemonstratethediagnosticcapabilitiesoftheQSAmethodonasmallcircuit[5][6].Inthispaper,severalweaknessesoftheresistance-basedmodelareuncoveredfromsimulationsofaproductionpowergrid(PPG).Acur-rent-ratio-basedmodelisproposedanddemonstratedtoimproveondefectlocalizationaccuracyoftheoriginalmethod[7].Thenewmethodrequirestheinsertionofcalibrationtransistors(CT),oneundereachofthesupplypadsinthedesign,thatpermittheshortingofthepowerandgroundsupplyrailsatpointsclosetothesubstrate.ThestateoftheCTsarecontrolledbyscanchainßip-ßops.TheIDDQsobtainedwhenoneoftheCTsisturnedonareusedtocalibratetheIDDQsmeasuredunderafailingIDDQpattern.Thecalibrationtechniqueisshowntoaddressseveralweaknessesofthepreviousmodelincludingnon-zeroprobecardresistanceandirregularsupplygridtopologies.Cur