第08讲-自动测试生成超大规模集成电路测试技术课件_第1页
第08讲-自动测试生成超大规模集成电路测试技术课件_第2页
第08讲-自动测试生成超大规模集成电路测试技术课件_第3页
第08讲-自动测试生成超大规模集成电路测试技术课件_第4页
第08讲-自动测试生成超大规模集成电路测试技术课件_第5页
已阅读5页,还剩137页未读 继续免费阅读

下载本文档

版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领

文档简介

Lecture8AutomaticTestPatternGeneration第八讲自动测试生成Lecture8AutomaticTestPatte1Contents

内容目录TestabilityMeasures/可测试性测度CombinationalCircuitATPG/组合电路ATPGSequentialCircuitATPG/

时序电路ATPGSummary/小结Contents

内容目录TestabilityMeasu21TestabilityMeasures

可测试性测度Needapproximatemeasureof:Controllability--Difficultyofsettinginternalcircuitlinesto0or1bysettingprimarycircuitinputsObservability--Difficultyofobservinginternalcircuitlinesbyobservingprimaryoutputs1TestabilityMeasures

可测试性测度N31.1Purpose

目的Uses:Analysisofdifficultyoftestinginternalcircuitparts–redesignoraddspecialtesthardwareGuidanceforalgorithmscomputingtestpatterns–avoidusinghard-to-controllinesEstimationoffaultcoverageEstimationoftestvectorlength1.1Purpose

目的Uses:41.2Origins

起源ControltheoryRutman1972--FirstdefinitionofcontrollabilityGoldstein1979--SCOAPFirstdefinitionofobservabilityFirstelegantformulationFirstefficientalgorithmtocomputecontrollabilityandobservabilityParker&McCluskey1975DefinitionofProbabilisticControllabilityBrglez1984--COP1stprobabilisticmeasuresSeth,Pan&Agrawal1985–PREDICT1stexactprobabilisticmeasures1.2Origins

起源Controltheory51.3TestabilityAnalysis

可测试性分析InvolvesCircuitTopologicalanalysis,butnotestvectorsandnosearchalgorithm.StaticanalysisLinearcomputationalcomplexity,Otherwise,ispointless–mightaswelluseautomatictest-patterngenerationandcalculate:ExactfaultcoverageExacttestvectors1.3TestabilityAnalysis

可测试性分61.4SCOAPmeasures

SCOAP测度SCOAP–SandiaControllabilityandObservabilityAnalysisProgramCombinationalmeasures:CC0–Difficultyofsettingcircuitlinetologic0CC1–Difficultyofsettingcircuitlinetologic1CO–DifficultyofobservingacircuitlineSequentialmeasures–analogous:SC0SC1SO1.4SCOAPmeasures

SCOAP测度SCOA71.4.1RangeofSCOAPMeasures

SCOAP测度范围Controllabilities–1(easiest)toinfinity(hardest)Observabilities–0(easiest)toinfinity(hardest)Combinationalmeasures:Roughlyproportionalto#circuitlinesthatmustbesettocontrolorobservegivenlineSequentialmeasures:Roughlyproportionalto#timesaflip-flopmustbeclockedtocontrolorobservegivenline1.4.1RangeofSCOAPMeasures81.4.2ControllabilityRules

可控制性规则1.4.2ControllabilityRules

可控91.4.2ControllabilityRules(Cont.)

可控制性规则(续)1.4.2ControllabilityRules(C101.4.3ObservabilityRules

可观察性规则Toobserveagateinput:Observeoutputandmakeotherinputvaluesnon-controlling1.4.3ObservabilityRules

可观察性111.4.3ObservabilityRules(Cont.)

可观察性规则Toobserveafanoutstem:Observeitthroughbranchwithbestobservability1.4.3ObservabilityRules(Con121.4.4DFlip-FlopRules

D触发器规则AssumeasynchronousRESETline.CC1(Q)=CC1(D)+CC1(C)+CC0(C)+CC0

(RESET)SC1(Q)=SC1(D)+SC1(C)+SC0(C)+SC0

(RESET)+1CC0(Q)=min[CC1(RESET)+CC1(C)+CC0(C),CC0(D)+CC1(C)+CC0(C)]SC0(Q)isanalogousCO(D)=CO(Q)+CC1(C)+CC0(C)+CC0(RESET)SO(D)isanalogous1.4.4DFlip-FlopRules

D触发器规则131.4.4DFlip-FlopRules(Cont.)

D触发器规则(续)CO(RESET)=CO(Q)+CC1(Q)+CC1(RESET)+CC1(C)+CC0(C)SO(RESET)isanalogousThreewaystoobservetheclockline:SetQto1andclockina0fromDSettheflip-flopandthenresetitResettheflip-flopandclockina1fromDCO(C)=min[CO(Q)+CC1(Q)+CC0(D)+CC1(C)+CC0(C),CO(Q)+CC1(Q)+CC1(RESET)+

CC1(C)+CC0(C),CO(Q)+CC0(Q)+CC0(RESET)+CC1(D)+CC1(C)+CC0(C)]SO(C)isanalogous1.4.4DFlip-FlopRules(Cont.141.4.5LevelizationAlgorithm6.1

分级算法Labeleachgatewithmax#oflogiclevelsfromprimaryinputsorwithmax#oflogiclevelsfromprimaryoutputAssignlevel#0toallprimaryinputs(PIs)ForeachPIfanout:LabelthatlinewiththePIlevelnumber,&QueuelogicgatedrivenbythatfanoutWhilequeueisnotempty:DequeuenextlogicgateIfallgateinputshavelevel#’s,labelthegatewiththemaximumofthem+1;Else,requeuethegate1.4.5LevelizationAlgorithm6151.4.6TestabilityAlgorithm6.2

可测试性算法ForallPIs,CC0=CC1=1andSC0=SC1=0Forallothernodes,CC0=CC1=SC0=SC1=GofromPIstoPOS,usingCCandSCequationstogetcontrollabilities--IterateonloopsuntilSCstabilizes--convergenceguaranteedForallPOs,setCO=SO=

0Forallothernodes,

CO=SO=WorkfromPOstoPIs,UseCO,SO,andcontrollabilitiestogetobservabilitiesFanoutstem(CO,SO)=minbranch(CO,SO)IfaCCorSC(COorSO)is,thatnodeisuncontrollable(unobservable)8881.4.6TestabilityAlgorithm6.162CombinationalCircuitATPG

组合电路ATPGElectron-beam(E-beam)testobservesinternalsignals–“picture”ofnodeschargedto0and1indifferentcolorsTooexpensiveTheATPGproblem:Givenalogicalfaultmodel,andacircuit,determineasmallsetoftestvectorsthatdetectallfaultsinthecircuit.2CombinationalCircuitATPG

172.1Functionalvs.StructuralATPG

功能和结构测试2.1Functionalvs.Structural182.1.1Compare

比较FunctionalATPG–generatecompletesetoftestsforcircuitinput-outputcombinations129inputs,65outputs:2129=680,564,733,841,876,926,926,749,214,863,536,422,912patternsUsing1GHzATE,wouldtake2.15x1022yearsStructuraltest:Noredundantadderhardware,64bitslicesEachwith27faults(usingfaultequivalence)Atmost64x27=1728faults(tests)Takes0.000001728son1GHzATEDesignergivessmallsetoffunctionaltests–augmentwithstructuralteststoboostcoverageto98+%2.1.1Compare

比较FunctionalATP192.2AlgorithmCompleteness

算法完备性Definition:Algorithmiscompleteifitultimatelycansearchentirebinarydecisiontree,asneeded,togenerateatestUntestablefault–notestforitevenafterentiretreesearchedCombinationalcircuitsonly–untestablefaultsareredundant,showingthepresenceofunnecessaryhardware2.2AlgorithmCompleteness

算法完202.3Algebras:5-Valuedand9-Valued

算法代数:5值和9值逻辑代数SymbolDD01XG0G1F0F1Meaning1/00/10/01/1X/X0/X1/XX/0X/1FailingMachine0101XXX01GoodMachine1001X01XXRoth’sAlgebraMuth’sAdditions2.3Algebras:5-Valuedand9-V212.3.1Higher-OrderAlgebras

高阶代数Representtwomachines,whicharesimulatedsimultaneouslybyacomputerprogram:Goodcircuitmachine(1stvalue)Badcircuitmachine(2ndvalue)Bettertorepresentbothinthealgebra:Needonly1passofATPGtosolvebothGoodmachinevaluesthatprecludebadmachinevaluesbecomeobvioussooner&viceversaNeededforcompleteATPG:Combinational:Multi-pathsensitization,RothAlgebraSequential:MuthAlgebra--goodandbadmachinesmayhavedifferentinitialvaluesduetofault2.3.1Higher-OrderAlgebras

高阶222.4TypesofAlgorithms

算法类型Exhaustive/穷举算法Random-PatternGeneration/随机码生成BooleanDifferenceSymbolicMethod/布尔差分符号方法PathSensitizationMethod/路径敏化方法BooleanSatisfiability/布尔可满足性2.4TypesofAlgorithms

算法类型E232.4.1Exhaustive

穷举算法Forn-inputcircuit,generateall2ninputpatternsInfeasible,unlesscircuitispartitionedintoconesoflogic,with15inputsPerformexhaustiveATPGforeachconeMissesfaultsthatrequirespecificactivationpatternsformultipleconestobetested

2.4.1Exhaustive

穷举算法Forn-in242.4.2Random-PatternGeneration

随机码生成FlowchartformethodUsetogettestsfor60-80%offaults,thenswitchtoD-algorithmorotherATPGforrest2.4.2Random-PatternGeneratio252.4.3BooleanDifferenceSymbolicMethod

布尔差分符号方法g=G(X1,X2,…,Xn)forthefaultsitefj=Fj(g,X1,X2,…,Xn)1j

mXi=0or1for1i

n

2.4.3BooleanDifferenceSymbo26Shannon’sExpansionTheorem:

F(X1,X2,…,Xn)=X2

F(X1,1,…,Xn)+X2

F(X1,0,…,Xn)BooleanDifference(partialderivative):

FjgFaultDetectionRequirements:

G(X1,X2,…,Xn)=1FjgBooleanDifference(Sellers,Hsiao,Bearnson)=Fj(1,X1,X2,…,Xn)Fj(0,X1,…,Xn)=Fj(1,X1,X2,…,Xn)Fj(0,X1,…,Xn)=1

Shannon’sExpansionTheorem:272.4.4PathSensitizationMethod

路径敏化方法FaultSensitization/故障敏化FaultPropagation/故障传播LineJustification/线验证2.4.4PathSensitizationMetho2CircuitExample

电路实例Trypathf–h–k–L

blockedat

j,sincethereisnowaytojustifythe1on

i10DD111DDDCircuitExample

电路实例Tr2CircuitExample(Cont.)

电路实例(续)Trysimultaneouspathsf–h–k–Land

g–i–j–k–LblockedatkbecauseD-frontier(chainofDorD)disappears1DDDDD1CircuitExample(Cont.30CircuitExample(Cont.)

电路实例(续)Finaltry:path

g–i–j–k–L–testfound!0DDD1DD10CircuitExample(Cont.312.4.5BooleanSatisfiability

布尔可满足性2SAT:xixj+xjxk+xlxm…=0

xpxy+xrxs+xtxu…=03SAT:xixjxk+xjxkxl+xlxmxn…=0

xpxy+xrxsxt+xtxuxv…=0......2.4.5BooleanSatisfiability

3SatisfiabilityExampleforANDGateS

akbkck=0(non-tautology)or

P(ak+bk+ck)=1(satisfiability)ANDgatesignalrelationships:Cube:Ifa=0,thenz=0azIfb=0,thenz=0bzIfz=1,thena=1ANDb=1zabIfa=1ANDb=1,thenz=1abzSumtoget:az+bz+abz=0(thirdrelationshipisredundantwith1sttwo)SatisfiabilityExample3Pseudo-BooleanandBooleanFalseFunctionsPseudo-Booleanfunction:useordinary+-integerarithmeticoperatorsComplementationofxrepresentedby1–xFpseudo—Bool

=2z+ab–az–bz–abz=0Energyfunctionrepresentation:letanyvariablebeintherange(0,1)inpseudo-BooleanfunctionBooleanfalseexpression:

fAND

(a,b,z)=z(ab)=az+bz+abz

Pseudo-BooleanandBoo3ANDGateImplicationGraph

隐含图ReallyefficientEachvariablehas2nodes,oneforeachliteralIf…thenclauserepresentedbyedgefromifliteraltothenliteralTransformintotransitiveclosuregraph

Whennodetrue,allreachablestatesaretrueANDingoperatorusedfor3SATrelations

ANDGateImplicationG352.5ComputationalComplexity

计算复杂性IbarraandSahnianalysis–NP-Complete(nopolynomialexpressionfoundforcomputetime,presumedtobeexponential)Worstcase:

no_piinputs,2no_piinputcombinations

no_ffflip-flops,4no_ffinitialflip-flopstates(goodmachine0or1badmachine0or1)worktoforwardorreversesimulatenlogicgatesa

nComplexity:O(nx2no_pix4no_ff)

2.5ComputationalComplexity

计362.6HistoryofAlgorithmSpeedups

算法历史AlgorithmD-ALGPODEMFANTOPSSOCRATESWaicukauskietal.ESTTRANRecursivelearningTafertshoferetal.Est.speedupoverD-ALG(normalizedtoD-ALGtimeTPGSystem2189ATPGSystem8765ATPGSystem3005ATPGSystem48525057Year1966198119831987198819901991199319951997

2.6HistoryofAlgorithmSpeed372.7FaultCoverageandEfficiency

故障覆盖率和效率Faultcoverage=Faultefficiency

#ofdetectedfaultsTotal#faults#ofdetectedfaultsTotal#faults--#undetectablefaults=2.7FaultCoverageandEfficie382.8TestGenerationSystems

测试生成系统CircuitDescriptionTestPatternsUndetectedFaultsRedundantFaultsAbortedFaultsBacktrackDistributionFaultListCompacterSOCRATESWithfaultsimulator2.8TestGenerationSystems

测试392.9TestCompaction

测试压缩FaultsimulatetestpatternsinreverseorderofgenerationATPGpatternsgofirstRandomly-generatedpatternsgolast(becausetheymayhavelesscoverage)Whencoveragereaches100%,dropremainingpatterns(whicharetheuselessrandomones)Significantlyshortenstestsequence–economiccostreduction2.9TestCompaction

测试压缩Fault402.9.1StaticandDynamicCompaction

静态和动态压缩StaticcompactionATPGshouldleaveunassignedinputsasXTwopatternscompatible–ifnoconflictingvaluesforanyPICombinetwoteststaandtbintoonetesttab

=

tatbusingD-intersectionDetectsunionoffaultsdetectedbyta&tbDynamiccompactionProcesseverypartially-doneATPGvectorimmediatelyAssign0or1toPIstotestadditionalfaults

2.9.1StaticandDynamicCompa412.9.2CompactionExample

压缩实例t1

=01Xt2=0X1t3=0X0t4=X01Combine

t1andt3,then

t2andt4Obtain:t13

=010t24=001TestLengthshortenedfrom4to22.9.2CompactionExample

压缩实例t423SequentialCircuitsATPG

时序电路ATPGAsequentialcircuithasmemoryinadditiontocombinationallogic.Testforafaultinasequentialcircuitisasequenceofvectors,whichInitializesthecircuittoaknownstateActivatesthefault,andPropagatesthefaulteffecttoaprimaryoutputMethodsofsequentialcircuitATPGTime-frameexpansionmethodsSimulation-basedmethods3SequentialCircuitsATPG

时序电433.1Time-FramesExpansionIfthetestsequenceforasinglestuck-atfaultcontainsnvectors,ReplicatecombinationallogicblockntimesPlacefaultineachblockGenerateatestforthemultiplestuck-atfaultusingcombinationalATPGwith9-valuedlogicComb.blockFaultTime-frame0Time-frame-1Time-frame-n+1UnknownorgivenInit.stateVector0Vector-1Vector-n+1PO0PO-1PO-n+1StatevariablesNextstate3.1Time-FramesExpansionIfth443.1.1ExampleforLogicSystems

实例FF2

FF1ABs-a-13.1.1ExampleforLogicSystem4Five-ValuedLogic(Roth)

0,1,D,D,X

A

BXXX0s-a-1DA

BXXX0s-a-1DFF1FF1FF2FF2DDTime-frame-1Time-frame0Five-ValuedLogic(Rot4Nine-ValuedLogic(Muth)

0,1,1/0,0/1,

1/X,0/X,X/0,X/1,XA

BXXX0s-a-10/1A

B0/X0/X0/1

Xs-a-1X/1

FF1FF1FF2FF20/1X/1Time-frame-1Time-frame0Nine-ValuedLogic(Mut473.1.2ImplementationofATPG

ATPG实现SelectaPOforfaultdetectionbasedondrivabilityanalysis.Placealogicvalue,1/0or0/1,dependingonfaulttypeandnumberofinversions.JustifytheoutputvaluefromPIs,consideringallnecessarypathsandaddingbackwardtime-frames.Ifjustificationisimpossible,thenusedrivabilitytoselectanotherPOandrepeatjustification.IftheprocedurefailsforallreachablePOs,thenthefaultisuntestable.If1/0or0/1cannotbejustifiedatanyPO,but1/Xor0/Xcanbejustified,thethefaultispotentiallydetectable.3.1.2ImplementationofATPG

A483.1.3ComplexityofATPG

计算复杂性Synchronouscircuit--Allflip-flopscontrolledbyclocks;PIandPOsynchronizedwithclock:Cycle-freecircuit–Nofeedbackamongflip-flops:Testgenerationforafaultneedsnomorethandseq+1time-frames,wheredseqisthesequentialdepth.Cycliccircuit–Containsfeedbackamongflip-flops:Mayneed9Nfftime-frames,whereNffisthenumberofflip-flops.Asynchronouscircuit–Highercomplexity!Time-Frame0Time-Framemax-1Time-Framemax-2Time-Frame-2Time-Frame-1S0S1S2S3Smaxmax=Numberofdistinctvectorswith9-valuedelements

=9Nff3.1.3ComplexityofATPG

计算复杂性4Cycle-FreeCircuits

无环电路Characterizedbyabsenceofcyclesamongflip-flopsandasequentialdepth,dseq.dseqisthemaximumnumberofflip-flopsonanypathbetweenPIandPO.Bothgoodandfaultycircuitsareinitializable.Testsequencelengthforafaultisboundedbydseq+.1Cycle-FreeCircuits

无环50Cycle-FreeExample

无环电路实例F1F2F3Level=12F1F2F3Level=1233dseq=3s-graphCircuitAllfaultsaretestable.SeeExample.3.2Cycle-FreeExample

无环电5Cycliccircuit

循环电路

Cyclicstructure–Sequentialdepthisundefined.Circuitisnotinitializable.Notestscanbegeneratedforanystuck-atfault.Afterexpandingthecircuitto9Nff=81,orfewer,time-framesATPGprogramcallsanygiventargetfaultuntestable.Circuitcanonlybefunctionallytestedbymultipleobservations.Functionaltests,whensimulated,givenofaultcoverage.Cycliccircuit

循环电路Cy5CyclicCircuitExample

循环电路实例F1F2CNTZModulo-3counters-graphF1FCyclicCircuitExample5BenchmarkCircuitsCircuitPIPOFFGatesStructureSeq.depthTotalfaultsDetectedfaultsPotentiallydetectedfaultsUntestablefaultsAbandonedfaultsFaultcoverage(%)Faultefficiency(%)Max.sequencelengthTotaltestvectorsGentestCPUs(Sparc2)s1196141418529Cycle-free412421239

03099.8100.0

331310s1238141418508Cycle-free413551283

072094.7100.0

330815yclic--14861384

2267693.194.8

2452519941yclic--15061379

2309791.693.4

285591918BenchmarkCircuitsCirc5AsynchronousCircuit

异步电路Anasynchronouscircuitcontainsunclockedmemoryoftenrealizedbycombinationalfeedback.Almostimpossibletobuild,letalonetest,alargeasynchronouscircuit.Clockgenerators,signalsynchronizers,flip-flopsaretypicalasynchronouscircuits.Manylargesynchronoussystemscontainsmallportionsoflocalizedasynchronouscircuitry.SequentialcircuitATPGshouldbeabletogeneratetestsforcircuitswithlimitedasynchronousparts,evenifitdoesnotdetectfaultsinthoseparts.AsynchronousCircuit

异5AsynchronousModel

异步电路模型ClockedFlip-flopsFeedbackdelaysSynchronousPIsSynchronousPOsSystemClock,CKFastmodelClock,FMCKCKCKFeedback-freeCombinationalLogicCCombinationalFeedbackPaths:FeedbacksetModelingcircuitisShowninorange.PPOPPIAsynchronousModel

异步电5Time-FrameExpansion

异步电路时帧扩展Time-framekTime-frame-k+1Time-frame-k-1CFMCKCFMCKCFMCKCCKAsynchronousfeedbackstabilizationPIPOFeedbacksetPPIPPOFeedbacksetVectorkTime-FrameExpansion

异573.2simulation-basedmethods

基于模拟的方法Difficultieswithtime-framemethod:LonginitializationsequenceImpossibleinitializationwiththree-valuedlogicCircuitmodelinglimitationsTimingproblems–testscancauseraces/hazardsHighcomplexityInadequacyforasynchronouscircuitsAdvantagesofsimulation-basedmethodsAdvancedfaultsimulationtechnologyAccuratesimulationmodelexistsforverificationVarietyoftests–functional,heuristic,randomUsedsinceearly1960s3.2simulation-basedmethods

基583.2.1UsingFaultSimulator

使用故障模拟器FaultsimulatorVectorsource:Functional(test-bench),Heuristic(walking1,etc.),Weightedrandom,randomFaultlistTestvectorsNewfaultsdetected?Stoppingcriteria(faultcoverage,CPUtimelimit,etc.)satisfied?StopUpdatefaultlistAppendvectorsRestorecircuitstateGeneratenewtrialvectorsYesNoYesNoTrialvectors3.2.1UsingFaultSimulator

使用593.2.2ContestAConcurrenttestgeneratorforsequentialcircuittesting(Contest).Searchfortestsisguidedbycost-functions.Three-phasetestgeneration:Initialization–nofaultstargeted;cost-functioncomputedbytrue-valuesimulator.Concurrentphase–allfaultstargeted;costfunctioncomputedbyaconcurrentfaultsimulator.Singlefaultphase–faultstargetedoneatatime;costfunctioncomputedbytrue-valuesimulationanddynamictestabilityanalysis.Ref.:Agrawal,etal.,IEEE-TCAD,19ContestAConcurrenttest60ContestResult:s537835PIs,49POs,179FFs,4,603faults.Synchronous,singleclock.Contest75.5%01,72257,5323min.*9min.*Randomvectors67.6%057,532--09min.Gentest**72.6%122490--4.5hrs.10sec.FaultcoverageUntestablefaultsTestvectorsTrialvectorsusedTestgen.CPUtime#Faultsim.CPUtime##SunUltraII,200MHzCPU*Estimatedtime**Time-frameexpansion(highercoveragepossiblewithmoreCPUtime)ContestResult:s53783613.2.3GeneticAlgorithms(GAs)

遗传算法Theoryofevolutionbynaturalselection(Darwin,1809-82.)C.R.Darwin,OntheOriginofSpeciesbyMeansofNaturalSelection,London:JohnMurray,1859.J.H.Holland,AdaptationinNaturalandArtificialSystems,AnnArbor:UniversityofMichiganPress,1975.D.E.Goldberg,GeneticAlgorithmsinSearch,Optimization,andMachineLearning,Reading,Massachusetts:Addison-Wesley,1989.P.MazumderandE.M.Rudnick,GeneticAlgorithmsforVLSIDesign,LayoutandTestAutomation,UpperSaddleRiver,NewJersey,PrenticeHallPTR,1999.BasicIdea:Populationimproveswitheachgeneration.PopulationFitnesscriteriaRegenerationrules3.2.3GeneticAlgorithms(GAs)6GAsforTestGenerationPopulation:Asetofinputvectorsorvectorsequences.Fitnessfunction:Quantitativemeasuresofpopulationsucceedingintaskslikeinitializationandfaultdetection(reciprocaltocostfunctions.)Regenerationrules(heuristics):Memberswithhigherfitnessfunctionvaluesareselectedtoproducenewmembersviatransformationslikemutationandcrossover.GAsforTestGeneratio6StrategateResultss1423s5378s35932Totalfaults1,5154,60339,094Detectedfaults1,4143,63935,100Faultcoverage93.3%79.1%89.8%Testvectors3,94311,571257CPUtime1.3hrs.37.8hrs.10.2hrs.HPJ200256MBRef.:M.S.Hsiao,E.M.RudnickandJ.H.Patel,“DynamicStateTraversalforSequentialCircuitTestGeneration,”ACMTrans.

onDesignAutomationofElectronicSystems(TODAES),vol.5,no.3,July200.2StrategateResults643.2.4SpectralMethods

光谱方法ReplacewithcompactedvectorsTestvectors(initiallyrandomvectors)Faultsimulation-basedvectorcompactionStoppingcriteria(coverage,CPUtime,vectors)satisfied?Extractspectralcharacteristics(e.g.,Hadamardcoefficients)andgeneratevectorsStopAppendnewvectorsCompactedvectorsNoYes3.2.4SpectralMethods

光谱方法Rep6SpectralInformation

光谱信息Randominputsresemblenoiseandhavelowcoverageoffaults.Sequentialcircuittestsarenotrandom:SomePIsarecorrelated.SomePIsareperiodic.Correlationandperiodicitycanberepresentedbyspectralcomponents,e.g.,Hadamardcoefficients.Vectorcompactionremovesunnecessaryvectorswithoutreducingfaultcoverage:Reversesimulationforcombinationalcircuits(Example5.5.)Vectorrestorationforsequentialcircuits.Compactionissimilartonoiseremoval(filtering)andenhancesspectralcharacteristics.SpectralInformation

光6SpectralMethod:s5378Simulation-basedmethodsTime-frameexpansionSpectral-method*StrategateContestHitecGentestFaultcov.79.14%79.06%75.54%70.19%72.58%Vectors73411,5711,722912490CPUtime43.5min.37.8hrs.12.0min.18.4hrs.5.0hrs.PlatformUltraSparc10UltraSparc1UltraIIHP9000/J200UltraII*A.Giani,S.Sheng,M.S.HsiaoandV.D.Agrawal,“EfficientSpectralTechniquesforSequentialATPG,”Proc.IEEEDesignandTestinEurope(DATE)Conf.,March200.2SpectralMethod:s5378674Summary

小结Testabilityapproximatelymeasures:Difficultyofsettingcircuitlinesto0or1DifficultyofobservinginternalcircuitlinesUses:AnalysisofdifficultyoftestinginternalcircuitpartsRedesigncircuithardwareoraddspecialtesthardwarewheremeasuresshowbadcontrollabilityorobservabilityGuidanceforalgorithmscomputingtestpatterns–avoidusinghard-to-controllinesEstimationoffaultcoverage–3-5%errorEstimationoftestvectorlength4Summary

小结Testabilityapprox684Summary(Cont.)

小结(续)TestBridging,Stuck-at,Delay,&TransistorFaultsMusthandlenon-Booleantri-statedevices,buses,&bidirectionaldevices(passtransistors)HierarchicalATPG--9Timesspeedup(Min)Handlesadders,comparators,MUXesComputepropagationD-cubesPropagateandjustifyfaulteffectswiththeseUseinternallogicdescriptionforinternalfaultsResultsof40yearsresearch–mature–methods:PathsensitizationSimulation-basedBooleansatisfiabilityandneuralnetworks4Summary(Cont.)

小结(续)TestB694Summary(Cont.)

小结(续)CombinationalATPGalgorithmsareextended:Time-frameexpansionunrollstimeascombinationalarrayNine-valuedlogicsystemJustificationviabackwardtimeCycle-freecircuits:Requireatmostdseqtime-framesAlwaysinitializableCycliccircuits:Mayneed9Nfftime-framesCircuitmustbeinitializablePartialscancanmakecircuitcycle-free(Chapter14)Asynchronouscircuits:HighcomplexityLowcoverageandunreliabletestsSimulation-basedmethodsaremoreuseful(Section8.3)4Summary(Cont.)

小结(续)Combin704Summary(Cont.)

小结(续)FaultsimulationisaneffectivetoolforsequentialcircuitATPG.Testscanbege

温馨提示

  • 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
  • 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
  • 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
  • 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
  • 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
  • 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
  • 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。

评论

0/150

提交评论