Professional Documents
Culture Documents
DesignForTest:DFTQ&APart2
0
More NextBlog
CreateBlog SignIn
DesignForTest
Wednesday,September11,2013
SearchThisBlog
DFTQ&APart2
Search
6.Ifweprovidecontrollabilityandobservabilitytotheportsofareaundertest,willit
improvethetestcoverage?Ifyes,howmuch%ageofimprovementcanwesee?
Labels
Ans:yes,wecanseeanimprovementincoverage.Thinkabouttheboundinglogicthatwe
usuallygiveformacros...
theimprovementincoveragewilldependontheinitialcoveragewithoutboundandthenooffaults
inthedesignundertest.
7.WhenboththeTransitiondelayandpath_delaytargetfortransitionfaults,whyweshould
havetwodifferentmethodstodetectthesame?
DFTQ&A(29)
PopularPosts
DFTQ&APart19
Ans:YesbothTransitionandPathdelaywilltargetthetransitionfaults,buttheyaredifferentin
thefollowingways
1)Transitiondelayisrelatedtoslowtoriseorslowtofallfaultsataparticularnode.Whereas
pathdelayisrelatedtoslowtoriseorslowtofallfaultsofaparticularpath.
2)Thereasonfortransitiondelayatanodeissomemanufacturingdefectatthatnode(more
resistivenode).Thereasonforpathdelayissomemanufacturingdefectthatisdistributedthrough
outthepath(moreresistivepath).Letmeexplainthisindetailwithanexample.
Letusassumethatinapaththerearesome10nodes,andthetransitiondelayateachnodeis
thatmuchlesswhichwon'tcreateanytimingviolationinthetransitionATPG.Butinpathdelay
thesedelaysateachnodemightgetaccumulatedandresultinnotmeetingthetiming.Soitis
musttohavepathdelaybesidesTransitiondelay.
DFTQ&Apart1
DFTQ&APart2
DFTQ&APart25
DFTQ&APart3
FollowbyEmail
Emailaddress...
Transitiondelayissimilartostuckatatpg,exceptthatitattemptstodetectslowtoriseandslow
tofallnodes,ratherthanstuckat0andstuckat1nodes.Aslowtorisefaultatanodemeans
thatatransitionfrom0to1onthenodedoesntproducethecorrectresultsatthemaximum
operatingspeedofthedesign.Similarlyaslowtofallfaultmeansthatatransitionform1to0ona
nodedoesntproducethecorrectresultsatthemaximumspeedofthedesign.Transitiondelay
faulttargetssinglepointdefects.
Submit
Archives
2010(6)
2013(53)
September(29)
ThePathdelayfaultmodelisusefulfortestingandcharacterizingcriticaltimingpathsinthe
design.Pathdelayfaulttestsexercisecriticalpathsatspeedtodetectwhetherthepathistoo
slowbecauseofmanufacturingdetectsorvariations.
DFTQ&Apart1
Pathdelayfaulttestingtargetsphysicaldefectsthatmightaffectdistributedregionofachip.For
example,incorrectfieldoxidethicknessescouldleadtoslowersignalpropagationtimes,which
couldcausetransitionsalongacriticalpathtoarrivetoolate.
DFTQ&APart4
Ifweattain100%coveragewithtransitionatpgtestthenwedontneedtodothepathdelaytest.
Butthisisnotthecaseinmostofourdesigns.ThenhowcanweGUARANTEEthatallthe
criticalpathsaretargetedbyTransitionATPG
DFTQ&APart7
sowegivethecriticalpathtotheatpgtoolandtellittogeneratepatternsforit
DFTQ&APart2
DFTQ&APart3
DFTQ&APart5
DFTQ&APart6
DFTQ&APart8
DFTQ&APart9
DFTQ&APart10
DFTQ&APart11
thisiscalledpathdelay.
8.WhatisBurnintest?Whyisitdone?
Ans:Burninistheprocessbywhichdeviceinfantmortalityfailuresare
DFTQ&APart12
DFTQ&APart13
DFTQ&APart14
DFTQ&APart15
DFTQ&APart16
acceleratedthroughapplicationoftemperatureandstressvoltagesfor
specificperiodsoftime.Thegoalofburninistostressthedeviceasmuch
DFTQ&APart17
DFTQ&APart18
DFTQ&APart19
aspossibletherebyacceleratingdeviceinfantmortalityrate.Sincescan
http://vlsidoubts.blogspot.in/2013/09/dftqpart3.html#gsc.tab=0
DFTQ&APart20
1/5
10/5/2016
DesignForTest:DFTQ&APart2
basedtesthashighercoverage,scantestvectorsareusedtostressthelogic
portionofthedevice.Thiscaneitherbedonebyperformingdevicelevel
ATPGbyapplyingexternaltestpatternsononchipLBIST.MemoryBISTis
DFTQ&APart21
DFTQ&APart22
DFTQ&APart23
DFTQ&APart24
DFTQ&APart25
usedformemories.
Burnintestisthetechniqueofestimatingthelifetimeofthechipbyprovidingstressintheform
temperature,voltageandcurrent.
Hightoggledatpgvectorsisgeneratedtosensitizetheinternalnodesofthechipsothatchipwill
getmorestress
DFTQ&APart26
DFTQ&APart27
DFTQ&APart28
DFTQ&APart29
October(2)
December(22)
.BurninmoniorpinisavailabeforeverySoCtochecktheinternalstatusintheformof
pulses.Chipwillgivepulsestillthedeviceburnout..
Followers
9.WhydowehavedifferentkindsofSDF'slikePrimeTimeSDFandVerilogSDFwhen
thereisanOpenVerilogInternationalstandard?
Followers(55)Next
Ans:TherearenodifferentkindofSDFlikePrimetimeSDForVerilogSDF.
Whenadesignissynthesizeditcreatesnetlist,whichismadeoflibrarycellsonly.Thelibrary
cellshavepathdelays(InputtoOutput)andtherearealsointerconnectdelayinnetlist.SDF
meansStandardDelayFormat.SoSDFbasicallycontainsthecellpathdelayandInterconnect
delaysbetweencells.TheSDFformatcanberead/understoodbyallSTA/simulationtools.
Generarally(1)theSDFcanbegeneratedusingSynthesis(dc_shell)/STA(pt_shell).ThisSDFsare
usedforinitialTiminganalysisandgatesimulation.(2)postroutetoolsalsogeneratesSDF(or
spefandthenyoucangenerateSDFfromspef)files.TheseSDFareusedforfinalTiming
Analysisandfinalgatesimulationforsignoff/tapeout.
Eachlibarycellcanhavemax,minortypicaldelay.soyoucangeneratedSDFbasedonthis
delayandhavedifferentSDFformaxdelayormindelayortypicaldelay,butyoucannothave
primetimeSDForVerilogSDF.
10.IfforaDFTproductionsetwehavebothPATHDelayandTransitionpatternswhich
schemeshouldbedonefirst?
Ans:ItsalwaysbettertodoPath_delayfirstandthenthetransitiondelay.
Follow
TotalPageviews
1 6 4 6 8 5
vlsi
doubts.blogspot.in
Search
byYakkalaSrikanth
Search
Path_delaystartflopandendfloparegivenbytheuser.(i.e)transitionthroughawelldefined
path(Moreeffective)
TransitionStartandendflopisdecidedbythetool,sotrasitionmayhappenthroughashorteror
longerpath(Lesseffective)
1)pathdelaybasicallytargetsforallthecriticalpathsinthedesign.
2)Generatethepath_delaypatternsforallthecriticalpathsinthedesignandwritedownthefaults
forthesame.
JavaAssist
MethodCallduringPolymorphism
Passinganarraytoamethod
instanceof
HashMapVsHashtable
3)Generatethetransdelaypatternbyloadingthefaultsonthepath_delay,suchthatthefaults
detectedinthepath_delayarenotRetargeted.
4)Themajorreasontofollowtheabovesequenceisintransdelaywearenotsureweathera
transitionhasreallyhappenedthroughthecriticalpath.
(A)IfwedothetransitionATPGfirst,wearenotsurewhetherthetransitionhashappened
throughthecriticalpath,Insuchcasewemaynothavepatternwhichmaycatchthefaultthrough
thecriticalpath,butthefaultmaybedetectedthroughsomeothershorterpathswhichmayadd
thefaultstothedetectedlist.
(B)Ifwerunapath_delayaftertheabovestep(A)thenwemayhaveapatternforthecritical
path,butitleadstothepatternredundancybecausewehavealreadyatransition
patternforthesamefault.
11.WhatdowemeanbyfaultsimulationinDFTterminology?
Ans:Faultsimulationconsistsofsimulatingacircuitinthepresenceoffaults(logicalfaults).
Comparingthefaultsimulationresultswiththoseofthefaultfreesimulationof
thesamecircuitsimulatedwiththesameappliedtest,wecandeterminethe
faultsdetectedbythattest.
http://vlsidoubts.blogspot.in/2013/09/dftqpart3.html#gsc.tab=0
2/5
10/5/2016
DesignForTest:DFTQ&APart2
Thereareseveraltypesoffaultsimulationtechniquessuchasserialfaultsimulation,parallelfault
simulation,deductivefaultsimulatione.t.
12.WhatismeantbycoverageinDFTterminology?
Ans:Thetermcoveragecanbebraodlyclassifiedas:
1.FaultCoverage:Thisisthetotalnumberofdetctedfaults(duingtesting)dividedbythetotal
numberoffaultsinthedesign.
2.Testcoverage:Thisisthetotalnumberofdetectedfaultsdividedbythenumberoftestable
faultsinthedesign.
Testablefaultsarethefaultsinthedesignthatcanbetestedandobservedbythepatternsused
duringtesting.
Pleasenotethat,heretheword"faults"referstomanufacturingdefects,asDFTdoesnotmerely
targetfunctionalfaults,butratherfocussesoncircuitstructure.
CoverageCalculationsareusedtomeasuretheeffectivenessoftestpatternsandtestgeneration
foragivensetoffaults.
Ex
ABasicANDGatehassixfaults
SA1/SA0InputA
SA1/SA0InputB
SA1/SA0OutputY
TotalfaultsforabasicANDgateis6
Coveragecalculationgoesonhowmanyfaultsweareabletodetectonthesame.
TestCoverageisapercentagedetectedofalldetectablefaultsandgivesthemostmeaningful
measureoftestpatternquality.
TestCoverage=DT+(NP+AP)*PT_credit)/(totalfaultsUD(AN*AU_credit).
FaultCoverage:thefaultcoverageisthepercentagedetectedofallfaults.itgivesnocreditfor
undetectablefaults.
FaultCoverage=(DT+(NP+AP)*PT_credit)/totalfaults.
DTDetected
DRdetectedrobustly
DSdetectedbysimulation
DIDetectedbyimplication
PTPossibltydetected
APATPGuntestablepossiblydetected.
NPnotanalyzed,possiblydetected.
UDUndetectable
UUundectableunused
UTUndectabletied
UBundetectabletied
URundettableredundant
AUATPGuntestable
ANATPGuntestablenotdetected.
NDnotdetected
http://vlsidoubts.blogspot.in/2013/09/dftqpart3.html#gsc.tab=0
3/5
10/5/2016
DesignForTest:DFTQ&APart2
NCnotcontrolled.
NOnotobserved.
13.WhatisIDDQTesting?Whyisitdone?
Ans:IDDQTestingcandetectcertaintypesofcircuitfaultsinCMOScircuitsthataredifficultor
impossibletodetectbyothermethods.
IDDQtesting,whenusedwithstandardfunctionalorscantesting,providesanadditionalmeasure
ofqualityassuranceagainstdefectivedevices.
IDDQtestingreferstotheintegratedcircuit(IC)testingmethodbaseduponmeasurementof
steadystatepowersupplycurrent.IddqstandsforquiescentIdd,or
quiescentpowersupplycurrent.MajorityofICsaremanufacturedusingcomplementarymetal
oxidesemiconductor(CMOS)technology.Insteadystate,whenallswitchingtransientsare
settleddown,aCMOScircuitdissipatesalmostzerostaticcurrent.Theleakagecurrent
inadefectfreeCMOScircuitisnegligible(ontheorderoffewnanoamperes).However,incase
ofadefectsuchasgateoxideshortorshortbetweentwometallines,aconduction
pathfrompowersupply(Vdd)toground(Gnd)isformedandsubsequentlythecircuitdissipates
significantlyhighcurrent.Thisfaultycurrentisafewordersofmagnitude
higherthanthefaultfreeleakagecurrent.Thus,bymonitoringthepowersupplycurrent,onemay
distinguishbetweenfaultyandfaultfreecircuits.
WhydoIDDQTesting?
Forfunctionaltesting,atesterappliesasequenceofinputdataanddetectstheresultsinthe
sequenceofoutputdata.Then,theoutputsequenceiscomparedagainsttheexpectedbehavior
ofthedevice.Anadvantageoffunctionaltestingisthatitexercisesthedeviceasitwould
actuallybeusedinthetargetapplication.However,thistypeoftestinghasonlyalimitedabilityto
teststheintegrityofadevice'sinternalnodes.
withfunctionaltestingonly,aninternaldefectcouldslidebyundetected.
Themethodologyforscantestingisallthesequentialelementsofthedeviceareconnectedinto
chainsandusedasprimaryinputsandprimaryoutputsfortestingpurposes.Usingautomatictest
patterngeneration(ATPG)techniques,youhavethecapabilitytotestamuchlargernumberof
internalfaultsthanwithfunctionaltestingalone.ThegoalofATPGistosetallnodesofthecircuit
toboth0and1,andtopropagateanydefectstonodeswheretheycanbedetectedbytest
equipment.
Usingbothfunctionalandscantestingyougreatlyincreasesyouroddsatfindinganinternal
defect,butwhatifthedefectisnotcontrollableorcan'tbeobserved?ThatiswhereIDDQtesting
canhelp.
14.Ifoneneedstodosynthesis/STAwithscanreplacedFF(notstitched)andneeddo
generatetimingandotherreports.WhatshouldbevaluesofSE,SIandSOpinssince
designisnotstitched?
Ans:WeneednotconstraintheSE,SIandSOpinsforsynthesis/STAofascanreplacedbut
notstitcheddesign.ButwewillnotbeabletodoanytestrelatedSTA.
15.Canyoubrieflydescribethepointstobeconsidered,whilereorderingthescanchain
inPhysicalDesign?
Ans:Scanchainreorderingneedstorespect3importantLogicalconstraints.
1.Clockdomaintimingconstraints
2.Userspecifiedscansegmentpositions
3.Minimizingclockdomaintraversals
eachactiveedgeofeachclockisconsideredtobeinaseparateclockdomain.Bothedgesofa
clockandclockswithdifferenttimingsmaybeusedtocontroledgetriggeredscanflipflopsofa
scanchain.
Inordertoconstructfunctionalscanchains,twoconsecutivescanflipflopsAandB(Aserially
drivingB)
1)mustbeclockedatthesametimeor
http://vlsidoubts.blogspot.in/2013/09/dftqpart3.html#gsc.tab=0
4/5
10/5/2016
DesignForTest:DFTQ&APart2
2)BmustbeclockedbeforeA.
Inthefirstcase,wesaythatAandBhavecompatibleclockdomains.
Inthesecondcase,wesaythatAandBhaveincompatibleclockdomains.
Theprecedencerelationshipbetweenscanflipflopsimposedbyclockdomaintimingsis
translatedatthescansegmentlevel.Captureandlaunchtimesforascansegmentare
respectivelydeducedfromthecapturetimeofitsfirstscancell(drivenbyitsscaninput)andthe
launchtimeofitslastscancell(drivingitsscanoutput).Therefore,theprecedencerelationship
betweenscansegmentscanbeestablished,andthusrespectedduringscansegments
reordering.Userspecifiedscansegmentpositionsarerespectedduringscanreorderingunless
theyviolateclockdomaintimingconstraints.
Thelastconstraint,minimizingclockdomaintraversals,takespriorityonphysicaldesign
informationbecausewewantourapproachtobeminimallyintrusiveintermofadding
synchronizationlatches.Onlyscansegmentswithcompatibleclockdomainsarereordered.
Reorderingasetofscansegmentswithcompatibleclockdomainsconsistsof:
1.identifyingandmarkingthesetofclusterscontainingthescansegments.
2.Determiningtheentryandexitpointsbetweenwhichthescansegmentsaregoing
tobereordered.
3.Orderingthepreviouslyidentifiedclustersbetweentheentrypointandexitpoints.
4.Reorderingscansegmentswithineachoftheorderedclusters.
PostedbySreeatWednesday,September11,2013
<! Place this tag in your head or just before your close body tag > (0)
Reactions:
Linkstothispost
CreateaLink
NewerPost
Home
OlderPost
Didyoufindthisposthelpful
Yes
0(0%)
No
0(0%)
Canbebetter
0(0%)
Votessofar:0
Pollclosed
Simpletemplate.PoweredbyBlogger.
http://vlsidoubts.blogspot.in/2013/09/dftqpart3.html#gsc.tab=0
5/5