MIL-HDBK-217 vs.HALT/HASS
MIL-HDBK-217vs.HALT/HASS
byBarryMaandMekonenBuzuayene,Anritsu
Forthelastthreedecades,MIL-HDBK-217hasbeenwidelyusedtopredictproductreliability.1Today,however,highlyacceleratedlifetesting(HALT)andhighlyacceleratedstressscreening(HASS)arebeingrecognizedaseffectivetoolstointensifyproductreliability.2ThemilitarystandardandHALT/HASScoverdifferentareasinthereliabilityworld.Isthereanycorrelationbetweenthem?
ManufacturersusuallymakereliabilitypredictionsbasedonfailuremodelsdescribedinMIL-HDBK-217,BellcoreTR-332,orsomeothermodelbeforetheproductismanufacturedormarketed.3,4Butwhenaproductisdeliveredtocustomersandthenfieldfailurereportsbegintoarrive,thepreliminaryreliabilitypredictionsometimesisnotvalidatedbyreal-worldfailurereports.
Somemanufacturershavesaidthepredictionmodelcouldbewidelyinaccuratewhencomparedwiththeperformanceinthefield.4,5Whatmakesthediscrepancybetweenthereliabilitypredictionandthefieldfailurereport?
ThePurposeofMIL-HDBK-217
Thismilitarystandardisusedtoestimatetheinherentreliabilityofelectronicequipmentandsystems,basedoncomponentfailuredata.Itconsistsoftwobasicpredictionmethods:
•Parts-CountAnalysis—Requiresrelativelylittleinformationaboutthesystemandprimarilyusesthenumberofpartsineachcategorywithconsiderationofpartqualityandenvironmentsencountered.Generally,themethodisappliedintheearlydesignphase,wherethedetailedcircuitdesignisunknown,toobtainapreliminaryestimateofsystemreliability.
•Part-StressPrediction—Usescomplexmodelscomposedofdetailedstress-analysisinformationaswellasenvironment,qualityapplications,maximumratings,complexity,temperature,construction,andanumberofotherapplication-relatedfactors.Thismethodtendstobeusedneartheendofthedesigncycle,aftertheactualcircuitdesignhasbeendefined.
Thegeneralfailuremod-elinMIL-HDBK-217andBellcoreTR-332isoftheform:
where:b=thebasefailurerate,describedbytheArrheniusequation
Q,…=factorsrelatedtocomponentquality,environment,andapplicationstress
TheArrheniusequationillustratestherelationshipbetweenfailurerateandtemperatureforcomponents.Itderivesfromtheobserveddependenceofchemicalreaction,gaseousdiffusion,andmigrationratesontemperaturechanges:
where:b=processrate(componentfailurerate)
E=activationenergyfortheprocess
=Boltzmann’sconstant
T=absolutetemperature
K=aconstant
Detailedmodelsareprovidedforeachparttype,suchasmicrocircuits,transistors,resistors,andconnectors.
TheMeritofHALT/HASS
HALTisperformedduringdesigntofindtheweakreliabilitylinksintheproduct.Theappliedstressestotheproductarewellbeyondnormalshipping,storage,andapplicationconditions.HALTconsistsof:
•Applyingenvironmentalstressinstepsuntiltheproductfails.
•Makingatemporarychangetofixthefailure.
•Steppingstressfurtheruntiltheproductfailsagain,thenfixit.
•Repeatingthestress-fail-fixprocess.
•Findingfundamentaloperationalanddestructlimitsoftheproduct.
HASSisperformedintheproductionstagetoconfirmthatallreliabilityimprovementsmadeinHALTaremaintained.Itensuresthatnodefectsareintroducedduetovariationsinthemanufacturingprocessandvendorparts.Itcontainsthefollowing:
•Precipitationscreenforlatentdefectstobeturnedintopatentdefects.
•Detectionscreentofindpatentdefects.
•Failureanalysis.
•Correctiveactions.
TheprecipitationanddetectionscreenlimitsofHASSarebasedonHALTresults.Usually,theprecipitation-screenlimitsarelocatedbetweenoperationallimitsanddestructlimitsandthedetectionscreenlimitsbetweenspeclimitsandoperationallimits,asshowninFigure1.3
Figure1.HassLimitsSelectedFromHALTData
HALT/HASShasbeenproventofindlatentdefectsthatwouldverylikelyprecipitateinend-useapplications,causingproductfailuresinthefield.Asaresult,theHALT/HASSprocesscaneffectivelyintensifyproductreliability.
WhyMIL-HDBK-217TurnsOutInaccuratePredictions
ThepredictiontechniquesdescribedinMIL-HDBK-217forestimatingsystemreliabilityarebasedontheArrheniusequation,anexponentiallytemperature-dependentexpression.Butmanyfailuremodesintherealworlddonotfollowtheequation.
Forinstance,mechanicalvibrationandshock,humidity,poweron/offcycling,ESD,anddielectricbreakdown—allindependentoftemperature—arecommoncausesoffailure.Evensometemperature-relatedstresses,suchastemperaturecyclingandthermalshock,wouldcausefailuresthatdonotfollowtheArrheniusequation.
Moreimportantly,thereliabilityofcomponentsinmanyelectronicsystemsisimproving.Consequently,componentfailurenolongerconstitutesamajorreasonforsystemfailure.But,theMIL-HDBK-217modelstilltellsushowtopredictsystemreliabilitybasedonpartfailuredata.
Figure2illustratesthenominalpercentageoffailuresattributabletoeachofeightpredominantfailurecauses,basedondatacollectedbytheReliabilityAnalysisCenter.6ThedefinitionsoftheeightfailurecausesinFigure2areasfollows:
Parts—22%:Partfailingtoperformitsintendedfunction.
Design—9%:Inadequatedesign.
Manufacturing—15%:Anomaliesinthemanufacturingprocess.
SystemManagement—4%:Failuretointerpretsystemrequirements.
Wear-Out—9%:Wear-out-relatedfailuremechanisms.
NoDefect—20%:Perceivedfailurethatcannotbereproduceduponfurthertesting.Thesefailuresmayormaynotbeactualfailures;however,theyareremovalsandcounttowardthelogisticfailurerate.
Induced—12%:Anexternallyappliedstress.
Software—9%:Failuretoperformitsintendedfunctionduetoasoftwarefault.
Toillustratethedisparity,considerthefollowing:Acircuitboardcontaining338componentswithsixcomponenttypesisusedinamobileradiosystem.4ThefailurerateoftheMIL-HDBK-217predictionis1.934failurespermillionhours,asshowninTable1.Thefieldbehavioroftheboard,however,shows19failuresinatotaloperatingtimeof4,444,696hours,resultinginafieldfailurerateof4.274failurespermillionhours.Thedeviation4.274-1.934=2.34failurespermillionhourswasnotcoveredbytheMIL-HDBK-217prediction.
Table1.ContributiontoFailureRateofEachComponentinMIL-HDBK-217Prediction
ComponentCeramic
CapacitorDiodeBipolar
ICResistorBipolar
TransistorTantalum
CapacitorFailureRate
Calculated
Failures0.0040.0090.050.0521.2250.5941.934
Actually,manyfieldfailuresarecausedbyunpredictablefactors,oftenthemainreasonsforreliabilityproblemsintoday’selectronicsystems.Butthoseunpredictablereasonscanbesuccessfullyprecipitated,detected,andeliminatedduringaHALT/HASSprocess.
Conclusion
Beforemakingareliabilityprediction,becertainofoneofthetwofollowingitems:
a.Thefailuremodesdescribedinthepredictionmodelaccountforthevastmajorityofsystemfailures.Ifnot,gotob.
b.PredictionismadeafterreducingunpredictabledefectsbyperformingHALT/HASS. FYR
Thereliabilityportionsofourcontractsoftenspendconsiderabletimeaddressingreliabilitypredictions.Areliabilitypredictionmayhavelittleornothingtodowiththeactualreliabilityoftheproductandcanactuallyencouragepoordesignpractices.Asoneexample,whenninecontractorscameinwithseparateradiodesignsandpredictions,subsequenttestingshowedthatthereliabilitypredictionsrangedfrom30percentto3900percentoftheactualvalues.Contractorsandsubcontractorsthatfrequentlyquotepredictionsmaynotunderstandtheengineeringanddesignconsiderationsnecessarytominimizeriskandtoproduceareliabledesign.Inmanycases,thepersonproducingthepredictionmaynotbeadirectcontributortothedesignteam.Thehistoricfocusontheaccountingofpredictionsversustheengineeringactivitiesneededtoeliminatefailuresduringthedesignprocesshassignificantlylimitedourabilitytoproducehighly-reliableproducts.Highreliabilityisnotobtainedthroughreliabilitypredictions.
AMSAAandATEChavedevelopedalistofreliabilitybestpractices.Thesepracticescanbeusedtohelpdiscriminatebetweencontractorswhenitcomestoreliability.Inadditiontotheengineeringmodelingandtestingactivitiesdiscussedabove,agoodcontractorwill:
·havedevelopedengineering-levelcharacterizationsoftheenvironment(i.e.shock,thermal,vibrations)thattheirequipmentwillsee;
·havedevelopedinnovativeapproachestoreliabilitytoincludedesigningforsimplicityandforpit-stoprepair;
·haveaclosed-loopandresponsivefailurereporting,analysis,andcorrectiveactionsystem(FRACAS)program;
·notviewreliabilityasjustaMeanTimeBetweenFailure(MTBF)butinsteadfocusonbuildingaproductthathasasignificantfailure-freeoperating
period
·haveanalyzedthepotentialforobsolescenceofparts;and,
·haveanalyzedtheirsupplychaintoincludeunderstandingthereliabilityrisksanddesign/manufacturingpracticesofthecomponentandsubassembly
suppliers.
~fromhttp://www.amsaa.army.mil/ 1&2楼内文可译成中文方便理解吗?
强壮论述与MTBF成正比 呵呵，学习了，感谢分享哈。 很不错的读物，加深了我对HALTandHASS的理解，谢谢楼主。 怎我看的,所有英文字全連在一起,沒有空格:Q
页:
[1]