后端設計教學課件_第1頁
后端設計教學課件_第2頁
后端設計教學課件_第3頁
后端設計教學課件_第4頁
后端設計教學課件_第5頁
已閱讀5頁,還剩58頁未讀, 繼續(xù)免費閱讀

下載本文檔

版權說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權,請進行舉報或認領

文檔簡介

第十二章后端設計第十二章后端設計OutlinesBackendDesignFlowFloorplanPlace&RoutePhysicalVerificationSignalIntegrityDFM/DFYOutlinesBackendDesignFlowStepsofBackend/PhysicalDesignSynthesisFloorPlanningPlacementScanchaininsertionandre-ordering(optional)ClockTreeSynthesisRoutingParasiticandnetlistextractionPoweranalysisSignalintegritycheckingFinaltiminganalysis(STAandsimulation)ECO(optional)LVS/DRCExportGDSIILVS/DRCusingsign-offtoolsStepsofBackend/PhysicalDesBackendFlowwithECOEngineeringChangeOrder(ECO)Achievedbyaddingsmallamountofcellsinlimitedarea,sizingbuffersandroutingtheconnectionsPreventdisturbingtheplacementandroutingoftherestofthechipBackendFlowwithECOEngineerKeepinmind:

Performance,Power,Size,ReliabilityItisnotimpossibletodevelop“plug&Play”toolsKeepinmind:FloorplanningBasedonnetlist,createareasoffunctionalityonyourchipDeterminetheplacementofblocksDeterminetheplacementofI/OpinsDeterminethepowersupplystrategyGivefeedbackonhoweasyyourfloorplanmightbetowire(Globalrouting)andhowbigthechipisFloorplanningBasedonnetlist,ChipFloorplanningConsiderationsChiplevelfloorplanningHighspeedblockissueLocationaffectthetimingperformanceAnalogblockissuecleanVdd/Vss;minimalspacingtodigitalblock;IOlocationDiesizeissuePinlimited;CorelimitedPower-GroundroutingissuePowerringwidthaccordingtopoweranalysisPowerstrip/meshspacingPinplacementandIOringissue(willbetalkedinnextclass)Padpitchvs.boundingrule;ESD;noiseisolation;ChipFloorplanningConsideratiDieSizeIssue–cont.Determinetheareaforstandardcells“Utilization”–70%?80%?90%?ExtraspaceforclktreesynthesisExtraspaceforscanchainLayersforroutingDieSizeIssue–cont.DetermineHardMacroPlacementMacrosaregenerallyplacedaroundtheperipheralI/Oring

Acontiguousareaforstandardcells.Higherfreedomforyourplace-and-routetoolsduringplacementandroutingofthestandardcellsThegoalofmacroplacementisto:Reducetiming-criticalpathsbetweenthemacrosandinterfacinglogic. Reduceinterconnectionsinthefollowingorder:ChipI/OtomacrosMacrotomacroMacrotostandardcellblocksHardMacroPlacementMacrosarePower/GroundDevelopmentIRDropandElectromigrationPower-netIRdropdegradesthesupplyvoltagelevelExcessivecurrentdensityinmetalwirecauseselectromigrationfailurewhichbreaksmetalconnectionMoresignificantIRdropeffectwhenVddgetssmallerHighercurrentdensitywhenmetalwirewidthissmallerPower/GroundDevelopmentIRDroPower/GroundDevelopment-cont.RingstructurePowerringsaroundalllayoutblocksMajorpowertrunksbetweenlayoutblocksDifficulttoguaranteetheworstIRdropStrapstructureSimple,easyforroutingMeshstructureEvenlydistributeofIRSpacingofPowerstripsconsideration

IRdropanalysisFixtheprobleminearlystageP/GStructuresPower/GroundDevelopment-cont.BeWareofMaximumWidthRuleMaximumwirewidthlimitduetothermalstressandlocaldensityrulesSlottingvs.“bus”ofthinwiresDisadvantageofslotting:SlotsmaynotbealignedwithcurrentflowTrueIRdropnotknownuntilafterslottingEspecialhappenforPower/GroundringsM1M1VS.GNDGNDGNDGNDCommonlyusedforpower/groundBeWareofMaximumWidthRuleMPlacementBasedonagivenfloorplan,determinethelocationofcellsinagivennetlistGoals&objectivesRoutabilityGuaranteetheroutercancompletetheroutingstep(Globalrouting)TimingMinimizeallcriticalnetdelaysMinimizediesizeMakethechipasdenseaspossibleSignalIntegrityPlacementBasedonagivenfloCheckfeasibilityofroutingafterplacementLogiceffort-forthosepathswithpositiveslack,reducecellsizeCongestionandFixBeforeAfterCongestionareasCheckfeasibilityofroutingaRoutingCompletepower/ground/clockrouting(clocktreesynthesis)Completedetailedwirerouting,conformwiringruleandorder)ImprovethedensityMinimizethelayerchangesImprovecriticalpathandmeettimingrequirementProducearouteddesignfreeofDRC/LVSviolationsRoutingCompletepower/ground/cGeneralRoutingFlowClockTreeSynthesisAddbuffers/inverters,minimizeclkskewanddelayPostPlacementOptimization(PPO)FixsetupviolationPre-RouteStandardCellsVDD/VSSrailsonmetal1VerifyPGconnectionandroutingRouteGroupNetclocksbusroutingPost-RouteCTOFixclkskewandinsertiondelayGlobalRoutingcriticalpathlongwire,interconnectionGeneralRoutingFlowClockTreRoutingflow–cont.TrackAssignment&DetailRoutingWireconnectionSearch&Repair(DRC/LVS)fixroutingviolation(unconnectednets,shorts)PostRouteOptimizationFixtimingCoarseLVS&DRCcheckingmetalwidth,notch&gapcheckingDataOutputstreamout:gds2formatverilogout:hierarchy(PT)/non-hierarchy(forHercules)parasiticout:spefformat(cellview)Routingflow–cont.ClockTreeSynthesisObjective:minimizeclockskewoptimizeclockbuffersClockTreeSynthesisObjectiveBasicCTSFlow

&

Concepts

BasicCTSFlow

&

ConceptsClockConstraintDefine:Clksource:rootpin,targetinsertiondelay,targettransitiontimeatclkportClkendpoint:Synchronouspin,ignore/excludepinDrivingcell,clkcell,delaycell:buffers,inverters,specialclkcellsDRC:maximumtransitiondelay,maximumnetcapacitance,maximumfanout,clknumberofbufferlevelsClockConstraintDefine:ClockSkewGlobalSkewandLocalSkewGlobalskewGlobalskewistheclockarrivaltimedifferencebetweenanytwoflip-flops.LocalskewLocalskewistheclockarrivaltimedifferencebetweentwoflip-flopsthatareadjacentthroughcombinationallogic.ClockSkewGlobalSkewandLocConceptofUsefulSkewUsefulskewisamethodofintentionallyskewingaclocktoimprovethetimingonacircuit.ItisalsocommonlyusedinECOWarning:Couldcauseproblemin

DFTscaninsertion

ConceptofUsefulSkewUsefulsUseCTSforHigh-FanoutNetSynthesisHigh-fanoutpins:rest,scan_enNeedtobalancehigh-fanoutpinstoguaranteethefunctionalityUsingCTStool:high-fanoutnetsbyinsertingabalancedbuffertreeTominimizebothskewandinsertiondelayButshouldavoidusinglargebuffersforpowersavingUseCTSforHigh-FanoutNetSyLargeSoCClockDistributionPartitionthedesigntoseveralblocksCTSforeachblockClktreenetworkattoplevelExternalclockIPCore

orModuleCoreInternalClockNetPLLGlobalClockNetLargeSoCClockDistributionPaHTreeforTopClockNetworkUsebigbuffertobalancedelayandclkskewEqualdistance,equalloads,equaldrivingabilityHTreeforTopClockNetworkUsClockDistributionCaseStudy:Pentium?SpinesKurdetal.,AmultigigahertzclockingschemeforthePentium?4microprocessor,JSSC2001ClockDistributionCaseStudy:ClockDistributionCaseStudy:

Intel’sItanium?HTreeClockingTametal.,ClockgenerationanddistributionforthefirstIA-64microprocessor,JSSC2000ClockDistributionCaseStudy:IssuesLargeamountofclockbuffersaddedonclocktreePowerconsumptionNoisetosupplylinesReducepowerconsumptionWidewirewidthsClockgatingcellplacementLimitationofusinglargeclockbuffercellsReducenoiseSpecialclockbuffercellswithdecouplingcapacitorIssuesLargeamountofclockbuExtractionWhencompletedetailedrouteWriteoutthehierarchicalnetlistandparasiticforbackannotationDatamanagementonhugefileofextractedparasiticdataAccurateRCandtimingmodelfornanometerdesignWidthandspacingdependenceResistanceshieldingLocaldensityeffectExtractionWhencompletedetailSDFBackAnnotationUsedincell-baseddesignflowPerformsdelaycalculationonparasiticRCsininterconnectwiresDSPF-DetailedStandardParasiticFormatSPEF–StandardParasiticexchangeFormatSDF-StandardDelayFormatusedforpost-layoutsimulationCanbeconvertfromPrimeTimeSDFBackAnnotationUsedincelPhysicalVerificationDRC-DesignRulecheckVerifythemanufacturingrules,example:InternallayerchecksWidemetalchecksMetalslottingneededforwidemetalLayer-to-layerchecksDFM/DFYExample:AntennaRuleCheckLVS–Layoutvs.SchematicsComparelayouttoschematics-everycellandnetPhysicalVerificationDRC-DeDRCTrendsandChallenge>75%timeonmetallayerandviacheckERC-typechecksincreasingRiseofpre-tapeoutDFMutilitiesDRCTrendsandChallenge>75%tLVSLayoutvs.Schematic(LVS)CheckphysicallayoutagainstfunctionalgatelevelschematictoensureallintendedconnectivityhasbeenmaintainedSteps:Extractthenetlistfromlayout(GDSII)ComparethenetlistwiththeoneafterroutingandoptimizationHints:

MostofLVSerrorsarecausedbymanuallayoutorcongestion“Virtualconnect”(connectedbytext)couldcauseakillerfailureLVSLayoutvs.Schematic(LVS)SignalIntegritySignalIntegrityistheabilityofasignaltogeneratecorrectresponseinacircuitSignalhasdigitallevelsatappropriateandrequiredvoltagelevelsatrequiredinstantsoftimeCrosstalk,IRDrop,ElectromigrationSignalIntegritySignalIntegriLayoutParasiticvs.CircuitPerformanceInterconnectparasiticresistors,capacitorsandinductorscauseextratimingdelayAdditionalpowerconsumptioncausedbyparasiticRCInter-wirecapacitancescausecouplingnoiseandwilldominateinterconnectwiredelaysParasiticresistancesinpowersupplycausevoltagedropandmaydegradecircuitperformanceHighercurrentdensityinpowernetmaycauseelectromigrationfailureLayoutParasiticvs.CircuitPInductanceEffectsInductivecouplingeffectissignificantforlonginterconnectsandforveryfastsignaledgerateInductivecouplingisnegligibleatshorttraceinterconnects,sincetheedgetraceislongcomparedtotheflighttimeofthesignalInductanceextractionandsimulationaremoredifficultthancapacitanceCLInductanceEffectsInductivecoCrosstalkAnalysisDefinitionAggressor:generatingcrosstalkVictim:receivingcrosstalkTimingsensitiveCrosstalkanalysisconsistingsignaltransitiontimingwindowcaneliminatepessimisticdelaycalculationThecrosstalkspikeisrelatedtocapacitancevalueandthevictimdriverimpedanceCrosstalkAnalysisDefinitionCrosstalkAnalysis–cont.TimingsensitiveCrosstalkAnalysis–cont.TimiCrosstalkPreventionPreventcrosstalkfromsynthesisstageMinimizethedrivingsizeonthosenon-criticalpathtoreducethenumberofaggressorsApplymaxtransitiontime(set_max_transition)inphysicalsynthesis/placementtoavoidlongnetsCrosstalkPreventionPreventcrCrosstalkPrevention–cont.FromroutingstageEffectivespacingbetweennoiseregionandquiteregionShieldingbetweencriticalpathsCrosstalkPrevention–cont.FrCrosstalkPrevention–cont.Fromroutingstage–cont.BufferinsertionInsertedbufferbreaksupthecouplingcapacitanceoflongwireCrosstalkPrevention–cont.FrCrosstalkPrevention–cont.Fromroutingstage–cont.BuffersizingIncreasethedriversizeofvictimDecreasethedriversizeofaggressorTrackreorderingTrackreorderingisbasedontimingwindowCrosstalkPrevention–cont.FrCrosstalkPrevention–cont.ForinductancecrosstalkCoplanarShieldsReferencePlanStaggerInverter/BufferCrosstalkPrevention–cont.FoElectromigrationEffectsTheelectronsflowthroughthewiresandcollidew/metalatoms,producingaforcethatcausesthewirestobreakCausedbythehighcurrentdensitiesandhighfrequenciesgoingthroughthelong,verythinmetalwiresMTTF(MeanTimeToFailure)increaseswhencurrentdensityandtemperatureincreaseCanbeeliminatedbyusingtheappropriatewiresizingElectromigrationEffectsTheelFixEMControllingcurrentdensitytolimitelectromigrationfailureisneededindesignandverificationLayoutoptimization:Increasethepowerlinewidth,layerIncreasethepowerpadsIncreasetheconnectionIssuesMoremetal(add8%costperlayer)Larger,slowerdesigns(growinxandy)FixEMControllingcurrentdeOtherConsiderationsESD(willbetalkedinnextclass)Packagevs.performance(willbetalkedinnextclass)DFM/DFYOtherConsiderationsESD(willDFM/DFY90nmandbelowtechnologieschallengesinyieldDFM–DesignforManufacturabilityDFY–DesignforYieldDFM/DFY90nmandbelowtechnoloDFMandDFYDFMisthemanagementoftechnologyconstraints(sizingrules)appliedtothelayoutAmanufacturabledesignhoweverisnotnecessarilyahigh-robustorhigh-yieldingdesign.DFY,aspartofDesignforManufacturability,concentratesonthedevelopmentandqualityofthecircuitdesigninthepre-andpost-layoutphase.DFYisthemanagementofdesignsensitivitiestothemanufacturingprocessandhelpstoguaranteehigh-yieldingdevicesDFMandDFYDFMisthemanagemeDFM/DFYMethodologyOptimalresolutionenhancementtechnology(RET)MaskandexposureOpticalProximitycorrection(OPC)PhaseShaftMask(PSM)YieldenhancementandoptimizationtechnologyDFMrulesimplementationToovercomelimitsofOPCYieldcheckingduringthelayoutstageSupportedbyEDAtoolsDFM/DFYMethodologyOptimalresWhyNeedRET?WavelengthusedvsprocessgenerationWhyNeedRET?WavelengthusedvDesignforManufacturingNotallthethingscanbedonebymaskandexposure:CorrectionsarenotcompleteSomedesignscannotbebuiltatallwithcertainRETtechnologiesOfthosethatCANbebuilt,somearemoremanufacturableafterRETthanothersDFM/DFY-drivenroutingOPC-drivenroutingPSC-drivenplacementDFMruleimplementationDesignforManufacturingNotaDFM/YRulesLimittheuseofminimalpoly-enclosedgates,minimallyenclosedviasandsinglycontactedlinesBetteryieldLessresistanceExample:ViaVoidrules-doubledviasDFM/YRulesLimittheuseofmiCurrentDFM/YDesignFlowSupportedLoadDesignPerformantenna

fixesAddcontacts/viaMetalFill&SlottingVerifyLVSandDRCCurrentDFM/YDesignFlowSuppWhyNeedDoubleVias?CopperprocessingcausesnewproblemsforviasVoidsinCumigrateunderthermalstresstowardsviasIfenoughvoidsmigratetoaviaitcancausefailureWorseat90/65nmduetoincreasedstressofsmallerviaVoidscanmigratelongdistances~10micronsVoidscanmigratearoundcornersWhyNeedDoubleVias?CopperprYieldvs.AreaYieldvs.AreaAntennaRulesAntennaruleshavenothingtodowithtraditionaldefinitionofantennaReallyacollectorofstaticcharge,notelectromagneticradiationAntennaproblemonlyhappensduringmanufacturingPlasma-basedprocessforetching,oxidedepositionPlasmaetcherincludeavoltageintofl

溫馨提示

  • 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會有圖紙預覽,若沒有圖紙預覽就沒有圖紙。
  • 4. 未經(jīng)權益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫網(wǎng)僅提供信息存儲空間,僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護處理,對用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對任何下載內(nèi)容負責。
  • 6. 下載文件中如有侵權或不適當內(nèi)容,請與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準確性、安全性和完整性, 同時也不承擔用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。

評論

0/150

提交評論