6_EMPro2013_and_EMIEMCDesignMethodologyusingS

EMPro2013&EMI/EMCDesignMethodologyusingSimulationsandMeasurementsPage1AvoidEMINightmaresAgilentEEsofEDA谢成诚[e-mail]•EMICOverview•IC,PKGandBRDEMICExample-1:IC–WhatimpactstheharmonicsExample-2:BRD–ImpactofgndstitchingExample-3:BRD–ImpactofCommon-modeNoise•SystemEMICExample-4:DDR2DigitalNoiseonRFReceiversensitivityExample-5:BRD–ImpactofShieldingExample-6:Connector–Impactof3DdiscontinuitiesExample-7:ImpactofICNoiseonPCB•EMIAnalysisFeaturesinEMPro2013Page2WhatisEMIC?GraphicSource:://ntuemc.tw/index.php?node=emc&content_id=3Source(Culprit,emitter)Receptor(Victim,Receiver)Transfer(CouplingPath)ConductedRadiated•PowerLines•SignalLines•Magnetic•Electric•PlanewavePage3ImprovingCHIPEMICperformanceNeedtomodelnon-idealICbehaviorandunderstanditsimpactonEMIC.-Rise/Falltimemismatch-Intra-pairdelay-Jitter(RJ,PJ,DCD,ISI)Page4Example-1:SimpleCKTsimulationintotestloadPage5StudyImpactofDriver-Rise/Falltimemismatch-Intra-pairDelay-PresenceofRandomJitterMatched/UnmatchedClock/DataPairsPSDPage6MatchedCLKPairUn-MatchedDataPairMatchedUn-MatchedUn-MatchedcasesshowstrongpresenceofevenharmonicsHowtoaddressIC&PKGEMI/EMCPage7ICPKGCLK/ADDR…HIGH-SPEEDDATA…Step1:IsolateCLK•Traceroutingclosetoslots-harshradiation•SizeofReferencePlane-patchantennaradiation•LackofGND-Stitching-hot-spotsthatbecomeradiatorsStep2:ModelHigh-SpeedDataSignals•Alloftheabove•Properreferencing•CouplingbetweenDataandCLKHIGH-SPEEDDATA…WhydoweneedFull-EMaccuracy?-Non-IdealReturnPathSlotsinreferenceplanesFinite-planeedgeeffectsNumberofGNDStitchingVias-ViasVia-to-ViaCouplingViaStubsAntipadeffect-NoiseCouplingfromReferenceplanetoSignalPage8TypicalEMICtoolsmakelotofassumptionsNOTmodelreal-worldeffectsBoardDesignforoptimumEMICSameflowasPKGminimize“hot-spots”BRDEMICperformanceishighlydependentonlayoutBadlayouthabits–SignalspassingbyoroverSlotsSignalschanginglayerswithoutproperGNDreferencingSignalschangingreferencingwithbaddecapsSSONoisepropagationonplanesduetolackofproperdecaps&perimeterstitchingviasonthePCBPage9Needtohelpcustomersisolateroot-causeofEMICproblemsExample#2:High-SpeedBoardPage10IsolatingtheproblemPage11InspectionoflayoutshowslackofGNDstitchingViasTakingCorrectiveAction…Page12-AddmoreGNDstitchingVias-ChangeDProutingtoToplayerandavoidVia-to-ViacouplingImprovedEMICresults…Page13MuchSmallerRadiationEmissionAntennaGainis-41dBSurfaceCurrent“hot-spots”aregonewiththeadditionofsufficientGNDstitchingViasandImprovedlayerroutingExample#3:DoesCommon-modeimpactEMIC?Page14Common-modecurrentinjectiontoconnectorPage15Example-4:DDR2DigitalNoiseonRFReceiverSensitivityPage16ClockFrequency:266MHzDataRate:533MbpsClocksignalcontainshighfrequencycomponentsaround2.45GHz,whichactsasthesourceofnoiseClockSignalSpectrumfptvConceptualApproachHighfrequencycomponentofDDRclocknoisemayinterferewithW-LAN,Bluetooth,andAntennaS-parameteranalysistechniquewillprovideinsightontheboard’sfrequencyselectivitytoDDRnoiseClockSignal(BroadbandInput)fpfpBoard‘sFrequencySelectivityResponsePCBStructure(Network)Page17CompleteBoardSimulationinEMProVerycomplexmulti-layerboardThousandsofviasDozensofICsHundredsofpassivecomponentsPage18TestScenarios3differenttestscenarioswithbroadband,2.45GHz,time-domainclocksourcesW-LANOutAntBluetooth2.45GS-parameterFielddataTime-DomainPCBCPUPage19S-ParameterAnalysis:DDRtoRFS-parameteranalysistechniquewillprovideinsightontheboard’sfrequencyselectivitytoDDRnoisePage20ThefrequencyresponseofDDR2toW-LANprovestheclocknoisecouplesmorethanothersNetworks’(PCBBoard)FrequencySelectivityW-LANBTAnt10+dBPage21FieldDataWith2.45GHzInputFielddatamayprovidesomeinsighttothepotentialcouplingpathsE-fieldplotsondifferentlayersat2.45GHzNoisecoupledtoW-LANPage22Time-DomainSimulationwithDifferentialClockSignalsDirectlyexcitedwithdifferentialclocksignal,Vpp=1VCPUPage23TimeDomainSimulationResultNoisevoltagesaretime-dependantandkeepincreasingduetomultiplepathsPage24Inatypicalhigh-speedsystem….BadlyroutedtracesinpackagesandboardsaretheculpritsofgeneratingEMIandhigh-speedconnectorstendtoamplifytheEMIproblems.Page25ConnectorsHigh-speedtracesMinimizingEMIonPackagesandBoardsgreatlyminimizedSystemEMIproblem.Example#5:ImpactofShieldingforFX-GraphicscardGoals:-ConfirmEmissionfromhigh-speedboard-What-ifanalysisforlow-costsuppressionmaterial-MetalShield-R4NSuppressionmaterial(fromNEC)Page26Far-FieldRadiationofTMDSSignalsPage27Analysis@770MHzConfirmMaxradiation@BottomofbrdduetoTMDSrouting:MaxAngle:144/16026uWattsTotal-PWRAntenna-Gainis-48dB5uWatts/SteradianMetalShieldMountedonBottomoftheCardPage28MetalShieldMetal-ShieldMountingonbottomofPCBconnectedtoheat-sinkscrewsNoMajorChangeofRadiationFieldTotalPowerincreasedslightly27uWattsMaxRadiation@BottomofPCB(157/161)Intensityincreasedslightly:6uWatts/SteradianAntenna-Gain-47dB(worse)R4NMountedonBottomoftheCardPage29SuppressionMaterialR4NMaterialplacedonBottomofthePCBMajorimprovementofRadiationfieldfrombottomofPCB:TotalPWRdroppedto13uWatts(50%)Antenna-Gaindroppedto-51.5dB(~3dB)Intensitydrop