Jitter-Analysis

JitterAnalysis7/27/2005JamesJaw1.JitterTypes/Theory–Overview-Cycle-CycleJitter,PeriodJitter,PhaseJitter,PhaseNoise,TIE-JitterSpecifications–FibreChannel,SONET,IEE802.3ae-Measurementdo’sanddon’ts-Thebestequipmentformeasuringdifferentjittertypes2.MeasurementMethods-Cursor-basedJitterandTimingAnalysis-Spectrumanalyser-Softwareaid-BERTAgendaWhatisJitter•1.Thedeviationofasignaltransitionfromitsidealpositionintime,or…•2.Thetimingvariationfromtransitiontotransition,or…•3.Astheshort-termvariationsofadigitalsignal’ssignificantinstantsfromtheiridealpositionsintime.PeriodJitter,Cycle-cycleJitter,andTIEPeriodJitterCycleToCycleJitterDigitalJitterMeasurementPeriodJitterJitterStatistics•TIEofthesignalinFigure3.2bismodulatedbythetrianglewavedepictedinFigure3.2c•Figure3.2disthetimetrend(jittervs.time)JitterSpectrum•Periodiccomponentsthatotherwisemightbehiddenbywidebandnoisecanoftenbeclearlydistinguished.•AspectralviewofjittercanrevealwhetherthenoiseinasystemisofconcernornotEyeDiagram•Givesinsightintotheamplitudebehaviorofthewaveformaswellasthetimingbehavior.•Simplytriggeringonawaveformwillresultinaneyediagramdifferentfromonerelativetotheunderlyingclock.•Needclockrecovery(insoftwareorhardware).•Oratriggerderivedfromanexternalclock-recoverycircuit.•Jitterisusuallydividedintorandomanddeterministicjitter.•DeterministicJitterisboundedandhasafinitepk-pkvalueregardlessofsamplesize.•Randomjitterisunbounded–thelongeryoulook,thewiderperiodvariationyouwillsee•Thisiswhypk-pkjitterisameaninglessspecificationunlessyouspecifyabiterrorrateassociatedwiththatpk-pkjitter.Random/DeterministicJitterJitterSeparation•AnanalysistechniquethatusesaparameterizedmodeltodescribeandpredictsystembehaviorCausesofJitter•PeriodicJitter–Deterministic–usuallyaverysmallcomponentofjitteronawell-designeddifferentialclock–Reflections–GroundBounce–thresh-holdshift–Mixingofdifferencefrequenciesonchipandoffchip(crosstalk)–Ringing–ReferenceClockfeedthroughonPLL’s•RandomJitter1–FlickerNoise–(1/fnoise).Onlyoccurswithdccurrentflowing.Arisesduetotrapsinthesemiconductor(duetocrystalimperfectionsandcontamination)wherecarrierswhichwouldnormallycontributetodccurrentflowareheldforsomeperiodoftime,thenreleased–ShotNoise–arisesbecausedccurrentflowisnottrulycontinuous,butratherconsistsofpulsesduetotheflowofindividualcarriers.–ThermalNoise–duetothermalexcitationofchargecarriersinaconductor/resistor.Thisnoiseisnotdependentondcbiascurrent–BurstNoise(“popcorn”noise)–notfullyunderstood,butrelatedtoheavymetalioncontamination.Golddopeddevicesshowveryhighlevelsofburstnoise.1Sourcesused:AnalysisandDesignofAnalogIntegratedCircuits,3rdEditionbyGrayandMeyer.AnalogIntegratedCircuitDesignbyJohnsandMartin.WavecrestSIA3000TIAProvidesRandomandDeterministicJitterSeparationPeriodicJitter•Jitterthatrepeatsinacyclicfashioniscalledperiodicjitter.•Typicallycausedbyexternaldeterministicnoisesourcescouplingintosystem,suchasswitching-power-supplynoiseorastronglocalRFcarrier.Itmayalsobecausedbyanunstableclock-recoveryPLLData-DependentJitter•AyjitteriscorrelatedwiththebitsequenceinadatastreamistermedData-DependentJitter.•AnothercommonnameisInter-SymbolInterference,orISIDuty-CycleDependentJitter•JitterthatmaybepredictedbasedonwhethertheassociatededgeisrisingorfallingiscalledDuty-CycleJitter.–-Theslewratefortherisingedgesdiffersfromthatofthefallingedges.–-Thedecisionthresholdforawaveformishigherorlowerthanitshouldbe.PuttingItAllTogether•SystemI–Arandomjitterwithastandarddeviationof0.053UI.•SystemII–Arandomjitterwithastandarddeviationof0.028UI.–Twouncorrelatedcomponentsofperiodicjitterwithp-pamplitudeof0.14UIBothTIEarethesame:0.457UITheBathtubCurve•Onecantellwhathorizontalportionoftheeyewillremaincompletelyfreeofsignaltransitions,foragivenconfidencelevel.•Itcantakealongtimetoaccumulateenoughdatatodirectlymeasuretheeyeopeningsnearthebottomofthechart.Sothemathematicalmodelcanbeusedtopredictperformanceonthebasisofamuchsmallersampleset.TimeDomainMeasurementsDigitalJitterMeasurementPeriodJitterPeriodJitterMeasurement•JEDECsuggeststhatyoumustcaptureatleast10,000cyclesforavalidmeasurement•Real-timescope+jitteranalysissoftware–Notasgoodbecausethenumberofperiodscapturedislimitedbyscopememorydepth•Samplingscope–Tektronix11801C,CSA8000•TimeIntervalAnalyzer–Wavecrest2070,2075,2077,2079,CSA80001PeriodJitterMeasurementusingaTEKCSA8000Thisshowshowthefirstedgefromthetriggeredgevariesintime.AhistogramIsbuilttoshowhowthisedgevariesvs.thenumberofsamples(hits).Pk-PkJitter:25ps#hits(samples):20,257StdDeviation:2.997psRMSPeriodJitter:4.5ps(random+deterministicrmsnumber)Peak-PeakJitter:51.04ps(11.3StandardDeviations)Notehowthepk-pkjitterincreasedbecauseofthemuchlongersampletime8432-01with113millionhits(48hrs)pk-pkjitter=51ps8432-01with20,000millionhitsPk-Pkjitter=25psCycle-to-CycleJitterMeasurement•Real-TimeScope+JitterSoftware–M1SoftwarefromAmherstSystems–Jit1,Jit2,Jit3softwarefromTektronix•Lookatcycle-cycledisplacementsoveranacquisitionofseveralthousandcycles.Thelargestpositiveandnegativecycle-cycledisplacementsovertheentireacqui