PDL Measurement 5988-1232EN

AbstractThemeasurementofPolarizationDependentLosshasgainedagreatdealofattentionamongcomponentmanufacturers.ThisApplicationBriefdiscussestwodifferentmeasurementtechniques,thePolarizationScanningtechnique,andtheMuellerMethod,andexaminespracticalimplementationdifficulties.PolarizationDependentLossMeasurementofPassiveOpticalComponentsApplicationNote2IntroductionFiberopticnetworksarethepresentandfuturemediumofchoiceforhigh-speed,high-volumedatatransmission.Thegrowthindemandforgreaterdatathroughputrequiresgreaterbandwidthandsmallerchannelspacing.Therapiddevelopmentoffiber-opticnetworktechnologyisdrivenbydramaticadvancesinthedesignandmanufactureofbothactiveandpassiveopticaldevices.Thetremendousneedforhigherdatatransmissionrateshasalwaysdriventhedevelopmentofnewopticalcomponentstothelimitsofexistingtechnology.Thedevelopmentandtestingofnewopticalcomponentshasbecomemorechallengingandcomplex,forexample:•Channelspacingisconstantlybeingreduced,sowavelengthdependentmeasurementsmustbeincreasinglyaccurate.•Thecomplexityofmulti-channeltestsystemsincreasesasthenumberofchannelsincreases.•Theextensionofopticaldatatransmissionintonewspectralregions,suchastheL-band,involvesthedevelopmentofbothopticalcomponents,andtheequipmentrequiredfortestingthem.•TheperformanceofDWDMsystemsisincreasinglyinfluencedbythepolarizationoflightwavesignals.Theincreasinglengthoffiberlinkshasfocusedattentiononnewtestparameters,suchaspolarizationdependentloss,whichisasignaldistortionthataccumulatesoverdistance.•Higherdatatransmissionrates(10Gbit/secor40Gbit/sec)requireshorterpulseduration.Inthefrequencydomain,thisresultsinabroaderspectrum.Hightransmissionqualityrequiresbroaderspectralareasoflowpolarizationdependentloss,toavoidattenuationvariationsfordifferentspectralcomponents.Inaddition,duetotherapidgrowthinthefiber-optictechnologymarket,manufacturersmustrampupproductionvolumesbyincreasingmanufacturingcapacity,andbyshorteningtesttimewhilenotcompromisingtestaccuracy.ThisApplicationNotefocusesontheevaluationoftwopolarizationdependentlossmeasurementtechniquesthataresuitablefordeploymentinthehighvolumemanufactureofpassiveopticalcomponents.Theadvantagesanddisadvantagesofeachtechniqueisdiscussed.Finally,atypicalappliedmeasurementsolutionisdescribedindetail.First,polarizationdependentlossisbrieflydefined,anditseffectsinfiber-optictransmissionlinksdescribed.PolarizationDependentLoss–DefinitionPolarizationdependentlossisameasureofthepeak-to-peakdifferenceintransmissionofanopticalcomponentorsystemwithrespecttoallpossiblestatesofpolarization.Itistheratioofthemaximumandtheminimumtransmissionofanopticaldevicewithrespecttoallpolarizationstates.PolarizationDependentLoss,PDL,isdefinedas:

∗=MinMaxdBPPPDLlog10Equation1:Definitionofpolarizationdependentloss.InFigure1,theeffectofapplyingallpossiblestatesofpolarizationtoanopticalcomponentisshown.Thepolarizationoftheconstant,andfullypolarized,inputsignalisvaried.Asthepolarizationoftheincidentlightvaries,theoutputsignalshowsacorrespondingchangeinpower.Polarizationdependentloss(PDL)MINMAXPPlog10PDL=DeviceundertestConstantpower,100%polarized,scanpolarizationTimePMINPMAXOutputInput•PDLmeasuresthepeak-to-peakdifferenceintransmissionforlightwithvariousstatesofpolarizationFigure1:PolarizationDependentLossofpassiveopticalcomponents.Theoutputpowervariationistheresultofthevariationinthepolarizationoftheincidentlightwavesignal.3CausesofPolarizationDependentLossThepolarizationdependenceofthetransmissionpropertiesofopticalcomponentshasmanysources.Someofthemostcommoneffectsare:•Dichroism•Fiberbending•Angledopticalinterfaces•Obliquereflection.PolarizationDependentLossinopticaltransmissionnetworksAlltheaboveeffectsappearinthestandardopticalcomponentsusedinfiber-opticnetworks.Atypicalstructureofafiber-optictransmissionnetworklinkisshowninFigure2.Thetransmissionlinkincludesanumberofdifferentpassiveandactivecomponents.ThemostcommonpassivedevicesthatexhibitPDLincludeopticalcouplers,isolators,wavelength-divisionmultiplexers(WDM)andphotodetectors.StateoftheartWDMlinkDataoutADMDatainMUXMUXλ1λn-1λnEDFAEDFADEMUXDEMUXλ1λn-1λnSplice/ConnectorTransmitterGainFlatteningReceiverFigure2:TypicalWDMlinkinfiberopticnetworks.Thepolarizationstatethatexhibitsmaximumloss(thatis,minimumtransmission)throughonecomponentisgenerallynotthesameasforothercomponentsinthetransmissionlink.Furthermore,thepolarizationstateisnotmaintainedalongafiber.Theevolutionofpolarizationalongafiberisofacompletelystatisticalnatureand,inconsequence,istotallyunpredictable.EvenifthePDLaxisofeverycomponentisaligned,thisdoesnotcorrespondtotheminimumormaximumeffectonpolarizationsensitivetransmission.SincePDLeffectsbuildupinanuncontrolledmanner,PDLcanleadtoadegradationofthetransmissionqualityofthefiber-opticlink,oreventoafailureoftheopticalsystem.Therefore,modernfiber-opticcommunicationsystemsrequirecomponentswithlowPDL.Consequently,themeasurementofPDLhasattractedenormousattentionfromcomponentmanufacturers.TheneedforPDLtestsolutionsisaccompaniedbytherequirementsofshortmeasurementtime,highaccuracyandhighreliability.Inthefollowing,twoPDLmeasurementtechniquesaredescribedandevaluatedfortheirsuitabilityformodernhigh-volumemanufacturing.MeasurementtechniquesInthecontextofpassivecomponenttestingduringcomponentmanufacture,twotechniquesfordeterminingthePDLofadeviceundertest(DUT)arerecommended:ThePolarizationScanningtechnique,andtheMuellerMethod.WhilethePolarizationScanningtechniqueisfoundsuitableforPDLmeasurementsatspecificwavelengths,formanywavelengthpointsinabroadwavelengthrangetheMuellerMethodshowsclearadvantages.Bothtechniquesdeserveamorein-depthtreatment.ThePolarizationScanningtechniqueThePolarizationScanningtechniqueisthefundamentalmethodformeasuringPDL.TheDUTisexposedtoallstatesofpolarizationandthetransmissionismeasuredwithapowermeter.ThemaximumandminimumtransmissionthroughtheDUTcandirectlybemeasured.ThePDLcanthenbecalculatedusingEquation1.DUT∆P(Sin)S0OpticalPowerMeterOpticalSource11896A01020304050600102030Polarizationscantime(s)MeasurementUncertainty(%)HP11896AP.ControllerRandomizeatrate=5Pwrmtravgtime=20msTypicalSetupforPolarizationScanningRelationbetweenPDLerrorandScanningTimeFigure3:SetupofaPDLmeasurementusingthePolarizationScanningtechnique.ThegraphshowshowthePDLmeasurementuncertaintydependsonthemeasurementtime.4ExposingtheDUTtoallstatesofpolarizationisallbutimpossible.Inpractice,anumberofpolarizationstatesaregeneratedatascanratethatissuitableforthepowermeteraveragingtime.Thelongerapolarizationscantakes,asthetransmissionthroughtheDUTisobtainedformorepolarizationstates,thesmallertheuncertaintyofthePDLmeasurement[1].ThisisdemonstratedbythegraphinFigure3.Atsomepoint,increasingthemeasurementtimedoesnotyieldsignificantlyimprovedmeasurementaccuracy.Here,wherethepolarizationcontroller'srandomizerateis5andthepowermeter'saveragingtimeis20ms,ameasurementuncertaintyof5%requiresapolarizationscantimeof10s.Increasingthemeasurementtimeto20s,(thatis,measuringovertwicethenumberofpolarizationstates)resultsinameasurementuncertaintyof3%,animprovementofonly2%.Consequently,improvingPDLmeasurementuncertaintymustalwaysbeconsideredinthecontextoftheaffectonmeasurementtime.AtypicalPDLmeasurementsetupemployingpolarizationscanningisshowninFigure3.Thesourceproducesnearlyfullypolarizedlight.The11896APolarizationControllertransformsthepolarizationbymeansoffourmotorizedfiberloops.Themovementofthefiberloopscausesavariationinthebirefringenceofthefiber,whichresultsinvariationofthepolarizationstate.Thedifferentrotationalspeedsofthefiberloopsgeneratepolarizationstatesinapseudo-randommanner.The11896APolarizationControllerprovideseightdifferentscanrates,wherethefastestscanisdenotedbyrate8.FiberlooppolarizationtransformationStackedfibercoilsInputE-fieldfastslowfastslowfastslowInputfieldateachcoildecomposesintothefastslowaxisofeachcoilReorientationofacoilchangesdecompositionofinputE-fieldFigure4:Fiberlooppolarizationtransformation,shownonanexamplewiththreefiberloops.Settingthecorrectpolarizationscanratewithrespecttotheaveragingtimeofthepowermeteriscritical.Thepolarizationscanratedictateshowrapidlythepolarizationofthelightwavesignalischanged.Afasterscanrategeneratesmorepolarizationstatesinagiventimeinterval,somightdecreasethedurationofameasurement.However,ifthepolarizationscanrateistoofastwithrespecttotheaveragingtimeofthepowermeter,resultsarefalsified.Atfasterscanrates,thepowermeteraveragesovermorepolarizationstates;amaximumorminimumtransmissioncouldbeaveragedout.ItisclearfromEquation1thatanerrorinmaximumorminimumtransmissionvaluedirectlyaffectsthePDLvalueobtained.Averagingtimeisalsocriticalintermsofnoise.Thesignal-to-noiseratioisproportionaltothesquarerootoftheaveragingtime.Clearly,choosingtheoptimumaveragingtimeisatrade-offbetweenthequalityofthemeasurementintermsofnoiseandthemeasurementtime.HowaveragingtimeaffectsthePDLresultsisdemonstratedinFigure5.Thethreemeasurementexamplesatvariousaveragingtimesshowthatwithasmallaveragingtime,suchas100µs,thequalityofthemeasurementisdegradedbynoise.Ontheotherhand,alongaveragingtimeprovidesnovisibleimprovementofthemeasurementresults.PDLvsAvg.time00.10.20.30.40.50.60.70.80.911.11.21.31.41.515311531.515321532.515331533.515341534.515351535.51536Wavelength[nm]PDL[dB]100us10ms1ms00.050.10.150.20.250.30.350.41533.51533.7515341534.251534.51534.751535100us10ms1msFigure5:PolarizationScanningwiththreedifferentaveragingtimes:100us,1ms,and10ms.TheMuellerMethodAdifferentapproachtothemeasurementofPDListodeterminetheMuellermatrixfortheDUT.ThetechniqueisthereforeknownastheMuellerMethod.5MuellerMethodDUTPn(Sn,λ)S0OpticalPowerMeterOpticalSource8169APolλ/4λ/2Snn=1..4Applicationoffourwell-knownpolarizationstatestotheDUTOpticalpowertransmissionmeasuredatthesepolarizationstatesonlyTransmissiondataareusedtocalculateMuellermatrixcoefficientsPDLisobtainedfromtheMuellermatrixcoefficientsSinglewavelengthPDLmeasurementwithinfewsecondsFigure6:PrincipleofaPDLMeasurementusingtheMuellerMethod.TheMuellerMethoddeterminesPDLbyexposingtheDUTtoonlyfour,butwell-known,statesofpolarization.ThefourpolarizationstatesarechosentobeLHP(linearhorizontalpolarized),LVP(linearverticalpolarized),L+45(Linear+45degrees),RHC(righthandcircular).ThePDLiscalculatedfromthetransmissionresults.ThisapproachwasfirstintroducedinReference[2].AdditionalInformationisprovidedbyReference[3].ThePDLmeasurementprocedurehastwosteps,areferencemeasurementandtheDUTmeasurement,asillustratedbyFigure7.Step1:MeasurethepowerPnattheinputoftheDUT(forcalibration)DUTPn(Sn,λ)S0OpticalPowerMeterOpticalSourceHP8169APolλ/4λ/2Snn=1..4(Sn,λ)S0OpticalPowerMeterOpticalSourceHP8169APolλ/4λ/2Snn=1..4nP~Step2:MeasurethepowerPnattheoutputoftheDUT~Figure7:MeasurementProcedureofMuellermethod:ReferenceandDUTmeasurement.First,theopticalpoweratthefourdefinedpolarizationstatesismeasured.Inthesecondstep,thesamefourpolarizationstatesareappliedtotheDUTandthetransmittedopticalpowerismeasured.TheMuellermatrixdescribesthepolarizationandpowertransmissionpropertiesoftheDUT.TherelationshipbetweenaninputStokesvectorandoutputStokesvectorofaDUTcanbewrittenas:inDUToutSMS∗=whereMDUTistheMuellermatrixofthedevice.TheMuellermatrixisa4x4matrix.Thefourfirst-rowcoefficientsoftheMuellermatrixdescribethepowertransmissionofadevice,whichissufficienttoobtainthePDL.Asstatedpreviously,thereferencemeasurementdeterminesthepoweroftheinputStokesvector.TheDUTmeasurementyieldsthetotalpowertransmittedthroughtheDUT.Whenmeasuredforthefourpolarizationstates,asystemoflinearequationscanbesolvedtodeterminethedesiredcoefficientsoftheMuellermatrix,asshowninFigure8.Fromthesecoefficients,themaximumandminimumtransmissioncanbederived,asshowninFigure9,fromwhichthePDLcanbecalculated,asshowninEquation1.




=00~~111,PPSin




−=00~~221,PPSin




=0~0~331,PPSin




=441,~00~PPSinLHP:LVP:L+45:RHC:21211122121111~~~~PmPmPPmPmP−=+=41441143133113~~~~PmPmPPmPmP+=+=InputStokesVectors(fromDUTinputpowermeasurement-Step1)OutputStokesVectors(fromDUTouputpowermeasurement-Step2)Figure8:Withthepowermeasurementresults,asystemoflinearequationscanbesolved.









−− − +=



114411332211221114131211~~~~21~~21mPPmPPPPPPPPPPmmmm21421321211maxmmmmT+++=21421321211minmmmmT++−=Solvingtheequationsystemform11..m14yields:minimumandmaximumtransmissionthroughDUT:Figure9:Calculationofminimumandmaximumtransmission.Thefourpolarizationstatesaresynthesizedbyan8169APolarizationController.Thepolarizationcontrollerconsistsofapolarizer,andtworetarderplates(onequarter-wave,andonehalf-wave).Allelementsarerotatable;theaxisofrotationbeingparalleltothedirectionoflightpropagation.6Thepolarizergeneratesalinearpolarizationstate,whichtheretarderplatestransformintoanyotherpolarizationstate.Desiredpolarizationstatesareobtainedbysettingtheretarderplatestospecificangles.Thepolarizationcontrollerexhibitsapolarizationdependentloss.ThePDLofthepolarizationcontrollerisspecifiedtowithin±0.03dB.Thereferencemeasurementrecordstheabsolutepowerateachofthefourpolarizationstates.Ifthesystemisunchanged,eachpolarizationstatehasthesameoutputpowerduringtheDUTmeasurementasduringthereferencemeasurement.Hence,powervariationacrosspolarizationistakencareofbythereferencemeasurement.AdvancedPDLmeasurementRequirementsformaximumaccuracyRegardlessofwhichmeasurementtechniqueisused,toachievethehighestpossibleaccuracy,itssetupmustmeetcertainrequirements:Zoom:InsertionLossofaTransmissionWindow,AWG-typefilter(averageof4pol.States,10measurements)0.000020.0000210.0000220.0000230.0000240.0000250.0000260.0000271.534E-061.5341E-061.5341E-061.5341E-061.5341E-061.5341E-061.5341E-061.5341E-061.5341E-061.5341E-061.5341E-061.5342E-061.5342E-061.5342E-061.5342E-061.5342E-061.5342E-06Wavelength[m]power[W]SeriesSeriesSeriesSeriesSeriesSeriesSeriesSeriesSeriesSeriesPowerStabilityWavelengthAccuracyFigure10:Evaluatingthemeasurementsetupincludeswavelengthrepeatabilityandpowerstability.•Thetunablelasersourcemusthaveastablepoweroutput.AnyvariationintheoutputpowerofthelasersourceisnotrecognizedinaPDLmeasurement,andmaybemisconstruedaspolarizationsensitivityoftheDUT.•Wavelengthaccuracyandwavelengthrepeatabilityplayimportantrolesinthequalityofameasurement.Wavelengthaccuracydeterminestheabsolutelocationofthefiltercurvealongthewavelengthaxis.WavelengthrepeatabilityisespeciallyimportantfortheMuellerMethod,wherethefiltercurveismeasuredfourtimesatdifferentinputpolarizationstates.AnydeviationinwavelengthbetweenthefourmeasurementresultscanseverelyaffectthefinalPDLresult.Powerstabilityandwavelengthrepeatabilitycaneasilybequalifiedbyrepeatedmeasurementofafiltertransmissioncurveatafixedinputpolarization,asdemonstratedinFigure10.Powerstabilityisbestevaluatedatthepeakofthefiltertransmissioncurve.Theoverlapattheslopeofmultiplefiltercurvesisavaluablemeasureofwavelengthrepeatability.AsamplemeasurementseriesisshowninFigure11.WavelengthAccuracy:SlopeofaTransmissionWindow,AWG-typefilter(averageof4pol.States,10measurements)0.000020.00002020.00002040.00002060.00002080.0000210.00002120.00002140.00002160.00002181.534180E-061.534181E-061.534181E-061.534182E-061.534182E-061.534183E-061.534183E-061.534184E-061.534184E-061.534185E-061.534185E-061.534186E-061.534186E-061.534187E-061.534187E-061.534188E-061.534188E-061.534189E-061.534189E-061.534190E-061.534190E-06Wavelength[m]power[W]0.5pm~0.5pmFigure11:Demonstrationofthewavelengthrepeatabilityof10wavelengthsweeps.Thedetectorinthetestsetupalsoplaysanimportantrole.Asstatedintheintroduction,photodetectorsareamongthecomponentsthatexhibitpolarizationdependence.Thus,itisessentialtousedetectorswithlowPDL.Asmentionedearlier,thePDLofdifferentcomponentscombinesinanuncontrolledmanner,sothePDLofthedetectorcansignificantlyaffectthePDLmeasurement.Moreover,spectralrippleofthepowerdetectorscandegradethemeasurementquality.ThelatestAgilentpowermetermodulesprovidetheflexibilityneededtomeettherequirementsofdifferenttestenvironments.Thechoiceofpowersensormoduleisdrivenbythemeasurementpriorities.Forthehighestaccuracy,thesingle-channelopticalpowersensors(81633A,81634A),oropticalpowerheads(8162xA),arepreferred.Thedual-channelpowersensors(81635A)provideaneconomicalsolutionwithslightlylowerperformance.TheintrinsicPDLoftheAgilent81635Adual-channelpowersensormoduleisspecifiedastyp.±0.015dB.However,thetwochannelsofeachpowersensorsavespaceinthetestenvironment.Wheretherequirementisforthehighestpossibleaccuracy,opticalheads,withtheirlowintrinsicPDL(typ.±0.002dB),providethebestsolution.Whenusedwithadual-channelinterfacemodule,opticalheadsmeetdemandsforthehighestaccuracywhileprovidinganeconomicsolutionintermsofthemainframe'smodulecapacity.7Usingsingle-channelpowersensors,withtheirlowPDL,canmeetarequirementforlowuncertainty.However,foragivennumberofchannels,twiceasmanymainframeslotsarerequiredthanforadual-channelsolution.Extramainframesmayberequiredtohostallthepowersensormodules.Notonlythedetector,butalsoeveryotherpassivecomponentinthesetupcaninfluencethePDLmeasurement.Therefore,toreducethemeasurementuncertainty,itisessentialtominimizethenumberofopticalinterfacesandcomponents.Openangledconnectors,forexample,haveanintrinsicPDLdependentontheanglebetweenthefront-endsurfaceandtheplanenormaltothedirectionoflightpropagation.Anopen8°angledconnectorexhibitsaPDLof0.019dB.UsedasthefinalconnectorbetweentheDUTandpowermetermodule,theintrinsicPDLoftheconnectorinfluencesthemeasurementresult.Thiseffectcannotbecalibratedout,soitisessentialtouseastraightconnectortothepowermetermodule.PDLoverWavelengthMostoften,thePDLofaDUTatdifferentwavelengthsmustbemeasured.Generally,thePolarizationScanningtechniquecanbeshowntobebestsuitedforPDLmeasurementatsinglewavelengths,andtheMuellerMethodforPDLmeasurementoverawavelengthrange,asshowninFigure12.PDLMeasurementatthreepointsinchannelwithpolarizationscanning.PDLmeasuredovertransmissionandattenuationbandsofoneormorechannelswithMuellerMethodandwavelengthsweepFigure12:MeasuringPDLoverwavelengthwithpolarizationscanningandMuellermethod.ThePolarizationScanningtechniqueexposestheDUTtomanystatesofpolarization,sothePDLcanbemeasuredonlyatonewavelengthatatime.ItisclearthatcapturingthePDLofaDUTatmanywavelengthscanquicklybecomeverytime-consuming.However,ifthePDLisonlyrequiredatcertainpoints,suchasthecenterwavelengthorthe3dBbandwidthwavelengthsofapassband,thePolarizationScanningtechniqueissufficientlyfast.ComparedtotheMuellerMethod,thePolarizationScanningtechniqueisrelativelyeasy-to-implementanddoesnotinvolveextensivemathematicalcalculations,exceptingEquation1.ThePolarizationScanningtechniqueisthepreferredsolutionforthiscase.The11896APolarizationControllerisspecifiedforoperationinabroadwavelengthrange(1250nm–1600nm).Thefiber-baseddesignofthepolarizationcontrollermeansthatwavelengtheffectscanbepresumedtobenegligible.Thelasersourceemployeddependsonthewavelengthaccuracyrequiredandtherangeofwavelengthsofinterest.Forexample,theAgilent81689Acompacttunablelasersourcecoversa50nmwiderange(1520nm–1570nm).Thelackofcontinuoussweepcapabilitydoesnotplayarole,becausethePolarizationScanningtechniqueonlyallowsthewavelengthrangetobecoveredinsteps.Furthermore,thetransmissionpropertiesoftheDUTaremeasuredonlyatspecificwavelengths,whichneednotbeequallyspaced.Incontrast,theMuellerMethod,inconjunctionwithacontinuouswavelengthscan,shouldbeusedwhereanentirechannel,orevenanumberofchannels,mustbecharacterizedforPDL.Inotherwords,wheretherearealargenumberofwavelengthpointswithfixedspacing.TheAgilent81680Atunablelasersource,designedforpassivecomponenttestintheC-band,iscapableofcontinuouswavelengthscan,whichdecreasesmeasurementtimewhenmanywavelengthpointsmustbemeasured.Inaddition,thislasersourcehashighdynamicrange,lowpowerfluctuationsovertime,aswellasoutstandingwavelengthaccuracyandrepeatability.Despitetheadvantageinmeasurementtime,usingtheMuellerMethodwithacontinuouswavelengthscanistricky.Asmentionedpreviously,the8169APolarizationControllersynthesizespolarizationstates