PicoQuant MultiHarp 150 - Timing Resolution

PicoQuantGmbH MultiHarpSoftwareV.3.1.0.0
2.2. Timing Resolution
ThemostcriticalcomponentforthetimingresolutioninTCSPCmeasurementsisusuallythedetector.However,
incontrasttoanalogtransientrecording,thetimeresolutionofTCSPCisnotlimitedbytheimpulseresponseof
thedetector.Onlythetimingaccuracyofregisteringaphotondeterminestheresolution.Thisislimitedbythe
timinguncertaintythatthedetectorintroducesintheconversionfromaphotontoanelectricalpulse.Thistiming
errororuncertaintycanbeasmuchastentimessmallerthanthedetector'spulseresponse.Thetiminguncer-
taintiesareusuallyquantifiedbyspecifyingthermserror(standarddeviation)ortheFullWidthatHalfMaximum
(FWHM)ofthetimingdistributionorinstrumentresponsefunction(IRF).Notethatthesetwonotationsarere-
latedbutnotidentical
1
.MicrochannelplatePMTs,canachievetiminguncertaintiesassmallas25psFWHM.
LowercostPMTsorSPADsmayintroduceuncertaintiesof50to500psFWHM,HPDslieinbetweenwithtypi-
caluncertaintiesof50..150psFWHM.Superconductingnanowiredetectorshavetiminguncertaintiesoftypi-
cally20to100psFWHM,someoptimizeddesignscanevenreachbelow10ps.
ThesecondmostcriticalsourceofIRFbroadeninginfluorescencelifetimemeasurementswithTCSPCisusu-
allytheexcitationsource.Whilemanylaserscanprovidesufficientlyshortpulses,itisalsonecessarytoobtain
anelectricaltimingreferencesignal(sync)forcomparisonwiththefluorescencephotonsignal.Thetypeofsync
signalthatisavailabledependsontheexcitationsource.Withgainswitcheddiodelasers(e.g.,PDL800–D)a
lowjitterelectricalsyncsignalisreadilyavailable.Thesyncsignalusedhereistypicallyanarrownegativepulse
of−800mVinto50Ω(NIMstandard).Thesharpfallingedgeissynchronouswiththelaserpulse(<3psrmsjit-
terforthePDL800–D).Withotherlasers(e.g.,Ti:Sa)aseconddetectormustbeusedtoderiveasyncsignal
fromtheopticalpulsetrain.Thisiscommonlydonewithafastphotodiode(APDorPINdiode).Thelightforthis
referencedetectormustbederivedfromtheexcitationlaserbeame.g.,bymeansofasemi–transparentmirror.
Thereferencedetectormustbechosenandsetupcarefullyasitcontributestotheoveralltimingerror.
AnothersourceoftimingerroristhetimingjitteroftheelectroniccomponentsusedforTCSPC.Thisiscaused
bythefiniterise/fall–timeoftheelectricalsignalsusedforthetimemeasurement.Atthetriggerpointofcom-
parators,logicgatesetc.,theamplitudenoise(thermalnoise,interferenceetc.)alwayspresentinthesesignals
istransformedtoacorrespondingtimingerror(phasenoise).However,thecontributionoftheelectronicstothe
totaltimingerrorisusuallysmall.ForthehighendvariantsoftheMultiHarprandomjitterofatimedifference
measurementislessthan45psrms.
Generally,itisalwaysagoodideatokeepelectricalnoisepick-uplowinallsystemcomponents.Uncorrelated
electricalnoisewillcausejustrandomjitterandIRFbroadeningbutcorrelatednoisecancauseevenmoredis-
turbingartifacts.Thisiswhysignalleadsshouldbeproperlyshieldedcoaxcables,andstrongsourcesofelec-
tromagneticinterferenceshouldbekeptawayfromtheTCSPCdetectorandelectronics.
ThecontributionofthetimespreadintroducedbytheindividualcomponentsofaTCSPCsystemtothetotalIRF
stronglydependsontheirrelativemagnitude.Strictlyspeaking,theoverallIRFistheconvolutionofallcompo-
nentIRFs.AnestimateoftheoverallIRFwidth,assumingindependentnoisesources,canbeobtainedfromthe
geometricsumoftheindividualcomponentsasanrmsfigureaccordingtostatisticalerrorpropagationlaws:
Astheindividualcontributionsaresquared,thetotalwillbedominatedbythelargestcomponent.Itistherefore
oflittlevaluetoimproveacomponentthatisalreadyrelativelygood.Ife.g.,thedetectorhasanIRFwidthof
200psFWHM,shorteningthelaserpulsefrom50psto40pswillhavepracticallynoeffect.Althoughitisdiffi-
culttospecifyagenerallowerlimitonthefluorescencelifetimethatcanbemeasuredbyagivenTCSPCinstru-
ment,asaruleofthumb,onecanassumethatunderfavorableconditionslifetimesdownto1/10oftheIRF
width(FWHM)canstillberecoveredviadeconvolution.
Afinaltime–resolutionrelatedissueworthnotinghereisthebinwidthoftheTCSPChistogram.Asoutlined
above,theanalogelectronicprocessingofthetimingsignals(detector,amplifiers,etc.)createsacontinuous
distributionaroundanytruetimevalue.Inordertoformahistogram,atsomepointthetimingresultsmustbe
quantized.Thisquantizationintroducesfurthererror,ifchosentoocoarse.Thequantizationstepwidth(i.e.the
resolution)shouldthereforebesmallcomparedtotheIRFwidth.Asaminimumsamplingfrequency,fromthe
pointofviewofinformationtheory,onewouldassumetheNyquistfrequency.Thatis,thesignalshouldbesam-
pledatleastattwicethehighestfrequencycontainedinit.Forpracticalpurposesonemaywishtoexceedthis
limitwherepossible,butthereisusuallylittlebenefitinsamplingthehistogramatresolutionshigherthan1/10
oftheoverallIRFwidthoftheanalogpartofthesystem.
1 IncaseofaGaussiandistributionther.m.sdeviationcorrespondstothestandarddeviationσandFWHM≈2,35σ.
Page10