5.2CoherentInputFrequencySelection
www.ti.com
ADCEvaluation
TypicalADCanalysisrequiresuserstocollecttheresultingtime-domaindataandperformaFourier
transformtoanalyzethedatainthefrequencydomain.AstipulationoftheFouriertransformisthatthe
signalmustbecontinuous-time;however,thisisimpracticalwhenlookingatafinitesetofADCsamples,
usuallycollectedfromalogicanalyzer.Consequently,userstypicallyapplyawindowfunctiontominimize
thetime-domaindiscontinuitiesthatarisewhenanalyzingafinitesetofsamples.ForADCanalysis,
windowfunctionshavetheirownfrequencysignaturesorlobesthatdistortbothSNRandSFDR
measurementsoftheADC.
TIusestheconceptofcoherentsamplingtoworkaroundtheuseofawindowfunction.Thecentral
premiseofcoherentsamplingentailsthattheinputsignalintotheADCiscarefullychosensuchthatwhen
acontinuous-timesignalisreconstructedfromafinitesampleset,notime-domaindiscontinuitiesexist.To
achievethis,theinputfrequencymustbeanintegermultipleoftheratiooftheADCsamplerate(f
s
)and
thenumberofsamplescollectedfromthelogicanalyzer(N
s
).Theratiooff
s
toN
s
istypicallyreferredtoas
thefundamentalfrequency(f
f
).DeterminingtheADCinputfrequencyisatwo-stepprocess.First,the
usersselectthefrequencyofinterestforevaluatingtheADC;then,theydividethisbythefundamental
frequency.Thistypicallyyieldsanon-integervalue,whichshouldberoundedtothenearestodd,
preferablyprime,integer.Oncethatinteger,orfrequencybin(f
bin
),hasbeendetermined,usersmultiply
thiswiththefundamentalfrequencytoobtainacoherentfrequencytoprogramintotheirADCinputsignal
generator.Theprocedureissummarizedasfollows.
f
f
=f
s
/N
s
f
bin
=Odd_round(f
desired
/f
f
)
Coherentfrequency=f
f
×f
bin
SLAU206B–September2007–RevisedApril200815
SubmitDocumentationFeedback