Accurately Measuring ADC Driving Circuit Settling …

1 ApplicationReportSLYT262A January2007 RevisedMay 2015 AccuratelyMeasuringADCD rivingCircuitSettlingTimeRajivMantri,Bha skarGoswamiABSTRACTM anymoderndataacquisitionsystemsconsistof highspeed,high-resolutionADCs.(1)CMOS-sw itched,capacitorbasedADCsare oftenchosenfor suchdesignsdue to theirlow cost and low an unbufferedfrontend directlycoupledto the effectivelyminimizenoiseand signaldistortion,it is necessaryto drivethe ADCwith a high-speed,lownoise,low-distortionoperat ionalamplifier.(2)To achieveminimaldistortionit is importantfor the op ampoutputto settleto thedesiredaccuracywithinthe acquisitiontimeof the op ampsettlingtimeis eithercalculatedfromthe frequencyresponsespecifiedin the datasheetor measuredby probingthe outputwithan oscilloscopethat has a limitationon differencebetweenthe op ampinputandoutputis amplifiedto limitedby the oscilloscoperesolutionor ,the settlingtimeof the op ampis affectedby the parasiticcapacitanceand inductanceintroducedby the anothermethod.

2 The differencebetweenoutputand inputis amplifiedto increasethe resolutionof the thesemethodsincludesthe parasiticcapacitanceand inductancepresentin the of SamplesfromB to A of Tables1 Comparisonof January2007 RevisedMay 2015 AccuratelyMeasuringADCD rivingCircuitSettlingTimeSubmitDocumenta tionFeedbackCopyright 2007 2015,TexasInstrumentsIncorporatedPattern GeneratorConversionStartAdjustable-Delay GeneratorSTHS4031Op AmpTS5A3159 Analog MUXC hannel 1 GroundedChannel 2 Voltage, VADS841116-Bit,2-MSPSADCMUX switches tochannel 2 heretsMUX Channel 1 Channel 2 ADefinitionof SettlingTimeSettlingtimeis the time elapsedfromthe applicationof an idealinstantaneousstep inputto the time atwhichthe closed-loopamplifieroutputhas enteredand remainedwithina includesa very briefpropagationdelay,plus the time requiredfor the outputto slewtothe vicinityof the final value,recoverfromthe overloadconditionassociatedwith slewing,and finallysettleto withinthe high-resolutionADCs,the specifiederrorbandis usuallyone fourthof oneleastsignificantbit (LSB)of the ADCusedhereis the TexasInstruments(TI) ADS8411,whichis a 16-bit,2-MSPS successiveapproximationregister(SAR)

3 Driverop ampis the TI showsthe instantaneousstep inputis generatedwith an analogmultiplexer(MUX)(the TI TS5A3159)byswitchingits two dc voltage,V, is appliedto channel2, and channel1 is connectedto ground;so this setupcan producea step inputrisingto V from0 or fallingto 0 fromV. Alternatively,the step inputcan be generatedfromany step step generatorshouldsettlemuchfasterthanthe op (connectedto ground)first. A long samplingtime is providedto makesurethatthe inputcapacitorof the ADCis fully analogMUXis switchedto channel2 fromchannel1 at instantA in Figure2. This diagramshowsthevoltageat pointS (Figure1) whenthe MUXswitchesoverto channel2 fromchannel1. The settlingtimeof the MUXis denotedby ts.

4 It is assumedthat ts is muchshorterthanthe op SettlingTimeEvaluationSetupFigure2. SettlingTimefor MUXC hannelChange2 AccuratelyMeasuringADCD rivingCircuitSettlingTimeSLYT262A January2007 RevisedMay 2015 SubmitDocumentationFeedbackCopyright 2007 2015,TexasInstrumentsIncorporatedSamplin gstartshereAMUX OutputSampling edge isshifted toward leftBOp Amp analogMUXis switchedat instantA, the inputof the op outputof theop ampstartschangingaftera very briefpropagationdelayafterinstantA. The op ampsettlingtime(tideal)is approximatelycalculatedfromthe slewrate and bandwidthspecifiedin the op methodproposedhereplotsthe op ampoutputfrominstantA to instantB (Figure3). The timedifferencebetweeninstantB and instantA is Averagingn SamplesfromB to A IncreasesAccuracyStep4 The first ADCsamplingedgeappearsat instantB, and n numberof readings(digitaloutputsfromtheADC)are n numberof readingsare averagedfor betteraccuracy(discussedlater).

5 Nextthesamplingedgeis shiftedto the left by 1 ns (Figure3) with the help of a patterngeneratorand anadjustable-delaygenerator(Figure1), and againn numberof readingsare way the samplingedgeis shiftedtowardthe left frominstantB to instantA in eachsamplingedgetheaverageis storedin an elementof an arrayis plottedagainstthe time to get the true pictureofthe op ampoutputsettling(Figure3). AchieveBetterResolutionInputof an n-bitADCshouldsettleto at leastn+2 bits, but measuredoutputis an n-bitdigitalcodefromthe resolutioncan be increasedby repeatedlysamplingthe sameinputand takingmultiple(n)readingsfromthe averageis takenon the n can be shownthat foreachadditionalbit of resolution,the numberof readingsshouldbe 4, so w extrabits of eachadditionalbit, the signal-to-noiseratio(SNR)increasesby In this casethe 16-bitADCshouldsettlewith at N + ,whereN is the for 18-bitaccuracy,so an extrabit (w) of 86* / 4* TypicalSNRspecificationof ADS8411 The numberof samples(n)

6 Neededfor eachreadingis 44 = January2007 RevisedMay 2015 AccuratelyMeasuringADCD rivingCircuitSettlingTimeSubmitDocumenta tionFeedbackCopyright 2007 2015,TexasInstrumentsIncorporated5952359 5255952759529595315953359535595375953959 54159543050100150 Time (ns)Output Code200250300680 pF1000 pFNo Capacitor3000035000400004500050000550006 000065000 Output Code050100150200250300 Time (ns)680 pF1000 pFNo CapacitorCRC'R' RC filter is usedat the outputof the op ampto filter the ADCsamplingcircuitalwaysconsistsof anotherRC (R , C ), as shownin showsthe settlingbehaviorwhenthreedifferentvalues of an externalcapacitorare usedfor TypicalNoiseFilterFigure5. InputSettlingTimeWithan ExternalCapacitorFigure6.

7 ExpandedScaleMagnifiesSettling-TimeBehav iorShownin Figure5 Figure6 is a zoomed-inversionof Figure5 to showthe outputcodeisbasedon 16-bitsampling,the resolutionof the measurementis morethan16-bitbecause65536sampleswerecap turedand averagedfor resultshowssignificantringingand underdampingofthe systemwhenno capacitorwas use of a bigger(1000-pF)capacitorsignificantlyinc reasesthe January2007 RevisedMay 2015 SubmitDocumentationFeedbackCopyright 2007 2015, 113 301 0 Time (ns)Output BiasCurrentA summaryof theseresultsis shownin Table1. Averagingthe outputdatacan improvethe resolutionofthe Comparisonof Edge-ShiftMethodVersusTraditionalMethodC apacitor,Accuracy*SettlingTimeMethodC (pF)(%)(ns) (R = 20 ) Effectsof ChangingFeedbackResistors3 Measurementof BiasCurrentFigure7 showsop ampsettlingbehaviorwith differentvaluesof differencebetweenthe settledvoltagesindicatesthe offsetvoltageshift causedby bias the biascurrentcan be calculatedas 3 A, whichmatchesthe typicalspecificationof the correctnessof this January2007 RevisedMay 2015 AccuratelyMeasuringADCD rivingCircuitSettlingTimeSubmitDocumenta tionFeedbackCopyright 2007 2015,TexasInstrumentsIncorporated==m938 Bias A301 ==m(59610 59595) (Offset Voltage)

8 938 settledvaluewith 0 in the feedbackis settledvaluewith 301 in the (offsetvoltage)= bias current resistor(usedin the feedback).(comparedto the datasheettypicalspecificationof 3 A).5 ConclusionThis is a practicaland simplemethodto accuratelymeasurethe settlingtime of an settlingbehavioris unaffectedby the measurement,becauseno additionalcomponentis usedfor methodcan be implementedas a built-inself-test(BIST)in the averagingofmultiplereadingsimprovesthe accuracyof the moreinformationrelatedto this article,you can downloadan AcrobatReaderfile at replace litnumber with the TI Lit. # for the KevinM. Daugherty,Analogto DigitalConversion:A PracticalApproach(McGraw-HillCompanies,1 993).

9 2. Ron Mancini,ed., Op Ampsfor Everyone, DesignReference(RevB) (SLOD006) ADS8411, THS4031, or (MonthYear)to A Changeddocumentformatto :Pagenumbersfor previousrevisionsmay differfrompagenumbersin the January2007 RevisedMay 2015 SubmitDocumentationFeedbackCopyright 2007 2015,TexasInstrumentsIncorporated7 SLYT262A January2007 RevisedMay 2015 SubmitDocumentationFeedbackCopyright 2007 2015,TexasInstrumentsIncorporatedIMPORTA NTNOTICET exasInstrumentsIncorporatedand its subsidiaries(TI) reservethe rightto makecorrections,enhancements,improvement sand otherchangesto its semiconductorproductsand servicesper JESD46,latestissue,and to discontinueany productor serviceper JESD48, latestrelevantinformationbeforeplacingor dersand shouldverifythat suchinformationis semiconductorproducts(alsoreferredto hereinas components )

10 Are sold subjectto TI s termsand conditionsof salesuppliedat the time of warrantsperformanceof its componentsto the specificationsapplicableat the time of sale,in accordancewith the warrantyin TI s termsand conditionsof sale of otherqualitycontroltechniquesare usedto the extentTI deemsnecessaryto supportthis applicablelaw, testingof all parametersof eachcomponentis not assumesno liabilityfor applicationsassistanceor the designof Buyers responsiblefor theirproductsandapplicationsusingTI minimizethe risksassociatedwith Buyers productsand applications,Buyersshouldprovideadequate designand doesnot warrantor representthat any license,eitherexpressor implied,is grantedunderany patentright,copyright,maskworkright,orot herintellectualpropertyrightrelatingto any combination,machine,or processin whichTI componentsor servicesare TI regardingthird-partyproductsor servicesdoesnot constitutea licenseto use suchproductsor servicesor a of suchinformationmay requirea licensefroma thirdpartyunderthe patentsor otherintellectualpropertyof thethirdparty,or a licensefromTI underthe patentsor otherintellectualpropertyof significantportionsof TI informationin TI databooksor datasheetsis permissibleonly if reproductionis withoutalterationand is accompaniedby all associatedwarranties,conditions,limitati ons.