数字滤波器外文翻译.docx
- 文档编号:9576677
- 上传时间:2023-02-05
- 格式:DOCX
- 页数:16
- 大小:37.85KB
数字滤波器外文翻译.docx
《数字滤波器外文翻译.docx》由会员分享,可在线阅读,更多相关《数字滤波器外文翻译.docx(16页珍藏版)》请在冰豆网上搜索。
数字滤波器外文翻译
中文5590字
毕业设计
(外文翻译材料)
2009年6月
DIGITALFILTERS
DigitalfilteringisoneofthemostpowerfultoolsofDSP.Apartfromtheobviousadvantagesofvirtuallyeliminatingerrorsinthefilterassociatedwithpassivecomponentfluctuationsovertimeandtemperature,opampdrift(activefilters),etc.,digitalfiltersarecapableofperformancespecificationsthatwould,atbest,beextremelydifficult,ifnotimpossible,toachievewithananalogimplementation.Inaddition,thecharacteristicsofadigitalfiltercanbeeasilychangedundersoftwarecontrol.Therefore,theyarewidelyusedinadaptivefilteringapplicationsincommunicationssuchasechocancellationinmodems,noisecancellation,andspeechrecognition.
Theactualprocedurefordesigningdigitalfiltershasthesamefundamentalelementsasthatforanalogfilters.First,thedesiredfilterresponsesarecharacterized,andthefilterparametersarethencalculated.Characteristicssuchasamplitudeandphaseresponsearederivedinthesameway.Thekeydifferencebetweenanaloganddigitalfiltersisthatinsteadofcalculatingresistor,capacitor,andinductorvaluesforananalogfilter,coefficientvaluesarecalculatedforadigitalfilter.Soforthedigitalfilter,numbersreplacethephysicalresistorandcapacitorcomponentsoftheanalogfilter.ThesenumbersresideinamemoryasfiltercoefficientsandareusedwiththesampleddatavaluesfromtheADCtoperformthefiltercalculations.
Thereal-timedigitalfilter,becauseitisadiscretetimefunction,workswithdigitizeddataasopposedtoacontinuouswaveform,andanewdatapointisacquiredeachsamplingperiod.Becauseofthisdiscretenature,datasamplesarereferencedasnumberssuchassample1,sample2,sample3,etc.Figure1showsalowfrequencysignalcontaininghigherfrequencynoisewhichmustbefilteredout.ThiswaveformmustbedigitizedwithanADCtoproducesamplesx(n).Thesedatavaluesarefedtothedigitalfilter,whichinthiscaseisalowpassfilter.Theoutputdatasamples,y(n),areusedtoreconstructananalogwaveformusingalowglitchDAC.
Digitalfilters,however,arenottheanswertoallsignalprocessingfilteringrequirements.Inordertomaintainreal-timeoperation,theDSPprocessormustbeabletoexecuteallthestepsinthefilterroutinewithinonesamplingclockperiod1/fs.Afastgeneralpurposefixed-pointDSPsuchastheADSP-2189Mat75MIPScan。
executeacompletefiltertapmultiply-accumulateinstructionin13.3ns.TheADSP-2189MrequiresN+5instructionsforanN-tapfilter.Fora100-tapfilter,thetotalexecutiontimeisapproximately1.4μs.Thiscorrespondstoamaximumpossiblesamplingfrequencyof714kHz,therebylimitingtheuppersignalbandwidthtoafewhundredkHz.
However,itispossibletoreplaceageneralpurposeDSPchipanddesignspecialhardwaredigitalfilterswhichwilloperateatvideo-speedsamplingrates.Inothercases,thespeedlimitationscanbeovercomebyfirststoringthehighspeedADCdatainabuffermemory.ThebuffermemoryisthenreadataratewhichiscompatiblewiththespeedoftheDSP-baseddigitalfilter.Inthismanner,pseudo-realtimeoperationcanbemaintainedasinaradarsystem,wheresignalprocessingistypicallydoneonburstsofdatacollectedaftereachtransmittedpulse.
Anotheroptionistouseathird-partydedicatedDSPfilterengineliketheSystolixPulseDSPfiltercore.TheAD772516-bitsigma-deltaADChasanon-chipPulseDSPfilterwhichcando125millionmultiply-accumulatespersecond.
Eveninhighlyoversampledsampleddatasystems,ananalogantialiasingfilterisstillrequiredaADCandareconstructionfilteraftertheDAC.Finally,assignalfrequenciesincreasesufficiently,theysurpassthecapabilitiesofavailableADCs,anddigitalfilteringthenbecomesimpossible.Activeanalogfilteringisnotpossibleatextremelyhighfrequenciesbecauseofopampbandwidthanddistortionlimitations,andfilteringrequirementsmustthenbemetusingpurelypassivecomponents.Theprimaryfocusofthefollowingdiscussionswillbeonfilterswhichcanruninreal-timeunderDSPprogramcontrol.
Asanexample,considerthecomparisonbetweenananalogandadigitalfiltershowninFigure2.Thecutofffrequencyofthebothfiltersis1kHz.Theanalogfilterisrealizedasa6-poleChebyshevType1filter(rippleinpassband,norippleinstopband).Inpractice,thisfilterwouldprobablyberealizedusingthree2-polestages,eachofwhichrequiresanopamp,andseveralresistorsandcapacitors.The6-poledesigniscertainlynottrivial,andmaintainingthe0.5dBripplespecificationrequiresaccuratecomponentselectionandmatching.
Ontheotherhand,thedigitalFIRfiltershownhasonly0.002dBpassbandripple,linearphase,andamuchsharperrolloff.Infact,itcouldnotberealizedusinganalogtechniques!
Inapracticalapplication,therearemanyotherfactorstoconsiderwhenevaluatinganalogversusdigitalfilters.Mostmodernsignalprocessingsystemsuseacombinationofanaloganddigitaltechniquesinordertoaccomplishthedesiredfunctionandtakeadvantageofthebestofboththeanalogandthedigitalworld.
Therearemanyapplicationswheredigitalfiltersmustoperateinreal-time.ThisplacesspecificrequirementsontheDSPdependinguponthesamplingfrequencyandthefiltercomplexity.ThekeypointisthattheDSPmustfinishallcomputationsduringthesamplingperiodsoitwillbereadytoprocessthenextdatasample.Assumethattheanalogsignalbandwidthtobeprocessedisfa.ThisarequirestheADCsamplingfrequencyfstobeatleast2fa.Thesamplingperiodis1/fs.AllDSPfiltercomputationsmustbecompletedduringthisinterval.ThecomputationtimedependsonthenumberoftapsinthefilterandthespeedandefficiencyoftheDSP.Eachtaponthefilterrequiresonemultiplicationandoneaddition(multiply-accumulate).DSPsaregenerallyoptimizedtoperformfastmultiply-accumulates。
FINITEIMPULSERESPONSE(FIR)FILTERS
Therearetwofundamentaltypesofdigitalfilters:
finiteimpulseresponse(FIR)andinfiniteimpulseresponse(IIR).Astheterminologysuggests,theseclassificationsrefertothefilter!
ˉsimpulseresponse.Byvaryingtheweightofthecoefficientsanthenumberoffiltertaps,virtuallyanyfrequencyresponsecharacteristiccanberealizedwithanFIRfilter.Ashasbeenshown,FIRfilterscanachieveperformancelevelswhicharenotpossiblewithanalogfiltertechniques(suchasperfectlinearphaseresponse).However,highperformanceFIRfiltersgenerallyrequirealargenumberofmultiply-accumulatesandthereforerequirefastandefficientDSPs.Ontheotherhand,IIRfilterstendtomimictheperformanceoftraditionalanalogfiltersandmakeuseoffeedback.Thereforetheirimpulseresponseextendsoveraninfiniteperiodoftime.Becauseoffeedback,IIRfilterscanbeimplementedwithfewercoefficientsthanforanFIRfilter.LatticefiltersaresimplyanotherwaytoimplementeitherFIRorIIRfiltersandareoftenusedinspeechprocessingapplications.Finally,digitalfilterslendthemselvestoadaptivefilteringapplicationssimplybecauseofthespeedandeasewithwhichthefiltercharacteristicscanbechangedbyvaryingthefiltercoefficients.
ThemostelementaryformofanFIRfilterisamovingaveragefilter.Movingaveragefiltersarepopularforsmoothingdata,suchasintheanalysisofstockprices,etc.Theinputsamples,x(n)arepassedthroughaseriesofbufferregisters(labeledz,correspondingtothez-transformrepresentationofadelayelement).Intheexampleshown,therearefourtapscorrespondingtoafour-pointmovingaverage.Eachsampleismultipliedby0.25,andtheseresultsareaddedtoyieldthefinalmovingaverageoutputy(n).
Sincethecoefficientsareequal,aneasierwaytoperformamovingaveragefilterisshowninFigure3.Notethatthefirststepisstorethefirstfoursamples,x(0),x
(1),x
(2),x(3)inaregister.Thesequantitiesareaddedandthenmultipliedby0.25toyieldthefirstoutput,y(3).Notethattheinitialoutputsy(0),y
(1),andy
(2)arenotvalidbecauseallregistersarenotfulluntilsamplex(3)isreceived.
Whensamplex(4)isreceived,itisaddedtotheresult,andsamplex(0)issubtractedfromtheresult.Thenewresultmustthenbemultipliedby0.25.Therefore,thecalculationsrequiredtoproduceanewoutputconsistofoneaddition,onesubtraction,andonemultiplication,regardlessofthelengthofthemovingaveragefilter.
Thestepfunctionresponseofa4-pointmovingaveragefilterisshowninFigure4.Noticethatthemovingaveragefilterhasnoovershoot.Thismakesitusefulinsignalprocessingapplicationswhererandomwhitenoisemustbefilteredbutpulseresponsepreserved.Ofallthepossiblelinearfiltersthatcouldbeused,themovingaverageproducesthelowestnoiseforagivenedgesharpness.ThisisillustratedinFigure5,wherethenoiselevelbecomeslowerasthenumberoftapsareincreased.Noticethatthe0%to100%risetimeofthepulseresponseisequaltothetotalnumberoftapsinthefiltermultipliedbythesamplingperiod.
Adaptivefiltersarewidelyusedincommunicationstoperformsuchfunctionsasequalization,echocancellation,noisecancellation,andspeechcompression.Figure6showsanapplicationofanadaptivefilterusedtocompensatefortheeffectsofamplitudeandphasedistortioninthetransmissionchannel.Thefiltercoefficientsaredeterminedduringatrainingsequencewhereaknowndatapatternistransmitted.Theadaptivealgorithmadjuststhefiltercoefficientstoforcethereceivedatatomatchthetrainingsequencedata.
Inamodemapplication,thetrainingsequenceoccursaftert
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 数字滤波器 外文 翻译