混合信号与数字信号处理设计技术（Mixed-SignalandDSPDesignTechniques）AD转换时的信噪比分析的很精彩摘要：《DSP算法综合与优化》聚焦数字信号处理算法的硬件化实现与自动化优化，围绕多字长定点设计展开系统研究。书中提出信号峰值估计、字长优化与饱和运算分析方法，建立解析噪声模型与面积代价模型，兼顾线性与非线性系统。通过启发式搜索与混合整数线性规划实现字长最优分配，结合调度与资源绑定完成多字长架构综合。全书覆盖FPGA/ASIC实现、自适应滤波、滤波器设计等典型场景，可在满足精度约束下最小化面积、降低功耗并提升运行速度，为DSP算法高效硬件实现提供完整理论与工程方法。SECTION1INTRODUCTIONWaltKesterORIGINSOFREAL-WORLDSIGNALSANDTHEIRUNITSOFMEASUREMENTInthisbook,wewillprimarilybedealingwiththeprocessingofreal-worldsignalsusingbothanaloganddigitaltechniques.Beforestarting,however,let'slookatafewkeyconceptsanddefinitionsrequiredtolaythegroundworkforthingstocome.Webster'sNewCollegiateDictionarydefinesasignalas"Adetectable(ormeasurable)physicalquantityorimpulse(asvoltage,current,ormagneticfieldstrength)bywhichmessagesorinformationcanbetransmitted."Keytothisdefinitionarethewords:detectable,physicalquantity,andinformation.SIGNALCHARACTERISTICS■SignalCharacteristics◆SignalsarePhysicalQuantities◆SignalsareMeasurable◆SignalsContainInformation◆AllSignalsareAnalog■UnitsofMeasurement◆Temperature:°C◆Pressure:Newtons/m2◆Mass:kg◆Voltage:Volts◆Current:Amps◆Power:WattsFigure1.1Bytheirverynature,signalsareanalog,whetherDC,AC,digitallevels,orpulses.Itiscustomary,however,todifferentiatebetweenanaloganddigitalsignalsinthefollowingmanner:Analog(orreal.world)variablesinnatureincludeallmeasurablephysicalquantities.Inthisbook,analogsignalsaregenerallylimitedtoelectricalvariables,theirratesofchange,andtheirassociatedenergyorpowerlevels.Sensorsareusedtoconvertotherphysicalquantities(temperature,pressure,etc.)toelectricalsignals.Theentiresubjectofsignalconditioningdealswithpreparingreal.worldsignalsforprocessingandincludessuchtopicsassensors(temperatureandpressure,forexample),isolationandinstrumentationamplifiers,etc.(seeReference1).Somesignalsresultinresponsetoothersignals.Agoodexampleisthereturnedsignalfromaradarorultrasoundimagingsystem,bothofwhichresultfromaknowntransmittedsignal.Ontheotherhand,thereisanotherclassificationofsignals,calleddigital,wheretheactualsignalhasbeenconditionedandformattedintoadigit.Thesedigitalsignalsmayormaynotberelatedtoreal.worldanalogvariables.Examplesincludethedatatransmittedoverlocalareanetworks(LANs)orotherhighspeednetworks.InthespecificcaseofDigitalSignalProcessing(DSP),theanalogsignalisconvertedintobinaryformbyadeviceknownasananalog.to.digitalconverter(ADC).TheoutputoftheADCisabinaryrepresentationoftheanalogsignalandismanipulatedarithmeticallybytheDigitalSignalProcessor.Afterprocessing,theinformationobtainedfromthesignalmaybeconvertedbackintoanalogformusingadigital.to.analogconverter(DAC).Anotherkeyconceptembodiedinthedefinitionofsignalisthatthereissomekindofinformationcontainedinthesignal.Thisleadsustothekeyreasonforprocessingreal.worldanalogsignals:theextractionofinformation.REASONSFORPROCESSINGREAL-WORLDSIGNALSTheprimaryreasonforprocessingreal.worldsignalsistoextractinformationfromthem.Thisinformationnormallyexistsintheformofsignalamplitude(absoluteorrelative),frequencyorspectralcontent,phase,ortimingrelationshipswithrespecttoothersignals.Oncethedesiredinformationisextractedfromthesignal,itmaybeusedinanumberofways.Insomecases,itmaybedesirabletoreformattheinformationcontainedinasignal.Thiswouldbethecaseinthetransmissionofavoicesignaloverafrequencydivisionmultipleaccess(FDMA)telephonesystem.Inthiscase,analogtechniquesareusedto"stack"voicechannelsinthefrequencyspectrumfortransmissionviamicrowaverelay,coaxialcable,orfiber.Inthecaseofadigitaltransmissionlink,theanalogvoiceinformationisfirstconvertedintodigitalusinganADC.Thedigitalinformationrepresentingtheindividualvoicechannelsismultiplexedintime(timedivisionmultipleaccess,orTDMA)andtransmittedoveraserialdigitaltransmissionlink(asintheT.Carriersystem).Anotherrequirementforsignalprocessingistocompressthefrequencycontentofthesignal(withoutlosingsignificantinformation)thenformatandtransmittheinformationatlowerdatarates,therebyachievingareductioninrequiredchannelbandwidth.Highspeedmodemsandadaptivepulsecodemodulationsystems(ADPCM)makeextensiveuseofdatareductionalgorithms,asdodigitalmobileradiosystems,MPEGrecordingandplayback,andHighDefinitionTelevision(HDTV).Industrialdataacquisitionandcontrolsystemsmakeuseofinformationextractedfromsensorstodevelopappropriatefeedbacksignalswhichinturncontroltheprocessitself.NotethatthesesystemsrequirebothADCsandDACsaswellassensors,signalconditioners,andtheDSP(ormicrocontroller).AnalogDevicesoffersafamilyofMicroConverters™whichincludeprecisionanalogconditioningcircuitry,ADCs,DACs,microcontroller,andFLASHmemoryallonasinglechip.Insomecases,thesignalcontainingtheinformationisburiedinnoise,andtheprimaryobjectiveissignalrecovery.Techniquessuchasfiltering,auto.correlation,convolution,etc.areoftenusedtoaccomplishthistaskinboththeanaloganddigitaldomains.REASONSFORSIGNALPROCESSING■ExtractInformationAboutTheSignal(Amplitude,Phase,Frequency,SpectralContent,TimingRelationships)■ReformattheSignal(FDMA,TDMA,CDMATelephony)■CompressData(Modems,CellularTelephone,HDTV,MPEG)■GenerateFeedbackControlSignal(IndustrialProcessControl)■ExtractSignalFromNoise(Filtering,Autocorrelation,Convolution)■CaptureandStoreSignalinDigitalFormatforAnalysis(FFTTechniques)GENERATIONOFREAL-WORLDSIGNALSInmostoftheaboveexamples(theonesrequiringDSPtechniques),bothADCsandDACsarerequired.Insomecases,however,onlyDACsarerequiredwherereal.worldanalogsignalsmaybegenerateddirectlyusingDSPandDACs.Videorasterscandisplaysystemsareagoodexample.ThedigitallygeneratedsignaldrivesavideoorRAMDAC.Anotherexampleisartificiallysynthesizedmusicandspeech.Inreality,however,thereal.worldanalogsignalsgeneratedusingpurelydigitaltechniquesdorelyoninformationpreviouslyderivedfromthereal.worldequivalentanalogsignals.Indisplaysystems,thedatafromthedisplaymustconveytheappropriateinformationtotheoperator.Insynthesizedaudiosystems,thestatisticalpropertiesofthesoundsbeinggeneratedhavebeenpreviouslyderivedusingextensiveDSPanalysis(i.e.,soundsource,microphone,preamp,ADC,etc.).METHODSANDTECHNOLOGIESAVAILABLEFORPROCESSINGREAL-WORLDSIGNALSSignalsmaybeprocessedusinganalogtechniques(analogsignalprocessing,orASP),digitaltechniques(digitalsignalprocessing,orDSP),oracombinationofanaloganddigitaltechniques(mixedsignalprocessing,orMSP).Insomecases,thechoiceoftechniquesisclear;inothers,thereisnoclearcutchoice,andsecond.orderconsiderationsmaybeusedtomakethefinaldecision.WithrespecttoDSP,thefactorthatdistinguishesitfromtraditionalcomputeranalysisofdataisitsspeedandefficiencyinperformingsophisticateddigitalprocessingfunctionssuchasfiltering,FFTanalysis,anddatacompressioninrealtime.Thetermmixedsignalprocessingimpliesthatbothanaloganddigitalprocessingisdoneaspartofthesystem.Thesystemmaybeimplementedintheformofaprintedcircuitboard,hybridmicrocircuit,orasingleintegratedcircuitchip.Inthecontextofthisbroaddefinition,ADCsandDACsareconsideredtobemixedsignalprocessors,sincebothanaloganddigitalfunctionsareimplementedineach.RecentadvancesinVeryLargeScaleIntegration(VLSI)processingtechnologyallowcomplexdigitalprocessingaswellasanalogprocessingtobeperformedonthesamechip.TheverynatureofDSPitselfimpliesthatthesefunctionscanbeperformedinreal-time.ANALOGVERSUSDIGITALSIGNALPROCESSINGToday'sengineerfacesachallengeinselectingthepropermixofanaloganddigitaltechniquestosolvethesignalprocessingtaskathand.Itisimpossibletoprocessreal.worldanalogsignalsusingpurelydigitaltechniques,sinceallsensors(microphones,thermocouples,straingages,microphones,piezoelectriccrystals,diskdriveheads,etc.)areanalogsensors.Therefore,somesortofsignalconditioningcircuitryisrequiredinordertopreparethesensoroutputforfurthersignalprocessing,whetheritbeanalogordigital.Signalconditioningcircuitsare,inreality,analogsignalprocessors,performingsuchfunctionsasmultiplication(gain),isolation(instrumentationamplifiersandisolationamplifiers),detectioninthepresenceofnoise(highcommon.modeinstrumentationamplifiers,linedrivers,andlinereceivers),dynamicrangecompression(logamps,LOGDACs,andprogrammablegainamplifiers),andfiltering(bothpassiveandactive).SeveralmethodsofaccomplishingsignalprocessingareshowninFigure1.3.Thetopportionofthefigureshowsthepurelyanalogapproach.ThelatterpartsofthefigureshowtheDSPapproach.NotethatoncethedecisionhasbeenmadetouseDSPtechniques,thenextdecisionmustbewheretoplacetheADCinthesignalpath.Ingeneral,astheADCismovedclosertotheactualsensor,moreoftheanalogsignalconditioningburdenisnowplacedontheADC.TheaddedADCcomplexitymaytaketheformofincreasedsamplingrate,widerdynamicrange,higherresolution,inputnoiserejection,inputfilteringandprogrammablegainamplifiers(PGAs),on.chipvoltagereferences,etc.,allofwhichaddfunctionalityandsimplifythesystem.Withtoday’shigh.resolution/highsamplingratedataconvertertechnology,significantprogresshasbeenmadeinintegratingmoreandmoreoftheconditioningcircuitrywithintheADC/DACitself.Inthemeasurementarea,forinstance,24.bitADCsareavailablewithbuilt.inprogrammablegainamplifiers(PGAs)whichallowfullscalebridgesignalsof10mVtobedigitizeddirectlywithnofurtherconditioning(e.g.AD773x.series).Atvoicebandandaudiofrequencies,completecoder.decoders(Codecs–orAnalogFrontEnds)areavailablewhichhavesufficienton.chipanalogcircuitrytominimizetherequirementsforexternalconditioningcomponents(AD1819BandAD73322).Atvideospeeds,analogfrontendsarealsoavailableforsuchapplicationsasCCDimageprocessingandothers(e.g.,AD9814,AD9816,andtheAD984xseries).APRACTICALEXAMPLEAsapracticalexampleofthepowerofDSP,considerthecomparisonbetweenananalogandadigitallowpassfilter,eachwithacutofffrequencyof1kHz.ThedigitalfilterisimplementedinatypicalsampleddatasystemshowninFigure1.4.Notethatthereareseveralimplicitrequirementsinthediagram.First,itisassumedthatanADC/DACcombinationisavailablewithsufficientsamplingfrequency,resolution,anddynamicrangetoaccuratelyprocessthesignal.Second,theDSPmustbefastenoughtocompleteallitscalculationswithinthesamplinginterval,1/fs.Third,analogfiltersarestillrequiredattheADCinputandDACoutputforantialiasingandanti.imaging,buttheperformancedemandsarenotasgreat.Assumingtheseconditionshavebeenmet,thefollowingoffersacomparisonbetweenthedigitalandanalogfilters.Therequiredcutofffrequencyofbothfiltersis1kHz.Theanalogfilterisrealizedasa6.poleChebyshevType1filter(rippleinpassband,norippleinstopband),andtheresponseisshowninFigure1.5.Inpractice,thisfilterwouldprobablyberealizedusingthree2.polestages,eachofwhichrequiresanopamp,andseveralresistorsandcapacitors.ModernfilterdesignCADpackagesmakethe6.poledesignrelativelystraightforward,butmaintainingthe0.5dBripplespecificationrequiresaccuratecomponentselectionandmatching.Ontheotherhand,the129.tapdigitalFIRfiltershownhasonly0.002dBpassbandripple,linearphase,andamuchsharperrolloff.Infact,itcouldnotberealizedusinganalogtechniques!Anotherobviousadvantageisthatthedigitalfilterrequiresnocomponentmatching,anditisnotsensitivetodriftsincetheclockfrequenciesarecrystalcontrolled.The129.tapfilterrequires129multiply.accumulates(MAC)inordertocomputeanoutputsample.Thisprocessingmustbecompletedwithinthesamplinginterval,1/fs,inordertomaintainreal.timeoperation.Inthisexample,thesamplingfrequencyis10kSPS,therefore100µsisavailableforprocessing,assumingnosignificantadditionaloverheadrequirement.TheADSP.21xx.familyofDSPscancompletetheentiremultiply.accumulateprocess(andotherfunctionsnecessaryforthefilter)inasingleinstructioncycle.Therefore,a129.tapfilterrequiresthattheinstructionratebegreaterthan129/100µs=1.3millioninstructionspersecond(MIPS).DSPsareavailablewithinstructionratesmuchgreaterthanthis,sotheDSPcertainlyisnotthelimitingfactorinthisapplication.TheADSP.218x16.bitfixedpointseriesoffersinstructionratesupto75MIPS.TheassemblylanguagecodetoimplementthefilterontheADSP.21xx.familyofDSPsisshowninFigure1.6.Notethattheactuallinesofoperatingcodehavebeenmarkedwitharrows;therestarecomments.更多内容见附件免责声明：本页面/内容部分素材来源于互联网公开信息，旨在传递更多信息，不代表本平台立场。版权归原作者或机构所有，如涉及侵权，请通过平台联系我们，我们将在核实后第一时间处理。本平台对转载内容的真实性、准确性不作任何保证，用户需自行判断并承担使用风险
生成制造
