TI DSP培训ppt1

Chapter1IntroductionDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide2LearningObjectivesWhyprocesssignalsdigitally?Definitionofareal-timeapplication.WhyuseDigitalSignalProcessingprocessors?WhatarethetypicalDSPalgorithms?ParameterstoconsiderwhenchoosingaDSPprocessor.ProgrammablevsASICDSP.TexasInstruments’TMS320family.Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide3Whygodigital?Digitalsignalprocessingtechniquesarenowsopowerfulthatsometimesitisextremelydifficult,ifnotimpossible,foranaloguesignalprocessingtoachievesimilarperformance.Examples:FIRfilterwithlinearphase.Adaptivefilters.Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide4Whygodigital?Analoguesignalprocessingisachievedbyusinganaloguecomponentssuchas:Resistors.Capacitors.Inductors.Theinherenttolerancesassociatedwiththesecomponents,temperature,voltagechangesandmechanicalvibrationscandramaticallyaffecttheeffectivenessoftheanaloguecircuitry.Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide5Whygodigital?WithDSPitiseasyto:Changeapplications.Correctapplications.Updateapplications.AdditionallyDSPreduces:Noisesusceptibility.Chipcount.Developmenttime.Cost.Powerconsumption.Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide6WhyNOTgodigital?Highfrequencysignalscannotbeprocesseddigitallybecauseoftworeasons:AnalogtoDigitalConverters,ADCcannotworkfastenough.Theapplicationcanbetoocomplextobeperformedinreal-time.Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide7DSPprocessorshavetoperformtasksinreal-time,sohowdowedefinereal-time?Thedefinitionofreal-timedependsontheapplication.Example:a100-tapFIRfilterisperformedinreal-timeiftheDSPcanperformandcompletethefollowingoperationbetweentwosamples:Real-timeprocessing990kknxkanyDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide8Wecansaythatwehaveareal-timeapplicationif:WaitingTime0Real-timeprocessingProcessingTimeWaitingTimeSampleTimenn+1Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide9WhynotuseaGeneralPurposeProcessor(GPP)suchasaPentiuminsteadofaDSPprocessor?WhatisthepowerconsumptionofaPentiumandaDSPprocessor?WhatisthecostofaPentiumandaDSPprocessor?WhydoweneedDSPprocessors?Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide10UseaDSPprocessorwhenthefollowingarerequired:Costsaving.Smallersize.Lowpowerconsumption.Processingofmany“high”frequencysignalsinreal-time.UseaGPPprocessorwhenthefollowingarerequired:Largememory.Advancedoperatingsystems.WhydoweneedDSPprocessors?Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide11WhatarethetypicalDSPalgorithms?AlgorithmEquationFiniteImpulseResponseFilterMkkknxany0)()(InfiniteImpulseResponseFilterNkkMkkknybknxany10)()()(ConvolutionNkknhkxny0)()()(DiscreteFourierTransform10])/2(exp[)()(NnnkNjnxkXDiscreteCosineTransform10122cos).().(NxxuNxfucuFTheSumofProducts(SOP)isthekeyelementinmostDSPalgorithms:Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide12Hardwarevs.MicrocodemultiplicationDSPprocessorsareoptimisedtoperformmultiplicationandadditionoperations.Multiplicationandadditionaredoneinhardwareandinonecycle.Example:4-bitmultiply(unsigned).1011x11101011x1110HardwareMicrocode1001101000001011.1011..1011...10011010Cycle1Cycle2Cycle3Cycle4Cycle5Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide13ParameterstoconsiderwhenchoosingaDSPprocessorParameterArithmeticformatExtendedfloatingpointExtendedArithmeticPerformance(peak)NumberofhardwaremultipliersNumberofregistersInternalL1programmemorycacheInternalL1datamemorycacheInternalL2cache32-bitN/A40-bit1200MIPS2(16x16-bit)with32-bitresult3232K32K512K32-bit64-bit40-bit1200MFLOPS2(32x32-bit)with32or64-bitresult3232K32K512KTMS320C6211(@150MHz)TMS320C6711(@150MHz)C6711Datasheet:\Links\TMS320C6711.pdfC6211Datasheet:\Links\TMS320C6211.pdfDr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide14ParameterstoconsiderwhenchoosingaDSPprocessorParameterI/Obandwidth:SerialPorts(number/speed)DMAchannelsMultiprocessorsupportSupplyvoltagePowermanagementOn-chiptimers(number/width)CostPackageExternalmemoryinterfacecontrollerJTAG2x75Mbps16Notinherent3.3VI/O,1.8VCoreYes2x32-bitUS$21.54256PinBGAYesYes2x75Mbps16Notinherent3.3VI/O,1.8VCoreYes2x32-bitUS$21.54256PinBGAYesYesTMS320C6211(@150MHz)TMS320C6711(@150MHz)Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide15Floatingvs.FixedpointprocessorsApplicationswhichrequire:Highprecision.Widedynamicrange.Highsignal-to-noiseratio.Easeofuse.Needafloatingpointprocessor.Drawbackoffloatingpointprocessors:Higherpowerconsumption.Canbemoreexpensive.Canbeslowerthanfixed-pointcounterpartsandlargerinsize.Dr.NaimDahnoun,BristolUniversity,(c)TexasInstruments2004Chapter1,Slide16Floatingvs.FixedpointprocessorsItistheapplicationthatdictateswhichdeviceandplatformtouseinorde