您好,欢迎访问三七文档
CircuitCellarApril2010ConverterPerformanceManagementDesignTipsforWorkingwithOn-ChipADCs片上ADC转换性能的优化片上多通道ADC设计技巧指南翻译:norman33第一版Whencreatinganapplicationthatusesamicrocontroller’sintegratedADC,youneedtoknowhowalltheparts(e.g.,sensors)canaffectthesystem’soverallperformance.Thisarticleincludesdesigntipsforgettingoptimalperformancefromanon-chipADC.当在您设计中使用mcu芯片的片上集成ADC时,您必须清楚的知道,在系统中所有的部件(例如:传感器等)都能够对整个系统性能造成影响。本文讲述了如何使一个片上集成ADC发挥最佳性能的技巧。Often,whendesigningembeddedsystemswithanalog-to-digitalconverters(ADCs)builtintothemicrocontroller,adesignerwillspendalotoftimecalculatingtheresolutionneededforthesystembutneglecttospendanytimeunderstandinghowotheraspectscanaffecttheanalog-to-digital(A/D)performance.Thisarticleexploressomeoftheseotheraspectsandhowtheycaninfluence(A/D)performance.Itisnotenoughtojustunderstandthespecificationsinthemanufacturer’sdatasheetandhowtheyaffectperformance.Sometimes,systemparameterscanbethelimitingfactorinperformance,andnottheADCitself.通常当在一个嵌入式系统中使用片上集成ADC的时候,设计者会花很多时间计算ADC在整个系统工作的分辨率,而电路的其他部分对ADC的影响则往往被他们忽略了。本文将讨论这些被他们忽略的部分和这些忽略的部分是如何影响片上集成AD性能的。我们要知道,要最优化片上集成ADC的性能,对于制造商提供的数据手册表面上的理解是远远不够的。时常,系统参数设置不当往往是造成整个系统性能不良的最大原因,而不是片上集成ADC自身的问题。ADCARCHITECTURESADC的构架ThetwomostcommonADCarchitecturesfoundintegratedintotoday’smicrocontrollersareSuccessiveApproximation(SAR)andSigmaDeltaconverters.当前片上集成ADC的主流构架有逐次逼近型ADC(SAR)和西格玛-德尔塔(Sigma-Delta)型两种。Figure1illustratesablockdiagramofasuccessiveapproximationconvertercommonlyfoundonmodern-daymicrocontrollers.图1用框图的形式表现了现在常见的逐次逼近型ADC的结构框架。Thebasicpremisebehindtheoperationofthisconverteristhattheanalogsignalbeingsampledislatchedintoasampleandholdcircuitandfedintoacomparator.这个构架的工作基本原理是模拟信号被采样然后和进入采样保持电路,然后送入比较电路。ThesignaliscomparedtotheoutputofanR2Rladderdigital-to-analogconverter(DAC).信号经比较电路后输入至一个R2R阶梯数模转换电路(DAC)ThedigitalreadingisconstructedsequentiallystartingwiththeMSBandworkingdowntotheLSBbyhavingtheoutputoftheDACsuccessivelycomparedtothesampledsignaluntiltheoperationiscomplete.输出的数字信号由MSB开始到LSB结束,由比较器持续对输入的采样信号和基准信号进行比较并逐位输出,直至整个操作结束。KeyadvantagestothistypeofADCarethatitisrelativelyfastanditiseasytomultiplexsignals.这类ADC构架的关键优点是,它的速度快而且多路复用容易实现。However,highresolutionSARADCscangetveryexpensiveduetotheanalogcircuitrythatcanbedifficulttomanufacturewithadigitalprocess.缺点是高分辨率的SARADC可能非常昂贵,由于在高分辨率下,此类构架ADC的模拟电路部分在和处理器一起生产的时候会非常困难。TheSigmaDeltaADCisspecificallydesignedforhighresolution.SigmaDelta构架的ADC本身就是针对需要高分辨率情况下设计出来的。AsshowninFigure2,the4-Vsignalisfedintoanerroramplifierandtheresultisfedintoanintegrator.如图24V输入信号送入一个误差放大器,此误差放大器的输出结果送入了下级积分电路。Theintegratoroutput,theslopeofwhichisproportionaltotheinputvoltage,iscomparedtothereferencevoltage.Theflip-flopgeneratesthefirstbit(MSB).积分电路的输出信号的斜率和输入信号成正比,此信号再和基准信号做比较后输入下级触发电路,由触发电路生成数字信号的最高位(MSB)。TheadvantageofthistypeofADCisthatitismostlyimplementedindigitalcircuitry,makingiteasytomanufacture,andwiththeuseofdigitalfiltering,noiselevelscanbekeptverylow.这种构架的ADC的优点是,几乎所有电路的实现都是基于数字电路的。所以此构架的电路很容易生产,如果配合数字滤波技术,电路整体的噪声水平能够控制在非常低的水准上。Thisalsoendsupbeingadisadvantageifhigh-speedconversionratesarerequiredduetotheneedforhigh-speeddigitalcircuitsinordertoachievethis.TheSigmaDeltaarchitecturerequiresmanysamplesinordertoconstructthedigitalsignalwhichcaninduceaphasedelaythatmaybeunacceptabletoapplicationswhereminimizingphasedelayisimportant.这个电路也有缺点,如果需要一个这个构架的高转换数率的ADC,它就需要非常高的采样率来输出数字信号,这个过程将产生相位延迟。在需要相位延迟最小化的设计中这将不能被接受。Figure3showsanidealtransfercurveforanADC.图3显示了一个ADC的传输特性曲线。Herewehaveastraightlinestartingat0countsat0Vextendinguptofullcountandthehighestvoltage.图上有一条从输出数字0-0v为起点延伸至满输出-最高电压的理想直线。Therealtransfercurveisgoingtodeviatefromthisidealcurve.现实的传输特性曲线会偏离这个理想的曲线。Itwillnotbe0countsat0Vandwillinfacthaveanoffsetknownastheoffseterror.现实的传输曲线不会为0-0V为起点。现实的起点为这个点加上一个偏移量,这个偏移量叫做偏移误差(offseterror)。Conversely,whenthevoltageisfullscale,theA/Dcountsmaynotbefullscale.相反的,当输入最高电压时,A/D输出的往往不是满量程的值。Inthatcase,wehaveafull-scaleerror.这个误差叫做满量程误差(full-scaleerror)。WhendesigningembeddedsystemsthatuseADCs,themanufacturer’sdatasheetwilloftendenotethefull-scaleandoffseterror.芯片生厂商的数据手册中一般都会标明偏移误差和满量程误差的值。Theseerrorscantypicallybecalibratedoutwithtechniquesthatarewellknown.这两种误差一般都可以用我们常见的办法去修正。Notethattheidealcurveisalinear,monotonicallyincreasingfunction.Therealcurveisindeedmonotonicbutisnotlinear.我们需要注意的是,理想曲线是一个线性的、单调递增的函数。现实的曲线是一个单调递增但是不是线性的函数。Thenonlinearityerrorbecomesmoredifficulttocalibrate,anddoingsoisoftenimpractical.要完全修正非线性失真是非常困难而且几乎是不可能的。SYSTEM-LEVELSYSTEM-LEVELSYSTEM-LEVELSYSTEM-LEVELPOVPOVPOVPOV着眼于系统层来讨论Nowwewilltakealookatsomethingstoconsiderfromasystemlevelpointofview.现在我们将在着眼于系统层上来讨论。First,itisimportanttoconsidertheparameterbeingmeasuredversustheactualresolutionoftheADC.首先,我们要考虑对于检测信号来说,ADC分辨率的选择问题。Ofcourse,oneoftenassumesthatmoreresolutionisalwaysbetter,butyoumaynotevenneedit.当然,我们经常认为分辨率越高越好,但实际上我们并不需要那么高的分辨率。Let’sconsidera10-bitADCasanexample.让我们假定有一个10-bit的ADC。Inthiscase,wecanresolveupto1,024counts.10-bit的ADC能够最高有1024个档位。Ifthemeasurementvariableistemperaturefrom0to250°C,ourexampleADCcanresolvedownto0.25ofad
本文标题:Circuit Cellar April 2010-Design Tips for Working
链接地址:https://www.777doc.com/doc-3611619 .html