文献翻译——利用AT89CX051微控制器进行模拟数字转换

第页1外文文献资料Analog-to-DigitalConversionUtilizingtheAT89CX051MicrocontrollersTheAtmelAT89C1051andAT89C2051.microcontrollersfeatureon-chipFlash,lowpincount,wideoperatingvoltage,rangeandanintegralanalogcomparator.Thisapplicationnotedescribestwolow-costanalogtodigitalconversiontechniqueswhichutilizetheanalogcomparatorintheAT89C1051andAT89C2051microcontrollers.1.RCAnalog-to-DigitalConverterThisconversionmethodoffers.Anextremelylowcomponentcountattheexpenseofaccuracyandconversiontime.Intheexamplepresentedbelow,resolutionisbetterthan50milliovolts,accuracyissomewhatlessthanatenthofaVoltandconversiontimeissevenmillisecondsorless.AsshowninFigure1,theRCanalogtodigital.ConversionmethodrequiresonlytworesistorsandacapacitorinadditiontotheAT89CX051microcontroller.Amicrocontrolleroutput(pin11),whichswingsfromapproximatelygroundtoVCC,alternatelychargesanddischargesthecapacitorconnectedtothenon-invertinginputoftheinternalcomparator(pin12).Themicrocontrollermeasuresthetimerequiredforthevoltageonthecapacitortomatchtheunknownvoltageappliedtotheinvertinginputoftheinternalcomparator(pin13).Theunknownvoltageisafunctionofthemeasuredtime.TheHP5082-7300LEDdisplaysshowninFigure1arenotrequiredfortheConversion,butareutilizedbythesoftwaretoimplementasimpletwo-digitvoltmeter.Theresultoftheanalog-to-digitalconversionisdisplayedinvoltsandtenthsofavoltonthetwodisplays.ThevoltmeterapplicationdoesnotutilizethefullresolutionoftheRCconversionsoftware,butservestodemonstratethemethodaswellasprovidingatoolfordebug.Thewaveformforatypicalcapacitorcharge/dischargecycleisshowninFigure2.Thedischargeportionofthecurveisidenticaltothechargeportionrotated第页2aboutthelineVC=VCC/2.Theequationsanddiscussionbelowapplytothechargeportionofthecycle,exceptwhereindicated.Thevoltageonthecapacitorasafunctionoftimeisgivenbytheexponentialequation:VC=VCC(1-e-t/RC)(1)WhereVCisthevoltageonthecapacitorattimet,VCCisthesupplyvoltageandRCistheproductofthevaluesoftheresistorandcapacitor.NotethatvoltageisexpressedinVolts,timeinseconds,resistanceinOhmsandcapacitanceinFarads.TheproductRCisalsoknownasthe“timeconstant”ofthenetworkandaffectstheshapeofthewaveform.Thewaveformissteepestwhencapacitorchargingordischargingbeginsandflattenswithtime.ThefirstproblemwiththeRCconversionmethodisthedifficultyofsolvingtheexponentialequationwithoututilizingfloatingpointcalculationsandtranscendentalfunctions.Onacompressedtimescale,theexponentialcurveappearsstraightovermuchofitslength,suggestingthatitmightbeapproximatedbyaline.Thisschemefailsduetothecontinuousvariationinslopeoverthelengthofthecurve,whichproducessignificanterror.ItalsodoesnotaddresstheproblemwherethecurverollsoffseverelyneartheasymptoteatVCC.Themicrocontrollerneednotsolvetheexponentialequationinrealtimeifalookuptableisusedtomappre-calculatedvaluestoeachsampledtimeinterval.Thisschemeallowsthedatatobeencodedandformattedasrequiredbytheapplicationwhilesimplifyingtheconversionsoftware.Symmetriesinthedatamaybeexploitedtoreducethesizeofthetable.ThesecondproblemwiththeRCconversionmethodisthesubstantialerrorwhichresultsfromvariationsincomponentvalues.Figure3showsanexaggeratedviewofthevariationinthevoltageonthecapacitorduetovariationsinthevaluesoftheresistorandcapacitor.Asshowninthefigure,thevariationinthevoltageonthecapacitordecreasesasthevoltageonthecapacitordecreases.Thesymmetryofthecapacitorcharge/dischargecyclecanbeexploitedtoreducetheeffectofvariationsincomponentvaluesonconversionaccuracy.ThisisdonebyutilizingthechargeportionofthecycletomeasurevoltageslessthanVCC/2andthedischargeportiontomeasurevoltagesgreaterthanVCC/2.Theworstcase第页3errorisreducedtotheerroratVCC/2.Beforecomponentvaluescanbeassigned,thetimeintervalatwhichtheComparatoroutputistobesampledmustbedetermined.Thesampleintervalshouldbeasshortaspossibletomaximizeconverterresolutionandminimizeconversiontime.Thesampleintervalislimitedbythetimerequiredtoexecutetherequisitecode,whichisdeterminedbytheclockrateofthemicrocontroller.Inthevoltmeterapplication,themicrocontrolleroperateswitha12-MHzclock,resultinginasampleintervaloffivemicroseconds.Thetimeconstant(RC)affectstheshapeofthecapacitorcharge/dischargewaveform.Thevalueofthetimeconstantmustbechosensothatthesteepestpartsofthewaveformareresolvabletothedesiredresolution.Thesteepestpartofthechargeportionofthewaveformoccursneartheorigin,whilethesteepestpartofthedischargeportionoccursnearVCC.Duetothesymmetryofthewaveform,thesametimeconstantmaybeusedformeasurementsmadeoneitherportionofthewaveform.Thefollowingshowsanexpandedviewoftherelationshipbetweenvoltageandsampletimeneartheorigin.Inthefigure,VisthedesiredvoltageresolutionoftheconverterandTisthesampleintervaldeterminedpreviously.ThecurvelabeledVCrepresentsthevoltageonthecapacitor,whichappearslinearatthisscale.Inthefigure,theslopeofthecurveisideal,causingsamplingtooccurnearthecenterofthevoltageintervals.Theslopeofthecurvemaybelessthanshown,butmaynotbegreater,orresolutionwillbelost.Notethatthefirstsampleisoffsetfromtheoriginby1/2ttocenterthesampleinthefirstvoltageinterval.Toobtaintheminimumvalueofthetimeconstantwhichwillproducetherequiredslopeatthefirstsample,solveEquation1forRC:RC=-t/1n(1-VC/VCC)(2)TheproductofthevaluesofRandCmustnotbelessthanthecalculatedminimumtimeconstant.Utilizingaresistorwithaonepercenttoleranceandacapacitorwithafivepercenttolerance(Rnorm-1%)(Cnorm-5%)4.99*10-4Inthevoltmeterapplication,theselectedvaluesofRandCare267kilohmsand2Nanfarads,respectively,yieldingaminimumtimeconstantofapproximately5.0210-4.AnadditionalconstraintisplacedonthevalueofR.Referringagainto第页4Figure1,notethe5.1kilohmpullupresistorconnectedtopin11ofthemicrocontroller.Thisresistorispresenttosupplementthemicrocontrollersweakinternalpullup,buthasthedetrimentaleffectofchangingthetimeconstantoftheRCnetworkduringthechargeportionofthecapacitorcharge/dischargecycle.Thisproducesanasymmetryinthecharge/dischargewaveform,whichcontributestoconversionerror.Tominimizetheeffectofdifferencesinthecapacitorchargeanddischargepaths,thevalueofRshouldbechosentobemuchgreaterthanthevalueofthepullupresistor.Inthevoltmeterapplication,theselectedvalueofRis267kilohms,whichexceedsthevalueofthepullupresistorbymorethananorderofmagnitude.Thetimeconstant(RC),whichisafunctionofthedesiredconverterresolution,determinesthedurationofthecapacitorcharge/dischargecycle.Themoretimerequiredforthecapacitortochargeanddischarge,thegreaterthenumberofsamplesrequiredinthemeasurementloopandthegreaterthenumberofentriesinthelookuptable.Ctothesymmetryofthecapacitorcharge/dischargewaveform,thedeterminedsamplecountmaybeusedformeasurementsmadeduringeitherportionofthecycle.FromEquation3:Tmax=-RmaxCmaxln(1-(1/2)VCC/VCC)=-(Rnom+1%)(Cnom+5%)ln(1/2)=-(1.01)(267103)(1.05)(210-9)ln(1/2)393s.Theminimumnumberofsamplesforhalfthecycleis:10-6)/(510-6)=79Tomaximizeaccuracy,voltagesfromzerotoVCC/2aremeasuredduringthechargeportionofthecapacitorcharge/dischargecycleandvoltagesfromVCCtoVCC/2aremeasuredduringthedischargeportionofthecycle.Asaresult,thetotalnumberofentriesinthetableistwicethenumberofsamplescalculatedpreviouslyforeachhalfcycle.Thelookuptablecontainsapplication-specificvaluescorrespondingtothecalculatedvoltageateachsample.Foreachhalfcycle,theNthentryinthetablecorrespondstothevoltageatt=(Nthesampleintervaldeterminedpreviously.Forthe第页5chargehalfcycle,thevoltageateachsampleiscalculatedbysolvingEquation1forthetimeelapsedsincethecapacitorbegantocharge.Forthedischargehalfcycle,thevoltageateachsampleiscalculatedbysolvingthefollowingequationforthetimeelapsedsincethecapacitorbegantodischarge:VC=VCCe-t/RC(4)Thesizeandcontentsofthetablemayvaryfromapplicationtoapplicationdependingonthesampleintervalandconversionresolution.Astheresolutionincreases,thenumberofentriesinthetablegrows.Inthevoltmeterapplication,withresolutionequalto0.05Volt,thelookuptablecontains158entries,whichistwicethenumberofsamplesperhalfcyclecalculatedabove.VoltagescorrespondingtosamplestakenduringthechargehalfcyclearecalculatedEquation1,whereNrepresentsthesamplenumber(0-78).Rto267kilohms,Cto2Nanfarads,andVCCto5.00-volts,Equation1becomes:V=5(1-e-N(.0093633)VoltagescorrespondingtosamplestakenduringthedischargehalfcycleareEquation4,whereNrepresentsthesamplenumber(0-78).Usingthesamevaluesasforthechargehalfcycle,Equation4becomes:V=5e-N(.0093633)Anabbreviatedlistofthevoltagescalculatedforthecapacitorcharge/dischargecycleisshownbelow.Theorderingofthevoltages,increasinginthefirsthalf,decreasinginthesecond,tracksthevoltageonthecapacitoranddefinestheorderingofthetableentries.N=0V=0.000N=1V=0.047.N=74V=2.499N=75V=2.523第页6N=76V=2.546N=77V=2.569N=78V=2.591N=0V=5.000N=1V=4.953.N=74V=2.501N=75V=2.477N=76V=2.454N=77V=2.431N=78V=2.409Asshownbythelist,thenumberofsamplesineachhalfcycleisgreaterthanrequiredtoreachthemidrangevalueof2.500-volts.Thisallowsfor“fast”cycleswhichovershootthenominalmidrangevaluebeforethelastsampleistakenineachhalfcycle.NotethatthedifferencebetweenthecalculatedvoltagesatsamplesN=0andN=1iswithinthedesiredresolutionof0.050-volt,butthedifferenceinvoltagebetweenadjacentsamplesdecreasesasNincreases.Thisreflectsthenon-linearrelationshipbetweenvoltageandtimeinthecircuit.Thecalculatedvoltagesshowninthelistarenotenteredintothelookuptable,butareusedtodeterminethevaluesofthetableentries.Inthevoltmeterapplication,thecalculatedvoltagesareroundedtotenthsofavoltandtheresultstoredinthetableinpacked-BCDform,twodigitsperbyte.Example:thetableentrycorrespondingto2.523-voltsis25hex,whichdisplaysas2.5-volts.Thevoltmeterprototypedemonstratedaccuracyof+/-onecount(0.1Volt),butaccuracyofsomewhatlessthanatenthofaVoltisaboutthebestthatcanbeexpectedfromtheRCprecisioncomponents,variationsincomponentvaluesmaycontributeanerrorof0.104-volt,asshownbelow.TocalculatetheworstcaseerroratVC=2.5-volts,firstdeterminethecorrespondingtatthenominalvaluesofRandCusingEquation3:第页7t=-RnomCnomln(1-VC/VCC)=-RnomCnomln(1-2.5/5.0)=-RnomCnomln(0.5).SubstitutefortinEquation1togetminimumVC:VCmin=VCC(1-e-t/(RmaxCmax)=VCC(1-e(RnomCnom/RmaxCmax)ln(0.5)=5(1-eln(0.5)/(1.011.05)2.399VAgain,formaximumVC:VCmax=VCC(1-e-t/(RminCmin)=VCC(1-e(RnomCnom/RminCmin)ln(0.5)=5(1-eln(0.5)/(0.990.95)2.607VTheresultsshowavariationof0.208-voltsat2.5-volts,oraworstcaseerrorof0.104-volts.Theworstcaseconversionerrormaybefurtherreducedbyutilizingcomponentswithtightertolerances.Conversionaccuracyandlinearityarealsoaffectedbythecharacteristicsofthecapacitor.Thecapacitorusedinthevoltmeterprototypeisapolystyrenefilmtype,whichnotonlyprovidesgoodaccuracy,butanalog-to-digitalconversionmethod.Evenusingminimizeserrorduetodielectricabsorptionandothereffects.Errorsourceswhichhavenotbeenexaminedinclude:comparatorlimitations;asymmetriesbetweenthechargeanddischargeportionsofthecycle;failureofthevoltageonthecapacitortoreachgroundorVCC;variationsinVCC.Thecontributionstoconversionerrormadebythesesourcescanbeexpectedtoincreaseerrortosomewhatmorethanthevalueduetocomponenttolerancesalone.2.SuccessiveApproximationAnalog-to-DigitalConverterThisconversionmethodoffersgoodresolutionandaccuracyandashortconversiontimeattheexpenseofincreasedcomponentcount.Successiveapproximation(SA)ADCsincorporateadigitalto-analogconverter(DAC),acomparatorandasuccessiveapproximationregister(SAR).第页8TheSARcontrolstheconversionbyperformingasearchforthebinarycodewhich,whenfedtotheDAC,willproduceanoutputmatchingthevoltagetobeconverted.ThecomparatorcomparestheDACoutputtotheunknownvoltageandreturnstheresulttotheSAR.TheSARbeginsthesearchwiththemostsignificantDACbit,whichcontrolsthewidestoutputvariation,andmovestowardtheleastsignificantbit,causingtheDACoutputto“zeroin”ontheunknownvalue.Theresultofthetrialisthebinarycodecorrespondingtotheunknownvalue.Inaneight-bitSAconverter,onlyeightiterationsarerequiredtofindthecorrectbinarycode,resultinginrelativelyfastconversions.Inthisapplication(Figure5),anAT89CX051microcontrollerwithanintegralanalogcomparatorperformstheSARfunctioninsoftware,reducingthecomponentcount.TheDACselectedfortheapplicationisanMC1408-8,eight-bit,currentoutputtypechosenforitslowcost.Seven-andsixbitversionsareavailableastheMC1408-7andMC1408-6,respectively.TheMC1408seriesisguaranteedaccuratetoof1.992milliamps.TherelativeaccuracyoftheMC1408-8isbetterthan0.19%,assuringeight-bitmonotonicityandlinearity.TheDAChasanoutputsettlingtimeof300nanoseconds.TheDACcontainsbinary-weighted,current-steeringswitcheswhichscaleaninputcurrentbytheappliedbinarycode.TheinputcurrentisderivedfromanLM336-2.5precisionvoltagereferenceandaseriesresistor.ThescaledcurrentoutputisconvertedtoavoltagebyanLF355Boperationalamplifierwiredasacurrent-to-voltage(I/V)converter.TheLF355BopampwasselectedfortheI/Vconverterbecauseofitslowinputoffsetvoltageandhighoutputslewrate.ThevoltageoutputoftheI/VconverterisfedintotheAT89CX051comparator,whereitiscomparedtotheunknownvoltage.Whentheprogrammedvoltageexceedstheunknownvoltagetheoutputofthecomparatorgoeshigh,whichisdetectedbysoftware.Asecondopamp,wiredasanon-inverting,unitygainbuffermaybeinsertedbetweentheunknownvoltagesourceandtheinputtotheAT89CX051comparatortoprovideisolation.TheLM336-2.5referenceprovidesanominal2.490-voltoutput(Vref).Theactualvoltagemayvaryfrom2.390-voltsto2.590-volts.Thereferencevoltageand第页9temperaturecoefficientmaybetrimmedusingthemethodindicatedintheLM336-2.5datasheet.Thenominalvalueofthecurrentreferenceresistor(Rref)connectedtopin14oftheDACis1240Ohms,yieldingareferencecurrent(Iref)of2.490V/1240Ohms(Vref/Rref)=2.008milliamps.Theeight-bitbinarycodeappliedtotheDACscalesIrefbyfrom0/256to255/256,resultinginacurrentoutput(Io)offromzero(Iref0/256)to2.000milliamps(Iref255/256)fullscale.NotethatthesignoftheDACoutputcurrentisoppositethesignofthereference(input)current.TheoutputvoltageisdeterminedbymultiplyingtheDACoutputcurrent(Io)bythevalueoftheI/Vconvertergainresistor(Ro).Nominalfullscaleoutputvoltageis2.000mA2500Ohms(IoF.S.Ro)=5.000-volts.Thecircuitdoesnotprovideadjustmentsforoffsetorgain.Offsetvoltageadjustmentsshouldnotberequired,duetothelowoffsetvoltagespecificationoftheLF355Bopamp.Iftheoffsetvoltagemustbeadjusted,addtheoffsettrimcircuitshownintheLF355Bdatasheet.ThegainmaybechangedbychangingthevalueoftheI/Vconvertergainresistor(Ro).Theresistorconnectedtothenon-invertinginputoftheopampshouldbeofthesamevalueasthegainresistorforinputbiascurrentbalancing.The1240Ohmresistorconnectedtopin15oftheDACandthe2500Ohmresistorconnectedtopinthreeoftheopampmaybeeliminatedwithonlyaslightdecreaseinperformance.TheMC1408-8DACrequirespowersuppliesof+5.0-voltsand-5.0to-15-volts;5.0-voltsupplieswereselectedtominimizepowerconsumption.TheLF355Bopamprequiresbipolarsuppliesbetween5.0-voltsand15-volts.-5.0-voltswasselectedforthenegativerailforcompatibilitywiththeDAC,butmaybereplacedwith-15-volts,ifdesired.Thepositivesupplywaschosentobe+15-voltstoallowthelimitedoutputswingoftheopamptoreachthefiveVoltupperinputlimitofthecomparator.ThespeedoftheA-to-DconversionislimitedbytheDACoutputsettlingtime,theslewrateandsettlingtimeoftheopamp,theresponsetimeandslewrateofthecomparatorandthetimerequiredtoexecutethesuccessiveapproximationalgorithm.TheDACoutputsettlingtimeandthecomparatorresponsetimearenegligiblecomparedtoopampdelaysandthetimerequiredtoexecutetheSAalgorithm,andsomaybeignored.Themaximumvoltage第页10stepinputtotheopampisfivevolts,whichrequiresonemicrosecondtoslewandfourmicrosecondstosettle(seetheLF355Bdatasheet).Thisdelayisaccommodatedinthesoftware;consultthelistingforadditionalinformation.Witha12MHzprocessorclockandtheresultingonemicrosecondinstructioncycle,aneight-bitconversioncanbeperformedinunder300microseconds.Theunknowninputvoltagemustbeheldconstantforthedurationoftheconversion.ObviousdisadvantagestothesuccessiveapproximationanalogtodigitalconverterpresentedherearetheneedforbipolarpowersuppliesandthelargenumberofmicrocontrollerI/OpinsrequiredtocontroltheDAC.The+15-voltsupplycouldbeeliminatedbyreplacingtheLF355Bopampwithasinglesupply,5-volt,functionalequivalentwithoutputsthatswingrail-to-rail.ThenumberofmicrocontrollerI/OpinsrequiredtocontroltheDACcouldbereducedsomewhatbysubstitutingasevenorsixbitDAC.TheparallelinputDACcouldbereplacedwitha(moreexpensive)serialinputDAC.Alternately,logiccouldbeaddedtoacceptserialdatafromthemicrocontrollerandpresentparalleldatatotheDAC.ThesoftwareforthisapplicationmaybeobtainedbydownloadingfromAtmelsBBS:(408)436-4309.Consultthecommentblockatthebeginningofthesourcecodefilefordetailedinformationonfeaturesandoper.第页11中文翻译稿利用AT89CX051微控制器进行模拟数字转换Atmel公司的AT89C1051和AT89C2051微控制器功能芯片上的Flash,销量低、工作电压范围宽而且带有不可或缺的模拟比较器。这篇短文描述了两种低成本的模拟到数字的转换技术,它利用AT89C1051和AT89C2051微控制器的模拟比较器。1.RC模拟-数字转换器这种转换方法提供极低的组件数量的准确性和转换时间。下面的示例中,分辨率大于50毫伏,准确性有些不到十分之一伏特和转换时间是7毫秒或更少。RC模拟到数字转换方法除了AT89CX051单片机外只需要两个电阻和一个电容。单片机的输出(P11),从近地VCC开始波动,时而充放电，电容器连接到内部的非反相输入比较器(P12)。单片机措施所需的时间为电容器上的电压匹配未知电压内部比较器的反相输入(P13)，电压测量时间的函数未知。hp5082-7300LED显示屏不需要如下的转换,但利用软件实现一个简单的两位数电压表。模拟数字转换的结果在两个显示器间显示为伏特和零点几伏。电压表的应用程序不使用完整的RC转换软件,但是用来证明方法以及提供的调试工具。波形显示了一个典型的电容充电/放电循环。放电部分曲线是相同的电荷部分旋转对线VC=VCC/2。下面的方程和讨论适用于收取部分的循环除外。电容上的电压作为时间的函数的指数方程:VC=VCC(1-e-t/RC)(1)VC在电容器上的电压在时间t,VCC是电源电压和RC是产品的电阻和电容的值。注意表达伏特电压,时间以秒为单位,在欧姆电阻和电容在法拉。产品RC也被称为“时间常数”的网络和影响波形的形状。波形是最大电容器充电或放电开始时,随着时间的推移趋于平稳。RC转换方法存在的第一个问题解决的难度指数第页12方程没有利用浮点计算和超越函数。规模压缩时间,连续出现指数曲线的长度,这表明它可能被一条线近似。这计划失败由于连续变化斜率曲线的长度,产生重大的错误。它还不解决这个问题,曲线辊严重的渐近线VCC附近。单片机不需要解决实时指数方程如果使用一个查找表pre-calculated值映射到每个采样时间间隔。这个计划允许数据被编码和格式所需的应用程序,同时简化转换软件。对称的数据可能被利用来减少表的大小。第二个问题与RC转换方法是大量错误组件值变化的结果。下面显示了一个夸张的电容器上的电压的变化由于电阻和电容的值的变化。在电容器上的电压变化会随着电容器上的电压降低。电容器充电/放电循环的对称性可以利用减少组件值变化对转换精度的影响。这是通过利用收取的部分周期测量电压小于VCC/2和放电部分测量电压大于VCC/2。最坏的情况下误差减少到VCC/2。组件值被分配之前,必须确定比较器输出的时间间隔采样。取样间隔应尽可能短,以最大限度地提高转换器的分辨率和减少转换时间。取样间隔有限执行必要的代码所需的时间,这是由微控制器的时钟频率决定。电压表的应用,单片机运行12-MHz时钟,导致一个示例间隔5微秒。时间常数(RC)影响电容器充电/放电波形的形状。时间常数的值必须选择这样的最陡的部分波形可解析所需的分辨率。最大的一部分费用部分波形发生在原点附近,而最大放电部分VCC附近发生的一部分。由于波形的对称性,同一时间常数可以用于测量波形的一部分了。下面显示了一个扩展的原点附近的电压和样品时间之间的关系。在图中,V是所需的电压转换器的分辨率和T是取样间隔决定之前。标记为“风投”的曲线代表电容上的电压,这似乎在这种规模的线性。在图中,曲线的斜率是理想的,导致采样电压间隔的中心附近发生。曲线的斜率可能少于所示,但可能不是更大,或决议将丢失。注意,第一个样品是抵消从原点by1/2t中心示例在第一电压区间。获取的时间常数的最小值在第一个示例中,将生产所需的斜坡解决方程1RC:RC=1-t/n(1-VC/VCC)(2)的产物R和C的值不能小于最小时间常数计算。利用电阻的百分之十一宽