課件說明教案pdaf module calibration acs training

INTERNALPDAFModuleL/RPDpixelsensitivitydifferenceTheconversionfunctionfromPDtoTheconversionfunctionfromlenspositiontoobjectdistance(forextensionfeature) CalibrationCalibrationCalibration4.Gettheoutput 3.Callthedllfromthedll Calibration

1Controlkitto

2.Grabtheraw5.WritetheoutputdatatoEEPROMinsensormodule

Calibration CapturerawdataofDNPlightboxwithmacrolensposition,justliketheshadingcalibrationIfsensorsupportLSCfunction,itmustbecalibratedandenabledfirstbeforestep1Callthedllwiththerawdata,thenwritetheoutputdatafromthedlltoEEPROMinsensorColorTemp.:D50Illumination:~LV10calibrationisok) central?areaexposureGchannelaverage=80LSB320LSB*Useexposuretimetocontrolbrightnesswith1x og*Useexposuretimetocontrolbrightnesswith1x oggainordigitalgaintoavoidgainnoise Dividepdimageintomax20x4CalculatetheluminancedifferenceratioofeachpairofR/LPDpixelsTheresultofSS2P8 Use2Bytestostoreonedifference10bitsforfloatingpoint,remainingfor20sampleinxdirection,4sampleinyThechangeofdifferenceratioarelargeralongxdirectionthanydirectionThus,itneedstotally20x4x2=160StoregridsizeandthevalueofLandRPDpixelsforflexibilityifthereissufficientmemory4+20x4x2x2=324 Capturerawdataofcalibrationpatternat15cmwithlenspositionsfrominfinitytomacroInfinityandmacrolenspositionsaredeterminedfrom100modulesamplesOrjustuseinf.andmacrolenspositionsofeachmoduleiftheyarealreadycalibratedindividuallyDivideequallythelenspositionintonnisconfigurableandrangesfrom5~10sampledlensPatternshouldoccupyentiresensorThussensorshouldsee~30diamondin

Noover&underexpo.EnvironmentsynctoINF,MACROcalibrationis

NormalizedTotalVCMstepto10VCMmin~maxStep=[0~Findthe umofVCMinfinitysteprangeandminimumofmacrosteprangeamong100Forexample:[infinitystep=300,macrostep=Infinitystep:objectdistance=200Macrostep:objectdistance=10TotalVCM10selectedlens

@INF

@Macrocentral?areaexposureGchannelaverage=80LSB320LSB*Useexposuretimetocontrolbrightnesswith1x og*Useexposuretimetocontrolbrightnesswith1x oggainordigitalgaintoavoidgainnoise Callthedllwiththerawdataandthelensposition,thenwritetheoutputdatafromthedlltoEEPROMinsensormodule Dividerawimagesintomax.8x8ApproximatetheconversionfunctionfromPDtoDefocusforeachblock fromPDtoDefocusDefocusmeansthelensmovementtoin-TheconversionfunctioncanbefitbystraightTheconversionfunctionisinvariantfordifferentobjectdistance,thusthecalibrationcanbedoneatonlyoneobjectdistance,ex.15cmSS2P8 fromPDtoDefocusHowever,theconversionfunctionhasspatialdifferenceinslope,thustheimageisdividedintomax.8x8blocks

SS2P8fromPDtoDefocusUse2BytestostoreoneThesloperangesfrom20~60,weuse10bitstostorefloatingpoint,remainingforintegerisenoughUse2Bytestostoreonein-focuslensItneedstotally2x8x8+2=130Storegridsize,samplenum,thelenspos.,andPDvalueforflexibilityifthereissufficientmemoryLenspos.=>2Bytes,PDvalue=>1Bytes,max10Lenspos.arethesameforeach4+1x10x8x8+2x10=664 Capturerawdataofcalibrationpatternat50cmwithlenspositionsfrominfinitytomacroInfinityandmacrolenspositionsaredeterminedfrom100modulesamplesOrjustuseinf.andmacrolenspositionsofeachmoduleiftheyarealreadycalibratedindividuallyDivideequallythelenspositionintonnisconfigurableandrangesfrom5~10sampledlensPutthepatternatthecenterof

Noover&underexpo.Environmentsynctoisok Callthedllwiththerawdataandthelensposition,thenwritetheoutputdatafromthedlltoEEPROMinsensormodule Estimatein-focuslenspositionsat15cm(instep2)and50cm(instep3)byPDcalculationoncentralAFwindowApproximatethelinearfunctionbetweenobjectdistanceandlensposition f Use2Bytestostoreonein-focuslensDACvaluerangesfromItneedstotally2StoresomedebuginfoandthePDvalueforeachlenspositionforflexibilityifthereissufficientmemoryPDvalue=>1Bytes,max1078+1x10=88 Size1L/RPDpixelsensitivitydifferenceratiomapTocompensatethesensitivitydifferencethatcanimprovePDcalculationaccuracy(+122TheconversionfunctionfromPDtodefocusToestimatelensmovementfromcurrentPDandAFwindowposition(+123TheconversionfunctionfromlenspositiontoobjectdistanceAnextensionfeaturetoestimateobjectdistance2(+12Claim2KBytesEEPROMrequirementfor VectorgraphicfileofcalibrationTheaspectratioofcalibrationpatternisDifferentpatternwidthforstep2andstepPatternwidthatstep2=Patternwidthatstep3= RequestmodulehousetoprovidecalibrationoutputdataasonebinfileandtheinputdataofeachstepandtheinifileUseparsingtooltounpacktheoutputbinfileandgenerateout3textfilesasoutput1.txt,output2.txt,andoutput3.txtParsingtoolBinfilenameisthefilenameofcalibration3textfileswillbegeneratedinpd_cali_data CheckCheckiftheinputrawdataisCheckiftheinifileisCheckiftheselectionoflenspositionisCheckiftheoutputdatais Plotoutput1.txtas3Dsurface.YoushouldseethefigurelikebelowR/Lluminance(from1.xto0.xalongx-axis)R-PDpixel

Plotfirstsectionofoutput2.txtas3Dsurface.YoushouldseethefigurelikebelowSpatialslope(largerslopeincornerblocks) Plotsecondsectionofoutput2.txtasscatterwithline.YoushouldseethefigurelikebelowLensEachlineisthepd-lenspos.conversionfunctionforeach8x8blockTheresultshouldbeapproximatedtolinearlines Plotoutput3.txtwithlenspos.asscatterwithline.YoushouldseethefigurelikebelowLesThelineisthepd-lenspos.conversionfunctionofcentralTheresultshouldbe approximatedtoalinearline Q1InStep1,ColorTempD50Illumination:~LV10,產(chǎn)線變異性要考慮DNP有差若沒有DNP，要怎辦？用LEDor光板，可達上述條件即可？OrorLVOK嗎？A1以上拍攝環(huán)境與Kshading時的環(huán)境條件align Q2:InStep1,拍pureraw吧？(因為是在modulevendor的demokit),只要KFullResolution嗎？之後其它resolution(即aspectratio改變，我記得binning是不能用要扣OB？(因為INI沒有可填WBgain不用乘？A2:Yes.拍Pureraw只需拍FullResolution,之後其他resolution可以basedonfullresolution推而OB/WBgain皆不需要apply到capturedraw Q3:InStep2&B.若100個samplesinformarco不含未來量測的informacro是否A10個lensstep我們希望盡可能covervalidVCMrange,所以不建議B