/* ======================================================================== *\ ! ! * ! * This file is part of MARS, the MAGIC Analysis and Reconstruction ! * Software. It is distributed to you in the hope that it can be a useful ! * and timesaving tool in analysing Data of imaging Cerenkov telescopes. ! * It is distributed WITHOUT ANY WARRANTY. ! * ! * Permission to use, copy, modify and distribute this software and its ! * documentation for any purpose is hereby granted without fee, ! * provided that the above copyright notice appear in all copies and ! * that both that copyright notice and this permission notice appear ! * in supporting documentation. It is provided "as is" without express ! * or implied warranty. ! * ! ! ! Author(s): Markus Gaug 02/2004 ! ! Copyright: MAGIC Software Development, 2000-2004 ! ! \* ======================================================================== */ ////////////////////////////////////////////////////////////////////////////// // // MCalibrationChargeCalc // // Task to calculate the calibration conversion factors and quantum efficiencies // from the fit results to the summed FADC slice distributions delivered by // MCalibrationChargeCam, MCalibrationChargePix, MCalibrationChargeBlindPix and // MCalibrationChargePINDiode, calculated and filled by MHCalibrationChargeCam, // MHCalibrationChargePix, MHCalibrationChargeBlindPix and MHCalibrationChargePINDiode. // // PreProcess(): Initialize pointers to MCalibrationChargeCam, MCalibrationChargeBlindPix // MCalibrationChargePINDiode and MCalibrationQECam // // Initialize pulser light wavelength // // ReInit(): MCalibrationCam::InitSize(NumPixels) is called from MGeomApply (which allocates // memory in a TClonesArray of type MCalibrationChargePix) // Initializes pointer to MBadPixelsCam // // Process(): Nothing to be done, histograms getting filled by MHCalibrationChargeCam // // PostProcess(): - FinalizePedestals() // - FinalizeCharges() // - FinalizeFFactorMethod() // - FinalizeBadPixels() // - FinalizeBlindPixel() // - FinalizeBlindCam() // - FinalizePINDiode() // - FinalizeFFactorQECam() // - FinalizeBlindPixelQECam() // - FinalizePINDiodeQECam() // // Input Containers: // MCalibrationChargeCam // MCalibrationChargeBlindPix // MCalibrationChargePINDiode // MCalibrationQECam // MPedestalCam // MBadPixelsCam // MGeomCam // MTime // // Output Containers: // MCalibrationChargeCam // MCalibrationChargeBlindPix // MCalibrationChargePINDiode // MCalibrationQECam // MBadPixelsCam // // // Preliminary description of the calibration in photons (email from 12/02/04) // // Why calibrating in photons: // =========================== // // At the Barcelona meeting in 2002, we decided to calibrate the camera in // photons. This for the following reasons: // // * The physical quantity arriving at the camera are photons. This is // the direct physical information from the air shower. The photons // have a flux and a spectrum. // // * The photon fluxes depend mostly on the shower energy (with // corrections deriving from the observation conditions), while the photon // spectra depend mostly on the observation conditions: zenith angle, // quality of the air, also the impact parameter of the shower. // // * The photomultiplier, in turn, has different response properties // (quantum efficiencies) for photons of different colour. (Moreover, // different pixels have slightly different quantum efficiencies). // The resulting number of photo-electrons is then amplified (linearly) // with respect to the photo-electron flux. // // * In the ideal case, one would like to disentagle the effects // of the observation conditions from the primary particle energy (which // one likes to measure). To do so, one needs: // // 1) A reliable calibration relating the FADC counts to the photo-electron // flux -> This is accomplished with the F-Factor method. // // 2) A reliable calibration of the wavelength-dependent quantum efficiency // -> This is accomplished with the combination of the three methods, // together with QE-measurements performed by David in order to do // the interpolation. // // 3) A reliable calibration of the observation conditions. This means: // - Tracing the atmospheric conditions -> LIDAR // - Tracing the observation zenith angle -> Drive System // // 4) Some knowlegde about the impact parameter: // - This is the only part which cannot be accomplished well with a // single telescope. We would thus need to convolute the spectrum // over the distribution of impact parameters. // // // How an ideal calibration would look like: // ========================================= // // We know from the combined PIN-Diode and Blind-Pixel Method the response of // each pixel to well-measured light fluxes in three representative // wavelengths (green, blue, UV). We also know the response to these light // fluxes in photo-electrons. Thus, we can derive: // // - conversion factors to photo-electrons // - conversion factors to photons in three wavelengths. // // Together with David's measurements and some MC-simulation, we should be // able to derive tables for typical Cherenkov-photon spectra - convoluted // with the impact parameters and depending on the athmospheric conditions // and the zenith angle (the "outer parameters"). // // From these tables we can create "calibration tables" containing some // effective quantum efficiency depending on these outer parameters and which // are different for each pixel. // // In an ideal MCalibrate, one would thus have to convert first the FADC // slices to Photo-electrons and then, depending on the outer parameters, // look up the effective quantum efficiency and get the mean number of // photons which is then used for the further analysis. // // How the (first) MAGIC calibration should look like: // =================================================== // // For the moment, we have only one reliable calibration method, although // with very large systematic errors. This is the F-Factor method. Knowing // that the light is uniform over the whole camera (which I would not at all // guarantee in the case of the CT1 pulser), one could in principle already // perform a relative calibration of the quantum efficiencies in the UV. // However, the spread in QE at UV is about 10-15% (according to the plot // that Abelardo sent around last time. The spread in photo-electrons is 15% // for the inner pixels, but much larger (40%) for the outer ones. // // I'm not sure if we can already say that we have measured the relative // difference in quantum efficiency for the inner pixels and produce a first // QE-table for each pixel. To so, I would rather check in other wavelengths // (which we can do in about one-two weeks when the optical transmission of // the calibration trigger is installed). // // Thus, for the moment being, I would join Thomas proposal to calibrate in // photo-electrons and apply one stupid average quantum efficiency for all // pixels. This keeping in mind that we will have much preciser information // in about one to two weeks. // // // What MCalibrate should calculate and what should be stored: // =========================================================== // // It is clear that in the end, MCerPhotEvt will store photons. // MCalibrationCam stores the conversionfactors to photo-electrons and also // some tables of how to apply the conversion to photons, given the outer // parameters. This is not yet implemented and not even discussed. // // To start, I would suggest that we define the "average quantum efficiency" // (maybe something like 25+-3%) and apply them equally to all // photo-electrons. Later, this average factor can be easily replaced by a // pixel-dependent factor and later by a (pixel-dependent) table. // // // ////////////////////////////////////////////////////////////////////////////// #include "MCalibrationChargeCalc.h" #include #include #include #include "MLog.h" #include "MLogManip.h" #include "MParList.h" #include "MRawRunHeader.h" #include "MRawEvtPixelIter.h" #include "MGeomCam.h" #include "MGeomPix.h" #include "MHCamera.h" #include "MPedestalCam.h" #include "MPedestalPix.h" #include "MCalibrationChargeCam.h" #include "MCalibrationChargePix.h" #include "MCalibrationChargePINDiode.h" #include "MCalibrationChargeBlindPix.h" #include "MCalibrationChargeBlindCam.h" #include "MExtractedSignalCam.h" #include "MExtractedSignalPix.h" #include "MExtractedSignalBlindPixel.h" #include "MExtractedSignalPINDiode.h" #include "MBadPixelsCam.h" #include "MBadPixelsPix.h" #include "MCalibrationQECam.h" #include "MCalibrationQEPix.h" #include "MCalibrationCam.h" ClassImp(MCalibrationChargeCalc); using namespace std; const Float_t MCalibrationChargeCalc::fgChargeLimit = 2.5; const Float_t MCalibrationChargeCalc::fgChargeErrLimit = 0.; const Float_t MCalibrationChargeCalc::fgChargeRelErrLimit = 1.; const Float_t MCalibrationChargeCalc::fgLambdaErrLimit = 0.2; const Float_t MCalibrationChargeCalc::fgLambdaCheckLimit = 0.5; const Float_t MCalibrationChargeCalc::fgPheErrLimit = 4.5; const Float_t MCalibrationChargeCalc::fgFFactorErrLimit = 4.5; // -------------------------------------------------------------------------- // // Default constructor. // // Sets the pointer to fQECam and fGeom to NULL // // Calls AddToBranchList for: // - MRawEvtData.fHiGainPixId // - MRawEvtData.fLoGainPixId // - MRawEvtData.fHiGainFadcSamples // - MRawEvtData.fLoGainFadcSamples // // Initializes: // - fChargeLimit to fgChargeLimit // - fChargeErrLimit to fgChargeErrLimit // - fChargeRelErrLimit to fgChargeRelErrLimit // - fFFactorErrLimit to fgFFactorErrLimit // - fLambdaCheckLimit to fgLambdaCheckLimit // - fLambdaErrLimit to fgLambdaErrLimit // - fPheErrLimit to fgPheErrLimit // - fPulserColor to MCalibrationCam::kCT1 // - fOutputPath to "." // - fOutputFile to "ChargeCalibStat.txt" // - flag debug to kFALSE // // Calls: // - Clear() // MCalibrationChargeCalc::MCalibrationChargeCalc(const char *name, const char *title) : fQECam(NULL), fGeom(NULL) { fName = name ? name : "MCalibrationChargeCalc"; fTitle = title ? title : "Task to calculate the calibration constants and MCalibrationCam "; AddToBranchList("MRawEvtData.fHiGainPixId"); AddToBranchList("MRawEvtData.fLoGainPixId"); AddToBranchList("MRawEvtData.fHiGainFadcSamples"); AddToBranchList("MRawEvtData.fLoGainFadcSamples"); SetChargeLimit (); SetChargeErrLimit (); SetChargeRelErrLimit (); SetFFactorErrLimit (); SetLambdaCheckLimit (); SetLambdaErrLimit (); SetPheErrLimit (); SetOutputPath (); SetOutputFile (); SetDebug ( kFALSE ); Clear(); } // -------------------------------------------------------------------------- // // Sets: // - all variables to 0., // - all flags to kFALSE // - all pointers to NULL // - the pulser colour to kNONE // - fBlindPixelFlags to 0 // - fPINDiodeFlags to 0 // void MCalibrationChargeCalc::Clear(const Option_t *o) { fNumHiGainSamples = 0.; fNumLoGainSamples = 0.; fSqrtHiGainSamples = 0.; fSqrtLoGainSamples = 0.; fNumInnerFFactorMethodUsed = 0; fBadPixels = NULL; fCam = NULL; fBlindPixel = NULL; fBlindCam = NULL; fPINDiode = NULL; fPedestals = NULL; SetPulserColor ( MCalibrationCam::kNONE ); fBlindPixelFlags.Set(0); fPINDiodeFlags .Set(0); fResultFlags .Set(0); } // ----------------------------------------------------------------------------------- // // The following container are searched for and execution aborted if not in MParList: // - MPedestalCam // // The following containers are searched and created if they were not found: // // - MCalibrationQECam // - MBadPixelsCam // Int_t MCalibrationChargeCalc::PreProcess(MParList *pList) { // // Containers that have to be there. // fPedestals = (MPedestalCam*)pList->FindObject("MPedestalCam"); if (!fPedestals) { *fLog << err << "MPedestalCam not found... aborting" << endl; return kFALSE; } // // Containers that are created in case that they are not there. // fQECam = (MCalibrationQECam*)pList->FindCreateObj("MCalibrationQECam"); if (!fQECam) { *fLog << err << "Cannot find nor create MCalibrationQECam... aborting" << endl; return kFALSE; } fBadPixels = (MBadPixelsCam*)pList->FindCreateObj("MBadPixelsCam"); if (!fBadPixels) { *fLog << err << "Could not find or create MBadPixelsCam ... aborting." << endl; return kFALSE; } // // Check the pulser colour --> FIXME: this solution is only valid until the arrival of the DM's // if (fPulserColor == MCalibrationCam::kNONE) { *fLog << endl; *fLog << err << GetDescriptor() << ": No Pulser colour has been chosen. Since the installation of the IFAE pulser box," << " you HAVE to provide the LEDs colour, otherwise there is no calibration. " << endl; *fLog << "See e.g. the macro calibration.C " << endl; return kFALSE; } return kTRUE; } // -------------------------------------------------------------------------- // // Search for the following input containers and abort if not existing: // - MGeomCam // - MCalibrationChargeCam // // Search for the following input containers and give a warning if not existing: // - MCalibrationChargeBlindPix // - MCalibrationChargePINDiode // // It retrieves the following variables from MCalibrationChargeCam: // // - fNumHiGainSamples // - fNumLoGainSamples // // It defines the PixId of every pixel in: // // - MCalibrationChargeCam // - MCalibrationQECam // // It sets all pixels in excluded which have the flag fBadBixelsPix::IsBad() set in: // // - MCalibrationChargePix // - MCalibrationQEPix // // Sets the pulser colour and tests if it has not changed w.r.t. fPulserColor in: // // - MCalibrationChargeCam // - MCalibrationChargeBlindPix (if existing) // - MCalibrationChargePINDiode (if existing) // Bool_t MCalibrationChargeCalc::ReInit(MParList *pList ) { fGeom = (MGeomCam*)pList->FindObject("MGeomCam"); if (!fGeom) { *fLog << err << "No MGeomCam found... aborting." << endl; return kFALSE; } fCam = (MCalibrationChargeCam*)pList->FindObject("MCalibrationChargeCam"); if (!fCam) { *fLog << err << "Cannot find MCalibrationChargeCam... aborting" << endl; *fLog << err << "Maybe you forget to call an MFillH for the MHCalibrationChargeCam before..." << endl; return kFALSE; } // // Optional Containers // fBlindPixel = (MCalibrationChargeBlindPix*)pList->FindObject("MCalibrationChargeBlindPix"); if (!fBlindPixel) { fBlindCam = (MCalibrationChargeBlindCam*)pList->FindObject("MCalibrationChargeBlindCam"); if (!fBlindCam) { *fLog << endl; *fLog << warn << GetDescriptor() << ": MCalibrationChargeBlindPix nor MCalibrationChargeBlindCam " << " found... no Blind Pixel method! " << endl; } } fPINDiode = (MCalibrationChargePINDiode*)pList->FindObject("MCalibrationChargePINDiode"); if (!fPINDiode) { *fLog << endl; *fLog << warn << GetDescriptor() << ": MCalibrationChargePINDiode not found... no PIN Diode method! " << endl; } // // Initialize the pulser colours // if (fCam->GetPulserColor() == MCalibrationCam::kNONE) { fCam->SetPulserColor( fPulserColor ); if (fBlindPixel) fBlindPixel->SetColor( fPulserColor ); if (fBlindCam) fBlindCam->SetColor( fPulserColor ); if (fPINDiode) fPINDiode->SetColor( fPulserColor ); } if (fPulserColor != fCam->GetPulserColor()) { *fLog << err << GetDescriptor() << ": Pulser colour has changed w.r.t. last file in MCalibrationChargeCam" << endl; *fLog << err << "This feature is not yet implemented, sorry ... aborting " << endl; return kFALSE; } if (fBlindPixel) if (fPulserColor != fBlindPixel->GetColor()) { *fLog << err << GetDescriptor() << ": Pulser colour has changed w.r.t. last file in MCalibrationChargeBlindPix." << endl; *fLog << err << "This feature is not yet implemented, sorry ... aborting " << endl; return kFALSE; } if (fBlindCam) if (fPulserColor != fBlindCam->GetColor()) { *fLog << err << GetDescriptor() << ": Pulser colour has changed w.r.t. last file in MCalibrationChargeBlindCam." << endl; *fLog << err << "This feature is not yet implemented, sorry ... aborting " << endl; return kFALSE; } if (fPINDiode) if (fPulserColor != fPINDiode->GetColor()) { *fLog << err << GetDescriptor() << ": Pulser colour has changed w.r.t. last file in MCalibrationChargePINDiode." << endl; *fLog << err << "This feature is not yet implemented, sorry ... aborting " << endl; return kFALSE; } fNumHiGainSamples = fCam->GetNumHiGainFADCSlices(); fNumLoGainSamples = fCam->GetNumLoGainFADCSlices(); fSqrtHiGainSamples = TMath::Sqrt(fNumHiGainSamples); fSqrtLoGainSamples = TMath::Sqrt(fNumLoGainSamples); UInt_t npixels = fGeom->GetNumPixels(); for (UInt_t i=0; iIsValid()) { *fLog << warn << GetDescriptor() << ": MCalibrationChargePINDiode is declared not valid... no PIN Diode method! " << endl; fPINDiode = NULL; } if (fBlindPixel) if (!fBlindPixel->IsValid()) { *fLog << warn << GetDescriptor() << ": MCalibrationChargeBlindPix is declared not valid... no Blind Pixel method! " << endl; fBlindPixel = NULL; } *fLog << endl; // // First loop over pixels, call FinalizePedestals and FinalizeCharges // Int_t nvalid = 0; for (Int_t pixid=0; pixidGetSize(); pixid++) { MCalibrationChargePix &pix = (MCalibrationChargePix&)(*fCam)[pixid]; // // Check if the pixel has been excluded from the fits // if (pix.IsExcluded()) continue; MPedestalPix &ped = (*fPedestals)[pixid]; MBadPixelsPix &bad = (*fBadPixels)[pixid]; const Int_t aidx = (*fGeom)[pixid].GetAidx(); FinalizePedestals(ped,pix,aidx); if (FinalizeCharges(pix,bad,"pixel ")) nvalid++; } *fLog << endl; // // The Michele check ... // if (nvalid == 0) { *fLog << err << GetDescriptor() << ": All pixels have non-valid calibration. " << "Did you forget to fill the histograms " << "(filling MHCalibrationChargeCam from MExtractedSignalCam using MFillH) ? " << endl; *fLog << err << GetDescriptor() << ": Or, maybe, you have used a pedestal run " << "instead of a calibration run " << endl; return kFALSE; } for (UInt_t aidx=0; aidxGetNumAreas(); aidx++) { const MPedestalPix &ped = fPedestals->GetAverageArea(aidx); MCalibrationChargePix &pix = (MCalibrationChargePix&)fCam->GetAverageArea(aidx); FinalizePedestals(ped,pix,aidx); FinalizeCharges(pix, fCam->GetAverageBadArea(aidx),"area id"); } *fLog << endl; for (UInt_t sector=0; sectorGetNumSectors(); sector++) { const MPedestalPix &ped = fPedestals->GetAverageSector(sector); MCalibrationChargePix &pix = (MCalibrationChargePix&)fCam->GetAverageSector(sector); FinalizePedestals(ped,pix, 0); } *fLog << endl; // // Finalize Bad Pixels // FinalizeBadPixels(); // // Finalize F-Factor method // if (!FinalizeFFactorMethod()) { *fLog << warn << "Could not calculate the photons flux from the F-Factor method " << endl; fCam->SetFFactorMethodValid(kFALSE); return kFALSE; } else fCam->SetFFactorMethodValid(kTRUE); *fLog << endl; // // Finalize Blind Pixel // if (fBlindPixel) if (FinalizeBlindPixel()) fQECam->SetBlindPixelMethodValid(kTRUE); else fQECam->SetBlindPixelMethodValid(kFALSE); else if (FinalizeBlindCam()) fQECam->SetBlindPixelMethodValid(kTRUE); else fQECam->SetBlindPixelMethodValid(kFALSE); // // Finalize PIN Diode // if (FinalizePINDiode()) fQECam->SetPINDiodeMethodValid(kTRUE); else fQECam->SetPINDiodeMethodValid(kFALSE); // // Finalize QE Cam // FinalizeFFactorQECam(); FinalizeBlindPixelQECam(); FinalizePINDiodeQECam(); // // Re-direct the output to an ascii-file from now on: // MLog asciilog; asciilog.SetOutputFile(GetOutputFile(),kTRUE); SetLogStream(&asciilog); // // Finalize calibration statistics // FinalizeUnsuitablePixels(); fCam ->SetReadyToSave(); fQECam ->SetReadyToSave(); fBadPixels->SetReadyToSave(); if (fBlindPixel) fBlindPixel->SetReadyToSave(); if (fBlindCam) fBlindCam->SetReadyToSave(); if (fPINDiode) fPINDiode->SetReadyToSave(); *fLog << inf << endl; *fLog << GetDescriptor() << ": Fatal errors statistics:" << endl; PrintUncalibrated(MBadPixelsPix::kChargeIsPedestal, Form("%s%2.1f%s","Signal less than ",fChargeLimit," Pedestal RMS: ")); PrintUncalibrated(MBadPixelsPix::kChargeRelErrNotValid, Form("%s%2.1f%s","Signal Error bigger than ",fChargeRelErrLimit," times Mean Signal: ")); PrintUncalibrated(MBadPixelsPix::kLoGainSaturation, "Pixels with Low Gain Saturation: "); PrintUncalibrated(MBadPixelsPix::kMeanTimeInFirstBin, Form("%s%2.1f%s","Mean Abs. Arr. Time in First ",1.," Bin(s): ")); PrintUncalibrated(MBadPixelsPix::kMeanTimeInLast2Bins, Form("%s%2.1f%s","Mean Abs. Arr. Time in Last ",2.," Bin(s): ")); PrintUncalibrated(MBadPixelsPix::kDeviatingNumPhes, "Pixels with deviating number of phes: "); PrintUncalibrated(MBadPixelsPix::kDeviatingFFactor, "Pixels with deviating F-Factor: "); *fLog << inf << endl; *fLog << GetDescriptor() << ": Unreliable errors statistics:" << endl; PrintUncalibrated(MBadPixelsPix::kChargeSigmaNotValid, "Signal Sigma smaller than Pedestal RMS: "); PrintUncalibrated(MBadPixelsPix::kHiGainOscillating, "Pixels with changing Hi Gain signal over time: "); PrintUncalibrated(MBadPixelsPix::kLoGainOscillating, "Pixels with changing Lo Gain signal over time: "); PrintUncalibrated(MBadPixelsPix::kHiGainNotFitted, "Pixels with unsuccesful Gauss fit to the Hi Gain: "); PrintUncalibrated(MBadPixelsPix::kLoGainNotFitted, "Pixels with unsuccesful Gauss fit to the Lo Gain: "); SetLogStream(&gLog); return kTRUE; } // ---------------------------------------------------------------------------------- // // Retrieves pedestal and pedestal RMS from MPedestalPix // Retrieves total entries from MPedestalCam // Sets pedestal*fNumHiGainSamples and pedestal*fNumLoGainSamples in MCalibrationChargePix // Sets pedRMS *fSqrtHiGainSamples and pedRMS *fSqrtLoGainSamples in MCalibrationChargePix // // If the flag MCalibrationPix::IsHiGainSaturation() is set, call also: // - MCalibrationChargePix::CalcLoGainPedestal() // void MCalibrationChargeCalc::FinalizePedestals(const MPedestalPix &ped, MCalibrationChargePix &cal, const Int_t aidx) { // // get the pedestals // const Float_t pedes = ped.GetPedestal(); const Float_t prms = ped.GetPedestalRms(); const Int_t num = fPedestals->GetTotalEntries(); // // RMS error set by PedCalcFromLoGain, 0 in case MPedCalcPedRun was used. // const Float_t prmserr = num>0 ? prms/TMath::Sqrt(2.*num) : ped.GetPedestalRmsError(); // // set them in the calibration camera // if (cal.IsHiGainSaturation()) { cal.SetPedestal(pedes * fNumLoGainSamples, prms * fSqrtLoGainSamples, prmserr * fSqrtLoGainSamples); cal.CalcLoGainPedestal((Float_t)fNumLoGainSamples, aidx); } else { cal.SetPedestal(pedes * fNumHiGainSamples, prms * fSqrtHiGainSamples, prmserr * fSqrtHiGainSamples); } } // ---------------------------------------------------------------------------------------------------- // // Check fit results validity. Bad Pixels flags are set if: // // 1) Pixel has a mean smaller than fChargeLimit*PedRMS ( Flag: MBadPixelsPix::kChargeIsPedestal) // 2) Pixel has a mean error smaller than fChargeErrLimit ( Flag: MBadPixelsPix::kChargeErrNotValid) // 3) Pixel has mean smaller than fChargeRelVarLimit times its mean error // ( Flag: MBadPixelsPix::kChargeRelErrNotValid) // 4) Pixel has a sigma bigger than its Pedestal RMS ( Flag: MBadPixelsPix::kChargeSigmaNotValid ) // // Further returns if flags: MBadPixelsPix::kUnsuitableRun is set // // Calls MCalibrationChargePix::CalcReducedSigma() and sets flag: MBadPixelsPix::kChargeIsPedestal // and returns kFALSE if not succesful. // // Calls MCalibrationChargePix::CalcFFactor() and sets flag: MBadPixelsPix::kDeviatingNumPhes) // and returns kFALSE if not succesful. // // Calls MCalibrationChargePix::CalcConvFFactor()and sets flag: MBadPixelsPix::kDeviatingNumPhes) // and returns kFALSE if not succesful. // Bool_t MCalibrationChargeCalc::FinalizeCharges(MCalibrationChargePix &cal, MBadPixelsPix &bad, const char* what) { if (bad.IsUnsuitable(MBadPixelsPix::kUnsuitableRun)) return kFALSE; if (cal.GetMean() < fChargeLimit*cal.GetPedRms()) { *fLog << warn << GetDescriptor() << Form(": Fitted Charge: %5.2f is smaller than %2.1f",cal.GetMean(),fChargeLimit) << Form(" Pedestal RMS: %5.2f in %s%4i",cal.GetPedRms(),what,cal.GetPixId()) << endl; bad.SetUncalibrated( MBadPixelsPix::kChargeIsPedestal); } if (cal.GetMean() < fChargeRelErrLimit*cal.GetMeanErr()) { *fLog << warn << GetDescriptor() << Form(": Fitted Charge: %4.2f is smaller than %2.1f",cal.GetMean(),fChargeRelErrLimit) << Form(" times its error: %4.2f in %s%4i",cal.GetMeanErr(),what,cal.GetPixId()) << endl; bad.SetUncalibrated( MBadPixelsPix::kChargeRelErrNotValid ); } if (cal.GetSigma() < cal.GetPedRms()) { *fLog << warn << GetDescriptor() << Form(": Sigma of Fitted Charge: %6.2f is smaller than",cal.GetSigma()) << Form(" Ped. RMS: %5.2f in %s%4i",cal.GetPedRms(),what,cal.GetPixId()) << endl; bad.SetUncalibrated( MBadPixelsPix::kChargeSigmaNotValid ); return kFALSE; } if (!cal.CalcReducedSigma()) { *fLog << warn << GetDescriptor() << Form(": Could not calculate the reduced sigma in %s: ",what) << Form(" %4i",cal.GetPixId()) << endl; bad.SetUncalibrated( MBadPixelsPix::kChargeSigmaNotValid ); return kFALSE; } if (!cal.CalcFFactor()) { *fLog << warn << GetDescriptor() << Form(": Could not calculate the F-Factor in %s: ",what) << Form(" %4i",cal.GetPixId()) << endl; bad.SetUncalibrated(MBadPixelsPix::kDeviatingNumPhes); return kFALSE; } if (!cal.CalcConvFFactor()) { *fLog << warn << GetDescriptor() << Form(": Could not calculate the Conv. FADC counts to Phes in %s: ",what) << Form(" %4i",cal.GetPixId()) << endl; bad.SetUncalibrated(MBadPixelsPix::kDeviatingNumPhes); return kFALSE; } return kTRUE; } // ----------------------------------------------------------------------------------- // // Sets pixel to MBadPixelsPix::kUnsuitableRun, if one of the following flags is set: // - MBadPixelsPix::kChargeIsPedestal // - MBadPixelsPix::kChargeRelErrNotValid // - MBadPixelsPix::kMeanTimeInFirstBin // - MBadPixelsPix::kMeanTimeInLast2Bins // - MBadPixelsPix::kDeviatingNumPhes // - MBadPixelsPix::kHiGainOverFlow // - MBadPixelsPix::kLoGainOverFlow // // - Call MCalibrationPix::SetExcluded() for the bad pixels // // Sets pixel to MBadPixelsPix::kUnreliableRun, if one of the following flags is set: // - MBadPixelsPix::kChargeSigmaNotValid // void MCalibrationChargeCalc::FinalizeBadPixels() { for (Int_t i=0; iGetSize(); i++) { MBadPixelsPix &bad = (*fBadPixels)[i]; MCalibrationPix &pix = (*fCam)[i]; if (bad.IsUncalibrated( MBadPixelsPix::kChargeIsPedestal)) bad.SetUnsuitable( MBadPixelsPix::kUnsuitableRun ); if (bad.IsUncalibrated( MBadPixelsPix::kChargeErrNotValid )) bad.SetUnsuitable( MBadPixelsPix::kUnsuitableRun ); if (bad.IsUncalibrated( MBadPixelsPix::kChargeRelErrNotValid )) bad.SetUnsuitable( MBadPixelsPix::kUnsuitableRun ); if (bad.IsUncalibrated( MBadPixelsPix::kMeanTimeInFirstBin )) bad.SetUnsuitable( MBadPixelsPix::kUnsuitableRun ); if (bad.IsUncalibrated( MBadPixelsPix::kMeanTimeInLast2Bins )) bad.SetUnsuitable( MBadPixelsPix::kUnsuitableRun ); if (bad.IsUncalibrated( MBadPixelsPix::kDeviatingNumPhes )) bad.SetUnsuitable( MBadPixelsPix::kUnsuitableRun ); if (bad.IsUncalibrated( MBadPixelsPix::kHiGainOverFlow )) bad.SetUnsuitable( MBadPixelsPix::kUnsuitableRun ); if (bad.IsUncalibrated( MBadPixelsPix::kLoGainOverFlow )) bad.SetUnsuitable( MBadPixelsPix::kUnsuitableRun ); if (bad.IsUnsuitable( MBadPixelsPix::kUnsuitableRun )) pix.SetExcluded(); if (bad.IsUncalibrated( MBadPixelsPix::kChargeSigmaNotValid )) bad.SetUnsuitable( MBadPixelsPix::kUnreliableRun ); } } // ------------------------------------------------------------------------ // // // First loop: Calculate a mean and mean RMS of photo-electrons per area index // Include only pixels which are not MBadPixelsPix::kUnsuitableRun nor // MBadPixelsPix::kChargeSigmaNotValid (see FinalizeBadPixels()) and set // MCalibrationChargePix::SetFFactorMethodValid(kFALSE) in that case. // // Second loop: Get mean number of photo-electrons and its RMS including // only pixels with flag MCalibrationChargePix::IsFFactorMethodValid() // and further exclude those deviating by more than fPheErrLimit mean // sigmas from the mean (obtained in first loop). Set // MBadPixelsPix::kDeviatingNumPhes if excluded. // // For the suitable pixels with flag MBadPixelsPix::kChargeSigmaNotValid // set the number of photo-electrons as the mean number of photo-electrons // calculated in that area index. // // Set weighted mean and variance of photo-electrons per area index in: // average area pixels of MCalibrationChargeCam (obtained from: // MCalibrationChargeCam::GetAverageArea() ) // // Set weighted mean and variance of photo-electrons per sector in: // average sector pixels of MCalibrationChargeCam (obtained from: // MCalibrationChargeCam::GetAverageSector() ) // // // Third loop: Set mean number of photo-electrons and its RMS in the pixels // only excluded as: MBadPixelsPix::kChargeSigmaNotValid // Bool_t MCalibrationChargeCalc::FinalizeFFactorMethod() { const UInt_t npixels = fGeom->GetNumPixels(); const UInt_t nareas = fGeom->GetNumAreas(); const UInt_t nsectors = fGeom->GetNumSectors(); TArrayF lowlim (nareas); TArrayF upplim (nareas); TArrayD areavars (nareas); TArrayD areaweights (nareas); TArrayD sectorweights (nsectors); TArrayD areaphes (nareas); TArrayD sectorphes (nsectors); TArrayI numareavalid (nareas); TArrayI numsectorvalid(nsectors); // // First loop: Get mean number of photo-electrons and the RMS // The loop is only to recognize later pixels with very deviating numbers // MHCamera camphes(*fGeom,"Camphes","Phes in Camera"); for (UInt_t i=0; iFit("gaus","Q"); const Float_t mean = hist->GetFunction("gaus")->GetParameter(1); const Float_t sigma = hist->GetFunction("gaus")->GetParameter(2); const Int_t ndf = hist->GetFunction("gaus")->GetNDF(); if (IsDebug()) camphes.DrawClone(); if (ndf < 2) { *fLog << warn << GetDescriptor() << ": Cannot use a Gauss fit to the number of photo-electrons " << "in the camera with area index: " << i << endl; *fLog << warn << GetDescriptor() << ": Number of dof.: " << ndf << " is smaller than 2 " << endl; *fLog << warn << GetDescriptor() << ": Will use the simple mean and rms " << endl; delete hist; continue; } const Double_t prob = hist->GetFunction("gaus")->GetProb(); if (prob < 0.001) { *fLog << warn << GetDescriptor() << ": Cannot use a Gauss fit to the number of photo-electrons " << "in the camera with area index: " << i << endl; *fLog << warn << GetDescriptor() << ": Fit probability " << prob << " is smaller than 0.001 " << endl; *fLog << warn << GetDescriptor() << ": Will use the simple mean and rms " << endl; delete hist; continue; } *fLog << inf << GetDescriptor() << ": Mean number of photo-electrons " << "with area idx " << i << ": " << Form("%7.2f+-%6.2f",mean,sigma) << endl; lowlim [i] = mean - fPheErrLimit*sigma; upplim [i] = mean + fPheErrLimit*sigma; delete hist; } *fLog << endl; numareavalid.Reset(); areaphes .Reset(); areavars .Reset(); // // Second loop: Get mean number of photo-electrons and its RMS excluding // pixels deviating by more than fPheErrLimit sigma. // Set the conversion factor FADC counts to photo-electrons // for (UInt_t i=0; i upplim[aidx] ) { *fLog << warn << GetDescriptor() << ": Number of phes: " << Form("%7.2f out of %3.1f sigma limit: ",nphe,fPheErrLimit) << Form("[%7.2f,%7.2f] pixel%4i",lowlim[aidx],upplim[aidx],i) << endl; bad.SetUncalibrated( MBadPixelsPix::kDeviatingNumPhes ); bad.SetUnsuitable ( MBadPixelsPix::kUnsuitableRun ); pix.SetFFactorMethodValid(kFALSE); continue; } areaweights [aidx] += nphe*nphe; areaphes [aidx] += nphe; numareavalid [aidx] ++; if (aidx == 0) fNumInnerFFactorMethodUsed++; sectorweights [sector] += nphe*nphe/area/area; sectorphes [sector] += nphe/area; numsectorvalid[sector] ++; } *fLog << endl; for (UInt_t aidx=0; aidxGetAverageArea(aidx); if (numareavalid[aidx] == 1) areaweights[aidx] = 0.; else if (numareavalid[aidx] == 0) { areaphes[aidx] = -1.; areaweights[aidx] = -1.; } else { areaweights[aidx] = (areaweights[aidx] - areaphes[aidx]*areaphes[aidx]/numareavalid[aidx]) / (numareavalid[aidx]-1); areaphes[aidx] /= numareavalid[aidx]; } if (areaweights[aidx] < 0. || areaphes[aidx] <= 0.) { *fLog << warn << GetDescriptor() << ": Mean number phes from area index " << aidx << " could not be calculated: " << " Mean: " << areaphes[aidx] << " Variance: " << areaweights[aidx] << endl; apix.SetFFactorMethodValid(kFALSE); continue; } *fLog << inf << GetDescriptor() << ": Average total number phes in area idx " << aidx << ": " << Form("%7.2f%s%6.2f",areaphes[aidx]," +- ",TMath::Sqrt(areaweights[aidx])) << endl; apix.SetPheFFactorMethod ( areaphes[aidx] ); apix.SetPheFFactorMethodVar( areaweights[aidx] / numareavalid[aidx] ); apix.SetFFactorMethodValid ( kTRUE ); } *fLog << endl; for (UInt_t sector=0; sectorGetAverageSector(sector); if (sectorweights[sector] < 0. || sectorphes[sector] <= 0.) { *fLog << warn << GetDescriptor() <<": Mean number phes per area for sector " << sector << " could not be calculated: " << " Mean: " << sectorphes[sector] << " Variance: " << sectorweights[sector] << endl; spix.SetFFactorMethodValid(kFALSE); continue; } *fLog << inf << GetDescriptor() << ": Average number phes per area in sector " << sector << ": " << Form("%5.2f+-%4.2f [phe/mm^2]",sectorphes[sector],TMath::Sqrt(sectorweights[sector])) << endl; spix.SetPheFFactorMethod ( sectorphes[sector] ); spix.SetPheFFactorMethodVar( sectorweights[sector] / numsectorvalid[sector]); spix.SetFFactorMethodValid ( kTRUE ); } // // Third loop: Set mean number of photo-electrons and its RMS in the pixels // only excluded as: MBadPixelsPix::kChargeSigmaNotValid // for (UInt_t i=0; iGetAverageArea(aidx); pix.SetPheFFactorMethod ( apix.GetPheFFactorMethod() ); pix.SetPheFFactorMethodVar( apix.GetPheFFactorMethodVar() ); if (!pix.CalcConvFFactor()) { *fLog << warn << GetDescriptor() << ": Could not calculate the Conv. FADC counts to Phes in pixel: " << Form(" %4i",pix.GetPixId()) << endl; bad.SetUncalibrated( MBadPixelsPix::kDeviatingNumPhes ); bad.SetUnsuitable ( MBadPixelsPix::kUnsuitableRun ); } } } return kTRUE; } // ------------------------------------------------------------------------ // // Returns kFALSE if pointer to MCalibrationChargeBlindPix is NULL // // The check returns kFALSE if: // // 1) fLambda and fLambdaCheck are separated relatively to each other by more than fLambdaCheckLimit // 2) BlindPixel has an fLambdaErr greater than fLambdaErrLimit // // Calls: // - MCalibrationChargeBlindPix::CalcFluxInsidePlexiglass() // Bool_t MCalibrationChargeCalc::FinalizeBlindPixel() { if (!fBlindPixel) return kFALSE; const Float_t lambda = fBlindPixel->GetLambda(); const Float_t lambdaerr = fBlindPixel->GetLambdaErr(); const Float_t lambdacheck = fBlindPixel->GetLambdaCheck(); if (2.*(lambdacheck-lambda)/(lambdacheck+lambda) > fLambdaCheckLimit) { *fLog << warn << GetDescriptor() << Form("%s%4.2f%s%4.2f%s%4.2f%s",": Lambda: ",lambda," and Lambda-Check: ", lambdacheck," differ by more than ",fLambdaCheckLimit," in the Blind Pixel ") << endl; return kFALSE; } if (lambdaerr > fLambdaErrLimit) { *fLog << warn << GetDescriptor() << Form("%s%4.2f%s%4.2f%s",": Error of Fitted Lambda: ",lambdaerr," is greater than ", fLambdaErrLimit," in Blind Pixel ") << endl; return kFALSE; } if (!fBlindPixel->CalcFluxInsidePlexiglass()) { *fLog << warn << "Could not calculate the flux of photons from the Blind Pixel, " << "will skip Blind Pixel Calibration " << endl; return kFALSE; } return kTRUE; } // ------------------------------------------------------------------------ // // Returns kFALSE if pointer to MCalibrationChargeBlindCam is NULL // // The check returns kFALSE if: // // 1) fLambda and fLambdaCheck are separated relatively to each other by more than fLambdaCheckLimit // 2) BlindPixel has an fLambdaErr greater than fLambdaErrLimit // // Calls: // - MCalibrationChargeBlindPix::CalcFluxInsidePlexiglass() // Bool_t MCalibrationChargeCalc::FinalizeBlindCam() { if (!fBlindCam) return kFALSE; Float_t flux = 0.; Float_t fluxvar = 0.; Int_t nvalid = 0; for (UInt_t i=0; iGetNumBlindPixels(); i++) { MCalibrationChargeBlindPix &blindpix = (*fBlindCam)[i]; if (!blindpix.IsValid()) continue; const Float_t lambda = blindpix.GetLambda(); const Float_t lambdaerr = blindpix.GetLambdaErr(); const Float_t lambdacheck = blindpix.GetLambdaCheck(); if (2.*(lambdacheck-lambda)/(lambdacheck+lambda) > fLambdaCheckLimit) { *fLog << warn << GetDescriptor() << Form("%s%4.2f%s%4.2f%s%4.2f%s%2i",": Lambda: ",lambda," and Lambda-Check: ", lambdacheck," differ by more than ",fLambdaCheckLimit," in the Blind Pixel Nr.",i) << endl; blindpix.SetValid(kFALSE); continue; } if (lambdaerr > fLambdaErrLimit) { *fLog << warn << GetDescriptor() << Form("%s%4.2f%s%4.2f%s%2i",": Error of Fitted Lambda: ",lambdaerr," is greater than ", fLambdaErrLimit," in Blind Pixel Nr.",i) << endl; blindpix.SetValid(kFALSE); continue; } if (!blindpix.CalcFluxInsidePlexiglass()) { *fLog << warn << "Could not calculate the flux of photons from Blind Pixel Nr." << i << endl; blindpix.SetValid(kFALSE); continue; } nvalid++; const Float_t weight = 1./ blindpix.GetFluxInsidePlexiglassErr() / blindpix.GetFluxInsidePlexiglassErr(); flux += weight * blindpix.GetFluxInsidePlexiglass(); fluxvar += weight; } if (!nvalid) return kFALSE; flux /= fluxvar; fluxvar /= 1./fluxvar; const Float_t photons = flux * (*fGeom)[0].GetA() / fQECam->GetPlexiglassQE(); fCam->SetNumPhotonsBlindPixelMethod(photons); const Float_t photrelvar = fluxvar / flux / flux + fQECam->GetPlexiglassQERelVar(); if (photrelvar > 0.) fCam->SetNumPhotonsBlindPixelMethodErr(TMath::Sqrt( photrelvar * photons * photons)); return kTRUE; } // ------------------------------------------------------------------------ // // Returns kFALSE if pointer to MCalibrationChargePINDiode is NULL // // The check returns kFALSE if: // // 1) PINDiode has a fitted charge smaller than fChargeLimit*PedRMS // 2) PINDiode has a fit error smaller than fChargeErrLimit // 3) PINDiode has a fitted charge smaller its fChargeRelErrLimit times its charge error // 4) PINDiode has a charge sigma smaller than its Pedestal RMS // // Calls: // - MCalibrationChargePINDiode::CalcFluxOutsidePlexiglass() // Bool_t MCalibrationChargeCalc::FinalizePINDiode() { if (!fPINDiode) return kFALSE; if (fPINDiode->GetMean() < fChargeLimit*fPINDiode->GetPedRms()) { *fLog << warn << GetDescriptor() << ": Fitted Charge is smaller than " << fChargeLimit << " Pedestal RMS in PINDiode " << endl; return kFALSE; } if (fPINDiode->GetMeanErr() < fChargeErrLimit) { *fLog << warn << GetDescriptor() << ": Error of Fitted Charge is smaller than " << fChargeErrLimit << " in PINDiode " << endl; return kFALSE; } if (fPINDiode->GetMean() < fChargeRelErrLimit*fPINDiode->GetMeanErr()) { *fLog << warn << GetDescriptor() << ": Fitted Charge is smaller than " << fChargeRelErrLimit << "* its error in PINDiode " << endl; return kFALSE; } if (fPINDiode->GetSigma() < fPINDiode->GetPedRms()) { *fLog << warn << GetDescriptor() << ": Sigma of Fitted Charge smaller than Pedestal RMS in PINDiode " << endl; return kFALSE; } if (!fPINDiode->CalcFluxOutsidePlexiglass()) { *fLog << warn << "Could not calculate the flux of photons from the PIN Diode, " << "will skip PIN Diode Calibration " << endl; return kFALSE; } return kTRUE; } // ------------------------------------------------------------------------ // // Calculate the average number of photons outside the plexiglass with the // formula: // // av.Num.photons(area index) = av.Num.Phes(area index) // / MCalibrationQEPix::GetDefaultQE(fPulserColor) // / MCalibrationQEPix::GetPMTCollectionEff() // / MCalibrationQEPix::GetLightGuidesEff(fPulserColor) // / MCalibrationQECam::GetPlexiglassQE() // // Calculate the variance on the average number of photons assuming that the error on the // Quantum efficiency is reduced by the number of used inner pixels, but the rest of the // values keeps it ordinary error since it is systematic. // // Loop over pixels: // // - Continue, if not MCalibrationChargePix::IsFFactorMethodValid() and set: // MCalibrationQEPix::SetFFactorMethodValid(kFALSE,fPulserColor) // // - Call MCalibrationChargePix::CalcMeanFFactor(av.Num.photons) and set: // MCalibrationQEPix::SetFFactorMethodValid(kFALSE,fPulserColor) if not succesful // // - Calculate the quantum efficiency with the formula: // // QE = ( Num.Phes / av.Num.photons ) * MGeomCam::GetPixRatio() // // - Set QE in MCalibrationQEPix::SetQEFFactor ( QE, fPulserColor ); // // - Set Variance of QE in MCalibrationQEPix::SetQEFFactorVar ( Variance, fPulserColor ); // - Set bit MCalibrationQEPix::SetFFactorMethodValid(kTRUE,fPulserColor) // // - Call MCalibrationQEPix::UpdateFFactorMethod() // void MCalibrationChargeCalc::FinalizeFFactorQECam() { if (fNumInnerFFactorMethodUsed < 2) { *fLog << warn << GetDescriptor() << ": Could not calculate F-Factor Method: Less than 2 inner pixels valid! " << endl; return; } MCalibrationChargePix &avpix = (MCalibrationChargePix&)fCam->GetAverageArea(0); MCalibrationQEPix &qepix = (MCalibrationQEPix&) fQECam->GetAverageArea(0); const Float_t avphotons = avpix.GetPheFFactorMethod() / qepix.GetDefaultQE(fPulserColor) / qepix.GetPMTCollectionEff() / qepix.GetLightGuidesEff(fPulserColor) / fQECam->GetPlexiglassQE(); const Float_t avphotrelvar = avpix.GetPheFFactorMethodRelVar() + qepix.GetDefaultQERelVar(fPulserColor) / fNumInnerFFactorMethodUsed + qepix.GetPMTCollectionEffRelVar() + qepix.GetLightGuidesEffRelVar(fPulserColor) + fQECam->GetPlexiglassQERelVar(); const UInt_t nareas = fGeom->GetNumAreas(); // // Set the results in the MCalibrationChargeCam // fCam->SetNumPhotonsFFactorMethod (avphotons); if (avphotrelvar > 0.) fCam->SetNumPhotonsFFactorMethodErr(TMath::Sqrt( avphotrelvar * avphotons * avphotons)); TArrayF lowlim (nareas); TArrayF upplim (nareas); TArrayD avffactorphotons (nareas); TArrayD avffactorphotvar (nareas); TArrayI numffactor (nareas); const UInt_t npixels = fGeom->GetNumPixels(); MHCamera camffactor(*fGeom,"Camffactor","F-Factor in Camera"); for (UInt_t i=0; iGetPixRatio(i); const Float_t qe = pix.GetPheFFactorMethod() / photons ; const Float_t qerelvar = avphotrelvar + pix.GetPheFFactorMethodRelVar(); qepix.SetQEFFactor ( qe , fPulserColor ); qepix.SetQEFFactorVar ( qerelvar*qe*qe, fPulserColor ); qepix.SetFFactorMethodValid( kTRUE , fPulserColor ); if (!qepix.UpdateFFactorMethod()) *fLog << warn << GetDescriptor() << ": Cannot update Quantum efficiencies with the F-Factor Method" << endl; // // The following pixels are those with deviating sigma, but otherwise OK, // probably those with stars during the pedestal run, but not the cal. run. // if (!pix.CalcMeanFFactor( photons , avphotrelvar )) { (*fBadPixels)[i].SetUncalibrated( MBadPixelsPix::kDeviatingFFactor ); (*fBadPixels)[i].SetUnsuitable ( MBadPixelsPix::kUnreliableRun ); continue; } const Int_t aidx = (*fGeom)[i].GetAidx(); const Float_t ffactor = pix.GetMeanFFactorFADC2Phot(); camffactor.Fill(i,ffactor); camffactor.SetUsed(i); avffactorphotons[aidx] += ffactor; avffactorphotvar[aidx] += ffactor*ffactor; numffactor[aidx]++; } for (UInt_t i=0; iFit("gaus","Q"); const Float_t mean = hist->GetFunction("gaus")->GetParameter(1); const Float_t sigma = hist->GetFunction("gaus")->GetParameter(2); const Int_t ndf = hist->GetFunction("gaus")->GetNDF(); if (IsDebug()) camffactor.DrawClone(); if (ndf < 2) { *fLog << warn << GetDescriptor() << ": Cannot use a Gauss fit to the F-Factor " << "in the camera with area index: " << i << endl; *fLog << warn << GetDescriptor() << ": Number of dof.: " << ndf << " is smaller than 2 " << endl; *fLog << warn << GetDescriptor() << ": Will use the simple mean and rms " << endl; delete hist; continue; } const Double_t prob = hist->GetFunction("gaus")->GetProb(); if (prob < 0.001) { *fLog << warn << GetDescriptor() << ": Cannot use a Gauss fit to the F-Factor " << "in the camera with area index: " << i << endl; *fLog << warn << GetDescriptor() << ": Fit probability " << prob << " is smaller than 0.001 " << endl; *fLog << warn << GetDescriptor() << ": Will use the simple mean and rms " << endl; delete hist; continue; } *fLog << inf << GetDescriptor() << ": Mean F-Factor " << "with area index #" << i << ": " << Form("%4.2f+-%4.2f",mean,sigma) << endl; lowlim [i] = 1.1; upplim [i] = mean + fFFactorErrLimit*sigma; delete hist; } *fLog << endl; for (UInt_t i=0; i upplim[aidx] ) { *fLog << warn << GetDescriptor() << ": Overall F-Factor " << Form("%5.2f",ffactor) << " out of range [" << Form("%5.2f,%5.2f",lowlim[aidx],upplim[aidx]) << "] pixel " << i << endl; bad.SetUncalibrated( MBadPixelsPix::kDeviatingFFactor ); bad.SetUnsuitable ( MBadPixelsPix::kUnsuitableRun ); } } for (UInt_t i=0; iGetNumPixels(); // // Set the results in the MCalibrationChargeCam // if (fBlindPixel) { if (fBlindPixel->IsFluxInsidePlexiglassAvailable()) { const Float_t photons = fBlindPixel->GetFluxInsidePlexiglass() * (*fGeom)[0].GetA() / fQECam->GetPlexiglassQE(); fCam->SetNumPhotonsBlindPixelMethod(photons); const Float_t photrelvar = fBlindPixel->GetFluxInsidePlexiglassRelVar() + fQECam->GetPlexiglassQERelVar(); if (photrelvar > 0.) fCam->SetNumPhotonsBlindPixelMethodErr(TMath::Sqrt( photrelvar * photons * photons)); } } // // With the knowledge of the overall photon flux, calculate the // quantum efficiencies after the Blind Pixel and PIN Diode method // for (UInt_t i=0; iIsFluxInsidePlexiglassAvailable()) { qepix.SetBlindPixelMethodValid(kFALSE, fPulserColor); continue; } MBadPixelsPix &bad = (*fBadPixels)[i]; if (bad.IsUnsuitable (MBadPixelsPix::kUnsuitableRun)) { qepix.SetBlindPixelMethodValid(kFALSE, fPulserColor); continue; } MCalibrationChargePix &pix = (MCalibrationChargePix&)(*fCam)[i]; MGeomPix &geo = (*fGeom)[i]; const Float_t qe = pix.GetPheFFactorMethod() / fBlindPixel->GetFluxInsidePlexiglass() / geo.GetA() * fQECam->GetPlexiglassQE(); const Float_t qerelvar = fBlindPixel->GetFluxInsidePlexiglassRelVar() + fQECam->GetPlexiglassQERelVar() + pix.GetPheFFactorMethodRelVar(); qepix.SetQEBlindPixel ( qe , fPulserColor ); qepix.SetQEBlindPixelVar ( qerelvar*qe*qe, fPulserColor ); qepix.SetBlindPixelMethodValid( kTRUE , fPulserColor ); if (!qepix.UpdateBlindPixelMethod()) *fLog << warn << GetDescriptor() << ": Cannot update Quantum efficiencies with the Blind Pixel Method" << endl; } } // ------------------------------------------------------------------------ // // Loop over pixels: // // - Continue, if not MCalibrationChargePINDiode::IsFluxOutsidePlexiglassAvailable() and set: // MCalibrationQEPix::SetPINDiodeMethodValid(kFALSE,fPulserColor) // // - Calculate the quantum efficiency with the formula: // // QE = Num.Phes / MCalibrationChargePINDiode::GetFluxOutsidePlexiglass() / MGeomPix::GetA() // // - Set QE in MCalibrationQEPix::SetQEPINDiode ( QE, fPulserColor ); // - Set Variance of QE in MCalibrationQEPix::SetQEPINDiodeVar ( Variance, fPulserColor ); // - Set bit MCalibrationQEPix::SetPINDiodeMethodValid(kTRUE,fPulserColor) // // - Call MCalibrationQEPix::UpdatePINDiodeMethod() // void MCalibrationChargeCalc::FinalizePINDiodeQECam() { const UInt_t npixels = fGeom->GetNumPixels(); // // With the knowledge of the overall photon flux, calculate the // quantum efficiencies after the PIN Diode method // for (UInt_t i=0; iIsFluxOutsidePlexiglassAvailable()) { qepix.SetPINDiodeMethodValid(kFALSE, fPulserColor); continue; } MBadPixelsPix &bad = (*fBadPixels)[i]; if (!bad.IsUnsuitable (MBadPixelsPix::kUnsuitableRun)) { qepix.SetPINDiodeMethodValid(kFALSE, fPulserColor); continue; } MCalibrationChargePix &pix = (MCalibrationChargePix&)(*fCam)[i]; MGeomPix &geo = (*fGeom)[i]; const Float_t qe = pix.GetPheFFactorMethod() / fPINDiode->GetFluxOutsidePlexiglass() / geo.GetA(); const Float_t qerelvar = fPINDiode->GetFluxOutsidePlexiglassRelVar() + pix.GetPheFFactorMethodRelVar(); qepix.SetQEPINDiode ( qe , fPulserColor ); qepix.SetQEPINDiodeVar ( qerelvar*qe*qe, fPulserColor ); qepix.SetPINDiodeMethodValid( kTRUE , fPulserColor ); if (!qepix.UpdatePINDiodeMethod()) *fLog << warn << GetDescriptor() << ": Cannot update Quantum efficiencies with the PIN Diode Method" << endl; } } // ----------------------------------------------------------------------------------------------- // // - Print out statistics about BadPixels of type UnsuitableType_t // - store numbers of bad pixels of each type in fCam // void MCalibrationChargeCalc::FinalizeUnsuitablePixels() { *fLog << inf << endl; *fLog << GetDescriptor() << ": Charge Calibration status:" << endl; *fLog << dec << setfill(' '); const Int_t nareas = fGeom->GetNumAreas(); TArrayI counts(nareas); for (Int_t i=0; iGetSize(); i++) { MBadPixelsPix &bad = (*fBadPixels)[i]; if (!bad.IsBad()) { const Int_t aidx = (*fGeom)[i].GetAidx(); counts[aidx]++; } } if (fGeom->InheritsFrom("MGeomCamMagic")) *fLog << " " << setw(7) << "Successfully calibrated Pixels: " << Form("%s%3i%s%3i","Inner: ",counts[0]," Outer: ",counts[1]) << endl; counts.Reset(); for (Int_t i=0; iGetSize(); i++) { MBadPixelsPix &bad = (*fBadPixels)[i]; if (bad.IsUnsuitable(MBadPixelsPix::kUnsuitableRun)) { const Int_t aidx = (*fGeom)[i].GetAidx(); counts[aidx]++; } } for (Int_t aidx=0; aidxSetNumUnsuitable(counts[aidx], aidx); if (fGeom->InheritsFrom("MGeomCamMagic")) *fLog << " " << setw(7) << "Uncalibrated Pixels: " << Form("%s%3i%s%3i","Inner: ",counts[0]," Outer: ",counts[1]) << endl; counts.Reset(); for (Int_t i=0; iGetSize(); i++) { MBadPixelsPix &bad = (*fBadPixels)[i]; if (bad.IsUnsuitable(MBadPixelsPix::kUnreliableRun)) { const Int_t aidx = (*fGeom)[i].GetAidx(); counts[aidx]++; } } for (Int_t aidx=0; aidxSetNumUnreliable(counts[aidx], aidx); *fLog << " " << setw(7) << "Unreliable Pixels: " << Form("%s%3i%s%3i","Inner: ",counts[0]," Outer: ",counts[1]) << endl; } // ----------------------------------------------------------------------------------------------- // // Print out statistics about BadPixels of type UncalibratedType_t // void MCalibrationChargeCalc::PrintUncalibrated(MBadPixelsPix::UncalibratedType_t typ, const char *text) const { UInt_t countinner = 0; UInt_t countouter = 0; for (Int_t i=0; iGetSize(); i++) { MBadPixelsPix &bad = (*fBadPixels)[i]; if (bad.IsUncalibrated(typ)) { if (fGeom->GetPixRatio(i) == 1.) countinner++; else countouter++; } } *fLog << " " << setw(7) << text << Form("%s%3i%s%3i","Inner: ",countinner," Outer: ",countouter) << endl; } // -------------------------------------------------------------------------- // // Set the path for output file // void MCalibrationChargeCalc::SetOutputPath(TString path) { fOutputPath = path; if (fOutputPath.EndsWith("/")) fOutputPath = fOutputPath(0, fOutputPath.Length()-1); } // -------------------------------------------------------------------------- // // Set the output file // void MCalibrationChargeCalc::SetOutputFile(TString file) { fOutputFile = file; } // -------------------------------------------------------------------------- // // Get the output file // const char* MCalibrationChargeCalc::GetOutputFile() { return Form("%s/%s", (const char*)fOutputPath, (const char*)fOutputFile); } // -------------------------------------------------------------------------- // // Read the environment for the following data members: // - fChargeLimit // - fChargeErrLimit // - fChargeRelErrLimit // - fDebug // - fFFactorErrLimit // - fLambdaErrLimit // - fLambdaCheckErrLimit // - fPheErrLimit // Int_t MCalibrationChargeCalc::ReadEnv(const TEnv &env, TString prefix, Bool_t print) { Bool_t rc = kFALSE; if (IsEnvDefined(env, prefix, "ChargeLimit", print)) { SetChargeLimit(GetEnvValue(env, prefix, "ChargeLimit", fChargeLimit)); rc = kTRUE; } if (IsEnvDefined(env, prefix, "ChargeErrLimit", print)) { SetChargeErrLimit(GetEnvValue(env, prefix, "ChargeErrLimit", fChargeErrLimit)); rc = kTRUE; } if (IsEnvDefined(env, prefix, "ChargeRelErrLimit", print)) { SetChargeRelErrLimit(GetEnvValue(env, prefix, "ChargeRelErrLimit", fChargeRelErrLimit)); rc = kTRUE; } if (IsEnvDefined(env, prefix, "Debug", print)) { SetDebug(GetEnvValue(env, prefix, "Debug", IsDebug())); rc = kTRUE; } if (IsEnvDefined(env, prefix, "FFactorErrLimit", print)) { SetFFactorErrLimit(GetEnvValue(env, prefix, "FFactorErrLimit", fFFactorErrLimit)); rc = kTRUE; } if (IsEnvDefined(env, prefix, "LambdaErrLimit", print)) { SetLambdaErrLimit(GetEnvValue(env, prefix, "LambdaErrLimit", fLambdaErrLimit)); rc = kTRUE; } if (IsEnvDefined(env, prefix, "LambdaCheckLimit", print)) { SetLambdaCheckLimit(GetEnvValue(env, prefix, "LambdaCheckLimit", fLambdaCheckLimit)); rc = kTRUE; } if (IsEnvDefined(env, prefix, "PheErrLimit", print)) { SetPheErrLimit(GetEnvValue(env, prefix, "PheErrLimit", fPheErrLimit)); rc = kTRUE; } // void SetPulserColor(const MCalibrationCam::PulserColor_t col) { fPulserColor = col; } return rc; }