class AliTPCCalibPedestal: public AliTPCCalibRawBase

          Implementation of the TPC pedestal and noise calibration

   Origin: Jens Wiechula, Marian Ivanov   J.Wiechula@gsi.de, Marian.Ivanov@cern.ch



 *                                Class Description                                  *


 Working principle:

 Raw pedestal data is processed by calling one of the ProcessEvent(...) functions
 (see below). These in the end call the Update(...) function, where the data is filled
 into histograms.

 For each ROC one TH2F histo (ROC channel vs. ADC channel) is created when
 it is filled for the first time (GetHistoPedestal(ROC,kTRUE)). All histos are stored in the
 TObjArray fHistoPedestalArray.

 For a fast filling of the histogram the corresponding bin number of the channel and ADC channel
 is computed by hand and the histogram array is accessed directly via its pointer.
 ATTENTION: Doing so the the entry counter of the histogram is not increased
            this means that e.g. the colz draw option gives an empty plot unless
	    calling 'histo->SetEntries(1)' before drawing.

 After accumulating the desired statistics the Analyse() function has to be called.
 Whithin this function the pedestal and noise values are calculated for each pad, using
 the fast gaus fit function  AliMathBase::FitGaus(...), and the calibration
 storage classes (AliTPCCalROC) are filled for each ROC.
 The calibration information is stored in the TObjArrays fCalRocArrayPedestal and fCalRocArrayRMS;



 User interface for filling data:


 To Fill information one of the following functions can be used:

 Bool_t ProcessEvent(eventHeaderStruct *event);
   - process Date event
   - use AliTPCRawReaderDate and call ProcessEvent(AliRawReader *rawReader)

 Bool_t ProcessEvent(AliRawReader *rawReader);
  - process AliRawReader event
   - use AliTPCRawStreamV3 to loop over data and call ProcessEvent(AliTPCRawStreamV3 *rawStream)

 Bool_t ProcessEvent(AliTPCRawStreamV3 *rawStream);
   - process event from AliTPCRawStreamV3
   - call Update function for signal filling

 Int_t Update(const Int_t isector, const Int_t iRow, const Int_t
              iPad,  const Int_t iTimeBin, const Float_t signal);
   - directly  fill signal information (sector, row, pad, time bin, pad)
     to the reference histograms

 It is also possible to merge two independently taken calibrations using the function

 void Merge(AliTPCCalibPedestal *ped)
   - copy histograms in 'ped' if the do not exist in this instance
   - Add histograms in 'ped' to the histograms in this instance if the allready exist
   - After merging call Analyse again!



 -- example: filling data using root raw data:
 void fillPedestal(Char_t *filename)
 {
    rawReader = new AliRawReaderRoot(fileName);
    if ( !rawReader ) return;
    AliTPCCalibPedestal *calib = new AliTPCCalibPedestal;
    while (rawReader->NextEvent()){
      calib->ProcessEvent(rawReader);
    }
    calib->Analyse();
    calib->DumpToFile("PedestalData.root");
    delete rawReader;
    delete calib;
 }


 What kind of information is stored and how to retrieve them:


 - Accessing the 'Reference Histograms' (pedestal distribution histograms):

 TH2F *GetHistoPedestal(Int_t sector);

 - Accessing the calibration storage objects:

 AliTPCCalROC *GetCalRocPedestal(Int_t sector);  - for the pedestal values, mean from gaus fit
 AliTPCCalROC *GetCalRocSigma(Int_t sector);     - for the Noise values, sigma from guas fit
 AliTPCCalROC *GetCalRocMean(Int_t sector);  - for the pedestal values, truncated mean
 AliTPCCalROC *GetCalRocRMS(Int_t sector);     - for the Noise values, rms from truncated mean

 example for visualisation:
 if the file "PedestalData.root" was created using the above example one could do the following:

 TFile filePedestal("PedestalData.root")
 AliTPCCalibPedestal *ped = (AliTPCCalibPedestal*)filePedestal->Get("AliTPCCalibPedestal");
 ped->GetCalRocPedestal(0)->Draw("colz");
 ped->GetCalRocRMS(0)->Draw("colz");

 or use the AliTPCCalPad functionality:
 AliTPCCalPad padPedestal(ped->GetCalPadPedestal());
 AliTPCCalPad padNoise(ped->GetCalPadRMS());
 padPedestal->MakeHisto2D()->Draw("colz");  //Draw A-Side Pedestal Information
 padNoise->MakeHisto2D()->Draw("colz");  //Draw A-Side Noise Information


 example: fill pedestal with gausschen noise
 AliTPCCalibPedestal ped;
 ped.TestEvent();
 ped.Analyse();
 //Draw output;
 TCanvas* c1 = new TCanvas;
 c1->Divide(1,2);
 c1->cd(1);
 ped.GetHistoPedestal(0)->SetEntries(1); //needed in order for colz to work, reason: fast filling does not increase the entries counter
 ped.GetHistoPedestal(0)->Draw("colz");
 c1->cd(2);
 ped.GetHistoPedestal(36)->SetEntries(1); //needed in order for colz to work, reason: fast filling does not increase the entries counter
 ped.GetHistoPedestal(36)->Draw("colz");
 TCanvas* c2 = new TCanvas;
 c2->Divide(2,2);
 c2->cd(1);
 ped.GetCalRocPedestal(0)->Draw("colz");
 c2->cd(2);
 ped.GetCalRocRMS(0)->Draw("colz");
 c2->cd(3);
 ped.GetCalRocPedestal(36)->Draw("colz");
 c2->cd(4);
 ped.GetCalRocRMS(36)->Draw("colz");


 Time dependent pedestals:

 If wished there is the possibility to calculate for each channel and time bin
 the mean pedestal [pedestals(t)]. This is done by

 1) setting SetTimeAnalysis(kTRUE),
 2) processing the data by looping over the events using ProcessEvent(..)
 3) calling the Analyse() and AnalyseTime(nevents) functions (providing nevents)
 4) getting the pedestals(t) using   TArrayF **timePed = calibPedestal.GetTimePedestals();
 5) looking at values using   timePed[row][pad].At(timebin)

 This functionality is intended to be used on an LDC bu the detector algorithm
 (TPCPEDESTALda) to generate a data set used for configuration of the pattern
 memory for baseline subtraction in the ALTROs. Later the information should also
 be stored as reference data.