#include <HBHEPulseShapeFlag.h>

Public Member Functions
void	Clear ()

	HBHEPulseShapeFlagSetter ()

	HBHEPulseShapeFlagSetter (double MinimumChargeThreshold, double TS4TS5ChargeThreshold, unsigned int TrianglePeakTS, const std::vector< double > &LinearThreshold, const std::vector< double > &LinearCut, const std::vector< double > &RMS8MaxThreshold, const std::vector< double > &RMS8MaxCut, const std::vector< double > &LeftSlopeThreshold, const std::vector< double > &LeftSlopeCut, const std::vector< double > &RightSlopeThreshold, const std::vector< double > &RightSlopeCut, const std::vector< double > &RightSlopeSmallThreshold, const std::vector< double > &RightSlopeSmallCut, const std::vector< double > &TS4TS5UpperThreshold, const std::vector< double > &TS4TS5UpperCut, const std::vector< double > &TS4TS5LowerThreshold, const std::vector< double > &TS4TS5LowerCut, bool UseDualFit, bool TriangleIgnoreSlow)

void	Initialize ()

void	SetPulseShapeFlags (HBHERecHit &hbhe, const HBHEDataFrame &digi, const HcalCoder &coder, const HcalCalibrations &calib)

	~HBHEPulseShapeFlagSetter ()

Private Member Functions
double	CalculateRMS8Max (const std::vector< double > &Charge)

bool	CheckPassFilter (double Charge, double Discriminant, std::vector< std::pair< double, double > > &Cuts, int Side)

double	DualNominalFitSingleTry (const std::vector< double > &Charge, int Offset, int Distance)

double	PerformDualNominalFit (const std::vector< double > &Charge)

double	PerformLinearFit (const std::vector< double > &Charge)

double	PerformNominalFit (const std::vector< double > &Charge)

TriangleFitResult	PerformTriangleFit (const std::vector< double > &Charge)

Private Attributes
std::vector< double >	CumulativeIdealPulse

std::vector< double >	mCharge

std::vector< std::pair< double, double > >	mLambdaLinearCut

std::vector< std::pair< double, double > >	mLambdaRMS8MaxCut

std::vector< std::pair< double, double > >	mLeftSlopeCut

double	mMinimumChargeThreshold

std::vector< std::pair< double, double > >	mRightSlopeCut

std::vector< std::pair< double, double > >	mRightSlopeSmallCut

bool	mTriangleIgnoreSlow

int	mTrianglePeakTS

double	mTS4TS5ChargeThreshold

std::vector< std::pair< double, double > >	mTS4TS5LowerCut

std::vector< std::pair< double, double > >	mTS4TS5UpperCut

bool	mUseDualFit

Detailed Description

Definition at line 28 of file HBHEPulseShapeFlag.h.

Constructor & Destructor Documentation

HBHEPulseShapeFlagSetter::HBHEPulseShapeFlagSetter ( )

Definition at line 23 of file HBHEPulseShapeFlag.cc.

References mMinimumChargeThreshold, and mTS4TS5ChargeThreshold.

 {
    //
    // Argumentless constructor; should not be used
    // 
    // If arguments not properly specified for the constructor, I don't think
    // we'd trust the flagging algorithm.
    // Set the minimum charge threshold large enough so that nothing will be flagged.
    // 
 
    mMinimumChargeThreshold = 99999999;
    mTS4TS5ChargeThreshold = 99999999;
 }

HBHEPulseShapeFlagSetter::HBHEPulseShapeFlagSetter	(	double	MinimumChargeThreshold,
		double	TS4TS5ChargeThreshold,
		unsigned int	TrianglePeakTS,
		const std::vector< double > &	LinearThreshold,
		const std::vector< double > &	LinearCut,
		const std::vector< double > &	RMS8MaxThreshold,
		const std::vector< double > &	RMS8MaxCut,
		const std::vector< double > &	LeftSlopeThreshold,
		const std::vector< double > &	LeftSlopeCut,
		const std::vector< double > &	RightSlopeThreshold,
		const std::vector< double > &	RightSlopeCut,
		const std::vector< double > &	RightSlopeSmallThreshold,
		const std::vector< double > &	RightSlopeSmallCut,
		const std::vector< double > &	TS4TS5UpperThreshold,
		const std::vector< double > &	TS4TS5UpperCut,
		const std::vector< double > &	TS4TS5LowerThreshold,
		const std::vector< double > &	TS4TS5LowerCut,
		bool	UseDualFit,
		bool	TriangleIgnoreSlow
	)

Definition at line 37 of file HBHEPulseShapeFlag.cc.

References i, Initialize(), mLambdaLinearCut, mLambdaRMS8MaxCut, mLeftSlopeCut, mMinimumChargeThreshold, mRightSlopeCut, mRightSlopeSmallCut, mTriangleIgnoreSlow, mTrianglePeakTS, mTS4TS5ChargeThreshold, mTS4TS5LowerCut, mTS4TS5UpperCut, mUseDualFit, and python.multivaluedict::sort().

 {
    //
    // The constructor that should be used
    //
    // Copies various thresholds and limits and parameters to the class for future use.
    // Also calls the Initialize() function
    //
 
    mMinimumChargeThreshold = MinimumChargeThreshold;
    mTS4TS5ChargeThreshold = TS4TS5ChargeThreshold;
    mTrianglePeakTS = TrianglePeakTS;
    mTriangleIgnoreSlow = TriangleIgnoreSlow;
 
    for(std::vector<double>::size_type i = 0; i < LinearThreshold.size() && i < LinearCut.size(); i++)
       mLambdaLinearCut.push_back(std::pair<double, double>(LinearThreshold[i], LinearCut[i]));
    sort(mLambdaLinearCut.begin(), mLambdaLinearCut.end());
 
    for(std::vector<double>::size_type i = 0; i < RMS8MaxThreshold.size() && i < RMS8MaxCut.size(); i++)
       mLambdaRMS8MaxCut.push_back(std::pair<double, double>(RMS8MaxThreshold[i], RMS8MaxCut[i]));
    sort(mLambdaRMS8MaxCut.begin(), mLambdaRMS8MaxCut.end());
 
    for(std::vector<double>::size_type i = 0; i < LeftSlopeThreshold.size() && i < LeftSlopeCut.size(); i++)
       mLeftSlopeCut.push_back(std::pair<double, double>(LeftSlopeThreshold[i], LeftSlopeCut[i]));
    sort(mLeftSlopeCut.begin(), mLeftSlopeCut.end());
 
    for(std::vector<double>::size_type i = 0; i < RightSlopeThreshold.size() && i < RightSlopeCut.size(); i++)
       mRightSlopeCut.push_back(std::pair<double, double>(RightSlopeThreshold[i], RightSlopeCut[i]));
    sort(mRightSlopeCut.begin(), mRightSlopeCut.end());
 
    for(std::vector<double>::size_type i = 0; i < RightSlopeSmallThreshold.size() && i < RightSlopeSmallCut.size(); i++)
       mRightSlopeSmallCut.push_back(std::pair<double, double>(RightSlopeSmallThreshold[i], RightSlopeSmallCut[i]));
    sort(mRightSlopeSmallCut.begin(), mRightSlopeSmallCut.end());
    
    for(std::vector<double>::size_type i = 0; i < TS4TS5UpperThreshold.size() && i < TS4TS5UpperCut.size(); i++)
       mTS4TS5UpperCut.push_back(std::pair<double, double>(TS4TS5UpperThreshold[i], TS4TS5UpperCut[i]));
    sort(mTS4TS5UpperCut.begin(), mTS4TS5UpperCut.end());
 
    for(std::vector<double>::size_type i = 0; i < TS4TS5LowerThreshold.size() && i < TS4TS5LowerCut.size(); i++)
       mTS4TS5LowerCut.push_back(std::pair<double, double>(TS4TS5LowerThreshold[i], TS4TS5LowerCut[i]));
    sort(mTS4TS5LowerCut.begin(), mTS4TS5LowerCut.end());
 
    mUseDualFit = UseDualFit;
 
    Initialize();
 }

HBHEPulseShapeFlagSetter::~HBHEPulseShapeFlagSetter ( )

Definition at line 102 of file HBHEPulseShapeFlag.cc.

 {
    // Dummy destructor - there's nothing to destruct by hand here
 }

Member Function Documentation

double HBHEPulseShapeFlagSetter::CalculateRMS8Max ( const std::vector< double > & Charge )

private

Definition at line 576 of file HBHEPulseShapeFlag.cc.

References alignCSCRings::e, i, RMS, python.multivaluedict::sort(), and mathSSE::sqrt().

Referenced by SetPulseShapeFlags().

 {
    //
    // CalculateRMS8Max
    //
    // returns "RMS" divided by the largest charge in the time slices
    //    "RMS" is calculated using all but the two largest time slices.
    //    The "RMS" is not quite the actual RMS (see note below), but the value is only
    //    used for determining max values, and is not quoted as the actual RMS anywhere.
    //
 
    int DigiSize = Charge.size();
 
    if (DigiSize<=2)  return 1e-5;  // default statement when DigiSize is too small for useful RMS calculation
   // Copy Charge vector again, since we are passing references around
    std::vector<double> TempCharge = Charge;
 
    // Sort TempCharge vector from smallest to largest charge
    sort(TempCharge.begin(), TempCharge.end());
 
    double Total = 0;
    double Total2 = 0;
    for(int i = 0; i < DigiSize - 2; i++)
    {
       Total = Total + TempCharge[i];
       Total2 = Total2 + TempCharge[i] * TempCharge[i];
    }
 
    // This isn't quite the RMS (both Total2 and Total*Total need to be
    // divided by an extra (DigiSize-2) within the sqrt to get the RMS.)
    // We're only using this value for relative comparisons, though; we
    // aren't explicitly interpreting it as the RMS.  It might be nice
    // to either change the calculation or rename the variable in the future, though.
 
    double RMS = sqrt(Total2 - Total * Total / (DigiSize - 2));
 
    double RMS8Max = RMS / TempCharge[DigiSize-1];
    if(RMS8Max < 1e-5)   // protection against zero
       RMS8Max = 1e-5;
 
    return RMS / TempCharge[DigiSize-1];
 }

bool HBHEPulseShapeFlagSetter::CheckPassFilter	(	double	Charge,
		double	Discriminant,
		std::vector< std::pair< double, double > > &	Cuts,
		int	Side
	)

private

Definition at line 680 of file HBHEPulseShapeFlag.cc.

References first, i, and MessageLogger_cff::limit.

Referenced by SetPulseShapeFlags().

 {
    //
    // Checks whether Discriminant value passes Cuts for the specified Charge.  True if pulse is good.
    //
    // The "Cuts" pairs are assumed to be sorted in terms of size from small to large,
    //    where each "pair" = (Charge, Discriminant)
    // "Side" is either positive or negative, which determines whether to discard the pulse if discriminant
    //    is greater or smaller than the cut value
    //
 
    if(Cuts.size() == 0)   // safety check that there are some cuts defined
       return true;
 
    if(Charge <= Cuts[0].first)   // too small to cut on
       return true;
 
    int IndexLargerThanCharge = -1;   // find the range it is falling in
    for(int i = 1; i < (int)Cuts.size(); i++)
    {
       if(Cuts[i].first > Charge)
       {
          IndexLargerThanCharge = i;
          break;
       }
    }
 
    double limit = 1000000;
 
    if(IndexLargerThanCharge == -1)   // if charge is greater than the last entry, assume flat line
       limit = Cuts[Cuts.size()-1].second;
    else   // otherwise, do a linear interpolation to find the cut position
    {
       double C1 = Cuts[IndexLargerThanCharge].first;
       double C2 = Cuts[IndexLargerThanCharge-1].first;
       double L1 = Cuts[IndexLargerThanCharge].second;
       double L2 = Cuts[IndexLargerThanCharge-1].second;
 
       limit = (Charge - C1) / (C2 - C1) * (L2 - L1) + L1;
    }
 
    if(Side > 0 && Discriminant > limit)
       return false;
    if(Side < 0 && Discriminant < limit)
       return false;
 
    return true;
 }

void HBHEPulseShapeFlagSetter::Clear ( )

Definition at line 107 of file HBHEPulseShapeFlag.cc.

 {
    // Dummy function in case something needs to be cleaned....but none right now
 }

double HBHEPulseShapeFlagSetter::DualNominalFitSingleTry	(	const std::vector< double > &	Charge,
		int	Offset,
		int	Distance
	)

private

Definition at line 483 of file HBHEPulseShapeFlag.cc.

References CumulativeIdealPulse, alignCSCRings::e, and j.

Referenced by PerformDualNominalFit().

 {
    //
    // Does a fit to dual signal pulse hypothesis given offset and distance of the two target pulses
    //
    // The only parameters to fit here are the two pulse heights of in-time and out-of-time components
    //    since offset is given
    // The calculation here is based from writing down the Chi2 formula and minimize against the two parameters,
    //    ie., Chi2 = Sum{((T[i] - a1 * F1[i] - a2 * F2[i]) / (Sigma[i]))^2}, where T[i] is the input pulse shape,
    //    and F1[i], F2[i] are the two ideal pulse components
    //
 
    int DigiSize = Charge.size();
 
    if(Offset < 0 || Offset + 250 >= (int)CumulativeIdealPulse.size())
       return 1000000;
    if(CumulativeIdealPulse[Offset+250] - CumulativeIdealPulse[Offset] < 1e-5)
       return 1000000;
 
    static std::vector<double> F1;
    static std::vector<double> F2;
 
    F1.resize(DigiSize);
    F2.resize(DigiSize);
 
    double SumF1F1 = 0;
    double SumF1F2 = 0;
    double SumF2F2 = 0;
    double SumTF1 = 0;
    double SumTF2 = 0;
 
    double Error = 0;
 
    for(int j = 0; j < DigiSize; j++)
    {
       // this is the TS value for in-time component - no problem we can do a subtraction directly
       F1[j] = CumulativeIdealPulse[Offset+j*25+25] - CumulativeIdealPulse[Offset+j*25];
 
       // However for the out-of-time component the index might go out-of-bound.
       // Let's protect against this.
 
       int OffsetTemp = Offset + j * 25 + Distance;
       
       double C1 = 0;   // lower-indexed value in the cumulative pulse shape
       double C2 = 0;   // higher-indexed value in the cumulative pulse shape
       
       if(OffsetTemp + 25 < (int)CumulativeIdealPulse.size() && OffsetTemp + 25 >= 0)
          C1 = CumulativeIdealPulse[OffsetTemp+25];
       if(OffsetTemp + 25 >= (int)CumulativeIdealPulse.size())
          C1 = CumulativeIdealPulse[CumulativeIdealPulse.size()-1];
       if(OffsetTemp < (int)CumulativeIdealPulse.size() && OffsetTemp >= 0)
          C2 = CumulativeIdealPulse[OffsetTemp];
       if(OffsetTemp >= (int)CumulativeIdealPulse.size())
          C2 = CumulativeIdealPulse[CumulativeIdealPulse.size()-1];
       F2[j] = C1 - C2;
 
       Error = Charge[j];
       if(Error < 1)
          Error = 1;
 
       SumF1F1 += F1[j] * F1[j] / Error;
       SumF1F2 += F1[j] * F2[j] / Error; 
       SumF2F2 += F2[j] * F2[j] / Error;
       SumTF1  += F1[j] * Charge[j] / Error; 
       SumTF2  += F2[j] * Charge[j] / Error; 
    }
 
    double Height  = 0;
    double Height2 = 0;
      if (fabs(SumF1F2*SumF1F2-SumF1F1*SumF2F2)>1e-5)
        {
      Height  = (SumF1F2 * SumTF2 - SumF2F2 * SumTF1) / (SumF1F2 * SumF1F2 - SumF1F1 * SumF2F2);
      Height2 = (SumF1F2 * SumTF1 - SumF1F1 * SumTF2) / (SumF1F2 * SumF1F2 - SumF1F1 * SumF2F2);
        }
 
    double Chi2 = 0;
    for(int j = 0; j < DigiSize; j++)
    {
       double Error = Charge[j];
       if(Error < 1)
          Error = 1;
 
       double Residual = Height * F1[j] + Height2 * F2[j] - Charge[j];  
       Chi2 += Residual * Residual / Error;                             
    } 
 
    // Safety protection in case of zero
    if(Chi2 < 1e-5)
       Chi2 = 1e-5;
 
    return Chi2;
 }

void HBHEPulseShapeFlagSetter::Initialize ( )

Definition at line 206 of file HBHEPulseShapeFlag.cc.

References HcalPulseShape::at(), CumulativeIdealPulse, HcalPulseShapes::hbShape(), i, and mCharge.

Referenced by HBHEPulseShapeFlagSetter().

 {
    // 
    // Initialization: whatever preprocess is needed
    //
    // 1. Get the ideal pulse shape from CMSSW
    //
    //    Since the HcalPulseShapes class stores the ideal pulse shape in terms of 256 numbers,
    //    each representing 1ns integral of the ideal pulse, here I'm taking out the vector
    //    by calling at() function.
    //
    //    A cumulative distribution is formed and stored to save some time doing integration to TS later on
    //
    // 2. Reserve space for vector
    //
 
    std::vector<double> PulseShape;
 
    HcalPulseShapes Shapes;
    HcalPulseShapes::Shape HPDShape = Shapes.hbShape();
 
    PulseShape.reserve(350);
    for(int i = 0; i < 200; i++)
       PulseShape.push_back(HPDShape.at(i));
    PulseShape.insert(PulseShape.begin(), 150, 0);   // Safety margin of a lot of zeros in the beginning
 
    CumulativeIdealPulse.reserve(350);
    CumulativeIdealPulse.clear();
    CumulativeIdealPulse.push_back(0);
    for(unsigned int i = 1; i < PulseShape.size(); i++)
       CumulativeIdealPulse.push_back(CumulativeIdealPulse[i-1] + PulseShape[i]);
 
    // reserve space for vector
    mCharge.reserve(10);
 }

double HBHEPulseShapeFlagSetter::PerformDualNominalFit ( const std::vector< double > & Charge )

private

Definition at line 435 of file HBHEPulseShapeFlag.cc.

References CumulativeIdealPulse, DualNominalFitSingleTry(), i, and gen::k.

Referenced by SetPulseShapeFlags().

 {
    //
    // Perform dual nominal fit and returns the chi2
    // 
    // In this function we do a scan over possible "distance" (number of time slices between two components)
    //    and overall offset for the two components; first coarse, then finer
    // All the fitting is done in the DualNominalFitSingleTry function
    //
 
    double OverallMinimumChi2 = 1000000;
 
    int AvailableDistance[] = {-100, -75, -50, 50, 75, 100};
 
    // loop over possible pulse distances between two components
    for(int k = 0; k < 6; k++)
    {
       double SingleMinimumChi2 = 1000000;
       int MinOffset = 0;
 
       // scan coarsely through different offsets and find the minimum
       for(int i = 0; i + 250 < (int)CumulativeIdealPulse.size(); i += 10)
       {
          double Chi2 = DualNominalFitSingleTry(Charge, i, AvailableDistance[k]);
 
          if(Chi2 < SingleMinimumChi2)
          {
             SingleMinimumChi2 = Chi2;
             MinOffset = i;
          }
       }
 
       // around the minimum, scan finer for better a better minimum
       for(int i = MinOffset - 15; i + 250 < (int)CumulativeIdealPulse.size() && i < MinOffset + 15; i++)
       {
          double Chi2 = DualNominalFitSingleTry(Charge, i, AvailableDistance[k]);
          if(Chi2 < SingleMinimumChi2)
             SingleMinimumChi2 = Chi2;
       }
 
       // update overall minimum chi2
       if(SingleMinimumChi2 < OverallMinimumChi2)
          OverallMinimumChi2 = SingleMinimumChi2;
    }
 
    return OverallMinimumChi2;
 }

double HBHEPulseShapeFlagSetter::PerformLinearFit ( const std::vector< double > & Charge )

private

Definition at line 619 of file HBHEPulseShapeFlag.cc.

References alignCSCRings::e, and i.

Referenced by SetPulseShapeFlags().

 {
    //
    // Performs a straight-line fit over all time slices, and returns the chi2 value
    //
    // The calculation here is based from writing down the formula for chi2 and minimize
    //    with respect to the parameters in the fit, ie., slope and intercept of the straight line
    // By doing two differentiation, we will get two equations, and the best parameters are determined by these
    //
 
    int DigiSize = Charge.size();
 
    double SumTS2OverTi = 0;
    double SumTSOverTi = 0;
    double SumOverTi = 0;
    double SumTiTS = 0;
    double SumTi = 0;
 
    double Error2 = 0;
    for(int i = 0; i < DigiSize; i++)
    {
       Error2 = Charge[i];
       if(Charge[i] < 1)
          Error2 = 1;
 
       SumTS2OverTi += 1.* i * i / Error2;
       SumTSOverTi  += 1.* i / Error2;
       SumOverTi    += 1. / Error2;
       SumTiTS      += Charge[i] * i / Error2;
       SumTi        += Charge[i] / Error2;
    }
 
    double CM1 = SumTS2OverTi;   // Coefficient in front of slope in equation 1
    double CM2 = SumTSOverTi;   // Coefficient in front of slope in equation 2
    double CD1 = SumTSOverTi;   // Coefficient in front of intercept in equation 1
    double CD2 = SumOverTi;   // Coefficient in front of intercept in equation 2
    double C1 = SumTiTS;   // Constant coefficient in equation 1
    double C2 = SumTi;   // Constant coefficient in equation 2
 
    // Denominators always non-zero by construction
    double Slope = (C1 * CD2 - C2 * CD1) / (CM1 * CD2 - CM2 * CD1);
    double Intercept = (C1 * CM2 - C2 * CM1) / (CD1 * CM2 - CD2 * CM1);
 
    // now that the best parameters are found, calculate chi2 from those
    double Chi2 = 0;
    for(int i = 0; i < DigiSize; i++)
    {
       double Deviation = Slope * i + Intercept - Charge[i];
       double Error2 = Charge[i];
       if(Charge[i] < 1)
          Error2 = 1;
       Chi2 += Deviation * Deviation / Error2;  
    }
 
    // safety protection in case of perfect fit
    if(Chi2 < 1e-5)
       Chi2 = 1e-5;
 
    return Chi2;
 }

double HBHEPulseShapeFlagSetter::PerformNominalFit ( const std::vector< double > & Charge )

private

Definition at line 368 of file HBHEPulseShapeFlag.cc.

References CumulativeIdealPulse, alignCSCRings::e, F(), i, and j.

Referenced by SetPulseShapeFlags().

 {
    //
    // Performs a fit to the ideal pulse shape.  Returns best chi2
    //
    // A scan over different timing offset (for the ideal pulse) is carried out,
    //    and for each offset setting a one-parameter fit is performed
    //
 
    int DigiSize = Charge.size();
 
    double MinimumChi2 = 100000;
 
    double F = 0;
 
    double SumF2 = 0;
    double SumTF = 0;
    double SumT2 = 0;
 
    for(int i = 0; i + 250 < (int)CumulativeIdealPulse.size(); i++)
    {
       if(CumulativeIdealPulse[i+250] - CumulativeIdealPulse[i] < 1e-5)
          continue;
 
       SumF2 = 0;
       SumTF = 0;
       SumT2 = 0;
 
       double ErrorTemp=0;
       for(int j = 0; j < DigiSize; j++)
     {
       // get ideal pulse component for this time slice....
       F = CumulativeIdealPulse[i+j*25+25] - CumulativeIdealPulse[i+j*25];
       
       ErrorTemp=Charge[j];
       if (ErrorTemp<1) // protection against small charges
         ErrorTemp=1;
       // ...and increment various summations
       SumF2 += F * F / ErrorTemp;
       SumTF += F * Charge[j] / ErrorTemp;
       SumT2 += fabs(Charge[j]);
     }
       
       /* 
      chi2= sum((Charge[j]-aF)^2/|Charge[j]|
      ( |Charge[j]| = assumed sigma^2 for Charge[j]; a bit wonky for Charge[j]<1 )
          chi2 = sum(|Charge[j]|) - 2*sum(aF*Charge[j]/|Charge[j]|) +sum( a^2*F^2/|Charge[j]|)
      chi2 minimimized when d(chi2)/da = 0:
          a = sum(F*Charge[j])/sum(F^2)
      ...
          chi2= sum(|Q[j]|) - sum(Q[j]/|Q[j]|*F)*sum(Q[j]/|Q[j]|*F)/sum(F^2/|Q[j]|), where Q = Charge
      chi2 = SumT2 - SumTF*SumTF/SumF2
       */
       
       double Chi2 = SumT2 - SumTF * SumTF / SumF2;
       
       if(Chi2 < MinimumChi2)
     MinimumChi2 = Chi2;
    }
    
    // safety protection in case of perfect fit - don't want the log(...) to explode
    if(MinimumChi2 < 1e-5)
       MinimumChi2 = 1e-5;
 
    return MinimumChi2;
 }

TriangleFitResult HBHEPulseShapeFlagSetter::PerformTriangleFit ( const std::vector< double > & Charge )

private

Definition at line 242 of file HBHEPulseShapeFlag.cc.

References TriangleFitResult::Chi2, i, TriangleFitResult::LeftSlope, mTrianglePeakTS, query::result, and TriangleFitResult::RightSlope.

Referenced by SetPulseShapeFlags().

 {
    //
    // Perform a "triangle fit", and extract the slopes
    //
    // Left-hand side and right-hand side are not correlated to each other - do them separately
    //
 
    TriangleFitResult result;
    result.Chi2 = 0;
    result.LeftSlope = 0;
    result.RightSlope = 0;
 
    int DigiSize = Charge.size();
 
    // right side, starting from TS4
    double MinimumRightChi2 = 1000000;
    double Numerator = 0;
    double Denominator = 0;
 
    for(int iTS = mTrianglePeakTS + 2; iTS <= DigiSize; iTS++)   // the place where first TS center in flat line
    {
       // fit a straight line to the triangle part
       Numerator = 0;
       Denominator = 0;
 
       for(int i = mTrianglePeakTS + 1; i < iTS; i++)
       {
          Numerator += (i - mTrianglePeakTS) * (Charge[i] - Charge[mTrianglePeakTS]);
          Denominator += (i - mTrianglePeakTS) * (i - mTrianglePeakTS);
       }
 
       double BestSlope = 0;
       if (Denominator!=0) BestSlope = Numerator / Denominator;
       if(BestSlope > 0)
          BestSlope = 0;
 
       // check if the slope is reasonable
       if(iTS != DigiSize)
     {
       // iTS starts at mTrianglePeak+2; denominator never zero
       if(BestSlope > -1 * Charge[mTrianglePeakTS] / (iTS - mTrianglePeakTS))
             BestSlope = -1 * Charge[mTrianglePeakTS] / (iTS - mTrianglePeakTS);
       if(BestSlope < -1 * Charge[mTrianglePeakTS] / (iTS - 1 - mTrianglePeakTS))
         BestSlope = -1 * Charge[mTrianglePeakTS] / (iTS - 1 - mTrianglePeakTS);
     }
       else
     {
       // iTS starts at mTrianglePeak+2; denominator never zero
       if(BestSlope < -1 * Charge[mTrianglePeakTS] / (iTS - 1 - mTrianglePeakTS)) 
             BestSlope = -1 * Charge[mTrianglePeakTS] / (iTS - 1 - mTrianglePeakTS);
     }
       
       // calculate partial chi2
 
       // The shape I'm fitting is more like a tent than a triangle.
       // After the end of triangle edge assume a flat line
 
       double Chi2 = 0;
       for(int i = mTrianglePeakTS + 1; i < iTS; i++)
          Chi2 += (Charge[mTrianglePeakTS] - Charge[i] + (i - mTrianglePeakTS) * BestSlope)
             * (Charge[mTrianglePeakTS] - Charge[i] + (i - mTrianglePeakTS) * BestSlope);
       for(int i = iTS; i < DigiSize; i++)    // Assumes fit line = 0 for iTS > fit
          Chi2 += Charge[i] * Charge[i];
 
       if(Chi2 < MinimumRightChi2)
       {
          MinimumRightChi2 = Chi2;
          result.RightSlope = BestSlope;
       }
    }   // end of right-hand side loop
 
    // left side
    double MinimumLeftChi2 = 1000000;
 
    for(int iTS = 0; iTS < (int)mTrianglePeakTS; iTS++)   // the first time after linear fit ends
    {
       // fit a straight line to the triangle part
       Numerator = 0;
       Denominator = 0;
       for(int i = iTS; i < (int)mTrianglePeakTS; i++)
       {
          Numerator = Numerator + (i - mTrianglePeakTS) * (Charge[i] - Charge[mTrianglePeakTS]);
          Denominator = Denominator + (i - mTrianglePeakTS) * (i - mTrianglePeakTS);
       }
 
       double BestSlope = 0;
       if (Denominator!=0) BestSlope = Numerator / Denominator;
       if (BestSlope < 0)
     BestSlope = 0;
 
       // check slope
       if(iTS != 0)
     {
       // iTS must be >0 and < mTrianglePeakTS; slope never 0
       if(BestSlope > Charge[mTrianglePeakTS] / (mTrianglePeakTS - iTS))
             BestSlope = Charge[mTrianglePeakTS] / (mTrianglePeakTS - iTS);
       if(BestSlope < Charge[mTrianglePeakTS] / (mTrianglePeakTS + 1 - iTS))
             BestSlope = Charge[mTrianglePeakTS] / (mTrianglePeakTS + 1 - iTS);
     }
       else
     {
       if(BestSlope > Charge[mTrianglePeakTS] / (mTrianglePeakTS - iTS))
             BestSlope = Charge[mTrianglePeakTS] / (mTrianglePeakTS - iTS);
     }
       
       // calculate minimum chi2
       double Chi2 = 0;
       for(int i = 0; i < iTS; i++)
          Chi2 += Charge[i] * Charge[i];
       for(int i = iTS; i < (int)mTrianglePeakTS; i++)
          Chi2 += (Charge[mTrianglePeakTS] - Charge[i] + (i - mTrianglePeakTS) * BestSlope)
             * (Charge[mTrianglePeakTS] - Charge[i] + (i - mTrianglePeakTS) * BestSlope);
 
       if(MinimumLeftChi2 > Chi2)
       {
          MinimumLeftChi2 = Chi2;
          result.LeftSlope = BestSlope;
       }
    }   // end of left-hand side loop
 
    result.Chi2 = MinimumLeftChi2 + MinimumRightChi2;
 
    return result;
 }

void HBHEPulseShapeFlagSetter::SetPulseShapeFlags	(	HBHERecHit &	hbhe,
		const HBHEDataFrame &	digi,
		const HcalCoder &	coder,
		const HcalCalibrations &	calib
	)

Definition at line 112 of file HBHEPulseShapeFlag.cc.

References HcalCoder::adc2fC(), CalculateRMS8Max(), HcalQIESample::capid(), CheckPassFilter(), alignCSCRings::e, HcalCaloFlagLabels::HBHEFlatNoise, HcalCaloFlagLabels::HBHESpikeNoise, HcalCaloFlagLabels::HBHETriangleNoise, HcalCaloFlagLabels::HBHETS4TS5Noise, i, HBHERecHit::id(), HcalDetId::ietaAbs(), TriangleFitResult::LeftSlope, create_public_lumi_plots::log, mCharge, mLambdaLinearCut, mLambdaRMS8MaxCut, mLeftSlopeCut, mMinimumChargeThreshold, mRightSlopeCut, mRightSlopeSmallCut, mTriangleIgnoreSlow, mTrianglePeakTS, mTS4TS5ChargeThreshold, mTS4TS5LowerCut, mTS4TS5UpperCut, mUseDualFit, HcalCalibrations::pedestal(), PerformDualNominalFit(), PerformLinearFit(), PerformNominalFit(), PerformTriangleFit(), TriangleFitResult::RightSlope, HBHEDataFrame::sample(), CaloRecHit::setFlagField(), and HBHEDataFrame::size().

Referenced by HcalHitReconstructor::produce().

 {
    //
    // Decide if a digi/pulse is good or bad using fit-based discriminants
    //
    // SetPulseShapeFlags determines the total charge in the digi.
    // If the charge is above the minimum threshold, the code then
    // runs the flag-setting algorithms to determine whether the
    // flags should be set.
    //
 
    // hack to exclude ieta=28/29 for the moment... 
    int abseta = hbhe.id().ietaAbs();
    if(abseta == 28 || abseta == 29)  return;
 
    CaloSamples Tool;
    coder.adc2fC(digi, Tool);
 
    mCharge.clear();  // mCharge is a vector of (pedestal-subtracted) Charge values vs. time slice
    mCharge.resize(digi.size());
 
    double TotalCharge = 0;
 
    for(int i = 0; i < digi.size(); ++i)
    {
       mCharge[i] = Tool[i] - calib.pedestal(digi.sample(i).capid());
       TotalCharge += mCharge[i];
    }
 
    // No flagging if TotalCharge is less than threshold
    if(TotalCharge < mMinimumChargeThreshold)
       return;
 
    double NominalChi2 = 0; 
    if (mUseDualFit == true)
      NominalChi2=PerformDualNominalFit(mCharge); 
    else
      NominalChi2=PerformNominalFit(mCharge);
 
    double LinearChi2 = PerformLinearFit(mCharge);
 
    double RMS8Max = CalculateRMS8Max(mCharge);
    TriangleFitResult TriangleResult = PerformTriangleFit(mCharge);
 
    // Set the HBHEFlatNoise and HBHESpikeNoise flags
    if(CheckPassFilter(TotalCharge, log(LinearChi2) - log(NominalChi2), mLambdaLinearCut, -1) == false)
       hbhe.setFlagField(1, HcalCaloFlagLabels::HBHEFlatNoise);
    if(CheckPassFilter(TotalCharge, log(RMS8Max) * 2 - log(NominalChi2), mLambdaRMS8MaxCut, -1) == false)
       hbhe.setFlagField(1, HcalCaloFlagLabels::HBHESpikeNoise);
 
    // Set the HBHETriangleNoise flag
    if ((int)mCharge.size() >= mTrianglePeakTS)  // can't compute flag if peak TS isn't present; revise this at some point?
    {
      // initial values
      double TS4Left = 1000;
      double TS4Right = 1000;
  
      // Use 'if' statements to protect against slopes that are either 0 or very small
      if (TriangleResult.LeftSlope > 1e-5)
        TS4Left = mCharge[mTrianglePeakTS] / TriangleResult.LeftSlope;
      if (TriangleResult.RightSlope < -1e-5)
        TS4Right = mCharge[mTrianglePeakTS] / -TriangleResult.RightSlope;
      
      if(TS4Left > 1000 || TS4Left < -1000)
        TS4Left = 1000;
      if(TS4Right > 1000 || TS4Right < -1000)
        TS4Right = 1000;
      
      if(mTriangleIgnoreSlow == false)   // the slow-rising and slow-dropping edges won't be useful in 50ns/75ns
        {
          if(CheckPassFilter(mCharge[mTrianglePeakTS], TS4Left, mLeftSlopeCut, 1) == false)
        hbhe.setFlagField(1, HcalCaloFlagLabels::HBHETriangleNoise);
          else if(CheckPassFilter(mCharge[mTrianglePeakTS], TS4Right, mRightSlopeCut, 1) == false)
        hbhe.setFlagField(1, HcalCaloFlagLabels::HBHETriangleNoise);
        }
      
      // fast-dropping ones should be checked in any case
      if(CheckPassFilter(mCharge[mTrianglePeakTS], TS4Right, mRightSlopeSmallCut, -1) == false)
        hbhe.setFlagField(1, HcalCaloFlagLabels::HBHETriangleNoise);
    }
 
    if(mCharge[4] + mCharge[5] > mTS4TS5ChargeThreshold && mTS4TS5ChargeThreshold>0) // silly protection against negative charge values
    {
       double TS4TS5 = (mCharge[4] - mCharge[5]) / (mCharge[4] + mCharge[5]);
       if(CheckPassFilter(mCharge[4] + mCharge[5], TS4TS5, mTS4TS5UpperCut, 1) == false)
          hbhe.setFlagField(1, HcalCaloFlagLabels::HBHETS4TS5Noise);
       if(CheckPassFilter(mCharge[4] + mCharge[5], TS4TS5, mTS4TS5LowerCut, -1) == false)
          hbhe.setFlagField(1, HcalCaloFlagLabels::HBHETS4TS5Noise);
    }
 }