// * provided "as is" without express or implied warranty. *
// **************************************************************************
-#include "AliHMPIDCluster.h" //class header
+#include <TVirtualFitter.h> //Solve()
#include <TMinuit.h> //Solve()
#include <TClonesArray.h> //Solve()
#include <TMarker.h> //Draw()
+#include "AliLog.h" //FitFunc()
+
+#include "AliHMPIDCluster.h" //class header
+
Bool_t AliHMPIDCluster::fgDoCorrSin=kTRUE;
ClassImp(AliHMPIDCluster)
+
//++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
void AliHMPIDCluster::CoG()
{
// Arguments: none
// Returns: none
Int_t pc,px,py;
- AliHMPIDDigit::Lors2Pad(fX,fY,pc,px,py); //tmp digit to get it center
- Float_t x=fX-AliHMPIDDigit::LorsX(pc,px); //diff between cluster x and center of the pad contaning this cluster
+ AliHMPIDParam::Lors2Pad(fX,fY,pc,px,py); //tmp digit to get it center
+ Float_t x=fX-AliHMPIDParam::LorsX(pc,px); //diff between cluster x and center of the pad contaning this cluster
fX+=3.31267e-2*TMath::Sin(2*TMath::Pi()/0.8*x)-2.66575e-3*TMath::Sin(4*TMath::Pi()/0.8*x)+2.80553e-3*TMath::Sin(6*TMath::Pi()/0.8*x)+0.0070;
}
//++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
TMarker *pMark=new TMarker(X(),Y(),5); pMark->SetUniqueID(fSt);pMark->SetMarkerColor(kBlue); pMark->Draw();
}
//++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
-void AliHMPIDCluster::FitFunc(Int_t &iNpars, Double_t* /*deriv*/, Double_t &chi2, Double_t *par, Int_t /* */)
+void AliHMPIDCluster::FitFunc(Int_t &iNpars, Double_t* deriv, Double_t &chi2, Double_t *par, Int_t /* */)
{
// Cluster fit function
// par[0]=x par[1]=y par[2]=q for the first Mathieson shape
// par[3]=x par[4]=y par[5]=q for the second Mathieson shape and so on up to iNpars/3 Mathieson shapes
-// For each pad of the cluster calculates the difference between actual pad charge and the charge induced to this pad by all Mathieson distributions
+// For each pad of the cluster calculates the difference between actual pad charge and the charge induced to this pad by all Mathieson distributions
// Then the chi2 is calculated as the sum of this value squared for all pad in the cluster.
// Arguments: iNpars - number of parameters which is number of local maxima of cluster * 3
// chi2 - function result to be minimised
// par - parameters array of size iNpars
// Returns: none
- AliHMPIDCluster *pClu=(AliHMPIDCluster*)gMinuit->GetObjectFit();
- Int_t iNshape = iNpars/3;
-
+
+ AliHMPIDCluster *pClu=(AliHMPIDCluster*)TVirtualFitter::GetFitter()->GetObjectFit();
+
+ Int_t nPads = pClu->Size();
+ Double_t **derivPart;
+
+ derivPart = new Double_t*[iNpars];
chi2 = 0;
- for(Int_t i=0;i<pClu->Size();i++){ //loop on all pads of the cluster
+
+ Int_t iNshape = iNpars/3;
+
+ for(Int_t j=0;j<iNpars;j++){
+ deriv[j] = 0;
+ derivPart[j] = new Double_t[nPads];
+ for(Int_t i=0;i<nPads;i++){
+ derivPart[j][i] = 0;
+ }
+ }
+
+ if(iNshape>6) {Printf("HMPID Error!!: n. of clusters in FitFunc %i",iNshape);return;}
+ for(Int_t i=0;i<nPads;i++){ //loop on all pads of the cluster
+ Double_t dQpadMath = 0;
+ for(Int_t j=0;j<iNshape;j++){ //Mathiesons loop as all of them may contribute to this pad
+ Double_t fracMathi = pClu->Dig(i)->IntMathieson(par[3*j],par[3*j+1]);
+ dQpadMath+=par[3*j+2]*fracMathi; // par[3*j+2] is charge par[3*j] is x par[3*j+1] is y of current Mathieson
+
+ derivPart[3*j ][i] += par[3*j+2]*(pClu->Dig(i)->Mathieson(par[3*j]-pClu->Dig(i)->LorsX()-0.5*AliHMPIDParam::SizePadX())-
+ pClu->Dig(i)->Mathieson(par[3*j]-pClu->Dig(i)->LorsX()+0.5*AliHMPIDParam::SizePadX()))*
+ pClu->Dig(i)->IntPartMathi(par[3*j+1],2);
+ derivPart[3*j+1][i] += par[3*j+2]*(pClu->Dig(i)->Mathieson(par[3*j+1]-pClu->Dig(i)->LorsY()-0.5*AliHMPIDParam::SizePadY())-
+ pClu->Dig(i)->Mathieson(par[3*j+1]-pClu->Dig(i)->LorsY()+0.5*AliHMPIDParam::SizePadY()))*
+ pClu->Dig(i)->IntPartMathi(par[3*j],1);
+ derivPart[3*j+2][i] += fracMathi;
+ }
+ if(dQpadMath>0 && pClu->Dig(i)->Q()>0) {
+ chi2 +=TMath::Power((pClu->Dig(i)->Q()-dQpadMath),2)/pClu->Dig(i)->Q(); //chi2 function to be minimized
+ }
+ }
+ //loop on all pads of the cluster
+ for(Int_t i=0;i<nPads;i++){ //loop on all pads of the cluster
Double_t dQpadMath = 0; //pad charge collector
for(Int_t j=0;j<iNshape;j++){ //Mathiesons loop as all of them may contribute to this pad
- dQpadMath+=par[3*j+2]*pClu->Dig(i)->IntMathieson(par[3*j],par[3*j+1]); // par[3*j+2] is charge par[3*j] is x par[3*j+1] is y of current Mathieson
+ Double_t fracMathi = pClu->Dig(i)->IntMathieson(par[3*j],par[3*j+1]);
+ dQpadMath+=par[3*j+2]*fracMathi;
+ if(dQpadMath>0 && pClu->Dig(i)->Q()>0) {
+ deriv[3*j] += 2/pClu->Dig(i)->Q()*(pClu->Dig(i)->Q()-dQpadMath)*derivPart[3*j ][i];
+ deriv[3*j+1] += 2/pClu->Dig(i)->Q()*(pClu->Dig(i)->Q()-dQpadMath)*derivPart[3*j+1][i];
+ deriv[3*j+2] += 2/pClu->Dig(i)->Q()*(pClu->Dig(i)->Q()-dQpadMath)*derivPart[3*j+2][i];
+ }
}
-// if(dQpadMath>0)chi2 +=TMath::Power((pClu->Dig(i)->Q()-dQpadMath),2)/dQpadMath; //
- if(dQpadMath>0 && pClu->Dig(i)->Q())
- chi2 +=TMath::Power((pClu->Dig(i)->Q()-dQpadMath),2)/pClu->Dig(i)->Q(); //
- } //loop on all pads of the cluster
+ }
+ //delete array...
+ for(Int_t i=0;i<iNpars;i++) delete [] derivPart[i]; delete [] derivPart;
+
}//FitFunction()
//++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
void AliHMPIDCluster::Print(Option_t* opt)const
//Arguments: pCluLst - cluster list pointer where to add new cluster(s)
// isTryUnfold - flag to switch on/off unfolding
// Returns: number of local maxima of original cluster
- CoG();
+ const Int_t kMaxLocMax=6; //max allowed number of loc max for fitting
+//
+ CoG(); //First calculate CoG for the given cluster
Int_t iCluCnt=pCluLst->GetEntriesFast(); //get current number of clusters already stored in the list by previous operations
if(isTryUnfold==kFALSE || Size()==1) { //if cluster contains single pad there is no way to improve the knowledge
(isTryUnfold)?fSt=kSi1:fSt=kNot;
new ((*pCluLst)[iCluCnt++]) AliHMPIDCluster(*this); //add this raw cluster
return 1;
}
-//Phase 0. Initialise TMinuit
- const Int_t kMaxLocMax=6; //max allowed number of loc max for fitting
- if(!gMinuit) gMinuit = new TMinuit(100); //init MINUIT with this number of parameters (3 params per mathieson)
- gMinuit->mncler(); // reset Minuit list of paramters
- gMinuit->SetObjectFit((TObject*)this); gMinuit->SetFCN(AliHMPIDCluster::FitFunc); //set fit function
- Double_t aArg=-1; //tmp vars for TMinuit
- Int_t iErrFlg;
- gMinuit->mnexcm("SET PRI",&aArg,1,iErrFlg); //suspend all printout from TMinuit
- gMinuit->mnexcm("SET NOW",&aArg,0,iErrFlg); //suspend all warning printout from TMinuit
+
+//Phase 0. Initialise Fitter
+ Double_t arglist[10];
+ Int_t ierflg = 0;
+ TVirtualFitter *fitter = TVirtualFitter::Fitter(this,3*6); //initialize Fitter
+
+ arglist[0] = -1;
+ ierflg = fitter->ExecuteCommand("SET PRI", arglist, 1); // no printout
+ ierflg = fitter->ExecuteCommand("SET NOW", arglist, 0); //no warning messages
+ arglist[0] = 1;
+ ierflg = fitter->ExecuteCommand("SET GRA", arglist, 1); //force Fitter to use my gradient
+
+ fitter->SetFCN(AliHMPIDCluster::FitFunc);
+
+// arglist[0] = 1;
+// ierflg = fitter->ExecuteCommand("SET ERR", arglist ,1);
+
+// Set starting values and step sizes for parameters
+
//Phase 1. Find number of local maxima. Strategy is to check if the current pad has QDC more then all neigbours. Also find the box contaning the cluster
fNlocMax=0;
- for(Int_t iDig1=0;iDig1<Size();iDig1++) { //first digits loop
+ for(Int_t iDig1=0;iDig1<Size();iDig1++) { //first digits loop
+
AliHMPIDDigit *pDig1 = Dig(iDig1); //take next digit
Int_t iCnt = 0; //counts how many neighbouring pads has QDC more then current one
+
for(Int_t iDig2=0;iDig2<Size();iDig2++) { //loop on all digits again
+
if(iDig1==iDig2) continue; //the same digit, no need to compare
AliHMPIDDigit *pDig2 = Dig(iDig2); //take second digit to compare with the first one
Int_t dist = TMath::Sign(Int_t(pDig1->PadChX()-pDig2->PadChX()),1)+TMath::Sign(Int_t(pDig1->PadChY()-pDig2->PadChY()),1);//distance between pads
if(dist==1) //means dig2 is a neighbour of dig1
if(pDig2->Q()>=pDig1->Q()) iCnt++; //count number of pads with Q more then Q of current pad
+
}//second digits loop
+
if(iCnt==0&&fNlocMax<kMaxLocMax){ //this pad has Q more then any neighbour so it's local maximum
+
Double_t xStart=pDig1->LorsX();Double_t yStart=pDig1->LorsY();
- Double_t xMin=xStart-AliHMPIDDigit::SizePadX();
- Double_t xMax=xStart+AliHMPIDDigit::SizePadX();
- Double_t yMin=yStart-AliHMPIDDigit::SizePadY();
- Double_t yMax=yStart+AliHMPIDDigit::SizePadY();
- gMinuit->mnparm(3*fNlocMax ,Form("x%i",fNlocMax),xStart,0.1,xMin,xMax,iErrFlg); // X,Y,Q initial values of the loc max pad
- gMinuit->mnparm(3*fNlocMax+1,Form("y%i",fNlocMax),yStart,0.1,yMin,yMax,iErrFlg); // X, Y constrained to be near the loc max
- gMinuit->mnparm(3*fNlocMax+2,Form("q%i",fNlocMax),pDig1->Q(),0.1,0,100000,iErrFlg);// Q constrained to be positive
+ Double_t xMin=xStart-AliHMPIDParam::SizePadX();
+ Double_t xMax=xStart+AliHMPIDParam::SizePadX();
+ Double_t yMin=yStart-AliHMPIDParam::SizePadY();
+ Double_t yMax=yStart+AliHMPIDParam::SizePadY();
+
+ ierflg = fitter->SetParameter(3*fNlocMax ,Form("x%i",fNlocMax),xStart,0.1,xMin,xMax); // X,Y,Q initial values of the loc max pad
+ ierflg = fitter->SetParameter(3*fNlocMax+1,Form("y%i",fNlocMax),yStart,0.1,yMin,yMax); // X, Y constrained to be near the loc max
+ ierflg = fitter->SetParameter(3*fNlocMax+2,Form("q%i",fNlocMax),pDig1->Q(),0.1,0,100000); // Q constrained to be positive
+
fNlocMax++;
+
}//if this pad is local maximum
}//first digits loop
//Phase 2. Fit loc max number of Mathiesons or add this current cluster to the list
// case 1 -> no loc max found
- if ( fNlocMax == 0) { // case of no local maxima found: pads with same charge...
- gMinuit->mnparm(3*fNlocMax ,Form("x%i",fNlocMax),fX,0.1,0,0,iErrFlg); // Init values taken from CoG() -> fX,fY,fQRaw
- gMinuit->mnparm(3*fNlocMax+1,Form("y%i",fNlocMax),fY,0.1,0,0,iErrFlg); //
- gMinuit->mnparm(3*fNlocMax+2,Form("q%i",fNlocMax),fQRaw,0.1,0,100000,iErrFlg); //
+ if ( fNlocMax == 0) { // case of no local maxima found: pads with same charge...
+
+ ierflg = fitter->SetParameter(3*fNlocMax ,Form("x%i",fNlocMax),fX,0.1,0,0); // Init values taken from CoG() -> fX,fY,fQRaw
+ ierflg = fitter->SetParameter(3*fNlocMax+1,Form("y%i",fNlocMax),fY,0.1,0,0); //
+ ierflg = fitter->SetParameter(3*fNlocMax+2,Form("q%i",fNlocMax),fQRaw,0.1,0,100000); //
+
fNlocMax = 1;
fSt=kNoLoc;
}
// case 2 -> loc max found. Check # of loc maxima
- if ( fNlocMax >= kMaxLocMax) // if # of local maxima exceeds kMaxLocMax...
- {
- fSt = kMax; new ((*pCluLst)[iCluCnt++]) AliHMPIDCluster(*this); //...add this raw cluster
- } //or...
- else{ //...resonable number of local maxima to fit and user requested it
- Double_t arglist[10];arglist[0] = 10000;arglist[1] = 1.; //number of steps and sigma on pads charges
- gMinuit->mnexcm("SIMPLEX" ,arglist,2,iErrFlg); //start fitting with Simplex
- gMinuit->mnexcm("MIGRAD" ,arglist,2,iErrFlg); //fitting improved by Migrad
- if(iErrFlg) {
+ if ( fNlocMax >= kMaxLocMax) { // if # of local maxima exceeds kMaxLocMax...
+ fSt = kMax; new ((*pCluLst)[iCluCnt++]) AliHMPIDCluster(*this); //...add this raw cluster
+ } else { //or resonable number of local maxima to fit and user requested it
+ // Now ready for minimization step
+ arglist[0] = 500; //number of steps and sigma on pads charges
+ arglist[1] = 1.; //
+
+ ierflg = fitter->ExecuteCommand("SIMPLEX",arglist,2); //start fitting with Simplex
+ if (!ierflg)
+ fitter->ExecuteCommand("MIGRAD" ,arglist,2); //fitting improved by Migrad
+ if(ierflg) {
Double_t strategy=2;
- gMinuit->mnexcm("SET STR",&strategy,1,iErrFlg); //change level of strategy
- if(!iErrFlg) {
- gMinuit->mnexcm("SIMPLEX" ,arglist,2,iErrFlg);
- gMinuit->mnexcm("MIGRAD" ,arglist,2,iErrFlg); //fitting improved by Migrad
-// Printf("Try to improve fit --> err %d",iErrFlg);
+ ierflg = fitter->ExecuteCommand("SET STR",&strategy,1); //change level of strategy
+ if(!ierflg) {
+ ierflg = fitter->ExecuteCommand("SIMPLEX",arglist,2); //start fitting with Simplex
+ if (!ierflg)
+ fitter->ExecuteCommand("MIGRAD" ,arglist,2); //fitting improved by Migrad
}
}
- if(iErrFlg) fSt=kAbn; //no convergence of the fit...
- Double_t dummy; TString sName; //vars to get results from Minuit
+ if(ierflg) fSt=kAbn; //no convergence of the fit...
+ Double_t dummy; char sName[80]; //vars to get results from Minuit
+ Double_t edm, errdef;
+ Int_t nvpar, nparx;
+
for(Int_t i=0;i<fNlocMax;i++){ //store the local maxima parameters
- gMinuit->mnpout(3*i ,sName, fX, fErrX , dummy, dummy, iErrFlg); // X
- gMinuit->mnpout(3*i+1 ,sName, fY, fErrY , dummy, dummy, iErrFlg); // Y
- gMinuit->mnpout(3*i+2 ,sName, fQ, fErrQ , dummy, dummy, iErrFlg); // Q
- gMinuit->mnstat(fChi2,dummy,dummy,iErrFlg,iErrFlg,iErrFlg); // Chi2 of the fit
+ fitter->GetParameter(3*i ,sName, fX, fErrX , dummy, dummy); // X
+ fitter->GetParameter(3*i+1 ,sName, fY, fErrY , dummy, dummy); // Y
+ fitter->GetParameter(3*i+2 ,sName, fQ, fErrQ , dummy, dummy); // Q
+ fitter->GetStats(fChi2, edm, errdef, nvpar, nparx); //get fit infos
if(fSt!=kAbn) {
- if(fNlocMax!=1)fSt=kUnf; // if unfolded
- if(fNlocMax==1&&fSt!=kNoLoc) fSt=kLo1; // if only 1 loc max
- if ( !IsInPc()) fSt = kEdg; // if Out of Pc
- if(fSt==kNoLoc) fNlocMax=0; // if with no loc max (pads with same charge..)
+ if(fNlocMax!=1)fSt=kUnf; // if unfolded
+ if(fNlocMax==1&&fSt!=kNoLoc) fSt=kLo1; // if only 1 loc max
+ if ( !IsInPc()) fSt = kEdg; // if Out of Pc
+ if(fSt==kNoLoc) fNlocMax=0; // if with no loc max (pads with same charge..)
}
new ((*pCluLst)[iCluCnt++]) AliHMPIDCluster(*this); //add new unfolded cluster
}
git://git.uio.no / u / mrichter / AliRoot.git / blobdiff