Convolution of two TH1F Histograms

[jtnattress]

I was wondering if there was a way to convolve two TH1F histograms? I was able to find this in the forum:

Can ROOT or RooFit do the convolution of two historgrams

I am a grudate student who use root to analysis the data. For my best knowledge, the RooFit can do convolution of two functions. And I need to get the convolution historgram of two sub-histograms instead of functions. Is there anyone who has experience or an idea about how to do this with ROOT or ROOFit easily? Thanks a lot.

but still was left wondering if it was possible.

I also found this older post regarding deconvolution:

Deconvolution Example

Hi; I remember a while ago, Rene posted an example on how to do spectrum deconvolutions, complete with a sample script and a root file. Does this still exist? If anyone has done deconvolutions, and can provide an example of how this is done, I would certainly appreciate it. Thanks, Mike

which was helpful.

I saw that ROOT has a built in class TF1Convolution, but it seems that only will accept functions as the arguments.

[Danilo]

Hi,

if you look closely to the code in the post you already identified, viewtopic.php?t=18405, the procedure does start with TH1F instances.

Cheers,
Danilo

[jtnattress]

Hi Danilo,

Thanks for the hint. I looked at the code more closely as you suggested and implemented some of the RooFit functions.

Here is my updated code:

/////////////////___Begin Code___////////////////////////
void fitConvolution()
{
  TH1F* h2 = new TH1F("h2", "Spectrum;time[ns];counts", 2000, 0., 2000.);
  TH1F* h3 = new TH1F("h3", "Spectrum;time[ns];counts" ,20000, 0., 60000.);

  TCanvas* c = new TCanvas("c1","rando",200,10,600,400);
  c->Divide(2,2);

  for ( Int_t i=0; i<10000; i++) 
  {
    Float_t x1, x2, x3, E1, E2, E3, 
            meanE1, meanE2, meanE3, sig1, sig2, sig3;

    double eVToJ = 1.55E-18;
    double ns =  1E9;
    double us = 1E6;
    double d = 8.5;
    double mN = 1.67E-27;
    
    meanE1 = 1.;
    meanE2 = 5.;
    meanE3 = 10.;

    sig1 = 0.5;
    sig2 = 1.;
    sig3 = 1.5;

    x1 = gRandom->Gaus(meanE1, sig1);//(mean, sigma)
    E1 = (d / sqrt(2 * x1*1E6 / mN * eVToJ) ) * ns;

    x2 = gRandom->Gaus(meanE2, sig2);
    E2 = (d / sqrt(2 * x2*1E6 / mN * eVToJ) ) * ns;

    x3 = gRandom->Gaus(meanE3, sig3);
    E3 = (d / sqrt(2 * x3*1E6 / mN * eVToJ) ) * ns;

    h2->Fill(E1);
    h2->Fill(E2);
    h2->Fill(E3);
  }

  for ( Int_t i=0; i<1E6; i++) 
  {
    double val;
    val = gRandom->Uniform(0.,10000);
    h3->Fill(val); 
  }

  RooRealVar x("x","x", 0, 2000);

  RooDataHist* RDHh2 = new RooDataHist("RDHh2", " ", x, h2);
  RooDataHist* RDHaccPulse = new RooDataHist("RDHaccPulse", " ", x, accPulseHisto);

  RooHistPdf* pdfh2 = new RooHistPdf("pdfh2", "pdfh2", x, *RDHh2, 2);
  RooHistPdf* pdfaccPulse = new RooHistPdf("pdfaccPulse", "pdfaccPulse", x, *RDHaccPulse, 2);

  RooNumConvPdf expTOF("ExpTOF", "ExpTOF", x, *pdfh2, *pdfaccPulse);

  RooPlot* frame = x.frame();
  expTOF.plotOn(frame,LineColor(kRed),LineStyle(kDashed));

  c->cd(2);
  h2->Draw();
  c->cd(3);
  accPulseHisto->Draw();
  c->cd(4);
  frame->Draw();
  c->Update();
}
/////////////____End Code____//////////////////////

I getting the following error:
[#0] ERROR:Integration – RooAdaptiveGaussKronrodIntegrator1D::integral() ERROR: number of iterations was insufficient

Any help would be greatly appreciated.

-Jason

[jtnattress]

I installed FFTW and performed the convolution with RooFFTConvPdf, which cleared the error I was receiving.