EMC resolution [message #15567] |
Fri, 11 October 2013 14:05 |
binsong
Messages: 10 Registered: December 2011
|
occasional visitor |
From: *in2p3.fr
|
|
hi,
i find some problems with the EMC energy resolution. Maybe there are some bugs.
I do the simulations for electron and photon with box generator with the latest trunk at both barrel region (80deg--100deg) and forward region (12deg--20deg).
i get the reconstructed energy(E_rec) by using GetEmcRawEnergy(),
then i plot (E_mc-E_rec)/E_mc for each simulation.
but i find the resolution sigma/E much smaller then the result in TDR:
electron_barrel: 1.18%
electron_forward: 1.58%
photon_barrel: 1.07%
photon_forward: 1.39%
and in TDR, the resolution for photon at 1GeV is more than 2%.
you can find all the four plots in attachment.
does someone know this problem?
Binsong
|
|
|
|
|
|
|
Re: EMC resolution [message #15619 is a reply to message #15567] |
Wed, 30 October 2013 15:23 |
Dima Melnychuk
Messages: 213 Registered: April 2004 Location: National Centre for Nucle...
|
first-grade participant |
From: *fuw.edu.pl
|
|
Some additional results on emc resolution from my side.
First of all after Jifeng put a latest version of digitization by default, (based on code for EMC feature extraction developed at KVI) the energy resolution is different.
I did study for barrel 30-130 degree range, 10 k events, 1 GeV photons.
For new digitization sigma=1.6%
For previous digitization sigma=1.2%
There are two options for non-uniformity, first is based on measurements and implemented by Christian Hammann, sigma=2.6%
And second one (linear non-uniformity) is based on fit to reproduce prototype data and implemented by Hossein Moeini, sigma=2.5%, so both options are very close.
You can use this option with
PndEmcHitProducer* emcHitProd = new PndEmcHitProducer();
TString nonuniformityFile=gSystem->Getenv("VMCWORKDIR");
nonuniformityFile+="/macro/params/EmcDigiNoniformityPars2.root";
emcHitProd->SetNonuniformityFile(nonuniformityFile.Data());
So non-uniformity definitely affect the result as well as newer digitization.
But you can see if you use non-uniform response the peak position is shifted and calibrartion/energy correction is necessary. And shift is in different direction for two implementations of non-uniformity. Hossein provided correction for his case but it was for older digitization and newer digitization shift a peak position itself a little bit. So it should be redone.
So I plan to redo energy correction for new digitization with both non-uniformity options and I hope rather soon.
2.6% resolution for 1 GeV photon is what Christian Hammann quoted before as a result in agreement with prototype data.
So when energy correction will be ready the question can be closed.
Dima
-
Attachment: energy1.png
(Size: 14.68KB, Downloaded 665 times)
-
Attachment: energy2.png
(Size: 15.18KB, Downloaded 640 times)
-
Attachment: energy3.png
(Size: 15.80KB, Downloaded 584 times)
-
Attachment: energy4.png
(Size: 16.12KB, Downloaded 703 times)
|
|
|
|
Re: EMC resolution [message #15671 is a reply to message #15567] |
Wed, 06 November 2013 12:35 |
Dima Melnychuk
Messages: 213 Registered: April 2004 Location: National Centre for Nucle...
|
first-grade participant |
From: *fuw.edu.pl
|
|
Hi,
I have just update emc energy correction parameters for the case of non-uniformity switched on with the latest version of digitization.
I also switched on the use of non-uniformity by default in all.par
The files with correction are in /macro/params/
emc_correction_hist_gamma_2.root
emc_correction_hist_gamma_3.root
emc_correction_hist_gamma_4.root
emc_correction_par_gamma_2.root
emc_correction_par_gamma_3.root
emc_correction_par_gamma_4.root
You use correction like
PndEmcAbsClusterCalibrator * calibrator1= PndEmcClusterCalibrator::MakeEmcClusterCalibrator(1, 3);
And then obtain calibrated energy like
cluster_energy_calibrated=calibrator1->Energy(cluster);
And here first number stands for the method applied (1 - correction from histogram, 2 - correction from parametrization)
Second number stands for version and here
1 - previous version of digitization, no non-uniformity (deprecated)
2 - current version of digitization, no non-uniformity
3 - current version of digitization, with non-uniformity
(should be used by default)
4 - current version of digitization, with non-uniformity calculated by Hossein (linear non-uniformity 1.5%, set from the file /macro/params/EmcDigiNoniformityPars2.root)
By the way the macro which produces the last non-uniformity file is /macro/emc/dedicated/fill_nonuniformity_param.C
Here some plots which demonstrate how applying non-uniformity change reconstructed emc energy. I do not know if it's by accident or by intention of Christian Hammann who provided non-uniformity parameters from measured data in case of barrel EMC correction is almost not needed, i.e. 1 GeV energy peak is centred around 1 GeV. But for endcaps correction is still needed anyway.
For energies in range 0-10 GeV energy versus polar angle
Reconstructed energy of 1 GeV photons
Reconstructed pi0 invariant mass
In this last case the corrected energy gives even slightly worse results.
In PndPidCorrelator.cxx
fEmcCalibrator= PndEmcClusterCalibrator::MakeEmcClusterCalibrator(2, 1);
should be modified to
fEmcCalibrator= PndEmcClusterCalibrator::MakeEmcClusterCalibrator(2, 3);
Dima
-
Attachment: e_vs_theta.png
(Size: 38.03KB, Downloaded 730 times)
-
Attachment: energy_1GeV.png
(Size: 17.17KB, Downloaded 684 times)
-
Attachment: mpi0.png
(Size: 25.53KB, Downloaded 668 times)
|
|
|
|
|
|
|
|
|
|
|
|
Re: EMC resolution [message #15871 is a reply to message #15567] |
Thu, 13 February 2014 15:16 |
binsong
Messages: 10 Registered: December 2011
|
occasional visitor |
From: *in2p3.fr
|
|
Dear all,
now the problem is clear. it is due to the nonuniformity.
To obtain a realistic resolution (validated by the PROTO60 results),
simulations must be done using following parameters in all.par:
Use_nonuniformity:1
Use-Shaped_noise 0
The first flag needs to be set to 1 to take into account the observed
non linearity of the light yield
in the barrel EMC crystals as a function of the interaction depth.
The effect is an increase of the sigma of a Nivosibirsk fit from
1.6 (Use_nonuniformity:0 ) to 2.5 % (with Use_nonuniformity:1) for
photons at 1 GeV with a flat distribution in theta from 30 to
130 deg.
With the second flag set to 0, a more precise description of the
electronic noise is obtained, while faster but less accurate results are
obtained with the flag set to 1. The effect is an increase of the
sigma of a Novosibirsk fit from
2.2 (Use-Shaped_noise 1) to 2.5 % (Use-Shaped_noise 0) for a photon
at 1 GeV
Before end of november 2013, the standard parameters were
"Use_nonuniformity 0" and "Use-Shaped_noise 1", leading to a too
optimistic resolution, by about a factor 2. Now, the standard setting is
"Use_nonuniformity 1" and "Use-Shaped_noise 0")
The raw energy includes both effects of the non-uniformity and of the
electronic noise. The cluster calibrated energy is deduced using
calibration
parameters adjusted as a function of theta and energy to reproduce the
pi0 mass.
By default, the calibration parameter file is the one corresponding to
Use_nonuniformity:1 and "Use-Shaped_noise 0".
The two attached figures are the results for photon with the new version of pandaroot.
[Updated on: Thu, 13 February 2014 16:54] Report message to a moderator
|
|
|
|