Home » PANDA » PandaRoot » Tracking » Momentum resolution and reconstruction efficiency of LHE tracking
|
|
|
|
|
Re: Momentum resolution and reconstruction efficiency of LHE tracking [message #10167 is a reply to message #10166] |
Thu, 11 February 2010 00:30 |
Johan Messchendorp
Messages: 693 Registered: April 2007 Location: University of Groningen
|
first-grade participant |
From: *xs4all.nl
|
|
Stefano Spataro wrote on Thu, 11 February 2010 00:21 | Hi,
LHE ideal track finder is an ideal track finder (no conformal mapping) inside lhe package
However, the SttHelixHit are coming from the (realistic) STT helix fit. This means that, if for some reason the stt helix fit failed, lhe will have no hits (or maybe wrong hits), and the track will be simply lost. This effect is more pronounced at low momenta, were for some unknown reason the Z coordinate had systematic deviations.
|
Ok, now I understand. As a proof of principle, would it be possible to have the option to take the MC-matched PndSttHits, which are - I guess - before the helix fit? I am trying to see whether we would could provide easily a more efficient track efficiency, albeit less realistic.
Johan.
[Updated on: Thu, 11 February 2010 00:38] Report message to a moderator
|
|
|
|
Re: Momentum resolution and reconstruction efficiency of LHE tracking [message #10172 is a reply to message #10169] |
Thu, 11 February 2010 09:47 |
Johan Messchendorp
Messages: 693 Registered: April 2007 Location: University of Groningen
|
first-grade participant |
From: *xs4all.nl
|
|
Stefano Spataro wrote on Thu, 11 February 2010 07:47 | As I as writing before, lhe is based in 3D points -> XYZ.
In the case of SttHit this information is not available, and the helix fit transforms straw hits into point of closest approach to the central wire, used by lhe for the finding.
Maybe one could write an "ideal" stt helix hit producer, which does not take the reconstructed PCA but the MC point -> points for the finder, hits for the fitter.
But I am not so sure if this is want we want/need.
|
I see. In any case we need to understand what the origin is of the loss in efficiency. I understand that most probable it is related to the (x,y,z) reconstruction for the STT that troubles the helix fit. It would, nevertheless, be nice to confirm it in some way or the other. What about comparing directly with the track finding code of Gianluigi? That is available now and should be easy to do, right?
Greetings,
Johan.
|
|
|
Re: Momentum resolution and reconstruction efficiency of LHE tracking [message #10173 is a reply to message #10172] |
Thu, 11 February 2010 10:05 |
StefanoSpataro
Messages: 2736 Registered: June 2005 Location: Torino
|
first-grade participant |
From: *to.infn.it
|
|
Lhe tracking does not substitute stt pattern recognition, but it runs after. It requires an already existing stt stand-alone tracking.
After stt pattern recognition, once you have a stt helix, then you use lhe to merge helixhit together with mvd and gem. This means that you are running two patern recognition algorythm, and two fits. Of course, if each algorythm has an efficiency of 0.9, the combined efficiency will be 0.9*0.9 -> 0.8. And you have to add the Z problem of the helix hit producer, which decreases the efficiency again.
For tpc one has only one pattern recognition, therefore the efficiency is higher. And there is no pre-pre-fit.
In this sense, the comparison cannot be straightly done, and one has to invent something different.
|
|
|
Re: Momentum resolution and reconstruction efficiency of LHE tracking [message #10174 is a reply to message #10173] |
Thu, 11 February 2010 10:43 |
|
Wow, you're answering fast. Ok, of course I did the same analyses with the TPC. The track reconstruction
efficiency is better like you said. It goes up to 90%, but expecially in the low momentum range you can see
the same behavior.
MVD+TPC+GEM, standard macros in tutorial folder
The blue criterion means that no track is produced. Very low momentum pions decay before a detector
is reached and as a result to less points (<3 MVD+TPC hits) are created for tracking. So this is should be
just a physical reason.
greetings,
David
[Updated on: Thu, 11 February 2010 11:12] Report message to a moderator
|
|
|
|
|
|
|
|
|
|
Re: Momentum resolution and reconstruction efficiency of LHE tracking [message #10183 is a reply to message #10181] |
Thu, 11 February 2010 15:19 |
Lia Lavezzi
Messages: 291 Registered: May 2007 Location: Torino
|
first-grade participant |
From: *pv.infn.it
|
|
Hi Alicia,
the cut for low momentum particles was put there because of some instabilities of GEANE for low momenta, but in principle GEANE can propagate also low momentum particles.
I will make some tests removing the cut to see what happens and to fix the crashes if/when they happen. Unfortunately I am working on several other things at the moment and I have to postpone a little this investigation (sorry!).
Quote: | Nevertheless even if the particle is losing energy
geane should be able to take care of that? or am i wrong?
So far i know instead of a helix, geane considers then a parabola(x,y,dx/dz,dy/dz).
|
Yes, GEANE follows the particle while it loses energy. The trajectory of the particle is made by pieces of helix (in each step GEANE tracks with an helix, but the different helices can have different radius of curvature, for example, to take care of the energy loss).
With GEANE you don' t have the track as a whole helix with fixed radius of curvature.
Concerning the "parabola", it is just a track representation (the parameters used to describe the track are the ones you wrote). You always can convert it to the helix parameters, but the important thing is that you have the description of the track in one point and the parameters don' t stay constant along the helix (while if you use the helix fit you have that the momentum, and so the radius, is constant along the track).
Ciao,
Lia.
|
|
|
|
Re: Momentum resolution and reconstruction efficiency of LHE tracking [message #10185 is a reply to message #10183] |
Thu, 11 February 2010 15:46 |
Anonymous Poster
|
|
From: *cern.ch
|
|
Hi,
GEANE can follow particles to lower momenta, as Lia says. I played a lot with that for another experiment. But the models with which GEANE can propagate the errors (multiple scattering, energy loss straggling) and the state vector (energy loss) are to be studies I think at these low momenta. I am very interested in the issue, since it has a big impact on another experiment (Belle II) where we will use GENFIT with Geane or with another track follower we are working on. Also there, the questions of how to treat material effects for low momenta is not clear yet.
Could the GEANE experts please maybe state their opinion on low momentum performance of the physics models inside GEANE? I would very much appreciate it!
Cheers, Christian
|
|
|
Re: Momentum resolution and reconstruction efficiency of LHE tracking [message #10188 is a reply to message #10185] |
Thu, 11 February 2010 18:51 |
Alberto Rotondi
Messages: 9 Registered: June 2007
|
occasional visitor |
From: *pv.infn.it
|
|
Hi Christian,
the physics contained in GEANE is the following:
Mean values: dE/dx is the same of GEANT3, that is valid down to 10 keV (Bethe Bloch, Tables and Bremmstrahlung). Multiple scattering has zero mean value. Magnetic field works for any energy. However, spiralization could create problems, depending on the particular geometry. This is, in my experience, the main problem, because the user has to manage missing planes, missing volumes and so on.
Errors: multiple scattering sigma should work for any energy.
At low energy the shape deviates from Gaussian, but only on the tails, and the effect should be small.
dE/dx sigma is calculated with the standard formula sigma^2=csi*Emax*(1-beta^2/2)
which works very well at low energy.
In conclusion, I think that the main problem is spiralization.
I do not exclude the existence of problems due to the
robusteness of the program in dealing with a complicated track
geometry, but this should be verified in practice.
Best regards
Alberto
|
|
|
Re: Momentum resolution and reconstruction efficiency of LHE tracking [message #10202 is a reply to message #10172] |
Sat, 13 February 2010 23:06 |
Gianluigi Boca
Messages: 177 Registered: March 2004
|
first-grade participant |
From: *gsi.de
|
|
Sorry for the late reply.
Yes it is possible now to use the real pattern recognition for
the Straw Tube system.
As I have already said in my previous message, one example is in $VMCWORKDIR/macro/stt/runreco.C
where one needs to substitute
PndSttTrackFinderIdeal* sttTrackFinder = new PndSttTrackFinderIdeal(iVerbose);
with
PndSttTrackFinderReal* sttTrackFinder = new PndSttTrackFinderReal(iVerbose);
This macro produces the famous Helix Hits with the statement
PndSttHelixHitProducer* sttHHProducer = new PndSttHelixHitProducer();
fRun->AddTask(sttHHProducer);
Notice that in this macro there is also Lia's fit to the tracks just after the
pattern recognition, with the statement
// trackfitting ....
PndSttTrackFitter* sttTrackFitter = new PndSttHelixTrackFitter(iVerbose);
PndSttFitTracks* sttFitTracks = new PndSttFitTracks("STT Track Fitter", "FairTask", sttTrackFitter);
sttFitTracks->AddHitCollectionName("STTHit");
fRun->AddTask(sttFitTracks);
From the study that I am going to show at the next EVO meeting, I conclude
that for muon tracks, the resolution on the Momenta component are in general better
after Lia's fit compard to the values obtained directly by the Pattern Recognition
(with the exception of the resolution on Pz at momenta < 1. GeV/c).
Therefore today I think that the best is to use the real pattern recognition and then Lia's fit.
Typically the resolution on the Transverse Momentum ranges from 1.5% for 300 Mev/c
total momentum muons, to 20% for 10 GeV/c muons.
Typically the resolution on Pz ranges from 5% for 300 MeV/c muons up to 34% for
10 GeV/c total momentum muons.
Consequently, I would expect a not so bad resolution on the J/Psi mass after
real pattern recognition + Lia's fit.
One comment on the resolution obtained in the J/Psi channel using the Ideal Straw Tube pattern recogniton.
Lia-Susanna-Alberto last year obtained a good resolution with the Ideal Pattern
recognition + fit with muon tracks of different total momentum.
Now David Pohl claims he obtains a quite worse result.
I think that in his study the J/Psi decays hadronically or maybe even in e+e- ??
That could make things worse.
I would suggest first of all to study the decay channel into mu+mu- and see what happens.
In that case the results would be directly comparable to those of the Pavia group.
Gianluigi
Johan Messchendorp wrote on Thu, 11 February 2010 09:47 |
Stefano Spataro wrote on Thu, 11 February 2010 07:47 | As I as writing before, lhe is based in 3D points -> XYZ.
In the case of SttHit this information is not available, and the helix fit transforms straw hits into point of closest approach to the central wire, used by lhe for the finding.
Maybe one could write an "ideal" stt helix hit producer, which does not take the reconstructed PCA but the MC point -> points for the finder, hits for the fitter.
But I am not so sure if this is want we want/need.
|
I see. In any case we need to understand what the origin is of the loss in efficiency. I understand that most probable it is related to the (x,y,z) reconstruction for the STT that troubles the helix fit. It would, nevertheless, be nice to confirm it in some way or the other. What about comparing directly with the track finding code of Gianluigi? That is available now and should be easy to do, right?
Greetings,
Johan.
|
|
|
|
Goto Forum:
Current Time: Sat May 04 16:31:21 CEST 2024
Total time taken to generate the page: 0.01041 seconds
|