GSI Forum
GSI Helmholtzzentrum für Schwerionenforschung

Home » PANDA » PandaRoot » Tracking » TpcRiemannTrackingTask - ideal PID now working - please rerun reconstruction
Re: TpcRiemannTrackingTask - ideal PID now working - please rerun reconstruction [message #12440 is a reply to message #12433] Tue, 02 August 2011 20:49 Go to previous messageGo to previous message
Sverre Dørheim is currently offline  Sverre Dørheim
Messages: 23
Registered: October 2008
Location: Munich
occasional visitor
From: *natpool.mwn.de
Hello Stefano, Lia

To study the problem a bit closer I took a look at the output of the fits in the TPC+MVD+GEM before the conversion to PndTracks.
index.php?t=getfile&id=6562&private=0
In this plot one can see the start-position of all tracks in the XY-plane, and one can clearly see some starting points which are way off. This is of course ALL tracks, also the tracks where the fits went wrong.
Looking at only the tracks with
track.getCardinalRep().getStatusFlag()==0, where the fit converged I get something like this, where all start values are within a reasonable range.
index.php?t=getfile&id=6563&private=0
The rough binning is to make the outlyers a bit more visible.

This led me to investigate if this flag is checked at any time later in the analysis chain and as far as I can see this information is lost when converting the GFTrack to a PndTrack, where the PndTrack gets a new flag which is:
-1 if NDF=0 or Genfit threw an exeption when calculated the momentum at the first and last detector plane
1 otherwhise

I did the same plot for all GFTracks where NDF!=0 resulting in:
index.php?t=getfile&id=6564&private=0
Where clearly can see that the worst outliers are gone.
Since only this flag is awailable atthe PndPidTrackInfo level I implemented this check before the extrapolation:
Bool_t rc=false;
if(track->GetFlag()==1){
rc = fPro0->Propagate(helix, fRes, fPidHyp*charge);
}
This was then tested on 1000 events, whith no crash in the PID-macr. This is probably the quickest fix as one only needs to rerun the PID-Macro.


Best regards
Sverre


Ps:
Looking at the statistics in the 3 plots:
All GFTracks:5113
statusflag 0:4132
ndf!=0 :4912

A lot more tracks get lost when cutting away events with statusflag!=0(ie something whent wrong) than just cutting away events with ndf!=0.

The effect of this should also be tested.
 
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Previous Topic: Update on eta_c reconstruction with STT and TPC
Next Topic: Run 951 GRID -TPC
Goto Forum:
  


Current Time: Fri Mar 29 02:34:57 CET 2024

Total time taken to generate the page: 0.01161 seconds