| Re: concept of stt reco [message #4949 is a reply to message #4838] |
Tue, 21 August 2007 12:20  |
Pablo Genova
Messages: 32
Registered: May 2007
|
continuous participant |
From: *PV.INFN.IT
|
|
Hi Piotr,
as Stefano pointed out - thank you Stefano - we were on holidays, so sorry for the late reply.
The currently up-to-date directory for straws is the stt one: stt1 contains the old Pavia code and stt2 the old Juelich code.
The stt package contains the merged code (stt1+stt2) and we are developing this package, so you should look in it to have the the most recent code.
Concerning the reconstruction in stt, it is divided in different steps and for each step a task is called (see in the macro file macro/stt/runreco.C).
The tasks are:
1) CbmSttFindTracks
2) CbmSttMatchTracks
3) CbmSttFitTracks
Tasks 1 and 2 come (almost) directly from stt2, and they perform an ideal pattern recognition: the MC thruth is used to assign points/hits to the tracks (in CbmSttFindTracks task, which calls CbmSttTrackFinder) and a matching between reconstructed tracks and MC tracks is performed (in CbmSttMatchTracks, calling CbmSttTrackMatch).
Then the track fitting is performed (step 3) in the CbmSttFitTracks task, which calls CbmSttHelixTrackFitter.
The real track fitting code is contained in CbmSttHelixTrackFitter: it performs a fit with a helix (as the name suggests), separating the problem in a circle fit in the plane orthogonal to the magnetic field and a straight line fit in z - track length plane.
In brief the fitting procedure is the following - for a more detailed explanation or visualization we can send you our presentations:
----- for the transverse momentum:
1 - pre-prefit: using the conformal mapping technique we fit the centers of the firing tubes to have a preliminary hypothesis for the track parameters.
2 - minuit: then we call Minuit to re-do the fit on the center of the tubes, using the results of the pre-prefit to initialize the parameters.
3 - intersection finder: we find the points on the drift circles which are the nearest to the reconstructed circle (obtained at point 2.). To do this we simply join the centres of the reconstructed circle and the drift circles and find the intersections between this segments and the drift circles (we can send you a plot to show this).
[By 'drift circle' we mean the circle with radius = drift radius measured by the straw, centre = wire centre]
4 - refit: these "new" points are fitted again using the conformal mapping technique.
By this procedure the parameters
R = radius of curvature
d0 = distance of the point of closest approach to the vertex
phi0 = corresponding angle
are found.
------ for the last two parameters:
z0 = z of the point of closest approach
tan(lambda) = tangent of lambda (lambda = dip angle of the track)
they come from the following procedure:
1 - zfinder: from the projection of the skewed tubes in the transverse plane and from the prefit in the transverse plane we infer the z coordinate of the hits. We use the hough transform to decide between the two hypothesis for each point (see the Dubna presentation by Pablo
http://panda-wiki.gsi.de/cgi-bin/view/Computing/PresentationsCMJuly2007
)
2 - zfit: a linear fit is performed in the z - track length plane.
On the issue of the "global tracking" (the last question) we do not know so much, since for the moment we mainly worked with the STT only.
We are currently working on the kalman filter method, for which the previously described fit is just the starting point.
Feel free to reuse our code as you prefer and to ask any question, Andrea and Pablo will attend the next Collaboration Meeting so we can also discuss directly.
Best regards, Pablo & Lia
|
|
|
|
GSI Forum
Search
Help
Members
Login
Home
