DchBgsModules
This page gives a brief description of the code to be contained within the package DchBgsModules. This package contains the so-called downstream modules which access fast Dch hits that have been stored in the event (via BgsDchSim code), perform fast
tracking, and output a reco-like track object.
- DcbmLoadVFTracking : This is a framework module to simply load the central tracking hits (eventually both Dch and Svt) from fast detector models (BgsDchLayerHit for the Dch info), call the algorithm to do fast tracking, and finally make a
list of reco-like track objects (TrkRecoTrk) in the event.
- DcbmVFTrackFit : The actual fast-tracking algorithm is implemented here. For a given particle, the information stored in the fast tracking hits (e.g. BgsDchLayerHit) is used to calculate the full 5x5 correlated error matrix and the
smeared track parameters (all calculated at the nominal interaction point). The relevant information contained in the fast hits includes: layer number (for geometry information such as stereo angle and angle the layer makes with the beam
axis), particle number, point of intersection with the layer (x, y and z), momentum of particle at point of intersection, intrinsic resolution of the layer, etc. The tracking philosophy follows closely what has already been done for the Aslund
simulation, using the trackerr code. Here is a summary of important issues:
-
-
- Only particles which pass minimum p_t and cosine cuts are considered! The idea is to do fast track parameterization only for those particles which penetrate to the Dch and are within the fiducial volume of the chamber (presumably there will be
no hit collection for generated particles which fail this test, anyway, so probably this is a redundant requirement that can be removed later).
- The B Field is considered to be uniform and in the z-direction.
- The fit begins with the last hit and moves inward, towards the interaction point. The error matrix is built up by considering the new information which comes with each hit. This information includes the intrinsic resolution
for the detector layer (this is most conveniently taken into account by adding the hit resolution to the information matrix, which is the inverse of the error matrix), the multiple scattering contribution to the error matrix (this is computed by
hand, for the moment, using knowledge of the amount of material between a given hit and the previous one) and the energy loss fluctuation contribution (this is small and is, for the moment, neglected). The Billoir method is used to
transport the matrix element from one hit to the next, until one reaches the first hit. Now one has the "final" covariance matrix. The amount of smearing to be performed on each of the five track parameters (d0, phi0, Omega (curvature),
z0 and tan(Dip)) is determined by: a) diagonalizing the error matrix and throwing Gaussian random numbers with width given by the eigenvalues of the diagonalization transformation, b) transforming these random numbers back to the
original, undiagonalized basis.
- The output of the "fit" is the error matrix and the smeared track parameters referenced to a single point, the position of the first hit in the hit list. What one really wants, from the point of view of reconstructed track objects,
are the error matrix and smeared track parameters wrt the origin. To do this, one must eventually incorporate the fast Svt hits, treating them in an almost identical manner as the Dch hits (there are some small differences to be taken into
account).
- Non-Active material must also be scored a fast hit and be used in this fast fit! A "hit" in non-active material (the Dch inner and outer walls, Svt supports, etc.) must contain all the same information as a detector hit--the only
difference is that the "intrinsic resolution" should be set at some arbitrarily large value (10^4 cm will do).
Release Information:
- First tagged version, V00-00-01, committed to cvs. Tested on Sun OS5 and AIX (some linking problems on AIX unrelated to DchBgsModules). The hits produced by the Dch layer model in BgsDchSim are used to make a list of
TrkRecoTrks. This list is placed in the event for use by other downstream modules and analysis.
- Works with release 6.5.2 and the following tags:
-
- BgsDchSim HEAD
- BgsFwkModule HEAD
- GenFwkInt V00-01-15
- DbiEvent V00-10-04 (I don't know if this one is necessary or not)
- V00-01-01
- This is the first tagged version for which the track smearing works and does something reasonable. Note that only Dch hits are used, and the track errors reflect this (the error on z0 is large, for instance). The code has been tested
within release 6.5.2 and using the same tags listed above (except BgsDchSim V00-00-09 was used instead of the head). This version is tagged to be included in the next release (whichever comes after 6.6.2). See the README file in
the package for more information about how to run and what the output should look like.
- Starting with this version, the track parameters and error matrix are expressed at the production vertex of the particle. This was done in order to treat daughters of long-lived decays such as K0s mesons correctly. The production
vertex is found by accessing the lists of GTracks and GVertices placed in the AbsEvent by the module BfwInputModule.cc in BgsFwkModule. Thus, the old G3Data structure is being used, which ensures compatibility with reco/Beta software. We may
decide to change this in the future.
Outstanding Issues:
- Need to incorporate the Svt, which means a list of Svt hits in identical format to that of the BgsDchLayerHits must be put into the AbsEvent. There already exists a method in the DchBgsModules package to use two lists and make the overall fit.
- As of yet, tracking inefficiencies are not taken into account. The easiest thing to do is to simply assign a probability to miss a hit and then throw out the hit measurement information accordingly. Once we decide to do this,
implementation will be simple.
-
The code is still in progress and the intent is to follow the example given by Andrei Salnikov in his DrcBgsModules package. Please contact carlod@slac.stanford.edu if there are any questions or
comments. In the future, I will try to keep this page reasonably up to date with information about the fast track fit.
For more information, look at the code and README file contained in DchBgsModules (once this package becomes available).
 This page created with Netscape Navigator Gold
|
