Hi All, In connection with the most recent brat upgrade several changes that Kris had made to the global trackfitting was also included. As it turn out this result in a bug which we have now resolved. It also brings up several not so pleasant features how the FsTrack class is actual designed . Included are some comments on how this is organized and how it affects in particular what is actual stored in the dst class. The notes end with a set of conclusion that kris and I came to discussing this. Notes: I started reviewing the ffs and Fstrack classes. I think it is really a mess to be frank. FsTracks: TObject has pointers to bfsTrack and ffsTrack has variables for fTheta, fPhi and fD1SwimStatus the ffsTrack that belongs to a fsTrack has fTheta, fPhi , fTrackVertex (from glbtrack) fD1SwimStatus fD1EntranceLine fD1ExitLine fD1fidX/y Thus several variables are duplicates - and certainly not identical after TrackFitting, this is particular the case after refitting of the FfsTrack which changes certain parameters. For now the trackfitting set fSTrack - fTheta, fPhi,fMomentum and trkbeamVtx delete track if it does not swim trhough D1. a refitted ffsTrack will set -ftheta,fPhi, fTrackvertex ,fMomenum for the ffs part. The Bdst Classes works as follows. a) It first fills the ffs info in the dst (ie. swimstatus, entranceline,fidX/y ftheta,phi, mom) b) the the ffstrack matches a bfstrack it then overwrites the following entries: momentum, theta, phi and trackvertex information; Thus some information is from the ffs and some from the fsTrack. My particular concern is that the variables D1Swimstatus ,fFidX comes from the FFS track only and will NOT reflects the best values from the resulting fit. - replacing the values with the best from the fits for the following variables (in addition to theta,phi,..) fidX/Y, d1entranceline, d1exitline while KEEPING all the fT1Track,.. so a refit can be done if so wanted seems to me the better approach. Proposal for modifying tracking fit-code: Criteria: It should be possible to refit the tracks after the gtr's are created. The final information store in the bdst should be the best and most accurate information available. Action: When the global fitting is done for a full fsTrack the following variables will be set - fsTrack: fTheta,fPhi,fTrackvertex - ffstrack that belongs to this fsTrack o fD1SwimStatus, fidx/y, D1entranceline,.. Only ffs tracks that do NOT belong to a fstrack is refitted. After this refitting not only the Theta, Phi and momentum is updated but also the fiducial distances from the tracking back. In principle all the fiducial distances for all the magnets should be recalculated to adhere to the philosophy of having the best information available at the dst level. The ffstable in the gtr thus would end up with two classes of tracks those where the full spectrometer information was used to deduce fiducial cuts entrance line and those w/o. This I do not think is a real objection. The dst already have this grouping because e.g. the final momentum depends whether or not a ffstrack belongs to full track. Action from your side: Kris will modify the BrFsGlobalTrackFitModule according to the outline here, and we plan to make this the official version of brat by tomorrow unless we hear from you. This in part because we want to finish the generation of gtr and dst for the run-5 pp data. A large fraction of these have to be redone from the ltr level since there was no Vdrift calibrations for t1 from about run 14300. ---------------------------------------------------------------- Flemming Videbaek Physics Department Brookhaven National Laboratory e-mail: videbaek@bnl.gov phone: 631-344-4106 _______________________________________________ Brahms-dev-l mailing list Brahms-dev-l@lists.bnl.gov http://lists.bnl.gov/mailman/listinfo/brahms-dev-lReceived on Thu Sep 15 17:09:13 2005
This archive was generated by hypermail 2.1.8 : Thu Sep 15 2005 - 17:09:26 EDT