Why we need the timing cut in SVT reconstruction:

    :
    • PRV0 will be run with very large backgrounds. The SVT reconstruction software has been extensively tested with the TDR background frames (the only ones that have been available up until recently). See, for example, this talk for a recent summary of the SVT reco performance in the presence of backgrounds. The performance seems to be quite good. We now have a few "new" background frames (1 nTorr and 3 nTorr are available). It appears that with these frames there is a real problem with how much time the SVT reco takes in the overall reconstruction chain. Ed Frank made the following summary of the top CPU consumers with the 3 nTorr frames. SvtTrackFinder is on the top of the list!


    • What can be done about it? We can use the correlation between the digis time stamp (or corrected time) and the time of event. There is, of course, correlation between these quantities for signal, and none for background, as shown here. One can then cut on the digis' corrected time minus the expected corrected time (from the time of event), as shown here. Using the cut value of 1 speeds up SvtTrackFinder dramatically (see this summary). It also speeds up the overall reco process by a factor of 6.


    • How is this done? The cut is used in SvtHitReco to pre-select digis in time for further processing (clustering and tracking). The time of event is obtained from the L3 trigger.


    • Do we lose anything by using this cut? These plots show the number of tracks SvtTrackFinder found with and without the cut on the corrected time, as well as the track parameter resolution. The resolution appears to degrade on a level of a few percent, and we are losing some tracks (at about the same level). We should work on trying to get those few percent back.


    • What's the current status? SvtHitReco V00-02-43 has the cut on the corrected time turned on by default, with the cut value of 2. This tag is intended to be used for PRV0 only. The code that does not affect PRV0 may not be optimal (such as hitreco.tcl) in this tag.


    • One final point to keep in mind. We are expecting to see even higher backgrounds than 3 nTorr (5 nTorr looks more like it).




      • Natalia Kuznetsova natalia@charm.physics.ucsb.edu 3/3/98