Segment pattern mask scan instead of requiring layers 2 or 3.
Routines affected: ana_segm.F sda_dinit.F analink.inc
-------------------------------------------------------------
I've checked in an improvement to the segment finding code in recsis. The
change involves the entry test condition for segment pattern searching.
Previously, we were initiating segment pattern searching if there was a
hit in layer 3, or a hit in layer 2. There were two classes of segments
that were not being found this way:
1) segments with no hit in layer 2 or3 (obviously)
2) steep segments with no hit in layer 3.
Class 1) is important because we are only requiring 3 or 4 hits per
segment and class 2) is important because we were missing some
steep segments that only missed a hit in layer 3.
The new technique is to use a mask that represents an .or. of all segment
patterns in the dictionary to sweep down the super layer looking for at
least min_hit_seg layers with hits that lie inside the mask. I believe
now we should be finding all segments that we can given know parameters
like min_hit_seg.
I've observed an increase of about 15% in the number of tracks found in
hit based tracking. Some of these are certainly multiple found tracks for
the same physical track, but I've seen a 16% improvement in the number of
single tracks.
# of tracks % improvement
per event with new segment
searching
1 16
2 33
3 11
4 10
5 0
this is based on a study of run 3741.
I've not had a chance to examine the quality of the new tracks found.
\ R A Thompson /
/ Dept of Physics and Astronomy University of Pittsburgh \
\ Pittsburgh, PA 15260 /
/ richardt@cebaf.gov 757-269-7475 \