I have mostly completed the transition away from the old imu
area to use of the Muon/MuonSim/MuonObjects/MuonGeometry areas.
This includes a revised stub finder, which uses MuonXZHit and
MuonSignedHit collections. It generates an internal collection
of BMU_Candidate objects, which are then transformed into a
algorithm remains pretty much the same as it always was. The
BMU hits are broken out into 8 groups, one for each layer/side.
The fitting routine uses a collection of 4 hits, which is managed
by an object that takes care of details such as 'no more data',
how to get the next collection (find the lowest phi and select the
next hit from that collection), and determines whether or not there are
enough hits in the phi window to qualify.
There are hooks to allow throwing out stubs with Z's too close to
the rear, but these won't be implemented until more studies are done.
I incorporated Bob Handler's time offsets into a calculation of
the Z-position of the hits. This process (done in the DtoE) determines
which of two hits is the original and which is the echo hit in the
ganged cell, and marks the latter as noise. Thus the hit collections
found with the new algorithm contain fewer hits than the old, which
did not distinguish them. This changes the results slightly for
real data, which contains more noise and accidentals than monte carlo.
Results from data suggest there are still a few issues to be resolved.
If both the inner and outer Z are found in a stub, they agree
to within 9.6 cm by the gaussian fit.
However, I see some structures in
phi vs impact parameter for 3-hit
fits which suggest that perhaps there are some crosstalk issues
remaining on the west side.