- I have made numerous modifications to the event display, which
are not yet in the repository. See fcdfsgi2 ~belling/newestRootEventDisplay.tar
Among other innovations I modified routines to make the RZ display
actually work, though not as automatically as we might like. Since
the operation of drawing is to tell an object to draw itself on all
available frames, and since we want the phi slice to only appear on the
RZ plot, I had to add a work-around in which the phi-slicing is turned
on when you request a slice using the slice entry, and then when the
updates are requested only the RZ windows are updated, after which phi
slicing is turned off again. A hack, but unavoidable under the
- Ken Bloom used the muon high Pt sample to create a sample of Z
candidates with a stiff track headed in what was supposed to be the
BMU direction. Using these he determined that our efficiency was
about 61%. This seemed excessively low, so using a file of the events
that missed I scanned the events to determine where the losses arose.
(Hence the first bullet point above.)
The first thing I noticed was that the fiducial program wasn't
perfectly accurate, and a number of the events were actually CMX events
or CMX events which skimmed the outer layer of the BSU/BMU. In
addition, it sometimes pointed at the gap where the plug support sits
and we use short chambers instead.
The second problem was that 3-hit stubs were not being found correctly
when there was a 4'th hit nearby. This problem I fixed (twice--once
again this morning).
The third problem was a well-known geometric issue with tracks
which pass close to the wires in two cells tending to go through cell
walls and inter-cell gaps in the other two cells. I dealt with this
by writing a Stublet-finder which looked for hits close to the wires
in matching cells, together with an in-time hit in the BSU associated
with these cells. This took a little time, of course.
The results are summarized in this
description. The executive summary is that once we remove the
non-fiducial events our efficiency is 70%, and after I made the code
changes our efficiency reaches 87%.
Adding the stublet capability to our stub-finder does not appear
to overwhelm us with garbage. I ran on a few random B sample files,
and found 10 stublets in 8181 events, to be compared with the 12217
BMUS ordinary stubs found.
I also ran Bob's algorithm on the sample of missing stubs in badimu.root.
His algorithm found 12 stubs, where mine found 39, including all those his
Since these are more-or-less golden events (Z's), I conclude that it
is premature to switch from my old algorithm to Bob's.
Modified 9-Jan-2003 at 12:29
Previous meeting notes
Next meeting notes
Main slide directory
if you have trouble accessing the information on this page.