Findtracks by identifying hits which lie along a straight line
Apply drift point calibration to each hit
Use single- or double-ended hits from planes in which only ONE strip
appears to be hit
Apply an attenuation correction to hits based on measured average response
as a function of position along length of strip
Apply a pathlength correction to hits:
Fill histograms per strip-end, reconstruct zeros for single-ended hits,
take 90% truncated mean
The problem:
When looking at path length corrected light output, the distributions do
not look flat:
Corrections shown here have been made just using ds/dz, which to first order
should be correct.
Stan's Comments:
...the method used to get strip by strip calibrations
together with elimination of all "corner-clippers", ie >1 strip in a plane
hits, has to
introduce biases and lead to incorrect answers. Let me give you three ways
(there may be others) how this causes problems:
1) Corner clipping hits in strips with lower light outputs have a lower probability
of being detected that for higher light output strips.
2) Tracks with the same energy loss but different incident angles wrt normal
have different probabilities for getting zero pes in an adjacent strip. If
due to geometry (eg vertical vs horizontal strips) the average pathlength
is different in different planes--> odd-even asymmetry.
3) The absolute calibration for all strips will be shifted downward since
you will be accepting some fraction of events that are corner
clippers and thus do not deposit all of their energy in the strip of interest.
3) is less important as long as the strip-to-strip calibration is carried
out correctly. Once the detector response has been flattened, the absolute
calibration can be achieved using a different method .
1) and 2) were expected and attempts have been made to correct for these
by calculating the average path length through a scintillator strip for a
particular set of track angles.
Factors affecting average path length:
Angles
Finite width of strip
Strip Light Output (->probability that any hit is observed)
Light Output of adjacent Strips (->probability of corner clipping)
Attempted to calculate average path length for our specific case with a tunable
parameter corresponding to typical strip light output:
Can also look at correction as a function of individual track angles, dxdz
and dydz:
Corrections good to +/-5% over angular range observed.
Alternatives
Stan's Solution:
...use all hits on a plane in the vicinity of a track.
Weigh the energy in different strips by the relative light output in each
strip, and assign total energy thus weighed to the strip with maximum energy.
This would have to be an iterative procedure, with all gains equal initially,
but it should converge very fast. Some cuts and/or corrections would be needed,
ie one would want to eliminate catastrophic energy loss events, correct for
edge effects, dead strips (if any), etc. I think that this procedure would have more or less the same problems
as have already been shown.
At the Far Detector:
Take only the largest hit in each plane. Doing this at CalDet and applying
a simple correction which accounts for finite strip width:
(ds/dz)*4.1/(4.1 + |dy/dz|)
This would introduce an even/odd plane asymmetry into the calibration.
What we eventually resorted to:
Realising that the pattern is more or less the same for even/odd planes,
green/clear readout and that light output looks fairly linear with pathlength,
the "92% solution" correction was applied. i.e. instead of correcting
by ds/dz, correct by 0.92*ds/dz:
