Pathlength Corrections at CalDet
The 92% Solution - ds ~ ( (ds/dz) - 1)*0.92 + 1
- Previously we have shown that using this empirical correction the
pathlength dependance can be removed to a few percent
- Jim tried to use this at the FarDet, but the 92% value did not work
- The reason is that the value is governed by the hit selection criterion
- At CalDet we only take hits that do not appear to corner clip i.e.
only one hit strip along track per plane
- At FarDet, all hits associated with track are used
Clearly at CalDet, we expect the correction to only be a small deviation
from ds/dz, hence 92%
Colours used throughout:
Even Planes - Red (uncorrected), Magenta (corrected)
Odd Planes - Green (uncorrected), Blue (corrected)
(n.b. for all plots, we do not sum the charge from both strip-ends, we
fill each strip-end individually. Also, we plot truncated mean
instead of mean)
Attempting to impose a similer hit selection as at FarDet, we include all
hits associated with track. We then see that 92% does not work, and that
50% is more realistic. Although it is clear that this is not a viable correction
for this hit selection.
Model Based Parameterization
We then tried to make similar plots to those shown by Jim in the previous
talk but with higher statistics. Using all hits associated with track, and
using the "model based parameterization" with the same a=2.
Back to No Corner Clippers...
Zooming into our Best Correction so far:
Conclusions:
Hit selection is very important when trying to arrive at a pathlength
correction
Need a coherent policy.
Applying correction based on removing corner clipping hits corrects to +/-
2% over full range of tracked pathlengths
Parameterized approach based on including all tracked hits corrects to +/-
6%, although restricting dsdz can improve this
Is it possible to demand only non-corner clipping hits at FarDet and still
retain good statistics?
Other things to try:
2D Fitting function:
