Study of swapped PMT signal cables using cosmic data
Update on March 24. All miscabled channels identified below have
been corrected.
The sector summary plots were updated to the 1.85M statistics for sectors
0, 1, 9, 10 and 11.
We have studied the single tube Cherenkov spectrum for some 660k events
taken from Jan.14 through Jan.31 in some detail.
A loose time cut of +- 25nsec for the difference between measured
hit time and most probable event time was applied. For each tube in sector
10, 11 and 0 all hits were added up and the difference between the measured
theta_C value and the expected (muon) theta_C value was plotted and fitted
with a double Gaussian. While the detailed study of resolution and background
per tube is ongoing, one interesting feature of these distributions, their
potential for detecting mis-cabled PMT signal cables, is dicussed here.
If the cables of two tubes are swapped, the oberved Cherenkov
spectrum should be shifted by, e.g., about 25mrad if two neighbouring tubes,
situated perpendicular to the Cherenkov ring, are swapped. One such example
for channels with (most likely) permuted cables is harness 2, channels
27 and 28 in sector 10, as shown below (entries versus theta_C(measured)
- theta_C(expected) [mrad]). Note that the mean values of the two channels
are shifted in opposite directions. A compilation of all 896 single-tube
theta_C distributions for each of the 3 sectors studied can be found as
PDF files:
The group that Stefan mentioned in an
earlier
HN (sector 10, harness 6, channels 16-31) which correspond to TDC chip
1, show up clearly as a very broad, "strange" distributions. It is not
obvious how those distributions can be explained by swapped signal cables.
The figure below shows the mean values of the narrow Gaussian as a function
of the channel for each harness of sector 10. Swapped signal cables have
a significant zic-zac shape (areas marked in blue), whereas the large fluctuation
are due to bad fits (the broad and narrow gaussians are exchanged) or bad
statistics. We have studied all those cases in detail.
Summary for all five sectors as GIF files:
After studying all Delta(theta_c) distributions for the sectors 0, 10 and
11 we obtained the following results:
| Sector |
Harness |
Channel |
Comment |
| 0 |
12 |
38 <> 39 |
swapped (incorrect label) |
| 10 |
1 |
11 -> 12, 12 -> 13, 13 -> 11 |
swapped |
| 10 |
2 |
27 <> 28 |
swapped |
| 10 |
3 |
46 <> 47 |
swapped |
| 10 |
6 |
24 <> 16, 25 <> 17, 26 <> 18, 27 <> 19, 28
<> 20, 29 <> 21, 30 <> 22, 31 <> 23, 16 <> 24, 17 <>
25, 18 <> 26, 19 <> 27, 20 <> 28, 21 <> 29, 22 <> 30, 23
<> 31 |
swapped in groups of 8 channels |
| 10 |
7 |
44 <> 45 |
swapped |
| 10 |
12 |
38 <> 39 |
swapped (incorrect label) |
| 10 |
13 |
34 <> 36, 37 <> 39, 40 <> 42,
43 <> 44, 45 <> 47 |
swapped |
| 11 |
3 |
24 <> 25 |
swapped |
| 11 |
6 |
4 <> 5 |
swapped |
To estimate the influence of the candidates for miscabling of the signal
wires, I cut against produced the delta(theta_C) = theta_c(meas) minus
theta_C(exptected) distribution for the same 660k events used in the original
study. The time cut was +-10nsec for the difference between expected time
and most probable event time (not +-25nsec as in HN 101). As before, the
algorithm used was HitChiSq. The "swapped cable"-cut was applied against
all listed tubes, except the known swapped HV groups in sector 11 (H4,5,
ch40-47). Of course, to really determine the influence of the miscabling
on our cosmic data set, they should be removed from the hitlist before
the ring fit. This analysis is just an attempt to get a rough estimate
of the effect. The result of the double-Gaussian fit to the data after
and before the cut and the "suspicious channels"-only is shown:
The background-to-signal ratio (defined as the ratio of the integrals
of the two fitted Gaussians) is reduced from 0.65 to 0.52 with this set
of cuts, corresponding to a relative reduction of the background by 20%.
Andreas and Joe
Last modified: Mon Jun 2 17:55:52 PDT 2003
|
