I generated ntuples to study the correlation between reconstructed
thetaC and number of bounces or photon exit angle with high statistics
and results on those issues will be posted at a later time.
This HN is concerned with an interesting side effect / first application
of those ntuples: hunting down more cable errors.
The ntuples were produced for about 1.5Mio DIRC tracks using Elf in
release 8.6.1a (with corrected DrcMaxLikel), conditions from recoboot,
running on xtc files from post-October-shutdown data. I use the nearest
solution algorithm. Quality cuts on the track require the momentum to
be between 0.3-9 GeV/c and the number of solutions to be between 15 and
300. The photon quality cuts are |deltaT minus meantime| less than
6nsec and |thetaC(photon) minus thetac(track)| less than 100mrad.
That leaves me with about 48Mio hits.
As usual, I fit the thetaC residual (thetaC of tube minus thetaC of
track) per tube. The function used here is Gaussian plus a line to allow
for asymmetric backgrounds.
The high statistic allows me to plot the sigma of the Gaussian per PMT
only for tubes where the PMTs are *not* in the sector opposite the
bar box that the track went through:
http://www.slac.stanford.edu/~jochen/cableerror/occu_nosec_edi.gif
The scale is chosen so that tubes with a sigma of more than 15mrad
appear as black dots. One notices immediately a set of 19 neighboring
pairs of tubes with large sigma values. The distributions for the
tube highlighted in the previous plot can be seen here:
http://www.slac.stanford.edu/~jochen/cableerror/sec08105758.gif
The top plot shows the thetaC residual for all accepted hits while the
other three show the thetaC residual for tracks going through bars in
bar box 7, 8, and 9, respectively. It can be seen that the distributions
are shifted away from zero when the bar sector is different from the PMT
sector with the sign of the shift flipping. This can be understood by
visualizing that the Cherenkov ring segment becomes steeper in the
sectors one or two bar boxes away from the one that the track passed
through, thus giving us some sensitivity to cable swaps within a given
row. That sensitivity did not exist in the studies performed previously
where the statistics limited us to more horizontal rings in the highest
occupancy area.
List of candidates for cable swaps within a row (30 total):
+--------+--------+--------+--------+--------+
| sec | har | chan | col | row |
+--------+--------+--------+--------+--------+
| 1 | 13 | 12 | 22 | 22 |
| 1 | 13 | 13 | 23 | 22 |
| 5 | 0 | 26 | 0 | 21 |
| 5 | 0 | 27 | 1 | 21 |
| 5 | 8 | 18 | 10 | 12 |
| 5 | 8 | 19 | 11 | 12 |
| 6 | 1 | 12 | 2 | 10 |
| 6 | 1 | 13 | 3 | 10 |
| 6 | 4 | 19 | 5 | 19 |
| 6 | 4 | 20 | 6 | 19 |
| 6 | 4 | 21 | 6 | 20 |
| 6 | 5 | 22 | 7 | 20 |
| 6 | 12 | 2 | 16 | 10 |
| 6 | 12 | 3 | 17 | 10 |
| 8 | 10 | 57 | 20 | 37 |
| 8 | 10 | 58 | 21 | 37 |
| 9 | 0 | 4 | 0 | 10 |
| 9 | 0 | 5 | 1 | 10 |
| 9 | 0 | 24 | 0 | 20 |
| 9 | 0 | 25 | 1 | 20 |
| 9 | 0 | 32 | 0 | 24 |
| 9 | 0 | 33 | 1 | 24 |
| 9 | 4 | 38 | 8 | 29 |
| 9 | 4 | 39 | 9 | 29 |
| 9 | 6 | 52 | 12 | 33 |
| 9 | 6 | 53 | 13 | 33 |
| 11 | 3 | 38 | 5 | 30 |
| 11 | 3 | 39 | 6 | 30 |
| 11 | 6 | 18 | 7 | 15 |
| 11 | 6 | 19 | 8 | 15 |
+--------+--------+--------+--------+--------+
List of swap candidates within a harness, resulting in
col&row swaps (8 total):
+--------+--------+--------+--------+--------+
| sec | har | chan | col | row |
+--------+--------+--------+--------+--------+
| 9 | 5 | 2 | 5 | 2 |
| 9 | 5 | 3 | 4 | 3 |
| 10 | 2 | 4 | 1 | 3 |
| 10 | 2 | 5 | 1 | 4 |
| 10 | 0 | 52 | 1 | 34 |
| 10 | 0 | 53 | 0 | 35 |
| 11 | 1 | 50 | 3 | 35 |
| 11 | 1 | 51 | 1 | 36 |
+--------+--------+--------+--------+--------+
A special case seems to occur in sector 2 where 4 tubes may
have been swapped in a more complicated way, resulting in shifts
of the thetaC residual mean of up to 30mrad:
+--------+--------+--------+--------+--------+
| sec | har | chan | col | row |
+--------+--------+--------+--------+--------+
| 2 | 2 | 49 | 4 | 35 |
| 2 | 2 | 50 | 5 | 35 |
| 2 | 3 | 51 | 6 | 35 |
| 2 | 2 | 51 | 6 | 36 |
+--------+--------+--------+--------+--------+
On the next ROD (currently scheduled for April-27) we are going to
have the BW doors open to unplug the HV cable of the two light emitting
tubes. That should give us a chance to also verify if the 42 candidates
listed above are indeed miscabled. If they are, the plan is to leave
them in their current state and change the database to correct the cable
errors for the entire data set consistently.
Joe
Created: 05/19/2000
