I first select hits with both |TDCZ| and |ADCZ|.LE.400, and also with
|ADCZ-TDCZ|.LE.200. This selects about 2/3 of the good hits and gets rid
of a lot of garbage. If the selected hits are less than 1/10 of the total
hits on that tank, I exclude the tank from my analysis. There is obviously
a serious calibration problem.
For each selected hit, I histogram the energy reconstructed at both ends
of the tank separately. Finally, if there is a high enough rate of
selected hits, I find the energy above which the hit rate is 3 mHz (2mHz
for verticals). If the total rate of selected hits is below these, I
exclude the tank from my analysis. The total number of tanks rejected for
a recent run (10838) was 150, less than a third of all tanks. Most of the
rejected tanks are verticals.
To reconstruct normalized pseudo energy, I use the formula:
IF (ENERGY_1.LT.30.0)
+ ENERGY_1 = 6. + 24.*((ENERGY_1 - E_FACTOR(IBOX,1))/
+ (30. - E_FACTOR(IBOX,1)))
IF (ENERGY_2.LT.30.0)
+ ENERGY_2 = 6. + 24.*((ENERGY_2 - E_FACTOR(IBOX,2))/
+ (30. - E_FACTOR(IBOX,2)))
where: ENERGY_1 and ENERGY_2 are the energy reconstructed for each
tank end normally.
E_FACTOR(IBOX,n) is the calibrated energy that gave the target
rate for this box and tank end.
Rejected tanks are reconstructed with 0 energy so they can be used for
rejecting multi-box hit, but they will not otherwise affect the burst
search.
For run 10838 with a live time of 23275 seconds, the following are my
statistics on hit rejection. Cuts are listed in the order they were done.
All E>6MeV E>8MeV E>10MeV E>12MeV
Total GC events: 252163 62713 23544 17910 16987
Multi box hits: 23474 17586 17052 16672 16377
TDC Over/underflow 23079 747 74 7 3
|TDCZ|>700 52108 10310 1761 324 68
|ADCZ-TDCZ|>200 48755 8880 1320 228 60
Trigger Cut: 201 149 146 142 142
ST wires Hit>=6 138 126 126 126 126
Near hits cut: 147 131 131 129 128
Energy Cut: 78450
Kept events: 24784 2934 282 83
Most importantly, the rate of clean 10 MeV background is only 0.012 Hz
and the background rate at 12 MeV is only 3.6 mHz. The Poisson probability
of a burst of 5 events at 10 MeV in 10 seconds is only 1.8E-7. Even a
burst of 4 hits, with a probability of 7.66E-6 would occur only once per
15 days. It would be possible to scan all the data for LMC sized bursts if
my method is valid.
The most important contribution to the reduced background rate is
throwing out the mis-calibrated tanks. The energy renormalization is also
important, but it primarily corrects a few tanks on the fringes. It's
amazing what good data we may have if we can weed out all the garbage.
Comments please...
Rich