rolling bhabha calibration issues
quick recap, status in Feb/Mar
-accumulated 3e6 bhabhas
-calibration startet, constants have been calculated, cut out works, output of non validated channels works, everything went fine...until storing the constants
-storing crate 3 failed: "Failed to store crate 3: CalErr: failed, code 42, too many channels failed validation"
-didn't store a full calibration because of erroneous constants file (constant for one channel was set to 0), temporal db not deleted
-deleted temporal db by hand, loaded 1st run 6 calibration into test condition, deleted the temporal db again (to have the same base for the next calibration)
-comparison of the first 3 crates (quickly done in March):
-left plot shows absolute difference between online and offline calibration, right plot shows relative difference
-good agreement, channels with large differences originate mostly from failed validation (or # of direct hits) cut outs
-so far so good...
-between Mar, 12th and Apr, 7th again 3e6 bhabhas accumulated, calibration started again
-constants were calculated, cut out of failed channels:
-5 Xtals with too low number of direct hits, all known from last offline calibration
-all constants stored - no crash
-29 non validated channels, 27 in theta ring 56 (as usual...) and 2 new ones: 14:104; 17:42 --> offline calibration
-15 dead channels, 3 known and 12 others: 33-36:39-41 masked???
-temporal db content deleted automatically
-everything went fine this time.........almost
-PC farm showed strange behavior after deletion of temporal db: every run crashes and then runs fine, storing the collected bhabhas in the temporal db
-NOT a problem of the bhabha calibration code, Andy (Salnikov) found a bug in the database interaction code and provided a new tag
-new tag will go into release 22.1.1a, until then temporal db needs to be deleted by hand after every calibration to prevent the farm showing the strange behavior again :-(
-comparison between online and offline calibration soon
-comparison of online computed constants with old calibration constants
-looks like the actual constants are too high, maybe because I used old source constants for calculating them?!?
-plotted relative change of the new constants:
-new constants would lead to a decrease -> agrees with the expectation if the coanstants are indeed too high
-new offline calibration on monday (if there appear no computing problems up to then)
rolling bhabha calibration is on a very promising way