The on-orbit asymmetry calibration is based on 1.5 months of nomSciOps data collected by August 14, 2008 ( ~ 500runs: 236249505 - 240374588).
The calibGenCAL executable genGCRCalib.exe used both gcr and digi root files. Only events and hits selected by GCRCalib are used. Addition selection cuts were applied:
...
- for LEX1 range resolution is ~4 mm for old calibration and ~ 2mm for new calibration
- for HEX8 range resolution is ~13 mm for old calibration and ~9 mm for new calibration
There calibration.
There is significant population of events having much worse position resolution and creating non-gaussian tails. The origin of these effect is not clear. My initial suspect was shaped readout noise, but there is no correlation of position deviation with GemDeltaEventTime, as show on the following plot:
Another possible reason could be the bad tracker reconstruction causing the track to clip crystal corner and significantly increase the error in the pathlength correction and calculation of the mean position of the track segment inside the crystal.