Background
Starting in January 2016 the readout timing of the pnCCD was incorrect for runs taken with the DAQ BEAM configuration. For runs taken with the NOBEAM configuration the timing was correct.
Affected experiments:
- amo1141
- amo0571
- amol941
- amok891
- amok991
- amol341
- amol311
- amo0651
Experiments not in this list either had no pnCCD in the data stream, or had correct timing settings.
Symptom and Workaround:
The pnCCD data has 4 panels of 512x512 pixels. Pairs of these panels are on a monolithic piece of silicon (1024x512) with no physical gap between them, however when the timing problem is present an artificial 12-pixel "gap" must be inserted using the psana geometry. Also, the outer 6 columns of each panel contains bad data and the pixels must be marked as bad in the psana bad pixel mask. Both of these workarounds must be done on a run-by-run basis, depending on the timing parameter setting.
The following event demonstrates the problem and the geometry workaround (from amok8916 run 154, event with timestamp 1467157117/339649197 and fiducials 118176). In the first picture the 4 blobs should be in a straight-line. The pnCCD hole is at the lower-left:
After inserting an artificial 12-pixel gap the 4 blobs do form a line:
NOTE: affected experiments should not re-deploy dark-run calibration constants: it could overwrite the fixed bad-pixel-mask.
Overview
Content Tools