The problem: gaps / magic numbers
After running the three pedestal alignments steps, as described in the calibration wiki, the pedestal distributions looks something like this:
We can see some clear gaps in the distributions at 240 on both links. Note that this is for low statistics, but no huge difference is shown for higher statistics. We expect this distribution to be Gaussian.
Odd "magic numbers", resulting in ADC gaps, has been a consistent occurrence in multiple scans. See this issue. Dialogues have been held around timing being the most probable cause of the issue, in combination with poor parameter alignment. Poor timing could cause certain bits to be "skipped", resulting in the shown behavior.
There are four delay parameters on each channel, that delay the bit readout timing for certain groups of bits. A preliminary study of these parameters has not shown improvements (when testing 12 configurations out of 4096), although an increased effect has been noted. A full study of all configurations (with an implementation in the calibration procedure) is needed to affirm/refute the delay parameters as the source of the problem. If no improvement is seen, it begs the question if timing is the issue.
The problem: gaps / magic numbers
After running the three pedestal alignments steps, as described in the calibration wiki, the pedestal distributions looks something like this:
We can see some clear gaps in the distributions at 240 on both links. Note that this is for low statistics, but no huge difference is shown for higher statistics. We expect this distribution to be Gaussian.
Odd "magic numbers", resulting in ADC gaps, has been a consistent occurrence in multiple scans. See this issue. Dialogues have been held around timing being the most probable cause of the issue, in combination with poor parameter alignment. Poor timing could cause certain bits to be "skipped", resulting in the shown behavior.
There are four delay parameters on each channel, that delay the bit readout timing for certain groups of bits. A preliminary study of these parameters has not shown improvements (when testing 12 configurations out of 4096), although an increased effect has been noted. A full study of all configurations (with an implementation in the calibration procedure) is needed to affirm/refute the delay parameters as the source of the problem. If no improvement is seen, it begs the question if timing is the issue.