Demonstrated in DPOPS 200511.6 (and almost every other run). The RTC elapsed time counter should be linearly monotonic, but there are quasi periodic spikes. In the initial extraction in this run, one spike in particular occurred in uDACS_A_elapsed at:
T = 1589224601.5 unix time
T = 69401.5 seconds since midnight UTC
Three successive values were:
115022208 or 0x8019DB06
121103358 or 0xFEE337D7
115221577 or 0x4924DE06
Upon examining the raw data file (00/00/27.dat in this case), it was discovered that middle value was actually:
115121906 or 0x06DC9EF2
which makes more sense.
What this means is that the erroneous values arise not during data acquisition or realtime logging, but during extraction. I suspect this occurs in association with when a buffer in rdr or bfr fills up due to the higher speed processing. I suspect this will be difficult to track down, since any instrumentation is likely to slow things down to the point where the problem no longer occurs.
Demonstrated in DPOPS 200511.6 (and almost every other run). The RTC elapsed time counter should be linearly monotonic, but there are quasi periodic spikes. In the initial extraction in this run, one spike in particular occurred in uDACS_A_elapsed at:
Three successive values were:
Upon examining the raw data file (00/00/27.dat in this case), it was discovered that middle value was actually:
which makes more sense.
What this means is that the erroneous values arise not during data acquisition or realtime logging, but during extraction. I suspect this occurs in association with when a buffer in rdr or bfr fills up due to the higher speed processing. I suspect this will be difficult to track down, since any instrumentation is likely to slow things down to the point where the problem no longer occurs.