Similar to  Fuyao's timing investigation alog68801, I checked that the DAQ framewriter's GPS time is internally consistent between the frame file naming and the data timestamps within the frame.

Details are in this wiki

An overview of the process:

1. On NDS0 I took a full frame file at random, in this case H-H1_R-1365966080-64.gwf. I extracted the data for the H1:CAL-PCAL_IRIGB_DQ channel, which gave 16384*64 data points.

2. I took the first 16384 data points, which span the first second of the frame. I converted them into a CSV file and plotted them (see attachment)

3. By eye, I decoded the HH:MM:SS encoded in the IRIG-B signal (called Decode_UTC, leap seconds have not been applied)

4. I converted the frame filename's GPS time to True_UTC (leap seconds have been applied)

5. Verified Decode_UTC = True_UTC + 18

Result

In this case, Decode_UTC = 19:01:20, True_UTC = 19:01:02 which is correct.