For multiple years now Keck and BICEP3 mounts have been scanning back and forth all day every day according to scan trajectory file az-2p8-1p4b.scan, where the file naming indicates velocity of 2.8 deg/s with turn-around acceleration of 1.4 deg/s\(^2\). The file contains a start, a body and an end, and the numbers are degrees of offset on the three axes (with only azimuth varying). When told to do multiple repetitions the control system loops the body resulting in only one zero value per rep. We nominally loose 12 seconds to accel/decel out of 52 seconds total for an efficiency of 40/52=0.769 The scan looks like this:
Loading some real data and looking at the samples selected for binning into the maps results in the following plot. The gap between the red blocks are indeed close to 6 seconds.
Clearly one could be a bit more aggressive and push closer to 4 seconds of dropped samples per turnaround without mixing scan speeds much. Maybe mixing scan speeds doesn't matter much and one could use even more of the turnarounds. In many years of BK we have never tried it - but it could be investigated the existing data.
It would be great if POLARBEAR and ABS could share some similar info about their real scan trajectories and data selection.