I’m also keen to see a periodic recentring option included in SGP, having suffered a lot of hours of wasted subs due to PhD walking the target off-centre when detecting intermittent clouds as the guide star. Here are my thoughts:
I would like to propose that the required plate solve could optionally use the most recently-download integrated image. This would save the time of having to do the “change to the Lum filter, take a short plate solve image, solve, change back to previous filter” routine. Integrated images are usually of much longer duration, and thus should provide a high SNR source for the plate solver, resulting in less chance of plate solve failures. If the plate solver was ASTAP then the time taken for confirming centring would only be a few seconds of additional time between integrated images, considering these images are almost always well-centred, so even checking every downloaded frame would not cost too much imaging time over the duration of a full night.
In terms of parameters, a centring tolerance in pixels would be required, but that parameter could perhaps be the same as the parameter currently used by SGP’s centring function. Another parameter could be a search radius (-r option in ASTAP I think). Because we only need to confirm whether the target centring is still within tolerance or not (a few 10’s of pixels usually), the search radius can be very small, and a small search radius will make the plate solve lightning fast too, so this parameter would be most likely independent of the equivalent in the Plate solving tab.
In terms of operation, if centring is confirmed to be within tolerance, the sequence simply continues. If plate solve succeeds but the centring is out of tolerance, immediately run a re-centring action and continue the sequence. If the plate solve fails … I’m not sure what to do here; it would depend on the failure mode:
(1) image centre is too far off target (outside the search radius of the plate solver);
(2) clouds prevented plate solve;
(3) mount is on-target, no clouds, but plate solve simply failed on that image (long focal length + narrowband image = not enough stars).
If (1), we would want the re-centring to occur immediately, but if (2) we would not want immediate re-centring; centring should be retried after a period of time has passed (e.g., retry every M minutes). If (3), just continue imaging. Unfortunately the cause of a plate solve failure cannot be determined. All of these three failure modes are rare events anyway, so perhaps an OK compromise would be to simply pass the image over to the blind solver. If it succeeds then re-centre (if necessary) and continue the sequence. If it fails, clouds are most likely the cause, so retry every Y minutes via the current centring function.
As mentioned above, there are still circumstances where an integrated image does not provide enough usable stars for plate solving (e.g., 6 minutes subs of Abell 1 @ 2.8m FL using a 3nm narrowband filter and ASI1600). For this reason, the option to recentre using the default recentring routine with a Lum filter should also be provided.
. I think you’re right about the time it would take in that ASTAP would take seconds. Now if SGP were to use multi-threading (maybe it does, I don’t know), plate solving could be spun off as a thread while other tasks are running like dithering or refocusing or whatever so that no time is wasted if centring is OK.