bfv Didn’t really notice this behavior when shooting 5 minutes subs of nebulae. Now that I’m shooting 2m subs of a galaxy I had to cancel my entire session because my main (off-center) target was slowly drifting out of the FOV altogether.
It is possible that once in a blue moon, dithering will move the center of the FOV quite far away, even though each dither has zero mean. (Dithering involves something called the Markov Chain in probability theory where each movement is independent of past movements, and the Drunkard's Walk problem tells us it is possible for the drunk to eventually walk off to infinity, but with very low probability. Not kidding -- check it out in Google :-)
However, for you to consistently see the center of FOV move far away suspiciously point to each dither not having zero mean, or it is not memoryless (i.e., the direction of dither is influenced by a previous dither). Or perhaps even worse, dithering is not obeying the 2 pixel maximum (that the OP had set). (The 2 pixels movement is determined by mount calibration -- so if calibration is wrong, it would lead to a dither problem too.)
You might want to find the PHD2 log (it should be the ASIAIR storage -- USB memory if you are using USB storage) and look at the dither commands to see if it shows any anomally. Or post it here so the rest of us can help.
Even if the movement is not completely random, but has a bias (for example, a random number with a bad sign), and your camera is more than 1000 pixels wide, it still should have taken more than 500 sub frames (couple of hours) to move a star out of frame, if like the OP, you have chosen a max dither of 2 pixels.
If your target moving faster than that, I would check the log for any anomally with calibration (i.e., the part where it figures out the mumber of arc seconds per pixel).
FWIW, unless I only take a dozen subframes, I have been dithering (even way before ASIAIR) only once every 3 or 4 sub frames. I also don't dither more than 2 pixels at a time.
If you are doing so, try not selecting the large (e.g., 10 to 30 pixels) dithers. Most cameras don't need that much. 2 pixels should be sufficient for most cases. That should keep things in check, even if there is a bug.
Anyway, check the log, it should tell us what is happening.
Chen
