OK, looks like confirmed Dither bug.
I took a look at last night's data (guided with FMA180 -- 180mm focal length, ASI678MM and Astronomik ProPlanet 645 IR-pass filter [near IR guiding with 200nm passband from 650nm to 850nm]), and plate solved the first 20 subframes. Image scale is 2.29" per pixel.
The max dither distance was set to 5 guide plate pixels, with a dither interval of 3 subframes.
The 5 pixel distance should limit the dither circle (and thus subframe offsets) to about 5*2.29" = 11.45 arc seconds.
This is what the plate solves gave (ordinate is arc seconds) as declination offsets from the original target declination.
The largest interframe dither is about 20 arc seconds (from -10 arc seconds to -30 arc seconds, between subframe 7 and subframe 8), or about the equivalent of 9 pixels -- greater than the 5 pixel limit that I had set. Who knows what the plate scale really is (probably much worse than the factor of 2.5x that I saw above), since the above is just a small measured sample. On top of that, just from these 20 subframes, the offset from the beginning target is over 30 arc seconds (13 pixels), again, already way larger than than the 11.45" dither circle that I have set.
The error will of course appear even more exaggerated when seen at the main camera, where the focal length is much longer.
Chen
