As we move toward breaking ground in version 2.5, we will be taking a hard look at algorithms used for auto focus. This will start, of course, with star (centroid) detection. As many of you know (especially if you own a scope with a secondary obstruction), SGPro does not handle detection well here (donut stars are detected as several smaller stars with a lower average HFD, then when the donut hole disappears, the HFD suddenly shoots up).
Anyhow… it would be very helpful to us if those of you who own gear capable of creating this type of data were able to share it with us for the development process. We provide an option to do this that is pretty painless, but you will need to upload and host them long enough for us to download them. Regardless of the success of the current method, we just need the data. To grab data from an AF run, go to AF settings, provide a path and check the box:
Three star fields progressively harder (fewer bright stars). I deliberately started each AF run 240 steps outside the focus point so that AF’s initial outward movement went right to donut land.
I’ll load some in a dropbox Ken - one thing I have found, your current AF routine is considerably more effective if I bin the AF exposures 2x2. I’m using a RCT, which has a large central obstruction (probably the worst case) and my unbinned resolution is 0.5arcsec / pixel.
I’ll try and do both 1x1 and 2x2 AF routines, weather permitting, to show the difference.
do you still need donut AFPacks? i seem to have a case where autofocus fails, however, i’m not 100% it’s simply because of the algorithm. could be the focuser slipping.
can the AFPack collection be turned on at the start of the night and will it then save all focus runs? the target rises too late at night for me to do anything manual.