As the title indicated, I am having troubles getting autofocus to work in areas with a lot of stars (Pickering’s Triangle, Crescent Nebula, etc.) Things go as expected when I am out of focus. But as I approach focus, the smaller stars get clustered together and gives false HFR spikes. This is really irritating when the spike happens at the known focus point.
I have tried a variety of remedies such as tweaking the step sizes, adjusting star detection size, and changing exposure times all of which had no impact. I’ve tightened the nStep focus motor, I’ve tightened the focuser. Nothing.
Another odd thing is that half of the images that autofocus captures are darker and tend to get better HFR numbers. The other half look brighter and noisier. I created a dark library. This had no impact.
Has anyone else encountered this? Is there a work around such as being able to set the region to a less dense area within the image. Or glob forbid, automatically slewing to a less dense region for autofocus and then slewing back.
I’ve linked a 2.5 minute video that shows what is going on. The upward spike in the middle of the curve is near ideal focus. iCloud
This is a problem with longer focal lengths too with globular clusters, imaging M13 i have to sit and watch every autofocus run to check for a good curve.
A function for target offset autofocus should be added to help with these problems.
Yeah, I have noticed the same thing at 731 mm FL and 530 mm FL. But I don’t have nearly as much of an issue as your movie shows. Its clear that SGP is setting a focus point clearly some distance from ideal focus. Mine is more of a flattening of the “V” (I may work on step size a little bit.)
So I went through my SGP log for the last session and I see where the min star size along with step size, number of steps, etc are logged. I also notice that there is a max star size logged! It seems to always be 20 times the min star size. I don’t know where to change this setting (or if it is simply a fixed factor of min star size). But 20X seems a bit large to me.
Further, since a large max star size is only needed for the top ends of the “V” curve, it would be nice for SGP to automatically factor this max star size down in the middle steps (example as a feature enhancement, for a 9 step program, max star size could be factored down on steps 3-6). This might help avoid close stars clustered into one HFR star reading that is essentially bogus.
Another thing I noticed in the log is that the is an entry for every auto focus point as well as the verification image stating “Calculating HFR with sample size: 3”. I don’t get this since there are 300 star detected. I don’t know if there is a setting for this and what happens if one of the high HFR bogus clusters is included in the calculation.
I would sure like to know if max star size and sample size for HFR calculation is independently settable anywhere.
Browsing for this issue on this forum someone had the idea to raise the exposure from 4 seconds to 10 seconds and to bin at 1x1 instead of 2x2. I tried this and it completely fixed the curve. I still think it is a software problem but for now it’s manageable.