Statistical view on SGP focus

Sequence Generator Auto Focus
1

I have been experimenting with SGP focus vs focusmax focusing.
I use my 300mm f/4 newtonian with a QSI660 camera.
I find that the SD of repeat focusing is about 4 times higher with SCP focusing than with FM focusing. I tried several combinations of SGP settings but ended up with 9 points 12 steps, Smaller steps gave much more variation. When I measure the CFZ of my scope it is around 8 steps. One step is 4 my (Moonlite focuser).
I’m quite sure that the precision of SGP focus could be improved it the software had an option that allowed using averaged HFR of several images at each step. I once evaluated FM and found that the number of samples at each datapoint was critical for reliable focusing. (http://www.focusmax.org/Documents_V4/Statistical%20view%20on%20FocusMax_00.pdf)
I realize that many stars are sampled in each image by SGP, but they may all be affected in somewhat the same way by seeing, i.e. they are not truly independent samples.
I think this option would give better focus for rigs with less than 1 arcsec/pix resolution.
I’m aware that focusing time will be lengthened by taking more samples at each point, but on the other hand time wasted on poor focus will be less.

Morten la Cour

2

Morten,

With my “slower” refractors SGP autofocus routine was good enough, at least comparable to the precision of a Bahtinov mask.

However I never managed to proper focus my 12inch F/4 newtonian or my Epsilon 160mm f/3.3 with SGP. No matter how much I tweaked the focus settings, the routine was not repeatable. I also found a similar mix of setting to be “better” 9 points with big steps in between. But that was not good enough for my systems. Granted, the Epsilon is very hard to focus even with a Bahtinov mask, the only way I can focus it is with FocusMax or with a GoldFocus mask.

For that particular reason I abandoned SGP and moved to another solution the plays with FocusMax.

I really miss all the integration and easy of use of SGP but sharp focus is paramount to have good images.

I still use SGP for planing mosaics tho.

Cheers,

Jose

3

HI Jose

I think some averaging might save the algorithm. But the software is too closed.

Morten