Autofocusing a fast focal ratio telescope problems

I am having problems generating V-Curves in SGPro for my (fast) F/3.8 Astro-Physics Riccardi-Honders Astrograph using the Starlight Express Focus Boss II electronic focuser. I have also noticed that there is no consistency in the focus point between different nights. Sometimes the focus point (tested with a Bhatinov mask) is 12,000 steps, others 16,500 etc. From examining the Half Flux Radius (HFR) manually through the RH305, to reach the CFZ I need between 10 and 20 steps to start seeing differences. I have no problems controlling the focuser from SGPro in temperature compensation mode where the delta is c. -25 steps each 1 degree (F) drop. I previously used exactly the same Focus Boss II on an F/7 refractor and had no problems generating V-Curves.

On a 7-step auto-focus run I expect the focuser to rack out 3 steps and work its way back in past the focus point. The first exposure sample is correct, but successive samples start going all over the place. My graphs look more like a W than a V. I have tried all sorts of combinations of step and sample size with no luck. The Astro-Physics 3.5” rack and pinion focuser cannot slip, so I am trying to eliminate other variables.

On the Astro-Physics User Group a comment was made that auto focussing in SGPro can be problematic at fast focal ratios and that a new beta is taking another approach. Does what I am describing seem familiar to anyone?

Your step size might be one thing to check if you havent done so already. At F3.8, the CFZ might be as little as 30 microns depending on the pixel size.

I’ve sometimes had trouble with strange shaped V Curves. One approach might be to start with a well focused position (using Bahtinov mask etc.) and try with different step sizes.

I’ve noticed that when you are significantly away from correct focus point, the HFR calculation isnt robust, i.e. very far away doesnt automatically have a worse HFR than just far away.

Regards,
–Ram

Thanks and yes done all that. I have tried a whole range of step sizes from 2 to to 100. Before doing so, I also used a Bhatinov mask to manually ensure I was in the CFZ before running the autofocus routine. I inspected the HFR in the image as well of course. Through manually examining stars HFR at different focus positions, I have a good idea that the step sizes I need are in the 10 to 20 range. (When the focuser is about 16,000 steps out from the zero point. Camera pixel size is 9 microns.

What is notable is that if I start from a point tested with a Bhatinov mask, the auto-focus run seems OK for the first sample on a 7 step run, the second sample will often get nearer focus as expected, the third will often go “backwards” out of focus, with a forth sample getting nearer focus again as if the software is trying to correct itself. This is the “W” type graphs I am getting. I know this seems like the step size is wrong, but when I am down to values of between 2 and 5 there is not much left to work with.

I have used SGPro with an F/7 telescope with exactly the same Focus Boss II unit and not had any trouble in generating V-Curves. It is only when I went to F/3.8 that these problems appeared. The only thing I am left with, is either (i) some mechanical slippage issue though the AP rack and pinion focuser cannot slip, or (ii) some software issue between the Starlight Instrument Focus Boss II and SGPro at certain tolerances.