Definitely willing and going to test whatever you come up with
As for the HFR calculation, ISTR that one of the last commits had improved that and its stability/reliability.
(Also nice might be a separate tool to measure backlash, though it would already get much easier with an unidirectional focus method by just reversing the direction)
Sorry I've never gotten back to you on this--been working on something different.
That said, I have gotten a crude linear focus algorithm working on my development raspberry pi.
However, the reason I'm writing, is I've found that the mainstream focus is working a lot better for me now, and I wonder if that's true for you others.
I've switched to using the full-field focus (there was a bug in that scheme, but Jasem fixed that month or so ago), with Annulus of 20% & 75% -- so it uses a "donut" area,
not the inner 20% (to avoid a galaxy or nebula I might be pointing at) and not the outer 25%. I've been using about 10s exposures with moderate gain (75 on the ASI1600).
I don't think there's anything magic about these numbers, and it might be fine without the annulus.
In any event, it's been OK for me, are you folks still having issues?
Yes, I had also noticed that. The curves IMHO look smoother, so somehow there had been some additional noise that is gone now.
Haven't tried the field modus lately as I was looking at field curvature issues, and wanted to make sure the focus is set at some 2/3 distance to the border. Though that should in principle also be possible with filed modus and a narrow donut around 66%.
So no urgent issues anymore. But if you have the linear mode in place already - maybe that could easily be extended to a backlash measurement? (Assuming with 'linear' you mean just scanning through one direction).