As Wouter said, he, Jo and I are iterating on improving this algorithm, but would be happy to get additional testers.
I think things are better with the latest nightlies (the most recent one will show "v3" on the menu where you select the algorithm: "Linear (Experimental v3)"

One of the weaknesses is that you need to select an appropriate step size--ultimately we should have a "wizard" for that, but until things are more mature, I think it should be done by hand.

Søren: As I am not familiar with your particular configuration I can't recommend anything, but probably each step size might be the kind of small steps you'd use manually when you were somewhat close to focus. If you go too small, though, the algorithm can get caught in the noise of the focus system.

Søren: I suspect you weren't using the latest release, because there was a bug in there a week or so ago that caused that last output line. Are you sure you're using v3?

Pit: join in!

Anyway whoever does want to provide feedback, please have logging to a file turned on with at least verbose, file and focus logging enabled.

Note, you don't have to waste a nice evening to do these tests. If there are clouds, so it's not the ideal evening to image, but there's a patch of clear sky, you can try a few autofocus rounds pointing at that patch.

Thanks,
Hy

Hy, I'd love to, but as I wrote in my last mail (did you actually get it?) I'm in Sweden at the moment, without a telescope
So I'm only following the discussion, but cannot do much testing...

Ekos focus module says "Linear (Experimental) v3".

The initial step is described, in the "mouse over help", as a step that give a noticeable change in focus, so this is what I typed in, i.e. much bigger than I would use to pin-point the focus. I will try to go for a smaller value next time.

I will keep my nightly updated and try again when weather permit and report back.

I have updated everything and it shows "v3" and I still get errors in the log:
[2020-01-22T19:28:27.334 CET WARN ][ org.kde.kstars.ekos.focus] - Focus GSL error: invalid argument supplied by user

I also get these:
[2020-01-22T19:28:27.354 CET DEBG ][ default] - WARNING: bool Phonon::FactoryPrivate::createBackend() phonon backend plugin could not be loaded
[2020-01-22T19:28:27.354 CET DEBG ][ default] - WARNING: Phonon::createPath: Cannot connect Phonon::MediaObject ( no objectName ) to Phonon::AudioOutput ( no objectName ).
[2020-01-22T19:28:27.355 CET DEBG ][ default] - WARNING: bool Phonon::FactoryPrivate::createBackend() phonon backend plugin could not be loaded
[2020-01-22T19:28:27.355 CET DEBG ][ default] - WARNING: bool Phonon::FactoryPrivate::createBackend() phonon backend plugin could not be loaded

The V-curve looks kind of strange, with two sets of points at the same two x-values. It also writes only
2020-01-22T19:39:27 Focusing inward by 1250 ms...
2020-01-22T19:39:24 Focusing outward by 2500 ms...

in the log window. It does not change these values (initial step is 250, but I think it would be better with a larger value. But then I get errors from fcusb like Error: Invalid range for FOCUS_TIMER_VALUE. Valid range is from 0 to 5000. Requested value is 20000").

And it does not converge.

Polynomial focusing seems to work OK.

Can you please include the log and I'll have a look?

Also let me know the kind of telescope and focuser you're using.

BTW, is yours a timer-based focuser (which focuser is it)?
I've recently discovered that there is an issue in the V3 algorithm with timer based ones.

Thanks,
Hy

The logs simply repeatedly displays the messages I have shown above, I dont think you will get much more out of it...

Yes, it is a timer based focuser, fcusb focuser, so that may be the issue.

Let me know if I should try something specific next time.

Got it. OK, I just sent Jasem the fix I had asked Jo to try out with his FCUSB. May as well have it in the nightly for others to test, as I'm pretty sure these FCUSB focusers won't work with the linear algorithm the way I wrote it.
If/When you see v3.1 as the linear focus version, you'll have my change and can test again. I'll try and email out on this thread when I see that this change has been pushed.

This is the same focuser I am using, Hy, so it might be informative to compare the logs from the different sessions. It is interesting that 'The V-curve looks kind of strange, with two sets of points at the same two x-values.' That is also what happened to me sometimes, but not all the time. Obviously, that would mess up the calculation of the curve.
Your revised version of the algorithm may already have taken care of that. I will test once the clouds are gone here, but currently it does not look promising for the entire next week.

Jasem pushed the change phabricator.kde.org/D26854 it looks about 12 hours before I'm posting this message. So, it's currently in the latest source code, and I'm not sure if it's in the nightly build now or will be tomorrow .
Let me know what you think.

The "two sets of points at the same x-values" issue is not surprising, given that for timer-based focusers like the fcusb, I believe the x-values were wrong, and often not changing without the most recent fix.
Hopefully this won't happen anymore. I don't have one of those focusers, so it's hard for me to test the change without your help.

Hy

Great! As I remember, the x values were in the order of 5000, so very far from the other focus algorithms. Sounds like you are on track.

I will test as soon as I get some stars. Next two weeks seem completely cloudy though. Fortunately, the metrologists are often wrong

Hy,
The latest focuser algorithm (v3.2) works GREAT with my analog 3D printed focuser based on the Shoestring Astronomy FCUSB focuser.
This is what I get with a 135mm Nikkor lens attached to a cooled ASI183MM. I have a deep red filter in the lightpath, which includes Ha, S2 and extends into the infrared.



This is what my rig looks like (taken this morning after a night of imaging using a 55mm Nikkor Macro lens).



Simple solutions, great results.

Happy to share the STL files for the 3D printed rig, if anyone is interested.

Jo

PS: Here another image of the Linear focus routine, v3.2, this time using SEP, not Centroid.

Thanks for writing this code, Hy!
It is very fast and works great with such a simple focuser. Takes care of common backlash issues.
Jo