×

INDI Library v2.0.6 is Released (02 Feb 2024)

Bi-monthly release with minor bug fixes and improvements

*** Help needed to test changes to the Ekos Focus Module ***

  • Posts: 460
  • Thank you received: 69
I believe the Mac nightlies can be installed from binary-factory.kde.org/job/KStars_Nightly_macos/. There are occasional hitches in finishing the compile run.
10 months 2 weeks ago #92806

Please Log in or Create an account to join the conversation.

  • Posts: 992
  • Thank you received: 155
Great! I just saw that the latest nightly in the MacOSX Binary factory was may 11. I'll download it tonight.
10 months 2 weeks ago #92809

Please Log in or Create an account to join the conversation.

  • Posts: 593
  • Thank you received: 276
Yes all feedback greatly appreciated!

Just to set the scene… this is beta testing and the changes are quite extensive so make sure you can revert to your “normal” config / version in case of issues.
10 months 2 weeks ago #92815

Please Log in or Create an account to join the conversation.

  • Posts: 1009
  • Thank you received: 133
Hi John,

first, thanks for that massive work!
I had prepared a version right after you released it, but only last night had the chance to look at it under starry skies.
Not too much to report yet, as I did not dig into the changes and new options and only used the 'classic' L1P.
One thing I noticed was on the first AF run (started manually). I got an outlier just at the proper focus position. After that, it had stopped updating the fit curve and interpolated focus position. Instead it continued further into the wing of the hyperbola (more than usual, 8 points after the min instead of typically 6), and only then plotted the fit and went to the (correct) focus position.
Cannot tell if that repeated like that in other (automatic) AF runs, in the manual run it was sort of an irritation.
All the other (automatic) runs succeeded, even one in really bad seeing conditions - likely due to the new outlier rejection. I'm sure with the old L1P it would have started another full run, here it completed with 16 points. Nice!

I got a bit irritated by the tooltips in the Assist tab, hovering over 'Help' it said it would reset stuff to defaults....

Also, a comment about the backlash compensation tooltip/help entry:
You say one should only use one of them, and set the other to zero. I disagree there, and the reason is BL accuracy and filter offsets: It is correct that for the AF run only one of them is needed, but ISTR that we agreed earlier that it does not harm to use both. The problem is that you do want to start close to focus for the AF run to succeed with high probability. The BL of my setup is 84 steps. That is about the same as the filter offset between the blue and the red channel, and almost as much as the focus range (5 times 20 steps) that I use. So IMO it is crucial that the filter offset is applied including (coarse) BL correction, and as I understand right now the only way of doing that is with the BL correction in the driver. My experience with that however is that it is not too reliable for short tunings. When I was measuring it, I did not get a mere parallelogram for the steps-vs-distance curve, it always was slowly creeping up until the linear regime started. So for AF I use the additional overdrive on top of the driver BL correction (...just like the guy wearing both belt and suspenders :S )

And a question about the final point of your list: Overdrive focusing for all focus movements. That isn't in ATM, correct? That would eventually remove the above constraint. I had wishlisted that some years ago, but got told that this would have to be in (all the) drivers, not in the general INDI code. So I'm curious how you (inted to) tackle this.

CFZ:
Very nice addition. It computes a CFZ of 13 steps for my system, and suggests to set the step width to that. I have mine set to 20 steps. Partly coming from the old 'Linear' that would do the refinement at half that step size. Do you think doing the full scan at that step size is necessary? I'd definitely have to increase the number of 'steps out', substantially increasing the time needed for an AF run. It was my feeling that about twice the CFZ is sufficient to interpolate the focus position to well within a CFZ (even more with the new weighting and outlier rejection)

Cheers,
Pit

PS: Will continue tests tonight, though only in twilight to save the moonless imaging time... :blush:
10 months 2 weeks ago #92835

Please Log in or Create an account to join the conversation.

  • Posts: 593
  • Thank you received: 276
Hi Pit,

Thanks very much for the feedback!

If the graph stops redrawing the curve after each point that's usually because the curve fitter fails to find a solution. AF will move to the next point and try again. That usually fixes things. The intermediate graphs are just there to show progress to the user; only the last one is used to calculate the focus point.

FYI, if you want to review focus runs from a previous session you can use the Analyse tab and click on the focus event and it will display the datapoints on the focus graph. Useful for reviewing an unattended session in the morning.

Good spot on the tool tips on the Focus Advisor button - they are wrong. I'll fix those.

AF Overscan. With this release AF Overscan is applied to all focuser movements that Kstars Focus knows about. If you move the focuser outside of Kstars Focus then Overscan won't be applied.

So if you only use Kstars Focus to move the focuser all moves with have Overscan applied: AF runs (all algorithms), Goto Position button, Focus In/Out buttons, Adaptive Focus moves, filter offset moves. So you only need to set AF Overscan. Hopefully that's a win!

If for some reason you do need backlash as well as AF Overscan then its fine to use both as we discussed. But I'm hoping that it won't be necessary for most folks to do this now.

As far as issues with the focus starting position when changing filters, check out "Adapt Start" in the document at the top of this thread. Basically, if you change filters from Red to Blue, then if you have "Adapt Start Pos" checked then the start position for the AF run on Blue will be last successful run on Blue (rather than the last run on Red). It will also "adapt" that last successful focus run on Blue based on temperature / altitude differences between the last good run and now (if you configure parameters for this). The document has more information.

CFZ. There is a view that it makes sense to have at least 1 datapoint within the CFZ during a focus run which means Inital Step Size <= CFZ. This is why Focus Advisor suggests Initial Step Size = CFZ. Personally on my equipment I set a step size of about 1.5 to 2 * CFZ. With accurate curve fitting I don't feel its necessary to have a very small step size, but happy to hear what others think. I can easier change this recommendation depending on what feel.

I did include a focus walk called "CFZ Shuffle" (again see the document for more details) but this basically will take half step sizes around the CFZ during the focus run. So for those that could get benefit from having extra steps near the focus point but don't want many extra focus points (that slow the focus run down) this might be an option to try.

Of course, the Focus Advisor is just applying generic logic. More experienced users will get better results by playing around with parameters but the idea is to give some robust parameter suggestions for folks that are struggling to get a good focus.

Once again, thanks for the feedback and good luck exploring the new functionality!
10 months 2 weeks ago #92838

Please Log in or Create an account to join the conversation.

  • Posts: 200
  • Thank you received: 28
Hi John,

First, thanks for all you've done on this.

Have been using the beta for a couple of nights now but, before I provide detailed feedback, I wanted to confirm my environment with you. I control my setup from a desktop M1 Mac running the latest OS. On my mount I have an RPi running the current release of StellarMate OS which I use primarily as an INDI server. When I switch back and forth between the latest Beta (I'm using the May 11 nightly build) and 3.6.4 Stable I'm not changing anything at the RPi end of things. Is that the right way to test this or should I have something different running on the RPi when running the beta on the Mac?

An example of why I ask. Last night, while running the Beta, my guiding errors were in the .75 arc seconds to 1.5 arc seconds range while under 3.6.4 they were running at half that level under similar/same focus settings.

Thanks!
10 months 1 week ago #92869

Please Log in or Create an account to join the conversation.

  • Posts: 593
  • Thank you received: 276
Hi Fred,

That config should be fine. My focus changes haven't changed Indi.

As far as guiding is concerned the focus changes shouldn't have made any difference. I had a quick look through the other merged MRs. There are a couple of small changes but nothing that looks significant. If you get a repeat or if other folks see the same thing then we can investigate it more.

Regards,
John.
The following user(s) said Thank You: Frederick Ruegsegger
10 months 1 week ago #92877

Please Log in or Create an account to join the conversation.

  • Posts: 200
  • Thank you received: 28
Hi John,

Had a much better night with the beta last night. In fact I used it all night on two targets. I did find one thing I can't explain.

For the first focus run I set the Measure in the Process tab to FWHM and it worked wonderfully. After an hour (I have things to refocus every hour) when it started to focus again, it failed three times (usually because I stopped the process). In looking at the graph, it was reporting numerous negative FWHM values and the log file showed this message quite often: "Linear: ******** No poly min: Poly must be inverted". After three failed autofocus attempts I switched the Measure to HFR and then autofocus completed successfully (after I reset the focuser start position). In fact it completed successfully 7 more times through the night with the Measure set to HFR..

I have included the entire log file via Dropbox (too big for here): www.dropbox.com/s/r0l6j9a29q7enr4/FR_17-25-21.txt?dl=0. Hopefully that will just work - obviously let me know if you have any problems downloading the file.

To help you locate the relevant section of the log file, here are the times of the first few autofocus runs:

- 21:02:59 - FWHM, completed successfully
- 22:07:32 - FWHM, failed
- 22:10:37 - FWHM, failed
- 22:12:46 - FWHM, failed
- 22:14:40 - HFR, completed successfully
- 22:39:15 - HFR, completed successfully
10 months 1 week ago #92891

Please Log in or Create an account to join the conversation.

  • Posts: 593
  • Thank you received: 276
Hi Fred,

Thanks very much for the feedback and log. Very helpful.

For the FWHM AF run at 21:02 everything in the log looks good. Ekos was able to fit Gaussians to a lot of stars at each datapoint and the curve fit looks as expected.

Strangely at 22:07 things are quite different. The reason curvefitting didn't work is that at every datapoint Ekos is only managing to fit a very small number of Gaussians, sometimes 1 or 2. So the data going to curvefitting is very low quality. Curvefitting then gives up because the data it gets isn't very hyperbolic.

So why were so few stars processed? Looks like there were plenty of stars available so I don't think it was something like passing clouds. I have a number of control checks that are designed to throw out poor quality star fits and think something is probably going wrong there.

I'll work on this now.
10 months 1 week ago #92903

Please Log in or Create an account to join the conversation.

  • Posts: 200
  • Thank you received: 28
Thanks John. Also, I can confirm clouds are likely not the problem. I was watching the autofocus process, including the frames from the camera - the skies were clear and the stars visible. Just going outside that evening and looking up provided some confirmation ; ).
10 months 1 week ago #92911

Please Log in or Create an account to join the conversation.

  • Posts: 593
  • Thank you received: 276
Hi Fred,

I've made a tweak. It works on the Sim (but then so did the original parameter settings). I'll let you know when its merged.

John.
The following user(s) said Thank You: Frederick Ruegsegger
10 months 1 week ago #92913

Please Log in or Create an account to join the conversation.

  • Posts: 593
  • Thank you received: 276
Hi Fred,
The change is merged.
The following user(s) said Thank You: Frederick Ruegsegger
10 months 1 week ago #92918

Please Log in or Create an account to join the conversation.

Time to create page: 0.696 seconds