It'll be easier to use Aberration Inspector when the Milky Way comes back and there are more stars.

Indeed, but earlier in this thread I posted an image of M52, where I could detect 5199 stars with the SEP profile in FITS viewer, and reproducing that profile as an autofocus profile I still ended up with 100 stars, on the same image.

Can you send the image please.

It's a bit big to send (205 Mb).
Here's a DropBox link: www.dropbox.com/s/fhsvxqiqdl5htsw/autofo...e_20-55-13.fits?dl=0

Thanks. Running with "your" SEP params (except the 200 restriction) I got 3441 stars so its not the parameter settings.

Must be something else going on.

If you put verbose logging for focus on the next session and we'll see what we get.

I've been trying to run the old session in the simulator.

The first point is that I can confirm the previous V-curve is screwed up -- until I looked at the verbose log for that session I had assumed that averaging over 3 frames would result in 3 images saved. But it's not like that. So when I tried to run the simulator on the saved images, also with average over 3, it got a bit confused.

Second point: I can't reproduce the result with the same settings (I took a screen-shot at the time). In fact I can get close but can't reproduce at all. But I think there's something throwing the conversion pixels to arc seconds off in the simulator (or the real one). That's because, though I've matched my system as far as possible in the simulator, the CFZ results are different, and it's in the calculation for CFZ Camera. In the simulator I get 695 steps, but in real it's 503 steps. The simulated CCD is the same as mine except that the CCD X and Y resolution won't go above 8192 in the simulator, while mine is 11664 x 8750. I can't see any other difference in there.

How is the CFZ Camera calculated?

Looks like only the last of n images (when Average Over n is selected) is saved. I'll fix that.

So the processing is done in pixels and the Arc-seconds option is just for display, The calc requires Focal Length (from the Optical Train telescope) and Camera Pixel Size (you can set in Indi for the sim).

The CFZ Camera = pixel_size * f^2 * A
f = focal ratio (calculated from Focal Length and Aperture from Optical Train)
A = Aperture (from Optical Train)

Returning to this topic (once the weather cleared up, though it's cloudy again now). This is with 3.6.9 beta, by the way.

I recorded a test run the other night, and I've put a set of 12 images, and logs, and screenshots in this folder: www.dropbox.com/scl/fo/aypzvgy8lc3v0p2a0...xt3oxfihe2xtrd4&dl=0

A small illustration of the problem is here, where I set the focus metric to # of stars. Note that the out-of-focus image is the image #11, since if I let it complete it knows it hasn't a solution and returns to the starting position, which, with the bug in the FLI PDF focuser is assumed position 0, a very long way from the start. So I made the screen grab at this point rather than wait for everything to disappear while it moves to 0.

Note that this is not the set of the 12 frames, since that one converged, and it doesn't have the earlier profile, which is set to 200 (not 100) stars, but with stars set to 0, as shown in the (slightly later) second image.



I cannot understand why I never get more than 100 stars.

FWIW, using the FITS Viewer's "Plate Solving tab" (we discussed previously) on the first image (autofocus_frame_23-07-03_1_1_1.fits)
with default align parameters I get: 195 detected (3464 unfiltered) -- where the min size filtering parameter is responsible from 3464 --> 195.
and with your parameters I get: 1033 detections (1621 unfiltered) -- where the max ellipse=1.3 is responsible for the filtering.
[Note--I'm using the align parameter settings since that's what the FITS Viewer's "Plate Solving" tab uses.]

Are you roughly able to reproduce that?
That is, if you can reproduce that, then the issue must be in how the focus module is using StellarSolver, as opposed to StellarSolver and its parameters.

Hy

Richard,

While you're at it, can you try one other thing for me? Just a hunch...

I see you have a custom/new profile you've named "Gutter-Profile" that you're using.
I have never tried to use a new custom profile. Perhaps there's a bug in using one of those?
Could you please also try to assign the same values to 1-Focus-Default and use that default profile?

Perhaps the truth is that the code is really using a different profile than the one you've been editing?
This is suspicious because I can see that the defaults for 1-Focus-Default is NumKeep = 100.

Hy

Richard,

My hunch was right. I was able to reproduce your problem when I tried to create a new focus profile.
In fact, the way I read the code, the system always uses the parameters in the first focus profile (1-Focus-Default) no matter how that menu is set.

So, I will discuss this with Jasem and John and we'll decide how/when to fix, but in the meantime to work around this, please just use 1-Focus-Default.

Hy

I submitted a fix which will be in the upcoming release (should be within a week) and currently in the latest software.
invent.kde.org/education/kstars/-/merge_requests/1124
Hy