Hi Rob

Some updates. I'm unable to solve the result.fits file that I sent yesterday with M27. But if I use one of the frames, as captured by KStars two months ago, it was solved on les that 2 sec. Same for other individual frames from other sessions. So, perhaps the stretching that I have applied to the result is invalidating the photometry, and confuses the matching algorithm. What is important for me, is that It works with individual frames as taken from the telescope, so it's seems reliable enough for my needs.

NOTE: I'm using Builtin Sextractor and FastSolving as defined by default on the current master branch.

Rafa

Cerro Torre,

I just ran a couple of tests on your M57 fits image. In fact, yes it doesn't solve right out of the box. I believe the issue is in the Scale information. When I unchecked "use scale", it solved in about 10 seconds using ParallelSmallScale on the internal Solver. It was about the same time for a totally blind solve.

I believe the issue is in the FITS Header. It says it is about 0.71 arc seconds per pixel. From the solution information below, the solver said it is 0.634 arc seconds per pixel. That is outside the range that KStars searches. KStars automatically sets the low point in the scale search to 0.9 * scale and the high point to 1.1 * scale. This image's scale is *just* beyond the range of the search

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Starting Internal StellarSolver Astrometry.net based Engine. . .
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Solve Log Odds: 106.697
Number of Matches: 16
Solved with index: 4108
Field center: (RA,Dec) = (283.396, 33.0297) deg.
Field center: (RA H:M:S, Dec D:M:S) = (18:53:34.946, +33:01:46.930).
Field is: (-29.1828, 19.4304) deg from search coords.
Field size: 54.8396 x 36.5765 arcminutes
Pixel Scale: 0.634873"
Field rotation angle: up is 89.5811 degrees E of N
Field parity: neg

Successfully solved with child solver: 5
Shutting down other child solvers
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Internal Sextractor w/ StellarSolver took a total of: 9.543 second(s).
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Rafa,

Did you update to the latest stellarsolver first? Based on your image, I found an issue with float images. I think I have corrected that. if you haven't updated please try that first.

Hi Rob,

Yes, I have the stellarsolver library installed but the problem building kstars persists:

I removed/reinstalled them today - no change in the build failure point.

Thanks!
Greg

Hi Greg,

I just took a look on the PPA and it is out of date. launchpad.net/%7Emutlaqja/+archive/ubunt...=backwards&start=375

It says it was built on 10/8/20. That would explain why it doesn't have things I added almost a week ago.

try this: github.com/rlancaste/stellarsolver

Here a link to my Dropbox: www.dropbox.com/sh/xu1uwz5jpqpa9lr/AAAyy...Fqic2AEGGc4-oaa?dl=0

FITS file xx.010 failed repeatedly to solve, file xx.215 worked right away and again during the meridian flip.

Makes no sense to me.

Jo

I disabled the scale option, but the scale range lies between 0.638705 to 0.780639 arcsec/pixel and it fails.

I would like to change kstars scale range in the source code. Do you have an idea where it is implemented? Class and method name?

Jo, I just ran some tests on yours,

This was the first image:

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Starting Internal StellarSolver Astrometry.net based Engine. . .
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Solve Log Odds: 177.889
Number of Matches: 28
Solved with index: 4112
Field center: (RA,Dec) = (43.7501, 60.4722) deg.
Field center: (RA H:M:S, Dec D:M:S) = (02:55:00.030, +60:28:19.980).
Field is: (98.0025, 113.7) deg from search coords.
Field size: 4.08117 x 3.08207 degrees
Pixel Scale: 3.15488"
Field rotation angle: up is 89.9464 degrees E of N
Field parity: neg

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Internal Sextractor w/ StellarSolver took a total of: 1.191 second(s).
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

And you were correct the second one did not solve. . . .until I unchecked "use scale". The issue again was the scale was wrong in the FITS Header.

in your first image, the fits header said the scale was 3.13 arcsec per pix, which is well within the range that KStars computes for the high and low values. The second image was basically identical except the scale in the fits header was 6.53, which is way outside the range. So a blind solve would solve this fine, but an informed solve would fail

Cerro Torre,

I believe the code you are looking for is in Ekos/Align.cpp. and is in the method calculateFOV, though it might be someplace else too.

Interesting! But now the real question: Why on Earth would the scale be different? Both images where acquired with the same camera, same exact exposure setting, i.e. same Capture file, same Scheduler file, same filter (Halpha), same target (same exact position), same telescope (NOTHING was touched). The only difference is the time at which they were acquired.

What would have led to the difference in the scale????? Kstars recorded that scale, not me.

PS: Just to clarify: What do you mean with 'first image'? 010 or 215? Because 010 did not solve for me, 215 did!

Greg,

Jasem also mentioned to me that this is the nightly stellarsolver build:

www.google.com/url?q=https://code.launch...jeynXHjYhhooqFA1n6_A

El Corazon, that is a really good question indeed. I don't know the answer, but it is worth investigating. At the moment I am very much focused on getting StellarSovler perfected in KStars, but I can mention it to Jasem.