Can't plate solve on Raspberry Pi, works fine on Windows

[Kevin Ross]

Hello.

I just re-imaged my Raspberry Pi setup (with KStars/Ekos/INDI). I previously had ASTAP working perfectly fine. Now on my new setup, I can't solve any images. I'm using version 0.9.553a. I tried H17 and H18 databases.

Just for testing, I tried loading a file into the ASTAP GUI program, and try to solve. No joy. I load the exact same image into the Windows version of ASTAP, it solves immediately.

Here's a link to an image. It solves in Windows, not in Raspberry Pi. It's a 90 second Light frame. I also can't solve 1 or 2 second images, when trying to sync the mount, for example.

http://nova.astrometry.net/user_images/4797709#annotated

Thanks for any help you can provide.

[han.k]

Hello Kevin,

The image is perfect to solve. The focal length and pixel size match with the solution.

The only thing what could possible go wrong is the command line from Ekos. I have noted that in the last version the Ekos is sending image width rather then height. This is mentioned at the INDI forum:
https://indilib.org/forum/ekos/9772-dither-messes-up-guide-exposure.html?start=36#72581

Ekos write the temporary image file to solve at /tmp , ASTAP will write an .ini log file at /tmp. Can you send me the .ini file so I can have a look to the command line used?

Regards, Han

Re the plate solving: I see you are using ASTAP. I reported an issue a while back with stellarsolver/ekos & ASTAP where ASTAP expects FOV degrees for the image height, but
is being sent the width degrees. This can be worked around by unsetting "Use Scale" or >>perhaps both unchecking "Use Scale" and "Use Position" in the align options Scale & Position
menu when using ASTAP. You can test on your own as well by using the offline astap program. For instance, on my system astap is installed as /opt/astap/astap. So you
can >>run "/opt/astap/astap -help" to see its options, and, for instance, capture a fits you're having trouble with, and running "/opt/astap/astap -f FILE.fits" to solve it "blind" without
scale nor position. You can see the args stellarsolver is sending astap from your log--where it's >>adding in the scale (that's the -fov 1.36103 in your log) and the position (-ra 19.1981 -spd 75.0158).

[Kevin Ross]

Unfortunately, I can find no such file. I just tried "Load and Slew" from Ekos, which will load the FITS image, do a plate solve, and slew to the coordinates. Ekos of course fails with "Solver failed. Please check you have sufficient stars in the image.".

The ASTAP GUI on Raspberry Pi also fails. Would it produce useful log files?

pi@telescope:/tmp $ ls -la
total 72120
drwxrwxrwt 19 root root     4096 Jun 22 11:54 .
drwxr-xr-x 19 root root     4096 Jun 15 21:40 ..
drwx------  2 pi   pi       4096 Jun 22 02:17 dhcpcd-pi
drwxrwxrwt  2 root root     4096 Jun 22 02:17 .font-unix
drwxrwxrwt  2 root root     4096 Jun 22 02:17 .ICE-unix
drwxr-xr-x  2 pi   pi       4096 Jun 22 02:26 kstars
drwx------  2 root root     4096 Jun 22 02:17 pulse-PKdhtXMmr18n
-rw-r--r--  1 pi   pi    2047680 Jun 22 02:34 solver01a73a9079d442b0b7a9ed27d483d77a.fits
-rw-r--r--  1 pi   pi   32785920 Jun 22 11:53 solver10d3b55c61f14fdabd96fcf68caf3738.fits
-rw-r--r--  1 pi   pi    2047680 Jun 22 02:33 solver1d61d885263743b1ae353d3ccc181ead.fits
-rw-r--r--  1 pi   pi   32785920 Jun 22 02:33 solver2c08f0ecddbd49649f0bfd94f8580dc9.fits
-rw-r--r--  1 pi   pi    2047680 Jun 22 02:33 solver399e703ae24c456f9569be9f8a717803.fits
-rw-r--r--  1 pi   pi    2047680 Jun 22 02:33 solver6269192133824a7594b79c10987fe7ca.fits
drwx------  2 pi   pi       4096 Jun 22 02:17 ssh-0kUigZUfWd3x
drwx------  2 pi   pi       4096 Jun 22 02:17 ssh-MAUcISYimY0z
drwx------  3 root root     4096 Jun 22 02:17 systemd-private-329099933f6a4ec29339d7fc770716eb-colord.service-GyZOEE
drwx------  3 root root     4096 Jun 22 02:17 systemd-private-329099933f6a4ec29339d7fc770716eb-rtkit-daemon.service-nbfMNl
drwx------  3 root root     4096 Jun 22 02:17 systemd-private-329099933f6a4ec29339d7fc770716eb-systemd-timesyncd.service-Yc6zNb
drwxrwxrwt  2 root root     4096 Jun 22 02:17 .Test-unix
drwx------  3 root root     4096 Jun 22 11:00 .vnc-0
drwx------  3 pi   pi       4096 Jun 22 02:17 .vnc-1000
drwxrwxrwt  2 root root     4096 Jun 22 02:17 .vncserver-license
drwx------  2 root root     4096 Jun 22 02:17 .vnc-vncservice
-r--r--r--  1 root root       11 Jun 22 02:17 .X0-lock
drwxrwxrwt  2 root root     4096 Jun 22 02:17 .X11-unix
drwxrwxrwt  2 root root     4096 Jun 22 02:17 .XIM-unix

The /tmp/kstars folder is empty

pi@telescope:/tmp $ ls -la kstars
total 8
drwxr-xr-x  2 pi   pi   4096 Jun 22 02:26 .
drwxrwxrwt 19 root root 4096 Jun 22 11:54 ..

[Kevin Ross]

I enabled debug logging in Ekos, and I now get this log file in /tmp:

12:03:17  /opt/astap/astap -o /tmp/externalSextractorSolver_7.ini -speed auto -f /tmp/solver2e371aca21df42b988243480c18d6eb0.fits -wcs -z 3 -fov 0.958149 -ra 12.3316 -spd 137.381 -r 15 -log
12:03:19  Creating grayscale x 3 binning image for solving/star alignment.
12:03:20  241 stars, 189 quads selected in the image. 182 database stars, 143 database quads required for the square search field of 5.1°. Search window at 100% based on the number of quads. Step size at 100% of image height
12:03:20  Using star database H17
12:03:20  No solution found!  :(

Unfortunately, it references /tmp/externalSextractorSolver_7.ini, but that file doesn't exist in /tmp

[han.k]

That log helps. The command line is fully correct. The only major abnormality is that the ASTAP searches with window a 5.1 degrees. This is fatal. It should be 0.95 degrees. But the command line clearly specifies -fov 0.958149.

Very strange. I think the log file /tmp/externalSextractorSolver_7.ini is deleted if you leave Ekos. You should be able to copy it when Ekos is still live. If the Ekos log is correct, it should work. This change from 0.95 to 5.1 degrees should never happen.

It would be nice if you could retrieve a /tmp/....ini file so we can check the command line recorded in this ASTAP file. Else I would suggest to renew ASTAP. This behavior should never happen. There has never been a bug like this.

The binning I would set at 0 or 2 rather then 3 but the image you send me solved without problems at binning 3.

Han

[han.k]

The ASTAP GUI interface should work on the Raspberry PI. Maybe ASTAP executable is damaged. You could just download install the ASTAP Debian package and install it. What operating system are you using? 32 or 64 bit.

Han

[Kevin Ross]

Ekos was still running while I went looking for that .ini file.

BTW, using the GUI on Raspberry Pi also does not work. Attached is a screen shot.

I just removed the current ASTAP installation, redownloaded the .deb file, and installed again. Same result.

ASTAP version ß0.9.553a, 64 bit, dated 2021-6-17

I'm using 64-bit (arm64) on Raspbian. On my previous Raspbian build, the one that was working, it was 32-bit Raspbian, and hence 32-bit (armhf) ASTAP. Not sure if that makes a difference. It was also a much older version of ASTAP.

[han.k]

Again this 5.1 degrees. Puzzling. I will startup my Raspberry Pi 64 bit to test it.

[han.k]

I see the same 5.1 degrees here .....

[Kevin Ross]

Well I'm glad it's not just me then! :)

[han.k]

Fixed. Download again the astap_arm64.deb This is version 0.9.553b.

I pretty sure the compiler was spinning and made a mistake. It was on "slow optimizations." Now it is working here.

Han

[Kevin Ross]

That works a treat! Much better! You da man!

Tested in Ekos, as well as standalone ASTAP GUI. All working again.

Just curious, is this just a Raspberry Pi issue? Just 64-bit?

[han.k]

I think this is a known compiler bug. It knows 4 levels of optimizations and level 3 seems a problem. Normally I use level 1. I can't remember why I put it on level 3 except for a smaller executable file.
I haven't installed the latest compiler on my Raspberry Pi, maybe I will do that the next time.