Odd that you have been working on the same thing I was. I am setting up a new RPi3 B+, which has been going very nicely. The 5Ghz Wireless support is very nice. Anyhow, I was trying to use Load and Slew to load an image I took in SGP. I grabbed all of the same indexes I use on my ANSVR implementation and grabbed all of the 41xx FITS as well. In testing with nova.astrometry.net, the 4108.fits was used to solve the same image.

broiler.astrometry.net/~dstn/4100/

The remote solver still failed. I noted that an informational message was given saying that PIXSIZE1 was not found in my FITS header. So, I went into Pixinsight, updated the header and added PIXSIZE1 and PIXSIZE2 values to match the pixel size of the camera used to take the image. Once I did that, the solver completed (although it took 45 seconds) and the simulator mount moved to the correct target. The solver also used the 4108.fits file to solve this image.

I then parked the simulator and loaded the same image, but now used ANSVR to solve it. It also completed successfully, and took 19 seconds to solve the image. That makes sense, since my desktop is much more powerful than the RPi3.

Then I loaded the same image (without PIXSIZE1 and PIXSIZE2 in the header) and tried it again. Remote failed, and ANSVR failed.

So I loaded the JPG into SGP, and ANSVR could blind solve it, but it took 100 seconds, and the command passed to ANSVR was identical in SGP and KStars.

Then I loaded the original FITS file into SGP, and ANSVR blind solved it in 6 seconds. Now, SGP read the fits header to get data (even for the blind solve):

exec /usr/bin/solve-field -p -O -U none -B none -R none -M none -N none -C cancel --crpix-center -z 2 -u arcsecperpix -L 4.0528092288503 -H 4.47942072662402 stars.fit

The strange thing is that my pixel scale is not 4 arc seconds per pixel. According to the FITS header of the same image, it is: 1.20455011136108.

SGP is using hints though, and you can see those here (again it is reading those from the FITS header)

In conclusion I think that the solver could be improved to extract more data from the header in order to make solving much easier and faster. How the folks at SGP knew to set app to that value, I have no idea but it sure worked lightning fast. Even though I had app selected in the Ekos options, it seemed to stick to AW (whatever that is).

2018-07-07T16:33:16 Using solver options: -O --no-plots --no-verify --resort --downsample 2 -L 51.3588 -H 94.0029 -u aw

Read More...