SGP and PlateSolve2

[Horwig]

Just upgraded from version 2.4.1.10 of SGP to 2.6.0.21.
I used to use Astrometry.net, but now want to give PS2 a go. 
Both star catalogues are installed. I loaded up some images taken with version 2.4. A couple solved within seconds, and I was impressed, but most didn't solve, and they eventually fall over to Astromentry. but a few times it will lock me out of sgp entirely.

The two images that solved were from entirely different equipment profiles, and the images that failed were from the same profiles. Am I doing something wrong?
Huw

[frugal]

Platesolve 2 needs to have really accurate pixel scales otherwise it gets confused about how far apart the stars really are. If you pixel scales are different in the second equipment setup, that would explain why Platesolve2 failed to solve them.

Effectively it gets it's speed by not looking too far outside the parameters you have given it, whereas Astronomy.net will scan the whole sky at a variety of pixel scales which has a higher chance of success, but will take longer.

[cardconvict]

I had this problem until I figured it out and like frugal has said you need to give it a accurate or very close pixel scale for PS2 to work. 

[Horwig]

How accurate do you call accurate? are we talking decimal places?

[cardconvict]

I know my imaging scale at 2x2 bin is around 1.86 so i just put in 1.8 and 1x1 is around 0.8 and I have put in 1 and still been good

[Horwig]

Hmm, I've put in 1.2 and 3.5, and Astrometry reports 1.204 and 3.48, so I'm not that far off. Must be something else

[MCinAZ]

21 hours ago, Horwig said:

Hmm, I've put in 1.2 and 3.5, and Astrometry reports 1.204 and 3.48, so I'm not that far off. Must be something else

How much sky area do your images cover?

If you could post an image at each scale, I could try solving them with a later revision of PlateSolve, so see if perhaps the algorithm has been improved.

[Horwig]

2 hours ago, MCinAZ said:

How much sky area do your images cover?

If you could post an image at each scale, I could try solving them with a later revision of PlateSolve, so see if perhaps the algorithm has been improved.

My images are relatively wide, from 85*57 arc minutes to 325*216 arc minutes. Thanks for the off of trying your version on my images, I'll PM you a wetransfer link.

 

Huw

[MCinAZ]

  Sky coverage should not be an issue. I've downloaded your files, however I've learned that my current installation of PS3 does not include catalog data. I have those files at home, but won't be able to install them until Monday.

  I note a slight discrepancy between the center coordinates in your FITS headers and those reported by Astrometry.net. This may be enough to confuse PS2. You might attempt a solution after clearing the coordinates loaded from the FITS headers. Don't know if that will remedy the issue, but its worth a try.

[Horwig]

1 hour ago, MCinAZ said:

 

  I note a slight discrepancy between the center coordinates in your FITS headers and those reported by Astrometry.net. This may be enough to confuse PS2. You might attempt a solution after clearing the coordinates loaded from the FITS headers. Don't know if that will remedy the issue, but its worth a try.

Yes, I noticed the discrepancy, so blind solved the images, and input the new co-ordinates into the hint window, but Platesolve still failed

[Horwig]

Well it's clear, and finally I have the kit working on real stars, and guess what PS2 solves in seconds, so why does it not on saved images?

[StevieDvd]

Any platesolver works on the same principles.  You pass an image along with a set of parameters about the image, if you use saved images that may invoke an image conversion on the fly and depending on the image type may or may not contain FITS headers.  A live image is similar but is likely to use the host program to pass across the parameters and image information  - usually a bit more fiocused on the image supplied and easier for the plate solve to resolve. You would of course want a live image solved the quickest.

If you were to give as much info for your saved images as the host program does  then you should see the same resolution speed.

I would expect the host program passes across more of the optional data by default since it already has the data and it speeds up the process. Whereas a saved image may only pass across the mandatory and some of the optional - or even just enough for a blind-solve.

Have seen a lot of this whilst debugging the process when I was attempting to have my local astrometry server and change astrotortilla to use that instead of it's hard-coded settings for cygwin.  Never did get it sorted to my satisfaction though it would solve when given the right parameters.

 

 

[MCinAZ]

After installing catalog data, I found that the M81+M82 and pacman images solved using PlateSolve-3.48 without issue. The other two solved after I manually entered approximate image dimensions. Curiously, when the pacman image is loaded, the image size dialog boxes are populated with approximately correct values. This is not the case with any of the other three files. Looking at your FITS headers, I don't see any obvious reason why the program behaves as it does.

[Horwig]

Thank you very much for taking the time to do this MCinAZ, that's very much what I was finding. The M81+82 solved every time, the Pacman would sometime solve, the others would not.

Now that I'm imaging with the currenet version, everything seems to work fine, but if I were to be going to add data to existing sequences, I'd need to solve with another solver, so not a big issue.

 

Thanks again

 

Huw