Solver and FITS headers.

[Paul Leyland]

Since running ANSVR on my telescope control system syncing the telescope
has become a doddle. The scope is pointed at a random bit of sky and a
5-30s exposure taken depending on how many stars are picked up. Plate
solving shows where the scope is pointing and the control software
(Maxim DL at present) then takes over from there. It is rare that
slewing is more than 5 arcmin away from the target and that error is
reduced to under 1 arcmin after another solve or two taken a good few
degrees away.

The main limitation is plate solving time. The server is configured to
use likely plate scales, corresponding to native image scale at typical
binning levels and the image is down-scaled as well. The supplied image
has target RA and Dec in its FITS headers but I have yet to find out how
to make use of them. Does anyone know?

It wouldn't work for the very first solve because the FITS headers will
not have usable coordinates, but subsequent ones could be made much more
efficient. Even for the first solution, I know roughly where I am
pointing (for instance, roughly due south, 45 degrees altitude, at known
latitude & longitude) and the UTC time. Again, any clues on how to
exploit this information? So far I have thought only of using
planetarium software to locate that piece of sky to 30 degrees, say, and
to solve the image "by hand". That is laborious and takes at least as
long as searching the entire database automagically.

No big deal but it would be nice to wait only 15 seconds instead of 150
seconds each time.

[Dustin Lang]

Hi,

I'm a bit surprised you're finding it taking 150 seconds.  What's the angular scale of your images, which index files are you using, what scale limits are you giving?

We don't currently have a mechanism for pulling estimated RA and DEC values from headers, but you could probably do this with a script before submitting the image.  The API that ansvr is providing does have an option to set the center_ra, center_dec, and radius.

cheers,

--dustin

--
You received this message because you are subscribed to the Google Groups "astrometry" group.
To unsubscribe from this group and stop receiving emails from it, send an email to astrometry+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/cf9f76e6-83c6-3ddf-77ef-02b53b30557f%40gmail.com.

[Bryan]

Paul

Can you provide a copy of one of your typical images used for plate solving?  The easiest way is probably to upload to the Internet nova.astrometry.net and then post the Job Number here.

Bryan

[Paul Leyland]

It sometimes take 5 seconds, it sometimes times out after 5 minutes. 150 is achieved a few percent of the time and seems a reasonable upper limit. I have never managed to work out why some take longer than others, assuming that at least 20 stars are visible. The field of view, at 16x13 arcmin might have something to do with it perhaps.

Thanks,

    Paul

[Dustin Lang]

Hi,

Interesting.  Most of the large-scale tests I did on Astrometry.net were on image from the Sloan Digital Sky Survey, 9'x13' images, so your images should be similar in that regard.  If there's any way you could dig out a typical or fast (5-second) and a failure or long-runner, we might be able to take a look and suggest things to try.  You can either send them to me privately, or post at nova.astrometry.net and provide links to your submissions.

cheers,

--dustin

[Dustin Lang]

Thing to check would be:

- do your scale ranges bracket the true scale?  The scale-low, scale-high limits are very hard limits -- the solver will outright reject any solution outside that range, regardless of how good it is

- looking at the source detection images (especially the -objs.png plot, if you make those), are the detected sources well centered on the stars?  Sometimes saturation can cause this to get messed up.  Usually --downsample 2 improves the situation until stars become more saturated.

cheers,

--dustin