Weird behaviour of wcs_world2pix in astrometry.net created WCS

[Keenman Alex]

Hello all,

I'm having a weird problem with a WCS created by astrometry.net and can not understand  how to solve it.

I have a Landolt field - can be loaded here (ftp://ftp.physics.uoc.gr/pub/tmp/blubb/SA104_R.fits)

Finding chart and coordinates of standards are at this page http://james.as.arizona.edu/~psmith/61inch/ATLAS/charts/c90.html

When I solve the field with the command:

solve-field --ra 190.59166666 --dec -0.6666666 --radius 0.1 --scale-units arcsecperpix -L 0.2 -H 0.3 --overwrite --no-plots --no-fits2fits SA104_R.fits

it does the job and creates SA104_R.new (ftp://ftp.physics.uoc.gr/pub/tmp/blubb/SA104_R.new).

World coordinates look correct. If I point to a star in the resultant file in DS9 i see RA and DEC very close to catalog values.

The problem is that when I search for x,y on the frame using world coordinates of standards e.g. wcs.wcs_world2pix(190.587499,-0.67444444,1) it outputs notably wrong x. For the given star SA104 334, x = 1105.119  while the actual position of the star is somewhere ~1135

All x coordinates seem to be shifted in one direction, so it doesn't look like a fitting error.

Do you have any ideas what is the reason of the error?

Thanks,

Alex

[Keenman Alex]

Do you have any ideas what is the reason of the error?

Also I see this warning of astropy all the time:

WARNING: FITSFixedWarning: The WCS transformation has more axes (2) than the image it is associated with (0) [astropy.wcs.wcs]

However WCSAXES = 2 in the header.

wcslint SA104_R.new outputs:

HDU 0 (PRIMARY):

  WCS key ' ':

    No issues.

[Dustin Lang]

Please file the FITSFixedWarning report with astropy -- that FITS file looks fine to me.

As for finding your standard stars: on that web page, it only gives coordinates to 1 second -- 60 pixels in your image!  Not surprising it's off by 30 pixels.

Comparing against the USNO-B1 coordinates, your field seems to be well solved:
  http://nova.astrometry.net/user_images/415849#redgreen

One other thing to watch out for is that not all FITS tools understand the SIP polynomial distortion corrections we produce.  They're small for your image (< 1 pixel, I think); you can turn off SIP with "solve-field --no-tweak [...]"

By the way, here it is in SDSS: http://skyserver.sdss3.org/dr8/en/tools/explore/obj.asp?ra=190.5852&dec=-0.6745

And with those coordinates I get the observed position:

> wcs-rd2xy -w SA104_R.new -r 190.58506428 -d -0.67456481
RA,Dec (190.5850642800, -0.6745648100) -> pixel (1136.1351972232, 1144.0002793254)

cheers,
--dustin

[Keenman Alex]

Thank you Dustin!

Please file the FITSFixedWarning report with astropy -- that FITS file looks fine to me.

Yes, i will report. 

As for finding your standard stars: on that web page, it only gives coordinates to 1 second -- 60 pixels in your image!  Not surprising it's off by 30 pixels.

actually the scale is 0.283 ''/pix i.e. 1'' = 3.5 pix

 

Comparing against the USNO-B1 coordinates, your field seems to be well solved:
  http://nova.astrometry.net/user_images/415849#redgreen

One other thing to watch out for is that not all FITS tools understand the SIP polynomial distortion corrections we produce.  They're small for your image (< 1 pixel, I think); you can turn off SIP with "solve-field --no-tweak [...]"

By the way, here it is in SDSS: http://skyserver.sdss3.org/dr8/en/tools/explore/obj.asp?ra=190.5852&dec=-0.6745

And with those coordinates I get the observed position:

> wcs-rd2xy -w SA104_R.new -r 190.58506428 -d -0.67456481
RA,Dec (190.5850642800, -0.6745648100) -> pixel (1136.1351972232, 1144.0002793254)

Yes it seems you are right - precision of my catalog is not good enough.

the difference in RA 0.41s = 0.41*15 * 1.0/0.28 = 22 pix 

 Thanks again!

cheers,
--dustin

[Dustin Lang]

As for finding your standard stars: on that web page, it only gives coordinates to 1 second -- 60 pixels in your image!  Not surprising it's off by 30 pixels.

actually the scale is 0.283 ''/pix i.e. 1'' = 3.5 pix

Yeah, I meant 1 second * 15 = 15 arcsec = 52 pixels

--dstn