[fitsbits] Plate coordinate tolerances

[Mark Calabretta]

On Mon 2004/08/09 13:14:20 MST, "K. L. Tah" wrote
in a message to: fitsbits at nrao.edu

>I was wondering if someone could tell me the accuracy with which one
>can determine positions on the fits images?

This depends on the particular image.  The FITS keywords CRDERia and
CSYERia may be used to record the random and systematic error in the
world coordinate for axis-i (though it is unlikely that you will find
these in your header).

>I'm trying to predict the path of  stars over a course of several years in
>to the future and have written a program that plots this on the fits
>image.
>I had first retrieved the fits images centered on the 2000 coordinates of
>star, and converted this to pixel coords using sky2xy. As a check, I
>extrapolated  paths back the epoch (which I'm
>assuming is when the pictures were taken and is given in keyword EPOCH).

EPOCH, now deprecated in favour of EQUINOX, was used to record the
equinox of (dynamical) equatorial coordinate systems, e.g.
Bessel-Newcomb/FK4 or IAU1984/FK5 - look for an RADESYS card to
determine which.  The date of observation would have been recorded in
DATE-OBS, or as a modified Julian date in MJD-OBS.  Of course it's
possible that whoever wrote this FITS file confused these times.

>If there are no errors, I would expect this to
>lie exactly on the center of the star in the image. However, I have
>found
>that these are off by several pixels in most cases (see, eg. GJ729
>attached). Is this to be expected? If so, is there any info in the fits
>header that might help me quantify this?

If the header has DATE-OBS or MJD-OBS, try plotting the star's position
at that time.  You might also try interpreting EPOCH, which I assume from
your attachment was given as 1992.4235, as the equinox of the equatorial
coordinate system, presumably IAU1984/FK5.

Without seeing the FITS header itself there's not much more I can say.

Mark Calabretta
ATNF