[fitswcs] TAN+poly & astrometric discussions in WCS paper-2

[Don Wells]

Peter Bunclark writes:
 > .. I contend that FITS WCS.. needs to be accurate enough to correlate
 > catalogs, images & maps, and do astrometry down to the requirements of
 > target acquisition, fibre positioning, etc.  It serves the astrophysicist
 > rather than the astrometrist.

I agree with this verbal summary of the scientific goals of our FITS
WCS effort. These goals imply numerical tolerances for requirements

I contend that our accuracy must be of order +/-50_mas in order to
properly support HST. Many modern groundbased imaging systems, with
excellent optics at excellent sites, need about +/-100_mas accuracy
for _astrophysical_ work, and whenever such imagery are compared with
HST imagery the accuracy requirement will be upgraded to the HST
level. MERLIN imagery has resolution comparable to HST, as does
high-frequency VLA imagery, and ALMA will also have such resolution.
Chandra's grazing incidence X-ray camera has a PSF with FWHM of order
500_mas, so it will be prudent to insist on accuracy of +/-200_mas
when comparing its imagery with any of the other systems I have
mentioned.  The recent availability of the Hipparcos data and
consequent adoption of the ICRF have eliminated the whole class of
problems with reference frame differences that we had when comparing
radio and optical imagery in the past, so there is no longer any
excuse for not achieving the accuracy levels specified above.

BENDXY: TAN+poly represents this distortion function to +/-7_mas RMS
        (with 33_mas worst residual), so the proposed FITS WCS
        _interchange_ notation is good enough for this difficult
        case. Pat will know the RMS of the original BENDXY _fit_ (my a
        priori guess is 100_mas); the mean error across the whole
        field is smaller, of course.

DEIMOS: TAN+poly represents the distortion function of this camera to
        23_mas RMS (164_mas worst) over its full field, and over the
        partial field to be used for imaging the worst case residual
        of TAN+poly is less than 2_mas. So, our proposed interchange
        notation passes this test too.

Please review the discussion of requirements and tolerances in the
above paragraphs and suggest additions/corrections as needed.

Perhaps some elements of the above science requirements discussion
should be included in the WCS paper-2.

-Don
-- 
  Donald C. Wells      Scientist - GBT Project        dwells at nrao.edu
                    http://www.cv.nrao.edu/~dwells
  National Radio Astronomy Observatory                +1-804-296-0277
  520 Edgemont Road,   Charlottesville, Virginia       22903-2475 USA