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

[Mark Calabretta]

On Wed 2000/01/12 09:00:47 -0000, Peter Bunclark wrote
in a message to: Mark Calabretta <Mark.Calabretta at atnf.csiro.au>
and copied to: Don Wells <dwells at NRAO.EDU>, fitswcs at NRAO.EDU

>Like many normal procedures, it's still sub-optimal for Schmidt plates.

The GSC - the reference should have been to Russell (1990) - and earlier
workers they mention were indeed fitting TAN to Schmidt plates.

>We established this in practice in the early days of APM (Automatic Plate
>Measuring machine in Cambridge, UK).
>I'll bet than none of these `normal' procedures involve doing the fit
>in ARC and transforming to TAN + radial for _interchange_.

In the general case, you're never going to devise an interchange format
for the "model approach" to plate solutions, as opposed to the "empirical
approach" unless by including on-the-fly compilable code in the header
(be my guest).  Fitting ARC to Schmidt plates and then converting to
TAN+poly is no different to doing a model fit for refraction etc. and
then converting.

Pat Wallace foresaw these arguments a long time ago and proposed pixel
regularization as a purely practical solution for data _interchange_.
TAN+poly is better in the cases where it's applicable.

You could perhaps look at it not as a problem but as an opportunity for
photographic astrometrists to compare their instruments via plate
solutions represented in a common interchange format.

>Don, you keep using the string _interchange_ - good try, but FITS is
>just so phenomenally successful it is in fact used for far more than 

Sorry, but we have never offered anything more than data interchange.
Naturally you're at liberty to invent your own in-house set of FITS WCS
header cards that only you understand (e.g. DSS).  But unless you're
also prepared to provide the software for interpreting them (e.g.
getimage) then you'd have to convert these to a standard format for data
interchange.

>interchange; so much so that producers of reduction packages incorporating
>astrometry (I'm thinking of IRAF and STARLINK in particular) are waiting
>for this !"$%^&*() WCS to finalise before they'll cut yet more code;

Noone is keener than Eric and I to see the end of this thing!

>and FITS is becoming a data-reduction normal format, and most folk
>will use either packages or things like wcslib building blocks, so my 
>point is, doing a lsq fit in ARC and transforming to TAN via an analytical
>patch to the polynomial terms is in practice several orders of magnitude
>harder than working in a coordinate system that appears in the FITS
>standard.

Several orders of magnitude?  Let's be realistic!  For ARC vs TAN it's a
simple analytic expression.  For more complicated cases I have a program
using LAPACK which I can make available.  It consumes trivial CPU time.

Cheers, Mark