[fitswcs] Start of the TPV and ZPX FITS Convention Comment Periods

[Frank Valdes]

> I am slightly confused by the "double request" to discuss the thing on 
> fitsbits and fitswcs. Anybody *relevant* (in fitswcs) not reading both ?

I'm sorry for the confusion.  My original request was to discuss this only on fitswcs.  But the original announcement was to fitsbits to alert the full FITS community.  All comments should just be to fitswcs.
> 
>> This is to announce the start of the Public Comment Period on the TPV
>> and ZPX FITS world coordinate system conventions developed at NOAO.
> 
> Assuming as usual that the conventions are already in use, there is no 
> sense in comments which require a change in the definition or 
> implementation of the convention, but just on the completeness of the 
> documentation.

While this is essentially true any comments which indicate there is a serious problem with FITS conformance or that the WCS will not correctly meet its purpose of mapping between FITS images and celestial coordinates in a camera with distortions would lead to a change.   This convention was put forward in advance of actual scientific data being recorded by the new NOAO camera.  There is no existing data using TPV, though I believe there is data using this same formulation but labeled as TAN as originally proposed by Calabretta and Greisen.  So if there is anything really serious these can be addressed before systematic use.  The argument against making serious changes is the effort already made to get buy in and implementations by several teams.

> 
>> The TPV and ZPX conventions build on the standard TAN and ZPN WCS
>> projections, respectively, with the addition of a general polynomial
>> distortion correction.
> 
> The two conventions seem to use two different formalisms to tackle a 
> similar problem.
> 
> (1) the TPV uses a set of PV keywords for the coefficients, which is a
>      practical and clear solution. Although it is perhaps a pity not to
>      have another kwd indicating the order of the polynomial (i.e. the
>      number of coeffs to expect), the documentation is clear enough in
>      defining the rules for missing PV kwd defaults.
> 
> (2) However the HTML documentation for the TPV is not clear in the
>      fact it does not state that CTYPE1/2 should be suffixed with --TPV,
>      which one can infer from the sample header listing.
> 
>      This key issue to tag the convention shall be EXPLICITLY STATED in
>      the HTML documentation
> 
> (3) in fact the HTML documentation for the ZPX does state that !
>      So in this respect it is better.
> 
> (4) the fact ZPX uses not a clear set of coefficients but codes
>      them into the long string given by the concatenation of the WAT
>      keywords (the rules for the decomposition are clearly described in
>      the documentation) is at variance with the simpler mechanism of the
>      PV keywords for TPV. A similar mechanism would have been more
>      elegant, but it is too late now to complain.
> 

You have correctly pointed out that the two formulations effectively accomplish the same thing.  The thing that links them is the use of radial terms about the tangent point.  There are other WCS, e.g. TNX, which just use x/y polynomials to describe distortions.  But because of the way most optical systems make use of optics that are azimuthally symmetric and often have a well defined radial distortion, the formulations that include a ZPN like component are more appropriate and lead to a more stable description.  In fact the ZPN WCS was included in the standard as the only one with a general polynomial structure because it was recognized that it would handle the typical behavior of optical cameras.  However, real data that has not been remapped will still have residual distortions which is the purpose of these WCS conventions.

It is true that the TPV formulation is simpler in a FITS keyword context and is a key reason for adoption by the Dark Energy Camera over the ZPX formulation. It is also likely to be more widely supported by the community.  Though I would still urge developers to also support ZPX and TNX since there is a large body of important NOAO archival data using these WCS.  DS9 has done this.

Frank