[fitswcs] Summary of status of fitswcs negotiations

[Don Wells]

Dear friends of FITS WCS,

I have added Perry Greenfield <greenfield at stsci.edu> to the fitswcs
mailing list so that he will get this message.

Three weeks ago I posted a list of four issues which needed to be
resolved. Yesterday I concluded that we appear to have resolved
them. Here is my review of the issues:

		   -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Don Wells wrote on 98-07-15:
 > (1) SPLITTING-G&C: I perceive that there exists a substantial
 > consensus that the G&C paper needs to be split into two papers, the
 > first with a title something like "Generalized World Coordinate
 > Conventions for FITS" and the second with the same title as the
 > current G&C paper..

It appears to me that there is no dissent on this question.  Eric
Greisen has stated that he is willing to do the work of splitting the
G&C text as a part of making the various changes which have been
agreed during our discussions, but that he wants to be *certain* of
his mandate before starting the work. I absolutely agree with him on
the latter point---there should be no more 'trial balloon' documents.
The next public version of the text should be the two documents which
will be reviewed by the FITS community as a whole and will then be
submitted to the regional committees for the formal approval process
leading to a formal vote of the IAU-FWG.

 > (2) MULTIPLE-WCS: It appears to me that there is a requirement to
 > support multiple WCS descriptions of the same data object. I recommend
 > that we do this by appending a character code [A-Z] to the WCS
 > keywords for secondary descriptions; the keywords without the
 > character would be the default operational ones. This notation would
 > require reducing the number of axes supported by the WCS keywords from
 > 999 to 99..

It appears to me that the discussions during the past three weeks have
shown that there are a variety of requirements for multiple WCS
descriptions, that the appended character notation is feasible, and
that no better alternative has been proposed. Entire sets of keywords
of the WCS versions would be swapped, without any defaulting of values
other than the default-value rules which apply to the active WCS (the
base set of keywords without the appended character codes).

 > (3) CD-MATRIX: I perceive that there exists a substantial consensus
 > that the lowest level linear transformation should be done with a
 > matrix of partial derivatives, of the sort called 'CD' in current
 > implementations. As far as I am aware, there is no longer a
 > science-driven requirement for the CDELTi/PCij notation which was
 > proposed in the G&C draft..

It appears to me that no science-driven requirement for CDELTi/PCij
has been asserted during the past three weeks, and that it has been
implicitly agreed that the CD-matrix partial derivative notation is
functionally equivalent to CDELTi/PCij. There are existing archives
which are encoded using the CDi_j version of the latter notation (see
NICMOS and STIS at http://archive.stsci.edu/cgi-bin/keyword_form) and
existing popular application software which supports it (see
http://www.cv.nrao.edu/fits/os-support/ms-windows/fitsview/ and
http://tdc-www.harvard.edu/software/saoimage.html), whereas there are
no archives coded with CDELTi/PCij notation, and very few application
programs. We must choose CDi_j for the interchange notation.
  
 > (4) PIXEL-CORRECTION-FUNCTION: Many data acquisition systems produce
 > image matricies which are slightly distorted relative to canonical
 > projection formulae, due to misalignments and to the use of complex
 > corrector optics..

We have made dramatic progress on this issue.  During the course of
the discussion (a) Doug Mink asserted that a DSS-like polynomial form
represents a variety of optical imagery well, (b) Mark Calabretta
posted to fitswcs (45_KB of Postscript on 07-17) an excellent
description of his implementation of a 'DSS' projection type for
wcslib, and I pointed out that all of the terms needed for WF/PC2 are
included in the DSS set. DSS uses a gnomonic (TAN) projection--not
ARC!--so this leads to the obvious conclusion that we should add
parameters for the DSS terms to the TAN projection. These terms would
be a function of the x-y coordinates in degrees relative to CRPIXi
which would be produced by the CDi_j linear transformation; the
resulting distortion-corrected x-y coordinates in degrees would then
be passed through the TAN projection. The DSS set of terms includes
two terms of radial ($(x^2+y^2)^k$) distortion in addition to the
usual $x^iy^j$ terms; these should support both transformation of ARC
to TAN and the radial distortions present in the wide variety of prime
focus cameras and field flattening optical elements which are used in
optical astronomy.  Probably there are no ground-based cameras which
are more optically complicated than WF/PC2 and the DSS representation
of a wide-field Schmidt camera, and so it appears that this proposed
TAN' projection is likely to support all of optical astronomy, both
ground- and space-based. Mark Calabretta has agreed to implement such
a function in wcslib. We will delete Appendix-A of G&C.  We will add
analogous distortion-correction terms to the spectroscopic projections
to be defined in a future agreement, and so the projection parameter
notation must be generalized.  We will replace PROJPii in G&C with
CPnn_iiV, where nn is the axis number, ii the parameter ordinal and V
the version code; for celestial projections the parameters would be
associated with the latitude-like axis.

(5) [New agreement] The recommendation that CDELTi and CROTA2 be
    written in headers in addition to the transformation matrix
    (CDi_j) will be deleted from the G&C text.

(6) [Technical detail pointed out by Bill Pence] The LONGPOLE keyword
    proposed in G&C probably needs to be reduced from eight characters
    to seven to enable it to include a version code.

	    -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

In preparing the above summary of our negotiations I have been forced
to infer the opinions of some of you who have been silent during the
discussions, generally by assuming that silence implies tacit
approval. Also, I know the opinions of many of you from prior
conversations over many years. If I have made mistakes in this
process, and some of you disagree with the above summary, please state
your disagreement now.  Please send such messages to me by private
Email if you are reluctant to post them to the whole fitswcs mailing
list.

If there are technical mistakes in the above summary, please post them
to fitswcs or send them to me by private Email.

If there are additional items which should be included in order to
make the mandate for Eric and/or Mark unambiguous and complete, please
post them to fitswcs or send them to me by private Email.

I have asked Doug Tody to confirm that the CDi_j notation which we are
discussing, which would replace CDELTi/PCij as used in G&C, has the
same semantics as his IRAF implementation. I expect that his answer
will be 'yes, of course', but the question is of critical importance,
because it provides a compelling reason to choose the CDi_j notation
as our canonical interchange notation for WCS.

I have asked Perry Greenfield and his colleagues to confirm that the
image geometry of ACS (see http://jhufos.pha.jhu.edu/) will be
supported properly by the proposed TAN' projection. As I noted above,
several HST instruments use the CDi_j notation; if Perry etal have any
reason to believe that the semantics of their usage are different from
the semantics of IRAF and of this proposed agreement, I ask them to
inform this task group.

I have asked the ESO members of this mailing list to comment on the
state of these negotiations. In particular, I would like for them to
confirm that the proposed WCS notations will support the anticipated
instrumentation of the VLT (see http://www.eso.org/projects/vlt/ and
http://www.eso.org/instruments/). I realize that it may be difficult
to fully answer such a question for instruments like ISAAC
(http://www.eso.org/instruments/isaac/fig1.html) until a spectroscopic
WCS agreement has been negotiated, but certainly the direct imaging
modes of the VLT instrument packages must be supported by our imaging
WCS interchange agreement.

I ask all of you to review the table column name notations tabulated
by Bill Pence in his posting yesterday 08-06 to see if there are any
obvious problems. His tabulation is an update of Appendix-B of G&C.

Ground-based optical cameras form their images through the refractive
atmosphere. In addition to any distortions which are intrinsic to the
cameras, the geometry of the celestial sphere which they are imaging
is distorted by the abberation of light. It is true that over modest
angles these physical effects can be compensated fairly well by small
contributions to the coefficients of the CDi_j linear transformation,
but higher-ordered terms would be needed to fully represent the
physics. The proposed TAN' agreement appears to me to provide
sufficient new terms to model the physics of abberation and refraction
with enough accuracy to satisfy a professional astrometrist. Can Pat
Wallace or any other fitswcs subscriber confirm my conjecture? If it
is true, then application programs which solve for the WCS of images
using reference stars could be arranged to include the higher-order
terms of the apparent place calculations in the regressions
implicitly. 

The VLT is not the only new large ground-based optical telescope which
is coming into operation. A large number of wide-field cameras will
soon be in use. Most of these have distortion properties which we can
be confident will be properly modelled by the proposed TAN'
projection, but I am concerned that a few of them might require us to
add additional terms to the DSS set. Cameras which have multiple axes
of symmetry, such as HST+WF/PCi (Holtzman etal, PASP 107, 156 [1995])
and VLT+VIRMOS (http://www.eso.org/instruments/vimos/fig1.html), are
known to be difficult cases. If any of you know of optical cameras
which should be reviewed for compatibility with CDi_j+TAN', please
post to fitswcs and talk to colleagues about the cameras. If any more
terms need to be added, let's do it now, rather than after we have
pushed this agreement through the formal FITS approval process.

The goal is to produce a summary statement of agreement on WCS issues
which can win nearly universal acceptance in the worldwide FITS user
community. When all of you are in agreement with the text, I intend to
post the final version of it to sci.astro.fits/fitsbits for public
review. The final version resulting from that review will provide the
mandate for Eric Greisen to change the text of G&C and for Mark
Calabretta to modify wcslib.

-Don
-- 
  Donald C. Wells         Associate Scientist         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