An idea for a table of WCS information

In DEIMOS a typical single exposure may need up to 24 different world coordinate system transformation. This is one for each CCD transforming from pixels to sky, slitmask, and dewar focal plane coordinates.

While it might be nice to place this information into FITS IMAGE HDU keywords, some of the world coordinate system (WCS) keywords themselves are already doubly-indexed. The 8-character limit does not provide enough keyword namespace to permit sensible encoding of two more indices. Such an attempt inevitably produces numeral-jammed keywords whose structure defies trivial parsing. Furthermore, the 8-character constraint creates field widths with uncomfortably small upper limits on the indices.

The archival FITS format for DEIMOS will store data from separate CCDs into separate FITS HDUs. This permits the WCS information to be stored in precisely the manner prescribed by the draft WCS standard document. However, for purposes of the DEIMOS simulators and quick-look reduction software it is desirable to store all WCS information in a single configuration file.

A possible solution comes with a phrase we have found ourselves using over and over in the DEIMOS project.

Don't use keywords to do the job of a table.
The information could be stored in either a FITS ASCII table or binary table and appended to the relevant FITS image files. There are some advantages to each method. These tables could also be stored in a relational database and expressed in a plaintext form suitable for manual editing.

Question: Should these tables also contain URIs in the sense used by the proposed FITS Grouping Convention? If they do then the image-based WCS rows could point directly to the images that they describe.

A FITS ASCII table containing WCS information

A FITS Binary table containing WCS information

An editable plaintext representation of the WCS table

Representative Values of the various columns

CCDPOS
An indication of the position within the detector mosaic occupied by the CCD in question. For DEIMOS this will typically be an integer in the range 1 to 8. The value may also be something like 'GUIDE' to indicate that the WCS applies to the TV guide camera. These values require further consideration.
AMPPOS
An indication of the position within the detector mosaic occupied by the amplifer in question. For DEIMOS this will typically be 1 or 2.
CTYPE
RA---TAN, DEC--TAN
Equatorial coordinates as in the WCS draft.
GLON-TAN, GLAT-TAN
Galactic coordinates as in the WCS draft.
ELON-TAN, ELAT-TAN
Ecliptic coordinates as in the WCS draft.
MASK_X, MASK_Y
Coordinates on the slitmask.
DEWAR_X, DEWAR_Y
Coordinates on the dewar focal plane.
The use of the tangent plane (-TAN) projection above is merely illustrative. The projection best suited for DEIMOS images has not yet been determined.
CUNIT
For transformations to the sky as described in the WCS draft this will be degrees ('deg '). For transformations to slitmask or dewar focal plane this will be meters ('m ').
RADECSYS
The values 'FK5 ', 'FK4', 'FK4-NO-E', and 'GAPPT ' are as in the WCS draft. Values which are specific to DEIMOS include
SLITMASK
transformation is to the slitmask coordinate system
DEWAR
transformation is to the dewar focal plane
In addition, for spectral purposes we will need a form of WCS which transforms from a FITS pixel position to a slitletID and wavelength. This will require an additional table not yet specified.
DOMAIN
This column is proposed to permit the transformations to apply from coordinate systems other than FITS pixels. Its implications and possibilities are not yet clear.

The DOMAIN is the coordinate system from which the transformation is being done. Without this mechanism it would be necessary to use FITS pixels as an intermediate step in any coordinate conversion. Possible values are

FITS_PIX
Transformation is from FITS pixels. This is the default.
CCDPIXEL
Transformation is from unbinned CCD pixels.
SLITMASK
Transformation is from slitmask coordinates.
DEWAR
Transformation is from dewar coordinates.
Note that there is currently no way to specify the units of the domain. Thus coordinates from each domain must be given in some established canonical system.

Back to the DEIMOS home page.
Steve Allen <sla@ucolick.org>
$Date: 1996/08/19 06:11:11 $