[fitswcs] I've got this FITS WCS matrix

[Malcolm J. Currie]

> The conclusion for this CDi_j matrix is that it cannot be decomposed
> into CDELTi and CROTAi cards in the AIPS convention.  This is apparent
> just by looking at the signs of the CDi_j matrix elements.

Indeed.  This doesn't help my problem.  It does partially answer the
question.  As far as the FITS definition is concerned, a solid-body
rotation can't be derived, and it's up to the application to decide what
to do.  I doubt that this arrangement was intended.  I can try Doug's
approach, once I have the full details, but that must surely use only
part of the information.  Failing that I have to assume a default
rotation angle for the instrument.

> Suppose that a CDi_j matrix is the product of a diagonal (scaling)
> matrix times a rotation matrix.

Like CDELTi and the PC matrix.

> Either the signs of the elements on the top row of the rotation matrix
> should differ xor those on the bottom row should differ (not both).

Yes.  I understood this from the rotation matrix.  It would be
obvious had the keyword order been CD1_1, CD2_1, CD1_2, CD2_2.  Being
used to Fortran order, viewing by inspection made it appear as if the
signs were the same in each row, but different between rows.  A case of
misdirection.

Thanks for spelling it out.  It could be useful when I contact the
Gemini people about it.

Malcolm Currie
Starlink Project