VO Spectrum FITS serialization
jcm at head.cfa.harvard.edu
Mon Oct 16 17:36:58 PDT 2006
Following discussions at ADASS I suggest one further change
to the FITS version of Spectrum (no change to the model itself)
This comes from Randy and Inga...
Problem: which column is the X axis data?
Current solution: it's the one with TTYPEn = 'SPCO'
Issue: SPCO sucks as a column name. We'd like to be able
to use 'WAVE' or 'FREQ' etc depending on the ucd
Proposed new approach: TTYPEn value should be the FITS Paper 3 CTYPEn
string, or at least the first 4 chars should be. So, 'WAVE', 'FREQ',
'ENERGY' are allowed. But, we still want it easy to find the X axis.
So we define two new generic keywords
giving the UCD and the UTYPE for each column. Then, the X axis is
the one with TUTYPn = 'Spectrum.Data.SpectralCoord.Value'
and what kind of X axis is it? TUCDn tells you that, and so
does TTYPEn if you prefer to use FITS rather than VO diagnostics.
At this point, SPCO_UCD and FLUX_UCD are no longer needed
(the Y axis is the one with TUTYPn = Spectrum.Data.FluxAxis.Value'
and its UCD is the corresponding TUCDn.
Anyone have a big problem with this?
(PS spotted a goof above: X axis UTYPE is Spectrum.Data.SpectralAxis.Value
in the latest version).
Also, some of us are meeting tomorrow morning at ADASS to discuss
Spectrum. Any feedback on the latest docs before 1500 UTC would be
particularly welcome (especially if it is positive!)
More information about the dm