Spectral 2.0 - Document update

[Petr Skoda]

Hi all

I did not want to repeat myself for the 100th time (but it seems its 
necessary) - so in summary

So far I did not see proper description (and singe working service) using 
proper SDM elements to describe normalized spectra since the time when I 
proposed NORMALIZED  in FLUXCALIB (in SSA) and wanted to cover this in SDM 
(the arith.ratio was there but I have noticed in the meantime that people 
have different idea about what is normalization to continuum.
With some hacks we succeed running DaCHS - delivering normalized spectra 
or even computing them on the fly.  It allowed a lot of interesting 
science ( more on this as reply to Igor's mail as well).

But so far we had to ignore the fact that the flux axis is simply faked - 
and even in DaCHS it needed hacks to display some reasonable axis 
description (as it is internally bound to SDM and so inherently 
insufficient to describe this properly).

So I was hoping that the work on SDM2.0 and SSAP v2 will correct this 
historic omission... Thats why I simply asked recently - supposing its 
naturally included .


> Thanks for the questions, I like to get input on use-cases which aren't
> (perhaps) covered by the current model.  The scope and coverage of the
> Spectral-2.0 model hasn't changed in some time.

That is what I am worry about !

Already in Heidelberg there was a quite cleanly demostrated by A Pallacios 
as I remember that Spectropolarimetry na basically all the high 
resolution spectral analysis thay wanted to do on Narval data was not 
possible due to the need of clearly describe the continuum normalized 
spectra - the VOSpec did not expect empty units and it did not work with
no Dimensional factors ....

In fact in SPLAT there is no proper solution as well.

I think that there is a good time to solve this issues by giving explicit 
recommendation how to do it properly - e.g as an special appendix of SDM.

The contiuum is explained as a ratio of two spectra - so the problem is 
how to reference the second one. Which PUBDID should it have - or is it 
something like the zero point deserving special keywords ?
How to serialize this in FITS ?

So the problem is

1) can we freely use EMPTY unit description  in FluxAxis ? What about ucd 
for this ?

What  Mireille suggests was exactly my good feeling   that it is clear and 
natural:

Char.FluxAxis.CalibrationStatus='NORMALIZED'
Char.FluxAxis.unit=NULL  or 'unitless'
Char.FluxAxis.ucd=phot.flux;arith.ratio;

BTW  will this work as well for ObsCore ucd ?

How will the NULL be expressed in Osuna-Salgado conversions like
SpectralDataset.fluxSI  ?

And in SDM2.0 Appendix B.1 is only
phot.flux.density;arith.ratio

  it is of course to discussion if density is required or not .

But the problem is

phot.flux.density;em.wl; spect.continuum

Continuum only
erg.cm**(-2).s**(-1).angstrom**(-1).arcsec**(-2

It is referring to continuum as a absolute flux-calibrated flux/intensity.

But in optical ground based - namely stellar astronomy it is more useful 
to reduce spectra to sum of ADUs on "flux axis" - in literature people 
sometimes write relative intensity. Then follows  a subjective 
black magic procedure the result of which is typically some polynomial 
which is divided into the observed (aperture extracted) spectrum.
The result is then called contiuum normalized (or "rectified") spectrum 
where the intensity out of spectral lines is put to  value of 1.0 .

so

2) How to describe the function - e.g. polynomial to refer to - it does 
not have dimension of absolute flux as suggested in ucd above but it is 
abstract fit of data which are already just in ADUs...

3) How to serialize and express in FITS keyword ?

Ususally people do the rectification by tweeking some polynomial 
parameters in IRAF or SPLAT or other tools and produce new spectrum that 
may have something like original name + some suffix or new name like 
normxxxx ....

Some resources are implicitly already normalized but the original 
extracted - ADU  spectra are not available (hidden as a by-product of 
pipeline) - example it used to be the HARPS or UVES spectra on ESO SSAP 
resources (working until last year) - part of FEROS and HEROS, NARVAL 
spectra etc 
...

IMHO one possible solution is datalink - we already try to use it to 
select between normalized and original spectra. In FITS it might be some 
keyword describing the polynomial coeeficients or second spectrum as a 
computed points - in another extension or as a second axis using 
degenerate NAXIS2

But IMHO this is so common (in stellar ground astronomy sure) and so 
specific case that desires explicit explanation in the documents.

>
> I'd like to spend some time with you in Sesto to understand the particulars
> of what would be needed to describe the continuum.  I think Mireille'
> suggestion for the UCD and unit sounds about right..

Sure I suppose to discuss this and in fact Markus must be involved as well 
- either as a DaCHS developer or as a registry head ...
(what IMHO many people want to do is to search in registry for "contiuum 
normalized spectra" to compare e.g spectral types ..., estimate emission 
heights ...


> There is this record in the FluxAxis ucd list for ucd, description, unit
> respectively.
>    "phot.flux.density;arith.ratio"   "Flux ratio of two spectra"  (empty
> string)

see above

*************************************************************************
*  Petr Skoda                         Tel   : (323) 649201, l. 361      *
*  Stelarni oddeleni                          (323) 620361              *
*  Astronomicky ustav AVCR, v.v.i     Fax   : (323) 620250              *
*  251 65 Ondrejov                    e-mail: skoda at sunstel.asu.cas.cz  *
*  Ceska republika                                                      *
*************************************************************************