relative fluxes

[Petr Skoda]

> Now, at ESO we have some spectra which are calibrated in a RELATIVE way
> (i.e. the ratio between any two points gives the right answer) but their flux
> is off by an undetermined constant multiplicative factor. Therefore the 
> spectra


If it is really just offset in flux it is similar
  problem as  with NORMALIZED.

When I asked for this some time ago the basic recommendation was to leave 
the string empty. It should be also added the normalization (in your case 
reference number - scaling factor) function as part of the description but 
I am afraid it was not 
studied in detail ( the only hint by Jonathan was to use
phot.flux.density;em.wl;spect.continuum  or BackgroundModel Object array
or the arith.ratio prefix .

In SDM section 4.6.5 is said the units should be blank.


> end up having "flux" values between e.g. 0.0 and 0.6.

I see 0.0 to 0.06 !!


> As an example, please see this preview:
> http://archive.eso.org/~amicol/tmp/relative_flux_spectrum.png

I am afraid your data are not flux calibrated at all. Just by first look 
it seems to be like output from reduction pipeline of some two-channels 
spectrograph - I guess echelle one (and my further guess is UVES) mand the 
vertical (Not FLUX) axis is just the value of extracted ADUs which depend 
on with of separated echelle orders, the optimal extraction settings etc 
...
The data are just divided by flat field and so I guess it should be stated 
as UNCALIBRATED  as there is the "unspecified coordinate-dependend 
function" - it is sometimes named "response function" or global 
sensitivity function.

I would put vertical axis units empty.


> We could hence set the units to be "erg/cm**2/s/Angstrom"
> but I would be very much afraid of the astronomers' reaction

No you cannot compare this - suppose with theoretical model.
As I was stressing many times you can only compare in VO data that are of 
similar nature - i.e. you may normalize it by fitting of pseudo-continuum 
) - 95% od stellar ground-based optical spectroscopy or in asbolute flux (
all space and part of ground based data (only several nights at best sites 
allow good absolute flux calibration).

Anyway I am glad that you have opened this questions again as I feel the 
whole VO spectral stuff is still very confusing for ordinary consumers - I 
tried recently look at many SSA services - in fact only space missions are 
comparable (overplottable) in SPLAT or VOSPEC - most of optical spectra 
from ground based projects are just in the instrumental (as your case) 
flux - it is UNCALIBRATED.

But you can do nice science on them - measuring the line positions and if 
you want you can normalize it yourself - or approximate just arround 
particular line and you get Equivalent with or just plot changes of 
profiles.


> They would normally not get to see the calibration status (RELATIVE),
> or even if displayed, they wil not understand the meaning of it, will they?


Even if they had seen the status they would not be able to overplot in 
VOSpec of SPLAT these spectra with other without further processing.


On the other hand - in the VO client you have to be able to select 
FLUXCALIB type and if you want to compare fluxes you use ABSOLUTE - it 
will give you all for building SED. IF you select ANY or RELATIVE you are 
expected to treat data yourself (e.g. zoomimng to different curves and 
measuring line positions..). But when You give NORMALIZED you should see 
the vertical scale from 0 to 1+ and all spectra should fit in this range.
AS I said it depends on your type of science (SED or line profile 
variations) and it has to be decided by scientist.

The VO should just help him to select the relevant data.

> Finally, whatever the outcome, it would be extremely nice if the SpetrumDM + 
> SSA,
> presumably within an associated note or tutorial,
> could describe this kind of scenarios and provide suggestions to the
> diligent but VO-unaware (or even VO-aware, but very much undecided, like me)
> data providers.

Nice idea - I think there should be some use cases of typical usage of 
different kind of spectra.


> The worst could happen is that different  data providers will
> describe the same situation in different ways, hence hampering 
> interoperability.

iT HAS BEEN THE CASE ALREADY

(just look at the mess when you query all SSA services)

Petr


*************************************************************************
*  Petr Skoda                         Phone : +420-323-649201, ext. 361 * 
*  Stellar Department                         +420-323-620361           *
*  Astronomical Institute AS CR       Fax   : +420-323-620250           *
*  251 65 Ondrejov                    e-mail: skoda at sunstel.asu.cas.cz  *
*  Czech Republic                                                       *
*************************************************************************