[Passband] diverting into SEDs

[Martin Hill]

Alberto Micol wrote:
> 
> I do not know for the radio domain, but certainly in the optical
> I would claim that the difference between an SED and a spectrum is
> that to describe the spectrum one needs a dispersion relation; i.e. not 
> all the spectra
> are regularly spaced in "fravergy". A dispersion relation is not 
> available for a SED;
> the SED points come from different catalog/images/publications etc.; the 
> spectrum comes
> from dispersing the light using a grating or a grism or a prism or a 
> combination of those,
> whose characteristics are known and described via the dispersion relation.

So far I see an SED as being a collection of 'SED Measurements', each 
measurement being a Flux, a Passband, some kind of Frame of reference, etc.  It 
may be that an SED could include a spectrum; if someone has for example measured 
an optical spectra and wants to combine it with an X-ray count.  But I suspect 
there will be many more things required in SEDs, so I would like to leave them 
for a bit...

> 
> 8.- Atmospheric conditions; airmass and seeing are needed to 
> characterise a SED measurement; should they end up modelled in the 
> bandpass?

Only in so far as they reflect passrate (throughput).  Note that later we can 
make 'atmospherisc characteristics' that include passband info.  But for now 
that is a different model.

> 
> 9.- When playing with SEDs (a la AVO demo) it is extremely important to 
> pass around the bandpass info; not only for describing the SED but to 
> allow for more sophisticated analysis, like for example
> when an SED is to be compared with a library of template spectra for 
> classification purposes.
> AVO and GAVO are putting together a system to actually classify SEDs of 
> planetary nebulae
> by matching the SED with models of stellar atmospheres. The models are 
> to be multiplied by
> the filter throughputs before the comparison can take place.
> The SEDs should hence always bring along a URI(?) defining (if not 
> pointing to) the
> bandpasses of each of the SED points.
> It would be nice to have a service able to resolve those URIs to return 
> all the bandpass
> characteristics and throughput.

Be careful here as you're starting to mix 'representation' with the model.  When 
we come to modelling SEDs we simply say that the SED must/could 'have' stellar 
atmosphere/filter/etc data.  Whether this 'haved' data is remotely accessed or 
passed along with the rest of the SED is a representation issue we need to sort 
out (later).

> BTW, I remember a Jonathan's document describing a bandpass datamodel.
> What is the status of that document? Is this exercise a way to continue 
> Jonathan's work?
> I should read it again now, after few years ...

It's very readable, but covers a lot more than bandpass, and I feel (Jonathon 
may disagree!) that it tends to mix up all the issues of modelling SEDs.  For 
now, I just want to have a look at a part of it, then work up.  Also other 
people have also done work, and I want to try and gather it all together from 
the different areas.

> 
> That's all I can think of, after a nice free day spent with a dozen of 
> screaming 7to9 years old
> kids, playing bowling and trying to make sure they get back to their 
> parents safely (for the record, I almost succeeded :-).

Almost?  Did you lose one or two?!


-- 
Martin Hill
www.mchill.net
+44 7901 55 24 66