Specview with VO interface released

[Philippe Prugniel]

Ivo Busko wrote:

>Hello, Philippe
>
>Overploting models is something that can be done right 
>now in specview. See this page:
>
>http://specview.stsci.edu/javahelp/Templates.html
>  
>
I have seen this feature in action during the AVO science demo!

>However, doing quantitative comparisons, as required by a 
>fitting algorithm, is another matter. There are three 
>parts to this problem:
>
>1 - we should be able to access model data and make it 
>available to the specview fitting code in an efficient way.
>It doesn't matter if we use a SSAP-type access, or a web
>services type of access, in both cases we should be able
>to access the data with queries that are based on the model
>physical parameters. A caching mechanism should probably
>be necessary as well in order to prevent duplicate retrievals
>from the remote site. 
>  
>
An alternative is to pass the spectrum to a remote service doing the fit.
In that case we just have to send one spectrum (no numerous retrieval 
for each set of parameters).
This also allows to make fits which are more complicated than least 
square minimization.

>2 - one must add a "spectral component" to the existing 
>library of spectral components in specview. 
>  
>
Do you mean that a spectral component could be, for example, a synthetic
stellar population specified by some physical parameters, and retrieved 
remotely?
(to make cz or flux adjustements do not require multiple access to the 
remote access
and would be useful, even if we do not make a fit of the physical params 
in specview)

>3 - the actual data being fitted should provide "resolution"
>or "bandpass" information for each data point, so the proper
>resampling and chi-squared computations can be performed. This
>is addressed by the Resolution and/or Accuracy elements of
>the SED data model, but implementation is still on the air.
>  
>
As far as I understood, describing the resolution of each point of a 
spectrum
is beyond the scope of SED-DM 1.0. But a precise description of the line 
spread
function (width AND shape) will be needed at some point...

>Can we populate those elements on an ad-hoc basis, using, say,
>the distance in wavelength between successive data points in
>a spectrum as a measure of the bandpass of each data point? 
>  
>
Probably we should do like that: guess the resolution from what we are 
giving
you (use sampling, or a general info about resolution, like FWHM).

Philippe.