[Heig] How to describe and access response functions in the VO ?

[BONNAREL FRANCOIS gmail]

How to describe and access response functions in the VO ?

Below I try to consider some aspects which are not specific to High Energy.
Response functions give us information on the way an observation chain 
react to an initial signal from astronomical space
This can be used to directly transform raw data and obtain some 
calibrated measurements.
Or degrade some assumed sky signals trying to match the raw observation. 
And select the best matching assumption as the calibrated measurement.

In both cases I think we can call that "calibration".

Now I think we can face two different kinds of response functions :

        - some of them are obtained "a priori" and can be used for many 
observations (let's talk about a flat field or a dark in optical imaging)

        - some of them are estimated  from the observation itself and 
part of it, but different from what is coming from the sky.

        In the first case, up to now, the classical way to describe and 
access it was through DataLink.
        Maybe we can wonder if we want a more direct discovery ?
        For example to be able to go back from such a generic response 
function
        to all the observations which made use of it ?
        In that case a specific table describing the response function 
with appropriate parameters
        describing the response could be defined and made accessible via 
TAP.
        A DataLink service could make the reverse link from the response 
to the observation (probably with a new semantics term)  ?

         In the second case the response is actually part of the 
characterisation of the dataset.
        Characterisation abstract model 
(https://www.ivoa.net/documents/latest/CharacterisationDM.html)
        is the basis of the characterisation class in ObsCore but 
contains much more.
        We have this feature of "details levels" in the characterisation.
        The first level is the simple location in the parameter space.
        The second level is made of the bounds
            (limits ensuring that outside them there are no relevant 
measurements in the dataset),
        The third level is the support which matches much more 
accurately where we have relevant data
        and where we do not.
        The fourth level was supposed to be the actual response for each 
axis estimated for the considered dataset.

         This not only applies to the coverage but also to the 
resolution (and in this latter case it's close to the psf).

           I already discussed these things with Catherine 15 years ago.

        But IVOA  never  succeeded to work out this level 4 in a 
consistent way.

         The important point is that it is PART OF THE DESCRIPTION OF 
THE DATASET.


         To discover and access this kind of description  can be done :

                   - by specific additional parameters  describing the 
response

                   - by providing links to the map of the response

          As we may have a lot of response maps for the same 
observational dataset it's better to have this in a specific table and 
make a join using obs-publisher did.

          Or to use DataLink if it is enough

Cheers

François