Various comments and typos

[AlanLoh]

General

(l. 85) 'They can be organized in the following way : catalogs of matrices of measurements for various fields and settings of the antennae pairs distances and orientation. Matrices are regulary sampled in time, wavelength and polarisation.'

'regulary'->'regularly', but I am not sure what 'regularly' means here.

(l. 93) 'Spectral and Time axis are regularly sampled but we can have gaps between time and spectral intervals. In addition, on the spectral axis different samplings may coexist.'

The two sentences are contradictory?

(l. 106) 'Moreover visibility data show some dependency of the spatial field of view and resolution with wavelength.'

This is not specific to radio visibilities (I used to work on Fermi/LAT gamma-ray data for which the PSF could go from 0.1 to several degrees). It's whenever the approximation of lambda/d ~ constant cannot reasonably hold.

UV plane

We either use meters (or equivalent) or 'lambdas' (=uv distance/wavelength).
Maybe it has already been discussed and addressed (if yes, sorry!), but here are my two cents, if we use 'lambdas' instead of meters:

• we no longer need s_maximum_angular_scale because it would be equivalent to uv_distance_min. (same for s_resolution_min but I guess we would like to keep it for symmetry purpose with s_resolution_max).
• it may better represent the PSF shape for a multi-wavelength observation

Typos

• s_fov_min (l. 238): Char.SpatialAxis.Coverage.Bounds.Extent.LowLim -> LoLim (to match other parameters)?
• uv_distance_max (l. 253): Char.UVAxis.Coverage.Bounds.Limits.LoLim -> Char.UVAxis.Coverage.Bounds.Limits.HiLim

Would be happy to discuss any of the above points :)

[BaptisteCecconi]

By the way, about the UV plane discussion, the document cited in the introduction of the Note is also using UV plane measures in wavelengths units (see, e.g., their Table 1)

[BaptisteCecconi]

About the "regular sampling" statements in the introduction, I don't know what is the ground of this. It seems to be an overstatement since we can think of radio interferometers where it is not the case. Furthermore I don't see why this restriction would be necessary.

[BaptisteCecconi]

About the varying field of view within the instrumental spectral range, as @AlanLoh said, this is n to specific to visibilities, and even not specific to radio astronomy.

[Bonnarel]

Hi Alan

General

(l. 85) 'They can be organized in the following way : catalogs of matrices of measurements for various fields and settings of the antennae pairs distances and orientation. Matrices are regulary sampled in time, wavelength and polarisation.'

'regulary'->'regularly', but I am not sure what 'regularly' means here.

OK see below

(l. 93) 'Spectral and Time axis are regularly sampled but we can have gaps between time and spectral intervals. In addition, on the spectral axis different samplings may coexist.'

The two sentences are contradictory?

This has already be rephrased by Baptiste. But anyway I can explain what was meant.
Most of the MS data we know are organized in observations/fields/spectral windows. We thought that INSIDE a spectral window for a given field, the time and spectral sampling was regular.

(l. 106) 'Moreover visibility data show some dependency of the spatial field of view and resolution with wavelength.'

This is not specific to radio visibilities (I used to work on Fermi/LAT gamma-ray data for which the PSF could go from 0.1 to several degrees). It's whenever the approximation of lambda/d ~ constant cannot reasonably hold.

Yes. this opens a discussion on two points : 1 ) How do we design extensions of ObsCore in the IVOA and 2 ) what is the extent of this proposal.
About the first point : it's true that with HEA data (gamma, X, etc...) we are facing the need of an extension and that some attributes useful for radio data may be also useful for them. But the effort for an HEA extension is not so much advanced at the moment. So I imagine finding out all the extension attributes needed for all domains could be a very long process. I think that having separated domain extensions with some logical binding between the concepts is better. This should not prevent to reuse some terms in other extensions. This would be the responsability of the TCG and the DM working group to insure that the same concept is tackled by the same attribute name.

About the second point : Even in the radio domain, some attributes discussed in this draft can be useful for single dish data. Fo example fov min and max. Or use of frequency. And also the spllitting of observations in different datasets for complex observations with different fields and inhomegeneous spectral windows. On the other side the uv plane characterisation attributes could be useful for optical interferometry. So wher do we stop the scope of this draft ? generic interferometry ? all radio domain ? radio interferometry ? First discussions with optical interferometry people doesn't show a strong requirement for them to describe their visibility data. On the other side restricting to the sole radio interferometry domain may require to duplicate many things in a single dish extension. So maybe an overall radio domain extension with optionnal interferometry attributes would be the best solution. This would require some significant change in the text.

Thoughts ?

UV plane

We either use meters (or equivalent) or 'lambdas' (=uv distance/wavelength). Maybe it has already been discussed and addressed (if yes, sorry!), but here are my two cents, if we use 'lambdas' instead of meters:

• we no longer need s_maximum_angular_scale because it would be equivalent to uv_distance_min. (same for s_resolution_min but I guess we would like to keep it for symmetry purpose with s_resolution_max).
• it may better represent the PSF shape for a multi-wavelength observation

Typos

• s_fov_min (l. 238): Char.SpatialAxis.Coverage.Bounds.Extent.LowLim -> LoLim (to match other parameters)?
• uv_distance_max (l. 253): Char.UVAxis.Coverage.Bounds.Limits.LoLim -> Char.UVAxis.Coverage.Bounds.Limits.HiLim

Would be happy to discuss any of the above points :)

[Bonnarel]

About the varying field of view within the instrumental spectral range, as @AlanLoh said, this is n to specific to visibilities, and even not specific to radio astronomy.

Hi Baptiste,
As I discussed in my answer to Alan, we should consider extending to the overall radio diomain. I'm not sure we will converhe if we try to have a single extension for many domains. This doesn't exclude having the same attributes in different extensions

[Bonnarel]

About the varying field of view within the instrumental spectral range, as @AlanLoh said, this is n to specific to visibilities, and even not specific to radio astronomy.

I already answered something to Alan. Can you point to examples where the spectral and time axes are sparsed as the uv plane is ? I thought we always got some homogenous chunks at some levels

[Bonnarel]

By the way, about the UV plane discussion, the document cited in the introduction of the Note is also using UV plane measures in wavelengths units (see, e.g., their Table 1)

In my opinion units in the uv plane should be given in lambda. Baselines in meter should be for the description of the facility and could be useful for a priori caharcterization limits if we don't have the uv plane description of each specific observation

[Bonnarel]

UV plane

We either use meters (or equivalent) or 'lambdas' (=uv distance/wavelength). Maybe it has already been discussed and addressed (if yes, sorry!), but here are my two cents, if we use 'lambdas' instead of meters:

As I wrote to Baptiste. I think the correct way to describe uv coverage is to use lambdas.
But instead or in parallel to this description we may describe the facility and the the baseline will be in meters. It is somewhat redondant I admit. But I don't think we will be able to get the uv coverage in all, the cases from the providers.
The angular expresssions (s_resolution and s_maximum_scale) are there for a priori comparison with standard imaging data. See below

• we no longer need s_maximum_angular_scale because it would be equivalent to uv_distance_min.

just by inversing (from lambda to angular units) correct ?

(same for s_resolution_min but I guess we would like to keep it for symmetry purpose with s_resolution_max).

I realize we may need s_maximum_angular_size_min and s_maximum_angular_size_max to take into account the dependancy from wavelength

• it may better represent the PSF shape for a multi-wavelength observation

OK . Interesting

Typos

• s_fov_min (l. 238): Char.SpatialAxis.Coverage.Bounds.Extent.LowLim -> LoLim (to match other parameters)?
• uv_distance_max (l. 253): Char.UVAxis.Coverage.Bounds.Limits.LoLim -> Char.UVAxis.Coverage.Bounds.Limits.HiLim

OK thanks

Would be happy to discuss any of the above points :)

Hope taht other co-authors will enter the discussion