Reference frames

[kettenis]

The list of possible reference frames includes the capitalized casacore reference frames, but I think some of those map directly onto the astropy ones (e.g. 'GEO' is the same as 'gcrs')?

Is 'icrs' indeed the most sensible choice for the default velocity reference frame?

The linked astropy documentation still has a warning that SpectralCoord is experimental (stating that it is new in Astropy v4.1). It is probably not wise to to base MSv4 on something that is expected to change at some point. Maybe clarification should be sought from the astropy developers about this?

[kettenis]

Also ITRF and ICRS have multiple realizations. The difference might be important for (geodetic) VLBI, although for current astrometric VLBI I don't think there is an expectation amongst users that things are labelled according to the actual realization of the reference frame that was used.

[Jan-Willem]

• From https://casadocs.readthedocs.io/en/stable/notebooks/memo-series.html#Spectral-Frames GEO definition is:
  Geocentric, referenced to the Earth’s center. This removes just the observatory motion.

  From https://docs.astropy.org/en/stable/api/astropy.coordinates.GCRS.html:

  A coordinate or frame in the Geocentric Celestial Reference System (GCRS). GCRS is distinct form ICRS mainly in that it is relative to the Earth’s center-of-mass rather than the solar system Barycenter. That means this frame includes the effects of aberration (unlike ICRS).

I remember having a conversion about this. @FedeMPouzols do remember what the objection was to this suggestion? I think part of the problem was that Earth’s center was not clearly defined in the CASA docs.

• In https://docs.astropy.org/en/stable/api/astropy.coordinates.SpectralCoord.html the SpectralCoord class is listed as experimental. It was introduced in Astropy 4.1 and Astropy 7 has been released. The only thing we use from this class is the following nomenclature (the class itself is not used in the schema):

  • The measures type is spectral_coord (named after the class).
  • The observer key that takes a reference frame string (the center of the reference frame is the observer's location). The reference frame strings are used by other non-experimental Astropy classes. Any of these naming conventions changing is highly unlikely.

• We could add a key for the realization. Since no MSv2 has this information 'unknown' would also be required as an option.

  • Example: observer='icrs', frame_realization='icrf1'/'icrf2'/'icrf3'

[FedeMPouzols]

Yes, I think that it was not clear whether the definitions are exactly the same. More than an objection I think that it was a lack of confidence that we can safely translate GEO=>GCRS (once done we'd forget about the hidden casacore "GEO").
There is also the caveat that GCRS in astropy has some flexibility and parameters that allow for example "This frame also includes frames that are defined relative to the Earth, but that are offset (in both position and velocity) from the Earth." which could also introduce ambiguity in the translation.