[daip] UVFITS antenna table

[Eric Greisen]

André Offringa wrote:
> Dear Eric,
> 
> I'm a post-doc in Australia working on MWA. We're seeing some issues 
> when converting CASA measurement sets to uvfits files and back, relating 
> to the antenna table and its format in the uvfits files. Randall Wayth 
> told me that you are the expert on uvfits files, so I was wondering if 
> you could help us.
> 
> The problem that we are seeing is that the AIPS and Miriad's frame of 
> reference for the antennas in the uvfits antenna table seem not to 
> correspond with what Casa thinks it is. The former two seem to use 
> geocentric, but rotated to the longitude of the array, while the latter 
> seems to assume they are in Geocentric (ITRF) without the rotation. This 
> has gone undetected for quite some while by us, because the antenna 
> positions are rarely used, but we're trying to do phase rotations etc. 
> and now are running into problems.
> 
> I found in one of your memo's (#117, "AIPS FITS File Format") that AIPS 
> indeed uses this longitude rotation in its revised version (since August 
> 28 2012), and Miriad calculates UVW's correctly from the rotated 
> positions too. However, in the FITS IDI reference ( 
> http://www.aips.nrao.edu/FITS-IDI.html ) it is specified as being always 
> geocentric. Both your memo and the FITS IDI reference specify a possible 
> "FRAME" keyword, but as far as I can tell this keyword seems not to be 
> adhered to very consistently. Casa for example writes 'ITRF' in the 
> keyword (which is not valid according to the IDI), while we saw that 
> ATCA and MWA data do not provide a FRAME keyword.
> 
> Is this problem known? Do you know what the correct way for these 
> packages is to be compatible to each other? (i.e., is this a bug in 
> Casa, or...?)

I have been on vacation so this response has been delayed.

The computation of u,v,w depends on the antenna coordinates but is not 
tied to the longitude of the array except when the array center is not 
zero.   For EVLA and VLBI the array center is zero and Greenwich is then 
the reference longitude.  I do not know from your note what CASA is 
doing (to or for) you - a display of the contents of the antenna file 
would help.  FITS-IDI was designed before we added the FRAME and XYZHAND 
keywords to the usual antenna file (both of these may not appear in 
software other than AIPS since they are new) and they are not in IDI.

In AIPS Memo 117 the red letters mean that I added text - the usage has 
been present since 1980 and was used with the VLA.  In that system the 
array center was not zero and so all antenna coordinates were relative 
to that center and X/Y was a rotated system.  When the center is zero 
the "longitude" of the array center is not defined and so is taken to be 
0.0.

I do not know what the "problem" is - I would need to see what CASA is 
giving you.  Note that CASA has no one assigned to FITS at all.  For a 
brief period a programmer (non astronomer, questionable abilities) was 
assigned to it and he did stuff but I have not gotten anyone to test 
what he did or at least to show me the results.

Eric Greisen