[daip] Re: VBGLU in AIPS

[Eric Greisen]

Greg Ball writes:

 > I am using VBGLU to reconstruct data which was correlated in 2 passes.
 > Both input files are in compressed format, as is the output.
 > 
 > I believe that VBGLU constructs the output file using the visibility 
 > time/baseline values found in the first input file; and that any 
 > visibilities in the second input file at *other* time/baseline values are 
 > discarded.   This just amounts to a little lost data, which is fine.
 > 
 > However, I'm not sure what happens when a time/baseline slot from the 
 > first input file is missing from the second input.
 > 
 > I would guess that the missing data is 'empty' and should have zero 
 > weight,  but I'm not sure that this is properly represented by the 
 > compressed data format.  Lincoln Greenhill and Alice Argon here at the CfA 
 > report that the half-empty visibililities have caused them trouble in the 
 > past...   however they were using AIPS version from a few years ago.

VBGLU is pretty complicated so I have not actually figured out what it
does for sure.  If you have a bug to report I can do that.

I suspect that you are right about discarding data when the second
file has data missing from the first file.  In the opposite case,
there is absolutely no problem in handling flagged data in compressed
format.  Lincoln may have had problems but his analysis of the cause
cannot be correct, independent of aips version.

Eric Greisen