[daip] Do the REFANT values stored in the CL table matter?

[Eric Greisen]

On 02/21/2019 10:50 AM, Michael Bietenholz wrote:
> Hi Eric:
> 
>    I'm trying to sort out a messy data set where more than one refant
> is required.  I realized that I don't really understand how refant
> works - do you know where the refant values stored in the CL table
> matter?  Is my thinking below correct?
> 
> 1) when visibility data from e.g. baseline 1-2 calibrated by
> UVGET,DATCAL etc, it needs to get gains for ant1 and ant 2 from the CL
> table.
> 
> Now if the CL table gain phases are formally relative to that of the
> refant, the CL-table gain of the refant itself should be identically 0.
> This doesn't seem to be true.  Also, DATCAL etc should only
> use the CL-table gain phases directly in caes where both refants
> are the same (not sure whether that is true).
> 
> Also - it seems common, for e.g. ANTPOS corrections, for the refant
> in the CL table to be "0".
> 
> 2) I'm thinking that the CL gain phases are taken as "absolute", and
> applied directly without worrying about the CL-table refant.  REFANT
> *does* matter to CLCAL, which tries its best to rotate all phases in
> the SN tables to the same refant, then applies them.
> 
> 3) in actual fact, the CL-table gains are of course also relative, and
> there is no change in the calibrated data if rotate all the CL-table
> gains by an arbitrary phase, but I think they are relative to some
> common point (not necessarily any one antenna)
> 
>      thanks!           michael

Yes reference antennas do matter a little.  The cal routines will not 
interpolate between 2 samples with different reference antennas.  In 
that case, it takes the closest one in time, so jumps half way between 
the 2 cal values.

The CL table builds over various processes.  If you change source 
position or antenna position, the relevant phases are simply written to 
the CL phases with normally 0 as refant and a separate phase in all 
antennas varying with time, source, and antenna.  This is from CLCOR 
basically.

It is important, when interpolating SN values to the CL table to have 
all the same refant (except for RLDLY which wants to put a constant in
all L delays).  If one puts a constant in all CL phases (right and left) 
then nothing changes.  If the constant is only in left, the the 
crosshands change.

Does this help?

Eric