[mmaimcal]Re: ACA Phase Calibration

[Mark Holdaway]

> Hi Mark,
> 
>     I fully agree with the conclusions, but someone should work on
> proper use of a WVR for the ACA.

Perhaps that is a job for a Japanese collaboration with Richard Hills
(here I am giving work to everyone but me!)

> There are two reasons for that:
>      - Phase calibrating the ACA will too frequently would INCREASE the
> phase noise on these short baselines.
>     (essentially you add up the phase fluctuation on the calibrator to
> the ones on the source, and they are uncorrelated
>      on short baselines and short timescales)

Ah, yes!  Thank you for this point.

>     - Yet, with the WVR, the maximum time between calibration is given
> by the WVR stability. One should look
>     whether there is any conflict with the above minimum time between
> calibrations...
> 
> More generally for ALMA, improper use of fast switching will increase
> phase noise on short baseline. Can we solve for phases on different

I think this is not correct -- when doing fast switching, the
short baseline's phase errors follow the structure function,
and the longer baselines have a phase error equal to the structure
function at the effective calibration baseline = (vt/2 + d).
This is an experimental result (ie, MMA Memo 126, which seems
to be unavailable from the web;  MMA Memo 262, figure 7 --
oops, that is for self-calibration, though.)  --  anyway, it is
possible that it would make things worse by sqrt(2), which might not
show up so clearly on the typical log-log plots as in Memo 126.


> timescales for short and longer baselines ? More precisely, can we do
> that without induly breaking phase closure ? There is a some research to
> be done in an optimal calibration technique from this point of view...
> 
> A related point: I did not get how you would increase the S/N on the ACA
> antenna gains by not solving for the 12-m antenna gains when they are
> cross-correlated together. How do you get the gain of these 12-m

For example, lets say you are doing a pointing calibration, you could
hold the 4 12m dishes fixed and just solve for the gains on the
ACA antennas -- if you don't need to solve for the 12m pointing,
you win big.  It is an unproven algorithm, and there may be some tricky 
details involved, but I think it could sometimes speed things up.

   -Mark

> antennas then ? And how does the error on this gain (which necessarily
> exists) reflects on the error on the ACA antenna gains ? Doesn't that
> cancels to some extent the improvement obtained in S/N ?  I believe you
> should detail your calculation there.
> 
>         Stephane
> 
>