[evlatests] Autophasing

[Barry Clark]

The way I had implemented autophasing was to simultaneously
adjust both delay (as sent to the Widar correlator) and phase
(by strobing phases into the L302 LO).  This was not working
very well.  In particular, there seemed to be an interaction
that I did not understand between the delay setting and the
phase setting.  To simply things, I have changed the code, in
an update to be installed shortly, to first set delays, and
then to set phases.  This works more slowly but better.  I
would recommend at least five scans or subscans on the first
visit to phase calibrator to make sure things get set up
properly.  In fact the really conservative thing to do would
be to first setup on a seriously strong calibrator to get the
delays set right, then do touchups on the nearby phase calibrator
with a couple of scans each visit.

Even then there may be problems.  Every now and then one
antenna does something totally wrong, which is then corrected
when the antsol for that subscan percolates through the system.
Antenna 27 seems to do this more often than others, but two or
three others did it as well.  I am suspicious of L302 hardware.

I will
a.  Look at the data taken today to better characterize the
above random misbehaviors.
b.  Explore another phasing algorithm which doesn't use the L302.
c.  Devise a couple of tests to check on L302 behavior.