[evlatests] Autophasing

[Barry Clark]

With my test Executor, autophasing of all subbands works, but not
very well.

In the implementation, there is a primary subband, which is used to
adjust the baseband delay for small delay and zero phase.  This
subband works as well as usual.

The phasing of the other subbands is done through adjustments to the
subband delays.  It seems to take a long time to settle down.  In
the example I was running, four 128 MHz subbands at C band, it was
taking about 10 scans to settle down.  But then it worked very well.
However, the phase stability was very good that morning, and there is
no guarantee that it would work as well if the atmosphere were less
stable.

Antenna 16 did not phase as well as other antennas, with frequent
errors of 10 degrees or so.  It is suggestive that it is the antenna
with the slowest delay rate.

Several times the phase of antenna 12 jumped by 112 degrees, and
immediately jumped back.  Antenna 12 is the telcal reference antenna,
and although it nominally should have zero phase and delay, rounding
errors may give it a very small one.  My suspicion is that going
from omitting the subband delay, which is done for delay and phase
zero, to putting in a very small value, results in a phase difference
of 112 degrees.  However, after I installed a trap to give a printout
if that happened, the jump did not occur again.

I collected some data that may help determine exactly what is going
on, but I haven't looked at it yet.

So, more work obviously needed, but I may commit what I have to get
most of the stuff, at least, into the standard Executor.