[evlatests] Numerology and a historical perspective

[Barry Clark]

The heartbeat frequency at the antennas is 19.2 Hz.  This was chosen,
rather than 20 Hz, to avoid a simple ratio relationship with the (then)
ubiquitous 60 Hz.  Heartbeats are synchronized such that a heartbeat
pulse occurs on an even TAI 10s.

The L302 synthesizers are driven by a 128 MHz signal.  To be synchronized
reliably, they must be commanded to sync at integer number of cycles of
this rail.  This occurs on every third cycle of the 19.2 Hz, every 
0.15625 seconds.  The Executor software selects appropriate heartbeat
pulses to synchronize the L302s.

The VLA correlator is commanded every 10 seconds, with commands taking 
effect on the even TAI 10 seconds.  EVLA scripts are in UTC; the .obs
files from which they are (usually) derived are in LST.  When making 
scripts for the EVLA Executor, obs2script caters to that peculiarity
of the correlator, forcing source changes to occur on the TAI 10s
(with a few interesting exceptions).

Widar has its own numerology.  Its heartbeat is 100 Hz, anchored to UTC.
When we run hand edited scripts that are not synchronized to 10s TAI,
we see phase jumps, almost certainly due to improper truncating when 
these various periods do not line up.

Possible fixes:

1.  Require observations to be synchronized to the shortest common multiple
of the intervals.  This is 1.25 seconds.  Problem:  this is too long for 
OTF mosaicing.

2.  Change the antenna heartbeat; make the 19.2 Hz change to 20 Hz on 
command.  Problem:  the L302s are sensitive to the line lengths of the 
heartbeat signal.  Problem:  getting the L302s to work reliably again 
after changing to 20 Hz might be traumatic.

3.  Software fix in Executor; the commands to L302s could be synchronized
differently from those to Widar.  Problem:  something will be doing the
wrong thing for a few milliseconds at source changes.

4.  Software fix in Widar.  The information to correct the phases is 
available in the message to Widar.  Problem:  as above,  something will 
be doing the wrong thing for a few milliseconds at source changes.