[evlatests] evla-only subarrays

[Ken Sowinski]

The Modcomps only record data when it is determined that at least
one antenna is on source after a source change.  This presents a
problem with EVLA only subarrays because the Modcomps insist on
thinking that EVLA antennas are never on source and thus data is
never recorded.  I have changed the behavior of the Modcomps to 
make this less of a problem.  The modified task is not yet a formal
part of the system but can be made available upon demand until a
software update is performed.  If I have sufficient confidence in 
it I will probably let it run routinely.

VLA only subarrays and mixed subarrays will work as before; data
will be recorded once it is determined that at least one antenna
is on source.  The change is that subarrays of only EVLA antennas
will be deemed to have an antenna on source immediately after a
source change.  There are two consequences, one trivial and the 
other of some consequence.  First, data will be written while the
antennas are slewing; if the EVLA flagger is running this data
will be flagged.  Second, pointing, delay and focus cycling will
commence as soon as data is written and so the part of the cycle
in which data is first unflagged will not be predictable.  This is
most important for pointing scans.  Pointing scans in observe
programs should allow enough time for two pointing cycles in order
to be sure of getting at least one valid solution.  Pointing scans
in the pointing model determination files will not work as reliably
as we are used to.  I recommend including at least one VLA antenna
in such programs.  Eventually the pointing solver in telcal will be
able to deal with this more gracefully, but I want to keep it compatible
with the Modcomp version while it is still under development.

The caveat decribed in the previous paragraph applies to raster scans
for mosaicking and holography as well.  Here it is potentially a 
more serious problem because the initial part of a track may be lost
to the slew time.  The obvious remedies are to include a dummy scan when
necessary to position the antennas, or to make the raster lines longer
than necessary to absorb the slew time.  This may be a sufficient 
argument to make the EVLA work like the old system for raster scans
by not starting to count till the antennas are on source.