[evlatests] L352 round-trip phase meters, etc.

[vdhawan at aoc.nrao.edu]

	I report on comparisons of EVLA fringe phases with the round
trip delay variations.  Since all four L352 units were on antenna 16,
the RT data refer to 16 only, but the phases of all EVLA antennas were
examined and I expect the conclusions are generally true.  The L352
meters will soon be dispersed to the 4 EVLA antennas, and I will
update as needed.

EVLA antennas were included in two types of routine VLA observations:

1. A science project using only one band (X) to avoid phase jumps from 
   band changes; a 10hr track using one calibrator; and a near-identical
   track a few days later [Apr 21, 24]

2. Two 'baseline' runs where VLA antennas get their positions refined 
   based on all-sky observations of many sources, also at X band
   [Apr 26, May01]

As usual, the fiber variation is mostly from diurnal temperature
change, about 0.007 turns/hr of round-trip delay at 512MHz = 7 ps/hr
of one-way delay = 20deg/hr of antenna phase at 8.5 GHz. This linear
trend is dwarfed in experiment 1 by a parabolic phase change with
time, due to offset antenna positions, of ~1000 degrees at EA13 over
10 hours. EA14 and 16 were a bit better, 18 (nearest the array center)
was almost flat.

I found rough position offsets for 13, 14 and 16 that flatten the
parabola, leaving a residual slope of 30-40 deg/hr on all these
antennas. This is a bit larger than the fiber rate of change
calculated above. Have I mislaid a factor of 2? Anyway, I vote for the
sign of fiber delay that minimizes this slope:
 
Antenna_Phase_corrected (degrees) =  
Antenna_phase_raw  -  180*F_MHz/512 * L352_round_trip (turns)

(I am uncertain of exactly what F to use; I used 8.45GHz and I expect
10% accuracy at X band.)

Applying the same antenna position offsets to the baselines data in
experiment 2, the phases are unwrapped and a slope of ~40deg/hr
remains, again a factor of 2 larger than I get from the L352.

(As an aside, Ken has put in a change of coordinates that has improved
things. More baseline tests soon.)

In the baseline runs, the antennas change elevation rapidly. The
change of L352 delay with elevation is easily seen, but is small and
well behaved, (~3 deg of phase at X band; < 20 deg @ Q) for elevation
change from 15-115 degrees. The spec is 0.7ps (13deg @ Q) for any 2
pointings, AFTER correction with the L352 meters). This was at night, no
temperature variations (e.g.  light/shadow) with elevation are
expected. I should try this in daytime perhaps.  There is no hint of
fiber delay glitch as the antenna(16) goes over the top, an
improvement since I last complained about it in October 2005, though
it was no showstopper then.

No delay change with azimuth is seen in the L352 data.

So, the fibers are very good. 

Not so the residual phases on EVLA antennas even after the best-fit
position is in. (Assuming it is legal to fit these positions in the
VLA frame - I thought about it, my head hurt, so I just did it.) 
There are 30-40deg phase errors (X-band), which increase in magnitude
at low elevation.  The worst VLA antennas look about the same as the
EVLA, but the best are under 10 degrees.  I cannot account for this
much from the fiber. Even if I allow that factor 2 above, the errors
do not correlate well with the fiber.

I see about the same size errors on EVLA antennas near and far, and
the errors appear somewhat systematic with elevation, but do not
correlate from one antenna to another (as I naively expected if the
atmosphere model was different on the EVLA).

I conclude that the fiber variations (including L352 measurement
system) are not limiting the phase stability spec right now.

Vivek.