[evlatests] Test meeting June 25

[Barry Clark]

EVLA Test meeting June 25, 2009

1.  S band.

E. Momjian reports that now, with three receivers installed, we can
now evaluate sensitivities.  He finds that the sensitivities are
better in the 3-4 GHz high band than in the 2-3 GHz part of the
band, with Tsys/efficiency in the former about 60, and 70 in
the latter.  Similar (though slightly worse) results seen in
the BD IFs, except that IF 24D behaves anomalously, with very
low sensitivity at the high frequencies.  R. Hayward noted that
the general shape of the curve was rather similar to the shape
of the efficiency curve that Srikanth had provided.  R. Perley noted
that Srikanth's curve showed quite high efficiencies, whereas very
preliminary estimates of efficiency on antenna 24 indicated values in
the low 50s%.

2.  C band.

R. Perley presented holography results produced by B. Hesman.  The
motivation of the holography was a similar effect to that above,
that the high, 6-8 GHz part of the band was more sensitive than
the 4-6 GHz part.  30*30 holography was done at three frequencies -
4385 and 4885 in the low band, and 6385 MHz in the high band.  A
fourth frequency, 7385 MHz, was observed, but the holography
failed.  In the low band, the images exhibit a small corregation
in phase which cannot be attributed to surface error; it is probably
due to standing waves to the subreflector.  This corregation is
not seen at 6385 MHz.  But the main difference between low and high
band shows up in the amplitude part of the illumination.  In the
low band, the illumination has a peak at perhaps 3m radius, and
dribbles off to the edge, dropping to about 0.8 of the peak (in
voltage).  The high band shows the same peak, but, after falling
off halfway out, rises to the edge, reaching about 1.15 of the
peak.  Since there is a lot of area at the edge of the dish, this
rise if possibly the cause of the better sensitivity at the high
band (although one would expect the system temperature to rise
also, due to more spillover).

3.  Polarization.

B. Sault looked at stability of the D terms with time, He looked
at a large number of frequencies across the entire frequency range.
He compared two observing sessions 11h or so apart, and also two
sessions about five weeks apart.  With the 11h spacing, any changes
in D terms were near the thermal noise limit, which was about
0.3%.  There were a couple of antennas with outlying results which
he believes are real polarizer problems.  The plot looking at
sessions separated by five weeks also looks remarkably similar,
with a couple of exceptions.  Antenna 13 had behavior problems at
at least a couple of bands, maybe more.  And the results at 6385
MHz were extraordinary.  There were changes of up to 2%.  The
five week interval spanned an array reconfiguration.  There seemed
to be no difference between moved and unmoved antennas - D terms
stick with the antennas, as they should.

4.  Widar debugging.

M. Rupen reports fairly intensive use of the 10 antenna Widar over
the past month or so.  Various small technical problems have been
found and fixed.  There remain a few one-off problems to be solved,
of no systemic significance.  There are a couple of systemic
problems.

On scan boundaries, one group of five antennas jumps in phase by
180d with respect of the other group.  This is slot dependent - the
two groups are those plugged into odd and even slots in the station
racks.  Careful tests have eliminated all possible causes.

When looked at with high time resolution, a fast wobble is seen
in amplitude or phase or both, with rates of a few Hz.  Exact
frequency and amplitude of the wobble clearly vary from time to
time; the systematics of this are not currently known.  This may
or may not be the same phenomenon which R. Perley reports in data
averaged to one second.  There he sees an effect only when the
wobble frequency passes through zero.  However, the very fast
wobble seems to have different systematics for subband zero than
for the other three subbands.

5.  Widar commisioning.

R. Perley has been working with a couple of large datasets in AIPS.
The bottom line is that X band makes good, high dynamic range,
maps, and that C band does not.  Description of processing:
bandpass done every one minute (one minute scans), gain calib
every second (one second integrations).  Then center 920 channels
added together, and fed to mapper.  The X band maps are noise
limited with a peak/rms about 160,000.  C band maps rms is about
ten times worse.  The background rumble in the maps is fairly low
spatial frequency, so problem arises, somehow, on short baselines.
Just to confuse things, one C band map (after a closure calibration),
is good.

Looked at bandpass stability.  There was a clear change of bandpass
when the T304 attenuators changed state - in well defined cases
the difference in amplitude is approximately a ripple with a 400 MHz
period and a peak-to-peak amplitude of about 2%.  Phase behavior
is stranger.