[evlatests] Friday evening tests -- Some Good News!
Rick Perley
rperley at nrao.edu
Sat Jul 8 12:47:17 EDT 2006
As noted by Ken yesterday, we decided to observe a strong calibrator
with known structure -- 3C273 -- for 1.5 hours in 'fast continuum'
mode. The primary goal was to try learn more about our dropout problem,
but also to enable imaging tests, as we had all five EVLA antennas
available and operating at X-band. We observed in 3 scans of 30 minutes
each.
As there seems a good chance the drop problem is now understood, Ken did
not interrupt any of these scans.
A) Stability
All 5 EVLA antennas operated, although as noted below, 24C is not in
good shape (as also noted by Ken).
It should be noted that we had 27 operating antennas!!!
Antenna 13: All four IFs have the 10-second drop, on all three
scans. The drop is by 0.76 in amplitude for all scans
and IFs. Phase behavior was as normal for the band, configuration, and
weather (a dark and stormy evening).
Antenna 14: Amplitude stability is outstanding for IFs A and C.
Better even than the best VLA antennas. A slight
perturbation in gains were seen coincident with the passage of a hot
cloud through the beam (as clearly seen in Tsys).
But the perturbation is small compared to most VLA antennas! On IFs B
and D, a few 'gain spikes' were seen -- very short duration and quite
rare. These are not matched by Tsys variations, so something else has
diddled the correlator
coefficient. Note that the overall stability was maintained, even as
the Tsys dropped from about 85K at the beginning
to about 40K at the end (as the receiver cooled down). This is good
news -- the system is working exactly as it should.
Phase behavior was as expected.
Antenna 16: The gain is wandering by significant amounts, both up
and down, on long (tens of minutes) timescales.
All IFs are doing it identically. Review of Tsys shows similar
variations. My speculation is that the Tsys is being erroneously
measured, and these values are then imposed (post-correlation) onto the
(proper) correlator coefficients.
Phase behavior as expected.
Antenna 18: IFs A and C are outstanding. Just like antenna 14.
IFs B and D have 'double drops', by 0.66 and 0.26.
Nothing is different here than has been reported before.
Phase behavior is as expected -- excellent.
Antenna 24: Very interesting ... IF A has drops, by factor 0.76,
for the first 30-minute scan only. IF C has the same drop behavior, but
the sensitivity of this IF is appallingly bad -- as expected from Ken's
report. However, a sudden change in sensitivity occurred about 60% of
the way through the run (but not associated with scan change) -- the
amplitudes rose, and the sensitivity improved (but no where near to what
is normal).
Phase behavior is fine! There are no significant long-term drifts,
so the X-Y baseline components for the master pad are pretty good. (I
can't measured Z component with a single source).
Once the drops were edited out (very easily done with the AIPS
super-program EDITA), the amplitude stability of the 'dropping antennas'
was outstanding (other than 24C, which clearly has other problems).
For all EVLA antennas, phase stability was excellent, with no
significant longterm drifts, and short-term behavior clearly caused by
weather.
B) Imaging.
As noted above, the source was selected because of its well known
structure and strong point-source, affording us the possibility to look
at imaging and closure issues.
As the observations were taken in 50 MHz continuum, signficant
amplitude loss was expected on EVLA-VLA baselines. This was seen, of
course. Standard calibration was done, resulting in large apparent
'excess amplitude' seen on EVLA-EVLA baselines.
A blind image, using all antennas, was made, which was 'pretty
good', but showed the 'closure' errors as a lumpy
north-south stripe of residuals (these show up this way because the
source declination is only +2 degrees).
A much better image, using the VLA-VLA baselines only, was then used
to determine the baseline-based 'closure'
corrections. A single solution, for the entire 1.5 hours, was made
using BLCAL. The typical VLA antenna corrections are 0.999 to 1.001
in real, +/- .001 in imaginary. The corrections for EVLA - EVLA
baselines are 0.95 in real, +/- .005 in imaginary. VLA - EVLA baseline
corrections looked the same as VLA-VLA baselines -- as expected.
The result of applying these corrections was spectacular! The image
using all 27 antennas showed no discernible
N-S 'closure' pattern, and rms noise of 0.36 mJy -- a dynamic range of
70,000. Not bad for 1.5 hours continuum data.
For comparison, an image using only the VLA antennas was made -- the
noise level was the same, as was the dynamic range.
I will point out here that the new capability of IMAGR to specify
antennas to be included in a map was a very great help to this study.
Thanks Eric!
The excellent image made with all antennas clearly demonstrates that
the EVLA 'closure' error is stable, and can be corrected for with very
high accuracy. Very good news indeed (but not surprising to me).
Users should not be dissuaded from wideband continuum observations --
but they need to be informed about how to do the extra calibration.
Fortunately, BLCAL is a very easy program to run.
Some disappointment may accompany the revelation that the noise
level did not improve with the addition of the
5 EVLA antennas. This is actually no surprise. We thermal noise level
is well beneath the observed noise in the map. These are due to
residual closure errors, likely small time variations which are on all
baselines. I also made to heroic efforts to flag bad data, or execute
any of the tricks we've learned to identify discrepant observations.
C) EVLA antenna sensitivity.
I computed the AIPS weights, to judge antenna sensitivity. The
results are mixed:
Antenna 13 is good on all IFs. Just like a VLA antenna -- as
expected.
Antenna 14 was very bad at the beginning, but steadily improved,
reaching normal (like VLA) status at the end. This behavior was
expected, as the receiver cooled down from about 85K at the begnning to
~40K at the end.
Antenna 16 is just awful, the worst in the array (save 24C -- a
special case). As noted before, Tsys is varying all the time, and is
high -- 100 to 140K. There is evidence that the Tsys variations are not
'real' -- in the sense that the correlation coefficients may show no
variations. If Tsys was really changing, the raw correlation
coefficients would reflect the changing noise. (I say 'may' as I don't
have the raw values, nor have I gone to the trouble (yet) to
cross-correlate the gain solutions with the Tsys values).
Antenna 18 is good on all IFs -- about normal.
Antenna 24 is very poor on IF A, and a total disaster on IF C. The
former problem is likely due to a very high Tsys (about 125K). The
latter cannot be blamed solely on the 100K system temperature...
More information about the evlatests
mailing list