[evlatests] X-band test run this morning.
Rick Perley
rperley at nrao.edu
Fri Jun 29 12:40:38 EDT 2007
A two hour slot of dynamic time this morning was used to make a deep
image of a well-resolved, but strong, source at X-band. I filled with
1.67 second averaging (my suggested future standard for continuum
observations like this).
A) Flagging Issues
All 27 antennas were working, and no antennas were flagged out.
(The problems with antenna 16 in yesterday evening's data were not
present in this dataset).
However, as reported before, antenna 13B is fringing very weakly,
and 13D is fringing not at all. These were not flagged by the flagger.
The initial (1.67 second) of unflagged data were generally a few
percent low in amplitude, mostly on EVLA antennas. The 'on-source'
algorithm should be tightened just a bit.
B) Stability Issues
Some antennas have temporal (many minutes) slow gain variations.
These are as likely to be VLA and EVLA antennas. I'll check later to
see if there is a connection with recorded Tsys. The effect is easily
calibrated for.
All phases look great. No winds. It seems to me that the EVLA
needs new baselines (order of cm, not meter!) more than VLA antennas.
Overall data quality is simply superb!!!
C) Imaging
The point of this test was too see how good an image could be made.
The data were taken in continuum, 50 MHz BW, so 'closure corrections'
were needed. I used a pure point-source calibrator (0102+584) to do
this, then made an image of that source to check. I used a single point
as the model. The resulting image shows that the source is not a pure
point -- there is a small detached secondary, which due to the closure
correction model being wrong, has in the image an unphysical negative
pair of holes at the 0.01% level, on opposite sides. Other than this,
the noise structure indicates the closure corrections are working well
for this source. The Dynamic Range (peak to rms) is 43000, which I
think quite fine given the total integration time on this source was
about 10 minutes.
I then moved to imaging the target source, 0102+584. This is listed
as a calibrator, and in fact has a remarkable range of quite complicated
structure at the resolution of X-band in A configuration. It was
quickly found that IF 'D' had a few hundred visibilities that extended
to thousands of Janskys. These appear to be due to some sort of
correlator error, as there is no preferred identification with either
VLA or EVLA antennas, and were highly localized in time. They were easy
to find and remove.
After a couple rounds of standard self-calibration and imaging, a
decent -- but not superb -- image was made. The noise in distant areas
looked fine, but in the vicinity of the source itself were much higher
errors which appear to be due to incorrect closure corrections. The
VLA-alone data provided no better image. The EVLA-alone was quite a bit
worse, but the data quantity is about half that of the VLA, and the
source quite complicated.
A UVSUB, using my best image showed that the errors are not confined
to any particular time or antennas, and are almost certainly due to
constant low-level multiplicative offsets (aka closure corrections)
which are different than those used for the calibrator. I then used the
best (but still flawed) model of the source to make updated closure
corrections. The image which employed these was much better, but still
not thermally noise limited. The dynamic range is about 10,000 -- I'm
sure I could iterate to a much better solution, but I don't think it
worth the effort.
We thus have some evidence that the multiplicative baseline errors
are not perfectly constant -- or that their determination using 10
minutes on a 2.3 Jy object is not sufficiently accurate. I'm thinking
of repeating the observation, using spectral line mode, to see if a
proper bandpass calibration removes these apparent errors.
More information about the evlatests
mailing list