[evlatests] [Fwd: Some WIDAR testing results]
Rick Perley
rperley at nrao.edu
Wed Oct 22 13:34:36 EDT 2008
apparently didn't go out -- I try again...
-------- Original Message --------
Subject: Some WIDAR testing results
Date: Wed, 22 Oct 2008 08:45:20 -0600
From: Rick Perley <rperley at aoc.nrao.edu>
To: evlatests at aoc.nrao.edu, Brent Carlson <brent.carlson at nrc-cnrc.gc.ca>
Michael took some WIDAR test data yesterday afternoon with the
4-station prototype configuration.
The primary goal was to see if phase continuity is retained between
a pair of close-spaced sources. We observed two perfect point-sources,
1310+323 and 1407+284, separated by about 10 degrees, at the standard
X-band frequency. Each source provides about 3 Jy, sufficient to
enable straight-forward calibration of phase and amplitude.
Observations extended over 25 minutes, cycling between the two
sources, with 1 minute on 1407+284 and 2 minutes on 1310+323. The
weather was perfect, with clear skies and light winds.
Six sub-bands at a single polarization were provided (Michael will
have to explain why there were not eight). Averaging time was 1
second. There are 1024 channels in each sub-band, providing 125 kHz
spectral resolution.
The first ~35 seconds of each scan had to be flagged as there was no
coherent signal during this interval -- this presumably represents the
actual setup and move time (the provided data are not flagged in any way).
The identification of the antennas in the AN table is incorrect:
Antenna 1 is actually 17 (on W48), antenna 2 is actually 18 (on N16),
antenna 3 is actually 23 (on E72), and antenna 4 is actually 26 (on N48).
FRING quickly established there were large and steady delays amongst
the antennas. Referenced to antenna 1, we have:
Antenna 2: -142 nsec
Antenna 3: +14 nsec
Antenna 4: +167 nsec.
I elected to correct for this via CLCOR, by utilizing the 'SBDL'
function. This 'unwraps' the frequency phase wind for each antenna,
using the values listed above for all sub-bands and times.
Antenna 3 (actually 23) gave perfectly zero amplitude (i.e. neither
signal nor noise) for sub-bands 1, 2, and 3.
BPASS was used to determine the mean bandpass function. I applied
the known time delay, and determined a single solution, averaging over
all observations. I have not yet determined the time-variability (by
differencing solutions determined for each observation separately) --
this will be done shortly.
CALIB was then run to determine the antenna-based amplitude and
phases for each observation. There was difficulty in determining
solutions for sub-bands 1, 2, and 3 -- very likely a result of there
being only three functioning antennas. No difficulties were encountered
for the remaining sub-bands.
Examination of the solutions showed that there is good phase
stability *within any single observation* -- with a single exception on
the first scan, which I think is due to some initialization condition,
as the scan is one minute longer than any other.
However, there is little phase continuity between scans. Changing
source nearly always results in a new phase, at least 90 degrees
different. Upon return to the same source, the phase is sometimes
similar, and sometimes not. For example, antenna '4' (at N48), for
source '1310' has plausibly connected phase for source 1310 for the
first 10 minutes, but upon return to this source for the succeeding four
observations, a completely different phase is found. The phase change
for each antenna is different. It is possible that these changes are
atmospheric -- but highly unlikely in my view.
I proceeded to calibration, in order to check closure and view the
noise statistics. For this, a 10-second calibration interval was chosen.
I used UVHGM to view the noise statistics. It is immediately
apparent that antenna '1' (aka 17) is noisier than the others -- the
operator confirmed that its sensitivity is notably poorer as determined
by 'stress test' results.
The gaussianity of the real and imaginary parts of the noise is
simply stunning! These distributions should be in a textbook! The
phase distributions are notably non-gaussian -- as they should be as the
SNR in single channels with 1 second averaging is no more than 2.
Examination of the calibrated phases showed normal distributions
about zero degrees and unit flux -- good, but not overwhelming evidence
that closure is being well maintained. (Only 'good', because the SNR is
not as high as we need to do an exact test).
Indirect, but strong, evidence that closure is good comes from
making an image using data calibrated from the solutions determined from
CALIB. A clean point is found, and the rms noise in the background is
only 0.05% of the peak. I don't think more can be expected from so
little data.
More information about the evlatests
mailing list