[evlatests] Notes from EVLA Test Meeting of March 25

[Barry Clark]

EVLA Test Meeting, March 25, 2010

1.  RFI

R. Perley reports on an oldish cross correlation spectra taken at
L band.  Strongest signals:  1030, 1090 MHz aircraft transponders.
Aircraft DMEs, from 1090 up to 1130 MHz.  L2 (military) GPS at
about 1372 MHz.  1244-1372 is fairly clear, with four moderately
strong, but narrow band, radars.  The 1372-1500 MHz subband has
an internally generated birdie at 1408 MHz (=11*128).  The
1500-1628 MHz subband is OK up to 1526, where the satellites
(INMARSAT, GPS, GLONAS, etc) take over, with signals to 25000 Jy.
The 1612 OH band, 1607-1617 MHz is reasonably clear.  The
1628-1756 MHz subband contains weather balloons (1670-1680 MHz,
but only a couple of hours twice a day, at 12h and 0h UT), and
meteorological satellites (1680-1710 MHz).  The 1756-1884 subband
was fairly clear except for three signals of about 30 Jy at about
1800 MHz.  In the final, 1884-2012 MHz band, the upper part of
the band is badly interfered with by PCS cell phones, 1930-1990 MHz.

He also reports on a total power scan of S band.  There is an
allocation for balloon transmissions at 2105 MHz.  Digital Satellite
radio is very strong signals in 2319-2347 MHz.  There are signals
of unknown origin at 2188 and 2199 MHz.  In fact, the whole lower
part of the band is pretty bad up to about 2490 MHz.  There are
strong interfering signals in the upper part of the band, 3600-
4200 MHz, which he mainly saw at low elevations (unclear why).

M. Rupen reported on problems at the bottom of C band (4000-4500 MHz),
which he attributes to interference.  He was observing with 2 MHz
channels, and was seeing strong time variability and large closure
errors, but did not see a spiky spectrum, probably due to the large
channel width and multiplicity of signals.

V. Dhawan reported that very similar behavior was seen at the upper
end of S band.  He commented that averaging over either frequency or
time seemed to make things better, and that perhaps observing at this
frequency should not yet be ruled out.

2.  Single dish tests on antenna ea24.

R. Perley reports on the prototype Ku band receiver.  At this band
it appears to be relatively easy to sort out the various contributions
to system noise.  There is extra spillover (past the edge of the dish)
of perhaps 1K near the zenith, down to perhaps as low as 50d elevation.
Then at low elevations we start to see spillover past the edge of the
subreflector, of about 2K, starting at elevations of perhaps 15d.
Receiver temperatures are remarkably low, from 6k to 10k, depending
on frequency.

At S band the situation is somewhat less clear.  It was necessary to
assume an atmospheric zenith opacity to be able to sort out spillover
(1.25% was used).  This band also shows an extra 1K at the zenith.
The low elevation is more interesting.  Spillover starts to rise at
about 20d, rising to about 10K at 12d, then stays at that level the
rest of the way to 8d elevation.

Comparing low elevation spillover at the various bands is interesting.
As noted, S band rises several K, as does K band.  Ka band does not
have extra spillover at low elevation.  Q band spillover appears to
be a function of frequency across the band.

3.  Blank field observation.

F. Owen looked at a known, fairly blank field at L band.  In the low
frequency subband, he had to discard the lower part of the subband
because of lingering interference problems, retaining only 78 MHz
of the 128 MHz correlated.  In the high frequency subband, he was
able to use most of it.  With these selections, he was able to get
very close to what the Exposure Calculator expected, an RMS of about
100 uJy.  He tried the reweighting task recently installed in AIPS,
and found it made little difference to the map RMS.

4.  Zero data.

There have been panicky reports lately of very large numbers of
true zeros found in the data.  M. Rupen investigated one case.
This was data taken for AK730, which, when loaded into AIPS was
reporting 24% zeros in the data.  Turns out this was loaded with
flags applied in CASA (now a standard option).  For reasons unknown,
when data are flagged in CASA and exported, the data are sent as
flagged zeros.  (A similar effect is seen within CASA if averaging
is used - if all data going into an average are flagged, the average
is a flagged zero.)  If one looks at unflagged data only, the zeros
drop from 24% to 5.6%.  Most of the remainder arises from
autocorrelations.  At the moment, the correlator is only supplying
half the autocorrelations.  (Eventually, it may be made to supply all
by time sharing.)  The others are in the exported dataset as zeros.
Also, only one of the two crossed hands is calculated (the two are
redundant for stationary signals), and at the moment, a recently
introduced bug is dropping that one.  The percentage of zeros in
unflagged crosscorrelations was actually about 0.4%.  This is a
well known problem in the correlator network somewhere, which has
not yet been deemed important enough to be investigated and fixed.

The above data were taken in OSRO 1 mode.  The situation is a bit
different in OSRO 2.  A hardware problem is causing dropped
correlations involving antennas ea17 and ea18.  This can probably
be eliminated by choosing a slightly different setup.

5.  Band switching problems.

V. Dhawan reports that a data set switching LCXK was showing
individual antennas with problems returning to where they left
a given band.  X band is by far the worst.  The magnitude of
the problem is a few percent, a few degrees of phase.