[evlatests] VLA Test Meeting, March 26, 2009

[Barry Clark]

VLA Test Meeting, March 26, 2009

1.  S Band receiver

R. Perley reports.  Most receiver tests have been done on antenna 24,
which has a proper analog detector system installed, but has no S band
feed.  Antenna 28 has the only S band receiver.  Analysis was done
mostly through W. Brisken's spectral display software (running in
the deformatters), taking 1000 MHz at a gulp.  For overlap, he took
three different frequencies.  Center band is nominally 2.5 GHz to
3.5 GHz.  The IF filters have a Strong roll-off at the low end, over
about 200 MHz.  (Since this filter is only in the system for the 8 bit
digitizers, this doesn't particularly matter.)  Standard measurements
(as for other low frequencies) were made:  a hot (ie ambient temperature)
load covering the feed, and cold sky, and a strong source (Cygnus A).
Derived system temperature has a bit of oscillation due to mismatch at
the hot load, but is pretty flat at 28 K across the 1 GHz center band.
The derived Tcal wanders around a little, but is basically pretty flat.
Estimating efficiency using Cygnus A yielded 44% (low end) to 40% high
end.  RP believes these very low numbers are due to a serious collimation
error, of order 3'.  Applying that correction would yield 50% to 55%.
These numbers also are lower than expected.  Doing proper pointing
requires interferometry, which would require three receivers, or using
the band overlap trick with the C band receivers.

As expected, we see fairly strong spillover effects, from the beam
looking past the edge of the subreflector and seeing the ground.
System temperature starts to rise at about 25d elevation, and more
than doubles by the 8d limit.

In the high part of the band, 3.3->4.3 GHz, there were no particular
surprises.  Receiver dies pretty thoroughly at 4.1 GHz, possibly due
to the OMT.

In the low part of the band, measurement was set up to cover 1.75->2.75
GHz.  Main feature here is the interference.  Digital satellite radio,
Sirius and XM, are at about 2.3 GHz, very large spikes.  Also, a strong
narrow birdie at about 2190 MHz.  Band pretty well cuts off at about
1.85 GHz on the low side.  System seems to behave fairly well in the
presence of Sirius and XM except for one measurement when it is speculated
that the antenna beam was close to or on the satellite, when the system
became wildly non-linear.

Danger of receiver burnout due to pointing at a satellite radar or
broadcast antenna was discussed.  We think Sirius and XM are probably
OK even if we point directly at the satellite.  We know little about
other satellite systems.  We think most interesting high power systems
are in higher bands than S band.

2.  Ka Band

We have eight systems up and are rapidly adding more.  L. Sjouwerman
reports (relayed by E Momjian), that he successfully used the system
for science, looking for the 36.2 GHz methanol maser in the circum-
nuclear disk, finding, and imaging, a new maser, with a six antenna
system.

3.  Timing blip

J. Robnett reported on the problem.  L302 MIBs in the antennas are
connected to the 19.2 Hz (52 ms) system heartbeat pulse, and use
time disseminated by NTP to resolve the actual time of each pulse.
An operating system upgrade on the infrastructure computers at the
site, which act as level 2 NTP servers, to incorrectly estimate their
clock rates.  So the time they were broadcasting had errors as large
as 128 ms, causing the MIBs to select the wrong heartbeat pulse to
use to set the LO phase.  Problem timescale is a few hours, so
everything seemed to be running satisfactorily after the upgrade
on a Wednesday maintenance day.  Problems with phase jumps in the
date were noticed on the Friday.  It still took several days, into
the next week, for the nature of the problem to make its way to James,
who immediately recognized the probable cause.

Longer term solution for reliability is to use a dedicated system
connected to a GPS receiver as the site NTP server.  This is probably
more immune to software "upgrades" than a general purpose system.
A second such system will be purchased for redundancy.  (? Do we
not need a second antenna as well?)

Another example of poor circulation of information about problems
was the antenna 19 story.  It was written up as weak fringes at 20 cm
and no fringes at X band.  In fact, there were no fringes at bands
other that 20cm at all, and once this was recognized, a week later,
the cause was immediately suspected, correctly, to be with the
subreflector rotation.  (The readout synchro shaft had slipped in
its coupling due to trying to rotate when the gears were locked by
ice.)

4.  Solar modifications to receiver systems

The EVLA project promised solar observing at L band and one other
band to be chosen later.  Receiver engineers would prefer that the
other band be one not yet fully designed - X or Ku.  Initial
indication was that solar observers might like Ku.

B. Hayward described plans for implementation at L band, based on a
design by Paul Lilie.  This involves using the cal coupler as, effectively,
a 30 db attenuator, to prevent saturation of the LNA front end.  Switchable
attenuators are still needed to get the strongest flares into a reasonably
linear regime, and to get the noise cal at a reasonable level for all
setups.  Bob is concerned about various calibration problems, the phase
and amplitude calibration of the various switched paths.  There are a
number of calibration strategies which might work, and it is unclear which
is easiest to implement.  It is very clear that the project lacks solar
expertise, and that we need a specialist to work on these problems, as
well as on finding realistic observing strategies for active sun
observations.

5.  Widar progress

M. Rupen reports on the progress in bringing up the Widar-partial that we
have recently installed.  There are currently eight station boards, two
baseline boards, and three crossbar boards.  This is currently configured
to produce eight antenna, single polarization, four subband operation.
Various neonatal problems have been found and addressed.  This include
rack fibering errors (swap of the fibers carrying odd and even samples),
a problem with pickup on network cables, a probably avoidable OS problem
of unknown etiology, and problems with synchronizing timecodes through
the system.  Latest accomplishment is getting the Correlator BackEnd
system to write the data out, so that it can be read into the postprocessing
systems.