[evlatests] Test Meeting April 23

[Barry Clark]

EVLA Test Meeting April 23, 2009

1.  S band
There are now two antennas equipped with S band receivers.  E Momjian
reports on tests with these antennas.  With only a single baseline,
it is not possible to attribute features to only one antenna.  With
that caveat, he showed sensitivity plots across the band.  There is
an apparent sensitivity change across the band - 3 to 4 GHz is more
sensitive than 2 to 4 GHz.  It was speculated, but not convincingly,
that this might be due to the influence of the strong interfering signal
from digital satellite radio (DSR) at 2.4 GHz.  As well as the strong
satellite radio signal, there was another pretty steady signal at
about 2200 MHz, and a collection of signals that came and went.  In
his first test at 22d elevation, EM saw the best Tsys/Eff = 60 K in
the upper band.  At a better elevation, about 45d, the best Tsys/Eff
was around 45 K.

K. Sowinski used Widar to look at the subband containing the DSR
signal.  His first look showed a lot of ringing around the signal.
He then adjusted the power out of the subband filter back to nominal
(an adjustment of two to three dB), and the ringing was not present
on his next look.  M. Rupen commented that similar ringing on masers
is well suppressed by Hanning smoothing.

2.  Tsys measurements.
R. Perley compared the results from three power measurement systems.
They were 1). A good, independent very linear detector (called,
historically, Milhaus), 2. The detector at the output of a T304
downconverter module, and 3. The FFT/detection system implemented
on the digital bit stream in the D351 deformatter modules.  This was
done at C band.  In the absence of a cold load large enough to cover
the entire feed, measurements were made of the cold sky (assumed to
be about 12 degrees), cold sky with the noise diode on, a hot load,
and Cygnus A.  The cold sky to Cygnus A measurements showed a very
linear relationship between the three systems, albeit with small offsets
(about 0.4 K equivalent for the T304 detector, about 1K equivalent
for the digitizer system).  The hot load measurements deviated
from that linear slope.  Compared to the expensive, high linearity
detector, the T304 detector read a few percent high, the digitizer
system read about ten percent low.  Not clear why the digitizer was
low - possibility of the amplifier in the digitizer module saturating,
or the digitizer overflowing.

3.  C band polarization.
B. Sault looked at polarization effects at C band, at multiple
frequencies over several hours.  In some antennas, the leakage
terms seem to be frequency independent (antennas 3 and 28 appear
to be particularly good), others show a spread of a couple of
percent as a function of frequency.  The time dependence of the
leakage terms was pretty small, but varied from frequency to
frequency.  But even the bad frequencies seem to wander around
by about +/- 0.2% only; the good frequencies do not change above
the measurement errors.  He did see a curious effect:  when looked
at on a 1 second timescale the right-left phase difference exhibits
variations much larger than noise, and much larger than expected.
He calibrated on a one minute timescale, so he was insensitive to
variations on longer timescales.  M. Rupen said he has seen short
term phase variations on Widar that could be the same phenomenon.

4.  Bounds of L band.
B. Heyward asked about what filters we should use to bound L band.
This is the filter immediately outside the dewar, and is used to
shape the bandpass to minimize unused signals and thus preserve
a maximum dynamic range in the following amplifiers.  The proper
procedure is probably to get a final configuration receiver on an
antenna (planned for the end of May), and to remove its filter
entirely and see where its sensitivity crosses that of S band
on the high side, and where excess interference is introduced on
the low side (the OMJ goes down to about 940 MHz, so frequencies
below that are not useful).  This is a complication, because if
we wait for this measurement, the ordering cycle for filters
may delay getting production version receivers installed.  It was
concluded that we probably have a small stock of 1-2 GHz filters
purchased for another purpose that could be used in the interim.

5.  Phase variation with elevation.
V. Dhawan presented the very confusing situation.  The situation
is complex for astronomical measurements because the phase versus
elevation curves of the instrument are close enough to being
cos(elevation) that at any one frequency, they can be removed
by changing the axis intersection defect (the so called K term)
in the phase equation.  In turn the K term is not well separated
from the other geometric terms describing the telescope location,
except by a very lengthy series of observations.  So a short
baseline run may yield a good fit at one band, but a totally
inadequate fit at another.  The various hardware tests have not
been very decisive in indicating the etiology of the problem.
The one problem decisively found, and fixed, was that an integrated
crystal oscillator module with both 512 MHz and 128 MHz outputs,
both of which were used in the system, did not have a stable phase
relationship between the two.  The hardware is extremely difficult
to diagnose in the lab.  The temperature control in the lab is
much worse than in the antenna vertex room racks, and attempts to
tilt racks in the lab were much bothered by phase changes due to
flexure of the cables driving the racks.