[evlatests] Sensitivity Before and After
Rick Perley
rperley at aoc.nrao.edu
Thu May 4 22:37:53 EDT 2006
Walter determined, via some algorithm he will have to describe, what
he hopes are optimum
settings for the input and output attenuation for the L304 module. The
overall goal is to
decrease the power into the T304 so as to maximize headroom, yet not so
low as to
compromise sensitivity by permitting the T304's internal workings to add
noise more than
~1%, then to compensate for this input attenuation by adjusting the
output attenuation so
the output 'cold sky noise' occupies the lowest ~2 to 3 bits of the
quantizer. There may be
other criteria.
We made a system sensitivity check both before and after Walter's
new settings.
The method was to observe in spectral line mode (Mode 4) with 6.25
MHz BW, giving
32 channels in each of the four parallel-hand correlations. Calibration
was done on 3C84,
and the noise distribution measured on cold sky. For this report, I am
using the 'aips weights'
as in indicator of improvement, or not, due to Walter's settings.
Referenced pointing was employed at X, K and Q bands. Solutions
were obtained on
13, 14, and 16 at X and K bands. Only antenna 16 gave solutions at
Q-band. Antenna 18
gave good solutions at X-band for the 'after' observation. I have the
offsets for Ken to
peruse.
Sensitivity Results:
A) L-Band
The antennas got on source too late for the 'before' observation, so
no calibration was
possible, and hence no deltas can be determined. For the 'after'
observation, the aips weights
show the same story as oft-reported: IFs A and C show poor sensitivity,
IFs B and D
are so-so (near the median). This latter is better than usual, but the
result needs to be confirmed
with real sky noise.
Antenna 13 is especially poor on the AC side.
Antenna 16 gave no fringes on the AC side.
B) C-Band.
Antennas 13 and 14 did not get on source in time for calibration in
the 'before'.
Antenna 16 was on source for both, and there was no change in
sensitivity.
The aips weights showed as usual that antennas 13, 14 and 16 are far
better than any of their VLA counterparts. This result was checked and
confirmed by plotting the actual blank sky noise: For 13, 14, and 16 in
IF 'B', the real and imaginary parts each showed a dispersion of 0.32
Jy. For a 'typical' set of VLA antennas (2,3,5,7 and 9), the value is
0.42. For the very best VLA antennas (11, 20 22 and 27) it is 0.35.
This is a good result, but not quite as good as I am expecting via our
single-antenna tests.
C) X-Band.
Antennas 13, 14 and 16 worked 'before', 13, 14, 16 and 18 (except IF
D) 'after'.
13: Three of four IFs got *worse* in sensitivity. The other
remained about the same.
14: The same as above. Degradation in 3 of 4 IFs.
16: Same story here. 3 of 4 went down.
18: The 3 IFs gave good values, but there are some very strange
effects seen in the
'on 3C84' observations which need to be followed up. Not today.
D) K-Band.
It is impossible to make any judgement, as the 'aips weights' are
changing by large
amounts over timescales of minutes, expecially 'before'. Clouds are the
likely culprit.
The median aips weight value, however, is significantly below the VLA
median.
E) Q-Band.
It is impossible to make any judgement based on weights. The clouds
(or whatever)
are causing huge temporal variations.
The 'before' data gave decent amplitude fringes on 13, 14 and 16
(except 13 B and D,
which were dead). The 'after' data gave no believeable fringes on any
EVLA antenna.
There is no clear reason, unless the attenuator settings through things
really far off.
More information about the evlatests
mailing list