[evlatests] L-Band Sensitivity tests
Rick Perley
rperley at nrao.edu
Sun Jun 25 14:13:32 EDT 2006
After consultations with Ken and Bob, an experiment was fashioned to
try shed light on the L-band sensitivity differences repeatedly noted
between IF pairs 1 and 2. In short, we have always noted that:
-- IF Pair 1 (IFs A and C) always have significantly poorer
sensitivity than IF Pair 2 (IFs B and D).
Given that we have always used the standard default frequencies, with A
and C set to 1465 MHz, and
B and D set to 1385 MHz, the difference could be due to a frequency
dependent antenna feed or Tsys dependency, or due to some IF channel
dependency.
To separate these effects, we observed a standard calibrator (3C286)
with four different combinations:
1) IFPair 1 = 1485, IFPair 2 = 1385
2) IFPair 1 = 1385, IFPair 2 = 1485
3) IFPair 1 = 1385, IFPair 2 = 1385
4) IFPair 1 = 1485, IFPair 2 = 1485.
Note that with this setup, we observe each channel twice at each
frequency.
We chose 1485 rather than 1465, in order to avoid the VLA
Polarizer's 'suckouts' located near
1465 MHz.
Observations were made in correlator mode '4', at 6.125 MHz BW,
giving 32 spectral channels
in all four parallel-hand correlations. Averaging was set at 5 seconds
-- as fast as the correlator can go.
Calibration used the convenient on-line images of 3C286 (nifty!),
which removes any resolution-dependent gain variations. Sensitivity
was judged via AIPS weights, which have proven to be a reliable
(although not guaranteed!) metric for sensitivity. AIPS weights are
nominally 1/sigma^2, and hence proportional to (efficiency/Tsys)^2. A
high value is a good thing.
Results:
First, some observations of stability:
Antennas 13 and 18 were regularly unstable -- in a manner exactly
consistent with the short-term regular dropouts previous described.
Since these dropouts occur every ten seconds (on the 'ones'), every
other 5-second integration for antennas 13 and 18 (IFs B and D only on
18) were low. These low values were flagged out so as not to bias the
sensitivity calculations. We have good reason to believe the unflagged
values represent good data.
All phases were stable over the duration of the scan (about 7
minutes each).
The ~30 second change in gain, at the beginning of the scan, noted
on Friday for antennas 16 (all IFs)
and 14 (IF D only) were not seen in this experiment -- surely because we
did not change band.
Antenna 18, IFs B and D provided no data at 1485 MHz at either
observation. Antenna 13, IFs B and D, provided no data on one 1485 MHz
observation, but did on the other. The data were truly dead (this
determined after filling the data with all flags off) -- not just weak.
I presume a LO failure.
Now, for the sensitivity results.
In short, the dependency is purely one of frequency. No clear IF
dependency is seen (either than
two clearly weak IFs: 14C and 16D).
These results are more easily seen in the following table, giving
the observed AIPS weights (multiplied
by 100). `X' means no data were obtained (as noted above).
IF 1385 MHz
1485 MHz
-----------------------------------------------------------|--------------------------------------------------------------------
13 14 16 18 |
13 14 16 18
-----------------------------------------------------------|--------------------------------------------------------------------
A | 320 304 310 334 | 150
200 200 254
320 315 310 345 |
172 215 205 279
--------------------------------------------------------------------------------------------------------------------------
B 305 300 295 350 | 185
215 197 X
340 320 306 370 | X
230 205 X
--------------------------------------------------------------------------------------------------------------------------
C 360 204 280 335 | 165
62 185 264
350 212 278 349 |
173 79 190 288
---------------------------------------------------------------------------------------------------------------------------
D 300 285 93 340 | 175
228 23 X
340 305 97 365 |
X 230 21 X
----------------------------------------------------------------------------------------------------------------------------
The small changes in weight between between the two observations at
any given frequency/IF
are easily accounted for by the change in elevation of the source.
From these results, we see that the sensitivity is significantly
better at 1385 MHz than
1485 MHz, for all antennas and IFs. When we swap frequency pairs (i.e.
observe at 1485/1385,
then 1385/1485 for IFpairs 1/2), the sensitivity moves with the frequency.
Some IF dependencies are clear: antenna 16D is clearly in bad
shape, with either a very
high Tsys, or very low efficiency, regardless of frequency. Antenna 14C
is also a poor performer,
but not as severely.
Antenna 18 is the best of the four EVLA antennas, at both frequencies.
For comparison, here are the *best* VLA weights:
IF 1385 1485
----------------------------
A 380 405
B 385 410
C 320 330
D 330 290
--------------------------
From this, we note that in general, the EVLA at 1385 is generally as
good, or better than
any VLA antenna -- particularly antennas 13 and 18. And (as noted
numerous times) at
1485 MHz, the EVLA sensitivity is far poorer than the best VLA (although
the sensitivity of
18 is about equal to a median VLA antenna).
Note that for the VLA, IFs A and B (RCP) are significantly better
than IFs C and D (LCP) -- at both
frequencies. This polarization difference is not seen in the EVLA
antennas.
I will next look at the sensitivity histograms, to see where the
EVLA antennas fall, and confirm
these results with actual noise histograms, to confirm the reality of
these conclusions.
More information about the evlatests
mailing list