[evlatests] L-Band Sensitivity, Tsys and Efficiency
Rick Perley
rperley at nrao.edu
Tue Aug 1 19:51:33 EDT 2006
I used an hour of 'Dynamic Time' yesterday afternoon to run some
more tests on the peculiar frequency dependence of L-band sensitivity.
Antennas 14, 16, and 18 were available (13 was 'under repair', and
24 does not have its LSC converter installed yet).
I observed the standard calibrator 3C147 in the usual spectral line
mode (12.5 MHz bandwidth, 16 channels per correlation, mode '4') at 3
pairs of frequencies: 1275/1375, 1440/1510, and 1670/1706 MHz
(AC/BD). I also observed a piece of blank sky, 5 degrees away from
3C147, to permit calculation of the true system noise.
Data quality from the EVLA antennas was simply outstanding. No
flagging needed in any IF. No drops or spins seen. Only one case of
misbehavior to report : Ant 14, LCP, (IFs C and D) was very weak at the
last frequency pair (1670/1706 MHz). No cause is known. The
subsequent report does not include these weak data.
The data were calibrated with the correct flux density of 3C147 for
each frequency. Bandpasses were calculated and applied. AIPS weights
were then calculated for the central channel. To check that these
weights can be safely used as a proxy for real sensitivity, I generated
noise histograms for both VLA-VLA baselines, and for EVLA-EVLA
baselines. In all cases, these noise histograms (which are beautiful
gaussians) confirm the validity of AIPS weights. (It's almost like
magic -- but I think this actually means the Tsys system is working well
for EVLA antennas).
I again remind readers that AIPS weights are proportional to
(antenna efficiency/Tsys)^2. A high number is good. The actual values
given below are not important -- the ranking, mean, and spread are.
Results:
I used the convenient AIPS program ANBPL to calculate the
antenna-based AIPS weights, applying all calibrations. I then made
'sensitivity histrograms', of all VLA and EVLA antennas. Some very
interesting results came from this:
At 1275 MHz: VLA antennas span a wide range in weights: from 8 to
20, with a median of about 14. EVLA antennas 14 and 18 are at the very
top -- weights of 21 to 23. Antenna 16 is a little worse -- weight =
18, which is still much better than the VLA median.
At 1375 MHz: VLA antennas span a range of 11 to 23, with median of
17 -- the array is notably more sensitive at this frequency. The EVLA
antennas however slip considerably in their ranking: 16 is in the
bottom third, 14 is barely in the upper half, but 18 is still near the
best.
At 1440 MHz, the VLA median weight is about 15 (down a bit from
1375), and the rankings of EVLA antennas continue to slip: 16 is at the
bottom (weight 11), 14 below the median (weight 14), 18 is now the 5th
best antenna.
At 1510 MHz, the VLA median is now at 13, antenna 16 remains at the
bottom, 14 and 18 are both in the middle of the distribution.
At 1670 MHz, the VLA median continues to drop -- now at 11, and all
the EVLA antennas are below the median. However (and this is different
than in past measures) -- the EVLA antennas are not below the bottom of
the VLA distribution.
Finally, at 1706 MHz, the VLA median is down to 10 (in other words,
the array sensitivity is now 75% of its best value near 1375 MHz), and
the EVLA antennas are all in the bottom half, but not worse than the
poorest VLA antennas.
Summarizing the above: The trend noted many times before of the
EVLA antennas have very good performance at low frequencies, with
decreasing sensitivity, (both absolutely, and relative to VLA antennas)
is confirmed. What is different in this test is that at no frequency
were any of the EVLA antennas found to be worse than the worst VLA
antennas. I can't explain why the relative sensitivities of the EVLA
antennas are better than we have seen in the past at these higher
frequencies.
I also observed the absolute flux density calibrator, 3C274 (aka
Virgo A). This source provides enough flux to raise the system
temperature by an amount (typically 20K) far greater than the noise in
these measurements and variability in ground pickup and atmospheric
emission, so an attempt can be made at determining antenna efficiency.
In the following, I analyzed only the RCP channel 'A' (1275, 1440,
and 1670 MHz). There are no reasons to believe the results will be
different in other channels.
The expected antenna temperature (that is, the change in system
temperature when going onto Virgo A) is 17.8, 16.0, and 14.1 kelvin at
1275, 1440, and 1670 MHz) assuming an efficiency of 0.43 (which was
measured by Bob Hayward and me two years ago on antenna 13), and the
published Baars et al. flux (232, 209 and 184 Jy, at 1275, 1440, and
1670 MHz, respectively). Virgo A has a large halo, contributing
significant flux on a scale comparable to the antenna beam -- I have
ignored this effect in what follows. I have also ignored the
contribution of 'cold sky' -- this should be small for Virgo, which
sits far from the galactic plane.
The observed change in Tsys must be corrected for the error in Tcal
-- the VLA system knows only the value of Tcal at the default frequency
of 1465/1385 MHz. The Tcal values have been carefully measured by the
front-end group, and I've made these necessary adjustments for each of
the frequencies, for each antenna. (I trust that the final EVLA system
will do all this for me!)
The resulting rise in Tsys (= Tant) can be interpreted in two
distinct ways:
a) Assume the efficiency is known, and use Tant to calculate Tsys
(cold sky), and Tcal.
b) Assume the Tcals are correct, and use Tant to calculate the true
efficiency.
I give both answers in the following.
At 1275 MHz: We have at this frequency an 'embarassment of
riches'. The system is better than we expect.
a) Assuming the efficiency really is 0.43 (as measured on
antenna 13), we find:
Antenna 14A: Tsys = 24 (rather than the system's claim of 33).
Antenna 16A: Tsys = 25 (rather than 31)
Antenna 18A: Tsys = 25 (rather than 32).
Applying a beam dilution factor will lower these system temperatures.
All of these Tsys values are better than expected.
b) Assuming the Tant values are correct (meaning the Tcals are
correct, and ignoring beam dilution), we find efficiencies of 0.58,
0.53, and 0.53 for 14, 16, and 18, respectively. These are much better
than the measured 0.43 -- but I note that this measurement was made on
antenna 13. Applying a beam dilution correction will (by the way) make
these efficiencies even higher!
I'll note here that all these Tsys values are 'back-end', and for
reasons not clear, they are higher than the corresponding front-end
values (for VLA antennas -- there is no 'front-end' Tsys on EVLA
antennas yet). The effect is not believed to reflect a true rise in
Tsys -- it is likely an error in the VLA's system. The scale of the
effect is I think 5 to 10%, which would decrease the efficiencies
derived above by the same factor. Not enough, however, to lower them to
0.43.
At 1440 MHz:
a) The system temperatures are 24, 26, and 22 K, for an assumed
efficiency of 0.43. The Tsys values given by the system are 24, 33, and
19, for antennas 14, 16, and 18, respectively.
b) The efficiencies are 0.43, 0.54, and 0.37 for the three antennas,
if we assume the Tsys values are correct. These values are disturbingly
different, but they reflect the variation in Tsys.
At 1670 MHz:
a) Tsys values are 24, 30, and 26, for antennas 14, 16, and 18,
rather than the listed values of 24, 30, and 26 K (after correction for
Tcal error).
b) Efficiencies are 0.46, 0.49, and 0.43, assuming the Tsys values
are correct.
The three antennas are in decent agreement, and all are performing
at about the expected level.
So: From these measurements, it seems the real mystery is why the
EVLA antennas are so good at low frequencies (where the efficiency is
expected to be decreasing). The next question is why the EVLA antennas'
performance is not improving at the higher frequencies.
More information about the evlatests
mailing list