[evlatests] Calibrating TCals
Rick Perley
rperley at nrao.edu
Mon Mar 6 19:13:07 EST 2017
*Upfront Warning* This is a long report. To save time, you might
want to come to the monthly 'Test-Engineering' meeting this Thursday
(March 9), where I'll review all the material, with nice plots. ('A
picture is worth a thousand words... ')
The February 14 test data provide a useful way of calibrating (or,
more correctly, adjusting) the Tcals.
The observations were of four sources, two rising and two setting,
plus 3C286 (calibrator). The primary purpose of the test was to
remeasure the antenna gain change as a function of elevation, for two
frequencies in each of the eight Cassegrain bands.
But the data can also be used to measure the antenna system
temperatures as a function of elevation, and to make comparisons between
antennas. Recall that our estimates of Tsys are given by:
Tsys = PSum/(2*PDif)*Tcal
The data were taken in D configuration, in ideal dry and calm
weather. We can then safely presume each antenna 'sees' the same
atmosphere and ground emission. Furthermore, to good approximation, the
atmospheric and spillover emission is independent of the antenna
efficiency. (For example, if the focus is wrong, the forward gain
declines, but the beamwidth widens such that the power seen through the
main beam remains nearly constant).
The difference in system temperature for an antenna between two
different elevations depends on the ground spillover and atmospheric
emission contributions (as the receiver contribution subtracts out),
times the error in the Tcal value. If all the antennas see the same
atmosphere and ground, then the differences between antennas is then a
measure of the errors in the utilized Tcal value.
Comparison of the apparent differences from some suitable mean then
provides correction values which can be used to adjust the Tcals.
To try these ideas, Eric write a special AIPS program ('ELFIT')
which determines the coefficients of a polynomial fit to the elevation
system temperature data. The program returns (amongst other things) the
system temperatures at 70 degrees and 10 degrees elevation, and the
difference between these. This was run for all antennas and both
polarizations for the two frequencies in each of the eight bands that
were in the test.
A spreadsheet was then utilized to determine the median Tsys
difference over all antennas, and the ratio for each antenna to that
median for both polarizations and both frequencies at each band. This
ratio is then the factor by which the Tcal for that
antenna/polarization/frequency is `wrong'. More correctly, the ratio is
that needed to make all antenna/polarization estimates of the system
temperature the same as the median value. There's no guarantee the
median is correct! (But, see below for one way to check).
Below is a small table with the basic results. Columns are:
Frequency in GHz
Med.Delta.Tsys = median of difference in Tsys between El = 70 and El = 10.
Avg.Err = average of absolute differences between the median Tsys
Difference and the individual antennas Tsys Difference, over all
antennas, for the IFs as listed. These are usually dominated by a few
antennas with large differences.
------------------------------------------------------------------------------------
Band Frequency Med.Delta.Tsys Avg.Err.
Q 48.0 109 9.8% in A, 17.2% in C
Q 42.0 58 5.4% in B, 6.3% in D
Ka 36.5 40 9.8% in A, 9.5% in C
Ka 29.5 31 9.7% in B, 23.8% in
D (due to ea27)
K 25.5 40 5.8% in A, 6.4% in C
K 19.5 32 6.3% in B, 6.5% in D
Ku 16.8 20 16.8% in B, 16.8% in
D (due to ea26)
Ku 13.2 18 16.1% in A, 17.8% in
D (due to ea26)
X 11.3 12 12.6% in B, 11.8% in D
X 8.3 11 13.8% in A, 9.6% in C
C 6.69 15 9.5% in B, 10.7% in D
C 4.44 18 10.1% in A, 9.7% in C
S 3.524 21 9.2% in B, 8.5% in D
S 2.625 25 9.6% in A, 9.3% in C
L 1.865 13 9.0% in B, 7.7% in D
L 1.465 13 7.2% in A, 8.1% in C
-------------------------------------------------------------------------------------------------
As noted, the average absolute errors are dominated by a few bad
antennas. Given below are the 'worst of the worst' -- those whose
Tcals are off by more than 20%. The factors given are the factor by
which the Tcal is wrong. (i.e., 1.10 means the Tcal is 10% too high).
Band IF Bad Guys
----------------------------------------------------
Q B none
Q D none
Q A ea01 (1.35), ea19 (0.58), ea24 (0.78), ea26 (0.76)
Q C ea02 (1.69), ea10 (1.29), ea15 (1.60), ea18 (1.30)
Ka B ea05 (1.25), ea08 (1.24), ea13 (0.76)
Ka D ea01 (0.80), ea08 (1.35), ea27 (3.83)
Ka A ea01 (0.71), ea02 (1.23), ea13 (0.72), ea25 (1.26)
Ka C ea01 (1.58), ea28 (1.22)
K B ea05 (1.24), ea23 (0.65)
K D ea23 (0.76)
K A none
K C ea20 (0.78)
Ku A ea16 (1.23), ea26 (2.89)
Ku C ea16 (1.32), ea26 (2.51)
Ku B ea05 (0.74), ea14 (0.74), ea19 (0.68), ea26 (2.30), ea27
(1.28)
Ku D ea11 (1.44), ea12 (1.33), ea19 (0.68), ea26 (2.36)
X A ea07 (1.37), ea10 (1.34), ea14 (1.25), ea16 (1.43)_, ea20
(1.77), ea23 (1.31)
X C ea10 (1.25), ea15 (1.31)
X B ea07 (1.25), ea14 (1.39), ea16 (1.38)
X D ea07 (1.27), ea15 (1.29), ea16 (1.25), ea17 (1.21), ea21
(0.77), ea24 (0.71)
C A ea08 (0.48), ea17 (1.23), ea28 (1.30)
C C ea08 (0.50), ea19 (0.77), ea28 (1.30)
C B ea08 (0.58), ea17 (1.22), ea24 (1.29)
C D ea08 (0.61), ea11 (0.77)
S A ea13 (1.32), ea15 (0.68), ea21 (0.77)
S C ea15 (0.67), ea16 (0.70), ea20 (1.26)
S B ea15 (0.69), ea20 (1.24)
S D ea10 (0.75), ea15 (0.72)
L A ea27 (0.76)
L C ea19 (1.22), ea27 (0.75)
L B ea19 (1.22), ea27 (0.75)
L B ea14 (1.25)
L D ea04 (0.79), ea19 (1.25)
-----------------------------------------------------------------------------------------------------------------------------
Finally -- If were were to employ all the factors, and modify the
Tcals by these values, we would expect these elevation Tsys curves to
become nearly identical. But are they right? We're only modifying
numbers to fit the median. Do we have absolutely -determined values to
compare these to? The answer is (mostly) 'YES'.
Bob Hayward and I spent many many days on ea 24 (and other
antennas), from 2005 through 2011, doing tip curves at all bands, which
were calibrated by thermodynamic standards (hot and cold loads). These
data are a much more secure determination of the true system
temperatures (and, more importantly, the difference between two fiducial
elevations). Below is a comparison of the data reported above, and the
data taken by Bob and me.
Band Freq. These Data P&H
-------------------------------------------------------------
Q 42 GHz 58 K 58 K
Q 48 GHz 109 102
Ka 29.5 31 30
Ka 36.5 40 33
K 19.5 32 28
K 25.5 40 34
Ku 13.2 18 17
Ku 16.8 20 20
X 8.3 11 11
X 11.1 11 11
C 4.40 18 17
C 6.69 14 13
S 2.624 25 23
S 3.524 21 20
L 1.465 13 13
-----------------------------------------------------------
Agreement is generally very good -- except at Ka (at the higher
frequency in A/C IFs) and K band (all four IFs), where the recent data
appear to give system temperatures significantly too high.
More information about the evlatests
mailing list