[evlatests] Results from the OTF/ACU tests
Rick Perley
rperley at nrao.edu
Fri Mar 20 18:15:32 EDT 2015
On Tuesday and Thursday this week two tests of the OTF mode
performance were run. These were done to judge the performance of the
OTF modes, and also to compare the new ACU antennas (14, 21) against the
old.
Executive Summary (for the impatient):
The OTF modes work very well indeed! It is clear that the limiting
performance is the antenna's residual pointing offsets -- typically 5
arcseconds, presuming referenced pointing has been done. Time offsets
(in the sense whether a particular antenna is behind or ahead of the
others in its motion) are typically 20 to 40 ms.
The pointing limitation should only be important at high
frequencies and fast scan rates, where beam crossing times can be less
than 1 second of time.
There is no evidence that ea14 or ea21 (the new ACU antennas)
perform differently than the others from this test. This is consistent
with the limitations being due to pointing residuals.
There is evidence that some antennas are much poorer pointers than
others. This means that even after referenced pointing, there appears
to be time-variable offsets remaining. Diagnosing this problem requires
much more attention be given to the referenced pointing methodology.
The Gory Details (only for the strong):
These tests were set up thusly:
1) Scans of 2 degrees, 6 degrees, and 10 degrees length were
arranged to terminate just beyond the strong source J2136+0041.
2) Scans in both RA and Dec were done.
3) The observations were done at meridian transit, so that the
motions were (almost) purely in azimuth and elevation.
4) Two scans, in opposite directions, done consecutively, were
made for each length. This is to permit separation of pointing errors
from timing errors.
5) Three different speeds were used for each length: 1, 2, and 3
arcmin/second (corresponding to 4, 8 and 12 times 'sidereal').
6) The observations were done at X-band. Note that the beam
crossing times (first null to first null) are approximately 12, 8, and 4
seconds of time for the three speeds. Two SPWs were used, at 8.2 and
11.0 GHz.
7) The delay/phase stepping time was 1.5 seconds.
8) Referenced pointing was done prior to the 10-degree scans, then
again before the 2 and 6 degree scans, which were executed
sequentially. The RA and Dec scans were done on different days.
9) The visibility dump rate was 10 Hz (0.1 seconds integration).
Thus, a total of 36 scans were made.
For each of these scans, the data were first calibrated in the
standard way, using on-source observations of the calibrator. Next,
FRING was run on every timestep when the source traversed the primary
beam. These solutions were then applied, after which CALIB was run to
determine the antenna beam amplitude. (NB Such a fine FRING is needed
to prevent delay errors from degrading the amplitude solutions).
Finally, I applied the special purpose program SNFIT to the CALIB
solutions. This program solves for the beam crossing time, and the beam
width using a quartic model to the logarithmic gains. Comparison of the
solution times between the two IF pairs shows the accuracy of the
crossing time is certainly better than 50 ms.
I attach a typical solution for an elevation cut, at a rate of 1
arcmin/sec. Shown are the four IFs -- the narrower pair are from the 11
GHz SPW, the wider pair from the 8.2 GHz SPWs. The solution for the
crossing time and width used only the data from 0 to -4 dB. The variance
in the post-fit residuals is ~40 milliseconds.
The crossing times for the 36 scans * four SPWs * 27 antennas
(=3888 numbers) were analyzed in a spreadsheet. The opposite
polarizations were averaged together to eliminate the squint, and the
mean for each of the 36 scans determined. From this, the deviation in
crossing time was determined for each antenna and scan.
A simple model was then applied: I assume that for each pair of
scans (backwards and forwards), the same pointing error and 'system
offset' applies. Hence, the deviation in crossing time can be written as:
T+ = Tp + Ts for one direction, and
T- = -Tp + Ts for the opposite direction.
The 'pointing offset' (Tp) in time units, and the 'system offset'
time (Ts) was then determined for each antenna, for each pair of scans.
The 'pointing offset' time was converted to angular offset using the
known rate for that pair.
Finally, the dispersion in these two values was determined in two
ways -- first over all antennas for a given pair of scans, and secondly
over all scans for a given antenna.
Attached are two summaries:
First: (MotionTable.dat)
The variance in Ts (time offset) and Tp (pointing offset, in
arcseconds) amongst all the antennas, for each of the travel distances
(2, 6, 10 degrees) and speeds (1, 2, 3 arcmin/s), for the azimuth and
elevation scans.
Second: (AntTable.dat)
The variance, for each antenna, over all the scans, separately for
elevation and azimuth, for Ts and Tp.
*Some Basic Conclusions*
The dominant component in the variation in beam crossing time is
the pointing. When an antenna is notably late (or early) in crossing a
source, it is nearly always equally early (or late) in the opposite
direction. When this effect is subtracted, the time residuals for all
antennas reduce to typically 30 milliseconds.
The typical offset in pointing, as found from the back-and-forth
method used here, is ~ 5 arcseconds -- which I consider about what is
expected for these daytime tests. There is some evidence that this
'pointing offset' becomes larger with faster scan speeds - this suggests
a failure in the simple model used to analyze the data, which assumes a
'clock' offset which is the same for opposite motions. There is no
indication that the 'clock offset' time grows either with scan length of
scan speed.
There is good evidence that some antennas don't point as reliably
as others. The 'AntTable.dat' file shows this well: In azimuth,
antennas 21, 26, and 8 have far higher scatter in their pointing offsets
than others. In elevation, it is antennas 25, 26, 16 and 5. On the
other hand, note that about half the antennas have really excellent
performance, with offset pointing std. deviation at 3 arcseconds, or
less. (This is found from the 9 transversals for azimuth or elevation,
so I think this conclusion is fairly secure).
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ea14-Soln.png
Type: image/png
Size: 44056 bytes
Desc: not available
URL: <http://listmgr.nrao.edu/pipermail/evlatests/attachments/20150320/895bf733/attachment-0001.png>
-------------- next part --------------
Antenna Elevation Dispersion Azimuth Dispersion
Time (ms) Pnting (arcsec) Time(ms) Pnting(arcsec)
-------------------------------------------------------------------
1 40 4.4 41 3.2
2 18 2.2 21 1.6
3 19 3.0 20 2.8
5 24 7.6 38 8.4
6 22 4.0 64 3.0
7 22 1.9 75 3.6
8 18 2.9 50 8.6
9 19 3.2 15 2.6
10 42 3.8 24 3.1
11 15 3.8 9 1.5
12 14 2.1 50 4.0
13 36 5.3 37 3.5
14 19 4.5 27 1.9
15 50 2.3 50 6.0
16 46 8.4 16 3.4
17 39 7.5 45 6.7
18 17 2.9 25 2.9
19 10 2.0 32 3.7
20 18 4.2 13 2.3
21 18 2.1 48 13.5
22 12 2.1 48 4.2
23 11 4.9 17 4.2
24 11 2.6 13 3.1
25 72 15.4 23 4.4
26 17 8.9 56 11.4
27 22 1.5 37 3.3
28 29 4.3 25 3.7
-----------------------------------------------------------
-------------- next part --------------
Azimuth Motion
2 Degree move 6 Degree move 10 Degree move
Speed time Disp Pnt Disp Time Pnt Time Pnt
-------------------------------------------------------------------------------------------
1 arcmin/s 58 msec 4.8 arcsec 35 msec 5.2 arcsec 65 msec 8.6 arcsec
2 arcmin/s 28 msec 4.9 arcsec 31 msec 5.7 arcsec 43 msec 8.3 arcsec
3 arcmin/s 28 msec 3.6 arcsec 36 msec 7.0 arcsec 40 msec 9.4 arcsec
------------------------------------------------------------------------------------------
Elevation Motion
1 arcmin/s 48 msec 4.2 arcsec 35 msec 5.6 arcsec 48 msec 3.5 arcsec
2 arcmin/s 23 msec 3.5 arcsec 27 msec 5.5 arcsec 29 msec 7.7 arcsec
3 arcmin/s 16 msec 4.3 arcsec 17 msec 7.5 arcsec 18 msec 9.3 arcsec
---------------------------------------------------------------------------------------------
More information about the evlatests
mailing list