Minutes, CA-Forum #4, Oct. 6, 1994
1. Digital Synch. demods. (George Graves)
Board production underway.
The wiring looms need to be scheduled.
George is still expecting completion by the december shutdown.
Antenna work required:
a. George will make the wiring link between the analogue units
and the digital units (the wiring that defines the bandwidth in
use). ... 1/2 day per antenna.
b. Ron Beresford's group will need to provide the control signal.
This defines the ON state of the sync demod. At present this is "blank".
It should be a separate event, so that we can retard the START
(the dynamic hold).
c. Five new D1 datasets will be needed.
2. Digital Interface. George will start on these in december
after completeing his SETI contribution.
3. "Dynamic Hold" Sampler statistics.
Warwick has will attempt to have these in place for the december
shutdown.
Associated work required:
a. Defining event. Dave McConnell
The ACC will need to provide an event which
controls the data taking interval. The interval starts at the
end of the HOLD; it ends some time before the end of the cycle;
(long enough before the end that the corrections to the sampler levels
can be calculated and implemented, and long enough that a reliable
number can be computed.
Thus the ACC (Dave McC) will need to provide two events to cover items
1 and 3; both start at the same time, buthave different end points.
b. Some backplane wiring is required: Ron Beresford.
- to get the event line to the sampler.
- to connect the sampler to its dataset.
c. CAOBS will need some attention to compute the dynamic hold
and distribute it to the ACCs and the correlator.
4. X-Y phases.
A lengthy discussion followed, inspired largely by the fine data
provided by Andrew Bish and Bob Sault's analysis. The notes which
follow are a summary of the discussion along with clarifying
interpolations.
I will be happy to receive comments, and perhaps we could return
to this matter at a later forum.
---------------------------------
- Terminology : We have two independent conversion chains observing
at sky frequencies F1 and F2.
Each conversion chain has two IFs, one for each polarisation, called
A and B (freq 1); VIS has C,D for freq 2; elsewhere they are A2, B2.
For reasons lost in time the receiver probes are labelled X and Y,
with the mapping A <-> X; B<-> Y
----------------------------------
A. The hardware.
At the antenna :
A switched noise diode is coupled into the feed throat with a probe
positioned at 45 degrees to both the X- and Y- receiver probes.
The diode plays two roles: it is used to monitor the gain (Tsys),
and it is used to monitor the relative stability of the two IF chains,
the X-Y products. The switching rate is of order 10 Hz.
At the correlator:
Two correlator cards are associated with each antenna, providing
a standard cross-spectrum for each frequency channel (ie, we get both
spectra, AB and CD). The cross-spectra are always available (in principle),
but in practice they have been omitted from most line observation
configuration files - only those which record all four polarisation
products will also include the x-y products; the correlations are
synchronous with the noise diode switching.
At the Calibration software (CACAL).
CACAL expects the array to be tracking a point source. It proceeds in
3 steps:
i. adjust the A-polarisation phases to be zero on all baselines involving
the reference antenna; that is, the A-pol phase of every antenna,
is offset relative to the reference antenna's A-pol phase.
ii. adjust the B-pols.
iii. Subtract the reference antenna's x-y phase from all the B-pol phases.
If the antenna/feed instrumental polarisation is zero, this stage
will allow us to form I,Q,U and V directly from the visibilities.
B. Theory.
The x-y phase does not provide a measure of the instrumental polarisation
which arises in the feed and antenna; it is likely also to be affected
by the specifics of the probe coupling into the waveguide.
For these reasons we should regard the x-y phase as a monitor (at best)
of the relative stability of the conversion chains.
The basic idea was that we would perform a careful calibration at the
start of an observation; we would also determine the x-y phases.
Thereafter, any movement in the x-y phases should reflect a movement
in the phase difference betwen the two conversion chains.
Since Bob Sault's calibration machinery in Miriad determines the "sky" x-y
phase (ie, the instrumental polarisation), we can hope to keep the measured
x-y phase honest whenever we return to the calibrator.
C. Questions:
- Does the system work .. is every x-y phase step associated with
a step in the sky (visibility) phase?
- If it does work, how do we make best use of the data?
D. Observations.
Does it work ?
Much work in recent times at narrabri has greatly improved the quality
of the data. Andrew's plots (at 1.4 and 2.6 GHz)
showed that the phase noise is low (~0.2 degrees p-p = thermal);
the spread between antennas is around +/- 2 degrees.
The spread we (tentatively) attribute to the antenna/feed instrumental
polarisation. That is, each antenna has its own "peculiar" offset
from its x-y phase; so the CACAL process gives zero on the reference
antenna, and the difference [offset(antenna i) - offset(ref-antenna)]
on antenna i.
Bob Sault has analysed carefully some recent runs and has shown:
a. x-y phase structure tracks faithfully structure in the visibility
phase.
b. The measured "sky" x-y phases are in good agreement with the measured
noise diode x-y phases; that is, they have the same structure, but
they differ by small (+/- 2 degrees) offsets.
c. But there are still a few anomalies, at the few degree level. That
is, during a recent test run CACAL was invoked in mid-observation. This
produced a change in the difference between the sky and noise diode
x-y phases. Analysis continues, but the suspicions are in the
direction of the visibility data (the closure error also jumped
when CACAL was run ??).
(The point here is that we expect sky and noise diode to track closely,
separated simply by the antenna/feed instrumental polarisation; any
changes should be slow, and not abrupt, coincident with a CACAL).
Thus at this stage it looks as if the x-y phase is a valid,
high signal/noise measurement, but that at the 1-2 degree level
there are still a few unresolved questions.
Bob's analysis shows that the difference ("sky" - noise diode) is small,
which indicates that the on-axis feed/antenna polarisation is indeed
at the 1-2% level.
What do we do with it?
To some extent it looks as if we have an excellent measurement
which we don't exploit. One problem (as it were) is that the
receivers and conversion chains are stable, so the occasional
calibration observations probably contain all the safeguards we
might need.
However:
a. Diagnostic.
Constant monitoring of the x-y phase (and amplitude)
allow us to look for problems even during the observation of a
weak field. The antenna 5C problem during the CMB/S-Z observations
was a good demonstration of this. ASSISTANCE knows about this and
should be believed.
b. Calibration.
An attenuator change can cause a small (1-2 degree) phase change.
It therefore makes sense to apply the x-y phases before we invoke
the self-cal machinery. But there is a problem: how do you know
which of A or B has changed? One possibility is to impose the
measured x-y data as a constraint in any self-cal. solution.
Since the phase variations are small their impact on continuum
(Stokes I) images is probably small; they become rather more
significant in polarisation images.
c. CACAL
CACAL uses only one of the 6 available x-y phases (the reference
antenna's data) - it would be marginally cleaner to use them
all.
An additional refinement could be to examine the ("sky" - noise diode)
differences; if these are stable, they could be applied as part
of the CACAL operation. (Again, the aesthetic return may not
warrant the effort, although it may have some implications
for our long-term aim to provide on-line fully calibrated
data).