[evlatests] EVLA Test Meeting, May 28, 2009

[Barry Clark]

1.  Polarization circularity.

R. Perley reported that the 6GHz receiver on antenna 8 was accidentally
installed 90 degrees from its usual location, effectively making the
parallactic angle 90 degrees different from other antennas.  Having
a different parallactic angle allows one to solve for the true circular
polarization purity of the whole array, rather than just the
differential leakage available when the receivers are installed with the
same parallactic angle.  He finds the polarization purity is typically
five to eight percent midband, rising to about 15% at the band edges.
The orthogonality is good, so the differential leakage terms are
manageable at all frequencies.  G. Moellenbrock pointed out that the
absolute polarization may be important for calibration and dynamic range
issues, since this may couple source polarization into the RR and LL
correlations.  He suggests that starting with a model for the absolute
polarizations could alleviate this.

2.  Polarization stability.

B. Sault reports that at C band (4.3 GHz to 4.8 GHz in 500 MHz steps),
the stability over a fun of a few hours was 0.1% for the good antennas,
0.3% for the bad antennas and frequencies.  He observed again two
weeks later, and tried using the D terms from the earlier observation.
Errors ranged up to about one percent, so at this band, sub percent
level polarization measurements may be done using old D terms.  It is
encouraging that the worst antennas were the oldest - 14, 16, 13.  Some
frequencies appeared a bit worse than others.

K band was a more complicated story.  He observed 18 to 26 GHz in 1 GHz
steps.  Variations up to about 3% were seen.  Errors appeared to have
a generally quadratic variation with frequency.  The polarization
pattern within the beam would give a quadratic variation with frequency
if the pointing were in error, but it would require an unrealistically
large pointing error of order 20" to explain the observed effect.
Because of the possible connection with pointing error, he tried
correlating the changes with the pointing steps produced by the
reference pointing observations (which were done every half hour).
There was no apparent correlation.

He tried making maps, and showed maps in Stokes Q.  A map made with
calibration every half hour looked much cleaner than a map made with
the same D terms for the entire observation, having no obvious 
artifacts.  However, the off source noise was about twice thermal.  This
was on 3C 84.  (RP confirmed similar results on another observation.
Maps on blank fields have the expected noise.)

W. Brisken asked what effect the subreflector rotation trick might have
on polarization.  (The 'trick' is rotating the subreflector to produce a
cubic term in phase across the dish to cancel the cubic phase term
induced by the gravity sag of the subreflector support structure.)
Nobody knew the answer.

3.  Phase variation signatures.

V. Dhawan has observations with redundancy, with AC and BD IF pairs set
to the same frequency.  Then one may look at variations in RR-LL phase
on one IF, which eliminates atmospheric variations and almost all active
electronic components, or RR phase in one IF minus RR phase in the 
other, which includes the difference between two L302 modules.  He
showed examples with variations of a few degrees in minute or so time
scales.  The RR-LL and RR1-RR2 phase variations are qualitatively
similar.  He states that variations are comparable at the high bands (Q,
K, Ka) and a factor of three or so smaller at the low bands (C, L).  He
accuses the effect of being variable reflections in LO paths which have
insufficient isolation, and will pursue experiments along that line.

RP notes that the RR-LL phases do not show any long-term systematic
effects.

4.  Interference scans

E. Momjian and B. Butler have concocted a script which scans the entire
spectrum using W. Brisken's D351 software to produce spectra in 1 GHz
chunks with 1 MHz resolution in fairly short order.  He also has a
higher resolution version to look at any 1 GHz chunk with 30 kHz
resolution.  In coming weeks there will be some work done on identifying
interfering signals.