[evlatests] EVLA L-Band polarization

[Rick Perley]

    I used 4 hours of available dynamic time last Saturday afternoon to 
take a close look at the EVLA (and VLA) polarization behavior at L-band. 

    The source observed was 3C287 -- 7.0 Jy, unresolved, and with very 
low intrinsic polarization -- less than 30 mJy, or 0.5%.  This ensures 
that the vast majority of the amplitude and phase on the cross-polarized 
correlations (RL and LR) are due to the antenna polarization leakage.  
The source is very near the zenith (dec = 25), so most of the 
parallactic angle rotation is concentrated around meridian transit, 
which occured 3 hours after the beginning of the run.  This provides a 
useful diganostic for inappropriate behavior. 

    The data were calibrated using standard methods.  It is emphasized 
here that the contributions to the parallel hand amplitudes from source 
and antenna polarization is 2nd order (i.e., it involves products of 
D*P, and D*D, where D is the antenna polarization, and P the source 
polarization).  Even with 15% antenna polarization, the influence upon 
the parallel hand gains is less than 0.1%, and can be ignored in 
comparison to what is seen in the X-hand correlations.  (Indeed, 
variations due to V, the circulation polarization, are likely to be 
greater, as this term enters directly). 

    A 'PCAL' was run -- both using the (well-behaved) VLA antennas only, 
and with all antennas, using VLA antenna 7 as a reference.  I also did 
an 'EVLA-only' solution, both with 14 and 16 as references -- this makes 
no change to the problems noted below. 

    Plots were made of the cross-polarization response, in both 
amplitude and phase, with and without the polarization solution 
applied.  For a well-behaved (i.e. stable) system, the expectations are:

    1) For data which are calibrated only with the parallel-hand gain 
solutions (that is, no polarization corrections applied), the RL and LR 
amplitudes should be fairly steady at the values given by the 
cross-polarization, with a ~60 mJy modulation centered at meridian 
transit.  The phases should also be steady, with a corresponding change 
at meridian transit.  The magnitude of the change will depend upon the 
strength of the antenna polarization -- the larger this is, the smaller 
the phase modulation.   Ionospheric Faraday rotation will also modulate 
the phases, but this will also be small, since the antenna polarization 
is much less than that of the antennas. 

    2) For data which are fully calibrated with both gain and 
polarization, the data will have the  phase rotation due to the 
antennas' parallactic angle rotation applied along with the antenna 
polarization, so the cross-hand amplitudes should be at level of the 
source polarization (30 mJy), and the phase steady, absent ionospheric 
variations.    Such ionospheric effects should be much more visible now, 
as the antenna's signal should be removed. 

    Results:

    A) VLA antennas

    The above description is a fair representation of the observed 
X-polarization signals on VLA to VLA baselines.  Most of these baselines 
show about 200 mJy of X-polarized amplitude -- about 3% of I, and are 
constant to a few tens of mJy over the run.  Some baselines show 
significant deviations from this however -- 7R9L changes by over 100 mJy 
through the run, and 7R28L changes by over 125 mJy.  The X-hand phases 
are also quite constant (10 or 20 degrees, typically), but again there 
are a few notable exceptions.
       After applying the polarization corrections, most VLA-VLA 
baselines show the expected 30mJy signal.  Most baselines showed a 
notable rotation in cross-hand phase, on both IFs,  as might be expected 
with ionospheric rotation.  As expected, the LR phase rotates oppositely 
to RL.  But not all baselines showed this.  It is noted that the phase 
rotation is centered about meridian transit -- it which case it looks to 
me to be due to incorrectly determined antenna polarization. 
      
    B)  EVLA antennas.

    The RL amplitudes are both very high (over 1.1 Jy -- 16%), and 
highly variable, particularly on some of the EVLA antennas.  Variations 
of over 200 mJy on a good VLA to some EVLA antennas are seen -- the 
variations are greater than 500 mJy on some EVLA-EVLA baselines!  
Similarly, phase rotation in the 'raw' (not corrected for polarization) 
data of over 100 degrees is seen on some baselines.

    After polarization calibration, the situation is little improved -- 
this is no surprise, as the large variations in the antenna polarization 
are not included in the software's model!  Full calibrated X-hand 
amplitudes are varying by up to 200 mJy (7 times the source's true 
polarization!), and phases are correspondingly unstable.  Virtually all 
baselines show the phase rotation noted above -- centered about meridian 
transit. 

    Not all EVLA antennas behave this badly.  Some are moderately 
stable.  Using EVLA antenna 16 as a reference, the polarization 
solutions for a few (not including 14!) seem fairly good.  But overall, 
the EVLA polarization performance is much worse than desired. 

    C) Images. 

    I made polarization images from these data.  Stokes 'I' images for 
VLA, EVLA, and EVLA+VLA, look just great, with noises at a level many 
tens of thousands below the peak.  Stokes 'Q' and 'U' look terrible in 
all combinations -- but are especially bad for EVLA-only data, where the 
noises are many times higher than in Stokes 'I'.  Based on the stability 
of the cross-hand visibilities, none of this comes as any surprise. 

    D) Further work. 

    It is probably worth doing this type of careful persual of the data 
at other bands, to establish current standards of normal behavior for 
VLA and EVLA antennas. 
    More important is to establish the origin of the high variability in 
most of the EVLA's antennas.  One astronomical test (which will not be 
popular amongst users) might be to remove the quadrature hybrid from an 
antenna, and see if better stability is found. 
    Other suggestions are solicited.

    Plots are available in my office for perusal by any interested party.