[evlatests] L-Band Polarization for the EVLA and VLA

[Rick Perley]

    I've reviewed the L-Band observations taken in September and October 
for the purpose or characterizing the EVLA's polarization behavior. 

    There were four runs made:

    Sept 29, HA = -3 to +1, source = 3C287.  Local time was 11 AM -- 3 PM.
    Oct 05, HA = -3 to +1, source = 0555+398.  Local time was  3AM -- 7AM
    Oct 14, HA = -3 to +1, source = 3c287, 3C286.  Local time was 10AM 
-- 2PM.
    Oct 19, HA = -2.5 to +4.5, source = 3C286, 3C287.  Local time was 
10AM -- 5PM. 

    3C287 is an ideal source for antenna polarization studies, as its 
native linear polarization is less than 0.4%.  0555+398 is almost as 
good -- polarization is about 0.9%.  Both are small enough that listing 
of RL or LR correlations (following suitable gain calibration) will 
clearly show the behavior of antenna polarization.
    3C286 is a highly polarized source (about 11% at L-band), and was 
not utilized in this study. 

    I carefully edited and calibrated the data.  'Closure' corrections 
were made, and to judge how well the calibration and editing were done, 
I made large-area images of both sources, using all antennas.  The 
dynamic ranges were close to 100,000:1 (!) -- certainly showing that 
data quality is good, and that the VLA and EVLA can be safely combined 
to make high fidelity images. 

    Plots of the RL and LR amplitudes quickly showed that all baselines 
to the EVLA antennas have a large and time-variable polarization -- as 
high as 20%!    In reviewing all the data, the following conclusions are 
made:

    1) Virtually all baselines to an EVLA antenna show polarization 
amplitudes above 8% most of the time. 
    2) EVLA to EVLA baselines are no better, or worse, than EVLA to VLA 
baselines. 
    3) VLA to VLA baselines are far better -- very few have polarization 
above 5%. 
    4) IF 2 (1365 MHz) is notably worse than IF1 (1465 MHz)
    5) EVLA antenna polarizations change slowly, on timescales of hours, 
by up to a factor of two or more.   The phases are more stable, but 
still vary by up to 1 radian. 
    6) VLA antenna polarizations are nearly constant -- even antenna 28, 
which is anomalously high. 
    7) The amplitude of the polarization, and the pattern of the 
variability, are nearly identical on each of the four days, and for the 
two sources. 
    8) The variability is not quite symmetric w.r.t. hour angle (or 
elevation or parallactic angle).  There seems to be symmetry about an 
hour angle of +0.5.  

    The important differences between EVLA and VLA antennas are the new 
IF electronics, and the replacement of the dielectric phase shifter with 
a microwave quadrature hybrid.  The variable polarization indicates 
there is some improper coupling between the R and L signal paths.  
Although this could conceivably be happening in the electroncs, EVLA 
requirements stipulate a very high degree of isolation between these 
channels.  Thus, suspicion here centers on the quadrature hybrid.   
These are located in the crygonic receiver, between the OMT and 
amplifiers.  Being in this environment, it seems hard to understand how 
they would be sensitive to (say) elevation.  However, if, for example, 
these are badly matched due to their being at cryogenic temperatures, 
some signal in one of the linearly polarized inputs would be reflected, 
to return with a different phase and amplitude, and possibly some 
component in the orthogonal mode.   This reflected component would be 
sensitive to such variables as temperature and elevation. 

    It's important that we diagnose and understand the origin of this 
unacceptably high cross-polarization.   If indeed the hybrids are 
responsible, and if the cause cannot be addressed, we face an important 
decision on whether the EVLA at the lower frequencies should utilize a 
linear, rather than circular basis. 
    Note that even if the problem can be addressed, we still have a 
significant issue with phase matching of the OMT output to the hybrid 
inputs -- especially difficult at C-band. 

    I have interesting plots to illustrate all the above, which I'll 
show at the next test meeting.