[evlatests] Polarization

[Rick Perley]

    I believe we have the polarization problem sorted out. 

    A 4.5 hour test was run last night, on 3C147, at 20cm.   Two 
continuum setups were used, with bandwidths of 50 and 12.5 MHz, plus the 
polarization spectral line mode, at 12.5 MHz BW.   The frequencies were 
set at 1485 and 1365 MHz.  (I also observed at 1745 and 1865 MHz -- 
reports on these will be made later). 

    Bottom Line:

    The residual polarization seen in continuum observations is 
completely absent in spectral line -- *** providing bandpass solutions 
are applied***.   When these are applied, beautiful noise-limited Q and 
U images were obtained, where the rms noise levels are 0.01% of the 
total intensity.  There are no hints whatever in the image of residual 
signal.  The antenna polarizations are constant, and (essentially) 
perfectly removed by the software. 

    Some More Details:

    The 'lumpy bumpy' polarization residuals that I've showed before are 
present, at exactly the same level, in the new continuum data, at both 
bandwidths.  And I can generate a similar image from the spectral line 
data, provided I form a 'channel 0' without applying a bandpass 
solution.   When the bandpass function (derived from the parallel-hand 
calibration) is applied, the polarization residual problem is completely 
removed. 
    This would suggest that the fundamental cause of the problem is in 
the bandpass mismatch between the VLA and EVLA.  If so, then calibrating 
and imaging using EVLA-only, or VLA-only, data should show a much 
reduced effect.  Curiously, this is not the case.  The intra-array 
baselines, despite have low parallel-hand closure errors, retain the 
polarization offset problem.  I do not have an easy explanation for this 
result.