[evlatests] VLA to EVLA Closure

[Rick Perley]

    The continuum closure problem for VLA-EVLA baselines is well 
documented, and has been measured in numerous tests at 50 MHz BW.  But 
no attempt has been made to measure the magnitude of the effect at other 
bandwidths.  The known cause of the problem -- a ~180 degree 'hook' in 
the VLA baseband phase over the bottom ~2 MHz -- leads us to predict the 
closure effect (or, speaking more precisely, the loss of coherence on 
VLA -- EVLA baselines) will dramtically rise as the bandwidth narrows, 
so the fraction of the bandpass within which the 'hook' occurs rises. 

    I took some software time to do this little test.  Observations of a 
clean point source (BLLac to the old-timers, 2202+420 to the more modern 
set) were made in continuum at all bandwidths between 50 and 0.78 MHz.  
(Yes -- nobody in their right minds should ever observe in continuum at 
the very narrow bandwidths -- but never mind ...).  I used X-band, as 
this gives us all 9 antennas.  5 minutes integration at each bandwidth. 

    Results:

    Exactly as expected. 

    50 MHz:  No results possible, as the EVLA antennas didn't get on 
source in time.  (This was the first observation). 
    25 MHz:  Typical closure is 10 to 12%.
    12.5 MHz:  15%
    6.25MHz:   15 -- 20%
    3.125 MHz: 30 -- 40%
    1.56 MHz:   35 -- 50%
    0.78 MHz:   ~50% or more! 

    Phase closure errors are negligible. 

    In all cases, the errors are expected to be stable in time, so that 
application of a single BLCAL solution should remove the 
baseline-dependent offsets.  It should also be remembered that there is 
a loss of SNR in this effect, which is not recoverable.  The loss is 
very small at 50 MHz, (perhaps a couple of percent), but will be very 
significant at the very narrow bandwidths, where the signal strength has 
been cut in half, or more, by the phase wrap. 

    Because the number of VLA antennas still outnumbers EVLA antennas, 
the standard calibration routines will ensure that the EVLA antenna 
gains estimated by the CALIB program will be low, resulting in a 
post-calibration amplitude which is quite a bit too high on the EVLA to 
EVLA baselines.  The 'closure' errors reported by CALIB will thus be 
dominated by the EVLA to EVLA baselines, even though the problem 
orginates in the VLA to EVLA baselines.  This situation will slowly be 
reversed as EVLA antennas outnumber VLA antennas.  In between, the 
reported errors will be seen everywhere.