[evlatests] 12.5 MHz spectral noise, and self-test

[Rick Perley]

    I ran a test this afternoon to see if the presence of self-test has 
any effect on the noise histograms of data taken in 12.5 MHz spectral 
line modes.

    Answer:  It does -- and the difference entirely accounts of the high 
noise seen in the tests of recent weeks.

    Details:

    I ran short observations of 3C147 in 50-MHz continuum, and 12.5 MHz 
spectra line in modes PA, PB and 4.  The observations were run twice, 
once in standard mode, and once with self-test turned off. 

    I have reduced only the PA mode data thus far -- the result is very 
clear.  With self-test turned on, the parallel-hand histograms of the 
imaginary part of the visibility (for calibrated data) have a width of 
1.0 Jy.  (Channel width of 780 kHz, averaging time of 1.66 seconds).  
With self-test turned off, the width of the histogram was divided in 
half -- to 0.47 Jy.  This is seen in both VLA-VLA baselines, and 
EVLA-EVLA baselines.  The effect can easily be seen even in the noise of 
the CALIB solutions.