[evlatests] 12.5 MHz Spectral Line Noise

[Rick Perley]

    Ken and I ran more tests yesterday afternoon on the curious recent 
discovery that the rms noise in 12.5 MHz BW spectral line observations 
is approximately double that expected for most products. 

    The goal of this test was to ascertain whether turning off the 
'self-check' feature eliminated the additional noise in all cases, and 
to determine whether there is any effect of 'self-check' for 25 MHz 
spectral line data.  We remind readers that use of self-check in 25 MHz 
BW spectral line data resulted in small correlator offsets, which are 
removed if self-check is disabled. 

    Short Answers:

    To #1:  Disabling 'self-check' removes the extra noise for all 12.5 
MHz BW spectral line modes. 

    To #2:  Self-check has no effect whatever upon 25 MHz BW spectral 
line sensitivity. 

    Details:

    Observations were made of 3C147 (for calibration) and of a nearby 
blank field (for determination of noise statistics) at 20cm in a number 
of modes:
          BW = 50 MHz:  Mode 1A
          BW = 25 MHz, 12.5 MHz and 6.25 MHz:  Modes PA, PB, and 4

    The file was run twice, once with the self-check feature turned off, 
and again, after an hour, with the self-check turned on. 

    Results:

    For all data taken at 50 MHz, 25MHz, and 6.25 MHz BW, there was no 
difference in correlation coefficient or in the width of the noise 
histograms, for any mode, between the self-test 'on' and self-test 'off' 
states, other than that expected due to the change in elevation between 
the two observations.  (I mention this, as the source elevation changed 
by about 15 degrees between the two observations, resulting in a notable 
improvement in correlation coefficient, and a parallel decrease in 
histogram width.  The improvement is most notable on VLA antennas, and 
very modest on EVLA antennas -- exactly as expected!) 

    For the 12.5 MHz BW observations, the observations are completely 
consistent with my prior results, which I summarize below:

    A)  There is no change in correlation coefficient between the 'on' 
and 'off' states. 
    B)  All histograms for all correlation products are of the expected 
width when the self-test feature is turned off. 
    C)  When the self-test feature is on, then:

             1)  The histogram widths for the RR and LL correlations in 
mode 'PA' are doubled in width.  The histogram widths are of the correct 
width for correlations 'RL' and 'LR'.
             2) For modes 'PB' and '4', all correlation products have 
histogram widths which are at least doubled.  (For mode 'PB', it seems 
to be exactly twice.  For mode '4', the factor appears closer to 2.2). 

    Ideas?