[evlatests] Correction on sensitivity 'loss' in 8-bit path

[Rick Perley]

    I reported yesterday that there was a significant loss in 
sensitivity at the bottom edge of the 1-GHz-wide IF when in 8-bit mode, 
attributing this to the well-known ~5-db rolloff. 
    The loss of sensitivity is real, but my explanation/association is 
almost certainly wrong. 

    A more detailed investigation, using UVHGM to plot the noise 
histograms as a function of channel shows that the spectacular loss in 
sensitivity is restricted to the bottom 10 channels *only* -- 20 MHz.  
The well-known bandpass rolloff covers well over 100 MHz.  Within that 
span, the sensitivity is completely unchanged for at least 70 MHz -- 
over most of which which the spectral power density is attenuated by 5 
dB or more. 

    I think it far more likely that the sensitivity loss at the very 
edge is due to the the same issue which we've had all along with subband 
1 at the lower edge, and is completely unrelated to the bandpass rolloff 
effect. 

    Note that this problem of increased noise in the bottom edge of 
subband 1 is not present with the 3-bit path.