[evlatests] More 3-bit stats ...

[Rick Perley]

    My report yesterday on the 3-bit noise yesterday for the various bit 
levels was incomplete, as I had not compared the 3-bit noise levels to 
the 8-bit 'truth'.  This is now done, but there are no surprises.  A 
couple of perhaps interesting trends, however ...

    To simplify the effort, I compared the histogram noises for the 012 
and 345 bit selections to the 8-bit noises, taken at the same time, and 
in the same frequency subband.  These two observations were taken at 
different elevations (10 degrees different) and at different times (80 
minutes).  There are four ratios of interest:

    1) A1C1/B0D0 at the low elevation
    2) A1C1/B0D0 at the higher elevation
    3) A1C1(high)/A1C1(low)
    4) B0D0(high)/B0D0(low)

    Results:

    1) This shows the usual degradation of the 3-bit system, compared to 
the 8-bit.  The ratios are a little higher than expected -- ranging from 
1.09 to 1.28.  However, the higher ratios are all associated with 
specific antenna-polarizations:  26A1 is easily the worst (average 
1.25), and 23C1 a close second (1.17 average).  06C1 is also poor (1.14 
average). 
    2) This is the 3bit/8bit degradation ratio at the higher elevation.  
Curiously, the values average a little less than those in (1) at the 
lower elevation, ranging from 1.03 to 1.30.  The same antenna-IF 
dependency noted in (1) is seen here also, but (for example), ea26A1 is 
now 1.15, rather than 1.25, while ea06C1 is 1.20, rather than 1.14.  'It 
goes both ways'. 
    3) This is the ratio reported yesterday:  The average ratio is 0.98, 
with a dispersion of .05.  The high end of the distribution is dominated 
by ea06C1, with an average of 1.07.  (This is reflected in the ratios 
given above in (1) and (2)). 
    4) This average is  1.01, with a dispersion of .04.  The difference 
in the averages in (4) and (3) reflect the difference between those in 
(1) and (2). 

    My bottom line:  There's no salvation here.  The noise is the same 
for any of these 3-bit selections, to within the measurement errors.