[evlatests] 3-bit sensitivity results for different bit levels

[Rick Perley]

    Ken ran an observation this morning with the 3-bit antennas, using 
the usual setup (A1/C1 set to same frequencies as A2/C2, and B0D0 (the 
8-bit path) set to the same frequencies as the center of the AC 
paths).   Ken arranged for the 012, 123, 234, and 345 bit levels to be 
utilized from the 6-bit high speed sampler.  The 8-bit system was kept 
unchanged for all four setups. 
    We observed 3C345 and a nearby blank field, with about 4 pairs of 2 
minutes each.   The elevation of the observation changed from 30 degrees 
to 38 degrees.  The reported system temperature from the 8-bit switched 
power monitor did not change more than 0.1K (except for ea18) throughout 
the experiment. 
    I did basic calibration of these data (delay, bandpass, gain 
calibration), and used the central channels, on the noise field, to 
determine the thermal noise levels for each baseline pair and 
polarization pair, for the A1/C1 path. 

    The results show that there is no change in sensitivity for any of 
the 3-bit level combinations for the 7 antennas as a whole.  But there 
are small and significant changes for individual antennas.  For example, 
baselines with ea26A1 are 7% more sensitive with bits 345 than with 
012.  However, baselines with ea06C1 are 6% less sensitive with 345 than 
with 012.  Overall, 345 shows a 2% improvement compared to 012, but the 
dispersion in the distribution is 5%.  (These statistics taken over 29 
baselines, not including ea18). 

    It's a wash.