[evlatests] 3-bit update

[Vivek Dhawan]

Today spent mucking with the slope equalizers and requantizer gains.

Bruce Rowen's code to measure gain slope works, it makes a plot that
agrees very well with the widebandCorrelator spectrum. The equalizers
were adjusted based on the measured slope (there is a factor 2 in the
units: slope/1GHz vs. slope/bandpass which is 2GHz). These changes
did affect the Tsys in small ways, usually decreasing it.

(A Tsys puzzle: after the optimizing, the Tsys looks reasonable, ~30K.
For some ant/pol the un-optimized Tsys is lower, ~10-20K, clearly
not real. This could be the same thing Rick sees as artificially
good SEFD - the mechanism remains mysterious.)

Then requantizer gains were set to give post-RQ rms of 2.7, these were
~450 for 3-bit, vs. the nominal 512. There is also a gain in the first
FIR filter stage which wanted to be ~27 vs. nominal 23. These changes
did not affect Tsys, as expected, since the Psys is measured before RQ.

Wind, antennas stowing etc.... had enough.

So, looking good but really need to get calibrated data on/off a source
of known flux.