[evlatests] A short summary of 3-bit results

[Rick Perley]

    After a couple of hectic weeks of testing, I thought a short summary 
might be useful:

    1) Noise measurements of the astronomical signal when observing 
blank sky show the 3-bit samplers give 20% higher noise than the 8-bit 
samplers.  This is a very stable and repeatable result.  At most, a 5% 
degradation is expected. 
    2) The 8-bit samplers noise values are at the theoretical level 
(~0.13 Jy rms for 2 MHz spectral channel width and 1 second averaging) 
for the accepted SEFDs at C-band (the band we have done all testing at). 
    3) Both polarizations give the same results, although there is some 
small chance the RCP is slightly higher (previous observations have been 
skewed by two of the samplers on that side misbehaving). 
    4) Turning off fringe tracking causes the noise values to drop by 
20%.  This happens for *both* 3-bit and 8-bit data paths, providing 
strong evidence that this effect is a gain change, not an improvement in 
sensitivity (since the 8-bit data are believed to be at theoretical 
values with fringe tracking on).  But if this effect is a gain change, 
why do the autocorrelation amplitude remain unchanged when the fringe 
tracking is turned off? 
    5) Comparison of the cross-polarization antenna autocorrelations to 
the same-polarization antenna autocorrelations, when the same signal is 
provided to both samplers shows the cross powers are about 92% of the 
auto powers. 
    6) All other aspects of 3-bit testing have given good results so 
far:  Bandpass shapes, phase and amplitude stabilities all look good.  
There are some concerns about closure, but more testing of this will be 
needed before definitive statements can be made.