[evlatests] sub-band to sub-band changes

[Rick Perley]

    A quick perusal shows that the PSum varies remarkably from subband 
to subband, most notably in the 3-bit data paths (but also seen in at 
least a couple of the 8-bit paths as well).   Adjacent subbands, at the 
same time, will have PSum values different by up to 30%.  I see no 
evidence for a trend in PSum values across the 16 subband IF -- evidence 
that the slope filters are working about right. 

    This subband to subband changes should be visible in the bandpass 
solutions (if done 'absolutely', rather than normalized).  I'll check 
this next.