[evlatests] sub-band power variations

[Rick Perley]

    After perusing a lot of spectra (both BPASS and autocorrelation), 
some conclusions can be summarized:

    1) The autocorrrelation spectra and the cross-power bandpass spectra 
are in excellent agreement -- this is as it should be. 

    2) There are occasional significant (many dB) offsets in the 
spectral power in a single subband.  In *all* cases, these are due to 
the presence of a strong RFI signal -- nearly always the 'phenomenon'.  
Dealing with this in a simple and automated way may be difficult ...  
This problem will certainly be worse in L and S bands, where copious RFI 
can (and do!) dominate subband power. 

    3) There are many instances of small individual subband offsets of 
~1 to 1.5 dB (20 -- 30%).  There is no ready explanation for these -- 
they are certainly not caused by RFI within the subband.  They also 
don't repeat from one setup to another.  These offsets are seen in the 
8-bit path spectra as well -- with about the same probability. 
    Question:  What tolerance should we expect to see from the 
requantizer gain?  Is 1dB about what is expected?  What is the harm of 
having powers deviate from the mean by this amount? 

    4) An ubiquitous feature in the 3-bit path spectra is the presence 
of a ~30 MHz ripple.  This is not seen, or is much weaker, in the 8-bit 
path.  The amplitude is small typically 0.5 dB pk-pk.  The ripple seems 
to be quite stable over time.  It is stronger in some antennas than 
others.