[evlatests] More on autocorrelation phase issues

[Rick Perley]

    I noted yesterday that the cross-hand autocorrelation spectra showed 
different delays on alternating subbands.  I've taken a closer look, and 
there is a clear pattern:

    A)  Both 3-bit and 8-bit data show the same effect.
    B) The apparent delay for subbands 1, 3, 5, and 7 are of the same 
magnitude as that for subbands 2, 4, 6, and 8, but of opposite sign.  
The cross-power amplitudes appear about as expected. 
    C) Only the autocorrelation spectra are affected -- cross spectra 
(from different antennas) are normal. 
    D) After applying the R-L delay correction derived from the normal 
cross-power spectra, and important clue emerges:

          -- In 3-bit mode, subbands 1, 3, 5, and 7 have correct (flat) 
phases in the autocorrrelation cross-spectra.  The odd numbered subbands 
have 'doubled' phase slopes. 
          -- In 8-bit mode, subbands 2, 4, 6, and 8 have correct phases 
in the autocorrelation cross-spectra.   The even numbered subbands have 
'doubled' phase slopes.