[evlatests] A note on 'cross-auto' correlations

[Rick Perley]

    I've noted before about a strange effect seen in the 'cross-auto' 
correlations:  RxL correlations from the same antenna.  The noted effect 
was that these were large-amplitude sinusoids for most baselines. 

    It has been determined that these are erroneous -- the result of a 
buglet in AIPS.  Eric will fix this. 

    The proper cross-auto correlation spectra show effectively the R-L 
delay in phase, and the cross-polarization in amplitude (provided the 
source itself has no linear polarization). 

    In the various tests today, wherein the RCP signal was fed into both 
RCP and LCP samplers, the output 'cross-auto' spectra are identical 
except for a reversal in the phase slope.  This is as expected.  But an 
unexpected result was also found -- these cross-auto amplitudes are 
significantly higher than the auto-cross amplitudes (RR and LL).  Since 
the same signal is feeding both sides, we expect the same amplitudes in 
all four cross-products. 
    Michael, in investigating this issue, plots the four cross-products, 
for a single channel, as a function of time.  The 'autos' (RR and LL) 
are quite stable, and do reasonable things.  The 'crosses' (RL and LR) 
have multiple states, the highest of which is occupied most of the time 
(and is *much* higher than the 'autos'), and various other states, all 
of lower amplitude, which are occupied rarely, but regularly.  
Explanations involving the 3-level lobe rotator are unconvincing (so far)...

    Stay tuned ...