[evlatests] An accidental discovery of how good fshift is ...

[Rick Perley]

    I used SPFLG to take a look at the autocorrelation spectra, to try 
understand why subband #9 showed the same total power variations, and 
the same gain suppression as subband #8, yet the cross-power spectra 
showed no RFI.  A remarkable story unfolds:

    The autocorrelation spectra showed that there is indeed a 2nd RFI 
spike, this one at 349 MHz.  And it's exactly the same in its time 
variability as the one at 339 MHz.  Indeed, each is exactly 5 MHz away 
from the 'channel 0' frequency of 344 MHz. 

    Ken provides the rest of the story:

    Subbands 1 through 8 are taken from one of the 128 MHz-wide 
'chunks', while subbands 9 through 16 come from the next 128 MHz chunk.  
The frequency of the boundary between them is 344 MHz.  The signal at 
339 is the real one, the signal at 349 MHz is its alias.  The reason 
that this latter signal doesn't show up in the cross-power spectra is 
due to the fshift.  Indeed, the observation that the cross-power spectra 
of subband#9 shows (almost) no effect from the alias is a nice 
demonstration of how effective the fshift is. 

    Because the aliased power is real, and unaffected by fshift, the 
gain settings, including the requantizers, and the switched power, all 
see it.  And this is why, for that period when the super-strong RFI was 
active, the PSums in subbands 8 and 9 were nearly identical. 

    P-band is a rough neighborhood to play in ...