[evlatests] Still more on spectral line noise

[Michael Rupen]

>    The constant correlator coefficients indicate that the actual SNR
> remains the same.  The simplest explanation for the oddities at 12.5 MHz
> BW, when in spectral line modes, is that somehow the actual integration
> time is significantly less than 3.33 seconds, or that the actual
> channelwidth is significantly less than the value claimed.


Right.  So I wonder whether it would be worthwhile to try (say) 10s
averaging, to see whether the noise remains off by the same factor (indicating 
loss of the same fraction of data) or becomes closer to the nominal value 
(indicating loss of a more nearly constant number of seconds in each 
integration).


The channel width is an interesting idea...it might be interesting to observe
an absorption line that has some structure over a few channels at 12.5 MHz
total bandwidth, to see whether the line at 12.5 MHz looks like it should
based on (say) 6.25 MHz observations.

          Michael