[daip] FRING question

[Walter Brisken]

Hi DAIP,

In doing some correlator comparisons (DiFX versus the JIVE SFXC software 
correlator, and with come comparisons to the old VLBA hardware correlator 
thrown in the mix too) we have identified that FRING reports substantially 
different SNR for the same data when different spectral smoothing is 
applied.  This brings up the question:  In FRING, how is SNR actually 
calculated?  Specifically, how is the noise calculated?

As a slight follow-up to our explorations, we tried running FRING before 
and after deleting the supplied CQ table which supposedly contains enough 
information to describe the spectral purity.  We think that this table 
(and the values of spectral smoothing) should be considered when 
calculating the FRING SNR.  Basically we find that when neighboring 
channels are correlated the reported SNR is artificially high.

We are still following some lines of evidence to confirm our suspicions. 
Some input from you would certainly help us figure this out.

 	-Walter