[Lowfer] WSPR noise bandwidth, was 74.5495 QRSS 60 as usual ...

[JD]

Jay, I know what the manual says... I've been reading it and the WJST manual 
and others, long before I started messing with WSPR myself.  But simply 
stating an arbitrary number in a document doesn't explain the point of 
decreeing a noise bandwidth that's unrelated to channel bandwidth.  That's 
not and never has been standard engineering practice.  As I say, any agreed 
mutually bandwidth may be fine for *threshold* comparisons, but that is 
simply not the same thing as signal-to-noise ratio.

Comparing one mode to another with an arbitrarily wide noise bandwidth also 
has no basis in information theory, where S/N is _defined_ with the same 
noise bandwidth as the transmission channel necessary for the specific 
signal.  Without the same bandwidth for the noise as signal, you're not 
exploring theoretical SNR limits at all...first, because that's not how SNR 
is measured, and second, because you're inflating the real number by an 
amount that is going to vary with both (a) the necessary communication 
bandwidth for a given mode relative to others, and (b) where in the noise 
passband other non-random signals may fall.  Just a broad illustration of 
the latter:  If the software doesn't even LOOK for signals to decode outside 
its primary 200 Hz analysis band--as WSPR doesn't--then anything else 
outside that band, be it random or coherent, makes no difference to the 
decoding.  So, why should the other 2300 Hz be counted as noise and allowed 
to affect the supposed SNR?  That doesn't provide a common basis for 
measuring detection efficiency at all.

The standard definition of S/N matters for the same reason the "m" in dBm 
does.  If a person starts using "dBm"  when he merely means a ratio in dB, 
or uses it to indicate power referenced to something other than 1 milliwatt, 
then it loses its usefulness.

73
John