[Lowfer] WSPR noise bandwidth, was 74.5495 QRSS 60 as usual till at least 0600...

[Garry Hess]

> Reporting an SNR that has no relevance to the software's
> real ability to extract signal is basically marketing hype.

...

>  But in any case, the full SSB channel width is not relevant to SNR.
>
> John

John,

WSPR isn't "magic", nor is it marketing hype. Signal-to-noise ratios 
require some reference noise bandwidth and the WSPR manual clearly 
states SNRs are referenced to a 2500 Hz bandwidth (cfm. 
http://physics.princeton.edu/pulsar/K1JT/WSPR_2.0_User.pdf, page 4). 
Yes, that makes the SNR for minimum decodable circumstances quite 
negative. So what? If one wants to use SNR in the 6 Hz detection 
bandwidth instead, fine; add 10 log(2500/6)= 26.2 dB. Common engineering 
practice is to normalize things to 1-Hz noise bandwidth. We just have to 
be careful comparing different systems to use the same noise bandwidth 
for both. The gold standard for such comparisons of course is the 
Shannon capacity curve versus SNR 
(http://en.wikipedia.org/wiki/Shannon%E2%80%93Hartley_theorem). Thanks 
to modern coding and compute power, the best schemes get mighty close.

73, Garry, K3SIW

-- 
73, Garry, K3SIW, EN52ta, Elgin, IL