That's exactly my findings.  Most of those in the know assume the
ear's effective bandwidth to be somewhere in the region of 30 - 100Hz.
  7dB S/N in a signalling bandwidth is pretty typical of what any
power detection software can manage blindly - it is a mathematical
calculation that isn't too difficult to do.  And since that value
falls out of maths, its quite reasonable to expect ears to follow the
laws of physics / maths too.  So in a signal bandwidth, ears / eyes
and signal detection software based on pure non-coherent power
detection are all roughley equal.


Andy


On 12 May 2012 12:04, Roelof Bakker <[email protected]> wrote:
> Hello Andy,
>
> Thank you for this little gem.
>
> I have been playing with it for a while.
> With a signal to noise ratio of -10 dB in 2500 Hz, I can copy a 10 WPM
> message with the output filter set to 2500 Hz.
> Estimating my brain ear system bandwidth as 50 Hz, the effective signal to
> noise ratio is 6.9 dB.
> This seems a bit on the low side, so the biological filter seems to be even
> narrower than 50 Hz.
>
> Setting the output filter to 20 Hz, results in armchair copy.
> The CW tone in this case is 300 Hz.
>
> 73,
> Roelof, pa0rdt
>