> 'DK7FC JN49IK 73'
> Eb/N0 = +1.1 dB   BER = 37.8%

Typo, BER = 37.3%

Overall S/N 17.6 dB in 6.0 uHz, which is -68.6 dB in 2500 Hz.
About -88 dB before sferic blanking.

Each received symbol has its complex amplitude normalised by the
mean of its neighbouring +/- 100 absolute symbol amplitudes.

Symbols are 80 seconds each, so that's a 4.4 hour window.
That seems to be about optimum (for this particular message)
with a broad max:  +/- 200 symbols is 0.3 dB weaker, +/- 50
symbols is 0.1 dB weaker.

Stefan wrote (5th Oct):

> f = 6470.00000 Hz
> Start time: 05.10.2016   21:00:02 UTC
> Symbol length: 140 s
> Characters: 15
> CRC 16
> Coding 8K19A

Unable to decode this one with the same trick.

Stefan wrote: (3rd Oct)

> f = 6470.00000 Hz
> Start time: 03.11.2016   06:00:00 UTC
> Symbol length: 30 s
> Characters: 11

and 5th Oct

> > A second transmission is running now
> Same message, symbol length, phase orientation
> and start time...

From the coherent sum of these two transmissions:

 'ROCK N ROLL'  Eb/N0 = -0.2   BER 39.5%

S/N 15.7dB in 19.7 uHz.  Constant reference phase and
using +/- 100 symbol normalisation (1.7 hour window).

--
Paul Nicholson
--