Markus wrote:

> wav from the DL0AO cardioid:
> http://df6nm.bplaced.net/VLF/fec_tests/dl0ao_6k47_06101811.wav

> Start offset accounting for ~ 26 s latency so should
> be around 2702 s.

Did not decode.

The message is there but Eb/N0 = -2.9 dB
carrier S/N: 24.68 dB in 10.0 uHz,
            -25.34 dB in 1Hz,
            -59.32 dB in 2.5kHz;

The S/N peaked at the suggested offset, and was slightly
higher at frequency offset -2.3 uHz.

I'm starting another run with a different setting for normalisation.

Stefan wrote:

> Can you tell the message please?

I copied

 "CONGRATS, YOU MANAGED TO DECODE A 100 CHARACTER MESSAGE
  ON THE 46 KM BAND.. 73+55 DE DK7FC IN JN49IK"

> 94% channel capacity and just 31.1 % bit error rate,
> how can this be?

Well we were a bit lucky with this decode, the 31.1%
would be the expected BER at +0.4 dB, but I report the
carrier Eb/N0 as this is is a genuine measure of S/N.

BER is a bit misleading with soft decision decoding, the
individual symbols are not used, the result is decided by
a correlation over the whole message.

Also, with longer messages the rate of the convolutional
code is higher.  Eg 100 chars, the inner code carries
608 bits encoding to 5024 symbols, rate 1/8.26.  If we
sent 10 chars with same CRC and code, it would be 68
bits encoded to 704 symbols, rate 1/10.35.  The difference
is because of the constant 'overhead' of the termination
symbols.  Shorter messages will tend to have lower Es/N0
and therefore higher BER for the same Eb/N0 as a longer
message.

... the other run just finished with no decode.  DL0AO is
a couple of dB short of decoding.

--
Paul Nicholson
--