Domenico wrote:

> a little signal, but decoded also here

Excellent result.   That's the sort of test I like to
see, where Eb/N0 = 0.7 dB and the program will have
had to search for the phase and it had to use some
list decoding.   That's a proper weak signal test.
I guess it should stand another dB weaker.

The 21:30 just decoded here with Eb/N0 = -0.6 dB and
the phase pattern was -30  -60  -30  -60, might be
propagation or my timing system having a hiatus.

The symbol timing of the 21:00 looked good.  I tried
some time shifts both ways and it didn't improve.

Markus wrote:

> It seems that non-existing symbols with null
> amplitude are not being counted as symbol
> errors, which severely biases the error rate.

Yes it looks like you're right, the zero symbols are
not counting as errors.  They contribute neither positive
or negative to the accumulated log-likelihood of the
path metric.  It's effectively an erasure code.  I'll
alter the BER arithmetic tomorrow, it probably should
consider only the non-zero symbols.  Might need some
thought.

Have to go out now but the rx continues recording.
--
Paul Nicholson
--