A decode from a near field test at VLF:

 Rank 6938 Eb/N0 -0.4 dB 8K19A 60 chars

The interesting thing is, the symbol period is 6.25 milliseconds
(3152 symbols for 19.7 second total duration) and the message
was sent by magnetic coupling between two undisciplined
soundcards.  Received against the natural VLF background.
The message was also 'unannounced'.

EbNaut here is sending a coded prefix.  The decoder monitors
a band and searches for prefixes, then measures the frequency
and timing accurately enough to decode the following message
coherently.

It relies on the undisciplined soundcard clock or RX LO being
stable over the message duration, say up to a minute or so.

The overhead of the prefix is at the moment 1.25dB, hoping to
bring that down a bit.  A delicate matter of balancing the
prefix energy and message energy.  No point in a detectable
prefix if the message wont decode, or vice versa.

In principle then, with a small Eb/N0 overhead, the decoder can
take a live untimed I/Q stream from SpecLab or your favourite
SDR and monitor a band a few Hz wide. The only thing you have
to know in advance is the symbol rate.  Coding in the prefix
supplies the rest.

No good of course for slow narrow band, but there might be
applications for short duration wide band messages coherently
detected and decoded with signals a dB or so from the limit.

--
Paul Nicholson
--