A plot of the daytime correlation of I/Q against the
known encoded message

 http://abelian.org/vlf/tmp/151204a.gif

shows significant spikes with each half-hour message,
up to 14:00.   Only the first two decoded:

 08:30  Eb/N0 =  2.7 dB
 09:00  Eb/N0 =  3.0 dB

Judging from the size of the spikes the remaining daytime
messages were all at least 6dB weaker, that's 3dB or more
too weak to decode.

My battery could be good for another night.  No point
in repeating the same tests.   Maybe try to max out the
capacity with 0.2 second symbols?  Here I might get a 50%
decode rate with that setting.

Or go for some DX with long symbols if there are listeners
further away?   One hour messages perhaps instead of
half-hour?   So far all LF tests have been less than 30 minutes
I think.  One hour messages with an undisciplined oscillator
will make the phase search have to work harder.

--
Paul Nicholson
--