Stefan wrote:

> ...in less than a day and without stacking!

You made a good choice of symbol duration.  The error map is
at http://abelian.org/vlf/tmp/170525e.png

We might do as well or better with a daytime slot 05:00 to
21:00 to avoid the night noise and two terminators.

Markus wrote:

> I should have mentioned that SpecLab FFT recordings are
> usually delayed by a four sample latency.

That helps, I can see some correlation now. The message is at

 Carrier Eb/N0: -7.5 dB
 Carrier S/N: 22.09 dB in 12.6 uHz,
             -26.90 dB in 1Hz,
             -60.88 dB in 2.5kHz

I tried resampling to a higher rate but that didn't help.

--
Paul Nicholson
--