Final result

 time   Eb/N0     T-Offset  Phase
 00:00  no decode, tried very hard
 00:30  -0.8 dB    +1.8   -30  -60  -60  -90 (rank 35975 BER 36.8%)
 01:00  10.8 dB    +2.9     0    0   30   30
 01:30   8.4 dB    +4.2   -90  -90  -60  -60
 02:00   9.7 dB    +5.7   -30  -30    0    0
 02:30   8.0 dB    +6.7    180  180 -150 -150
 03:00  11.6 dB    +7.9    180  180 -150 -150
 03:30  12.4 dB    +9.1     0    0  30    30
 04:00  no decode

Most phase patterns show a slight increase in phase lead,
so the frequency is a bit higher than 137.777000 this time.

Wonder what happened to the signal before 01:00, it was very
weak, 00:30 was right on the limit.

Joe I hope you can fix the Arduino clock.  Is it losing RTC
interrupts or something?  Or just a slow xtal?  Do you have a
Raspberry Pi?  I can quickly put together a keyer program.
It would be fun to do a really weak signal test with longer
symbols but the clock will drift too much by the end of the
message.

Markus wrote:

> Actually, each station operator is using his own
> unique home-grown setup:

That's a wonderful summary, thanks.  W4DEX and other
VLF stations could also go on the list too.

Various mentions of previous experiments with coherent signals.
I've been following those up and found some interesting reading
and good results.  But coherent operation never caught on for
some reason.  Maybe too fiddly and expensive?   We're lucky
today because PCs are better and GPSDOs only cost a day's
wages not a month's.

I'm enjoying dabbling with this and it revives my interest
in amateur radio.  Hopefully we see a few more stations adding
themselves to Markus' list.

--
Paul Nicholson
--