Having decoded the 20:00 UT 22 character message, I
re-encoded it and used the symbols to phase modulate the
received signal.   The result is an unmodulated carrier
which I can take measurements from.

Here's the reconstructed carrier spectrum:

 http://abelian.org/tmp/lf151024a.gif

The peak is at 137502.239997 Hz +/- 1uHz.

Phase measured in 10 intervals was

 interval_start      degrees_lead
 2015-10-23_20:00:00 159.9
 2015-10-23_20:02:12 158.3
 2015-10-23_20:04:25 157.9
 2015-10-23_20:06:38 152.0
 2015-10-23_20:08:51 102.4
 2015-10-23_20:11:04 153.5
 2015-10-23_20:13:16 154.8
 2015-10-23_20:15:29 167.1
 2015-10-23_20:17:42 169.1
 2015-10-23_20:19:55 136.0

Then I repeated the process, but for the undecoded 23:00 message
(knowing from Markus' decode that the message was 'IZ7SLZ')

But I get nothing above noise.  For the other hourly
transmissions 00:00 to 06:00 I did the same, but nothing
shows above noise, not even a hint of anything there.

The noise floor didn't change much overnight.

I'd expect to see a reconstructed carrier about 10dB above
noise if the power reduction was 10dB.   I'm guessing that
the 10W tx power is less than 10W, probably less than 5W.

That's one possibility.   Alternatively, the tx oscillator
lost its GPS lock and the carrier began drifting.

The carrier 20:35-20:43 had an offset of 1,338 mHz which
suggests the tx oscillator became unlocked.

We need to measure the phase of some long distance carriers
to see if the LF band can use very narrow bandwidth coherent
BPSK.   So far, we know it is good for 20 minutes at range
2022 km.

--
Paul Nicholson
--