It looks like the peak on Marlton's signal is spurious
and is occurring due to cross-talk in a 2-channel
vorbis link.

I've been doing some trials.   The cross-talk on large
signals is 70dB down or more, but not so the background noise.

The background noise (and anything buried in it) becomes quite
significantly correlated after passing through the variable
bit rate encoding.   Will test the constant bit rate mode, that
will hopefully behave differently.

This cross talk is enough to explain the peaks seen last
night.   Now I concentrate on analysing the raw data in
Mike's raw signal database, which is not contaminated by cross-
talk.  The recorded data is fine, so I start again from scratch.

--
Paul Nicholson
--