In article .invalid,
Adrian Tuddenham wrote:

That's what I was thinking, but why only on the left channel? Both
recorders behave the same way.


Spread in components, differences in layout causing parasitic caps an
inductances
to vary, take your pic.

Possibly there's some occasional switching noise on a digital or power
trace, which is coupled more effectively into the left-channel analog
trace going to the ADC? The left-channel trace might actually be
acting to shield the right-channel trace from some of the noise.

That could be the sort of thing that is happening - but I wonder why it
comes on after about a second's delay?

Demonstration at:
www.wrightshardware.co.uk/Other/DR0002_0276.wav
...beware loud clicks!

Looking at that file, I see three sets of noisy impulses... two
negative-going and one positive-going. They're present in both
channels - similar but not identical waveforms in the two channels.
There's a very sharp leading edge (rise time is two or three samples)
and then a much slower fall, with some ringing. The signal takes
about 40 milliseconds to fall back to zero, and then overshoots.

This doesn't look like a one-sample error on the digital-storage side
of things - it's not just a brief impulse. It appears to me that
something is spiking the input to the ADC pretty hard, and that this
glitch is getting into both ADC inputs. The fact that the pulse and
ringing is slightly different, between the two channels, suggests that
it's not just a single sample of bad data in the ADC being blurred out
by the ADC's filters.

Is this .WAV exactly as it came out of the recorder, or did you adjust
it (e.g. gain-boost for visibility) in any way?