Let's try an experiment.

Capture a short sample 6 channels 96k/sec int2 format,
directly into a file.

No input, all inputs open circuit. Max alsamixer gain.

Hopefully capture where at least one channel is showing
a long 'event', eg 100mS where the noise drops.

Maybe it doesn't appear when there is no input tone.

If it does, we can measure the noise during the event
and compare with my Octo.  We can work out whether or not
the noise level during the event is the 'correct' one.

 vtcard -d hw:0,0 -b32 -c8 -r96000 -u -v test.vt,i2

-u turns off timestamping.   Just ctrl-C to stop it
after 20 seconds or so then look at it with vtsgram
to see if any of the channels has a good noise event.

--
Paul Nicholson
--