Linus Torvalds e6a1991e7f Documentation: say which meter the reverb's cpu figures came from ...

They were taken with the audio load meter timing its idle spin against a
1MHz timer, where one tick is 4.8% of a sample period - so each of the
four figures is a rounding of that.  The meter counts cycles now and
re-measures the reverb at 22.1% against the 22.4% quoted, which is
inside the old instrument's resolution and changes nothing the section
says.

Not re-measured beyond that.  The before-and-after rows would mean
reflashing three superseded builds to move numbers by a few tenths, and
the point of those rows is the shape - that correctness cost 28.3% and
control-rating the LFOs gave most of it back - which is far larger than
the error.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

2026-08-11 12:27:11 -07:00

..

compressor.md

Spread [COMPRESSOR]'s Ratio across its knob instead of piling it at one end

2026-08-10 18:44:54 -07:00

klon.md

Documentation: what [KLON] actually measures like

2026-08-10 17:06:04 -07:00

reverb.md

Documentation: say which meter the reverb's cpu figures came from

2026-08-11 12:27:11 -07:00

signal-chain.md

Documentation: what the noise gate does, and what its knobs reach

2026-08-11 07:32:08 -07:00