It doesn't have to be; what matters is that's it being scheduled into
a space that fragments the bandwidth the audio device needs.
Even if the audio device only needs 3% of the bandwidth and the mouse
.0001%, the mouse drops that USB frame into low-speed compat mode and
blocks all higher speed devices from using any bandwidth during that
timing period. If the audio device requires a tiny, uninterrupted
piece of scheduling every period, it's just lost any chance of
On top of this, the Linux host schedulers are not particularly
effective about avoiding these conflicts.
[sorry for some ambiguity above; I don't remember if 1.0/1.1 drops a
full frame to low speed or just the microframe].
On Tue, Dec 10, 2013 at 8:28 PM, Bob van der Poel wrote: