Re: [Jack-Devel] Jack on ARM with low-latency USB

PrevNext  Index
DateThu, 28 Nov 2013 22:25:53 -0800
From Gabriel M. Beddingfield <[hidden] at gmail dot com>
ToFabian Schlieper <[hidden] at fabi dot me>, [hidden] at jackaudio dot org
In-Reply-ToFabian Schlieper [Jack-Devel] Jack on ARM with low-latency USB
On 11/21/2013 08:17 AM, Fabian Schlieper wrote:
> Hi everyone,
>
> I'm trying to build a low-latency audio processor on ODROID-X2 with an
> external USB sound card (Behringer UCA-202).
> I can't get jack working stable with a period size of 128 samples.
[...snip...]
>
> It's running without audio glitches, but after some time (or when other USB
> devices are heavily active) the sound gets disturbed and keeps crackling until
> jack is restarted.
>
> I know I could increase the period size, but I want to have low-latency, Is
> 128 period size even realistic with USB sound?

Don't know if this relates, but...  On the OMAP4 SoC I've seen similar 
problems with the EHCI controller with respect to audio reliability.  In 
that set-up, the CPU has to respond to very frequent interrupts in order 
for the controller to work properly.  This often lead to XRUNS.

But at this latency, it could be *anything* causing your glitches.  Any 
driver holding a spin lock while doing some "background" task will do 
it.  I recommend tracing with `trace-cmd'[1] and inspecting them with 
`kernelshark'.  If you can get a `trace-cmd' capture that includes the 
XRUN's you'll get some clues about what is happening.

-gabe

[1] At the very least, do:
     # trace-cmd record -e "sched:*" -e "irq:*"
PrevNext  Index

1385706375.26050_0.ltw:2, <52983371.1040101 at gmail dot com>