[LAU] linux-rt & presonus 1818VSL

[Robin Gareus]

On 07/03/2013 10:09 AM, Jeremy Jongepier wrote:
> On 07/02/2013 08:27 PM, Robin Gareus wrote:
>> On 07/02/2013 07:49 PM, Harry van Haaren wrote:
>>> On Tue, Jul 2, 2013 at 5:53 PM, Fero Kiraly
>>> <fero.kiraly at gmail.com> wrote:
>>>> I have some troubles - read xruns - with this setup:
>>> There's a lot more to removing Xruns from a system than just
>>> the kernel and JACK settings (although they are very important
>>> :)
>> 
>> * linux 3.2.35-2 PREEMPT RT (debian's RT kernel recompiled with
>> VSL1818 clock selector fix added) * jackd 1.9.10 (git) * Rui's
>> rtirq script * jackd with -p64 -n2 (actually jackdbus)
> 
> Hi Robin,
> 
> So with and USB2.0 audio interface like the 181VSL using -n2 yields
> a stable setup too, no need anymore to use -n3?
> 

I don't know if that's USB-2.0 or something else.. - The same goes for
the statement that "USB devices should perform better if the sample
rate is divisible by the period size". I cannot find any evidence that
supports that statement.

As with all complex systems, I trust measurements more than theory:

  http://robin.linuxaudio.org/tmp/vsl1818latency.png

(x-axis are permutations of jackd's -p, -n, --sync parameters,
no x-runs with either configuration but it's just jackd + jack_delay,
no load).

[Surprisingly]? the latency increments are not quantized to
milliseconds (as the USB protocol implementation would suggest it should).