Re: [LAU] Latency and USB interfaces

From: Robin Gareus <robin@email-addr-hidden>
Date: Wed Jan 08 2014 - 19:35:46 EET

On 01/08/2014 05:54 PM, James Stone wrote:

[..]
> This discussion about the Scarlett 18i8 has got me thinking about
> latency with USB devices. Is there an accepted buffer size that USB
> audio interfaces should be able to use with jackd now? I do seem to be
> getting xruns with 128/2 (even occasionally with 256/2 on program
> startup/shutdown), but I have not made any particular effort to try to
> tune my system for low-latency use.

try 128/3 or 96/2 .. See the excellent recent explanation from Clemens
on that subject:
http://lists.linuxaudio.org/pipermail/linux-audio-user/2013-December/095401.html

[..]

> Also, how meaningful is the reported jackd buffer size in terms of
> actual latency?

jack buffer size is only meaningful within jack.

Additional latency introduced by hardware depends on the hardware.
As with any complex system where you cannot precisely know what is
happening: measure it, with jack_delay (or it's son jack_iodelay which
is included with jackd these days).

Results for some USB devices:
 http://robin.linuxaudio.org/tmp/vsl1818latency.png
 http://wiki.linuxaudio.org/_detail/wiki/jack_latency_tests/latency-ua25-corrected.png?id=wiki%3Ajack_latency_tests
created with http://gareus.org/gitweb/?p=latentor.git

You'll find some patterns and there usually is a relationship between
the jack-buffersize and the actual round-trip latency. But it's
different for different hardware (and to some extend also
software/drivers,.. the whole stack)

ciao,
robin
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@email-addr-hidden
http://lists.linuxaudio.org/listinfo/linux-audio-user
Received on Wed Jan 8 20:15:18 2014

This archive was generated by hypermail 2.1.8 : Wed Jan 08 2014 - 20:15:18 EET