[ardour-users] experience with soundcard over firewire

Mark Knecht markknecht at gmail.com
Thu Apr 13 12:03:41 PDT 2006

On Thu, 13 Apr 2006 11:42:37, Matthew Polashek <matt at tinysongs.com> wrote:
> When I bought my new laptop for interactive music I did a bunch of testing of the
> USB-Audio driver and latency.  It is really slow.  10 ms was the smallest I could
> get with pd and jack with the CCRMA kernel.  I ended up buying a Mac.  Core audio
> is pretty good with USB but still there is significant latency.  I just picked up
> an m-audio Firewire 18/14 and it's really great with latency.  It looks like
> several people have the m-audio stuff working on freebob.  For the most common
> and well supported interface, go for the RME multiface or digiface via PCMCIA.
> Matthew Polashek
> matt at tinysongs.com
> www.tinysongs.com
> _______________________________________________
> ardour-users mailing list
> ardour-users at lists.ardour.org
> http://lists.ardour.org/listinfo.cgi/ardour-users-ardour.org

Here are the contents of a recent FreeBob letter from Pieter on the
subject. I consider it early but generally positive.

- Mark

Pieter Palmers 	
to freebob-devel
	 More options	  Apr 10 (3 days ago)
Hi all,

I've been coding a latency measurement setup for the ieee1394 stack this
weekend, and I've gathererd quite some insight in the linux1394
operation wrt latency. I used 2 computers: one was a packet
sender/receiver and the other one bounced the received packets back to
the sender as soon as possible. Then some statistics were gathered, e.g.
round trip time.

It took me a while but I got pretty good results:
round trip times of about 2.5ms, being 10 ISO cycles. The processing of
delay of the bouncer is about .25ms.
All using userspace realtime threads and the stock 2.6.14 kernel
drivers. The downside of this is that CPU usage is horribly high (but
what did you expect).

Based upon my newly gained experience I changed parts of the streaming
code, and this allowed me to achieve about 300frames of roundtrip
latency at 48000Hz sample rate (qjacklam measured) for my quatafire 610.
This corresponds to 6.25ms roundtrip latency. For comparison: recently
there were some posts with measurements of the latency of a RME
multiface, and that was about 200frames (4.3ms). A Delta88 gets 188
frames. We are getting to competitive figures ;)

I'm not going to commit these changes to CVS yet, I'll incorporate them
when I can do a more thorough rewrite. They are more hacks than anything


More information about the Ardour-Users mailing list