Minimum reasonable latency Was: Re: ZynAddSubFX was: Re: [linux-audio-dev] some new soundfiles on-line

Jens M Andreasen jens.andreasen at
Thu May 19 12:14:06 UTC 2005

On Thu, 2005-05-19 at 11:30 +0200, fons adriaensen wrote:

> Don't suggest that to a concert master - they tend to be sensitive ;-)

My mistake, it is of course the other way around. The band following the
lead from an acknowledged master of the arts. The lead will have to be
somewhat pragmatic though ...

Now, to get this thread back on the rails ...

There is a set of scenarioes regarding realtime latencies and
soundcards, each with their own requirements and pitfalls. For starters
we have:

  The Producer, recording real sounds to harddisk and overdubbing as
needed. Here the main obstacle is disk-io and unpredictable kernel

  The Guitarist, processing the output of an instrument in realtime. The
roundtrip from sample-in to sample-out is what bothers. Getting a
reliable RT priority saves the day.

  The Synthesist, calculating a sound from midi trigger input in
realtime. This is almost like "The Guitarist", but with only half of the
roundtrip buffer worries. 


  Combinations of the above like, say GigaSampler played from a keyboard
while pulling in samples from hd as needed ... or live radio pulling in
jingles/commercials at the end of that witty punchline processed with
the reverb from hell ...

In order to say what kind of buffer size is relevant or optimal, you'll
also have to specify what kind of scenario you are targetting. 

A "synthesist" (that's my scenario), who can't hear *anything* before
the end of the pipe, would be very sensitive to jitter (the deviation,
or "sloppyness", between triggers and actual sound) and would therefore
require the smallest of small buffers.


More information about the Linux-audio-dev mailing list