Re: Free Looper coming with Tiger!

[Jeff Larson <Jeffrey.Larson@Sun.COM>]

Kim Flint wrote:
 > How accurate is the sync? I would imagine this area is very difficult
 > or impossible to do well for software running on top of a non-realtime
 > OS, in a non-realtime hardware system architecure. Has anybody tested
 > it seriously?

This isn't related to AULooper per-se, but I've seen this
"non-realtime OS" issue come up up a few times, and I would like
understand what people perceive the problem to be.

There are no realtime OS's in widespread use, at least not by
musicians.  Yet there are a rather large number of PC/Mac users that
seem to be accomplishing various forms of MIDI/audio sync acceptably.
Yes, there is latency.  Yes, there there will be a subtle quantization
in the processing of MIDI events.  No one argues that dedicated
hardware can do a better job at this, the question is whether software
can do it well enough.

Regarding sync among plugins running within the same host application,
it is relatively easy to maintain sync among them since they are all
ultimately fed from the same clock managed by the host, and they all
generally do their MIDI processing during the same audio interrupt.

Having the host be a slave to an external MIDI clock is admittedly
harder but it has been done many times.  I think it can be fairly said
that you will have more more detectable sync problems if you try to
run a piece of software as a sync slave in a predominantly hardware
rig, than you will in a predominantly software rig where the computer
is the sync master.

I'm sure there are those that can hear the difference between a
"jittery" clock generated by a non-realtime OS and a "rock solid"
clock generated by hardware.  But I can't help but be reminded of
audiophiles that can hear "dramatic differences" between brands of
speaker cable.  Is this a problem that most software-based musicians
can actually detect?

Jeff