MSP and Laptop Audio

[cstecker@ovenguard.com (Chris Stecker)]

Wow.  I didn't expect quite this thread on MSP here, although I can
understand it.  As far as audio quality, I have to agree that laptops leave
much to be desired.  I've got a Powerbook 520 ("CD-Quality Audio") and a
Quadra 950 (plain old mac audio), and the quadra sounds much better, even
just off the board.  Add a sound card, and it actually starts to sound
good.

For MSP, there are also weird timing and interrupt(?) issues with several
of the powerbook models, including the 5300 and 2400, resulting in very
slow operation.  None of the desktop models are affected in this way.

Latency depends upon the audio interface used.  The shortest, around 23 ms,
is achieved using the sound manager, and audio from the CPU.  Some sound
cards can achieve this as well (PT Direct I/O, Korg 1212), but most are
worse, and some are a lot worse.  There is also a fundamental low-end limit
to latency, based on the sample buffer size (number of samples processed
and output as a chunk).   This is adjustable down to 64 samples (~1.4 ms)
in MSP, though not all sound cards can go so low.  This is probably one
place where dedicated hardware will show a real advantage.  I assume
something like the EDP could process single samples (which allows control
changes to happen at any time during the data stream).  I could be wrong,
though.  Am I?  It seems software devices, where you have to deal with an
entire general-purpose computer, will always suffer from this overhead, but
higher clock speeds, and possible higher sampling rates might help (64
samples at 192 kHz is about 0.3 ms).

For more traditional music performance, these latency issues can be
serious, especially where they interfere with audio-rate interactions
between devices.   However, my EDP is usually set to loop between 4 and 40
seconds.  Audio in-out latency of 20ms or so should hardly matter at all.
Setting loop points by button-press, as in the EDP, may be made difficult
due to control latency (time from button-press to change in audio).  I'm
not sure if the system would work like this or not, but adding 20ms to a
loop would mean desynchronization of a full second after about 50
repetitions, which wouldn't be good.  Loop points could still be set by
calculating loop lengths, with graphical controls, etc.


-Chris