[sdiy] Sequencer interrupt latency?

[WeAreAs1 at aol.com]

In a message dated 4/12/04 10:04:54 PM, thescum at surfree.com writes:

<<  If you're used to acoustic instruments 
with near-instantaneous event-to-sound translation, then you might be 
more attuned to noticing when an electronic instrument takes a little 
while to start playing a note.

But even with acoustic instruments, you're dealing with some acoustic 
delay between the source and your head, of roughly 1 millisecond per 
foot.  I never felt that the acoustic delay from standing 25 feet 
away from a guitar amp was much of a problem... >>

Most musicians don't have the ability to physically play their instrument as 
in-time as even the laziest, sloppiest electronic triggering device.  That is 
to say, when you record them playing, for instance, straight eighth notes or 
sixteenth notes, and later analyze the timing of their notes relative to a 
metronome reference, you will see a very wide amount of timing variance from note 
to note.  It's very rare that players actually play within 50 milliseconds of 
the metronome that thy're listening to.  This is true even if the stuff they 
played sounds perfectly in time and "in the pocket".  That's one of the cool 
things about human feel.  (BTW, these observations come from years and years of 
experience dealing with this stuff in the recording studio.)

On the other hand, latency of the sound coming from the instrument that you 
are playing with your own hands is usually very easy to detect.  That's one of 
the reasons I don't have any personal interest in using any of the current 
crop of plug-in-based virtual instruments.  Even with the best of them, on the 
fastest of computers, I still feel physically disconnected with the sound that I 
am playing.  Some MIDI modules feel this way, too, but usually not as bad.

I used to work as a tech for Jim Cooper (J.L. Cooper Electronics), who 
designed the Oberheim OBxa (and the SEM, and the Four-Voice, and the OBx, and the 
Maestro Phaser, etc.).  He told me that the OBxa, in the best case, had a 
minimum 50 millisecond latency for single notes that you played from its own 
keyboard.  This got worse, of course, if you played a big chord.  I'm guessing that 
this was probably also true of most other poly synths of that era (Prophet V, 
Jupiter 8, etc), most of which had similar processing horsepower.  Keep in 
mind, this had nothing to do with MIDI.  I think it is probably true for a lot of 
modern synths we use that MIDI propagation delay is insignificant compared to 
the internal delay of the tone generation process (that is, the delay that 
happens once the MIDI note has been received).  

I have found that when MIDI sequencing very note-heavy, heavily quantized, 
dense sequences on one multi-timbral sound module, the module will start to 
choke and stutter if it receives a lot of notes on exactly the same clock.  
However, if you edit your data so that the notes are all offset from each other by 
just one clock (I use Digital Performer, which has 480ppq clock resolution), 
then the module is able to breathe, and it plays the notes as if they all hit on 
the same beat.  It actually sounds tighter and feels better than if you 
quantize all the notes to the same clock.  This is not because of a MIDI data 
logjam, it's because the module's tone generator just can't get it together to 
respond to all those notes simultaneously.

The point of all this is:  If you can get your sequencer to fire off triggers 
within 10 milliseconds of your metronomic timing clock, then you've really 
got nothing to worry about.

Michael Bacich