Thoughts on Software Generated Envelopes

[Eli Brandt]

> can someone hear the difference between a 0.1 msec attack and a .3
> msec attack?

One way to think about this might be to consider a train of gaussian
pulses, of varying width.  As you narrow the pulses, the spectrum will
become pretty much flat over the audible range.  At that point, you've
got all the "snap" you're going to get.

This is pessimistic in that you don't usually set both attack and
decay to the minimum, so the body of the sound normally covers up some
of the attack.  It's optimistic in that a non-gaussian will widen the
spectrum a little faster.

This may be bogus; I haven't really thought about it.  You probably
do need listening data...

--
   Eli Brandt
   eli+ at cs.cmu.edu