[sdiy] Mono vs Poly

[Richard Wentk]

At 10:48 10/07/2003 +0200, jhaible wrote:
> > How many polyphonic flutes do you know?
>
>Pipe organs ?

Er, no. It should be obvious why. :-)

Paul is right. Flute vs Organ is just an early case of non-keyboard vs 
keyboard instrument.

Synth culture to date has been based on the creation of whiz-bang sounds 
which mostly are shaped automatically with envelopes and LFOs. Polyphony is 
really just a substitute for high levels of manual expression. Triggering 
lots of evolving sounds with varying time-offsets creates the illusion of 
complexity and control where neither really exists. (Well, that and you can 
play chords too, but that's just by the by. :-) )

If you take the time to list in detail the 'expression channels' - i.e. all 
the different things you can do to change the sound - on any non-keyboard 
instrument, you end up with a huge list. The fact that musicians don't care 
about *exact* reproducibility suggests that it's the existence of these 
channels that makes these instruments so interesting to listen to. And so 
hard to play well.

But it's not just performance. The timbral changes you can create with 
these controls are far more complex than those you can get from an analogue 
synth. Acoustic sounds - even simple acoustic sounds like a flute - are 
still far more complex and playable than anything one of today's standard 
synths can do.

And there's also the ensemble effect. Real instruments in a group resonate 
with each other and interact acoustically with the space they're in.

Finally there's also the physical production issue. Any speaker (except for 
exotica like Quad ESLs) is, to a good approximation, one or more 
decorrelated point sources in a box. The fact that you're dealing with 
boxes and point sources means you're already a long way from something like 
a vibrating string, which is more like a ribbon source. Any acoustic 
instrument produces a far more complex collection of wavefronts than 
something coming out of a speaker. [1]

Anyway... the point of all this is the 'quality' of a performance depends 
to some extent on information density, in the sense of being able to hear 
that something just changed. The more changes are happening the more likely 
it is that listeners will find a performance interesting.

There's a point where you can go too far and create changes that jar, so 
what seems to work best are performances that add ornamentation and extra 
richness to a simpler but well-defined and clearly audible structure. I 
think this is why processes like chorus and echo work so well, and why 
they're so good at taking a simpler sound and making it richer. It's the 
acoustic equivalent of adding wood veneer to chipboard. It may not be as 
solid as real wood, but it kinda sorta looks similar enough to fool people 
enough that they'll buy it.

So... the point of the exercise isn't just to add extra channels of 
performance information, but to add channels and synthesis techniques that 
have the same kind of wide-ranging flexibility. No, you won't be able to 
get that same level of detail and play them polyphonically, but for me the 
polyphonic problem has been solved. Polysynths don't seem to be evolving 
much, and that's because they have nowhere to evolve to.

But monosynths have the potential to go much further than they have. The 
key (ho ho...) is to ditch the idea of automated performance using LFOs and 
ADSRs and replace the keyboard and mod wheels with more interesting, 
playable and responsive interfaces. And also synthesis techniques which use 
more organic approaches to sound production, especially when played in 
ensembles.

This means forgetting the idea that monosynths are the poor relations of 
polysynths. Potentially, they're not at all.

Richard

[1]. Speakers actually aren't bad when they're reproducing a single 
monophonic source. Once you start trying to play any kind of ensemble 
through them they immediately lose the definition that acoustic instruments 
have in a physical space. An interesting experiment is to play a recording 
of a string quartet using four discrete channels feeding four separate 
speakers vs a stereo mix. Once you've tried the former you'll wonder how 
you could ever listen to the latter again.