AW: EG design

[Haible Juergen]

	Dear Grant,

	> I am not following you reasoning here. 

I don't think my thoughts are that much different from yours, but
maybe I was not so clear in my arguments.

	>If you go by textbook acoustic
	>theory, volume (the psychoacoustic perception of amplitude) is
perceived
	>exponentially. 

Absolutely right. That's why I said an exponentially increasing 
voltage could be of some use when you model some sound source
that moves in from a distance.

But I don't think the attack phase of many instruments behaves
like this. A bowed string, a trumpet, a flute - do they really build
up their amplitude exponentially? 

	>Just as pitch (the psychoacoustic perception of frequency)
	>is perceived exponentially. I can prove this to my satisfaction
by feeding
	>a manually adjustable linear control voltage to a linear VCA.
All volume
	>change is bunched up at the top end and the first 70% of
control travel has
	>no apparent response. If I switch the VCA to exponential
response, then the
	>control action is spread out over the whole control travel.

I totally agree here, too. For *manual* control over the whole range
you want exponential response. (Or a "logarithmic" - misleading 
word! - potentiometer.)

	> With the caveat that the attack portion is an inverted
exponential
	> response. This sound just fine, lots of Minimoog owners will
tell you so!

This was exactly my point. It just sounds fine, and you get it for
free with a common RC-shaped envelope and a linear VCA.
But it would be hard to get this behaviour with an exponential VCA.
This was exactly my point.

	> In the case of a linear envelope (which is a more generally
useful
	> controller for all parameters) ...

Agreed for general use like panning, crossfading etc.

>  ... the exponentiation need to be supplied by
> the VCA. 
> 
But (sorry I repeat myself) during the attack phase the VCA
would do exactly the opposite of what it should do (if we have
the "Minimoog behaviour" in mind)..

	>You can simulate both exponential and inverse exponential
envelopes from a
	>linear EV by feeding the rate control input with either
positive or
	>negative polarity from the output. The Wiard envelator has this
patch
	>normalized to the rate control inputs. All you have to do to
simulate a
	>Moog EV is turn up the modulator inputs.

Surely a very good solution to have all the choices in one module.
No doubt about this, and no doubt that your Wiard Envelator
will be a great module ! (saw it at your page - impressive, just
as most of your stuff!)

No, I really don't have anything against universal modules like
envelopes and VCAs that can be switched from linear to exponential.
(Or VCAs which have linear and expo CV inputs available at the same
time.)
My caveat was about the combination (pure) linear envelope +
(pure) exponential VCA. You can't expect to get the beloved
"Minimoog behaviour" out of this.

	> I don't want to be dogmatic - if it sounds good, throw out the
book. 

I don't want to be either. And I haven't actually done research
how the volume of real instruments really builts up. Surely
has to do with force vs. friction, so an initially exponential law
would be damped into something much more linear. Would
be interesting to get deeper into this aspect, for sure!

JH.