AW: EG design

[Haible Juergen]

	>Are you saying that they use, say a 15volt supply to charge the
timing
	>cap. in the attack phase then initiate the decay phase when the
cap pd
	>has reached, say, 5volts. This would appear to approximate a
linear
	>rise as the exponential curve has not significantly rounded
off.

Yes. You basically have an almost linear slope in the first part,
in any case. Then there is some rounding - how much rounding
will depend on the speciffic EG design (V_asymptode / V_peak).

What does this mean to *level* in combination with a linear VCA?
Say your VCA has 70dB of SNR. You'll have a linear increase
of signal voltage from -70dB to maybe -6dB during less than one 
half of your total attack time. The second half (or a little more, if
you have more rounding) is used to go from -6dB to 0dB.

This behaviour is very good to give a sound some directness -
most natural instruments have contours like this. (You surely
know about the trick of making fast attack / fast decay envelopes
more "punchy" by holding the peak level for a few ms before
the decay begins ...)

An exponential VCA would produce an almost exponentially rising
voltage (= linear rising level). This might have its uses as well, if
you want to model some sound source coming near to the listener
from a distance. Maybe. But otherwise it's not of much use.
I am not sure (but almost), that some manufacturers of d*g*t*l
multisegment envelopes fell into this trap. You end up using
3 segments to do a piecewise linear approximation of a log
function, in order to compensate for the exponential DCA
response ...) 

	> But
	>what about AD envelopes, eg. Moog Rogue, these sound OK and
they charge
	>the cap.to full pd.

I don't know at the moment - don't have schematics here. I think the
"cheapo Moogs" used dual 555 timer IC's, didn't they? So asymptodes
of Vs and peak voltages of 2/3 * Vs would be expected. You can change
these thresholds with external components, of course, but I am just
wondering.

BTW, if you consider the stuff about the level increase from above,
linear or nearly linear truncated RC-type contours only make a very
little difference. Everything is fine (for natural sounding stuff), as
long
as you
(1) don't have linear decay / release slopes
(2) don't use an exponential VCA.

	>What about VCF sweeping, is it worth having a linear CV input
for
	>exponential EGs?

I haven't thought much about envelope shapes on VCF modulation
yet. Good point !

Linear inputs always leave the choice if you *add* a linear 
modulation signal (but then it's effect gets weaker and weaker
when you increase the manual cutoff frequency, or keyboard tracking),
OR *scale* some other CV by your modulation source. Which ends
up needing an additional VCA. (Avoiding tons of VCAs with making
multiplications by adding logarithms is the main reason why the
V/8ve system was so sucessful - think of it!)

	JH.