[sdiy] VC ADSR

[UrosM]

hello ppl

> This is something I've been brainstormin on lately :
> I decided to use a uC and LUT to generate log / lin slopes,
> and even parabolic & other exotic shapes...
> 
> I've even considered VC-morphing between those slopes.
> This is feasable via software, but the downside is the number
> & type of controls to make it easy to use : the front panel of each
> VCADSR would feature more pots than the rest of the modular...

VC morphing is much more feasable if your doin' it completly
analog . Simple brute force solution : lets asume control inputs
of vcadsr accept 0-5V CV and that you have 0-5V morph CV .
Make one unity gain inverter with opamp and feed morph CV to it .
Note output of this opamp as CV1 . Make inverting unity gain
summing amp with two inputs . feed one input with CV1 and
other input with consatant 5 volts . Note output of summing amp
as CV2 . Make inverting unity gain summing amp with two inputs 
and use lets say 1k resistors for it . Take two 1M pots and 
conect one between one summing input and CV1 and other
between other summing input and CV2 ( ! pots are used
here as variable resistors not variable voltage dividers ,
therefor one leg of pot is left unconnected ) .
These two pots will give you Attack1 and Attack2 values .
Make same pots+summing amp arangements for decay , sustain
and release . Feed outputs of summing amps to coresponding
VC inputs of ADSR . Now you can with morph CV make 
continous changes between two sets of values for A,D,S and R .

As for parabolic , you only need linear AD envelope generator
and simple sine shaper as one from VCOs ( cos uncle Fourier
will tell you that a-x*x is very very very close to cosx
for segment -Pi/2<x<Pi/2 ). And better yet , you could
normaly change A time and D time and thus make
assymetrical parabolic envelope . Or you can feed
complete adsr to sine shaper or you can feed env or lfo
to full wave rect. or trough DC coupled state variable
filter . This is what modular analogs are all about .
Modelling nonlinearity in analog domain by actually
solving some math equation by circuit is very wrong aproach .

And finaly , i think there should be no problem at all
with switchable lin/log envelope . If we look at env generator
it can be simplified as follows : you feed some sort of gate 
signal to either first order LPF and you get log ( or 
should I say "log" ; why the hell people call this
log when all segments are desribed with expo function )
or feed it to integrator and you get linear segments .
So we need something that acts as VConepoleLPF or 
(one switch away) VCintegrator . We actually do have
such circuit . Take a look at first order cell in
SSM2040 . It is integrator with unity gain feedback around
it wich turn it into first order low pass . You just
need switch that cuts feedback path and you have
core element for lin/log vcadsr .

regards
urosh 
(have done complex integrations all morning on horrible math exam)