[sdiy] Soft ADSR ways?

[Magnus Danielson]

On 14/02/11 00:30, David Moylan wrote:
> Isn't it possible to use a difference equation?  For instance you figure
> out the difference between the current envelope value and the asymptote
> value, the divide (preferable by a power of 2, so you can just bit
> shift) then add that back to the envelope val?

For linear slopes the state y progresses from input x with

y(n+1) = y(n) + x*b

where x is the target level, b is a speed factor and y is the 
instantaneous value.

Similarly for an exponential slope

y(n+1) = y(n) * (1-a) + x*a

which is better described as

y(n+1) = y(n) + (x-y(n))*a

since it becomes evident that the difference between current value y(n) 
and target valye x steers the rate of change, and not only the a factor.

A combined variant could be

y(n+1) = y(n) + (x-y(n))*a + x*b

per sample:
2 additions
1 subtraction
2 multiplications

Vary a and b according to taste.

About the same thing can be done in analogue.

Cheers,
Magnus