David G Dixon dixon at mail.ubc.ca
Sun Jun 29 00:13:49 CEST 2014

```> To put it another way, the decay phase is clearly defined
> since it decays exponentially from the end of the attack
> level down to wherever the sustain level is set to.
> Similarly the release phase decays exponentially from
> whatever the sustain level was set to down to zero or
> silence.  Both the decay and release phases have defined
> start and end points and defined time-constants, but the
> end-point for the attack phase seems less well defined.  For
> example, it might typically consist of charging a capacitor
> from 0V towards 10V with the classic exponential charging
> waveform, but it theoretically takes forever to get to 10V.
> So at some point we have to say "that's enough waiting" and
> switch over to the decay phase of the envelope.

No, it doesn't take forever.  10V is the threshold for the comparator that
switches from A to D, but the "target" for the attack phase is actually the
rail (+15V).

> Various analogue synths seem to set this comparator threshold
> at different levels.  Set the threshold too low and terminate
> the attack phase after less than a time-constant of the
> exponential growth and the attack shape starts to tend
> towards a linear attack.  Conversely set the threshold too
> high and there is going to be very little perceived variation
> in volume towards the end of the attack phase, particularly
> as we perceive volume logarithmically!

The threshold can be adjusted to change the curve shape, but for
"snappiness" one wants a lower threshold.  I have found 10V (with a 15V
rail) to be pretty snappy.

> To me as an engineer, it would seem sensible to set the
> threshold level for switching to the decay phase at 63% of
> the level that the capacitor is charging towards.  So for the
> 0-10V envelope case above, the capacitor would charge from 0
> to 6.3V during the attack phase before moving into the decay
> phase.  That way the envelope would linger in the attack
> phase for exactly one RC time-constant at whatever the attack
> time-constant is set to.

Yes.  In eurorack, with its 12V rail, the typical threshold is 8V.

> Is there any convention for this, or has anyone even given
> the implications of this any thought!?!?

2/3 is very close to 63%, so 10V for a 15V rail, and 8V for a 12V rail, are
good choices.  That's what I've always done.

```