In part 2 of this series on synth envelopes, we’ll look at some case studies – actual implementations of synth envelopes, to align them with the theory from part 1.
Instrumentation, Test Cases, and Devices under Test
I tried to make the instrumentation as simple as possible: I’ll be using patches that have a (white) noise source as the only signal source. Filters are removed/turned off or, if that’s not possible, completely “open” and unmodulated. The amplifier is modulated only by the amplitude envelope, and that’s the envelope we’ll be looking at. The resulting output will be recorded and analyzed qualitatively and, as far as possible, quantitatively.
The test cases are as follows:
- Curve Shape: Using the Attack stage, setting decay and sustain to zero, the shape is qualitatively analyzed for a sufficiently long Attack duration (around 1 second) and for the synth’s standard curve shape, if that can be set.
- Duration: With Decay and Sustain as before, he duration of the attack stage is measured for minimum, maximum (if there is a maximum, if not, for a reasonable value (say: 1 second). If the amplifier doesn’t reach full level for that setup, Sustain is set to full instead (typically relevant for the minimum Attack time).
- State Machine Behaviour – Gate off during Decay: For a sufficiently long Decay setting and minimum Release, the key is released during the decay stage. It is examined if the envelope then jumps to the Release part or simply completes the Decay stage.
- State Machine and Timing Behaviour – Gate on during Release: The synth is set to monophonic behaviour, and a sufficiently long Attack and Release time is used. During the Release stage, a new gate on trigger is applied. It is examined if the envelope starts the attack stage from zero level or the level in the Release stage, and how the duration of the Attack stage is affected (i.e. if it is shorter if not starting from 0 level).
- State Machine Behaviour – Gate on during Sustain: Sustain is set to a value considerably below full level, and the synth is set again to monophonic behaviour. Attack and Decay are set to medium values. During the Sustain stage, a new trigger is applied. It is examined if the envelope retriggers (i.e. runs through Attack and Decay) or just remains in the Sustain stage.
Finally, for our devices under test:
- Arturia Minibrute: a contemporary all-analogue synth. Meaning the envelope generators are analogue as well.
- Korg Wavestation KEX: from the digital era well before the advent of plugins. Which is also the only synth here from one of those big synth makers of old.
- Clavia Nord Modular G2: a great example of the “virtual analogue” variety. Here, the standard “ADSR” envelope is used.
- Steinberg HALion 5: the VSTi variety of a contemporary synth workstation. Again, the standard envelope (with standard settings) is used.
A classic ADSR envelope hardwired to the amplifier, with four (unlabeled) sliders. There’s also a switch to switch from “slow” to “fast”. That’s analogue!
The settings used for the first test case can be seen below
As we can see, the envelope goes up in a straight line. That’s what’s called “linear” behaviour.
At the Attack=Decay=0 setting, the level didn’t come up to maximum. Next issue: when using white noise (due to the fact that this synth’s noise isn’t white, rather has a considerable HF rolloff), it wasn’t possible to measure this. A repeated measurement using a square wave shows this to be around 0.9ms.
State Machine Behaviour – Gate off during Decay
As clearly visible, the envelope jumps to the Release stage for gate off during Decay.
State Machine and Timing Behaviour – Gate on during Release
In that scenario, the Attack portion is resumed from the current level (i.e. from the middle of the Release section, and does not reset to zero). The duration is shortened (as around 600-700ms vs 1000ms for the full rise shows).
State Machine Behaviour – Gate on during Sustain
The level jumps to considerably below the Sustain level, then the Attack section starts from that level. There seems to be limited reproducibility.
An interesting find is while we’ve identifed the shape as linear for the Attack portion before, this and the preceding test case show that both the Decay and Release section look more exponential.
Perhaps the most noteworthy finding is that gate on/off transitions always trigger a jump to the Attack and Release portions, respectively. There is, however, a noteworthy difference between the Release->Attack and Sustain->Attack transition triggered by a new gate on: while in the former case, the Attack section picks up from the level we were at, the Sustain->Attack transition sees a reduction in level.
This last case is also the one where the envelope acts not exactly reproducible.
Finally, the minimum duration of below 1ms allows for what people tend to call a “snappy” envelope.
And with that – look forward to the three remaining contestants!