[sdiy] Best & fastest envelope follower schematic.. anybody ?

Tom Wiltshire tom at electricdruid.net
Thu Dec 16 19:28:18 CET 2021

Hiding under here somewhere is a deeper philosophical point about whether the abstraction we call the "envelope" of a signal actually exists or not.

We mostly assume it does because we explicitly create the envelope separately in our synths and then apply to a constant audio signal. But actually, synths are just about the only thing that works like that! (Theremin is the only other example I can come up with). For most instruments, the creation of the audio also creates the level of the audio, and there is no distinction between the signal and the envelope at all. The hammer on a piano creates both the audio signal and the envelope of that audio. Blowing the reed of a clarinet does the same thing. Plucking a string does the same thing. Hitting a drum, bell, gong, or vibraphone does the same thing.

Certainly, the process of multiplying a constant bipolar audio signal by a unipolar envelope and producing a shaped bipolar audio "ping" as a result is a one-way process. Information is lost, since we no longer know what both waveforms looked like, only what their product is. In its most basic form, if I give you the number "24", you don't know what two numbers I multiplied together to get that result. Might have been 3 and 8. Might have been 2 and 12...etc. That's where we are with the envelopes.

So, quite apart from the "needing to see the future" element of this being impossible, there's also the "it doesn't exist" and "not enough information even if it did" elements of this being impossible!!

Despite all this, even quite simple circuits do a usefully successful job. Have a look at the Mutron:


This uses IC3b as a half-wave precision rectifier, with R14, R15, and C9 to set attack time and decay time and provide basic RC filtering. It's hard to do it in many less parts, though other designs are similarly minimal.

> On 16 Dec 2021, at 17:35, Richie Burnett <rburnett at richieburnett.co.uk> wrote:
>> ... Like in a frequency shifter, you need to do phase shifters at each octave in your range, so it gets parts heavy fast. In the end, it’s probably not worth it. 
> Surely it's only the bottom few octaves that cause a problem by causing excessive ripple with their outrageously long periods. (Envelope ripple with short periods can be removed easily with a gentle filter that still has a moderately fast response time.) So does the dome filter or Hilbert transform really have to conform over the entire audio range?
> I must admit I have found the Hilbert transform useful for finding the envelope of an amplitude modulated sinusoidal carrier, but it doesn't necessarily give you the "envelope" you expect when fed with signals that are spectrally more complex.
> -Richie,
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at synth-diy.org
> http://synth-diy.org/mailman/listinfo/synth-diy
> Selling or trading? Use marketplace at synth-diy.org

More information about the Synth-diy mailing list