Detect envelope from PWM signal?

[earthtonesaudio]

I have a project (still in the idea stage) in which I'd like to detect the envelope of a signal after it's been pulse-width modulated.  Of course it would be easy for half-wave (diode, resistor, cap), but I'd rather do full-wave rectification.

Essentially for zero input (50% duty cycle) I'd want the low control voltage, and as the duty cycle moves away from 50% (in either direction) I want the control voltage to rise.  I feel like there should be a way of doing this with just a few simple logic gates...

[earthtonesaudio]

I think I figured out how to do this:
Start with interleaved PWM signals, then XOR each one with a corresponding square-wave derived from a "zero signal" voltage.  Then OR those two outputs, and the result would be a duty cycle that increases whenever the audio signal's absolute value increases.
...That's a lot of work for a rather small gain in performance, so I think for now I'll stick with the quick & dirty half-wave envelope detector.

[earthtonesaudio]

Another idea:



The PWM section is not ideal, because the frequency varies along with the duty cycle.  However this quirk can be exploited by converting the signal to a variable frequency square wave, and then feeding a frequency-to-voltage converter.  The F-V puts out a small signal with zero signal input to the buffer, but when the signal swings above OR below the quiescent point, the F-V puts out a larger signal, so you get full-wave rectification.  The F-V is over-simplified and probably incorrect but I'm just trying to convey the general idea.

[Jaicen_solo]

Awesome! but ermm, what's it for..? 

[Cliff Schecht]

That's super duper secret!!

[earthtonesaudio]

That's super duper secret!!

Awesome! but ermm, what's it for..? 

Cliff is right, because... even I don't know yet!

This latest incarnation started as an attempt at another Crash Sync/Uglyface type of noisemaker, using a CMOS chip with multiple Schmitt trigger gates (4093 or 40106)... then I got a bit carried away with the possibility of putting envelope control over ...something.  Couple that with my dislike of optocouplers and a pie-in-the-sky notion of doing as much as possible in the quasi-digital/switching domain (to save power compared to biasing things for linear operation).


Interestingly enough, I kept getting rather clean audio out and I have no idea why, so it's been tabled for now.