Simple Envelope Extractor

[rockgardenlove]

What's the best envelope extractor I can build with simple components?  Say TL072 and 2n3906?

Thanks!

[Paul Perry (Frostwave)]

It depends how much filtering you need after it.
Incidentally, searching for a simple envelope follower brought up this: http://www.monzy.org/urinecontrol/
There is a simple envelope follower schem there, but a bit too fuzzy to read...

WARNING: page not suitable for adults!

[rockgardenlove]

Hmm, well, I'm just trying to get a very generic follower that I can use for all sorts of stuff.  Auto wah, phase shifting, volume modulation, etc.  I want to come up with some sort of modular setup that I can patch together with little 1/8" jacks to make crazy sounds. 

[StephenGiles]

Simple = lots of ripple = useless I'm afraid.

[markusw]

Considering that you probably will use the envelope follower for various things in your modular approach I'd suggest a good (unfortunately not simple) envelope follower.

http://www.edn.com/article/CA265499.html

This one is supposed to be a very good one with very low ripple....

Markus

[A.S.P.]

more less simple solutions

[gez]

It depends how much filtering you need after it.
Incidentally, searching for a simple envelope follower brought up this: http://www.monzy.org/urinecontrol/
There is a simple envelope follower schem there, but a bit too fuzzy to read...

WARNING: page not suitable for adults!

Guy next to you at urinal frantically fiddling with genitals: 'It's not what you think, it's just my joystick!'

[Paul Perry (Frostwave)]

On the other hand, most wahs use only a half wave rectifier! not even an 'ideal' one at that (see Dr Q and all the derivatives).
One problem I had when designing my own commercial envelope filter (the funk.a.duck) was, that if you apply 'enough' filtering of the envelope then you lose the attack.
I guess if one samples the envelope for a short time & stores the maximum value and outputs it while repeating the process, you would get a pretty sharp low pass filter.
Personally, I was suprised how 'rough' an envelope follower could be and still be very musically acceptable.
Really, polishing up the envelope rapidly turns into a diminishing returns experience, at least for a stompbox.
My bottom line: full wave rectifier.. anything that looks like the first circuit here: http://www.physics.udel.edu/~nowak/phys645/More_opamp_circuits.htm

[StephenGiles]

more less simple solutions

Daft as it must seem, I never got this one to work. Of course it may just have been duff breadboard connections! I have had a project on hold until lighter evenings which we now occasionally have, but I must get back to that Harry Bissell circuit this summer, whenever it arrives.

[Mark Hammer]

Complexity will depend on what part of the envelope or degree of responsiveness is most important to you.  Half-wave rectifiers (HWR) work just fine if you set the time constants such that they don't reveal the ripple inherent in HWR.  That, in turn, will depend on your playing style.  Ripple shows itself quite plainly if one is playing lower notes and likes to let chords ring for a while.

At  the same time, although you can reduce audible ripple considerably by forcing a slow decay with a larger value cap, you deprive yourself of the opportunity to have smooth short decays.

All things considered, a FWR is not orders of magnitude more complex.  Build it.

[moosapotamus]

What about trying the Anderton envelope follower in EPFM? I've built a number of the other projects in EPFM. But, I haven't built the envelope follower, so I don't personally know how well it performs. But, it looks realtively simple and pretty flexible, too. The other EPFM projects I have built came out great.

~ Charlie

[Mark Hammer]

I've got one of those sitting around for use as an experiment tool.  It's just another HWR, except that by using an LDR with a slightly sluggish fallback some of the envelope ripple that would normally show up during the decay of strummed chords is "smudged" so that it is much less audible.

[Paul Perry (Frostwave)]

FWIW, I once built a fairly complex low pass filter (set at about 10hz) in the hope of filtering out the ripple but leaving the useful part of the envelope. Wasn't any practical advantage.  (again, this was for a filter. Maybe for another application....)

[R.G.]

A reasonably simple envelope extractor uses double full wave rectification.

You make a precision full wave rectifier from a dual opamp. Now make another one. Feed the output of the first FWR to the the input of the second FWR without filtering it to DC, and through a series capacitor to remove the DC it does have. Filter to DC after the second one. The lowest frequency in the result is now four times the lowest frequency in the input signal, and filtering time constants can be much smaller, giving lower ripple and better tracking.

It's not as complicated as the Bissell envelope generator.

[Mark Hammer]

The general goal of solid envelope-follower design is to maximize responsiveness while minimizing ripple.  The resulting rectified envelope voltage should be as close to smoothly-changing DC as you can get.  When one thinks of "smoothing", one inevitably turns to lowpass filtering, and virtually all higher quality envelope extractors have some lowpass filtering built in in some fashion.

One thing I've always wondered about is the possible profitable use of highpass filtering in attaining this goal.  I understand the pros and cons and rationales for using lowpass filtering, but at least some of the ripple that occurs is in the subsonic-to-audio zone.  For instance, the sorts of beats that can occur from adjacent notes or from strings that generate their own sideband products when the string gets deformed in some way.  Most envelope followers I've seen for guitars or general-purpose seem to have largely unrestricted bandwidth at the low end, and all filtering appears to be tailored towards keeping the "chatter" of the envelope signal to a minimum through lowpass filtering.  I'm wondering if higher envelope-follower performance might not be achievable through the use of, say, a 2-pole (or greater) highpass filter on the input, prior to rectification so as to weed out the under 60hz warble, in tandem with whatever lowpass filtering is needed, post-rectification.

Aside from the obvious concerns about use of such a technique for bass, and differential sensitivity to the lowest notes, is there any problem in such an approach that I failed to notice?  Has anybody ever tried this out?

[A.S.P.]

A reasonably simple envelope extractor uses double full wave rectification.

You make a precision full wave rectifier from a dual opamp. Now make another one. Feed the output of the first FWR to the the input of the second FWR without filtering it to DC, and through a series capacitor to remove the DC it does have. Filter to DC after the second one. The lowest frequency in the result is now four times the lowest frequency in the input signal, and filtering time constants can be much smaller, giving lower ripple and better tracking.

It's not as complicated as the Bissell envelope generator.

That frequency-quadrupling is the principle in the "Elektor 2x NE571"-extractor/follower that Paul "Frostwave" Perry once pointed out, too;

"THAT" signal processors seem to go the way which "puretube" thought to have (re-)invented a while ago: create a Hilbert/Dome-shifted additional 90° signal;

both reduce the "ripple-factor".

Both are not what "rockgardenlove" was looking for.

[StephenGiles]

I still think this is the way to go
http://img.photobucket.com/albums/v517/uncle_boko/msynthfollowermod.gif

[A.S.P.]

to eliminate the ripple, you need to get rid of the "to zero" parts of the signal... 

[puretube]

This was a nice idea...
more over there, too...
not simple either,
but my new attempt:
"Where is the most ripple?" somewhere in the low frequency signals...
So: just shift the complete signal frequency range in the rectifier-side-chain (be it "2-wayed" or "Domed", or cascaded afterwards...)
up a (few) hundred Hz,
and such lift the lowest "to be detected" frequency above the danger-zone,
where it can be smoothed without lagging side-effects...
(no need to go into RF territory).