MXR envelope filter, circuit analysis

[markusw]

Hi all. I have just completed it and it sounds great (thanks to tonepad for the great layout and to Mark Hammer for "advertising" this filter here). By the way, for bass its really fine to have a second 1nF cap to tune the thing down one octave.

After browsing through this forum and after reading M. Hammers article on VCFs I now have a quite good idea how VCFs work. For the Mutron everything is (more or less) clear. But for the MXR unit I do not really understand how all those Hex inverters and the CMOS switches give this great sound. I can somehow imagine that on versus off time of the switches controls the resistance (and also the filter) but I really would like get it in detail. Is there a circuit analysis like for the Mutron?

Thanks in advance.

[Mark Hammer]

Thanks for the flattery.

There is no thorough analysis presently, although it works pretty much the way any other envelope-controlled filter would with only a few small differences.

1) There is an envelope follower (you can spot the traditional diode+cap confuguration with D2/C9 in the Tonepad layout) but it controls a high-frequency clock formed by IC2A/B/C/D rather than directly controlling some element like an FET, transistor, OTA, or LED.  The clock is constantly switching the CD4066 on and off at a constant ultra-sonic rate.  The envelope follower voltage varies the mark/space ratio such that the amount of "on" time, relative to "off" time changes with picking strength.  It is this feature that permits Steve Giles' clever and simple mod for reverse sweep by using the "leftover" invertor section.

2) There are two cascaded filter sections, formed by IC1e/f.  I'm still perplexed by just what sort of filter these make, since they look like they ought to be lowpass but sound like they are bandpass in many respects.  Conceivably this is due to some blending of straight signal in the output via R12.   Each switch section acts like a variable resistor in a manner sort of like the LDRs in a Mutron, except that these are not LDRs and they produce a changing resistance based on how much time they spend being maximum off resistance.  The range of the filter is partly determined by the cap in the feedback loop of the invertor.  For whatever reasons, Fransisco misunderstood and thought this only meant one invertor rather than both.  Consequently, earlier A-Gua layouts produce a less obvious range-shift by only doing so on one, rather than two filter sections.  This also has an impact on the effect of the resonance control.  He has, however, corrected that from what I understand, but I have not seen it yet.

3) The resonance/emphasis mod essentially applies variable feedback around the first filter stage.  I find this interesting since the two filter sections really are in series.  There IS an overall feedback loop through R12 from the second filter section (IC1f) to the input section (IC1c).  I wonder if it is possible to introduce emphasis for BOTH filter sections for super resonant filter sweeps or even to play with that loop.

4) Presumably,  IC1d fills the same general role that one of the op-amps plays in the Doctor Quack, providing high gain for the audio signal to yield a reasonable envelope signal.  I'm speculating that varying R11 would permit one to play with the sensitivity range a bit for those instances where pickups or playing style seem not to evoke much sweep.

5) R17 (1meg) sets the decay time of the envelope signal by providing a path for C9 to discharge.  Faster decay times should be feasible by reducing its value, and longer attack and decay times would be feasible by increasing the value of C9 itself.  On the other hand, one of the things that the EF does differently than other filters is that it extracts the envelope signal and then applies additional filtering to *that*., which may be a big part of what produces the (to my ears) extremely pleasing envelope control on this unit.  IC2E acts as a kind of lag generator permitting the envelope signal to be as fast as it is, or slower rise time than it is.  That seems to get around a lot of the envelope-ripple issues that arise when one doesn't use something like an LDR to smooth things out.  If you've ever A/B'd a Doctor Q (or clone) and an MXR-EF you will immediately notice how fairly ripple free the EF is, even at faster settings.

6)  As some build reports on Tonepad have suggested, R14 and R16 form a voltage divider that feeds a bias voltage in to more or less set the initial sweep point.  This could probably even provide fodder for a mod than included a tuning pot in addition to overall range-switching.  Based on build reports, though, you're probably looking at something like a 50k pot with fixed resistors on each end to accomplish the same overall current going to the invertors in the HF clock.

One of the things that has never been discussed, and I'm wondering if it is possible, is some sort of mod to the clock/switching circuit that would permit either:
a) footpedal control
b) LFO modulation as is found in the MXR/Dunlop Auto-Q
c) alternate envelope shapes like exponential rise

There.  That's probably what you were looking for, wasn't it?

[Transmogrifox]

2) There are two cascaded filter sections, formed by IC1e/f. I'm still perplexed by just what sort of filter these make, since they look like they ought to be lowpass but sound like they are bandpass in many respects.

This is just a state-variable biquadrature filter topology realized with CMOS inverters instead of opamps.  I spent a good deal of time trying to track down the schematic for reference and ran into a bunch of dead links and couldn't view it, however, if my memory serves me right, it looked like they took the low-pass output.  The bandpass output follows the first integrator stage, lowpass follows the second integrator stage, high pass precedes both integrator stages.

The high-pass output on the state-variable filter sounds very cool at a high resonance...it's a kind of a synth-wah sound.  I don't know if high Q can be accomplished with CMOS inverters or not--I'm not sure what the gain is, but I know the bandwidth and slew rate are generally fast, however, upping the Q with uncompensated amplifiers may cause the thing to become an oscillator.

If I remember right, the high pass output on the MXR EF is the output of the first CMOS inverter.

One of the things that has never been discussed, and I'm wondering if it is possible, is some sort of mod to the clock/switching circuit that would permit either:
a) footpedal control
b) LFO modulation as is found in the MXR/Dunlop Auto-Q
c) alternate envelope shapes like exponential rise

Certainly!
a) Just connect a DC voltage accross a pot and the wiper goes to where the EF output would normally connect to the circuit.
b) No different than a.
c) I have thought about this kind of thing before, except for a string swell.  Just set an input "Envelope Detection Threshold" comparator to trigger whatever envelope you want.  Use some synth module topology...maybe an envelope generator chip, and you have your exponential rise or whatever, triggered by a certain dynamic threshold.

[gez]

I don't know if high Q can be accomplished with CMOS inverters or not--I'm not sure what the gain is, but I know the bandwidth and slew rate are generally fast, however, upping the Q with uncompensated amplifiers may cause the thing to become an oscillator.

At 9V the gain of CMOS inverters isn't that great, as a result it's possible (well, I had no problems) of running these things with a 'high' Q (the low gain limits what is possible) without the circuit oscillating.

As much as I love CMOS I wasn't that impressed with SVFs built from inverters, I much prefer the "sharpness" of op-amp designs.

[Transmogrifox]

If the gain isn't very great, then that in itself limits the filter possible Q, so getting as high of Q's as say, a wide-bandwidth, high slew rate op amp like the OP275 or TSH22, is not possible...however, does a guitar filter circuit ever need a higher Q than what a CMOS inverter topology can acoomplish?  I don't know that...never tried.

[gez]

however, does a guitar filter circuit ever need a higher Q than what a CMOS inverter topology can acoomplish?  I don't know that...never tried.

It's not really a question of high Q, just 'higher'.  I've constructed all sorts of filters from CMOS inverters and after a while the outcome becomes pretty predictable - they all sound like the 'real thing' just not as good!  I found SVFs to be the same.

The only way I've managed to get good results (and by good I mean as good as or better than a op-amp/bipolar version) with any type of CMOS filter is to run them at lower voltages to up the gain, buffer the outputs and take the feedback network of the filter (or at least part of it) from the buffer.  Either that or use a nonbuffered chip for the higher gain.  The lower output impedance, when buffered, makes the filter more efficient and the higher gain enables Q comparable with an op-amp's (obviously the open-loop gain wont be as high, but as you said there's a limit to how high you can set closed-loop gain before oscillation kicks in).

Doing the above can yield excellent results, but it's something I only do with simple filters.  Buffering the outputs of a CMOS SVF would simply be a false economy, what you save in biasing resistors you loose in FETs/Darlingtons and source/emitter resistors - higher parts count (and higher current consumption) to get the same result you could have got with op-amps.

The MXR circuit isn't bad, it's just that I found it to be lacking (the filter part) compared to a well setup op-amp circuit.

[markusw]

Wow!
Thanks Mark for your exhaustive analysis!. You ought to write a book. It´d be a bestseller for sure (at least in the DIY stompbox community).
And thanks to gez and transmogrifox for your infos.
I will have to work through it.

If I have got it right, it should be possible to modify it a la Mutron to be able to select either LP, BP or HP?? And adding a send/return loop as well as some more controls like pots to vary decay (instead of R17?) and blending should give some kind of Meatball/MXR hybrid. Considering the great responsiveness of the MXR this should be extremely fine.

[earthtonesaudio]

I know it's been a long time, but I was wondering about the highlighted part of the quote here.  How do you implement that, Gez?

I have a 4069UB running at about 5 volts (lowered the voltage using diodes and series resistance), but what about the buffer connection?  Do you just take the output of your filter directly into the buffer, or is there a coupling cap between them?

And for the negative feedback, seems like you just take the leg of the filter off the output of the inverter, and stick it on the output of your buffer.  Am I way off?

Thanks for the help, and for revisiting an old topic!

however, does a guitar filter circuit ever need a higher Q than what a CMOS inverter topology can acoomplish?  I don't know that...never tried.

It's not really a question of high Q, just 'higher'.  I've constructed all sorts of filters from CMOS inverters and after a while the outcome becomes pretty predictable - they all sound like the 'real thing' just not as good!  I found SVFs to be the same.

The only way I've managed to get good results (and by good I mean as good as or better than a op-amp/bipolar version) with any type of CMOS filter is to run them at lower voltages to up the gain, buffer the outputs and take the feedback network of the filter (or at least part of it) from the buffer.  Either that or use a nonbuffered chip for the higher gain.  The lower output impedance, when buffered, makes the filter more efficient and the higher gain enables Q comparable with an op-amp's (obviously the open-loop gain wont be as high, but as you said there's a limit to how high you can set closed-loop gain before oscillation kicks in).

Doing the above can yield excellent results, but it's something I only do with simple filters.  Buffering the outputs of a CMOS SVF would simply be a false economy, what you save in biasing resistors you loose in FETs/Darlingtons and source/emitter resistors - higher parts count (and higher current consumption) to get the same result you could have got with op-amps.

The MXR circuit isn't bad, it's just that I found it to be lacking (the filter part) compared to a well setup op-amp circuit.

[gez]

How do you implement that, Gez?

http://www.diystompboxes.com/smfforum/index.php?topic=64846.new#new

[earthtonesaudio]

Great. Thanks!

[gez]

I have a 4069UB running at about 5 volts (lowered the voltage using diodes and series resistance),

A nice little trick that avoids having to use a current hogging regulator (though the one I use in the linked-to schematic is micropower and a beauty), is to stick a 3V3 zenner in series with a Si diode (this can double as a protection diode) between V+ and the Vdd pin of the chip.  This will drop 4V, but only if a reasonable amount of current flows through the zenner.   You can get away with just 2 stages of a 4049, but not a 4007.  The more stages you run, the higher the current draw and the more accurate the drop.  It's a nice way of conserving current and is more consistant that using a resistor (current draw varies from chip-to-chip).

Sometimes a little RC filter is required too, as zenners can introduce noise.

[earthtonesaudio]

I like the simple voltage regulation idea.  So I'm guessing that's a forward-biased zener, right?

[gez]

I like the simple voltage regulation idea.  So I'm guessing that's a forward-biased zener, right?

No, zeners are used reverse-biased (it's the breakdown voltage that is specified on the can).  If forward-biased, they have the usual forward-drop for silicon.  A 3V3 zener, if enough current flows through it, will drop 3.3V.  The extra Si diode drops a further 0.6 - 0.7V.  So, a total drop of 4V; leaving 5V (from a 9V supply) for the chip. 

The supply to the chip will drop as a battery flattens, but its not really critical.  I still prefer to use 5V micropower regulators, though.

[Paul Perry (Frostwave)]

I'd stick with the regulator too, unless you are usign a battery, or sharing a 9v regulated supply.
Because, if the power is from a lightly loaded wart & instead of the nominal 9v is actually 11v, then you have 11v - 4V = 7V, not 5!

[earthtonesaudio]

Cool.  I have an LM317 adjustable voltage regulator that I plan on using for the Valve Caster that Dano posted on here a while ago.  I suppose that would work, but what's a good regulator to use that doesn't hog power?  The one in Gez's schematic is a LP2950CZ.  Are they a good buy for a hobbyist (meaning, less than a dollar)?

[gez]

The one in Gez's schematic is a LP2950CZ.  Are they a good buy for a hobbyist (meaning, less than a dollar)?

http://www.lynxmotion.com/Search.aspx?txtSearch=LP2950CZ-5.0

http://www.lynxmotion.com/images/data/lp2950.pdf

ESR & Rapid for UK forumites.

IIR, the data sheet advises not to use a ceramic cap on the output of the regulator, but I do exactly that and they work fine (suck it and see).

[earthtonesaudio]

Thanks, I'll have to put a couple of those in my next round of bench-stocking.

[gez]

If you're interested, I added a sound clip in the COSMOS thread.

[Mbas974]

...resuming old post 

I've built the tonepad EF, (as stated here) it has too many ripple free.

Is there any mod to have it more silent ??
Thanks in advance.

[Mark Hammer]

...resuming old post 

I've built the tonepad EF, (as stated here) it has too many ripple free.

Is there any mod to have it more silent ??
Thanks in advance.

I am not sure what you mean.  What is your first language?  Maybe if you say it in that language, someone else here can translate it into English better.

If "more silent" means less hiss, then you will need to try out different 4069 chips to find one that is quieter.  The 3k6/100nf network at the output already gives a treble rolloff at 442 hz, so you don't want to make it any lower.