Understanding the MXR Envelope Filter

[slacker]

Slacker, what voltage are you seeing at Vs?

And thanks for your patience, everyone.  I'm determined to get this, but I am groping around sort of half blind here.

I get about 4.5 volts for Vs. I think it's correct, if you jumper it so the chips are running off 9 volts then the filter looses all its "quack". It is a bad way to get 4.5 volts as it relies on the current draw of the chips, like Stephens Giles pointed out though it's probably the cheapest way. I suppose the current limiting could be deliberate though.
What exactly was the problem you had when you built this before, did the filter not pass any sound at all or was it just that there was no sweep from the envelope follower.
Have you got any 40106 or other schmitt trigger chips? If you have you can make a simple clock out of one of those, which would let you test the switches and envelope follower. Let me know if you have and I'll do a quick schematic.
It's definately worth getting it working because it sounds amazing 

[slacker]

Couple more things you might like to try.
To test just the envelope follower disconnect the clock by removing R26 and C10. Then get an LED and a current limiting resistor, 5-10k should work, and connect them between point G and ground. The LED should flash as you play. On mine if you remove the LED and attach Point G to the 4066 control pins the filter will sweep even with the clock disconnected. It gates horribly but it does sweep.
If you connect the control pins of the 4066 to Vs through a resistor you should get sound through the filter. If you connect the control pins to ground you'll get no sound at all.

[mdh]

What exactly was the problem you had when you built this before, did the filter not pass any sound at all or was it just that there was no sweep from the envelope follower.
Have you got any 40106 or other schmitt trigger chips? If you have you can make a simple clock out of one of those, which would let you test the switches and envelope follower. Let me know if you have and I'll do a quick schematic.

The exact problem was that the effect basically behaved like a bad noise gate.  With the threshold pot turned fully one way (don't remember which direction), it passed little or no signal.  As I turned it in the other direction, it started acting like a gate, but there was a big spike (pop!) when the gate opened and closed.  I ended up building the whole circuit twice on different breadboards, and the first one behaved like I described, but I didn't bother to twiddle the threshold pot on the second attempt.  It's still on the board, though, so maybe I'll get to it tonight.  I'll also check the original build and see if the clock is oscillating.  Maybe I'm barking up the wrong tree by concentrating on the oscillator, but then again, maybe not.  It seems that if the amplitude is off, it could screw up the switching.

I don't have any Schmitt trigger chips, but I can make a Schmitt trigger out of two inverters, right?  I saw this in one or more of the 4069 data sheets.

Thanks for the confirmation on the voltage.  I guess I must have just missed Mr. Giles' comment, even though I thought I scoured the various threads for information about voltages for this circuit.

... and you posted more while I was replying... this place is the gift that just keeps giving.  I'll try those things, too, but like I said, I probably won't have much time to work on it this week.  Thanks again!

[slacker]

The exact problem was that the effect basically behaved like a bad noise gate.  With the threshold pot turned fully one way (don't remember which direction), it passed little or no signal.
Maybe I'm barking up the wrong tree by concentrating on the oscillator, but then again, maybe not.  It seems that if the amplitude is off, it could screw up the switching.

Thats basically what mine does with the clock disconnected, so it does sound like the clock isn't oscillating or the pulses aren't strong enough.

I don't have any Schmitt trigger chips, but I can make a Schmitt trigger out of two inverters, right? 

Not sure, if you haven't got the right chips you might as well carry on with the original clock. It's just that you can make an oscillator out of 1 gate of a 40106, which works as well as the original clock and would be simpler to debug.

Thanks for the confirmation on the voltage.  I guess I must have just missed Mr. Giles' comment, even though I thought I scoured the various threads for information about voltages for this circuit.

I just meant his post about MXR stripping things down to the bone for maximum profit, not specifically about the voltages, sorry for any confusion.

[mdh]

Thats basically what mine does with the clock disconnected, so it does sound like the clock isn't oscillating or the pulses aren't strong enough.

That's really useful to know.  If you try to measure an AC voltage at the output of the clock, do  you get anything?  My DMM will report a frequency there, but it doesn't seem to be able to measure the voltage.  I don't know if this is because of the high frequency, or just because the meter is loading the circuit.  I guess I really need a scope for this kind of stuff.

[Transmogrifox]

Thats basically what mine does with the clock disconnected, so it does sound like the clock isn't oscillating or the pulses aren't strong enough.

That's really useful to know.  If you try to measure an AC voltage at the output of the clock, do  you get anything?  My DMM will report a frequency there, but it doesn't seem to be able to measure the voltage.  I don't know if this is because of the high frequency, or just because the meter is loading the circuit.  I guess I really need a scope for this kind of stuff.

It may be as simple to solve as making C10 larger (like 22pF, or 33pF).  As far as testing the AC amplitude, you can make a peak detector out of a pair of op amps.  This would be like what you see in Harry Bissell's envelope follower, except you manually reset the peak with a switch or jumper wire.  You could measure the peak AC voltage by measuring the DC on the peak detector output.  You could then reverse the peak detector diodes and measure a negative peak.  You wouldn't be able to see the waveform, but you could tell how far it was deviating either way. 

As for making an oscillator with variable pulse-width control, this can be done with a single opamp, a capacitor, diode, a handful of resistors and a pot.
That way you could at least test the filter section.  You may also simply want to use a dual-ganged 500k pot to replace the switch in the filter section just to test that part.  Then tackle the clock.

[mdh]

I had a chance to play with this a little bit tonight.  I built a peak detector on the breadboard (http://www.uoguelph.ca/~antoon/gadgets/741/741.html, a little more than halfway down), and checked the amplitude at pin 9 of IC2.  With the diode oriented as shown in the schematic, I measured about 2.5V.  Increasing the value of C10 to 20pF seemed to increase the peak voltage at first, but after a little twiddling of the attack and threshold pots, the change seemed to go away.  I'm not sure if this should be the case, but in any case it seemed like a little more signal was passing with the larger cap, and I was getting a kind of (faint) misbiased fuzz sound.

I guess this is really just a little progress update.  I'm having a surprisingly good time debugging this thing, and I'm learning a lot.  Thanks for holding my hand through this.

[Transmogrifox]

Just for starters, I'm curious if the filter works at all.  Replace the switches (IC3) with 22k resistors (try a few different values in the 1k-50k range) and see if it sounds like a stuck wah.  If it doesn't work with resistors like a stuck wah, then you may as well quit worrying about the envelope follower and clock section until you can get the filter to work.

I'm glad you're enjoying this.  It's much more fun when you can enjoy the process as much as the outcome.  Obviously if you were only interested in the end result, you'd have bought the IC's that work with the filter a long time ago    Kudos on taking up the challenge

[mdh]

Thanks for the continued support, Transmogrifox.  I didn't see your last post until just now, when I came to post a little bit of an update.  I did manage to get ahold of some Toshiba 4069s, and the PCB build works now.  But I am still interested in figuring out whether I can make the circuit work with the TI ICs (I actually picked up a big pile of TIs when I got the Toshibas).  Now that I have a working version of the circuit that I can make measurements on, it should be a little easier.

I've done too many little experiments to describe here, but I think I'm making progress.  I think I'll rig up a way to sweep the filters with an LFO so that I can easily test different chips in the filter section.  Hopefully in the end I'll be able to provide some answers.  I have quite the assortment of 4069s and 4066s now, so I should be able to establish which chips play well together, and hopefully I'll be able to figure out why.

[rogeryu_ph]

Guys,
I search and found this thread. Currently also i'm on this project with the same  problem. The sound is too low need to crank my amp just to hear the sound and what i could hear is like tremolo sound and not sweeping also with popping sound which when attack knob is turn the pop sound become slow? My VR is 4.78v. I did change R16 into 62k already. R1 = 1k C2 = 1uf tantalum with neg going to output is this correct? I don't have mkt cap can I use NP electrolytic cap? One strange is even the circuit has no power the I still get fair amount or little bypass sound!!  Bypassed is ok. I already change all the out board wiring and DPDT and switch craft still strange!!

Anyone guys

mdh, hi what was your project outcome?

Transmogrifox,

Just for starters, I'm curious if the filter works at all.  Replace the switches (IC3) with 22k resistors (try a few different values in the 1k-50k range) and see if it sounds like a stuck wah.

R18 and R25 resistor is the one I should replace with 22k? (i'm using Tonepad Schem)

Thanks guys,
Roger

[Paul Perry (Frostwave)]

I just took a look at the original MXR EF schematic....... if my analysis is correct, no wonder people are having trouble!!
Because it seems that the detected envelope voltage AND the ultrasonic triangle voltage are both being fed to an inverter pin, so that they are being combined (as in a passive mixer - but a passive mixer that is depending an various characteristics of the oscillator inverter) - and if the result is above whatever the threshold of the inverter is (this can vary quite a bit, and changes with battery voltage as well) then the output changes & the frequency determining resistors are switched.

Now there is quite a lot to go wrong there, it's just plain dodgy to my mind!

[George Giblet]

> the ultrasonic triangle voltage are both being fed to an inverter pin,

Paul,  I analysed this ages ago and it's not dodgy.    As I recall it's is actual how the PWM works!

In many PWM circuits the voltage control varies the threshold where the tri-wave is sliced and this varies the PWM.

In the MXR ckt the detecting inverter has a fixed threshold and you vary the DC offset of the tri-wave so the (fixed threshold) inverter slices the tri-wave at different points (as the control voltage is varied).   If you see that (?) you will see that it works and that there is nothing dodgy going on - but yes it's a bit harder to see/understand.

[StephenGiles]

I had no problems at all breadboarding this a few years back and even used  4049 for the 4069

[Paul Perry (Frostwave)]

George, I agree with your analysis, but I can't help feeling that to have to depend on the detecting threshold of the inverter is a bit rough, because 1, the threshold varies between units (especially between manufacturers) and 2. the threshold varies with Vcc.
And the ultrasonic triangle oscillator probably DOES have 'issues', because - from posts - people have had trouble when they used different manufacturer's inverters.

Personally, I think the MXR envelope filter is a very creative design & gives very good results - the thing that concerns me here though, is whether it is an easy design to get going. I don't believe it is. If I wanted to make a hundred of them, I would have no trouble, once I secured the 'right' inverters & tweaked the circuit to suit. It's like a lot of the early EH designs, where the primary design aim appears to have been, how few parts can we use? (note in manufacturing, you pay for the parts twice - once when you buy them, and once when you pay to have them soldered in place.)

If I were designing a PWM envelope filter to build for myself, I'd be thinking a couple of 555s instead. (mind you, I could be wrong - wouldn't be the first time). But if I were wanting to manufacture a PWM envelope filter, everything would be happening inside a PIC.

[George Giblet]

> Personally, I think the MXR envelope filter is a very creative design & gives very good results

It's a great design from an innovative perspective.

> people have had trouble when they used different manufacturer's inverters

Some people are obviously having problems but I'm not sure what is going on here.

As far as stability and repeatability goes take the following extremes:  regulate everything and make everything precise in *absolute* terms, or, make the circuit self scaling in that it is independent of supply voltage.  (In the second case can only make the frequency independent of supply, not the output level).

From what I remember the thresholds of CMOS gates were a fairly predictable from unit to unit for a given supply voltage - that excludes Schmitt triggers which are extraordinarily variable!   The threshold was a relatively constant percentage of the supply.  That should make the MXR EF behaviour relatively independent of supply.

With oscillators like the LFO in flangers and phasers the opamps have constant staturation voltages on the output stage which make their behaviour deviate from self scaling.  The same goes for CMOS oscillators where the input clipping diodes are used to discharge the timing caps.  If you double the supply the saturation voltages and diode voltages don't double.  If MXR EF design doesn't have any traits like that.

On the MXR EF, if the supply voltage varies the behaviour seems like it should be relatively constant.  For example if the supply voltage varies the PWM inverter threshold will vary, however, the bias point of the prior envelope gain stages should also shift down in proportion and all should be well.

That's why I can't explain why people would have a lot of trouble with this circuit!

As a side note if you look a the 555 it's threshold are derived from the power rails and will vary with the supply voltage.  However, because the capacitor charge circuit is derived from the supply rail the thresholds vary in the same proportions and that's why 555 is stable with voltage variations.  *if* the threshold on CMOS inverters is relatively constant fraction of the supply voltage then it should be resonably stable with the supply in the same way.   

If you wire up a 555 with the a resistor from the output back to pins 2 and 6 and a cap to ground it will operate as a Schmitt oscillator.  As usualy the threshold points will move up and down in proportion to the supply but this circuit has nowhere near the robustness to supply variations.  The reason is the 555's bipolar transistor outputs now feed timing resistor and the saturation voltages don't scale up and down with the supply voltage.

(Sorry for the long winded drivel.)


Ed: The 40106 mentioned above is a CMOS schmitt trigger which are known to have widely variable thresholds.   Also it *won't work* properly in the PWM threshold inverter positions or any of the positions where the inverters are operating as amplifiers.  The 40106, or any other schmitts,  should not be used in this circuit!!!!

[Paul Perry (Frostwave)]

George, I appreciate your response. It is exactly the kind of reply I was hoping to see when I switched the computer on tonight!

[slacker]

Ed: The 40106 mentioned above is a CMOS schmitt trigger which are known to have widely variable thresholds.   Also it *won't work* properly in the PWM threshold inverter positions or any of the positions where the inverters are operating as amplifiers.  The 40106, or any other schmitts,  should not be used in this circuit!!!!

You're right it shouldn't be used to replace any of the 4069s in the circuit. The only reason I suggested it is because you can make an oscillator using 1 gate of a 40106, 1 cap and 1 resistor that works just as well as the more complicated oscillator used in the original. This could be used to debug the circuit to try and identify whether the issue was with the "clock" or the switching.
Like Paul pointed out this circuit seems to be a bit touchy as to whether it works or not but no one ever seems to have got to the bottom of it. The only solution that seems to work is switching to different manufacturer's 4069s, but that's not solving the underlying problem.

[slacker]

One strange is even the circuit has no power the I still get fair amount or little bypass sound!!  Bypassed is ok. I already change all the out board wiring and DPDT and switch craft still strange!!

That's normal with no power to the circuit the signal can still pass through C3, R7, R12, C2, and R1 so you will get some sound.

Just for starters, I'm curious if the filter works at all.  Replace the switches (IC3) with 22k resistors (try a few different values in the 1k-50k range) and see if it sounds like a stuck wah.

R18 and R25 resistor is the one I should replace with 22k? (i'm using Tonepad Schem)
[/quote]

No what you need to do is remove IC3 and put the 22k resistors between pins 10-11 and 8-9 of where IC3 was. If you've used a socket just plug the resistors into the socket. IC3 basically acts a a variable resistor so replacing it with normal resistors will make the filter work. If this solves you volume problem then you've got the same problem as mdh had and the only fix might be to try some different 4069s.

[mdh]

Mine worked perfectly well once I ran out and got 4069s and a 4066 from manufacturers other than TI.  I believe the 4069s are from ST Micro and the 4066 is a Motorola.  I've been meaning to break the circuit down into its constituent parts and test them with chips from different manufacturers, but I haven't had the time, and it went down on the list of priorities after I got it working.  So far, I think this thread has the best information on what the problem might be.

Are you using TI chips, Roger?

[StephenGiles]

My 4049 was a CMOS I also experimented with inverting the CV, using 2 filters in parallel and using an LFO. But where all the posted notes on this are I don't remember. I recall that Mark Hammer did something, but it may have been over at Ampage.