Envelope controlled filter question

[Fancy Lime]

Hi everyone,

so I decided it's time to graduate to envelope filters. The very funky kind. Designs goals:

-- Low pass filter with resonant peak (mode 1), band pass (mode 2) and maybe also notch (mode 3, not a priority. Will only be implemented if the design makes it easy to do so).

-- Up and down modes.

-- Full control over Q, range, attack and sensitivity

A natural starting point here seems to be the Mu-Tron III. And since it heavily features in R.G.'s auto-wah article, I even understand whats going on there. However, I would like to avoid optocouplers if possible. Mostly because they are expensive and a bit hard to get. Therefore, I'm leaning toward the filter control methods of the Doctor Q (or Dr. Quack) or Tim Escobedo's Phuncgnosis. Thing is: I don't quite understand the filter control on those. Can someone point me in a direction to learn a little something about:

1) How to design Voltage Controlled Resistors using BJTs. That's what the Dr Q does and applies to a multiple feedback bandpass, if I understand that correctly. Could not find useful material on that (that I understood) but I might have bees searching for the wrong terms.

2) How does the Phuncgnosis filter control work? What are the diodes doing in the filter section? Complete mystery to me at this point. How does the up-and-down control work?

Deadastronaut, you there? Your Phuncgnosis mods seem to improve things. Can you shed some light on this?

Thanks a bunch,
Andy

[Mark Hammer]

Use of commercial packaged optoisolators is fine to avoid if they are deemed expensive or hard to get.

That said, there are cogent reasons for use LDRs as one's control element.  One of the conundrums of envelope controlled anything is envelope ripple during the decay phase of a plucked note.  As much as we would like to place faith in our envelope-extractors, they are imperfect.  Or, if they are good and smooth, they are not responsive enough for our liking.  One of the things that LDRs do is smooth things out a bit, such that we can aim for a faster response, but not have to suffer the jittery, sputter, ripple that leads many to ask "Why does my <filter> distort?".

A second consideration is that, while getting a bandpass filter to sweep only requires one control element, be it an LDR, bipolar transistor or FET (or diodes in the case of the Phuncgnosis), getting a lowpass with a resonant peak will require two control elements in order to sweep a state variable filter.  And, especially since those elements will not be connected to ground, that points towards use of optocouplers, be they commercial or homebrew.

The other alternative is that one use an entirely different approach, such as PWM of a control element or OTAs.  The MXR Envelope Filter uses the former.  And, while it is a SVF, only one filter mode is used in the stock pedal, even though others are available if one taps at other points.  I've loved that pedal ever since I bought one in '77 or so, mostly because it was for a long time, the only commercially available one that had variable attack time.  Decay time is easy to implement, and while it does play a role in matching the feel of the sweep to a song, attack time is probably more critical for sweep/song matching, and the MXR circuit does a nice job of that.

[digi2t]

I recently traced out the Frostwave Funk-a-Duck. I can post the schematic if it will help you out. Envelope controlled MS-20 style filter.

No opto's.

[Kipper4]

3 modes
Defo with an SVF.

Up down mode things can get messy but no impossible within limits.

Nsl32 are a cheap and cheerful and very useful opto. I couldn't guess how much they would cost in your part of the world, but its probability worth googling.


I'll go find the phungnosis scheme.

[Transmogrifox]

1) How to design Voltage Controlled Resistors using BJTs. That's what the Dr Q does and applies to a multiple feedback bandpass, if I understand that correctly. Could not find useful material on that (that I understood) but I might have bees searching for the wrong terms.

The use of a BJT as a variable resistor operates it in saturation where it behaves most like a diode.  A diode exhibits a resistive behavior depending on the current through it.  More current, lower resistance.

This is only true for very small signals as the behavior is very nonlinear and it gets to sounding like distortion when the signal levels are very large.

The MFB filter is well-suited to this use since the voltage across the resistive element can be small.  With the dual-integrator loop of the SVF, the voltage across the device can get relatively large and you end up with distortion.

2) How does the Phuncgnosis filter control work? What are the diodes doing in the filter section? Complete mystery to me at this point. How does the up-and-down control work?

Phuncgnosis works the same way as the Dr Q, only it uses LED's and uses the envelope detector current to change the small-signal resistance in the LED's.  The up and down control works by swapping polarity of the LED's.  In the down direction, the LED chain is biased ON, and the envelope detector works to shut it OFF (higher resistance).  In the up direction, the LED is biased OFF and the envelope detector works to turn it ON (lower resistance).

As Mark mentioned, PWM is an option.  You could apply the analog switches directly to the Mutron circuit, or just build an MXR filter and take the additional outputs.

Another option is the OTA-based state variable filter.  The LM13700 datasheet has an typical application circuit with low-pass and band-pass outputs.  It has to be extended somewhat to get the high-pass output (and thus notch), but this isn't hard to do.

Basically you replace each integrator stage in the Mutron design with an OTA stage.  The OTA stage is the integrator.

[Kipper4]

This phungnosis

http://www.diale.org/escobedo.html

I suppose it's not out of the question to use the charge pumps envelope detector.

Lob a led driver on the end of it. Add attack . Decay pot. A must for me. Attack is upto you.

The up down switch. Not sure if it makes a difference.
Not certain how it works but I guess the diode ladder versus the led provide different bias points in relation to the +input at from.


The mfb filters could be improved with a pot instead of 5M. Or series value and pot..

Rich

[Ben Lyman]

in my Phuncgnosis the up/down switch works great but when it's in one of the modes (reverse of normal) it adds enough extra noise that it's unbearable with a heavy distortion pedal.

[Mark Hammer]

If one aims for very fast attack and decay, then there is no practical difference between LDRs and any other sort of control element, since the changes happen too fast for ripple to be audible.  The Guyatone WR-2 Wah-Rocker, that Guthrie Govan uses exceptionally well and expressively, has fast time constants, and sounds ideal.  But if the intent is to have a choice of faster AND slower sweeps, using solid-state devices with slower sweeps shows up their shortcomings - they are TOO responsive to the little ripples in the envelope, and have more time to show it off.

I made myself a clone of the old DOD 440 Envelope Filter, with homebrew dual vactrol, and it sounds very vocal.  A different sort of filter with a pleasant sound.  There was a mod published in DEVICE (posted on my site) that shows how to increase the resonance.  The DOD FX-25 is an OTA-based filter that can provide lowpass and bandpass outputs (though the stock unit only provides bandpass) and comes with variable resonance.  The envelope-extraction lends itself to mods readily.  The bandpass and lowpass outputs have highly discrepant output levels, so any switching between the two requires some compensation for level.  But not bad sounding.  The 440 is smoother-sounding to my ears, though.

[Fancy Lime]

Boy, that escalated quickly. I mean, that really got out of hand fast.

Thank you so much, everybody! You lost me with the terminology for a bit there. Praised be the almighty Google. Envelope filters are one gargantuan barrel of worms, I'm starting to think. That's going to be so much fun to dive into. Strictly figuratively speaking.

@Transmogrifox
Thanks for the explanations! That makes sense now.

@Kipper4
The NSL-32 is indeed the only optoisolator I can get for a reasonable price. A pair of those would set me back 8€, the price of a 1590B Hammond enclosure in these parts. I can get other optoisolators for 20€ a piece. Or LDRs for 8€ a piece. So nothing I would not pay if necessary but something I'd like to avoid if possible without sacrificing quality.

@Mark
I was dimly aware of there being some sort of "speed-problem" but not of the specifics. Thanks for the clarification!

@All
By popular suggestion I took a good hard look at the LM13700 data sheet. Let me see if I can summarize here: One can use an operational transconductance amplifier (OTA) such as the LM13600 or LM13700 to make a lowpass or bandpass in much the same way the mu-tron does with LDRs and op-amps. Tim Escobedo's "9V MS-20 Sallen-key LP VCF" would be such an example (albeit static). The DOD Fe-25 and Mad Professor Snow White (and Frostwave Funk-a-Duck?) use this principle method.

The MXR Envelope Filter sounds mighty interesting, especially because I agree with Mark that Attack is an important contol to have. Unfortunately getting my hands on a 4066 is not going to happen locally. So, out of sheer convenience, my first attempts are probably going in the direction of a LM13700 based thingamajig, tonally geared toward Mu-Tron and Moogerfooger 101 LP. The big question there: is there a way to control attack at all? I have not seen an attack knob on any of the schematics other than the MXR and I cannot think of an obvious way to do it. One would need a variable resistor that reacts slowly to a trigger signal... aaaaand we're back to LDRs, damn. How else to postpone the rise of the envelope (uh, Rise of the Envelope, nice band name)?

Cheers and thanks,
Andy

[ElectricDruid]

If you're diving into the barrel of worms, you should have a read around "state variable filters" (SVF). Here's a good start:

http://sound.whsites.net/articles/state-variable.htm

Here's the typical schematic, complete with three outputs. You can get a notch by mixing the LP and HP outputs.



The gist of this is that R6 and R7 set the cutoff frequency (in association with the two caps of course). At this point you decide how you want to do the actual implementation - you can use OTAs/VCAs to act as the resistors, or you can use LDRs in a vactrol, or you can use a PWM scheme. It's all the same filter, just with different control elements.

Of course, you don't *have* to use a SVF, but for three outputs simultaneously, it can't really be beaten.

HTH,
Tom

PS: Incidentally, about the notch...

http://www.edn.com/design/analog/4437967/Three-op-amp-state-variable-filter-perfects-the-notch

[Fancy Lime]

Hi Tom,

yes, I'm aware of the SVF concept although I did not know its name before starting this here thread. The concept itself and how it works is fairly self explanatory from the Mu-Tron III schematic. Thanks for the links, great to have some additional theory. Need to get some more background to get this thing right.

Andy

[Mark Hammer]

The big question there: is there a way to control attack at all? I have not seen an attack knob on any of the schematics other than the MXR and I cannot think of an obvious way to do it. One would need a variable resistor that reacts slowly to a trigger signal... aaaaand we're back to LDRs, damn. How else to postpone the rise of the envelope (uh, Rise of the Envelope, nice band name)?

The general convention is that there will be a small resistance in series with the rectifier output (e.g., 47-330R), followed by an electrolytic cap to ground and often (though not always) a fixed resistor in parallel with the cap.  The small-value resistor determines how quickly the cap will charge up.  The cap also determines how long that charge-up will take.  The series resistor is the bucket used to fill the barrel, and the cap is the barrel.  A smaller resistor provides more current, filling up the barrel faster.  A bigger cap makes for a bigger barrel that takes longer to fill.

The parallel resistor is a hole in the barrel that drains it.  The barrel is leaking anyway, but the hole speeds up the leaking.  Designers will generally trade off series-resistor and cap values to achieve different sorts of charge-up/drain-off speeds.  Although increasing the series resistor will slow down the charge-up, unless there is a suitable buffer stage inserted immediately after the cap, increasing the series resistance beyond a certain point will yield an outcome of very poor sensitivity.  In other words, you can't simply increase Attack time without needing to turn up the Sensitivity.  The MXR unit is one of the few that allow independent Attack-time adjustment.  Although, even there, one can need to turn the Threshold up a smidgen at longest attack times.

I used to "believe" in mucking with the attack resistance, but found that the range of resistances needed to achieve a meaningful range of attack times  was unfeasible, if I followed stock designs.  I had much better luck by simply adding a variable resistance in parallel with the cap, to vary drain-off (decay) times.  It had much less impact on useful sensitivity, but still yielded a pleasing variation in feel without having to change the design very much at all.

[digi2t]

Here's the Funk-a-Duck;

[Fancy Lime]

@digi2t
Interesting schematic, thanks! Not awfully obvious to me what does what, though. I'll have a deeper look and compare to the demos on youtube as soon as I find the time.

@Mark Hammer
Thanks for the explanation! The "bucket and barrel network" after the rectifier you are referring to looks something like this in the Mu-Tron.

---330Ω-----------
          |      |
         4µ7   47k
          |      |
          v      -9V

If I'm not mistaken, replacing the 47k to ground with a 100k or so pot (plus maybe a 22k resistor in series with the pot) would act as a Release control (draining the barrel with a spigot), correct? If changing the series resistor (lets call it Rs) has unwanted side effects, I would imagine that this has partially to do with the fact that it forms a voltage divider with the resistor to ground (Rg). If Rs becomes large (increasing attack time), then the peak voltage available at the output of the network goes down, influencing sensitivity. So we want to keep Rs<<Rg in all settings to keep interaction with the sensitivity knob reasonable. Changing the capacitor (within certain margins) on the other hand should have no effect on the output voltage. Have you fiddled with that? A pot for fine tuning Attack would be great but a switch with 2-5 settings that switches between capacitors and allows some discrete Attack values as presets would still be a lot better than nothing. I don't mind have to readjust the Sensitivity after changing Attack and Release, but of course if the interaction is too strong it becomes difficult to find the desired settings.

Also: I'm wondering what would be the effect of adding a second such network right after the first. I mean, for rising voltages, the Rs-C combination acts as a delay from a voltage over time perspective, correct?. The C-Rg combination does the same for falling voltages. So should a second network not have the effect of smoothing out ripples. One may want to design the first one with faster and the second with slower reaction. The effect may be too small to be worth the effort but has anyone actually tried? If it works that may be a bit of an improvement (and nice switchable option) for fast Release times on bass frequencies, where we otherwise get the "space wobble" effect (which can be nice too). Is that worth the try?

Cheers,
Andy

[ElectricDruid]

Adding a second RC filter combination after the first does do what you think it does, but the losses through the filter start getting serious, since the second stage loads the first stage.

If you think of the 330R/47K as a voltage divider, you're ok, because most of the voltage gets through.

If you stick another 330R/47K across the 47K of the first divider, you've now got basically half of 47K (two 47Ks in parallel) so the output level drops significantly.

In reality it isn't this simple (I've completely ignored the caps, for example) but you get the principle.

But try it. You might have enough signal, or enough sensitivity in the following filter control input.

Tom

[Fancy Lime]

Hi Tom,

thanks for the clarification. I'll play with that a bit and see what sticks. The values will certainly need major tweaking in this case. I was planning to start with something around 330Ω for Rs1 + Rs2 in series and 47k for Rg1 and Rg2 in parallel to keep close to the original voltage divider values. So the Rs have to becom smaller and the Rg bigger. Maybe equal values (~165Ω and 100k) are a good starting point. We'll see.

I'm not 100% sure but I think the first Rg in such an arrangement may be unnecessary (well, more unnecessary than the additional Rs and C) since it is parallel with the other Rg and only separated from it via a comparatively small resistor.

Cheers,
Andy

[Mark Hammer]

Your instincts regarding the voltage-dividing properties of Rs/Rg are correct.  But one also needs to be concerned with the current-limiting effect of Rs.  Imagine that the circuit directly drives an LED.  The sort of series resistance required to appreciably increase the charge-up time of the cap would also drastically reduce the current driving the LED, reducing its brightness.

As I hope I conveyed earlier, if there is something else in between the envelope-modifying components (Rs, etc.) and the control element it is driving (as is the case for the MXR Envelope Filter), in order to buffer or otherwise compensate for that current loss, then one can tinker with attack time at will.  But most commercial designs won't do that, largely because it adds parts, cost, space.  One would have to "invent" the added circuitry to adapt to an existing design and insert it.  That's certainly not impossible, but only a small fraction of us here are equipped to do all that on our own.  A great many prefer to hunt for a circuit and accompanying layout that meets their needs and just make that.  That's not wrong, but it means that "I just wish it could do more X" requests will require a bigger step beyond.

[Fancy Lime]

Hi Mark,

[...] That's certainly not impossible, but only a small fraction of us here are equipped to do all that on our own.

I agree and I sure am not one of them who can do that all on my own. Good thing I'm not alone here, isn't it? Seriously though, I'm not the social type, so the main reason I am on this forum is to exchange ideas. I'm getting really useful input here and I hope to give others some useful ideas too because in my experience, most great ideas are the result of dialog and quite few are a divine afflatus to an isolated genius.

A great many prefer to hunt for a circuit and accompanying layout that meets their needs and just make that.

Well, not me I don't. I started the whole DIY thing partly to see if I could and partly because I was unhappy with any and all available bass fuzzes at the time. But by now its mostly driven by the quest for something new. The whole cloning and modding is just a way for me to learn how things work before trying to improve on them and making something new. If I just wanted an envelope filter that sounds the way I want it, I'd simply build a Mu-Tron III clone with a few mods and additions, that would get me 95% to what I want. But I'll be damned if I won't try to get those last 5% and stray from the beaten path where it is useful. I'm not a believer in traditions for traditions sake and often interesting things happen when one tries to do things a little differently just to see what happens. That's a big part of the fun of this game for me. Having a good sounding stompbox is nice and all but having developed and built one that is (subjectively) better than the others doing similar things is what I enjoy most. As you said, that's difficult to do alone. And that is exactly what is great about this forum and the support and discussion input I get here. Thanks, I really appreciate that.

Cheers,
Andy

[Mark Hammer]

It's still DIY if one does hunt down a suitable circuit and layout.  Heck, it's still DIY if you send away for a pre-etched board and parts kit.  It's just a bit more DIY if one DIAYs (A = all) it. 

Some folks will stick with tradition because there is something they always wanted to own and use, but they either couldn't find one or the price was prohibitive.  I don't fault them for not having an original; I credit them for their good taste.  And most will eventually decide to wander beyond the perimeter of what has already been done.  Sometimes they only think they have wandered outside that boundary.  But that's okay, even if they are repeating something that has already been done to death.  Their heart is in the right place.  Sometimes, folks get possessed by the B.U.M. syndrome (blind urge to mod), purchasing a perfectly decent pedal that resulted from hundreds of hours of R & D, and immediately asking "What mods can I do?".  Again, that may be wasted effort on their part, but their heart is in the right place if they feel it's a good thing to want to go beyond the known.

One of the aspects that has distinguished this forum for the two decades of its existence in various forms/platforms is that folks spend 95% of their effort in collaboratively solving problems, and much less of it either mouthing off or criticizing others for their choices, as can sadly be the case when there is nothing to work on together.  It makes for a much nicer ambiance.  We're glad you found it, and glad we found you, too.

[Kipper4]

Here here.
For me this is just my bag. I've built two mutron iii and have parts and pcb for another.
Good luck getting the vcl vactrol for it. Those are getting scarce..

I've been waiting for a topic like this.
I'm too busy doing landscaping in the garden to make pedals until winter arrives. U.K.

Maybe I can learn some about the implementation and use of OTA's.
Bring it on.

Mark Hammer, Dino, Tom. These guys are the best. I've learn so much from them.

There's a very helpful magazine PDF about OTA's on the internet somewhere.
I'll look it up for ya.

Rich.
edit
Would'nt you know it i cant find the magazine article on the net. I'll dig mine out and send it to you asap.

Did I say already R.G Keens Technology of the auto wah should help you with knowing what does what in an envelope filter.