How to design a BBD flanger?

[Fancy Lime]

Hi there,

I've been looking into bucket brigade effects lately. A fun rabbit hole, that one. One thing I have difficulties figuring out, though, is why different flanger designs sound so different. Ok, they are different, so they ought to sound different. What I mean is: I have yet to figure out how to get even a rough a feel for how it is going to sound by simply looking at the schematic, which normally works rather well for me with most effect types. But telling which flanger is going to sound more nice and mistress-y or more extreme and low-altitude-jetplane-fly-over-y is still a bit of a tough one.

One important factor is obviously the maximum possible feedback. Another would be the maximum possible filter sweep (if it is called that in a flanger), aka the minimum and maximum possible delay times of the BBD. What delay time range am I aiming for in a flanger anyway? What else is important for the "extremity" of the flane? I noticed that many designs, especially the 80's Japanese ones, run their BBDs at 5V although as far as I can tell they should be specified for up to 10V. What's that all about? Seems a strange thing to do unless you want the BBD to saturate, which would probably sound fine, I guess, but take away from the depth of the flange, no?

Background is this: I find LFO controlled effects musically relatively useless for my purposes, with the exception of tap-tempo tremolo. So what I want in a flanger, is to be able to control it with an expression pedal. I am, quite frankly, absolutely flabbergasted that there seem to be no more than a few expression controlled flangers on the market. Makes them orders of magnitude more useful in my opinion. Anyway, i have learned to accept that my opinions are rarely the most popular. Still, I want that expression flanger, with an envelope control option, too. But I want it to go from not-very-subtle to tone-destroyingly-ridiculous. For me, flanger is an "effect effect". To be used sparingly but spectacularly. So any pointers to how to design the delay stage, what BBD to use, etc. are highly appreciated.

Thanks,
Andy

[ElectricDruid]

Completely agree about "fun rabbit hole". Been lost down here for a while, in fact. I thought going up might get me back to ground level, but instead I just got higher

You should use a something-3207 BBD, because those are the ones you can get. And they do a decent job - there are tons of classic flangers with those inside. Chasing some unobtainium chip to get "that sound" is a mug's game in my view. The sound of a flanger *isn't* going to depend on the BBD, except in a few rare cases perhaps.

The delay time is important, but the sweep *range* is perhaps more important. You want as wide a sweep as you can get. 1:20 is pretty good. 1:40 is much better. At one end, you'll be limited by how low the clock can go before you get audible whine. Don't go too low, because flangers are one thing that really benefits from a nice rich full-range signal, so it'd be a shame to have to filter everything above 2KHz because you're trying to get rid of clock noise. I'd aim for 8-10KHz or so, maybe more. That's perhaps a bottom end of 20-30KHz clock freq. At the other end, you'll be limited by how much drive you can provide to the BBD to overcome the clock input capacitance. This is why you see flangers with multiple paralleled CMOS sections - more drive current means you can go faster. I found the BBD chips themselves have a limit, and the signal loss starts to become noticeable, which then hits the depth of the flanging effect.

You might also see some highpass filtering in the feedback paths on flangers. Cutting the bass out of the feedback lets you turn it up more before it runs away and gives a more powerful flanger sound.

I don't see much reason to run the BBD at 5V unless your clock circuit also runs at 5V for some reason. My own Flangelicous design _does_ run the BBD at 5V, but that's to save level-shifting the clock which is provided by a PIC, which can only run at 5V. Otherwise, more headroom is better.

Another question is "to compand or not to compand?". Some did, some didn't. I haven't build one that does yet, but I suspect the combination of a slightly compressed/uncompressed (mangled?) sound and an improved S/N ratio could be better than not. Most flangers are a bit noisy especially with the resonance way up.

I also agree that envelope control or expression control is a good idea. The Flangelicious can be wired for expression control pretty easily (just add a stereo jack), but adding envelope control to it would have been too many more parts.

Finally, I still don't know what a given circuit will sound like, even having read all the flanger schematics I can get my hands on. Perhaps after simulating them and looking at stuff like filter responses and working out clock speeds and so on, you might have a better idea. But I'd say that's a forlorn hope, really.

[Mark Hammer]

After reading Tom's post through and through, read the relevant parts of these documents:
http://hammer.ampage.org/files/Device1-3.PDF
http://hammer.ampage.org/files/Device1-9.PDF
http://hammer.ampage.org/files/hypertriangleclock.gif
http://hammer.ampage.org/files/Hyperflange.PDF
http://hammer.ampage.org/files/PAiA_Phlanger.PDF
http://hammer.ampage.org/files/Anderflange1.PDF

That ought to shape your thinking in a productive way.

[anotherjim]

It's a rabbit hole I try to avoid.
I think...
Flanging requires a 50/50 wet/dry mix.
Flanging requires wet and dry to have identical sonic content.
Flanging requires short (shorter than Chorus) minimum delay to moderate (longer than chorus) delay.
Companding and BBD anti-alias filtering can counter the above requirements.
The closer to zero delay the sweep can go, the closer you get to the magic thru-zero sound.
A 1024 stage BBD needs clocking twice as fast as the classic 512 stage to reach the minimum. However, it can give better quality at the longer delay sweep with less anti-alias filtering.
Feedback is useful, but not essential. Feedback polarity only matters at the short delay end of the sweep.

[mth5044]

When I was looking at Flangers, I found DrAlx's posts on this forum to be incredible. There's one where he documents the progress of (I think) a TZF type with a lot of awesome info from others as well. Highly recommended!

[Fancy Lime]

Hi folks!

Wow, this generated more interest than I expected.

@Tom

You should use a something-3207 BBD, because those are the ones you can get

Yes, if I decide to design something partially or mostly from the ground up, I would most likely go with a coolaudio V3207 controlled by a V3102. This should also eliminate the need for an additional buffer for feeding the BBD as opposed to CD4013 controllers like in the modernized Electric Mistress or MXR 117 workalikes (see Madbeanpedals Current Lover and Collosalus, which use a CD4049 for that). As you say:

At the other end, you'll be limited by how much drive you can provide to the BBD to overcome the clock input capacitance. This is why you see flangers with multiple paralleled CMOS sections - more drive current means you can go faster.

I wonder if that may be the reason for limiting the MN3102/MN3207 combo in the Boss BF-2 and Ibanez FL-9 to 5V. Does the MN3207 need less current to go faster at 5V than it would at 9V? Both chips are rated 4-10V, so that can't be the issue. Interestingly, the Pearl FG-01 uses the same MN3102/MN3207arrangement but runs them at unregulated 9V. From watching youtube demos I cannot tell a difference in how high they go but then again, I would not expect to even if there is one.

Another question is "to compand or not to compand?".

Yeah, I'd like to avoid it unless it turns out to be really necessary. But that will depend on other choices which need to be made earlier.

@Mark
Thanks! Of those I had only read the last one so far. Hope to get around reading the rest in the coming days. Does indeed look very helpful.

@Jim

It's a rabbit hole I try to avoid.

Bwak Bwak Bwaaak!    Never admit to avoiding rabbit holes unless you want to be dared to a flange-off.

The closer to zero delay the sweep can go, the closer you get to the magic thru-zero sound.

I read about that. Unfortunately, I never heard how it sounds. Do you know or have any examples that show the effect. To get to true through zero, we need a second BBD and the additional control infrastructure along with it. So it would be nice to know in advance, if it is worth the hustle.

@Matt

When I was looking at Flangers, I found DrAlx's posts on this forum to be incredible. There's one where he documents the progress of (I think) a TZF type with a lot of awesome info from others as well. Highly recommended!

Thanks! I did not know those posts. Very interesting and educational indeed, although most of it is more complex than I am aiming for at the moment. But things usually end up more complicated than they started, so...

Cheers and thanks for the input,
Andy

[Fancy Lime]

Another thing: What delay times am I aiming for anyway? The MN3207 and V3207 do 2.56ms to 51.2ms, so 1:20. If I calculate that correctly that would mean that the lowest notch can go from about 40Hz to about 800Hz. That seems... tame. So does that mean I want to go with two BBDs, set one to minimum delay time and vary the other to get 40Hz to (theoretically) infinity? In other words, is through zero the way to go?

Andy

[Mark Hammer]

The choice point is really deciding what you want it to be best at: medium-speed swirl that does a slightly better version of rotating speaker than typical chorus pedals do, or slow dramatic Alex Lifeson sweeps.

Dramatic slow sweeps are going to require a few things:
- as wide a bandwidth as you can manage, so that notches can be heard way up high at shortest delay times
- a wide sweep ratio, of 30:1 or more; that's something that is achieved by aiming for shorter minimum-delay times, since a delay-range of 1.5 to 45msec is not going to get you much in the flanging zone
- an LFO waveform that will complement the range and speed of sweep; for long slow sweeps, the ear prefers a hypertriangular waveform that does the upper part of the sweep quickly and slows down as it reaches the lower part of the sweep

The old Eventide Instant Flanger added another feature, referred to as "theta processing" that used several fixed phase-shift stages to space out the notches at the bass end a little further apart.  But there we're moving beyond the pedal universe into rackmount.

[anotherjim]

To get the short delays you have to clock a 1024 stage chip faster than the book minimum. the classic BBD flangers had 256 stage devices. To get the longer delay end you might need to fit anti-alias filtering that reduces the notching effect of the higher frequency content - making going close to zero delay at the fast end irrelevant.
The flanging sound is the changing series of peaks and notches when the delayed sound is mixed with the dry. At the short delay end it is similar to phasing. For any particular frequency, a 180deg phase shift from the delay should cause complete cancellation when mixed 50/50 with the dry, while at 0deg or 360deg it adds to a +6dB boost. The cancellation notch can be infinitely deep and far outweigh the addition boost.
Where it differs from phasing is that the phase shift is time delay generated and continues beyond 360deg into the short echo range. In the longer delay range, the comb filtering sounds like a down-a pipe-effect as the notches spread throughout the signal bandwidth. In fact, at the longer delay times wet and dry are no longer spectrally similar with dynamically changing input such as music. So the notching and peaking is

If the highest harmonic from a guitar was 5kHz, then 0.0001ms delay is required to cancel it. How fast would a 1024 stage BBD need to be clocked to do that? Then you can also catch it at 0.0003ms delay.. and so on. However, since the signal is dynamically changing, the longer the delay is, the less chance there is of the wet being identical amplitude to the dry when the mixing happens - so notching won't be as deep.

Thru-zero is the sound of proper tape flanging when the second (wet) tape playback starts in perfect time sync with the first. When the wet tape is slowed, the highest frequencies in the audio will be notched first. It can be replicated with a pair of delays with one treated as the dry signal. If they are modulated in anti-phase to each other, then every cycle includes periods when there is no difference so you have zero delay.

[Fancy Lime]

To get the short delays you have to clock a 1024 stage chip faster than the book minimum. the classic BBD flangers had 256 stage devices. To get the longer delay end you might need to fit anti-alias filtering that reduces the notching effect of the higher frequency content - making going close to zero delay at the fast end irrelevant.

Interesting. So we can go shorter than the datasheet says by overclocking? Tell me more! I think I might keep the anti-aliasing as unaggressive as possible. I don't mind a bit of noise and messy bit-crushed sound on the low end of the sweep. The question is, if the notching is affected by that too much. Then again, the flanging effect is more noticeable at mid to high frequencies anyway and makes the lower frequency mostly go soft and wobbly. So one may think about focusing on the higher end or even add a switchable high-pass to one of the signals (the one with the longer delay), so keep the bottom tighter. The bigger problem that may force more aggressive anti-aliasing is the clock feed through. The 3207 wants 10kHz at the longest delay times, meaning heavy filtering above 5kHz will be required.

Also: Is there any 256 or 512 stage BBD on the market anymore? Not counting 25$ NOS chips that need a 24V negative supply and only work on the first Thursday after a full moon.



@Mark
Yes, I am definitely going for the dramatic jetplane in a tunnel sound as heard from Alex Lifeson, Eddie van Halen or (sometimes) Lennie Kravitz. Chorus sounds or Leslie simulation are not what I am looking for with this.



Huh, and to think I started this mental exercise thinking "I'll just take a commercially available pedal or PCB and mod that to my liking. Looks like that won't happen...

Andy

[Mark Hammer]

Let's clear this up.  NOTHING sweeps down to .0001ms.  That's 1/10,000 of a millisecond!!    The Boss BF-1 goes down to half a millisecond, and the BF-2 goes down to 1msec.  The A/DA Flanger, considered by many to be one of the very best ever, goes down to .4msec and up to 14msec ( https://www.adaamps.com/Media/Original%20FLNGMAN.pdf ), which is a 35:1 ratio.

All of these use a 1024-stage BBD, although earlier A/DA units. used a dual 512-stage unit (the MN3010), and the Boss HF-2 Hi-Band Flanger uses and MN3204, which is a 512-stage BBD.

The datasheets for the Panasonic chips pitch a maximum clock frequency well below the capacity of the BBD.  However, that recommended maximum frequency presumes one is driving it directly with the MN3101/3102 clock chip.  That chip provides a terrific compact and efficient clock output, but does NOT provide any means to overcome the high input capacitance of the clock pins on the BBD.  If one uses a buffer/driver in between whatever provides the clock, and the BBD clock pins, you can push those things into the stratosphere.  I've seen an MN3007 get successfully clocked up to 1.5mhz - fifteen times higher than what the datasheet says is maximum.

While the virtue of the Reticon chips is that their clock pin capacitance is low enough to tolerate high frequencies without needing additional buffer/drivers, commercial flangers like the A/DA or DIY projects like the PAiA Hyperflange and John Hollis Ultra Flanger using the Panasonic BBDs employ buffer/drivers to achieve shorter delay times.

While using lower-capacity BBDs, like 256-stage chips is helpful for achieving very short delay times, they are much less helpful for achieving the longer ones required for wide sweeps.  If one was aiming for A/DA territory, a 256-stage chip would be great for nailing 0.4msec (400usec), but getting 14msec out of it would require a low-enough clock frequency that heavy lowpass filtering would be needed.  This defeats the objective of getting as wide a bandwidth as you can.  Better to use a higher-capacity BBD, and additional circuitry to push it to higher clock speeds.

[Fancy Lime]

Let's clear this up.  NOTHING sweeps down to .0001ms.  That's 1/10,000 of a millisecond!!    The Boss BF-1 goes down to half a millisecond, and the BF-2 goes down to 1msec.  The A/DA Flanger, considered by many to be one of the very best ever, goes down to .4msec and up to 14msec ( https://www.adaamps.com/Media/Original%20FLNGMAN.pdf ), which is a 35:1 ratio.

All of these use a 1024-stage BBD, although earlier A/DA units. used a dual 512-stage unit (the MN3010), and the Boss HF-2 Hi-Band Flanger uses and MN3204, which is a 512-stage BBD.

The datasheets for the Panasonic chips pitch a maximum clock frequency well below the capacity of the BBD.  However, that recommended maximum frequency presumes one is driving it directly with the MN3101/3102 clock chip.  That chip provides a terrific compact and efficient clock output, but does NOT provide any means to overcome the high input capacitance of the clock pins on the BBD.  If one uses a buffer/driver in between whatever provides the clock, and the BBD clock pins, you can push those things into the stratosphere.  I've seen an MN3007 get successfully clocked up to 1.5mhz - fifteen times higher than what the datasheet says is maximum.

While the virtue of the Reticon chips is that their clock pin capacitance is low enough to tolerate high frequencies without needing additional buffer/drivers, commercial flangers like the A/DA or DIY projects like the PAiA Hyperflange and John Hollis Ultra Flanger using the Panasonic BBDs employ buffer/drivers to achieve shorter delay times.

While using lower-capacity BBDs, like 256-stage chips is helpful for achieving very short delay times, they are much less helpful for achieving the longer ones required for wide sweeps.  If one was aiming for A/DA territory, a 256-stage chip would be great for nailing 0.4msec (400usec), but getting 14msec out of it would require a low-enough clock frequency that heavy lowpass filtering would be needed.  This defeats the objective of getting as wide a bandwidth as you can.  Better to use a higher-capacity BBD, and additional circuitry to push it to higher clock speeds.

And with that, I am happy to bestow this years "Most Enlightening Post Award" upon Mark Hammer!

Read the datasheets, they said! The datasheets will give you all the info you need, they said! DATASHEETS ARE A LIE, I say!

But seriously, I was just about to ask how come that the age-old SAD1024 can be clocked with 1.5kHz - 1.5M Hz to give 340µs to 340ms and "modern" chips can only deliver a fraction of that range. So buffering it is then! Looks like I might order a current lover board for tinkering with after all.

Thanks a ton,
Andy

[Mark Hammer]

One has to think about the clock pins the way you think about long cables and their capacitance. 

Cables can behave like a single-pole lowpass filter; that is like a series resistance and capacitor to ground.  (There is likely more to it than that, and I'm sure the real EEs here will soon remind me of that, but we'll start simple.)

How do you fight cable capacitance such that you don't lose high end?  Well, you push more signal current through the cable, and provide a lower output impedance from the last thing before the cable. 

Though they never quite reach pure square wave state, the clock pulses are supposed to at least be nice crisp trapezoids, with minimal gap between falling and rising edges of each complementary pulse.  Crisp edges make for a smooth and relatively seamless handoff between the "tick" and "tock" sides of the BBD. 

The same way that cable capacitance will have a more pronounced effect on the tone of your Telecaster, or acoustic guitar, than it will on the tone of your EB-0, the input capacitance on the BBD clock pins won't start to have much impact on the crispness of the clock pulses until they reach a certain speed.  So, lower clock frequencies require no assistance whatsoever to produce nice seamless transitions/handoffs from stage to stage within the BBD.  It is that "unassisted" context that the datasheets refer to.  No great sin, since they want to sell the clock chip and the BBD as a sort of package.  But if one wishes to make the BBD do "tricks" and achieve higher clock speeds, they will need assistance.  Just as you wouldn't need much to run a 20ft cable from your bass to the amp, but would turn to buffer/drivers if you were running a cable from an acoustic, to the sound-desk 100ft away.

Given that, the limits of the Boss BF-2 are predicated on using an MN3102 without any assistance, and simply living with the minimum delay time that results in: 1msec.

[ElectricDruid]

Further to what Mark said, I can add a few details on the '3207 specifically. I've got both V3207 and MN3207 in my workshop, and they work the same - there's no audible difference.

I found that it was relatively simple to clock it to 500KHz (so 1 msec) but that above that I started see the output level droop which it starts to hit the 50/50% wet/dry balance. It is probably the case that with more work on the clock this can be overcome too. The clock input capacitance tends to round the clock pulses, and you need a nice square pulse to get a clean handover from one cell to the next, so more clock drive would probably help with the droopy levels too, but I haven't tried it.
It's possible that the levels droop because there is no longer time for the charge to completely be transferred from one cell to the next, I dunno. If so, that would be a genuine upper limit. But I doubt it at only 500KHz.

[Fancy Lime]

Hmm, so after some more reading and deliberation, more questions appeared. Since I have very little time at the moment, developing from scratch is not really my preferred option. So I thought I'd order the Madbean Current Lover and try to squeeze the maximum dramatic effect. The design seems to best lend itself to mod for using an expression pedal and/or envelope control. Schematic here:
http://www.madbeanpedals.com/projects/_folders/FilterMod/schematics/CurrentLover.gif

By replacing the Range control with an expression controlled Pot or an envelope controlled vactrol, I should be able to use the alternative controlled when the switch is in "manual" position, no?

Also: I'm wondering if I might want to add a Balance trimpot instead of R16 or in line with, in order to get the optimal balance between direct and delayed signal and thus the deepest notches. Would be easy to add in the send-return loop. Would that be a good idea?

Cheers,
Andy

[ElectricDruid]

Schematic here:
http://www.madbeanpedals.com/projects/_folders/FilterMod/schematics/CurrentLover.gif

That schematic is weird. There's all sorts of choices on there that seem odd to me. For example, IC3C and 3B don't do anything. Take them out. What changes? Nothing. They'd have been better paralleled with the other stages.

And that output mixer. A passive mixer, followed by an inverting op-amp stage with make up gain. If it had been non-inverting I could see the point, but if you've got an inverting amp there, why not use it as a mixer and avoid the losses?

And the power supply. There's C21 and C22 to bypass the power supply, but no resistor to aid with filtering, so then the separate supplies to all the different bits of the circuit are given their own RC filtering. The individual RC filtering is perhaps a good idea, but it all seems a bit haphazard.

Also *things left over*! Argh! I hate things left over! We've got half a 4013 and an op-amp wasted in this circuit, plus some other bits that look like they were only used because they were there already rather than because they were necessary.

By replacing the Range control with an expression controlled Pot or an envelope controlled vactrol, I should be able to use the alternative controlled when the switch is in "manual" position, no?

Yes, it looks like that's how it is intended.

Also: I'm wondering if I might want to add a Balance trimpot instead of R16 or in line with, in order to get the optimal balance between direct and delayed signal and thus the deepest notches. Would be easy to add in the send-return loop. Would that be a good idea?

Yes, could be. If the mixer were more sensible, I'd say not, but here it might make a difference.

I might also think about replacing R11 and R12 with a 10K pot so you can get the best trim to avoid clock noise. Given that the output filtering is "sketchy" this might help.

[Fancy Lime]

That schematic is weird. There's all sorts of choices on there that seem odd to me. For example, IC3C and 3B don't do anything. Take them out. What changes? Nothing. They'd have been better paralleled with the other stages.

Yeah, I was wondering the same thing.

And that output mixer. A passive mixer, followed by an inverting op-amp stage with make up gain. If it had been non-inverting I could see the point, but if you've got an inverting amp there, why not use it as a mixer and avoid the losses?

Wouldn't we get a n inverting active mixer by simply jumpering (is that a word?) R19?

Also *things left over*! Argh! I hate things left over! We've got half a 4013 and an op-amp wasted in this circuit, plus some other bits that look like they were only used because they were there already rather than because they were necessary.

Right? Sometimes there is no (sensible) way to not have thing left over but here I was wondering, why not just use an MN3102 as clock and driver (while keeping the CD4049 as a buffer)? Would save a lot of parts and presumably make the clock less finicky to set. Sure, it would diverge even more from the original mistress, but so do many other parts of the circuit. This is more a workalike than a clone at this point anyway. Hm.

I might also think about replacing R11 and R12 with a 10K pot so you can get the best trim to avoid clock noise. Given that the output filtering is "sketchy" this might help.

That one I did not have on the radar, thanks.

Andy

[Scruffie]

Schematic here:
http://www.madbeanpedals.com/projects/_folders/FilterMod/schematics/CurrentLover.gif

That schematic is weird. There's all sorts of choices on there that seem odd to me. For example, IC3C and 3B don't do anything. Take them out. What changes? Nothing. They'd have been better paralleled with the other stages.

While I've never bothered to scope and actually confirm it, the logic behind that move stems from the A/DA flanger clones back in the day where supposedly it was found that having all 3 paralleled actually began to distort the square wave at higher clock frequencies. Probably not an issue here but it's just the done thing.

And that output mixer. A passive mixer, followed by an inverting op-amp stage with make up gain. If it had been non-inverting I could see the point, but if you've got an inverting amp there, why not use it as a mixer and avoid the losses?

That output filter acts as a divider so the voltage is off, you'd have to move the 470R/47nF pair and then it wouldn't be a 'true' mistress clone. However there is a problem with that output booster, it's cutting bass.

And the power supply. There's C21 and C22 to bypass the power supply, but no resistor to aid with filtering, so then the separate supplies to all the different bits of the circuit are given their own RC filtering. The individual RC filtering is perhaps a good idea, but it all seems a bit haphazard.

While it's not done here so your statement stands, a small bypass cap directly at the clock power pins sans resistor (frequency and current draw) is very wise. The diode has some resistance though so it's not a total waste.

Also *things left over*! Argh! I hate things left over! We've got half a 4013 and an op-amp wasted in this circuit, plus some other bits that look like they were only used because they were there already rather than because they were necessary.

Use the spare 324 op amp as your V.Ref buffer, use the current V.Ref buffer to replace the transistor BBD output buffer. No clue what you could use the spare 4013 half for though.

[Mark Hammer]

Subbing an expression-pedal control for the Range pot?

Keep in mind that the earliest Electric Mistress pedals were in big boxes, and produced concurrently with the EHX Hot Foot, that was used to mechanically rotate pedal-pots with your foot.  I used to have one back in the late '70s.

The target applications were for producing manual flange-sweep with the EM, and for producing the first "whammy" effects by manipulating the delay-time control on the Memory Man.

Sadly, by assuming use with big-box FX that had widely-spaced controls, and lots of heft, the Hot Foot's design made it largely incompatible with smaller pedals from other manufacturers, that had controls too close together to permit attaching the flexible shaft, and would flip over from the torque the shaft produced.  Many years later, Tone in Progress produced their Third Hand controller that was essentially a redesign of the Hot Foot that overcame the Hot Foot's problems.

[Fancy Lime]

One thing I need to mention now: I've been reading through the articles Mark suggested only to find on the bottom of page 13 of this one :
http://hammer.ampage.org/files/Device1-9.PDF
none other than the mufferfuzzin' Space Jockey from the Derelict. Can we have a moment of solemn contemplation about the fact that Alien is 40 years old this Year, please? And so are the stompbox concepts we still build! Need to find someone to design me some H.R. Giger-esk designs for some stompboxes.

Also: I'd like to take back the insults I hurled at datasheets earlier. I took a look at the old datasheet of the MN3102 clock generator:
https://www.experimentalistsanonymous.com/diy/Datasheets/MN3102.pdf
after previously contenting myself with the one for the supposedly identical V3102, which seems to be more or less coolaudios reissue. Turns out the MN3102 datasheet contains a lot more useful information. Figure 3 actually solves the mystery, why Boss and Maxxon would run their MN3102-MN3207 combos at 5V, despite the loss of headroom that that entails. The maximum possible clock speed with the capacity of a MN320x series chip attached directly goes up as supply voltage goes down! While only 10V, 9V and 8V curves are shown and all end at around 1.5MHz, we can still see, that at 5V we can expect a very short delay with a 1024 stage MN3207. I would assume that the switching should be at least a factor 3, maybe 5 faster than the 900kHz at 9V, although extrapolations of this kind are to be taken with a standard overseas shipping container of salt.

All of that makes me wonder: Maybe I should give the Ibanez FL-9 a go. Reportedly, the Joyo Classic Flanger is a more or less exact replica of that one. Although the sources for this info are highly shaky and the whole theory may be based on little more than the knobs having the same utterly weird names, for all I know. Considering how sacrilegiously cheap the Joyo is, I might just buy that one and mod it and see how it goes. Assuming it is indeed the FL-9 design, then it has a Manual control (called "Delay Time" cause that won't confuse guitar players at all on something that isn't a delay) and the Width knob is positioned such that a switch can be inserted here to disconnect the LFO and/or insert an envelope follower here.

The Third Hand seems to not be available for purchase anymore. So sticking everything in a wah enclosure is probably the way to go. Does anyone know a good long-stroke wah enclosure and where to buy it? Preferably in Europe? Ok, that's pushing it, I know. Seems impossible to find wah enclosures anymore. And even broken units of shitty old wahs go for rather crazy money on ebay these days. Or we go back to the external pedal control. Have we agreed on a universal standard yet? 5V variable voltage, was it?

Andy