Clock for BBD

[POTL]

I noticed that my favorite flanger sound with a non-original watch is in such circuits as Ibane FL303, MXR117, EHX Electric Mistress and other.
I think this can be useful with other delay effects.
After looking at the many available schemes, I saw only 2 popular options (I do not take into account the strange diy projects in the Little Angel flock).
1) CD4047 in EHX circuits
2) CD4013 + CD4049 for EHX, DOD, Ibanez, MXR.
I'm interested in the advice of experienced people.

1) the modulation of the CD4047 (Small clone & DMM) seems somewhat wild, asymmetrical and ugly with great depth.
It seems like the LFO works in standard mode, like other manufacturers, is it a feature of the CD4047 or am I missing something in the LFO circuits?
2) Why does the CD4013 require buffers in the CD4049, and the CD4047 does not require them?
3) Is it possible to tune the CD4013 to both delay and modulation, like the CD4047?
4) Is it possible to replace 4049 with 4069, it has slightly fewer contacts (and case size), but judging by the data sheets its output resistance is worse, will it work worse or the difference will be imperceptible?

[Mark Hammer]

Panasonic BBDs have a somewhat higher input capacitance on their clock pins than comparable Reticon chips.  That input capacitance means that the higher the clock frequency you want to drive them with, the more current you need.  Here is how to think about it more easily.

A typical lowpass filter sets a rolloff equal to 1/2pi*R*C.  If we hold the value of the capacitor, the lower the value of the input resistor, the higher the rolloff frequency is, because the rolloff is set by how quickly the cap can be charged up, and more input current means the cap charges up faster.

What this means for us is that the need for buffers and current drivers will depend on what clock frequency we are aiming for.  Not much different at all than the challenge of cable-capacitance removing more high-frequency content if the cable is 100ft long, compared to 6 inches long; if you want those high frequencies, then that 100ft cable will have to be buffered. 

The Panasonic clock chips have limited current-drive capability.  A delay line will aim for a low clock frequency in order to increase delay time, so a basic MN3101/3102 is quite sufficient.  A chorus, using a 512-stage device, will aim for a clock frequency that delivers a delay range from a few msec to double-digits.  Again, an MN3101/3102 delivers sufficient current to overcome the input capacitance nicely and deliver sufficiently crisp clock pulses to the BBD.  Chorus with a 1024-stage BBD needs a faster clock, but as thousands of chorus pedals out there can attest, the required clock frequencies do not exceed the current delivery of the MN3101/3102.

When we get to flanging, now we move into different territory.  The Boss BF-2 illustrates the limits of the MN3102/3101 here.  The BBD itself can be clocked much faster to achieve very short delays, but the 3102 clock can only push it to a shortest delay time of 1msec.  If one wants anything faster from the clocking, the current drive has to be greater, in order to overcome the capacitance on the input pins of the MN3207.

So, to sum up, the higher the clock frequency one is aiming for, the more current drive there has to be to provide a crisp clock pulse, in the face of the input capacitance on the BBD.  Panasonic BBDs have about 700pf per 1024 stages.  So if one was trying to use a 2048-stage MN3208 for flanging, you would likely need additional current drive even to match the puny delay times of the BF-2.

Long story short, a CD4047 on its own is clearly sufficient to achieve the delay times suitable for chorus, but flanging would want more current.  In other words, if you want to go faster, you have to give the engine more gas.  One thing you will see sometimes is that the inverter sections on the 4049 are doubled or tripled to achieve more current drive.  A single invertor section may well provide the needed buffering, but not the current drive.  Those flangers one sees that ONLY use a 4013 are generally Reticon SAD-1024/512-based flangers and those chips have much lower input capacitance (110pf per 1024 stages, compared to 700pf for Panasonic), so the need for additional current drive is not there, unless one is aiming for exceptionally short delays.

[ElectricDruid]

Have you got links to example schematics with the 4013+4049, POTL?

I would guess (without having seen them) that they use a couple of inverter sections to generate the clock, and then use the 4013 flip-flop to produce the biphase output (since that chip has both Q and ~Q outputs). Since you've also got some inverter sections left over, you might as well use them...

The 4047 includes a flip-flop on the chip. The 'Osc' output is from the oscillator directly, and is twice as high as the Q and ~Q outputs. Since it has Q and ~Q outputs already, no extra chip is needed.

So really, it's that the 4047 doesn't need the *4013*, not that it doesn't need the 4049. I think I've seen a 4049 design that uses *just* that chip too (no 4013), now I think about it. I'll have to dig through some old flanger schematics. And just to be clear, it is pretty much always *flangers* where this starts to be an issue. Like Mark said, for chorus and delays, the clocks are low enough the current drive doesn't matter much.

The relative linearity of the clock modulation (and whether you even want it linear) is another story completely and will heavily affect the sweep of a flanger and therefore it's overall "character" or "sound". That depends on the details of how the modulation of the clock is achieved as well as the clock itself, and there are plenty of ways I've seen that done.

[POTL]

Hello
This is a standard variation of the circuit for simulating old reticon pedals.
A similar idea was used for a copy of the MXR flanger.
Therefore, it can be applied to other pedals with the CD4013.
A fairly simple solution that does not require any components other than the chip itself.

If we talk about flangers, the sound of this modulation is much nicer and more musical than that of the CD4047 (DMM & Small Clone).
Real performance can be heard on the clone Electric Mistress - Mooer E-Lady.
Here is a comparison example, the sound is even richer than real EHX pedals.

CD4013 can be found with tar chorus MXR, in DOD690, Ada Flanger, Ibanez FL303 all these effects were used by reticon chips.

The CD4047 looks attractive because it saves space on a single to-16 chip, but I have not heard other modulation / delay effects besides EHX, and their modulation sound is rather strange.
At large and medium depths, it resembles d-tune.

At the same time, the only modulation effect that I really like is the Electric Flanger, which just sounds harmonious and works on the CD4013.
I don't really understand how the watch works, how they transmit a signal from LFO to BBD, but after listening to many devices and comparing the circuits, I made the following conclusion:
1) MN3101 / MN3102 is the quietest sound, minimal depth and usually low speed (but maybe this is the work of LFO), a classic example of the Japanese sound BOSS / Ibanez / Maxon, plus DOD effects.
2) CD4013 - mainly MXR, early DOD / Ibanez are in EHX, an excellent range of effect depth and speed of LFO is musical in the whole range of work.
3) CD4047 sounds from moderate to extreme values, with increasing depth, the beauty of the sound is lost.

[Scruffie]

The clock is just a clock, it tells the BBD how much to delay and could be made from anything, it doesn't have a sound or impart one, it simply functions within given parameters.

How it's modulated is what produces the 'sound'. You could inject the mistress VCO circuitry in to a 4047 and it would sound identical to one with a 4013.

The 4049 is because the standard CMOS clocks can't drive a 3007 over about 800kHz which most of these circuits want at their top end.

The memory man and small clone both use a diode as a modulation element, its relationship with the LFO is not entirely linear, that may explain why you don't like it but it was cheap and did the job.

The designs are just totally different, you can't compare a small clone chorus to an electric mistress flanger, it sounds more like you just don't like chorus and prefer flanger. Have a listen to an A/DA Flanger, that uses a 4047.

A basic modulation circuits sound is made up of three things, its filtering, its delay time range and the shape of the LFO that modulates that delay time, they all go in to determining what it will sound like. How well that's achieved is down to the designer.

[POTL]

Mark thanks!
I'm trying to digest information 

Scruffie
Thanks

[ElectricDruid]

That Current Lover schematic is pretty interesting. Let's have a bit of a look at it.

The LFO IC5A/IC5D over on the left is pretty standard. Nothing unusual there.

It's followed by a gain stage IC5C (used for the matrix mode) and a LPF R31/C19. The lowpass is interesting. It's set at 4.8Hz, which suggests to me that it's designed to smooth out a triangle waveform at higher speeds and probably to reduce the depth too (stop things getting too warbley).

Now, the clock circuit is where things get really interesting. This is based on Q2 and IC6, a LM311 comparator. They could have used an op-amp, but I suspect they wanted to go to higher speeds, so a proper comparator became necessary.
I don't understand the detail of how the clock circuit works, or (most significantly) what response it will have to modulation. I might simulate it and play with it if I get a chance some day.
The output from that is taken to the 4013 which divides the clock by two and produces Q and ~Q outputs. Now, this is one of two steps in this circuit which is totally unnecessary. The other is feeding the flip-flop outputs through a single inverter before feeding them through a pair of inverters. You could ignore those two inverters IC3C and IC3B and nothing would change. If we don't have a good use for them, it would make more sense to parallel them up with the others and increase the current drive a little more. But back to the flip flop..why is it unnecessary?
Well, we could get an inverted "~Q" version with one of those spare inverters we've got kicking around. "Ah!", I hear you cry, "but then one phase would have one extra propagation delay compared to the other". This is true, but in fact, that's exactly what happens inside the 4013 flip-flop too. Check out the datasheet diagram of how the Q and ~Q outputs are generated:

https://www.ti.com/lit/ds/symlink/cd4013b.pdf

So one output has an extra inverter compared to the other anyway - so there's no benefit to producing the ~Q output with the flip-flop over the inverter. It could go.

Ok, now everyone can jump in and tell me why I've got it hopelessly wrong and why those bits are totally necessary. I'd love to know.

[POTL]

often came across posts where people did not understand how to use the second half of CD4013, found such a scheme, DOD suggests parallelizing them.
What do you think?

[Mark Hammer]

Strikes me more as precautionary rather than necessary.

[DrAlx]

I don't understand the detail of how the clock circuit works, or (most significantly) what response it will have to modulation. I might simulate it and play with it if I get a chance some day.

It linearly maps the control voltage applied to Pin2 of the 311 to VCO period.
So a triangular CV applied to the VCO will give a delay that changes linearly in time.
So you get a fixed pitch shift throughout the upward/downward sweep (the frequency sweep is hyperbolic).
There is no pitch "warble" unless you feed it a CV that is not varying linearly.

The 9V Mistress smooths the CV at high sweep rates so it is no longer triangular and as a result you can get a non-musical pitch warble for some settings of sweep rate and sweep range.
The original 18V EM did not do that because it did not filter the triangle wave.  So large fast sweeps on an original 18V don't warble, but rather ping-pong between upshifted and downshifted notes.

Explained here:
    https://www.diystompboxes.com/smfforum/index.php?topic=116836.msg1083340#msg1083340

[ElectricDruid]

Alex, if I'd seen your analysis before, I'd forgotten, so thanks very much. I find it a bit difficult to believe I missed such a thing, since it's so much right up my street, but I may have done.

POTL, I agree with Mark. Paralleling the two halves of the 4013 for a *delay* seems like it was only done because the other half was there, rather than because it was strictly necessary. Still, it should work - you're doubling the output drive just like you are with the 4049 inverters in other circuits.

[azone]

Ok, now everyone can jump in and tell me why I've got it hopelessly wrong and why those bits are totally necessary...

Old but useful thread. I've had to go down the rabbit hole of designing a BBD chorus. The reason the CD4013 is necessary is because the pulse width of the cheap linear VCO changes with frequency (amplitude too but not enough for triggering of 4013 to fail until higher freq's). Since the 4013 is only flipping and flopping on the rising edge it's not a problem. If the raw VCO is only sent to a buffer/driver/schmitt trigger it's detecting threshold on both rising and falling edges so PW changes. This is not good for the BBD IC. It wants to see a constant pulse width, ideally 50%.

I ended up designing about (10) BBD clocks and it was not easy to design a cheap clock with an extended linear region. I was solely interested in a linear clock, time w/ respect to voltage. so x ms per volt (not exponential 1V/oct or other response). This is the response for all the old roland juno's and boss chorus pedals (dimension C & D as well), and is responsible for the desirable character of these choruses. It is not so difficult to design a nice cheap linear clock for a MN3009 or even an MN3007 at longer delay times, but it was a little challenging to design one for a BBD that you want a wide sweep range at higher clocks. The best cheap solution ended up being based on the Electric Mistress LM311 VCO. If you keep the resting voltage of the modulation right around 5V and limit the range to about 2.2V to 7.8V you can get a very linear response. For example, depending on selectable timing caps, we got ranges from 70kHz to 700kHz (~0.7ms to 7ms for 1024-stage w/ 33pF timing cap) and 36kHz to 170kHz (3ms to 14ms for 1024-stage w/ 100pF timing cap) with >90% accuracy.

Regarding current drive to the BBD it is true that the 700pF input capacitance of the MN3007 is a problem at higher clocks. A buffer is necessary. I was able to get the MN3101 to work w/ some additional external circuitry but it wasn't really worth it. I like the MN3101 for its waveforming - keeping the pulse width perfect to the spec the BBD wants - and it's generation of VDD for the BBD . There's also no reason you can't buffer the output of it so you get a sharper clock at high frequencies and don't exceed the max 200mW power the device can handle, but it just wasn't worth the extra parts and complexity.

To summarize the cheapest lowest parts count solution for a 95% linear VCO (V vs time) over a range of 60kHz to 700kHz is the Electric Mistress VCO modulated from 2.2V to 7.8V with a 5V resting voltage. The clock generation is best with a 4013/4047 and a 4041 buffer. The 4041 is better than the 4049 or 4050. Using all buffers in parallel in either case the rise time is decreased from about 220ns to 150ns using the 4041 instead of 4049/4050.

 

[cabintech]

Been messing a lot with BBD clocks recently but I wonder if anyone can give me an idea of how fast a BBD clock might commonly be modulated? I understand the clock rate itself might range from a few kHz to 800kHz or more, but when the clock is modulated with a triangle or whatever, what kind of frequency is that modulation signal? I know LFOs are a common source of modulation, but are they really slow (a few Hz) or a few kHz, or even more?

A fast changing clock rate seems problematic since it would cause very rapid pitch shifting of the delayed signal, but what kind of modulation rates are actually used in common things like chorus, reverb, and flangers?

[ElectricDruid]

It's an LFO, and "Low Frequency Oscillator" in this musical context generally means sub-audio (otherwise it'd just be an "Oscillator"!).

Typical rates for chorus are between about 0.1Hz and 10Hz, with many covering less ground. The Maxon Stereo Chorus CS-9 only goes from 0.3Hz to 3Hz according to my notes, for example. For chorus, you're looking at rates you might think of for vibrato.

Flangers tend to sound good with big, slow sweeps, so often the LFO range is wider and lower, down to 0.05Hz (1 sweep every twenty seconds). They sound less good with fast sweeps, so the fast end of the range is rarely higher than for a chorus.

[cabintech]

Thanks, that's kind of what I thought, but not versed enough in BBD circuits to know for sure.

When the BBD clock frequency changes it causes a shift in the delayed signal pitch in proportion to the speed of the change. Is that pitch shifting actually a desirable side effect? If you could make a BBD that did not change pitch when the clock frequency is changed would that be a useful thing?

[Eb7+9]

To summarize the cheapest lowest parts count solution for a 95% linear VCO (V vs time) over a range of 60kHz to 700kHz is the Electric Mistress VCO modulated from 2.2V to 7.8V with a 5V resting voltage. The clock generation is best with a 4013/4047 and a 4041 buffer. The 4041 is better than the 4049 or 4050. Using all buffers in parallel in either case the rise time is decreased from about 220ns to 150ns using the 4041 instead of 4049/4050.

Thank-you for that reporting

Btw, a friendly reminder and hint here: one can only go so far (to maintain a solid clock shape at higher frequencies) by paralleling drain circuits ...

[ElectricDruid]

When the BBD clock frequency changes it causes a shift in the delayed signal pitch in proportion to the speed of the change.

It turns out that's a simplification, but it's true some of the time. If it was totally true, a triangle wave modulation would make two distinct pitches *and nothing else*. That's not the case. Here's why:

The delay is the last X (for example, 512) clock periods added together. We can think of the clock frequency as "stage shifts per second", in which case we need 512 shifts before the signal comes out. So the delay is an integration of the area under the clock frequency curve. Since the integration doesn't happen at a single point, but rather over a period of time (and a varying period of time at that - it's the *area* of the integration that stays fixed) there are *four* cases for the pitch shift due to a triangle wave modulation of the clock period:

Going up
Going over the top
Going down
Going around the bottom

The over-the-top and round-the-bottom parts produce a slew between the stable pitches produced in the middle of the slopes.

Is that pitch shifting actually a desirable side effect? If you could make a BBD that did not change pitch when the clock frequency is changed would that be a useful thing?

I think the pitch shifting is desirable. Or at least, "has become so" because it's associated with a desirable sound. It's an intrinsic part of the character of BBDs.

Digital delays that alter the delay time by moving read and write pointers in a delay buffer instead of changing the sample rate don't do it. Instead, if you just jump the pointer to the new location, you get a glitch in the audio. As a consequence, you finish up having to do some sort of crossfading or interplation to get back to a good effect. These delays are definitely useful and widely used. However, a naive implementation of a chorus using a digital delay line of this type won't sound like a BBD because it doesn't behave the same way.

Edit: Said period when I meant frequency. Fixed.

[cabintech]

Thanks, that is really helpful. I knew something else had to happen when the clock frequency changes suddenly (e.g. top and bottom of the triangle) but I could not quite get the math concept; the integral makes perfect sense.

Seems like it should be possible for a digital delay to do this, the integral is simple math as is the pitch shifting. The other major BBD "effect" of course is the effect of slower sampling rates for longer delay times. I guess some view that as good "crunchy" sound or some such term for "low fidelity" :-).