Flanger designs. clock modulation, filtering etc.

[Brian Marshall]

First off as a disclosure I am working on a flanger which we plan to build for profit in the near future.  I am not looking for any specific information, but have made some observations that I thought I would share.  However random thoughts and musings are welcome.

First off I am using a mn3102 and mn3207.  Nothing too exotic or hard to find.  I drew up an initial idea without reference to any prior work, other than what I remember from looking at schematics months or years ago.  I tried several different ideas to get my clock to respond to a voltage over a wide range in a pleasant way, and started out initially with fairly minimal filtering.  

the clock
So first off, the datasheet for the mn3102 is pretty sparse when it comes to modulating a clock.  All of the ideas used were similar to ideas I have used with CD4047's in the past using transistors.  All of my original ideas used all three control pins (5 6 and 7) but being driven by a symmetrical wave form they all seemed to spend their more time in the high end.  It worked, but it didn't really sound right to me.

So I went off looking at other examples.  ADA, Boss, DOD, Pearl, etc.  Some don't even use a regular clock driver chip, the ADA being an obvious example using the CD4047.  We'll basically ignore those ones for now.
The first thing i noticed about all the 3102 based circuits was that all of them left pin 6 unconnected, and added an external transistor use (essentially) as an inverter.  Another transistor was used to drain current in some way to slow down the clock.  I tried most of them, but it seemed that in most cases the LFO was manipulated in some way before it really got near the clock for a non-linear response.  This was probably workable in the original designs because the flanger worked with a fairly limited range.  

Eventually after gaining some insight from all this prior work I came up with a fairly simple solution based on way typical older DOD flangers worked.  This seemed the most unlikely at first, because the DOD phasers basically use a diode hack to make the LFO voltage non symetrical.  From the basic idea I added a limiting resistor so the clock couldnt go to low and a very small capacitor to send some HF "feedback" to the current control transistor's base.  After trying it with several different chips and transistors I think you could probably get away with out any need for trim pots, but I suppose only time will tell.  If any one is interested, and i get some time, I will post a drawing of the clock circuit at some point.

filtering

As I said my original idea had very little filtering.  I was skeptical that a flanger would need the same level of filtering you'd normally see in an echo pedal since the clock runs in a much higher range, but all my memories for repairs old flangers and looking at schematics lead me to believe this was not the common practice.  In spite of my clock issues in the beginning the actual flanging sounded pretty darn good in my opinion.  I did end up adding one mutiple feedback filter which both brightened up the effect and dealt with some clock noise at lower delay times.  I started to wonder what really was the point of all this extra filtering, especially after plugging in some of the flangers i own which seem to all be much more subtle.

Then I took another look at the ADA schematic.  It is considered one of the best by many vintage gearhead types.  In contrast to others there is remarkably little filtering save for some upper end bleed off and some high end emphasis in the feedback/regen signal.

It does have a gate though... probably the reason they didn't worry about clock noise.

Anyway, just some general observations I thought others might find interesting/useful.

[Valentinych]

Eventually after gaining some insight from all this prior work I came up with a fairly simple solution based on way typical older DOD flangers worked.

It would be interesting to see your circuit VCO.

[Fender3D]

Hi Brian,
"random thoughts and musings are welcome" so...

the clock
maybe it's just me, but I've never heard a good sounding flanger with MN31xx.
Great flangers such as ADA, EHX or MXR didn't used them when those chips weren't available and didn't used them when available later either.
But, again, maybe it's just me.
You get the strongest flanging effect when you have a wide Delaymax/Delaymin ratio (30-35 and more) then you should consider your delay line (1024 stages in your case). It's a common opinion that, with guitar, you'll barely notice the flanging effect with less than 0.4 ms delay. Within these data you can start designing a schematic...
Personally, my ideal clock has a fmin and a fmax external adjustable controls where LFO sweeps. (but maybe this is more a synth thing than a pedal...).
With MN31xx you'll be seriously limited, at least in fmax.
Another point is that with just 10V max supply (or worse 5V) you must find a way to properly modulate it to achieve your desired dmax/dmin ratio.

filtering
I agree with you with the minimal filtering.
I use the ADA gate in my flangers and chorus.
Maybe a switchable filter, so you can choose when higher bandwidth is better than noise reduction...

[slacker]

Interesting stuff, thanks for sharing.

It does have a gate though... probably the reason they didn't worry about clock noise.

As far as I know the gate in the ADA is to kill the flanging below a certain input level so that when you aren't playing the pedal doesn't flange any hiss and noise coming from your guitar or whatever pedals you have in front of it. I don't think it's there to hide clock noise.  

[Mark Hammer]

First, welcome back Brian!     It's been a long time.  I think I speak for many when I say we are proud of your accomplishments as "one of ours" (Brian is the brains behind Subdecay pedals.  You can read more here: http://www.effectsdatabase.com/interviews/brands/subdecay).

Sometimes I think designers in the major companies approached flangers and choruses as kind of the same circuit, with only a few minor tweaks, but really they are only distant cousins.  Choruses don't really "need" bandwidth to work well, where flangers are absolutely desperate for top end, since the impact of the effect is most pronounced when there are the greatest number of notches.  And to have more notches, you need to have bandwidth in both the signal, and the circuit it passes through.

Apart from the limited use of lowpass filtering in the A/DA, I think one of the other aspects of the design that doesn't get talked about much is their use of a diode-based "soft limiter" (the Boss BF-2 uses the same strategy).  The intent is to assure that the input+feedback signal does not exceed the handling capacity of the BBD, but one of the things it does is introduce a little more harmonic content.  Not a lot, but some.  Of course, if you apply steep filtering to it, you won't hear it, but if you don't filter very much it can add some liveliness.

If you haven't already (though I doubt it), read the flanger-related documents on my site (hammer.ampage.org); particularly the documentation for the PAiA Hyperflange, and the relevant articles in those scanned copies of DEVICE.  One of the things they emphasize is the sweep ratio.  And for a whole host of reasons, using an "unassisted" 3207/3102 pair simply won't take you there.  The principal reason is the input capacitance on the clock pins of the BBD.  It,s about 700pf per 1024 stages, if memory serves, and the 3102 simply can't deliver the needed current to make those clock pulses crisp and square under those circumstances.  That's why you see CMOS chips pushing the best analog flangers, whether as buffers for a 310x-based clock generator, or as the clock generator themselves.  Feeding a higher current clock signal permits the circuit to sweep over a much broader range, and well above the 100khz upper limit the typical datasheets propose.  The MN3207 can probably be clocked quite high - up to 1mhz - if the clock signal is well managed.  And that's where you get the great flanger sounds.

[Brian Marshall]

Hi Brian,
"random thoughts and musings are welcome" so...

the clock
maybe it's just me, but I've never heard a good sounding flanger with MN31xx.
Great flangers such as ADA, EHX or MXR didn't used them when those chips weren't available and didn't used them when available later either.
But, again, maybe it's just me.
You get the strongest flanging effect when you have a wide Delaymax/Delaymin ratio (30-35 and more) then you should consider your delay line (1024 stages in your case). It's a common opinion that, with guitar, you'll barely notice the flanging effect with less than 0.4 ms delay. Within these data you can start designing a schematic...
Personally, my ideal clock has a fmin and a fmax external adjustable controls where LFO sweeps. (but maybe this is more a synth thing than a pedal...).
With MN31xx you'll be seriously limited, at least in fmax.
Another point is that with just 10V max supply (or worse 5V) you must find a way to properly modulate it to achieve your desired dmax/dmin ratio.

I've never owned an ADA, but I'm considering trying to pick one up as this circuit is developed for reference. Currently I have a vintage electric mistress and a couple old DOD flangers.  It's funny you mention EHX as I really find the electric mistress understated.  I have a boss BF2 and HF 2 around somewhere, but havent seen them in a long time.
I don't have a frequency counter so I can't tell you the exact delay times i am getting, but the range is pretty wide the way i have things set up currently.  Wide enough that it will probably need to be scaled back a bit.
The MN pair I am using is simply because I have had them laying around for a couple years.  I bought them knowing I can get them from several different places and that there are modern equivalents being made.  That is kind of important for a production pedal.
I've not tried driving the BBD with anything else just yet, and as i understand the CD4046 and CD4047 (which I have way more experience with) usually needs to be buffered to drive these BBDs without buffering.  I notice in the ADA schematic they use an MN3010.  Shorter delay line... and it is driven directly by the CD4047.
I'm probably going to order some other BBDs to play around with.  I may not end up using them for production but at least I can hear and see the difference.

filtering
I agree with you with the minimal filtering.
I use the ADA gate in my flangers and chorus.
Maybe a switchable filter, so you can choose when higher bandwidth is better than noise reduction...

I was considering adding a tone control of some sort in the feedback line... That might be counterintuitive though because it would basically do nothing with the feedback at zero.

[Brian Marshall]

If you haven't already (though I doubt it), read the flanger-related documents on my site (hammer.ampage.org); particularly the documentation for the PAiA Hyperflange, and the relevant articles in those scanned copies of DEVICE.  One of the things they emphasize is the sweep ratio.  And for a whole host of reasons, using an "unassisted" 3207/3102 pair simply won't take you there.  The principal reason is the input capacitance on the clock pins of the BBD.  It,s about 700pf per 1024 stages, if memory serves, and the 3102 simply can't deliver the needed current to make those clock pulses crisp and square under those circumstances.  That's why you see CMOS chips pushing the best analog flangers, whether as buffers for a 310x-based clock generator, or as the clock generator themselves.  Feeding a higher current clock signal permits the circuit to sweep over a much broader range, and well above the 100khz upper limit the typical datasheets propose.  The MN3207 can probably be clocked quite high - up to 1mhz - if the clock signal is well managed.  And that's where you get the great flanger sounds.

I haven't done much reading just yet other than datasheets and a few schematics.  The ampage site is a great resource that I always forget about.
I've just been trying to find a way to get a decent sweep out of the 3102.    Our control voltage is going to be from an MCU, and I'm looking for a really wide range that won't need much if any manipulation for a nonlinear response.
As I mentioned in a previous reply I've yet to try using anything else to drive the BBD.  I'm going to try a CD4047 today or tomorrow with a similar control scheme that we use in the noisebox, albeit at a much higher frequency.  I'm going to have limited space to work with when drawing up a PCB, so that is a concern... but somehow I fit an 8 stage phaser with similar features in to a 1590BB 

[Mark Hammer]

The old PAiA Hyperflange uses a two-position bass cut switch for the feedback path.  That helps to remove much of the talking-inside-a-10-gallon-oildrum tone, and also lets one increase the feedback level up to near oscillation.  I find I can improve the hi-feedback sounds of phasers by cutting bass too.  You are certainly correct about its non-impact at zero-feedback settings, but the point is to make higher feedback settings more usable.  For me, the more interesting concern is where that rolloff ought to be to strike a good balance between effect intensity and avoidance of oscillation, and also what sort of slope we're talking about.

Does it NEED to be variable?  I don't know.  Personally, I don't think so.  It may not even need to be switchable.  The first issue Small Stones had a cap that "reshaped" the LFO waveform with increasing LFO speed, such that it was hypertriangular for slow speeds, and increasingly triangular for faster ones.  It was a seamless transition that the user was completely unaware of, and didn't need to switch anything.  Maybe that's what good design looks like.  I am presently unaware of any commercial flangers that carry out a similar LFO reshaping with speed changes, though they may well exist.

As for what "fits" into a 1590BB, is there anything SIPs and SMT can't do?  

BTW, is there anything out there that does the "pillowy" envelope-controlled flanging that Frank Zappa used to use his Wasatch for?

[Valentinych]

Brian, I think those messages could interested you:
http://www.diystompboxes.com/smfforum/index.php?topic=91981.msg803554#msg803554
http://www.diystompboxes.com/smfforum/index.php?topic=93466.msg804021#msg804021

Edit:
In addition, I can tell you about my development of the LFO on microcontroller ATtiny13. The device allows you to generate signals of sinusoidal, triangular, hipertriangular and circular shape in the desired frequency range (0.04 -10 Hertz). And if necessary I can program any form of signal.

[Brian Marshall]

The old PAiA Hyperflange uses a two-position bass cut switch for the feedback path.  That helps to remove much of the talking-inside-a-10-gallon-oildrum tone, and also lets one increase the feedback level up to near oscillation.  I find I can improve the hi-feedback sounds of phasers by cutting bass too.  You are certainly correct about its non-impact at zero-feedback settings, but the point is to make higher feedback settings more usable.  For me, the more interesting concern is where that rolloff ought to be to strike a good balance between effect intensity and avoidance of oscillation, and also what sort of slope we're talking about.

Does it NEED to be variable?  I don't know.  Personally, I don't think so.  It may not even need to be switchable.  The first issue Small Stones had a cap that "reshaped" the LFO waveform with increasing LFO speed, such that it was hypertriangular for slow speeds, and increasingly triangular for faster ones.  It was a seamless transition that the user was completely unaware of, and didn't need to switch anything.  Maybe that's what good design looks like.  I am presently unaware of any commercial flangers that carry out a similar LFO reshaping with speed changes, though they may well exist.

As for what "fits" into a 1590BB, is there anything SIPs and SMT can't do?  

BTW, is there anything out there that does the "pillowy" envelope-controlled flanging that Frank Zappa used to use his Wasatch for?

It appears to me that most analog flangers do some speed related shaping simply with some low pass filtering on the LFO.  Granted that is subtle.

I suppose the filtering does not need to be variable, but I'd been playing around with the feedback signal.  Currently I've got an HPF around 180hz and have tried several different areas and ideas for low pass filtering.  Making it adjustable can make it sound warmer, or what some might call 'more vintage.'  Of course with zero feedback it does nothing.  I tried moving the LP filtering elsewhere, but it didn't have the same effect.

Other than ICs we've not really done much with SMT.  Id hate to try to place resistors by hand, and a pick and place isn't anywhere in our near future.  I've looked in to having others do it for us, but the minimum quantities and up front costs are a bit much at this point.

[Brian Marshall]

So I am about to pull the 3201 and try some other ideas.  I thought I would draw it up for the notebook in case i ever needed to refer back to it.  So here it is in all of its sloppiness.



So if you basically ignore Q2 and the control voltages you have Q1 acting as an inverter creating a loop which oscillates.

It all probably looks familiar if you have looked at DOD based designs in the past.

For most of my testing R7 was omitted.

Added is:
R1, which limits how slow the clock can go.
C2, which sends the clock signal to the control current transistor.

C2 is kind of the tricky part to explain.  Initially I had C2 going from pin 5 of the IC to the base of Q2.  I don't remember exactly what the problem was, but it wasn't working very well.  So instead I saw that I already had the signal reduced by half at the base of Q1, so I tried that and I got much closer to the desired results.

Follow me for a minute here.
Clock speeds go lower as the current to Q2's base gets higher.
So we have some clock signal going back in to our control current, which is essentially making Q2 give Q1 less current just as it switches on.  
This seems like it would affect the clock frequency, but not necessarily do anything to sweeten up the sweep, but lets take a look at Q2 again.
As the CV goes up and down so does the current at the base.  As the current goes up so does the impedance from base to emitter goes down, thus making the clock signal to the base of Q2 smaller and smaller.
But then we have to look back at Q1.  This is the part I did not think of when I tried this initially.  C2 is creating a LPF at the base of Q1.  When the current at Q2's base is low C2 has little effect, but as the current rises impedance falls. The LPF starts to slow down the clock a bit more.

The battling effects of C2 across both transistor bases seems to even out in a way that makes the sweep a bit more even.... it does however make the slow end of the clock go much slower.
If the control voltages get to high the clock will fall low enough that it's not really flange-ey anymore.  More like a chorus with some extra predelay.  It can sound rather interesting, but not what you'd expect necessarily out of a proper working flanger.

After a bit of thought I see the reliance of the impedance on Q1's base as a bit of a potential problem... As we all know transistors can vary quite a bit, and I suspect a Trim pot, either at the emitter or the base may be necessary for reliable results.

Anyway, I thought I'd throw it out there.

[12Bass]

Just noticed this thread recently and am curious to know if any headway has been made.

Regarding filtering:  Although the A/DA design employs fairly lax filtering, I reduced it even more in my SAD1024A build.  I also employed high quality op amps and capacitors in the signal path.  The delay path on mine covers the full audio bandwidth and sounds surprisingly hi-fi when isolated.  When combined with the straight signal, the flanger produces healthy notches which go up past 20 kHz; after careful calibration, at their maximum, I measured over 34 dB of attenuation.  Given a wide bandwidth signal, the resultant flanging is quite pronounced and dramatic, even without added regeneration.  I'm fairly sure that it sounds more "alive" than the stock circuit. 

An interesting observation is that, while processed music and instruments sound quite natural, the relaxed filtering produces easily audible aliasing noise if fed high frequency test tones.  If I had to guess, I figure that the aliasing is sufficiently masked by normal signals that it doesn't call attention to itself; I can't detect aliasing, even if I strain to hear it.  But feed in a 10 kHz tone and sweep it up and down, and there are some nasty byproducts....   

There are a couple of things that I'd like to change about the A/DA design.  One is that headroom is limited, even though I've set the BBD bias as optimally as possible.  The other is the waveform of the sweep.  IMO, it rushes too quickly through the midrange, then lingers a bit longer at the top and bottom of the sweep.  I've been meaning to experiment with adding a hypertriangular circuit to the LFO to see if that helps. 

[12Bass]

Looks like the end result: http://www.subdecay.com/starlight-quantum-series-flanger