Another Stab at Explaining the Wah -- And an interesting inductorless wah

[Transmogrifox]

http://cackleberrypines.net/transmogrifox/TransmogriNotes/skwah/index.html
[EDIT]Here are some mp3 demos:
http://www.cackleberrypines.net/transmogrifox/TransmogriNotes/skwah/demo/
[/EDIT]

I decided to just start uploading notes & sim on stuff I'm poking at.  Was kind of a fun write-up anyway.  It's sort of like an extension to "The technology of the wah".  You would need to read that first as a pre-req because I did little to explain the inner workings of the wah.

I trust somebody will learn something and maybe even try the SKF wah presented there.

I also expect I may learn something from somebody who considers it and adds things I haven't considered here.

Here's the circuit for those who don't like to follow links:

If you want to know more about it, you have to follow a link.

...or ask a question.  I really didn't explain much about HOW it works.  It was focused mostly on what it does and how its salient features compare to the classic inductor wah.

[B Tremblay]

Nice work! Thank you for sharing it.

[Gus]

Interesting. It is good to see something different.

I like the output follower, looks like a white follower.  I noticed Q5 biasing resistor R55 is connected to Q1 collector and not +9 directly.  when I have some time I will study this circuit some more.

Have you simmed a higher open-loop compound pair at Q1 location? or is it better to have one limited open-loop gain transistor for more distortion this being a guitar effect?

[dschwartz]

Wow, you have somethinggoing on here!!

Have you build it?? I tried to analize the schem but i have no idea what is doing..it is not a twin t filter, doesnt look like a mfb, or a sallen key,,,

Do you mind explaining it a bit?

[Transmogrifox]

Have you simmed a higher open-loop compound pair at Q1 location? or is it better to have one limited open-loop gain transistor for more distortion this being a guitar effect?

I can't say I understand what you mean by "compound pair"...maybe the  npn-pnp "69" pair type follower (like fdeck HPF)?  Or something else?

I have tried several different combinations of transistors, and interestingly this is the highest gain with lowest output impedance I was able to get with 2 transistors.  The "69" type follower (like fdeck HPF) seemed a good amplifier, but if you give it gain, the output impedance is high enough to defeat the usefulness of the gain in terms of driving the resonant tank.   Maybe a different RC selection could get there.  It definitely works well when the variable element can go up to 1 Meg.  Then R36 doesn't need to be so small, then you can actually get gain with a higher output impedance.  R36 ultimately is what limits the gain.

The "69" pair does work well as long as there is a constant current source in the location of Q5.  That is how I came to this final configuration.  I thought if I just drive the base of the CCS, then I can do away with the extra pnp transistor.  As it turns out it seems to perform the same with or without it. 

What I needed was a good solid push-pull drive, and I think this does it well enough. 

I would be interested in any other ideas you have.  I don't think I have entirely exhausted every combination of feedback amplifier here.

Have you build it??

Yes.  It sounds really good.  [EDIT] Got demos uploaded:
http://www.cackleberrypines.net/transmogrifox/TransmogriNotes/skwah/demo/
[/EDIT]

I tried to analize the schem but i have no idea what is doing..it is not a twin t filter, doesnt look like a mfb, or a sallen key,,,

Do you mind explaining it a bit?

It actually is a Sallen-Key filter.  It's arranged as what is commonly the high-pass realization of the filter, but the high pass input is grounded.  As implemented it is a low pass filter.

Instead, the signal is injected into the point that normally has a resistor tied to a 4.5V bias voltage.  My earlier implementations included a JFET in place of Q1 and therefore R1 and C2 were not used, and the pot fed the signal directly onto the JFET gate. 

Unfortunately I found the JFET does not have enough gain to get enough resonance on the filter, so a BJT was needed for its high transconductance.  Another thing that would work is a TMOS (2N7000) biased at 100 mA, clearly a power hog (and probably noisy too).

It looks a lot more complicated than it actually is because  the arrangement around Q5 makes it looks like something interesting is going on there.  Really, Q5 is what makes this into a sort of low-quality op amp where the output of the circuit would be the op amp output, and the emitter of Q1 looks like the inverting (-) input.  R48 is a feedback resistor.

If you read the link provided you will actually see a snip of this implemented with an op amp.  Maybe in that configuration it will look more obviously like a Sallen-Key high pass with the input grounded, and signal injection at a different point.

R50 may also add some confusion in the way of being able to identify this as a Sallen-Key topology.  This resistor essentially degrades the quality of C17 and works as a location to insert a pot where you can decrease the Q without changing the resonant frequency.

[PRR]

> what you mean by "compound pair"

I can't say I ever heard of the "69", but I know just what it is.

There are many types of BJT "compounds". Cowles 1966 has a summary. Zen.co.uk has a quickie of the most-pop compounds.

> a Sallen-Key high pass with the input grounded, and signal injection at a different point.

Mighty twisted.

If I am reading it right, I would say "gyrator inductor plus capacitor". This rigs a series C-L tank. For high Q the source Z must be mighty low and the C-L node resistance must be very high; also the gyrator amp output impedance must be low.

Yes, that is a (non-optimized) White Cathode Follower. Zout is reduced about hFE times. A similar scheme at the input would be pretty, though I'd think the Sziklai would work fine at input or for both. (Not to mention op-amps, though the low series R may need sumthing beefier than TL072.)

[Transmogrifox]

I just coined the term "69" trying to think of some visual way to describe it -- case in point, you knew what I was talking about

Thanks for the link.  Based on the summary my term "69" refers to the Sziklai pair.  This is where I started but wasn't getting the Q I was aiming for without a constant current source bias.

What I ended up with was a different implementation of the same thing combining the amplifier transistor and current source into one device, and looks like I reinvented something that has been well known longer than I have been alive .

I agree with your assessment of "gyrator inductor plus capacitor" is correct.  The feed resistor acts as a parallel resistor decreasing Q as it decreases simulated inductance.  That is one way to look at it.

A Sallen-Key high pass realization is a capacitor in series with a grounded gyrator inductor.  It doesn't seem twisted to inject signal at a different place in the circuit and still call it a Sallen-Key:

The RC & amplifier network doesn't change by grounding C1 and injecting Vin at R2.  That is all I have done with the circuit under consideration.

The "twist" is using a feedback resistor and tapping "R1" from the op amp inverting input in order to recover high frequency gain at the output simply by taking advantage of the fact that the impedance of the network decreases with increasing resonant frequency.

So whether you call it a gyrator-capacitor filter or a Sallen-Key, the transfer function looks the same and http://sim.okawa-denshi.jp/en/OPseikiHikeisan.htm accurately predicts the resonant frequency and Q of this network.

FWIW an OP275 worked fine in this configuration when I prototyped the op amp version.  The discrete transistor version is much lower noise as predicted by simulation.

[PRR]

> looks like I reinvented something that has been well known longer than I have been alive .

The Ancients stole all our good ideas.

Here is a compilation of Eric White's patent application from 1940, Broskie's analysis, and Cavelli's analysis.
http://www.pearl-hifi.com/06_Lit_Archive/02_PEARL_Arch/Vol_04/Sec_19/991_White_Follower_Optimization.pdf
3MB PDF file

Yeah, "cathode" means these papers do not apply directly to a BJT implementation. And my "hFE better" snap-opinion may be wrong. {EDIT: only about 2X better.} But the WCF, in any functional form, is a Heavy Feedback amplifier with significant gain going to lower output impedance. (And slightly better output drive.)

Your implementation is somewhat off from a clean WCF. The 680r between transistors has no equivalent in a plain WCF. This may do something for you in this application. {EDIT: some quick sims don't find a whole lot of improvement using "optimum" theory, nor any real harm from the added 680r.}

[PRR]

> It doesn't seem twisted

By seeing it as a gyrator plus C and R, you can write the resonance in terms of L, C, R(series), R(shunt); and change of response is relatively straightforward.

However since you already have found/computed values which give a very pretty set of curves, the need for simple synthesis is past.

[Transmogrifox]

Here is a compilation of Eric White's patent application from 1940

Really interesting read.

Yes, I see your point that viewing it as a gyrator greatly simplifies the practical aspects of dealing with this circuit analytically.  It sort of sweeps the "negligible" terms under the rug.  As an aside, the gyrator model is how I originally selected my "starting point" C & R values for the circuit before I started simulation.  The rest is just a natural evolution based on empirical results.  I have about 3 prototypes of this where I simulated, built, listened, modified simulation, simulated, built, listened, etc...until I liked it.

Intuitively your conclusion about the 680r makes sense.  In a small-signal model, it's a resistor in series with a voltage-controlled current source, so it has no bearing on the final feedback current going into the summing node, and negligible effect when other physical model parameters are taken into account to make the "ideal current source" less ideal.

The nonlinear behavior is the interesting piece of the 680r.  The currents coming from Q5 are of a higher magnitude as frequency increases (as impedance at the summing node decreases).  This results in more distortion on the high end and mimics the way the inductor wah has typically less distortion on the low sweep range, increasing distortion as it sweeps up.  That's the freebie that comes out of this topology.

[Transmogrifox]

I might also add that the 680R is very much intentional to recover the high frequency response of the filter. 

Notice R48=1.75k vs R48 vs 680R.  The filter naturally rolls off the high end, but the feedback R can be adjusted for a constant gain response at center frequency.

Below R48=1.75k

Vs R48 = 680R

[deadastronaut]

sounds really great, nice work man..

[Transmogrifox]

Thanks for the props.  I have my proximity sensor circuit working, and this circuit working...now to marry the 2 and end up with a finished proximity wah.  Right now it's just a mix of little proto boards loosely wired into circuitry on the breadboard and a copper clad board (for proximity plate) on a 1x4 board propped up on a 2x2.

I also got the active electronic bypass working with proximity sensor and volume levels balances so it comes on as you approach it and sounds natural to transition between clean and wah.  It will be fun to have it all in a final enclosure.

All it takes is for a kid to step in the middle of the breadboard and the joy is over 

[jatalahd]

This is a really elegant circuit and it also sounds good. Nice work! I joined this thread to get the latest updates, hopefully I will have time some day to actually build this and try it out.

A little off topic: The White follower stuff on this thread was new to me and got me interested on the subject. After some searching (...well anyone can find it, but anyway...) I found a thesis work from 1965, which analyses a "White emitter-follower", more directly related to the design by T-fox and it also includes a version where the feedback resistor has been used (p. 18 - 19):

http://trace.tennessee.edu/cgi/viewcontent.cgi?article=3833&context=utk_gradthes

There the circuit is drawn a bit differently, but more clearly for understanding it. When the feedback resistor is used it is just like the standard series-shunt feedback pair, where RE1 is taken infinitely large (taken out of the circuit). This considerably increases the loop-gain, which in this feedback mode directly down-scales the output impedance by a very decent amount (almost by a scale factor of 100).

Another not that meaningful off-topic note is that when injecting the filter at that point, it will not be a pure low-pass filter, but a sum of band-pass and low-pass. This might cause the low-frequency content to get dampened more, as is seen at least in some of the simulation graphs. But it is also possible that I made a mistake when analysing the filter part.

But as said, a really neat little circuit it is. Thanks for sharing it.

[Transmogrifox]

Another not that meaningful off-topic note is that when injecting the filter at that point, it will not be a pure low-pass filter, but a sum of band-pass and low-pass. This might cause the low-frequency content to get dampened more, as is seen at least in some of the simulation graphs.

What you see in the simulation graphs is due to the 1-pole input high-pass which was set to mimic the typical inductor wah response.  If you make the input capacitor much larger (like 10 nF) you see it's 0 dB response up to where the filter starts peaking.

It pretty much overlaps perfectly in simulation with an RLC lowpass network,  in which there is an R parallel with the L instead of series.  For example, a series RLC gives 40 dB/decade above the cutoff, but this filter gives 20 dB per decade as does an L||R series C lowpass, .... as does the classic inductor wah... and that is the reason for injecting this filter in this way.  It comes a lot closer to what the inductor wah does above cut-off than a pure 40 dB/decade LPF would do.

I suppose the 20 dB/decade (6 dB/octave) response may be rightly compared to bandpass behavior as a 2nd order bandpass does go to the 20 dB/decade slope.  A bandpass + low pass doesn't have such a flat response on the low end.  This is the response of an L||R series C low pass.

[amz-fx]

Great job on the circuit!

Here is an idea to try...  since the input impedance of the classic Cry Baby is much lower than what you have here, modify the input of this circuit to be similar.

Since we know the input Z of the CryB is about 120k, change R45 to 120k and then increase C19 to 0.068uF to let the lows through. I would add a 1uF capacitor from VbB to ground to filter the bias supply too.

This might upset the frequency balance that you have worked out for the circuit, but it is worth a try, IMO.

Best regards, Jack

[Transmogrifox]

Great job on the circuit!

Thank you for the encouragement.  It's been several iterations fiddling in Spice, building, then fiddling in Spice.  I have about 3 different iterations of this laying my circuit bones collection...and it has been a lot of fun 

Here is an idea to try...  since the input impedance of the classic Cry Baby is much lower than what you have here, modify the input of this circuit to be similar.

It would certainly be fun to try that.  I honestly have never played through an inductor wah that was not buffered.  Mine is (was) a GCB-95 with input buffer so it didn't occur to me to fiddle with the effect of loading the pickups in a manner similar to the classic wahs.

Building further on that idea one may extend it to using feedback to the input in which the loading becomes reactive as it is in the unbuffered Crybaby.  I don't have enough interest to take it that far myself, but maybe an idea for somebody reading this thread.

I would add a 1uF capacitor from VbB to ground to filter the bias supply too.

I agree.  Omission of this was laziness on my schematic drawing but I wasn't lazy when I made the prototype.  My prototype currently in my wah shell uses a 1 uF just as if I had read your mind before you commented (I have read enough amz notebook posts I probably did read your mind in some sense).  This of course brings the high pass cut-off just a little higher than what simulation plots indicate.

The end product is as quiet as the inductor wah, as far as I can tell.  My amp's high gain channel is composed of 3 cascaded 12AT7's so the gain is high, but still moderate compared to what it was before I replaced the 12AX7's.  That said, it takes max gain setting on 3 12AT7 stages to make the "swoosh" audible.

[thomasha]

So, I built this pedal and it sounds awesome!
I need to make a comparison between this and my colorsound just to show how it sounds better.
Palm muted playing sounds much better.

I made a layout for it to fit in a 1590a like my other wahwah. Check out>
https://www.4shared.com/office/NSk921Alca/Inductorless.html

here some pictures of it finished:



and a small soundsample>

cheers!

[Kipper4]

Love It tomasha.
I'm a paisley freak and proud.
I'm going to view the demo now.
Edit
Yep. Nice vocal filter voicing.
Is that switch gonna take a beating. Is the switch for bypassing?
Good job man.


Rich

[deadastronaut]

sickest wah ever.... love it. 

is that a clear plate under it?.

if so doesnt that affect the response slightly?.