ESR Graphic Fuzz Power Supply

[antonis]

You may consider 100nF + 220R in parallel with wiper-lug1 resistance..

For wiper at 50%, at 1k6Hz effective wiper-lug1 resistance is about 980R (and NOT 5k) where at 160Hz is about 3k4.. (more close to 5k)..
It is clear for very high frequencies actual Pot setting should be considered negligible..
(almost zero output volume variation..)

[Rob Strand]

Yes, it's a 10k lin pot. Maybe I will change it to a log one for being able to dial in lower volumes.

A log pot would help.

Doesn't mean loading down the op amp results in less gain? Maybe what I hear is the low frequencies now have less gain than when the op amp only sees the 100n + 220R network and a relatively high impedance in parallel. Although I would expect the low frequencies to distort less instead of being less present.

But if thats true the volume pot should have a great influence on the bass content. I would try that, but unfortunately I haven't found a way to do that yet without busting my ears.

When we are talking small effects it can be quite hard to narrow down the cause.    Thing we normally consider have no effect may have a small effect.  In isolation those effects can be difficult to pickup and it's only when you do detail A/B testing that is becomes clear there is a subtle effect.

Opamps aren't like transistor stages you *really* have to load them down to have an effect.    Something like the 100n+220R network where the load is 100's of ohms to k's of ohms we might expect something.   

The other effect you might be experiencing is the level pot setting in the original and modified circuits is difficult to set to the same *equivalent* position.    it's actually harder than you think.

If we wind back a bit to more tangible engineering causes.   If you are using a tube amp the output impedance of the pedal interacts with the overloading preamp tube.    Keep in mind that pedal can push out about +/-4V.   As far as making the modded volume control work like it did before,  perhaps try adding a 10k between the pot wiper and the output jack.   That makes the output impedance more like the original.

Another thing you can do is add place 2.2k to 10k from the *opamp* output to ground.   A switch to switch the resistor in would be ideal here so you can easily do AB testing.    If you can't hear a difference then resistive loading isn't the cause.

A far more repeatable approach would be to put say 22k between the level pot wiper and the output jack, similar to the previous mod, then simply add a small capacitance across the output jack and ground to take out the fizz.   Start at say 100pF and increase the cap until you get to tone you want.

[pokus]

Another thing you can do is add place 2.2k to 10k from the *opamp* output to ground.   A switch to switch the resistor in would be ideal here so you can easily do AB testing.    If you can't hear a difference then resistive loading isn't the cause.

I put in a 470R right after the op-amp with the volume pot in the lug3 to output cap(not the effect out) setting. Now I have that brighter sound the original pot wiring was creating, but I'm also able to control the volume of it a lot better. Making that resistor value higher, the sound comes closer to what I experienced without it and reaches no-difference at about 10k.

So the op amp loading seems to have a huge effect in this circuit. The volume pot in the Graphic Fuzz therefore shouldn't be (intended or not) just controlling the volume. Maybe this is also very op-amp dependent. Would the addable offset-null-control do anything here?
I still don't know which exact sound I like better. The original one has a lot of bite and sounds like that early raw vintage fuzzes, at least at quite low volume regions. But also kind of harsh and not that great when multiple strings are in action. The different volume pot wiring definitely smooths things up a bit and brings back bass content. Maybe with the setting of the resistor after the op-amp you can have/dial in best of both worlds at a reasonable volume.

[Rob Strand]

So the op amp loading seems to have a huge effect in this circuit.

That must be it.  You did well persevering with that one.  Certainly not clear what was going on.

If the sound match is with 470R, you must have had the output Level pot set quite low.

Looking at original circuit, for the Level control to work it relies entirely on loading down the opamp.

Trying to make more sense out of it,  if you look at some older Fairchild and National Semiconductor datasheet, they have current limit info and an internal circuit,

https://www.alldatasheet.com/view.jsp?Searchword=LM741MH

The current limit is typically 25mA (10mA to 35mA), so with 470 ohms, 25mA*470 = 11V.   That's more than the 4.5V swing.  So it's not even current limit that's changing the sound.   Nonetheless it's not far off so it's likely the internal drive currents are getting to the point where they are running out of steam and are just on the edge with the heavy load.   In other words the outright loading is changing the sound.

[pokus]

Thank you guys for leading me through this.
I think i will stick with the "wrong" volume pot wiring as the tone can be better altered elsewhere in the circuit.
Overall the Graphic Fuzz stays a very weird but also great sounding pedal to me.

[pokus]

It still keeps getting weird 
I boxed this thing up (veroboard). On my breadboard the oscillation was really subtle, shining through a bit but mainly coloring the tone.
Now it's really wildly oscillating whenever it's on. Really don't think that all of this is coming from the wires.

My voltage readings of the 741 in the box (on breadboard) are:
0V (0V) at the non-inverting input, as it is tied to gnd.
0.7mV (1.4mV) at inverting pin
50mV (12mV) at output pin
The other readings are the same. About +4.7V on +pin and -4.7V on -pin. Also -4.7V on both null pins.


Tried the null control for the 741 by connecting pin1 and 5 with a 10k pot to V-. The best result, in regards to the oscillation, I'm getting is when pin1 is tied directly to V- through that pot. But it's still very present. The voltage on the output pin than drops to about 35mV, on the inverting pin it rises to 6.7mV.
It also seems like it's a lower oscillating frequency then. When pin5 is connected to V- it's a higher frequency.
Tried this with different chips, but the sound on breadboard always is subtle and in the box it's very present. It's hard to tell, but apart from that I don't think there's any difference in the sound.

Anything that could be done to fix this? Or any idea why this occurs?

[Rob Strand]

Anything that could be done to fix this? Or any idea why this occurs?

Perhaps the place to start would be to increase the 2x100nF caps across +/-4.5V to say 100uF - maybe add the 100uFs across the 100nFs temporarily.

The cause could be the 100n + 220ohm (and any load resistors) on the output is feeding into the ground.   It's a low impedance and is the most likely cause of trouble, especially with the relatively light 2x100nF supply caps.   The precise ground wiring and the wiring of the 100nF input cap could tip it over the edge.

Some other things to try would be:
- Add a small resistor say 1k between the opamp - input and the 100nF cap.  You could also try 1k between the input jack and the 100nF cap.   Resistor values may need to be tuned to find a balance between no oscillation and affecting the sound.   
- another possibility is to add a 220 ohm *after* the 470 ohm load you added and before the volume pot.
- and another is to add a small cap across the 1M pot, perhaps 22pF to 100pF ; many need to juggle values for sound vs oscillation trade-off.

[pokus]

The 100uF caps somehow cut my tone completely. 200nFs in parallel with the 100nFs really lowered the oscillation a bit, but also the frequency changed there and the overall sound was a lot duller. (on my breadboard the 100uF caps added oscillation  )
That's why I focused on the feedback cap across the filter pot. It's really a sound vs. oscillation trade-off like you mentioned. I found it best with a small 10pF cap, which cuts the oscillation almost completely, but keeps the nature of those high frequencies and the filter control. Keeping pin1 tied to V- also helps.
As it's already on veroboard I haven't managed to try the other options with some added resistors yet.
I left out the 470 ohm resistor, so what the 741's output sees is just the 100n+220R network and the 10k pot (wired as 10k path to gnd).

[antonis]

The 100uF caps somehow cut my tone completely.

 

Did you place them as Rob suggested..??
(across +/- 4.5V)

[Rob Strand]

The 100uF caps somehow cut my tone completely.

 

Indeed.

The 100uF caps somehow cut my tone completely. 200nFs in parallel with the 100nFs really lowered the oscillation a bit, but also the frequency changed there and the overall sound was a lot duller. (on my breadboard the 100uF caps added oscillation  )

Those results show we are dealing with a crazy beast.   It's like it relies on the ground flapping about relative to the power rails.    Very hard to understand and even harder to fix without affecting the tone.   Normal thinking probably isn't going to get us to a solution.

So maybe we need to try a sightly different cap scheme.    Try putting a single cap across the power rails of the opamp, perhaps 10n, 100n, 10u, 100u.

As it's already on veroboard I haven't managed to try the other options with some added resistors yet.

We might have to try all the options (or combinations of option) since it's not looking easy to fix.

[antonis]

Rob, I think we're confused a bit..

It seems to me that OP refers on 100nF output cap (the one in series with 220R) when you point on supply rails decoupling..

[Rob Strand]

It seems to me that OP refers on 100nF output cap (the one in series with 220R) when you point on supply rails decoupling..

I didn't think of that.

Pokus's post has "100nFs" with an 's' so maybe he did put the 100uF's across the two 100n's on the power rails (for clarity the 100n's across the 2x10k's).

Maybe he can confirm.

[pokus]

I put one 100uF from V-(-4.5V) to ground, in parallel with the 100n and another one from V+(+4.5V) to ground, also in parallel with the already existing 100nF.
It also felt weird to me that it completely cut my tone. No difference if there's one(from V+ or V- to gnd) or two( from V+ to gnd and V- to gnd). Greatly smaller values, like the 200n I mentioned, are only cutting the highs. So maybe that's why the 100uF cuts the bass, too (although I got no plan why it does so here  ). With 1u there's still a little bass content left.

The opamp output still sees the 220R + 100n. I did put an LED with a resistor across V- and V+, but that was after I had experienced the tone cut - at least I think so. As with the small feedback cap it sounds and behaves like my breadboard build, I don't think there's anything wrong there. Apart from being that weird.

[antonis]

Just to make things clear..

Placing a single cap between + & - bipolar supply rails instead of one cap between each rail and GND is often desirable to prevent each of two caps injectingrail noise to ground.. 

Try to delete both +/- 4.5 to GND caps and put a single cap between supply rails (with GND left uncoupled..)

P.S.
More a more "sensible" gain circuit, connect IN to 1M pot "open" lug and cut the connection between 100nF cap and pot wiper..
(of course, for max gain setting, signal will still "see" zero impedance so Rob's suggestion for a 1k or so input resistor is always advisable..)                                                                                           
                                                                                                                                                   

[Rob Strand]

I've been trying to decipher the Graphic Fuzz in order to come up with some more ideas stabilizing the circuit.

So, up front, I don't understand why the bass is cut with the big caps.   It could be some non-linear effect but it's hard to visualize at the moment.

So here's what I've come up with so far.

Some circuit bending trying to understand what it's doing,  linear case only,
 


From that we can see how the load affects the response.   I guess that something     It does not explain the bass cut with big caps.




So as far as stabilizing the circuit is concerned.   The small feedback cap across the 1MEG still holds.   However, perhaps there's a better scheme.    Perhaps adding a 470R or 1k in series with the 100n cap at the input of the opamp will help stability.     Maybe need to experiment.   The way I see it is the 100n is adding lag to the feedback.   By putting a resistor in series with the cap the phase shift at high frequencies is removed and that should help stability.

Given this is a rough model and totally I'm ignoring opamp common mode effects, I wouldn't be surprised if the resistor made things worse.   Nonetheless we have to start somewhere since we aren't dealing with a normal circuit!


EDIT:
The "circuit bending" part seems OK superficially but it might not be correct.   The reason is i've tried to keep things at a high level bt representing to opamp supply path as a voltage source.   In reality it probably acts as a current source which changes the required "circuit bending".

[Rob Strand]

From what I can see that pedal was a real pedal,



I can see three 10k's on the PCB.   

My first guess is extra 10k is simply in series with 1M pot.  However, in reply #4 of this thread there is an extra 10k between the opamp+220R+100n and the 10k pot,
https://www.diystompboxes.com/smfforum/index.php?topic=83300.0

(Knowing what we know now and the fact there is a PCB pic,  PRR's mod in reply #7 are probably changing the intent of the pedal.)


Schematic credited to Jerms (2009_03_21),


Confirmed correct by retracing PCB.
A few details,
- IC = 741CP.
- 1M pot is called Filter, probably A taper,  clock-wise = high resistance
- 10k pot is called Boost, probably A taper
- all caps disc ceramics
- Input socket is a square plastic type with a separate set of contacts to switch the battery 0V ("-4.5V")

Perhaps we can trace it from this bottom shot,


EDIT:  I retraced the board and confirmed the jerms schematic is correct.
EDIT: The difference between this correct schematic and the one at the top of the thread is the 10k between the opamp output + 220R+100n and the output level pot (Boost pot).

[Rob Strand]

Here's a few more points:

As far as fixing the oscillation I'm pretty sure adding resistor in series with the 100nF input cap to ground is the way to go.    There is some engineering sense behind this instead of magic and mystery.

There is a minimum value for the resistor which reliably fends off oscillations.  The minimum is around 47R to 100R but you need to check both 470ohm loads and 10k loads at the output.   A higher value around 470R seems more reliable and will prevent oscillation independent of the load.  With very low values you might find the oscillation stops but if you change the load resistance the oscillation might kick-off again.  That means you need to increase the value of the series resistor.

The peaks in the response (see plots a couple of posts back) are affected by the load resistance.   The resistor in series with the 100nF cap also affects the peaks.      You may need to delicately adjust the value to find a balance between oscillation and tone.

So here's the problem:   I'm fairly certain the presence of the peak itself is an indication the circuit is close to oscillation.  If you are using those peaks to EQ your tone then you are treading a very fine edge between oscillation and getting the sound you want.  There seems no way out of this.       The peaks are around 3kHz so that means they will be quite noticeable.   A better design is perhaps to setup the circuit to have no tendency to oscillate then add a post EQ to get the EQ profile you like.

As far as my "circuit bending" stuff goes.  I tried a few other ideas and they give similar types of behaviour.    Even if the analysis so far doesn't quite represent reality  I do have a feeling the closeness to oscillation vs EQ is coming into play.

[ElectricDruid]

I appreciate your determination, all.

I get annoyed with circuits like this. It worked once, for someone, somewhere, with some chip they had in a drawer. And then they published it and tens/hundreds/thousands of other people have wasted hours trying to get the bl00dy thing working ever since. That's a badly designed circuit in my view, and I lose patience with that.
 
What you're doing is the basic design work that the designer of this circuit should have done originally. It's quite possible, given the surfeit of talent on this forum, that something good will come out of it. But it won't give any credit to the original designer, and it still p1sses me off that you have to bother.

Sorry, I'll stop moaning now...

[Rob Strand]

I get annoyed with circuits like this. It worked once, for someone, somewhere, with some chip they had in a drawer. And then they published it and tens/hundreds/thousands of other people have wasted hours trying to get the bl00dy thing working ever since. That's a badly designed circuit in my view, and I lose patience with that.
 

Agreed, the one day it worked circuits are a total pain,  especially when they have design problems.   (One of my favourite oscillation cases on the forum was the oscillating SHO.    The lead inductance to the CRACKLE pot causes the MOSFET to oscillate with some weird symptoms - solution is posted somewhere on the forum.
http://www.luciferstrip.com/fuzz/sho-schematic.jpg )

FWIW, Earth Sound Research weren't known for their engineering.   A large section of the business was copying Peavey amps, right down to the PCB in some cases.

After pondering the resistor in series with the input cap, it also occurred to me the guitar cable capacitance at the input jack could set the things off oscillating in the same way the 100n input caps does.  Maybe a resistor from IC pin 2 + 100n input cap  to the input jack is wise as well.   100R might do it perhaps upto 1k.  1K is better but it might change the sound so the 100R could be a compromise.   I mentioned this somewhere above but now it seems like it would be a wise addition.

[pokus]

Placing a single cap between + & - bipolar supply rails instead of one cap between each rail and GND is often desirable to prevent each of two caps injectingrail noise to ground..

Try to delete both +/- 4.5 to GND caps and put a single cap between supply rails (with GND left uncoupled..)

I put a 100uF cap from V- to V+. All that changed was a reduction of some hissing noise, not affecting the oscillation. That's also great, as the circuit was a bit noisy when rolling back the guitars volume and increasing the boost/volume pot on the pedal.

I can see three 10k's on the PCB.   

My first guess is extra 10k is simply in series with 1M pot.  However, in reply #4 of this thread there is an extra 10k between the opamp+220R+100n and the 10k pot

That makes perfect sense. The volume pot therefore shouldn't be affecting the sound in the original circuit, other than what I've experienced with the original volume pot wiring right behind the opamp. Don't know why this is missing at the schematic I posted at the beginning.

As far as fixing the oscillation I'm pretty sure adding resistor in series with the 100nF input cap to ground is the way to go.    There is some engineering sense behind this instead of magic and mystery.

Just to be clear, you suggest a small value resistor from the 100n cap at the input to gnd, just like the network on the opamps output?
It should be very easy to try this on my breadboard build, but there's almost no oscillation (whyever?!), so it could be hard to hear an effect.
As Rob said, normal thinking maybe won't get us to a solution, I made my breadboard build oscillate again with the 100uF from V+ to gnd (again: whyever?!)   
Now I connected the 100n input cap to a 470 ohm resistor and then to gnd.     Oscillation is gone!     It stays unhearable until about 100R, then comes in gradually (I'm not stacked with every value). There might be a small effect on the frequency response, but really nothing I could define exactly. I didn't change the load, but with the 10k in the real original circuit it should always be 10k minimum.
Great work and a big thanks for your effort on this one.


While we're at it. When my guitars volume is down completely, there's a lot going on for an undefinable time. Crackles and that kind of stuff. Not super loud, but you can hear it. That's on both, the breadboard and the boxed one.