Tune'o'Fuzz: My new silicon Fuzz Face with fully tuneable transistor gain!

[jasperoosthoek]

The last few days I've been trying something new (I hope): I've read somewhere that you can lower the gain of a transistor by adding a resistor somewhere (could not find that information back). But I don't know if anyone tried this before:

The gain of a transistor is normally defined to be some relation between the collector current and the base current (either DC or small signal AC). So if you increase the base current and keep the collector current the same the gain should be lower, right?

So what I did was: I connected a dummy diode parallel to the base and emitter of the (working) transistor. The base and emitter of a transistor are a diode and this just creates an alternative path for the current to flow making the total base current larger. In order to have a diode characteristic as similar as possible to the transistor I used a dummy transistor as a diode (Only using the base and the emitter). A trim pot tunes the amount off current through the dummy transistor. This one is important because current from the collector to the emitter of the working transistor makes the base go up a bit (Hybrid-pi model).

This is the circuit I came up with:

I connected a 6 pin IC socket to the B, C and Es of the transistors so I could tune the gain of the transistors with my DMM in the circuit. To be able to do that the switch jumpers have to be removed so the transistors are completely disconnected and the gain can be tuned.

When I started building I only had two 2n3904s with a gain of 280. With the trim pot in place the gain could be tuned between 0 and 160. The new 2n3904s I got from the electronics store had a somewhat lower gain of 190 which could be tuned between 0 and 105. A 50k trim pot would have allowed higher gains for this one.

Actually, I don't know if the Hfe values I chose by the DMM will still remain the same for the voltages in the Fuzz Face circuit as the collector-emitter voltage might be different from the measurement in the DMM. I couldn't check that as my DMM can only do one function at the same time. I cannot guarantee that these compound transistors really behave like a lower gain transistor...

So the results:
I tuned the first transistor to Hfe=70 and the second to Hfe=130 and the collector voltage of the second transistor to 4.5 volts. It sounded and responded very much like how a fuzz face should. Fuzzy and distorted. But the distortion was not completely what I had hoped for: It lacked the smoothness of my germanium FF. Maybe I'll install a 'smooth' control http://fuzzcentral.ssguitar.com/axisface.php as in other silicon fuzz faces. Apparently even a fuzz face with silicon transistors with real low gains is still not completely smooth. My Tune'o'Fuzz oscillated with a wah in front, something my Ge FF doesn't do (that one only screams).

So I started playing around with the trim potmeters, changing to different gain settings and listening to the guitar notes played through it. There are a lot of different  sounds that you can get out of this circuit! Very interesting. But I couldn't dial into a real germanium type sound. It could be that I need real low gain silicon transistors for that but more likely the reason is that the non-linear distortion characteristics of germanium is just different: That's where the FF sound comes from, not the linear part. Different material, just a different way electrons behave. This can also be done with high gain germanium transistors, maybe you need different trim potmeters.

Anyhow, this was a very good learning experience for me. I hope you guys find this interesting and maybe try out this circuit too .

EDIT:
Did I just re-invent the wheel? http://www.diystompboxes.com/smfforum/index.php?topic=21245.msg132860

[azrael]

Haha, yes indeed. Looks like a good experience, though.

Try small caps across the B-C junction. Like maybe 47-220 pF. This should soften clipping a bit, make it more like a Ge.

[Derringer]

EDIT:
Did I just re-invent the wheel? http://www.diystompboxes.com/smfforum/index.php?topic=21245.msg132860

yes, but sometimes something new is learned with every re-invention

so ... nice job and thanks for sharing!

when you adjust the gain of the trannies, make sure you recalibrate the bias on the C of Q2 ... it will change as the tranny gain is altered

[jasperoosthoek]

I couldn't imagine that it had never been tried. Nice that I spotted some problems myself that were also in the other thread : Can you compare Hfe measured with a DMM to Hfe in the circuit for instance?

I've already have an idea for a new Tune'o'Fuzz schematic where you can choose the gain at the correct operating levels (CBE voltages and base currents). That might be a used as a tunable Fuzz platform with digital panel meters inside. So, flick the switch and simply choose Hfe, or change the dials to hear what sounds right and have an instant indication of the values. But I don't think I will be building such an elaborate thing myself.

[amptramp]

The nice thing about this circuit is that the Vbe voltages of the amplifying transistors and the diode-connected transistors track over temperature.  Typically, hfe rises with temperature, but this circuit may stabilize that.  If you connect the collector to base of the diode-connected transistors, it could be even better!

[jrod]

Great experiment, man! Looks it was probably a lot of fun to play around with and great learning experience.

I do have a question: Why use a transistor as a diode and not just a diode?

[jasperoosthoek]

That should work in theory, I even tried it. The problem is that the IV curves do not match: The real diode will start to conduct at much lover voltages. This means you need a very large resistor to match them. The result is that the large signal HFE (IceDC/IbeDC) will be smaller but the small signal Hfe (IceAC/IbeAC) will not be affected by the diode. I'm still starting to understand more about this circuit. Maybe the way I check the gain of these transistors it wrong.Also the parallel transistor will result in a lower input impedance. Adding a series resistor helps, just like the Smooth control in the axis (link above). That does not change gain at all but does change the sound.

[jrod]

Oh, I see! Thank you.

[Caferacernoc]

Haha, yes indeed. Looks like a good experience, though.

Try small caps across the B-C junction. Like maybe 47-220 pF. This should soften clipping a bit, make it more like a Ge.

I hope you try the small caps and report back.

[jasperoosthoek]

I'll give it and the smooth pot a try soon! I had an idea to make 'tracking' better: Add a fixed series resistor to the base and connect the trim pot to the other side of that resistor. That way both trannies have their own series resistor that helps with making their characteristics similar (tracking): making the diode iv curve less steep.
Knowing the input impedance (r-BE) of silicon and germanium would help a lot. I want to keep that resistor fixed so the total input impedance is as similar to germanium as possible: A fixed smooth pot.
The problem now is that even with low gain the input impedance is too low causing oscillation () and ugly clipping at the input.

[PRR]

> 2n3904s with a gain of 280. ...gain could be tuned between 0 and 160.
> The new 2n3904s ....gain of 190 which could be tuned between 0 and 105.

At what current??

"Whatever the multi-meter does" is not an all-purpose answer.

We could also ask about voltage, but for Vce from 0.2V to 20V the hFE will not change much.

>  I cannot guarantee that these compound transistors really behave like a lower gain transistor...

They don't.

I had my doubts. While, as Ron says, the tracking is a nice thing, basically this is NOT the "same" as a low-gain transistor. It is a transistor with a huge base-emitter leak. (Which is quite different from hot Germanium where the collector-base leak is overwhelming.) I did not believe this would be "same effect" at ALL currents.... but I ran out of fingers to think on.

So I asked my idiot to plot it:



The transistor current is forced to vary from 10uA to 10mA, the usual zone where we work 2N3904s. The base current is measured. The "hFE" is computed and plotted.

The top curve is "without the Jasper diode", plain 2N3904. This reminds us that a transistor's hFE is NOT a "constant", that ti varies with current. A moment's thought will show that it must be zero gain at zero current, and it probably falls off at very high current. For this particular (modeled) '3904, hFE varies from 70 at 10uA to 170 at 10mA.

A multi-meter may read "hFE" at 1mA, or 100uA, or whatever was handy internally. Or it may feed a 1uA base current and read collector current: real easy but bogus.

While the stock '3904's hFE varies a little with current, the diode-resistor mod's hFE varies a LOT with current. In a linear amplifier, current should not change a lot. And in overdrive, the peak current probably does not increase a lot. But the minimum current goes FAR lower than the idle current. Where a real transistor's hFE will fall slightly, this hacked transistor's hFE will fall a LOT. That must have some impact on "how it distorts". I'm not going to ponder that tonight.

The bottom curve is with the resistor set nearly-zero. The hFE will not go to "0" as reported on your meter. Depending on different things, but assuming somewhat-matched (from the same bag) transistors, hFE will be 1 or slightly higher.

There is a mistake in the plot. I computed "hFE" as Ie/Ib. We usually compute Ic/Ib. Ie is Ib higher than Ic. When hFE is like 100, the difference is 1%. When hFE is really 1, this mistake says "2". So subtract 1 from everything.

With Jasper's open-collector connection, minimum hFE is about 1.7. When connected Trans-Diode as Ron suggests, the minimum hFE is just about 1. These results could be expected from a study of current-mirrors.

What you really want is a current mirror with an adjustable ratio. If you know your ratio and have a chip-factory, and want hFE=70, you make 71 identical transistors, use one for the diode and parallel the other 70 for the output. (Although such a large ratio raises other problems.) Widlar would know a better way but he isn't talking. There's "exact" methods but so complex that they spoil the whole idea of a "simple fuzz".

The other path: do you have a free variable? The input current is fixed: this plan depends on pickup loading. But all other currents are "free", may be varied as you like (withing reason). With higher hFE, the first collector current can/should be higher, the second collector current even higher. Taking numbers above, (270*270)/(70*130)= 8.7 times higher. Your typical 5.6K resistor becomes 640 ohms. That does not matter, the output impedance is unimportant. The 33K becomes more like 8K. The battery dies 8.7 times sooner, which is a problem often made moot by wall-wart power. The 1K reverse-audio pot becomes 115 ohms rev-audio, a very distressing value/part.

[jasperoosthoek]

First of all, thanks for your elaborate answer!

At what current??

"Whatever the multi-meter does" is not an all-purpose answer.

I couldn't agree more. My biggest worry also.

We could also ask about voltage, but for Vce from 0.2V to 20V the hFE will not change much.

Well, if it doesn't behave like a real transistor than hFE will change a lot, as you showed in the similation.

I had my doubts. While, as Ron says, the tracking is a nice thing, basically this is NOT the "same" as a low-gain transistor. It is a transistor with a huge base-emitter leak. (Which is quite different from hot Germanium where the collector-base leak is overwhelming.) I did not believe this would be "same effect" at ALL currents.... but I ran out of fingers to think on.

It is indeed a transistor with a huge base-emitter leak. But if it mirrors the current you should be alright. A big concern I have is the dependence of the base voltage on the collector voltage and current: there is no such thing in the second (piggyback) transistor.

So I asked my idiot to plot it:

The transistor current is forced to vary from 10uA to 10mA, the usual zone where we work 2N3904s. The base current is measured. The "hFE" is computed and plotted.

Thanks for that! I've been thinking, I might be improved a lot if you add a series resistor to the base of the working resistor. That will probably help a lot with tracking. If I'm correct it will make the IV curve of the working transistor less steep.

The top curve is "without the Jasper diode", plain 2N3904. This reminds us that a transistor's hFE is NOT a "constant", that ti varies with current. A moment's thought will show that it must be zero gain at zero current, and it probably falls off at very high current. For this particular (modeled) '3904, hFE varies from 70 at 10uA to 170 at 10mA.

Isn't the first transistor biassed at about 2uA?

The other path: do you have a free variable? The input current is fixed: this plan depends on pickup loading. But all other currents are "free", may be varied as you like (withing reason). With higher hFE, the first collector current can/should be higher, the second collector current even higher. Taking numbers above, (270*270)/(70*130)= 8.7 times higher. Your typical 5.6K resistor becomes 640 ohms. That does not matter, the output impedance is unimportant. The 33K becomes more like 8K. The battery dies 8.7 times sooner, which is a problem often made moot by wall-wart power. The 1K reverse-audio pot becomes 115 ohms rev-audio, a very distressing value/part.

Good idea 

Do you know the value of the input impedance of a germanium transistor and a 2n3904 at 2uA. Maybe just an in the ball park value? I could use it as a fixed resistor in front of the working base to increase tracking.

[jasperoosthoek]

So I did an input impedance measurement. Just a very simple measurement with a sine wave generator set to 100mV and a 3.3k resistor connected between the output of the signal generator and the input of the fuzz. I adjusted the trim pots on the piggy back transistor to the maximum value to turn them off and set the collector of Q2 to 4.5 volts.

I checked the drop in signal with Fuzz knob turned to the left and to the right. Then I repeated the measurement for my germanium FF. These are the input impedances:

Si: Fuzz turned down: 3.9k (54mV)  Fuzz turned up 8.9k (73mV)
Ge: Fuzz turned down: 5.9k (64mV)  Fuzz turned up 8.9k (73mV)

I was kind of surprised that the input impedance of the Si and Ge fuzz (turned up) was completely the same! The lower input impedance of the Si fuzz with the Fuzz knob turned down just means that the transistors with more gain result in more feedback. Also I checked it for different frequencies. The input impedance became a bit bigger at 10kHz but nothing significant.

To me this means that the input impedance is not causing the change in sound.

[brett]

Hi

Where a real transistor's hFE will fall slightly, this hacked transistor's hFE will fall a LOT.

Great analysis ! Thanks for the graph    
The first piggyback pedal (somewhere in the shed now) had tiny 1k piggyback resistor in both pairs.  Results were low hFE, low input impedance, but a very nice fuzzy distortion.  At the time, RG Keen and I wondered how low and high the dynamic hFE was swinging and whether this might have been part of the excellent tone.  (We observed that Ge devices swing all over the shop, too).  Clearly, the hFE "knee" is more like a ski slope for piggybacked devices.

To me this means that the input impedance is not causing the change in sound.

Do you mean: not causing the difference from Ge?  That makes sense.  But there would be a big difference from non-piggybacked Si (with input Z much higher at about  40k or 50k and almost eliminating pickup load and bloom).

thanks again.  Get piggybackin' !

[brett]

Hi
another piggyback application to "throw" into the piggybacking story.  Posted some time ago.  Input section built and tested, and output section built and tested, but not together (yet).

It's a variable-pigged, tone-controlled input fed to a variable-gain differential distortion.  Build it, mod it, love it, hate it, sell it, patent it - whatever.

http://www.aronnelson.com/gallery/main.php/v/BinOfBrett/Fuzz+with+no+name+schematic.jpg.html
cheers

[jasperoosthoek]

Do you mean: not causing the difference from Ge?

That's what I meant.

Did you read my previous post? The input impedance was small, in the 10k region but very similar to Ge. Of course it will be voltage dependent.

The way the input is connected directly to the base means that the input impedance will decrease for positive voltage swings (for an NPN transistor). This might cause asymmetric clipping which could be harder or different for silicon than for germanium. It might be even lower when you add a piggy back transistor. A scope measurement might to check that. I could do that another time.

I love the Fuzz you just posted! Seems like you can tweak it a lot with all the controls. "Piggy-ness" is a good one. For your eyes only this video where Gary Busey explains that piggy backing can also be done with goats: http://www.youtube.com/watch?v=ZMURD17Jeok

[Gus]

I have been posting for years part of the FF is the imperfect input summing node. If it was perfect the input would be 0 ohms.  The input resistance is reduced by the feedback/bias resistor
Look at an inverting opamp circuit
The FF Q2 emitter is in phase with the Q1 collector shifted by the VBE drop.  The feedback/Q1bias resistor from Q2 emitter Q1 base is feedback that REDUCES the input resistance, if the transistor had gain like an opamp it would reduce it to close to 0 ohms.  Base of Q1 -input emitter of Q2

Think of the first gain stage of a FF type as an limited openloop gain opamp or read about feedback theory

This is another reason to select Q1.  If you look at the distortion FF like circuit I like, note the 10K and 100 ohm resistors and then work out the open loop Q1 gain at the operating points.

[brett]

Hi
thanks Gus.  So many of us focus on particular features, and forget to be holistic.  Like many, I am guilty of underestimating the feedback effect on input impedance and the dynamic hFE.  In fact, I've never done the maths or simulated the changes in feedback and Zin with changing nominal hFE of Q1.  Being rather lazy, I've assumed that Zin is proportional to the hFE of Q1 (and that the feedback effect is small - maybe 10 to 20% reduction in Zin ??). 

Would someone now care to do the simulations (input = 100mV p-p, 880 Hz) for a FF with Q1 of either a Ge (hFE=70), Si (hFE=200) or piggybacked Si (hFE=200 each, Rpiggy = 10k).  What is the Zin? (step Zsource =2k to 12k, step =2k and find the half voltage point). 
Fix Zsource = 10k (or whatever an average pickup is).  What is the average voltage gain? What is the minimum and maximum of dynamic gain (-(Vout-Vaverage)/Vin)?  I *think* these are at the mnimum and maximum input voltages, respectively (PNP, positive ground circuit).  In any case, one will be a big number and the other small.
cheers   

[jasperoosthoek]

I hope it will be similar to my measurements of input impedance. I measured that the input impedance dropped more than 50% due to the feedback loop. It went from 8.9k (no feedback) to 3.9k (full feedback). That was for the silicon FF with 20k piggy back transistors. hFE from DMM was 190 for the two Q1s and 280 for the two Q2s. Also measured at 100mVAC at a very wide range of frequencies (~500 to 5kHz).
The Ge FF only dropped about 34%. Reasonable but not huge difference. The FF does indeed load the pickup a lot: 6k to 9k depending on the Fuzz/feedback setting. That explains the dull sound you get when you turn down the guitar volume just a notch from being maxed out.