Circuit analysis help: Experimental fuzz!

[tatou]

I've been breadboarding an experimental fuzz, combining some ideas from the fuzz face, muff fuzz, bazz fuss, and even the big muff.

Here is the current status of what I've hacked together (my "question areas" are boxed in red):


Quick overview:
* It's essentially a fuzz face topology (with tweaked values) running into a modified AMZ presence tone control
* The fuzz is controlled by an input pot rather than the normal fuzz control (e.g., R5 as a pot)... I found this to sound better & be more reliable
* A switch toggles between a single LED in the Q2 feedback loop vs. silicon hard clipping
* The amplifier output comes directly from the Q2 collector, unlike a typical fuzz face

It's sounding great, but I still have a few questions about how parts of the circuit work, and hope some of you can help me understand my own frankenstein a bit better!!

D1: WHY DOES ONLY ONE DIODE DO ANYTHING?
At first, I wired up two LEDs in a symmetrical typical clipping fashion, as in the big muff, etc. But I noticed only one LED lit up. With further experimentation I realized only one diode going from collector to base on Q2 appeared to do anything at all. The other direction simply has no perceivable effect. This is how things are wired in the bazz fuss, but I suspect you can actually still use symmetrical clipping in a single-transistor circuit like that (e.g., I've seen it in the muffer or 5th gear overdrive). But I'm wondering if there's something particular going on with the coupling of transistors in the FF topology (rather than having two distinct "stages"? Could it be that the feedback loop via R3 is somehow affecting the voltage swing in such a way that only one diode ever matters? If so, is this not a clipping diode in the conventional sense? It really does sound different... and the hard-clipping diodes to ground work in both directions. What gives?

Regardless, that single LED sounds great. It's a harsh, sputtering bumblebee-type fuzz, whereas toggling to the hard clipping network has a creamier distortion feel (and no diodes at all is a beefier, more open and uncompressed fuzz, and of course more output). The tone controls help balance out some of the tonal differences between those three "clipping" options.

R5: THIS SEEMS TO BE THE ONLY WAY TO GET A GOOD SOUND OUT OF THIS CIRCUIT
There's no equivalent to this resistor in any other schematic I've seen... but if you jumper it and go straight from the Q2 emitter to ground via R8 and C2, the pedal sounds... starved. Like the battery is nearly dead or something. The fact that R5 is even there is a happy accident, a breadboarding vestige from experiments getting a fuzz control to sound good in this part of the circuit. If any of you have any thoughts, please let me know!

D2-D5: SUBLE MULTIPLE CLIPPING DIODE QUESTION
I drew things here on the schematic as I accidentally breadboarded it up... with the anodes of D2/D5 touching the cathodes of D3/D4. In this way, it seems like some current can skip D3/D5 without going all the way to ground, but then I couldn't hear any difference when I "fixed" it (D4->D5 and D3->D2 directly, with no point touching in between). The only other place I've seen this is in some (but not all) Timmy schematics floating around. Does it matter to have this connection point between diode pairs in opposite directions? Why or why not?

C10: HOW DO YOU REALLY CUT HISS?
Adding this cap to ground seems to help cut a little bit of high-end hiss/noise out of the circuit (although probably not even enough to bother with the parts count)... However, it doesn't appear to me to be a proper low-pass filter. Any thoughts on why this works or what I should be doing instead are welcome.

The rest of the circuit I feel like I understand pretty well. It's a little low on output... if you dial the fuzz & tone knobs back and crank the volume to max, the effect is on par with unity bypass volume. With everything cranked there's quite a bit more headroom. Based on my research, a fuzz face adds 18db of gain and a BMP-style tone stack attenuates 7-15db of that, and with diodes on that cuts into things a bit more.

Any other thoughts you have about this circuit and how to improve it are welcome. So far it's proven to be a fun, versatile little beast!

Thanks!
(And apologies for cross-posting to other boards)

[highwater]

D1: The other diode is built-in to Q2, as the PN junction between the base and collector. Do you have any germanium diodes? They would turn-on at a lower voltage than the one in the transistor, and might make a difference (or might screw-up the biasing, I'm not sure... easier to try it than think about it IMHO. Either way, you won't let any magic smoke out).

R5: It's essentially the same as having a FF-style pot that's not *quite* turned all the way up. It's possible that the thing is oscillating without it, or maybe it just doesn't sound any good at full blast.

D2-D4: The only possible difference I can think of with vs. without the link is that the effective capacitance to ground from the inactive diodes might be slightly difference... but it's doubtful, and even less likely to be enough to matter. Also, with the link, you can use an SPST switch to short-out one pair of diodes to give two levels of clipping, without the link you'd need a DPST switch to do that... which is probably why some of those Timmy schematics ended-up with it.

C10: That's still a proper low-pass filter. The only problem is that the 'R' part of "RC filter" is the output impedance of the FF stage... which means that the math is more complicated if you need to select a precise cutoff frequency (meh, just try different values until it sounds good), and the cutoff can change if you make changes to the preceding stage (not a problem in this usage).

[tatou]

That was fast! Thanks, highwater! These explanations are very helpful.

I completely forgot that transistors can be wired as diodes. I measured the B->C forward voltage to be v_f = .75 on a 2n5088 transistor, whereas LEDs are about 1.6 so it makes sense that they wouldn't produce any clipping in that direction. I tried a Ge diode as you suggested (v_f = .45), and that time you could hear the difference! I also tried a Si diode (v_f = .62) and you could also hear it, although less pronounced.

I also tried the Si diode instead of an LED for D1, and it produced no sound until you turn up the fuzz knob about 1/3 of the way... and then it's a sputtery buzzsaw sound. I guess I just got lucky picking an LED at first and liking how it sounded, tailoring the various resistor values around it to create the "bumblebee" sound.

Also interestingly: 2n5088/2n5089 both sound really good in this circuit, but a 2n3904 ends up not sounding as aggressive... I guess that's a function of the hfe for the various transistors?

[j_flanders]

R5: THIS SEEMS TO BE THE ONLY WAY TO GET A GOOD SOUND OUT OF THIS CIRCUIT
There's no equivalent to this resistor in any other schematic I've seen... but if you jumper it and go straight from the Q2 emitter to ground via R8 and C2, the pedal sounds... starved.

Although not exactly the same, the Vox Tonebender and Joe Bonamassa FF have an 'extra' resistor at Q2E:
http://www.geofex.com/article_folders/fuzzface/fftech4.gif
https://tinyurl.com/r9tk3g6

If it sounds starved the bias is probably off. Most people tweak the bias with a different value resistor or trimpot at Q2C but you did it at Q2E.
You could remove the 330r and tweak the 10k to get rid of the starved sound.

[tatou]

Thanks! Yes, I'm planning to play around with different ways of biasing the circuit today.

I also discovered this morning that there are a few similarities between what I've done here and the EQD Dirt Transmitter! (Minus clipping diodes.) Also going to try messing around with how they wired things up...