Finally got around to a Superfuzz.

[digi2t]

Hi all,

Well I had Jimi's praise of the Superfuzz ringing in my ears like a 3 hour rock concert for the longest time, so finally, I got it on the breadboard.

Mind you, I decided to try it differently. I went a different route, and tried a 4 transistor array chip, and an LM394 can for the octave section. Before some of you get in a huff about it being not the cheapest way to go, I'll give to Caesar what rightfully belongs to Caesar, you're all correct, it's not. Not by a longshot. With that said, here's a mugshot of the suspect on the breadboard;



I'm using the Univox schematic. A word of warning though; I initially used the schematic from here; http://cabbages.orconhosting.net.nz/graphics/joes_rubbish/fx/superfuzz/superfuzz.png
I found that the 10nF cap before the tone switch was too big, and didn't correspond with other schematics. In fact, with a 10nF, there's a big volume drop when switching tones. A 1nF should be spec'd here.

I'm using a TPQ2222 array chip, and an LM394 can for the octave section. I decided to keep the trimmer for the octave section, just in case I wanted to adjust the octave. Truthfully, I didn't know what to expect. Metal film resistors all around, metal film caps, run of the mill electros, and a pair of 1N60 diodes for the clipping. Here's the customary video of the test;

Like I mention in the video, it's a shame I don't have it mic'd up better. The "arpeggio" effect on the decay in the mid-scoop position is really nuts. I don't know if it's because of the 4 transistors on the same chip, the super-matched pair, or a combo of the two.

I have a MPQ3904 on the way, as well a THAT300. Between them, and the TPQ2222, there's a range of between 75 to 150 gain, with the MPQ3904 being the lowest. It'll be interesting to hear the difference between them. The gains of the TPQ2222 transistors on the breadboard all sit around the 120 to 122 range, while the LM394 transistors come in at 455 for the can, and 525 for the chip.

Like I mentioned off the top, definitely not the cheapest way to build a Superfuzz. The LM394 set me back about $8, and the TPQ2222 was about $10. The THAT300 runs about $10 as well. The MPQ3904 is much cheaper. I picked up five of them for $8, so that's a bit over a buck and a half each. Definitely "booteek" prices, for sure. What else can I say, I just had to try it. No regrets here. This setup just rip.

Here is a vero I whipped up as well, using the TPQ / MPQ type chip, and LM394. Mind you, you can still use the same vero with individual transistors, just respect the EBC pinouts. Not verified till built, but I've back traced it, and it should be fine;



Cheers,
Dino

[Kipper4]

You tease.
I want to hear it now. 

[Mark Hammer]

Waaayyyy back in my late 70's Electronotes "preferred collection" (that Steve Bragg over at Empress Effects borrowed from me and I still have to get back from him), bernie Hutchins notes that single transistor phase-splitters (Q3 in your linked-to schematic) provide "better" (which I interpret to be better matched) performance if the collector resistance is juuuuuuust a teeensier bit higher than emitter resistance.  Obviously there is no really guarantee of that with 5% resistors, unless one measures them and selects for that.  It may be that adjustment of R15 or R16 can provide the sort of matching needed to get optimal octaving, and it may be that the trimpot compensates for all of it.  I have no way of knowing.  I'm just offering up another element to consider.

I had one until recently, that used a toggle to lift and engage each of two sets of clipping diodes (GE and Si).  Plenty of distortion in their absence.  Note that, much like the Tychobrahe and Mayer Octavias, you tend to lose the octaving when the gain is cranked too high.

The Z-Vex Octane is essentially a modded Superfuzz, but with the addition of a variable tone control instead of two switch positions.  There are likely differences due to component tolerance, but I made myself a Superfuzz to Univox specs, and an Octane using the madbean layout.  The Octane has a more pronounced octave.

Keep in mind that setting it to the  bridge pickup, picking near the bridge, and setting the pedal gain just right, gets you a nifty electric sitar sound.

[digi2t]

single transistor phase-splitters (Q3 in your linked-to schematic) provide "better" (which I interpret to be better matched) performance if the collector resistance is juuuuuuust a teeensier bit higher than emitter resistance.

Can you define "teensier bit"? 1K, 2K? Or does it really depend on the individual circuit, and I need to test? I'm kind of think that the trimmer would compensate for that though. I did set it dead center (5K), and moving either way, I start to slowly lose the octave. I guess with the LM394, I can forego the trimmer all together.

Yeah, lifting the diodes is nuts. It's like it just hogged back a line of coke, but on the other hand, then I lose that magical "arpeggio" decay. I've never experienced anything like it from a fuzz. It's not in your face or anything, but subtle. Like harmonics cascading down, as the chord decays away. It's mesmerizing. The other thing that pops to mind is whether the diode shunts between the emitters and bases in the LM394 are playing any role here. Hmmm......

I pulled the LM394, and used a set of matched PN2222's. These were more inline with the gains of the TPQ2222, at 126 each. The octave is very nice, but isn't as, well... BLAM! as the LM394. Yeah, yeah, I know, not very scientific. But, with the LM394, cranking the drive gets me into a sound that reminds me a bit of the Ampeg Scrambler. With the PN2222's, the drive pots range is from "crunchy fuzz" at one end, to "great fuzz with pronounced octave" at the other. With the LM394, it's like the "great fuzz with pronounced octave" gets moved over to the 2/3 of the dial, and the rest is much more octave still, with some chords sounding like they're being run through a ring modulator, a la Scrambler. It just seems to add more color to the unit overall.

I should add that using even hotter transistors for the octave section is not good either. I tried a LM394 chip, which generally has higher gains than the cans (525 in this case), and the octave got a bit too splatty for my liking when the drive was dialed up. I wish I could find some LM394B's or C's. According to the datasheet, these are in the 300's range, and might be better still. Gonna keep my eyes peeled for them. In the meantime, I'll experiment with other matched sets of transistors. You never know.  Hey! I still have those 2SC1583's kicking around. Hmmm....

[Mark Hammer]

I wish I could define "teensier bit" , but I can't until I get my binder back from Steve.  Although I guess it bears consideration that Bernie Hutchins is a synth guy, not a guitar pedal guy, so he may be thinking about phase-splitters in a somewhat different way.  Having said that, I would gather that we are likely talking tweaks to the collector resistance that are well within the normal 5% tolerance; think 500Rtrimmer on a 10k resistor.  I gather it's the sort of thing you want a dual-trace scope to assess.

Can you forego the trimmer?  Maybe, and maybe not.  The 394 may have matched devices inside, but the phase-splitter may not send the identical signals to them.

I wouldn't feel bad about what transistors you use for it.  I'm not anal retentive about "authenticity".  If it sounds good, that's good enough, and I'll sub if it can be improved or if there is a tone I want that I haven't quite got yet.

I do find, though, that just about any phase-splitter-based octave fuzz tends to produce more obvious and audible octaves when the gain is held back a bit.  In those instances like the Superfuzz, Fender Blender, and Foxx, where there are clipping diodes, the octaves tend to get lost in a wash of harmonic content.  That may get improved by lifting the diodes, or sticking a cap to ground in parallel with the diodes to trim back on the high end.

[Thecomedian]

Whats those little black nodules on the ends of your wires into the pockets of the breadboard?

[digi2t]

Whats those little black nodules on the ends of your wires into the pockets of the breadboard?

Rubber, molded around the crimp of the pin to the wire. Prevents two adjacent jumper wires from shorting together.

[soggybag]

Why not LM3086 or LM3046? These have 5 NPN types in an 14 pin DIP. Two of the transistors emitters share a pin. But this works for the octave section.

[duck_arse]

digi2t,

back when eti was using them in car alarms, we couldn't get the lm394 in aust, at least not where I worked. we had to *settle* for the lm194, which is why I have a pair of lm194h's. and no car to build an alarm for.

is that a pile of j201's in the top right of yr breadboard pic, just out of interest?

[digi2t]

Why not LM3086 or LM3046? These have 5 NPN types in an 14 pin DIP. Two of the transistors emitters share a pin. But this works for the octave section.

Those were first on my list, but I accidentally discovered the TPQ series in my research. I think I went with them more out of an esthetic reason than technical. Four transistors for the buffering/amping/fuzzing etc., and two just for the octave section, allowing for matching. Just seemed like a neater set up in my minds eye. No real fancy EE reasoning involved, sorry to disappoint.

digi2t,

back when eti was using them in car alarms, we couldn't get the lm394 in aust, at least not where I worked. we had to *settle* for the lm194, which is why I have a pair of lm194h's. and no car to build an alarm for.

is that a pile of j201's in the top right of yr breadboard pic, just out of interest?

Well, I guess they can be right at home in a Superfuzz now. According to TI`s datasheet, the 194 and 394 are essentially the same, the 394 being the tougher of the two.
What you see there on the bread board is PN2222`s. They`re plastic package equivilents to the 2N2222 cans. Same spec`s, but I`ve found that in some applications, they are much quieter. I find the metal cans to be antennae sometimes, introducing noise into the audio. I had them out because the circuit wasn`t working at first. Thinking that my TPQ wasn`t working, I threw them in to test. Finally, it was a resistor in the wrong hole.

Out of curiousity, I tried a 2SC1583 for the ocatve section. I couldn`t really «try» it, volume-wise (late at night), but in the head phones, it sounded quite nice. More to come on that in a couple of days.

[duck_arse]

I thought the 194 was the mil-spec part. that would be a bit tuffer than the consumer version, surely.

[digi2t]

I thought the 194 was the mil-spec part. that would be a bit tuffer than the consumer version, surely.

Aw CRAP!, I read it wrong. Thanks for setting me straight. You are correct, the LM194 is the tougher of the two. Just the way I read the sentence in the datasheet, dyslexia set in.

[pinkjimiphoton]

DINO, THIS IS COOL AS F*CK!!!!!!!!!!!!!!!!

[digi2t]

DINO, THIS IS COOL AS F*CK!!!!!!!!!!!!!!!!

You know it brother. Don`t know what took me so long, but here I is. Video coming tomorrow of the circuit with a 2SC1583 in the octave section.

[Mark Hammer]

For me, I have yet to find a better - in the sense of patently obvious - octave-fuzz than the Foxx Tone Machine.  While it combines the two outputs of the phase-splitter in a different way than some other units, like the Superfuzz under discussion here, its chief difference lies in the network between the emitter of the second transistor and the base of the first.  If you isolate it from everything else, it is essentially a midscoop filter, of the sort one would see at the end of the Superfuzz.

So you have to wonder whether it presents something folks could play around with to improve the octaving even a little more.

Commander Keen would likely know the answer to this one, via the relationship between the two-tranny input stage of the Foxx and Superfuzz, and the basic structure of the Fuzz Face (feedback from E of Q2 to B of Q1).  Does the midscoop network in that feedback loop accentuate, or diminish the midrange gain for those frequencies that might constitute note fundamentals?

[digi2t]

Commander Keen would likely know the answer to this one, via the relationship between the two-tranny input stage of the Foxx and Superfuzz, and the basic structure of the Fuzz Face (feedback from E of Q2 to B of Q1).  Does the midscoop network in that feedback loop accentuate, or diminish the midrange gain for those frequencies that might constitute note fundamentals?

Now that you mention it, I have a GGG FTM. I should get it out, and do a side-by-side. Honestly though, I was never a big fan of my FTM. I found it quite noisey/hissy. I used it a bit after I built it, and quickly developed a love/hate relationship with it, the latter taking precidence. I`ll dig it out and try it again. I don`t recall it having the agressiveness of tone that this Superfuzz circuit I`m playing with now. I think the FTM was quite smoother, fuzz-wise. Octave-wise, yes, I believe the FTM is great. Even better than the Octavia. The octave fuzz that I like best though, is the WTF? (What the Fuzz?) octave fuzz. Simple circuit, based on Escobedo`s Push me Pull you, but rather than using a 3904 in the gain stage, it uses an LM386. It suits me in two ways; my love for 386 fuzz (see: Uglyface), combined with just the right touch of octave.

With the Superfuzz, it`s not so much the octave that I`m chasing, but the fuzz. That chainsaw, rip your face off, fuzz is great. I do realize though, that the octave section is an integral part of it`s character. So, rather than trying to experiment in making the octave section better, I`m trying to improve the octave section so as to improve the fuzz. If that makes any sense. That`s why I was pleasantly surprised with the 2SC1583 in the octave section. Albeit that I didn`t really get to taste it in a «cranked» enviornment, the initial headphone offering sounded promising.

Also, with the 1583, the 10K balancing trimmer is very effective. I decided to check my voltage at the base of both transistors, and found that they were off. I carefully trimmed them, so both sides would be equal. It so happened that I had the headphones on while I was doing it, and when I really hit the perfect balance between the two (1.888v / 1.888v), the circuit got quieter. This was like hitting a sweet spot of sorts. The tone, and octave was really sweet too. I`m getting back to it tomorrow morning, this time through the Ampeg, with some volume. Then I`ll really be able to compare to the LM394.   

[Mark Hammer]

Nice!  Thanks for doing the R&D on this.

And again, I'll put in a plug for the ZVex Octane tone control, which is essentially the variable midscoop that I put in the Shin-Ei FY-2.

[digi2t]

Alright folks, here's part 2. I'm using the 2SC1583, and I think I'm in love... big time. Although the octave isn't as crazy as the LM394 can, the fuzz is great. Especially that "arpeggio decay" that I talked about before. Listen in at 11:04, when I hit that E chord, and let  it ring. Depending on your speakers, you might be able to hear it cycle through harmonic notes as it decays. Headphones might help here. My friend had a listen, and came up with an interesting analogy on the decay. The way he put it was, "It's almost like staring at an airplane propeller, as it starts to spin, and speed up. You know, the way it appears to turn one way, slow down, spin backwards, speed up, slow down, change direction again. This is like an acoustic version of that illusion."

It's just absolute fudgenuts!

Just one thing I should mention. In the videos I talk about the trimmer being set to have the voltages dead even on the two bases in the octave section. After more prodding with my DMM, I have to make an amendment here;

I have to say for the 2SC1583, after playing with it some more and checking my DMM, the best setting for it is just off center a bit, at 1.885v and 1.905v (9.05v supply) at Q4/Q5 base. At dead center (1.892 / 1.892), the circuit is a bit noisier, especially when you roll the guitar volume right off. The best way I found to set the trimmer, for the either dual transistor, is to hook up the DMM at the collectors of Q4/Q5, and gently nudge the trimmer, while watching the voltage. The sweet spot for either, is where I hit the highest voltage on my DMM. Either side of that sweet spot (or "hump"), the voltage drops away. Even with a multi-turn trimmer, it can be quite touchy, but it works great.

If you set the LM394 this way, be prepared for more noise than the 2SC1583, but dead even base voltages produces a gate the initial attack of chord strikes, like a Scrambler. If the LM394 is set with dead even voltage at the bases, it doesn't sound as good. Using the peak voltage method yielded the best results sound-wise. I'm guessing that the "more noise" is from the metal can. I've noticed that metal can transistors are consistently noisier than their plastic counterparts, I guess this is no exception.

So...
2SC1583 = Great fuzz, not as pronounced octave, but better note and chord definition. Wonderful decay with harmonic overtones in either tone switch setting, any pickup, anywhere on the neck.
LM394 = Great fuzz, much more octave overtones, overpowering past 5 on the drive, a la Ampeg Scrambler. Note definition is quite good, but can get very sloppy on chords, especially on high drive settings. Great decay, but less harmonic overtones.

Any way you decide to go, the trimmer can be a valuable asset in getting a great sound out of the Superfuzz. I would say that it's still very important to match the transistors for Q4/Q5, but the trimmer can help hone the sound to the point where your ears will say "PERFECT!".

Here's a vero using the 2SC1583;


Cheers,
Dino

*EDIT* - Today I received the two other transistor arrays, the THAT300, and the MPQ3904. I tried both, and I've scratched the THAT300 as a possible IC. Although the gain seemed to be alright (113 to 118), it would produce this bizarre squeaking noise as the note would die away.

The MPQ3904 works great. Just as well as the TPQ2222. All the great harmonic decay artifacts of the 2222. The 3904 is a tad hotter than the 2222 (170 to 180), but sounds just as good as the 2222. Either one get my three thumbs up.

[digi2t]

Well, I took your POV Mark, and sure'nuf, you've sold me on it. I removed the scoop switch, as well as the 47K / 10K divider resistors. I lifted the tone scheme from the Madbean "Poindexter" schematic, and put a B10K pot between the 100nF cap and ground. I kept the 22K though, because I found it a touch muddy in full scoop with a 10K / 10K combo. 22K / 10K seems to retain some of that "brassiness" at full scoop. All I can say now is....

WOW!!!

Just ripped myself some new earholes. There's more top end as well at the other end of the dial. Here's the final product in action on the breadboard;

and a vero to reflect the changes;



My only question left pertains to the placement of the two resistors. Mark, I've seen your FY-2 scoop control, and I noticed that you used a 10K / 15K combo, with the 10K being before the 15K. I'm using the 22K before the 10K in this case. I tried them the other way around too, but I find that the mid scoop just doesn't sound the same. I think I've been staring at this sucker too long, and I'm not quite sure what's going on here. Does switching them around raise / lower different frequencies?

[Mark Hammer]

The "first" resistor and cap to ground form a lowpass filter, according to the usual rules ( F = 1/[2 * pi * R * C] ).  Nothing really magic or obligatory about it, so select the value of the first resistor according to where you want/need/prefer the lower bound of the scoop to be, and the second resistor to balance out how much bottom-vs-top there is.

With 22k/100nf, the rolloff starts around 72hz, and around 160hz with 10k/100nf.  I would imagine a higher rolloff, without a suitable 2nd resistor to tone things down, would make the result kind of woofy, or what some might describe as muddy.  Differences between the scoop resistance values on the FY-2 and the Superfuzz likely reflect not what is "correct", but rather what suits the timbral qualities of what each circuit produces.  YMMV.

If you wanted to go stark raving nuts, then what you do is replace a 22k/10k resistor pair with a 20k pot, wiper to the 100nf cap, and a 5k1 on either side of it, to move the rolloff of the LPF part around.  That, in conjunction with the variable bleed-to-ground through the 100nf cap will gve you plenty of tonal variation, though maybe more "tweak maintenance" than one normally wants in a stompbox fuzz.

Somebody, whether Jack Orman, or RG, posted some stuff a while ago on adjustable midscoop filters.  Check Jack's labnotes at AMZ first.