Superfuzz - tuning the octave effect

[yeeshkul]

I guess for the best octave response you need to pair the transistors in the octave-up stage, and tweak the base-to-ground 22k resistors (or the whole 470/22k dividers). Is there any way to do this on the scope?

[R.G.]

Sure. Apply a signal of the same approximate size as you'll have while playing, adjust for equal peaks at the octave output, not alternating peaks.

It is possible that there are two competing effects there, in that the most equal peaks will be different for very small signals and larger (100mV in) peaks. You may need to adjust for some compromise at both smaller and larger signals.

[PRR]

Isn't that what this trimmer does?


Although if you use two modern transistors from the same batch, it is hard to expect more than 0.005V offset error, while the trim shown gives +/-0.3V range. It could perhaps be a 1K or even 500r pot.

[Derringer]

yep

if you're building one, build it with that 10k trimmer and tune the octave with it
you don't even need a scope really, you can tune it by ear

just play some double-stops around the 12th - 15th frets and turn that trimmer to get the most pronounced octave effect

[PRR]

Assuming same-batch transistors and matched resistors, it should come out really excellent without trim.



Error of zero, 1mV, even 3mV is about all the same, 10mV will give a fine octave but with some fundamental leakage.

The main error is the resistor tolerance. 5% resistors could give 0.1V error. 1% parts is better, but still ~~20mV error. You might tack-up your resistors and supply, without the transistors, and measure across where the transistor bases will go. You want less than 3mV difference, preferably 1mV. Swapping several "22K" resistors for best balance may be more stable than a trim-pot.

[yeeshkul]

Thank you PRR
I've built about 5 units and i guess i had been only lucky, as they worked fantastic with the transistors from the same batch (paired for Hfe).
I have never measured the resistors, i didn't realize the 5% tolerance may do such a mess. This would be my second time this week (the first time it was Ibanez PT9)  i'm about to fiddle with a trimpot, trying to find a needle in a haystack. I'll do it via the fixed resistors this time, or a small 1k trimpot as you say.

[R.G.]

This kind of gets back to my overworked concept of defining what "best" means. As Mark Hammer observed, there is some support for making that trimpot be a front panel control, not to make the pairing of full wave rectification peaks perfect, but to make them variably imperfect, down to not at all. This lets you dial the octave down or out.

You'd do this by making one side of where the trimmer sits be fixed and the other variable.

[yeeshkul]

R.G. that's what i have done first - i replaced the 22k res with 25k trimpot. However the output freq. did not get doubled by tweaking it (i added 4k7 in series to the trimpot just to be sure ...). I have no idea what can be wrong now. The base signals are equal now - the trimpot allows that kind of adjustment indeed, but it has no influence on the output signal shape, the way i would expect. I am doing it on my scope.

PS: by the way when you turn the trimpot (= trimpot wired the way it replaces one of the 22k res) down (not all the way - just by ear), the superfuzz produces very musical, rich and mushy old-fart type fuzz sound. I really like it a lot - something i expected Tonebender to be like but have never kicked that sound out of it.

[R.G.]

R.G. that's what i have done first - i replaced the 22k res with 25k trimpot. However the output freq. did not get doubled by tweaking it (i added 4k7 in series to the trimpot just to be sure ...). I have no idea what can be wrong now. The base signals are equal now - the trimpot allows that kind of adjustment indeed, but it has no influence on the output signal shape, the way i would expect. I am doing it on my scope.

Let's look at how the circuit works. First, turn the 10K trimpot to exactly halfway. That gives the equivalent of a 27K resistor to ground from each transistor base. There's also a 100K to +9. Let's make the somewhat bold and rash assumption of infinite current gain on both transistors. When we do that, the current in the emitter rises until it starts to back-bias the base, where it shuts off. Each transistor (assuming perfectly matched resistors and transistors) then has a base *voltage* of (27/127)*9V = 1.9V. The joined emitters will rise until the voltage hits 1.9V minus Vbe, or about 1.3V. That's through a 1.8K, so the current in the emitter is 1.3V/1.8K, or about 722uA, 3/4 of a milliamp. We said the transistors had infinite hfe, but if they only have hfe of 100, then the base currents are only contributing 7.2ma; this hardly affects the emitter current at all, so we'll ignore it. Besides, transistors with 200 or 500 or more are easy to put in there. Each transistor is contributing exactly half.

But what if the resistors and transistors are not identical? Then one transistor/resistor set will try to force the emitters a little higher than the other one will. It pulls up the emitter a little more, and this has the effect of further turning off the "weak" side. It could happen because the 100K and 27K voltage divider on the high-side base gives a bigger base voltage because of the tolerances, or that the high-side transistor has bigger gain or lower Vbe.

We can correct this imbalance by tweaking the 10K trimpot with the grounded wiper to turn on the low-side transistor a little bit. "Little" is the operant word here. We want just enough to make the low-side transistor and high-side transistor come into equal currents. When we get that balanced point, any signal coming into the bases pulls one base high and the other low. This turns on the transistor where the signal is going high, and the joined emitters follows the high-going signal. The base on the low side is pulled low by the signal and by the high-side transistor pulling up on the joined base, so it turns more off. When the signal reverses, the signal goes high on the side that was low, and turns off the transistor that previously went high. This means that if the balance is perfect, the signal on the emitter always goes positive from the no-signal position and follows the waveshape up, never down. The signal at the emitter looks like a full wave rectified version of the signal.

If you deliberately unbalance the two transistors, one is conducting signal to the emitter most of the time, and the unbalanced-low one only turns on or signals bigger than the unbalance voltage.

If you replaced one of the 22Ks with a 25K trimpot, that will unbalance it, all right. It can turn it one transistor all the way off, so it never conducts. This makes the stage into a follower, and there is no rectifying and hence no octave effect.  In fact, it's not necessary to turn one all the way off. Simply replacing one of the 22Ks with a 27K to ground and the other with a 22K in series with a 5K to ground would give the equivalent of the 10K trimpot from the stock circuit, but crippled to only lower one side.

Notice that the bases being equal voltage and equal signal do not mean it's balanced for octave signals. There isn't anything in that circuit as shown to let you balance it at DC by anything other than watching the signal at the emitter on a scope. That's because the collectors are joined to a single 10K resistor. If you replace the 10K with two 5K's to 9V, one on each collector, you can now look at the difference between the collector voltages and balance what now is obviously a differential amplifier.

The 10K trimpot lets you unbalance to either side, so matched transistors are not needed. Some iterations of this circuit left the trimpot out. There have been at least five versions of pedals as I recall that use this circuit for octave.

[yeeshkul]

There we go! In the end it was a shameful mistake (Q5 sideways) that made it so very unbalanced 
RG thank you for the last post, it puts a lot of light on the most important stage.

I am posting a blend of two the actual scope screens - the blue/red sine waves are signals at the bases of Q4/Q5 and the gray signal is the mixed output of double frequency.
The Expander was turned down a more than half way, the output was tweaked by 1k trimpot (viper to the ground, legs on 22k base resistors as pictured above).

[R.G.]

There we go! In the end it was a shameful mistake (Q5 sideways) that made it so very unbalanced 

No embarassment needed. We've all done it.

I am posting a blend of two the actual scope screens - the blue/red sine waves are signals at the bases of Q4/Q5 and the gray signal is the mixed output of double frequency.
The Expander was turned down a more than half way, the output was tweaked by 1k trimpot (viper to the ground, legs on 22k base resistors as pictured above).

That is showing only the smallest bit of what this thing can do. When it's balanced perfectly. the waveform looks much like just the two halves of the red and blue waveforms above the middle/zero line, the round parts all up and the pointy-parts all down.  But the gray waveform will have a nice touch of second harmonic. Maybe not an in-your-face octave, but should sound nice.

[PRR]

> base currents are only contributing 7.2ma...

Typo: uA.  (perhaps a /2 error also.)

Since I had the idiot compute it all, let's have the theoretical voltages and currents:



There will be some DC shift with signal. (It is in fact a rectifier and the emitter cap will charge-up some.)

I notice that my sim shows fairly low collector voltage; Jan's 'scope shows top-flattening which my sim did not show and I am guessing may be a high collector voltage?

But asymmetry here can also be unbalance in the phase-spliter before it. The "cathodyne" is well balanced if the loads are balanced; here the two bases cut off and on. For mathematical perfection we might want a buffered driver.

> watching the signal at the emitter on a scope.

At emitter, should be teeny due to large cap (which might perhaps be larger, but WTH). The octave signal is strong at the collector.

> dial the octave down or out.

Going by sim(!): for small unbalances, the octave is pretty constant but the fundamental leaks more as balance gets worse. At 30mV offset (275uA/400uA collector currents) the 1st and 2nd partials are about equal, 3rd about 10db down (depending on level). It gets pretty goofy from there until the offset exceeds the signal voltage, where it becomes a straight amp.

> gray waveform will have a nice touch of second harmonic

My eye says 100% 2nd harmonic with 5%-10% of fundamental leaking. How it will sound is complicated. If there is a 3rd above a few percent, your ear will follow the 3rd 2nd series and assume a 1st. If the 3rd and all odd harmonic is well suppressed, the ear may assume the 2nd is the fundamental. Since the relative balances vary with attack and decay of the string, it isn't at all like playing a half-length string.

[yeeshkul]

So i decided to make some more screen shots and sound samples. I used 1kHz/400mV signal as the test signal for the scope.
I used a 5k trimpot between the 22k resistors, viper to the ground as pictured above. 1k is enough to tune the octave, but 5k lets the more visible changes happen.
The red curve is the signal voltage at Q4 base, the blue curve is the mixed output measured at the clipping stage.
The blue curve (octave stage output) seems to be dependent on the level of the signal (pickup, Expander pot).

To get as much octave as possible i've done this:
- switched to the neck pickup
- played above the 12th fret
- turned the expander down to 1/4
- kept the pick short or used finger-picking
- switched the superfuzz tone switch to the "tight" sound


1. Trimpot is unbalanced - the 2nd harmonic is tuned to have the same level as the fundamental (1st harmonic)
Expander at 1/4




sound sample

2. Trimpot is almost in the middle position - the 2nd harmonic raises way above the 1st harmonic - maximum octave
Expander at 1/4




sound sample

3. Trimpot is almost in the middle position - the 1st, 2nd and 3rd harmonic are almost equal
Expander all the way up




sound sample