Making fuzzes self- oscillate

[Seven64]

I have built a few Devi fuzzes per a document of her's I came across on the internet.  On each schematic, she indicated both part substitutions for variations in the circuit to achieve different pedals, and oscillation points for each pedal. 

I do not understand why these oscillation points are where they are at.  I am trying to make a Univox Square Wave self-oscillate, but after 5 tries I cannot get it to oscillate like the devi pedals.  I have varied the spots on the "jumper" trying to feed output back into the input, but it either doesnt pass any sound, or more recently just passes a clean sound (when i actually ran the effect output back into the effect input)



This is the schematic I am using.  Can anyone point me in the right direction, but more importantly try to explain it in a way I can adapt this knowledge to other circuits?!

Thanks!

[R.G.]

Oscillation requires three things:
1. A way to feed the output back to the input, even with some losses along the way - a feedback path.
2. Enough gain from the input to the output to make up for the losses of the feedback path, so that the signal from the output being fed back to the input is at least large enough to make itself again by the circular path from input to output and back to input.
3. A total phase shift from input to output and back to input so that the feedback drives the input in a direction that reinforces itself.

If you think about it, these are just common sense. If there is no way for the output to drive the input, it *can't* drive itself to oscillate. If the amount of output being fed back is not big enough, it will run down as the oscillation is not big enough to recreate itself. If the direction of the output opposes the signal that would make it through the gain, the oscillation again dies out.

There are two main kinds of oscillators: relaxation oscillators and gain-phase or "Nyquist" oscillators. Relaxation oscillators have a feedback that is in phase with the input at all frequencies, plus some way to force it to not lock up forever. LFOs are made this way in many cases, but it's not much good for audio.

The other kind, gain-phase oscillators, have feedback that's negative at most frequencies, but the total phase shift at SOME frequency gets around to positive, and that's where the thing oscillates.

It's possible to have both kinds mixed up in the same circuit, where what happens depends on how much gain is available. Or, I suppose, different kinds that depend on where you take how much signal to feed back to the (or AN...) input.

So - how to get circuits to oscillate; find an input, find an output which may not even be the final output of the whole mess, just a place that "outputs" enough signal to feedback, and in the phase you want; then establish some feedback path, and juggle gain and how much/little feedback gets there to make it sing.

One interesting consequence of gain is that (1) all amplifiers are feedback amplifiers, as there is some amount of stray capaciitance from every conductor to every other conductor in the universe, and (2) all amplifiers will oscillate by some path if you increase the gain enough.

Those are the general principles. You have to know the gains from point A to point B, and you need to dither about how to get enough signal back to make it oscillate.

It's also very helpful to make the mental separation between DC conditions and AC conditions. Changing resistors in your circuit that have DC across them usually changes the DC operating point, which may lower the gain, limit the output amplitude, or even make it not amplify at all if one or more devices are turned off or fully on by the change in DC conditions. If you take your feedback through a capacitor (well, OK, transformers work for that too) to block DC = changes, you're only messing with how much signal gets fed back, not the DC operating conditions. This last is generally better unless you're working with something like an opamp where you have ironclad control of the DC operating points, at least theoretically.

[mistahead]

Pull a signal from around your diodes (I'd start on the diode side of that 4.7 before the clipping), put it through a voltage divider and push in back into the dry signal path after your input filter cap (right before Q1 at that junction).

You'll definately fulfill the feedback network and minimum gain requirements that way... use a pair of clips and lock on onto the feeback return I described and probe around for a good send, then swap - clip it on around the diodes and probe for a feedback return. Do it with a 25 / 50K pot L1 feedback source, L3 GND, L2 feedback return.

The two controls you really want to hit are reducing the resistance in the amplification feedback network to very low amounts, and the other I have a complete blank on suddenly... ummm... RG - Bail me out! lol.