Unwanted octave down bursts - help

[blackieNYC]

I am curious as to the source of an octave-down feature in a fuzz.  In general, and I actually have an unwanted one showing up occasionally. It's gotta go.
I imagine an octave down is created with a half wave rectification.  I'm wondering what is achieving this in my circuit.  Perhaps a bias imbalance can cause this?
Forgive me, I lost 5 of my most recent schematics (hand drawn) behind some cabinetry which I now need to dismantle.  I can't adhere to the debugging protocols, but some of you may be able to tell me what capabilities I may have going that could create this. 
The pedal works great.  For passages without this occasional occurrence it sounds great.  If I back off on the distortion a good deal, it goes away.  but from moderate distortion upwards, there is an octave down which occurs briefly, just after the loudest notes played.  More pronounced with the lower notes - an amplitude related issue then. But with the pre-boost and the attack (a 1Meg pot instead of below) up full, the octave doesn't show up on most notes.  So there appears to be a threshold factored in, but again, I have to back off substantially to fully prevent these bursts. t have a Jordan Bosstone circuit with the diodes removed.  I love the sound, and I chose to give myself the option of hitting it a little harder. 

To drive the  thing a little, I have a AMX mosfet booster up front, then a notch filter using something a little like the Geofex EQ with a fixed Q, but I believe the circuit I used was just a notch filter with no peak boost capability http://www.geofex.com/article_folders/eqs/paramet.htm

[bluesdevil]

Are you getting the low octave "bursts" using this circuit alone?

[DDD]

Just solder 0,5 uF caps in parallel with the second and third 0,022 uF ones, and look for the changes.
Then, playing with the caps, you'll find the ''weak point''.

[aron]

I loved the octave down. I talked about it before, but reduce the value of the coupling caps and it will go away like DDD said.

[DDD]

It seems that second 22 nF cap to be the most critical one.

[R.G.]

I suspect that the circuit is going into a semi-synchronous "latchup". The second transistor is driven hard enough to be at the edge, and the whole Bosstone circuit is oscillation prone. This would fit the observed change in behavior with different coupling caps changing how much drive gets through and the frequency where it starts almost latching.

Similar effects were used in organ relaxation oscillators to get 'octave down' in the very early days of electronic organs. Circuits that are at or near oscillation at some frequency easily "pull" or get into lock step with other oscillators nearby on either harmonics or sub-harmonics. In this case, the guitar can hit it where it's sensitive for some notes.

It's a theory, anyway.

[DDD]

The same octave-down effect can be observed sometimes in Harmonic Percolator stompboxes (with ALU de-coupling cap).
Moreover, there are some topics on the Net reporting that Electra fuzz can do the same thing under some conditions. The Electra fuzz consists of one BJT and couple diodes, so it's a surprise.

[blackieNYC]

I'll try tacking on the larger caps - you're saying that whichever spot gets worse with a larger cap is the weakness, and it's there that I should reduce the cap.  Fine, but that's making me think I'm just using low end rolloff to get rid of the problem.  So overall, this fix will cause me to lose some low end - I was hoping there would be another solution.  That doesn't involve "organ relaxation".  Maybe if the 2nd transistor is driven hard by the original Bosstone design, and now I'm boosting the input and clipping the first, I should reduce the input to the 2nd.  Time to go back and read up on fuzz theory again.
Aron- the octave down is brief and unpleasant.  If I could get it to happen constantly that would be cool, but I'm sure there are better ways and more efficient circuits to do that.  
Oh- this circuit alone, meaning isolate the Jordan.  Yes, I can bypass the op amp in the simulated inductor, and the MOSFET. If they make the problem go away, my design has sorta gone t*ts up. Sure wish I'd noticed this in prototyping.  Probably didn't turn up loud enough.
If low end reduction is the only way to get rid of this, suppose I could test with a graphic eq in front, and see what kind of low end rolloff it takes to get rid of the octave-downers, but I think I won't like the results.  I was hoping someone would tell me it's a biasing thing and a resistor change would magically make everything perfect.

[R.G.]

Try reducing the 18K base-emitter resistor on the PNP. This will reduce the drive to the PNP without changing much else. That PNP is being driven very hard by the collector current of the NPN, This might cut back on the internal current gain without losing you much outside the circuit.

[Mark Hammer]

As I recall (and as I built into mine), increases to the value of the input cap tend to result in those "octave outbursts".  If one can learn to tame them by playing appropriately, they can be fun.  But like any boisterous child, you only want the outbursts when you want them.; otherwise they're just plain annoying.

[DDD]

Increasing capacity of the second 0.022 uF cap (from two 560k to ground) may stop the octave effect without any LF loss.
But the effect may be unpredictable. It worth to try increasing the cap's value 10 to 100 times in any case.