Yet another TB MK II question: *low*-frequency oscillation with too much bass

[midwayfair]

This circuit is driving me nuts and I can see why the design was scrapped for the MKIII. 

I soldered up a Tone Bender variant I've been working on, and it sounded decent when I breadboarded it a couple months ago, but when I made a layout I decided to include a bass control after stage 1 that blends a 100uF (the stock value) with a 10nF. Why? At full bass and full gain, it starts oscillating with certain transistor combinations, and I was really hoping to use a silicon in Q3 for bias stabliity (after having used the requisite germanium on the BB). And it's a low-frequency hum, nothing like the squealing in an oscillating silicon fuzz face, which I *think* is what people have described in . It intermodulates with the signal and does other weird things, like extra hum when I turn down my guitar. I've added quite a bit of power filtering, so it shouldn't be coming from there. I've tested with silent single coils and humbuckers, so I'm pretty sure it's not just amplifying guitar hum. It sounds like it could be amplifying thermal noise.

My voltages are in the right range according to multiple posts around here on the subject: .5v on Q2 (or thereabout), 8V on Q3.

I'm going to try some more things out tonight, but I'd be interested in knowing two things before I socket myself silly:
1. WHY is leakage necessary here when it isn't in the Fuzz Face; and does it have to be leaky in both Q2 AND Q3? Does a lack of sufficient leakage cause what I'm hearing? (I got best/quietest results with a low-gain Ge hfe ~50 in Q2 than with the Ge in Q3, but only when the Si Q3 was *even lower* gain - 2N3440, hfe 43.)
2. Why does merely turning up the bass control introduce the sound? Are FETs secretly much noisier in the low frequency range than I'm used to hearing and it's just being brought out here?

I'm using the classic schematic, but I made a slew of changes and I haven't redrawn it:
-My input stage as a FETzer valve with a gain of ~2 (SK170). The 100K pulldown is 1M and there is a 33K in series with a 100nF input cap (in place of the 4.7uF).
-blend between 10nF and 100nF cap after Q1. If I decrease the bass, the oscillation disappears.
-Trimmer on Q2 bias (Q3 bias is stock). Adjusting the bias can make the sound disappear ... sometimes.
-100R limiting resistor between Q3 emitter and the fuzz pot -- which I'd like to avoid if possible.
-Reverse bias on the base of Q2, plenty of power filtering, polarity protection, all that stuff ...

[duck_arse]

maybe try increase the 5uF on the attack pot?

[edit:] and can you edit a couple of letters from the post-title?

[R.G.]

I'm going to try some more things out tonight, but I'd be interested in knowing two things before I socket myself silly:
1. WHY is leakage necessary here when it isn't in the Fuzz Face; and does it have to be leaky in both Q2 AND Q3? Does a lack of sufficient leakage cause what I'm hearing? (I got best/quietest results with a low-gain Ge hfe ~50 in Q2 than with the Ge in Q3, but only when the Si Q3 was *even lower* gain - 2N3440, hfe 43.)

Q1 must be leaky to bias itself up at all. Your problems may be with Q1. Add a quite-high-value resistor from -9V to Q1 base, or increase the 100K pull-down on the base-emitter to make the best use of whatever leakage you have. Since you're not using Q1 as shown in the schematic (and this makes it more difficult to help you with this, btw) and some of your changes are "all that stuff", we're down to guessing what you do have. This is important because you're probably getting motorboating back through the power supply, at a very wild guess.

2. Why does merely turning up the bass control introduce the sound?

It increases the gain above unity at the frequency where the feedback gain from input to output and back to input has a phase shift of N*360 degrees.  This is a high-attenuation path, so it takes more gain to make it oscillate. Google "Nyquist criteria".

Are FETs secretly much noisier in the low frequency range than I'm used to hearing and it's just being brought out here?

They have very high input impedance, letting the available gain go to quite-low frequencies, which I suspect is one thing you wanted, but it also enables more loop gain down at those frequencies. FETs are neither inherently more noisy or quiet than bipolars (although all of today's parts are better than germaniums used to be, for other reasons). But they have their applications, and their noise characteristics are different, specifically in the way they interact with source impedance.

I'm using the classic schematic, but I made a slew of changes and I haven't redrawn it:

See above subtle rebuke.  

[midwayfair]

Thanks for your help, RG. I'll try to clarify some things below ...

It increases the gain above unity at the frequency where the feedback gain from input to output and back to input has a phase shift of N*360 degrees.  This is a high-attenuation path, so it takes more gain to make it oscillate. Google "Nyquist criteria".

The phase shift here is created by the negative feedback resistor? (I guess it has to be ... since it's output from the emitter of Q3, which is reverse phase, back to Q1.)

EDIT: I wonder ... if I put my FET in backward (and said FET is symmetrical), would that do bad phasey things? I'm pretty sure it's oriented correctly, but I don't have it in front of me.

Q1 must be leaky to bias itself up at all. Your problems may be with Q1.

To clarify Q1:
Chop of everything between 5uF (now 100nF) and .1uF in the schematic.
Substitute this:


(Minus the 220pF and with 100nF in place of 47nF).

Also forgot to mention ... negative ground, NPN transistors.

Q1's output is fine according to the audio probe. A FET has higher output impedance, right. Shouldn't there be LESS bass content? And when I breadboarded the original, Q1 was amplifying about as much as the FET. I just wanted an input stage that didn't require leakage.

some of your changes are "all that stuff", we're down to guessing what you do have. This is important because you're probably getting motorboating back through the power supply, at a very wild guess.

Sorry about that!  
Power input is: 1N5817 > 47R > 100uF to ground, 100nF to ground. Voltage at Va is 9.17.
The reverse bias germanium diode on Q2 is base to ground, but it made no difference in behavior when I added it.

See above subtle rebuke.

I was a bit burned out this weekend ... I layouted and schematic myselfed silly over the last week and really was hoping to just throw together a fuzz in a couple hours as a diversion on Sunday. Instead it turned into A Thing.

Here's what's bugging me:

It works and doesn't oscillate with the transistors I have in there now, but the gain ranges are wildly different from (and much lower than) what I expected. It's very confusing to me that it's oscillating. Schematically, the pedal right is no different that putting a boost in front of an NPN Fuzz Face (daisy chained), which I've done many times, and yet this build won't accept the same gain ranges or transistor compliment in Q2 and Q3 as a Fuzz Face without oscillating uncontrollably. In such a case, the bass between stages would be even greater. I've built the Skreddy Screwdriver, which uses a MOSFET in Q1 with otherwise kinda similar topology and that doesn't oscillate despite the higher Q1 gain and even higher bass content. So what is different in this instance?  

Maybe this is one of those problems that goes away when the circuit is in a grounded enclosure, but I really don't want to put it in one.

[midwayfair]

Okay, I figured it out. It was two things: not enough power supply filtering (I grabbed a 47uF instead of 100uF), and the size of the drain resistor I used on the FET (anything smaller than about 5k ends up picking up feedback by connecting the 9v rail through the drain resistor, the 100k feedback resistor, and the caps on the input and output). I got it to stop now but I might mess with it a little more later.

[R.G.]

Okay, I figured it out. It was two things: not enough power supply filtering (I grabbed a 47uF instead of 100uF), and the size of the drain resistor I used on the FET (anything smaller than about 5k ends up picking up feedback by connecting the 9v rail through the drain resistor, the 100k feedback resistor, and the caps on the input and output). I got it to stop now but I might mess with it a little more later.

Classic feedback through the power and ground.