I have a tried and tested layout (by GGG)
Yeah. Actually, that's my layout, sold under license.
and have rechecked all components of the input buffer on the board. They all have the correct value according to the GGG PDF, even the 330p capacitor. At least, they were all new components and the color codes and descriptions match. Also, Mich P reported the same problem on page one of this thread. He also fixed it with changing the transistors. This means that at least two people have reported this now.
OK. Two of them have had this problem
a discrepancy though: The resistor at the base of Q3 has to be a 3.3k
according to the information on your site. However in the GGG layout it is a 3.9k
. Maybe not a spectacular difference but this does increase the feedback slightly. Was it was put there on purpose to have more similar gain from both sides of the phase splitter?
I am aware of the Nyquist theory. But, of course, the theorem is based on a perfect differential amplifier with some voltage gain A which is completely independent on frequency voltage and whatnot.
The idea of a perfect amplifier in the theory is just to allow you to model imperfections external to the "perfect part" of the model for separating them out. Imperfections in the amplifier do not invalidate the theoretical application. Rather the perfect amplifier part lets you model imperfections in the amplifier you actually have and correct or modify the imperfections.
So I lowered the "gain".
Yes. However, lowering the gain of any oscillating amplifier will stop always it from oscillating at some point. Obviously with a gain of less than one, no amplifier can oscillate. But simply fixing things by lowering gain also sacrifices the advantages that higher gain and feedback give you, at the very same time. Keeping the gain high AND making it stable at the same time is an advantage.
Since the circuit in question has worked properly and not oscillated with all higher-gain devices, this is a big suggestion that lowering one transistor's current gain is not the single and only way to fix it. It's like stopping your teenagers from talking or texting all their waking hours. You can do that by tying them to a chair out of reach of the phone/computer. But that may possibly not be the best of all possible ways to stop the behavior.
On the other hand, if it produced results you liked, that's fine too. If it works and you're happy with it, great. Making any one pedal work to your satisfaction is fine. But generalizing from one - or two! - pedals to every pedal is an easy to misunderstand how the electronics works.
I admit I used the term gain too freely. I know what hFE means, I've spend some hours explaining this stuff to undergraduate students . I'm not an expert but at least I have some idea what we're talking about .
OK. That's good!
Also true, but as you say this assumes that the circuit is designed well and that local feedback is used.
Yep. That was my point.
But just take a look at the emitter of Q2. What's that capacitor doing there? I smell a complete lack of local feedback there.
Yep. That smell is the deliberate use of a bypass cap to eliminate local feedback. My point is that not all circuits do this, or do it well. And not all circuits which have overall feedback need local feedback at every point. IC opamps for instance don't necessarily use local feedback on every point.
There is no substitute for knowing the details.
Clearly a part of the circuit where the voltage gain of the combination of Q1 and Q2 is directly related to the bare transistor characteristics (hFE is some sort) .
Yep. That's where the raw voltage gain for the overall feedback around all three of the first transistors is developed, Q3 being a combination buffer/phase splitter. And in this circuit the combination of gains in Q1 and Q2 is what's making the raw voltage gain. And the parasitic capacitances of those two devices plus Q3, plus oddities in layout and wiring, make the parasitic components which would make the amplifier unstable if it's not properly compensated by gain/phase compensation techniques. The whole point of overall feedback from the resistors in the emitter of Q3 and the gain from Q1 and Q2 is to form an overall feedback loop around all three. It can be done so it makes the individual characteristics of all three transistors not matter much as long as there's enough gain, and as long as the compensation gets the feedback loop gain under unity by the time the phase shifts add up to enough to make the feedback turn to positive, as Nyquist said.
Of course the combination of the 47k and 330p should stop feedback
Yeah. Actually, it's a single dominant pole to cut the gain under unity before the other phase shifts make the thing oscillate. I found in simulation runs that it also makes for a resonant peak that's not quite oscillatory out at a couple of MHz. A 30pF cap from Q2 C to Q2 B is much more effective, and does not have the resonant peak. But both sound the same.
but I can imagine that if Q2 is completely open that the 'resistance' of collector/emitter pair will be much smaller. Definitely the transistor is loaded capacitively for positive input swings.
It's capacitively loaded for both positive and negative swings.
I think that you or probably anyone else here would have never designed that buffer like that.
I'm confused - Q2's not a buffer. Q3 is. No, I'd not do a circuit like this, given today's state of knowledge about components and circuits. But Mieda-san didn't have that advantage some forty-plus years ago.
I spent some time today performing circuit simulation of the input buffer. I could not change the gain freely in my spice program. I was not able to get the circuit to oscillate, even with darlington transistors so in 'theory' it should be stable. So that's one point for the Univibe.
Yeah, I get pretty much the same results. My simulator lets me turn up the gains on transistors at will for a given run, and I couldn't get the thing to oscillate until I put in current gains of nearly ten thousand. That is another thing that makes me think that changing the 2N5088 for a 2N3904 may have lowered open loop gain, and may have stopped the oscillation, but that it could be something about the transistor that was replaced or the solder joints or PCB condition that made the first one oscillate, not simply that it was a high gain device. It is quite difficult to ensure that simply because something got "fixed" when you changed something that it was the part *type* that you replaced that fixed it. Could have been a bad/nontypical part you replaced, could have been a number of other things.
But at least two people have reported oscillation of the input stage, both with the same solution.
OK. And more than two have reported all 2N5088s work fine. I've also found that all 2N3904s work fine. The sheer number of counter examples lead me to believe that there is more to it than using all 2N5088s is a problem, or all high gain devices is a problem. Not that you're not seeing and reporting correctly - just that I suspect that there's more going on than is apparent.
I think that maybe stray capacities play a role.
I agree 100%.
Just things that are harder to predict. My unit also sounded great except for distortion/clipping at relatively high input levels. That's how I found out. It could have slipped under the radar undetected...
Good that you caught it then.