If that is the case, then put the second channel probe on the output and see if the signal disappears.

I thought that was mentioned in post #32. With the scope on the input the noise was eliminated.
The input impedance of most scopes is 1M with some capacitance.

 

Whoa. That worked! I used a 1n2 instead of 1n. Grey one is 100n.

 

I thought that was mentioned in post #32. With the scope on the input the noise was eliminated.
The input impedance of most scopes is 1M with some capacitance.

Not the way I read it. He said that the signal was present on pins 1 and 2, but not pin 3, but he didn't say that the signal on pins 1 and 2 disappeared when he connected the probe to pin 3.

 

Whoa. That worked! I used a 1n2 instead of 1n. Grey one is 100n.

View attachment 348918

View attachment 348917

So the difference is 200pF. What happens if the cable is a metre longer?

 

Whoa. That worked! I used a 1n2 instead of 1n. Grey one is 100n.

Should give a shout out to crutschow.
Problem could have been solved hours ago.

 

What make is your TL072? Back in the 1990's when I used to work in professional audio, we always insisted in Motorola or Texas TL072s, because the SGS-Thomson ones were not unity-gain stable.

Texas Instruments. Most all our parts we’re using for these are from Mouser. I’ve tried TL072, LM358, TLV2372, OPA1642, OPA2134. The TL072 sounds the closest to me like my guitar with a 10 foot cable plugged straight into my amp. Some of the others are close, some make it sound thin, etc. TL072 seems to be a good balance of transparent sound and inexpensive … and also allows

 

Should give a shout out to crutschow.
Problem could have been solved hours ago.

Blame me. I was being lazy. Didn’t wanna have to desolder stuff. :-| Sorry @crutschow .

 

Try adding 10kΩ in series with the input followed by a 1nF capacitor to ground.

Awesome. I read this, but it wasn't *as easy* as some of the other suggestions. Thanks!

So, please explain what's going on here. I don't understand it.

 

Should give a shout out to crutschow.
Problem could have been solved hours ago.

Steady on now! You've added a tiny amount of capacitance and this one works. Will the next one work? Bear in mind that the tolerance on that capacitor is ±120pF, more than half of the extra capacitance you added. Unfortunately, I think you have a sticking-plaster, not a long term solution.

 

What does the tiny amount of capacitance actually do? What’s the theory/practice behind this?

…and chances are we could potentially solve it by somehow improving our input trace layouts?

 

What does the tiny amount of capacitance actually do? What’s the theory/practice behind this?
…and chances are we could potentially solve it by somehow improving our input trace layouts?

It's a low pass filter.
I suspect noise from the screen is the culprit.
Can you disconnect power to the screen while still using the buffer amp?
And yes a different layout might help as well.

 

It's a low pass filter.
I suspect noise from the screen is the culprit.
Can you disconnect power to the screen while still using the buffer amp?
And yes a different layout might help as well.

https://forum.allaboutcircuits.com/...ating-with-floating-input.206626/post-1982638

In post 50 I mentioned that I removed the ESP32 entirely with the relays latched in buffer bypass mode. The noise oscillation still happened.

 

It's a low pass filter.

1n2 and 10k make a 13.3 kHz LPF. Right?

 

1n2 and 10k make a 13.3 kHz LPF. Right?

Correct, appx 15.9Khz at 1nf.

 

A LPF is really what’s stopping the oscillation or it’s it the small capacitive load?

 

I would say it's the filter but try it without the 10K resistor.

 

Confirmed. Kills the oscillation with just a 1nF cap from input to GND. Confirmed in both of my PCB builds.

Lazy non-solder test:

 

Buuuut, look at the simulation of the input impedance in LTspice...

First one is 1nF only. Second screenshot is 1nF with 10K resistor.

 

Also confirmed that now my guitar doesn't sound as bright in buffered bypass vs true bypass. I don't think this solution is gonna fly.

 

The lowest cap I have handy is a 10pF MLCC. I put it temporarily in place of the 1nF and the buffer sounds pretty good. I'll have to make some recordings and compare them more surgically, but this is how a 10pF cap impedance simulates: