weird interaction between 2 pedals - can't find the reason

[intripped]

Early morning test:
No change with battery.
Maybe a bit less noise, but still scratchy pot on guitar.

[Rob Strand]

With this ordering:

Guitar -> BE -> BOR -> amp

The BOR has an enormous gain.   Any noise/whine from the BE is going to be amplified by a significant amount.

You might be able to do this experiment to prove the noise/whine actually comes from the BE

    Guitar -> BE -> [ANOTHER HIGH(ISH) GAIN PEDAL]  -> amp

If it's something like a tube screamer then you might have to listen very carefully and perhaps crank your amp.

The ultimate test would be to connect:

     Guitar -> BE -> amp

Crank the amp and treble up full and listen carefully.   You may or may not be able to hear something.  The point is though if you do hear something you know it comes from the BE.

What to do?

You might be able to add shields or ground planes around the parts of the BE circuit.  If the problem comes from a layout issue where the ground  itself has the noise then you will have some serious layout debugging of the BE to do.   If you find the problem you may or may not be able to fix it.   It might require an entirely new layout!!!

[R.G.]

That's where I went first, Rob. He says that it's only with the single-stage boost section of the BOR, not the whole thing, and only in the middle of the "crackle OK" travel that it happens.

[Rob Strand]

That's where I went first, Rob. He says that it's only with the single-stage boost section of the BOR, not the whole thing, and only in the middle of the "crackle OK" travel that it happens.

Thanks.  I just passed by and wrote something.  Looks like everything has already been checked out, including using the overdrive. I'll just get back on the bus now and leave you guys alone. 

[R.G.]

Don't get back on the bus just yet. This one might need all the eyes - and brains - we can muster. There's something hidden here that's still not out in the open.

[patrick398]

You've certainly got the cavalry helping you on this one so not sure how much use i am but we might hive mind our way through it 
I've built a few BORs and never noticed any background noise.
If you think it's the boost section generating this noise itself it might be worth either disconnecting the OD part of the circuit or at least grounding the output in bypass just to rule out the OD completely.
You could also throw together a SHO very quickly and using the same guitar, cables, power, and amp, see if the noise is still there

[Rob Strand]

Don't get back on the bus just yet. This one might need all the eyes - and brains - we can muster. There's something hidden here that's still not out in the open.

Well  you guys have covered a lot of ground already and I'm not sure what I can add. 
The OP has done a good job working through things as well.

the most important thing that i discovered in my tests, is that if i turn the Boost pot at max, the noise completely disappears!
and i don't mean "it reduces to almost zero", it does actually disappear: while i turn the pot i hear the noise going up in level, until it becomes very loud, and after a sort of "thump" or strong motorboating, silence is back. Everything works, volume level is very high (as it should be with the gain pot at max) but there's no more noise.

The key lies here for sure.  It's a true cause and effect and also a solution.
I'm assuming the "Boost" pot is the one marked "GAIN" on the schematic.

My guess is the output impedance of the BE might be high and this promotes coupling inside the BOR (and hence feedback causing oscillation) from internal wiring back to the BOR input.     When the boost pot is set to high gain the Miller effect increases the input capacitance of the first stage which then helps shunt the HF coupling back to the input and so the oscillation stops.

For removing hidden RF from the signal lines, insert 47pF to 100pF CERAMIC capacitors from the signal line to ground, ideally also inserting 10R to 100R series resistors before the caps. This adds some resistance to turn the RF energy to heat and the capacitance shunts remaining RF to grround. It's a central tenet of low noise design to limit your amplifier's bandwidth to only the expected signals, not dramatically more or less.

Your recommendation here covers it.     I guess the only thing left is to go more:
-  so maybe 1k to 10k in series with the input.  Also make sure the resistor is at the PCB input, or retrofitted onto the PCB; not at the input socket or foot-switch.
- Then, add capacitance to the first stage MOSFET gate to ground. maybe 100pF.
  Perhaps not a solution, but in order to identify the problem, keep increasing the cap 1n, 10n, 100n!!!
  to see if the problem can be removed.  After that back track to a workable solution.
  The idea would be to add enough capacitance to match the MOSFET capacitance on full gain, which we know fixes the problem. With 1k or 10k we might not need the capacitor at all.
---------------
Maybe 1k is a good start (10k is too high) and perhaps a 220pF to 470pF, then try 1n if it doesn't work..

[intripped]

Thank you guys, for all your answers and hints!
I relly appreciate.

I'll not be at home this week, so I can't do any test, but I keep on reading the forum.

[intripped]

sooooo guys!!! 
BIG NEWS!

i think i've found the cause:
it's the zener diode 

i tried a breadboard version of the Boost (aka super hard on). the schematic is exactly like the krankosaurus' from madbean that i previously posted. i'm using a 9.1V zener diode 0.5W.

with this setup: guitar - Boost - amp
with boost active, if i turn the guitar's volume knob to zero, it makes THAT noise.

by accident (lucky accident) i discovered that the cause of this noise is the diode.
when i breadborded the circuit for the first time, i made a little error and connected the diode in a different way: instead of putting it between gate and source, i put it between gate and ground - the result was a working boost, without issues.
i was even more confused, because i was expecting the same behaviour as the boost in the pedal, but then i saw the error, and connected the diode as in the schematic.
the result was THAT noise; the same issue i have experienced many times with the two clones i've built.

so now guys, the questions are:
WTF? is the schematic or Mr. Zvex wrong?
can i connect the diode between source and ground? do i get the same protection for the mosfet? if not, could you suggest a correct way for doing it?
why the Overdrive circuit, which has the same diode at the first mosfet, doesn't behave this way?
should i change the connections in the overdrive circuit as well?

thanks again for all the help

[intripped]

Update:

I've subbed the zener with a 1N4148 and the noise with just guitar and Boost is gone, even with the diode between gate and source as in the schematic.

BUT 

I've tried again the Belle Epoch in front of the breadboarded Boost.
Well, there's still noise, THAT noise, when both the effects are active.

The only way to stop it is to remove the diode, zener or 1N4148, from the circuit.

...?


PS: this thing is definitely driving me nuts

[Rob Strand]

The only way to stop it is to remove the diode, zener or 1N4148, from the circuit.

Did you try adding the series input resistor and the cap to ground?

Diodes and zeners have capacitance.  The diode capacitance might be feeding high frequencies from the source to the gate forming some sort of unintentional oscillator.

Here's a diode scheme which doesn't connect the diodes between G and S,

https://hotbottles.files.wordpress.com/2012/03/zvex_sho.gif

[Rob Strand]

See,
https://www.microsemi.com/sites/default/files/micnotes/APT0402.pdf

"Since  adding  a  gate-source  zener  diode  does
not  effectively  restrict  high  frequency  noise
and  parasitic  oscillation,  it  is  best  to  leave
them  out."

See the text under figure 11.

[intripped]

The only way to stop it is to remove the diode, zener or 1N4148, from the circuit.

Did you try adding the series input resistor and the cap to ground?

Diodes and zeners have capacitance.  The diode capacitance might be feeding high frequencies from the source to the gate forming some sort of unintentional oscillator.

Here's a diode scheme which doesn't connect the diodes between G and S,

https://hotbottles.files.wordpress.com/2012/03/zvex_sho.gif

No, I didn't.
Tomorrow I'll do other tests, and I'll try this as well.

I've already seen that schematic variant, but if I'm not wrong that is the ver1, and I've read that the first version of "super hard on" had some noise issues and also some mosfets fried...

Let's see how it goes with res + cap at the input

[Rob Strand]

Let's see how it goes with res + cap at the input

Connecting the GS protection diodes before the RC filter could actually help kill any oscillations.

[R.G.]

If messing with conditions at the gate of one of the MOSFETs affects the noise, then it's a good working theory that what is actually happening is that the MOSFET is self-oscillating at some bias conditions. The "crackle OK" knob affects gain and bias  both, and if it's only whee-ing at some middle settings, it makes sense that there is a self-oscillation point.

Gate protection diodes are an unfortunate necessity. If the gate gets more than about 15-20V from the source, the gate insulation will punch through and the MOSFET becomes a DED (Darkness Emitting Diode) in sub-microseconds. The general practice with small signal MOSFETs has been to put protection diodes on any MOSFETs that go to wires off-board at a minimum. That idea is based on the thought that off-board wires are the richest source of electrical transients.

The commonest way used to prevent MOSFET self-oscillation is to put a damping resistor as close to the MOSFET gate as it can be put. These are usually in the range of 10 ohms to 1K. This seems to damp the ringing on the gate-source. This doesn't change the audio response of the MOSFET noticeably, as it's in series with the ridiculously high resistance of the MOSFET gate. It lops off some RF response due to its interaction with the gate capacitances, but that's not an audio concern.

Gate protection still needs to be thought about. The zener being a 9.1V device is a bit unusual. Gate protection zeners are usually more in the 12V to 15V range. The reference to leaving out gate protection zeners in the quoted article seems to be a little inapplicable. The application is on paralleled MOSFETs in switching applications, not amplification. The "transeint" they were protecting against is drain-gate voltages fed through drain-gate capacitances from paralleled devices. Their conclusion was that series impedances, resistors and ferrite beads, was a better solution.

In the pedal in question, there aren't any paralleled MOSFETs, and the application is only linear amplification; this seems to be substantially different from the quoted article.

Removing the gate protection diode seems to change things a lot. Given what the OP found, I'd say that there are a couple of ways to protect the devices. First, you could just leave gate protection out. That would work mostly, until some big enough transient came through to wreck the gate of the input device(s), and then a new MOSFET would be needed. That might be OK, as long as the pedal owner knows it. It's probably better to get some series protection on the gates - I personally would have put spaces on the PCB for series resistors at the gate of all the MOSFETs; but I tend to be very conservative about places where dragons are known to live.
Second, you could pull up the gates of the MOSFETs and stick one lead of a 1/8W resistor into the hole, then solder the gate of the MOSFET to the top of the resistor in mid air. That ought to damp the self oscillation. That ought to let you leave the zeners on the PCB, although I personally would change those out to 12V or 15V zeners to make sure they stay off in normal circumstances.
Third, you could remove the zeners and put snubbing diodes at the inputs. I would use 1N4148/1N914 types, one with anode to ground and cathode to signal; the other with anode to signal and cathode to power supply. this lets the diodes conduct to the power supply and ground for trasients outside the power supply. With luck, the power supply will drag along the source and drain of the MOSFETs and keep things in check.

[intripped]

..Today I haven't had the time to work on this, maybe I'll try something later this evening, and report back.

Planned tests:

- I have some 12 and 15V zener around, I will try them.
- low value resistor (10R/1k) at the gate
- 1-10k res + small cap (100pF) at the input (signal/ground)
- I think I have a ferrite bead as well
- snubbing diodes

Let's see how it goes with res + cap at the input

Connecting the GS protection diodes before the RC filter could actually help kill any oscillations.

Could you please explain with more details?

Thanks a lot RG and Rob

[Rob Strand]

Could you please explain with more details?

High frequencies that pass through the diode path will be attenuated by the filter before getting to the gate, effectively removing the diode from the whole equation.  It's possibly overkill but when you are tracking down evil problems it's best to try things that remove any suspect causes.

[Rob Strand]

In the pedal in question, there aren't any paralleled MOSFETs, and the application is only linear amplification; this seems to be substantially different from the quoted article.

It's true the article even mentions a push-pull oscillator.  However, often one evil cause will pop-up in other areas.  It's no secret MOSFETs can form HF oscillators.

A unique characteristic of the current problematic circuit is the fact it has a pot wired to the source.
The gain pot wiring could well be forming a parasitic inductance in the source.   The GS capacitance adds more feedback to the system and finally an oscillator is borne.  All speculation of course.

Look at the pattern here:
https://lab.whitequark.org/notes/2014-11-30/three-point-oscillator/

[PRR]

> The zener being a 9.1V device is a bit unusual. Gate protection zeners are usually more in the 12V to 15V range.

5V, 15V, not critical. Whatever you find in your drawers.

In BIG switching you need more than whatever Vgs is needed to deliver the big power. This can sometimes be 10V when you need many Amps.

BS170 data suggests that 4Vgs will pass >200mA, the proposed circuit can't pass over 2mA (9V/5.1k), so a 4V Zener is ample. OTOH we are assured 20V is safe. Even with ample leeway, anything 5V-15V is fine.

In another thread elsewhere, guys were killing $15 power MOSFETs "for no reason" and I advocated for a Zener in EVERY case even if the reason was not clear. (In this case a combination of shut-down events was tossing 400V spikes at the Gate.)

And yes, in Audio we should plan for a Gate resistor, and be heavy-handed about it. 1K is not too large. (MHz switchers have different constraints and Rob's link is informative for such stuff.)

[intripped]

hello guys!
i've eventually found the time to do some more tests. i know, i'm very slow, i'm sorry for that.

things are getting interesting, since i've managed to get rid of THE noise in some different ways:
it seems that any little modification to the original circuit just cures the problem.

i've tried:
- 12V, 15V, 5.1V zeners
- ferrite bead at the input
- ferrite bead at the output
- 10k resistor at the input (after the input cap, before the two 1M resistors)
- 100R resistor at the gate (after the two 1M resistors)
- 47pF capacitor (connected to ground) before the input cap
- 47pF capacitor after the input cap

changing the zener does not affect THE noise: it makes no difference. problem still present.

BUT: any other modification cures the problem.
i've not tried res + cap, but i'm quite confident that it would work as well, since just the resistor or just the capacitor is enough.

***i've not tried yet and applied one of these mods to my cloned BOR***

what solution do you think is the best?
the one i prefer is the ferrite bead, just because it doesn't alter the original circuit, and is very effective; but i don't know which is the best position: IN, OUT, both?
or maybe after the two 1M resistor, just at the gate of the mosfet? (i haven't tried it here yet... is it doable?) 

what do you think about the real cause of this issue? some kind of mosfet' self oscillation? or something else?