Parasitic oscillation problem untraceable

[TriodeLuvr]

I thought the cathodyne provided the least gain of the tree PI types?
In ascending order:
1 cathodyne
2 LPT
3 paraphase

The cathodyne is the least only if we don't consider the other half of the tube. If we think about the 12AX7 driver as a whole, then the cathodyne PI will have one half of the tube with a gain of roughly 50 driving the other half (the actual cathodyne stage) with a gain of about unity. The LTP exhibits gain of about half that amount or a little more.

Jack

[TriodeLuvr]

I only have a 1000uf and a 4700uf 63V electrolytic cap .. should I try the latter?

Those are BIG caps for this purpose. What you gain with larger caps is a longer time constant. If this is a bias problem, that means a sustained overdrive will take longer to create the fizz. However, as the signal falls away, the unwanted distortion will stay with you longer because bias will take longer to normalize.

About Jack's suggestion on lowering the 6V6 cathode resistor and fire up the bias to 100% ...makes sense in a way to avoid blocking distortion or is it just going to make things worse? This amp is so damn loud already. I even added a 1/4 power switch to be able to play it at home

Just so we're on the same page regarding terminology, blocking distortion is what happens when a too-large signal draws grid current on positive peaks, causing the grid of the power tubes to move negative. That can have a similar effect, but it's a totally different cause, and the smallish coupling caps that are typical in guitar amps tend to minimize it as a problem.

What I was referring to earlier is cathode bias shift, and it's specific to self-biased Class AB1 amplifiers. At idle, the DC voltage on the cathodes is determined by the current through the tubes and the value of the cathode resistor. However, when the tubes are driven with a signal, they draw more current, because this is AB1 operation. That additional current creates more positive DC voltage at the cathodes, pushing the tubes in the direction of cutoff (the grids become more negative, relative to the cathodes). Depending on the specific combination of resistor value, signal level, RC time constant and signal duration, the result can be significant crossover distortion. The tubes will operate closer to Class B than AB1, and they will continue to do so until average current falls and the cathode capacitor discharges.

I want to stress that this might or might not have anything to do with the problem you're having. It's just one of the things to nail down, given the nature of what you're experiencing. This is why I suggested monitoring the DC voltage at the cathodes while you listen to the problem. If you have a DVM, clip lead it to the cathodes and ground, and watch what happens when you smack the amp. You might find that certain cathode voltage levels correspond to what you're hearing.

Not sure how practical this might be for you, but another approach would be to temporarily ground the cathodes and apply fixed grid bias. In a pinch, you could use three 9V batteries in series and a 100K pot. The batteries will last many hours in that configuration, much longer than you would need to do a listening test.

Jack

[JMPGuitars]

If it were in front of me, I would correct the LTP PI, raise the power tube cathode capacitor substantially, and possibly add a snubber if needed.

Bieworm, Make your PI match this schematic but don't mismatch the plate resistors: https://mhuss.com/Plexi6V6/Plexi6V6.gif

I will try the huss plexi power section you suggested. I intended to do that in the first place, but eventually didn't for some reason.
I only have a 1000uf and a 4700uf 63V electrolytic cap .. should I try the latter?

About Jack's suggestion on lowering the 6V6 cathode resistor and fire up the bias to 100% ...makes sense in a way to avoid blocking distortion or is it just going to make things worse? This amp is so damn loud already. I even added a 1/4 power switch to be able to play it at home

I would use the 1000µF cap. The higher one will make the amp way too stiff.

You're not that far off from Huss' PI. Your cathode resistor is too high, and that could be part of the problem. You might also need to raise the tail resistor. You can try 15k to 27k and see what happens.

BTW- the PI sweetness is what you currently have setup. 100K on both plates is what gives you the desired imbalance. 82K + 100K is an attempt at balancing (even though the values aren't perfect).

[JMPGuitars]

I'm wondering why you suggested this.

Some people believe that a slightly unbalanced PI (by not compensating one of the load resistors) provides a sweeter sound.

Yup. The (almost) perfectly balanced PI is more for hifi than guitar.

[Bieworm]

You're not that far off from Huss' PI. Your cathode resistor is too high, and that could be part of the problem. You might also need to raise the tail resistor. You can try 15k to 27k and see what happens.

You mean lower the cathode resistor from the PI, no? I'll go for the 470R then...
Do I need to swap the coupling caps too from .01 to .022? Is there a problem leaving the .1uf cap to ground as is?
The amp has plenty of bass, so maybe leaving the .01uf caps in isn't a bad idea? Unless I'm missing something...
And about the cathode resistor of the 6V6's.. leave it at 83% or check if the near 100% is better?

[JMPGuitars]

You mean lower the cathode resistor from the PI, no? I'll go for the 470R then...

Correct, 470R.

Do I need to swap the coupling caps too from .01 to .022? Is there a problem leaving the .1uf cap to ground as is?
The amp has plenty of bass, so maybe leaving the .01uf caps in isn't a bad idea? Unless I'm missing something...

You can leave the caps alone on the PI. They're fine.

And about the cathode resistor of the 6V6's.. leave it at 83% or check if the near 100% is better?

I wouldn't recommend changing that. I think the problem (or at least most of it) is in the PI. In fact, I would also recommend lowering the grid stopper resistors on the power tubes once the issue is sorted out.

[Bieworm]

Thanks Josh

I just bought what I need. Even the snubber circuit. Except the 8M2 were only available in 1/4 watt.. I hope that's ok.
That company is probably going to provide me with the siglent and take back the utd.
Also bought a 8R 50W for a dummy load. That signal generator is driving me insane

Once this gets sorted out I'm going to try lowering the B + and see..

[JMPGuitars]

Thanks Josh

I just bought what I need. Even the snubber circuit. Except the 8M2 were only available in 1/4 watt.. I hope that's ok.
That company is probably going to provide me with the siglent and take back the utd.
Also bought a 8R 50W for a dummy load. That signal generator is driving me insane

Once this gets sorted out I'm going to try lowering the B + and see..

1/4 watt is not okay. 8.2M is available in 1W Carbon Film, and you can also find it in 1W metal film. I use the carbon film - 500V working, 1000V surge.

I use a speaker motor loaded attenuator as my dummy load. I can still hear the thing, but it's quiet enough.

[Bieworm]

Thanks Josh

I just bought what I need. Even the snubber circuit. Except the 8M2 were only available in 1/4 watt.. I hope that's ok.
That company is probably going to provide me with the siglent and take back the utd.
Also bought a 8R 50W for a dummy load. That signal generator is driving me insane

Once this gets sorted out I'm going to try lowering the B + and see..

1/4 watt is not okay. 8.2M is available in 1W Carbon Film, and you can also find it in 1W metal film. I use the carbon film - 500V working, 1000V surge.

I use a speaker motor loaded attenuator as my dummy load. I can still hear the thing, but it's quiet enough.

Then the snubber circuit has to wait until I order stuff next time in Germany. Let's get this PI to operate properly in the meantime

[crgfrench]

I thought the cathodyne provided the least gain of the tree PI types?
In ascending order:
1 cathodyne
2 LPT
3 paraphase

Or am I wrong?

That sequence only includes the one PI triode in the cathodyne. When you include the other one in the tube (which you would in reality) as a gain stage, the tube generates 2x the gain as the same tube operating as LTP.

[crgfrench]

The cathodyne is the least only if we don't consider the other half of the tube. If we think about the 12AX7 driver as a whole, then the cathodyne PI will have one half of the tube with a gain of roughly 50 driving the other half (the actual cathodyne stage) with a gain of about unity. The LTP exhibits gain of about half that amount or a little more.

Exactly.

[Bieworm]

Thx for that clarification

[crgfrench]

Does that mean that with the LPT, which has a shared K resistor too, can lead to positive feedback too?

I'm not sure about "LPT" but for a Long Tail Pair PI ("LTP"), that shared resistor sets the bias and there's nothing to worry about.

[TriodeLuvr]

Yup. The (almost) perfectly balanced PI is more for hifi than guitar.

Unbalancing the PI creates even-order harmonics. That's OK, but it's not related to his problem and it might make troubleshooting more difficult. If he builds the phase inverter exactly according to the original Marshall design, it can be eliminated as the culprit.

As an aside, it appears to me that the original 82K/100K design is already unbalanced. The correct values for a balanced PI appear to be 91K/100K. Maybe 100K/100K simply unbalances the PI in the opposite direction.

https://www.ampbooks.com/mobile/amplifi ... iled-pair/

This could be confirmed with a distortion analyzer. I'll try to remember to do that when I get my amp built.

Jack

[crgfrench]

Unbalancing the PI creates even-order harmonics.

How does a slightly lower amplitude on one side introduce harmonics?

[TriodeLuvr]

Unbalancing the PI creates even-order harmonics.

How does a slightly lower amplitude on one side introduce harmonics?

The imbalance creates an asymmetrical waveshape. Asymmetry is composed of even-order harmonics.

[JMPGuitars]

Unbalancing the PI creates even-order harmonics. That's OK, but it's not related to his problem and it might make troubleshooting more difficult. If he builds the phase inverter exactly according to the original Marshall design, it can be eliminated as the culprit.

As an aside, it appears to me that the original 82K/100K design is already unbalanced. The correct values for a balanced PI appear to be 91K/100K. Maybe 100K/100K simply unbalances the PI in the opposite direction.

That is correct. I don't think it matters either way for his issue.

91K is correct, but 82K supposedly became the standard because it was a value more readily available. Honestly it's kind of silly. But also consider that they were using mediocre carbon comps whose values would stray up. So they could have found 82Ks that ran high fairly easily. If it's 10% tolerance, and straying up, then they could be very close to balanced if the 100Ks are accurate enough.

[TriodeLuvr]

LOL, you're absolutely right. It's easy to forget sometimes about the real tolerances with this type of circuitry. I'm old enough to remember when most schematics had a note at the bottom, something like "All voltages are +/- 20%." Modern circuitry with regulated supplies and semiconductor junctions is really OCD in comparison.

Jack

[Bieworm]

Well I think there is some progress. There is still fizzy on the OD but it's creamier. Perhaps it's just the nature of the amp?
The changes I made:

PI:
RK 470R + 22k tail

6V6's:
RK 330R + 1000uf cap (91.7%)
1k5 5W screen resistors

All in all sounds pretty good, but I'll have to play some more. It's pretty late now...

Here's the voltage chart.. is the negative grid voltage normal on V1? Nothing's plugged in the input.

Fwiw I checked the voltage swing on the cathode of the power tubes. Idee is 21.1V and with a good strum it rises up to 30V

16097961712413564128678924108376.jpg

Sound clip:
https://www.dropbox.com/s/er24jamf71c5q ... 4.m4a?dl=0

[TriodeLuvr]

Here's the voltage chart.. is the negative grid voltage normal on V1? Nothing's plugged in the input.

Both V1 grids should be grounded by the input jacks when nothing is plugged in. If you didn't use switched jacks, a small negative voltage might be normal. 1M to ground is probably high enough to allow the grid(s) to develop voltage due to contact potential. IOW, the 1M might be acting as a grid leak resistor, not just a grid resistor (the two terms are frequently misused).

Fwiw I checked the voltage swing on the cathode of the power tubes. Idee is 21.1V and with a good strum it rises up to 30V

At 30V, the amp is deep into Class B. That's almost certain to be the problem. You're hearing crossover distortion.

Jack