ok so the problem came back after a few minutes of playing, and I continued to search for the problem, I drew a schematic here:
https://oshwlab.com/evalc2636/problematic-amp(the 6N2P and 6N23P tubes, pin 9 is the shield between the 2 triodes, not the heater tap like shown in the schematic, so it's fine!)
I've changed:
the 2 main rectifying diodes for the main B+ supply from the old germaniums to some CO15... something (I forgot what was the actual name for them but I remember them being rated for 1.5kV and like 2amps), that didn't fix it, if anything it made it slightly worse perhaps.
changed the grid leak resistor and the ac ground capacitor on the PI NFBL side.
changed the CF resistor, the plate resistor for the stage driving it as well.
changed the coupling capacitor into the PI.
changed the NFBL presence pot.
changed all grid leaks on the output tubes.
changed all the coupling capacitors on the output tubes.
I made sure the output impedance is correct.
I grounded the place where the 470, 2 1Meg's, and 10k connect on the PI.
I grounded the NFBL circuit.
I measured all plate voltages, grid voltages, cathode voltages, everything seems normal (except the NFBL side of the PI which I'll explain down below)
I swapped tubes, tried a 6N23P instead of a 6N2P in the PI, it kinda made it a bit louder but still the same.
I moved all the shielded wires around to see if maybe there's some feedback somewhere somehow in some wire.
I moved the screen supply from B+, to B+1.
I tried biasing the output tubes super cold, and really hot,
still, after doing all that, nothing changed, still the same poping thing.
when it pops the volume gets lower, and then after around 1-2 seconds it starts fading back in, I then started measuring voltages to see which voltage related to what I was hearing, and that was the NFBL side of the PI, which the
cathode voltage jumped really high,
bias voltage was doing something super weird and my multimeter just showed overload on DC volts or AC volts (I measured the bias from the 10k to grid, not from the ground to grid because the 1Meg input impedance of my meter would probably mess with the readings a lot), the bias seemed super low,
low being 0volts, and it was the same on both sides, but the NFBL loop side jumped all around, it never went positive really, it was always negative bias.
the
plate voltage jumped high when it popped, and then slowly dropped exactly in relation to the volume fading back in,
the voltage on the 10k, 470, and 2 1Meg's, junction had exactly the opposite behavior of the cathode voltage, when it poped it dropped, and when it came back it increased.
this led me to believe it was the 470ohm cathode resistor, which changed that. still the same...
now I'm starting to think it's some sort of feedback loop within the NFBL side of the PI, but how could this be, even if I ground the NFBL network out?
the NFBL line goes right next to the plate wires for the output, the signal
could have coupled over... but I doubt the signal would be nearly great enough to cause any oscillations if the other end of that wire (where the pots connect) is grounded.
I don't have an oscilloscope and I don't have any place I can borrow one from.
sometimes the amp pop's and then stays in some sort of state where it hums at 50hz, and it draws a ton of current from the wall, I would think that would be a clear sign of ultrasonic oscillation happening like one of you suggested was the problem already, the question now is tho, where is it coming from?
and that has only happened like 2 or 3 times, which
could be just a coincidence and I just had the instrument cable close to the OPT or something...
although now I'm thinking... maybe it is in my inputs because the junction of the grid and 22k get grounded not the junction of the 22k and 1Meg like shown in the schematic (couldn't find a 6.3mm jack symbol that was the same as the ones I have in the amp)... it could be causing some ground loop, and maybe it picks up something from the output since the 22k isn't there anymore and it won't cancel out ultrasonic frequencies.
the input wires going to the input jacks are both shielded, and both shielded wires are in another shield, all shields there are grounded at the input JACKS.
so I'll try wiring the inputs like normal tomorrow. meanwhile, feel free to suggest checking anything else.
for some reason, every amp I build is cursed with problems like this
EDIT: so I tried just plugging in a random cable in the other input to break the contacts that ground the grid, that didn't fix it!
it's weird that these problems occur only if I have all 4 tubes in, if I remove any 2 it works like normal, I doubt it's oscillations in the output tubes because that PI side related to the behaviour exactly, meanwhile the output doesn't seem to do anything in relation to what I'm hearing, the voltage on the grid could be jumping up because of a spike in voltage on the PI plate, the screen and plate voltages relate to what the bias does, not what I hear from the output, so I really doubt it would be that.
I'm was thinking it was maybe within the power supply, but none of those caps are open, so it shouldn't be able to oscillate at a high frequency, the amp itself has nearly no audible hum from itself.
perhaps it's some feedback from the CF to the 1st stages? but again it would need to have open filtering capacitors to oscillate ultrasonically within the power supply I think...
maybe it's a grounding issue, but I've tested nearly every ground, and everything seems ok.