Tube Amp connections : 0V vs EARTH vs HEAT vs CURRENT

[systemtruck]

Scope arrived and I was able to run some quick tests between WORK work. I had already reconnected the OT and ran these tests before seeing your message about current readings @CJ

Only had time to try the Grid1’s and the Plates, and the tests showed no signs of oscillation at either. Didn’t have time to test Grid2’s but will later.

I made sure to zoom in on the time span and also mV scale so i didn’t miss any high frequency waveforms. The grid’s basically are flat lined when the guitar volume itself is down, which is pretty nice to see, and on the Plates i can see the 60Hz power cycle which i understand gets canceled by the other side.

But one important thing is that no waveforms increase during the B+ collapse. So i don’t think that some high frequency oscillation is ramping up and causing the EL84’s to self overload, draw current, red plate, until rails collapse. I was really hoping to see that. I also used the Auto scaling function a number of times just to make sure i wasn’t somehow missing something. The scope does a pretty good job bringing everything into perspective with the press of that one button.

But, i was able to very clearly watch the distorting waveforms when hitting one string on the guitar heavily, and watch the clipping slowly disappear as the sound also cleared up. This happens on the plates as well as the input audio grids. Pretty much the same clipping. It’s possible that one side clips a bit more than the other, but not much i don’t think. I may measure for that later when i have more time. I did use 82k and 100k as different plate resistors for the PI, since other designs out there mention the nice sound of even harmonics that can introduce. But this is irrelevant from the issue. The clipping was the same way before i did that. But at least it might explain a slight difference in the point each clips.

It sure just looks like plain old diode clipping to me, but I’m not that well versed in the appearances of clipping waveforms. And again, this clipping does not exist in the PI outputs when the PI is disconnected from Grids. Is it possible that the EL84 stage design is just biased to only ever handle very small signals, and i just need to redesign the balance of things to re-center the waveform better?

But the AC15 design is very similar, as are other EL84 bias designs. So I suppose that’s the ticket. Maybe it’s just some bad solder joint of mine that happens to be causing clipping and also DC current draw issues, or it’s TWO sets of badly clipping tubes that also have current draw issues.

Here are the EL84 plates during clipping:



And here are the EL84 input grids during clipping:

 

[ccaudle]

The third picture shows positive grid voltage, that is always going to crunch pretty badly.

The increasing DC voltage on the grid is still the biggest concern. The bias level (i.e. voltage at the control grid, nominally 0V but possibly a few mV because of grid leakage current, and the voltage at the cathode) should be constant. You can't ever have a stable amp if the output stage bias keeps running away.

 

[systemtruck]

The third picture shows positive grid voltage, that is always going to crunch pretty badly.

The increasing DC voltage on the grid is still the biggest concern. The bias level (i.e. voltage at the control grid, nominally 0V but possibly a few mV because of grid leakage current, and the voltage at the cathode) should be constant. You can't ever have a stable amp if the output stage bias keeps running away.

Well what’s interesting about that is the fact that this clipping is present from the get go. Right after power up. So the clipping is not even depending on any kind of rising DC buildup over several minutes.
It’s almost like there is DC offset in the tube itself in some way that there shouldn’t be. And this is the case with both pairs of tubes i have. ( A brand new JJ set arrives in a few days. )

Also, i recall some measurements before… perfect 0V DC at one grid, and like 0.04V DC at the other grid. This is soon after power up, with no audio going into the grids. Doesn’t that mean there would be no clipping at that point? And clipping would only later develop and get worse if DC slowly accumulated?

I’ll do some more grid DC measurements when I can hop back into the shop.

 

[ccaudle]

The third picture shows positive grid voltage, that is always going to crunch pretty badly.

Sorry, I was interleaving doing something else, and what I wrote before isn't really correct. For the pre-amp stages that would happen, but this is a push-pull output stage, so you expect one of the tubes to be turned off during part of the cycle. If I am reading the pictures correctly the input to the grid still goes down to -25V, which seems excessive amplitude. More experienced guitar amp people will probably know right away if I am off base with that

This might be a useful reference to compare to your current configuration:
https://valvewizard.co.uk/pp.html

 

[systemtruck]

I’m leaning towards the bias issue being a bad tube, in my new matched set as well as my old set. But here are tests with my newer matched pair…

- When measuring DC at the grid, the “good” tube holds about 0.008 V indefinitely. Also, when a heavy strum is played on guitar, the grid jumps down to around -13V and promptly returns to “0”. The bad tube starts off the same way and behaves the same when playing too, dropping down to -13V, but within a minute or so it starts slowly climbing and climbing. I have been shutting it down around 3V, and it was only getting faster. At first i tried adding a new grid leak resistor to the bad tube just in case the old grid leak was intermittent under stress, but the problem persisted. So I switched the tubes, and the issue follows the tube. So it’s not the circuit nor the design.

- When disconnecting one EL84 and monitoring DC on that grid pin, voltage rises a little during power up but settles down to about 0.001 V, and flutters in the sub-mill volt range indefinitely. This happens identically at both tube locations when each is disconnected and the other is left in. (I kept the rail down at 310V by using the variac) So i don’t think there is a bypass caps leakage problem coming from the PI causing any grid issues.


As for distortion….

- Scope on just one (of either) grid to 0V visually shows clipping, which occurs rather synchronized with speaker output clipping sound. The audible and visible clipping fade away at the same time as a note fades out.

- Scope across both grids together, as in one positive/negative signal, looks perfectly smooth and balanced. Even when audible clipping occurs at the speaker output. It’s a night and day difference with visual monitoring just one or other grid… this positive/negative waveform at the pair of grids is perfectly smooth at all times, even during severe audible clipping at speaker output.

—-

I think it makes sense to wait for the brand new JJ tubes to arrive and see what happens. It is alarming how this can happen on the tube market though, with a matched pair tested for current and gm. I guess that only accounts for first power up, not after some run time.

But i do have one question… Is a Push Pull system supposed to already have the waveform chopped up at the output of the PI? Or do the EL84’s take care of that business internally and each grid should appear as if they are Class A full wave? Because I’m confused as to why the EL84 grids, ie the PI outputs, would already have chopped waveforms individually. I had been under the impression that any PI is just built to send out full waveforms just in reverse polarities.

 

[Cqwet Dbdfte]

Sorry, I was interleaving doing something else, and what I wrote before isn't really correct. For the pre-amp stages that would happen, but this is a push-pull output stage, so you expect one of the tubes to be turned off during part of the cycle. If I am reading the pictures correctly the input to the grid still goes down to -25V, which seems excessive amplitude. More experienced guitar amp people will probably know right away if I am off base with that

This might be a useful reference to compare to your current configuration:
https://valvewizard.co.uk/pp.html

-25V on negative peak seems ok for a tube biased at -12V.
and 0V at max + peak.
With cathode bias and negative supply as reference these numbers shift positive by 12.
The driver stage may be able to provide more voltage swing than the power tubes can handle
The phase splitter should provide unclipped swing on both phases up to 0V on the power tube grids.
If there is an expectation of more voltage swing, maybe a better amp is needed.
EL84s are not exactly power houses

 

[systemtruck]

-25V on negative peak seems ok for a tube biased at -12V.
and 0V at max + peak.
With cathode bias and negative supply as reference these numbers shift positive by 12.
The driver stage may be able to provide more voltage swing than the power tubes can handle
The phase splitter should provide unclipped swing on both phases up to 0V on the power tube grids.
If there is an expectation of more voltage swing, maybe a better amp is needed.
EL84s are not exactly power houses

I have faith that this is not normal for the staging. The PI and output circuit is remarkably similar to the Vox AC15, and would be pretty much identical if i just lowered the grid leak resistors (which i tried and didn’t solve anything). The kind of clipping distortion I’m getting here is not similar to tube break up ala AC15 or any amp. It happens well before that breakup, and which does also happen with this amp if turned up.

I found this article, which i suppose could be happening.
https://www.18watt.com/storage/18-watter_buzz_info_311.pdf
Some kind of self induced EL84 crossover distortion that is caused by an input signal causing extra bias upon itself due to the contact with its bypass caps or something, which seemingly can remedied with a zener+diode setup strapped around the grid leaks. Pretty interesting stuff.

Will update when I’ve put in this pair of JJ EL84’s. Who knows at this point !

 

[systemtruck]

Adding a note that when measuring the post bypass caps output of PI, when disconnected from the EL84’s, the individual waveforms look round and great, not clipped.
So the difference is this: when circuit is connected, each individual P.I. output waveform is visibly clipped. When the circuit is disconnected, each individual P.I. output waveform is perfectly normal.

Something about the EL84’s grids/load is causing this, even just after power up when the EL84 grids are still sitting very close to 0V.

Looking forward to solving this one some day haha.

 

[ccaudle]

when circuit is connected, each individual P.I. output waveform is visibly clipped

Measured at which point? Where the coupling caps from the PI connect to the grid stopper resistors, or where the grid stopper resistors connect to the grid?

 

[systemtruck]

Measured at which point? Where the coupling caps from the PI connect to the grid stopper resistors, or where the grid stopper resistors connect to the grid?

I eliminated the grid stoppers recently but haven’t updated the schematic. The grid stoppers have seemingly altered nothing. Things have clipped with them in and with them out, and at different values when they’re in. I think an AC15 has 1.5k? Maybe I’ll put that in for good measure.

 

[systemtruck]

Updated schematic with notes about scope showing clipping.
Tomorrow new EL84’s arrive, but I’m preparing for the worst and assuming this will still occur.

 

[CJ]

It is normal to see those grid waveforms as your signal reaches the level of the bias voltage. And that clipped waveform is going to travel upstream and make it look like it is a phase inverter problem but it is not.
What does the output waveform look like on the secondary if the output xfmr when the distortion takes place?

How are your filter caps? If you are using the original fisher cans then that might he part of your problem.

 

[merlin]

You need a sinewave generator. Even if it's just the headphone output from your phone and you play a sinewave vid on youtube, it's easier to understand a circuit than whatever random waveforms you're using at the moment.

 

[Cqwet Dbdfte]

Better yet use a triangle wave to spot compression. A large square wave is useless as it is basically a DC waveform that will hide bias issues, and have no use for audio signalling, other than finding time domain issues with smallish levels.

 

[systemtruck]

It is normal to see those grid waveforms as your signal reaches the level of the bias voltage. And that clipped waveform is going to travel upstream and make it look like it is a phase inverter problem but it is not.
What does the output waveform look like on the secondary if the output xfmr when the distortion takes place?

How are your filter caps? If you are using the original fisher cans then that might he part of your problem.

You need a sinewave generator. Even if it's just the headphone output from your phone and you play a sinewave vid on youtube, it's easier to understand a circuit than whatever random waveforms you're using at the moment.

Better yet use a triangle wave to spot compression. A large square wave is useless as it is basically a DC waveform that will hide bias issues, and have no use for audio signalling, other than finding time domain issues with smallish levels.

I set up my software to generate a signal out of interface and somehow it doesn’t have triangle as an option, so i used a sine wave…
The level sent was guitar level basically, and the preamp pot was kept the same throughout this. This is me adjusting the master volume pot just prior to the PI.

Below are 4 arbitrary levels with increasing volume/distortion, starting clean. Yellow is a single P.I., Blue is speaker.

One thing that i will add as a reminder is that this distortion of each single P.I. output would not occur anywhere close to this early if it were disconnected from EL84 grids.

The graph of this scope in coordination with the probe 1X/10X factors aren’t quite mathing out to me yet, something seems weird with subdivision lines and instructions are effectively non existent. So take the scale note of 20V with a heavy grain of salt. Or just ignore it. My current guesstimate is that each grid lines is a factor of 4V and that there is no 5th line to view the whole scale range, lol.

The filter caps are all new, Nichicon UCY series. First few are 500V, the rest downstream are 450V. I suppose one could have been zapped enough to be partially functioning. Another thing to test.

All the bypass caps are new, MKT1813 series.

New EL84’s arrive soon today. Looking forward to popping them in.

I did stumble across this quote from someone on the interweb.. “Leaky tubes can tend to run away with themselves. Tubes that start to red plate for no reason usually will have leakage issues. If you have a high impedance voltmeter, you can actually watch the voltage at the grid climb. More of a problem with cathode bias since it usually involves higher value grid resistors but in extreme cases it can run off with fixed bias too.”

That does align with my one tube’s behavior. There was a point recently where the grid voltage of that EL84 ran off NEGATIVE instead of positive. Other than that it’s been running off positive. But it did the same exact thing, just negative. Shut it down around -3.5V. So at this point I’m just fully assuming i was sold a bad set of “matched” Mullard tubes, probably by accident because tube testers aren’t putting anything to work for any duration of time and it takes minutes for this runaway to occur.

But how much that relates to this audio distortion remains to be seen. I don’t think it relates, unless somehow both of these tubes are bad, because each P.I. looks the same (the other side looks just like these below), and there is basically 0V DC at each grid during these tests. It’s minutes later that the DC runaway occurs.

 

[CJ]

Tube sockets can become conductive and leak hi voltage into the bias circuit, but less likely on the 6BQ5 due to the pinout.

Square wave can be useful for checking ringing in audio transformers and transient response in amplifier circuits.

You have dual trace capabilities with the scope? Normal test us to compare input with output for sine wave sweep tests.

 

[systemtruck]

Not sure this reveals anything all all, but I ran a quick square wave test at low volumes just to see how the performance looked (and again, with the total circuit connected).
Yellow is direct monitoring of source before input to amp.

I’m not sure what a guitar amp should look like, but it’s basically coming out as a perfect ocean wave shaped sawtooth. It looks to me what a capacitor hanging on the signal would do to a square wave, taking some time to charge/discharge after each transient.
I am betting that the preamp output will not look this altered. And I’m also 99% certain the PI output would not look like this disconnected from the EL84’s.

This is as clean as it gets at the PI output on the left, and this is as clean as it gets on the speaker on the right.



Then when you crank the master volume in amp, the PI distorts like this on the left, and the speaker distorts like this on the right

 

[Matador]

Looking at the speaker output, the power amp is letting go of the speaker during zero crossings: isn't this kind of crossover distortion indicative of the amp trying to run in AB2? I'm wondering if there is just too much input signal driving the grids too positive

 

[ccaudle]

the grid voltage of that EL84 ran off NEGATIVE instead of positive

That would turn the tube off, not red plate. Is that what you saw occur?

New EL84’s arrive soon

The current tubes you have are what, 60 years old?

 

[systemtruck]

Ok I think I’m getting closer….

I feel like the clipping has to be speaker or load related. Or perhaps faulty wire inside output transformer bell that can be fixed. But wouldn’t all my OT tests have revealed that? Or the transformer is somehow toast yet passing all the tests I’ve tried.

I got the brand new JJ EL84’s. They don’t red plate which is good, and they remain stable at B+ indefinitely. So I can at least check that off the list. For now. I have to get this other distortion thing solved because I feel like it’s possibly making the job for tubes really really difficult and slowly damaging them.

Listen to this. Just a few chords, I’m sitting playing quietly, phone recording on the floor by cab.

https://www.dropbox.com/scl/fi/7bqz...ey=hfsunt60f4fl0mhb23zs21554&st=0ipo29t7&dl=0

When I switch between using a Webber dummy load vs using a speaker, the tone of distortion is at a very different “level”. And by level I mean the amp knobs have to be turned up quite a lot more to get the nasty clipping when the Webber is used. Mind you, not loud enough to drive the EL84’s. Talking like 10% to get the clipping with speaker, and 35-40% to get the clipping with dummy load.

A speaker clips big time at very quiet levels, as heard above. There is like zero chance I am hitting the grids too hard. This is practically a AC15 and I have played plenty of 5 and 10 and 15 Watt amps. You’d know in person if you were here. The guitar sound is voice conversation level and it sounds like a cheap fuzz pedal already.

When I put a scope on the speaker out, the tone of the fuzz changes a bit, definitely noticeable. So maybe that’s something of note.

I almost feel like slapping some kind of moderate valued power resistor across the speaker connection or adding some extra series resistor/wire somewhere.

This clipping reminds me so much of a long time ago in a studio I used an unbuffered unbalanced line out of a vintage console preamp that was modded but not properly, feeding a standard balanced line in. The Ring was shorted to ground as it should be, but still the load matchup was so messed up that the signal would spastically clip just like this at moderate to high levels.

I also resoldered the speaker jack pins to their wires from transformer and 0V, thinking maybe it was a bad solder point.

No idea yet but this is all pointing to transformer or load issue. My Webber is good, I’ve used it lots. My speaker is also good, it’s a 1950’s P12Q that I restored last year and works wonderfully with other heads.