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

[systemtruck]

That tube is sourcing grid current into it's grid leak resistor, if you've disconnected the coupling caps and it still happens. If you are saying the pin 2 of a "bad" tube is charging upwards of 12V then it's sourcing north of 20uA which is a LOT.

If you want to be triple sure this is the root cause, I would hard ground pin 2 of both tubes with an alligator clip lead from pin 2 directly to your audio 0V bus bar. The tubes should DC idle under those bias conditions all day, as no amount of grid current can overwhelm your 0~0 ohm wire.

I don't know of any underlying mechanism for this (generally grid current is a function of running the cathode and grid at similar potentials, but that isn't the case here).

With input grids grounded to bus bar, (and with input clipped again post bypass caps so that I didn’t damage PI stage with such a low load), sure enough, it completely stabilizes and stops red plating. I let it go for a long time and it’s fine. This is using the newly aquired matched pair, where one used to red plate. Now both tubes glow less and glow evenly as well.

I’m really starting to think both of my issues, the red plating and the PI distortion when EL84’s are connected, are Fisher design issues. First of all they’re running the tubes at much higher voltages than everyone says is ok to do. So maybe that itself causing this buildup of DC because it’s simply overloading the tubes internal capabilities?

I feel like maybe a compromise can be made with the grid leak value….

http://www.r-type.org/pdfs/el84.pdf

This seems to be the Mullard EL84 datasheet that is right for these tubes but maybe it’s not the same generation. My tubes are from the 50’s or 60’s.

I couldn’t figure out from the datasheet what the recommend grid leak resistor value should be. But maybe what I can do is just try lowering it down from 490 but not so low that it causes problems either in the EL84 nor in the PI stage loading.

Maybe I’ll try adding a pair of same value in parallel via these alligator clips. That’ll half the grid leak resistance to around 250K each. I can let it sit again, and if it doesn’t red plate after several minutes, maybe it works.

 

[systemtruck]

The new “matched pair” installed, one having the red plate issue..

With PI outputs disconnected, the following grid leak resistor value was tested (parallel stuff, so weird values here..)

245k, it red plates
196k, it red plates
114k, it red plates, takes longer of course
65k, red plates, takes even longer, several minutes

Will there be a stopping point before just plain shunting to ground? Kind of seems like no, and that eventually the DC will build enough to rock the boat.

I suppose one theory is that the EL84 design is not supposed to handle this level of voltage, and that only the strongest ones will bear it but those will eventually give way.

But if that were the case, Fisher would never have been able to support this product line, so I don’t think that’s the situation.

 

[Matador]

The maximum grid resistance to cathode is listed in the "Limiting Values" section, labelled Rg1-k, and it's 300k. 220K is a common value, but keep in mind you might need to adjust the values of the PI coupling caps to keep the frequency cutoffs the same (if you halve the grid resistance, you need to double the coupling cap value, all other things being equal).

The limiting longevity factor is almost always not plate voltage, but rather screen (dissipation). Most robust designs almost always strive to limit screen dissipation instead of plate dissipation. I might suggest swapping in 470R or 1K screen grid resistors and seeing if you can stay below 300V max on the screens, even if the plate voltage is over spec.

Although I don't know of any second or third order effect where screen voltage changes or modulates grid current, but maybe you've discovered one.

 

[systemtruck]

The maximum grid resistance to cathode is listed in the "Limiting Values" section, labelled Rg1-k, and it's 300k. 220K is a common value, but keep in mind you might need to adjust the values of the PI coupling caps to keep the frequency cutoffs the same (if you halve the grid resistance, you need to double the coupling cap value, all other things being equal).

The limiting longevity factor is almost always not plate voltage, but rather screen (dissipation). Most robust designs almost always strive to limit screen dissipation instead of plate dissipation. I might suggest swapping in 470R or 1K screen grid resistors and seeing if you can stay below 300V max on the screens, even if the plate voltage is over spec.

Although I don't know of any second or third order effect where screen voltage changes or modulates grid current, but maybe you've discovered one.

Gotcha, thanks!
I already inserted some extra resistance before the second rail and the screen grid voltage now sits around 290V whenever the first rail is 340. The downstream rails were already too high also, so this helps everything.

I added another post just before yours here, with my numbers after lowering the grid leak incrementally. No luck! The tubes that want to red plate will indeed red plate eventually, they’ll just be slowed a bit. However, maybe there is a value that is low enough where it will allow audio and not be TOO low. But what is too low for a grid leak? I had it down to 65K and they still red plate slowly.

I suppose at this point I just have to throw my hands up in the air and conclude that I bought a matched pair that had a dud in it, and that my old tube pair also had a dead one.

So I’ll stick with the cold one from each set and just roll with it, and see if I can adjust the PI circuit to stop the ugly low level clipping.

 

[CJ]

One tube red plating is normal.

It is just like thermal runaway in a transistor.

Usually the tube with the higher gm will be the one to run away.

As it starts to lose its bias, it draws more current, which makes it hotter, which makes it draw more current, rinse and repeat.

New tubes can not take some of the old school plate voltages. So you will need to lower your plate voltages, or drop your screen voltage. But if you drop your screens too much, your sound will suffer. So maybe a little of both, lower B+, lower screen.

And you need to check this amp with a scope, stage by stage, to map out the gain structure.

Ultrasonic Oscillation is another topic that should be crossed off the list. But usually when this problem persists, both tubes will red plate at the sane time.

You want to hear a weird story about current hogging? One time my boss went to check the current draw on a large generator circuit. Big 1/0 cables were being used, two for each leg of the circuit, very low resistance. When he put his clamp on meter across one of the cables he got very low current. When he put it on the other cable that was in parallel, he got very low current. WTF? Turns out the insertion loss of the meter was enough to force all the current through the other cable.

This is kind of what us going on in your circuit, the higher plate resistance of one of your tubes is forcing current into the other tube. However the imbalance will not he as radical as the 1/0 cable thing.

Here us a good article by tube God Eric Barbour>

 

[systemtruck]

One tube red plating is normal.

It is just like thermal runaway in a transistor.

Usually the tube with the higher gm will be the one to run away.

As it starts to lose its bias, it draws more current, which makes it hotter, which makes it draw more current, rinse and repeat.

New tubes can not take some of the old school plate voltages. So you will need to lower your plate voltages, or drop your screen voltage. But if you drop your screens too much, your sound will suffer. So maybe a little of both, lower B+, lower screen.

And you need to check this amp with a scope, stage by stage, to map out the gain structure.

Ultrasonic Oscillation is another topic that should be crossed off the list. But usually when this problem persists, both tubes will red plate at the sane time.

You want to hear a weird story about current hogging? One time my boss went to check the current draw on a large generator circuit. Big 1/0 cables were being used, two for each leg of the circuit, very low resistance. When he put his clamp on meter across one of the cables he got very low current. When he put it on the other cable that was in parallel, he got very low current. WTF? Turns out the insertion loss of the meter was enough to force all the current through the other cable.

This is kind of what us going on in your circuit, the higher plate resistance of one of your tubes is forcing current into the other tube. However the imbalance will not he as radical as the 1/0 cable thing.

Here us a good article by tube God Eric Barbour>

Great story! I suppose the solution would have been to have TWO identically performing current metering rigs and have both cables monitored at the same time? They’d have to be clinically identical though.

At one point i had 82k grid stoppers going to my EL84’s, and the sound was quite dark. I reduced those to 41k which are still in place and the tone is still fairly dark. So I’m curious, doesn’t that rule out ultrasonic oscillation within the EL84 stage? I thought about ultrasonic oscillation and ruled that out due this heavy filter in place. I mean as it stands i think you’d have trouble passing 7k through it haha.

I’m going to do one more output stage test and move forward working on the PI, ( which I’m probably going to switch to become a long tailed pair.)
My test is that i will put in the two “cold” non-red plating tubes, so one old tube that was in the amp and one newly bought tube that was from a matched pair, and i will record some audio. Then, I’ll put in the matched pair and record audio of that before its hot tube starts red plating. Then I compare the two recorded sounds. I should have a solid minute or so to play the matched pair before it red plates.

Basically i want to be able to determine if my two cold tubes will perform similarly to my matched pair. If so, I’ll just use the cold set and move forward and accept the fact that for some reason these two tubes don’t red plate with each other in place. Interestingly, because i am powering up using a variac i have been dialing in a perfect 340V at the EL84 plates each time, no matter what the tube configuration. When i change the EL84 configuration the rail behaves a little differently. So in theory i can do this test rather accurately.

So if it seems my two cold tubes are a decent set to commit to, at that point I’ll begin changing the PI to see what sounds best. I think ill just jump right into the example schematic from @merlin here : https://valvewizard.co.uk/acltp.html

The goal of changing the PI is to lose this horrible breakup sound that occurs when the output stage is hooked up. This cathodyne stage is reacting very poorly to the EL84 stage, and it happens no matter if the red plating happens, or it’s new tubes, or old tubes, etc. So i think it’s just the PI that’s causing it. When the connection is broken from PI to EL84’s, the PI output performs much better but the same breakup occurs after you bring up the volume a bunch. So the PI is collapsing when connected to the output circuit.

With trepidation, I’d like to officially conclude that there have been two issues with my amp..
A) the voltages are so high that EL84’s have a tendency to red plate no matter what condition they’re in, and this has NOTHING to do with the ugly clipping
B) the PI design sucks for the application i want to use it for. It was designed for consumer radio/phono amplification in the 1950’s. Perhaps it was always just going to be amplifying very quiet sources and only max volume would ever cause any kind of breakup.

I do have one question though… what audibly happens when “your sound will suffer” if i lower the screen grid voltage too much on its own? I do also want to note that I’ve experimented with lowering the whole rail system, since i have the variac on tap, and the red plating still occurs even when i lower the whole thing by say 10-15%. So say a 300V plate V instead of 340V. I could try lower though.

 

[Cqwet Dbdfte]

I visually re-inspected all points in the PI and EL84 pins…all looks good.

New tests:
Removed EL84’s, tested all pins to ground.
Pin 2 measures 510K to ground on both sockets. This is my grid leak resistors.
Pin 3 is 225R, and that is because i recently added a 100R to the 125R to see if that helped anything. Maybe that’s a bit high in the long run, so will lower. But this has the cathode bias sitting at 12V instead of 9.5V.
Pins 7 and 9, the hot DC pins, each have very high 1-2M impedance and slowly sinking impedance due to the capacitors.

Then I put the EL84’s back in, and I snipped the connections from the pair of bypass caps coming from the PI, added a pair of 470k resistors in series coming out from each cap, connected this balanced signal to a TRS input on audio interface, and powered everything up as normal and measured DC post caps just to see if there is some leakage coming through the bypass caps that had nothing to do with the EL84’s. There may be a little bit but i don’t know enough about cap technology to say it’s a problem… they both rise up to around 20mV during power-up surging, but settle back down quickly to around 5mV to 12mV. Both caps do this. And they remain there for a while. I thought maybe that would be enough leakage to feed the grid and cause a surging DC situation, but even with the connection from PI to EL84’s completely severed, the one EL84 still red plates. And when it red plates, interestingly these detached cap outputs settled down a bit further to almost 0. I think one side settled lower than the other, in fact in went negative maybe -0.005, and the other cap remained positive around 0.005. These are brand new MKT1813 caps. Is that sufficient? I am guessing that the reduction here during red plating is due to the overall B+ collapsing during red plating.

But again, this test means the EL84 red plating is not directly related to any tiny DC output from the PI. Maybe it’s not helping and maybe it’s slightly exacerbating it, but it’s not the root cause. And yes, the 470k grid leak resistors were still in place when i ran this EL84 test that red plated. The only thing detached from the ELI84’s was the audio input lead post the bypass caps. My currently existing 41k grid stoppers were just floating mid air.

I also listened to the audio of the PI output. Just like the last time i did this, the sound seems excellent. Mind you this time i had 470K resistors feeding a 15k audio input, so this made quite the audio level drop due to voltage divider there, so I had to turn up the preamp and output volume of the amp to get a good level. Just like last time, it was pretty much perfect until it does eventually start clipping in the PI itself but this in theory is WAY beyond the point where the EL84’s would be driving with their amazing drive sound so it is not even close to the same clipping level. The normal clipping happens even at low volumes.

The only thing i can possibly conclude right now is that not only is one of my old tubes bad, one of my new “matched pair” tubes is ALSO bad, just by some bad stroke of luck. Like can a tube cause it’s own DC surging/leaking?

But even using both “cold” tubes as a pair, although it stabilizes the situation and nothing red plates, it does NOT solve the early clipping level….

As I mentioned this yesterday, i hooked up the one OLD cold tube and the one NEW cold tube, and by that i mean the tube from each pair that didn’t ever red plate, and the system becomes stable and no tube red plates even after sitting for several minutes (red plating has been occurring around 1.5 minutes i would estimate). So….

The PI distortion is still there. But this has me wondering, maybe i have two usable tubes on hand now, and maybe i need to adjust the circuit that somehow sucks by design? Like if I have two tubes that i can pop in the EL84 sockets and the DC situation doesn’t red plate, I’m one step closer to this functioning. However, what would then be adjusted to stop the PI from clipping at low levels?

Or maybe i just need to abandon this PI design and change it to a long tailed pair or something else? I hear cathodyne PI’s are known for clipping ugly. But my audio test proves that the cathodyne early clipping is caused by the load the EL84’s are putting on it, or something like that.

Here is the latest schematic and i think i got GroupDiy’s platform to insert it visually now..



View attachment 146255

Grid "leak" of 0.5 meg may be optimistic. Try a lower value, 10k, for a test, if your tube turns red with this, the tube would be faulty. Grid stopper should be like 1k and on the tube socket, to prevent HF response. Some use a ferrite bead.
Check your circuit with a scope for HF oscillation.
The screen grids may act like a small "plate", circuit designs differ.
Measure the grid 1 voltage. It should be 0V.
My preference for a screen supply is a well regulated voltage source, not an RC network.
A high value grid stopper is not a valid proposition.
Old school tube circuits had no provision for driving grids (loads) into a positive region, where the load (Power tube) could have good linerity.
Using a MOSFET active current source on the driver tubes plates and connecting the load to the FET's sources could allow a lower source impedance and positive grid voltages.
However, an weeny EL84 would not be a candidate for high power anyway.
The perfect phase splitter does not exist, but a well designed transformer maybe close, but brings its own set of design tradeoffs.
The classic long tailed pair could be considered, but may need a high negative voltage source, or use semiconductors to arrange a common current source for the phase splitter, LM334 could work, have not tried.