• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Pentode push-pull screen question - voltage in relation to plate

I'm currently going through running ltspice simulating Rob Robinette's EF80-based push pull output JCM800 Marshall amp, and I have a simple question. Usually you'd have the pentode screen grid at a lower voltage than the plate but the sim indicates that the design (with a 1K screen resistor) is higher than the plate. So even if the ltspice model is wrong, the resistances and currents means that the voltage drop for the screen is smaller than the voltage drop over the transformer at idle.

However if a PP output transformer is 22.5:8, the resulting grid is 1/4 of that from my understanding so around 5K6 in terms of plate resistance, the screen resistor plus the choke is 1270R (1K+270 DCR). The diagram below gives the schematic but ignore the voltages noted on the diagram, I've given the measurements below:

Screenshot 2023-05-26 at 09.32.49.png

Note the diagram and the components follow Rob's design, although I have a separate pre power supply and so there's less current through the choke (but even an extra 5mA at 270R is only going to drop 1.35V).

B+4 = 324Vdc
B+5 = 327Vdc
Up = 288Vdc, Ik = 8.6mA
Ug2 = 322Vdc, Ig2 = 1.7mA

So you can see that the grid is sat above the plate which is not healthy from all my self-taught knowledge. Is there some difference between pentode screen use on HiFi vs Guitar amps that I'm not aware of?

If I add 20K to the screen resistors, it reduces the screen voltage below the plate and the screen current remains about the same (within spec). The plate swing drops and from my knowledge that would also make the kink more pronounced as it's not particularly negative below the plate). However I've never heard of 20K+ screen resistors (this is my first pentode output push pull).

Any wisened thoughts would be welcome.
I think your schematic is incorrect, but it is hard to tell for sure because the power supply is obscured. The B+ after the rectifier goes to the OPT center tap, then there is a choke then there is a supply to the screen. I am looking at Rob's page, comparing with the schema for JCM800 6V6.
Last edited:
Yes, I looked at that too. It looked to me that the screen supply is after the choke, and the centre tap of the OPT from before the choke. The choke has resistance, and the screens have stopper resistors, so how do you see a problem? Is it just simulated and your OPT model in LTSpice the reason for it?
The choke only has 270R. With the screen 1K, that's 1270R but the output transformer is around 5K6. Hence in DC operation the EF80 plate sits at 288V but the screen sits at 322V. So yes - it could be the ltspice model causing a problem.

The concern I have is the pentode screen sits 34V above the plate at DC idle. Everywhere I've read, the pentode sits below the plate voltage for pentode operation and at/very near for triode operation. The datasheet (Frank's datasheet here) indicates both Va and Vg2 max as 300V which would suggest that this is outside of the safe operating limits.

I know Rob runs his EF80 plate voltage at around 270V (the B+5 being 289V) which could be why, the +30V screen above plate ensures that the Vg2 is under the 300V limit.

So this boils down to this question: is a screen 34V above a plate at idle safe?

(I suspect that it's not meant to operate at +30V but the original Rob EF80 plate ~270+30 is less than 300V and so it doesn't cause a problem even if the screen is above the plate).
Last edited:
Very good point. The full specs are here https://www.hammfg.com/files/parts/pdf/125B.pdf and that DCR is 296R across the primary so half that for each side so 148R.

So my ltspice model is wrong for DCR but how do I then model additional impedance (I suppose put a 148R resistor in series with each winding)?

Don't get me wrong - I suspect that Rob's works, hence things weren't adding up in my head.
Removing the 5k6 and adding a DCR of 148 results in a screen below 2.4V which is how I would expect it to be :)

Now to test out the impedance itself (ie through the inductance turns ratio).

I normally avoid transformers, but this has allowed me to challenge myself a little.

Welcome From the Instruments & Amps threads, To the Tubes / Valves threads!

Using a 10k series resistor to the screen will drop the screen voltage, reduce the quiescent screen current; but it will also severely drop the output power, unless you use a bypass cap from the screen to ground.
Caution: the bypass cap can supply Lots of current, when the tube is clipping (more on the clipping below)
Pentodes were made to operate with a constant screen voltage.
Normally, when the plate voltage swing took the plate voltage down to 50V, the screen constant voltage at 300V, caused the screen to draw Very Large screen current.
That was OK for transients, like a loud music passages (no clipping allowed, most Hi Fi nuts do not want to listen to clipped Philadelphia Orchestra playback).
But you are talking about a Marshall (electric guitar) amplifier.
Hard clipping causes the Push plate to spend 50% of the time at 50V, and the screen to spend 50% of the time at Very Large screen currents (then the Pull plate and screen to spend the other 50% of the time at Max currents too).
Yard on that Marshall for Max Distortion: Clip, Clip, Clip, until the screens glow Red, Red, Red, and melt into the control grid, or suppressor grid, or in the case of a beam power tube into the beam formers.
Short, Short, Short. Ouch!

Triode wired came along, the screen voltage completely follows the plate voltage swing. Just like a push pull Williamson amp with triode wired KT66 Beam Power tubes.

Then along came the Ultra Linear circuit, where the screen voltage reduced with signal, typically 40% as fast as the plate voltage dropped.
Just like a push pull Dyna ST70 with EL34 tubes in Ultra Liner.

There are more details there than meets the eyes . . .
Last edited:
  • Like
Reactions: 1 user
Yup, I was half-and-half where I should post - the subject of pentode, screens and push pull along with ltspice is across a number of sub-forums.

Thank you - this is my first OT based amp, and my understanding on the screen was basically - either static, triode strapped or ultra linear so I couldn't work out the logic why a plate would be above the plate (it didn't make sense hence when logic deduction removes the options, those that remain are your causes).
I know the g3 also drops power but changes the load line characteristics quite considerably.

Thank you for your help.
Just consider some aspects that characterize instrument amplification vs hi-fi amplification: screens are at the same voltage of the plates and must drop in voltage while playing (Mesa Boogie in one of his patents gives also RC times to respect for the screen supply) and the loadline has a very low Raa that hits the g1=0 curve way on the right side of the pentode knee, passing through areas where plate dissipation is even twice the maximum allowed (that becomes ok considering that half of the time the tube is off). PSU filtration is on the edge of hum in alot of famous designs.

There are some exceptions in all of those characteristics, as usual.
I know that Rob covers the point about the overdrive and that the load line goes over the knee although for now I'll build the almost standard model with enough adjustability to allow me to alter caps etc, although that's probably a discussion for the instrument forum.
I do know that the Marshall simply made a copy of the grounding scheme and that it's possible to drop that noise floor too with some attention.

Guitar Amplifier:

For just one minute, forget about the plate dissipation when the guitar amplifier is run at 25% or 30% of rated power out.
Instead, consider the following:

1. The most often plate watts over-dissipation occurs because of too much quiescent plate voltage x too much quiescent plate current.
An EL34 with 400V on the plate and 62.5mA plate current, is right at the EL34's maximum 25 Watts plate dissipation rating.
Now, Either change the quiescent plate volts to 500V, Or change the plate current to 80 mA, and you are running the EL34 Too Hot!
Hmm, that sounds like what some guitar amplifier designers do to get more power from the amplifier.

2. Now, really wail hard on those guitar strings, and turn the amplifier gain all the way up . . .
The plate Volts spends 50% of the time at 50V, and 50% of the time at 350mA.
50V x 0.35A x 50% = 8.75 Watts plate dissipation. Cool!
Have you ever heard of an efficient switching power supply . . . that is the EL34 amplifier I have just described when the rock star really wails on the amplifier.
The only danger to the EL34 is the Melting Screens that are at 300V, and are drawing 100mA or more 50% of the time.
300V x 0.1A x 50% = 15 Watts on the screen, OUCH!
In that case, the screen is dissipating almost 2X as much as the plate is dissipating.
That is the Screen Sag and Short scenario . . . good luck!
Last edited:
Nowadays nonody plays its own amp at full blasting, and idle current is always between 60 to 70% as a general rule.

The fact is that EL34 runs between 460 and 500V on the plates with loadlines of 3.4 kOhm. Plates I would say that never reach 50 V in guitar amps, usually stay around 100 V in AB1. 6L6GC runs at same voltages and 4 kOhm.

You are correct.

But consider the following 6L6GC setup, with the g1 signal that swings to 0V, when the guitar player runs the amp into hard clipping:

Set the screen at 400V, Plate at 425V, 4k p-p primary, 4k in AB1 is 1k with one tube cutoff.
The plate current is 325mA x 1k = 325V drop, plate is at 100V.
So far, so good.
But . . . with the screen still at 400V, and the plate at 100V, the screen current will be off the tube curves.
Ouch; Sag and Short Screen scenario.
  • Like
Reactions: 1 user
Few Marshalls of the JCM800 series work at 400 V B+, most of them at 460V and above.

This, together with the fact that the lowest note on standard 6 strings guitars is 82 Hz (it can go down to 50-60Hz when using a dropped 7 string), frequency at which the speaker is entering into resonance and its impedance is 2-3 times the nominal one, makes indeed the plate go down to even less than 50 V. Screen voltage drops thanks to screen current, weak PSU node and (if well calculated) saves the screens… but not always.

I now remember a famous italian guitarist that stopped to use an early Marshall from the 60’s because they used very low screen resistor values: marvelleus sound at low to medium volume, few minutes of exstasis at full volume, then EL34 melted. Not what you want when you are on tour.

…maybe an octocoupler with the led side placed across a small value screen resistor (so that the led will start conducting when the voltage across that resistor will pass a certain limit, so when the screen current will pass a certain limit) and the resistor side to limit input signal, could help to get the low-screen sound adding safety to the amp? An RC on the led side could help to make the signal reduction like a compressor sound and not an abrupt cut.
There are few designs that exceed parts specification's maximum ratings, that can not be compensated for by putting in extra correction circuitry.
Increased complexity to fix the problem that should have been designed out in the original design.
Parallel tubes, bigger tubes, etc.

There are exceptions to the above rules . . .
The Original Tacoma Narrows Bridge . . . no compensation possible.
She does not Gallop anymore.