screen grid resistor and electron potential

[Andy Le Blanc]

I was at the bench, trying out loading and speakers, etc...

One of the speakers revealed an odd blur of distortion in the low mids.
So I looked down and thought to my self that I should swap out the 470r
screen grids for something more appropriate to the tube type.

But I figured I'd try and grab a measurement while I was at it.

So I start poking around at idle conditions checking voltages about the
push-pull circuit and the power side, and what stands out is the voltage
difference measured between the plate and g2 at the pins on the tube socket.

With the 470r as screen grid resistors the difference between the plate
and g2 was 1.7v, g2 being positive with respect to the plate.

After replacing the 470r with 1k, the voltage difference between the plate
and g2 was reduced to .484v, g2 being positive with respect to the plate.

In a beam power tube, g2 potentials accelerate the electron beam over
the hump of the virtual cathode, it affects power sensitivity and stability.
Changing the screen grid resistor effects the electron potential across the device.

The voltages are effected by the bias point, the g2 v remains steady as
the plate v rise and fall with the current as the bias is adjusted.

The up shot is that you have an easy measurement to quantify how your
screen grid resistor choice is affecting the tone you hear.
Its just the voltage difference between the P and g2 pins.

After looking back to the power supply to see how your bias point assumptions
will be utilizing the available current, you can select your g2 resistors
using a repeatable measurement. If you find a killer tone for a load a tube type
You can measure it and repeat it consistently

[Colossal]

I always enjoy your commentary Andy; very cool. Thanks for posting these observations.

[Andy Le Blanc]

The difficulty with much of the DIY approach is how to use available technical information
and get beyond qualitative observations too make rational design choices.
The only novelty here is that a simple measurement can be used to put the
kybosh on a host of subjective discussions. For the cloner it gives a hard
number to describe the mysterious, without struggling with arithmetic modeling.

With the right measurement criteria I think most circuit conditions can be modeled,
with simple tools and with good results.

G2 and bias conditions are basic to the operation of a tubed power side.
But, what is the most accessible way for the DIY to utilize or envision
what is actually going on in a circuit.

I'm beginning to think that despite the tube considerations, bias has much
more to do with the power supply than the tube.

I'm beginning to think the screen grid resistors have a larger tone implication as well.
There is an interaction with the bias point.
The g2 is a fixed voltage where bias is variable, finding an optimal
operating condition for different design expectations is the trick

[Cliff Schecht]

I've found that especially in small SE output stages, getting the screen voltage where it "wants" to be is critical to getting the best distorted sound possible and also achieving a very dynamic feel in the amp (I'll hold off on the dynamics discussion for a second).

If I'm relying on the power stage to develop a good amount of distortion, I tend to not drop the screens too much as this can make things sound too, for a lack of better term, gainy and overdistorted. But I do like some voltage drop so that the output stage acts like a true pentode instead of a triode. The "pentode sound" is what I'm looking for in a SE amp; loud, proud and full of the right mix of even and odd harmonics when the grids are driven "properly" (i.e. not overdriven too hard).

Lower screen voltages make the grid even easier to overdrive which can be a good or a bad thing depending on the tube type. Smaller tube types don't take very much signal to distort them and so not dropping the screen voltage too much keeps the power stage gain down and helps the amp clean up easier.

But this is also very dependent on tube type as Andy says. I found anything bigger than an EL34 (read: KT88) is actually pretty hard to overdrive (lots of "grid charge" on those big power tetrodes) and so dropping the screen voltage down a good amount (at 530V, I like 325-350V for g2 IIRC) makes the tube easier to overdrive and get that nice wide-open pentode sound that I love so much. But that large dropping resistor after the plate also has a big influence on how the amp is going to feel. Read below..

There is always the trade-off of how an amp is going to feel overall when you set the screen voltage. I like building single channel amps and so I want to squeeze as many different sounds out of that one channel as possible. I do this by choosing the right PT for the job and setting up my dropping resistors so that I get a lot of voltage sag when I dig in and the preamp distorts easier (not to be confused with tube rectifier "compression" type sagging on PP stages). So with this I'm already forced to drop a decent amount of voltage after the plate just because I want that dynamic feel (everything is a trade-off, you have to find the BALANCE), but I may lose some clean headroom with doing this.

The other solution to this "problem" of finding balance between the best sound and the best feel is channel switching. But IMO this is a complicated solution to a somewhat simple and elegant problem. Different people like different things though, so what I like may sound like ass to others..

I've done quite a bit of thinking and tinkering with this topic, hope this all makes sense .

[Andy Le Blanc]

More bench time, I played with a simple tube series regulator between the power
supply and screen grid resistors. Not enough current to make adjustments effectual.
But the insertion did make the plate 40v positive with respect to g2.
Wasn't a bad tone either, but I could do the same without the bother of the
extra tube, with in the power rail and or screen grid resistors.

So I simplified the notion and dressed the 12au7 I used as the simple regulator
to operate as a diode with the control grids dressed to the plates, and inserted
that between the power supply and the screen grid resistors.

This decouples the screen grids from the power rail and creates around a
2v drop at the screens keeping the plate slightly positive with respect to
g2 under static conditions.

The design center for a 12au7 class A amp is 5.5w max plate dissipation.
But as a diode with no AC signal I suspect it should fine, there was no
sign of excessive heat, the envelope of the tube was very much cooler
than one being used as a gain stage. Voltage present was a good 100v
less than the 550v max for the device.

[Gaz]

I'm really interested in what you did here, Andy, mainly because I've had troubles over-dissipating the screen grids of Pentodes before, even with large screen resistance, and conservative voltages (i.e. Plexi circuit with 420vdc B+, and 1k screen resistors). Do you have a schematic you can post? Just trying to make sense of the 12AU7 circuit.

[Andy Le Blanc]

I've cooked a few things my self....

So... I guess whats more important is that you can measure the "difference",
between the plate and g2, under static and dynamic conditions.

the diode is fun, I found in an old text book, they replaced the "crystal"
with an "electronic diode" to replicate very early radio.

you tie the grids to the plate, the plate goes to your positive power rail
the cathode to the screen grid resistors.

double up both sides of a 12au7 to get the most watts handling you can get out it.

the rational i used was to fix the bias at 50% of the PT rating and then look
closely at voltage difference between g2 and the plate under static conditions

[Gaz]

Okay, the wiring makes perfect sense now - that's really neat! So how do the tubes behave under load with this arrangement, compared to without the diode? Sorry if that's a really general question...

[Ian444]

But this is also very dependent on tube type as Andy says. I found anything bigger than an EL34 (read: KT88) is actually pretty hard to overdrive (lots of "grid charge" on those big power tetrodes) and so dropping the screen voltage down a good amount (at 530V, I like 325-350V for g2 IIRC) makes the tube easier to overdrive and get that nice wide-open pentode sound that I love so much. But that large dropping resistor after the plate also has a big influence on how the amp is going to feel.

There is always the trade-off of how an amp is going to feel overall when you set the screen voltage. I like building single channel amps and so I want to squeeze as many different sounds out of that one channel as possible. I do this by choosing the right PT for the job and setting up my dropping resistors so that I get a lot of voltage sag when I dig in and the preamp distorts easier (not to be confused with tube rectifier "compression" type sagging on PP stages). So with this I'm already forced to drop a decent amount of voltage after the plate just because I want that dynamic feel (everything is a trade-off, you have to find the BALANCE), but I may lose some clean headroom with doing this.

Cliff, and Andy, very interesting posts. Can I ask a few questions?

Cliff, do you use a screen grid voltage regulator to get EL34 screens down to around 325-350V, or is that the voltage that you want the screens to sit at under a good load via dropping resistor?

I often wonder when one biases an EL34 for example, at quiescent conditions, so the screen voltage will naturally be close to the plate voltage, that bias setting then becomes quite cold when the amp makes power and the screen voltage drops considerably. Is my thinking correct, or is this a non-issue?

Last question, I have an experimental small output tube (6P1P similar to 6AQ5) Liverpoolish amp, at quiescent the output tubes cathodes sit at 18V, under full output, about 11W, they shift up to around 24V. This would indicate too much overdriving of the output tubes? Is that why the RC dropping networks in the Liverpool/Express use such large resistor values, to prevent this by making the pre stages have less gain as the amp approaches full power? Is this part of the "balancing act"?

Thanks,
Ian.

[Andy Le Blanc]

I saw diodes used in PS first in GA 6/99 glass shard.

By a Michael Kornacker, who sites GA 6/95 article by Jerry Bouncer.
Both site Takao Kawabata

Electronics Circuit Designer's Casebook, Electronics, McGraw-Hill,
Takao Kawabata, "Single Diode Reduces Ripple in D-C Power Supply",
p. 82.

The notion pops up again GA letters 4/00 by a Lance Cochrane,
who suggest they might be used as a "poor man's regulator" on screen grids.

The notion is to isolate and decouple sections of a PS to reduce stereo crosstalk
and increase stability keeping separate channels from modulating each other
during transients and overload conditions, it does reduce audible distortion
even as I've tried to use it. I do like the simple diode dress of a twin triode
Might even give each screen its own, to insure that theres enough watt
capacity to handle the worst. but so far its working well

I wouldn't have noticed the sonic differences whether the g2 was slightly
+/- relative to the plate under static conditions if I hadn't played with it.
The 2v drop across diode was just enough and the amp was the better for it.

[Cliff Schecht]

But this is also very dependent on tube type as Andy says. I found anything bigger than an EL34 (read: KT88) is actually pretty hard to overdrive (lots of "grid charge" on those big power tetrodes) and so dropping the screen voltage down a good amount (at 530V, I like 325-350V for g2 IIRC) makes the tube easier to overdrive and get that nice wide-open pentode sound that I love so much. But that large dropping resistor after the plate also has a big influence on how the amp is going to feel.

There is always the trade-off of how an amp is going to feel overall when you set the screen voltage. I like building single channel amps and so I want to squeeze as many different sounds out of that one channel as possible. I do this by choosing the right PT for the job and setting up my dropping resistors so that I get a lot of voltage sag when I dig in and the preamp distorts easier (not to be confused with tube rectifier "compression" type sagging on PP stages). So with this I'm already forced to drop a decent amount of voltage after the plate just because I want that dynamic feel (everything is a trade-off, you have to find the BALANCE), but I may lose some clean headroom with doing this.

Cliff, and Andy, very interesting posts. Can I ask a few questions?

Cliff, do you use a screen grid voltage regulator to get EL34 screens down to around 325-350V, or is that the voltage that you want the screens to sit at under a good load via dropping resistor?

I often wonder when one biases an EL34 for example, at quiescent conditions, so the screen voltage will naturally be close to the plate voltage, that bias setting then becomes quite cold when the amp makes power and the screen voltage drops considerably. Is my thinking correct, or is this a non-issue?

Last question, I have an experimental small output tube (6P1P similar to 6AQ5) Liverpoolish amp, at quiescent the output tubes cathodes sit at 18V, under full output, about 11W, they shift up to around 24V. This would indicate too much overdriving of the output tubes? Is that why the RC dropping networks in the Liverpool/Express use such large resistor values, to prevent this by making the pre stages have less gain as the amp approaches full power? Is this part of the "balancing act"?

Thanks,
Ian.

I actually ended up using what was more like a voltage divider in the power section. It was a series 22k 10W resistor followed by a 68k 10W resistor to ground (the screen connected to the 22k/68k junction). This was a neat trick and I found that the resistors didn't get too hot and so it stayed, without the 68k resistor the screen was about 100V higher and the amp seemed to lose some of its sparkle. I haven't played around with this as much with PP designs though as all of this was on a SE design.

Why do you think the amp becomes cold? As the screens draw current and drop voltage, I think the power output would increase a bit (like moving from a triode to a pentode) and the plate current shouldn't change very dramatically (quiescently, of course the dynamic current changes!). If the power stage does become cold (i.e. from being pushed too hard in a PP amp) then this is a different story as you will get the dreaded crossover distortion (exacerbated by negative feedback too) and high frequency dynamic buzzing sound that sounds almost like a PWM happening.

Does the output signal (at the speaker) start looking like it has crossover distortion when pushed too hard? That bias shift is indicative of the grids charging from too much signal and causing a bias shift. The fix is a simple pair of back to back diodes on each grid (google "antifizz diodes" for more info on this).

Andy, this is an interesting discussion. Perhaps you could post a schematic and/or some oscope screenshots to help us visualize what's going on? I think I get what you are doing though. My thoughts are that the screen should always sit at least a hair lower than the plates if you aren't going for triode-strapped operation. The screens present a form of internal negative feedback on the power stage and where they sit has a big effect on how the power stage will sound (as I think you noticed in your first post). As G2 gets higher in potential it accelerates electrons less and less until you get to a point (when G2 is higher than P) where G2 is actually stealing current from the power stage and reducing power output. IIRC the relationship with the NFB is that the lower the screen voltage, the less internal negative feedback you get (and more power stage gain, again good or bad depending on what you are looking for).

[Ian444]

Cliff, thanks for the answers and info. Plenty of food for thought. I'll do some more scoping and see if I can't learn something.

[Andy Le Blanc]

If you have a voltage swing +/0/- , you have a signal or modulation.
I was very surprised that screen grid voltage of only + 0.5v relative to plate
caused an audible distortion, the g2 voltage started at + 1.5 with a 470r
moved to around + 0.5 with 1k sgr and then the insertion of the diode before
the sgr created a 2v drop that moved the point to - 1.5v relative to plate.

Look back to the original condition, g2 positive relative to the plate.
the plate becomes more positive with signal, which swings its relationship
to g2 relative to frequency and signal strength, through out and on both sides of the push pull cycle.

The distortion "color" shows up relative to bias, the "pentode tone" might very well be
distortion created by a poorly conceived power side.

In order to maintain a tight control on the relation ship between g2 and the plate
every bias point could have its own set of screen grid resistors.

having a diode in the screen circuit looks like it will prevent "other" issues
that happen when the amp is operated in overload for an extended period.

Lets say the anode is responding to a large signal, or a condition where the
PS can't respond or maintain current, the anode will drain the PS (sag)
until the g2 voltage drops too, a diode will change the time constant
of the PS section, preventing the condition from effecting the screens

or... say you have the diode in the screen circuit, the sag will happen
but a condition where the screen will conduct is prevented as the plate
voltage is starved, the screen won't become the plate, exceeding its watt limitation.

[Andy Le Blanc]

A little more time on it today, and couple interesting notes...

I dressed each screen grid its own 12au7 diode before a screen grid resistor.

separating the g2 revealed strongly the variations in tubes, which showed up
as different voltage measurements between the plate and g2.

Quit a time spent digging thru the bench tubes looking tubes with a similar screen characteristic

The extra tubes ran up the heater current and forced me to fall back on a
power tube type with a lesser heater rating.

which I biased to the same 50% of the PT current rating...
the difference became a power tube biased to 40% of rated plate dissipation
for one biased to 100% of its rated plate dissipation...

The tone was different,but the amp responded the same, surprisingly predictable.

It seems that keeping the plate 1-3v+, enough to get around tube variation
avoids some distortion/modulation issues that happen as the plate
swings positive with signal in reference to the screen grids.
I started to hear a difference, it started to get darker around 5-7v+.
Its worth keeping an eye on.

Choosing a bias point based on the PT rating is interesting.
The results regardless of the valve are consistent so far.
As long as the valve will handle the dissipation, they become interchangeable.

disposable,

The PT and the OPT (loading) become the corner stones in the design.

[Cliff Schecht]

I'm surprised you found a lower G2 voltage to be darker, I would think the opposite would be true. I guess it depends on how the plate resistance and damping factor play into the equation..