Screen voltage compared to Plates

[rschiller]

For lower powered two 6V6 amps over the years that do not have a choke, Fender has used a variety of values in the first drop resistor. Generally standardizing in the Princeton-Rev fender used a 1k. Earlier amps such as Harvard 5F10 used 470r, Tremolux 5E9A 2k5 and Deluxe 5E3 4K7. I don’t know if bias scheme matters though Tremolux and 5E3 with higher values are cathode biased whereas other amps are fixed bias. A larger drop resistor from reservoir cap to screen cap will lower the screen voltage as compare to plates.

What are the design parameters here, if any, and any difference in tone, efficiency of tube operation, touch sensitivity or ?? as the screen voltage becomes lower from the plate voltage?

[Roe]

you loose power, attack, gain, and aggression as you lower the screens voltage significantly

[pdf64]

Anode current is determined by the relative magnitudes of the control grid and screen grid voltages.

Have a good look at the 'average plate characteristics ' charts on p4 (screen grid at 250V) and p7 (screen grid at 400V). For a given anode and control grid voltages, consider how much higher anode current is with the screen grid up at 400V compared to 250V.


With signal, the AB1 loadline is chosen such that peak anode current occurs when the Vpeak of the signal voltage = the bias voltage.
Look at the upper plate characteristics chart p6, control grid voltage =0, and consider how anode current at the knee increases as screen grid voltage increases.

GE 6L6GC https://frank.pocnet.net/sheets/093/6/6L6GC.pdf

[nuke]

What are the design parameters here, if any, and any difference in tone, efficiency of tube operation, touch sensitivity or ?? as the screen voltage becomes lower from the plate voltage?

If you review the old tube data sheets, you'll tend to see in most cases the screen has a lower maximum voltage than the plate. Most guitar amps just ignore those limits altogether in the quest to get more power output. Watts sold amplifiers back in the day. If you said you can make 22 watts and your competitor said 18 watts, you'd probably outsell the other guy. Besides what guitar player ever turned the amp down?

In beam power and power pentodes ( 6V6 and 6L6 vs EL84 and EL34 for instance), the plate current for any given control grid input is governed by the screen voltage. The plate can even draw enough current through the transformer to become lower than the screen (ie, negative with respect to the plate). Thus the screen acts like a "virtual" plate. Let's us get a lot more clean power than say, a triode does. (which you can also create by tying the screen and plate together).

As the plate voltage drops under higher-current peaks, the screen current goes up. At some point in the curve as the plate current increases and the plate becomes more negative with respect to the screen, the suppressor grid or the beam-forming plates can no longer prevent electrons from returning from the plate. Screen current thus grows at an increasing rate, causing the screen to heat and even driven hard enough, to deform and melt. (a common form of power tube failure).

The screen grid resistor helps to limit the current input to the screen. But if you work on amps, you'll see these resistors are often overheated or failed due to misuse or from damaged screen grids in the tube.

Back in the day, tubes were mass produced and relatively cheap and easy to come by. Even drug stores had them. But a pair of output tubes is $60 USD or more today.

Reducing the screen voltage to something sane reduces the output power. But, we also need the screen supply to be "stiff" meaning well bypassed, so that the screen supply rail is not sagging with response to screen current draw. We also need a minimal amount of resistance between the screen grids and the supply rail.

So my current philosophy is to set the power supply voltage divider to a point that's within the datasheet limits for the particular output tubes, with a larger filter capacitor than would have been historically used in vintage amps, and perhaps, a smaller screen current limiting resistor. In some cases, maybe just removing the screen resistors entirely.

This will give you the same sonic performance as running them as vintage amps did, but with reduced RMS power at the speaker. They'll have the same sparkle and transition into overdrive, but it will happen at a lower power level at the speaker.

The payoff is improved tube life, potentially greatly improved. Some well-designed (not guitar amps) tube amplifiers from back in the day, could get thousands of hours of life from their output tubes. The Bendix Red Bank data sheet for the 6094 (electrical eq of a 6AQ5 or 6V6) indicates a 10,000 hour life expectancy when used conservatively and significant degradation of expected life when the power levels are increased.

In the modern era of expensive tubes of questionable quality and reduced stage volumes, it may make sense to reconsider the operating point of the screens in our output tubes. If you really need 60 watts or 100 watts, then go with the old way. But if 45 watts instead of 60 watts is a reasonable power target, operating the screens at a reduced voltage might be a great reliability improver.

What harms the sound is a screen that connected to a saggy power supply that dips a lot when the screen current goes up.

[Helmholtz]

What harms the sound is a screen that connected to a saggy power supply that dips a lot when the screen current goes up.

I tend to disagree.
The stiffer the screen supply, the more pentodes/tetrodes sound like triodes.
Being a blues/rock player I love the screen compression of vintage amps (think JTM45) due to sagging screen voltage and I hate triode (mode) outputs.
So its a matter of playing style/personal preference.

Agree on too high screen voltage issues, though.

[B Ingram]

... As the plate voltage drops under higher-current peaks, the screen current goes up. ... Screen current thus grows at an increasing rate, causing the screen to heat and even driven hard enough, to deform and melt. (a common form of power tube failure). ...

This situation only happens (in an adverse way) when the screen is at 300-400v and the plate voltage has been pulled down to <100v.

The situation is worse with true pentodes (EL34, EL84, some 6BQ5s) and less severe with beam power tubes (6550, KT88, KT66, 6L6, 6V6, some 6BQ5s).

The screen grid resistor helps to limit the current input to the screen.

Assume the voltage applied to the screen comes from an ∞µF capacitor, so it's absolutely unchanging.

When screen current goes up, screen dissipation goes up. But "higher screen current" mostly happens in a blip around peak plate current, when the plate voltage is pulled momentarily down to ~100v or less.

High Current x Same Volts = More Watts Dissipated ---> too many watts?

We need a way to reduce screen dissipation, but Screen Current is the consequence of things happening inside the tube we mostly do not control. What we can change is the voltage at the screen, by inserting a resistor between screen & filter cap. Now Screen Current x New Resistor = Voltage Reduction ---> Screen Volts drop somewhat when screen current "blips" high.

Turns out the screen dissipation is a matter of Average Heat over the entire signal cycle. We don't need the "momentary screen volts x screen current" to be less than the screen dissipation rating at all instants. We just need the average dissipation over time to be reasonable. So amp designers tend to use the smallest resistor that gives that acceptable "average dissipation over time."

[B Ingram]

What are the design parameters here

Anode current is determined by the relative magnitudes of the control grid and screen grid voltages.

Amp Designers start by deciding an amount of output power they want the amp to have. They might also have a rough mandate for total supply voltage based on the limits of the tubes, and available power transformers.

The screen voltage sets the amount of Peak Plate Current possible from the output tube(s) used.
Peak Plate Current and Output Transformer Primary Impedance determine the Power Output the power section can deliver.
So we could say that if reaching for a lot of Output Power, that a stable & high Screen Voltage is important.

... Fender has used a variety of values in the first drop resistor. Generally standardizing in the Princeton-Rev fender used a 1k. ... Tremolux 5E9A 2k5 and Deluxe 5E3 4K7. I don’t know if bias scheme matters though Tremolux and 5E3 with higher values are cathode biased whereas other amps are fixed bias. ...

5E9 Tremolux and 5E3 Deluxe were cathode biased, and they're barely out of Class A. They make relatively little power output from their pair of 6V6s (12-15w).

• These require a Peak Plate Current of only ~double the idle plate current of 1 tube. That means a high screen voltage is not needed to achieve a high peak plate current. Idle bias (of 15-19v) is relatively low.

• An AA1164 Princeton Reverb is not a great example (because it doesn't push for "high power"), but it has a higher B+ voltage, higher screen voltage and strives to generate a larger output power. It also winds up using a larger bias voltage (like 34v or so) to keep the 6V6s from overheating by turning them off part of the signal cycle.

• A Deluxe Reverb uses basically the same B+ and bias as a Princeton Reverb, but has a lower OT Primary Impedance. That allows the 6V6s pull more current before Plate Volts drop too low, and this amp manages 22+ watts before running out of poop. Screen Volts suddenly become more important in this amp for its higher power output, so a low-resistance choke is used the keep screen voltage from being pulled down when driven.

... Fender has used a variety of values in the first drop resistor. Generally standardizing in the Princeton-Rev fender used a 1k. ... Tremolux 5E9A 2k5 and Deluxe 5E3 4K7. I don’t know if bias scheme matters though Tremolux and 5E3 with higher values are cathode biased whereas other amps are fixed bias. ...

Folks soemtimes find their cathode-biased amps run too hot, and they jump to raise the cathode resistor value (thereby increasing the bias voltage, and lowering the plate current).

For a while I've pointed out the screen voltage in these amps is much higher than needed by a Class A design (where peak plate current capability only needs to a little bit over double the idle current). So instead I recommend increasing the power supply dropping resistor feeding the screen to bring that voltage down. This lowers plate current without needing to increase the cathode resistor value.

The knock-on effect is that the power stage distorts when the drive signal into the power tube(s) equals/exceeds the bias voltage. So raising the cathode resistor (which raises bias voltage) makes the power section harder to drive, and instead lowering the screen voltage makes the power section easier to drive.


IMHO, YMMV, and all that, as amp-design is a matter of balancing a lot of trade-offs to push towards one outcome and away from other possible outcomes.

[nuke]

This situation only happens (in an adverse way) when the screen is at 300-400v and the plate voltage has been pulled down to <100v.

The situation is worse with true pentodes (EL34, EL84, some 6BQ5s) and less severe with beam power tubes (6550, KT88, KT66, 6L6, 6V6, some 6BQ5s).


Assume the voltage applied to the screen comes from an ∞µF capacitor, so it's absolutely unchanging.

When screen current goes up, screen dissipation goes up. But "higher screen current" mostly happens in a blip around peak plate current, when the plate voltage is pulled momentarily down to ~100v or less.

High Current x Same Volts = More Watts Dissipated ---> too many watts?

We need a way to reduce screen dissipation, but Screen Current is the consequence of things happening inside the tube we mostly do not control. What we can change is the voltage at the screen, by inserting a resistor between screen & filter cap. Now Screen Current x New Resistor = Voltage Reduction ---> Screen Volts drop somewhat when screen current "blips" high.

Turns out the screen dissipation is a matter of Average Heat over the entire signal cycle. We don't need the "momentary screen volts x screen current" to be less than the screen dissipation rating at all instants. We just need the average dissipation over time to be reasonable. So amp designers tend to use the smallest resistor that gives that acceptable "average dissipation over time."

The screen grid is a metal structure and like any other metal structure, it will be damaged by heat when the rate of power going into it, exceeds the rate at which heat is removed from it and the temperature exceeds some critical point where it deforms and fails. Doesn't matter if it is electric current, a torch or a laser beam, or whatever. Carefully made metal structures will fail if you get them hot enough. I think we all agree on that.

I fully understand the dynamic nature of AC circuits too. Take my word for it.

I have also used beam power tubes as linear pass elements in regulated power supplies as well as in audio amplifiers.

6L6GB-screen-curves.jpg

If we examine this graph from the 6L6GB data sheet, the conditions are:

Control grid at 0v (ie, no signal, but no bias, basically wide open or maximum signal peak without grid current).
Each curve depicted is a particular, constant screen supply voltage.
Plate voltage range of 0 and 600v.

The x-axis is the plate voltage.
The y-axis is the screen current in relation to the plate voltage, for each curve drawn at a constant screen voltage.

Like I said, the screen current is relatively constant while the plate is higher than the screen voltage.

However if you look at the slope of the curves as the plate voltage is reduced, the slope (derivative) of the curve begins to show increased current as plate volts approach the screen voltage. The rate of change (second derivative) increases as the plate becomes negative with respect to the screen, until it reaches the very sharp knee at the left side of the graph.

And yes, these are essentially static conditions depicted, not audio, and G1 is 0v.

So, yeah, but we know the screen current is also affected by the G1 voltage. All true.

Now you mention the "cycle". If all we ever did was amplify clean sine waves, that would be really nice of course. Little short peaks on the top and bottom of the wave form, but the average as you say, is much lower, and with any good fortune at all, the screen is not damaged. But with clipping and asymmetric waveforms, things can get ugly and heat value of the signal may be a lot worse than a sine wave.

Since we're all stupid guitar players, and we turn these things up loud and clip the signal, sometimes really hard. That will increase the amount of time the output tubes are operating at high plate current, which of course, means lower plate voltages, perhaps negative with respect to the screen and then we get screen heating.

And that's why I've replaced so many burned screen resistors and output tubes for the last several decades in people's guitar amps, where the screens are typically supplied approximately the same voltage as the plate supply. ie, 450v plate and 440v screen at the typical 50 watt 6L6GC AB1 push-pull. With maybe a small screen resistor in series with each output tube screen.

[B Ingram]

If we examine this graph from the 6L6GB data sheet, the conditions are:

Control grid at 0v (ie, no signal, but no bias, basically wide open or maximum signal peak without grid current).
...
Like I said, the screen current is relatively constant while the plate is higher than the screen voltage.

... the curve begins to show increased current as plate volts approach the screen voltage. The rate of change (second derivative) increases as the plate becomes negative with respect to the screen, until it reaches the very sharp knee at the left side of the graph.

And yes, these are essentially static conditions depicted, not audio, and G1 is 0v.
...

The data sheet graphs are being misused and/or misinterpreted, so wrong conclusions are drawn.

In the graph you posted from Page 6 of the G.E. 6L6GB data sheet, screen current doesn't skyrocket until the plate voltage is down around 50v. That as little as 1/5 of the screen volts, so it's not a matter of "the plate voltage is somewhat negative of the screen."

6L6GB-screen-curves-1.jpg

The real question is, "Will our tube have that plate voltage?" And we only know that if we plot a loadline on the Plate Curves and investigate what happens with signal.

• Since the graph used up to now has "Ec2 = 250v" let's use the condition on Page 2 of the data sheet for "Class A Amplifier, Pentode Connection." This is a single-ended 6L6GB with 250v plate, 250v screen, -14v bias, and 72mA zero-signal plate current.

• Let's drop a Blue Dot at 250v plate and -14v on Ec1 (which is at 72mA plate current).

• We note the data sheet condition used a 2.5kΩ load impedance. We can draw a loadline by dividing 250v / 2.5kΩ = 100mA, and adding this to the idle current to get a total of 172mA. Now we can draw a Green Load Line through 0v, 172mA and 250v, 72mA to investigate the result of applying signal.

• If the positive-going signal has reached +6v, the resulting Ec1 voltage is -8v. We see loadline intersects "Ec1 = -8v" at about 162v, and so we look at the dashed "Ec1 = -8v" curve for screen current. We see the curve crosses "162v plate" at about 10.5mA. These observations are marked with Purple Lines/Dot.

• If the signal keeps increasing to 10v peak, the combined signal + bias leaves -4v on Ec1, and I've dropped an Orange Dot on that point of the loadline. This happens at about 95v on the plate, and I've placed a dashed "screen current at Ec1 = -4v" curve halway between the -8v and 0v curves. Reading screen current from the dashed -4v curve at 95v, we see about 16.25mA of screen current. These observations are marked with Orange Lines/Dot.

• If the signal keeps increasing to 14v peak, the combined signal + bias leaves -0v on Ec1, and I've dropped an Red Dot on that point of the loadline. This happens at about 37v on the plate, so reading current from the dashed 0v curve at 37v, we see about 27.5mA of screen current. These observations are marked with Red Lines/Dot.

Screen Current vs Plate Load.png

• The operating condition on Page 2 of the data sheet tells us zero-signal screen current is 5mA. Screen current rose along the half of the loadline to 10.5mA, 16.25mA, and 27.5mA.

• However, the other half-cycle of signal would cause screen current to fall. It should reach a minimum at -14v x 2 = -28v, which we can see with be maybe 1.25mA.

• For screen dissipation, the average screen current over the signal cycle is what matters. Rather than work through the other half of the loadline, let's guess screen current too-high and say it falls to 4mA, 3mA and 2mA.

• Average Screen Current = (27.5mA + 16.25mA + 10.5mA + 5mA + 4mA + 3mA + 2mA) / 7 = 9.75mA worst case.

• Screen Dissipation = 250v x 9.75mA = 2.44 watts ---> less than the data sheet limit

.


Now we really should have assumed equal-angle intervals of signal input, and found screen currents for those voltages. Had we done that, there would be less time at the peaks, and the Average Screen Current would have come in lower than estimated above. In fact, that's what is shown on Page 2 of the data sheet, where Maximum-Signal Screen Current is only 7.3mA average.

Max Signal Screen Current.png

We also see this confirmed on the last page of the data sheet, where Power Output and Distortion are plotted for different Load Impedances, but for our same condition: 250v Plate, 250v Screen, -14v Bias. Note also this graph assumes the output tube is fully driven, as 9.9v RMS x √2 = 14v. We read the Screen Current for 2500Ω Load Impedance, and see a bit less than 7.5mA Screen Current.

Power vs Load Graph.png

Now you mention the "cycle". If all we ever did was amplify clean sine waves, that would be really nice of course. Little short peaks on the top and bottom of the wave form, but the average as you say, is much lower, and with any good fortune at all, the screen is not damaged. But with clipping and asymmetric waveforms, things can get ugly and heat value of the signal may be a lot worse than a sine wave.

Since we're all stupid guitar players, and we turn these things up loud and clip the signal, sometimes really hard. That will increase the amount of time the output tubes are operating at high plate current ...

Be careful going there...

Over-driving the output tube grid causes its impedance to fall from ∞Ω to around 1kΩ, which clamps the output of the phase-inverter/driver, and prevents the grid from being drive positive of the cathode much.

Meanwhile, the output tube grid passes grid-current, charges the coupling cap, shifts output tube bias more-negative. The output tube doesn't block the negative-going part of the signal, so cathode current is in cutoff longer, which means screen current is 0mA longer. Screen dissipation might go either way as it remains at higher screen current longer, but also zero screen current longer.

Those things are best sorted by using an o'scope to monitor voltage drop across a 1Ω series screen resistor, as it is driven with square-waves of various sizes to determine worst-case average screen current.

[nuke]

I think you're mischaracterizing my words.

I'm not misusing anything.

[rschiller]

Nuke posted: "And that's why I've replaced so many burned screen resistors and output tubes for the last several decades in people's guitar amps, where the screens are typically supplied approximately the same voltage as the plate supply. ie, 450v plate and 440v screen at the typical 50 watt 6L6GC AB1 push-pull. With maybe a small screen resistor in series with each output tube screen."

While my primary purpose in initially posting this message is tone characteristics, and I appreciate the robust though sometimes conflicting responses, I'm unable to fully interpret tone wise. Nuke suggests that if an amp uses a drop resister to the 2nd B+ screen drop rather than a choke that a small value will leave the screens at a close B+ potential to the grids and why the screen resistors (typically 470r on Fender amps) and the tubes themselves fail. This suggests useing a large value resistor such as the 4k7 on the venerable 5E3 which can drop the screen voltage 25v to about 45v below the plate B+ depending on the size of the bias resistor. With a 250r bias resistor on the 6V6s typically vintage transformer put 360 volts on the plates and about 335v on the screens. Using say a 470r resistor and retaining a 5Y3 pushes the plate voltage up to about 385v. The screens do not rise up the same percentage.

When I was repairing amps I also came across a lot of burned 470r screen resistors. But back in the day Fender used carbon comp resistors which are not as heat tolerant and mounted them flat on the pin sockets where they would be most susceptible to the heat of the power tube directly underneath and I suspect this has a lot to do with burned out screens. What if Fender had mounted the 470r on the circruit board or minimally mounted the 470r on pin 6 and bridged out away from the socket and then back to pin 6? We then have to address the heat the 1k5 grid stops are dealing with heat as they also lays flat. They could be bridged up except now we may get too close the the bridge filament winding that Fender has always used. Perhaps laying filmanet windings flat against the chassis allows screen and grid stop resistors to be bridge up and bridge out for better heat dissipation.

How all this plays out technically as to Screen voltage vs plate voltage is another issue which I admit to still not fully understanding. \
However, the market place for about 70 years has voted and the 5E3 with its relatively large 4k7 screen drop is the most cloned amp in existence without a close 2nd. I suspect that a fixed bias version of a 5E3 might benefit from a Princeton-Rev resistor drop scheme with the first drop 1k.

[nuke]

The whole point of a beam power tube (sometimes called a beam tetrode), or the related, but different, power pentode, is that the screen grid functions as a virtual plate, and we can get better linearity of the plate as its current increases (and plate voltage decreases as a result of current).

The beam forming plates, or the suppressor grid in power pentodes, prevent electron reversal back to the screen when plate current is high and plate potential is low, perhaps even low enough that the plate is negative with respect to the screens.

Beam_Power_Tube_Cutaway_Spbg57.png

The way it works is the electrons are accelerated by the screen's positive charge toward the plate. The beam forming plates bunch the flow together and keep it from spreading. At high plate currents and low plate potential, the beam-forming electrodes bunch up the space charge between the grids and the plate forming a cloud of negatively charged electrons in the space, preventing secondary electrons from returning to the screen, without a third grid. (as in a true pentode).

What does all this mean for an audio amplifier?

We want a screen that maintains a constant voltage to keep the plate curves nice and flat for a wide range of currents, enhancing power output and lowering distortion. Yay.

Like all real world devices, the screen isn't perfectly aligned with the control grid and the beam forming is not perfect. So, we have to put some limits on the screen so we don't burn it up with too much current. (that means you, guitar playing hippies, especially).

We can limit the current with a series resistor, or we can limit it by connecting it to a supply voltage that is significantly less than the plate, or perhaps both.

Power is what damages screen grids. It's just a loop of fine wire, and if it gets hot enough, it will deform and melt. Duh.

If we look at the data sheets, especially the old books, when tubes were all the rage, we will see that many guitar amp designs kinda ignored the maximum ratings. Mostly in the quest for more watts. The higher the screen voltage supplied, the more power you get. But you run closer to the edge, tempting fate, and a failed tube and a burned 470-ohm resistor.

If you want to get the maximum power from your output tubes, even if you shorten their useful lives, then do what the great guitar amp makers did (connect them on power supply after the choke with a small resistor in series with each screen). Be happy, go about your life. Just do what they did, it isn't wrong.

You can get even more power if you run in ultra-linear mode with an output transformer that has screen taps at 42% of the plate windings. That works great too, especially if you want lots of clean, high-headroom watts.

But here we are today, when a pair of 6L6-ish(?) foreign tubes will cost you $80 USD, and we less often get to play our 50-watt 2x 6L6 amps at the volumes god intended, we might, just maybe, consider running them with a more conservative operating point, by designing the power supply with a screen node, with lower voltage. Tubes last longer and the waitresses at the dive bar you play at may appreciate hearing the drink orders and not needing earplugs.

Where do you get a lower voltage for the screen?

The conventional Fender/Marshall/Vox is usually rectifier -> capacitor input filter -> plate_node -> choke -> capacitor -> screen_node -> dropping resistor -> capacitor -> preamp_plate_node

Or in cheaper versions the choke is substituted with a dropping resistor.

You could increase the resistance between the plate_node and the screen_node. Effective, but sometimes, dropping 100v or more with just a resistor, might introduce enough resistance in the divider to make that supply not stiff enough to do a great job.

Some rare amps, like the Silvertone 1484, power the screens from different tap on the voltage multiplier incorporated on the two-winding power transformer. This actually works really well. Tubes last longer, the amp sounds really good, only downside is the power output is reduced.

Or another approach I've been experimenting with is a second set of rectifiers diodes feeding a choke input filter for the screen supply (and the preamps down stream) while the plate supply is on its own rectifier feeding a capacitor input filter. This gives a high plate voltage, and a well-regulated lower voltage to run the screens with.

It still sounds like a great tube amp, with less power. Which, ain't a bad thing in reality.

[rschiller]

And thanks to "Nuke" for his last post. As we all grow older and perhaps played way too loud at way too many gigs (I made a living playing guitar for 25 years) , smaller powered amps with the best possible tone is the mission. I should have clarified my initial post that given 3 parameters: maximum wattage, maximum tube life and maximum tone (which is always subjective), we cannot have all 3. So tone and tube life become important. Who needs 22 watts from a pair of 6V6s? 12 watts is fine.

[GrayDigger]

Perhaps a slightly different perspective?

If designing an amp from the ground up, after the basic hardware parameters are sketched out in pencil (tubes, power supply, output transformer primary), much of the fundamental “personality” of the amp can be distilled through more detailed design of the power section itself (assuming you get to turn it up enough to actually push the power section). Considering the type and “temperature” of bias (fixed, cathode, combination), where a PP amp will live between the extremes of Class B to Class A, and the B+ voltage regime for the power tubes will fundamentally influence the feel and tone of the amp, right?

It seems the power section often draws the short straw in design. Perhaps – as alluded to by nuke and B Ingram – we have become indoctrinated as “slaves to the knee” by vintage priorities. If max output is not a primary design goal, I would offer that manipulating screen voltage can be such a powerful – if subtle – tool that fundamentally affects touch sensitivity, stiffness, squish, compression, and distortion characteristics of the power section, provided it receives sufficient signal from the splitter. Manipulating Eg2 for the chosen OT primary Z, we can fundamentally influence our amp’s character and harmonic content by dictating the position of our design LL above, at, or below the knee. Playing with the screen grid stopper resistor for example, provides us much freedom to influence the musicality of the amp to pick dynamics (e.g. squish and compression). For me anyway, it makes designing the power section so much more fun.

[Smitty]

I know this is a design discussion and I enjoyed it tremendously. Thanks!

On a practical note, we must also remember to instruct end users to let the amp cool before moving it. Screen deformation can occur buy moving an amp when the tubes are hot. Once the screen grid has sagged out of alignment it will draw more current. Over time, repeated impact to even hotter screens... Well, you get the picture.

Spread the good word. Tubes are rugged when cold and fragile when hot.