exceeding max grid circuit resistance for power tubes...

[mxrshiver]

...seems to be the norm in guitar amps! what gives? how close to you adhere to that rating in your own builds, and how do you divide up that resistance between grid stoppers, grid leaks, and the bias circuit?

just about every commonly used power tube i know of besides the EL34, has a max grid circuit resistance value of no more than 100K with fixed bias. but most grid circuits i see are comprised of 220K grid leak resistors, and anywhere from 15-100K to ground in the bias circuit... considering the bias load resistance is shared by all power tubes, and the grid leaks are shared by parallel pairs when there are 4 power tubes... that calculates to a grid circuit resistance per tube of 235K at the absolute best (2 tubes, 15K bias), and at the worst with 4 power tubes and highest bias circuit resistance, 840K! exceeding the maximum rating by anywhere from 2.35, to 8.4 times.

my understanding is this limit is there to protect the power tubes when the grid is driven positive past 0V, ensuring there is a path that is low impedance enough for reverse grid current to flow, so that grid current limiting occurs and keeps the grid from being driven further positive. seems pretty crucial. i appreciate how datasheets are just the beginning of a tube's actual capabilities, and plenty of tried and true circuits use these values and have exceeded others as well... and how much it matters depends on whether or not the power amp is even being overdriven. but i just wonder how much this actually contributes to redplating and/or premature wear.

i'm designing a push-pull output section with 2x KT88's driven by a 12AT7 LTP, trying to make it stay tight with minimal blocking and crossover distortion, but a low end that'll take you off your feet, and a sustainable circuit for the KT88's that will make them go the distance under very heavy overdrive (yes, the homie i'm designing it for DOES play that loud lol). looking for less phase inverter distortion, more power amp distortion, and the ability for the LTP to really overdrive the living hell out of the KT88's without any adverse effects.

KT88 plates are at 580V, screens at 450V with 1.5K screen stoppers. LTP has normal 82K/100K anode resistors coming from a 440V supply, center biased with a 40V tail. bias circuit has 20K max resistance to ground, so that presents 40K to each grid circuit... technically only leaving 60K left per tube, divided between the grid leak and grid stopper, to still meet the maximum rating.

grid leaks provide the nice center biased AC loadline i like when they're at about 47K, leaving only 18K for the grid stoppers. i'd love to make the grid stoppers large as hell to control blocking distortion as much as possible though.

how large would you be comfortable with making the grid stoppers here? if i was to go with the conservative side of traditional design, meaning resistance 2.35x in excess of the recommended value, i could bump the grid stoppers up to about 150K, which would be awesome. but i really am expecting this power section to get heavily overdriven, more than those traditional designs likely accounted for, so i'm not sure how much i should push it...

last random question. does one take the bias circuit resistance to ground into account when calculating the AC loadline impedance for the LTP? i'm using such low resistance grid leaks that if that is the case, it would actually make a significant difference. my guess is that it does not factor in, because the out of phase AC signals cancel each other out after the grid leaks, making the negative bias supply a virtual ground of sorts for AC signals.

[pdf64]

The thermionic effect doesn't just apply to the cathode. Any hot metal element in a vacuum will emit electrons.
The control grid is necessarily in very close proximity to the hot cathode surface, hence will become hot and will itself emit a few electrons. As cathode current increases, the cathode and anode dissipation will get increase, the control grid gets hotter, and emits more electrons. That creates a positive grid current that across the grid circuit resistance, generates a grid voltage that opposes the bias voltage.
Because that grid current is a fixed characteristic of the valve, and we want the valve to tolerate high dissipation, then to limit the grid voltage resulting from grid current, the grid circuit resistance limit has to be imposed. Otherwise a redplating desth spiral is risked.

My understanding is that the valve  limiting values apply as a set, on the basis that the anode is dissipating is at its limit. eg at lower anode dissipations, anode voltage etc can be increased.
If the equipment designer restricts the anode maximum dissipation to a lower level, many of the other limits can be relaxed somewhat.
Hence if a sensible loadline and cooler idle dissipation is used, anode, screen grid and control grid idle values can be in breach of their limits without ill effect.

With screen grids at 450V, KT88 in AB1 fixed bias will need a large negative bias voltage, and hence a large p-p maximum signal voltage from the preceding stage for it to reach full AB1 power output. A typical 12AX7 LTP might struggle to achieve that into the load formed by the 47k grid leaks.

[pdf64]

...seems to be the norm in guitar amps! what gives? how close to you adhere to that rating in your own builds, and how do you divide up that resistance between grid stoppers, grid leaks, and the bias circuit?

Grid stopper + grid leak = A.
Then any resistances shared between 2 or more grids need their value multiplying by the number of shared grids. That effective resistance B is added to the A of each grid.

just about every commonly used power tube i know of besides the EL34, has a max grid circuit resistance value of no more than 100K with fixed bias. but most grid circuits i see are comprised of 220K grid leak resistors, and anywhere from 15-100K to ground in the bias circuit... considering the bias load resistance is shared by all power tubes, and the grid leaks are shared by parallel pairs when there are 4 power tubes... that calculates to a grid circuit resistance per tube of 235K at the absolute best (2 tubes, 15K bias),

So assuming no grid stopper, that A+B would = 250k
I can't think of any successful where the bias output resistance was 100k? Ok maybe some versions of the AC50 (but they're EL34), but no 4 valve output stages.

my understanding is this limit is there to protect the power tubes when the grid is driven positive past 0V, ensuring there is a path that is low impedance enough for reverse grid current to flow, so that grid current limiting occurs and keeps the grid from being driven further positive.

Dunno about all that, see my previous post.

looking for less phase inverter distortion, more power amp distortion, and the ability for the LTP to really overdrive the living hell out of the KT88's without any adverse effects.

The LTP should be designed to have sufficient linear region signal output available that its outputs should only clip when the output valve control grid enters grid current clipping.
That's the most significant clipping mechanism for typical 'through the knee' loadline fixed bias AB1 amps.

does one take the bias circuit resistance to ground into account when calculating the AC loadline impedance for the LTP?

No, the signal output currents flowing through the grid leaks will meet and null each other out, creating 0V AC at their node with the bias supply output.

[mxrshiver]

you rock so much, thanks @pdf64!! really helped put it in perspective and connect some dots from different things i'd read. i also reached out to Merlin, who pointed me towards a GEC datasheet for the KT88 where they specify a max grid circuit resistance of 100K for bias above 35W quiescent dissipation (anode and G2)... OR a max of 220K for bias below 35W. what an awesome tube!! 220K is so much easier to deal with. and what an awesome datasheet!! wish i could find 6L6GC, 6550, or 6V6 datasheets that state the same.

i feel like i have a better idea of the relationship now... i think it's a lot simpler than i was making it. at higher cathode currents, the grid heats up more and emits electrons, which get gobbled up (technical term) by the plate or screen. the grid needs a low enough resistance path to ground to complete that circuit and source new electrons. if it can't source enough new electrons quickly enough, the net electrons at the grid begins to drop, meaning that the grid grows more positive, leading to greater cathode current, more grid current, and thermal runaway.

thanks for touching up my math on that 'best case' scenario, that would indeed add up to 250K!

and yes, it's very strange to me that so many 4x 6L6GC amps continue to be made with grid leaks and bias circuits that add up to 500-850K total grid circuit resistance per tube. even if the split in max grid circuit resistance based on overall dissipation noted by the GEC KT88 datasheet, could be extrapolated to 6L6GC, so 220K per tube being okay below about 70% dissipation... that still leaves a lot of amps exceeding that by 2-3 times. looking at the EVH 5150III, using 4x 6L6GC with 220K grid leaks, and a 25K pot and 82K resistor in the bias circuit... makes an equivalent grid circuit resistance of 768-868K per tube! they use a reasonable bias of about 19.5W per tube including the screens, so about 60% plate dissipation... but is that really low enough dissipation to allow for exceeding the max grid circuit resistance by almost 4 times?!

in Eddie's tweaked 5150III 100S Stealth (imo the best sounding 5150 ever made), they changed the bias circuit to a 10K pot and 27K resistor, making the per tube resistance 548-588K, still about 2.5x over the max... and they also made the bias closer to 70%, so it really seems about equivalent. i just don't really get how these are sustainable designs... maybe the max allowable grid circuit resistance really does just about double with every 10% decrease in dissipation, and maybe no one ever has the Master high enough that it would make a difference. but it'd be great to see some data on it. it seems this isn't the mentality the 5150 was designed with per se, but for my own designs, i like to set the circuit up to be 100% safe and sustainable even with the settings cranked all the way up.

[pdf64]

... i also reached out to Merlin, who pointed me towards a GEC datasheet for the KT88 where they specify a max grid circuit resistance of 100K for bias above 35W quiescent dissipation (anode and G2)... OR a max of 220K for bias below 35W. what an awesome tube!!

Great, I hoped Merlin would comment on this :)
However I suggest to be cautious about applying that specific info to KT88 from other manufacturers / eras. eg the 1959 GEC info has 35W and 220k limits https://frank.pocnet.net/sheets/084/k/KT88.pdf

wish i could find 6L6GC, 6550, or 6V6 datasheets that state the same.

I think it's reasonable to apply the principle to other valve types.

Bias vary trem pushes grid circuit resistance way over the limit, but is probably mitigated by the lowish HT voltage (which given the standard 4k-ish OT will limit dissipation) https://el34world.com/charts/Schematics/files/Fender/Fender_custom_vibrolux_manual.pdf

i just don't really get how these are sustainable designs... maybe the max allowable grid circuit resistance really does just about double with every 10% decrease in dissipation, and maybe no one ever has the Master high enough that it would make a difference. but it'd be great to see some data on it. it seems this isn't the mentality the 5150 was designed with per se, but for my own designs, i like to set the circuit up to be 100% safe and sustainable even with the settings cranked all the way up.

Bear in mind that with AB amps, maximum anode dissipation occurs at medium power outputs.
Full power squarewaves into a resistive load should be pretty easy on the anode. Though the screen grids then take a beating.
See Aiken's anode dissipation simulations https://www.aikenamps.com/index.php/idle-current-biasing-why-70-percent