Why should I run screen resistors?

[Gtr_tech]

My Traynor Custom Special still has the 20w 1k shared screen R and I've had no problems with EH and Ruby STR EL34s. I use it for bass and it has tone for days. I did add control grid stoppers tho. If it ain't broke....

The PV "Vintage" I wrapped up recently had no screen R's other than a couple 100r on just one pair of 6L6s. Ruby STRs *hated* this setup....dancing sparkles on the scope wave indicating oscillation. So I redid the screen setup to the std 470r for each screen....troubles gone.

*Some* kind of screen resistor is pretty much mandatory for modern tubes. Control grid stoppers too for that matter...

[g1]

I'd say the 100W Selmer should have had four screen stopper resistors, but then they might as well just have used four individual screen resistors.

Lost me here, is this a typo? I'm assuming you meant grid stoppers, but if that is the case you seem to imply that with 4 screen stoppers we can get away without 4 grid stoppers?

Talking of strange screen grid supply arangements, mention should be made of the JTM45, 3 resistors, shared then split, 1k to 2 x 470 ohms.

To throw another wrench into the works, we also have arrangements (some Peavey, Traynor, Marshall) where there are screen resistors on 2 out of 4 power tubes, both with or without a common resistor off the screen node:
http://music-electronics-forum.com/t27919/

[Steve Conner]

I meant that, if they wanted to have one screen resistor shared amongst four tubes for tonal/voltage dropping reasons, they'd also have to add four individual stoppers for stability.

So then it becomes tempting to make the individual ones bigger, and lose the shared one. Looks like this is what John Chambers does to the 100W Selmers.

The ones where 2 out of 4 tubes have screen stoppers sound like an exercise in "Muntzing". The designer experimented to find the bare minimum of stopper resistors needed to stop oscillations.

The trouble tends to kick off when two tubes have their electrodes connected directly together, as they can then oscillate as a push-pull pair, using the inductance of the connecting wire and the interelectrode capacitances as the "tank circuit". Interposing one resistor in the wire to damp this parasitic tank circuit is probably enough.

[Wombaticus]

My take is that it's best practice to fit screen grid resistors, as (already mentioned) they are beneficial in suppressing oscillation, limiting current and acting as a crude fuse (given a type and suitable power rating); the issue what value to use.
In the case of a 'no choke' B+, where there's already significant resistance in the screen supply, eg >2k, 470 should be fine and as the node resistance increases could probably be reduced to a minimal value, eg 100 ohms.
As an example, for the 5E3 RI, Fender have fitted 470 ohm screen grid resistors, which weren't in the original design.
I'm a bit sceptical about how much surge current a (regular sized) cap, fed from a saggy supply, could put into the screen grid, and how much harm it would really do; rather my feeling (based on not much!) is that it's a sustained heavy overdrive condition that's more of a killer for the screen grid, and that a saggy supply is better mitigation for it than large value screen grid resistors, in regard of detrimental impact on tone.
Pete.

PDF, you're putting the cart before the horse. As I pointed out, the PSU cap at the screen node can source very considerable current very rapidly should it be called on to do so -- enough to kill you dead. Forget about other resistances upstream of the screen supply in the PSU -- they're trumped by the cap.

As for choice of screen grid resistor, this has more to do with your choice of tubes, and the operating point/load line you have selected. Load lines that pass far below the knee will be much harder on the screens....i.e. the screens will draw much more current as the voltage on the plate approaches zero.

It's worth noting that beam tetrodes and true pentodes have very different screen current characteristics. In general, a true pentode will draw quite a bit more screen current than a more-or-less equivalent beam tetrode. Again, as a very general rule, this is why you will see amps that use 6V6s or 6L6s getting away with 470R or 560R screen grid resistors, while it would be unwise to use less than 1K for most EL34 based designs. Of course, there's good arguments for erring on the high side with both types of tubes, and it's certainly possible (for both tube types) to contrive operating points/loads that would require rather more than the "standard" 1K. For those that really care to, Merlin details a graphical method for actually working out the required size of screen grid resistors on his site.

[leadfootdriver]

JJ doesn't publish a plate dissipation line for their KT77. Does anyone have the earlier published version of this tube WITH the curve?

[exclamationmark]

JJ doesn't publish a plate dissipation line for their KT77. Does anyone have the earlier published version of this tube WITH the curve?

I just fire up paint and mark points every 50 volts then connect the dots. I have a whole library of them, haha... but unfortunately not for the kt77.

[overtone]

JJ doesn't publish a plate dissipation line for their KT77. Does anyone have the earlier published version of this tube WITH the curve?

Not quite sure what you mean by earlier published version but here is the original MOV KT77. The JJ may of course be way off this... maybe of some help anyway: kt77-mov77.pdf

The late seventies KT77 amps that I am still abusing all use a 1k5 on each G2. 460V on G2 and the Plate is at about 465V. 3k6 OT for two output tubes and 1k8 OT for the quads. The very late versions actually added a 1k between the B+ node and the screen resistors. I kind of think that was in anticipation of the KT77 going out of production and having to be kind to the EL34.
best, tony

[leadfootdriver]

Not quite sure what you mean by earlier published version but here is the original MOV KT77. The JJ may of course be way off this... maybe of some help anyway: [ATTACH]16977[/ATTACH]

The late seventies KT77 amps that I am still abusing all use a 1k5 on each G2. 460V on G2 and the Plate is at about 465V. 3k6 OT for two output tubes and 1k8 OT for the quads. The very late versions actually added a 1k between the B+ node and the screen resistors. I kind of think that was in anticipation of the KT77 going out of production and having to be kind to the EL34.
best, tony

I read somewhere that JJ lowered the plate dissipation to 25 watts from like 30 or so watts because of reliability issues. So, I was just assuming that there was an early version data sheet with the 30 something watts plate curve out there somewhere.

[leadfootdriver]

I just fire up paint and mark points every 50 volts then connect the dots. I have a whole library of them, haha... but unfortunately not for the kt77.

I'm not from your planet.

[overtone]

I read somewhere that JJ lowered the plate dissipation to 25 watts from like 30 or so watts because of reliability issues. So, I was just assuming that there was an early version data sheet with the 30 something watts plate curve out there somewhere.

have not seen an earlier data sheet from JJ.
I set up the old KT77s for 25W, which helps the lifetime rating -at least there is a note to that effect at the end of one of the old MOV / GEC data sheets.
Best, tony

[Gingertube]

For your info and possible interest.
The shared single screen resistor (for a pair of output tubes) was Mullards way of trying to keep the screen voltage stable in a Class AB amp. As one of the push pull sides draws more screen current the other side draws less, thus keeping the total screen current reasonably constant and hence the screen supply reasonably constant. It will give a more "HiFi'ish" sound which for modern amps where the sound is mostly tailored/constructed in the preamp is no bad thing.
Cheers,
Ian

[Wombaticus]

I'm not from your planet.

You mean planet long division? It's not hard to plot the max dissipation line you are after. Just pick a voltage somewhere on the X axis. Divide that voltage by the max plate dissipation to get the current that will give you that dissipation at that voltage. Mark that voltage/current coordinate. Pick another voltage and repeat. After plotting three or four, you can just connect the dots to get the curve you're after....

[Steve Conner]

The dissipation curve is nowhere near as important as it is with transistors, anyway. Instantaneous peak power can blow up a transistor in milliseconds, so you have to keep the load line under the dissipation curve at all points.

But the plate of a tube is a big lump of metal that averages the dissipation with a time constant of several seconds, so it's OK to intrude above the curve, as long as you would be below it on average. It's not easy to tell the average power by inspection, though, hence the curve isn't very useful.

[Wombaticus]

It's not easy to tell the average power by inspection, though, hence the curve isn't very useful.

???

How can you possibly say that it's not useful? It's a fundamental design consideration, and knowledge of the where this curve lies is rather a key starting point for thinking about the load impedance/anode voltage/bias point triumvirate when designing a power section. Yes, the load line is actually elliptical, and yes, parts of the load line can under some circumstances make incursions into the over-dissipation region. But you still need to (or should) know where that region lies...

[Steve Conner]

You don't "design" a tube output stage, you just get the Philips or GE datasheet and pick the nearest set of example conditions to the transformers you managed to scrounge. If it doesn't work, you fiddle with the screen resistors and blame "modern tubes". It's not rocket science.

Even if you wanted to use some weird tube whose datasheet doesn't specify audio service conditions, or God forbid build a tube hi-fi amp, the factors you mentioned make it loose enough that you can just kind of squint at the characteristic curves and imagine where the overdissipation region would be.

To restate the point about the average dissipation: It's OK for part of the load line to be in the overdissipation region, since other parts of the load line are far away from it, so it averages out. But how much is it allowed to poke in there? Do you know how to calculate how much? Graphically? For a sine wave? A clipped square wave? What about taking idle dissipation into account? I rest my case.