Constant current source and screens

[anson]

I was thinking about screen of pentodes (and tetrodes). Their role, as far as I understand, is to increase the gain of the pentode by putting a high voltage between the grid and the plate, so as to increase the current to the plate. But it's not really necessary for them to conduct any current themselves (as far as I know) to fulfill that role, it's just a side effect that happens with the typical 1k screen resistor to B+topology.
I'm my amp, often times they are quite happy and dim at idle, but the amp is cranked then (presumably) AC current goes through them and they glow red.

So I was wondering about putting a constant current source, such as a mosfet or pentode, as is sometimes described for use in the tail of a LTP, so that a constant DC current is drawn but no AC is passed.

I'm imagining a pair or quad of screens hooked up to such a constant current source (instead of the typical 1k resistors connected to B+). Then the screens will not conduct AC current as the output valves get cranked as they normally would, but would still fulfill their duty of increasing current to the plates, but not soaking up any additional current themselves.

Any ideas? Is this crazy?

[BackwardsBoB]

Screens draw somewhere around 10% as much current as plates, and this current varies dramatically as plate voltage falls under high current.

[Steve Conner]

The screens have to be maintained at a steady voltage in order to do their electron-enticing job. A constant current source wouldn't do that: the screen voltage would jump around like crazy with the signal, and there would be weird distortion and unpredictable power output, especially with beam tetrodes, where the screen current is small and varies a lot from one tube to the next, like the base current in transistors.

However, I think there's mileage in having a screen circuit with an electronic current limiter. That is to say, a current source with its output current set to the maximum current the screens can safely eat. In normal operation, the current source would saturate and the screens would get a constant voltage. But in abnormal cases, like a too-high or missing speaker load, or just plain massive overdrive, it would transition into constant-current mode and the screens would be protected. The effect wouldn't be too audibly unpleasant, either.

Maybe a power stage like this could survive getting cranked into an open-circuit load.

Also, because of this protective effect, maybe you could do without screen resistors, or go down to 10 ohms or whatever value is needed to stop parasitic oscillations.

[redelephant]

All experimentaion is a good idea But u'll end up with lower power.

The screen does conduct current, and it varies with plate voltage. The screen is there to shield the grid from seeing the plate swinging up and down, greatly increasing gain (reduces feedback from plate) and reduces miller effect, both of which makes the tube very easy to drive. Try floating the screen and the tube is effectively turned off.

The screen will attract more electrons when the plate voltage goes lower than the screen voltage. So as the plate swings low, the screen current increases. When the plates really go low and the screens are at a potential several hundred volts more than the plates, the screen current can get very high, and without a series resistor that drops the screen voltage, the screens can melt. So in normal pentode operation the screens must be able to vary the current thru them.
If u limit the currentswing, or make it constant, the screen voltage will follow the plate. This will make the pentode act more like a triode and gain will be greatly reduced. During positive going plate swings, the plate needs no current and teh tube acts like a normal pentode. Possibly a cool effect in a SE amp, but in a PP I'm not sure.
A choke acts like a passive CCS, but will allow the screens to swing positive with the plate. Will that turn it into some kind of triode?

How about a CCS with some various resistor values u can switch in parallel to make a blend of op-modes? Or just a switch that places a normal value resistor in parallel with the CCS so u have normal op, and a low power op?

N

[redelephant]

I see Steve posted while I wrote...didn't mean to double up...But interesting to see we have slightly different views

[pdf64]

Yes I'm with Steve and think that current limiting the screens has big potential benefits, eg tube life.
When I took some measurements, for 6L6GC in a regular fender arrangement, the screen grids just take 2-5mA (ac+dc rms) as the signal level increases, up the point of clipping. Then it rises significantly, to over 50mA when heavily overdriven, if there's 2x impedance mismatch (eg 8ohm load on 4 ohm output).
This can result in the plates dissipating over 20watts - that level of overdissipation must have a negative affect on tube life.
Increasing the screen grid resistors to a point where they have any impact on this (>3k3) reduces clean output power and affects the tone.
I doubt that the situation is much different for true pentodes.
So a current limit of 5 -10mA might be the way forward.

[anson]

Thanks for all of your input. At first I was thinking that the CCS would somehow keep the screens at a constant voltage AND constant current, and I percieved all the current going through the screens as an unfortunate waste of current, since none of that signal gets turned into sound. Well, it's apparent from your posts now that the Voltage cannot be maintained constant as well. None of that screen current gets turned into sound, but maby it just has to be that way. I like Steve's idea of using a CCS to limit the current of the screen to what is safe for the tube. That's pretty cool.

[Old Tele man]

+1, screen needs a "constant voltage" rather than "constant current" source...that is, unless you're intending to operate the tube in a 'variable-Mu' configuration.