Solid state current limiters instead of screen resistors

[Roberto]

Hi all,

as you have probably noticed, I'm asking suggestions on different point of view on well known stuff.
This time I would like to ask if anybody has ever used transistors as current limiters for screen grids, instead of the classical resistors.

The reason is that I like the dynamic that the amp has without screen resistors, but at the same time the tubes are really prone to blow.

So I thought about using current limiters to let the screen to see a very low resistance until the critical current is reached, then it is limited.
This should be a win-win solution, but evil is in details, and what I don't know is if anyone else has done similar tests, and what is the sonic reaction on transitories.

Has anyone some experience on it?



Thanks in advance for your help.

[R.G.]

I can do some guessing. The screen grids have an effect somewhat like a second, less-in-control grid. They can be used as an overall gain control element. With no screen resistors, they let the tube have a fixed internal operating point, but they also can easily overheat from current, as you note. A screen resistor lowers the screen voltage with increasing current, effectively lowering the internal tube gain.

This is itself somewhat nonlinear, as the screen grid current skyrockets when the tube is "saturated", with the plate at a lower voltage than the screen grid is being held; the screen voltage attracts electrons more when the plate voltage is significantly lower than the screen grid voltage. Screen resistors offer a way to let the screen "follow the plate down" as the plate voltage goes lower.

A current clamp would be more abrupt. The screen voltage would be relatively constant right up to the point the clamp kicks in. I'm guessing that you'd see more abrupt "clipping" at the point where the current limiter kicks in. It's action would be to fix the screen current, which fixes the ratio of screen grid voltage to plate voltage, perhaps in a squirrelly way I haven't fully thought out yet.

It's an interesting experiment. Good on you for thinking of it, and it's worth a try!

You could consider using MOSFET power devices in your limiters, as their gate current is so low that it makes the limiter a bit easier and lower power to use.

[Roberto]

You could consider using MOSFET power devices in your limiters, as their gate current is so low that it makes the limiter a bit easier and lower power to use.

Thanks for your intervention R.G.,
what about setting the current a bit more lower than what is obtained with standard screen values, plus a small cap to let the RC (set around 100 ms, or so) soften the "abrupt clipping" behaviour?

[R.G.]

There's more there than I can do in the "circuit simulator" in my head.

It's a good idea to set the current limit around what you'd get with sufficient screen resistance.

It's also a good idea to make the limiter have some kind of time delay. I don't know how that would sound, though.

I think someone (wink, wink, nudge, nudge) needs to build up a couple of current limiters to try out in a listening test.

I'll be glad to provide what technical assistance I can.

[Mike Sulzer]

I wonder if limiting current alone is the right way to go. It may be that what you need to limit is time averaged dissipation, which involves current and voltage and some analog computing. Somethng like what some integrated circuits have. Crazy idea: fool some IC into acting as an analog for what the screen is experiencing and then use its output to control higher voltage, higher dissipation descrete devices.

[R.G.]

KEWL!!! I can use a PIC to do it!!

[Roberto]

Mike,

it's true that an Hall Effect sensor together with a scaled-down voltage-reading and an Arduino could be a solution, but I've always been a big customer of Occam's barber shop

I think the simple RC can do a simple approximation of the same concept.

[Roberto]

Thanks R.G.,

Do you have a suggestion for the mosfet to use and for the circuit?

[Roberto]

In case of an Arduino implementation, I'd do it this way:
a counter increases the value of the variable if the current value is exceeded (being B+ constant, the power is function of the current), and decreases if not. This variable then proportionally controls the resistance seen by the screens.

This way the reaction is smoothed in both ways, and is self controlling on time.

[Mike Sulzer]

KEWL!!! I can use a PIC to do it!!

That sounds like more fun. You could program in dynamics, such as sag. An advanced model would sample the guitar as well and introduce touch sensitive effects.

[Roberto]

I've found this thread that explores similar paths:
protecting output tubes- screen voltage vs screen resistors

[Roberto]

After some research I've found this:
http://www.mouser.com/ds/2/205/98729-63438.pdf

and some related readings, just pour parler:
https://www.w8ji.com/screen_supplies.htm

[Old Tele man]

They're probably far more expensive "fuses" than simple carbon resistors...which screen grids are known to need/use, especially with today's poorer quality tubes.

[eschertron]

Once we've modeled the tube behavior and use external components to keep the tube within an acceptable operating envelope, we can use the model to simulate the tubes' behavior and eliminate those fragile and unwieldy tubes altogether. Isn't that the logical conclusion to this line of thought?

[Roberto]

Isn't that the logical conclusion to this line of thought?

This is one of the possible conclusions, but not the only one. My persola tests have been done on the different RC times for the supply of the screens, and I have to say that you can get quite a good transient and a false perception of power above the capabilities of the amp itself. I like to have controls on my tube power amps (presence, depth, variable NFB), and vary this with an Arduino could be something more to add to the response of the amp itself. Do you think a capacitance multiplier could do the job of a screen power supply, together with a mosfet to vary the R component? Then a potentiometer could vary the two together following an RC constant, or just vary the R or the C.

Thanks.
Roberto