• E3443.png

    WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.​

Is high voltage Mosfet capacitance multiplier bypass cap necessary?

Does a high voltage Mosfet capacitor multiplier supply for a valve power amp ( PP class A) need a bypass cap, say, to discourage oscillation? If so, is there any way of determining safe start-up conditions so as not to breach the Mosfet S.O.A., for instance, scaling the bypass cap with the time constant of the filter in the gate circuit to limit current into what is, to all intents and purposes, a short circuit at t=0? Merlin Blencowe has a neat trick in his Preamp book for soft current limiting using the gate to source protection Zener and a resistor, but I think I would prefer that this limiter was not constantly strained at switch on. I've seen quite large bypass caps on BJT cap multipliers which would surely destroy a Mosfet in a microsecond or less.

Many thanks for any ideas and/or experience.


2008-03-12 9:59 pm
I've seen quite large bypass caps on BJT cap multipliers which would surely destroy a Mosfet in a microsecond or less.
it is actually the other way round, BJTs are much more limited with respect to SOA than mosfets;
the trick is to have the RC time constant at the gate or base such that charging current of the total capacitance on the source or emitter output never violates the SOA at turn on;
one problem remains though, you need a bleeder R not only on the output cap but also on the gate cap such that the gate cap depletes faster than the one on the output;
otherwise if you cycle the power switch on - off - on, (short off period) the cap on the gate is still fully charged while the output cap is already depleted, so the mosfet stays fully on and - bang - switches against a dead short ...
with a BJT the base current does that job, but not with a mosfet which has zero dc gate current ...
Sorento, it's difficult for me to see how this would be implemented. I would never cycle a power supply containing transistors, but, in case of power interruption: the valves are still hot so they will likely discharge any bypass cap. The gate cap, say 10uF, charges through the upper resistor of the potential divider supplying the gate but starts to discharge through the lower one which is likely to be a meg-ohm more or less. Have I made any incorrect assumption?


2007-10-10 5:21 pm
In most cases, you do not need a large bypass capacitor at all. Small ones are sometimes needed, a tenth of a uF to a uF or so. You can always add current limiting to the cap multiplier, like any other power supply. For tube amp supplies a handful of ohms of output impedance is not the end of the world, so a basic 0.7 volt vbe-based current limiter is easily added. You can also put a resistor or other two terminal current limiting in the DRAIN - that will limit the short circuit current too. In that case, be sure to use proper gate clamping to avoid gate-drain overvoltage. UF4007’s are fine here. You’d have a vgs clamp anyway (or should). Limiting is good practice for any power supply that uses transistors, and this is no exception.

And no, MOSFETs are not necessarily better for SOA compared to bipolars. They advertise that mosfets have purely thermally limited SOA but that is an outright lie. Many olde types like the IRF240 family were very good and we’re better than the bipolars of the day. And so are modern bipolars like the MJL4281, which blow away most mosfets, including the IRF240. The best switching types of either are positively awful and my advice is to run your own testing on any power supply circuit before incorporating it into an amplifier. Put it through a torture test or two to simulate any potential failure mechanism in your amp - if you blow mosfets it just means you have some work to do.
SpreadSpectrum, I presume you had at least some local bypassing since you were using RF valves?
None whatsoever. The Mosfet source voltage was pretty solid. I have scope captures somewhere in my Corona thread.

I did have to provide some inductive isolation between the output tube plate and the output transformer in the form of 3 ferrite beads (in series). There were little oscillations on transients and the beads cured that.
SpreadSpectrum, it was suspicion of minor oscillation which prompted me to start this thread; I had more hiss than expected although there is no distortion in the audio.

Gerrittube; I actually don't know what linear operation means-does it simply mean non-switching? I'm using a switching Mosfet rated for 800V 5A for a class A 300V supply. Its operating point is about 12V drain to source and 120mA. Is this really likely to be a problem?
Yes, I even put a series resistor before the reservoir caps to protect them from enormous inrush current from a big toroidal transformer.
Amp hiss is totally gone now; it must have been gas in the NOS output tubes which has now cleared after a few hours of use. I think past history of blowing up Mosfets made me immediately jump to the conclusion that they were the guilty partly (although I had taken all the anti-oscillation precautions).
SpreadSpectrum; I like your LC filter, which, it's good to see, is more than critically damped. I wouldn't want to be blowing up too many of those monstrous 8N150L devices. After reading comments on this thread about using switching Mosfets for DC service I have learned, for the first time, that Mosfets exhibit a kind of secondary breakdown too. The fact that you are using a, hopefully, indestructible 700W device to dissipate perhaps 10W gives me pause for thought.

Rod Coleman

Paid Member
2004-07-25 6:55 pm
As others have said, there is usually no need for a cap on the output. But like any transistor follower circuit, the active device can self-oscillate, if the supply wiring looks too inductive. It is always best to include a cap on the input. Connect directly to the FET Drain. Something like 220nF and preferably stacked-construction (low loop inductance). If you are a long way from the DC supply, add an electrolytic too.