More MOSFET Follies

[R.G.]

I've just posted at GEOFEX that I'd intended to do for a long time. It's a current limiter intended to stop rectifier current peaks from exceeding X amperes, where X is settable within a wide range by a resistor value.

This came up in thinking about the Brimistor/thermistor in old JMI Vox amps. The function of the Brimistor is to limit current surges. This does not do the same thing as a thermistor heating up, but it does clamp the current to a level that's safe for the rectifiers and caps. It works OK at start up and for going off standby, as well as being OK to leave in circuit all the time.

See Tube Amp Current Clamp.pdf

The circuit as shown with R2 = 0.47R limits current through the MOSFET to about 1.3A. Even at power on, this is all that can flow into the first filter cap. It's a less-soft form of soft start, but is effective, I think. Switching out of standby does the same thing. The normal current pulses into a filter cap run less than that by my estimation for a 100W amp running at 480V on the plates. But it's adjustable with R2.

In normal operation, a tube rectifier can only conduct as much as its cathode will emit, and this is one form of tube rectifiers softening current pulses. There are others. This circuit mimics the current limit. Not perfect, but should be better than no limiting at all.

The power device needs a heat sink to something like 5-8 C/W. The model says it dissipates 8.5W, but it could be more. Of course, the heat sink is hot to B+, so it needs care in positioning and safeguarding from curious fingers, like other high voltage points in the amp.

[loudthud]

Did you try putting the circuit in the ground leg? Then you Drain of the FET would be at ground. Maybe combine with the zener dropping idea some people like for their vintage knock-off amps.

[R.G.]

Did you try putting the circuit in the ground leg? Then you Drain of the FET would be at ground. Maybe combine with the zener dropping idea some people like for their vintage knock-off amps.

Yeah, good idea. I was functionally fixated on the current coming from the + side.

[trobbins]

It's always a balance of competing returns. Active limiting would negate any downstream fusing that may be in place, such as B+ fuse, or cathode fuses.

Perhaps a tweak to go in to foldback from long duration current limit so that FET doesn't have to sustain high power for continuous short circuit conditions?

[Steve Conner]

Yes, this circuit might well have a problem with the MOSFET sacrificing itself to save the fuse.

My old Toaster regulator used some fairly aggressive foldback to allow it to survive a short circuit indefinitely.

[Gingertube]

Did the same thing when I power scaled a 4 x 6V6 Git. Amp. Put a pot across the 5 Watt current sense resistor, wiper to the transistor base to make the current limit adjustable. Call the pot the "SAG" control, others who have done the same thing called their pot the "COMPRESSION" control.
Watch the current limit you achieve. At max "SAG" on my build, B+ dropped from +345V to +120V, obviously too much current limit at maximum "SAG". Its been running for 3 years now. I guess it has never suffered a continuous short.

Currently looking at doing the same thing but with the MOSFET configured as a gyrator. In that case I want the current limit to only work at switch on, to limit in-rush current to the cap. Therafter peak charge pulses are limited by the Simulated inductance behaviour of the gyrator.

Gyrator calcs covered by Merlin on pg 83 of his Power Supplies for Tube Amps book.

Lots of ways to "skin the cat" - none of them pleasant for the cat.

Cheers,
Ian

[R.G.]

Good points. I'll do some thinking about the "shorted output" issue. A couple of approaches spring to mind, like a thermistor shutting down the MOSFET based on MOSFET temps, perhaps coupled with using a MOSFET with a big Idmax.

I think temperature is the real issue in the MOSFET's survival, as the actual current available from the transformer is not huge by MOSFET standards.

@gt: You're right about the current limit being bigger than suspected. I did some simulation of the circuit for various loadings and current limit settings. For a 200W DC power out from a 460Vdc nominal output, the current limit had to be up at about 1.3A to get plausible sag of about 30-40V under change from kinda mostly no load to full load. This was much larger than I expected, but made sense when I thought about it.

Thanks guys - I'll think about the obvious fault issues a bit.

[R.G.]

I did some thinking.

The way most MOSFETs die is through sheer overheating since they don't have the minority-carrier hot spot death of bipolars. That being the case, a very good way to protect them is to sense temperature and stop when they overheat. I speculate that one could just put a 75C thermal switch on the MOSFET heat sink, using the switch to open the gate pullup resistor. There are semiconductor solutions to this problem, but they involve a clot of circuits and do much the same thing as the thermal switch. The only real trick in doing this is to get the switch to open fast enough if the temperature of the die in the MOSFET is increasing rapidly.

There's plenty of voltage available for enhancing the gate voltage, so I'd probably also add a gage-source pulldown resistor to ensure that the gate is pulled off and not held up by gate-source capacitance.

More simulation to do. I'll try to see what happens when the output is shorted.

[trobbins]

The thermal impedance variation with time of the FET comes in to benefit if you want to include foldback. Foldback certainly does add circuit complexity - but probably allows a more rugged solution, and one requiring less heatsink, than either aiming for continuous s/c operation, or thermal transfer to a bimetallic switch (which is likely to be a steady state condition as far as heatsinking and Tj-h is concerned). Tuning the foldback requires some design limits to be determined as to what would be the max normal current limit time required for start-up, or during 'heavy thrashing' - hmmmm (?).

[R.G.]

Foldback certainly does add circuit complexity - but probably allows a more rugged solution

That's correct. I'm a big fan of electronic limits to overloads, including foldback limiting. And you are correct - the thermal time constants make a bimetal switch hard to transfer to, although this can be improved with some odd mounting techniques.

My own predilection would be to make the MOSFETs proof against anything *except* a short (soft or hard) on the output, and use the time before they die to let the protection fire.

The problem I can forsee is that probably less than 6 of the simple versions will ever be built unless someone cops it for their supermagicthunder amp brand. If I make it complex enough to add foldback, that would likely go even lower.

One could back out to reviewing this at the system level too. The difference in reliability of an amp with one of these inserted could be estimated. The failure rate of the basic amplifier goes UP whenever a part is added whose failure would cause the amp as a whole to fail.

That is, painting the top panel (added component: layer of paint) does not lower the reliability of the whole amp because the paint failing in any reasonably forseeable way does not make the amp as a whole fail. Adding a modification to the amplifier power stage probably lowers the reliability of the entire amp because failure of the mod does cause the amp to fail.

Viewed in that light, we can look at the system reliabilty. Overall, the failure rate goes up, because the MOSFET can fail, as can the gate-related parts. That could result in two major failure modes: open and short. Let's leave aside "burst into flames and burn down the building and maybe the city" as something that good construction practice will make unreasonably unlikely.

Shorted is a non-failure of the amp as a system at all, because that reverts it to what it was before the addition. Open makes the amp non-functional, but the failure is soft: one can get the amp to play again with a jumper wire; that's how the amp was before the addition. Full repair requires replacing the MOSFET and the associated parts, and full replacement is probably advisable. Fortunately, it's not costly.

However, there are some failure modes of the amp itself which would be quite damaging to the amp, requiring costly repairs, that the existence of the current clamp converts into soft, non-disaster failures. In that light, while the current clamp adds to the possibility of some failure, however small, it converts some costly failure modes and operations on the basic amp into non-failures, or less expensive failures. That's the value, I think.

It's half a loaf, but sometimes half a loaf is better than none. Sometimes it isn't.

[trobbins]

From my perspective, the loss of an OT is top of the unwanted outcomes list - followed by the loss of the PT - as either they are very expensive, or irreplaceable for some vintage amps. So I'm completely with you that converting part failure modes to part non-failure modes, no matter whether the system has 'failed' or not, is the key attribute I would be looking for.

From the perspective of a user that just either gets their amp repaired, or buys another amp, then the perspective is very different (ie. maximise total system availability, but biased towards higher MTBF).

[Steve Conner]

According to the Mil-Hbk-Whatever method of calculation, adding components degrades reliability. Bob Pease argued the opposite: if the added components are a protection system, reliability can be improved.

I come down on Pease's side: complexity is cheap and I'm not scared of it. If you're trying to sell circuits to the hobbyist community, you need the opposite approach: they will pick a simple circuit that doesn't work over a complicated one that does work, every time.

Oh and painting the chassis might degrade reliability by making it run hotter.

[trobbins]

If adding high-rel parts could justify a reduction in base failure rate of a low-rel part then that would easily equate to a better system failure rate. The basic fet has a 0.012 FIT, wheras a low stress resistor could be 1% of that.

Reliability assessment (eg. via Hdbk-217) is often just one half of the issue. MTTR plays its role in meeting a suitable availability. But it is certainly horses-for courses - I guess some performance amps get new tubes every night.

[Gingertube]

These are expensive ($7 in 1 off, $3 in 25 off) BUT are what I used on the MOSFETs in the Laser Diode driver I did for the day job.

Normally open, 90 degrees C switch.

Airpax 67F090
http://www.farnell.com/datasheets/29932.pdf

They screw down under the MOSFET mounting bolt.

Cheers,
Ian

[Steve Conner]

I like those Airpax TO-220 thermal switches too! I've used them on a few projects.