Exceeding Vgs rating

Take 30 of these transistors and drive them to failure. See where they actually fail. That may help in your decision process if you think you can get away with exceeding the rating. Do it at cold and hot temp too

Failure voltage with bipolars is sometimes much higher than rated. I presum e the same is true of fets. You could always feed microamps through it to s ee what voltage it breaks down at harmlessly. A lot of leeway is needed if silicon is to survive well enough in use, don't push it too far.

NT

Is there some reason you can't pick a fet with higher Vgs?

George H.

Good call. Regardless of the Vgs rating, if the drain goes to an outside circuit, I have found a zener from gate to source is needed to protect against ESD. In this case, it might also be useful to avoid exceeding the Vgs rating. Although JL has measured the actual Vgs destruction voltage and found it exceeded the rating by a wide margin. ISTR around 70 volts. So a zener clamp might be good insurance.

It's trickier than a one-shot test, actually: at high voltages (above rating), ionic diffusion occurs, and Vgs(th) drifts. (Back in the day, one of the difficulties in realizing MOSFETs was getting the process clean enough so that mobile sodium ions weren't forever moving around in the oxide, shifting Vgs(th) depending on Vgs history. There are still ions and defects present today, but they aren't as mobile -- except in a strong electric field.)

And yeah, temperature matters too, so you could test them above Vgs(max) while monitoring gate leakage current and checking Vgs(th) periodically.

High temperature is probably the most important, but there might be something about lower temperature giving lower leakage, allowing higher voltages and therefore stresses? Not sure.

Tim

The few mosfets that I've tested failed at Vgs around 70 volts. There could be long-term effects.

Gate-protected ones zenered in the low 40's.

Test a couple to destruction. It's fun to blow things up.

I'd estimate that a couple of volts over Vgs_max will have no effect on reliability.

Overvoltage on non-protected mosfet gates seems to be destructive, even through megohms of source impedance. That happens at roughly 6 * Vgs_max.

I recently tested some EPC GaN fets for Vds limits. They fold back at roughly 2.2 * Vds_max, nondestructive if the current is limited.

Might not be available. Numbers like 8, 10, 12 volts max are common on low voltage, low Ron fets.

C-B breakdown in BJTs is pretty benign unless it results in dumping a lot of energy into the junction. MOSFET gate breakdown is terminal.

Cheers

Phil Hobbs

Agreed! The manufacturers cover their asses with lots of margin. ...Jim Thompson

Not really. I just like to keep the option of using a 12V type open. As JL says, low-V, Low-Ron types with 12V or lower Vgs rating are more common than those with, say, 20V Vgs(max).

That's what I thought. I understand that no one can guarantee a product's behaviour outside the manufacturer's specs. An informed estimate would have been welcome even if it was negative.

You guys talk like ignoring specifications is acceptable practice. IT IS NOT! Testing a few samples under controlled conditions is NOT reason to exceed specifications.

Have you never underestimated the consequences of a decision? Have you never had a vendor make a design change that still meets the original specification? Have you never had a part go obsolete and get replaced by an "equivalent" part? Have you never had a purchasing manager switch parts on you without even telling you so he could save a buck and get a bigger bonus?

I've never attended a seminar on "it won't cause a disaster if the transistor fails".

Grin, I've done all of that.. or had it happen to me. Still... I'm hardly an expert, but there are many specs that are listed on the conservative side. Maybe it's too hard to measure, leakage currents less than ~1uA. Or just lazy, (All LED's have a 5V Vrev.) And then other specs that you better beat by a large margin.. voltage rating on tants. I'm not sure how you learn this stuff, other than word of mouth, or doing it yourself.

George H.

it's done all the time in bottom end products. Why else do you think consumer products often have such short lives? I would be in less of a hurry to do it in mil hardware. Everything in engineering is tradeoffs & compromises.

NT

You're just exhibiting your inexperience. If you read the fine print, VGSmax is usually specified as "guaranteed by design"... that is, it's NOT measured during manufacture, so there's a HUGE margin. ...Jim Thompson

The C-B breakdown has the damaging high field buried in the device, far from surfaces and mobile contaminant ions. B-E breakdown, as used in the negative generator trick, makes near-surface field stresses, and some devices age badly when you do that. It used to be, you could buy oscillator transistors that were durable in this respect, but modern transistors aren't oscillator-rated.

And, a MOSFET gate is always at a surface, so there's no burial-like protection. How tight are the specs on protective diodes? For the MOSFETs that have them, can they ever be substitutes for Zener clamps? JL says 'zenered at 40', but you'd really want thresholds at 10V or so, if that didn't bollix the device during recombination times.

Acceptable to who?

It's a judgement call. Sometimes pushing parts results in serious performance improvements. Sometimes context, like duty cycles or cooling, influence how close to abs max you can run... under or over.

Some parts, like tantalum caps, should almost never be run at abs max voltage. Some parts are perfectly reliable at 3x abs max.

Too often, important things are not to be found on a data sheet. So we test parts to destruction and make up our own rules.

That is not allowed in our company. Engineering assigns acceptable manufactuers parts to company stock numbers.

It's impressive how rare parts failures are nowadays. The early days of transistors and mosfets and TTL were horrible.

Most small bipolars leak picoamps. Some cheap RF transistors have fA c-b leakage when used as diodes... their leakage isn't even specified!

I've seen LEDs and optocouplers that have reverse zener voltages in the 35v sort of range. The 5V spec is probably to permit multiplexing.