Mosfet protection diode -- needed or not?

[jymaze]

For the record, I still have to have a mosfet fail on me and I handled hundreds of them. The are more rugged than it appears: They are rated 20 volts, but I read destruction does not occur up to 3 time that value in some International Rectifier application note. Above 20 volts in a non-sustained manner, anomalies may appears in the oxide layer, but it may take a lot of these to lead to failure as long as the over-voltage is transient. The last time, I used an IRF510 as input stage and I can tell you they have the reputation to be pretty bulletproof, except for massive electrostatic discharges (which are probably diverted just fine by almost any reversed diode before the gate capacitance has time to charge).

I think a good way of protecting, besides all these diodes to ground/gate, is to put a resistor (10k is fine) at the gate so transients don't charge the gate  as fast. This resistor, combined with the gate capacitance, makes a low pass that protect fairly well too against voltage spikes. You have to watch for the bandwidth, but for audio/guitar signal, it should not pose problem.

I think it is good not to overthink it and become paranoid with mosfets: the destruction is more likely to happen before they are actually in the circuit, when you handle them. Once they are soldered, the risk is really decreasing because of all the other components around, including in priority the biasing resistors that provide a much better path to ground than the gate, even without protection diodes.

I still socket them though...

[gritz]

Voices of pragmatism at last!  

Series gate resistor sound good always. TBH a reversed biassed 4148 diode twixt gate and +9V and a sensible decoupling caps twixt +9V and ground (which it should have anyway) should be enough surely? If a transient pulls the gate north of +9V then the diode conducts and dumps the charge to the caps, etc, while the gate resistor limits current. Any transient big enough to get past that might well mean that our original poster has been struck by lightning...

[mordechai]

Just so I'm clear, following R.G.'s observation of the viability of an NPN transistor breaking at 5-7volts...if an LED has a reverse voltage of UNDER 9volts then it will work, but if it's over 9volts, it won't offer protection?

[gritz]

Just so I'm clear, following R.G.'s observation of the viability of an NPN transistor breaking at 5-7volts...if an LED has a reverse voltage of UNDER 9volts then it will work, but if it's over 9volts, it won't offer protection?

It's like this: if the mystery LED's reverse breakdown voltage is greater than the gate / source rating of the mosfet (which is +-20V for  BS170) then it might as well not be there. The trouble is that there's no way of knowing the reverse breakdown voltage of a particular LED without actually testing it (by e.g. putting it in series with a high value resistor and applying an increasing voltage in the wrong direction until current flows).

[R.G.]

For the record, I still have to have a mosfet fail on me and I handled hundreds of them.

I have killed a few, but I have also handled many of them with no ill effect. I mostly take minimal precautions, such as wearing cotton clothes and removing my shoes and socks before handling.

The are more rugged than it appears: They are rated 20 volts, but I read destruction does not occur up to 3 time that value in some International Rectifier application note. Above 20 volts in a non-sustained manner, anomalies may appears in the oxide layer, but it may take a lot of these to lead to failure as long as the over-voltage is transient.

The gate-source voltage is stated on the datasheets as a minimum: they will withstand without damage at least this much. The unstated implication is that they might, possibly, withstand much more. Or they may be only marginally damaged, upsetting some other parameter. Conservative design (that is, design that ensures that none of the parts are inadvertently overstressed) says to never go over the datasheet limits. I've seen people use parts outside the datasheet boundaries many times, with varying results. If you 100% test incoming parts to the limits actually used over the datasheet limits, you can be sure that your parts work even if over the boundaries guaranteed by the maker.  If you statistically test incoming parts to destruction, you can do estimates of where the actual boundaries are, at least for wafer-lot size batches. I've seen guys get into very hot water when they specified something over the datasheet boundaries, testing incoming parts, and found that only, say, 5% of incoming parts worked. Now you have to buy 20 times as much of that part as you need to "test in" the 5% that work. Bosses very angry. Accountants even angrier.

IR's application note almost certainly says that they're good for some more than the datasheet, but that IR will not guarantee anything above the spec. The social and economic consequences for IR from that app note are all good - more adventurous use of their parts and more parts sold for properties they will not guarantee. Talk about win - win!

But on a one-part-at-a-time basis, you only have to get one that works. And if failures are unlikely, you can probably skip protection altogether and simply buy a couple of spare replacement MOSFETs for if one ever fails. Might be cheaper. Tape them inside the box.

I think a good way of protecting, besides all these diodes to ground/gate, is to put a resistor (10k is fine) at the gate so transients don't charge the gate  as fast. This resistor, combined with the gate capacitance, makes a low pass that protect fairly well too against voltage spikes. You have to watch for the bandwidth, but for audio/guitar signal, it should not pose problem.

It's a defacto requirement to put a gate stopper on a MOSFET as an amplifier anyway. The gate stopper and gate capacitance cut of VHF/UHF/OMGHF response of the device and prevent oscillations you can't see on your ordinary scopes when the 90 degree bend of the wire to the gate tunes in a super-frequency radio source.

If the transient energy impressed on the resistor+gate is small enough and the pulse short enough to keep the gate from being overvolted by the amount flowing into the gate through the resistor, this works. So it's good for those 20kV/50nS spike things. A 100V/1mS spike may still give problems.

I think it is good not to overthink it and become paranoid with mosfets: the destruction is more likely to happen before they are actually in the circuit, when you handle them. Once they are soldered, the risk is really decreasing because of all the other components around, including in priority the biasing resistors that provide a much better path to ground than the gate, even without protection diodes.

Yep. Paranoid is not good. Enlightened self interest in reliability is the way to do. There are many ways to go about this. The single 12V zener works all the time, requires no thought at all, and requires a very, very odd and unusual transient to get past it. No thought or overthinking needed.

[jymaze]

One Zener= Fine. You have more chances to win lottery than kill your mosfet.

One LED/1N400= Fine if you want to live just a little dangerously (not much though) and have no Zener at hand (like me). Probably nothing will ever happen anyway because of the polarizing resistors being there.

One double-Zener+ resistor+ another diode to kill the capacitive leak= Complete over-the-top paranoid over-thinking.