How to protect MOSFET driver from gate over voltage

Question

Below is part of my circuit. The problem I'm facing is my transistor fails a bit too often, mostly due to my stupidity but also in my learning experience. I have a bunch of transistors which I got for very cheap (close to nothing) so I don't mind blowing them up. Apparently, depending on how they fail, most times 36V appears on the gate (pin 1) which causes my MOSFET driver to also die. These I don't have too many of, and are $2 each so a little bit too expensive for me to want to blow them like popcorn.

First I thought I can put a schottky diode on the gate path so that it will block the high voltage when the transistor inevitably fails, like this:

But then I realized that this will prevent the gate from discharging. I could probably put a ~100Ohm resistor after the schottky to ground which will allow the discharge to take place but will also waste ~120mA on each high pulse which is ~1.5W (ouch).

I'm learning that zener diodes can be used for overvoltage protection, so I thought I can put a reverse biased zener between the gate and ground like this:

But then the TC4420 is rated at an absolute maximum of 20V and max 18V supply voltage. I have no idea how the internals work, but I'm assuming that's the maximum it can also output, thus can tolerate on the line. If I put a 20V zener, that might work but is going to put the chip on it's limit and since no element is perfect the voltage could still reach more than 20V.

Is it possible to combine 2, say 15V, zeners such that when the voltage starts increasing (up to 39, which is pretty much instant), the first one will clamp at 15V, then the second at 30V and thus the chip will see 9V in the end?

Answer 1

The first step is to monitor transistor voltage and current, and simply avoid driving them to destruction.

There is only one failure mechanism that you can't directly see on the oscilloscope (overheating). Get good instruments and use them religiously (100MHz+ scope, 10x voltage probes, and a current probe isn't a bad idea either).

This does require some experience, of knowing what load conditions are, and how to control them. On the upside, you don't need to use power transistors to figure this out: a control works at any power level. So, burn a few of your cheap (potentially fake) transistors, figure out their ratings (if they indeed fail the datasheet), and study control schemes with them. Then move up to bigger projects, 100W, 500W, few kW -- while also developing an understanding of dynamics of the switching loop (stray inductance and capacitance), and how they affect higher power and higher speed systems.

As for gate drive, I sometimes use the following while evaluating prototypes, where transistor choice and design parameters are somewhat flexible, and I'm likely to either install poor combinations, or just goof up the controls entirely, something missteps, and kablewey.

^{simulate this circuit – Schematic created using CircuitLab}

The gate resistor (20 ohms, here) is divided in two, and a TVS placed inbetween. This way, when the transistor fails three-way shorted,

1. Voltage from D shorting to G, drops across R2, limiting current;
2. D1 limits voltage to a reasonable level (20V say);
3. R1 limits current into the driver (1-2A say).

In the process, R2 may end up blown, and D1 shorted, but the driver is much more likely to survive. Mind, still not guaranteed, it may still take damage. TC4420 for example "latchup free" to 1.5A reverse current (meaning, gate current pushing the output above V_DD). And with an 18V limit, the clamping voltage can't be too high (or supply capacitance too low) to deal with this.

Also, note TC4420 recommends several µF on its supply pin. It's a strong chip. It is rare you would need that much drive, even on earlier generation transistors (where ratings such as 300nC Qg, 70mΩ Rds(on), 500Vds(max) were typical; modern equivalents are now 50-100nC, with Vds of 600-700V becoming more common than 500V). Mainly because, at the power level implied by such a device (low ~kW), you're going to have speed limitations due to loop inductance and such. Especially at lower voltages.

Anyway, in regards to this, current backflowing through R1 causes supply voltage to rise, and using enough capacitance is necessary to avoid overvoltage.

Answer 2

Is it possible to combine 2, say 15V, zeners such that when the voltage starts increasing (up to 39, which is pretty much instant), the first one will clamp at 15V, then the second at 30V and thus the chip will see 9V in the end?

Yes it's possible, but the zeners will also be pulling current through the internal high side mosfet, so you will also be pulling the 12V rail down (for whatever current it can source, if it can source a lot of current, you may find some components heating up or being destroyed). If you wanted to use the lower scheme, it would also be better to put a series resistor on the gate to limit the current into D1, but you will also reduce switching time. The best thing to do if you are worried about over voltage is regulate the 12VDC line.

The IRLZ44N can handle a gate voltage of ±16V relative to the source, so a 12V rail is fine, in most designs its better to have a higher gate voltage because it lowers Rdson.

It's probably not the gate or exceeding Vgs that's killing hte IRLZ44N, it could be too much current in the gate. Put a scope on the gate just to be sure, set the trigger for something beyond the rail - say 14V - and put the scope ground on the ground of the FET. If it never gets triggered then it's not Vgs that is being exceeded.

It's likely that you are exceeding the current of the gate, so just a series resistor will suffice to limit the current into the gate 'capacitor'.

Another thing to check is Vds, which can't be more than 55V, if the load is inductive then the voltage on the drain of the fet could be more than 55V. Another thing is ESD, ESD can kill fets very fast because the gate is only nanometers thick and it's not hard to blast a hole with a little voltage.

Answer 3

I actually went with the first (schottky) solution, with an added resistor like so:

If you decide to do this, ideally you would want R1 resistance to be as small as possible, so that the MOSFET gate can discharge as quickly as possible.

In my case I picked a 5W 40Ohm resistor which resulted in 12v (output voltage) / 40 Ohm = 0.3A or 12V * 0.3A = 3.6W of wasted power on each high pulse. My frequency is relatively low at 30kHz, if you're switching at a higher frequency you will likely need a lower resistance.

This seemed to save my driver IC, but is only a temporary protection and I will be removing it once I make sure transistors no longer blow up.

Sharing this for future lost souls!