Can gate drive voltage be too high?

[Seekonk]

I have a low frequency application under 1KHZ driving a FET switching 150V.  I have a regulated 18V supply I would like to use for the driver chip.  Normally you see circuits a few volts over 10V gate drive. Have never seen any discussions about getting close to the gate limit. I don't care about added delays or drive current, only possible added transition heating which is minimal anyway.  Any thoughts of a downside, prefer not to add any voltage dropping just for this.

[jmelson]

For standard-voltage FETs (not the "logic level" ones) up to 20 V is pretty safe, as long as there are no overshoots.

A suitable gate resistor should make sure even if the gate driver does produce some transients, they don't appear at the gate, due to the capacitance.

Jon

[ajb]

Well, as long as you don't exceed the rated gate voltage or end up switching *too* fast (or making Rds(on) too low, if that matters in your application), there shouldn't be much downside.  Note that not exceeding the maximum gate voltage needs to account for overshoot/ringing, and that can be affected by switching speed as well, especially if the layout is non-optimal, so a setup that works at 12V or whatever may not work acceptably at 18V.  But the same basic rule about getting close to any of the absolute maximum ratings of a part apply here, which is that leaving yourself more headroom is going to make the whole thing easier to design and more reliable, all else being equal.  If the Vgs(abs max) is ~20V, which is a common value, running at 18V is getting awfully close to the limit especially taking into account supply voltage tolerances and the potential for ringing.  Maybe it's acceptable if this is a hobby project, but for commercial applications it might be a bigger risk than you'd want to take.  Exceeding an absolute maximum rating might not immediately make the thing blow up, but it does mean that any lifespan or performance values given in the datasheet can no longer be guaranteed.

[Zero999]

Increasing the voltage will speed up the turn-on time, at the expense of the off time.

[bdunham7]

At 1kHz I would expect that the amount of time spent in transition would be fairly low so there'd be no need for ultra-fast switching.  Why not use a gate resistor and a gate clamp where the gate clamp voltage is a bit below 18 volts and conducts a bit of current during the on time?

[T3sl4co1l]

I have a project with a row of UC3842s supplied from 18V or so.  They get rather hot, because gate charge is so much higher at that voltage.

Gate power is Qg(tot) * Vgs(on) * Fsw.  This is mostly dissipated in the driver and series resistor.

Tim

[TimNJ]

For Rds(on), it depends on the exact MOSFET, but it seems most are tuned for minimum Rds(on) @ 10V. The incremental improvement from 9V to 10V is usually very small (albeit not always negligible)...so I think the difference between 10V and 18V is usually even smaller, almost nothing...but again, it varies.

I've seen some SMPS control ICs, like synchronous rectifier controllers, with a maximum gate drive output of 7 or 8V. In some applications you will get higher efficiency with lower gate drive voltage, as the dynamic (i.e. gate charge) properties become more important than the static (i.e. Rds(on)) properties.

For low frequency switching, I'd probably lean towards a higher gate drive voltage.

[T3sl4co1l]

Mainly varies with rating, and somewhat with technology.  In my experience, newer SuperJunction types are fully saturated by 8V or so -- they don't draw additional current, even under fault conditions (Vds > 100V, well above the resistive region), at higher Vgs(on).  This is I think unusual, and probably unique to the technology.

Most classic MOSFETs keep increasing with Vgs, at least by 15V, not that you'd have any use for that outside of pulsed applications as the voltage drop is so high.  Likewise in most IGBTs, high Vge(on) is problematic under fault conditions, as the die is so small, only a few microseconds of fault are permissible before destruction occurs.

Most industrial modules (especially IGBTs, maybe not so much MOSFETs?) specify not only 15V on, but -15V off.  I think that's the most extreme case.

These are all high voltage examples (Vdss > 300V).  In the other direction, low voltage devices offer higher transconductance and logic-level inputs.  I don't think there are many (any?) "logic level" types over 250V or so, but, below that they are reasonable.  Typically a logic-level type is distinguished by lower Vgs(th) of course, as well as higher Qg -- no free lunch, that means you need maybe a fourfold stronger driver to get the same switching speed.  In still lower voltages (under 100V say), most devices are rated for 5 and 10V operation anyway, and low voltage devices (30V or less) may even be rated for 3.3 or 2.5V.  Very low voltages (<= 10V) are even available for 1.8V, or 1.2V I think.  Such low voltage devices of course carry a Vgs(max) of 12, 8, even 6V or less.

There is some wear effect of high Vgs(on) and low Vgs(off), near ratings, but I think it's normally specified to a reasonable lifetime (maybe a decade at max temp and rated voltage?), so it's not fatal to run near it, but it's preferred to run lower.

The failure mode beyond that, I believe is more to do with flux (voltage * time) spent at the extreme, as tunneling current is drawn through the gate oxide (some ~uA even at room temp), which can trap charges in the oxide, shifting Vgs(th) and eventually causing breakdown.  Sudden breakdown can occur anywhere from 30V to over 80V, for a typical 20V rated gate.

Tim

[Siwastaja]

If the absolute maximum rating is Vgs=20V (some parts have 30V rating though), any ringing on gate - or your 18V supply inaccuracy - brings you over the edge. If you have a double diode package or a zener in your BOM already, that could be used to bring a tad of extra safety margin at little cost, aiming somewhere around 15-16V.

[AndyC_772]

Bear in mind that, if you're switching the gate rapidly, transmission line theory dictates that the voltage at the far end of the trace from the driver (ie. the MOSFET gate) will double at the moment the incident wave hits the end of the trace and reflects back. If you're already close to the rated maximum Vgs, this effect can and will cause the actual gate voltage seen by the device to exceed its limit.

https://cawteengineering.com/design-focus-high-speed-digital-design-and-termination/

[T3sl4co1l]

MOSFET gates are pretty damn far from transmission line impedances, so it's hard to get a situation where this applies exactly.  The low frequency equivalent, when the transmission line is electrically short but relatively high impedance, manifests as series inductance, which has a similar effect but at wavelengths far longer than the electrical length of the line.

In that case, setting R_G >= sqrt(L/C), where L is the total gate drive loop inductance, and C is the equivalent gate capacitance (C = Qg(tot) / Vgs(on)), affords the damping necessary to avoid overshoot.

In short (ha), place the gate driver near the gate, and you'll be fine.

Tim

[Seekonk]

Sorry for not getting back sooner, I'm in the middle of a move.  Hats off for willing to run a 3842 hot.  This is semi hobby of an existing board that uses a 18V zener as an over protect. It normally uses an IR2153 half driver which regulates to 15V. I have 2,000 18V zeners which are close tolerance.  The issue comes when a IR2101 replaces that doing the same function on the board.  Prefer to underdrive to 2500pf of FET to actually slow down transitions with 27 ohm gate resistors to prevent serious EMI on drain. Gate could become potentially long with external leads. Hadn't noticed any serious ringing before.  Thanks everyone for their insight.