MOSFETs and PWM

You don't need a gate resistor at all. Just run your logic level to the fet's gate. There are some second-order reasons to use a gate resistor (kill rf oscillations during the slew, slow down the edges, protect the driver if the fet dies) so 1K, close to the fet, wouldn't hurt.

Make sure your logic high is enough to turn the fet on hard, not merely "gate threshold" voltage; that's on the datasheets.

Be aware that an open-gate fet could be on or off, so make sure your driver circuit keeps the gate low when it should. Disconnected cables can get interesting.

Most fets like this are fairly esd hard, so don't go overboard on handling. I take no esd precautions when working in my lab, and have never blown a mosfet. My manufacturing people do all the strap/mat stuff, and they don't blow fets either!

John

In this application, I'd say yes.

You will likely not need R, there is no current flowing. What you will need is active drive, high and low. FETs are voltage controlled, the only current to flow into the gate (base) when you turn it on is what is required to charge up the tiny gate capacitor, after that the current flow ends. To turn them off, the gate capacitance has to be discharged, just letting it float won't turn it off.

Allot of MOSFETs require high voltages to turn them on (usually you have to pull the gate higher than the highest voltage being switched. For example, if you are using 12V to power the relay, you would have to pull the gate higer than 12V. This can really suck when you want to use a microcontroller to do the switching. Enter the "logic level" FET. They only need 4 or 5 volts to turn them on hard, even when switching higher voltages. Much more convenient.

I've had good luck with NTE-2985 parts. 5V of drive and the on resistance is .05 ohms max. 60V breakdown voltage and 30A max continuous current. I'm sure there are many many others out there, but I accidentally happened upon these whilst looking for something else.

Not so. The curves of drain current versus drain voltage, at fixed gate voltage, are essentially flat. What is true is that fets with higher max drain breakdown specs do *tend* to need more gate drive voltage, because the oxides are thicker. I've never seen a fet spec that went above +10 gate voltage, even to switch 600 volts, and there are lots of mosfets around that will switch 50 volts with 3.3 volt gate drive.

John

Oops, I stand corrected. :-) After you make the point about 600V, it does seem ridiculous to expect that much gate voltage to turn one on. As for the

3.3V that switch 50V, how long have those been around, or are they just more of the "logic-level" collection? While you're here, is this all that "enhancement mode" means; logic level/low gate drive?

Don't think that's correct Anthony. Typically the gate threshold for a 30A MOSFET is 2-4 volts. For example take a look at a BUZ11

maybe you're defining this as a logic level fet too. ;)

Vgs(th) for the NTE part is 2V max. .05 Ohms Rds(on) is at 5V gate drive. At 4V drive, the on resistance is .07 Ohms max. Sounds like logic-level to me. ;-)

It takes more voltage to turn a mosfet on than it takes to turn on a darlington, but the gate looks like a small capacitor, rather than two diodes in series to ground, so instead of putting a current limiting resistor between the logic output and the gate, put it between the +5 supply and the logic output, so that the logic high voltage will be the full logic supply. 2.2k to 10k should help get the fet turned on more than just the TTL output would do.

And you will have to find a mosfet that is called a logic level type. This means that its on state resistance is specified with a logic level gate drove. Most other mosfets have the on resistance specified with 10 volts of gate drive.

--
Regardless, you\'ll need to know how long it takes to charge up the
gate capacitance between when the switch is "ON" and how long it\'ll
take to discharge it when it\'s "OFF", since during that time the FET
will be required to dissipate extraordinary power.

If I was actually going to try and switch 30A with it, I guess I'd use a gate-driver of some type. I have some in my shack that are made by Microchip (IIRC) that I used for tinkering with some IGBTs. They have very fast switching times and will supply much larger currents (>3A) during the transition. The NTE transistors that I mentioned seem to work fine being driven by a PIC at 5V. I was only switching a couple of amps though without a heatsink.

[snip]

Wow! Lots of really excellent advice from everyone!

I think I'll put the 1k in for driver protection. For peace of mind, if nothing else. ;)

The FDC655 lists 33 mOhms Rds(on) at 4.5v, which seems more than adequate. I think I'll go with it.

If you draw the "saturation" curve of a mosfet, Id versus Vd, there's some drain voltage where it stops being resistive and goes into constant-current mode. That voltage tends to be about the same as the applied gate voltage. To get it to saturate below X volts, you need at least X volts of gate drive. You can see that effect in the curves on the datasheets. Maybe that's what you were referring to. But you can switch a bazillion volts of load with just a 5 or 10 volt gate swing, if the final drain current isn't too high.

Logic-level fets are fairly new, last 10 years maybe. Classic n-channel mosfets fets were barely turned on at 3-4 volts, and not fully enhanced til 6-10 volts; p-channels were worse. The new parts use IC technology (cellular layout, polysilicon gates, thin oxide) to get low thresholds, at the expense of breakdown voltage.

Enhancement mode means the drain-source channel is inherently nonconducting, "normally off", and applying a gate voltage makes it conduct. Depletion mode devices are already conducting, "normally on" and require a negative voltage to turn them off. Most depletion devices are jfets or gaasfets, although a few people (like Supertex) make depletion mosfets, which can be very handy in certain apps.

John

The drain current at "gate threshold" is only 1 mA. The Rdson at 10 amps is spec'd with +10 on the gate. So this is of course not a logic-level fet, because it wouldn't be prudent to run a beast like this with just 3.3 or 5 volts on the gate.

John

As you can see, I'm not much on analog that's for sure. :-) I was thinking that I had read somewhere that conventional FETs worked the way I described, and it sounds like you might be saying that they sort of do if you need them "really turned on". ;-) And I was under the impression that "ordinary" P-channels would have to have the gate pulled below ground for the same reasons.

Logic level FETs sure are convenient for guys like me that like to use PICs. I also like the high-side drivers that are available, pretty slick features.

Ok, so it really doesn't have anything to do with what makes the FET turn on better at lower gate voltages?

You are referring to the BUZ11 right?

I'm a complete amateur WRT MOSFETs, so I'm finding this discussion interesting. I hope you'll bear with me if this question sounds dumb:

If you're driving the gate from 5V CMOS logis (eg; a micro), can you get around this by putting in a series diode after the logic output, & a large (1M+) pullup resistor from the MOSFET gate to your 12V (or whatever) relay supply?

--
   W  "Some people are alive only because it is illegal to kill them."
 . | ,. w ,      
  \\|/  \\|/              Perna condita delenda est
---^----^---------------------------------------------------------------

Nope. A diode doesn't give voltage gain. You can shift the gate swing up by one diode drop, so it swings 0.6 low to 5.6 high. but that doesn't help a lot.

John

Yup. The typical curves show it conducting pretty hard, 8 or 9 amps, with +5 on the gate, but that's not guaranteed. If your load is just a couple of amps, it's probably safe to go with +5 drive. But I wouldn't bet a production run on typical performance curves.

At 3.3, it might conduct just a few mA, if that.

John

Come on, the BUZ11 isn't a beast, It's a beauty. It's quite versatile and cheap too.(84cents at digikey). One of the least expensive N- Mosfet's available. Here's a PWM using a 555 and a BUZ11.

Just wanted to be sure. The NTE part I mentioned turns on hard at 4V drive. They aren't the cheapest things, but so far I like them ok for a jelly-bean part. Do you have a favorite cheapie logic-level part in this current range?