MOSFET body diodes

I have blown fets in h-bridges driving motors, at very low currents. Seems the substrate diodes had a snap recovery characteristic that blew the gates out somehow. These were older power fets, and I think some newer parts are rated to survive this.

John

If you've got a half-bridge, the slow substrate diode can cause conduction at full supply voltage through the opposite MOSFET during the reverse recovery time. That can significantly increase dissipation in the MOSFET.

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

That's not an option in H-bridge and half-bridge designs where the voltage across / current through the motor can be reversed. The use of a paralleled Schottky diode across each FET not only fights the rail-rail shoothro problem Tony mentioned, but also stops the severe reverse-recovery-time snapoff-EMI problem dead in its tracks.

--
 Thanks,
    - Win

I'm controlling a small DC motor with a MOSFET (irf7341). All diagrams I see have additional Schottky diodes in parallel with the MOSFET. I'm wondering if the body diode isn't enough in this application. The power level is small; I'm anticipating no more than 3A average drive current.

You might have guessed my education here consists almost entirely of application notes and datasheets. Is it just the reverse recovery time that indicates the need for an additional diode? The irf7341 has a t-rr of 90ns max, which seems quite fast. What am I looking for to guide the design? What search terms can I use to Google more info?

Thanks. Mike.

This is an interesting theory, but I find it hard to believe. Any parasitic Lstrays in series with the drain or source would tend to cause the drain voltage to increase, and source voltage to decrease. The gate however is a relatively large capacitor, and it's voltage cannot change instantly. Any change in gate voltage must be supplied with current through the gate driver circuit. The really important thing however, is that the body diode snap off event would try to increase the gate source voltage (for an N-channel device). As the gate source voltage were to increase beyond the threshold voltage of the MOSFET, the device would begin to turn on, thus inherently limiting the peak dI/dt event, and consequently, the maximum gate-source voltage reached.

Far more likely in my mind is the high speed snap off will cause the drain-source voltage to increase sufficiently that it will cause drain-source avalanche. The drain-source capacitor is much smaller than the gate source capacitor, so the drain source voltage is free to change much more readily by stray L's than the gate source voltage. Note that the peak avalanche current could be as high as the peak reverse recovery current. Depending upon how good or bad the body diode is, and how fast the reverse recovery event is forced (presumably by the speed at which the opposing MOSFET in a half or full bridge turns on), the peak diode reverse current may conceivably be higher than the load current.

This has some interesting implications. This means peak MOSFET drain-source avalanche current during reverse recovery could be very high, and is quite unpredictable given the inadequate reverse recovery information provided by most MOSFET manufacturer's datasheets. Even the best datasheets leave me wanting, it seems reverse recovery characteristics are notoriously undercharacterized by both the MOSFET and diode industries.

If the MOSFET is sufficiently avalanche rugged, the MOSFET should survive totally unharmed. On the other hand, most MOSFETs (even avalanche rated MOSFETs) don't have unlimited peak avalanche capability. As International Rectifier's application note AN 1005 suggests:

It would seem MOSFET destruction during avalanche can occur either due to thermal failure, or too much peak current.

Avalanche ruggedness is a very nice feature in a MOSFET, even if your intended application isn't supposed to cause normal avalanche conditions. An RC snubber between drain and source can't necessarily always prevent avalanche under these conditions since such a circuit has it's own parasitic Lstray which would limit how fast the current can ramp up during snap off. Additionally, there are practical limits as to how close an RC snubber could be placed, some unclamped Lstray will still exist between MOSFET drain-source since some Lstray is built right into the MOSFET package.

Note that when a discrete MOSFET fails it almost always fails (unless the failure is so spectacular as to fuse or blow the package apart) with all three terminals shorted together with relatively modest resistance. This could lead some to erroneously attribute the MOSFET failure with gate oxide failure. Note that the gate oxide is very thin and naturally must cover the entire die in order to access all of the MOSFET cells. This means that if even one tiny localized part of the die should fail (explode for instance), the gate oxide will probably be locally damaged, thus producing the gate-drain-source shorts without ever having to subject the gate-source capacitance to excessively high voltage.

There's also the Grehkov drift step-recovery effect: a diode that normally would be a mediocre, soft snapper (classic mode, dc forward bias followed by rapid current reversal) can become a very nice, very fast step-recovery diode if it's only forward biased for a few hundred ns. The charges don't have time to spread out much, so are swept away very cleanly when the current reverses.

This can play nicely into the typical anti-shoot-through gate drive timing of a fet totem pole:

current happens to be flowing into the motor,

bottom n-fet is on, conducting "backwards",

drop the bottom n-fet gate drive,

wait a couple hundred ns for anti-overlap,

during which time the drain swings negative and the substrate diode conducts,

then the upper fet turns on,

fet-fet current builds up big time,

lower fet substrate diode snaps off,

zowie!

John

Google "Grekhov diode" (I think that's the right spelling) and "drift step-recovery diode". Ordinary power diodes can do remarkable things, if you get lucky.

Here's one I did...

There's a second Grekhov effect, a superfast delayed avalanche thing, but that's different.

John

A diode in parallel with the MOSFET is in the wrong place to perform the function of providing a path for the inductive current when the switch turns off. For that you need a diode in parallel with the motor.

-Robert Scott Ypsilanti, Michigan

DOH! No need for bloodletting. This is a diversion for me... I'm building a pocket size Segway starting with parts I already have on hand: a wad of

16F684 and 16F690 (I know, but wait till New Years to draw and quarter me; I want to see if I can't make it work); a dual axis accelerometer; and assorted big and small DC motors. It can grow from there based on early failures.

The starting points are Microchip's AN964, "PID Control of an Inverted Pendulum", and a few dozen other datasheets and app notes, which together form roughly my entire knowledge of making things spin and move. As to H-bridge, versus single or multiple FETs or types, I have two irf7341Q and a dozen or so ZHCS1006 Schottkys left over from something else. (The point of the question was quite simple. Tacking a few SOT23's to the 50 mil proto-board had me wishing this wasn't strictly necessary; studying the datasheet left me no wiser.)

I hope you won't stop taking my questions seriously. I'm happy to grope along in the dark finding my own way as I have been, but the deeper discussions of how/why/better are of very keen interest to me. There's little point in dummying it down for my sake; eavesdropping in on the real discussions is among the better ways to learn. (I'm pretty good in my own chosen profession; I just happen to be a newbie dabbler in yours.) Anyway, it's intended as a Christmas gift for my machine shop Elmer, and a pleasant way to spend my evenings and weekends. If it falls flat, he'll have to pretend joy over a bottle of wine instead, which shouldn't be hard at all.

To answer your question, I have "two irf7341Q. A dozen ZHSC1006 Schottkys." I'm not wed to using them, or in any particular fashion, except what works well, let alone best. They're in the part box; Digikey is less than 3 days lead time if I want to play with something else.

I only found, just now, what looks like the early announcements of the "real" thing. So, did they manage a toy size with PICs?

Yikes. Bigger would be easier to control, too, at least in relation to processor speed.

Yes, that's what I want. Had been wondering what to use for its self-determinism.

Let me ask it a different way: the irf7341Q seems to have a fairly robust diode.

Under what motor drive conditions would it NOT require a separate, parallel Schottky? Conversely, what motor drive conditions would indicate one?

Thanks.

Curious. What's the body diode for, then? This would be the root of the original question.

I don't understand that part, and it seemed a bit controversial. Most of the discussion went way over my head.

Sorta like a unicycle, but with two coaxial wheels, so only 1 DOF initially, forward and back. The balance problem is similar to that of an inverted pendulum. Bi-directional drive is an inherent requirement.

The second DOF is yaw, but can be addressed independently of the first one of balance. At this stage of considering motor control, it only implies independent drive for each of the two wheels.

John There is difference in Gate Source drive between IRF540 ad IRF540N which was standardised for many modern MOSFET.

It's a "feature" of the MOSFET manufacturing process.

In general, if the datasheet have specifications for it, it is safe to use it - although I would use some sort of Transorb/RCD snubber on the MOSFET's. People who put shottky's in bridges do not want to use the parasitic diodes and they do not want the snubbers either.

PS:

You need bidirectional drive to get the thing to balance.

Hi John,

Win mentioned this in a post a few months back - seems the snap-off can be sub-ns, and the resultant enormous dI/dt drops enough V across stray L to over-voltage the gates.

If we guess Lstray = 10nH and Vsplat = 30V, Idead = 30V*1ns/10nH = 3A. So with sub-ns snap-off, its probably fairly easy to do....

Cheers Terry

"Mike Young"

** So you are using only one mosfet or one type of mosfet ?

Try to post unambiguously.

Ambiguous posts lead to flame wars.

......... Phil

"Mike Young"

** Just answer the damn question.

We have NO idea what "diagrams " YOU have looked at !!!!

( snip several hundred superfluous words that DID NOT answer the question)

......... Phil