Small Schottky diodes at very high currents

One aspect of the measurements I don't understand, for the 350A flyback scope plots, 2nd page. Why did the D1 voltage (green) level out, at about a volt, after falling for 1us, while the current was still 150A (blue)? Shouldn't it have been 5-volts, and continued falling at that point? It's too bad I can't go back into the lab to check out that, and other issues.

Beware pulsing diodes at high currents, even for short durations. There's an ultimate failure mode in there, perhaps electromigration?

Once had a badly ringing inverter (industrial, 5kW, SOT-227 FETs), which I clamped by wiring diodes to the opposite rails. Very brief pulses, 60ns or so, but around 100A peak. Started with 8A diodes. Poof. Ran for a few cycles, then failed shorted. (The transistors were fine because I implemented a desat detector, of course.)

Tried 12A diodes. Ran for a few seconds -- enough that if they were cooking off, they should've heated up some. Nope, stone cold. Hrm.

Tried 30A diodes. No failures for the rest of that round of prototyping.

The peak forward voltage was something like 60V, only a little of which is attributable to stray inductance.

Those were PN not schottky, but you're well into the guard ring forward-bias region so it should be equally relevant. I don't know how many cycles you'll get on that poor diode; if you're just doing single shots, it might be okay for a while.

(The moral of the story was actually to _not_ minimize inductance: that was patently impossible, as I was already using a 4-layer board, and the remaining ~15nH of loop inductance was in the devices themselves. The next revision opened up ~100nH of loop area in the PCB, putting an RCD snubber beside the transistor pairs. Current was also reduced, using an H-bridge rather than 2+2 in parallel for a half bridge. Total capacity went up, because we didn't have to use 1200V transistors anymore -- which at the time were atrociously bad, as SJ wasn't introduced yet in ~2010, or not in devices of this size anyway.)

Tim

Thanks Tim, I appreciate your stories.

One imagines that electromigration takes time? But yes, we've seen failures. My fav SMC diode has been doing well, zero failures, even at 900 amps. But Rob Legg has had the opposite experience, starting with a different part. So we've selected some candidate types, and are preparing a series of tests to explore robustness and identify good ones. We're looking at both SMC and D-Pak, but if necessary, I will have to modify the PCB to make room for larger packages.

It may also be a matter of di/dt. Examine my scope waveforms, you'll see that it takes 2us to reach the 300A level. This is intentional, to protect the LED. And it's actually a lower di/dt than I've observed in cases of half-bridge and H-bridge diode conduction.

Moreover, unlike the case of a half-bridge, etc., we're not struggling to keep clamp-diode capacitance low. The SL44 has 2000pF of capacitance at 5 volts. It would probably be a horrible choice in a bridge. The SL44 is a 40V part, good for most high-current instances, but we're looking for another candidate to handle higher-voltage applications, to 100V, which BTW are usually at lower currents.

Schottky construction, even high-voltage types, is a factor. They don't suffer from reverse saturation current, and subsequent dangerous high dV/dt snap-off spikes. Maybe that was a factor in your failures?

Wikipedia says low-voltage Schottky's may use overlap metallization to spread out the field gradient, rather than guard rings, for protection at high voltages. So I checked, and my SL44 does use a guard ring. But I haven't seen any sign of reverse-recovery snapoff. However this may an issue as we select a 100V part.

Hi Win, I was starring at page two... Well first it looks like D1 cuts out at >~0.5V kinda what you would expect. As far as why current is still flowing (why diode voltage isn't higher.) I can only guess the current is flowing somewhere else. Some 'slow' (us) recovery current into the big ass cap?

George H.

Exactly. There's some diode-drop voltage, so clearly there's at least 25 to 50A flowing in D1, but at 150A D1 should still have about 3 to 3.5 volts across it.

The circuit at that point is pretty simple: a current loop consisting of D1, R2 R3, the LED, and the wiring inductance that's driving the current. The Rogowski probe was wrapped around the LED return wires.** Both D1 and the LED have high capacitance, but at 150A current levels t = C V / I is sub-nanoseconds!

OK it looks like the 6000 uF cap to ground is still in the circuit too? But I'm not really sure how the circuit knows where ground is once the FET is off. (It's most likely that I don't understand the circuit.) So what's the jump in the diode voltage when the fet turns off?

George H.

The 6000uF 4mR capaci5or bank is charged to 30V, and provides high LED current when the FET is on. When the FET is off, the flyback goes up to about 30V, plus voltage-drop across D1, now conducting. The "30V" level is adjusted to be 0V in the Excel version of the scope waveform, but I don't know exactly where that is, because the 6000uF cap has lost some of its charge running the LED. So diode D1 is ON from 1 to 3us, but I don't know exactly what its voltage was then, because there wasn't a probe on the capacitor bank. Haha, I needed a five-channel scope. And I can't repeat the experiment to get the answer, because nobody is allowed in the building now.

It's not apparent from this trace, but the negative voltage at the end of diode conduction is charge stored on the LED (and other stray) capacity.

If the trace lasted long enough, you'd see this voltage discharge, (through the LED leakage, it is assumed) to the starting level, exhibited on the left-hand point in the display.

In early test stages with other emitters, resistors could be added across the diode, just to manipulate the time constant of this specific discharge, in order to assess the capacitance's volume and guess at it's location. The lamp was the 'usual suspect'.

The Osram parts have a built-in reverse-bias protection diode, which complicates things, but whose current might show up in actual LED current traces, if D1 isn't doing the job. The reverse protection is at a relatively higher voltage than a schottky would permit, under the same conditions, but D1 and the internal protection diode aren't in identical locations. (20mA @ 1.2V typ).

This reverse protection isn't a feature of all makes and models of emitting sources.

Whatever current isn't seen on R1, is going wherever stray loop inductances will permit it, generating node voltages that intentional and stray capacitances will allow. The semiconductors are just hanging on for dear life, relatively speaking . . . ;-)

These are difficult component voltages to record, differentially, and current measurements that don't alter circuit conditions, or bounce common-mode strays all over the place, are no joke either.

When you place a D1 type of diode into the DUT location, to test for surge integrity, it can get to become a question of " who's on first? ", when it comes to assigning stray effects.

RL