Triacs in Parallel

Triacs are like SCRs, in that when the current drops close to zero, they cut off. You might need to do something to prevent this. You will also need to do something to balance the current anyway. How will you limit the current to 8 amps in either device? Any irregularity will cause more current to flow in one, than the other.

Probably, but you could try it. Make sure both have hard gate drives and a common heat sink.

At high currents, they will become more ohmic with positive TCs so should share reasonably well. The zero-TC current should show up on the data sheet.

Big potted power diodes are often a bunch of little diodes in parallel.

One assumes they're all matched for barrier height profile, though. And set so close together for minimum temperature difference? I'll give it a go; got nothing to lose after all.

Buy identical parts, of course.

We've made deals with distibutors to buy 20 parts off a reel so we can test them, and then buy the rest of that exact reel.

Measure their voltage drops or temperatures to estimate sharing.

Okay..... but these aren't adjustable characteristics, so what do I do with the resuts?

See if it works.

OK. I shall do so. Would you expect it to have a greater or lesser chance of success than doing the same thing with - say for example - regular power diodes? BTW, The maximum load for these two triacs in parallel will be 13 Amps between them. That's the worst case scenario which leaves 3 Amps to spare (not sure if that's a good thing or a bad thing, though).

Worst case draw will be 13A, so with one of them flat-out, the other only needs to handle the other 5A. I'm hoping such an imbalance as that won't occur in practice (unless it's intentional because the 'natural sharing' didn't work out).

Why would anyone expect them to have the same ohmic behavior? Semiconductors have large variations compared to many devices. This is exactly an area where people typically care about the worst case, and not at all about matching.

If the data sheet specifies the ohmic behavior enough to reliably match parts, then fine. But I bet a call to the manufacturer says, "Don't do that!"

Even if it passes your test, why would you expect it to continue to pass with parts from different lots or with even small process changes?

If this is for some home project, then fine. But build in some sort of protection, like a fuse in line with each triac.

Why would you need to match from different lots? You just need to match within the same lot, or perhaps pre-sort and pre-test them.

Yeah, with a tiny bit of isolation, that'll work. For paralleling diodes (and preventing current-hog problems) it takes a bit of series inductance for each diode. That should work for triacs, too (you don't get normal triggering on slower device #2 if device #1 has acheved low-terminal-voltage already, because triggering is less effective if main terminal voltage drop is low).

That's an interesting idea, using a bit of inductance to limit the current through each triac to <8A. Damn sight more practical than power resistors! Well done. I might just use that.

How many henries?

At what frequency?

Oh dear. I haven't done the sums yet and I have a feeling it's going to end up implausibly huge. The frequency is 50Hz.

Or infinite if there is any DC component in the current.

I don't think I can spare that much wire.

There's reason to believe the current-hog positive feedback that happens in diodes (and transistors) isn't present in thyristors, so my thought on the matching was only to limit dI/dt during the (zero-voltage switch) trigger, which might last 100us, not during the whole of an 8300us half-cycle. It's even OK if the inductor saturates, the next half-cycle will see that sorted out.