We return now to the highly specialized extreme toaster controller board, last seen languishing following a hardware issue...
OK, back story. A couple of weeks ago, one of the other engineers on the project, who's using some of the same little power-supply components as I am, inquired about short-circuit behavior. Seems his little power supply, when shorted, had a penchant for lighting up its smoke-emitting inductors.
So I set up to do a short-circuit test, and on my board, the regulator behaves as expected, i.e., drops to a very low duty cycle. Presumably, his issue has something to do with his (vendor-blessèd) use of the part in a perverting an inverting configuration, which involves some mind-warping trickery.
Removing the shorting-related mods, I found that the main (5V) regulator was still at the crazy-low duty cycle. Eh?
Um. The 5V bus shows 8Ω to ground. Not a dead short, like a solder splash or a stray lead clipping, oh no. Sounds like either a dead chip somewhere, or maybe, just maybe, a scrawny little lead-free tin whisker.
And there are only three of this board in the known universe. At last report, the production model was on hold pending some decisions regarding the higher-level product architecture. The other two lab boards are firmly tied down. I gotta fix the one I have.
So, poking around. Remove the cap where I'd attached the shorting point; no stray solder under there. Lift the output lead of the regulator. Disconnect the power lead from a tacked-on supplemental amplifier boardlet. Start lifting other 5V power leads... oh, crap. I done broke off the input lead of the precision 2.5V reference regulator.
And that's where I'd left it, apart from ordering a replacement reference chip and spares for all the 5V-powered parts on the board.
Today, it's time to revisit the situation. There are several 5V-powered components scattered around the board. Some have things inconveniently piggybacked on them (this is, after all, the lab board; there's a fair bit of rework). Lifting pins is gonna be awkward.
Inspiration! The kind of inspiration I wouldn't normally follow through on without Boris egging me on, but maybe I'm a bit sleep-deprived or something.
Set the lab supply for 5V. Connect it to the 5V rail. Start cranking up the current limit.
A little before 1 Amp, with around 4.5V showing, suddenly the current drain dropped to nearly nothing. There was a tiny puff of Magic Blue Smoke; alas, I didn't see whence it came. But! A quick sniff narrowed it down to the warmup timer circuit, which means it's either U10 (level translator for 5V gate drive) or U17 (buffer amp).
I dunno how I would have fried either of those parts, but it looks like I should have this thing fixed in nothing flat. Well, given the fiddly little parts, and the previously-removed components that need reinstalling, maybe the rest of the afternoon.
Then I can get back to dealing with those less-than-urgent firmware issues that nonetheless do need to be resolved before product release (which I don't think is going to be last October at this rate).
Update: Guess what I didn't order? Right, that level translator. Oh, well: I can work around its absence for a while. Got spares of lots of other parts now, though.
Update 2: Rippin'-slippin'-dang-fang-rotten-zarg-barg-a-ding-dong! Re-install bits of 5V supply, plus the 3.3V regulator. Cautiously apply input power: it works. Get it put back together, modulo the level translator. Re-apply input power, incautiously. Nothin'. Well... a hint of voltage on the 5V rail, but it's definitely in shutdown. Remove last two components. Still not working. Use the bench supply for 5V: 100 mA, and all's well. Reconnect bench supply to input, and crank up the voltage. Yup, the supplies come up.
Some fiddling later: the 5V supply now works if I cautiously apply input voltage. If I just crank it up to 15V and turn it on, no worky. In fact... it works up to just about 12V input, and then turns itself off and sulks. Whaaaaa? It's designed for 15V input. It always used to work at 15V input. It's supposed to work up to 25V (limited by capacitor ratings).
With the regulator running happily on 11.7V input, if I touch the ENABLE pin with a voltmeter, (a) the meter shows 4.75V, and (b) the regulator shuts down until I cycle input power. According to the datasheet, anything above 1.3V on ENABLE means "turn on". Also, there's not supposed to be a permanent sulk mode.
Guess I need to poke it with a scope instead of a meter, and try to get some clue what's going on. Could be the chip got somehow damaged, or maybe there's a screwy leakage path.
Update 3: Sulk mode definitely looks like current limiting. But why? Current limiting is based on voltage across the MOSFET. Inadequate gate drive? Is the boost capacitor bad? Check waveforms: boost capacitor is fine. Voltage on BOOST pin... well, the datasheet isn't terribly informative, and I didn't observe it back when things were behaving themselves, but it's sitting at somewhat below the supply voltage, and pinging up to, well, roughly supply voltage above that, as expected. I'm not exactly sure what's supposed to appear at that pin... it seems to be plenty high during the ON time.
Oh, wait: that was in sulk mode. When it's actually running, the BOOST pin sits a couple of volts above the input, and pings up from there. This makes more sense, though I'm thinking it should sit more than a couple of volts above the rail.
So, yeah: looks like the chip got damaged somehow. Wonder if I've got any spares, or if the client has any available.
Update 4: Client had spare chips. Surprise! The new chip also cuts out at 12.mumble volts input. So... um... well, the other thing that could trigger sulk mode would be maybe the over-voltage protection feature, but that never tripped before. So: look at the voltage at the filter cap. Got a bit of a pedestal during the ON time, increasing in height as I crank up the input voltage. And, depending on where exactly I probe it... yes, the great wide PCB traces appear to be having a significant filtering effect. It's like the inductor is Not Doing Its Job. Also, it's getting a mite toasty. (The bringup log shows that, when I first installed that inductor, its temperature rise was 4°C, so something has changed for the worse. Also, I thought the pulse pattern looked lumpy, though I couldn't remember what it looked like before.)
Which gets back to that other engineer's original question: "Do your inductors exhibit smoke-bushing failure during a short circuit?"
Now, since that inductor was a rework item, and I never buy small fiddly rework parts by the one-each, I should have at least one spare in my parts bin. Assuming I can find the parts bin for that project....
Oh, foo faddle. Did I give the spare(s) to the client, to rework their copy of the board? Probably, though given my casual approach to record-keeping on small parts, I don't know for sure.
And so, tomorrow A.M., it's off to Anchor and/or Halted in search of a reasonable substitute. Basically, anywhere from 47 to 100 μH and capable of handling 250 mA or so. (Actually, I might have something that qualifies just sitting in one of my other parts bins. Maybe I'll check that first.)
Update 4: Found some 56 μH parts at Anchor that even fit the footprint! With some actual inductance in the circuit, the pulse pattern looks sane, and nothing gets noticeably warm.
So... somehow... that inductor had gotten toasted, along with a downstream chip. Maybe I need to compare notes with the other engineer.
Recent Comments