C16 exploded, carnage inside
 

Damn it stinks in here!

Just released a ton of smoke from C16. New MS3X build according to Frank's 99-05 instructions, finished it last week, did all the voltage tests on the bench from a stim+dc wall wart, everything was fucking perfect. Finally got around to flashing the firmware, things were going fine, I sit down to let it finish and hear a pop. Run over, pull USB cord, brain isn't working, takes me a few extra seconds to pull the DC power. Smoke stops.

There was lots of crap on the bottom of the MS3 board, but a quick douse with some alcohol and it's all gone, except for some hard baked-on crap around these two through-holes.

http://getfile9.posterous.com/getfil...egasquirt3.jpg

I got most of it off after taking this picture, but I wanted to know what the traces look like here for those who haven't fucked their shit up. Under closer inspection than I can photograph, it looks like there's a connection between the two plated holes here. These are the holes between the B and the 1 of "USB 1" and the cpu on the top side.

Seems OK to me?

And here's the C16 carnage, for your enjoyment.
http://getfile9.posterous.com/getfil...egasquirt1.jpg

Now, I'd like to think I got the orientation on C16 right because I'm an -----retentive jackass and this is the first time I can remember ever letting smoke out of an electronics project after building everything from silly little Heathkits to small ham radios, but there's a first time for everything, and I'm willing to admit when I'm a dumbass.

But really? I must have screwed with this for 20 minutes on the bench measuring the voltages in step 24 several times, plus setting the trimpots. Why the fuck would it blow now? I'm powering the stim from DC power supply from an old crappy wireless router, it says 12V, produced 15V at the stim. Still within the realm of what the MOV should handle, and I was seeing a perfect 12V on the CPU connector during testing... What the fuck?

So that's a pisser. Besides C16 and C17 (which got lightly toasted), what do I need to replace? What are the chances the CPU board is OK, given the damage and aborted flash?

Thanks guys :)

ouch!

I've done worse. Probably just c16 is bad and you should replace c17, this typically happens when you accidentally wire 12v to ground inside the CPU somewhere... but myself and reverant have both had it happen with brand new builds, and once powered up in the car, they pop and we replace and change nothing else and it's fine. IIRC reverant said he uses 36v rated ones, vs the ~18v or something ones that come in the kit.

.id have to go over the ms3 daughterbaord i have sitting here to see if those pins you mention so in fact be touching...that could be the cause.

Am I the only one who opened this thread expecting to see something much more dramatic, related to this:

https://www.miataturbo.net/attachmen...ine=1359122345

If you could do that, I'd appreciate it. I'm going to try to get some better macro shots of that area, but I've got a flight to catch and may not get to it until Monday...

Hope those pins are normal and this is just a really crappy wall-wart.

those two via's dont touch on the daughterboard i have here. possibly the issue.

Funny, I thought of this: https://www.miataturbo.net/attachmen...ine=1359124781

Yes, I was quite disappointed when I opened this one up.

I've been using 33uF, 35V for the C16 for more than 100 builds so far, no problems yet. 25V is too low.

Best shot I have of it so far.

http://getfile0.posterous.com/getfil...caled.1000.jpg

That's just dust between the two closest pins of the 4-pin usb connection.

yeah, those shouldn't touch like that.

id email this to DIY.

email sent. I'm afraid I won't be around again until Sunday night, but I'll be interested to see how this plays out.

it looks like leftover flux or something as well. shouldn't be like that at all, unless that is just part of the remains of c16...

Same here, brand new build, powered on the stim and C16 just popped. I wasn't in the room but luckely my wife had the sense to pull the plug when she heard the bang and saw the smoke :).
They should use higher rated caps in the kit.

I haven't gotten a chance to sit down and clean up the solder bridge yet, but I'm ordering the replacement caps now.

BOM spec's 22uF, 25V. Rev (or anyone else) - was your 33uF a typo, or a necessary change in capacitance to go along with the change from 25V->35V?

In other words: do I want a 22uF 35V or 33uF 35V?

Also: I went back in through my build process in my mind and realized I definitely installed these caps correctly.

It's shown as 33 µF on the schematic, as it its twin, C17:

https://www.miataturbo.net/attachmen...1&d=1359397294

Judging by their position in the circuit, however, the value of these capacitors is not terribly critical. They're just filtering and stabilizing the linear 5v regulator. In fact, I pulled up the datasheet for the LM2937ET-5.0, and it is extremely vague on the point. The sample schematic shows 0.1 µF for C16 and 10 µF for C17, with a note that, for C17 "The minimum output capacitance required to maintain stability is 10 µF (this value may be increased without limit)."

Long story short: 22 µF vs 33 µF is a toss-up. Rev says he likes 33 µF, and that agrees with the schematic, so go for it. The capacitance selection is relatively independent of the voltage rating.

You want 35V for C16. 33uF is preferred over 22uF, but not terribly important. I use 33uF/35V on the input (C16), and 22uF/25V on the output (C17).

Why is 25v too low of a rating? Input is ~12 VDC.

Not adverse to recommending replacing C16 with a 35v rated component -- but want to understand better.

Input is around 14V, and with various flybacks/spikes, and a capacitor that happens to be a little low on the 25V spec (it happens), you get a blown C16 if you are lucky, and if you are unlucky, it blows, shorts and you instantly fry several 12V traces on the mainboard, rendering it useless.

I've had a 0.6% failure rate on these so far.

I've decided to increase the rating to 35V to prevent it. One blown capacitor every 100 ECUs sold costs me more in damages (a full mainboard and shipping under warranty - not to mention the damage dealt to my reputation) than upgrading the infamous C16 to 35V.

Business as usual. :)

So I finally got to sit down with this today. The vias were in fact bridged and it looked and acted like solder to me. A few minutes with the iron and some wick did the trick. They are no longer connected electronically.

Now waiting on new caps from digikey.

IIRC, the daughter card schematics are not open-source, but does anyone know what those two vias went to?

Looked over the MS3 board a little more carefully.

Should these pins be bridged like this? There are some other pins on the opposite side that aren't quite bridged, but that definitely have solder trails that are nearly touching.

http://getfile0.posterous.com/getfil...caled.1000.jpg

Nope!

Yes, that's normal, see Megasquirt MSEXTRA / MS3EFI • MS3 Processor Problem (View topic)

I can look again, but it appears one goes to JP2 on the FLEX pin and the other to the CPU.

If it was a power and a ground I could see c16 blowing up from it.





the CPU pins are supossed to be like that, those are grounds.

A backwards-inserted 'lytic cap can see reverse voltage and will heat up, stew, and finally vent and/or explode, often leaving behind telltale capacitor confetti and the unmistakable stench of disemboweled electrolytic guts.

I've had cans rocket past my scalp more than once. (Rocketing was more common in the old days) I've since learned to power up a new circuit for just a few seconds, power down, then check for hot parts with a thermal camera. Before the advent of the latter I'd use a finger, carefully.

I like to run the MSes I build on the jimstim for a good long while if I cant test on my car, both units I've ever built that lost C16 only blew once installed on a car, and pretty much instantly roasted.

One happened on my car, I heard a pop during cranking (it would not quite catch) and smelled burning.

When I was building some of my early circuits, my dad let me reverse-wire some electrolytics just to see how far they'd shoot. Fun.

This board ran perfectly on the JimStim, without CPU, for several minutes while I was checking voltages and adjusting pots. Then it ran again for quite a while with the MS3 + X board while I screwed with R11. Interestingly, no matter how I screwed with R11, the voltage on the lower leg of R17 refused to move. It was (IIRC) 2.6v, I wanted 2.75v to be in the middle of the 2.5-3v range, but after a while I gave up. Perhaps this was a result of the bridged vias? Hmm.

Then again, if one of the vias was an output and was suddenly configured, and the other represented a short to that configuration, that could have done it. Of course that would mean a bunch of current probably went through my CPU...ugh.

clean up the solder and you can power up on the jimstim without c16 inplace. if you cant connect, you know the CPU is fried.

but I have a feeling it's still okay; when the CPU fries, it's instant. Blowing up c16 takes some time for the circuit ot get hot and explode things.

Tantalums, which C16 is, are even worse. Those fuckers are like a little firecracker when they go off.

Still happens to us here in the lab from time to time when bringing up a new board for the first time. Nearly all of our stuff these days runs off a single 48v supply and has local regulators to make 1.8, 3.3, 5, +/- 19, and whatever else it might need. One board in particular we have hanging on the wall from 2003, which we nicknamed Vesuvius. That one was ugly.

Thanks Scott, I may give that a try.

I don't think I've said it explicitly yet, but thanks to everyone who's chipped in here, even if I didn't reference you personally. Y'all are impressive electronics geeks.

Oooohhh it's a tant. I had one such fucker pop and fly. Then in the corner of my eye I saw some motion. Said fucker landed in a trash can and set some paper on fire. After I put the fire out I actually found the little shit in there, looking up at me, mocking me.

I don't like tantalums and will avoid using them when possible. They are also poor in the life and reliability department. If you need low ESR and some capacitance use a polymer cap (aka "solid electrolyte" formerly known as "OsCon"). If you don't need so much capacitance use a multilayer ceramic ("MLCC").

I see the value is about 35 uF. You would use an X5R (MLCC) cap instead of a tant. They're available in a 1206 package:
C3216X5R1E476M160AC TDK Corporation | 445-8047-6-ND | DigiKey

Or you can get this through hole guy:
http://www.digikey.com/product-detai...108-ND/2603953

Ceramic caps can take brief overvoltages no prob.

Maybe you got a batch where somebody pissed in the electrolyte?

Tantalums are sensitive little bastards- they get butt-hurt real quick if you over-volt them or subject them to reverse polarity. So we blow a lot of 'em up in the lab while prototyping and bringing up new products for the first time.

But in normal operation, they have proven to be exceedingly reliable. I can't recall ever having any serious problems with them. We've been using 'em in our audio consoles and routers for at least 15 years. A lot of our stuff is now old enough that it's starting to come back in for repair with dried-out electrolytics, but I don't think we've ever seen a systemic issue with tantalums.

FWIW, I don't see any ESR spec on that unit you posted.

Ceramics have no meaningful ESR (i.e. zero) for the majority of applications. (high current high frequency SMPS output stages are one exception, in some cases).

What you gotta watch out for is their capacitance changes hugely with voltage. For power supply bypassing, it's not a problem.

Tants need voltage derating to increase their life.
IIRC they are sensitive to overheating and like to explode, from ripple (AC) current.
CTTOI if the OP powered his board from a bench supply which for some reason was oscillating, that tant ate lots of ripple current and expressed its displeasure.

For my own education I looked up some 100 uF / 16V tantalums. The low-ESR ones form Vishay can only take 1 A, with 100 mΩ ESR. In contrast the mid-perf polymer Nichicon caps are 14 mΩ and can take >3A, with their bigger hi-current ones, capable of 5.6A with 7 mΩ. These solid electrolyte polymer caps make the highest perfromance 'lytics look wimpy, and have much longer life to boot.

Success! Got my new caps from DigiKey today, got them installed and tested everything out very, very carefully. Worked perfectly! It took the firmware no problemo and I've loaded Frank's 01 base MSQ (thanks Frank!).

Only issue I have now is the endplate fit. Even though my MS3 board is fully seated in the layers of standoffs, the USB port is maybe 1/8" too high to fit through the endplate. Is this common? I figured I'd borrow a file at work tomorrow and trim it up, but that seems a shame.

Very relieved that this worked out, very unhappy that QC failed so miserably on this card.

Again, thanks to everyone!

that's how everyone I've built has been, maybe 20 or so? I just push down on the usb port as I slide it fully closed. It's always about 1-2mm off.

picking this thread back up. Is DIY going to use higher voltage caps?

Actually I talked to Ben over the weekend, he said he'll push to get higher rated caps when their current supply run out. however, he also said they are really only seeing this in our community, because gaycar.

latest victim:

https://www.miataturbo.net/attachmen...ine=1368050180

https://www.miataturbo.net/attachmen...ine=1368050180

Wow, that looks toasty. Looks like a lot more collateral damage than I experienced. Did the board stay plugged in for a while?

Is that mine. hehe

no, yours was the worst i've seen :)

it blew c16, c17, d9, c18, and u6. I think I replaced c22 for good measure too.

The last one that happened on one of mine, fried the entire V3.0 mainboard.

that's crazy. I wonder what's causing it. Seems to be a MS3x only issue; possibly expander board issue?

Plain and simple: voltage spikes. 25V is too low.