Today I heard a loud bang that tripped the circuit breaker in my server room. It must have been really loud because I could hear it 2 rooms away through 2 heavy doors and it was like a firecracker going off right next to me.

Long story short, it narrowed it down to one PSU from one of the computers. It smelled like burnt rubber and was really hot even after ~40 mins of being turned off when I finally got to testing it. All the rest of the tech was thankfully fine.

Its an old server PSU, like 10+ years old so not really surprised it blew up. Its a 800W unit made by HP but I couldn't find any model identification on it.

The weird thing is, I opened it up to really make sure this is the thing that failed but on the inside, it looks totally fine. Tested the fuse - all good, all the caps look good, no charring anywhere. After around 10 mins of looking inside the burning smell had faded away too. Still though, its the only thing that won't turn on. I got the rest of the computer it was attached to back up and running with a replacement PSU.

At this point I'm just curious - what could possibly create such a bang and not leave a trace afterwards?

  • 29
    Normally, the biggest bangs come from large electrolytic capacitors, but they generally leave a rather obvious mess when they let go. – Dave Tweed Jun 12 at 13:21
  • 8
    One candidate is rectifier diodes (or bridge rectifier module) failing dead short circuit. If the breaker tripped fast enough, perhaps it saved the fuse. I have been lucky enough once or twice in the past that ONLY the diodes failed; so a relatively cheap easy repair. But I wouldn't count on it. – Brian Drummond Jun 12 at 13:34
  • 27
    Also, it's possible that a component explosion is not what you heard. The component failures could be secondary to an arc that was caused by a surge coming in on the mains, and it was the arc itself that created the bang. It just happened to be this particular unit that broke down, sacrificing itself to save the others. – Dave Tweed Jun 12 at 13:45
  • 6
    I once had a PSU do this. It did not take out the fuse. It did not trip the 30A breaker on the mains circuit. It did take out a 150A fuse in a box high on a corridor wall, which had been there since the 1930s and which generations of electricians had forgotten about. Half the building was in darkness for the rest of the day. After which I replaced the PSU and the PC was fine. I don't know what failed, but it failed in two stages. Somebody brought the PC to me because it would not power up. I plugged it in to diagnose and it went bang, ear-ringingly loudly. – nigel222 Jun 12 at 16:07
  • 8
    since all the answers are mostly guesswork, perhaps a picture or two of the PSU board could help.. – WooShell Jun 14 at 8:55

13 Answers 13

Lead batteries, used in a UPS, can explode violently due to hydrogen gas buildup.
Mechanical damage will be evident, since the battery encapsulation will have failed.
I put my money on this, if it can be heard from rooms away.

Diodes and traces can explode without much mechanical damage or residue. Yet they can sound like a small lightning bolt depending on the fault current capacity or in other words, the energy let through by the protection circuitry.

Electrolytic capacitors can rocket off a board, but they tend to poof a lot of smoke and crud. However, in a server the airflow might dissipate this quickly.

Tantalum and ceramic capacitors go up in flames. Not much bang.

Resistors will often show burning of the PCB first. Otherwise they've exploded on a surge condition, and they will be scattered around the enclosure, similar to diodes.

Fuses will only explode when selectivity or breaking capacity is improperly allocated.

  • I feel your answer is the most robust (more so than mine), because yours lists many components and their level of auditory action. – Bort Jun 12 at 17:26
  • "Diodes and traces" - For diodes I agree but 'traces'? Really? How is it possible? – DannyS Jun 12 at 22:04
  • 10
    @DannyS: With enough current passing through it, a PCB trace will happily act like a fuse. – myersjustinc Jun 12 at 22:14
  • 5
    Perhaps I've misunderstood, but I believe op was talking about a PSU (power supply unit - internal to a computer case), not a UPS (uninterruptible power supply). – matt_rule Jun 13 at 12:17
  • 2
    @matt_rule I know. Did you check them? – Jeroen3 Jun 13 at 13:58

Judging by the sheer volume, my bet is on an electrolytic capacitor. Those can build up a lot of pressure and violently burst under the right (wrong) conditions.

I know you said the caps "look good", though electrolytic failures are not always obvious upon first glance. Sometimes they vent on top, with just a small slit. Sometimes they vent from below (making it hard to notice). The bottoms can have a "plug" that can pop out of place and it's hard to discern from above. There is not always going to be charring/discoloration or visible fluid leakage.

I would check again to look under the caps (if possible). Perhaps remove the through-hole caps and inspect them from beneath. Check them with a multimeter to be certain they haven't failed.

  • 1
    Afaik capacitor cases are specially cut in order to NOT allow building up pressure. – Agent_L Jun 14 at 14:24
  • 2
    "I know you said the caps look good, though electrolytic failures are not always obvious upon first glance". I've seen several bits of tech exhibit this behaviour, where the capacitors have ruptured in a non-obvious way. On closer inspection you can see the slits have ruptured. It's worth noting that when they rupture in this way, they do not go with a bang, otherwise they'd have blown more obviously/violently, so it's probably not what happened in OP's case. – Doctor Jones Jun 14 at 14:40
  • 1
    @Agent_L - Yes, caps are specially designed and cut for a more controlled rupture. Take a fuse for example. They are meant to open, but with enough voltage, they will continue to conduct for extended periods before exploding. I have found overvoltage on caps to cause a slow degrade, but reverse voltage, even at half the magnitude of the positive voltage rating can make a great firecracker simulator in a short time – Bort Jun 14 at 17:08
  • 1
    @Bort Right, but venting immediately is still the exact opposite of "building up a lot of pressure". – Agent_L Jun 15 at 12:20
  • 1
    @Agent_L - What causes the immediate venting? It must be a lot of pressure to vent that quickly, no? My thinking is: Pressure requires time to propagate. If pressure builds up quickly, it can accumulate faster than the safety mechanisms can react. – Bort Jun 15 at 13:15

As other people have said, electrolytic caps are the usual culprit here. I've had a big one light off like a Roman candle, six inches in front of my nose, while I was fault-finding a board. A 4-foot plume of smoke went up right in front of my eyes, and I'm very lucky I wasn't leaning over the board a bit more.

If silicon has failed though, it can actually be hard to see. A chip that's gone bang generally has a small but significant pit in the centre where the device has blown out. You often have to really look for this though, because it's often not immediately obvious.

  • 14
    I have a 200A, 400V SCR I blew about 40 years as an apprentice. You really can't see any damage to it now but it clearly measures as a three terminal link and I remember it vaporizing at least two of the tracks on my PCB. It now serves as a trophy, nobody wants to own, for the best recent explosion in the lab. If you work with power electronics the occasional bang is par for the course. With luck we learn as we go on but the engineer who never made a mistake has not made antthing – Warren Hill Jun 12 at 17:52
  • 3
    I was once in the same room as a 5MW IGBT that failed catastrophically at just a bit over half power. That was intense. Even silicon devices can make quite the party when they go. But the power involved was ofc not breadboard size. (It weighed close to 20kg) – Stian Yttervik Jun 12 at 19:57
  • @StianYttervik I started my career at a place which built HVDC links and static var compensators, but I didn't stick around there long enough to have any good war stories of blowing stuff myself. One of my friends there did have to persuade a python to leave a wiring channel when commissioning a site in India, which isn't something you usually find on a project risk register - it took care of the rats nibbling the insulation though. – Graham Jun 12 at 21:36
  • I've had someone wire up an OpAmp round the wrong way before. tiny component, big bang. I'd compare it to a starter's pistol. – Baldrickk Jun 14 at 8:23

I'm inclined to think a MOSFET has failed.

The MOSFETs are the hottest parts in any SMPS system and can catastrophically fail when overheated. Unlike most materials, whose electrical resistance increases as temperature increases, the resistance of silicon-based semiconductors, including MOSFETs, will actually start to decrease when their temperature reaches about 160 °C, and continue to drop as temperature increases beyond that point.

This unusual behavior means that when a MOSFET overheats, it enters a feedback loop where the lower resistance causes more current to pass through the MOSFET, making it even hotter. This is called thermal runaway. The device eventually fails catastrophically, with the temperature increasing so rapidly that it will often explode, potentially even causing a fire. A video of a MOSFET exploding in 20× slow motion (600 fps recording played at 30 fps) shows how this can happen.

Due to this resistance drop, a MOSFET will typically fail short as it undergoes thermal runaway, potentially drawing enough power to trip the circuit breaker before it self-destructs completely.

Possibly a part failed short and allowed large currents to flow. You may find that a trace has been cleanly vaporized from the PCB and/or a fuse has blown.

Large currents cause things to move and vibrate from the forces and temperatures and can cause sound. This is particularly true in commercial/industrial situations where huge fault currents are available, in some cases exceeding the ability of ordinary fuses to cleanly break the current, which can cause the fuse itself to explode, which definitely causes a large bang (and glass shrapnel).

Capacitors mainly, I worked at a place where faulty Chinese capacitors (a company stole an electrolyte formula, but not the whole thing) would bring down power supplies on a weekly basis. We'd get some pretty worried people because the sound was loud and then their computer would shut down. You can tell most of the time because it looks like there is shredded up paper inside the power supply.

Anything can melt really, but most of the time it's capacitors that degrade and go out with a bang. Other components usually fail at design time (like not sizing a switching regulator resistor or inductor for the appropriate currents, but even then these typically melt during failure from what I've seen)

I've also seen transistors blow up several times.

I did have a relay blow up in my face from stupidity and almost took my eye out.

I'll bet if you examined your supply and looked at all of the caps, you'd find one that wasn't quite like the others and that would be the offending one. If you can't see it it doesn't mean a component hasn't failed, I'd get a meter out and start testing components to see which one failed. I would also look on the bottom of the PCB, which might tell more than the top.

A MOV can also make a loud bang. I once had a power supply where the MOV failed with explosion, smoke and hot rubbery smell - whether or not it was doing its job or had some kind of defect I don't know. Anyway the manufacturer didn't seem surprised but sent me spare MOVs, I replaced it, no problems.

  • 1
    Large MOVs can be like small hand grenades. – Robert Endl Jun 14 at 8:42

As partially mentioned in other posts, semiconductor parts can explosively fail too, not just the usually suspected capacitors.

One of the reasons is that the actual semiconductor chip is connected via extremely fine wires inside the case, with the whole molded into a plastic block. If there is a really serious current surge, this wire can suddenly vaporize within the confines of the hard plastic, probably creating a plasma arc from metal ions torn out of the wire ends. The pressure and thermal stress can get the better of the plastic encapsulation, which tends to be heavy on fillers and duroplastic on semiconductor parts, so it will not simply melt but rupture.

If a tiny length of bond wire seems unlikely to be able to create such a bang - read up on what an EBW detonator is and what it can do :)

I had one of those HP 800 PSU's (from a Proliant G4 or G5, I forget which) fail with a loud bang too. Scared the hell out of the people in the offices near the serverroom.

It looked OK on first inspection, but I later found out that the problem was mostly hidden beneath one of the larger components on the PCB.
One of the 12V PCB traces had actually snapped, leaving a 1 millimeter gap with some visible burn-damage. The copper was just gone. Vaporized I presume.
The remaining traces left en right of the gap were ripped free from the PCB over a distance of about 8 mm on one side and 4 mm on the other.
As these traces can already carry as much as 65 A in these PSU's under normal load it seems likely that some instability caused the trace to be fed with even more power at which point it failed.
The sound was probably the super-heated vaporized copper and air bubble rapidly expanding faster than the speed of sound and causing a miniature sonic boom.

Wire arcing will explain it

Consider the lowly mains breaker. Typically they have a thermal trip that will pop at 110% of circuit capacity, but take a half hour to do it. They also have a magnetic trip that trips the breaker in a cycle or two, but that will not operate below 1000% of breaker rating (so it doesn't trip on inrush from motors starting, PSU caps charging, etc.) Of course it will also trip on a higher current flow, say 5000%. So let's contemplate that one.

5000% of a 20A breaker is 1000A. Our mains power is what, 120V? That's 120kw, or 120,000 joules/sec. Now a .44 magnum, Dirty Harry's weapon, is 1150 joules and he got 6. Or was it only 5? Well, your short is banging off 100 or 120 of these a second, though hopefully the breaker is going to trip after just a couple of shots.

Anyway, that would explain the noise pretty definitely.

Maybe a tangent, but my coffee machine (a very nice Italian prosumer espresso machine) had some kind of electric failure recently which meant there was some short somewhere, probably in the cabling.

This showed itself through a veritable lightning bolt and an incredibly loud bang, which most certainly would be heard rooms away. And obviously the breakers tripping.

It was so loud that I noticed that I actually developed a bodily reaction during "testing" (i.e., trying to figure out whether it was a one-time fluke, or a repeated thing, after it happened again after a few days); i.e., I was only physically able to switch it on when wearing those headphone-like sound suppressors.

Long story short; it turned out it was just a short at some of the cables. You don't actually need any part to "explode". The symptoms were the same as yours; i.e., it did smell at the beginning, but the smell went away quickly, and there was no noticeable burn/char signs inside, anywhere.

So, without knowing about your machine, I'd say don't rule out a honest-to-god short in the 220V/110V path somewhere.

Bridge rectifiers. A lot of switching power supplies use a bridge rectifier on the wall outlet side to convert the AC to DC. This is the first thing to see a surge, assuming no filters, and frequently has a robust package. When it goes, there can be a short in the power main until it vaporizes internally.

Get yourself a IR thermometer with a narrow angle and you can check the temp of components with it. If it isn't a resistor and is hot, its lifetime is decreasing.

And it could have been the circuit breaker arcing when it tripped, you didn't mention its rating.

  • Or better yet, a thermal camera. Basic smartphone-based thermal imagers can be had for as little as $200. I personally use a Seek RevealPRO, which retails for $700, but punches well above its weight. The software's not as advanced as what FLIR makes and the image quality isn't quite up to the same standard, but those devices easily run several thousand dollars. Even so, 320x240 resolution is nothing to scoff at. – bwDraco Jun 13 at 17:45

Maybe it is some conductor, like a piece of copper wire or a bug. If a copper wire suddenly came into contact with 2 rails in the PSU that are not supposed to be shorted, an electric arc may be produced. For example, if a copper wire comes in from the fan, and it touches the chassis and the 12V rail, then all the energy in the capacitor will be discharged. The sound may be loud. And the temperature may be high enough to vaporize copper.

protected by W5VO Jun 15 at 14:54

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?

Not the answer you're looking for? Browse other questions tagged or ask your own question.