As an electronics hobbyist, I am quite familiar with the smell of fried electronics and burnt plastic. But I usually smell them immediately after I made a mistake, not waking up to that smell in the morning. I shot out of bed looking for the source, thinking one of my projects died overnight. But it wasn’t one of my projects. The smell was coming from my TrueNAS CORE server, which was surprisingly still running. After I properly shut it down and disassembled its power supply, this is what I saw:
Not good, something got hot enough to melt a hole.
The heat source appears to be this row of surface mount resistors. While a resistor is expected to turn electricity into heat, they’re not supposed to get this hot.
There are no components on the other side of the circuit board behind those transistors. Just a few identifiers “2960323904” and “DC-3266”.
The scorch marks left inside the power supply enclosure imply they got hot enough to start burning, which was thankfully contained within the enclosure and did not turn into a house fire. Hooray for electrical fire safety regulations! However, I’m concerned at the fact that it caught fire at all, as I would have the power supply to shut down when something goes wrong. Every one of my previous failed power supplies would shut down and refuse to run again, but this one kept running merrily along even after it had (briefly) caught on fire!
There were no signs of damage beyond this vertical riser circuit board. I saw nothing I recognized as a fuse, though I’m sure there are form factors I don’t recognize. [UPDATE: I took apart another computer power supply and found it died from a blown fuse. The fuse was very hidden and definitely not designed to be user accessible. Once I found the fuse in that other power supply, I was able to return to this power supply and find its non-user-serviceable fuse as well. This fuse did not blow despite the fire.]
No sign of overheated traces on the circuit board.
This power supply had been running quietly and reliably for years, powering my TrueNAS CORE server. My Kill-A-Watt meter indicated a steady-state power draw of roughly 80-90W, which is a tiny fraction of this power supply’s advertised 650W capacity.
I don’t remember how long I’ve had this power supply, but I am confident it is well out of warranty. The only date stamp I saw was on the back of the cooling fan, with a manufacturing date of December 13th, 2011. This puts an upper bound on the age.
Random side note: I think “Protechnic” is a perfectly reasonable name for an electronics supplier company. Given this episode of combustion, though, I also note it is just one unfortunate letter off from “pyrotechnic“.