There are some building management product PCBs that look interesting.
first up, they are certified products, and from an install in a ceiling of a commercial building (in a bathroom) . There are 2 units from one bathroom that have both failed this way, but other units in the building haven't. they are powering LED lighting ballasts (if that's the correct term?) looks like about 300mA is drawn from each device when on and the devices are nominally rated for 800mA continuous load.
Secondly the mains circuit here is a SSR mains switch... not much to it really, but there's an SSR switching active that has a MOV to neutral, and a snubber (50R, then 33nF mains rated film cap) also to neutral.. switched mains also has an NTC in series to the output of the device, to limit switch on current. the NTC is right up against the snubber cap.
the PCB is a 4 layer job with what seems to be prepreg/core/prepreg stackup. measures as a nominal 1.6mm thickness spec.
obviously inner layers all clear of metal around the mains circuit and mains only on top/bottom
The PCB has two mains areas.. the other mains area is for mains input, where there is a standard mains to DC encapsulated module for the control circuitry. this area seems to have no burning or anything else. the input filtering cap for the mains to dc module is the same manufacturer and model and voltage (but different size) as the snubber cap.
Only the output area has this failure in 2 units I've seen.
As for the damage -
The PCBs have been pretty much burned through, like I've seen underneath components like MOVs that had obviously died in the past... but the burn through is not under any parts.. it's NEXT TO a part (a mains rated cap that's in there as a load snubber alongside a 50R resistor)
The mains rated cap is in bad shape - in both cases split open at the bottom and swollen and the cover is melted and generally looking pretty badly wrecked (though on the board with the less split open cap and less PCB burned, this cap still maintains its 33nF value.. on the worse of the 2 boards it's down towards 0nF)
The SMT resistor right next to it which makes up the rest of the snubber is fine.. the board damage has delaminated its pads and tracks.. but the chip resistors themselves from both units measure 50R..
the NTC is completely fine. looks like nothing went wrong at all. not even any burn residue on it.
With the burn area, you have exposed fibreglass weave on each side that looks almost silver, and if you push into the weave with a ballpoint pen, the inner layer is all crumbly and black - it breaks apart and falls out of the weave in small chunks and dust.
both boards are burned their worst just near the node where the 50R, 33nF join. they both have an actual small penatrating hole.
Anyway - the best understanding I can manage right now, comes from the cap being the most heavily damaged part on the board, and less damaged than things that were actually on and next to the burning PCB.... so I'm expecting that it's the snubber cap failing for some reason and ejecting burning debris onto the board which is burning the board initially and causing a bunch of conductive residue to coat the board.. this is then conducting enough between active and neutral to heat up and slowly/progressively carbonise the PCB all around that area?
of course this relies on the cap doing something I've never seen a film cap do before, and something I'd really not expect from a mains filtering cap.. also it leaves me with no cause for this 400V rated part to fail like that in the first place... I don't really expect this circuit will be creating anything particularly interesting across the snubber cap when turning on and off a LED ballast through an NTC.
Mar 17.2020, 10:38:59
Do you have a good (non-burned) specimen?
We guess would be: Somehow an arc has formed between two traces through some series impedance (otherwise the upstream fuse would have been blown out) and was sustained long enough to burn through the board. You'd need quite a lot of heat and time to burn through a four layer board. We don't thing a capacitor spitting fire was the root cause (maybe except for the well known RIFA bombs).