| Electronics > Projects, Designs, and Technical Stuff |
| high energy sustained medium impedance faults? |
| (1/2) > >> |
| coppercone2:
On a PCB with convenctional parts other then large mains transformers, is it common, likely or even possible to have a sustained medium impedance fault on a high power rail like a ATX supply? 1) Electromechancial devices like pumps, fans, motors, solenoids, speakers, transformers (ok it seems clearly possible with mains transformers but how about like a DC DC converter transformer/inductor or ground referenced inductor filter. 2) Circuit elements like transistors, IC's, capacitors etc. I mean a high energy fault where something is going to smoke and glow red hot or whatever for a long period of time before it melts open rather then tripping a fuse or unintelligent current sensor. It would mean something fails and settles to like ~ 1 ohms rather then going dead short and can survive 400W of power going through it for a significant period of time instead of vaporizing in a explosion. Large transistor like IGBT maybe? Big diodes? processor core? I never really saw this happen that I can recall. I did see small signal stuff like this fail but the supply was really current limited and I don't think it would survive even a few amps for any period of time. |
| Siwastaja:
It indeed can, and these faults are really pain in the ass, dangerous, problematic and often overlooked. It's hard to protect against if the fault current is around the expected operating currents - but now all the power is concentrated in the failed spot unlike in proper operation. I have personally had a 1210 ceramic cap fail nearly-short after about 1000 hours of runtime, due to cracking thanks to improper soldering (no mechanical stress due to mounting), dissipating around tens of watts, glowing bright yellow for a fairly long time (>10 seconds) - I don't know how long it would have been if I wasn't around to turn the thing off. Fusing was properly implemented, but it couldn't possibly help because it wasn't a short, but around the expected operating current (or a bit less). If there was combustible material near said capacitor, it would have started a fire. No fuse, or a thermal fuse, could have done anything. This is part of the reason PCB materials have fire safety ratings, and plastics used in electronics have fire retardants. Yet, even properly rated PCB materials start to burn, given enough temperature. Sometimes, components desolder themselves, stopping the current, but they may as well stay in place. Fuses are a great, easy protection against most failures causing fire hazard, but not all - the rest require much more system-level engineering. |
| coppercone2:
at what voltage did you see that happen? 10's of watts is much different then 100's. Do you think its still possible for this to happen with +100W? i take it you had like 8A of current going through that part? I am used to things blowing fuses or exploding. A serious measure might be to light seal the assembly and put shut down photo-detectors in it with bypass pins for maintenance reasons. |
| Siwastaja:
--- Quote from: coppercone2 on December 20, 2018, 10:35:46 am ---at what voltage did you see that happen? 10's of watts is much different then 100's. Do you think its still possible for this to happen with +100W? i take it you had like 8A of current going through that part? I am used to things blowing fuses or exploding. A serious measure might be to light seal the assembly and put shut down photo-detectors in it with bypass pins for maintenance reasons. --- End quote --- This was at 10V, the caps were rated at 50V (X7R). Fuse was 35A, which is a tight fit for the typical running current between 30-35A. Of course something similar can happen at any power level. Cracking is a typical failure mode in MLCCs; other components have their own failure modes. |
| coppercone2:
can this happen to a crimp too? or are they designed against this mode? What happens with those capacitors that makes em glow, do you get like a carbon resistor forming with the dielectrics burning? But what is burning in a ceramic cap? What is the nature of these medium impedance failure modes? (rather then just shorting out to miliohms). Can this happen with aluminum oxide if you get the right thickness with an aluminum on aluminum connection (the oxide when very thin is partially conductive)? can you get stuff like small contact area metals melting and you have a molten pool surrounded by insulating oxide acting as insulator/pressure vessel and the molten pool acting as a conductor with surface contact? I have a hard time imagining the thermal profile of such a conductive junction where it can conduct, be so hot, but not vaporize. What does the topology look like? Are there any detailed studies? maybe some oxide/metal sponge/composite/mixture forms? maybe with the capacitor you just get metal expansion along the fracture so that there is partial good metal on metal contact with a particular junction area? i imagine in most cases it would look like a wedge. |
| Navigation |
| Message Index |
| Next page |