04-08-2014, 02:11 PM
OK, here's my understanding:
Film resistors are laser-cut into a spiral to achieve the required resistance value. The pitch of the spiral varies with value; low value resistors only require a course spiral, whereas higher values are much fine. If you could "unroll" the carbon film from the ceramic substrate and lay it out flat, you'd see that low value resistors are a large, wide block of carbon, but high value resistors are a long, very thin carbon track...
When a high value resistor has a lot of voltage across it, then between adjacent parts of the film that are separated by the narrow laser-cut section, there will be a voltage differential. This causes erosion in the track - leading to the track becoming thinner wherever this takes place. As the track thins out, the resistance value gradually rises. This leads to greater voltage differentials between "turns", which accelerates the erosion process. Ultimately, after a period of gradually rising resistance, it usually fails open-circuit.
This is the classic failure mode of the startup resistor in many switched-mode power supplies. But I've seen it in countless applications - e.g. video output stages, high voltage linear power supplies, nixie tube drive circuits, CRT bias circuits. As a rule of thumb, I have my "100 rule" that I mentioned above; anything that's 100k or more, with 100V or more across it. But I've seen it with 47k and ~50V in the past.
This "voltage stress" has nothing to do with V-squared-over-R power dissipation - this is a separate mechanism. It can take many years to develop. And most film resistors are rated at 250 volts (within their power envelope)...
Anecdotally, it's possible that metal-film resistors might fare better than carbon film, but that's only because most failures I've seen have been carbon film. Metal glaze resistors are much better at high voltages - e.g. Vishay VR37.
Often, it's cheaper to just use 2 or more resistors in series - each sees less voltage, and will have a proportionally wider carbon track as well.
Film resistors are laser-cut into a spiral to achieve the required resistance value. The pitch of the spiral varies with value; low value resistors only require a course spiral, whereas higher values are much fine. If you could "unroll" the carbon film from the ceramic substrate and lay it out flat, you'd see that low value resistors are a large, wide block of carbon, but high value resistors are a long, very thin carbon track...
When a high value resistor has a lot of voltage across it, then between adjacent parts of the film that are separated by the narrow laser-cut section, there will be a voltage differential. This causes erosion in the track - leading to the track becoming thinner wherever this takes place. As the track thins out, the resistance value gradually rises. This leads to greater voltage differentials between "turns", which accelerates the erosion process. Ultimately, after a period of gradually rising resistance, it usually fails open-circuit.
This is the classic failure mode of the startup resistor in many switched-mode power supplies. But I've seen it in countless applications - e.g. video output stages, high voltage linear power supplies, nixie tube drive circuits, CRT bias circuits. As a rule of thumb, I have my "100 rule" that I mentioned above; anything that's 100k or more, with 100V or more across it. But I've seen it with 47k and ~50V in the past.
This "voltage stress" has nothing to do with V-squared-over-R power dissipation - this is a separate mechanism. It can take many years to develop. And most film resistors are rated at 250 volts (within their power envelope)...
Anecdotally, it's possible that metal-film resistors might fare better than carbon film, but that's only because most failures I've seen have been carbon film. Metal glaze resistors are much better at high voltages - e.g. Vishay VR37.
Often, it's cheaper to just use 2 or more resistors in series - each sees less voltage, and will have a proportionally wider carbon track as well.
