22
\$\begingroup\$

How do components fail?

General rules with an answer per component type would be valuable.

We can work as a community to build up a single question that holds valuable information about how components fail.

\$\endgroup\$
  • 5
    \$\begingroup\$ Place the answers you have already made in an answer, not in the question. \$\endgroup\$ – Kortuk Dec 21 '10 at 2:40
  • \$\begingroup\$ Good question! Will have to talk about s/c protection and the like. \$\endgroup\$ – tyblu Dec 21 '10 at 12:27
  • \$\begingroup\$ @tyblu, lets keep this one about how components fail, make it one page you can go to to keep track of how they fail. \$\endgroup\$ – Kortuk Dec 21 '10 at 16:53
  • \$\begingroup\$ They do fail often. \$\endgroup\$ – kinokijuf Feb 7 '12 at 18:05
14
\$\begingroup\$

Switches and pushbuttons: failure to make contact.

What you've listed looks like the severity part of an FMEA (Failure Mode and Effect Analysis), at least at component level. While it's not impossible, it's a hell of a job to account for every possible component failure if your design has, say, over a hundred components. One failing component may cause an avalanche of other components failing. Most failures aren't subtle.
You'll experience that adding components to cope with other components failing only adds complexity; you'll have to do an FMEA for these components as well!

An alternative approach, FMEA-wise, may be to start from occurrences. What's the MTTF (Mean Time To Failure)? Most components are quite robust; tens of thousands of POH (power-on hours) are feasable. (A notable weaker component is the Al elco, but even there are solutions). Anyway, an IC usually doesn't short just like that. So, while component failure may be caused by aging, most failures are caused by external factors, like overvoltage on the grid, or user error like misconnecting. Try to reduce these risks. Power spikes may be handled by overvoltage protection diodes. Misconnection can be avoided by using different connectors so that they can't be switched. Color code wires and use matching colors on connectors.

Bottom line: it may be more important to know why components fail than how they do.

\$\endgroup\$
9
\$\begingroup\$

PCBs: cracks in vias

The story:
my brother had one of Philips' first CD players. One time it stopped working, but when I looked into it it worked again. This happened a few times. Trying to find out about the circumstances when it happened my brother said that the last time there was a thunderstorm. A lightning strike may do bad things to electronics, though in those cases the device doesn't start working again all by itself.
One day I was discussing the problem with a colleague when the conversation was overheard by a product manager (I was working for Philips Audio at the time). PM said that only after much searching they had found the cause of this problem: the PCB was made from some cheap material (I don't recall which, it may have been FR-2) which tended to expand when there was much moisture in the air, like during a thunderstorm. As a consequence the few vias on the board would crack open. When the air became drier again the PCB's thickness returned to normal, restoring the vias. That was one reason why I couldn't find anything. Another was that touching the PCB with a multimeter's probe caused enough pressure to close the cracks (these are microcracks!).
The remedy: soldering a wire in each via. Design solution: use FR-4 for PCBs.

Like I already said in my other answer it's important to know why the vias crack; it's no good just knowing how they do.

\$\endgroup\$
  • 2
    \$\begingroup\$ Apparently someone found it a good idea to make all answers CW. While I agree that the list provided by the OP should indeed be CW, other answers are too specific for that. It could be that the question was already CW when I posted this answer, I don't recall. \$\endgroup\$ – stevenvh Dec 22 '10 at 11:42
6
\$\begingroup\$

MOSFETs: Short circuit usually (with a bang), eventually leading to open failure due to melting of device

Resistors: Almost always open circuit

Capacitors (Electrolytic): Reduction in capacitance, leakage of electrolyte, eventually leading to open circuit

Capacitors (Ceramic): Reduction in capacitance - eventually failing open, though severe over-voltage can lead to failing closed (Citation needed).

LEDs: Gradual dimming then failing open

Zeners: Fails shorted in 90% of cases but can fail open due to extreme overheating (device can split into two pieces).
Sometimes Zener become little resistive in the reverse region. When this happens some current flow before the zener voltage.

\$\endgroup\$
  • \$\begingroup\$ I've killed plenty of ICs and I've never personally had one melt or burn. Typically because my power supply has some sort of current limit in it or other feature to prevent fire. \$\endgroup\$ – Nick T Dec 21 '10 at 14:59
  • \$\begingroup\$ @Nick T, Probably applies for most circuits, but LiPos and computer power supplies can pack a punch. \$\endgroup\$ – Thomas O Dec 21 '10 at 16:49
  • \$\begingroup\$ lets make this multiple answers, one per components type, I think IC may be a bit too broad. \$\endgroup\$ – Kortuk Dec 21 '10 at 16:49
  • 1
    \$\begingroup\$ @Kortuk, Why? It would just create clutter. \$\endgroup\$ – Thomas O Dec 21 '10 at 16:53
  • 2
    \$\begingroup\$ this is a short list, no citations, and no real detail. If you have MOSFET on one, Capacitors(Electrolytic) on one and so forth real detail can be added, citations can be added, and we can have a very nice question that can be used as a general reference by others, as it is you are just putting in your opinion. \$\endgroup\$ – Kortuk Dec 21 '10 at 16:54
5
\$\begingroup\$

electr.CAP - short is possible due to deformation => explodes.

ICs: internal wires fail open, internal safety diodes shorts, gate latchups(might be not fatal), degraded performance due to semiconductor degradation (when working at >100C), soft errors due to radiation. Power ICs might explode (I got hit by one) when failing under load.

\$\endgroup\$
4
\$\begingroup\$

Resistors

Failure Modes

Resistor failures are considered to be electrical opens, shorts or a radical variation from the resistor specifications. The failure modes experienced vary with the type of construction. A fixed composition resistor normally fails in an open configuration when overheated or overly stressed due to shock or vibration.

Excessive humidity may cause an increase in resistance. A variable composition resistor may wear after extensive use, and worn away particles may cause high resistance short circuits. Wirewound resistors may experience open windings due to overheating or stress, or short circuited windings due to accumulation of dirt, dust, breakdown of the insulation coating or high humidity. Film resistors fail for the same reasons as wirewound and composition, but have also failed due to changes in resistive material characteristics resulting in reduction and increase in resistance value.

Electronic Components - Resistors. (1978). FDA Inspection Technical Guides. Retrieved from http://www.fda.gov/iceci/inspections/inspectionguides/inspectiontechnicalguides/ucm072904.htm

\$\endgroup\$
2
\$\begingroup\$

Electronic system reliability is an ugly problem, but you can get a idea of how it's done in the aerospace business by reading MIL-HDBK-217. Mil standards can be found at The DOD Website ASSIST. The Wikipedia entry: Reliability Engineering has a good overview.

\$\endgroup\$
2
\$\begingroup\$

Ceramics can also fail short circuit, which can be exciting if they are decoupling a high-current supply...

Kemet's capacitor failures page

Syfer's appnote on capacitor cracking

AVX's appnote on cracking

\$\endgroup\$
  • \$\begingroup\$ edit this into an answer about ceramics. Make a nice solid answer about ceramics. Find references if you can! \$\endgroup\$ – Kortuk Dec 21 '10 at 16:49
1
\$\begingroup\$

TVS: Fails shorted in 90% of cases but can fail open due to extreme overheating (device can split into two pieces)

\$\endgroup\$
  • 1
    \$\begingroup\$ Diodes in general, tend to fail short. \$\endgroup\$ – Robert Endl May 31 '16 at 16:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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