13
\$\begingroup\$

What happens when a ceramic bypass capacitor fails? I am picking out parts and is wondering is it worth the extra few cents for rated capacitors?

Since generally most IC can function without a bypass, if a bypass cap fails there should be no problem, right? Or when a ceramic bypass fails does it form a short circuit?

\$\endgroup\$

2 Answers 2

22
\$\begingroup\$

Surface Mount Multi-Layer Ceramic Capacitors (MLCC) primarily fail in the cracking regime.

That is the ultimate failure mode of a capacitor is with a body crack.

MLCC are made of very fine intermeshing metallic layers embedded in a ceramic substrate.

A large portion of these failures result in short circuits (low impedance path) through the part. The severity of this outcome in a decoupling situation cannot be understated. The ceramic can take a lot of power and I have seen even 0402 shorts take out entire PCBs sustaining multiple amps through them.

Occasionally the component fails in such way that a short is not produced and simply the capacitance is dramatically reduced. This is not the usual failure mode and is largely a matter of luck and failure trigger (sudden crack will fail open more often than a gradually propagating crack).

The variables that contribute to component failure have dozens of possible sources, vibration, electrical stress (ripple), dielectric type, temperature, PCB placement, reflow temperature, depaneling process component size (plate density)... Component quality also plays a role because substrate voids and hot spots can concentrate stresses.

In principle the component should last a very long time in a stress-minimized environment with good PCB and electrical design, and high quality MFG.

NP0/C0G parts are particularly suitable for reclaiming due to their much better dielectric (no piezo effect) and temperature (no capacitance change) characteristic.

Ripple and overvoltage are issues, but MLCC are typically rated for nearly 100% overvoltage and significant ripple. The concerns there are largely mechanical due to piezo effect.

Failure of one capacitor in a decoupling bank may further degrade others and cause cascade failures.

\$\endgroup\$
9
  • 2
    \$\begingroup\$ Keep in mind that after desoldering it may take up to 24hours for the capacitance to stabilize, typical for X7R to a value with some tolerance from where it was . this can make verifying parts to be hard. The qualification process (official test) usually has a wait period before reverifying capacitance after soldering or overvoltage excursion (~200% of rated voltage). \$\endgroup\$
    – crasic
    Nov 12, 2019 at 20:52
  • \$\begingroup\$ my pcb has a bit of modularity to it, having read what you have said i am thinking of adding a ressetable fuse on each module in the cenario of a bad cap failing i can easily identify where the failure is \$\endgroup\$
    – Jake quin
    Nov 13, 2019 at 22:55
  • 1
    \$\begingroup\$ I would opt for a fast blow smt fuse A PTC is generally not fast enough, a 2A rated poly fuse will trip in 1 minute at 2A and 10s at 4A , not fast enough to protect other parts from a hard short. A fast blow of appropriate rating will blow in under a second . \$\endgroup\$
    – crasic
    Nov 13, 2019 at 23:10
  • 1
    \$\begingroup\$ If these are modules, having current limiting ("Crowbar circuit") or active current monitoring with cutout on the supply feeding the modules would be a clean, professional solution . A fuse could still be there sized to blow in the absolute worst case. I have found through pain that a fuse is a good idea and enables better failure investigation compared to when you get a crater and burned traces. Also you can use a Fuse in R&D/Testing and then populate with jumper for production if you worry about false trips. \$\endgroup\$
    – crasic
    Nov 15, 2019 at 17:42
  • 1
    \$\begingroup\$ Extra few cents make a huge difference between a cheap fake part from a "seller" or from a no-name manufacturer, and a very high quality part from a Name-Brand manufacturer. Many cheap fake parts are removed from defective old equipment. Buy cheap then get cheap. \$\endgroup\$
    – Audioguru
    May 17 at 18:17
20
\$\begingroup\$

Ceramic capacitors can fail in a couple of ways.

They can be mechanically damaged - too much physical stress (pressure on the part or the board is bent a little too much) can cause a crack. The capacitor will then develop short circuits between layers. It acts more like a resistor in that case. If you are really unlucky, enough layers will short that the whole thing acts like a short circuit. Depending on how much current the power supply can deliver, it will either go "bang" or just sit there and overheat.

The other way they can fail is through over voltage. Too much voltage can cause the dielectric (insulating) layers to break through and become short circuits. Too much voltage can also cause the capacitor to flex, resulting in a stress crack. Then you get the same result as above. The capacitor becomes more like a resistor, or shorts out entirely - bang or burn.

Neither failure mode is usually a problem. Stay under the rated voltage, don't bend your PCB or hammer on the capacitors and you should be fine.

\$\endgroup\$
4
  • \$\begingroup\$ if we say that capacitor is not subjected to over voltage or mechanical stress, and it went bad how would that fail ? is it more like the first one you mentioned where it becomes like a resistor and just eat power? \$\endgroup\$
    – Jake quin
    Nov 12, 2019 at 20:23
  • 3
    \$\begingroup\$ I highly disagree, every MLCC MFG out there has a 50 page document on MLCC crack failures because they tend to be catastrophic. Bulk TH ceramic caps are another question but MLCC failures are a major quality challenge. Largely tend to fail short due to high density of plates. \$\endgroup\$
    – crasic
    Nov 12, 2019 at 20:43
  • 2
    \$\begingroup\$ Reflow and PCB depaneling stress can cause early failure of ceramic caps outside of dropping, hammering, or other physical damage. The physical stress required can be very subtle especially for the smallest parts. \$\endgroup\$
    – crasic
    Nov 12, 2019 at 20:46
  • 3
    \$\begingroup\$ I've seen several supply MLCCs failing in short. On a big board, it can be really challenging to find the failed component. My technique of choice: feed the supply net with an external power supply with proper current limit to avoid frying traces and then hunt for the bad capacitor with a thermal camera. If you're lucky that the overheating was not enough to damage the PCB substrate you may be able to recover the board. \$\endgroup\$
    – joribama
    Nov 13, 2019 at 8:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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