Reliability of Ceramic Capacitors

[electricuwe]

Ceramic Capacitors with X7R or Z5U ceramic are widely used for decoupling the supply of integrated circuits. Does anyone have data on failure mechanism and failure rate of such capacitors.

 

[uWaveFDL]

ATC has a large selection of data and app. notes check them out. I'm sure they have the info your looking for.

http://www.atceramics.com/

 

[electricuwe]

To uWaveFDL,

thank you very much for providing the link to atc. I searched the website extensivly but I didn't find approbiate information. My application is not microwave or RF but just some simple comparators and buffers operating at moderate frequency (a few kHz)supplied from 30V DC.

 

[uWaveFDL]

OK, Try this article from EDN. It may contain better info.
Also, there is a large list of mfgr's at the bottom of the article. I would guess several would have the specific info you are looking for.

http://archives.e-insite.net/archives/ednmag/reg/1995/010595/01df1.htm

 

[electricuwe]

Thanks,
the EDN article seems to cover my problem better. I have already visited the websites of AVX and Kemet, which contain a lot of information on the different topics concerning ceramic capacitors. But according to the information there my capacitors should last at least a thousand years.

 

[Thester]

Failure mechanisms of ceramic chip capacitors based on manufacturing defects include voids in the dielectric, misalignment of the electrodes and dielectric, among other things. You can eliminate a lot of manufacturing defects by applying 2.5X the rated voltage at 125 deg. C for a few hours, as any weak parts will just explode! This should only be done in controlled conditions, like a metal walled heating chamber with insulated test fixtures.

Most problems are application related. Ceramic chip capacitors will crack if the printed circhuit card they are soldered to is flexed. Avoid using thin pcbs, add stiffeners like bus bars or heat sinks, keep the perimeter of the card small. You can guess that you might be cracking the chip capacitors if they are failing with resistive short circuits. It might help to use smaller sizes, as that decreases the dimensions that get flexed and also the stress.

Try using parts with voltage ratings greater than you really need. Sometimes you will have high voltage spikes in inductive circuits that exceed the dielectric strength of the capacitor.

 

[buzzp]

I have found that most caps are rated in peak voltages too, rather than rms. Big difference.

 

[electricuwe]

I'd like to add some additional information:

We use small leaded capacitors. So the mechanical problems described by thester shouldn't be the problem.

The rating of the capacitors is 50 V DC, so operation on 30 V DC shouldn't be a problem

The capacitors tend to fail to short circuit

 

[yesway]

Basically the failure rate is, as stated on manufacturers sites, very low. Much less than 1 in a thousand should show failure during say a few years, if operated at suitable conditions.
Many things may shorten life, primarly temperature. Increasing the temp 10'C can decrease life time by 50%. Operating 50V caps at 30V is just reasonable, so problems could still occur. How much ripple do you have? Is there lots of noise, or high energy spikes? I am a little surprised if you use Z5U, that indicates quite high values of capacitance, above 470n? Perhaps the cap cant handle the surge that arise at power up? Maybe you could split the caps to a tantalum + a 1-10n ceramic? Also Z5U drifts wildly with temperature, could be part of the problem though not likely.
During soldering, the caps are more vulnerable to mechanical stress, you should check the mounting + soldering process. E.g are the holes the correct pitch/size, so the component is easily mounted?

 

[electricuwe]

Yesway,

The capacitors we use are 100nF, Z5U ceramic, they are paralleld with 10uF Al-electrolytics and the are used to decouple the supply of a line driver with a BUF634,driving up to 20m of a shielded cable. The system usually is powered up only once and operates continously so this shouldn't be a problem.

 

[yesway]

Okay, then it sounds strange. There must be something really wrong to break the caps in such a configuration. Can you test if the same results occur if there is no cable attached/no signal driven? Sounds quite likely that the cable loads the driver in such a way that the cap sees lots of HF energy.