4
\$\begingroup\$

In comparing ceramics vs tantalums one of the oft cited disadvantages is the tendancy for ceramics to exhibit a piezoelectric (i.e. microphonic) effect. Old-school technicians talk about ceramic disk capacitors "singing along to the music" in audio circuits.

I am having great difficulties trying to quantify this effect. The closest I have is a Kemet paper.

So, is this effect still significant in modern X75/X5R capacitors (for example)? Does it vary by package size and manufacturer? What frequency range will cause problems? What sort of noise (microvolts? millivolts?) can be expected to be generated and at what frequencies?

Most papers discuss audio circuits, but primarily I am interest in applications with vibrations under 200Hz (sometimes much less).

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

The answer finally comes as two blog posts from the TI Precision Hub blog.

It isn't a laboratory-grade test, but the post looks at the effect of dropping a known mass from a known height on a PCB with differing capacitors. Accordingly, it's dealing with a "unit impulse" test rather than low frequency vibration, but the results are nonetheless interesting.

A standard X7R capacitor caused a spike of 120mV on the output. Owch! Special "soft-end" ceramics exhibited about 80mV, whereas tantalum and film capacitors caused no measurable effect. Missing from the test were C0Gs, which I would have really like to seen.

This adds to documents from AVX (and more) as well as more qualitative documents from ADI, AT Ceramics as well the aforementioned Kemet paper.

This isn't quite over the frequency spectrum I was interested in, but I think it's as close as I am going to get without attempting an experiment myself. If someone has this information, I'll un-accept this answer and happily give them my vote.

\$\endgroup\$
3
\$\begingroup\$

In a practical sense modern ceramic caps can certainly be a source of audible noise or mechanical vibration. This generally becomes noticeable when they are used in power supply circuits and the load transients are in the high audio range.

Mounting the cap to a PCB amplifies the sound as the dimensional changes in the cap couple to the PCB as a small warp causing it to act as a speaker. The physical size of the capacitor is definitely a factor, as is the dielectric.

To mitigate this, if possible ceramic caps should be placed in pairs on opposite sides of the PCB in the same exact X-Y location. (I.e. if you need 200uF use 2 100uF caps in parallel, one on the top side and one on the bottom.)

If the caps are just for decoupling on DC lines or there is no large ripple in the audio band there is usually no issue.

Some manufacturers claim to have ceramic cap products that mitigate this, either by mounting the cap inside a case with poor acoustic coupling or by choice of dielectric. I've tested some of those and was not impressed with the results, but they may have improved over the last several years.

\$\endgroup\$
1
  • \$\begingroup\$ I've just asked a shop who sold me a laptop charger to replace it because it was doing an irritating noise, which I believed to be caused by a capacitor charging and discharging. They replaced it by one exactly equal, but the new one doesn't make that annoying sound. \$\endgroup\$ – sergiol Feb 28 '15 at 12:10
0
\$\begingroup\$

(This started as a comment but grew too long.) Nice Kemet paper. Thanks. (Figure 1 is interesting a "Moore's law" in ceramic cap size.) I don't know about the piezo behavior of ceramics. I would expect that size change of the ceramic would be independent of frequency for these low frequencies. (You squeeze it and get a voltage.) The comparison of 100Hz and 1k Hz in the paper seems to show this. As for quantifying, that sounds hard since a lot will depend on the layout and the acoustic coupling between the cap and the pcb. And then the coupling of pcb to outside world.

For the size effects one might guess that the large dielectric materials would show a larger piezo effect. (I have no idea if that is true.) But then they will be smaller is size and so the deformation caused by motion of the pcb will be less. It's hard to know which would be the dominant effect.

\$\endgroup\$
5
  • \$\begingroup\$ Piezo can get pretty attenuated at low frequency. How low would depend on the input impedance of the upstream amplifiers, and to some extent, the nature of the crystal itself. Essentially, a piezo crystal is like a cap, with charged plates separated by an insulating crystal, but the crystal has a finite resistance, as does the upstream amp, so the charge leaks away. \$\endgroup\$ – Scott Seidman Sep 12 '14 at 13:10
  • \$\begingroup\$ Yeah that makes sense, @ScottSeidman. I'd model the piezo as a voltage source in parallel with the cap. And sure if there is some leakage current it will be less.. but that would be a fairly low frequency.. a long time. \$\endgroup\$ – George Herold Sep 12 '14 at 13:17
  • \$\begingroup\$ en.wikipedia.org/wiki/Piezoelectric_sensor -- lots of great models. In fact, at high frequency, you have to model the inertia of the crystal as an inductance to show proper resonance. \$\endgroup\$ – Scott Seidman Sep 12 '14 at 13:27
  • \$\begingroup\$ @ScottSeidman, Re: models. Ah right. (looking at wiki) there's a series capacitor between the voltage source and bulk capacitance. For the (one) peizo transducer that I've used the low frequency time constant was about 1 second or so. (It's been years since I looked at it.) \$\endgroup\$ – George Herold Sep 12 '14 at 13:52
  • \$\begingroup\$ Yup, the series cap Ce is a term corresponding to the elasticity of the crystal. These are pretty interesting models, with some of the electrical elements being electrical elements, and some being electrical analogues of mechanical elements. \$\endgroup\$ – Scott Seidman Sep 12 '14 at 14:00

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.