Audible noise from coils or capacitors?

[Timur Born]

Hello everyone!

Which part of a graphic-card's (or mainboard's) power-section (voltage regulartors) does emit audible noise, coils or capacitors? While I know both are capable of producing audible sound I wonder if there is a way to pinpoint down which of the two is responsible by deduction? These things are soldered closely together and the sound is reflected from all kind of surfaces, so ears and microphones can hardly pinpoint the general area, not so much the individual components.

It's noise reaching from hundreds to thousands of Hz, depending on load. With graphic-cards the fundamental frequency matches the frame-rate. So when the GPU displays 8000 frames/s then a 8 kHz tone is emitted from the GPU's power-section, at 200 frames/s you get 200 Hz.

Are both coils and capacitors capable of emitting sound in that range? Which one is more likely?

Thanks and advance! 

[amyk]

It's more likely to be the coils: https://en.wikipedia.org/wiki/Magnetostriction

[wraper]

It's more likely to be the coils: https://en.wikipedia.org/wiki/Magnetostriction

Nope, mostly capacitors. Class 2 and 3 MLCC ceramic capacitors experience piezoelectric effect and emit the sound, especially crappiest types like Z5U, Y5V. You can put a few kHz AC on them and listen how they sing.
http://product.tdk.com/capacitor/mlcc/en/faq/pdf/31_singing_capacitors_piezoelectric_effect.pdf

[Timur Born]

But are the capacitors used on high end graphic-cards and motherboards ceramic ones? One of the loudest graphic-cards I tested yet was an AMD W9100, these babies are expensive stuff.

[T3sl4co1l]

Both are well known sources.

In a PSU, I would suspect coils more than capacitors, because capacitors have little ripple across them, and therefore little structural change, while the inductors carry full switching ripple.  The majority of bulk capacitance in those applications is also aluminum polymer, leaving even less for ceramics to deal with.

Tim

[Timur Born]

I just found an image of the AMD W9100 with the cooler taken off. Can people in the know identify the caps and make some judgment calls based on that?



Noise from the W9100's voltage regulators was louder than on my own Nvidia GTX 780 (Ghz Edition) even when the load and frame-rates were the same. These puppies are sold for around 3500 EUR now and are expected to suffer through long hardships of high load scenarios. I don't know what the TDP numbers for the W9100 are, my GTX 780 caps at 250 Watts.

As far as I understood capacitors can be made less noisy by design. On the other hand I read that capacitors can bend the PCB underneath them and thus create audible noise!? One workaround seems to be to place caps on opposite sides of the PCB, but that surely isn't too viable for PC motherboards and graphic-cards.

Can coils be made less noisy either by design or by implementation? What about higher count phase-designs vs. lower counts? Does it matter as far as noise is concerned, both audible and induced into ground lines?

[wraper]

But are the capacitors used on high end graphic-cards and motherboards ceramic ones? One of the loudest graphic-cards I tested yet was an AMD W9100, these babies are expensive stuff.

There are hundreds-thousands of them. If you looks on the back of the GPU PCB, there is shitload of them. Vast majority of higher end GPUs have ceramic + polymer caps or almost  fully ceramic.

[Timur Born]

Well, yes, but the noise is generally traceable to the voltage regulator sections. So instead of all the little brown caps spread all around the PCB I would expect the big ones or the coils to be the source?! Obviously I don't really know, I just know the general areas.

It's even harder to identify these things on motherboards while loading the internal GPU of a Haswell CPU. Noise is really low volume then. I'm not even sure how to quantify in measurements what my ears take up by sticking my head right on top of it.