Would noise from MLCCs be cancelled out if adjacent capacitors had different pol

[demute]

I have a DC voltage with a high ripple current at 10 kHz which produces an audible noise from the MLCCs on the PCB. I can't change the frequency and I need to stick to MLCC class II capacitors for different reasons.

In this article:
https://www.newark.com/wcsstore/ExtendedSitesCatalogAssetStore/cms/asset/pdf/americas/common/nic-components/MLCC-Ringing-Singing-NSPH-SMT-FilmCapacitors-May2015.pdf
they say MLCCs mounted at the same place (e.g. same xy-coordinates) but on different sides of the PCB tend to cancel each other out. Is the same also true if the polarity of adjacent MLCCs would be reversed?

More specifically, assume I have different layouts where many MLCCs are placed close to each other, all connected to + and - of the DC-voltage:

      Layout 1               Layout 2               Layout 3
[+-] [-+] [+-] [-+]    [+-] [-+] [+-] [-+]    [+-] [+-] [+-] [+-]
[-+] [+-] [-+] [+-]    [+-] [-+] [+-] [-+]    [-+] [-+] [-+] [-+]
[+-] [-+] [+-] [-+]    [+-] [-+] [+-] [-+]    [+-] [+-] [+-] [+-]
[-+] [+-] [-+] [+-]    [+-] [-+] [+-] [-+]    [-+] [-+] [-+] [-+]

Assume 1206 capacitors placed as close to each other as possible, e.g. the wavelength of the 10 kHz is large (34 mm) in comparison to the distance between MLCCs.

Layout 2 is obviously much easier to route compared to layout 1 and 3, but that's not relevant for the discussion here. I'd just like to know, if all else being equal, would any layout above produce less noise over the others due to cancellation av sound waves?

[tom66]

You would have to assume the resonances and the vibration of each capacitor is very similar, which might be optimistic. 

You could consider using some kind of flexible epoxy over the capacitors to damp the noise.

[Kleinstein]

The capacitors itself are rather small and don't couple very much to the air. The main thing to worry about is the coupling of sound to the PCB and only than to the air. Here is may be about getting the capacitors longer / shorter depending on the voltage and this way bending the PCB. Chances are the capacitors are random in orientaion anyway - MLCC usually don't have a defined inner / outer contact or defined orientation. So turning them around would not change anything. The polartiy could also well be set by the DC bias. So chances are the different patterns shown would not have a significant effect.  Polulation on both sides would be more promissing, at least to reduce the more bending modes.

The idea of making the PCB stiffer with additional epoxy is also good - I don't think it has to be extra flexible as even the harder epoxy is still relatively acoustic absorbing. A main part would be from just making it stiffer and thus less amplitude (in µm) from the same force.

[AVGresponding]

Stiffening the board locally with epoxy would also raise the resonant frequency, which might help.

[Bud]

You would have to assume the resonances and the vibration of each capacitor is very similar, which might be optimistic. 

You could consider using some kind of flexible epoxy over the capacitors to damp the noise.

Conformal coating.

Edit: But that will make repairability a nightmare.

[voltsandjolts]

Some ideas maybe:
https://www.electronicdesign.com/technologies/analog/article/21154968/samsung-reduce-the-acoustic-noise-effect-from-class-ii-mlcc-vibration
https://e2e.ti.com/blogs_/b/powerhouse/posts/how-to-reduce-acoustic-noise-of-mlccs-in-power-applications

[demute]

Thanks a lot for all your feedback, especially the links provided by @voltsandjolts . I have now a better understanding of the vibrational modes and I see that reversing the polarity will not help. However, applying less solder paste, having a thicker PCB, making PCB cut outs around the capacitor terminals and placing the capacitors on both sides of the PCB will probably make the noise go away. 

[wraper]

You can buy diffrerent MLCC with required parameters and try which are the most silent. They are not equal at all. Some may be barely audibele, others whining like hell with everything else equal.

[SiliconWizard]

You would have to assume the resonances and the vibration of each capacitor is very similar, which might be optimistic. 

You could consider using some kind of flexible epoxy over the capacitors to damp the noise.

Conformal coating.

Edit: But that will make repairability a nightmare.

Coating helps dampening high-freq vibration for sure but it may not disappear completely. Most of the noise comes from the PCB itself - the cap transmitting vibration to it since it's mechanically coupled.

One common way of handling it otherwise is to put small slots around the capacitor in the PCB. That may not be easy or possible if the PCB is very small/dense.

[demute]

That's interesting. You mean that they can be very different while they still are class II MLCC:s? The capacitors I currently use are 10 uF / 100 V in 1210 package, or 10 uF / 63 V in 1206 package. So given the capacitance, I don't have the luxury to go for C0G capacitors for example.

[wraper]

That's interesting. You mean that they can be very different while they still are class II MLCC:s? The capacitors I currently use are 10 uF / 100 V in 1210 package, or 10 uF / 63 V in 1206 package. So given the capacitance, I don't have the luxury to go for C0G capacitors for example.

Class 2 is a very vague specification with subclasses within it. It does not specify actual materials used. All it specifies is maximum capacitance change within temperature range. But your issue is more related to piezoelectric effect and effective capacitance/voltage dependence which is not specified in this classification. Not to say mechanical resonance plays a role, so it's totally possible that worse capacitors will be quieter at certain frequencies that those which are overall better.

[jongeske]

Hi, this is my first post so if I did this wrong I apologize:
I was struggling with MLCC noise on a charge pump circuit for an audio application.  I got the noise down about 10dB by going to the minimal capacitance value the IC needed.  I had read that the capacity of the cap was a factor in the number of layers and more layers tended to be louder.  The rest of the improvement was then made using new ZRA and ZRB caps from Murata which have an integrated interposer.  With these two changes, all the noise was gone (audible and on the spectrum analyzer).