Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

[SOLVED] MLCC piezoelectric effect (high-pitched noise) elimination techniques

Status
Not open for further replies.

d123

Advanced Member level 5
Joined
Jun 7, 2015
Messages
2,505
Helped
494
Reputation
992
Reaction score
525
Trophy points
1,393
Location
Spain
Activity points
27,148
Hi,

Success! :ROFLMAO:

I'm really pleased to say that yesterday I had my first experience of hearing some X7R MLCC 1206 capacitors making that irritating, high-pitched whine/squealing noise ('singing' is not the adjective I'd use for that sound) on a very simple 10 kHz, 10% duty cycle LED driver circuit I had just finished making for yet another torch variant (bicycle light). ...To experience it is more meaningful for me than often reading about - and therefore, abstractly understanding - it, in this case.

I solved it by reluctantly changing Fosc to ~20 kHz, the piezoelectric effect is no longer audible, to human me, at least.

On looking for solutions I read these articles:

Reduce acoustic noise from capacitors

How to reduce acoustic noise of MLCCs in power applications

Capacitors piezoelectric effect

I see that choices are:
1) Make Fosc > 20 kHz
2) Use two capacitors that are half the value of required capacitance value on either side of the board (assuming you have a double-sided PCB) to cancel out each other's flexing, instead of one on one side.
3) Make cut-outs at the ends of each capacitor.
4) Use one of the three types of 'anti-noise' capacitors available. (...$$$! :unsure:)
5) Avoid using X7R at frequencies in the audible range.

Has anyone tried #3? Is it effective?
Are there any other methods not mentioned in the articles?

I know it's an ignorant question, but all the same, I'd like to know why, if compatible with the design, using either:
a) an RC, or preferably just a C, that is the same frequency as Fosc between Fosc source and offending MLCCs
OR
b) A large value capacitor between Fosc source and the singing capacitors to swamp out the ripple voltage
are evidently not methods that could work?

The frequency makes the capacitor plates vibrate, so why wouldn't swamping with a large value capacitor (relative to Fosc) act like a low pass filter and stop Fosc disturbing the MLCCs?

Is it the board/signal path, or only the MLCC capacitor that causes the vibration?

What I read all said that the capacitor vibrates, so then the PCB does too and acts as a loudspeaker for the capacitor's piezoelectric effect.

Thanks.
 

Not all capacitor families use piezoelectric dielectric materials. Unfortunately the PZT & similar are some of the highest density of C*V.
 
Hi,

Thanks. If I had 1% of your ability, talent, experience and knowledge, that would be a huge improvement. I looked up PZT yesterday. None of the stuff directly said it was used in X7R, so I'm a bit lost as to whether that's what you meant - 'unfortunately' makes me guess so. In another thread here, Klaus mentions using NP0 to the OP because they are not ferro-electric and so don't squeal.

--------------------------

Apart, just wanted to add to the list in post #1 that option 6) I read about is putting capacitors close to edge of board to 'reduce noise', but my noisy ones are as close to the edge as possible, so not so sure of that approach as much use, which is why option 3) interested me with regard to effectiveness.
 

You have not stated the voltage on the caps, current supplied by the caps, the rise time of the current in the caps, the data sheet for the caps - the pcb layout, how many caps - or any other highly relevant information to your predicament.

There are very easy ways to overcome this at 10kHz - but condition dependant - we don't know the conditions.
 

Hi,

Not sure if you're serious or trolling. I'm a hobbyist who knows they are scorned on this forum, hope you're not wasting my time trying to ridicule me but are genuinely interested in this topic. I'll send you the PCB and a detailed thesis, if you wish, and you need it all to provide a general discussion :confused: on effective methods, especially 3) that I suspect is as probably as 'effective' as 6).

This is not a predicament, the circuit functions as desired and shift up to 19.62 kHz has silenced the caps, I was just glad to finally hear the noise as I've never had a crummy, home-made 1oz/35um FR4 copper-clad (looks like cheap copper foil layer, not the nice, shiny copper-pour types) with those capacitors do that, presumably due to unwise combination of frequency (10 kHz) and the X7Rs.

Without getting the packet (sorry), I think they're Panasonic GRM something-or-other 1206 X7R, average quality type, 21.1uf (10uF, 10uF, 1uF, and 100nF). It's a crummy prototype 3.2V to 6.6V AA battery-powered circuit, current is about 180mA into six white LEDs, but duty cycle is only 10%. If I'd known you'd ask this, when it was still 10 kHz I'd have taken photos of the 180mV ripple RC curve rise and steep fall, and taken a note of the rise time (I had a good measure with DMM on DC to gauge the average currents and voltages and poked around with o'scope, and even made some notes of voltages and time-related aspects), like a bad sawtooth I saw on the oscilloscope before and after the caps.

From memory, Vcaps was (when batteries were totally new) 6.16V +- ripple.

Mosfet output took approx. 1us rise and approx 2us fall with an approx 8us ON time.

Here's a photo of the PCB. DSC_0473.JPGDSC_0474.JPG
DSC_0476.JPG
DSC_0479.JPG

First thing I've done in months, so skills are rusty.

That's a lot of info. from my side, some that seems embarrassing and irrelevant to the question even, so now please tell me the secret of combining audible frequency range and unsuitable capacitor types and them not whining horribly because 10 kHz is a useful frequency. Thanks.
 

Attachments

  • DSC_0476.JPG
    DSC_0476.JPG
    1.2 MB · Views: 118
Last edited by a moderator:

Hi,

The MLCCs are Murata:
100nF X7R 1206 overview on Murata website

1uF X7R 1206 overview on Murata website

10uF X7R 1206 overview on Murata website

The reason for the paralleled 220R resistors is that a packet of 45 (quantity) 100R 1206, 100V, 0.25W thin film resistors has literally evaporated from inside a box that was inside another box, go figure how that is at all possible, supposedly only I had access to my belongings in the houses I've rented in recent years... :mad:. I live alone. Not a nice surprise to discover the other day.

As you can see from the links, the capacitors are general purpose.
 

Ho hum... And so much for that special secret solution. Ho hum...

Point of thread was to collate some information on the topic here, so for anyone interested in the topic and looking for a quick round-up of the causes and the/some possible solutions:

These articles discuss solutions:

Reduce acoustic noise from capacitors

How to reduce acoustic noise of MLCCs in power applications

Capacitors piezoelectric effect

A summary of the choices are:
1) Make Fosc > 20 kHz
2) Use two capacitors that are half the value of required capacitance value on either side of the board (assuming you have a double-sided PCB) to cancel out each other's flexing, instead of one on one side.
3) Make cut-outs at the ends of each capacitor.
4) Use one of the three types of 'anti-noise' capacitors available. (...$$$! :unsure:)
5) Avoid using X7R at frequencies in the audible range.
6) Place capacitors close to edge of board to 'reduce noise'.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top