Localizing acoustic noise source on PCB

[mtwieg]

I've got a large PCB with several DC-DC converters with a pulsed load, and some boards are making an acoustic clicking sound during large pulses. Nothing spectacular, but localizing which component is the source is proving difficult because there are about a dozen inductors passing the pulsed current, all placed in a dense group. Component size ranges from 6x6mm to 20x20mm. I'm pretty certain it must be an inductor, since only current is pulsed, not voltage.

Often I try to identify culprits by placing a finger on each component and listening. Often the culprit will sound different when its pushed on. But in this case that's not working, perhaps because it's more than one component.

Anybody know any better tricks for localizing acoustic noise from inductors? In my head I'm imagining a magic "listening stick" like this, only much smaller:


It also occurred to me to wire up a tiny mems microphone chip and use it like a pickup loop to sniff out the source. But I assume that direct coupling of pulsed magnetic fields from the power supply would confound the measurement. Need to keep the actual electronics some distance away.

 

[schmitt trigger]

I once used a miniature MEMS microphone to locate a switch's scratching sound. The technique works.

But you may be right, the electromagnetic pulse may couple into the microphone's signal lines, confusing the reading.

Could you 3D-print something similar to what you show on the photo, only shorter?

 

[KlausST]

Hi,

several times I used a small flexible plastics tube.
One end close to the ear, the other end for scanning the PCB.


Klaus

 

[mtwieg]

I once used a miniature MEMS microphone to locate a switch's scratching sound. The technique works.

But you may be right, the electromagnetic pulse may couple into the microphone's signal lines, confusing the reading.

Could you 3D-print something similar to what you show on the photo, only shorter?

I tried using a plastic drinking straw, but it didn't seem to be effective, perhaps because it's plastic. I think these sorts of things are usually made of wood or metal, like a musical instrument.

Or maybe there has to be diaphragms on each side, like a stethoscope...

Hi,

several times I used a small flexible plastics tube.
One end close to the ear, the other end for scanning the PCB.

I have some airline tubing around, like what's used for aquarium pumps. Was it something similar to that?

Being flexible would certainly be nice, though I doubt it would transfer sound any better than my straw did...

It's also possible that I'm not "sniffing" on the correct surface of the component. Or perhaps the deformation creating the sound is anisotropic, and the sound is only emitted in certain directions...

 

[Easy peasy]

solid plastic tube, acryllic, or hard ABS ...

- - - Updated - - -

Also, phone with a freq vs amplitude app display, move the u-phone around - this works really well for > 12kHz - which some people find hard to hear

 

[KlausST]

Hi,

I have some airline tubing around,

Yes. I think mine had a bit wider diameter. Don't try to put it inside your ear, don't hurt yourself.

Not only inductors cause noise. Check the ceramics capacitors, too.

Klaus

 

[d123]

Hi,

Don't want to waste your time, obviously... Would a contact microphone, the way a stethoscope is used, placed on each inductor one by one help to locate the clicking one? It might show a definite pulse that matches the timing of the click which is stronger than when placed on other parts of board? Or who knows, crazy idea - an MLCC on flexible pcb and using piezoelectric effect, detect pulse by change in MLCC change in voltage? Other naïve, newbie suggestion is worse and probably wrong understanding - mini inductor (with capacitor, maybe) waved over each inductor and used to couple signal, or just a THT capacitor with one end connected to a resistor, other lead touching each inductor top in case it senses pulse and generates a voltage...? Shame an ordinary LED (maybe in combination with other components) couldn't do something useful, like when placed under light bulb develops a Vf. Sorry for silly suggestions.

 

[treez]

what you could do is , one -by-one, unsolder each inductor, and have it re-connected on the end of a long twisted pair flying wire back to the pcb, but have it away from the pcb.....indeed, dunk it in some insulating gunk so it is damped...then listen, with each inductor in turn removed in this way.

 

[mtwieg]

I'll give airline tubes and acrylic rods a try today, thanks. Unfortunately, the board has to be mounted inside a chassis, where there's no room for large devices like a phone.

Hi,

Don't want to waste your time, obviously... Would a contact microphone, the way a stethoscope is used, placed on each inductor one by one help to locate the clicking one? It might show a definite pulse that matches the timing of the click which is stronger than when placed on other parts of board? Or who knows, crazy idea - an MLCC on flexible pcb and using piezoelectric effect, detect pulse by change in MLCC change in voltage? Other naïve, newbie suggestion is worse and probably wrong understanding - mini inductor (with capacitor, maybe) waved over each inductor and used to couple signal, or just a THT capacitor with one end connected to a resistor, other lead touching each inductor top in case it senses pulse and generates a voltage...? Shame an ordinary LED (maybe in combination with other components) couldn't do something useful, like when placed under light bulb develops a Vf. Sorry for silly suggestions.

A stethoscope is pretty much exactly what I'm trying to duplicate. But needs to be much smaller, and preferably not require electronics directly in contact with the thing under test. I want it do detect sound only, and not be sensitive to any EM fields.

what you could do is , one -by-one, unsolder each inductor, and have it re-connected on the end of a long twisted pair flying wire back to the pcb, but have it away from the pcb.....indeed, dunk it in some insulating gunk so it is damped...then listen, with each inductor in turn removed in this way.

In this case, I'm not able to do any rework to the board (ISO 9001 and all that).