When I can hear LEDs switching, what am I hearing?

[meshtron]

I've got my first product getting close to launching which is great!  It's a high-side 12VDC PWM driver and I've been testing with a variety of LED light modules and light bars.  I notice as I change frequencies, and depending on the light, I can physically hear the LED making noise when it's running PWM (not when we get to DC).

I'm just curious what component(s) or process(es) are actually physically creating the noise.  I've also noticed that most lights have some "preferred" frequency where they make essentially no noise, but will emit noise at higher or lower frequencies (and for some reason most of what I've tested wants to be between 300 and 500Hz).

My intuition tells me that running lights at a frequency that produces the least amount of audible noise would be "better," but I realized not knowing even what the noise is from makes that feeling a bit moot.

[tooki]

On behalf of everyone who is like me, sensitive to low-frequency PWM flicker: please, for the love of god, please don’t use anything under 1kHz, with several kHz being preferred. 300-500Hz are readily noticeable to us (especially at low duty cycle) and is highly annoying and distracting.

With that said: as I understand it, the usual noise-making components are multilayer ceramic capacitors (piezo effect) and coils (inductors, transformers, etc; magnetostriction IIRC).

This is a reason why PWM frequencies in the “couple of kHz” range are fairly rare: it can easily lead to audible whine.

When I built the PWM dimmer I use at home, I designed it to be above 20kHz, and after component tolerances, it ended up at around 26kHz.

Unfortunately, thanks to switch mode power supplies, which sometimes whine on their own a bit, the interaction between the power supply’s switching and the PWM frequency can create audible beat frequencies and/or harmonics even if both the power supply’s switching frequency and the PWM frequency are both well outside the audible spectrum. You’ll have to test the combination.

[meshtron]

On behalf of everyone who is like me, sensitive to low-frequency PWM flicker: please, for the love of god, please don’t use anything under 1kHz, with several kHz being preferred. 300-500Hz are readily noticeable to us (especially at low duty cycle) and is highly annoying and distracting.

With that said: as I understand it, the usual noise-making components are multilayer ceramic capacitors (piezo effect) and coils (inductors, transformers, etc; magnetostriction IIRC).

This is a reason why PWM frequencies in the “couple of kHz” range are fairly rare: it can easily lead to audible whine.

When I built the PWM dimmer I use at home, I designed it to be above 20kHz, and after component tolerances, it ended up at around 26kHz.

Unfortunately, thanks to switch mode power supplies, which sometimes whine on their own a bit, the interaction between the power supply’s switching and the PWM frequency can create audible beat frequencies and/or harmonics even if both the power supply’s switching frequency and the PWM frequency are both well outside the audible spectrum. You’ll have to test the combination.

Interesting - thanks for the info.  Output frequency is user-configurable and I'm sure some use cases (for things besides LEDs) will want to be below 1kHz, but I can certainly steer people who ARE using them for LEDs towards 1kHz and up frequencies.  In my use cases, the audible noise isn't of any real concern, but I'm certain I'm going to get questions about it.

Sounds like, reading about the piezoelectric affect a bit, it's not representative of anything "wrong" necessarily so my thinking about the quietest frequency being "best" doesn't hold any water.

In some of my early testing, I definitely saw temperatures on the switching IC (DRV8144) climb at high frequencies, but I'll do some more now that I have all the other issues sorted and I can record both actual temperature and actual output waveform data to look at.

[tooki]

Yeah, audible noise isn’t necessarily a sign of a malfunction.

As for higher temperature: yes, higher frequency = more switching losses, which is one reason many engineers choose low PWM frequencies. (Unaware that the 60Hz minimum for flicker-free images on a CRT doesn’t apply to LED lighting…)

[MarkT]

How much current is involved?  Current carrying conductors with PWM will vibrate at high current due to their magnetic fields.  Audability would depend on mechanical compliance and resonances.  The force between two 3.5A carrying conductors 5mm apart is about 1 newton per metre.

[meshtron]

How much current is involved?  Current carrying conductors with PWM will vibrate at high current due to their magnetic fields.  Audability would depend on mechanical compliance and resonances.  The force between two 3.5A carrying conductors 5mm apart is about 1 newton per metre.

3A at the low end, 20A at the high end.  So, that could be part of it too?  Definitely "sounds" like it's coming from inside the lights and resonating through their housing, but hard to say and there may be wiring inside there as well.

[richnormand]

I usually use a Tygon flexible tube to find the culprit.
One end at your ear the other to do a close scanning of the components.
Works wonders.

[sicco]

Might be photoacoustic effect?
https://en.wikipedia.org/wiki/Photoacoustic_effect

[MarkT]

3A at the low end, 20A at the high end.

20A will do it nicely, quite a lot of force in fact.

[AnalogTodd]

3A at the low end, 20A at the high end.

20A will do it nicely, quite a lot of force in fact.

Being in IC design (power in particular) I've worked on plenty of parts where they have been in an open package and you could hear the bond wires move as you did load steps. That was down at currents below 1A, so 3-20A loads will definitely cause conductors to move. Movement is usually caused by fields in the conductors either attracting or repelling each other. Talk to people who do really high power (thousands of amps) and you can get cables in cable trays overhead jumping with sudden load steps.

[themadhippy]

I can physically hear the LED making noise

Make it a feature,retaining the audio experience of light of old .Wasn't uncommon to hear lights  ringing/whistling  when run at anything less than 100%,the 1kw par64 is especially good at it.

[MathWizard]

Tubes up to the ear works ok sometimes. I made a little 1-2 BJT audio amp for an electret-mic, I need to mount it to something to use as an audio-probe. I have a couple of PSU's that buzz a bit, I want to see if gluing some parts down helps or not.

[LinuxHata]

If these are led strips, and they're glued onto metal surface, of course they will sound.
Implementing soft switching, instead of hard PWM, will reduce it a lot.

[AndyC_772]

Do bear in mind that audible noise can be a sign that a power supply is unstable, and it's generating harmonics below the switching frequency that are more clearly audibe than the switching frequency itself.

More than once I've been handed a board with "it's making a whining noise, but otherwise working" and it's turned out to be a PSU fault affecting stability.

[MarkT]

Talk to people who do really high power (thousands of amps) and you can get cables in cable trays overhead jumping with sudden load steps.

My can-crusher (4 x 10mF 400V caps (3.2 kilojoules, 16 coulombs) will also squeeze the copper windings of the coil so hard with magnetic force the wires get crushed together and are no longer circular in cross section...  Can't figure out if the residual heat in the coil is mainly due to electrical dissipation or mechanical deformation!!

[DavidAlfa]

Coils and capacitors vibrate.
DC/DC converters rarely operate under 25-50KHz, not only allows smaller inductors, also removes human hearing range, specially around 1KHz which is the most sensitive frequency for us.

[meshtron]

Do bear in mind that audible noise can be a sign that a power supply is unstable, and it's generating harmonics below the switching frequency that are more clearly audibe than the switching frequency itself.

More than once I've been handed a board with "it's making a whining noise, but otherwise working" and it's turned out to be a PSU fault affecting stability.

Good point.  In this case we're running (testing) from 12V batteries, so I'm confident it's coming from the light itself.

[MathWizard]

Talk to people who do really high power (thousands of amps) and you can get cables in cable trays overhead jumping with sudden load steps.

My can-crusher (4 x 10mF 400V caps (3.2 kilojoules, 16 coulombs) will also squeeze the copper windings of the coil so hard with magnetic force the wires get crushed together and are no longer circular in cross section...  Can't figure out if the residual heat in the coil is mainly due to electrical dissipation or mechanical deformation!!

So what crushes the cans tho ? Is that charge fed to a motor or some plates that form a capacitor and squeeze together under electric force ? I could calculate that to see if it makes any sense, but I'm busy.

I just looked up electric can-crushers, they seem expensive. I don't need one, I gave up sugary drinks in my 20's and beer some years ago.

[Gyro]

The can acts as a shorted turn and so experiences severe deformation forces when the coil is energized with a very high current pulse. It's not a practical method of flattening cans, as in the domestic electric motor powered can crushers that you're thinking of.

[5U4GB]

With that said: as I understand it, the usual noise-making components are multilayer ceramic capacitors (piezo effect) and coils (inductors, transformers, etc; magnetostriction IIRC).

If you want to find lots of information on this, the magic Google terms are "coil whine" and "capacitor whine".