PWM dimming a switch mode led driver gives whining sound

Hello,
Have just been running switch mode LED driver demo boards from ti.com and linear.com. I usually do Analog Dimming, so am not used to listening out for “whining” during PWM dimming.

However, all of these boards “whine” when PWM dimmed….the whining sound varies as the PWM dimming frequency is varied (kind of as you’d expect).

The feedback loop is stable in all cases.
The chips are LM3405,LT3756,LTM8040.


(LM3405 is hystetetic so is intrinsically unstable but you know what I mean)

LTM8040:
http://cds.linear.com/docs/en/datasheet/8040fa.pdf
LM3405:
http://www.ti.com/product/lm3405
LT3756:
http://cds.linear.com/docs/en/datasheet/375612fb.pdf


I’ve removed the shielded, surface mount power inductors, and connected them back into the circuit with twisted pair, so that they cannot resonate against the circuit board, but the whining is still there. (i didnt remove the inductor on the LTM8040, becuase its integrated into the chip)

Do you think it could be the ceramic caps?

I wonder if I change to unshielded through hole inductors it will get rid of the whining.

In a way, I’m not jolly well surprised that the circuit is whining, the current is being slammed on and off at an audio frequency. (that’s PWM dimming)
Any ideas to get rid of this whining?

MLCC ceramic caps are indeed electrostrictive. They do whine.

Would you consider potting or encapsulating the assembly?

thanks, ill start by squeezing some silicone into the inductor, and if that doesnt help, will try and do something about the ceramic caps....i believe non-whining ceramic caps are for sale, but at a price...i am not sure if potting compound is restrictive enough to stop ceramic caps whining , as the ceramics dont appear to be vibrating much when you look at them.

Here is a ceramic cap which supposedly reduces the audible whining, its stacked into a frame which damps out the noisy vibration

**broken link removed**
(KPS seires capacitors from kemet)

Does anyone know by how much these capacitors curtail the audible noise?....does it make them non-hearable?

The best way would be to order samples and perform an A/B comparison with the standard cap.

Besides piezoelectric behaviour of high Er capacitors, you should also consider magnetostrictive properties of ferrites. Both will sound. I can't follow the idea that you really need to use special capacitors and inductors for a run-of-the-mill LED driver.

If your application is sensitive in this regard, e.g. home LED lighting, you preferably fix the circuit instability or shift all emissions above 25 kHz. If some driver products are falling through the selection process, give the manufacturer a respective feedback.

Potting the driver circuit may be a suitable means for various reason, but shouldn't be dictated by a badly designed IC (or circuit).

Thanks, as you know, in the case of LTM8040 we cannot fix the circuit, as for one reason it seems to be operating perfectly well, and the other reason is that the entire led driver smps is integrated into a chip, and not accessible.
I think what is happening here is that we have circuit current being slammed on and off at an audio frequency, and as you'd expect, there is whining sound, its not much, but enough to be a nuisance in a quiet environment. Id be surprised if potting would solve this problem, as potting compound is quite rubbery and I doubt it would clamp the vibration? I even de-soldered the single input capacitor on the ltm8040 circuit and soldered it back on two "legs", but the circuit still whined.

There are many types of potting materials, some of them rubbery (like urethane), other hard as rock (epoxies), and everything in between.

A simple and quick experiment you can do: place a drop of silicone RTV on top of the offending components. See if it works for you.

If it does help, make sure that you use electronic-rated RTV. The plain RTV that is used as caulking for bathrooms contains acetic acid (thus the vinegar smell) and will corrode your components.

thanks, don't these "flexiterm" capacitors give reduced audible noise?..
**broken link removed**

I really don't know....you really have to perform an A/B comparison between capacitors.

And of course, when you do the evaluation, you will need a calibrated sound meter. Make sure that the meter allows you to select both the "A" and "C" weighting curves.
The "A" curve, which is intended to determine speech interference, has some pretty steep rollof filters that could filter the whine frequency.

Since the whine will also be fairly low level, most likely you'll have to perform the test on an anechoic chamber, or at the very least on a very quiet room.