2

I've developed a PWM dimmer to dim 100W of LED strips. As I dim it to low PWM values, the power supply begins to hum. I have tried to use also higher PWM frequencies of up to 30kHz, but the problem persists. I have also tried to add big caps (470uF resp. 2200uF) to the power supply before the MOSFET (marked in schematics below) to smooth out the load the power supply sees. But it doesn't have any effect.

What can I do to mitigate the noise?

Edit: I'm using a Meanwell LRS-100-12 power supply and 12V LED strips. I also tried it with shorter LED strips (i.e. ~25W), but it also hums. Measuring the PWM signal when the noise is the loudest. Measurement is done as described in answer: https://electronics.stackexchange.com/a/619725/45406: pwm signal

schematics

board

JoLau
  • 121
  • 6
  • What's the ESR and ESL of the capacitors? (I realize who I start to sound like :-) ) What if you add even more of them, does the noise disappear? – winny May 24 '22 at 14:23
  • So it only hums at low duty cycles, which is low average current? – Aaron May 24 '22 at 15:11
  • If you can't silence it with caps as winny suggests, then you may need to add an LC filter in series with the positive side of the LED strip. – Aaron May 24 '22 at 15:11
  • @winny I've tried two of the 2200uF (EEU-FS1E222B) in parallel. They have an impedance of 0.015 ohm. – JoLau May 24 '22 at 15:15
  • @Aaron yes, in low duty cycles it's most noticable – JoLau May 24 '22 at 15:16
  • 1
    Can you generate the PWM of the two channels with a phase shift of 180°? This would equalize the load a bit. – Jens May 24 '22 at 16:49
  • 100W, at what voltage? Just to know what currents are being switched. Also you say the power supply is the part that hums, but you have not told anything about the power supply. What voltage, current, and watts it is rated for? Hopefully it's not an ATX computer supply. Is it a linear supply, switch mode power supply, or what? – Justme May 24 '22 at 17:17
  • @Justme just added more information about the power supply to the question. – JoLau May 24 '22 at 18:37
  • Do you have any plastic or ceramic capacitors with enough voltage rating laying around? Also, what happens if you change your PWM frequency? – winny May 25 '22 at 07:06
  • @winny unfortunately I've only got electrolytic caps. If changing from 300Hz to 20kHz the sound gets just higher pitched. Between 20kHz and 30kHz I can't really hear a difference. – JoLau May 25 '22 at 07:49
  • Thanks. You are in a bit of a tight spot here. Do you have access to an inductor which can take your full load current without saturating? Loose core? What voltage are you running at? – winny May 25 '22 at 08:12
  • A bit of a crude and lossy solution, but if the edges are the problem, what if you install fairly large gate resistors, say 470 ohm, to your power MOSFETs? – winny May 25 '22 at 08:23
  • @winny Thanks for your suggestions. what kind of inductor would I need? I'm using 12V strips and 12V power supply. Where would I need to put the gate resistor? I've already tried to use 330ohm pull down resistor (i.e. R13 in the schematics), without any luck. – JoLau May 25 '22 at 09:24
  • In series with the gate, from the ESP32 output signal. As for inductor, a not insignificantly large one. I can't post schematics as a comment, so I'll post it as an "answer". Don't accept it unless it really did solve your problem. – winny May 25 '22 at 09:33
  • @JoLau Do you resolve the issue? I have exactly the same problem, I tried to use 4 different SMPS, and I always get the same results. The other day, I was walking in Paris, I went across a electronic store. I saw the owner, asking him advice for the noise, and he show me a Mean Well PSU, a closed one. And he told me, he had never hear a single noise from it. Sadly I didn't write down the model but I think that it is the LPV-120-12. So I think we are on something when *bobflux* is talking about dipped in resin circuit. – ZigDuDim Sep 20 '22 at 22:38
  • @ZigDuDim unfortunately not, at the moment I just have to live with the noise. I also tried a closed one from Meanwell with integrated dimmer, with the hope that meanwell would be able to tune the dimmer and AC/DC converter to work together. But it also had noise. Please let me know if you find a solution. – JoLau Sep 21 '22 at 10:49

4 Answers4

4

PWM-ing the output of a switched mode power supply (SMPS) is nothing different than testing its transient response (e.g. 0-to-100% then 100%-to-0), but way faster. Load transients are normally performed with pure resistive or pure capacitive loads. But in your case, forcing a load transient with a partially inductive load creates some trouble for your power supply:

  • The ground rules for the power supply change due to the lengthy coppers
  • The power supply will see higher peaks due to the inductance

EDIT: The text I removed (strikethrough) above apply to power supplies with remote sense inputs which are connected to the load with very long cables.

Load transients with very low duty cycle may not allow the converter to recover quickly due to the very low load duration. If the converter's response is not fast enough the output may start to oscillate. And also, by PWM-ing with very low duty cycle, you might be forcing the supply to hiccup which sometimes results in buzzing (gapped transformers are happy to buzz - magnetostriction. Some smartphone wall chargers are a good example to that: As the charge level approaches to 100% the charger starts to buzz).

  • Increasing the output capacitance further at supply side usually helps.
  • Placing an electrolytic capacitor with high capacitance and relatively high ESR at load side to tame the resonant peaks may also help.

Another solution could be to decrease the PWM frequency to leave enough time to the converter to recover, but decreasing further will make the overall PWM audible due to the lengthy coppers.

Rohat Kılıç
  • 26,954
  • 3
  • 25
  • 67
  • To be specific: cable inductance is on the order of 0.3uH/m. Maybe not much, but it also adds up! – Tim Williams May 24 '22 at 16:50
  • I think, the power supply will see _lower_ peaks at an inductive load compared to a capacitive one. At turn on an inductor would soften the current ramp and at turn off the inductor is just disconnected (and back EMI not properly handled) here. This would be the same as turning off a resistive load. – Jens May 24 '22 at 16:59
  • I'm not sure this is the culprit as it also happens with shorter strips (~25W/12V). – JoLau May 24 '22 at 18:39
  • @Jens you're right. Good spot. Higher peaks could have been seen if remote sense inputs were involved. – Rohat Kılıç May 25 '22 at 06:47
3

If you have lots of LEDs (100W is a lot), you can reduce the current ripple with multichannel.

Instead of using just two MOSFETs, you can use as many as you have PWM channels on the ESP32 (so, 8 or even 16 channels). This means smaller cheaper MOSFETs, and something like 74HCT245 as 3V3 to 5V translator to drive them if they don't like 3V3 gate drive.

Then you simply offset the t=0 point of each PWM waveform so they don't switch all at the same time.

Right now, you only have two MOSFETs, but you can still shift one by a half PWM period. This will halve ripple current.

I use Meanwell power supplies for my LEDs, I never noticed any noise from them. But I didn't use the open frame ones, rather the ELG and other encapsulated models. Maybe the fact the whole circuit is dipped in resin dampens the noise.

bobflux
  • 70,433
  • 3
  • 83
  • 203
  • Thanks for you suggestion. Unfortunately splitting the LED strips up is not really an option, as it wouldn't work with design of my lamp. – JoLau May 25 '22 at 09:26
  • Ah well. Note it also makes the flickering frequency higher and less noticeable. Anyway, I stuck my ear on the XLG-200-24 power supply that powers my big living room light. I hear a very faint 5kHz tone from the PWM frequency of the aliexpress LED controllers I used, but no 100Hz hum at all. The 5kHz tone is only audible if my ear is stuck on it. So perhaps these potted supplies do reduce audible noise. Maybe stray 50Hz magnetic field from the transformer makes the steel shield vibrate on your open frame supply? Perhaps it could be damped or something. Try sticking a finger on the shield... – bobflux May 25 '22 at 10:04
  • I mean, press a finger on the shield to dampen the vibrations, then you'll know if that's the problem. – bobflux May 25 '22 at 10:05
2

Rather than fixed with PWM, it ought to be a variable f that increase due to hysteretic feedback % with series L.

Your low PWM rate must be stimulating a resonant pole in the regulator with sub-harmonics with overshoot and surge currents may be inducing resonance in the ACDC core causing hum. This is common with Buck then PWM with alias and even chaos white noise when used with a boost regulator.

It can be fixed several ways e.g. with ultra-low ESR cap after L to add a zero to cancel pole and PFM modulation instead of PWM with avg current feedback.

It must be a weakness in that ACDC regulator.

I can't read your schematic at the moment with a blank overlay on iOS on jpg, but I do not see any meaningful measurements or layout photos.

Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182
0

Not an answer, but some suggestions to try out:

  1. Try a different power supply.
  2. If you can live with the buzzing but the voltage ripple creates problem elsewhere, try this:

schematic

simulate this circuit – Schematic created using CircuitLab

  1. Heavy handed CLC solution: CLC filter

Feel free to play around with the values in your simulation software of choice but you will notice when doing so that the higher PWM frequency and the slower slope (bigger R2 gate resistor above), the less filtering is needed. Please adjust to include real ESR values for the capacitors. That 100 uH, 9+ A capable inductor will be large...

winny
  • 13,064
  • 6
  • 46
  • 63
  • Thanks for your response. Is it possible to share your LTSpice model, such that I can play around with it myself? – JoLau Jun 02 '22 at 09:05
  • @JoLau Deleted/didn't save it yesterday when I restarted my computer but it's only 8 parts so you can reconstitute it in your simulator of choice in two minutes. – winny Jun 02 '22 at 09:29