1.5KHz , 3W boost converter whining sound.

[grizedale]

Hello,

We have a 1.5KHz self-oscillating switch mode DCM boost converter (11V to 100V) which makes a continuous whining sound which is really quite loud.

Since the switching frequency is within the audible range, is the whining something that we just have to put up with?

Its a led driver and the customer have expressed a dislike for the whining.

I am wondering which components are the most likely causers of the sound? Is it the inductor? , or is it likely to be the 470pF ceramic capacitor that is from the switching (drain) node to ground?

The LEDs are SMT , and i am wondering of their lenses could be vibrating to cause sound?

 

[mister_rf]

Usually the inductors make the noise in switching regulators. But why are you using such a low frequency?

 

[goldsmith]

Dear grizedale
Hi
The core of each inductor and transformer can be cause of noises ! that you can hear with your ear ! there are some ways to decrease it's sound !. such as a little gap space , or piece of glue on your core . or keeping it with some special materials .
Best Wishes
Goldsmith

 

[grizedale]

it is low freq because, well, thats how they did it, lower switching losses, well damped switching transition is possible which avoids drain ringing , which is bad for this type of converter and reduces its power throughput

 

[emontllo]

Why are you using such a low frequency? typically, DC/DC converters for commercial applications use switching frequencies > 20 kHz. Operating at 1.5 kHz it is almost impossible to mitigate the audible noise unless you use expensive noise absorption solutions. I would firstly suggest increasing switching frequency.

Regarding noise sources:

- Inductive components are like "speakers" and cause mechanical vibration. Using some silicone or adhesive to damp the vibration might reduce some noise. Compact, or epoxy packed inductors may be less noisy.

- Ceramic capacitors have piezoelectric behavior and therefore they cause mechanical vibration with the ripple voltage. Electrolytic or film capacitors do not have piezoelectric properties and will be less noisy.

 

[mtwieg]

At such a low frequency, some audible noise from the inductor is probably unavoidable. But you might be able to diminish the volume. First, make sure the inductor is not saturating. After that you might try changing the frequency slightly, just in case 1.5KHz happens to be near some mechanical resonance in the circuit. If that doesn't help, then either using another inductor, or potting the existing one, is probably the only option (aside of course from operating at ultrasonic frequencies).

It's also possible that it's a capacitor making some of the noise, especially if it's a poor dielectric.

 

[grizedale]

thanks, but we wont be potting it as thats too expensive.

I think that 1.5KHz is in the sensitive part of human hearing (?) , and i wonder if i make the frequency about 10KHz, then that wont be so audible, because in reality, human hearing only goes up to 6Khz.

Most music is between 1-4KHz.

Phone call audio is 1 - 4KHz.

And guitar string frequencies (non fretted) are all less than 1 KHz.

So do you agree that at 10KHz, no one, except dogs will be able to hear it?

 

[mtwieg]

Many people can here 10KHz clearly, especially younger people. One of my coworkers routinely works high power at 10KHz and it is quite audible to most people.

If the ringing is due to lorentz forces in the inductor, then increasing frequency to 10KHz may actually make it louder, even if your hearing is less sensitive at 10KHz. Only one way to tell. But if you're willing to increase it to 10KHz, you might as well just go for 20KHz. I doubt the loss in efficiency will be severe, especially if you optimize component values.

 

[Mr.Cool]

20kHz = can't hear noise

 

[goldsmith]

Dear Mr.Cool
Hi
I designed an SMPS with 500 KHZ frequency of PWM . the core was a bit noisy ! because of low components of PWM . if you see the spectrum of PWM , you'll understand what i'm trying to say . i agree that at high frequencies the noise will attenuate . but if your power be high , thus , the noises will be more powerful ! . of course it depends on material of your core too . did you worked with LM2576 ? it's frequency is around 52 KHZ . but you can hear noises from your core !
Best Wishes
Goldsmith

 

[ALERTLINKS]

Can you find a silent flashgun using lo frequency in its boost converter for charging high voltage capacitor? Everybody knows, how it sound like. From low to higher frequency, a siren type ranging from around 200Hz to well above !0KHz. I could detect th charge state only from emitting sound. When battery is fresh its frequency shift is much higher until its near inaudible.

 

[Mr.Cool]

goldsmith - yes i know exactly what u speak. at much higher switching frequency you are not hearing the carrier wave, but actually you are hearing the harmonics of the fundamental, i.e. low frequency noise.

low frequency noise is generated from the DEAD TIME of your switching pattern. reduce or eliminate the deadtime (and if you're very clever you can eliminate it with NO LOSS in dc bus utilization) and you will elminate the low frequency harmonics. some day the subwoofer audio industry will pick up on this idea and run with it...

 

[grizedale]

Hi,

problem now known....when we desolder smt inductor and put it out on twited flying wires, and put blob of blu tak on it, then noise stop....but then we have to put lc filter at input, becuiase the noise was in the supplying psu too, as its inductors did not like the audio freq waveform.

so prob is known, and now we must effect fix...so presume we must make f(sw) higher then 1.5khz

 

[goldsmith]

Hi Mr.Cool
Dead time ? a single switch converter has not any dead time ! of course snubber can reduce that voice , but when i have played with compensation network the sounds eliminated ! ( in feedback path ) . ( don't forget the spectrum of PWM , it has low side components and high side components . like a mountain ! hence dead time ( if we use a two switch such as push pull ) can't has enough effect to create those voices .)
Regards
Goldsmith

 

[Mr.Cool]

yep, does not really apply to one switch topology.

 

[ZekeR]

grizedale, I can hear clearly to 16kHz. My fiancée can hear to 14kHz. Many of my students could hear to 18kHz, and we were shocked when one could only hear to 13kHz. It is very annoying for young people when engineers (who are generally older and deafer) create products that emanate noise within their audible range!

You should send the switching frequency above 20kHz, perhaps 25kHz or more, because component variation in production will cause some products to dip down in frequency.

I once built a 14V-100V boost LED driver, and I was running 300kHz. If I remember correctly, the efficiency hovered around 85%. The benefit you obtain from faster fsw is being able to use (much) smaller components, which can reduce costs (not to mention, making the product sleeker). I had other reasons to maximize switching frequency as well, other than form factor and price. However, I was using new and exotic inductors with low core losses and fast rise/fall times; with standard ferrite cores and EMI-conscious rise/fall times, you could probably push for 100kHz to 150kHz in this application. (You'd have to do the math to find the optimum efficiency point, as I'm not particularly inclined to.)

Question: Why do you have a 470pF capacitor attached to the switch's drain?

goldsmith, the noise you experienced when running at 500kHz cannot be attributed to the switching spectrum. Rather, it is likely that the control loop was small-signal unstable (as indicated when you said that the noise subsided when you changed the loop compensation). It is correct though that the switching regulator will downconvert signals at or near the switching frequency into baseband noise (such as low-frequency input-referred noise riding on the ramp comparator, or beat frequencies from nearby noise sources). Periodic loads placed on the switcher's ouput, such as PWM dimming, also can make audible noise if they're not done properly.

Hope this helps.