PWM Mdimming of switch mode led drivers gives whining sound

Flyback said:

Thanks, I took the inductor off and connected it back at the end of a long twisted pair...then I listened to the board and the inductor separately..it was the board, ie the ceramic caps that were making the majority of the noise..the inductor was just a quiet swishing noise, the ceramic caps were whining like bumble bees...when I change the frequency of the dimming signal, I can play a little wee tune!
Amazing isn't it that these led driver datasheets tell nothing of this problem. of audio noise.

Click to expand...

Hi,

That's very unusual. Perhaps you can try changing out the caps and see if the noise goes away. What size are they, uf and voltage rating?

they are whatever's on the demo boards......couple of uF and rated to double the voltage that they see.
we have 6 demo boards and they all whine like hyenas...not very good if its for a reading light in a library

Can you replace the ceramics with tantalums? Ceramics are piezo electric devices. Hit them with a screwdriver and they will generate a voltage. Hit them with a large transient and they will sing. Tants do not have this problem. Worth a try..

thanks, I will look for some ripple rated tants, as I remember, the old ones never used to be too great in ripple rating....but I suppose we can just massively derate them

Hi,

Yes something has got to work, i've had circuits for years that never made a sound regardless of the frequency in the caps.
We know there is some sensitivity thought through work in audio circuits. They can act like microphones, maybe not good ones, but still pick up audio noise via sound waves.

I missed post #8 where Flyback said, "BTW, the LTM8040 is switching at 500khz". Then later he said he is varying the duty cycle of the PWM at audio frequencies but did not say why.
A dimmer does not vary the brightness of a light at an audio frequency: on, off, on, off, etc very quickly. Instead a dimmer either dims with continuous narrow pulses, or brightens with continuous wide pulses, or slowly ramps the widths up (to become brighter) or down (to become dimmer).

thanks, we can pwm dim the ltm8040 at >20khz, but as you know, there is some (not much) loss of resolution when pwm dimming at higher frequencies.

Flyback said:

thanks, we can pwm dim the ltm8040 at >20khz, but as you know, there is some (not much) loss of resolution when pwm dimming at higher frequencies.

Click to expand...

I do not understand what you are saying. You do not dim at a frequency. You either make the LED continuously dim, make it continuously bright, slowly ramp up the brightness or slowly dim the brightness. There is no frequency.
The PWM simply makes the very high carrier frequency of the pulses.

Hi guys.

Just registered to clear a misunderstanding.
The lt3756 led driver has a 100k..1M pwm frequency. it supports many topologies.
there are two ways of adjusting the led brightness:
1. 'analog dimming' using the ctrl leg of the chip. this plays with the control loop, adjusts the duty cycle.
2. 'pwm dimming', the name is unfortunate. there is a pwm leg on the chip which enables/disables the main control loop.

by using this 'pwm dimming' you use a digital signal with whatever duty cycle you see fit and turn on and off (kinda) the chip.
the documentation recommends at least 100Hz, suggests the shortest pulse should be at least 6 main cycles long.
so suppose you running at 400kHz, and use the ctrl method to achieve 1:10 (thats about the max the chip can reliably do this way),
and choose to use a 100Hz digital signal to chop the loop some more,
you'll have 400k/100=4k cycles to play with, but the minimum number of cycles you should let it run once turned on vie pwm leg is 6,
so 4k/6=0.6k+ additional diming can be achieved for a total of ~1:6k+

of course the 100Hz *might* be audible.
i have a proto/breadboard sitting in front of me, with an lt3756 in a msop16 mse package, had some fun soldering wires to its legs -.-
it usually does not whistle (400kHz main loop with ctrl + 10Hz pwm dimming), although very sensitive to part and wire placement.
yesterday i have moved around a few things and it started to resonate.
i think it's time i get the pcb made, i tried to be careful with the design.
at least there are no 6+cm long wires with 3A 400kHz load on them >.<

BR
béla

edit: white power leds slightly change color, when dimmed using the analog method. the 'pwm' method comes to help, feeding led with constant current for constant color.

white power leds slightly change color, when dimmed using the analog method. the 'pwm' method comes to help, feeding led with constant current for constant color.

Click to expand...

Thanks, I know in theory they change color, but is there a video anywhere where we can see the color difference between pwm dimmed and analog dimmed leds?.....I cannot see any difference in the ones I play with.
I find the led drivers that make noise are the ones with capacitor banks comprising just a few ceramic caps.....add a few big electrolytics and the noise much reduces.

Hi, thank you for the tip.

The problem was the bad grounding. As the datasheet recommends, the few tens of milliohms range Rsense and other stuff should be well grounded, and all the ground shall be connected in a star like topology close to each other. I'm working on a proto board so I had some wires all over the place. Whatever, I solved it.
I have a 4.7u ceramic on the input, and a 3x1u ceramic on the output. No other filtering capacitor, no electrolitics. It is working fine with 100Hz PWM also.
Of course when I move my hand, I can see the stroboscope effect, but I hear no audible sound, no bad resonances whatsoever.

As for color change, I can't see the color shift either. I believe this is because the human eye is a wonderful and very sensitive device, it instantly accomodates to the (small) changes in color. We have a very nice auto white balance I guess. Other than that, when you reduce the light output, it is such a change alone, I think the eye is first trying to get used to the changing light level, it has no time to worry about color changes. You can't really compare 1:10 light luminance levels for small color changes.

I think these color changes are more visible when you compare different leds together. For example if you try to drive different led makes in series (they have to have similar color and light output), and when you start to reduce the current, their color might wander in different directions. One can go blueish, other can go yellowish, thats what I mean. Even if they are not different makes, every led is a bit different, so if you build a home lighting for example, when you have some leds probably close to each other, and you want to dim them, the color shifts might be easy to notice and it might be a bit annoying or lame.

Best regards,
béla

Flyback said:

Thanks, I know in theory they change color, but is there a video anywhere where we can see the color difference between pwm dimmed and analog dimmed leds?.....I cannot see any difference in the ones I play with.
I find the led drivers that make noise are the ones with capacitor banks comprising just a few ceramic caps.....add a few big electrolytics and the noise much reduces.

Click to expand...

Hi,

The human senses are not very good at discerning absolute levels of any kind because they are so adaptive. They are very good at discerning comparative differences however.

Take one LED and run it with an analog circuit, take another LED of the same make and model and run it using PWM, and run these both at the same time. Hold them near each other and see if you can see any difference then, or shine them on a white surface and compare that way. You can even test for equivalent brightness this way if you make a comparative light meter out of a pure white candle. If one is even a little dimmer than the other you'll be able to see it.

Thanks

If one is even a little dimmer than the other you'll be able to see it.

Click to expand...

So when comparing analog dimmed and pwm dimmed leds, suppose we can indeed tell a difference between them......then which one would look nicest?.....which one would somebody wanting domestic or office lighting choose?

Hi,

It depends on the application, so for normal office work it probably wouldnt matter.

There are other variables here like LED type, and what the workers are doing. For example, some LED types are not suitable for PWM dimming at all because the manufacturer does not recommend pulsing at all. Another variable is just how much does the LED change color if it does at all, and how much does that color change affect the particular work the workers are doing.

The way it used to be thought of was that if you pulse the LED you get the intended color even when dimming because the level of the current when the LED is actually 'on' is the same as when it is run at the intended operating current. But that has changed because some manufacturers specify that you should not pulse the particular LED. So the only choice there is the analog method, even though a buck would qualify as analog because the LED itself would still only see analog. The color change would have to be investigated, if there was a color change at all, and it's effect on the work environment relative to what is being done in that environment. For example, if the workers were just reading reports from some other company or something then it might not matter as they would still be able to read the print, but if the color change was enough and the workers were inspecting the color coding of some sort of parts, then it might not work out very well at all if they depended only on the PWM LED lighting