This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM3409HV: 3kHz noise

Part Number: LM3409HV
Other Parts Discussed in Thread: LM3409

Hello

Were using the LM3409HV with a shunt FET. Our circuit is very similar to the circuit in AN-1953 LM3409HV Evaluation Board (see below). Five of these drives a bank of LEDs. 

The problem is that when the driver should be "off", i.e the PWM signal is low, the controller is still switching, generating a 0.7% duty cycle 2.9 to 3.1 kHz signal to the switching FET (Q1). This creates a lot of piezoacoustic noise in the capacitors, especially when there are 5 LED drivers close together. And the ear is very sensitive to 3 kHz. We can't ship a product this noisy so I have to get rid of the sound.

When the PWM signal is held low, the EN pin is at a steady 5 volt. Is this what causes the LM3409 to keep switching, and could you suggest a way to stop the LM3409 switching when PWM is held low?

Thanks

/Morten



  • Hello Morten,

    I'm sure we can solve this. There is a bit to consider, including the fact that the EVM in question isn't exactly ideal for shunt FET dimming for all conditions.

    In any case, the point of shunt FET dimming is to get very fast current rise and fall times to increase the overall contrast ratio. The way to do that is to keep switching and regulating the inductor current so that when the FET turns off you can get very fast current rise times. If you do not need such fast rise times you can just use EN or UVLO which will stop switching when pulled low but will be limited in rise time by (Vin-Vout)=Ldi/dt.

    But if you need very fast rise/fall times that shunt FET dimming provides you want to keep the inductor current slewed up. BUT, ideally you want the current ripple the exact same when the FET is on as it is when the LEDs are on. That requires tuning the R11 value for your application (a loop of wire in series with the inductor, a current probe, and some tweaking). That will likely change the frequency and maybe will minimize the noise.

    Other noise sources of course are ceramic caps and/or the inductor. You can always get rid of it with some experimentation. First, do not use an output cap with shunt FET dimming (C8). That will cause noise and likely damage things over time. But the input caps can cause noise too if ceramic. But you can get rid of it by using aluminum electrolytic, tantalum, or by trying different configs. For example, if you replace two 1210 caps with 3 or 4 1206 caps at the input you will not only change the frequency at which they make noise but will also reduce the rms current in each which reduces stress. It takes some playing around. But if it's the inductor you can usually just try different ones and something will not make sound at the frequency of concern. If tuned right in the first place neither may matter.

    Worst case some folks just pot the circuit board. But that's not usually necessary with some experimentation.

    If you really want to stop switching with the shunt FET on for some reason you can just pull EN or UVLO low when the FET is on, but that will limit the rise time the next time you turn the shunt FET off and reduce the advantages of shunt FET dimming.

    What kind of PWM frequency and dimming range is required for your application? Maybe there are better ways to dim for your app. Thanks.

    Regards,

    Clint

  • Thank you Greg!

    Sorry I was not clear. I don't want the switcher to stop during the off-time of the PWM. That would as you point out mostly defeat the purpose of the shunt FET. I want the switcher to stop when the PWM signal goes "permanently low" if you will. When the light is off.

    The product is a professional light for photo and video, We color-mix 5 colors to produce a 3000-10000 kelvin white with a CRI above 95 over the range plus the full color gamut. Not all LEDs are lit at all times, for example at 5000 kelvin white the red LED is not active at all. The LEDs that are not active starts generating this 3kHz noise. And when when the light is turned all the way down, all five are making noise in unison.

    A previous design fed the PWM signal to the the EN pin and the light was silent. But due to rise times the output was not linear at low duty cycles and we need precise dimming down to well below 1% to maintain correct color or color temperature. 

    The main source of the noise seems to come from the input caps. The inductor also made sound but I was able to find a different brand that was silent. I'll try your suggestion for changing the sizes of the input caps and tune R11.

    Ideally I would like to get at the root cause for the noise. Pulling UVLO low would require an IO expander and firmware control if we were to do this from the MCU (I don't have five unused pins). The perfect solution would be local to each switcher and pull UVLO low when the PWM signal went away. Sending the PWM to an RC filter feeding the gate of a FET might be able to turn the UVLO off, but it would also have to come on very quickly when the PWM pulses appear again. I can't quite see how to implement something like that, or it is is possible at all.


  • Oops, I meant Clint! Sorry.
  • Hello Morten,

    I see your issue. Ideally to stop switching and get the fastest response you would need to pull up/down UVLO. Using an RC is a good idea to accomplish that, but it would delay the current a little next time you start up (it would be like driving EN directly except there is no startup delay after it has been pulled down for a while). So UVLO works better than EN timing wise, but the first pulses you would still need to slew the inductor current up.

    In either case the ceramics generally are the culprits and with some experimentation with those (or switching cap technologies) you should be able to get rid of any audible noise.

    Regards,

    Clint