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.

LM5023 Acoustic noise / Loop response

Other Parts Discussed in Thread: LM5023, PMP7991

Hi, I'm working on a 40W flyback SMPS controlled by LM5023. I got good thermal and regulation results at full load. However, I have some questions regarding LM5023 funcionality.

1) When the supply is working at very low or zero load, an acoustic noise appears. When the load is put above 200mA the supply works quiet, but around 1A the noise comes back and then disappear as the load gets higher than 1.2A. My question: could this be a compensation loop problem? If so, how can I improve loop response? Is there a way to get rid of audible noise at low load since LM5023 has a pulse skip function?

2) A was looking as TI's documentation  around LM5023.

This document is about the evaluation board: 

In this image compensation network parts are spotted in red and the SS circuit in blue.

So I found this other document about a reference design:

I marked in this schematic too compensation network in red and SS circuit in blue.

My questions:

a) what are the behaviour differences between this two kinds of compensation network?

b) what is the role of the resistor in parallel at the SS pin?

Thank you,


  • Yuri, I will get you connected to an AE to help answer your questions.
  • Hi Yuri,
    For the audible noise, it could be caused by vibration of different parts on the board. The part could be capacitors, transformer etc..
    You could find where is the audio from by hearing from a paper tube on the components.
    If it is from transformer, you must varnish the transformer or use different ways to varnish the transformer.
    If it is from capacitor, you may change other types of capacitors.

    As you noted, the unstable of loop could make the audible noise worse.
    A basic way to judge the loop is to measure the gain and cross frequency of the loop by using frequency response analyzer. If it is unstable, modify the R and C in the control loop.
    I suggest you to measure the loop response under different conditions.

    And from your description, in very light load, the audible noise seems caused by the burst mode. So varnishing the transformer should improve it a lot.
    But for the audible noise around 1A, it could be loop unstable or working abnormal. Please check the output ripple , loop response, the driving signals to try to find some clues. and you can also try add more output caps to see whether it help.

    Regarding to your questions
    a) what are the behavior differences between this two kinds of compensation network?
    b) what is the role of the resistor in parallel at the SS pin?
    Here are my comments
    a) In PMP7991, RX5, CX2, R12 added a Pole and Zero to the loop compared with LM5023-2 EVM. Fp0=1/(2pi*CX2*RX5), Fzo=1(2pi*CX2*R12)
    b)the internal source for SS pin is 22uA, the suggested using ways are only to add a cap. I don’t think you need to copy the ways of PMP7791. The 100k resistor will bypass some of the sourcing current and making a slow Soft start. When SS reaches 2.2V, all of the sourcing current will go to the resistor. So the soft start waveforms would be flat when it is near 2.2V. If you need a modified ss shape , you can use the added resistor.

  • Hi Kening,

    Thank you for your answer.

    I finally found what part was making the hiss. It was the capacitor between the primary and secondary grounds. I was using a 4n7 2kV ceramic disk type. I replaced with a 220p 500V MLCC type and the hiss decreased a lot.

    Do you have any hint about a good replacement for this cap?

    Another question: Is it to possible to shape the loop response in order to avoid the controller entering the burst mode?


  • Hi Yuri,
    for my understanding, the cap between primary and secondary Grounds should be Y-cap. there is requirements for Y cap. If it is, You cannot change it to other types such as MLCC.
    However, there is still something you could try.
    1. decrease the pin length as short as possible.
    2. use two y-cap in series with fixed assmble.
    3. decrease the voltage according to the y-cap by improve transformer.

    for burst mode, you cannot avoid it by shape loop response.