• State TI Thinks Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 3 answers
  • Subscribers 63 subscribers
  • Views 696 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

Original question:

LM3150: LM3150 - squeaking noise

LM3150: LM3150-frequency jitter and squeaking noise

Petr Zkrat
Intellectual 410 points
Part Number: LM3150
Other Parts Discussed in Thread: LM5166,

Hi,

getting back to this issue however can't open the previous thread https://e2e.ti.com/support/power-management-group/power-management/f/power-management-forum/986269/lm3150-lm3150---squeaking-noise

I finally got a board that was squeaking from the very first moment so could test some more on it. 

Reason seem really be that there is quite a jitter in the switching frequency. In light load it is about 200-300ns, in heavy loads can go to 1us. Some boards squeak even at light loads, but higher number is hearable at the very high loads. 

I tried to adjust the Type III ripple network with C17=1nF as suggested in the reference thread and things got significantly better, the ripple is reduced to < 500ns and squeaking went away. 

In the previous thread Orlando suggests to increase C17, but should then also R4 and C4 be adjusted? (this is an inherited design and there is no record if there was some reasoning behind these).

I also tried to use values suggested by webench - R4=100k, C4=6n8, C17=100pf. The jitter is even lower here, however the switching frequency jumps from about 200kHz to 275kHz. What is the reason behind it?

I am aware of SNVA776A, however that does not seem to address the C17 value wrt to the sw frequency.

Thanks

Petr

  • 0
    TI__Guru 73072 points

    Hi Petr,

    130pF coupling cap is probably too low and discharge may occur with the effective resistance at FB (hence why 1nF works better). Please ensure there is adequate triangular, noise-free ripple voltage at the FB pin. Also, it is critical to have the FB trace as short as possible to reduce noise pickup. 

    The FB resistances are relatively high here and should be located very close to the IC with short connection path to GND.

    Regards,

    Tim

  • 0 in reply to Timothy Hegarty
    Intellectual 410 points

    Thank you very much Tim.

    1) What is the recommended ripple for LM3150? I can see that for example the LM5166 has the hysteresis in datasheet, however I cannot find same information for LM3150.

    2) We did address FB loop on the latest prototype currently being released. Hand proto did improve things on low load (still audible at high load), 1nF improved the jitter to about 500ns. Is there any rule what is acceptable?

    3) With regards to webench, simulation, it seems to recommend R4=100k, C4=6n8, C17=100pF however the simulated switching frequency is 100kHz off from reality, however it seems that simulation frequency is spot on to frequency given by Ron. Is that just a limitation of simulation model?

    4) Is it better to keep the values then as 402k / 1n5 and 1nF for c17? Any particular concerns, things to verify? (apart from FB node injected ripple).

    Many thanks

    Petr

  • 0 in reply to Petr Zkrat
    TI__Guru 73072 points

    Hi Petr,

    The LM5166 has a hysteresis for PFM operation, not COT. With COT operation, the falling FB voltage reaches the FB threshold, which triggers the high-side FET to turn on.

    In general, the ripple should be triangular (in phase with the SW voltage) and about 50mV pk-pk amplitude. You can use the LM3150 quickstart calculator to assist with component selection: https://www.ti.com/tool/LM315XQUICK-CALC

    Also, take a look at app note snva874 for more detail: www.ti.com/lit/snva874

    Regards,

    Tim

  • 0 in reply to Timothy Hegarty
    Intellectual 410 points

    Hi Tim,

    I based the things above based on SNVA776A. There is referenced the 4mV hysteresis (also found in datasheet) and based on that, 20mV is chosen as a conservative target value. Or am I misunderstanding anything?

    "For the LM5166, the 20-mV term is conservatively selected based on the feedback voltage comparator hysteresis of 4 mV (typical). Note that other COT buck converters may have different comparator hysteresis, therefore the term may need to be adjusted accordingly." The 20mV is also referenced later in the stage where type 3 comp network is discussed. 

    The referenced excel sheet does not seem to address type 3 compensation, however I will try to read through the stability document you posted.

  • 0 in reply to Petr Zkrat
    TI__Guru 73072 points

    Hi Petr,

    The 4mV hysteresis is an additional feature on the LM5166 COT mode operation as it helps high duty cycle operation when the ripple amplitude may be low. Basically, the high-side FET stays on until the FB voltage increases by 4mV (this on time can be much longer than the one-shot timer). The LM3150 does not have this - it just turns on the FET when the FB voltage reaches the threshold and turns off after the COT on timer has expired.

    Regards,

    Tim