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.

LM25145: LM25145

Part Number: LM25145
Other Parts Discussed in Thread: LM61440

Hi,

I designed in the LM25145 using Power Designer.  I need 12V input, 3.3V output and 4A.  I copied the schematics almost exactly, except I used a 6.8uH vs the 6uH in the Power Designer and I put a 47uF (6.3V) and a 100uf (6.3V) cap, both ceramics as the datasheet states you can use low ESR caps on the output.

When I apply power, the output almost looks like the switching node. I don't like to turn it on too much because the noise greatly exceeds 3.3V.  I then soldered a tantalum cap (68uF, 25V) and the supply works perfectly.   I removed the tant and replaced with 2 more 100uF ceramics and the output shows the noise.

I have checked all component values thinking that maybe the CM installed a wrong value, but they are all good. 

Do you have any idea why this circuit behaves this way?

 

Thanks,

JohnPowerSupply.docx

  • Hi John,

    Please check the compensation using the LM25145 quickstart calculator. Send the file for review if you further help. Note ceramic output capacitance must be derated for applied voltage.

    PS: 3.3V/4A normally uses a converter with integrated FETs, e.g. LM61440. This is an easier/smaller solution relative to a controller.

    Regards,

    Tim

  • Just looking at the WEBENCH schematic, Rcomp2 and Ccomp3 create a pole that cancels the output cap ESR zero. I'm not sure why it was recommended for a ceramic output cap, but this explains why it works with a tantalum cap that likely has high ESR. Rcomp should be < 1kΩ when using a ceramic for Cout.

  • Hi Tim,

    Yes, the LM61440 would be easier but it was and is still not available.  Lately allocation is driving designs and one of the reasons I chose this part because we could purchase it.  MANY TI parts aren't available at all.  I will try RCOMP2 to be < 1K and see how that works.

    Thanks!

    John

  • Thanks, John. Check what it recommends with the quickstart calculator as well. The target crossover frequency can be set at ~10% of the switching frequency.

  • Hey Tim,

    I removed the tantalum and just have 100uf and 47uf ceramics which is what is on the board. I then replaced the RCOMP2 from the 35.7K value recommneded by webench to 49.9 Ohms and everything started to work properly.  I had one board that I had changed RT from 10K to the webench recommendation of 39.2K and it seemed to have more ripple so I went back to 10K and that cleaned that up.  

    I was also reading another forum issue on the part and you mentioned that 47uF is too small of an input cap but again, Webencch recommends that.  So it seems that particular webench design has some issues.

    Anyway, I've attached the LM25145 calculator. Is there a reason why you have a specific spreadsheet for this part?

    If you can take a look at the spreadsheet and see if everything looks ok, I think we can close this.  Although I did fry two boards (the output of the regulator fried some downstream devices, even my FPGA) I want to get a couple of fresh boards and make this change prior to powering it up so I will be 100% confident.  

    Thanks!

    John

    7026.LM(2)5145, LM5146 contorller design tool - revA1.xlsm

  • Hi John,

    300kHz is a good tradeoff of efficiency vs. size (1MHz with RT = 10kΩ seems too high). Keep the current limit at ~1.5x the full-load current, so 6-7A. The loop crossover frequency can be 10% of Fsw. Just ensure that the Cout and its ESR are accurately characterized as this affects the compensation calculations. In particular, the 100uF and 47uF ceramics should be derated for voltage of 3.3V (check the cap datasheet).

    Regards,

    Tim