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.

TPS23753A: Gate switching frequency too high ?

Part Number: TPS23753A
Other Parts Discussed in Thread: TPS23753

Hello, 

I recently developed a prototype that convert a POE input to a 5V, based on the Eval board schematics. Even if the output voltage seems to be OK, I have some doubts on the switching frequency of the MOS. 

I used a 38kohms resistor on FRS pin, but I measured that the maximumum frequency can be 2.55MHz, is it normal ? I guess we may have EMC issues at this frequency... 

Note that I see another period at 340kHz. 

The attached capture shows the gate voltage (Cian), the blue is the drain voltage. The top of the windows shows the unzoomed traces (to see the lower period). 

Best regards,

Théo

  • Hello Theo,

    The 2.55MHz appears to me to be some sort of interference. Can you provide a schematic so I can review please? Thank you!

    Regards,

    Michael Pahl 

  • Hi, thank you for your reply.

    Here is a part of my schematics. Note that I haven't other component on the board yet and I only use resistors as a load (1W). 

    (Component R17,R18,C20,C21,C22 have the same value as the eval board and APD is grounded). Do you have any idea ?

    Best regards,

  • Hello Theo,

    Are you able to share the datasheet for your transformer? I would like to check on that. 

    Additionally, I will test the TPS23753A EVM so we can compare results with the same waveforms. I will post my results here. Can you confirm what the pink cursor is?

    Lastly, I am curious to see how the circuit performs at higher load levels? Can you try 90% load? I am wondering if the circuit is in pulse skipping mode. 

    Thanks!

    Regards,

    Michael Pahl 

  • Hi, thank your for your time. 

    The transformer is a Wurth 749119150. The pink curve is the voltage of CS pin (so the current). 

    I loaded at 2.5W and the frequencies/waveforms are the same (I have to order other power resistors...)

    Note that if I short BLNK pin to RTN, my waveforms are much better (and stable), but it still switching at a frequency of 700kHz (see capture).

    I also tried other FRPS resistors with BLNK=RTN, it decreases the frequency but still different from the theoretical value (100kohms => 290kHz). In that case, the frequency is increasing at low load, especially when we reach the discontinuous conduction mode. 

    Note that I produced this proto with only two layers and no clean ground plane, may it be an issue ? 

  • Hello Theo,

    This is very interesting. The CS pin should not have 30V swings like in the first picture. 

    The BLNK pin is used to help with the overshoot on the CS pin. Increasing that should help eliminate those spikes. Below is a capture from the TSP23753A-EVM-004. This is with 1.45A load (7.25W). I measure about 130kHz. 

    This is a peculiar issue and I want to try to order some steps to take in order of ease to do:

    First, let's try decreasing the CS cap (C4 on your shcematic) from 100nF to 220pF. 

    Second, please review the following design guide: http://www.ti.com/lit/an/slva305c/slva305c.pdf

    This paper covers many calculations for the TPS23753 in a flyback design (7W). I think it is worth the time to double check some values, and the paper also covers how to design a snubber, which I believe would help. 

    Third, please review the following paper for EMI compliance: http://www.ti.com/lit/an/slua469/slua469.pdf

    This paper covers how to design a layout and other practical guidelines for EMI compliance in a PoE flyback design. I will say, designing a board with only two layers is difficult to do. Our EVMs typically have more layers, and the ones that have only two layers have one of those layers almost completely dedicated to grounding and they still have lower EMI performance. So if you do not have a dedicated ground plane, it could play a factor in this issue. But that is not definitive. 

    If you are able, you could copy either the layout or the components of one of our EVMs as a way to eliminate one as a problem source. 

    Regards,

    Michael Pahl 

  • Hi Michael,

    Thanks for your feedback.

    I did not populate C7 (CS capacitor). 

    Anyway, I think I found the problem that may come from the poor stackup:

    I placed my power components on one layer, exept the driver that is on the other layer.

    Even if routes from the TPS23753A to the BLNK and FRS resistor are short, they are just under the main power path, especially the MOS drain, that I guess is a source of HF noise.

    So, for testing, I soldered my MOS straight, and added a copper foil (GNDX) on the opposite side of the resistors and I finally got the expected switching frequency. 

    => I won't have this problem with my final PCB since it will be a 6 layers, with inner ground planes. 

    Does it seems coherent to you ?

    I will keep going with my tests and will check the snubber and the other parameters. 

    Best regards

  • Hello Theo,

    Yes this does seem consistent to me. FRS and BLNK play a role in the gate drive, and the drain of the MOSFET is one of, if not the "noisiest" parts of the circuit. 

    I think from here, adding the resistor back onto BLNK will help with the overshoot on the CS pin (pink trace). Once you complete your tests and other parameters I think we can definitively conclude the source of the noise. It also sounds like you have plans to address the noise in your final layout. 

    Let me know the results of your tests, or if you have any other questions. If you feel that we answered your questions, please indicate so by marking the solution on the forum. Thank you!

    Regards,

    Michael Pahl 

  • Hi Michael

    All the other parameters seems OK, so we can conclude that the noise was comming from the MOS drain and it was coupled with the FRS pin.

    We can close this thread ;)

    Thank a lot,

  • Theo,

    Excellent to hear, glad we were able to come to a conclusion!  I will close out this thread. Thank you for marking it resolved.

    Regards,

    Michael Pahl