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.

WEBENCH® Tools/TPS54360: Hot Plugging Burnout with TPS54360

Part Number: TPS54360

Tool/software: WEBENCH® Design Tools

I'm experiencing repeated catastrophic failure of the TPS54360 when it is plugged into an active power source.

In particular, I have the TPS54360 setup to convert 48V DC to 6.5V DC for power over Ethernet. When the Ethernet cable is connected with the system off, turning on the power supply results in normal operation. However, when the Ethernet cable is connected after power is already supplied to the system, the TPS54360 fails dramatically. This happens with some regularity, though not every time the device is hot plugged. I have been able to reproduce this behavior.

I'm using a recommended WEBENCH configuration for the TPS54360 and a Mean Well supply for the 48V. Any recommendations about the potential problem (and how to solve it) are appreciated.

  • Patton,

    Check for overshoot on VIN. You would see this sort of issue if the VIN voltage goes above the abs max spec (65V in this case).

    Hot plug will be worse than turning the supply on for a situation like this. The supply ramps up as it charges its output caps when you turn it on. But hot plugging immediately connects those charged caps to the input of your board which will cause worse transients.

    If this is the case you can remedy it by reducing inductance from Vsupply to VIN which can include shortening the connecting wires, move the VIN caps closer to the IC, and more. Let me know what you find and we can carry on from there.

    -Sam

  • Thanks for the quick reply Sam.

    I agree, it seems like an overshoot problem at VIN. I don't have an oscilloscope on hand to be able to verify that at the moment, but I think operating on that assumption is correct.

    The wires are already quite short between VIN and Vsupply (just a few feet) and the VIN caps are immediately adjacent to the VIN pin (just a millimeter or two). Perhaps larger caps would mitigate the issue? Currently using the WEBENCH recommended 2 x 2.2 uF.

    I'm happy to share photos of the board, but I don't see an option to upload attachments (possibly because I'm a new member) I also have a video of the most recent failure if that might be helpful.

    I just reviewed the video of the failure, and the board survived several hot plugs until the last one. On the last hot plug, the board was unplugged for only for 2-3 seconds (much shorter than the other tests). Perhaps the VIN caps retained their charge and where thus unable to protect from the initial onrush of current? Would a resistor across the VIN caps allow them to dissipate quickly and protect from the surge?

    Thanks,

    Patton

  • Patton,

    You can attach files by clicking the, "Insert Code, Attach Files and more..." button underneath your reply text box on the right. That will take you to a dedicated page with more formatting options.

    Please do share pictures of the board and the video. It would also be useful to see the schematic and layout if possible.

    2x2.2uF may be okay in theory but these are probably ringing with the parasitic inductance which causes worse overshoot. I'd recommend more capacitance or a bulk capacitor - electrolytic has more ESR than ceramic which will dampen this ringing effect.

    I would expect this issue to be worse when the caps are fully discharged (more time to charge means more energy in the parasitic inductance which means higher ringing overshoot). Your situation might just be that it happened to be the last straw and it happened to be unplugged for less time. I don't think adding a resistor to discharge the input caps will help this (plus it will hurt efficiency from that extra power dissipation).

    -Sam
  • Ah, I see the option now. See a number of files attached.

    I will add a large electrolytic capacitor the board and see if that resolves the issue. Is there a good rule of thumb for the proper size? I don't have any 100V on hand, so I'll have to place an order.

    Let me know if you see any other issues in the schematic or video that I missed. I removed the audio from the video as it just involves the fan and me swearing at the end. The failure in the video occurs at 40 seconds.

    Also, it's important to note that there is a resistor across the 6.5V line - the remainder of the board beyond the power supply has been disconnected to isolate this issue. On the schematic, this corresponds to just the power supply portion of the schematic up to C7.

    Patton

    Note: On this board, there is a linear regulator to bring the 6.5V down to a clean 5V. In this test, the linear regulator has been replaced with a large resistor. On the next draft of the board, we've decided to change TPS54360 to 12V for general use with a linear regulator for the 5V controller. Let me know if you believe this might affect operation significantly.

    Board Schematic.pdfBoard Top.pdfBoard Bottom.pdf

  • Reposting the video at a human scale.

  • Patton,

    Thanks for sharing this info. I got concerned when the video went black. I'm glad you're okay!

    You can add ~10 to 100uF bulk cap on the input. You already have 4.4uF from the ceramic so 15-22uF is probably good. But it doesn't hurt to order a bunch of caps to try them out.

    I'd also look out for inductance from the CAT5 cable to VIN of the IC (and back from GND pin of IC to CAT5 GND) but your layout looks alright in that regard. It may help to connect the IC thermal pad directly to the GND plane without thermal reliefs. Just a solid connection through the whole thing. But I don't expect that would cause your issue.

    A CAT5 cable will have a few hundred nH per meter. How long is that ethernet cable? That may be a significant source of inductance as well.

    -Sam
  • Quite alright. :) Just surprised at the end as it was my 10th or 15th test.

    The inductance in the wire is quite small - it's only about three feet long. However, in the final use case, we're planning to use wires up to 30ft long, so I will try a test with a longer wire once we solve the issue at this length.

    I'll pick up a few capacitors this evening and grab an oscilloscope for the next round of testing. I'll let you know what I learn!

    Thanks again,
    Patton
  • After doing some digging, it looks like an inrush current limiting circuit is the next best option if the larger capacitance doesn't do the trick.

    I mocked up this circuit which should prevent inrush until the capacitor is mostly charged (delaying the full power to the board about 0.05 seconds). Would this circuit interfere with the operation of the TPS54360? Are there other unintended consequences we should consider?

    Patton

  • Patton,

    My preferred option is to use a TVS diode in parallel with the electrolytic cap. This will keep VIN from overshooting too much and it won't hurt your efficiency.

    But inrush current limiting is an option. The only issue I could foresee is if the TPS543460 turns on and tries to draw current before the inrush limiter is letting through the max current. Then it'll pull VCC (VIN) down and it may never start up. You can reduce this likelihood by adding UVLO with a resistor divider on EN so the TPS54360 doesn't turn on until VIN reaches a specified VIN.

    -Sam
  • Ah, excellent suggestion! And much simpler to fit onto our already crowded board.

    I will perform a few more tests with the larger cap and the TVS diode to see if it solves our problem.

    Thanks again,

    Patton

  • Patton,

    Sounds good! Keep us posted. I'll click, "TI Thinks Resolved" and you can confirm if your problem is fixed.

    -Sam
  • For anyone viewing the thread in the future, the combination of a 47uF Electrolytic Cap and a TVS diode prevented the voltage spike and seems to have resolved the problem.