• State Resolved
  • Locked Locked
  • Replies 15 replies
  • Answers 2 answers
  • Subscribers 64 subscribers
  • Views 1554 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

TPS54494: Part Damage at almost full load currents on 3.9A CH1 and 1.9A CH2 on EVM

Rushi Gajjar
Intellectual 765 points
Part Number: TPS54494

Hi,

I have recently got the TPS54494-EVM for the evaluation of the Dual Channel buck and got it working with little success. There are no discrete level alterations on the EVM board done. All tests are performed with Resistive Load and at Room Temp=25C.

The observations are as follows:

Case1 (No Load)

Constant Vin=12V

Vout1=3.3V. Iout1=No Load

Vout2=1.5V, Vout2=1.5V. Iout2=No Load

The Output voltages are as expected.

Case2 (Single Channel Load)

Constant Vin=12V,

Vout1=3.3V. Iout1=No Load

Vout2=1.5V, Iout2=~1.9A (Rload=800mR)

Performs as expected.

Case3 (Loads on Both channels)

Constant Vin=12V, Vout1=3.3V. Iout1=~3.9A (Rload=800mR)

Vout2=1.5V, Iout2=~1.9A (Rload=800mR)

Both the channels are connected to resistive loads at the same time and upon powering up the Evaluation kit, the TPS54494 IC gets smoked.

Case4 (Loads on Both channels)

Replaced the TPS54494 on EVM Kit.

Constant Vin=12V, Vout1=3.3V. Iout1=3.9A (Rload=800mR)

Vout2=1.5V, Iout2=~1.9A (Rload=800mR)

The IC damaged badly. (Smoke, IC Bulged)

Case5 (Connecting loads one by one)

Replaced the TPS54494 on EVM Kit.

Constant Vin=12V

Reducing the Load Currents.

Step1: Kept the Ch1 on No Load. Vout2=1.5V. Iout2=~1.1A (Rload=1.3R) Vout1=3.3V. Iout1=No Load. The IC performed as expected.

Step2: Engaging the (1.3R, 2.5A) Load on Ch1. The IC damaged badly. (Smoke, IC Bulged) Checking the Impedances on the Output voltage rail: Ch2=21.5K, Ch1=<1Ohm.

I have made sure that in any case, the output rails are not getting shorted to each other, keeping both the loads separate.

Simulating the same on the Webench Simulation, the Simulator shows only one channel in the design. But, EVM guide shows that it's capable of driving simultaneous full load currents. Seeing this level of failure, Is the TPS54494 designed for the simultaneous Dual Channel operation of total 6A (Ch1=4A, Ch2=2A) on both the channels? Why the thermal protection or OCP didn't trigger and directly caused it to damage badly?

  • 0
    TI__Mastermind 19080 points
    Hi Rushi,

    Seems the first test on original EVM triggered Over Temperature Protection.
    Then you changed the device, did you do it manually? If so, thermal pad below the device may not be connected well.
    Another concern is the load resistor, it's also a thermal source, which will improve the ambient temperature significantly.
    BTW, what do you mean by IC damaged badly? Could you tell more detail, such as VIN short to SW, or to GND? Thanks!
    I will do the same test on EVM after a while.

    Regards,
    Sam
  • 0 in reply to Sam(MPAA) Zhang
    Intellectual 765 points
    Hi Sam,
    It was done manually, but I have made sure that the thermal pad gets soldered well with the PCB (using the solder paste). I would like to rule out this possibility as it happened on the stock assembled PCB (Original EVM) as well (described in Case 3).

    The Load resistor is a trusted 250W air cooled load which was placed relatively far from the EVM (also compensated to the wire resistance); This would definitely not affect the ambient temperature.

    During the tests, As I have used the Ti's EVM board,
    I have made sure not shorting the the VIN to SW or any such mis-handling. By the damage I meant that the IC smokes and starts getting bulged from the top.

    Sam,
    In case 1, how does Original EVM seem to be triggering Over Temperature Protection? Did you mean in Case 3? If yes, It would have protected the IC from getting smoked.
  • 0 in reply to Rushi Gajjar
    TI__Mastermind 19080 points
    Hi Rushi,

    I didn't find TPS54494-EVM at hand...
    Is there any possibility that the problem is caused by the power supply in previous stage? Can you help to probe the VIN pin when you add full loading? I mean sample the waveform at the moment full loading is applied. Thanks.

    Regards,
    Sam
  • 0 in reply to Sam(MPAA) Zhang
    Intellectual 765 points

    Hi Sam,

    Meanwhile, can you please clarify the questions I had asked in the first post? Copied here:

    1. Is the TPS54494 designed for the simultaneous Dual Channel operation of total 6A (Ch1=4A, Ch2=2A) on both the channels?
    2. Why the thermal protection or OCP didn't trigger and directly caused it to damage badly?

    Question 2 here might be related to Vin as well. So checking Vin behaviour will help us. As asked by you, Let me check the Vin behaviour at the time of loading the EVM board and furnish it here.

    Regards.

  • 0 in reply to Rushi Gajjar
    TI__Mastermind 19080 points

    Hi Rushi,

    1. TPS54494 is designed for operating with dual loading at 4A and 2A. It has the capability to output 4A current on CH1, and 2A current on CH2, simultaneously.

    2. I'm suspecting the part had been damaged by some reason before OTP is triggered. Let's find out why.

    Do you have another EVM? If so, please do the VIN test on a new EVM.

    Thanks.

    Regards,

    Sam

  • 0 in reply to Sam(MPAA) Zhang
    Intellectual 765 points

    Hi Sam,
    I have been doing extended tests with the EVM. The major observation which I found was that the 0Ohm (R5) resistor connecting to VBST1 pin to the Capacitor was having a dry solder which stopped the supply to high-side NFET gate drive circuit during operation, which caused the Output to go unstable. I am still wondering why the IC didn't help itself with Thermal Shutdown. Have you seen such a failure mode where the VBST pin not connected to capacitor damages the Buck badly? Awaiting Response.

  • 0 in reply to Sam(MPAA) Zhang
    Intellectual 765 points

    Sam,

    Awaiting your response to the previous query. 

    "Have you seen such a failure mode where the VBST pin not connected to capacitor damages the Buck badly?"

    Need to know whether it's a legit failure mode. 

    Please be closing. Thanks.

  • 0 in reply to Rushi Gajjar
    TI__Mastermind 19080 points

    Hi Rushi,

    When the VBST pin is not connected well, logic of the high side will be in wrong status. Part behavior may be unpredictable, so it's really hard to judge the failure mode.

    Any way, I will do the same test with you once I get the EVM (probably in about 2 weeks), and give you some feedback.

    Please connect the VBST pin well, and if you have any further question on the part, don't hesitate to contact me. Thanks!

    Regards,

    Sam

  • 0 in reply to Sam(MPAA) Zhang
    Intellectual 765 points

    Thanks, Sam.

    Please do the test and update me here. As a part of failure analysis, multiple observations and listing repercussions will help us take different measures in the design. Thanks.

  • 0 in reply to Rushi Gajjar
    TI__Mastermind 19080 points
    Hi Rushi,

    I've received the board and performed full loading test. I kept it running for 30 minutes. It worked well and the IC temperature went up to about 90 degree.

    Regards,
    Sam
  • 0 in reply to Sam(MPAA) Zhang
    Intellectual 765 points

    Hi Sam,

    Thanks. I have been running the buck on different ambient temperatures and with various loads on both channels.

    1. Did you see that reaching 90C at Room Temperature of 25C?
    2. Could you please check removing the R5 (0R on VBST1 pin) and behaviour of the buck on full load? If the buck goes unstable, it should ideally thermal shutdown instead getting damaged. But What I observed was described in the first question. Please share the behavior.

  • 0 in reply to Rushi Gajjar
    TI__Mastermind 19080 points
    Hi Rushi,

    1. Yes, it was at room temperature of 25C.
    2. I removed R5 and did the test again. With no loading, it works normally, but when applied 3A loading, the part got damaged. HS MOS was short. It's definitely a wrong configuration, because there would be no charging path for Cboot. The voltage across Cboot is very important as it's the power supply for the high side logic.
    If there is any further question, please let me know.

    Regards,
    Sam
  • 0 in reply to Sam(MPAA) Zhang
    Intellectual 765 points

    Hey Sam, Thanks for the experiment

    As also observed by you, Buck is not protecting itself from the Thermal Cutoff/Shutdown in this scenario and directly causing the failure, it should be highlighted to the team working on the part design revision to at least protect the damage of the part in the upgraded design if any. Even if there is minor abuse to the board (such as flexing) can crack open the Ceramic capacitor connected to VBST and hence the device failure can cause serious issues.

  • 0 in reply to Rushi Gajjar
    TI__Mastermind 19080 points
    Hi Rushi,

    Actually, I think the disconnection of boot capacitor could cause the part damaged.
    I don't think we can make a damaged part working normally to protect itself...
    So please make sure the boards are designed and soldered well, and our part will work well.
    Any further question about this part, please let me know.
    Can we close this case now?

    Regards,
    Sam
  • 0 in reply to Sam(MPAA) Zhang
    Intellectual 765 points

    Hey Sam,

    I didn't mean to convert damaged part into working. I meant that the normal working part in a normal scenario should at least protest itself from getting burnt/damaged in this discussed failure mode if TI puts a well-thought IC design specifically for this TPS54494. Now, the VBST capacitor becomes supercritical in the part selection process because the whole reliability of the system now depends on a single capacitor connected on a VBST pin. Any product designer wouldn't want their product to put on a fire just because the power-buck failed to protect itself from Thermal Shutdown and include costly measures for its protection. I would definitely like to raise this to the TI design team to consider this failure on a serious note.

    Yes, we can close this issue.