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.

LMZM23600: Unusually low efficiency causes devices to overheat, fail

Part Number: LMZM23600
Other Parts Discussed in Thread: LMZM23601EVM

I'm working on a space-constrained design with limited same-layer area for thermally sinking the 5V fixed version of the LMZM23600. We've now produced several dozen boards, each with 2 modules, and on ~10% of them, one or both of the modules run extremely hot, sometimes to the point of triggering thermal shutdown.

We're drawing nominally 160-170 mA (at 5V) from each module, and nominally supplying 35-50 mA at 28V, for efficiencies around 80% and on-module dissipations around 200 mW. We're nominally seeing ~20C temperature rises above ambient, for thermal resistances of 80-100 C/W, all of which feel reasonable and in spec. On the failing boards, we see no change in output current but dramatic increases in input current, 75-100 mA at 28 V, for dramatically lower efficiencies which heat the modules up much more (75-130 C above ambient). This doesn't happen on first power on but worsens over time, with failures occurring within ~10 hours of operation. Despite all of this the modules have no visual evidence of damage (other than darkened conformal coat due to the heat) and seem to be regulating effectively (until they overheat and trip their thermal protection), and actively cooling a module with compressed air allows it to continue functioning nominally. What could be causing this and what are useful tests to perform?

SCH:

PCB:

  • Hi Sasha,

    First I'd like to comment that the input capacitor should be placed closer to the VIN and GND pins (similar to how Cout is placed) of the device to reduce the high di/dt loop. This should significantly reduce the parasitic inductance because of PCB layout and result in lower output voltage noise.

    Aside from that, I would recommend you send in a few good and bad boards for FA. It sounds like the device may have been damaged which results in a higher input current and lower efficiency yield. Additionally, it would help if you can provide the top mark of the part so that my team can track the lot history of this device. 

    FA process: http://www.ti.com/support-quality/additional-information/customer-returns.html

    Regards,

    Jimmy 

  • Hi Sasha,

    May I get an update on this? I am interested to know the date code and top marking of these failed power modules. 

    Regards,
    Jimmy

  • Hi,

    All of the converters I have inspected (including a number that have not failed) have the codes "L4Q 928 4K A." I suspect all of our boards have components from the same reel.

    Please also advise on other tests we could perform to help troubleshoot this issue.

  • Hi Sasha,

    How many of these board have this issue? Ideally I'd like to have a good board and bad board (with the power module still mounted) sent in for TI's FA process to help further investigate the failure mode. This way the process can identify the device and inductor component's integrity to see if it may have been damaged,.

    Regards,

    Jimmy

  • We've seen three boards that we know to have failed, with three others which have not failed but which telemetry review shows to have abnormally low efficiency, consistent with the other boards which have failed. We've measured the inductors and they appear to be intact and functional, without a meaningful change in resistance.

  • Hi Sasha,

    Can you also check the switching waveform of the three boards that show abnormally low efficiency? Ideally the switch node should be a square wave with no oscillation. 

    The switch node can be easily probed by touching your oscilloscope's probe tip to the exposed top terminal of the inductor. Please make sure to not short the SW to GND during operation as this will damage the part.

    Regard,

    Jimmy

  • The failed boards are currently out for x-ray but I scoped what I believe to be a functioning board and got this waveform on one of the converters:

    The converter is somewhat lightly loaded and we have it in auto-PFM mode; please advise on if the above waveform is typical of this type of operation.

  • Yes this waveform is typical for light load operation. I've also taken a LMZM23601EVM configured for 12VIN/5VOUT at 80mA and got a similar switch node characteristic.

    At higher loads the switch node will square off and adjusts to your set frequency.

  • Hi Sasha,

    Did you get any results back from your X-ray analysis? The customer quality engineer will be returning from his time-off the later half of this week and can then check on the date code to trace back to any device issues during the assembly process.

    Please contact me directly at j-hua@ti.com to get the most up to date response.

    Regards,
    Jimmy

  • Hi Sasha,

    May I get an update on the analysis? If possible I would like to have a few failure boards sent in for TI FA as well to confirm failure mode. 

    Regards,
    Jimmy

  • Hi Sasha,

    Can you give me an estimate time on when you can provide the x-ray analysis? Also do you plan on sending in some failed boards for TI's FA?

    If not, what else regarding the LMZM23600 device do you have questions about?

    Regards,
    Jimmy

  • Hi Jimmy,

    I've attached some x-ray images below (we just got them back) labeled based on whether or not the converter failed. My teammates and I have reviewed and there doesn't appear to be an obvious difference between the failed and intact converters. However, we did notice some fairly significant voids forming under the module; do you know if this could exacerbate the problems we are seeing, and if so do you have any recommended layout/solder process changes to help eliminate the voids?

    Additionally do you have any data on how conformal coating can affect the operation of these modules?

    Thanks,
    Sasha

    TI_XRAY.zip

  • Hi Sasha,

    Solder voids under the power module would result in discontinuity/ no connection conditions that would generally result in a non-regulated output voltage. However, the issue you described sounds more like a potential EOS damage on the either on the part itself or on the exposed inductor since the output voltage is regulating, be it at a higher input current and lower efficiency. If not enough solder paste has been applied, there may be solder voids under the module. I would recommend following the solder paste example that I've attached below. 

    I will have to get back to you after talking with the apps team on your conformal coating question. 

    Regards,

    Jimmy  

  • Hi Sasha,

    I would like to again push for a TI FA on some of your failed boards to help come to a root cause analysis. As it is right now it is hard for me to determine what the problem could be without having it sent in for internal screening of the part. 

    As for the conformal coating question, can you provide what the type of conformal coating and the number of the conformal coating just so I can provide this information to a SMT expert.   

    Regards,

    Jimmy 

  • Hi Sasha,

    I have not heard back from you in a while. Inactivity of the thread will cause it to close automatically. Are you able to send in some failing units for TI FA as this will significantly help us come to a root cause analysis. Also can you provide an answer to my conformal coating question as this information is what the SMT expert will want to know before looking into this.

    May I contact you at the email address you provided in your E2E profile so that we can retain a direct form of communication and have any follow up conversations. The system will assume the problem has been resolved on your end due to inactivity. 

    Regards,

    Jimmy