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.

LM63615-Q1: Converter input short failure

Part Number: LM63615-Q1

We are using LM63615DQPWPRQ1 to supply two 12V axial fan. Converter input voltage is 24V. Total measured nominal output current for fan's is 600mA. Fan startup maximum current is 1000mA for short time period. Design done according WEBENCH tool. All products tested with production tester after manufacturing. Devices shipped to field over 5k. Currently failure rate is 2% and failure modes pointed to LM63615-Q1 to burned open or short. Fault distribution data shows that converter failure occured 10 to 140 days after installation. Most of devices still working without issues after 500 days.

Could you help or explain what might lead to this kind of failure?

I can provide more information if needed.

  • Hello

    Please provide images of your PCB layout.

    Altium files or .brd files or simple pdf images are OK for now.

    Thanks

  • 24V to 12V layout capture.

    12 hour full load thermal test image. Ambient 22 Celcius.

    Device is also HALT tested at 85C 1000hours, without issues.

  • Hello,

    Thank you for sharing the PCB layout and thermal images. In terms of your layout, feedback resistors R203 and R202 should be located closer to the FB pin in order to make the connections to FB and AGND as close to those pins on the device as possible. Ultimately I'm not sure if this is what is leading to the failure you are seeing. 

    Do you have details of when and how the device failed, perhaps the conditions (ambient temperature, input voltage, load current) at which the failure occurred? Please share any waveforms you've collected that show the failure occurring. I'm particularly interested in seeing the SW node, input voltage, and output voltage waveforms, as well as the load current. 

    Regards,

    Harrison Overturf 

  • It is true that FB resistor divider is placed in different way. FB net length is 6mm and it goes bottom side of PCB. PCB consist of four layer and middle layers are filled with GND copper.

    We try to replicate converter failure by adding a "antenna" wire between resistors to 1V FB net. The assumption was that disturbance in the FB net cause the failure. Result was that converter works with no issues or extra heat on full load. Conclusion is that FB disturbance is not root cause for this failure. 

  • Hello,

    Thanks for providing this information. Do you have any more details of when and how the device failed, perhaps the conditions (ambient temperature, input voltage, load current) at which the failure occurred? Please share any waveforms you've collected that show the failure occurring. I'm particularly interested in seeing the SW node, input voltage, and output voltage waveforms, as well as the load current. 

    In addition to that: what is providing the 24V supply? I'm interested in knowing the slew rate (dV/dT) of the 24V input voltage seen by the device. Is it possible that the input voltage seen by the converter exceeded the absolute maximum voltage rating of the part? 

    Are you able to place a known good unit of LM63615-Q1 on a board that has failed? If you are able to do this, is the part able to operate as expected? 

    Regards,

    Harrison 

  • Hello,

    All the devices are indoor conditions, ambient temperature 20C to 30C. 

    I have six failed units and I have removed converters from all to ensure that board 12V input or output are not in short circuit. I have not placed new converters on those six failed ones for few reason. Converter stock is zero in all common distributors, component is hard to solder due to center heat pad and I have no doubts about functionality as output and input are open (no shorted capacitors). But boards are stored for further investigations.

    24V supply voltage input filter:

    SW node with 1,5A output current

    Output and input with 1,5A output current:

  • Hello,

    I understand your reasoning to not put on the known good units onto the boards that have failed. 

    Have you measured the pins of the failed ICs to confirm if any pins are shorted together? 

    To confirm if the boards that you've removed the IC's from are indeed working, can you apply 12V at the desired output (between test points TP211 and TP212) and confirm that the circuitry downstream works as intended? 

    What does 24V input voltage waveform look like at the input to the buck when the board is powered on? I want to confirm that the voltage is not overshooting above the absolute maximum device rating when the board is powered on. 

    Regards,

    Harrison Overturf 

  •    Hello,

    First, resistance from Vin Pin 12 to GND pin 15 is 1 Ohm, measured from removed IC.

    Second, I supply 12V from bench power with jump wires to coil L204 second leg and circuit work as it should be. Fans start to spin at full RPM and no extra current consumption.

    24V Input measured in AC mode from last capacitor C212 close to Vin pins:

    Then I measure 24V input in a DC mode when powering the device:

  • Hello,

    It looks like the input voltage remains well below the absolute maximum input voltage rating of the device at power up. And the AC voltage waveforms taken across the input capacitor don't show signs of significant voltage spikes during steady state operation. 

    In your application, does the LM63615-Q1 remain on constantly or does it turn on and off regularly? I know you said the device is powering downstream fans so I wonder if there is a large voltage transient produced at the output of the buck when the device is turned off. If the buck output voltage exceeds the input voltage during normal operation, then this could cause damage to the device. Can you confirm this situation?

    Regards,

    Harrison Overturf 

  • Actually converter is always on. Most of the time two fan running on idle speed and converter is in pulse skipping/burst mode.

    24V Input AC ripple with idle load.

    Maybe 10% of time fans in the output running at full speed, total output current 0,6A.

    Power cut on Mains AC side is possible, but 12V output startup and power off during power cut look very clean

  • Hello,

    Thanks for sharing these, I understand that the converter is always on, but I want to confirm that in a situation where the input lines are cut that the output voltage does not exceed the input voltage. The output voltage shown in the last image seems to decay slowly, can you confirm that the input voltage always remains higher than the output during this situation? Ideally, I'd like to see the same waveform but with the input voltage included as well as the output voltage. 

    Regards,

    Harrison Overturf 

  • To be honest, I have never test this before, but here is the results about power cut.

    Yellow is 24V input from TP209

    Blue is 12V output from TP211

    Case 1. Power cut on full 630mA load (fans full speed)

    Case 2. Power cut on idle 21mA load (fans idle speed)

  • Hello,

    Thanks for testing this, as far as I can see the output voltage remains at or below the level of the input voltage during this turn-off test. My concern was that if the output voltage was every greater than the input voltage, it could create a condition where the body diode of the internal high side MOSFET (HSFET) would conduct. This body diode isn't meant to conduct so this condition could cause damage to the HSFET, this condition isn't present in the above images however. 

    Can you please share the inductor part number as well as the saturation current rating of this component? Have you performed short circuit testing (at the output) on this design? 

    Regards,

    Harrison Overturf 

  • We are using Bourns SRP7028A-6R8M

     

    I did overcurrent test with artificial load for 24V-to-12V converter. Green is output current and blue is output voltage.

    Converter is able to supply 1,9A current without issues and with 2,0A load output voltage drop and short circuit protector step in. Measurement result seems to be aligned with datasheet parameters.

    Output current (rms) with fans stay under 700mA, but output current is not very smooth.

    There are no feedforward capacitor across feedback Rfbt for improving transient response, but Rfbt value is 100k as recommended and there are no huge amount of capacitance in the output. Is it possible that 2% converter breakdown caused by FB transient response?

    I calculated that possible Cff values might be between 20pF to 50pF, but this will need layout change and quite hard to verified to root cause in this failure mode.

  • Hello,

    I can't comment on the root cause at this point because we haven't identified how the part is failing in the first place. 

    Let me consult some of my colleagues about this situation and get back to you in 1-2 business days. 

    Regards,

    Harrison Overturf 

  • Hello,

    I'd like to get some more information regarding the kinds of tests that you are performing during production. I'm interested in knowing the signals/voltage levels that are applied to each of the test points around the device.

    For instance, is there a point in testing the design where a voltage is applied to TP211 when TP209 is held low? 

    Regards,

    Harrison Overturf 

  • I need to gather more information about production tester, but according to my knowledge, Test Points are not used in the tester fixture at all. Converter functionality tested by a real fan connected to 12V output connector. 24V supply enabled and fan speed is read from the fan RPM feedback line by microcontroller on board. MCU communicate to tester via serial bus if the fan is spinning or not. I'm not sure if the tester measures 12V supply level from fan connector, but I will check that. There is only one fan in use and I think that tester will switch it for both fan output one at a time, but I will check that also.

    In a meanwhile I managed to get one new converter chip. I solder it to a one failed unit and unit start to work after the repair. I measured that input ripple is similar <200mVpp as other ones. I will keep power on for that unit and follow its status daily.

    I also check from another failed unit that input filter C207 100uF electrolyte capacitor capacitance is still 100uF and it is according to RLC meter. Then I removed whole C207 from one prototype and did some functional test and everything work as always. Even the 24V input ripple did not increase.

  • Hello,

    Yes, please gather more information on the production tester and how it interacts with the system. 

    In the meantime, it is good that you were able to secure one brand new unit and confirm that it functions on the board. If possible, can you probe the VIN, SW, VOUT, and IOUT and place a trigger on VOUT falling edge to capture the behavior if another failure occurs? Based on your previous data we don't expect a failure to occur for at least 10 days, correct?

    Have you been able to solder a failed device onto a known good board? 

    Regards,

    Harrison Overturf 

  • Hello,

    About production tester. There are only on fan inside the tester and relay for switching it to Fan1 and Fan2 output.

     

    Test sequence:

    1. 24V supply connected and activate CAN communication to MCU

    2. 24V supply current measured on sleep mode

    3. Connect K7

    4. Relay K5 and K6 connect fan to FAN1 output

    5. FAN PWM to 0% (idle, very slow speed)

    6. MCU measure RPM

    7. FAN PWM set 50%

    8. MCU measure RPM

    9. FAN PWM set 100%

    10. MCU measure RPM

    11. Relay K5 and K6 connect fan to FAN2 output

    12. FAN PWM to 0% (idle, very slow speed)

    13. MCU measure RPM

    14. FAN PWM set 50%

    15. MCU measure RPM

    16. FAN PWM set 100%

    17. MCU measure RPM

    18. FAN PWM set 0% (idle, very slow speed)

    19. Relay K5 and K6 connect fan to FAN1 output

    20. Disconnect K7

    Continue other test.

    Repaired board is currently on long term test and input, output, SW, output current is monitored with oscilloscope. I think we are not able to find 2% failure rate with this one setup no matter how long we wait.

    Failed converter on to a known good board I have not tested and I don't see it very useful test because some but not all converter packages are cracked.

    There are also some marks on failed converters about high current and heat on lead frame.


    -Mika

  • Hi Mika,

    Thanks for the detailed explanation of the test procedure.

    Step 11 is of interest, during this step the 12V is applied to the Fan but relay 6 opens which opens the GND connection. Without the GND connection the 12V will essentially be floating since there is not a GND connection. Also, just before this step the fan had been operating at 100% meaning full load current, and then this current is abruptly cut to 0. 

    Can you investigate what happens to the output voltage during the transitions between steps 10, 11, 12? Please provide images of VOUT, SW, IOUT during this condition. 

    Regards,

    Harrison Overturf

  • I have only Vout capture about fan hot swap with SW (ac mode).

    Voltage transient stay below input voltage with good margin.

    We are currently planning modifications to the production tester and add separate fans without any relays. 

    -Mika

  • Hello Mika,

    Thanks for sharing these, I've included an image with my assumption of the tester states:

    Does the device temperature increase during these transition points? 

    Regards,

    Harrison Overturf 

  • This capture is only from sequence 11 relay switch. Did you notice the Time/DIV 200us. 

  • Hello Mika,

    Thanks for clarifying. In the area I labeled as '10' the output voltage drops almost 4V in less than 40us (I'm assuming around 10us). Based on i=C*(dV/dT) this would require around 5.5A of current for the 3x4.7uF output capacitors voltage to discharge by that amount. Do you have the ability to probe the inductor current during this transition? 

    Also the SW node waveform approaches -5V when the device is switching, the absolute maximum rating is -6V for transients less than 10ns. This amount of negative SW voltage is not expected during normal operation. 

    Can you zoom in on the inductor current, output voltage and SW voltage during this transition sequence?

    Regards,

    Harrison Overturf 

  • Here is the biggest transient (label 10) zoomed from fan hot swap. Setup is not actual production tester, but similar. Real production tester located at Electronic Manufacturer.

    Yellow is SW node, blue Voutput and green is coil current.

    SW node seems to undershoot, but when I measure it separated, it looks better.

  • Hi Mika,

    Sorry for the delay, I've discussed this with my team and we have a couple questions that I'm hoping you can answer.  

    Can you remove a bad device and please send images of the DAP so that we can check for evidence of proper solder coverage? If the DAP is not soldered correctly then the part could be damaged. If you could send images of multiple devices this would be better. 

    If you take a known good board and good device and purposefully solder a device onto the board without solder on the DAP then this would approximate what would happen if the part doesn't have proper solder coverage. 

    Additionally, can you cycle the EN input while the device operates in steady state as it powers the fans? Ideally I'd like to see EN high, then EN low, then EN high again. The time to hold EN low as well as the number of high-low cycles can be varied but start by holding EN low for around 10ms. I would like to see waveforms of EN, VOUT, SW, and Iout, I'll be looking to see if there are any large transients on the output voltage. 

    Regards,

    Harrison Overturf

  • We have already suspected the production issues and investigated thermal pad soldering.

    Seems that groove and package epoxy prevent solder wetting over the center of the pad.

    I remove one unit and clean all solder away from thermal pad. Then I place polyimide (kapton) tape over the center pad by preventing the thermal and electrical connection. Surface temperature before modification was 40C (fans full) and after the modification temperature rise to 60C (fans full). That modified sample has now been running for over a month with no problems. 

    I test Enable pin by dropping it for 1ms, 10ms and 15ms, multiple times.

    Also 80% PWM tested for EN with 1ms off and 3ms on.

    Seems that there are no transients in output or SW node and supply start smoothly.

  • Hello, 

    Thank you for sharing the results of these experiments. Let me discuss these results internally and get back to you. 

    Have you been able capture a device failing or been able to induce a device to fail on your end? 

    Regards,

    Harrison Overturf 

  • No I haven't. It is very, very hard to hunt 1% failure rate by monitoring.

    We need to make new layout revision to design due to other reasons. Do you have some risk free improvement to implement in schema or layout? We could add it to this update sprint with minor effort. 

  • Hello,

    Understood, here are some changes you can make to the layout for the next revision:

    • Include the feedback resistors on the top layer next to the FB pin of the IC. Ensure that the trace connecting the center point of the resistor divider to the FB pin of the IC is as short as possible to minimize any noise coupling to this trace. 
    • Also include some vias to GND directly under the DAP of the IC, similar to Figure 11-2 in the datasheet. This will help with thermal management.
    • Consider including a solid GND plane on the layer beneath the IC.

    Regards,

    Harrison Overturf