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.

TPS22920 design issue

Other Parts Discussed in Thread: TPS22920, TPS22970

The form did not accept the TI p/n for some reason. We’re experiencing a repeating problem on a new design. I have attached a schematic

 

The design uses two (U38 & U57) load switches (TPS22920YZPR) from TI, they share a common enable.  Each switch supplies 3.3V to two QSFP cages.   The cages are empty.   The only DC load is the recently added 332 ohm bleeder resistor.

 

We have had several U38 & U57 fatalities.   We thought we found and fixed the problem.  The 3.3V input would ramp down quickly.  The unloaded (cages are empty) output bulk caps discharged back through the body diode of the pass FET to the power input pin and the enable input pin.  We added bleeder resistors to the output and confirmed the output ramped down faster than the input – so the body diode wasn’t conducting anymore at power off.

 

We added the bleeder resistors and changed U38 & U57 in case the ICs had been stressed by repeated power cycles.    Well, the new U57 now dies after being powered up for a couple of minutes. 

 

Here’s a snippet of schematic.  U38 & U57 circuits are identical.    The VSW_SYS_3V3 input rail has 100’s of uF in bulk decoupling.TPS22920YZPR.pdf

  • Here’s the scope capture of the power supply monitor detecting the failure and shutting down the two load switches:

    Here’s the confirmation that the ENABLE to U57 (and U38) operates at power-up (this shot was taken after the shot above).  Everything looks good:

    I measured the unpowered output to GND resistance of four switches on two boards:

     

    U57 on bd 17 < 1 ohm to GND.         U57 dies.

    U38 on bd 17 = 330 ohms to GND.   That is the bleeder resistance (332 ohms nominal). This is the expected resistance value.

     

    U57 on bd 14 = 4 ohms to GND.      So U57 is dying?

    U38 on bd 14 = 3.9 ohms to GND.   So U38 is dying?

    Thanks,

    Ken

  • Hi Ken,

    This last post didn't have any attached scope shots, could you please re-upload?

    If my understanding is correct, then these two TPS22920 load switches were initially failing because VIN would drop quickly and VOUT would drop slowly. This caused VOUT>VIN and reverse current was being put through the body diode, resulting in failure. To fix this, you put a pulldown resistance on the output of the switches.

    Now, U57 is failing after being on for some time. A few questions on my end:

    1. What is the failure mechanism of the device? Is it the increased pulldown resistance or is the device not turning on anymore?
    2. How are you measuring the output pull down resistance? The device is not meant to withstand a DC load on the output, only transients from discharging the load.
    3. Is the inductance on the output causing any negative transients? It would be helpful to have a scope shot of the output during power up and power down.

    Thanks,

    Alek Kaknevicius

  • Hi Alek,

    Yeah, I could not figure out how to post my scope captures.  This web site isn't too intuitive.

    What is the failure mechanism of the device?   After minutes the device appears to turn itself off slowly and the output voltage drops off.  I can cycle system power and the device restarts fine.  But then it turns itself off.

    How are you measuring the output pull down resistance?  I power off the system and use my DMM to measure resistance across the output caps / bleeder resistor.

    Is the inductance on the output causing any negative transients?  No.  The input ramps up smoothly in about 1 millisecond.  Since the only DC load is the bleeder resistor, there is no DC inductor current and no inductive kickback at shut-off.

    Thanks.  I'll keep trying to post my graphics.  

    Ken

  • Alek, Any update for us?

    Ken

  • Hi Ken,

    Have you monitored the ON pin voltage to make sure that the voltage on this pin has not dropped below the VIL level? If the ON pin is high and VIN is high, then it seems odd for the device to begin turning itself off.

    When you say you "power off" the system, do you also remove the VIN voltage? VIN is necessary to enable the QOD and measure the RPD resistance.

    As for the scope shots I don't see anything concerning.

    Thanks,

    Alek Kaknevicius

  • Hi Alek,

    The ON pin is driven from 1.8V logic.  It is marked on the scope shot I previously posted.  It looks good to me. 

    Yes, I turn off the main 12VDC supply.  The 12VDC feeds the 3.3V VIN switching supply and the 1.8V switching supply to the control logic is off.

    We replaced the IC and left it to run overnight.  It shut itself off sometime overnight.  When we cycle power now, we see the ON pin only rises to about 0.5V.   The 100 ohm resistor in series with the ON is dropping over 1V or 10mA. 

    Thanks,

    Ken

  • Alek,

      Do you have any specific rework instructions for this part? We need to be confident that we are not thermally damaging the parts when they are reworked. I can see from another supplier (same function & package) that they provide a max spec on their datasheet of 300C for 10-seconds. I could find no similar guideline from the TI website.

    Regards, Ken

  • Alek,

    Another engineer on the project found this note on the datasheet about cracking.  We'll switch to the backside coated version.

    Ken

  • Hi Ken,

    Apologies for the delay in response, and I'm glad to see that you have found the TPS22920 with backside coating. We also have a "next generation" version of this device called the TPS22970 which would also be a suitable replacement here. This device is similar to the TPS22920 but also comes with added thermal protection.

    One other variable to take into consideration is the pick and place (nozzle) force. For an 8 ball CSP, we recommend that the force is limited to 240gf or below.

    Thanks,

    Alek Kaknevicius