TPS25982: Unable to turn off despite pulling EN low pt 2

Part Number: TPS25982

Following up on this thread

"""""""""""""""""""""""""""""""""

Hi Jared,

One possible reason for this temporary latch up is negative transients at output. It is possible that uC is not catching them if its sampling is slow. These transients can be in microseconds time range.

  • Is this same board from the previous thread?
  • Is there any specific location which is failing, or different locations are all failing?
  • Is there any specific board which is failing or failing from different boards?
  • Did you try adding some output capacitor. This will help reducing the transients.
  • What were the take aways from the previous attempt in solving this issue. Did your colleague ("Aratrik") found anything which can help in solving this issue?
  • Fail rate of 15% (very high) signifies that there is some application related issue and not device defect. If you feel otherwise, you can share details (lot number) of failing IC. I get check if there is any issue with the lot. Also, can you share number of temporarily failed devices.
  • Did you try this with our EVM board? Does your board contain test points through which you can probe the signals and use oscilloscope to capture them.

Please understand that I am asking many questions is just because these types of fails are very difficult to replicate otherwise, I would have tried replicating this on bench.

Best Regards,
Arush

"""""""""""""""""""""""""""""""""

  • Is this same board from the previous thread?
    • This is relating to a different board than the previous thread. The previous thread's board (Design “A”) only had the fault occurring at or below -30*C (Issue A), but this board (Design “B”) has failures occurring at all temperatures we've operated it in (Issue B).
  • Is there any specific location which is failing, or different locations are all failing?
    • Yes, it is a single loadswitch of a set of 40 (for issue B)
  • Is there any specific board which is failing or failing from different boards?
    • We’ve seen failures on 2 of 4 instances (for issue B)
  • Did you try adding some output capacitor. This will help reducing the transients.
    • We made a test board with loadswitch configs from Issue A and Issue B. These loadswitches either have both output schottky and output cap, just output schottky, just output cap, or neither.
      • We have observed failures on loadswitches which have just output capacitors
        • Adding a schottky to this one did not prevent reoccurrence
      • We have observed failures on loadswitches with just output schottky
      • We have not (yet) seen failures with both schottky and cap
      • We have not (yet) seen failures with neither schottky or cap
  • What were the take aways from the previous attempt in solving this issue. Did your colleague ("Aratrik") found anything which can help in solving this issue?
    • Our current solution is to simply replace the loadswitch until the issue is no longer observed. (both issue A and B)
    • We’ve found that for Issue A, cycling at 1 second intervals will not excite the failure. 5 second cycle period does.
  • Fail rate of 15% (very high) signifies that there is some application related issue and not device defect. If you feel otherwise, you can share details (lot number) of failing IC. I get check if there is any issue with the lot. Also, can you share number of temporarily failed devices.
    • We also feel its unrelated to lot number. We've used devices from a number of lots and have not seen observable differences of behavior.
    • For Issue A, we’ve seen 13 failures of loadswitches across 20 boards with 10 loadswitches each. We’ve never seen failures on 4 of the 10 loadswitches on a board. The 4 which have never failed drive on-PCBA loads while the other 6 drive off-board loads.
      • An untracked quantity of replacements failed after replacement. This is probably about 30-50% of total replaced.
    • For Issue B, we’ve seen 2 failures of loadswitches across 8 boards with 20 loadswitches each. The boards are used in pairs and has only occurred with a specific load (loadswitch 3 of board ‘2’ of the pair, never board ‘1’)
  • Did you try this with our EVM board? Does your board contain test points through which you can probe the signals and use oscilloscope to capture them.
    • We have removed ICs which have failed and soldered them to the EVM board, where we continued to see the failure occurring (Issue A)
    • We have not used the EVM board with Issue B
    • We do have testpoints for this purpose
  • Other thoughts/comments
    • Issue B has only been observed but never successfully reproduced in the lab. It can be prevented from occurring by disconnecting the load’s chassis connection from the rest of the system’s chassis connection. The loadswitch ground is tied to chassis at the board enclosure
      • Current waveform of chassis on turn-on event
  • Current waveform of first stuck cycle, Issue B
    • ~1.88A while on
  • Voltage waveform of issue B