L7983 - Low-side MOSFET failure (LX shorted to GND)

[Max_L1]

Hello,

We are investigating a recurring failure of an L7983 used in a medical device.

Application:

• VIN ≈ 40 V
• VOUT = 5 V
• L = 68 µH
• Cout = 2.2 µF
• LCM mode
• Very light load conditions possible

Failure mode:

• 5 V output lost.
• After removing the output inductor, LX remains shorted to GND.
• Resistance LX-GND ≈ 2.5 Ω.
• High-side MOSFET appears intact (no abnormal resistance between VIN and LX).
• Failure seems limited to the internal low-side MOSFET.

Field observations:

• The failure is always preceded by a battery-related error.
• Batteries are often found degraded or unable to charge correctly.
• Similar failures have also occurred during production.

Tests performed:

• Startup without battery.
• Startup with low battery conditions.
• Long-duration operation with low battery voltage.
• Heavy battery discharge during operation.
• Various startup sequences.

No test has reproduced the failure so far.

Measurements:

• LX peak ≈ 50 V.
• Undershoot down to approximately -1.7V
• No obvious overcurrent event detected.

Questions:

1. Have you seen failures where only the low-side MOSFET becomes shorted?
2. Could repeated reverse-current events or output energy fed back into the regulator progressively damage the low-side MOSFET?
3. Are there known failure mechanisms associated with LCM operation at very light loads?
4. What stress tests would you recommend to reproduce this type of failure?

Thank you for any suggestions or similar experience with the L7983.

[Peter BENSCH]

Welcome ​@Max_L1, to the community!

Thank you very much for the detailed description of the fault.

Your indication of -1.7V for the undershoot already shows a massive level of stress well outside specification, since according to the datasheet, table 2, only -0.3V is permissible at pin LX. If you cannot reduce the undershoot with suitable layout measures, you still have the option of placing a suitable Schottky diode from GND to LX. If you can insert a picture of the layout section around the L7983 here directly, preferably also as a GERBER file, it could be reviewed.

Regards
/Peter

[Max_L1]

Hi ​@Peter BENSCH ,

Thanks for your feedback.

I repeated the measurements using a spring ground instead of the standard ground lead, and the undershoot is significantly reduced. Adding a snubber also helps reduce it further.

I agree that adding a Schottky diode from GND to LX could improve the signal even more and is something we may evaluate.

One interesting observation is that I see a very similar behavior on the ST evaluation board as well. The undershoot level is slightly the same and the overall waveform shape and behavior are comparable.