nRF5340 fails to recover from intermittent AC power loss in high voltage mode (custom board)

[Cagney]

Hardware:

• nRF5340 on Fanstel BT40F module

• Custom board, not a DK

• High voltage mode: VDDH supply 5V, VREGHVOUT set to 3.3V

• DC/DC mode

• Power source: 120VAC mains → AC-DC converter → VDDH

Software:

• nRF Connect SDK v3.1.0

• Zephyr v4.1.99

• MCUboot, Bluetooth LE, USB CDC-ACM enabled

Problem: The device fails to recover from AC power outages that allow VDDH and VDD to dip but not fully arrive at ~0v. My opinion currently is we enter an unknown state when voltage dips as described. I would like the device to recover in any situation of power loss.

Possible solutions considered:

My current understanding is that POR and BOR are not adaptable to my problem as they are set in hardware.

Also, I am aware of POFCON but knowing the power is dying and firmware will end up in an unusable state doesn't seem helpful.

Trying to avoid adding new hardware but a voltage supervisor seems like a solution.

Questions I have:

Are there any firmware solutions I am not considering?

Also based on the image below, does the voltage drop on VDD and VDDH not being quick lead you to any ideas? Is that normal?

Image below shows AC 120v  supply, VDD, VDDH and processor activity as gpio toggle.

We cycle power off in longer increment of duration to find window of power off failure.

[Cagney]

Update:

I made a mistake in posting the 3.3V rail from the DC power supply. However, in high voltage mode the processor relies on internal component to create VDD at 3.3V and this is isolated from external 3.3V. Measuring the processors VDD shows much different issue. Notice the 3.3V does not fully recover.

[Kenneth]

Hello,

Intermittent power drop on the nRF5340 can cause the device to enter an unresponsive state. If VDD/VDDH dip without reaching a clean power‑on reset, internal PMU and regulator state machines may end up in an undefined condition that is not recovered by software reset. In this state, the device may consume power but execute no code, and only a full power on reset can reliably restores operation.

As a side note: This have been improved on the nRF54L-series (which have a level triggered power on reset, and don't rely on edge triggered power on reset).

Kenneth

[Cagney]

Thank you for response. This turned out to be an issue with my testing setup rather than the processor. However, the documentation says "The power-on reset (POR) generator initializes the system when the VDD supply voltage is above the
power-on threshold." 

But I think there is a very small blurb on one of the web pages that points out the rise time requirement you reference as edge detect. correct?

[Kenneth]

Looking more into the details I can find that the nRF52-series do define such a requirement, ref:
https://docs.nordicsemi.com/bundle/ps_nrf52840/page/recommended_op_conditions.html 

However there is no such requirement for the nRF5340, ref:
https://docs.nordicsemi.com/bundle/ps_nrf5340/page/chapters/recommended_op_conditions/doc/recommended_op_conditions.html 

So maybe you are right here, that there is indeed no such requirement for the nRF5340.

Kenneth