ESP32-WROOM-32U EN pin pulled low internally?

[ZeroxCorbin]

Hello,
I have attached a schematic of a board design using an ESP32-WROOM-32U module that, after being powered on for about five days, stops functioning and becomes completely unresponsive. No output from serial at boot.

Measuring voltage at the EN pin over the five days shows a slow drop in voltage until it reaches zero (0) volts. It remains at zero volts after a power cycle. The 3.3v LDO, on the main board, stays at 3.3v when measured before the 10k resistor.

The only way to recover the module is to heat it up using a hot air gun to around 200 C. Yes, heat it up and let it cool down. Then it starts working again. We have tried many other ways to recover the module without success.

The problem does not occur on power up, however, we scoped the 3.3v LDO on the main board and the rise is as expected, smooth. A second probe on the EN pin shows a small RC delay and a smooth rise.

We have tried various sample code to test the board and they all eventually fail. Some use Bluetooth and some use WiFi. Some do not activate the antenna. But they all eventually fail.

We have no other ideas to try. It seems to be a design flaw in our board somewhere, but we can not figure it out. Any suggestions are very welcome

Thank you for any advise you may have.

[ESP_Sprite]

Interesting. Do you only have one prototype of the board that displays this behaviour or are there multiple ones that all do the same? Also, how were these PCBs produced? Manually assembled or by machine? I'm thinking there may be something odd going on like flux residue that gets conductive with moisture ingress, or perhaps a C9 that is cracked but 'comes good' when heated.

[estratos]

What you mean with a C9 that is cracked?, I have a custom board that have this behavior too

[ESP_Sprite]

What you mean with a C9 that is cracked?

I literally mean that that (presumably ceramic) capacitor has a physical crack in it, either due to out-of-range temperatures at soldering or due to stresses to the PCB that were conveyed to the capacitor, causing it to crack right in two, but it's still held in place by the solder pads. What then happens is that the internal layers can get 'misaligned' and make contact, causing a short. As this short effectively is caused by the halves of the capacitor rubbing against eachother, the short is very sensitive to temperature changes.

Things like this don't happen often, but they're known to happen. Great fun if a decoupling cap shorts like this.