3 times over perhaps 6 months of development, I've had the ESP32-POE lock up part way through boot. Hardware reset would reset the board, but it would stop at the same place. The only way to recover was a power cycle.
Has anyone else seen this, or have ideas why this might be happening?
I asked the Espressif community and one Espressif dev said he had seen this and suggested checking if reset times and power supply ramps are within spec.
We are designing a product (ATS) where it will not be easy to power cycle the ESP-32. If hardware reset would work, that would be adequate with external watchdog.
How does the lock up occur? Describe what happens before the boot exactly? Do you attach power to the board or how exactly do you initiate the boot?
During this lock up what exactly is connected to the board and where exactly?
Are you sure the whole board locks up and not just some part of it (like for example, is it possible that only Ethernet locks up)?
What kind of software are you using? Did you try to replicate the issue with other, basic software loaded instead?
Power is via USB port.
The Olimex is connected to a custom baseboard with a handful of components.
The problem happened during normal development -- flash, reboot, etc. Resetting or reprogramming the board did fix it, only power cycle.
The Zephyr serial console locks up. I did not test anything else since that is primarily how we interact with the system.
We are using Zephyr, which is supported on the ESP32 modules.
I cannot replicate the issue as it has only happened a few times over a very long time span, but experience tells me if we deploy more units, statistically it will show up if there is an issue.
Thinking of:
- verify power supplies are stable
- power cycle testing, where we cycle power and verify the unit comes up in an automated way
> The Olimex is connected to a custom baseboard with a handful of components.
Is this custom baseboard powered from the ESP32-POE or it has own power supply? Does it have any own ground? Sometimes if the custom board has own power supply it can parasitically power the ESP32-POE holding it in incorrect state.
Where exactly is the custom board connected? Which ESP32 pins are used? Maybe one of the ESP32 bootstrap pins gets affected by the custom baseboard and this affects the boot process.
> The Zephyr serial console locks up. I did not test anything else since that is primarily how we interact with the system.
If you can find a way to replicate the issue reliably it is good idea to test with other software tools, this would be good way to exclude any possible problems related to Zephyr.