ESP32-DevKitC-32UE Intermittent Boot Failure

[epsiloncomp]

Group,

I have a ESP32-DevKitC-32UE mounted to a custom PCB as the controller module. The core is powered by +5V regulated to the 5V pin and GND. The PCB runs at 5V logic and I have a TXB0104 and TXS0104 chips for 5v/3.3v interface.

The system also uses SPI and I2C. Both of these are buffered via the the TXB/TXS chips above. No pullups on SPI. I2C uses 2.4K pullups on the 3.3V side.

On one hardware application, the module fails to boot and initialize. Pressing the BOOT switch causes the 32UE to run successfully. Removing the module from the installed equipment and placing on a test bench removes the error and it always boots and runs. The board has 24 inputs connected to 3 (three) MCP23017 IO expanders via 27K pullups. Connecting or disconnecting a single input may also remove the error and the system will boot (when installed). This has no effect on the test bench.

Also changing to a new 32UE module may or may not remove the error.

Custom board power requirement when running are ~55ma at 12 VDC. The 5V regulated comes from a RECOM DC/DC module. Supplies 5V regulated @ 500ma, which should be sufficient power.

I cannot watch the debug messages via serial as plugging in the USB cable always removes the boot failure.

Programming via Visual Micro/Visual Studio 2019.

All of the following pins are NC. I assume that PU/PD exists in the module.

MTDI
GPIO0
GPIO2
MTDO
GPIO5

---

I am missing something simple and obvious. It reminds me of a floating pin, that is intermittent. Changing the physical environment changes the behavior.


Thanks David

[crgarcia]

Did you try adding some caps?

I have a project with ESP32 and MCP23017, powered from an external 5v PS and it works well:
https://github.com/crgarcia12/electroni ... /tree/main

I am not sure if this will solve your problems, but I added capacitors to the MCP in case it needs sudden current, and also one next to the dev board because it has a very little capacitance.
I exaggerated in my project, simply because I had those in stock, but I would consider adding a fat electrolytic in the 5v rail

I am saying all this because that could be the reason for the USB to solve the problem, it might be a more stable PS than the RECON

[ESP_Sprite]

Might be some weirdness with one of the bootstrapping pins having slightly too much capacitance... can you try adding an 10uF capacitor between EN and gnd?

[epsiloncomp]

Great suggestions. So I put my logic analyzer on 5v and 3.3v. These is a dip in the 5v when the ESP32 initializes.

And no, I do not have enough capacitance on the 5v rail. My bad, but easily resolved. Since I am building a REV2 of this board, I just may add a 2nd 5v regulator just for the ESP32.

Thanks a million for the quick replies. David

[epsiloncomp]

Group,

I added a 100µf cap to the 5v power plane, next to the 5v pin of the ESP32. The system now boots. I believe this will solve the issue.

Adding a cap next to the EN pin, just in case. Laying out the new PCB at this moment.

Many thanks, David

[epsiloncomp]

Well, I celebrated too soon. Still has the same behavior. I am adding a 10µf cap to the EN pin and GND.

DE