ESP32-P4-PC: USB-JTAG instability and LT8912B I2C issue

[Rochus]

While tying to migrate the Oberon System 3 to the P4-PC board I have hit the following road blocks (not sure whether this is the right place for this or it was already reported elsewhere):

Rev.C, behaviour reproduced on two separate boards

Not using any of your software or drivers

# 1. USB Serial/JTAG interface drops and re-enumerates continuously

Only started after several flashes and tests

/dev/ttyACM* node then appears and disappears every few seconds for some time with stable phases (~10 minutes) in between

# 2. LT8912B I2C interface strange behaviour when HDMI is connected

The LT8912B is on GPIO7/SDA + GPIO8/SCL

Once a monitor is attached, the LT8912B cannot be (re)configured

If the monitor is attached/detached after a working init, the LT
often loses its HDMI transmitter state and the display stays blue until the LT is re-initialized

[LubOlimex]

What is the hardware revision of the board that you use?

[Rochus]

As noted, Rev. C. I bought them on 19.02.2026, order W-RK180226. What other information do you need?

[LubOlimex]

First it is worth testing with our code and see if the same thing exists with it. I specifically recommend testing with this demo that I published like two weeks ago (since it changes a lot of settings and timings around the HDMI):

https://github.com/OLIMEX/ESP32-P4-PC/tree/main/SOFTWARE/ESP-IDF/p4_board_demo

If you don't want to build it you can simply use the binary at 0x0. See if one or both the issues remain.

This hardware revision is pretty good it hugely improved start up issues that occurred with older hardware releases, due to different regulator and logic. Some boards from older hardware revisions had bigger issues especially when ESP32-P4 chip is erased. It is described somewhere in the forum (but for ESP32-P4-DevKit). So this is probably related to either the software or the way Linux handles DTR and RTS or there is silicon bug in the ESP32 chip itself, maybe check the errata sheet for ESP32-P4 v1.3 (the module your board uses), and specifically this issue:

https://docs.espressif.com/projects/esp-chip-errata/en/latest/esp32p4/03-errata-description/esp32p4/unauthorized-ahb-access-may-block-subsequent-psram-or-flash-transactions.html

If the issues remain with the code above, let me know how you test exactly (e.g. step by step) so I can reproduce and test it myself here.