Custom ESP32-P4 PCB issues

[igkren]

Hello.

I am experimenting with the P4 SoC for a potential future product. I made a custom PCB following the Hardware design guidelines and the EV board schematics.
For some reason the P4 SoC does not work. After some analysis I found that the P4 SoC gets hot (to around 70 degress celsius) along with the HP domain buck (that converts 5V to 1.2V). Further testing showed that the buck does not receive the HP feedback signal from the P4 SoC and the HP buck outputs ~0.9V.
Also, I have exposed pins 52, 53 and GND to enabled flashing via USB peripheral. When I connect with a laptop the device is found but then lost again and then found again in an endless cycle.

Been playing around with this for a couple of days and was not able to find the culprit. The schematics seem fine, checked them multiple times. Any ideas what might be the cause of this? How to narrow down the issue? As far as I can tell everything else works as expected (namely the 3.3V buck and related circuits)

[Sprite]

I'd suggest you start with posting the schematic and possibly PCB design here.

[igkren]

Thank you for the reply. I will attach the relevant MCU schematics here.
Couple of notes:

• This is for my work so I cannot share the whole project for confidentiality reasons
• The problem is somewhere within the power circuit (specifically the HP 1.2V domain)
• For context, the MCU also has an Ethernet PHY (has it's own buck, same circuit and IC as buck_3v3.png) connected and 2 UART chips
• The PCB main power rail is 5V that comes from an externel PSU
• Ethernet PHY's and MCU's 3.3V buck work fine and are stable. The same buck used for 1.2V HP domain does not properly. I mentioned before that the MCU does not appear to give out the FB signal to the HP buck
• I suspected hardware chip revision could be the issue. My PCBs came with ESP32-P4 v1.3 (manufacturing code FEF0). This chip version should not have any HP buck feedback resistors (divider network). In buck_1v2.png you can the I had those resistors. I removed these divider resistors on another PCB (to isolate permanent damage from the divider network when it should not have been there) and then powered it for the first time but the problem did not disappear.
• I tried desoldering the MCU and the HP buck no longer overheated (of course if was disabled since the MCU did not give out the EN signal)
• Manufacturer of the boards was JLBPCB. Maybe that is relevant for manufacturer errors. I inspected all 5 received boards under a microscope and everything seems fine.

If you need more information let me know. I have access to lab equipment and can measure certain things if needed.