ESP32-P4 HP Core voltage issue

[nkp123]

Hi there,
I am facing an issue with stability of my custom ESP32-P4 + ESP32 board.
It seems to only work reliably with HP Core voltage between 1.20 and 1.27v. These thresholds varies a bit from board to board, however the range is much smaller than one permitted by documentation (1.0 - 1.3v). Tests were done using external power supply.
This makes using FB_DCDC as feedback for the DCDC converter impossible to use, as it targets ~1.1v. Even setting feedback manually is not a solution there because voltages can drift a bit (eg. from temperature changes).

In general outside 1.20v - 1.27v range the ESP32-P4 experiences random issues:
- do not even start executing main() function altogether, hanging after:

Code: Select all

I (168) esp_image: segment 5: paddr=00038574 vaddr=4ff0e880 size=01c90h (  7312) load
I (175) boot: Loaded app from partition at offset 0x10000
I (176) boot: Disabling RNG early entropy source...
I (189) cpu_start: Multicore app

- crashes on exception:

Code: Select all

Guru Meditation Error: Core  0 panic'ed (Load access fault). Exception was unhandled.

- reboots without any message
- hangs without any message

This is for esp-idf blink example. More sophisticated programs can crash on other exceptions, eg. PMP load access fault, invalid instruction. I have seen silent memory corruption as well.

Things I have checked so far:
- noise on HP core voltage line - ~40mv peak to peak
- noise on 3.3v line - ~60mv peak to peak
- capacitors - 100nF near each HP voltage input pin, 100nF + 1uF + 22uF near the DCDC converter

In general I am basing my design on Espressif's ESP32-P4-Function-EV-Board design, with unnecessary components removed and ESP32-WROOM classic module instead of C6 one. Both 1.0 and 1.3 P4 revisions are affected.

Do you have any idea what can cause such issues?

Best regards,
Norbert

[sreejithrnair]

Hello,

Is Espressif planning to produce ESP-32-P4 SoM? Currently ESP-32-P4 only has SoCs and Eval Boards. Will they make SOMs for ESP-32-P4 in the immediate future?

Regards

[nkp123]

@sreejithrnair It would be better probably if you have asked that question in a separate thread

With further testing, by experimenting with HP Core clock frequency I was able to lower it to minimum, which is 40 MHz - at that frequency I was able to run LED blink program with HP Core voltage down to 1.08v - still higher than specified in the datasheet. I'm slowly running out of ideas.

[nkp123]

Another observation - it looks like when I use DCDC converter and I add additional capacitor (which by documentation is not needed) between FB_DCDC and ESP_VDD_HP pins the processor starts to work fine. On the oscilloscope the voltage output of the DCDC converter gets more stable after that.
Are there some circumstances where this capacitor is needed when using FB_DCDC pin from ESP32P4?

As a side question - as I haven't seen that in the datasheet - in what HP Core voltage range I should expect the P4 to stably work at 360MHz clock? Maybe 1.0-1.3v range that I found earlier is for different HP Core clock frequency than max.

I am attaching reference schematics. The capacitor in question is C53:

FB_DCDC.png (73.45 KiB) Viewed 508 times

Best regards,
Norbert