Is 5-6V safe on non-ADC IO pins?

[kbaud1]

We powered an ESP32 with 3.3V, and we programmed every IO pin to slowly cycle high/low in open drain output mode with pullups and pulldowns disabled. We then checked for a voltage drop across a 3.3K resistor connected from 5V to each pin. Of course we saw about 5V when the pin switched low, but when the pin switched high the results varied. Some pins (pins 27 through 37 sequentially on the VROOM) pulled no current through the resistor. The other pins pulled the resistor down various amounts from 0.9 to 1.35V. In looking at the pin functions in the datasheet, the pins that pulled no current exactly correlate with the pins that have no ADC or RTC IO function. This leads us to guess that pins with ADC function probably have intentional or parasitic diodes to the power rails that the other pins don't have.

We increased the voltage on the resistor to the non-ADC pins and found a diode effect to eventually kick in. Even with varied current, the pin V was 6.3V (3V above the 3.3V supply). This seems like a strange V drop, but it's what we measured from all the pins.

Now we're wondering what over voltage/current is safe on these non-ADC pins. Obviously the ADC-capable pins can't be subjected to voltage much higher than the supply (perhaps 0.3 to 0.6V over) without current flowing into the pins in a potentially risky way. But what about the other pins that seem to pull no current at up to 3V above the supply? It seems they could see 5V and still have a safety factor of about 2? Or could the non-ADC pins actually be at higher risk of damage due to a lack of some kind of lower voltage clamping mechanism?

[ESP_Sprite]

Short answer: No. Datasheet says voltage on any IO pin can be maximum Vdd+0.3V. Going above that is undefined: it may work, it may not work, it may blow up your chip, it may install a virus on your PC, it may feed chocolate to your dog. In other words: do so at your own risk; we at Espressif are not responsible for any damage to your ESP32, dog or computer.

(Leaving the jokes aside, from what I understand there's actually some quantum tunneling-y effects going on that can degrade the chip if voltage is applied that is over those specs. It may be okay for a hobby project, but I wouldn't do it in production.)