Unnecessary use of 74AHCT125N buffers for the Para2P adapter?

[xaviergr]

I am quite confused with the various BlueRetro adapter schematics and the heavy use of the 74AHCT125N buffer chips. Why do we need the buffers at all if you can program the ESP32 GPIOs as open drain outputs?

Before discovering this project I was using RetroPad and my own sketch with an Atari 2600, and driving the controller port directly from the ESP32 seems to be working fine. Unfortunately in RetroPad, latency can become a big problem due to not being able to isolate the polling core from interrupts. Thankfully BlueRetro solves this issue nicely and there are no lag spikes due to activity coming from the other core.

Am I missing something? Could it be that there are other consoles that use the Para2P design that require the extra buffers? The project is great by the way, I am just interested to know more about the design decisions.

Also for inputs, ESP32 GPIOs are 5V tolerant so the hardware adapter design can be further simplified by skipping some of the input buffers. SNES clock/latch signals to the ESP32 don't need the extra input buffer for example (tested with BlueRetro too).

[darthcloud]

At the end of the day you do what you want, but the ESP32 is not 5V tolerant.

From espressif:

The official answer is that none of the IO-pins are tolerant, as you can read in the ESP32 datasheet, please do not go making products that involve feeding any 5V source into the chip without proper level shifting. Inofficially, however, there's not much that seems to go wrong if you use a proper current limiting resistor (10K or something) between the 5V source and the I/O pin, although this still can lead to long-term degradation. Just don't use that trick for production or something you need to depend on.

https://esp32.com/viewtopic.php?t=18327

[xaviergr]

At the end of the day you do what you want, but the ESP32 is not 5V tolerant.

Sure, I can't argue much about the 5V tolerant point since I cannot find any official source on the matter. Although there is a counterpoint according to this and this, where it is mentioned that ESP32 is 5V tolerant (no official sources though, just forum posts about the espressif CEO confirming it in a facebook group post).

In the end you are right though, for production it's better to be safe than sorry.

What about the first point point about open drain outputs, do you see any problem with that approach? By the way I am not saying that the current design is problematic or anything of that sort. Just curious to know if there could be a potential problem not using the output buffers.