Wrong measures with ESP32

[supcik]

I tried the HX711 with CircuitPython and I got some strange results (on an ESP32 chip). So I analyzed the signal and this is what I found.

Sometimes, the clock signal (in orange) stays too long (almost 80µs) up at the start of the sequence. This is longer than the maximum allowed of 50µs (T3)

In this case, the data stays at 1, and the result is 0xFFFFFF and this is easy to detect.

However, sometimes, the clock stays high in the middle of the sequence:

The data line stays up after the long pulse, and the measure is wrong (and not easy to detect).

I understand that Bit Banging a signal with a time constraint might be challenging with CircuitPython. This could perhaps be solved by disabling the interrupts (https://docs.circuitpython.org/en/stable/shared-bindings/microcontroller/#microcontroller.disable_interrupts) and doing active polling instead of using time.sleep().

Let me know if you want me to investigate further and come up (if possible) with a PR.

[FoamyGuy]

Thanks for reporting the issue @supcik. If you are interested and willing to look in to this further please feel free and definitely submit a PR if you come up with way to improve things.

I would probably lean towards active polling by waiting with a while loop and checking time.monotonic_ns() if possible instead of disabling interrupts. But honestly I have no experience with that disable interrupts API, I do not know what all other stuff it would effect.

I do think ultimately code running under CircuitPython has no realtime guaruntee so there could still be cases where we do not meet the timing requirements i.e. if garbage collection were to occur I think it could get the pin stuck in it's state for extra time.

[supcik]

Thank you for your reply. I will do some experiments to come with a solution.

[supcik]

I was able to produce a clean signal. The clock (in white) respects the constraints given by the datasheet, and I tested with many thousands measures without a single error.

I plan to propose a PR next week during the hacktoberfest. Would you mind tagging the repository with the "hacktoberfest" topic ?

Thank you in advance.

[FoamyGuy]

We have hacktoberfest added as a topic on the repository itself so all PRs during october will count toward hacktoberfest.