A PIO driver for DHT sensor
Find a file
hodasemi cf98999e96
All checks were successful
Check Build / Test build (push) Successful in 35m25s
Rework timeout
2024-08-12 09:38:44 +02:00
.cargo First commit 2023-09-08 23:09:15 +02:00
.forgejo/workflows Rename action file 2024-08-12 08:22:11 +02:00
.vscode Convert to embassy_rp 2024-03-13 16:46:52 +01:00
src Rework timeout 2024-08-12 09:38:44 +02:00
.gitignore First commit 2023-09-08 23:09:15 +02:00
Cargo.toml Remove example 2024-08-06 17:15:52 +02:00
LICENSE Create LICENSE 2023-09-09 14:23:18 +02:00
memory.x Add example 2023-09-21 18:45:47 +02:00
readme.md fix typo 2023-11-05 15:11:23 +01:00
readme_fr.md Add readme section about Dht22Type2 2023-11-05 15:09:18 +01:00
renovate.json Add serde 2024-07-10 22:01:44 +02:00
rust-toolchain.toml Add toolchain file 2024-08-06 08:15:41 +02:00

DHT PIO Rust Library

crates.io MIT GitHub

french version

Why?

The DHT (22 or 11) uses a 1-Wire protocol, which is not compatible with the Dallas Semicondutors protocol of the same name. The Raspberry Pico (like other microcontrollers) has no dedicated peripherals for this protocol.

Numerous crates exist for using DHT via a digital pin, but after testing several of them, they don't work reliably. The main problem is the implementation of the embedded_hal by rp2040_hal. Manipulating the state and direction of a pin takes too long (I measured between 2µs and 6µs depending on the action requested). This is due, among other things, to the impossibility of placing a pin in open drain, which requires "simulating" this feature

The PIO ❤️

The RP2040 chip (used for the Pico) has a rather atypical peripheral called PIO (Programmable Input/Output), Chapter 3 of the DataSheet. In simple terms, the idea is to be able to run a small program (max. 32 instructions), which executes independently. It can manipulate GPIOs and share information with the main program.

The PIO is programmed using an assembler called pioasm, with just a few very basic instructions. What's interesting is that each instruction takes (usually) 1 cycle to execute. What's more, it's possible to divide the clock at which the program executes. In our case, we divide the main clock of 125 MHz by 125, giving us one instruction per microsecond.

Usage

First, create and retrieve the PIO objects

let (dht_pio, dht_sm, _, _, _) = pac.PIO0.split(&mut pac.RESETS);

To create a new object:

  • DHT22
    let mut dht = Dht22::new(dht_pio, dht_sm, pins.gpio0.into_function());
    ```
    
  • DHT11
    let mut dht = Dht11::new(dht_pio, dht_sm, pins.gpio0.into_function());
    ```
    
    

Read data:

let dht_data = dht.read(&mut delay);

NB: read return a Result<DhtResult, DhtError>.

DHT22 type 2 🧐

It seems that there are two versions of DHT22. I haven't found anything really conclusive, but what is certain is that not all DHT22s have the same data format... In one case the format is the same as presented in (almost) all datasheets, i.e. the most significant bit is set to 1 if the number is negative, but the binary representation of the absolute temperature value is not changed. For example:

  • 0000 0000 0110 1001 = 105 or 10.5°C
  • 1000 0000 0110 1001 = 32873 or -10.5°C

This is how the Dht22 struct will "decode" the data coming from the sensor. However, I've come across sensors that don't work like this at all. But in a (ultimately) more logical way. Since the data is represented in two's complement. In this case, use Dht22Type2. For example:

  • 0000 0000 0110 1001 = 105 i.e. 10.5°C
  • 1111 1111 1001 0111 = 65431 i.e. -10.5°C

To simplify, if your sensor is a DHT22 but the values don't seem consistent (negative values), then try "Type 2" (and if nothing really works, open an issue 😉 ).

Support

Board

For the moment, the crates have only been tested on a Raspberry Pico.

DHT

DHT22
DHT11

TODO

  • Finish Readme
  • Add CRC read
  • Check CRC
  • DHT11 support
  • Test DHT11
  • Document code

Thanks

Geir Ertzaas (grukx), for actively discover (too many?) bugs.