AstroMech is an audio protocol for Arduino that is capable of sending small chunks of data via sound. It is the open source counterpart to https://chirp.io/. We use this audio protocol to exchange WiFi credentials between a Raspberry Pi WiFi Access Point and an ESP32.

AstroMech is a fairly simple one way frequency modulated (FM) audio protocol. There is a sender broadcasting an audio signal that can be consumed by one or multiple receivers.

Each frequency has a hexadecimal symbol assigned to it. Currently there are 16 distinct frequencies defined, which means that each sound encodes four bits. The protocol works asynchronously, thus sender and receiver agree on a fixed baud rate which is currently set to 20. This means that the sender makes 20 sounds per second. The protocol operates in the audible range of 600 Hz to 2850 Hz. Each symbol is separated by 150 Hz in order to avoid overlapping of the symbols. Each incoming sound is analyzed by a Fast Fourier transform (FFT) in order to calculate the matching frequency bucket. Because of the hardware demands of FFT, this library is more suitable for powerful hardware platforms like the ESP32. In addition to the frequencies assigned to hexadecimal symbols there are other frequencies used by the protocol.

In order to make the protocol more reliable a Reed Solomon code is appended to the payload. This makes it possible to correct some errors in transmission and to detect wether the transmission was successful or not.

The hardware setup is fairly simple. There is a receiver (Adafruit Feather ESP32) and a sender (Raspberry Pi).

The Adafruit Feather uses the MCP3008 to sample the audio signal of a microphone. Most built-in ADCs aren't sampling fast enough (sampling rate should be at least twice as fast as the highest detectable frequency). This is why most of the time you find yourself including an additional ADC converter like the MCP3008, which is fast enough for our purposes. We have tested it with a SparkFun MEMS microphone (https://www.sparkfun.com/products/9868), but an electret microphone (e.g. https://www.sparkfun.com/products/12758) should suffice as well. The following Fritzing illustrates the hardware setup on the receiver side.

We use a Raspberry Pi with an active speaker connected to cinch connector. It is recommended to use an active speaker since the volume of passive ones might not be loud enough to be reliably detectable. Also the audio driver of the Raspberry Pi can be used in order to create a sine wave signal. This is necessary because a square wave signal can't be detected by the Fourier transform.

Simply connect an active speaker to the Raspberry Pi and a push button to GPIO pin 14 and GND.

As a prerequisite to run the script you need to install play for the unix command line and reedsolo for python:

sudo apt install sox -y && pip install reedsolo

Run python RaspberryPi/main.py "This is my message to you" to emit the string as a sound when pressing the button. If you leave the argument it will emit a "Hello world" sound.

This library is fully self contained (includes all dependencies). This is because the Arduino does not support a sophisticated dependency management. Since the library is self contained you can simply download this repository as a zip file and include it into your Arduino libraries. As of now you can find this project in the managed libraries of the Arduino IDE.

This library is also compliant with platform.io. Just add the following to your platform.ini:

lib_deps =
  AstroMech

Check out the example, it is pretty much self explaining. To make the library reusable for other ICs or even the internal ADC, you need to pass the analogRead function which the library should use. Here is an example of how to define this function for the MCP3008:

Adafruit_MCP3008 adc;

int analogReadFP(uint8_t pin){
  return adc.readADC(pin);
}

When initializing AstroMech you have to pass it to the constructor like this:

... new AstroMech::AstroMech(analogReadFP, ANALOG_PIN, DEBUG);

The debug flag indicates if AstroMech should print its internals to the console.

Check out the macros defined in AstroMech.h. Those can be used to tweak the protocol according to your needs (e.g. different speaker or microphone). The following table explains the most important macros:

Ultrasonic has not been tested yet. Though, it should be fairly straight forward to implement. All the frequencies have to be shifted into the ultrasonic range. The hardware has to be replaced as well. You will need to pick a microphone with a frequency range in the ultrasonic spectrum. Also the speakers must be able to play ultrasonic sounds. I haven't tried to sample the > 40 kHz with the MCP3008. That has to be tested as well. I would be glad for any feedback on that topic.

With our speaker we experienced that, if the microphone is too close to the speaker, the receiver detects a stop bit right after the start bit. This issue could be mitigated by lowering the priority of the stop signal.

If there is too much background noise, like music, the transmission is likely to fail. This is due to the audible frequency range. This can either be fixed by implementing an ultrasonic setup (drawback is that you can't hear it anymore), or by adjusting the algorithm to adapt the threshold to the background noise. Any pull requests or forks on that topic are appreciated!