Audio playback via USB Serial or PROGMEM on pin 10 of the Arduino MEGA2560 board.
This project contains Rust code to handle host-side serial communication and audio encoding as well as generation of C++ header files containing audio to be stored into the atmega2560's PROGMEM.
Each playback mode (STREAM_PCM, STREAM_DFPWM, STORED_PCM, STORED_DFPWM) has a host-side utility.
Sample rate is determined by the WAVE file's sample rate. Multi-channel audio is averaged into mono.
The CLI written in Rust can be installed to be available anywhere in your shell like this:
git clone https://github.com/DasEtwas/arduino_audio_stream
cargo install --path . --force(use cargo uninstall arduino_audio_stream to uninstall respectively)
An Arduino IDE cpp file containing arduino audio playback code can be found here audio_stream.ino.
Using the streaming modes, arduino_audio_stream can be used to stream pcm u8 or DFPWM encoded audio.
Example streaming myaudio.wav as pcm:
arduino_audio_stream myaudio.wav arduino-ide/audio_stream/sounddata.h --format pcm --mode stream --port /dev/ttyACM0 (corresponding arduino mode: STREAM_PCM)
This will open /dev/ttyACM0 and stream the chunked audio in PCM format to the arduino.
NOTE: Generally, PCM allows for higher sample rates than DFPWM because decoding takes a while, and should be preferred for realtime playback for higher quality.
NOTE: The streaming may be baud-rate sensitive. The higher the better.
To store the file "myaudio.wav" in a header file besides the given example project, consider the following example usage:
arduino_audio_stream myaudio.wav arduino-ide/audio_stream/sounddata.h --format dfpwm --mode storedAfter enabling the #define STORED_DFPWM in audio_stream.ino, compile and upload. Sample rate and data length are included automatically.
NOTE: To achieve storing any proper length of continous audio on the device, one needs to use DFPWM which has an 8:1 compression ratio and low sample rates (eg. 16kHz or less). Higher sample rate sounds better up to 24-32kHz. Example: To fit 2:20 of audio onto the mega2560, I used a sample rate of 12900Hz and dfpwm encoding.
Especially when using DFPWM, adding low-pass filtering and amplifier on pin 10 is recommended. The more bass the signal has, the more pronounced are squeaking artifacts (due to DFPWM being more sensitive at higher amplitudes).
audio_stream uses Timer 2 for 62500Hz PWM for analog output on pin 10, of which the OCR2A register is modulated in 8-bit mode to achieve audio output (duty-cycle modulation).
Timer 4 is used to call the Output Compare Match A interrupt at the sample rate by setting it to CTC mode and setting OCR4A.
DFPWM stands for Dynamic Filter Pulse Width Modulation. It uses 3 input parameters for a basic LPF and signal response. Each bit corresponds to one output sample.
A 1 bit tells the predictor to increase amplitude, whereas a 0 bit tells it to decrease amplitude. Multiple of the same values (0 or 1) following eachother
cause an increase in response (sensitivity). A silent signal is just repeated ones and zeroes.
The best parameters where chosen by trying out every one by bruteforcing after analying noise using the RMSE of the input and DFPWM output. (RESP_PREC = 10, RESP_INC = 1, RESP_DEC = 1 are found to have the least RMSE and coincidentally are also the values in >1.7.10 minecraft versions of the minecraft mod Computronics mod including the DFPWM codec :P)
- https://www.maxpierson.me/2009/05/29/generate-real-time-audio-on-the-arduino-using-pulse-code-modulation/
- https://forum.arduino.cc/index.php?topic=485507.0
- https://wiki.vexatos.com/dfpwm (great minecraft mod btw, I used to listen to noisy audio all day in game)
MIT