This software implements a polyphonic, multi-channel audio player on the Raspberry Pi. Its purpose is to trigger asynchronous audio events with low latency from a variety of trigger sources, including MIDI, serial and ethernet. It is intended to be used for custom performance musical instruments, interactive art installation, museum exhibits, pinball machines, etc. It is not streamer softare.
If you're familiar with my embedded products on SparkFun, it is most like the Tsunami. Being hosted by a Raspberry Pi, however, means it has more speed, more memory and more control interface capabilities. Unlike Tsunami, MIDI Notes can trigger multiple sounds, and each sound can have individual pitch offset and be sent in any combination to any or all of the audio outputs.
Octaphon is being developed in conjuction with a custom Raspberry Pi multi-channel audio HAT.
This HAT uses the RPi I2S and I2C ports on the RPi GPIO header to provide 8 channels of output (as well as 2 channels of input, not yet used by this software.) Because I don't currently have official linux Alsa ASoC drivers for this HAT, the Octaphon software contains some dedicated code in the Mixer module to utilize 8 channels of output. Using a custom device tree overlay file, the HAT presents to linux as a stereo board, and the mixer module uses standard Alsa API functions to manage and play audio. But the HAT drives the I2S clocks at 4X the audio sample rate, while the mixer multiplexes 4 sets of stereo channels into the output buffers, and a novel mechanism implemented in both the mixer and the HAT firmware keep the channel assign- ments correct.
Because the mixer code uses Alsa functions, it would be relatively straightforward to modify Octaphon to support standard stereo HATs as well as other multichannel interfaces such as XMOS based USB Audio 2.0 devices.
Unlike the WAV Trigger and Tsunami, Octaphon loads sets of up to 128 sound files into RAM. This is done both on startup and also can be done at runtime to change the sound selection. Because of this, Octaphon will support looping over a specified address range, which need not be the beginning and end of the file.
Octaphon is intended to be run headless as an embedded audio player. I've implemented a GUI intended to be displayed on an 800X480 GUI LED panel, showing signal level on 8 outputs as well as timing/performance information. Of course, it also displays fine on a standard desktop used for development.
Some of the compile-time options are presented in the config.h file. These include:
POLYPHONY - The number of sounds that can be played simultaneously. Currently set to 32.
NUM_OUTPUTS - The number of outputs supported by the mixer. This is also the maximum number of sounds that any single MIDI note can trigger. Currently set to 8.
MAX_SOUNDS_PER_BANK - The number of sound files that can be loaded into memory at any one time. Currently set to 128.
At present, Octaphon implements a MIDI interface to trigger sounds. The MIDI input devices are enumerated and presented in a drop-down combo box and can be selected from the GUI.
An Excel-generated csv file is used to define a set of sound files to be loaded as well as function definitions for MIDI Note events. The format of this csv file is ongoing, but at the moment it supports three functions. Functions lines are denoted by the function code in the first column - all other lines in the file are ignored and may be used for documentation.
#PATH This command specifies an absolute path to the directory that contains sound files.
| Command | Absolute path to sound directory |
|---|---|
| #PATH | ~/Sounds/ |
#FILE This command loads a mono sound file into memory
| Command | Idx | Filename | Loop Start | Loop End |
|---|---|---|---|---|
| #FILE | 3 | track001.wav | 389644 | 408377 |
#NOTE This command defines the action for a specific MIDI Note
| Cmd | Note | Idx | Tuning | dB1 | dB2 | dB3 | dB4 | dB5 | dB6 | dB7 | dB8 | Loop |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| #NOTE | 3 | 12 | -2 | 0 | 0 | -8 | -10 | -99 | -99 | 0 | 0 | 0 |
This command specifies a sound to be played using the sound index (Idx) assigned in the #FILE command. The sound will be played at nominal pitch if Tuning=0, or plus or minus a pitch offset of the specified number of semitones. Gains for each output are in dB, with 0 being full scale (no attenuation) and -99 being completely muted in that output. If Loop=1, the sound will loop over the loop points specified in the corresponding #FILE command.
There can be up to NUM_OUTPUTS #NOTE commands for a given MIDI Note number. This allows a single MIDI event to trigger up to NUM_OUTPUTS sounds, with each sound having a different set of output gains.
Octaphon is still very much in the early stages. Here are some things on short to-do list:
-
Volume fader, used to provide attack, release, crossfades and generally smooth volume changes.
-
Looping.
-
Samplerate conversion (semitone offsets) for each note event.
-
Serial control
-
Ethernet control (likely using OSC)
Octaphon is written for linux in C++ using the Juce framework. For ease of development, I have Juce installed on a full desktop version of the Raspberry Pi OS, along with the Code::Blocks IDE. I have two exporters defined in the Octaphon.jucer project - one for Code::Blocks and one for the linux command line make.
To begin using just the linux command line, first install Juce, then the required linux packages. Build the Projucer and then open the Octaphon.jucer in Projucer. Save the project, then open a terminal to the resulting /Builds/LinuxMakefile folder and just type make. If you also have Code::Blocks installed you can navigate to the /Builds/CodeBlocksLinux folder, open the project and build from within CodeBlocks.
The build creates a single executable which can be from the command line.