Linux-Based Embedded Platforms

The Raspberry Pi

The Beagle Bone

Linux-Based Embedded Platforms

• Small computers based on a CPU (Central Processing Unit)
• Complex architecture with a BIOS, a boot sequence, etc.
• Large computing power, RAM and ROM
• Based on an Operating System (OS): Linux
• Ease of use because of the OS
• There exists special Linux distribution for this type of board towards audio applications (e.g., Satellite CCRMA, etc.)
• Operating system introduces audio latency
• A full OS is ran to carry out tasks that are often fairly simple

Microcontrollers

Microcontrollers Features

• Little computing power (typically <200MHz)
• Little RAM (typically <1M)
• Simple processor architecture
• No BIOS
• Analog inputs
• Many GPIOs
• Very low power
• Run a firmware directly from built-in ROM
• Programmable via USB

The example of the Teensy 3.6

A solid alternative to the Cortex M4:
the ESP32

ESP32-Audio-Kit Wifi + Bluetooth Module

ESP32 TTGO T-Audio

and more recently...

The Teensy 4.0

Embedded CPU-Based Computers for
Bare-Metal Audio Applications

Embedded CPU-Based Computers

• Large computing power (typically >1GHz)
• Large RAM (typically >512M)
• Advanced processor architecture
• BIOS
• No analog inputs
• Few GPIOs
• Higher power consumption than microcontrollers
• Run a kernel from external ROM (SD card)
• Not natively programmable via USB

Study case: the Raspberry Pi

• Their exists C++ bare-metal environments for the RPI such as Circle
• Allows for furthers C++ optimizations with Faust generating C++ code targeting specific processor architectures
• To write a kernel to the SD card, it must be taken off of the PI or accessed through a programmer connected to the board's GPIOs (not convenient)
• The PI doesn't have any analog input because it is based on a CPU
• The PI doesn't have an audio codec (only a PWM DAC)

Digital Signal Processors (DSP)

• Microprocessor optimized for the operational needs of digital signal processing
• Most commonly used solution for professional audio applications
• Programmable in C++ and more and more DSPs are floating-point compatible
• Analog Devices is a leader on the market with SHARC Processors

Analog Devices SHARC Audio Module

Faust and the SHARC Audio Module

• The SHARC Audio Module can be programmed directly using Faust
• Very powerful solution for high-level low-latency audio
• Toolchain is hard to master and is not very "plug-and-play"

Field Programmable Gate Array (FPGAs)

• Extremely low level
• Idea is to create a custom electronic circuit adapted to a specific computational task
• High level of parallelization
• Computational power is not limited by the clock of the chip but by its physical size: what we can fit in it
• Programming paradigm is completely different from other chips
• We're working on a Faust solution for the Zybo Z7 (Xilinx 7-series FPGA)

Digilent Zybo Z7

The DadaMachines Doppler: an Open Source FPGA solution

Low-Level Audio System

• i2c is used to configure the audio codec
• i2s is used to transmit audio information

Inter-Integrated Circuit (I2C)

• Used for low-speed serial communication between peripherals
• System of address (each peripheral has one)
• Used to configure audio codecs using a system of register/value (see the source code of SGTL5000 in the Teensy Audio Library)

Inter-IC Sound (I2S)

• Serial bus interface standard used for connecting digital audio devices together
• Separates clock and serial data signals: simpler than i2c
• Audio data are interlaced and coded on signed integers (supports any bit depth and sampling rate)
• Separate line for word select (audio channel)
• AudioStream.cpp of the Teensy Audio Library is a good example of how i2s works