FPGA-powered real-time processing system for unparalleled audio performance designed for easy integration.
Polyphonic unlocks high-performance, low-latency multichannel audio processing, making FPGA technology accessible like never before. By bridging the gap between raw computing power and creative applications, our platform simplifies the development of spatial and interactive audio systems.
Current low-latency audio systems struggle to reach sub-millisecond levels. Leveraging ultra-low latency audio converters and FPGA processing, Polyphonic’s platform achieves an impressive 11μs latency from analog input to analog output.
More...Thanks to its high parallelism, FPGA enables massive, perfectly synchronized multichannel processing, handling thousands of channels in real-time.
More...
Programming FPGA-based DSPs has never been easier with the Syfala compiler and the Faust language.
Syfala allows easy programming of FPGA systems using the high-level Faust programming language or C++.
• Compiles Faust DSP code to FPGA.
• Supports ultra-low latency audio processing.
• Intuitive and easy-to-use command-line interface.
Faust is a widely-used, accessible audio programming language for sound synthesis and audio processing with a strong focus on the design of synthesizers, musical instruments, audio effects, etc.
Faust Website
Control your audio program with an auto-generated GUI, a physical controller, or even an embedded Linux to seamlessly integrate Ethernet, MIDI, and more.
More...Polyphonic ensures seamless integration with audio systems through custom audio interfaces.
• 32 input & output channels, expandable up to 512
• Powered by Xilinx Zynq FPGA
• Compact, like a small audio interface
• A true plug-and-play solution
• 32 output channels per board, expandable to 224.
• Cost-effective and easy to integrate.
• Compatible with Digilent’s Zybo FPGA board.
• 32 input & output channels, expandable to 512.
• 10μs latency at 768kHz.
• Xilinx/AMD Ultrascale+ FPGA compatible.
Program your own application across a variety of domains, from active noise cancellation and radio studios to live performances and 3D audio. Polyphonic is redefining what’s possible in real-time audio processing, unlocking new possibilities and pushing the boundaries of audio innovation.
Power 3D audio environments, such as Wave Field Synthesis (WFS) and Ambisonics, enabling realistic and immersive sound experiences.
Deploy scalable multichannel systems (up to 512 channels), ideal for concert halls, VR/AR applications, and large-scale speaker arrays.
Enhance interactive audio installations in museums, theaters, and interactive exhibits.
Achieve ultra-low latency for live audio effects, real-time mixing, and dynamic sound synthesis.
Leverage FAUST-based high-level DSP compilation, allowing musicians and sound engineers to develop custom DSP algorithms without hardware expertise.
Seamlessly integrate into massive multichannel setups for high-fidelity spatial audio processing.
Implement real-time noise cancellation in environments such as smart offices, automotive cabins, and high-fidelity headphones.
Integrate adaptive acoustic control for environments requiring sound shaping, such as conference rooms, recording studios, and high-end cinemas.
Explore new frontiers in sound processing with a highly flexible and configurable FPGA-based audio platform.
Gain full control over parameters such as sample rate, bit depth, latency, and buffering size to test and refine new algorithms.
To learn more about:
This project is funded by the Inria Startup Studio and was born from the Syfala project, a research initiative led by the Emeraude Team in collaboration with the Center of Innovation in Telecommunication and Integration of Service (CITI) at INSA-Lyon, and GRAME – Centre National de Création Musicale (CNCM).
Detailed Latency Measurements:
• Bela: Tested in our laboratory with sampling rate = 22.005 kHz, bitwidth = 16-bit, 2 block samples (maximum frequency tested).
• Speedgoat: Tested in our laboratory with sampling rate = 48 kHz (maximum frequency tested), bitwidth = 16-bit, no reconstruction filter.
• Polyphonic: Sampling rate = 768 kHz, bitwidth = 24-bit, with reconstruction filter.
• L-ISA and Stagetracker: Manufacturer data.
The computational power of FPGA enables zero buffering and operates without an OS, eliminating the two main sources of audio system latency.
This allows the Polyphonic system to achieve ultra-low latency of just 2 samples for the entire digital processing. Combined with ultra-low latency ADC/DAC, it ensures the lowest possible latency based on the sampling rate.
Not only does this provide the best possible latency for a given sampling rate (Polyphonic systems can reach less than 200µs at 48kHz with reconstruction filters, compared to 500µs for Speedgoat without reconstruction filters), but it also supports high sampling rates up to 768kHz, further reducing latency for demanding audio applications.
Using adapted protocols and interfaces, modern FPGAs can physically handle up to a few thousand channels on a single chip. The current Polyphonic hardware limitation is 512 channels due to design choices but could be easily increased in the future.
The most common method for real-time audio processing in large multichannel systems relies on a computer paired with a multichannel audio interface. While modern CPUs offer significant computational power, most USB audio interfaces are limited to 32 outputs and are costly. Though parallelizing multiple interfaces is possible, surpassing 64 channels remains challenging and can add significant latency.
Ethernet-based audio interfaces can scale more efficiently, but this approach is expensive, and desktop computers may limit computational performance.
FPGAs offer a powerful solution with high computational efficiency and flexible interfacing, enabling high-channel audio processing without impacting latency. Their inherent parallelism makes them ideal for spatial audio applications like Wave Field Synthesis (WFS), where independent computations occur on each channel. Additionally, their numerous GPIOs allow direct integration with multiple audio converters, making them a scalable and efficient alternative.
Syfala generates various controller interfaces and automatically configures them to suit your DSP needs based on the controller used in your Faust program.
A software interface generated using the Faust graphical system can be hosted on a computer connected via USB.
An embedded Linux system can run directly on the board, supporting various standard protocols and interfaces.
A tailored board can be designed to meet specific needs, enabling quick and easy control of desired parameters.