Software

Software & Firmware Architecture

The software stack for the Audio Research Codec is designed to facilitate real-time multichannel audio processing, supporting research into Active Noise Control (ANC) and Active Vibration Control (AVC). The system is built around a Teensy 4.1 microcontroller, leveraging its DSP capabilities and high-speed I/O to stream, process, and output up to 16 channels of audio with minimal latency.

The software is split into distinct functional layers, ensuring separation of concerns and modular development. These layers include low-level hardware configuration, a real-time audio processing pipeline, and a command-and-control interface for live system monitoring and configuration. The software architecture allows researchers to dynamically tune processing parameters, log real-time performance data, and seamlessly swap between different filtering algorithms.

Hardware Interface Layer (I2C + TDM Control)

At the foundation, the firmware provides low-level register control over the TLV320ADC5140 ADCs and ADAU1966A DAC using the I2C bus. Configuration routines initialize:

• Audio data format (TDM16), clock polarity, and slot assignments for seamless multichannel transfer.

• Input channel enabling, gain, and impedance settings for both single-ended and differential inputs.

• Clocking and synchronization between the ADCs, DAC, and processor to ensure sample-accurate timing across all channels.

The low-level read/write functions enable direct interaction with codec registers, including debugging utilities that allow register reads/writes via a serial command interface, making system-level debugging easier in the lab.

Adaptive Filtering & FX-LMS Algorithm Layer

In addition to the static signal routing, the software supports real-time adaptive filtering using the Filtered-X LMS (FX-LMS) algorithm, a core method for both ANC and AVC systems. The FX-LMS implementation dynamically updates filter coefficients based on error microphone feedback, allowing the system to "learn" the optimal cancellation signal in real-time.

The adaptive filter coefficients are updated each sample, with the reference path accounting for system delay and secondary path dynamics. This layer is designed to be swappable, allowing experimentation with multiple adaptive filtering approaches, including:

• Standard LMS (Least Mean Squares)

• Normalized LMS (NLMS)

• Filtered-X LMS (FX-LMS) for systems with complex secondary paths

The filter processing is integrated directly into the Teensy Audio processing callback, ensuring it operates within the 44.1 kHz audio interrupt cycle.