NXP Advances AI Audio Processing with SAF9xxx Audio DSPs

NXP Semiconductors is improving AI audio capabilities in automotive infotainment and marking a significant improvement in automotive audio processing technology in general with the latest SAF9xxx family of audio DSPs. Utilizing Cadence’s latest generation high-performance Tensilica HiFi 5 DSPs, combined with dedicated neural-network engines, the SAF9xxx audio DSPs enable efficient implementation of next-generation, high-quality, self-learning audio and speech applications.
 
The new NXP SAF9xxx family of audio DSPs introduces the latest-generation technology for automotive, AI and machine learning (ML) applications, integrating a neural network engine and hardware-based accelerators. Designed to support the ever-increasing demand for AI-based audio in software-defined vehicles (SDVs), the new audio digital signal processing (DSP) solution launches with two devices.

The NXP SAF9000 is an automotive radio and audio single-chip solution enabling car infotainment for the software-defined car. It includes five build-in tuners (including DAB, HD-Radio, DRM, CDR, and AM/FM) for multistandard analog and digital radio processing, two advanced audio processing cores, an embedded MCU and comprehensive connectivity including PCIe and Gigabit-Ethernet. Based on a new software control approach, the SAF9000 allows to cover worldwide radio and audio requirements for vehicles with a single hardware variant.

The SAF9100 device is a dedicated automotive audio one-chip solution enabling car infotainment for the software defined car. It includes two open core latest generation audio DSPs that can satisfy the more demanding audio processing demands for entry level, mid-range and high-end audio applications. In addition, the SAF9100 has an embedded MCU and comprehensive connectivity including PCIe and Gigabit-Ethernet, supports AI/ML accelerators and includes NXP's own hardware-based biquad accelerators. The NXP SAF9100 machine learning capabilities with neural network integration creates the opportunity for advanced features, such as the ability to recognize and distinguish the driver from passengers by identifying their voice pitches and accents. Emergency vehicle sirens can also be detected with machine learning algorithms, and used to provide drivers with audiovisual information about the location and movement of an emergency vehicle.

"The NXP SAF9xxx DSP is a ready-made infotainment solution. Designed with AI/ML processing capability in mind, it provides more functionality, better performance, and more features to support the demanding audio and radio requirements of both current and future software-defined vehicles. And its scalable architecture makes it simpler and more affordable for OEMs to deploy in their vehicle design," explains John van den Braak, Senior Vice President and General Manager for Audio Systems at NXP.

Developers will be able to rely on the SAF9xxx family of audio DSPs for applications that require low-latency audio optimizations, enabling real-time active cancellations of engine, road, wind, and tire noise. The Gigahertz-DSP performance is complemented by a large, fast-access onboard memory and a comprehensive set of analog and digital I/Os that support the most demanding audio features, while Gigabit-Ethernet supports fast and direct connection to the vehicle network.

The new SAF9xxx family is set to debut at the Audio Engineering Society 5th International Conference on Automotive Audio on June 26-28, 2024, in Gothenburg, Sweden. 

A large audio IP partner ecosystem is ready to kickstart new audio developments using the SAF9xxx family, including an extensive range of audio software and algorithms from DSP Concepts, Arkamys, BdSound, Goodix, and Dirac. More details on the SAF9xxx family can be found on NXP’s website, including the software-defined radio option at nxp.com/saf9000 and the audio DSP solution at nxp.com/saf9100. 
www.nxp.com