NXP SAF9000 and SAF9100 Automotive Audio DSP bring AI to car infotainment

NXP has recently launched the SAF9xxx Automotive Audio DSP family, which currently includes SAF9000 and SAF9100 AI audio DSPs. Built around Cadence’s latest generation high-performance Tensilica HiFi 5 DSPs, these new chips not only feature AI and ML capabilities, but also include features like driver’s voice pitches and accent recognition, noise cancellation, voice recognition, emergency siren detection, and more.

Additionally, the SAF9000 chip includes a software-defined radio option with up to five integrated tuners (controlled by an integrated Arm Cortex-M7 core) that covers all major global broadcast radio standards, including DAB, HD Radio, DRM, CDR, and AM/FM into a single chip solution.

NXP AI Audio Proce with SAF9000 Audio DSPs

SAF9xxx Automotive Audio DSP specifications

  • DSP – Tensilica HiFi 5 DSPs with dedicated neural network engines
  • Integrated Controller – Arm Cortex-M7 core for tuner and audio control middleware
  • Radio Features (SAF9000 only)
    • Supports AM, FM, DAB, DAB+, DMB, HD Radio, DRM for AM, DRM for FM (DRM+), CDR
    • Single hardware variant covering all standard radio use cases (incl. PD and MRC)
    • Unique wideband tuner concept for receiving full bands
    • Five built-in tuners
    • Tuner configuration (number and standards) controlled via software
  • Audio Features
    • Two high-performance floating-point audio DSPs (HiFi5) with neural network accelerators
    • Extensive range of audio algorithms supported via a vast 3rd party ecosystem
    • A high number of integrated sample rate converters
    • 12 configurable audio ADCs
    • ASIL-A audio support
    • Fully programmable digital audio input/output port for various digital audio interface formats
    • Support for DSP Concepts Audio Weaver Tool
  • Connectivity
    • Optional 2nd PCIe interface or Gbit Ethernet SGMII interface (supporting 100BASE-T1 and 1000BASE-T1 PHY)
    • PCIe 3.1 compliant interface supporting data rates of 2.5 Gb/s and 5.0 Gb/s
    • Two general-purpose SPI ports
  • Package – Compact 9×9 mm package

As I mentioned earlier, the company has released two new DSPs the SAF9000 and SAF9100 the main difference between the two is that the SAF9000 features a radio module whereas the SAF9100 doesn’t have one.

NXP SAF9000 and SAF9100 AI audio DSP Block Diagram
Block Diagram of SAF9000(with Radio) and SAF9100(without Radio) AI audio DSPs

The Cadence Tensilica HiFi 5 DSP core supports fast-access memory and handles all the AI functionality and neural network algorithms. The Arm Cortex-M7 MCU in conjunction with the DSP handles all the security, safety, audio I/O, audio processing, and other generalized functionalities. The chip also supports GbE Ethernet for communication.

In terms of software, the chips utilize Audio Weaver by DSP Concepts, which is a software platform used to develop, debug, and deploy complex audio processing systems with AI and ML plugins from Bose, DTS, and Intelego. These enhance noise cancellation and audio experiences in electric vehicles.

This new chip reminded me of NXP’s RT600 Crossover Processor or the high performance of NXP i.MX RT1050 processor but with much more improved audio capabilities and AI twist. As of now, the chip is not available for purchase since it is in the preproduction stage, and NXP states that “Specifications and information herein are subject to change without notice”. More information about this SAF9xxx Automotive Audio DSP can be found on NXP’s press release page and the respective product pages.

Share this:

Support CNX Software! Donate via cryptocurrencies, become a Patron on Patreon, or purchase goods on Amazon or Aliexpress

ROCK 5 ITX RK3588 mini-ITX motherboard
Notify of
The comment form collects your name, email and content to allow us keep track of the comments placed on the website. Please read and accept our website Terms and Privacy Policy to post a comment.
Khadas VIM4 SBC