The SHARC-FX core delivers 5X higher performance and supports AI workloads for software-defined vehicle audio systems.
Cadence’s Tensilica IP Group and Analog Devices, Inc. (ADI) collaborated on the latest generation of SHARC automotive audio processors. The work incorporates Cadence’s Tensilica DSP architecture into ADI’s next-generation ADSP product family for automotive audio processing, voice interaction and in-vehicle system applications.
ADI’s ADSP processor family is used in automotive infotainment and audio systems. Applications include:
- Automated-driving alerts and Acoustic Vehicle Alerting Systems (AVAS)
- Road Noise Cancellation (RNC) and Engine Order Cancellation (EOC)
- Personal Sound Zones (PSZ) and 3D audio rendering
- Voice processing, AI assistance and hands-free communication
Building on the ADSP-SC59x/2159x family, ADI is introducing the ADSP-SC84x/2184x platform. The platform uses a 16nm process, compared with 28nm in the previous generation, and includes an LPDDR4 memory interface, additional on-chip L2 memory, a dedicated hardware security module (HSM) and two optional connectivity cores. These features support power efficiency, operating frequency, system integration, cybersecurity and connectivity requirements in software-defined vehicles.
The ADSP-SC84x family introduces the SHARC-FX processor core, developed through a collaboration between ADI and Cadence. The core uses a DSP architecture with support for multiple data types, VLIW SIMD processing, cache operations and an expanded instruction set for signal-processing and AI workloads. According to ADI, the SHARC-FX provides up to five times the performance of the SHARC+ processor used in the ADSP-SC59x family.
The SHARC-FX core is intended for applications such as seat-specific in-cabin audio, AI-based noise cancellation, voice assistants and generative AI audio. Its instruction set supports AI models alongside conventional DSP workloads, including real-time neural-network inference at the edge. Optimized neural-network kernels for AI and machine-learning frameworks such as Google’s LiteRT can be offloaded to the SHARC-FX core.
The development environment includes Cadence’s compiler and tools, ADI’s EZ-KIT evaluation system, CrossCore Embedded Studio, SigmaStudio+ and third-party frameworks such as DSP Concepts’ Audio Weaver.
For more information, visit cadence.com.
