DSP IP boosts automotive radar/lidar performance

February 27, 2019 //By Christoph Hammerschmidt
DSP IP boosts automotive radar/lidar performance
Cadence has developed a new IP for its Tensilica ConnX DSP. Under the name B20, this IP can speed processing of automotive radar and lidar sensor systems up to ten times. The secret of the speed boost are a deeper pipeline and significant instruction-set enhancements.

Cadence Tensilica’s ConnX B20 DSP provides a faster and more power-efficient solution for the automotive and, by the way, also for 5G communications markets, including next-generation radar, lidar, vehicle-to-everything (V2X), user equipment (UE)/infrastructure and IoT applications. With an enhanced instruction set architecture (ISA) and clock-speed increase, the ConnX B20 DSP processes the radar/lidar processing chain up to 10X faster and of parts of the communication processing chain up to 30X faster compared to the existing Tensilica ConnX BBE32EP DSP. Further expanding the ConnX family, Cadence has introduced the ConnX B10 DSP, which delivers half the vector width of the ConnX B20 DSP for applications requiring less parallelism.

Both the trend toward higher resolution imaging radar with more antennas for autonomous driving applications and the emergence of 5G mobile equipment and infrastructure demand significantly more processing power. In addition, consumer/industrial radar, automotive V2X and 5G IoT applications all require low-energy processing solutions.

Up to ten times faster processing of lidar and radar:
The Tensilica ConnX B20 DSP IP

Software compatible with the rest of the ConnX DSP family (the ConnX BBE16/32/64EP), the ConnX B20 DSP has a 512-bit vector width up to 128 MACs, can load 1024 bits of data each cycle and achieves 1.4GHz or greater frequency in 16nm process technology. Customers have a variety of algorithm acceleration options to reduce cycle counts, including higher precision with either 32-bit fixed-point operations (including operations to optimize MAC, FFT and FIR) with native complex support, or single-/half-precision vector floating-point (VFPU) operations with half precision at 2X the single-precision throughput.

Vous êtes certain ?

Si vous désactivez les cookies, vous ne pouvez plus naviguer sur le site.

Vous allez être rediriger vers Google.