The International Wireless Industry Consortium (IWPC) organized a face-to-face interactive workshop, themed Automotive Sensor Architecture, in Munich from May 31st to June 1st local time. Calterah was invited and delivered a speech titled “SoCs Used in Edge Radar and Satellite Radar”.

At this event hosted by BMW, renowned global OEMs, top Tier-1s, and companies in the upstream IC supply chain gathered, focusing on automotive sensor architecture, to explore the pros and cons of centralized and distributed architectures for ADAS in the dynamic landscape of rapidly advancing autonomous driving.

With the continuous development of autonomous driving, the collaborative performance of sensors plays a crucial role in ensuring vehicle safety. Multi-sensor architectures can help reduce the perception limitation and uncertainty of individual sensors based on the information gathered from different sensors, thus forming a more comprehensive perception of the environment or targets, which helps to optimize the all-around performance of multi-modal perception systems.

During the speech, Zheng Wang, Calterah Automotive Product Line Director, stated that as the radar typology evolves from distributed ECU to domain controller and centralized computing architecture, a multi-sensor architecture is required to jointly sense the environment with a Central Computing Unit to make comprehensive judgments and coordinate operations. He shared two types of collaborative working architectures for automotive radar, Satellite Radar and Edge Radar. Satellite Radar refers to the radar architecture where data from various radar sensors is coordinately processed by the Central Computing Unit after preliminary pre-processing; while Edge Radar involves independent signal processing by each radar sensor.

Wang also compared the two architectures. Satellite Radar can support more complex algorithms based on the high computing power of Central Computing Unit, facilitate sensor fusion, as well as reducing SoC cost and complexity. However, it comes with higher system-level costs due to its requirements for interfaces of higher data rates, making it unsuitable for cost-efficient vehicles without Central Computing Units and Ethernet interfaces. In addition, Satellite Radar is in the prototyping stage and has not been tested by the market. In comparison, cheaper and simpler interfaces can meet the requirements of Edge Radar, which is also supported by a mature ecosystem and offers proven performances.

The comparison was concluded with Wang’s summary of the development trend of these two architectures: “Satellite Radar is cutting-edge, but it would take a long way to become popular. Hence Edge Radar would dominate the market for quite a period. And the period would be further extended if the performance of radar SoCs continues to improve and the cost for imaging radar could be reduced.”

Calterah’s next-generation mmWave radar SoC family Andes can be employed for Edge Radar. In terms of the RF front end, Andes SoCs come with 7-bit transmission line based phase shifters and optimized floorplan for high isolation. For signal processing, Andes integrates cutting-edge DSP and Calterah proprietary RSP (radar signal processor), supporting high-throughput data processing, highly computing-intensive algorithms, and parallel signal processing for greater computation efficiency. Moreover, a flexible number of Andes SoCs can be cascaded in use. Its “Safety Island” design can meet various levels of cybersecurity requirements. Another Calterah SoC product, Andes-Satellite, which is yet in the development stage, is designed for Satellite Radar and will have better RF front-end performance and a lighter signal processing module.

Higher-level autonomous driving requires more advanced automotive sensing. Samples of the next-gen mmWave radar SoCs Andes are expected to be available in Q3 2023 for forward-looking radar and 4D premium and imaging radar, providing greater external perception capabilities for vehicles and facilitating intelligent driving.