Companies are now sampling a new SoC that networks car sensors, enabling data fusing and over-the-air updates for new vehicles.
Jan 17, 2024NXP has released a new version of its radar automotive products at the 2024 CES, aiming to enable a new generation of software-defined vehicles that can receive updated features as they come available as well as network their onboard sensors.
The new SAF86xx chip makes it possible for auto companies to offer over-the-air updates to a car’s radar software long after the car is purchased.
Additionally, the SAF86xx —when used with an NXP microprocessor—comes with a networking feature that allows a vehicle to gain improved information about situations around the vehicle by synthesizing data from numerous sensors.
The SAF86xx is a radar System-on Chip (SoC) using 28 nanometer (nm) RF CMOS technology that targets new architectures for software defined vehicles, explains Matthias Feulner, senior director of ADAS at NXP Semiconductors. RF CMOS is a process particularly suitable for monolithic integration of ICs for radar sensors.
The 28nm RF CMOS enables extended detection range for cars and motorcycles in highway scenarios as well as offering more reliable detection and separation of small objects like pedestrians or concrete blocks.
Additionally, NXP is enabling distributed aperture radar technology in partnership with San Jose technology company Zendar.
Sensors Gets Smarter
In recent years, the automotive industry has been working to provide continuous feature enhancement for cars, making them mobile computing devices similar to the driver’s smart phone. Initially the focus was on deploying independent sensors that helped provide safety related functions like blind-spot detection and automated emergency braking.
Today, as the vehicles of 2025 and later are being designed, there's a shift towards more convenience-oriented offerings, Feulner explains. Those may include advanced comfort features like highway piloting, automated park assist and urban piloting, for example.
“That goes along with the expectation of vehicle users to have an experience very much like you have with a mobile phone,” says Feulner.
Working with Auto Manufacturers
In the past year or so, NXP explored—with auto manufacturers—how vehicle architectures will evolve and the requirements to support that evolution.
Already companies have begun working on a 2027 and 2028 cluster of new cars and trucks that will qualify as software defined vehicles. That means that more functions are implemented in software rather than hardware and vehicle software would need the ability to be updated over time.
When SAF86xx -based sensors are connected to NXP’s radar processor—radar features can be hosted on the company’s S32R—the car’s system can receive software upgrades and new features. The processor connects to sensors via Ethernet and can send and receive to the cloud through a communications gateway, via cellular connectivity.
Sensor Fusion for Greater Intelligence
In the coming years, most mid-range vehicle will have five radar sensors—the four corners of the car and one on the front. Premium vehicles could be up to 10, adding rear and side facing sensors.
With NXP’s 86xx chips in sensors, a vehicle could use those networked sensors, to employ sensor fusion, and see around the vehicle with a holistic 360-degree view.
NXP collaborated with German technology company SmartMicro to demonstrate the radar sensor data fusion from all of a car’s sensors, showing detailed mapping of the vehicle’s environment.
That includes information such as identifying elevation of items that come into view, to help with object classification: from trees, a curb, a roadside barrier, or a bicycle.
“That level of detail is needed to support advanced comfort features [for hands-free vehicles] such as automated pilot functions,” says Feulner.
Extending Radar and Aperture
The solution enables collaborative sensing mode in which data from more than one physically separated sensor, fused in the S32R radar microprocessor, creates what is a larger, virtual antenna or aperture.
That enables usage of compact standard radar sensors for high-resolution sensing, avoiding the need for growing antenna size, which is limited due to need for integration into the front of the car, typically hidden behind the bumper.
NXP teamed with Zendar—a San Jose company—using the company’s Distributed Aperture Radar (DAR) fuse sensor data to gain more detail about objects that are within the field-of view of those sensors, thereby helping classification of a pedestrian or a street sign.
Hosted Software Features, Upgrades and AI Based Features
Employing NXP’s new system solution for software-defined radar, allows hosted software-based features on the S32R as part of an ADAS controller.
“By doing so it’s easier to upgrade them over time through over-the-air updates,” says Feulner.
One of those hosted software features could be AI functionality for radar-based object classification, where neural networks are trained with the detail-rich point cloud data delivered by the newer generation of radar sensors.
Using Data from Cars
Training of that network continues with new data after the vehicle is delivered to a customer and over time the new improved network can be downloaded to the vehicle through over-the-air updates to make object classification more accurate.
The radar system can stream low-level radar sensor data, compressed range FFT data, at up to 1 Gigabit over an Ethernet connection.
The SAF86xx is being sampled by customers today, and the early adopters may be robotic taxis and autonomous vehicles.
- SAF86xx enables networked sensors throughout cars, as well as updates of software to add new features to cars already on the road.
- The new products are being sampled now and can be expected in new vehicles around 2027.