A look at what makes the radar systems useful and their pitfalls
MADISON, Wis. — Now that regional New Car Assessment Programs (NCAP) are demanding such features as adaptive cruise control (ACC) and emergency brake assist (AEB) for their five-star safety ratings, NXP Semiconductors is urging the automotive industry to get cracking with radar systems.
To accelerate radar integration in advanced driver-assistance systems (ADAS), NXP rolled out Tuesday (Oct. 2) a radar solution combining its S32R processors, RF transceiver and antenna design on a new reference platform. Developed in partnership with Colorado Engineering, the platform meets “the stringent functional, performance, and safety requirements of the industry,” claimed NXP.
Integrated sensors in a vehicle (Source: NXP Semiconductors)
The new system was designed to demystify the intricate “art” of radar that typically requires big automotive OEMs to fine-tune antenna and analog designs. NXP hopes that its “out-of-the-box” automotive radar system can serve Chinese car OEMs who still need several years to catch up with automotive incumbents in the rest of the world.
In a recent phone interview with EE Times, Kamal Khouri, vice president and general manager for ADAS at NXP told us, “Radar has become the sensor of choice” to enable ACC and AEB. “Cameras can’t measure velocity, while radars can,” he explained. “By bouncing off signals, radars can also see around the corner. On the other hand, lidars that use no moving parts are still very expensive.”
However, it is well known that traditional radar lacks resolution and can’t distinguish nearby objects. Radars are also notorious for sounding false alarms and they consistently fail to process information fast enough to be helpful on the highway.
Khouri made clear that NXP does not believe radars will replace cameras. “The combination of cameras and imaging radars offers redundancy, thus making vehicles safer,” Khouri said.
Inside new radar solution
So, what does NXP’s new radar solution entail?
The reference design, dubbed RDK-S32R274, combines NXP’s S32R27 processor, TEF810x CMOS transceiver, FS8410 power management IC and a radar software development kit. NXP has added expansion and antenna modules that can be optimized to create a customized development platform for specific customer applications.
At the heart of the radar solution sits a scalable family of Power Architecture-based processors — S32R27 and S32R37 — which Khouri described as “the first chips dedicated to processing radar algorithms.”
NXP's S32R block diagram (Source: NXP)
According to Roger Keen, ADAS radar segment manager for automotive microprocessors, NXP’s radar processing IP runs on NXP processors, in addition to auto-grade software for ACC and AEB. Board and antenna design modules designed for the company’s radar solutions are “hardened as automotive qualified systems.”
With auto-grade radar SDK offered by NXP, developers who used to “hand tune” their own radar processing IP to specific hardware can now make function calls to the NXP’s radar system, Keen explained.
The S32R27-based solution is designed for extended applications such as ACC and AEB. S32R37, with less processing power than S32R27, is code-compatible and highly optimized for operations like blind spot detection.
The S32R27 version is priced at $14-$17 (1,000-unit distribution pricing). The S32R37-based solution costs $10-$12.
NXP isn’t alone in the charge for car radar integration. Ian Riches, executive director of global automotive practice at Strategy Analytics, counted both NXP and Infineon among leaders in automotive radar.
Meanwhile, Texas Instruments, a recent radar market entrant, launched its catch-up campaign with the 2017 introduction of millimeter-wave radar chips built on standard in-house RF CMOS technology. TI told us its radar chips offer “less than 5-cm resolution accuracy, range detection to hundreds of meters, and velocity of up to 300 km/h.” Even more important, TI differentiates its radar chip by combining mmWave-sensing devices with a 76- to 81-GHz mmWave radar with a microcontroller (MCU) and digital signal processor (DSP) cores on a single chip.
TI took this approach because a higher level of integration can reduce footprint, power consumption and cost without performance loss. Cédric Malaquin, technology and market analyst for RF devices and technologies at Yole Développement, told us that although NXP took the first step by developing its RF-CMOS transceiver, TI went further by integrating the DSP in its radar chip. Malaquin said that DSP integration gives TI’s radar solution an almost 60% footprint reduction. The DSP is key to “the signal processing chain to detect and classify an object.”
NXP, however, defended the company’s two-chip solution (radar chip + microprocessor), stressing that this approach offers customers much more scalability and flexibility for radar integration.
NXP's radar solution: Antenna side (Source: NXP)
NXP’s Keen said, “Think about thermal management under 110°F Arizona weather.” Keeping transceiver chips farther away from a microprocessor, for example, makes it easier to manage thermal controls when radars need to be installed in bumpers, he explained.
NXP's radar solution: processor side (Source: NXP)
Keen also added that NXP’s approach – using processors specifically designed for radar processing IP – has enhanced performance per watt for radar solutions. Pressed about the benchmark used for performance-per-watt analysis, NXP said that NXP teams have gleaned answers “from public data” and “what we’ve heard in confidential meetings with customers.” But Keen added, “Although this gave the best performance per watt we’ve ever seen, we’ll reserve the more expansive industry claim for third-party testing.”
Asked to compare TI’s radar chips with NXP’s radar solutions, Strategy Analytics’ Riches observed, “Essentially, the TI approach could potentially offer lower cost, but also slightly less flexibility.”
Radar suppliers and market research firms are bullish about the growing demand for automotive radars.
Different radar applications require many different radar modules. NXP told us, “Typically, for blind spot detection, two radar modules in the two rear corners of the vehicle. In more advanced requirements (Cross traffic detection), an additional two radar modules are required for the front corner of the vehicle.”
For long-range radar applications, a single module is typically mounted somewhere in the front bumper, according to NXP.
Strategy Analytics forecasts that from 2018 to 2022, it expects “a cumulative 375 million automotive radars to be fitted to light vehicles.” Riches believes that 107 million radars will be installed in 2022.
NXP's estimates on the radar market by application (Source: NXP)
Similarly, NXP estimates the shipment of 109.2 million units of radars – ranging from corner radars to high-end corner radars and long-range/medium range front/rear radars all included – in 2022, leading the adoption of radar technology to 50 percent of all new cars.
The newest trend among new radar solutions is how best a radar system can generate a high-resolution “image” that can both locate and identify/classify objects in the field of view, according to Strategy Analytics’ Riches.“Today’s radars used on vehicles do not have the resolution to do that over a sufficiently wide field of view to generate a true image," Riches said.
This objective can’t be accomplished with radar chips alone. Riches explained, “The antenna design is hugely important here, and that’s one of the reasons why we have seen start-ups like Metawave receive funding from the likes of Infineon, Denso, Toyota AI Ventures, Hyundai Motor Company and Asahi Glass (amongst others).”
The virtues of radar technology are well known, most notably its ability to work in all weather conditions. Automotive experts believe radar can team with vision sensors as critical sensing technologies in highly automated vehicles.
Strategy Analytics’ Riches explained:
Fundamentally, they operate at very different wavelengths. Cameras (obviously…) use visible light, and thus are weakest when it is dark, in very high-contrast lighting situations (e.g. exiting a tunnel) or in heavy rain/snow. Lidars emit light outside of the normal visible spectrum, but have the greatest challenges in bright sunlight, which gives the system a lower signal to noise ratio. High-resolution lidar technology is also currently costly and less mature as an automotive application than either cameras or radars.
In contrast, he noted that radars is “immune to lighting conditions, while it can penetrate rain and snow conditions very well.”
Radar is not, however, an end-all, be-all solution. “The key weakness of radar to date has been its resolution: it’s been good at saying there is 'something' out there, but not very good at telling what it is," Riches said.
Put simply, radar technology can be lousy for “making an informed decision as to whether to carry on driving (e.g. an overhead street sign has been detected) or apply emergency braking (a fire truck is parked ahead partially in-lane).”
This explains why current automotive radars tend to sometimes filter out and ignore stationary objects. The radar cannot tell if an object is something you don’t want your car to hit, Riches stressed.
Indeed, user manuals are full of warnings for drivers whose vehicles are fitted with radars. Riches gave us a few examples.
Taken from the manual for a Skoda Superb (which used RADAR-based ACC):
“The ACC does not react when approaching a stationary obstacle, such as traffic jams, vehicle breakdowns or vehicles waiting at a traffic light.” (page 236)
The manual for the Volvo XC90 has similar warnings:
“Distance Alert is active at speeds above 30 km/h(20 mph) and only reacts for vehicles ahead moving in the same direction as your vehicle. No distance information is provided for oncoming, slow moving or stationary vehicles.” (Page 289)
“Pilot Assist does not brake for people, animals, objects, small vehicles (e.g. cycles and motorcycles), low trailers as well as oncoming, slow or stationary vehicles.” (Page 310)
Riches concluded: “You’ll find similar text in many other user manuals from many other brands…. The imaging radars aim to fix this problem.”
— Junko Yoshida, Global Co-Editor-In-Chief, AspenCore Media, Chief International Correspondent, EE Times