High resolution imaging radar enables cars to sense the environment in all weather and lighting conditions to long, mid and short ranges as well as in any azimuth, elevation, and Doppler. It tracks velocity, and detects distance better than sensors now on the market.
The two recent incidents related to self-driving cars in the United States demonstrate the urgent need for improved sensors and related ADAS (Advanced Driver Assistance Systems) technologies. Arbe Robotics, a startup with roots in Israeli military radar technology development, is among the companies answering that need, as it begins rolling out a high-resolution automotive imaging radar chipset based on the 22FDX® technology from GLOBALFOUNDRIES.
Arbe Robotics’ imaging radar provides a high resolution of 1° azimuth and 1.25° elevation, at distances exceeding 300 meters and at a wide field-of-view of 100°. The company said its advanced technology allows the detection of small targets, such as a human or a bike even if they are somewhat masked by a large object such as a truck. The imaging radar can determine whether objects are moving, and in what direction, and alert the car in real-time about a risk.
While other car sensors can fail when it is raining, if there’s fog, and due to blinding lights such as a sudden reflection. Arbe’s radar is completely oblivious to all those factors. The custom designed radar processor creates a full real-time 4D image of the environment, and classifies targets using their radar signature.
“The performance we can show leapfrogs the existing radars,” said Avi Bauer, vice president of research and development at the Tel Aviv-based company, founded in 2015. In a previous role he benchmarked the available process technologies – ranging from silicon germanium (SiGe) to bulk CMOS – and found them all lacking. The fully-depleted SOI technology of 22FDX met the needs of both the radar front-end device and the processor. Having both chips made on 22FDX will make it easier to combine them into a single-chip solution as the company’s next-generation offering.
Bauer said that at his previous job, “we hit the glass ceiling with respect to efficiency due to the limitations of bulk CMOS,” including power handling. Bauer said that CMOS, at 28nm design rules, falls short both on integration and long-range radar power. Silicon germanium – used today for long-range radar – performs well but is power hungry and has low density. Moving to a largely digital RF design on a 16nm FinFET process would be too expensive and risky.
“With SOI the design is more straightforward, and (voltage) biasing allows you to do things that cannot be done in standard CMOS,” Bauer said. For the transmit and receive modules, SOI’s higher resistivity substrate benefits the passive components – inductors and capacitors – and allows good isolation. “High Q passives are important. At 22nm, SOI allows better performance overall.”
By avoiding the high mask counts and expensive design tools required for FinFET-based designs, Bauer said the 22FDX process meets the company’s power, performance, and density objectives, while remaining on a Moore’s Law cost-per-function curve. Speed and transistor density are important: high-resolution imaging radars generate enormous amounts of data, which must be processed close to where the sensing is happening, at very low latencies. Arbe developed a custom processor for the radar data analysis, Bauer said, and uses an off-the-shelf processor for memory and other control functions.
To LiDAR, or Not
Bert Fransis, a senior director at GF, said that with a high-resolution imaging radar system which can “see” under all weather conditions, ADAS vehicles “would have something of a winner compared to LiDAR.” Fransis said he believes that high-resolution imaging radar eventually will largely supplant deployment of LiDAR (Light Detection And Ranging), the laser-based sensors often seen on the top of today’s ADAS test cars. The ADAS companies could combine CMOS image cameras and high-resolution imaging radar and “significantly cost reduce what a vision system for a car would look like.” The rotating LiDAR modules mounted on the roofs of test cars cost $10,000 or more, and only work well on a clear day, and even then at relatively meager 20 Hz frame rates.
Today’s LiDAR modules “don’t work in foggy, snowy weather. They only provide high resolution under severe constraints,” Fransis said.
Phil Amsrud, senior analyst for automotive electronics and semiconductors at IHS Markit, said there are innovations going on in the LiDAR arena, ranging from MEMS-based and all-solid-state LiDAR, which are likely to keep LiDAR in the “sensor fusion” packages of many car companies. “Looking at the data we have now, LiDAR is going to have a much longer life than just as a science experiment on test vehicles. There is so much effort going into new technologies with fewer moving parts, so many partnerships underway, that we believe LiDAR will be used in production-intent vehicles. It still fits into the sensor fusion mentality, and I see all of these technologies running in parallel.”
3D Plus Velocity Equals 4D
LiDAR may well continue to be deployed by certain car companies, even as Arbe Robotics and other companies push radar’s effective distance to the 300-meter-plus range, and to higher resolution imaging. It claims to be the first radar company to provide high-resolution 4D pictures (3D + Velocity), at a wide dynamic range for real-time obstacle detection.
Shlomit Hacohen, vice president of marketing at Arbe Robotics, said the company is providing prototypes to customers now, and will move to general availability by early next year. “Our imaging radar is a true enabler of road safety, as it works in all weather and lighting conditions. It tracks velocity, and detects distance better than any other sensor in the market,” she said.
Today’s radars support safety systems, including adaptive cruise control, blind spot detection, and automated emergency braking. “However, with the current radars on the market you need to trade off resolution and field of view,” Hacohen said.
The Arbe Robotics systems can be configured for rear, side, or front-view detection. The company touts its ultra-high resolution of 1° azimuth, 1.25° elevation, and Doppler resolution of 0.1 m/s. It supports a wide field of view of 100° azimuth, 30° elevation, and a real-time-refresh rate of 40 FPS (frames per second).
The company has patented its post processing technology, which reduces power consumption by pointing the camera and LiDAR only to the areas of interest.
MRAM Under Consideration
I asked Bauer if Arbe Robotics plans to use the eMRAM (embedded magnetic RAM) technology developed by GF, and he said it is under consideration for Arbe Robotics’ next-generation, single-chip design. “As a stand-alone system in single device, we probably need to take a look at eMRAM. Today, we are already on the edge, and adding another feature like eMRAM would add risk. But we are looking seriously at it for the next generation.”