TI Puts CMOS mmWave Radar in Production
Article By : Junko Yoshida, EE Times
Automotive radar sensors to hit the market by end of 2018 or early 2019, says Texas Instruments
MADISON, Wis. — The limitations of automotive radar systems are well-known. Traditional radar lacks resolution and can’t distinguish nearby objects. Radars are also known to sound false alarms and they consistently fail to process information fast enough to be helpful on the highway.
However, automotive experts also recognize the redeeming virtues of radar technology, most notably its ability to work in all weather conditions. They believe that radars can team with vision sensors as the critical sensing technologies going into highly automated vehicles.
Knowing both its drawbacks and advantages, a bigger question is where radar goes from here.
Texas Instruments hopes to answer this question with millimeter-wave radar chips built on standard in-house RF CMOS technology. Introduced a year ago, TI’s radar chips offer “less than 5-cm resolution accuracy, range detection to hundreds of meters, and velocity of up to 300 km/h,” according to the company.
Sameer Wasson, TI’s general manager of radar & analytic processors, recently told EE Times that after a year of pitching the company’s radar chips, his team is seeing substantial traction for both automotive and industrial applications.
Pointing out that TI’s AWR1642 mmWave sensors (built for the automotive market) are already in mass production, Wasson said that he expects to see TI’s radar chips inside OEMs’ vehicles “at the end of this year to mid-2019.” Even more exciting to Wasson are “revelations in the industrial side of applications” for the company’s radar chips. TI’s IWR1642 mmWave sensors (designed for the industrial sector) are finding applications that let them go inside everything from smart buildings to factory floors and transportation systems.
DSP plays a central role
Analysts at Yole Développement predict that TI is poised “to change the [radar] technology landscape very quickly.”
How so?
Cédric Malaquin, technology and market analyst for RF devices and technologies at Yole, told us that the key lies in the integration structure of TI’s radar solutions. TI’s mmWave-sensing devices integrate a 76- to 81-GHz mmWave radar with a microcontroller (MCU) and digital signal processor (DSP) cores on a single chip.
Obviously, the higher level of integration can reduce footprint, power consumption, and the cost of radar chips without performance loss. NXP, for example, took the first step by integrating the MCU in its RF-CMOS transceiver, noted Malquin. But TI has gone further by integrating the DSP as well.
The integration of the DSP turns out to be critical. It gives an almost 60% footprint reduction by improving power consumption, noted Malquin. Furthermore, the DSP is central to “the signal processing chain to detect and classify an object.”
Indeed, Wasson noted that the DSP inside TI’s mmWave sensors makes it possible to classify and track objects and count people, for example. “The DSP enables users to place machine learning at the edge,” he said.
Yole’s Malquin added that the integration of the DSP inside one component with the MCU and the transceivers leads to less interconnection losses and quicker processing.
The DSP used inside TI’s mmWave radar is a 600-MHz user-programmable C674x DSP. The same radar chip incorporates a 200-MHz user-programmable ARM Cortex-R4F processor.
(Source: Texas Instruments)
Radars find broader auto applications
Basic ADAS functions such as blind-spot detection and adaptive cruise control are well-understood applications easily addressable with 24-GHz corner radar and 77-GHz front radar sensors. Wasson said that more significant is a rapid expansion of TI’s mmWave radar applications “well outside of usual ADAS features.”
For example, the digital processing capability inside the mmWave sensor can filter out noise, said Wasson, allowing TI’s radar chips to detect very small movements, even the breathing that indicates the presence of a person or animal inside a vehicle.
Wasson noted that “child occupancy detection” is likely to become a feature in the Euro NCAP roadmap. This, he believes, will open the door for TI’s radars in body, chassis, and in-cabin applications. As tier ones and OEMs look for the right sensing technology to enable such detection possibilities, Wasson noted that radars are much better-positioned.
Radar, for example can “see” through a blanket to determine whether a child is underneath. TI’s radar chips can even distinguish between a person and a static object like a duffel bag, explained Wasson, because their on-chip digital signal processing can detect a heartbeat.
A camera can’t do any of this.
In addition to sniffing out life inside the car, the inclusion of MCU and DSP inside a radar’s front-end chip also helps “detect free space and obstacles near doors and trunks, intruder alert, and smarter automated parking,” according to TI.
Wasson stressed that TI is the only company offering radar solutions integrated with front end, DSP, and MCU onto a single chip. It’s also alone in volume production with mmWave radar chips. “Other silicon vendors — our competitors — talk about their CMOS-based mmWave radar chips, but we haven’t seen them on the market. If they want to be in 2020 vehicles, they need to be in mass production by now.”
Imaging radars
Notably, TI is extending its radar into imaging. In traditional radar applications, radar emits radio waves, receives reflections, and uses this information to generate data on the position and motion of objects.
Imaging radar uses the returning waves to create an image. When radio waves reflect off objects, the radio waves change in ways that convey data about the objects, including how far the waves traveled and the kinds of objects at the end of the trip. Using the acquired data, a computer can create a 3D or 2D image of the target.
TI demonstrated at the Consumer Electronics Show earlier this year one such imaging radar. By cascading four single-chip AWR1243 radar front-end devices, TI showed that it can deliver angle resolution as high as 1.6 degrees and detect cars beyond 250 meters.
The DSP integrated into the AWR1243 is capable of interpreting data acquired by radio waves into a point cloud, said TI’s Wasson. An example is shown below.
Images that TI’s cascading sensors can produce. You can also see the video demo here. (Source: Texas Instruments)
Stéphane Elisabeth, project manager for RF and advanced packaging at System Plus Consulting, told us that TI also has in its portfolio simple radar chip transceivers without MCU and DSP. Compared to its competitors’ offerings, “one of its advantages is the simplicity of the die stacking,” he said.
“In fact, depending on the application (imaging, mid-range radar, or long-range radar) the transceiver could by stacked in waterfall effect — a cascading manner — to provide larger or stronger sensing.”
TI radar chips use standard Flip Chip Ball Grid Array (FCBGA) packaging, providing a low-cost sensor solution with better output power. Competitors, on the other hand, use advanced packaging techniques such as Embedded Wafer Level Ball Grid Array (eWLB), added Elisabeth.
Imaging radar is becoming a hotly pursued technology. Advantages include its ability to operate in the presence of obstacles that obscure the target and to penetrate ground (sand), water, or walls.
A number of startups engaged in the development of imaging radars are responding to RFQs issued by tier ones and OEMs, noted Wasson. These include startups such as Vayyar Imaging and Arbe Robotics, both based in Israel.
Vayyar Imaging reportedly created a radar system capable of seeing through walls and mapping everything from individual objects to entire factory floors. Arbe Robotics is said to be improving the simultaneous localization and mapping (SLAM) algorithms in its radar system. The software runs inside an embedded processor that builds a point cloud 50 times per second and localizes the vehicle 25 times per second with a degree of resolution between objects.
Both startups provide their imaging radar solutions in a black box in much the same way that Intel/Mobileye offers customers its vision-processing solution, said Wasson. In contrast, TI provides a more flexible imaging radar product solution that can be adapted to customer needs.
Radars go industrial
TI is confident that its mmWave sensors will transform industrial applications for radars.
For building automation, TI noted that developers can use the reference designs for “people-counting and -tracking” by using a mmWave radar sensor. “By monitoring both the large and small movements of people through the collection range, velocity, and angle data, developers can enable more intelligent building systems, including heating, ventilation, and air conditioning (HVAC); lighting; elevators; and more,” according to TI.
Wasson also mentioned his recent meeting with an engineering team in the U.K. who was working on a traffic-monitoring system to be installed at intersections. “By using our radars out of the box, the team quickly built a traffic-monitoring system that exceeded their expectations,” he said.
TI’s traffic monitoring, object detection, and tracking reference design uses mmWave radar to track vehicle and machine movement with a sensor-equipped intelligent transportation system, according to the company. Similarly, radars can be applied to factory or farm to streamline different processes.
Wasson believes that there’s no end to potential radar applications. TI wants to make sure that its users’ experiences — including some once viewed as niche applications — become mainstream radar uses.
— Junko Yoshida, Chief International Correspondent, EE Times