Here Comes High-Res Car Radar

A new crop of device makers are developing chips based on high-resolution radar technology for assisted and autonomous driving in cars.


A dozen or so startups are developing high-resolution radar chips that use various modulation schemes and processes, such as CMOS, FD-SOI and even metamaterials.

In theory, high-resolution radar could boost the capabilities of today’s radar for cars, as well as eliminate the need for a separate LiDAR system. But the technology is still in the research stage and has yet to be proven commercially.

“To some extent, these (high-resolution radar chip) companies are flying under the radar,” said Mark Granger, vice president of automotive at GlobalFoundries. “They will mount a challenge against LiDAR.”

In a car with assisted- or self-driving capabilities, a vehicle may incorporate three sensor types—cameras, LiDAR and radar. Using pulsed laser light to measure distances, LiDAR (light imaging, detection, and ranging) can accurately identify objects, but it’s expensive and has some limitations in weather conditions.

Radar uses radio waves to detect objects and is less expensive than LiDAR. But it also has trouble discerning actual objects, which is why companies are racing to develop high-resolution or imaging radar. Today’s radar chips work for current applications, but there is a need for better radar in more advanced vehicles.

“With the trend to autonomous driving progressing rapidly, the need for high-resolution radar is also increasing,” said Michael Knebelkamp, director of product marketing at NXP. “Hardware solutions are available and being tested now, and the software and algorithms are still being developed.”

NXP, for one, is beginning to offer solutions for high-resolution radar, which are being tested in vehicles. It’s too early to tell if high-resolution radar will pan out, let alone displace LiDAR. And high-resolution radar, which isn’t a new technology, has its own issues. For years, defense/aerospace, meteorology and other high-end sectors have used high-resolution radar. But bringing this technology into the automotive market is challenging. In automotive, chips must meet stringent cost, power and safety specs.

“(High-resolution radar is) not an easy technology. The know-how of radar mainly comes from the defense industry companies,” said Kobi Marenko, chief executive of Arbe Robotics, one of several new startups in the high-resolution chip market. “We don’t see hundreds of companies doing it. There are around 10 startups that are trying to do high-resolution radar. And I assume two or three will survive.”

What is radar?
Generally, the automotive industry divides a car into five main domains—body, connectivity, fusion/safety, infotainment, and power train.

Fig. 1: Semiconductors used pervasively in automobiles. Source: UMC

Fusion/safety consists of cameras, LiDAR and radar. These technologies are targeted for both advanced driver-assistance systems (ADAS) and autonomous driving technology. ADAS involves various safety features in a car, such as automatic emergency braking, lane detection and rear object warning.

The fusion/safety and other domains consist of a multitude of chips. In automotive, the chip requirements are different than other industries. “You can’t have parts fail because it impacts safety,” said Robert Cappel, senior director of marketing at KLA-Tencor. “So you’re seeing a much different level of quality and yield. There is also a focus on latent reliability defects. A part may pass a test but fail over time within a car. The requirements are changing.”

Today, in some high-end models, a vehicle typically incorporates cameras and radar. Generally, LiDAR is too expensive for most vehicles.

For those cars, OEMs use short- and long-range radar. Adaptive cruise control and automatic emergency braking use long-range radar (LRR). For LRR, the radar module is located in the front center of the car. Front-facing LRR operates at millimeter-wave frequencies at 77 GHz at a range from 160 to 200 meters.

Located in the back corner of the car, the SRR module is used for lane detection. As vehicles move toward more advanced ADAS functions, SRR radar is evolving from the 24 GHz to the 79 GHz band.

Fig. 2: Radar for autonomous cars. Source: NXP

In simple terms, radar transmits electromagnetic waves in the millimeter range. The wave signals bounce off objects and are then reflected back. The radar system then captures signals to discern the range, velocity and angle of an object.

For automotive radar, the industry uses a mmWave technology called frequency modulated continuous wave (FMCW). “FMCW radar transmits a frequency-modulated signal continuously in order to measure range as well as angle and velocity,” according to Karthik Ramasubramanian, radar systems manager at Texas Instruments.

FMCW radar is good at detecting objects at nearby wide angle views at acceptable resolutions. Generally, the technology is relatively inexpensive and isn’t affected by weather conditions.

It also has some drawbacks. “Radar does not precisely reproduce the obstacles and features in the surrounding area,” said Alex Lidow, chief executive of Efficient Power Conversion (EPC), a supplier of gallium-nitride (GaN) chips. EPC supplies GaN devices to Velodyne, a supplier of LiDAR systems.

“Radar uses microwaves that bounce off certain objects (such as metal or concrete), but goes through other types of objects (such as humans, animals, and plastic),” Lidow said. “In addition, since the radar beams are not columnated, the resolution is poor. And it gets worse as the distance from the transmitter increases. Because of these shortcomings, radar units need cameras to help understand the surroundings. This means there is a need for fast graphics processing and a lot of deep learning.”

For basic safety features, FMCW radar is suitable. But as OEMs move toward more advanced ADAS features, they want sensors with more and faster capabilities.

For this, OEMs can go down several paths. First, they could add more radar modules to the car, but this doesn’t solve the resolution-limited issues with FMCW.

Second, OEMs could add LiDAR to the mix, which also had some tradeoffs. “For the far out view, LiDAR today is the best technology for having high-resolution, meaning the distance as well as the velocity and the ability to measure it accurately at 250 meters,” said Bert Fransis, senior director of product line management at GlobalFoundries. “Radar systems can detect that there is something there. But they wouldn’t be able to detect that it’s a person or a dog, for example.”

LiDAR has some drawbacks, as well. “LiDAR uses GaN and other technologies to generate the laser and optics around it. It’s very expensive,” Fransis said.

So for now, there is no one technology that does everything. Some OEMs might use all technologies—cameras, LiDAR and radar—for redundancy purposes. Others have different philosophies. For example, Tesla doesn’t use LiDAR because it’s too expensive. Instead, Tesla’s vehicles incorporate cameras, ultrasonic sensors and radar. Ultrasonic sensors measure the distance to an object via sound waves.

Over time, though, high-resolution radar could play a significant role “There is innovation taking place in radar. (The idea is to) bring radar to that same high resolution as LiDAR, so that you might not even need LiDAR going forward. Then you can just have a car full of low-weight and low-cost radar systems that have the same high resolution,” Fransis said. “High-imaging radar is a hot topic in the industry. The big car makers are all looking at it. That’s taking today’s radar, which is low-resolution FMCW radar, to the same resolution that LiDAR can give you, but at a substantially lower cost structure.”

So how does today’s radar and LiDAR compare in resolution? And how much further does radar need to progress to match or surpass LiDAR?

“There are numerous performance metrics for these sensors that can be compared,” said Marcus Monroe, a technical marketing specialist at National Instruments. “Radar is seeing improvement in power, field-of-view, and angular resolution, which are the metrics that are most relevant when comparing radar with LiDAR.”

Power and field-of-view are well understood. “Angular resolution is the minimum angular separation at which two equal targets can be separated when at the same range,” Monroe said. “It is determined by the beam divergence, which is a function of wavelength and aperture size. Thus, improved angular resolution allows for the detection of smaller objects as well as discrimination between several objects that are close together.”

Fig. 3: High-resolution radar vs. LiDAR. Source: NXP

Based on these metrics, radar has a way to go to close the resolution gap with LiDAR. High-resolution radar will narrow the gap, but LiDAR is also advancing, meaning radar is chasing after a moving target.

“Radar is undergoing continuous improvement. New antenna designs and advanced processing algorithms are giving radar new capabilities, which allow it to be used in areas where it was not previously used, such as pedestrian detection,” he said. “LiDAR is also undergoing its own evolution in cost reduction, movement to solid-state and new continuous waveform versions.”

So which technology—LiDAR or radar—will ultimately prevail? “We predict that both these sensors and cameras are going to continue to play equally important roles in the ongoing autonomous vehicle revolution,” Monroe said.

Different approaches
Meanwhile, a number of companies have recently emerged in the high-resolution radar chip front. These companies include Arbe, Autoliv, Echodyne, Metawave, RADSee, Steradian and others. In addition, NXP is pursuing it. Imec, an R&D organization, also is working on it.

Vendors are taking various approaches. Some are developing souped-up versions of FMCW. Others are developing multiple-input, multiple-output (MIMO) technology.

Bringing this technology into the cost-sensitive automotive market isn’t going to be easy, though.
NXP and others are helping to bring the technology up to speed.

“NXP is working closely with partners to validate performance and refine next-generation products to continuously drive radar sensors to the maximum performance,” NXP’s Knebelkamp said.

Where is this all heading? “Today’s radar in the car is mainly used for ADAS in adaptive cruise control. In adaptive cruise control applications, what the radar needs to do is just follow the car that is right in front of you,” Arbe’s Marenko said. “There is really no need for high resolution. You need to see that there is a large piece of metal in front of you to detect the range and maybe the velocity.”

But as OEMs add more ADAS features in the vehicle—such as collision avoidance and others—they will require better radar.

In fact, collision avoidance is one of the first applications for Arbe’s high-resolution radar chip technology, according to Marenko. “We started with a vision of developing high-resolution radar that can actually give camera-like pictures, but with all of the advantages of radar. It works in any weather and reaches 250 meters,” he said. “It can really detect the Doppler effect and the relative velocity of objects. It generates pictures at 50 times per second. The refreshment rate is higher and the detection is done earlier.”

Used in aviation, meteorology and radar guns, Doppler radar produces velocity data about objects at a distance.

Arbe, a fabless chip maker, is developing a three-chip solution, which includes a transmitter, receiver and a radio processing unit. Specified at 79/80 GHz, the device is based on a 22nm FD-SOI process from GlobalFoundries.

Arbe’s technology takes some of the elements of military radar and implements it in an FMCW scheme. Specifically, the technology uses some aspects of synthetic-aperture radar (SAR). Using successive pulses of radio waves, SAR is used in military systems to create 2D/3D images of objects.

In systems, SAR radar looks sideways. That won’t work in automotive, where radar needs to look forward.

FMCW is limited, so the industry requires some new innovation. “Our novelty is a new method of FMCW that will make it work with (a number of) antennas,” Marenko said.

For this, a chip requires more channels than conventional radar devices. “When you have (more) channels and data, you cannot process it in real time with today’s processors. It’s around 90 gigabits, so it’s a crazy amount of data,” he said. “We developed a dedicated chip that is tied to a new method of signal processing. It is a radar accelerator.”

To help achieve its goals, the company is using a 22nm FD-SOI process. “This helps to increase the amount of channels. We are able to have more receivers and increase the resolution,” he said. “We are also able to increase the power output of the chip. So we can get to a longer range.”

Another startup, Steradian Semiconductor, is developing a 28nm CMOS-based, mmWave imaging radar device based on a version of FMCW.

Steradian claims it can improve the resolution of radar by a factor of eight at a lower cost point. “Our technology transforms the radar from a conventional few-obstacle tracking device to an all- weather terrain mapping device,” said Gireesh Rajendran, chief executive at Steradian. “We are using FMCW modulation at the physical layer in the evaluation platform. Our radar imaging/terrain mapping algorithm, though, will be agnostic to the physical layer modulation scheme.”

Then, in another approach, Imec is developing automotive radar devices based on MIMO. Imec has developed a 79 GHz radar device and is now working on 140 GHz technology. Both are based on 28nm CMOS.

MIMO uses multiple antennas. In operation, a transmit antenna radiates a waveform. Then, each receiving antenna receives these signals. “High-resolution digital radar with a large MIMO array looks promising,” said Wim Van Thillo, program director at Imec. “The challenges that we are tackling are power consumption and cost.”

Still, 140 GHz technology has several advantages over 79 GHz. “The Doppler shift is 2X higher at 140 GHz than at 79 GHz,” Van Thillo said. “(It has) a finer range resolution at a larger bandwidth with a higher angular resolution for a given antenna aperture or a smaller antenna aperture for a given angular resolution.”

Meanwhile, startup Metawave—a spinoff from Xerox’ PARC—is developing a different solution that combines elements of SAR military radar, metamaterials and artificial intelligence.

The technology, dubbed Metamaterial Frequency-Adaptive Steering Technology (M-FAST), resembles MIMO, but it’s different. Metawave’s radar steers a directive RF beam that can determine the location and speed of road objects in all-weather conditions.

Metawave is pursuing analog radar. “They exist in the military. In the military, they use radar to do missile tracking, for instance. Instead of having an omnidirectional beam, they have a highly directional beam that they are steering dynamically,” said Bernard Casse, CTO at Metawave. “The automotive industry departed from that simply because of cost. Analog radar is expensive and power hungry, because they use what we call phase shifters. The fact that you are using microwave integrated circuit phase-shifters (is why) the cost is going through the roof.”

Metawave has devised a scanned array technology. It has the performance of military-based phased arrays, but without the cost, complexity and power consumption. “We rely on metamaterials to do it,” he said. “We have these engineered structures that are capable of mimicking phase shifters.”

The company also found a way to overcome the issues with SAR. In SAR, the radar moves in a mechanical fashion. “In our case, we are not moving the radar mechanically. The beam itself is moving. We are electronically steering the beam. By electronically steering the beam, we are capable of reconstructing the scene. That’s means we are stitching all of the different snapshots taken by the beam,” he said.

What about LiDAR?
Amid the emergence of high-imaging radar, LiDAR is making progress. For example, Velodyne recently introduced a LiDAR system with 128 laser beams. The system is 70% smaller with twice the range and four times the resolution of the previous 64-beam model.

There are other innovations. UMC, for one, is working on technologies to enhance LiDAR imaging. UMC is not making the photodetectors for LiDAR. “Our 28nm/22nm RF team is working on mmWave-based LiDAR imaging that can fulfill the specifications required for 5G as well as deliver the cost advantages that these sensor products demand,” said Steven Liu, vice president of marketing at UMC.

“Where we come in is the back end of LiDAR, where it will use the mmWave to communicate (with the vehicle) at a higher speed than 5G,” Liu said. “Its operation is above the 28 GHz transmission range to be able to communicate from the car to the recognition system with low latency and rapid response. This is a major issue in supporting ADAS and LiDAR, as the 28 GHz mmWave support has limited range and requires very near 5G base stations that support 28 GHz and above signals. High transfer rate fiber optic transmission will connect these base stations. This is anticipated for 2020 and beyond as these technologies are still under development.”

LiDAR suppliers are also driving down the costs. The same is true for radar. “(Over time), both radar and LiDAR will be cheap,” EPC’s Lidow said. “The automotive industry has a way of grinding down the costs.”

Related Stories
China’s Ambitious Automotive Plans
Radar Versus LiDAR
Foundries Accelerate Auto Efforts
LiDAR Completes Sensing Triumvirate
Electric Vehicles Set The Pace
Rethinking Car Design
Connecting The Car
Automotive’s Unsung Technology
Autonomous Cars Drive New Software
Self-Driving Cars Rattle Supply Chain
The LiDAR Gold Rush
Rethinking Verification For Cars
Test More Complex For Cars, IoT


Spadule-Kram says:

> Lidow said. “In addition, since the radar beams are not columnated, the resolution is poor.”

Did you mean “collimated”?

Leave a Reply

(Note: This name will be displayed publicly)