MENU

NXP to focus on all CMOS radar

NXP to focus on all CMOS radar

Market news |
By eeNews Europe



NXP's booth at MWC

NXP’s booth at MWC

In briefings, Clemmer casually mentioned, without elaborating, that making today’s “big and clunky radars” small is one of the keys to next-generation advanced drivers’ assistance systems. It turns out that the small radars Clemmer referenced aren’t from Freescale’s, a company known for its fine 77GHz packaged radar front-end chipset using SiGe technology.

Clemmer was talking about an RF CMOS-based 80GHz radar front-end transmitter chip — currently a working prototype — developed at NXP.

Called Dolphin, NXP’s 80GHz chip uses digital CMOS process technology, an accomplishment long believed impossible.

Lars Reger, vice president strategy, new business, and R&D for automotive at NXP, told EE Times that the working prototype is currently in the hands of “our lead customers [Tier Ones and OEMs] under non-disclosure agreement.”

Asked about the tiny radars Clemmer cited, Reger said, “This isn’t a story about small radars. It’s about up-integration. We’ve found a path to integrate front-end radar transmitter with a baseband — all in CMOS.”

Keeping the front-end chip in a process technology like BiCMOS would make it hard to advance integration, said Reger. Just as NXP has won the global car audio market by integrating FM, AM, satellite radio chips with silicon tuners — all in CMOS, “Our goal is to do the same up-integration with radar chips,” he added.

During his interview with EE Times, Clemmer said NXP’s AM/FM car radio chips are used on “27 out of 28 car audio platforms of choice” used by Tier Ones and car OEMs.

From GaAs to SiGe, now to CMOS?
The first commercial radar systems of the late 1990’s were based on GaAs chips. But then Infineon started developing systems based on bipolar process SiGe chips. So did Freescale. Those SiGe radar chips are already designed into radar collision warning systems.

Reportedly, automotive radar developers have already warmed up to SiGe radar chips and begun switching from GaAs. But here’s a big question: Will the new millimeter-wave sensors made in plain CMOS prompt them to switch again — this time from SiGe to CMOS?

That’s the big market shift NXP is betting on. And certainly NXP isn’t alone thinking along these lines.

About a year ago, IMEC announced a 79GHz radar transmitter implemented in 28nm CMOS and designed for automotive radar systems. At that time, IMEC, which developed it in collaboration with Vrije Universiteit Brussel, called it “the world first,” explaining, “With an output power above 10dBm, the transmitter front-end paves the way towards full radar-on-chip solutions for automotive and smart environment applications.”

Asked if NXP’s Dolphin was spun out of IMEC’s development, Reger said no. He said it’s an internal project three years in the making.

How NXP gave birth to Dolphin
The idea of the development of an 80GHz radar transmitter chip came from a team of NXP engineers who developed a 60GHz chip for wireless HDMI, said Reger. That chip was able to cover 15 meters.


Reger, responsible for automotive R&D, said, “We don’t necessarily get funding for everything we want to innovate. Sometimes, my job is to steal good ideas from other division within our own company.”

The idea from the wireless HDMI team was compelling. But Reger knew that going from 60GHz to 80GHz would be a big jump. “I told them, ‘Guys, shouldn’t we do a test chip first?’”

On Christmas Eve, 2013, Reger got a phone call from the team. “It was exactly midnight,” he said. “I was informed of the chip’s tape out. I told my family how excited I was, but also said they probably wouldn’t understand…”

RF CMOS-based 80GHz radar front-end transmitter IC

RF CMOS-based 80GHz radar
front-end transmitter IC

By the end of March, the team got the new chip, called Dolphin, back from the factory. Early April, 2014, they put together demo boards. By May, Dolphin, which was working “way better than expected,” according to Reger, was shown to the management board. “Everyone on the board joined the demo; by jumping out in front of the radar system, testing Doppler effect, checking out how it works. The whole demo turned into a toy for boys.”

Dolphins prototype modules have been designed into Tier Ones’ systems for several months now, for further testing. 

Meanwhile, NXP is keeping Dolphin’s details close to the vest. Reger is neither talking of the exact geometry used for the 80GHz RF CMOS front-end chip nor the timing for launch of production chips. But he’s confident of the technology, and he sees big traction for it from automakers.

Today’s high-end vehicles typically feature a two- or three-chip single SiGe radar system, used in adaptive cruise control.  But expectations are high among automakers for building cars with lots of high-resolution short range automotive radar for various applications.

Examples include collision warning systems (front and side), collision mitigation systems (front and side), vulnerable road user detection — cyclists and pedestrians for example, blind spot monitoring (rear), lane change assistance and rear cross-traffic alerts.

Radar or vision? 
As to the future of ADAS, the auto industry isn’t choosing radar over vision or vice versa.  Euro NCAP isn’t mandating either radar or vision. Nor is it asking carmakers to have both.

A carmaker can rely on a more advanced radar system combined with a lighter vision system or, conversely, choose to go with a more advanced vision sensor with a lighter version of radar.

Vision technologies excel in tasks like detecting lane markings and other road information, such as reading traffic signs, reliably detecting pedestrians, and lighting functions such as controlling the high/low beams.

On the other hand, vision technologies can’t handle some jobs, like seeing through snow and fog. Dirt renders vision sensors blind. Unlike radar, vision technology can’t see very far. Long-range radar (LRR) can comfortably handle between 30 and 150 meters, and short-range radar (SRR) can detect objects within 30 meters.


 

The automotive industry is looking for both solutions as a package. Freescale is doing exactly that. Just last week, Freescale unveiled at the Mobile World Congress its S32V microprocessor.

Inside the automotive vision SoC is CogniVue’s second-generation APEX Image Cognition Processing technology. The SoC additionally supports the fusion of vision data captured by the S32V device. Fused in are other data streams, including radar, LiDAR and ultrasonic information to enhance resolution and image recognition, Freescale said.

Meanwhile, Freescale has its own radar solution. Its MR2001 is a high-performance 77GHz radar transceiver chipset “scalable for multi-channel operation enabling a single radar platform with electronic beam steering and wide field-of-view to support [multi-range] applications across automotive safety, communications infrastructure, and industrial systems,” according to Freescale.

In vision Freescale and NXP have solutions using different vision algorithms experts. Freescale works with CogniVue and NXP is partnered with Mobileye.

In radar technologies, the merged entity is likely to profit from each other’s diverging technology and market experience (Freescale’s SiGe-based radars; NXP’s nascent efforts for CMOS radar front-end transmitter chip).

Work has only begun
NXP’s Reger acknowledges that the work has only begun on a single-chip all CMOS radar transceiver in future.

Lars Reger, VP, strategy, new business, and R&D for automotive at NXP

Lars Reger, VP, strategy, new business, and R&D for automotive at NXP

Aside from integrating the radar front-end chip with MCUs to make a complete system-level solution for ADAS applications, antenna developments also need to come along to shrink the module. 

Obviously, CMOS is lower cost, better integrates digital circuitry and benefits from technology scaling, compared to a SiGe bipolar process, but some say that’s not enough. The maximum available gain at millimeter wave frequencies is known to be lower for CMOS, and its low supply voltage reportedly limits output power.

NXP’s Reger, however, noted that the team is working on “the best radar illumination,” to make more powerful, accurate and high-performance millimeter wave sensors that work for various range applications including mid and short-range.

The team is also working on a new scheme to connect multiple CMOS radar front-end chips via automotive Ethernet, so that they work as one.  

NXP’s Dophin operates on 80GHz band, “plus or minus a few GHz,” making it work between 77GHz and 81GHz, according to Reger.

About the author:

Junko Yoshida, Chief International Correspondent, EE Times

Related articles:

Dutch startup shrinks 60GHz radars, increases precision

Let’s aim above 100GHz

Imec implements 79 GHz radar transmitter in plain 28nm CMOS

 

If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News

Share:

Linked Articles
10s