Startup developed a single-chip solution that will dramatically decrease the cost and increase the performance of radar systems in cars.
Radar used to be slow, clunky, and expensive. That’s changing quickly as tech companies begin bringing single-chip or single-package solutions to market, setting the stage for a price/performance battle in markets ranging from automotive to robotics.
Uhnder, an Austin, Texas-based startup, is the first to roll out a single-chip radar solution that it claims is roughly 100X less expensive than LiDAR, based on a 28nm RF process. This is the just the beginning of what ultimately will drive this and other safety-critical sensing technology into many more markets
“Radar has been in automotive since 1995,” said Manju Hegde, Uhnder’s CEO. “But the goal has been to put everything on a single chip. If you take an analog approach, you are linearly modulating chirps. But if you can make that happen in digital, you can have digital code modulation. And once you’re doing digital modulation, the analog portion becomes much simpler.”
Uhnder’s chip is a uses a 28nm RF process with a fast DAC and ADC, fixed-function configurable pipelines, and a CPU subsystem that uses DSPs and a variety of Arm cores.
“From day one, the goal was a very tight cost target,” said Curtis Davis, COO at Uhnder. “We have a lot of interest from automotive OEMs for 2022 production and later. When you get to level 2 (ADAS) with a lane-change assistant, all of the OEMs will be doing at least three radars per car. Some will have six radars. By the time you get to Level 4, you will have all of that plus LiDAR in front. If there’s a piece of dry plastic or wood in front of a vehicle, you want to be able to see that at 135 meters so you have enough time to react. A camera will not pick that up. So you will need multiple devices, overlapping coverage, and more sensors.”
There has been much talk over the past year about single-chip radar and LiDAR solutions, as well as multi-chip solutions in advanced packaging. What wasn’t clear, though, was that radar can be scaled. Two chips are better than one, according to Hegde, and that can be used to improve resolution and speed of results, which means more data can be analyzed at the source than with analog systems.
“What we’re heading toward is cognitive radar,” said Hegde. “You want to determine the extent of the objects, so this also can go into other cognitive features in a car.”
All of these efforts are expected to drive down the cost of this technology at a rapid pace. With six radar systems in a car, that very quickly adds up to millions of these chips per year. Now the question is what else gets cut from the bill of materials as more of these technologies begin to follow classical scaling approaches.
Fig. 1: Radar can detect traffic as well as objects in the road, an accident, and a stalled car in the tunnel while traveling at highway speeds. Source: Uhnder
Radar has a storied history dating back to its development in the 1930s. At the outbreak of World War II, it was used to identify German bombers heading toward Great Britain in what became known as the Battle of Britain. Its use in the automotive sector is more recent, but it is only in the past few years that this effort has really kicked into high gear. There are multiple types of radar, ranging from Imec’s 140 GHz technology to imaging radar, which is used to generate 2D and 3D images. But the ability to put everything on a single chip is a big step forward, and it could have broad implications in coming years for how and where this technology is deployed.
Related Stories
 |
 |
 |
 |
 |
 |
 |
 |
 |
Leave a Reply