https://arstechnica.com/?p=1644163
Lidar sensors work by bouncing laser light off surrounding objects to produce a three-dimensional “point cloud.” The first modern three-dimensional lidar was created for the 2005 DARPA Grand Challenge, a pivotal self-driving car competition. Today, many experts continue to see lidar as a key enabling technology for self-driving cars.
That original 2005 lidar, made by a company called Velodyne, contained a vertical array of 64 lasers that spun around 360 degrees. Each laser had to be carefully aligned with a corresponding detector. This complexity contributed to prices as high as $75,000. Today, high-end lidars still cost tens of thousands of dollars.
There are now dozens of startups trying to build cheaper lidar. Many of them try to reduce costs by using a single laser beam that’s scanned in a two-dimensional pattern.
But other lidar companies are taking things in the other direction: building lidars with thousands of lasers. A company called Sense is selling a lidar with 11,000 lasers for around $3,000 each. Another company called Ibeo is working on a lidar that will have more than 10,000 lasers.
To be clear, Ibeo’s new lidar isn’t out yet, so we don’t know how well it will perform. And Sense’s current lidars aren’t close to matching the performance of Velodyne’s best lidar. They have a range of 15 to 40 meters, compared to more than 200 meters for some Velodyne units.
But Sense CEO Scott Burroughs says he’s only getting started. The company is working on a new sensor with a range of 200 meters that’s due out next year. That could make it competitive with today’s high-end lidar sensors. For its part, Ibeo has deep connections to the automotive industry that could allow it to score big deals with conventional automakers.
Micro-transfer printing
Both Sense and Ibeo are using a low-cost type of laser called a vertical-cavity surface-emitting laser (VCSEL). VCSELs can be fabricated using conventional silicon semiconductor techniques, allowing thousands of them to be created on a single silicon wafer. We previously covered another startup, Ouster, whose lidar is based on VCSELs.
Sense’s lidar has dramatically more lasers than Ouster’s. To achieve this, Sense uses a technique called micro-transfer printing.
It’s not too hard to fabricate thousands of VCSELs on a single die. But if you shipped a chip with 11,000 tightly packed lasers, you could have a couple of problems. Having so many lasers in a tight area could create a lot of heat. And you could also have issues with eye safety. VCSELs operate at a frequency that can damage the human retina, so if someone pointed 11,000 lasers at their eyes, it could cause permanent injury.
Sense has a clever solution for these issues: spread the lasers out. After fabricating thousands of VCSELs on a traditional silicon wafer, Sense moves them onto a new heat-conducting ceramic substrate, spreading them out in the process.
XCeleprint
This is where micro-transfer printing comes in. The technique uses a rubber stamp with a grid of tiny bumps on the bottom. When one of those bumps touches a tiny VCSEL chip, it can pick it up using electrostatic forces.
XCeleprint
The bumps are arranged so that one out of every n chips—in both the horizontal and vertical direction—is picked up from the original wafer and placed on the new substrate. Then for the next lidar unit, the stamp picks up another set of chips one slot over. In this way, a single silicon wafer can produce 11,000-laser assemblies for many lidar units.
Sense is aiming to boost its lidars’ range
Instead of scanning a scene sequentially, as many other lidar sensors do, Sense uses its 11,000 lasers to illuminate an entire scene in a single flash. The sensor then measures how long it takes the return flash to bounce back from various directions.
Flash lidars like this tend to have poor range because illuminating an entire scene means that light gets wasted in the space between pixels. Sense is essentially taking a brute-force approach to this problem, using a lot of light to illuminate the scene. Spreading out the lasers helps deal with the heat and eye safety issues that this approach would otherwise raise.
Still, Ouster CEO Angus Pacala notes that Sense’s approach has a significant downside: high power consumption. “More electrical power means bigger sensors,” he told Ars. “Bigger sensors means more cost and a more difficult integration.”
Sense’s current products deliver shorter ranges than leading lidars despite drawing more power. Sense’s lidars consume significantly more power—25 to 35 watts—than longer-range rivals like Ouster (14 to 20 watts) or Velodyne (8 to 12 watts). And the Ouster and Velodyne lidars are 360-degree spinning units; you’d need multiple fixed lidar units from Sense to get the same 360-degree coverage.
An important question is whether Sense can extend the range of its lidars without further increases in power consumption. Burroughs is aiming to release a 200-meter lidar in 2021. It will have even more than 11,000 lasers—though they exact number hasn’t been determined yet. A key challenge will be to achieve greater range without an equally dramatic increase in power consumption.
Single-photon avalanche diodes are getting trendy
Ibeo
One way Sense plans to do this is by using an array of single-photon avalanche diodes (SPADs) to detect reflected laser light in its next-generation lidar sensor. This is another parallel to Ouster, which uses SPADs in its own lidar. In an interview with Ars Technica in 2018, Pacala said his long-term vision was to use two-dimensional arrays of VCSEL lasers and SPAD detectors to build lidars that work a lot like cameras—which sounds a lot like the product Sense is aiming to introduce next year.
As the name suggests, SPADs are sensitive enough to detect a single photon. And like VCSELs, they can be fabricated using conventional silicon processes—allowing them to be cheap at scale. Their greater sensitivity may help Sense achieve longer range for a given amount of laser light.
Interestingly, Ibeo is also planning to use SPADs for its next-generation lidar.
Ibeo is not a startup. Some of its lidars were used in the 2005 DARPA Grand Challenge, but the company’s participation tends to get overlooked because its lidars had only four scanning lines as opposed to Velodyne’s 64. Ibeo scored a major coup a few years ago when it got a contract to supply lidars to Audi—the first time lidars were installed in production cars. Ibea counts ZF, a major “tier 1” auto supplier, as a minority shareholder, which could help it score additional automotive contracts in the future.
In a Thursday interview, Ibeo operations director Mario Brumm told Ars that Ibeo’s next-generation lidar, due out later this year, would feature an 128-by-80 array of VCSELs coupled with a 128-by-80 array of SPADs. Ibeo is pursuing a modular design that will allow the company to use different optics to deliver a range of models with different capabilities—from a long-range lidar with a narrow field of view to a wide-angle lidar with shorter range. Ibeo is aiming to make these lidars cheap enough that they can be sold to automakers for mass production starting in late 2022 or early 2023.
An obvious question here is how Ibeo will deal with the heat and eye-safety issues Sense is solving with micro-transfer printing. One possibility is that by using highly sensitive SPADs, Ibeo can reduce its lasers’ power output enough to avoid power and eye safety issues. It may also help that Ibeo has a one-to-one connection between lasers and detectors, leading to fewer “wasted” photons. In our conversation, Brumm told me that low power was a priority for the company’s automotive customers.
On the other hand, it might turn out that it’s difficult to get this approach to work without Sense’s micro-transfer printing technology—and that both Ibeo and Ouster will struggle to make solid-state flash lidars without it.
via Ars Technica https://arstechnica.com
January 16, 2020 at 12:54PM
