Electronics engineers may best know what typically happens in an emerging, rapidly moving technology space. First, innovation: Start-ups bring new solutions amid rounds of funding. Next is the often-protracted shakeout period. OEMs evaluate prototypes based on performance, quality, and cost; the technologies are refined. Product announcements are made. The “best way of doing it” becomes clearer.
Then come the Tier 1s, bringing their proficiency in manufacturing and cost engineering. They take the most valid architectures, knock most if not all the cost out of them, and create a cost-effective solution, at scale. Those dynamics are well along in automotive lidar, observed Tom Jellicoe, leader of the autonomous driving team at TTP, an independent technology company in Cambridge, England. “There are still new-device architectures coming out,” he noted. “The industry has not settled on one type. There is room for innovation.”
Of the 80 to 100 players in the mobility-perception arena who claimed to be “lidar companies” before the pandemic, a more realistic total of around two dozen are operating at various levels in 2021, Jellicoe surmises. “None are large corporates poised to dominate lidar any time soon,” Jellicoe, an opto-electronics expert, said. “But there are early signs of a couple of them moving in.” He cited Bosch, which launched its lidar product at CES 2020, and lighting systems specialist Osram. Bosch’s in-house development contrasts with most Tier 1s.
Aptiv (with Innoviz, LeddarTech), Continental (Aeye), Valeo and ZF (Ibeo) have partnered with or acquired lidar start-ups to help grow their sensor technology aimed at both enhanced SAE Level 2 ADAS applications (unofficially called “Level 2+”) and autonomous SAE Level 4. “I think that anyone now who has not raised tens of millions of dollars will struggle to get into the market,” Jellicoe asserted. Two of a handful of lidar start-ups he deems “heavyweights,” Innoviz and Luminar, enjoy high valuations because they’ve been approved on series-production vehicle programs.
Israel-based Innoviz’s product portfolio includes InnovizPro, a solid-state lidar with a 150-m (492-ft) range and 20-fps frame rate. Its lower-cost InnovizTwo, to debut later this year, is aimed at ADAS applications. Luminar, based in Florida, has demonstrated range beyond 250 m (820 ft) and a 120-degree field of view (FoV). Both companies have gone public recently via SPAC mergers – Special Purpose Acquisition Companies formed for the purpose of raising investment capital through an initial public offering (IPO). SPACs have become popular among automotive tech start-ups, bringing high valuations to some promising organizations.
“They’re among the guys raising the really big cash because they’re getting the orders,” Jellicoe said. “And once their vehicle applications expand beyond those of their initial customers, they’ll have volume.” Funding has not been as acute an issue at Alphabet’s Waymo autonomous-vehicle development group, who announced in 2020 that it would begin selling its proprietary near-range, 360-degree lidar technology to industry customers in 2021.
OEMs want 10x cheaper
Lidar development is in the process of transitioning from what Jellicoe calls ‘Generation 2’ devices, of which many are in use today, to higher performance and potentially lower cost, solid-state Generation 3. The Gen-2s have served as the gateway to OEM series production applications. They’re defined by their ability to scan in two dimensions. Instead of using a stack of up to 64 lasers and detectors as was typical on the first-generation, mechanical-scanning lidars, the Gen-2 units are internally simpler. They use up to five lasers and are generally more robust than the “spinning bucket” types. Their optical beams moving laterally and vertically to measure the entire scene.
“The two-dimension-scanning lidars are currently the sweet spot of the technology for vehicles,” Jellicoe explained. “They’ve worked through the automotive development cycle and qualification. They don’t have the cost and complexity of the previous generation. And they’re pretty much what the OEMs need in terms of performance, with several offering 200-meter [656-ft] range and the field-of-view [FOV] and resolution the vehicle makers want.”
Cost for the Gen-2 devices, however, still averages around $1,000 per sensor. “The OEMs want to knock a zero off of that; they want it ten times cheaper,” Jellicoe said. Some recent product debuts have pushed specific-application sensors under a grand, notably Velodyne’s $500 Vellarray.
Following Bosch’s entry into the lidar space, its arch-rival Denso also joined the battle in a collaboration with Aeva, a Silicon Valley start-up. Founded in 2017, Aeva is one of four lidar companies to become publicly traded via SPAC mergers this year. Founded by two former Apple engineers, Aeva makes a lidar sensor that gives a three-dimensional view of the road. It also can measure the velocity of the objects it is sensing – its fourth dimension.
Aeva's sensor works on the frequency modulated continuous wave (FMCW) principle that is now under development by Mobileye and others. FMCW lidars were conceived for use in fiberoptic communication. “FMCW is very promising technology,” Jellicoe said. “By providing that additional information, they can be super-useful. The challenge is in the lasers you need.”
A conventional lidar works on time-of-flight; its simple laser produces a single-wavelength pulse which is then measured in time upon return. By comparison, FMCW lidars feature tunable lasers that can change wavelength, and can be modulated. They’re designed to the cost-point and size requirements of the optic-comms industry. They also cost in the “many thousands of dollars,” he said. The challenge for FMCW lidar technology in autos is getting the laser to the size and cost point for integration into an automotive sensor. “If someone can crack that, it’ll be really exciting. FMCW would be Generation 4,” Jellicoe offered.
Apple lidar: gamechanger?
In the rapidly evolving lidar industry, the Gen-2 makers currently have the momentum. They’re driving cost reduction by putting their devices in vehicles. In the process, they’re working out refinements in the time-honored product-development tradition of over-engineering the initial unit, then optimizing it over time.
A tricky thing to get right is beam steering, using a MEMS mirror or another component. “Getting it right makes the rest of the design easier, putting all those lasers on a mirror,” Jellicoe said. “If you focus your R&D on improving the beam steering, you drive down system costs.” The mirrors in a MEMs (Micro-Electro Mechanical Systems) lidar measure up to 10 mm (0.393 in) across. Their high resonant frequency helps protect them from certain vehicle vibrations, also an attribute of solid-state lidars.
Solid-state Gen-3 lidars, without moving parts, use three microchips. One contains the lasers that send the light beams; another carries the detectors that receive the beams. A third chip handles control, with optics on the front end. “Elegant and simple,” Jellicoe said. “And because they’re based on semiconductor components, they’re easy to assemble, manufacture and qualify. The price will drop massively with volume,” he said.
That’s one reason why Apple’s entry into lidar has raised the interest of Jellicoe and other AV tech experts. “One of the big challenges in bringing down the cost of lidar to a level where it can be widely adopted is scale,” observed Sam Abuelsamid, principal analyst, e-mobility, at Guidehouse Insights. In a recent SAE Autonomous Vehicle Engineering column he noted that the building blocks in Apple’s lidar are the same as in automotive lidars – same type of laser, semiconductors, and detectors. Apple is using flash lidar in mobile devices such as its iPhone 12 that start at $800, meaning the cost of those sensors is likely in the tens of dollars.
“When Apple enters a market like this, vendors of the relevant components quickly ramp-up capacity, driving down costs for everyone – and not just in the consumer-electronics sector,” Abuelsamid explained. “With obvious demand for lidar in high volumes, Apple’s endeavors soon will benefit all who want to use or develop lidar sensing.”
Jellicoe observed: “If the brightness of the lasers and the resolution of the detectors starts scaling in the same way it did in mobile phone cameras, when megapixels were added on an almost monthly basis, then the Gen-3 devices will be good value very soon.”
A new category: Microflash
Another Israeli start-up, Opsys Tech, has been impressing AV development teams with the performance and potentially low cost of its lidar, aimed at driver-assist and autonomous applications. “We’ve created a new category of laser scanning that we call ‘microflash,’” said company executive chairman, Eitan Gertel.
Opsys’ technology employs the same amount of light as a flash lidar, but narrows it to one pixel – 1,000 times smaller than what a flash lidar can do. Its patented electronic architecture uses Vertical Cavity Surface Emitting lasers (VCSEL), a type of semiconductor laser diode as its pulsed light source, and a single-chip solid-state photodetector (SPAD) receiver.
“We measure light 16 times faster than competitor lidars,” Gertel told SAE International. “We scan the field of view at 1,000 frames per second, but automotive guys have no real use for 1,000 fps, so we average the data down to 30 fps.” He said the Opsys Tech lidar achieves four times the range of flash lidar while surpassing flash lidar’s resolution and scanning rate. Gertel claimed tests have demonstrated range of greater than 200 meters and 0.1° x 0.1° resolution, “while maintaining full scanning rate and range across the full FoV.”
Gertel said he expects Apple’s high-volume use of VCSELs in its new iPhone lidar to enable Opsys Tech’s device to price out at $1,000 per vehicle, or $250 per sensor. “They appear to have a strong combination of range, resolution, high scanning rate and prospects for a significant cost reduction,” TTP’s Jellicoe said. “They’re very experienced in scaling products and know what they’re doing.”Continue reading »