top of page

中文

Are we nearing a shakeout in Lidar? How VCSEL SPAD based LiDAR will help

Updated: Feb 14

Vcsel spad based lidar is best for autonomous driving

Introduction


LiDAR development has reached the 3.0 stage and is moving even beyond. Developers such as Israeli startup Opsys believe the path is set for sub-$200 sensors, and with automakers lining up suppliers for coming new-generation ADAS features, winners in the sector are beginning to emerge.


If startup Opsys succeeds, credit an assist from Apple – and get ready for the potential first real culling of the herd in the LiDAR sector.


The Israeli company is among a handful of LiDAR developers piggybacking on core technology employed by the consumer electronics giant for its iPhones. This has the potential to bring LiDAR costs down to mass-market levels and establish the technology roadmap for several vehicle generations to come.


But even before considering the benefits of an Apple bump, Opsys’ design approach, using VCSEL SPAD based LiDAR appears compelling enough to make the startup a contender in the still crowded LiDAR space by the middle of the decade.


Although a shakeout in the sector is inevitable – and even may be near – there remain an estimated 70-LiDAR – Audi’s Traffic Jam Pilot models were the first. But applications are expected to proliferate quickly over the next few years as automakers introduce higher-level advanced driver-assistance systems across their lineups and autonomous mobility fleets begin to hit the streets.


What separates these 70-plus LIDAR players today and will determine their success is the type of technology they offer, the funding backing them and the strategic tie-ups they have or haven’t formed. Velodyne was the early leader in the LiDAR field, but a number of companies since have caught up or edged past, reeling in key strategic investors and supply contracts with lower prices, and it’s still unclear which technologies and companies will win out in the long run.


State of the Art


LiDAR technology already has entered stage 3.0, where a handful of developers are racing to be among the first to offer the sensors at a cost the market can absorb, says Tom Jellicoe, head of autonomous technology for UK-based R&D consulting firm Technology Partnerships.


First-generation LiDAR were large, spinning units mounted to the roofs of self-driving test cars and costing $60,000 or more. They’ve been obsolesced by much smaller second-gen units based on MEMS (microelectromechanical systems) semi-solid-state technology that can be packaged more easily into production-vehicle grilles and headlamps.


Although they have mechanical mirrors that change angles to guide a handful of laser beams to scan the scene, overall these lidars have fewer moving parts, improving quality and reliability. Costs, halved in just the last couple of years, now hover around $600 per unit, clearing the way for application on luxury vehicles.


For example, BMW will use Innoviz’s LiDAR and Volvo will employ Luminar sensors for vehicles due in 2022. Valeo, which supplies Audi, also reportedly will supply Hyundai Genesis models as early as next year, while Ibeo (controlled 40% by ZF) and AEye and its stakeholder Continental say they have near-term deals with undisclosed OEMs. Gen 2 lidar is “obviously the sweet spot right now in terms of OEM investment and orders,” Jellicoe points out.


Is VCSEL SPAD based LiDAR a Game Changer?


But even as applications are beginning to the take hold for these second-generation systems, lidar has moved to the next development stage with full solid-state technology. Solid state eliminates all mechanical parts, meaning even lower cost and better quality and reliability. However, the technology is not without its challenges.


The first wave of solid-state lidars were based on edge-emitting laser technology. The process to make those lasers is complicated, because the semiconductor wafer must be diced up into small cubes so the light is emitted off the edge, and they require special coating as well. Because of that manufacturing complexity, scrap rates approach as high as 80%-90%.


But several start-ups – including Opsys – now are focused on VCSEL-based (vertical-cavity surface-emitting laser) technology. VCSEL semiconductor lasers are produced the same way as most other integrated circuits, by layering components on top of the wafer, with no dicing or special coating needed. The use of conventional processes greatly increases throughput and lowers cost: VCSEL yield rates are said to be in the 85%-95% range.


Jellicoe identifies four companies as being at the forefront with VCSELs: Ibeo, Sense Photonics, Ouster – which started out with mechanical lidar and has made the transition to VCSEL development – and Opsys.


Using VCSELs, “we can put thousands of lasers on a tiny chip,” says Opsys Executive Chairman Eitan Gertel. Five years ago, using edge-emitting technology, this would have been impossible, because every laser would have cost $1-$2,he adds.


Flash lidar faces a serious challenge from regulators, however. Concerned that repeated, intense bursts of laser light will damage retinas, the U.S. Food and Drug Admin restricts the intensity of the light that can be emitted during that single pulse. When spread over the lidar’s entire field of view, that amount of light becomes too diffused for the lidar to detect objects at long distances.


That’s why many flash lidars today are focused on near-field applications of up to 40-50 m (130-165 ft.), meaning automakers still might need to rely on second-generation MEMS technology for the type of long-distance detection required at freeway speeds. OEMs believe they need a range of at least 200 m (656 ft.) to detect objects early enough for a vehicle traveling at 70 mph (113 km/h) or more to actively avoid a collision.


The Microflash Solution


Opsys says it has solved that distance problem with what it calls microflash technology. Unlike more conventional flash lidars that trigger all lasers at once or in sequenced blocks, the Opsys lidar rapid fires its couple-thousand lasers one at a time, meaning each flash – focused on gathering a single pixel of image data – can be of maximum intensity, allowing the sensor to “see” much further down the road.


The Opsys lidar will detect objects at that required 200-m distance or even beyond, Gertel says. The developer showed off its 2.0 device to potential customers at CES in Las Vegas in January 2020 and made its smaller 2.5 version available for testing in November. An even tinier 3.0 rendition that meets automotive ASIL B requirements, and unlike most other lidars doesn’t require cooling, will be ready for testing in pre-production cars by the end of the year.


Inside the Opsys box are two chips, each with two transmitters and a receiver that support an array of lasers. Each chip covers about 300 pixels of the scene ahead. The lasers that are controlled individually by sophisticated software scan the scene at 1,000 frames per second – a rate said to be 30 times faster than the human eye and brain can absorb and about 30 times the speed automakers currently are demanding.


Because the Opsys lidar scans so quickly, each pixel can be examined several times a second to ensure accuracy, Gertel says. “We try to make sure it is a real point,” not simply bad data, he says.


The Apple Effect


Opsys is targeting its lidar sensors at $200 each to start, meaning to completely surround a vehicle for full autonomy would cost as little as $1,200 per vehicle – “and we give you more than 10X the points (scanned)than anybody else,” Gertel says.


He sees a further 50% reduction in size over the next few years, plus another 30% improvement in performance. Gertel also predicts cost could be cut another30% as a result of the Apple effect.


Because Apple is using VCSEL chips for its iPhone’s face-recognition system and other apps, the semiconductor industry is rapidly tooling up to meet demand. As that capacity grows, VCSEL costs should continue to come down precipitously, Gertelpredicts.


Once Apple decided to use this technology “nearly everybody in the world invested a ton of money trying to develop capacity,” he says. “That brought laser cost down by two orders of magnitude – so from a $1 to a penny. That’s what is enabling this family of lidars to work.”


That could move VCSEL-based lidars into the pole position when it comes to the race to land mid-decade sensor contracts. “The way the auto industry buys their products is not who has the best performance,” Jellicoe says. “Whoever can give the best price gets the contract. Cost is absolutely king in lidar.”


The Opsys Market Approach


Founded in 2016, Opsys spent its first year under the radar, developing its technology and filing more than a dozen patents to protect its intellectual property. It wasn’t until then that the company went looking for outside funding, drawing $9 million from venture capitalists and at least one major strategic investor, Hyundai Group. A Series B round in early 2020 landed an additional $22 million in backing.


The plan is to use contract manufacturers to produce the lidar units using tooling and processes owned and developed by Opsys. But the company also will work through a preferred Tier 1 supplier if that’s the direction an automaker wants to go.


The company already has won some business, with a program in place for an unnamed Asian automaker for 2022 aimed at Level 2+ and Level 3 ADAS applications.


Asked how big the market could get for lidar, Gertel says that based on an average of $700-$1,400 per vehicle, “even if only 30% of the cars get this level, it’s huge numbers. And that’s only automotive. Minimizing the size another50% we think allows us to have super integration and be able to address many other markets too and many other applications.”


What happens next in the lidar sector is anyone’s guess. But VCSELs have a chance to be the next big leap that can bring lidar costs down to radar-like levels – and if they work as advertised, a market coalescence around flash technology would seem likely. That battle between MEMS-based Gen 2 and flash-based Gen 3 sensors could play out in the next 24-36 months as OEMs define their ADAS plays for the next five years and start to place bets on which lidar technology to rely on.


But that likely won’t be the end of the story. There’s 4.0 lidar technology on the horizon that works like Doppler radar but uses laser light instead, Jellicoe points out. These sensors not only measure distance, but they can also detect velocity


Blackmore Sensors & Analytics is among companies developing this type of frequency-modulated continuous wave lidar. It was acquired by autonomous- vehicle developer Aurora in 2019 with an eye toward using the sensors on Level 4 autonomous vehicles.


Aeva, a California startup headed by two former Apple engineers and backed in part by Porsche, is another company pursuing the FMCW lidar route. It drew more than $2 billion in an initial public offering in November.



Comments


bottom of page