有了这项技术，普通汽车可以直接改造为自主驾驶汽车

听说过氮化镓晶体管吗？你一定是在有关车载激光雷达的文章中看到这个生僻词的。但它在无人驾驶汽车中发挥的至关重要的作用。

未来的自动驾驶汽车将依赖于激光传感技术，而该技术的基础则在于新型非硅高速晶体管和芯片。

上述观点来自斯坦福大学物理学博士、创业家艾利克斯•利多（Alex Lidow），他同时担任位于加州艾尔塞贡多的高效功率转换公司（Efficient Power Conversion，简称EPC）的CEO。EPC从事的业务是：采用比硅的运算速度更快、效率更高，但成本却较低的材料制造晶体管和芯片。

没错，硅——所有计算机的物理基础，以及谷歌和苹果等技术巨头得以崛起的催化剂——现在终于出现了鲜为人知的竞争对手。它就是新型半导体材料：氮化镓（GaN）。由氮化镓制造的晶体管（大量晶体管集合成芯片）具有速度快、尺寸小、能耗低、成本低等优点。

今天，整个计算机产业都是由硅所主导。然而，对探测并处理巨量信息的能力提出严苛要求的自动驾驶汽车技术已经成了氮化镓晶体管的早期试验田。

谷歌公司等致力于研制全自动驾驶汽车（非计算机辅助驾驶）的企业正在把目光投向激光雷达技术。激光雷达可在汽车周围发射一束激光束，进而创建周围环境的高分辨率3D地图。激光地图与摄像机和智能软件相互协作，足以为自动驾驶汽车提供上路自动行驶所需的全部信息。

EPC的芯片和晶体管以极高速度发射激光束——从而使车辆能够迅速判断周围物体的距离和位置，并创建高精确度3D地图。

“氮化镓是自动驾驶汽车激光雷达系统的关键。”利多说。在不同应用环境下，氮化镓晶体管的运算速度要比硅晶体管高100-1,000倍。须臾之间的运算速度高低就决定了自动驾驶汽车能否成功探测到周围物体并做出有效反应。今年早些时候，在自动驾驶模式下，一辆没有采用激光雷达技术的特斯拉Model S未能在晴朗天气下探测到正在逼近的白色卡车，造成车内人员死亡。

利多说，他的公司现在为所有采用自动驾驶激光雷达的车企供应氮化镓芯片或晶体管。某些车企——例如特斯拉——没有采用激光雷达，而是使用其他种类的雷达系统。

但是利多认为，包括特斯拉在内的所有自动驾驶汽车都将最终采用激光雷达技术。

激光雷达过去曾经由于价格高昂而无法用于普通汽车。第一辆谷歌地图测绘车和自动驾驶车都曾经采用了成本高达数万美元，甚至更高的激光雷达系统。但是今天，有多家企业正在试图把激光雷达的成本降低到数百美元的水平。

在晶体管科技领域耕耘了数十年之后，利多于2007年创立了EPC公司。1977年，他曾经发明了能够在不同设备之间切换电子信号的硅晶体管。“我注意到硅的时代即将画上句号。”利多说。利多和一位台湾商业伙伴合伙出资创办了公司，并且拥有公司几乎全部股份。

2014年，利多称EPC公司氮化镓晶体管的生产成本终于实现了低于硅晶体管的目标。“氮化镓同时具备了高性能和低成本，这是历史性的突破。”利多说。

EPC公司2016年的收入较之2015年几乎实现翻倍，有专家预计2017年的收入将达到2016年的3倍。该公司氮化镓晶体管业务的规模占其业务规模的15-20%，其他业务则包括用于数据中心服务器、无人机、无线充电、医疗设备、无线基站等应用领域的芯片和晶体管。

随着包括谷歌、特斯拉和苹果（据说）在内的科技巨头纷纷开始发力无人驾驶汽车技术，激光雷达已经成为当下的热门话题。如雨后春笋般出现的初创公司们正在努力降低激光雷达技术的成本，其中包括计划将激光雷达系统的成本压低到区区250美元的硅谷初创公司Quanergy Systems。

如果激光雷达的成本降至上述水平，无法自动驾驶的普通汽车只需像新增一部传感器一样做简单改造就可实现自动驾驶。“如果成本真的降下来，无需刻意购买自动驾驶汽车，所有汽车都能加装自动驾驶功能。”利多说。

利多并不想知道一个自动驾驶汽车大行其道的世界是什么样子，他最关注的是硅退出历史舞台后的世界如何运转。他在一通从南加州打来电话里说，“硅有硅谷，氮化镓也能创造只属于自己的地名。” （财富中文网）

作者：Katie Fehrenbacher

译者：郑立飞

In the future, self-driving cars will require laser-based sensing tech, and these systems will need new types of high-speed transistors and chips that can beat out silicon.

That’s the assertion of Alex Lidow, a Stanford PhD physicist, entrepreneur, and CEO and founder of Efficient Power Conversion (commonly called EPC), a company based in El Segundo, Calif. that makes transistors and chips out of a material that operates more quickly and efficiently—and costs less than silicon.

Yes, silicon—the backbone of all things computing and the juice behind the rise of tech giants like Google and Apple—has a little-known competitor. It’s called gallium nitride, or GaN, and the semiconductor can create transistors (the things that go on chips) that are fast, small, energy efficient, and low cost.

While silicon has dominated computing to date, it turns out that the massive amount of information that needs to be detected and processed on the fly by self-driving cars is a perfect early application for GaN transistors.

Some companies that are looking to make fully autonomous cars (not just computer-assisted driving) like Google are relying on Lidar technology, which fires out a laser light beam around the car to create a high-resolution 3D map of the surrounding environment. That laser map combined with cameras and smart software, is enough information for a car to drive itself on a road without a human driver.

EPC’s chips and transistors fire the Lidar system’s laser at a blazing fast speed—fast enough to help the car rapidly determine the distance and placement of surrounding objects as well as to create an accurate-enough 3D map.

“GaN is crucial” for the self-driving Lidar car system, says Lidow, as it can operate 100 times—or even a 1,000 times—faster than silicon, depending on the application.

Mere seconds can mean the difference between a car detecting and reacting to an object or unfortunately not. Earlier this year, Tesla’s autonomous car tech, which doesn’t use Lidar, failed to detect a white tractor trailer driving toward a Model S against the backdrop of a bright sky. The accident resulted in a fatality.

Lidow says that his company sells GaN chips or transistors to every car company that is using Lidar for self-driving cars today. Some companies, like Tesla , rely on radar systems instead of Lidar.

But Lidow thinks all self-driving cars, even Tesla, will eventually need to adopt Lidar.

Historically, Lidar has been prohibitively expensive to use on cars. The first Google mapping vehicles and self-driving cars employed ultra-expensive Lidar systems that cost tens of thousands of dollars or more. But today, a handful of companies are trying to push the cost of Lidar systems down to as little as several hundred dollars.

Lidow founded EPC in 2007 after decades of working on transistor technology. In 1977, he invented a silicon transistor that switches electronic signals in devices. “I noticed silicon was coming to the end of its valuable life,” says Lidow. Lidow and a business partner in Taiwan funded and almost completely own the company.

By 2014, Lidow says EPC’s cost to make GaN transistors was finally lower than the cost of making silicon ones. “For the first time in history, GaNis higher-performance and lower-cost,” says Lidow.

Today, EPC almost doubled its revenue in 2016 versus 2015, and expects to triple its revenue in 2017 over this year. The company’s Lidar transistors represent between 15% to 20% of its businesses today,while chips and transistors for data center servers, drones, wireless power, medical devices, wireless base stations, and other tech make up the rest of the business.

Lidar is becoming a hot topic for tech companies as everyone from Google to Tesla to reportedly Apple race to build autonomous car tech. Startups are emerging to lower the cost of Lidar tech, like Silicon Valley startup Quanergy Systems, which is buildinga Lidar system that’s supposed to cost $250.

If Lidar can get that cheap, then regular cars—those that don’t drive themselves—could employ the tech as just a better sensor. “If it’s cheap enough, it’ll be on all cars, independent of autonomy,” says Lidow.

For Lidow, he’s not just focused on what a world with self-driving cars would look like, but more so on how the world will operate after the fall of silicon. He says, speaking on the phone from southern California, “Silicon can have its Valley, we’ll take GaN Beach.”