ON Semi’s Image Sensors and SiPMs Give AutoX a Leg Up on Autonomous Driving

July 20, 2021 by Adrian Gibbons

Navigating with ON Semiconductor technology onboard, AutoX’s self-driving platform is the first to receive a license to operate autonomously on China’s roadways; however, what tech is it using?

The road to autonomous driving often feels like a pipedream. Even when breaking down vehicle automation into six levels, from 0 to 5, ranging from no automation to humans reading a book (or taking a nap), the road still looks blurry. 

Today, North American society still sits firmly at level 1, where driver assistance technology, like adaptive cruise control, can be employed by the driver. 


The Society of Automotive Engineers (SAE) levels of automation.

The Society of Automotive Engineers (SAE) levels of automation. Image used courtesy of Synopsys 


However, in China, AutoX, a forerunner of level-4 autonomous vehicle technology, is the first to successfully receive a license to operate "high automation" RoboTaxis. The AutoX platform contains ON Semiconductor technology, enabling 360° vision using twenty-eight 2D image sensors and four LiDAR sensors based on SiPM technology. 

In this article, All About Circuits takes a look at AutoX, along with its mission and technology. Why has it partnered up with ON Semiconductor, and what is SiPM technology concerning LiDAR systems?


AutoX Seeks to Provide Universal Access to Transportation

Since its founding in 2016, AutoX has deployed more than 100 RoboTaxis in China while expanding into California (USA), being the second corporation to receive a driverless permit. 

The company's technology is said to provide 'superhuman safety,' providing a universally positive impact on transportation. 


The AutoX RoboTaxi has advanced 360-degree vision.

The AutoX RoboTaxi has advanced 360-degree vision. Screenshot used courtesy of AutoX


The AI model for AutoX’s platform is data-driven, derived from both real-world and virtual-world testing. As of last year, their team had deployed test platforms to twelve cities for a ‘breadth and depth’ test to accommodate different road topologies, weather patterns, and driver styles.

Additionally, using an in-house technology called xSim, their engineers generate high-quality test scenarios for their system using deterministic simulation data. 


A subset of an xSim simulation.

An xSim simulation. Screenshot used courtesy of AutoX


AI training may be a critical element of the level-4 high automation driving paradigm; however, detection comes before data. To that end, Jianxiong Xiao, founder and CEO of AutoX, says that ON Semiconductor technology is the obvious choice for AutoX's platform. 


Eliminates Blindspots in Autonomous Vehicles

Safely driving in dense urban environments often requires significant amounts of sensor data, especially as you move up the autonomous levels. Utilizing twenty-eight AR0820AT 8.3 Megapixel image sensors and four LiDAR arrays from ON Semiconductor, AutoX claims to eliminate blind spots.


The AR0820AT is used in Autox's RoboTaxi.

The AR0820AT is used in Autox's RoboTaxi. Image used courtesy of ON Semiconductor 


The AR0820AT is a CMOS-based image sensor using a 2.1 µm backside-illuminated (BSI) pixel. It offers video capture operating at 40 fps with a resolution of 3840 x 2160.

ON Semiconductor's second technology onboard the AutoX platform uses silicon photomultipliers (SiPM) and is available in various formats and families

SiPM is at the heart of ON Semiconductor's LiDAR technology, but how does it work?


A Brief Technical Overview of SiPM

SiPM technology is another form of a solid-state photodetector; however, it is dissimilar to single-photon detection image sensors as it produces a real-time analog waveform and does not hold a charge. 


Structure of a SiPM microcell.

Structure of a SiPM microcell. Image used courtesy of Hamamatsu


The detector works as a series combination of an avalanche photodiode (APD) and quenching resistor (Rq). The presence of a photon produces a gain of 10e5 to 10e6 in developed charge carriers (either holes or electrons). 


Equivalent model of a SiPM (two microcells).

Equivalent model of a SiPM (two microcells). Image used courtesy of Hamamatsu


When a microcell is exposed to a photon 's' is closed, and the junction capacitance Cj is discharged through the Rs creating a current pulse. After a while, the circuit quenches the avalanche due to lack of charge conversion, and the current pulse dissipates. 

One significant advantage of this technology lies in the high-gain potential, especially because of the constant presence of uncorrelated 'light noise.' By using a histogram and direct time of flight measurement results, SiPM technology can correlate a signal from a real-world object of interest. 

AutoX and ON Semiconductor appear to be on a successful road trip with their autonomous platform, but they are not the only players in this race. Audi had made significant strides and then canceled it. 


A Race to Autonomous Vehicle Technology Level-5

In 2020 Audi canceled its Traffic Jam Pilot system, the first-to-be-deployed European level-3 autonomous system. Its reason? Legislation.

Currently, in the event of an accident involving autonomous vehicles, the liability falls on the automaker regardless of the maintenance state of the vehicle (as maintained by the owner). It appears that this is not a financial burden that automakers wish to bear in European and North American markets.

Despite the legality setbacks, there is no doubt that this technology could be coming in the future, and along the way, legislation will have to catch up. While other companies continue to push forward the autonomous revolution cautiously, it seems that AutoX, with licenses to operate in the greater areas of China and the state of California, has the lead.