“The ADAS front-view camera is required to detect farther vehicles and pedestrians, and at the same time perceive more refined content for the algorithm platform to recognize and process. The maximum detection distance of products with 2 million pixels and below is 120m, which is far from enough to accurately identify distant objects when driving at high speed. Therefore, the entire industry is actively developing and deploying 8-megapixel in-vehicle cameras with a maximum detection distance of up to 250m.

“

OmniVision CIS series special ADAS articles (2) Image sensing technology under the trend of high resolution

introduction:

The combination of consumers, policies/regulations, and the market continues to promote the penetration rate of ADAS. In March 2022, the L2-level supporting volume of domestic new cars was close to 416,000, with a penetration rate of nearly 30%. ADAS becomes standard, how to upgrade CIS?

OmniVision Automotive CIS series special ADAS articles will be divided into 5 issues, each with a hard-core knowledge point, to help you understand the application, technology trends and automotive products in the field of ADAS in detail.

The last issue of popular science explained why the image sensor chip is difficult to get on the car, and how OmniVision provides all-round guarantees in terms of reliability and functional safety. This issue will focus on image sensing technology under the ADAS high-resolution trend.

In terms of resolution, the initial automotive image sensors are mainly VGA products to 2 million pixels. The “Analysis Report on the Technical Demand for Image Sensors in ADAS and Autonomous Driving to 2023” released by the British market research company Smithers Apex in 2018 pointed out that the image sensor needs to achieve a monocular resolution of 8 million pixels.

VGA-2 million pixel stage

Since the birth of the first commercial vehicle equipped with an in-vehicle camera in 1991, the early automotive CIS was mainly based on VGA products, which were mostly used for rear view to assist drivers in reversing. The rear view camera does not focus on resolution, low light performance and compact size, so the image sensor in this field has always been dominated by VGA to 1M products.

In 2006, the 360-degree panoramic view entered people’s field of vision, and began to enter the stage of rapid development and initial application. At this stage, the CIS needs to reach more than 1M to adapt to the wide-angle lens and reduce deformation.

For VGA-2M products, OmniVision entered the game early, with complete products and solutions that have been verified by customers for a long time. Since proudly entered the vehicle market in 2005, it has closely followed the market demand, and has jointly defined and developed vehicle CIS with old European and American luxury car brands and Tier-1 first-class material suppliers. Top supply chain partners produce all vehicle-spec products to ensure advanced and reliable production process and technology.

In 2005, the car cameras on the market were dominated by CCD, and CMOS accounted for a small proportion. In response to customer demand for smaller, more cost-effective CMOS image sensors, OmniVision released the first VGA CMOS image sensor for vehicles. In 2008, at the beginning of the demand for 360-degree surround view, OmniVision released the industry’s first 1-megapixel vehicle-mounted CIS OV9710.

In the following ten years, OmniVision has continued to optimize and iterate the VGA-2M image sensor according to customer needs. The products have been proven by the market for their high quality and high reliability. The VGA product OV7955 released in 2012 is still recognized by the market ten years later and has a lot of shipments. The OV1065X and OV1064X series released in 2016 provide different color fitter (RCCC/RCCBRGGB) solutions, which are the mainstream 1.3M and 1.7M ADAS solutions in the industry. Compared with similar products of competitors, OmniVision image sensors have prominent advantages in market verification and supply chain.

high resolution stage

The ADAS front-view camera is required to detect farther vehicles and pedestrians, and at the same time perceive more refined content for the algorithm platform to recognize and process. The maximum detection distance of products with 2 million pixels and below is 120m, which is far from enough to accurately identify distant objects when driving at high speed. Therefore, the entire industry is actively developing and deploying 8-megapixel in-vehicle cameras with a maximum detection distance of up to 250m.

Based on the mature design and on-board experience of vehicle-regulated products, and relying on the leading technologies in high-resolution markets such as mobile phones and security, OmniVision launched the first high-end 8-megapixel product OX08B40 with LFM in the world in 2019, helping vehicles quickly , Accurately identify road conditions such as road signs and traffic signals, filling the gap of domestic 8 million pixel vehicle CIS products.

/tips/

The development cycle of automotive products is extremely long, generally at least 2-3 years. Compared with consumer-end CIS, automotive-regulated products have stricter requirements on quality and reliability, and carry more functions. OmniVision’s high-resolution products are produced by top supply chain partners with advanced processes and techniques to ensure reliability. The OX08B40 is the only 8MP CIS product in China that complies with ASIL-C functional safety and AEC-Q100 level 2 certification, and it is also the most mainstream ADAS 8MP CIS product used by automakers around the world. It is reported that domestic friends have not yet achieved mass production of similar products.

Why are there very few 8 million vehicle products in the industry? First of all, market demand has only appeared in recent years. Secondly, high resolution means more pixels. The other side of the coin is that products in the era of intelligence are getting smaller and smaller. How to achieve a smaller size, ensure the imaging effect, and carry more functions while increasing the pixels is a difficult problem.

In order to solve this pain point, OmniVision continues to innovate the vehicle image sensor technology, and has gone through four generations of OmniPixel®3-HS, OmniBSI™, OmniBSI™-2, and PureCel®Plus processes, and the pixel has developed from 4.2um to 2.1um at the same time. , which greatly improves Pixel performance and adds more features.

The OX08B40 series adopts OmniVision’s latest PureCel®Plus-S pixel architecture. The technology uses a stacked architecture and introduces features such as Buried Color Filter Array (BCFA), Deep Trench Isolation (DTI) to enhance sensor technology.

1) Stacked architecture: This technology divides the imaging array and processing functions into two layers, enabling smaller die size and additional functions.

2) BCFA: BCFA significantly improves the collection latitude of light at different incident angles while making the design more compact.

3) DTI: Crosstalk is reduced by setting isolation between pixels within the silicon wafer for better tolerance of the chief beam angle (CRA).

With the above innovative architecture, smaller chip size and lower power consumption can be achieved without sacrificing performance – better low-light performance, imaging quality, dynamic range, and more functions and algorithms, such as HALE (HDR) and LFM combination algorithm) and network security functions.