ON Semiconductor: Intelligent Perception Promotes the Development of Machine Vision

Industrial automation is closely related to artificial intelligence and machine vision. Machine vision technology can be said to be a branch technology of artificial intelligence. It is the world's intelligent "smart" eye image sensor, which affects the development and expansion of many applications, including industrial automation, robotics, security/surveillance, intelligent transportation systems (ITS), smart homes/buildings, smart retail, virtual reality (VR)/augmented reality (AR), etc. ON Semiconductor, the global leader in image sensors, has inherited more than 40 years of imaging experience and has more than 2,000 imaging patents. It provides a wide range of intelligent sensing lineups, including electronic rolling shutter front-illuminated (FSI)/back-illuminated (BSI) CMOS sensors, global shutter CMOS sensors, CCD sensors and image signal processors, with optical formats ranging from 1/13 inches to 35 mm and resolutions ranging from 100 K to 50 million pixels, meeting different application requirements and promoting innovation. At the same time, ON Semiconductor is also a leading global semiconductor supplier, providing semiconductor chips supporting power control modules, logic units, communication modules, 5G and wireless networks, industrial LANs and other fields.

Key imaging techniques

High dynamic range (HDR), global shutter, near infrared enhancement (NIR+), RGB-IR, and power scalability are all key imaging technologies that enhance the performance of embedded systems and machine vision systems. The higher the dynamic range, the more high-contrast, detailed images can be presented. The global shutter eliminates motion artifacts to the greatest extent. Key global shutter parameters include global shutter efficiency, readout noise, and sensitivity. ON Semiconductor's global shutter series products have best-in-class performance, enabling a wide range of applications, such as barcode scanning, 3D depth mapping and other demanding application scenarios. In addition to eliminating motion artifacts, they also have low noise, enhanced depth range, and fast capture. NIR+ technology can achieve 3 times better sensitivity than standard silicon processes. The use of RGB-IR can present richer and more realistic picture colors. Power scalability is a key advantage for battery-powered cameras, which can have different power consumption at different frame rates and continue to operate, achieving low-power energy-efficient design, such as low-power mode at 1 fps. Additionally, LED flicker reduction (LFM) technology in automotive applications eliminates high-frequency LED flicker in traffic lights and car lighting, contributing to safer driving.

Various machine vision application requirements

In the field of automation, robots require cost-effective global shutter solutions, assembly lines require high-speed global shutters, embedded/intelligent applications focus on deep learning and artificial intelligence, and flat-panel detection requires ultra-high resolution; for intelligent transportation systems (ITS), low-light performance and global shutter are important; for scanning applications, a cost-effective global shutter is required; in the field of motion analysis, an ultra-high-speed global shutter solution should be selected.

ON Semiconductor offers a wide range of vision solutions that use the key imaging technologies mentioned above to address the different needs and challenges of various embedded and machine vision systems.

Figure 1: Advanced product lineup for machine vision and edge AI

4.5 μm ITCCD Series: High Resolution and True Global Shutter

CCD technology can meet the image uniformity requirements of many industrial imaging applications; at the same time, the market requires continuous pixel development to achieve higher resolution. For example, ON Semiconductor's 4.5 μm ITCCD sensor KAI-43140 is suitable for demanding applications such as inspection and aerial photography. It adopts a 35 mm optical format and provides 43 million pixels. It uses the same package as KAI-29050 / KAI-29052 to simplify upgrades. The newly launched KAI-50140 is the highest resolution ITCCD image sensor on the market, providing 50 million pixels, providing the key imaging details and high image uniformity required for smartphone display inspection, circuit board inspection, mechanical assembly inspection, and aerial reconnaissance. KAI-50140 uses an aspect ratio of 2.18:1 to match the format of modern smartphones, reduce the number of image captures required to detect the full display, and use the same pin definition as the popular KAI-29050, KAI-29052 and KAI-43140 image sensors. The 4.5 μm series ITCCDs offer more than 50% higher resolution than the previous generation 5.5 μm series at the same frame rate while maintaining key imaging performance such as high dynamic range (>60 dB), suppressed smearing, CCD-level image uniformity, and electronic global shutter.

For the initial camera design, designers had to design and develop each image sensor from scratch, which was time-consuming, labor-intensive and costly. ON Semiconductor's ITCCD series shares a common architecture, enabling a single camera design to support a variety of different image sensors, and can be used in existing camera designs with only minor electrical changes, accelerating the time to market for new devices.

PYTHON series: 2 PCBs support 8 resolutions

ON Semiconductor's PYTHON series supports all 8 resolutions in the series (from VGA to 25 Mp) with only 2 PCBs, providing high-performance imaging and simplified upgrades for demanding industrial applications such as machine vision, intelligent transportation systems (ITS), security, surveillance, medical and science.

X-Class Platform: Enables a single camera design to support multiple product resolutions and different pixel capabilities

The X-Class platform is a further evolution of the PYTHON family, providing a scalable single device for multiple resolutions and pixels, enabling a single camera design to expand not only the number of pixels in the image sensor, but also the type of pixels used - whether it is a global shutter, rolling shutter, increased dynamic range, or other different functions. As long as the pixels are placed in the common high-speed and low-power frame used by the X-Class platform, the single camera design can be fully utilized to provide the required support, adopting a high-bandwidth, low-power architecture, which is ideal for applications such as factory automation, intelligent transportation systems (ITS), broadcast imaging, smart retail, and robotics.

Figure 2: X-Class Platform – Evolved Design Features

The platform's XGS series offers unmatched performance characteristics, outperforming the competition in image quality, power density, frame rate, supported platforms, 29 x 29 mm2 compatibility, customer support and price.

As the resolution increases, if the pixel size remains unchanged, the image sensor chip size will inevitably become larger and larger, and the lenses that match it will be expensive, which will increase the overall cost. Therefore, the market needs to reduce the pixel size. The first two devices of the X-Class platform, XGS 12000 and XGS 8000, use advanced 3.2 μm global shutter pixels, which can improve the resolution at the same optical size. The advanced pixel design ensures that the image quality is not affected by smaller pixels, and even provides better low noise and high performance. At the same time, the global shutter ensures that moving objects are captured without motion artifacts.

AR0237 RGB-IR: Enhanced Night Vision

The enhanced NIR sensitivity of the RGB-IR version AR0237 enhances night vision capabilities, allowing the camera to switch between day and night without special color filters. It is especially suitable for security applications. It has 2.1 million pixels, a resolution of 1928 x 1088, and can achieve 1080p/60 fps output. It provides excellent low-light performance and HDR, supports long and short staggered exposure with a high dynamic range, and uses infrared difference spectrum technology to avoid color distortion. It is also suitable for imaging applications in robots, AR/VR wearables, and smart buildings.

Global shutter AR0144 enables fast, clear and accurate imaging in static and dynamic scenes

The 3.0 μm pixel AR0144 uses an innovative global shutter design and is extremely compact. In static applications such as barcode scanners, the AR0144 can be easily designed into handheld products to ensure power saving, time saving and convenience for the initial scan. In dynamic applications such as drones, the sensor's low power consumption can extend the charging interval, and its imaging performance can better map the environment and avoid serious collisions. In 3D cameras and AR/VR designs, the AR0144 can provide a more cost-effective and small-size solution for capturing dual images for 3D mapping or depth perception. The AR0144 is a 1/4-inch, 1-megapixel CMOS sensor.

AR0431: Integrated NIR+ capability, simultaneous depth mapping

Enhanced NIR+ enables the AR0431 to achieve better imaging even in the darkest lighting conditions, helping security and IoT camera designers create highly sensitive night modes. Based on 2.0 um BSI pixels, it provides multiple low-power modes with a frame rate of up to 120fps, which is ideal for applications requiring slow-motion video shooting. Its low power consumption advantage enables it to provide 4-megapixel, 30 fps output with only 125 mW; in low-power monitoring mode, power consumption drops to 8 mW, making it ideal for battery-powered applications. The AR0431 uses Super Depth technology to simultaneously acquire color images and depth maps from a single image sensor, eliminating the need for another sensor that originally provided depth mapping capabilities, and realizing simple 3D models in AR/VR.

AR0522: High dynamic range combined with enhanced low-light performance

As one of the devices in ON Semiconductor's high-performance rolling shutter series, the AR0522 offers color and black-and-white color filter array (CFA) configuration options. It is 1/2.5-inch, 5.1 million pixels, based on a 2.2 μm BSI pixel technology platform. It is specially developed for applications that need to capture high-resolution, high-quality video in low-light conditions. Its sensitivity in near-infrared wavelengths is approximately twice that of the AR0521 image sensor.