Talking CMOS Image Sensors with SmartSens’ Leo Bai

CMOS image sensors are going into applications from machine-vision systems to smartphones. I talked with Leo Bai, SmartSens AI BU General Manager, about how the company’s image sensors are being used in machine-learning applications.

What are some examples of current mainstream technologies in CMOS image sensors?

Currently, there are two mainstream technologies in CMOS image sensors: FSI (front-side illuminated) and BSI (back-side illuminated).

FSI complies with the traditional semiconductor manufacturing process, so the FSI process is more mature and typically has lower production costs and higher yield. With pixels, light enters between the front metal wiring and then focuses on the photosensitive area. For larger pixels, FSI’s performance meets the requirements because the ratio between the height of the pixel's optical stack and the pixel area is small, and the photosensitive area is relatively guaranteed. However, the sensor’s performance and application scenarios will be limited.

With the rapid development of imaging technology in different application fields, people always want to use high-resolution CMOS image sensors with excellent performance in a small package size. That means the pixel size is reduced, the fill factor gets smaller, the optical path is elongated, and the metal wiring reflection absorption loss is larger—all of the factors that diminish performance.

The main advantage of BSI is that it separates electrical components from light, which allows the optical path to be independently optimized, avoiding absorption, reflection, and flare of the FSI metal wiring layer. The optical stack in the BSI pixel is also greatly reduced. Compared with FSI, BSI has a large fill factor of almost 100%, so BSI can achieve higher QE (quantum efficiency). However, the BSI manufacturing process is more complex and difficult than FSI.

In the early development of BSI, yield was a big challenge. With the advanced development of semiconductor technology, though, the BSI technique is becoming more and more mature, with a much higher yield.

What are the main applications for CMOS image sensors?

There are a variety of applications for different CMOS image sensor technologies.

The CMOS image sensor with BSI technology is suitable for applications that need high resolution with limited optical size, as well as small pixel size, high sensitivity, and low-light performance. For example, these applications include high-end security and surveillance, ITS (Intelligent Transportation System), FA (factory automation), and cell phones.

The manufacturing process of FSI is mature and low cost. The CMOS image sensor with FSI technology is more sensitive to system cost, rather than high performance. For example, these applications include smart-home and low-end video-monitoring products.

What are the customer benefits and technological advantages of using CMOS image sensors over CCDs, for example?

CMOS image sensors and CCDs are two different types of image sensor technologies used for digital imaging. CMOS image sensor technology has had great success due to its advantages over CCD.

First, CMOS image sensors are more cost-effective and power-efficient than CCD. CMOS image sensors are also easier to use, which means that users do not need to use external AD and design complex analog circuits. In addition, CMOS image sensors have a higher frame rate, which is a great feature not only for the FA market, but also for consumer products such as sport cameras and cell-phone cameras.

Can you share a few AI and machine-vision specific applications/case studies?  

Factory automation is one of the fields that has a high demand for AI. High-performance, industrial-grade global-shutter CMOS image sensors combined with BSI pixel design can provide better signal-to-noise ratio, higher sensitivity, and greater dynamic range. Such technology can be widely applied to improve factory-automated functions, such as inspection, quality control, optical character recognition, and robot arm guidance.

In other intelligent sensing applications such as drones, global-shutter CMOS image sensors are used for obstacle avoidance and optical flow. CMOS image sensors are also used for recognition-based AI applications like vSLAM in robotic applications, which makes gesture recognition, face recognition, and iris recognition more accurate and faster.

In addition, CMOS image sensor technology can be used beyond industrial applications. From barcode scanning (e.g., 1D or 2D code) at a grocery store to a lane-departure warning from an automobile, high-performance CMOS image sensor technology is greatly improving user experience and efficiency.

Furthermore, single-frame HDR Global Shutter technology is better for image-recognition-based AI applications than conventional CMOS image sensors that use Multiple-Exposure HDR technology. Combined with a DVP/MIPI/LVDS interface, single-frame HDR Global Shutter technology can be adapted to various types of SoC platforms.

What are the trends you foresee in the CMOS image sensor industry?

With the advanced development of emerging technologies like AI and machine learning, customers are trying to provide new product forms and better user experiences to meet market demands.