From CCD to CMOS
According to the consulting machine forecast, from 2021 to 2025, the global shipment of CMOS image sensors will continue to remain at 8. 5% compound annual growth rate is expected to reach 116% in 2025. 400 million. Not long ago, Sony announced the successful development of the world's first double-layer transistor pixel stacking CMOS image sensing technology, and said that the new architecture has increased the saturation signal level by about 1x, broadening the dynamic range and reducing noise, which can significantly improve imaging performance.
And how much do you know about image sensors?
Avoid missing out on in-depth information and reports on the sensor industry, as well as sensor technology, sensor and test technology, IoT sensor technology... and other sensor knowledge, please pay attention to the sensor expert network public number, set as a star, view the previous content.
Image sensors, or photosensitive elements, are devices that convert optical images into electronic signals, and they are widely used in digital cameras and other electro-optical devices. Early image sensors used analog signals, such as video tubes.
With the rapid development of digital technology, semiconductor manufacturing technology and network, the market and the industry are facing the advent of cross-platform video, audio and video, communication integration era, image sensing technology has also been greater development. In recent years, with the rise of artificial intelligence, image sensors, as one of the input signals, have supported many important applications such as face detection, industrial anomaly detection, gesture recognition and so on.
1
Classification of image sensors
Image sensor products are mainly divided into three types: CCD, CMOS and CIS sensors.
CCD is a high-end technical component applied in photography and video, CMOS is used in lower image quality products, its advantage is that the manufacturing cost is lower than CCD, power consumption is much lower, which is why many of the market using USB interface products do not need external power supply and the price is cheap. CiS is a contact image sensor. It uses contact-type photosensitive elements for photosensitivity, and the scanner made with CIS technology has the advantages of small size, light weight, low production cost, etc., and is widely used in fax machines, scanners, banknote clearing and zeroing and other fields. However, CIS technology also has shortcomings, such as can not be made into a high-resolution scanner, the scanning speed is relatively slow, so the current widely used in the market is mainly CCD, CMOS two kinds of image sensor devices.
Although there are large differences in technology, the performance gap between CCD and CMOS is not very large. CCD can be divided into two types: linear type and surface type, of which the line type is used in image scanners and fax machines, while the surface type is mainly used in digital cameras, camcorders, surveillance cameras and other image input products. In general, CCD sensors have the advantages of high resolution, low noise, wide dynamic range, good linear characteristic curve, high photon conversion efficiency, large area sensitivity, wide spectral response, low image distortion, small size and light weight, low power consumption, good charge transmission efficiency, not affected by strong electromagnetic fields, and can be mass-produced.
Compared with CCD, CMOS sensors adopt the CMOS process most commonly used in general semiconductor circuits, which have the characteristics of high integration, small power consumption, fast speed and low cost, and have developed rapidly in recent years in terms of wide dynamic and low illumination. CMOS is mainly a semiconductor made of two elements, silicon and germanium, and realizes the basic function through the transistors with negative and positive electricity on the CMOS. The current generated by these two complementary effects can be recorded and interpreted as an image by the processing chip.
Among analog cameras and standard definition network cameras, CCDs are the most widely used and have long dominated the market. CCD is characterized by high sensitivity, but low response speed, not suitable for high-definition surveillance cameras used in high-resolution progressive scanning mode, so after entering the era of high-definition surveillance, CMOS is gradually recognized, high-definition surveillance cameras generally use CMOS photosensitive devices.
2
Compete in the app market
Since the beginning of the 1990s, CMOS image sensing technology has been valued in the industry and has obtained a large number of R&D resources, continuously squeezing the market space of CCD image sensors in applications. CMOS image sensor chips have become one of the important components of global integrated circuits, and their main advantages can be summarized into the following three levels:
Cost: CMOS image sensor chips generally adopt standard process processes suitable for large-scale production, and the unit cost in mass production can be much lower than CCD;
In terms of size: CMOS sensors can integrate image acquisition units and signal processing units on the same substrate, and the volume has been greatly reduced, making it very suitable for mobile devices and various types of miniaturized equipment;
Power consumption: CMOS sensors maintain the advantages of low power consumption and low heat generation compared with CCDs.
Global image sensor market size (shipment caliber)
(Source: Frost & Sullivan)
It is understood that with the continuous innovation of image sensing technology and the gradual expansion of the application scale of downstream industries, the global image sensor market size has shown a continuous growth trend in the past decade. From 2016 to 2020, global image sensor shipments increased from 46% in 2016. 900 million rapid growth to 8.26 billion in 2020, with a compound annual growth rate of 15.5% during the period. According to Frost & Sullivan, the broader and deeper demand for image sensors in various application areas is expected to continue to help it maintain a certain growth rate. Global image sensor shipments are expected to reach 120.5% in 2025. 800 million, the compound annual growth rate between 2021 and 2025 is expected to reach 7. 9%。
In recent years, various head CMOS image sensor designers have continuously promoted back-illuminated and stacked technologies, so that CMOS image sensors have gradually penetrated from the smartphone market dominated in the past to many market segments such as automotive, security surveillance, medical, VR and industrial.
Thanks to the rapid development of security surveillance, automotive electronics and other emerging fields and the widespread popularity of multi-camera mobile phones, the overall shipment and sales of CMOS image sensors have continued to expand. From 2016 to 2020, global CMOS image sensor shipments increased from 41. 400 million rapid growth to 7.72 billion, during the period of compound annual growth rate of 16. 9%。 It is expected that from 2021 to 2025, global shipments of CMOS image sensors will continue to remain 8. 5% compound annual growth rate is expected to reach 116% in 2025. 400 million.
Global CMOS image sensor market size (shipment caliber)
3
Artificial intelligence and image sensors
With the advent of the era of artificial intelligence, corresponding chip products and industries have also emerged in a corresponding new direction. Among the various branches of artificial intelligence, machine vision is undoubtedly the most widely used direction. Machine vision is the use of machine learning/artificial intelligence methods to analyze visual signals, and the results of the analysis are directly produced through artificial intelligence. Therefore, machine vision naturally needs an image sensor as an input signal, and with the gradual development of machine vision and artificial intelligence, the combination of machine vision and image sensor chips has become an "intelligent image sensor", which conforms to the context of technological development.
Image sensors provide an input signal for the most important application of machine vision. In traditional machine vision chip solutions, image sensing and artificial intelligence algorithms are run on different hardware, with image sensors providing image signals and processors or AI-accelerated chips executing artificial intelligence algorithms. However, such an approach will cause waste of energy in mobile terminals or IoT smart devices that emphasize low power consumption and energy efficiency ratio, and it will be difficult to deal with some application scenarios that require always-on. This can be solved by artificial intelligence-enabled image sensors.
Specifically, such intelligent image sensors integrate the operation module of artificial intelligence algorithms in the image sensor module, so the results of the machine vision algorithm can be directly output, and the method of running machine vision in this way is often referred to as "in-sensor computing". Through in-sensor computation, the operating unit of the machine vision algorithm is changed from the SoC to the image sensor, so when running the machine vision algorithm, the SoC does not need to be in a wake-up state or running the operating system, but can be in a low-power standby state. On the other hand, because the entire system is relatively simple, there is no additional overhead such as running the operating system, and there is an accelerator chip module tailored for the machine vision algorithm, so the efficiency ratio can be very high. Finally, in terms of interface, it is usually possible to wake up the SoC in an interrupting manner by the intelligent image sensor when it detects an important output, and only needs to transmit the results of the operation of the machine vision algorithm, so the overhead of data transmission is also greatly reduced.
Conclusion
We can see that with the changes in the general environment, different stages, the integration of different technologies has also brought new developments to image sensors, from the half of the CCD, to the latecomers of CMOS, and the birth of "intelligent image sensors", image sensor technology continues to empower the industry at the same time, but also continues to enrich and expand its own connotation, to achieve more and more extensive scene applications.
The content of the star = the content you care about, according to the principle of WeChat public account recommendation, without the sensor expert network public account home page [set as a star], will not be able to get the latest information of the sensor industry in time!
Source: Intelligent Manufacturing Vertical and Horizontal 2022.02
Disclaimer: The content of this article is the personal opinion of the author and does not represent the views or positions of the Sensor Expert Network. For more views, please leave a comment.