What is the Difference Between Transducers and Sensors?

CMOS (Complementary Metal-Oxide-Semiconductor), Chinese name is complementary metal oxide semiconductor, it is an important chip in the computer system, which saves the most basic information of the system boot. CMOS manufacturing technology is no different from general computer chips. It is mainly a semiconductor made of two elements, silicon and germanium, which coexist with N (charged) and P (charged) grades on CMOS. Semiconductor, the current generated by these two complementary effects can be recorded and interpreted as an image by the processing chip. Later, it was found that CMOS can also be used as an image sensor in digital photography after processing. CMOS sensors can also be subdivided into passive pixel sensors (CMOS) and active pixel sensors (CMOS).

CMOS sensor

Today's CMOS image conversion technology not only serves "traditional" industrial image processing, but is also being accepted by a growing number of new consumer applications for its superior performance and flexibility. In addition, it ensures high safety and comfort when driving a car. Initially, CMOS image sensors were used in industrial image processing; it remains a vital part of new automation solutions designed to increase productivity, quality, and economics of production processes.

According to the forecast of market research company IMS Research, in the next few years, the annual growth rate of the European industrial image processing market will reach 6%. Among them, the market share of intelligent solutions integrating software functions in cameras will continue to expand. . In Germany, according to data provided by its National Machine Tool Suppliers Association VDMA, the image processing market grew by 14% in 2004. The market research company In-Stat / MDR also pointed out that, as far as the secondary market of image sensors is concerned, its annual growth rate will be as high as 30%, and this situation will continue until 2008. The most important thing is that the growth rate of CMOS sensors will reach seven times that of CCD sensors. The rapid popularization of camera phones and digital cameras is the main driving factor for this demand.

Obviously, people are so optimistic about the growth prospects of CMOS image converters based on the fact that, compared with the CCD technology that has monopolized this field for more than 30 years, it can better meet users' continuous demands for new image sensors in various applications Improved quality requirements, such as more flexible image capture, higher sensitivity, wider dynamic range, higher resolution, lower power consumption, and better system integration. In addition, CMOS image converters have created novel applications that have not yet been economically implemented. In addition, there are several "soft" standards that are beneficial for CMOS sensors, including: application support, radiation resistance, shutter types, windowing, and spectral coverage. However, this distinction is somewhat arbitrary because the importance of these standards will vary depending on the application (consumer, industry, or automotive). ^[1]

As we know from analog photography, taking a picture of a complete scene is quite common, and so is a camera phone. However, for industrial or automotive applications, the situation is quite different: there are occasions when a high full frame data rate is not required. For example, in surveillance cameras, as long as a change in a scene can be found (because this change may indicate a suspicious situation), a lower resolution is completely acceptable. On this basis, it is necessary to use the full resolution to collect more detailed information. The actions that follow will only be played in a certain part of the camera's field of view, and in the captured scene, only this part is the focus of the monitoring staff.

For a CCD image sensor that only provides a full frame image, only a separate evaluation circuit can provide two observation angles, which means an increase in processing time and cost. However, the working principle of CMOS image sensor is similar to that of RAM. All memory bits can be read separately. Although the sub-sampling of the CMOS sensor provides a lower resolution, the frame rate is higher; the windowing allows random selection of a region of interest. ^[2]

A prerequisite for the widespread use of the latest CMOS sensors is their higher sensitivity, shorter exposure times, and shrinking pixel sizes. One measure of pixel sensitivity is the product of the fill factor (the ratio of the light-sensitive area to the entire pixel area) and the quantum efficiency (the number of electrons generated by the photons that bombard the screen). CCD sensors have a large fill factor due to the inherent characteristics of their technology. In the CMOS image sensor, in order to achieve a noise index and sensitivity level comparable to that of a CCD converter, people have equipped the CMOS image sensor with an active pixel sensor (APS) and the fill factor has been reduced because the pixel surface is equivalent A large area is occupied by the amplifier transistors, leaving less space available for the photodiode. Therefore, an important development goal of today's CMOS sensors is to expand the fill factor. Cypress (FillFactory) through its patented technology can greatly increase the fill factor, this technology can turn the largest part of a standard CMOS silicon chip area into a photosensitive area. As the pixel size becomes smaller, it becomes more difficult to increase the fill factor. At present, the most popular technology is to change from the traditional front-side illumination (FSI, Front Side Illumination) to the back-side illumination (BSI, Back Side Illumination), amplifier, etc. The transistors and interconnects are placed on the back, and the front is left for the photodiode, so that a 100% fill factor is achieved (as shown in the diagram on the right).

In modern CMOS image sensors, an important development trend is that its spectral sensitivity has been extended to the near-infrared region NIR (to a wavelength of about 1,100 nm). Automotive applications equipped with the IM-001 CMOS image sensor will improve fog penetration and night vision capabilities. As industrial image capture technology begins to use more light sources located in the NIR, and biotechnology is also using interesting phenomena in this spectral region, the newly developed IBIS 5-AE-1300 sensor has a NIR sensitivity of 700 to 900 nm .

In consumer-oriented image capture technology, another trend is to continue to increase resolution. By mid-2005, about 70% of mobile phone cameras already had VGA format resolution (640 × 480 pixels); but in 2006, multi-megapixel sensors will occupy 50% of the market share, and by 2008 , Its market share is expected to further climb to more than 90%. To this end, Cypress has developed a 3-megapixel image sensor for cellular phones. This product uses Autobrite technology, which can perform 12-bit analog / digital conversion and provides a wide dynamic range of 72dB. The 10-bit analog / digital converter has a dynamic range of only 60dB. Frame rates up to 30 frames per second in progressive scan mode allow recording of live video programs.

This trend is also evident in the industrial and commercial sectors: Cypress has launched a 13-megapixel / 35mm image sensor for Kodak digital cameras, and a 6.6-megapixel IBIS 4-6600 sensor is being launched. Proof of excellence in an automatic reading aid for the visually impaired-it provides excellent resolution on a complete standard A4 page.

Relying on technology to achieve system integration Due to the accelerated digital convergence of traditional discrete functional devices such as cellular phones, digital cameras, MP3 players and PDAs (ie, becoming a compact consumer electronics product), people increasingly want at least partial autonomy The sexual subsystem can provide a very wide range of functions in a device. This trend will also have an impact on professional measurement technology: With a portable inspection tool that includes a digital camera, PDA user interface and WLAN networking capabilities, the range of applications for optical testing and monitoring will be effectively expanded. As a platform technology, CMOS is in line with this development trend: CCD image converters still need external logic circuits to implement control and analog / digital conversion functions, while CMOS standard logic devices can integrate sensors, controllers, converters and evaluation Logic circuits are all integrated into one chip.

A typical example is the image capture circuit of the CYIWCSC1300AA chip made specifically for demanding consumer applications. It is based on the 1.3-megapixel image sensor CYIWOSC1300AA and one for providing error interpolation, black level adjustment, lens correction, signal inter-segment correction, color mosaic repair, color correction, automatic exposure, noise suppression, special effects and gamma correction, and many more Functional additional signal processor. Integrating more system functions (up to autonomous photoelectric sensor systems) is feasible, depending on economic goals and constraints such as market capacity and development costs.

John Morse, senior market analyst at IMS Research, said: "The industrial image processing market is changing very fast, not only at the technical level, but also in the context of recent mergers of manufacturers. We believe this trend will continue. "If so, then the same applies to Cypress: By acquiring MIT (Massachusetts Institute of Technology) established SMal Camera Technologies in 1999, Cypress has extended its reach to the consumer and automotive sectors. ; And the merger of FillFactory, a company that was spun off from IMEC, a well-known European microelectronics and nanotechnology research center headquartered in Leuven, Belgium, in 1999, made Cypress further into the industrial field.

The CMOS image sensor market is booming and is about to become a large-scale market. It still relies heavily on customer-specific designs to meet a set of customized requirements in terms of specifications and system integration. However, it will increasingly provide a universal standard solution. Increasing resolution, frame rate, and sensitivity, as well as falling costs, are expanding its applications. The important link.

CMOS sensors are divided into passive and active based on pixel structure.

Noise

This is the number one issue affecting the performance of CMOS sensors. Such noise includes fixed pattern noise (FPN), dark current noise, thermal noise, and the like. The reason for the fixed pattern noise is that the output signal produced by the same light on two different pixels is not the same. This is how noise is introduced. To deal with fixed pattern noise, double sampling or correlated double sampling can be applied. It is a bit like introducing differential pairs to suppress common-mode noise when designing analog amplifiers. Double sampling is to first read out the charge integration signal generated by light, temporarily store it and then reset the pixel unit, and then read the output signal of this pixel unit. Subtract the two to get the image signal. Both samples can effectively suppress fixed pattern noise. In addition, the correlated double sampling requires a temporary storage unit, and as the pixels increase, the storage unit also increases.

Dark current

Physical devices cannot be ideal. Like the sub-threshold effect, due to impurities, heat and other reasons, even if no light is irradiated to the pixel, the pixel unit will generate charges, and these charges generate dark current. It is difficult to distinguish between dark current and charge generated by light. Dark current is not exactly the same across the pixel array, it causes fixed pattern noise. For pixel units with integration function, the fixed pattern noise caused by dark current is directly proportional to the integration time. The generation of dark current is also a random process, which is a source of shot noise. Therefore, the magnitude of the dark current generated by the thermal noise element is equal to the square root of the number of dark current electrons in the pixel unit. When a long-term integration unit is used, this type of noise becomes the main factor affecting the quality of the image signal. For dark objects, long-term integration is necessary, and the pixel unit capacitance is limited, so it is dark. The accumulation of current electrons limits the maximum integration time.

In order to reduce the influence of dark current on the image signal, firstly, a cooling method can be adopted. However, just cooling the chip is not enough. The fixed pattern noise caused by dark current cannot be completely overcome by double sampling. The effective method now adopted is to subtract the reference dark current signal from the obtained image signal.

3. Pixel saturation and overflow blur

Similar to the amplifier, there is an upper limit on the input due to the limited range of the linear region. For a CMOS image sensor chip, it also has an upper limit on the input. If the input optical signal exceeds this upper limit, the pixel unit will be saturated and cannot perform photoelectric conversion. For pixel units with integrating function, this upper limit is determined by the capacity of the optoelectronic integrating unit: For pixel units without integrating function, the upper limit is determined by the maximum current flowing through the photodiode or triode. When the input optical signal is saturated, overflow blur occurs. The overflow blur is due to the saturation of the photoelectrons of the pixel unit and then flows out to the adjacent pixel unit. The overflow blur is reflected in the image as a particularly bright area. This is somewhat similar to overexposure on photos. The overflow blur can be overcome by adding an automatic bleeder tube in the pixel unit, which can effectively discharge the excess charge. However, this only limits the overflow, but cannot make the pixels truly reproduce the image.

There are many manufacturers investing in CMOS R & D and production. There are more than 30 in the United States, 7 in Europe, about 8 in Japan, 1 in South Korea, and 8 in Taiwan. The world's leading manufacturer is Agilent (HP), with a market share of 51%, ST (VLSI Vision) 16%, Omni Vision 13%, Hyundai 8%, Photobit approximately 5%, these five companies together The market share reached 93%.

Sony

Sony is the world's largest manufacturer of CCD sensors and the first company to invest in 12-inch wafers and launch a 6-megapixel CCD. About 30-40% of Sony's CCD sensors are for private label products, and others are sold to Canon, Sanyo, Casio, and Taiwan's Xinhong, Purier, and Quanxun (merged with Canon of Taiwan).

Sony's product technology blueprint shows that in 2003, except for the 8-megapixel ICX 456, no other miniature products were available. Product size will remain roughly at the current level, replaced by enhanced photography functions and support for progressive scan, such as ICX455 / 465 for 5 megapixels, ICX451 / 481 for 3.3 megapixels, and 2.1 million megapixels ICX461, etc., enable high-end products to reach data transfer rates of more than 30fps.

Most of the high-end products market is still occupied by Sony, coupled with the market is still in short supply, the company is not eager to reduce costs, but once Sony's most advanced technology (pixel size 2.6 ~ 2.8um) reaches maturity Stage (finished product rate is over 50%), the company is bound to further apply this process to other products (currently only 1 / 1.8 inch, 5 million pixel products use this process), there may be 1 / 2.7 inch, 4 million pixel products came out.

OmniVision

OmniVision was established in 1995 (hereinafter referred to as OV). In June 2002, it was ahead of other peers and was the first to launch the 2.1 million pixel OV2610 shocking the market. Although there are not many products currently using this sensor in mass production, this has shown that CMOS sensors can begin Entered the mid-to-high-end digital camera market that originally belonged to CCD sensors; OV data shows that currently Taiwanese merchants such as Tianhan, Ming and Hongyou have started to use the company's OV2610. Looking forward to 2003, OV will launch 3.3-megapixel, 1 / 2-inch products between the first quarter and the second quarter, which will be produced by TSMC 0.18mm process, which will expand the application range of CMOS sensors again. In the mobile phone market, CMOS camera modules have become the largest products for mobile communication applications.

In terms of low power consumption products, OV also glimpsed V7640 in 2002, which can run in a 2.5V environment and is the lowest power consumption chip in current VGA products. In terms of newly planned products in 2003, OV plans to launch 1.3 million pixel, 1 / 4-inch, and VGA, 1 / 7-inch products in the second half of the year. It is hoped that before CCD manufacturers launch low-power 1.3 million pixel products, Seize the market first.

Agilent

Agilent's main products are the second-generation CIF (352 * 288) HDCS-1020 and the second-generation VGA (640 * 480) HDCS-2020, which are mainly used in emerging information such as digital cameras, mobile phones, PDAs, and PC cameras. Among home appliances, in addition, another successful strategy of Agilent in 2000 was to strategically enter into the field of optical mouse products with Logitech and Microsoft. However, this is a very low-end CMOS product, and it is not intended to capture images. Therefore, when doing global statistics of image sensors, this number was not added together, but this move shows that Agilent's planning intention to enter optical components based on CMOS technology.

Aptina

Photobit achieved great success in 2000. In 2001, Photobit pioneered the development of the PB-0330 product type CMOS image sensor. This product features a single-chip logic-to-digital converter. It is the second-generation 1 / 4-inch VGA (640 x 480). It also introduced PB. -0111 CMOS image sensor is the second generation 1/5 inch CIF (352 x 288). Photobit launched these two products, mainly aimed at digital cameras and PC Camera, which have been booming in recent years. DigiVision CIF (352 x 288) is positioned in the mobile phone market. And VGA (640 x 480) image sensors with different resolutions, the marketing scope is intended to cover the low-end and high-end markets. Photobit was later acquired by Micron. After that, Micron separated the image sensor department and established Aptina.

panavision

High-quality industrial CMOS image sensor, the main product is DYNAMAX-11. This new sensor contains global electronic exposure shutter technology, which greatly improves the application of industrial imaging indoors and outdoors. This newly released DYNAMAX-11 image sensor is suitable for industrial imaging fields such as machine vision, security monitoring, intelligent transportation, life sciences, biomedicine, scientific imaging, high-definition video recording, and television broadcasting. This newly released DYNAMAX-11 image sensor contains 3.2 million pixels with a pixel size of 5.0um × 5.0um.

Other companies

The most distinctive feature is Sanyo. The company is committed to improving the power consumption of CCD sensors, with camera phones as the main application target. The Sharp J-SHxx series, which was first introduced by J-Phone earlier, uses Sanyo's CIF-grade CCD sensors, Sharp, Toshiba and other mobile phone manufacturers also plan to gradually introduce Sanyo's VGA products between the fourth quarter of 2002 and the first quarter of 2003. Matsushita and Sharp's product plans are similar to Sony's. The main difference is that Matsushita plans to launch smaller 4 million pixel (1 / 2.7 inch) and 1.3 million pixel (1/4 inch) products.

Experts believe that the global CMOS image sensor market will grow significantly in the fields of PC cameras, mobile communication markets, digital cameras, and video camera markets in the early 21st century. In the next few years, between 1.3 million pixels and 2 million pixels Among the products below, CMOS sensors will become the mainstream. Cameras with miniaturized and low-power CMOS image sensors as the core are becoming the mainstream of consumer products, and the above fields will bring huge development to the image sensor market.

On August 28, 2009, Sony's Autumn Digital Imaging Product Launch Conference was grandly held in Beijing. Sony announced the launch of a total of ten new digital imaging products in three product lines. Among them, DSC-TX1 and DSC-WX1 applied the new image sensor for the first time

The sensor architecture can consist of a two-part, a quarter, or a pixel array. The output can be a parallel analog output, or a 10-bit digital output or a digital serial LVDS output. Each output can be sampled at up to 50 million times per second, which enables throughput of 5.5 billion pixels per second. To date, this image sensor is the one with the highest continuous pixel throughput. The image quality reaches at least 10-bit accuracy, so after the camera is digitized, the data throughput can be 55Gbit per second. Such high-speed applications usually require 6 transistor snapshot pixels, and require higher sensitivity and dynamic range. The sensitivity of an image sensor largely depends on the pixel size, and large pixel sizes require large areas of custom image sensors for specific applications. Internal multiplexing technology can support higher frame rate random windows. If the window size is reduced to a smaller ROI (circle the target area), the frame rate of the fastest device can reach 170,000 frames per second. Most sensors use a 0.25 process.

At present, CMOS is a favored technology for high-speed imaging. In the current market, we can find three development trends of high-speed image sensors, one is to the direction of extremely high speed, the other is to the direction of on-chip feature integration, and the third is to the general high-speed image sensor.

The combination of resolution and frame rate plays an important role. At present, we can launch a 1024 × 1024 pixel image sensor with a working speed of 5,000 full frames per second. If the analog-to-digital conversion is 10 bits, this means that the total data rate on the camera can reach 55Gbit per second. In order to achieve extremely high data rates and high image quality on the sensor, especially for this highly sensitive application, we must not only design the correct electronic circuit, but also ensure that the entire circuit layout achieves a good balance. That is to say, the power line should achieve excellent distribution, and all optical and stray light sensitive responses of each line node in the layout should be well controlled. A low-power module design is required to ensure that the overall power requirements are met.

Another trend in the high-speed imaging field is the on-chip integration of high-speed ADCs, timing generators, LVDS transmitters, and correction algorithms. Such image sensors are generally inferior to the above-mentioned image sensors in terms of speed and sensitivity, but have advantages in terms of ease of use and system integration functions. The third image sensor emerging on the market is the general high-speed image sensor. Older (simple) general-purpose image sensors with analog output or without a timing generator are being replaced by faster, more complex image sensors. This new image sensor allows us to design a universal high-speed camera in a short time. ^[6]

A common phenomenon now is that when smartphones update and iterate, they will always bring new camera solutions, which will provide a vast and unlimited market for CMOS image sensor suppliers. According to the latest market research report released by Yole Development, the CMOS sensor industry will prosper in the next 5 years. The compound annual growth rate from 2015 to 2021 is expected to reach 10.4%, and the total market value is expected to reach $ 18.8 billion. ^[7]