From photoelectric conversion to image generation: essential differences in imaging principles

In many application scenarios of industrial automation, machine vision, and scientific research, industrial cameras, as the core equipment for acquiring image information, directly affect the accuracy and reliability of the entire system in terms of their performance. The core component that determines the performance of industrial cameras is the image sensor, among which CCD (Charge Coupled Device) and CMOS (Complementary Metal Oxide Semiconductor) are the two mainstream technology routes. Although both are based on the same principle of photoelectric conversion, which utilizes the photoelectric effect of semiconductor materials to convert photons into electrons, there are fundamental differences in signal processing and transmission methods.

The design concept of CCD sensors is to centrally process photoelectric signals: when light shines on the pixel array, each pixel generates a charge packet proportional to the light intensity. These charge packets require a complex transfer process - under precise clock pulse control, pixel charges are shifted line by line to a single output node (or a very small number of output nodes) at the edge of the chip, where charge to voltage conversion and signal amplification are performed. This design ensures that all pixel signals pass through the same signal path, ensuring a high degree of consistency in signal output.

In contrast, CMOS sensors adopt an innovative architecture of distributed processing. On CMOS chips, each pixel not only contains a photodiode, but also integrates independent miniature amplifiers and analog-to-digital conversion circuits. This design enables each pixel to convert charges into voltage signals on-site, and directly read them through a network of intersecting row and column wires. Although this structure greatly improves read speed and reduces power consumption, the performance differences between millions of miniature amplifiers inevitably bring about signal consistency issues.
This fundamental difference in signal transmission has led to a series of performance differences between the two technologies in industrial camera applications. Understanding the difference between CCD's "sequential shift and centralized output" and CMOS's "parallel conversion and distributed reading" is the foundation for grasping all subsequent differences between the two.

Comparison of Five Core Performance Factors: Noise, Power Consumption, Resolution, Sensitivity, and Cost

2.1 Noise Performance and Image Quality

CCD sensors have the advantage of noise control due to centralized signal processing. Due to all pixels sharing the same (or very few) output amplifier, amplification differences between pixels are avoided. This design, combined with mature PN junction or silicon dioxide isolation layer technology, effectively reduces the generation of fixed pattern noise, thereby providing purer and more consistent output in image quality. Especially under long exposure or low light conditions, CCD sensors can still maintain low noise levels, making them highly favored in precision measurement and low light imaging applications.
In contrast, each pixel of a CMOS sensor is equipped with an independent signal amplifier. Although this design improves the read speed, the small performance differences between millions of amplifiers result in fixed pattern noise. This noise is manifested as fixed pattern interference on the image, especially in uniformly illuminated scenes. However, with the advancement of CMOS technology, modern industrial grade CMOS cameras have significantly improved this issue through correlated double sampling (CDS) and digital correction algorithms, and some high-end products have approached or even reached the image quality level of CCD.

2.2 Energy Efficiency and Power Consumption Differences

In terms of power consumption, CMOS exhibits significant advantages. CMOS adopts an active image acquisition method, where the charge generated by the photosensitive diode is directly amplified and output by the adjacent transistor. The entire sensor only requires a single power supply, and the typical power consumption is only 1/8 to 1/10 of similar CCDs. This feature makes CMOS the preferred choice for energy sensitive applications such as portable devices, embedded systems, and multi camera arrays.
The high power consumption of CCD is due to its passive charge transfer mechanism. It requires three sets of power supplies with different voltages (usually 12-18V) and a complex clock control circuit to drive the shift transfer of charges. This not only increases the complexity of power supply design, but also brings about heat dissipation issues - when working at high resolution or high frame rate, the temperature rise of CCD will further increase thermal noise, forming a vicious cycle. Therefore, industrial systems using CCD cameras often require additional heat dissipation devices.

2.3 Resolution and Pixel Design

When comparing sensors of the same size, CCD typically provides higher resolution. This is because the pixel structure of CCD is relatively simple, almost the entire pixel area can be used for photosensitive, and the proportion of photosensitive area (fill factor) can reach over 95%. And each pixel in CMOS requires the integration of additional transistors and circuit components, which reduce the effective photosensitive area in these 'non photosensitive regions'. For example, for sensors with a 1/1.8 inch specification, CCD can achieve a resolution of 1628 × 1236 (4.40 μ m pixels), while CMOS typically has a resolution of 1280 × 1024 (5.2 μ m pixels).
However, CMOS technology is gradually narrowing this gap through back illuminated (BSI) and stacked designs. Back illuminated CMOS uses a flip chip to direct light onto the photosensitive area from the back, bypassing the circuit layer on the front and significantly improving the fill factor. Stacked CMOS separates and manufactures the photosensitive layer from the processing circuit layer before bonding, further optimizing space utilization. These innovations enable modern high-end CMOS industrial cameras to provide resolutions of over 20 million pixels, meeting the vast majority of industrial inspection needs.

2.4 Sensitivity to Light and Low Light Performance

In terms of sensitivity, CCD sensors maintain their traditional advantages. Due to the larger effective photosensitive area within the pixel, CCD can capture more photons in low light environments, providing better signal-to-noise ratio performance. Test data shows that the human eye can recognize objects under 1Lux illumination (equivalent to a full moon night), and the sensitivity range of CCD is 0.1-3Lux, while traditional CMOS requires 6-15Lux illumination to work effectively - this means that in low light environments below 10Lux, traditional CMOS can hardly capture usable images.
This difference is particularly critical in special applications such as industrial endoscopes, night vision monitoring, and astronomical observations. However, modern CMOS has significantly improved low light performance through large pixel designs (such as pixel sizes above 3 μ m) and advanced micro lens array technology. Some high-end CMOS sensors have even achieved quantum efficiency (QE) beyond CCD through back illuminated technology, achieving photon conversion efficiency of over 95% at specific wavelengths.

2.5 Manufacturing Cost and Economic Considerations

In terms of cost structure, CMOS has an overwhelming advantage. CMOS sensors use the same manufacturing process as standard semiconductor integrated circuits and can be mass-produced in wafer fabs that produce computer chips and storage devices. This process compatibility significantly reduces unit costs. At the same time, the high integration of CMOS enables camera manufacturers to develop "chip level cameras" - integrating sensors, processors, and interface circuits on a single chip, further simplifying the assembly process and peripheral circuit requirements.
In contrast, the manufacturing process of CCD is unique and complex, with only Sony and DALSA, Panasonic and a few other manufacturers have production capacity. Its charge transfer mechanism is extremely sensitive to manufacturing defects: a single pixel failure may result in the inability to transmit the entire row of data, significantly reducing the yield rate. In addition, CCD cameras require additional supporting circuits (including timing controllers, analog-to-digital converters, and signal processors), which collectively drive up the price of the final product, making the cost of CCD industrial cameras typically 1.5 to 3 times that of CMOS cameras of the same specifications.