Machine Vision Sensors: CCD vs CMOS Explained

Charged-coupled devices (CCDs) and CMOS active pixel sensors (CMOS sensors) are the primary image sensor technologies. They rely on lenses and other optical components to deliver the image to be captured.

This FAQ reviews the differences between CCDs and CMOS sensors, some key specifications of image sensors, the concept of quantum efficiency, and some related optical elements.

CCD sensors are being replaced in many applications by CMOS sensors. Cost and power consumption are primary driving factors; CMOS sensors are much less expensive to produce and consume as little as 5% of the energy required by a CCD. But CMOS sensors also have lower performance. They are more susceptible to noise and can have lower light sensitivity. The bottom line is that CMOS sensors are improving and suitable for a growing range of machine vision applications, but CCDs are still preferred in most high-performance vision applications.

Resolution measures the number of pixels and is a baseline specification for image sensors. It’s generally stated as the number of X- and Y-axis pixels. While it may seem that ‘more is better’ in terms of resolution, it’s usually advisable to determine the resolution needed to ensure the required levels of accuracy and repeatability. Too many pixels can place unnecessary demands on image processing electronics.

Quantum efficiency

Quantum efficiency (QE) is an important metric for measuring the sensitivity of both CMOS and CCD sensors. It’s the ratio of the charge created (number of electrons) for a given number of photons hitting the sensor. QE changes with wavelength, and the QE specification is the efficiency at the optimal wavelength, often in the range of 500 to 600 nm (green/yellow). For example, a sensor with a QE of 90% at 550 nm may have a QE of only 25% at 200 nm. And QE curves are not necessarily smooth; they can have multiple peaks.

Sensor size and pixel size

Sensor size is a secondary consideration. It’s related to resolution, but the relationship is evolving as the pixels become smaller and more densely packed. There’s a wide range of image sensor sizes used for different applications (Figure 1). Industrial machine vision systems use smaller sensors while television and movie production video cameras use larger sensors. For a typical ½” sensor in an industrial imaging system, here are some examples of the relationship between pixel size and resolution:

Using relatively large 9.9 µm pixels can deliver a resolution of 640 x 480 pixels (0.3 megapixels, MP).
Smaller 5.5 µm pixels can boost the resolution to 1280 x 1024 pixels (1.3 MP).
Shrinking the pixels down to 3.6 µm can deliver a resolution of 1600 x 1200 pixels (2 MP).

Figure 1. Common optical sensor sizes (Image: Edmund Optics).

Optical considerations

Lenses are needed to focus the image on the sensor. A few key specifications for lenses in machine vision applications are resolution, field of view, depth of field, and working distance (Figure 2).

The resolving power of the lens and the resolution of the vision sensor need to be matched. Using a lens with too high or too low a resolving power can unnecessarily increase system cost or not use the full capabilities of the sensor, respectively.

The field of view (FoV) is the number of scenes the camera captures. For example, a wide-angle lens captures a larger area with a larger FoV. Sensor size is also related to a vision system’s field of view (FoV). For a given primary magnification, determined by the lens, larger sensors support larger FoVs.

Figure 2. Relationships between key lens performance specifications (Image: Cognex).

Depth of field (DoF) is the range of distances in which the image remains in focus. A shallow DoF can be useful for inspection systems tightly focused on a specific plane where the object will appear. A wider DoF is needed in applications like an autonomous mobile robot. Barcode reading is an application that can benefit from a moderately wide DoF.

The working distance is the distance between the object and the front of the lens. It’s related to FoV and DoF. A longer working distance generally results in a wider FoV, but a narrower DoF, and vice versa. Long working distances are used in applications like high-temperature thermal imaging, where the camera must be a safe distance from the measured object. In contrast, a short working distance can be used for close-up inspections of miniaturized devices.

Summary

CMOS image sensors are used in many industrial machine vision systems, while CCD sensors are still used in the most demanding applications. Regardless of the imaging technology, the performance of the lens system must be matched to the sensor capabilities and needs of the application.