How do you think the new GigE standards will influence the machine vision industry?
Respond or ask your question now!
Many of these same technology differences between CMOS and CCD sensors we discussed last month apply when using line-scan sensors, although a few of them become less important. The sensor's speed, for example, is less of an issue because there are fewer pixels involved. Similarly, the differences in fill area disappear even when electronic shuttering is required. The additional transistors that reduce fill factor for a CMOS area sensor can be placed alongside the pixel array, outside of the active area, in a line sensor.
There are application considerations that rise in importance with line sensors, however. One is color operation. For a sensor to produce a color image it needs three arrays, one each for red, green and blue components of the image. Because a CCD sensor needs transfer and readout circuits adjacent to its pixels, the three color arrays are spaced apart. A CMOS line sensor, on the other hand, can position its circuitry at a distance from the pixel arrays, allowing the pixels to be positioned more closely together. This closer spacing improves image resolution and helps minimize motion-generated image artifacts.
Still, unless color operation is required, simple line scan sensors do not demonstrate as significant a technology difference as is seen with area sensors. The most distinguishing feature of the technologies is their relative maturity. Because CCD sensors have a longer history, the challenges and solutions for their use in machine vision applications are better understood.
For TDI line scan sensors, however, technology differences become more significant. These sensors target a specific type of application where the image is formed through the steady, linear movement of an object across the sensor's field of view (see Figure 2). The output signal of a TDI sensor is the result of binning together as many as 100+ linear pixel arrays, so the binning considerations of sensor technology become paramount. Because CCD sensors add pixel data together without added noise, they achieve an N-fold increase in S/N ratio with N lines, while CMOS sensors add pixel data with noise, yielding a square root of N increase in S/N ratio. With a 100-line TDI design, CCD sensors thus have 10x the S/N ratio of an equivalent CMOS sensor, giving the CCD a much greater dynamic range.
Begin with the Application
It is clear, then, that making a machine vision sensor selection should begin not with the fabrication technology, but with the application. Developers should first determine whether an area sensor, a line sensor, or a TDI sensor is the best fit for their imaging needs. Once the sensor type is selected, developers should then determine their requirements for frame rate or imaging speed, the light sensitivity and dynamic range they will need, and whether or not to use color imaging or non-visible light for illumination. By prioritizing these requirements, a comparison of technologies will then show the relative strengths of CCD and CMOS technology in the given application.