AOI: Do You Need Color? -- Black-and-white AOI systems may offer distinct advantages in viewing components and boards.

When optically inspecting components on substrates, color is better than black-and-white, right? On the surface, that premise seems to make sense. For example, a brown capacitor on a green board is likely to show up as gray-on-gray when using a black-and-white camera. However, with a color camera, an operator should be able to see a distinction between the brown capacitor and the green board. The assumption is that automated optical inspection (AOI) systems with black-and-white cameras are at a disadvantage. However, the answer is not quite that simple. In fact, just the opposite is more apt to be true.

CCD Color Camera Technology

Today's AOI systems use charge-coupled device (CCD) cameras, in which the incoming light is received by a piece of silicon with light-sensitive cells-the CCD-instead of film. Each individual cell, or photodiode, is an element of the CCD array and is generally known as a pixel.

The camera in an AOI station contains literally tens of thousands of pixels. The actual number depends on the size of the array; 640 pixels high by 480 pixels wide is one example. As a photon of light hits the pixel, an electron is released. As long as this impingement continues, electrons accumulate in the pixel and are ultimately counted, processed and converted into an image on a video monitor.

The reality of AOI technology is that CCD chips are inherently black-and-white devices because a CCD cannot detect color, only the brightness of the light striking it. Thus, the pixel array is essentially a two-dimensional plot of light intensity versus position. In dealing with this fact, two types of color cameras have been developed: the one-chip camera and the three-chip camera.

A third type of camera is also touted as being a color camera. However, instead of the camera shooting in color, the images themselves are colorized. The process is much like the adding of color to old black-and-white movies, and the results are just as unsatisfactory. The images are unnatural looking, and the colors, in effect, degrade the ability to see detail.

One-chip camera

The most common one-chip type is the color mosaic camera, which incorporates a single CCD chip, much like a black-and-white camera. However, color imaging requires three color maps; one each for the colors of red, green and blue. Accordingly, the surface layer of the CCD is coated with a dye that acts as a filter to separate the pixels into the three colors.

The filter material is applied such that each individual pixel is covered by a separate color filter of either red, green or blue in a prescribed pattern. The pattern employed most often is the G-R-G-B Bayer pattern. It is biased toward green colors because the human eye is more attuned to green than to either red or blue.

As shown in Figure 1, with the Bayer pattern, one-half of the pixels measure green, while one-quarter of the pixels measure red and the remaining pixels measure blue. Other patterns can be used, and the actual percentage of each can vary with the design of the camera manufacturer. For example, some manufacturers provide one-third of pixels with red filters, one-third with green filters and one-third with blue filters.

The problem with color mosaic cameras is that, instead of each pixel having measured values for all three filters, only one color is measured per pixel, with no data being provided for the other two colors. Where the data are unavailable, algorithmic interpolation is required, which simply means the missing values for each pixel are estimated based on the surrounding pixels instead of being measured (Figure 2). Thus, in a typical filter pattern, many more pixels are interpolated instead of being measured. The net effect is to produce an image with less image sharpness (resolution) than would be the case with all measured values.

Three-chip camera

This type of camera represents a step up in quality and cost, but it still has inherent problems. As the name suggests, three-chip cameras incorporate a CCD detector for each of the three colors. The incoming light is separated by a prism, with the red, green and blue beams being directed toward the appropriate chip. With this type of color camera, the resolution is the same as that of a black-and-white camera.

However, the accuracy of the image depends on the mechanical alignment of the three CCD detectors. For today's machine vision inspection systems, CCDs must be aligned within 0.3 microns. Such alignment must be maintained under varying temperature conditions and, often, under incidences of high acceleration. High acceleration and deceleration of the camera occur as the platform moves rapidly across the PCB in imaging components. Such speed is necessary for maintaining throughput, particularly for in-line inspection stations.

Also, with three-chip CCD cameras, light sensitivity is somewhat reduced, which can be a concern, depending on the component being inspected and its location on the board.

With either the color mosaic or three-chip color camera, optic quality is critical. For each lens system, different wavelengths of light will come into focus in different planes because of the lens system's design. When designing a lens system, the optical designer must accommodate the various aberrations of the lens system to maximize the resolution and contrast of the image. Such aberrations include curvature of field, spherical aberration, lateral chromatic aberration, distortion and coma. Due to these complications and other effects, lens systems are rarely able to focus the red, green and blue light beams at the same time.

However, black-and-white cameras are unaffected by any of the above limitations, because no need exists to individually filter pixels or separate light with a prism.

Finally, with color cameras, image analysis may often only be applied in one of the three elementary color planes: red, green or blue. This process is commonly called color pass through because the image is passed through to the operator via a color video monitor, but it is only analyzed by the system with customary gray scale correlation techniques on only one of the three color channels. While the picture on the screen might look nice, the accuracy of component inspection is not improved.

Seeing the Components

When inspecting surface-mount components, the most important consideration is the ability to obtain adequate contrast between the component and the substrate. With black-and-white cameras, the process comes down to selecting a source of light that provides the desired result.

With color cameras, regardless of the type, the effort is a little more challenging. Adding to the difficulty is the fact that colors vary, depending on the nature of the light and its stability. Various sources of light produce very different effects, which can impact component visibility and measurement. For example, daylight and tungsten light have continuous spectra; the light is a blend of all colors with no discernable gaps between colors.

However, despite this basic similarity, the light sources differ in color temperature, as measured in degrees Kelvin. Tungsten light has a Kelvin temperature of about 3,200 K, while daylight has a Kelvin temperature of approximately 5,600 K. Objects derive their color from the wavelengths they reflect, so illuminating components with different color temperatures automatically results in a different appearance when viewed by the camera.

Accordingly, the camera operator must be adept at recognizing component shapes using the light source provided with the AOI system. Also, the operator must understand that the stability of the light source is crucial to identification and measurement; any variation in color temperature can impact the viewed image.

Black-and-white cameras are not as sensitive to varying color temperatures. The operator simply selects on the machine an LED color that bathes the viewing area and achieves optimum contrast between the component being inspected and the board.

What about the advantage of seeing the component in one color and the board in another color? With color cameras, thousands of colors and shades may fade into each other. The color camera captures more than 16 million possible colors, with a color image providing over 65,000 times more information than a black-and-white image. As a result, the image processing tools require far more time to detect simple features in a color image than in a black-and-white image.

Finally, the white balance must be considered. White balance is the adjustment of a camera so that it shows white parts of the image with no color tint. For example, a person directly viewing an object will see it as being white if it is white, regardless of the lighting conditions. However, for color cameras, white light occurs only with an equal combination of the three color channels, red, green and blue. Thus, fine-tuning the white balance on a color camera involves precise adjustment of the three CCD outputs, which can be critical to accurately depicting the true colors of the image. If the white color, which is the reference color, is off, then the other colors are off as well. Accordingly, differences in color temperature and even the ambient lighting can adversely affect the balance. However, with black-and-white cameras, white is white.

Conclusion

Unless an AOI system is required specifically to inspect a color feature such as color bands on a through-hole resistor, the use of color is not likely to improve the accuracy or speed of the effort. Moreover, color systems require considerably more processing power and complex algorithms to verify such basic AOI tasks as component presence, position and polarity. For many machine vision applications, black-and-white cameras offer distinct advantages in terms of contrast, balance and resolution in viewing components and boards.

Marc Stabile is senior product engineer and Glenn Wyllie is technical sales engineer, ViTechnology, LLC, Haverhill, MA; e-mails: [e-mail]; [e-mail].

http://www.circuitsassembly.com/

Copyright [copyright] 2001 CMP Media LLC