With apologies to Shakespeare, Phoenix-based TLC International was recently faced with a question: To see, or not to see? Among other things, the company makes high-accuracy mechanical glass cutters used to carve out windows for cell phones, substrates for microdisplays, mirrors for cars and other flat-glass applications. While designing its new cutting tool, the manufacturer considered whether to add machine vision, pattern recognition and fully automated target acquisition. In the end, said TLC President Ernest K. Linden, the choice was made to stay blind in the machine sense and to count on the eyes of operators.In designing its versatile computer-controlled glass scriber, the Phoenix-600, TLC International decided to forgo the use of machine vision, pattern recognition and fully automated target acquisition.The reasoning behind this decision says as much about the state of machine vision as it does about the intended applications of the Phoenix-600 glass scriber. "When we designed our equipment, I had two goals," Linden explained. "One, I wanted to eliminate as many moving parts as possible, because every moving part contributes to the possibility of something going wrong. And I wanted to keep the cost as low as possible for the customers."Workplace studies indicated that machine vision and automated alignment would not be significantly faster than the speeds that human operators achieve in a multipurpose glass shop in which many shapes are cut over a given day. Machine vision also would be more expensive and, perhaps, require more care and feeding.In addition, when working with sheets of glass, there is the issue of indium tin oxide. The alloy may be sputtered into a transparent conductive film that sits atop a sheet of glass, yielding a conductive glass product that can be used as a substrate for a liquid crystal display or a plasma display panel. The coating is becoming increasingly common, and its reflectivity and the resulting glare make it difficult for machine vision systems to acquire a target. On ITO-coated glass, the postimage processing and abstraction performed during pattern recognition can present a problem. What is displayed to the tool operator may not be what is on the glass, and an operator may have a hard time sorting out what is happening. In such cases, Linden said, capturing the actual image using a CCD camera offers a distinct advantage. "With our camera, I'm seeing real time. So I have a great deal more latitude in what I can look at and see...."The Phoenix-600 is a third-generation design that features a tungsten carbide wheel housed in a proprietary cutting-head assembly. The head incorporates a CCD camera from Genwac Inc. of Orangeburg, N.Y. The sensor is a 1/2-in. interline-transfer CCD, and the camera has 570 lines of resolution.The operator uses the camera to visually verify and adjust the alignment of the head to targets on a sheet of glass. Coarse alignment is performed using mechanical pins and stops, and fine alignment is accomplished with a trackball to move the head. Once finished with one target, the head moves to a spot where a second target should be, and the operator again verifies and adjusts the alignment. When the location of the glass is known, the tool cuts according to a preprogrammed pattern, such as one downloaded from an AutoCAD file.The Phoenix-600's head, not its 24-in.-sq platform, moves in both the X and Y directions. It also can pivot and rotate, enabling the tool to cut curves as well as corners. The accuracy of the cut is 2 mils, or approximately 50 µm, when cutting rectilinear patterns. For more complicated shapes, including those with curves, the accuracy is 3 mils. The speed when cutting rectangularly is up to 20 in. per second. In other situations, the speed depends upon the shape and size of the cut. The maximum thickness of the glass is 1/8 in.The CCD-based setup also enables onstage measurement and inspection so that operators can evaluate scribed glass for quality and tolerance. The use of a single camera mounted on the head, as opposed to other designs that employ two cameras mounted on a gantry, means that the work surface is unobstructed.Although TLC opted not to go with machine vision in this instance, Linden believes that robotic eyes do have their place. In particular, he noted that machine vision makes sense in a dedicated, large-volume setting in which the same cuts are performed over and over.