Teaching the wirebonder pattern recognition.

Wire bonders have pattern recognition capabilities essential to modern electronics manufacturing. Pattern recognition refers to computer identification of substrate or die features important for precise position of bond wires. A PR system will include a camera head, a camera control unit and a bonder controller that includes a microcomputer, a control circuit and a servo motor drive unit. Some wirebonders also come with proprietary gray-scale pattern recognition to improve yield and throughput rates.

To summarize the process, a feature of interest is identified, the area that it resides in specified, and the visual properties of the feature are defined; i.e., the machine is taught to recognize a pattern. Here are some steps that can be followed to ensure proper PR recognition.

First, a reference system referred to as a surface must be defined. Part of this requires simply setting the surface Z-height. This is performed by focusing the system camera on the surface, using a motor drive to move vertically on the Z-axis. The system has a relative focal scale that typically ranges from zero to 200 units. This Z-height is different from the common Z-height programmable home value. This axis is independent of the bond head Z-axis. The bonder also will ask for a range of locations for the bonds. This is defined by three points on the corners of the bonding surface.

Next, assist points are added. Assist points are features, such as global and local fiducials used to reference positions. Pattern recognition is an automatic feature once a program is running; however, the user is required to teach a pattern to the machine. In this example, the pattern to be taught is a “plus sign.” Once a feature of interest is identified, lighting is used to increase contrast between the background and feature. Usually for gold surfaces, oblique lighting levels are increased. Lighting from an oblique angle is useful on the diffuse surfaces such as gold thick film surfaces and other gold-plated surfaces, whereas lighting from a vertical source is useful for polished die surfaces.

Figure 1 shows several different PR modes, including Single Point, Dual Point, Edge, Chain and Line Find. The differences between these modes depend on the type of feature to be recognized. It is advised to try Single Point recognition first, as this setting will work most of the time. An edge finder will examine the contrast due to difference in the finish of the surface.


Image enhancements can aid the PR process (Figure 2). Contrast setting between gray scales level can be set either automatically or manually. Typically, if the image appears good to the human eye, the computer can recognize the feature.


Figure 3 shows the definition of the search area of interest. This region is defined by the user and is indicated by the outer dashed box. It represents the area that will be searched. Next, multiple models can be created (inner box) for the pattern. In this example, two models have been created for the same feature. The number of pixels also can be set. After being defined, the model then can be tested, and the best matches are displayed and ranked with a quality score. The process also determines the match robustness. If the robustness is too low, it can be improved by other techniques such as the Dual Point method.


Because yield and throughput are important criteria, wire bonders have high pattern recognition system search speeds.

PR enables a wire bonder to easily identify fiducial marks and other alignment features. In this way, areas where a wire bond will be placed easily can be referenced and determined. It is even possible to perform a pattern recognition step between each bond wire. This is used when each location is critical, and a dependency on alignment fiducials may not be possible. PR also permits the wirebonder to find the position of each die in MCMs to maximize throughput.

The American Competitiveness Institute (aciusa.org) is a scientific research corporation dedicated to the advancement of electronics manufacturing processes and materials for the Department of Defense and industry. This column appears monthly.

• Prev
• Next