Worldwide energy demand is rising, and so is the price of that energy as fossil fuel resources are consumed. These trends have led to growing worldwide interest in alternative energy sources such as solar photovoltaic (PV) cells. The key to success in the solar cell market, however, is lowering their cost to manufacture by increasing production volume – and machine vision systems are key to increased production. A combination of technology advancements and governmental stimulus has done just that, creating a tremendous opportunity for solar cell manufacturers. The advancements have indeed helped lower the cost of solar PV systems while government-funded initiatives have provided the capital for increased research and production investment. The result is a rapidly growing market for solar cells and increased competition to serve that market. The appeal of solar PV systems is the abundant availability of solar energy. Extrapolated to the whole of the Earth, this energy is equivalent to more than 10,000 times the global annual demand for primary energy. The challenge, of course, is to make the best possible use of this potential. The efficiency – the amount of solar energy converted to electric power – must be as high as possible. However, the theoretical maximum of 30 percent is far from being reached, even under laboratory conditions, forcing the use of large solar cell arrays to gather sufficient amounts of energy. For these large arrays to be cost-effective energy sources, therefore, the solar cells must be as inexpensive as possible. Solar cells usually are made of polycrystalline silicon and require many of the same manufacturing processes as integrated circuits. There are significant differences in their production, however. One is size: Solar cells are typically 150 x 150 mm or larger, which only high-end processor integrated chips (ICs) approach in scale. To get acceptable fabrication yields with such large cells, the raw silicon wafers must meet exacting demands on purity and freedom from defects. Early detection essential Solar cell wafers also are subject to different handling stresses than IC wafers. The front metallization layer that forms the conductor grid on a solar cell is applied using a silk-screen process that creates pressure stresses on the 0.2-mm-thick wafer. This can create defects in the finished cell that can adversely affect power production, including chipping, cracks and broken edges. Because solar panels require many cells to work together efficiently, defects in individual cells must be reliably identified before assembly into the panel, where they would compromise overall panel performance. Machine vision using high-resolution cameras is proving to be the only viable way to ensure early detection of these defects (Figure 1). Manual quality assurance is next to impossible because the fragile wafers are difficult to handle. High resolution is needed because cracks as small as 120 µm long are enough to compromise performance. The challenge is to provide such resolution at the high throughput rate required to achieve the production efficiencies needed to keep cell costs down. Figure 1. Machine vision systems are becoming essential tools for lowering the manufacturing cost of solar photovoltaic power panels. The Image Processing and Intralogistics department at Eckelmann AG of Wiesbaden, Germany, is currently developing such an inspection station – the E.SEE-Waferinspect – for a customer who wants to equip several wafer production lines with the system. The system not only looks for flaws such as chips and edge defects, but also must examine the wafer surface for contaminants, measure the wafer dimensions to an accuracy of 50 µm and measure the angle of chamfered corners. The production lines have a throughput of 3600 wafers per hour, leaving the machine vision system with just under a second to complete each inspection. To achieve the required 50-µm measurement accuracy on a 150 x 150-mm wafer, the camera needed a resolution of ~5 million pixels. Handling that much data at the production speed required a high-speed output port on the camera that allowed it to be positioned a large distance from the host image-processing computer. Based on previous good experiences with camera integration, product quality, service and support from IDS, Eckelmann decided on the UI-5480-C from IDS’ uEye series. One camera, dual imaging The UI-5480-C has a high-speed CMOS color sensor with 2560 x 1920 pixels and can capture and deliver full-resolution images at a rate of 15 fps. It also offers an area-of-interest function for limiting the field of view the camera sends to the image processor, enabling the image processor to work with a square partial image that matches the wafer shape and eliminates distracting background. The camera provides a Gigabit Ethernet port, supporting cable lengths up to 100 m to allow flexibility in camera positioning. Figure 2. Dual lighting systems allow this solar cell wafer inspection system to work with images optimized for different inspection tasks. To detect defects and make the needed measurements in the time available, the inspection station takes a divide-and-conquer approach by breaking the task into two parts. The camera is installed in a metal housing that is open at the bottom and equipped with built-in lighting. During production, a rotary table positions the sawn silicon wafers below the camera for in-line inspection (Figure 2). The camera acquires two successive images, each with different lighting. While the camera is taking the second image, the image processor analyzes the first one. The second image’s processing takes place while the station sets up for and captures the first image of the next wafer to be inspected. Because the image capture and analysis are performed in parallel, this solution saves valuable inspection time. The two types of lighting allow the processing software to work with images optimized for different types of inspection. The first image is acquired using diffuse red LED backlighting, which makes through-cracks readily visible. The colored light also helps simplify the image processing used to detect and measure the cracks. Red hides the grain boundaries on the polycrystalline silicon to prevent the vision system from confusing them with defects. Despite the color, however, image processing uses only gray-scale operations for speedier computation (Figure 3). Figure 3. The red backlighting helps through-cracks to appear in high contrast to the wafer surface, simplifying their detection using only monochrome processing. The second image is taken using diffuse incident white LED lighting, which provides good contrast between the normal wafer surface and any impurities or defects. Cracks that do not pass completely through show up clearly under the diffuse lighting. Besides detecting surface defects and contamination, the system uses the white-light image for measuring wafer size and the chamfers at the corners. Design for flexibility Eckelmann’s corporate philosophy is to flexibly customize standard components to meet the individual customer’s needs. As a result, the E.SEE-Waferinspect system is designed to be modular so that solar cell manufacturers can readily add other measurement functions such as grain size and wafer thickness. An additional software module working with the white-light image measures grain size. The addition of another camera expands on the divide-and-conquer approach to support measurement of wafer thickness. When the inspection station is completed, the company will have put more than two man-years of development into camera control and image processing. Thus, a key element supporting Eckelmann’s development of the inspection system was the software support that came with the IDS uEye camera. The camera’s software development kit (SDK) provided a ready-to-go interface for the image-processing software library used in the wafer inspection system, greatly facilitating the integration of the camera with the analysis program. Besides providing interfaces for ActiveX, DirectShow and various libraries, the uEye SDK also features a direct programming interface (API) for accessing drivers in C++, C# and VB. Eckelmann’s E.SEE-Waferinspect system demonstrates only one role for machine vision in solar cell manufacture: finished wafer inspection. Many other opportunities exist. Machine vision can speed incoming wafer inspection, surface treatment during processing, application of metallization and final test of assembled panels. By speeding production and catching defects early to avoid waste and rework, machine vision is helping solar PV systems fulfill their promise of providing a cost-effective, renewable energy resource. Meet the author Daniel Seiler is a member of the marketing department at IDS Imaging in Obersulm, Germany; e-mail: info@ids-imaging.com.