Lighting and software improve AOI results
By Steve Scheiber, Contributing Technical Editor -- Test & Measurement World, 10/1/2007
The ability of automated optical inspection (AOI) techniques to find board faults depends as much on issues such as lighting and contrast as it does on the capabilities of the inspection systems themselves. In addition, using the same architecture for both benchtop development systems and for systems in production simplifies the implementation of an inspection strategy. Josh Petras, product manager at YESTech, has devoted a lot of his energy to these issues, incorporating the principles into the company’s latest line of AOI systems.
“Lighting has to be flexible,” said Petras. “A flexible lighting system enables better defect detection and faster cycle times. Techniques such as 'fusion lighting,’ which detects specific features by incorporating multiple colored lights from different angles and color filters, are invaluable when dealing with ever-changing PCB complexities.”
![]() |
|
Fig. 1 Incident lighting of several colors at different angles produces an aggregate image. Courtesy of YESTech. |
Petras compared this type of analysis to the information on a topographic map. “A topographic map allows you to visually extract 3-D information from a 2-D image. The fusion lighting concept works the same way. Since structured lights are projected from different angles onto various surfaces of the board, the reflected light is also colored. Low-angle surfaces reflect mostly red light, while medium angles and steep angles reflect primarily green and blue, respectively. The software color filters allow the inspection algorithms to isolate specific color spectra that relate directly to their slope and orientation in three dimensions.”
Figure 1 illustrates a fusion lighting configuration that includes both incident white light and red, green, and blue lights that come in from different angles. The colored lights produce the aggregate image in Figure 2a. Figure 2b shows the result of applying a red filter, rendered as a gray-scale image. Figure 2c shows the same board area through a green filter, and Figure 2d through a blue filter. Examining the image as though illuminated only by red light reveals joints with insufficient solder. The leads themselves show up better in blue light. The white light version of the image allows the identification of part markings and the examination of part orientation.
![]() |
|
Fig. 2 Filtering the aggregate image in (a) with a red filter produces (b). A green filter reveals the details in (c), while a blue filter shows image (d). Courtesy of YESTech. |
“We are not necessarily looking for a high frame rate with this technique,” remarked Petras. “We would rather combine the information into fewer images taken at higher-resolution and then perform the analysis in software.
“We are not the first company to employ lighting at different angles to enhance image contrast, but we believe our unique software approach enables better results while simplifying the overall user experience. Other companies rely on multicolored light structure for all of their inspection points. We include white-light sources independent of the RGB composite, which is more effective for the white-light part of the analysis. We wanted to take advantage of the power of today’s off-the-shelf PCs by putting more of the throughput burden on the software.”
To verify that the new configuration improves results, YESTech compared inspections performed with the multicolor lighting scheme to similar inspections from their previous AOI systems that relied primarily on white and red light. Although those systems could use color filters, they didn’t take advantage of the feature very often. “The new arrangement produces images with more information for our inspection algorithms, and ultimately a more consistent and robust test,” said Petras. “Based on internal benchmarks and customer feedback, detection has gone up and false calls have gone down substantially, especially on very dense boards.”
Petras also advocates using benchtop and production systems that share an architecture. “The standardization allows an engineer to capture an image on the in-line system, then return to a desktop PC to create the inspection program. The benchtop version of the system can also work well in a repair station or quality cell. The overriding consideration is yield management.
“A common architecture makes managing the operation, as well as collecting, archiving, and analyzing the resulting data much easier. You can build a database, create historical reports, or implement Web-based statistical control. Critical information can be extracted at various points in the process.”
He added, “The common architecture helps ensure data consistency and improves the quality of the subsequent analysis.”






















