Machine vision inspects PCB coatings
FireWire cameras team with off-the-shelf software to check printed-circuit boards.
By Earl Yardley
Conformal coatings are protective, dielectric materials designed to conform to the surface of an assembled printed-circuit board (PCB). Commonly used conformal coatings include silicone, acrylic, urethane, and epoxy. While these coatings protect circuitry from moisture, fungus, dust, and corrosion, they also prevent damage from board handling during construction, installation, and use, reduce mechanical stress on components, and protect them from thermal shock. Designed to resist abrasion, these coatings also minimize the electromigration of metal between conductors.
In the fabrication of PCBs, electronics manufacturers apply these coatings using automated systems that consist of an x-y-z platform that moves a dispensing head to key points on the surface, in a process known as selective conformal coating. To allow operators to manually inspect PCBs, many of the standard conformal coating materials are available with a 365-nm-wavelength UV trace. Using standard UV lamps, inspectors can manually check individual boards to see if the coating is placed correctly around the component, whether it has migrated into incorrect areas of the board, and whether the correct amount of coating has been applied.
Due to the different levels of trace versus raw material within the compound, the fluorescent nature of such coatings can vary widely. Because of this, inspectors must be properly trained to understand the nature of both the compound used and its fluorescent properties to properly inspect each board. As with all manual inspections, such checks are subject to the individual accuracy and reliability of each operator.
One UK-based OEM manufacturer of PCBs approached Industrial Vision Systems (IVS) to automate the process of inspecting these coatings. IVS, the sister company to NeuroCheck, designs and manufactures machine-vision systems for the automotive, pharmaceutical, printing, and electronics industries.
In the design of its conformal-coating checking system, IVS chose a standard 1000-mm PCB conveyor system and canopy. In this way, the system can be adapted to any size board and multiple PCB types. Under control of a Micrologix PLC from Allen-Bradley, PCBs are individually stepped along the systems conveyor and into the inspection station using the Surface Mount Equipment Manufacturers Association machine interface standard interface. This equipment-interface specification standardizes the communication mechanism for single transfer manufacturing systems of surface-mounted PCBs and allows equipment from a number of manufacturers to be integrated more easily.
FIGURE 1. In the design of its conformal coating system, IVS uses two FireWire cameras and a PC interfaced to a PLC to control the flow of boards under the inspection station.
Because of the nature of the UV traces added to the conformal coating, no off-the-shelf UV lighting could be used to properly illuminate the board. As a result, IVS built two custom industrial lighting rigs to illuminate both sides of the PCB. To image each board, two 1392 × 1040 FWX 14c FireWire cameras from NeuroCheck were mounted above and below the conveyor belt. Digitized images were then captured using an industrial PC interfaced to the system’s PLC (see Fig. 1).
Three separate checks must be made on each PCB. The first, UV Trace Runout, ensures that that no conformal material has deviated into any other no-go area of the board. The second, UV Trace on Component, determines if the coating is in the correct position around the component and that the correct amount of material has been used. Finally, a RGB Color Match checks whether the color level of the coating is within specification and that the material has been mixed correctly. To perform these inspections, the host PC runs the NeuroCheck machine-vision software.
FIGURE 2. Using NeuroCheck machine-vision software, the system can perform three critical inspections that determine the proper position, amount, and color of the conformal coating.
To check whether the conformal material is in the correct position on the board, captured images are contrast-enhanced using a software-based look-up table. Defined regions of interest (ROIs) are then determined from a known good template of the board. Because the amount of coating that can be tolerated in any specific region of the board is known, the regions of interest can then be highlighted by thresholding the image and counting the number of pixels within each (see Fig. 2). In a similar manner, the software can be programmed to determine the correct position and amount of coating.
Color levels
To compute the color level of each coating, the UV trace within each ROI is color-matched using NeuroCheck’s Color Matching Wizard. Each color value is represented by three values for the three basic colors: red, green, and blue. These values fall into the standard gray-level range from 0 to 255. The similarity between two colors is expressed as
where c1, c2 represent the two colors and rN, gN, bN their respective red, green and blue values and
is the normalized Euclidean distance between the colors.
Taking an RGB color image as input, the software creates a gray-level image of the same size as its input image. The gray level of every pixel in this image indicates the similarity of the corresponding pixel in the original image to one or more known reference colors.
In the design of the IVS system, the Distance Mode function of NeuroCheck was used. In this mode, the brightness of the resulting gray-level pixel indicates its similarity to a single reference color. For each pixel in the source image, the similarity of its color to the reference color is computed. The gray level of the corresponding pixel in the result image indicates the degree of similarity. A similarity below the required selectivity value is coded as black. A similarity of 1.0, which means that the pixel has exactly the reference color, is coded as white.
FIGURE 3. Taking an RGB color image as input, the software creates a gray-level image of the same size as its input image. The gray level of every pixel in this image indicates the similarity of the corresponding pixel in the original image to one or more known reference colors. Areas of pure color with full brightness are shown on the left and three areas of pure color with medium brightness on the right (left). Since all the colors are pure colors, except for the background, the similarity scores are identical for the blue and green areas when a red reference is used (middle). In Hue-only processing mode, the medium red is treated identically to the bright red, whereas the background becomes much darker (right).
Figure 3 shows areas of pure color with full brightness (on the left) and three areas of pure color with medium brightness (on the right). Since all the colors are pure, except for the background, the similarity scores are identical for the blue and green areas when a red reference is used. In “Hue only” processing mode-which is not incorporated into the IVS system-the medium red is treated identically to the bright red, whereas the background becomes much darker, because it is treated as white and so loses the similarity to the pure color in overall brightness. By using color FireWire cameras with this software, the RGB settings of the camera can be dynamically changed during the inspection process, allowing the system to compensate for the varying levels in UV trace within the material.
After each board is inspected, the host PC uses the pass/fail data to trigger the Micrologix PLC using a 32-bit digital I/O card from Meilhaus. Should a board fail inspection, a reject mechanism can be triggered to eject the board from the conveyor system so that it can be returned for rework.
Human interface
IVS used NeuroCheck to develop a graphical user interface to display images and the results of each inspection (see Fig. 4). As each board is inspected, the operator is presented with an image of the board and the highlighted ROIs. The results of each pass or fail on each board is displayed, as well as a running total of the number of boards inspected and passed by the software.
Because the information is stored in a standard database format, it can be networked to existing database systems to provide information about the integrity of each system in the PCB production process. This information can be transferred from the system PC using a standard Ethernet interface.
While color algorithms such as those developed by NeuroCheck can check for specific placement, position, and conformal coating integrity, they also can be used in other applications to locate fiducial marks in parts-inspection applications. Rather than using template-matching algorithms, a color-based algorithm could locate and then reposition products for further inspection. Alternatively, they could also perform a pure color-matching function in print-inspection applications where a color hue or intensity must be correctly maintained.
Earl Yardley is director, Industrial Vision Systems, Kingston Bagpuize, UK; www.industrialvision.co.uk.
Company Info
Allen-Bradley
Milwaukee, WI, USA
www.ab.com
Industrial Vision Systems
Kingston Bagpuize, UK
www.industrialvision.co.uk
Meilhaus
Puchheim, Germany
www.meilhaus.com
NeuroCheck
Remseck, Germany
www.neurocheck.com
Surface Mount Equipment Manufacturers Association
www.smema.org