Vision system targets wafer inspection
Andrew Wilson, Editor, [email protected]
In semiconductor-manufacturing processes, robotic-based systems perform a number of operations, such as transferring wafers between cassettes, carriers, and magazines. “At each stage in the production process,” says Stanley Hack, president of ConsulTech Engineering (Manlius, NY, USA; www.consultechusa.com), “wafers must be checked for damage and for proper positioning in their carriers. The most efficient, economical, and accurate way to perform these checks is with machine vision. Unfortunately, the cleanroom manufacturing environment is not suitable for machine-vision system development.”
Hack and his colleagues have designed and developed an automated wafer-inspection testbed using off-the-shelf lighting, frame grabbers, display controllers, andx-y positioning systems. Hack explains, “The wafer-inspection testbed simulates the positioning systems and the vision component geometry used by machine-vision systems for inspecting wafers within the cleanroom environment.” Wafer cassettes, carriers, or magazines are positioned in the testbed and then moved vertically using a z-axis stage from Velmex (Bloomfield, NY, USA; www.velmex.com) under control of a PC (see figure). To properly illuminate each of the wafers as they pass through the camera’s field of view, the wafer rack is backlit with a 16-in. red LED line light from Volpi Manufacturing (Auburn, NY, USA; www.volpiusa.com). “As the wafers are moved in the z-direction,” says Hack, “they are imaged using a 2k × 1 CCD Piranha Camera Link-based linescan camera from DALSA” (Waterloo, ON, Canada; www.dalsa.com).
“Using a linescan camera,” says Hack, “only the edges of the wafers are in the optical field, simplifying the analysis by eliminating top and bottom wafer-surface image artifacts. If an area-array camera had been used, the top and bottom surfaces of the wafers would be partially imaged.” ConsulTech Engineering uses a Meteor CL frame grabber card from Matrox Imaging (Dorval, QC, Canada; www.matrox.com/imaging) in its testbed to capture the linescan imagery and store it in the PC memory The PC also controls thez-positioning stage that holds the wafer cassette and to accept signals from the stage’s shaft encoder.
“To visualize the images as they are captured,” says Hack, “the system incorporates a Millennium graphics card, also from Matrox, which is interfaced to a Comark (Medfield, MA, USA; www.comarkcorp.com) flat-panel touch-screen display.” The display shows the wafers in the cassette as they are being captured by the camera. The display also shows annotations for each wafer, describing its condition and position within the cassette. “The image-analysis software is written in C++ using ConsulTech Engineering ImageTech software library,” says Hack. “Matrox MIL Lite library controls the operation of the Meteor CL frame grabber.”
As a scan is being performed, images of the edges of the wafers appear dark on a light background. After performing a histogram of the light/dark/light regions surrounding each wafer edge in the digitized image, each region is adaptively thresholded and a straight line is fitted to the center of the wafer edge in the region. “By examining the thickness deviation of each edge from the mean and checking whether the center line is on- or off-axis, the system accurately determines whether any of the wafers are out of alignment or damaged,” Hack says. Since images are captured continuously, each light/dark/light region surrounding a wafer edge is processed as the carriage moves the wafers across the field of view of the camera. During image processing, the system creates a map of any misplaced or defective wafers in the cassette.