Camera architecture enables fast capture of high-resolution images
In automated printed-circuit-board (PCB) assembly manufacturing, the pick and place of ICs is performed in conjunction with a machine-vision system. Such machines are often required to handle ICs with edges that measure from 0.5 to 50 mm and pitch distances as low as 0.2 mm. Such components must be placed with an accuracy that does not exceed 25 mm (worst case) at sometimes less that 100 ms.
Because the parts are fed into pick and place machines with a positional accuracy of ۫ mm, part position has to be determined after the pick operation at a higher accuracy than the final accuracy to meet the demands of the placement system. In the case of a placement accuracy of 24 mm, for example, the position of the IC has to be measured with an error of less than 5 mm. Using subpixel computations, a physical image resolution of between 10 to 15 mm is therefore needed at the pixel level.
To meet such stringent demands, current pick-and-place systems take multiple images of different component regions with CCD cameras. This operation results in multiple camera movements and image acquisitions and requires multiple image-processing functions to be performed on the required images. In such applications, the use of high-resolution cameras is impractical because they require greater than 40-ms acquisition times and are relatively expensive. Fortunately, camera architectures are now employing novel electro-optical and electronic techniques to address these demands.
At last year`s meeting of the Optical Engineering Society (Maynooth, Ireland), Claudio Laloni and his colleagues from Siemens (Munich Germany) presented a camera specially designed to tackle such machine-vision applications. Developed in conjunction with IMEC (Leuven, Belgium), the camera uses a combination of two identical 2048 ¥ 684 CMOS-based image sensors that detect the same image by using a beamsplitter.
Placed at 90 to each other, the two sensors` 3:1 aspect ratio is used to reduce the number of pixels that need to be addressed by one-third that of sensors with symmetric aspect rations. In addition, Laloni used a staggered pixel structure in the design of the sensor that effectively doubles the number of effective pixels along the sensor row.
"The dual-sensor design with its asymmetric pixel shape and staggered pixel structure means that the camera sensors only need approximately 2000 ¥ 600 pixels to produce an effective resolution of 4000 ¥ 4000," says Laloni. In pick-and-place applications, the camera can resolve surface-mounted devices with a lead pitch of 0.3 mm in a 50 ¥ 50-mm field of view with a 25-mm/pixel resolution. If two subsequent rows are used during measurement, this resolution increases to 12.5 mm/ pixel.
