Throughput needs drive vision-system camera choices

Mark Butler Product Manager Dalsa Digital Imaging Courtesy of Dalsa Digital Imaging.

Q: What camera features do electronics manufacturers look for today?

A: A lot depends on the application—pick-and-place has different needs from AOI, x-ray, or layer-upon-layer bare-board inspection. Some companies choose analog cameras for pick-and-place, for example, often using as many as 20 to ensure proper component position. Cameras mounted inside production equipment can provide a troubleshooting tool for the equipment itself.

Q: What is the most significant factor driving manufacturers’ choice of camera?

A: Everyone wants high-quality images and wants cameras to be small and easy to use. But our customers’ number one priority is throughput—how fast they can inspect their boards.

A few years ago, we introduced a 2-Mpixel camera that took 30 fps and a 4-Mpixel camera at 15 fps. With the larger number of pixels, the camera can see more of the board in each image and can cover the board with fewer images. As parts and other board features continue to shrink, however, you need higher resolution to maintain the same field of view. That’s why we introduced the 4M60 camera that produces 4-Mpixel images at a full 60 fps.

Q: Why has higher resolution become such an imperative?

A: No matter how small features become, inspecting them requires the same number of pixels. If you need 5 pixels to detect a bad lead, as the leads get smaller, you still need 5 pixels, but they cover a smaller area. Either you have to increase the camera resolution or cope with a larger number of images to cover the entire board surface.

Q: What are the consequences of needing higher resolution?

A: Most high-speed camera imaging chips use “rolling shutter” CMOS technology, which acquires the image in a series of rows—up to the imager’s maximum. The technique exhibits time-domain effects. If either the object under inspection or the camera is moving, each row is acquired at a different moment in time and therefore a different point in space. Straight lines can look slanted. The image is less precise than it could be.

We designed a special CMOS chip with a global shutter that captures the entire frame at once to avoid that effect. It can also integrate one image while reading out another, increasing the camera’s effective frame rate.

Q: Do you lose throughput while repositioning the camera?

A: While the camera is in one place, the system takes a complete set of images, changing color, lighting, incidence angle, and so on. You perform image analysis while the camera moves to the next location, in say 200 ms. That way you can maximize your efficiency and throughput. In other cases, the board moves rather than the camera itself.