Converting analog to digital with GigE

Many magazine and Web articles have compared the capabilities of digital vision buses such as Gigabit Ethernet (GigE), Camera Link, and FireWire. Yet, rather than learning how digital buses compare to each other, many manufacturers are more interested in learning how these buses compare to their analog counterparts. When manufacturers select a digital bus, they are generally replacing outdated analog equipment, and they want to know what improvements a digital bus can offer.

Compared to analog systems, digital systems can be less expensive and easier to set up. Eric Carey, manager of the smart-products group at Dalsa, recently explained how his company helped a customer adopt a GigE-based system to replace its analog setup.

The customer uses machine vision to inspect assembled PCBs in-line, primarily looking for component presence or absence and alignment. The analog installation included four monochrome cameras—two looking down at the board, and one on each side at an angle—connected to a frame grabber (Figure 1).

Fig. 1  The analog setup included two cameras above the board under test, along with a camera on either side. Strobe lights provided illumination, and a frame grabber digitized the images. Courtesy of Dalsa.

The image data was then sent from the camera over an analog cable to a PC in a control room 20 m away. There, analysis software compared the images with known-good versions, identifying any deviation above a set threshold. The system processed a maximum of 2000 images/min, not counting board-positioning time.

The manufacturer wanted to replace the equipment with more up-to-date equivalents, both to reduce costs and to set the line up for future migration to color imaging. Modernizing the production line also had marketing advantages when the company presented its production capabilities to customers and prospects.

Although acquisition costs for analog and digital cameras were about the same, not needing the frame grabbers for the digital installation cut the cost dramatically. In addition, the line resolution necessary for the application meant that moving to analog color cameras would have required RGB rather than composite color cameras. Each analog RGB camera would have needed its own frame-grabber input for each color (red, green, blue), so an analog color system would need a total of four frame grabbers along with the accompanying cabling and software. Because in the new setup the cameras themselves digitized the acquired images, migrating to RGB color simply meant replacing the cameras and the software.

Digital imaging provided additional benefits. Because there is no cabling (and therefore no signal loss) between the image acquisition and the digitization step, direct digital images tend to have higher resolution and experience less noise. Filtering out the noise in an analog signal reduces resolution at image edges.

For the digital system, Dalsa used GigE Vision as the communication protocol because of its 1-Gbps bandwidth and the fact that it could carry signals all the way to the control room without the need for repeaters. In this configuration, the cameras group image data in packets, each packet generally representing a small number of lines of data from an area-scan or line-scan image.

The system labels the packets as it creates them to ensure an accurate reconstruction of the image at the other end. Packets do not always arrive at the receiver in the correct order, because an interrupted or garbled packet must be re-sent.

Dalsa selected monochrome VGA cameras with a 10-ìm² pixel size to permit resolution comparable to the analog setup. The company replaced the PC interface with an Intel network card.

In higher-bandwidth data situations, the system might have required a network card for each camera, but GigE Vision's bandwidth proved sufficient to allow using an Ethernet switch that included a port for each camera and one port connected to a single network card at the PC. The switch was placed a few feet from the cameras, minimizing cable lengths on that side, and an Ethernet cable connected the switch to the PCs in the control room. Figure 2 diagrams the new setup.

Fig. 2  The digital setup did not require the frame grabber. Each of the four cameras digitized its own images before sending them through an Ethernet switch and Ethernet cables to the PC in the control room. Courtesy of Dalsa.

An Ethernet solution adds a small image-acquisition latency compared to a frame grabber. Because the camera starts to expel image packets as soon as each packet is full, however, this effect does not generally cause any problem. The added latency in this case was less than 1 ms.

Any control that must be performed in real time should be handled directly at the camera. The new system uses the trigger pin of the cameras for that purpose. This approach keeps the delay between the trigger pulse and the start of the exposure below 20 ìs. To prevent any false pulse detection on the optocoupler camera inputs, the integrators set the trigger-pulse debouncing to 2 ìs. That is, the cameras are set to ignore any pulse shorter than 2 ìs.

An optimized GigE Vision packet processing driver reassembles the streaming packets into image files, a task that the frame grabber performed before. In this case, the extra step increased the load on the CPU by less than 6% for simultaneous acquisition by the four cameras. Since the PC in the control room contained a dual-core hyperthreading microprocessor running at more than 3 GHz, the extra load did not unduly burden the image-processing step.

Because Dalsa provided the image-control software for both the analog and digital camera installations, the impact of the change on the image-processing software was minimal, which reduced the system setup time. The engineering effort focused on replacing the hardware components.

The migration to the digital system presented a particular challenge, because when the project began, the GigE Vision specification had not been finalized. Close collaboration between Dalsa application engineers and the team operating the original analog system facilitated the updates and maintained software compatibility.

"This system was an excellent opportunity for GigE Vision, because of the cable length required to reach the control room," said Dalsa's Carey. "Customer cooperation allowed field-testing this new technology before we started system deployment. Field testing required some training to understand networking concepts and their impact on the vision system. The digital changeover proved a huge success."

Egri, John, "Ethernet vs. Camera Link," Test & Measurement World, May 2006. p. 41.

Scheiber, Steve, "Camera Link and GigE improve image speeds," Machine-Vision & Inspection Test Report, Test & Measurement World, February 2006. p. 52.

Titus, Jon, "Tech Trends: A need for speed," Test & Measurement World, February 2006. p. 16.

Titus, Jon, "Tech Trends: Vision meets Ethernet," Test & Measurement World, April 2006. p. 20.