Single Point Machine Vision

Multi-camera vision appliances are embedded solutions designed primarily to satisfy inspection needs, and in today’s market they have become rather sophisticated. They generally provide ease-of-use, performance, and flexibility required to meet the diverse requirements of industrial applications, while accommodating the needs and limited experience of end users.

Dalsa IPD (the IPD stands for Industrial Products of Dalsa) has been supplying machine vision for 25 years, and claims to be the only company to own all of the technology ingredients—including cameras, frame grabbers, software, vision appliances, and even the semiconductor fabrication of the CCD vision sensor chips. While the company has built its reputation on multi-camera solutions, its new BOA product follows the latest trends in electronic control devices: packing everything into a single, simple box. Steve Geraghty, director of Dalsa’s Industrial Products group, walked us through the technology.

Geraghty is not unaware that several competitors have supplied diminutive machine vision packages for at least the past 10 years. But, he claims, his company has done its homework and come up with a substantially different product. Everything this new camera has to offer is inside the 44 mm metal cube: the CCD sensor, light control, all the processing (with three different processing chips), I/O and Ethernet connections, and developer and operator application interfaces. All of this is packaged inside an IP67-compliant box that is so small, a golf ball would barely squeeze inside. The IP67-rated housing means that the camera can be deployed directly in harsh, washdown environments without the need for a separate protective enclosure.

“There are manufacturers today who have a variety of options available that offer a range of machine vision capability at various price points,” says Geraghty. “Each of them, in my opinion, provide only a part of the solution; [for example,] most still have to connect to a PC where the software is running.”

The BOA stands by itself, Gerahgty says, and all the software is embedded inside the box. It can do everything a machine vision inspection system is required to do: capture an image, process it, analyze it, and communicate immediately with an automation system if the part needs to be rejected. No external controller, PC, or server is required.

Technology inside

Geraghty adds another interesting claim to the one of supreme integration. He says Dalsa is the first to implement multiple processing engines. There are, in fact, three processing chips inside each BOA: a DSP, CPU, and an FPGA.

The DSP is for algorithm development and optimization; the CPU is for interfacing with operators to configure or monitor the inspection and for communicating the results with the factory environment. The FPGA is used to get the image into the camera and manipulate its quality.

The back side of the camera is the business end as far as external connections are concerned. Despite there being three M12 connectors, only one of them need be used for many typical applications: the 100 MB Ethernet connection. A second connector gives access to BOA inputs and output. It can be used, for example, to trigger the camera, or send output pulses to control the direction of automation equipment. At the top is a dedicated connector for power and strobe control of an external lamp. The remaining space on the back is filled with a set of LEDs for status indication.

Dalsa engineers have come up with a clever way to power their little box, too. In the M12 cordset to the Ethernet port, there are eight conductors: four for Ethernet signals, and four unused. Two of the four unused wires can be accessed for power supply. The camera works on 12 to 30 volts dc. Geraghty hastens to add that this is not Power over Ethernet (PoE), but simply a matter of taking advantage of the unused cable capacity. And, for many customers, it’s a big advantage because it means they have to run only one cable to the camera.

Why 100 MB Ethernet, and not Gigabit Ethernet? Or Firewire? Or USB? In its multi-camera vision appliances, Dalsa supports all the standard communications options, right up through the multi-gigabit CameraLink, but the company sees 100 MB Ethernet as the emerging factory automation standard throughout the world. So it figures this is the bandwagon to jump on.

But why not GigE, gigabit Ethernet? Why not go for high speed? This is machine vision, after all. “We’re not providing images real-time as we connect,” says Geraghty. “We’re just providing the results of the inspection, and 100 MB is fine for that purpose.

“We can send pictures over the Ethernet wire and the performance is, actually, not too bad. We can send 10 to 15 images per second, but this is usually done during the set up period. During normal production runs, this is not normally done.”

Rather than looking at thousands of photos, users can take another approach. “The camera has pretty good memory on board that allows customers to store results and failures, and look at them at any time. They can dial into the camera at any point and monitor what is happening.”

Where’s the light source?

Critics may be forgiven for noting, almost immediately, that unlike many other small machine vision appliances, BOA has no lighting system. There’s no bank of LEDs circling the lens or sticking out of the camera. “We elected not to do that for two reasons,” Geraghty says. “First, we wanted to keep the size of the camera small. And, secondly, our research has shown that in most applications, you can’t used the integrated lamp and you have to purchase and use an external lamp [anyway]. So, we’ve found that the integrated lamp, in many applications, is wasted.”

Another key technological point for the BOA is having the application software embedded in the camera, which means it doesn’t depend on an external computer for support. If the operator wants to interface with the internal workings — for example, to configure the software for an inspection application — he or she only needs to connect via a Web browser. “This eliminates all problems of [software] revision control,” says Geraghty.

The embedded software application is iNspect, which IPD also sells as a PC-based package for its multi-camera systems. Developed for both first-time users and machine vision experts, it accounts for about half the company’s software revenues. The other half comes from a more advanced software package called Sherlock, which will be ported to run on the BOA in the future, says Geraghty.

iNspect software is for end user manufacturers in any industry who either use or want to use vision to improve quality inspection or increase productivity, but who don’t want to become deeply involved in programming languages.

“A lot of times with automation applications, we find that customers don’t think about machine vision upfront,” says Geraghty. “They end up wanting to shoe horn it in after the line is up and running, because they’ve discovered they have quality problems, or something like that.” The small form factor of the camera allows it to be easily integrated to existing lines — attached to the end of a robot arm, for example.

The BOA also targets machine builders looking for vision platforms that can be customized to meet their needs, and small equipment manufacturers who want a private-brand platform to accelerate their time to market. For both of these customer segments, Dalsa provides support for combining proprietary algorithms with iNspect and running both on the BOA platform. Alternatively, customers may only want to purchase the hardware platform and BOA-specific software libraries to create their own exclusive package.

Another feature worthy of note is a PC-based software emulator that works exactly the same way as the embedded iNspect software. This is a tool primarily intended for vision system installers: They can run the camera at a new installation, capture some images on a Flash memory drive, then transfer them to the PC to develop the application with the emulator. Once finished, the application can be sent back to the camera.

Dalsa will initially offer a 640 x 480 pixel monochrome BOA camera for US$1,995, which includes the full complement of tools. A color version will be ready later this year. Although Dalsa makes its own machine vision sensor chips at semiconductor fabrication facilities in Canada and The Netherlands, these are mostly large format, high-end products: sensors with at least 1000 x 1000 pixels. To keep the price of the camera under US$2,000, the company is bringing in its 640 x 480 CCD sensor chips from an outside source. Higher resolution models, says Geraghty, are in development.

Do you have experience and expertise with the topics mentioned in this content? You should consider contributing to our CFE Media editorial team and getting the recognition you and your company deserve. Click here to start this process.