GenICam shows promise

A key part of building a machine-vision system is choosing the right interface for the application: Should it be Camera Link, FireWire, GigE Vision, or some other option? In this exclusive interview, Matthew Slaughter, vision marketing engineer at National Instruments, assessed several options and also offered praise for the GenICam standard.

Q: What is the most important challenge facing today’s camera standards?

A: Any worthwhile camera standard can make image acquisition easier, faster, and cheaper. While FireWire certainly provides these benefits, its maximum cable length of 4.5 m represents a serious limitation for many applications. In those applications, older analog camera technology still predominates. GigE Vision permits cable lengths up to 100 m. In the near future, I expect some combination of Firewire and GigE Vision to finally show analog the door.

Although often overshadowed by GigE Vision, the most significant camera standard in recent years may be GenICam. Sponsored by the European Machine Vision Association (www.emva.org), the GenICam programming interface formed the basis for GigE Vision and has since spread worldwide.

Matthew Slaughter Vision Marketing Engineer National Instruments Courtesy of National Instruments.

Q: What does GenICam add to the mix?

A: The FireWire specification defines register architectures that each camera has to follow. Such discipline is great for software interoperability but can prove too restrictive for camera manufacturers, who can’t adequately differentiate their products’ features. With GenICam, the cameras define themselves in XML, permitting them unique capabilities that still work with anyone’s software. For example, a LabView customer can use the most advanced features of any GenICam camera without needing a specific driver for that camera.

Q: Is GenICam limited just to GigE Vision?

A: At the moment, only GigE Vision requires GenICam compliance, but GenICam XML files are appearing for FireWire cameras, and several software packages support GenICam for both FireWire and GigE Vision.

Q: How does GigE Vision compare with the other standards?

A: GigE Vision requires the system CPU to acquire and build images from packets. FireWire and Camera Link do not require nearly the same CPU overhead. To reduce this workload, companies have developed several optimization levels into their drivers. Low-level, chip-specific camera drivers can strip off image packets before the CPU ever sees them.

Q: What changes do you see as we move forward?

A: For one thing, the 680 Mbytes/s that Camera Link offers is already too slow to cope with the next generation of sensors. Either Camera Link will have to be expanded, or the industry will need a new standard to handle the higher data rates.

One high-speed option that could be tethered or cabled is PCI Express, which would let cameras connect directly into a computer’s backplane, bringing them one step closer to the processor and onboard memory. Fiber-optic cabling could allow data to travel long distances at high speeds.

We also see more intelligence moving into the camera itself. Many cameras already contain FPGAs, so why not perform image processing there as well?