Machine vision: high-speed, long-distance communications
CoaXPress (CXP) is a high-speed point-to-point serial communication standard for the transmission of video data, scalable over single or multiple coaxial cables. It has a high-speed downlink of up to 6.25Gb/S per cable for video, images, and data, plus a lower speed, 20Mb/S uplink for communications and control.
Broken down to its simplest terms, machine vision is the ability to use a camera, computer, and an illumination source to make a decision. Each component—camera, computer, and illuminator—has witnessed rapid changes over the past decade, transforming the imaging landscape and making machine vision more powerful, scalable, and affordable.
While we are all well aware of the seemingly daily developments in PC architecture and the migration from analog to digital cameras, there is a less frequently mentioned component to machine vision: the interface technologies between the camera and the computer.
Cameras need to transmit data to the computer so that it arrives intact in a relatively short time frame. When analog cameras were in fashion, and some of them still are, this interface was a BNC coaxial connection between the camera and a data acquisition board known as a frame grabber. The frame grabber was usually a PCI interface that fit into the motherboard that allowed data to transfer seamlessly between camera and computer memory.
Just as cameras, computers, and illumination technology have developed, so has the frame grabber—now most common in a PCIe slot on the motherboard. Modern frame grabbers are capable of carrying interfaces such as LVDS, RS422, and Camera Link. Frame grabbers are also being developed to handle the newest of machine vision interfaces, CoaXPress.
The CoaXPress interface was recently adopted by the combined agencies of the Automates Imaging Association (AIA), European Machine Vision Association (EMVA), and the Japanese Industrial Imaging Association (JIIA), collectively known as G3. Merging the simplicity of coaxial cable with serial data technology, CoaXPress permits high-speed image data to be transferred over long distances in area-scan and linescan applications, even when captured by the largest, fastest, and most bandwidth-hungry of today’s CMOS sensors. This kind of application goes far beyond the capabilities of established interfaces.
CoaXPress (CXP) is a high-speed point-to-point serial communication standard for the transmission of video data, scalable over single or multiple coaxial cables. It has a high-speed downlink of up to 6.25Gb/S per cable for video, images, and data, plus a lower speed, 20Mb/S uplink for communications and control. Power is also available over the cable ("Power-over-Coax") and cable lengths of greater than 130 m may be achieved, much farther than twisted pair.
Industry analysts widely expect CoaXPress will be the fastest growing interface in machine vision over the next 3-4 years. This owes to a range of advantages including: low cost of standard 75 Ohm coaxial cables, ultrafast data transmission rates, and achievable longer distances. For even higher data rates, multiple CXP channels can be used. CoaXPress can be expanded to four ports, resulting in 25Gb/s of data throughput. Despite the long distances, CoaXPress permits trigger events to be sent from host to camera and vice versa with minimal latency and jitter.
Because it uses standard coaxial cable yet delivers impressive bandwidth, CoaXPress is drawing interest not only from the machine vision community but also from integrators in the medical, military, and security fields where legacy coaxial cable is often deployed. By “repurposing” already installed coaxial cables, CoaXPress eases the migration from analog to digital video.
To better appreciate these advantages let's look at an application example: license plate readers. In Europe, the standard license plate is large enough that it is easy to read, 18- to 20-in.-long plates, 4-in. tall, and large bold, unhindered fonts. The U.S. license plate is 12-in. x 6-in., and apart from up to seven alphanumeric characters, there is also a state name and sometimes an image on the plate. This leads to a lot of confusion at low resolutions, for example deciphering whether the plate of a car that drove through a toll booth without paying was ABC123 or A8CIZ3. The processing center would have to verify the car make and interaction with the registry, which in turn dramatically reduces efficiency. If the resolution and data delivery were sufficiently quick, it would be a simple read. However, while a higher resolution camera would offer the solution, an interface is needed to bring the data back to the processing center extremely quickly—a distance typically further than twisted pair cable can handle even with repeaters.
The CoaXPress interface will be of major benefit here. Several analog traffic cameras are on U.S. highways. This data is coming back at 11Mb/S over a coax cable from the camera to the computers. Consider the volume of data coming back at even one CXP link (650Mb/S). This would enable images of traffic on routes to be crystal clear and also allow for better information dissemination about traffic backlogs, accidents, or breakdowns. These images would be clear enough to be seen on a smartphone app. And because CXP offers Power-over-Coax, there would be no need for additional cabling work.
Additional CXP applications
CoaXPress will play a vital role in many aspects of machine vision in the coming years. As designers sit down with this new interface at their disposal, more ideas will be formed and CoaXPress will enable machine vision to perform more efficiently.
- Donal Waide is director of sales for BitFlow, Woburn, Mass. Edited by Mark T. Hoske, content manager CFE Media, Control Engineering, Plant Engineering, and Consulting-Specifying Engineer, firstname.lastname@example.org://%0Dwww.bitflow.com