CameraLink is a standard that grew out of the frustration within the industry when having to create custom cables for every LVDS camera/frame grabber combination. CameraLink is administered by the AIA (Automated Imaging Association) and is an official standard. The main purpose has been to standardise connectors, reduce the number of wires in cables and to offer scalability in terms of the number of outputs, bit depth and speed, so that it could be used for any application that previously used LVDS. CameraLink is based on the Channel Link (trade name of National Semiconductor) hardware technology for data transmission and can also be implemented in FPGA.
To provide scalability, the CameraLink standard provides three different levels of implementation, 'Base', 'Medium' and 'Full'. These different levels of implementation are dependent on specific camera requirements, such as data transfer rates, number of bits per channel (the dynamic range) and the number of channels or taps. The data transfer rate is in the order of 240 MB/s per chipset; so for the 'Full' imple mentation using three chipsets this gives a total throughput of roughly 674 MB/s or 850 MB/s in the new extended DECA mode which utilises unused connections from the Full configuration to deliver another two taps of data.
The latest CameraLink implementations on the camera and frame grabber use FPGA designs instead of dedicated chipsets for generation of the CL protocol, giving the advantage of extended cable lengths and reduced costs.
A key point to consider when using CameraLink is that the standard does not define the timing of the camera, or the camera communication. It only provides a physical interface that incorporates RS232 for camera control and Channel Link for data transmission. The standard has continued to evolve as the size of cameras has become ever smaller. The original connector was too large for the modern generation of small compact cameras, and there is now a choice of standard MDR connectors and miniature SDR connectors that were ratified in 2006 as part of the standard. As these miniature connectors require smaller cables, tests have shown that cable lengths are shorter for similar speed cameras. As cable quality can vary, the standard does not define cable specific lengths for these small connectors.
As this standard was developed specifically for vision applications, the connectors include locking mechanisms and the cables can be supplied with varying levels of flex for use in robotic applications.
Additions to the standard include Power over CameraLink (PoCL) and CameraLink Lite. PoCL allows a frame grabber to supply power to a camera via the cable, and is intended for use with smaller cameras as the power is limited. CameraLink Lite is a reduced version of CameraLink Base using fewer bits, allowing even smaller cameras to be built, thanks to the proposed connectors which are even smaller than Mini CameraLink.
It should be noted that the pinout of CameraLink Lite is not compatible with either CameraLink or Mini CameraLink and seems not to have gained traction in Europe.
The CameraLink interface uses a 'point-to-point' architecture which means that each camera requires its own interface channel on the computer. The advantage with this system is that there is no possibility of contention and the data is delivered deterministically, i.e. the time for the data transmission is predictable and constant. For cameras with less than 80 MHz signalling rate, cables up to 10 m are available. Innovative and low attenuation cable solutions enable longer cable length or support of higher signal rates. As most cameras run at slower rates, it is often possible to go up to 15m, however for longer lengths, two types of repeaters can be used: copper wire technology for intermediate lengths and fibre optic for very long distances of up to several kilometres.