Greg Blackman on the advances being made in machine vision cabling solutions, including the technology that forms the basis of the Camera Link HS interface standard
Ten or more years ago Camera Link was pretty much the only dedicated machine vision interface standard. The cabling for Camera Link is fairly bulky and can only reach 10 metres, but for machine vision needs it did and still does work very well. Machine vision though is a fairly niche area, so often the most cost-effective solution of transmitting data from a camera to a PC is to adapt other more consumer-based interface standards, like Gigabit Ethernet and USB, to the requirements of industrial imaging, hence GigE Vision and USB3 Vision.
In terms of the cabling, one of the trends happening in the industry, according to Ron Folkeringa, business manager at cabling provider Intercon 1, is a shift towards fibre optic cabling, making the most of the technology developed by the telecoms market.
One of the newer standards and the successor to Camera Link, Camera Link HS (CLHS), has a protocol based on fibre optic cabling, where cables can reach kilometres in length if necessary. CLHS’ fibre optic lane runs at 10Gb/s in the X-protocol.
‘The SFP+ modules [small form-factor pluggable (SFP) transceiver] can be bought online for $20 in China. Fibre cabling might be around $10 for 5 metres. You’re looking at 1.2GB/s through a $50 cable [for CLHS]; fibre cabling is cheaper than copper at longer distances and higher bandwidths,’ commented Michael Miethig of Teledyne Dalsa and chair of the Camera Link HS committee.
CLHS was first developed by Teledyne Dalsa as HS Link, before being offered to the machine vision community to be developed into a standard.
‘Copper always used to be the cheapest solution, but at 10Gb/s fibre becomes the only way to go beyond 10 metres,’ Miethig continued. ‘There are now SFP+ single lane solutions that rival the cost of copper at 10 metres. You’ll see those prices come down as well. For short distances, copper still has some advantages, but I anticipate this will change in a couple of years.’
The CLHS copper cable costs around the same as a Camera Link cable, but it can carry 2.1GB/s of bandwidth per cable compared to 850MB/s in two Camera Link cables. The CLHS copper cable reaches 15 metres, as opposed to a maximum of 10 metres in Camera Link. Because the standard uses the CX4 or Infiniband cable for the M-protocol, plug-in copper-to-fibre adaptor modules can extend the 15 metres to 100 metres.
One of the design goals for CLHS, Miethig stated, was to make use of the telecom industry’s technology and cabling, because it is low cost, proven, and is available for long periods of time. ‘We didn’t want to develop unique physical layer technologies that would only be used in machine vision because our volumes are so low,’ he said.
CoaXPress, another new machine vision standard addressing the higher bandwidth applications, is also based on existing, proven cabling solutions, in this case simple coaxial cables.
The CLHS M-protocol CX4 cable is designed for 5Gb/s using 8b/10b encoding, which is an encoded physical layer. Camera Link is not coded, so sending a long string of zeros will result in problems with Camera Link cables, noted Miethig. Because of the 8b/10b encoding in the M-protocol, the DC levels balance so data can be sent further and cables can handle higher bit rates.
Machine vision connectivity solutions require asymmetric bandwidth, said Miethig, so a low bandwidth link going to the camera and lots of bandwidth coming back from the camera. The second requirement is real-time control.
According to Miethig, CLHS probably has the best real-time performance of any packet-based protocol at 3ns jitter on trigger. ‘In real time you don’t have to schedule anything,’ he said, adding that the latencies of the cable now become important. Version two of CLHS, which is under development at the moment, will be able to synchronise multiple cameras with various cable lengths to nanosecond precision. CLHS also has the ability to split the data across multiple processing units, because cameras can create so much data that the PCs can’t keep up. This is important in web inspection and other high bandwidth applications.
When installing a machine vision system, the cabling is just as an important factor as the camera, lens and lighting. ‘What needs to be considered is the length of cable that is needed and the cost associated with the technology required to achieve longer distances,’ commented Folkeringa at Intercon 1.
Another consideration noted by Ray Berst, president of Components Express, which supplies machine vision cabling, is the environment the cables will be subjected to – whether the cable will have to move or be static, or whether there are concerns about electromagnetic interference, for example. Components Express has recently released a 25 metre USB3 Vision copper cable solution, which extends the relatively short reach of USB3 Vision cabling.
‘Equally important is to determine the appropriate cable for those applications that are high flex,’ Folkeringa commented. There are many types of flex motions and not all high-flex cables are suitable for a particular type of motion. A torsional motion, for instance, is different to a rolling flex or tick-tock motion. ‘It is important that your cable supplier understands the type of application so that they can offer the best solution that will withstand those types of forces over millions of flexes,’ he said.
In terms of technology advances, Folkeringa noted that now cable assemblies are available with integrated signal conditioning, which could include signal, pre-emphasis and equalisation. He added that there is more demand for robotic cable assemblies that incorporate multiple devices and motor connections within the same cable.
In terms of Camera Link HS, Miethig noted that the upgrades planned for version two include adding multiple buffers whereby each buffer in the frame grabber can have a different pixel type and bit depth. A 3D camera can therefore send 3D data to one buffer and the raw image to a second buffer, for example. The speed will also increase on both M- and X-protocols and the new version will add more robust frame header and frame trailer packets.
In the wider machine vision cabling market, Berst predicted that many standards like USB3.1 and Thunderbolt will require active cables as a standard, while Folkeringa commented that he expects to see new methods for achieving longer distances, as well as wireless applications being developed. ‘Bandwidth and data rate is currently a major limiting factoring in utilising wireless technologies,’ he said. ‘I think we may see some novel solutions that overcome these barriers in the next 10 years.’