- This article is filed under:
- » View All
New Cabling Techniques Solve Some, but Not All Vision Quandaries
by Winn Hardin, Contributing Editor - AIA Posted 07/17/2000
Camera, cabling, chipset and frame grabber companies face the same problems as the rest of the electronics industry - how do they make their products faster and simpler, while lowering cost. Traditional wisdom has been to incorporate consumer electronics into industrial vision applications whenever possible to capitalize on economies of scale. It was true with the explosion in PC-host vision systems and it would seem to be the same when it comes to cabling between the camera and the host controller.
Today, companies are evaluating four options. These options include the consumer-market driven USB 2.0; the IEEE 1394 or Firewire standard first developed by Apple Computer; the emerging Channel Link with its associated simplified chipset and protocol - again, a technology developed originally to communicate between PCs and flat panel displays; and finally, one of the oldest digital cabling techniques developed by Cypress, HOTLink.
Because industrial vision represents a growing, but niche, application in the larger electronics market, device manufacturers and vision integrators must decide which of these 'square' options can fit best in their 'round' holes.
RS-170 and faster versions of industry-standard parallel analog communication techniques were developed far before the electronics industry began to seriously consider standards. In many cases, the choice of the standard 16-pin connector was as much a product of luck as design. Printer applications drove much of the cabling pin configurations based on the dominance of one company at the beginning of the PC era.
Today's cabling scenario is not much different. Consumer applications still drive the market. Luckily, consumer and industrial vision applications share some common needs, mainly a need for simplicity and higher bandwidth. However, with four technologies vying for the number one slot, a player in the industrial vision market faces a difficult choice if he or she hopes to set their company on the correct road this early in the game.
As late as 1999, 1394 was still heralded as the clear winner - a technology on the cusp of mass market acceptance. Intel intended to incorporate 1394 chipsets into his motherboard, offering PC-based vision systems the economies of scale they needed. As microprocessor speeds increased, pundits began to wonder whether the frame grabber market would be eliminated as companies switched to digital cameras and the world's largest chip maker promised to provide virtually free PC-end connectivity through 1394 chipsets.
Fast-forward a few months, throw in a royalty lawsuit with Apple and Intel becomes the champion of USB 2.0, a protocol with faster bandwidth. Announcements in recent months and even the last few days have added weight to the USB 2.0 standard. In the meantime, National Semiconductor's Channel Link standard has gained significant acceptance within the vision community, as evidenced by more than a dozen camera, cabling and frame grabber manufactures creation of a Channel Link vision standard. So, the question remains, which option is the best?
1394 - do you really need an ethernet?
Firewire, or 1394, was originally designed to connect many consumer products through a singular cabling technology. Offering upwards of 50 MB/s with plans to go to 100 MB/s, the digital cabling alternative is smaller, easier to install and offers automatic network reconfiguration based on a rather significant software stack resident on the PC. Like Channel Link and USB 2.0, incorporating 1394 into cameras or other peripherals requires a relatively simple and inexpensive codec chipset inside the peripheral as well as inside the PC.
Prior to Intel's announcement in March of 1999, Firewire seemed the clear victor. However, when consumer acceptance of firewire connectivity did not develop as quickly as anticipated, the chipmaker altered its plans to make 1394 chipsets standard to all Intel motherboards. While the chipmaker still claims that 1394 has a place in the home networking arena, vision companies have been slower to adapt the technology. While Basler (Exton, PA) and Sony Electronics (Park Ridge, NJ) are among the camera manufacturers to at least partially adopt the 1394 standard, other company officials caution that the bandwidth is still below the needs of high-resolution, high-speed applications.
'Firewire doesn't have the bandwidth we need,' explained Mike Miethig, camera technology manager at Dalsa (Waterloo, Ontario). 'It just doesn't cover our product line at Dalsa and it's a very complicated standard that's designed to be a whole bunch of things to many different people. There's also a lot of software overhead and the standard is about 200 pages.'
USB 2.0 on the motherboard
Taking the place of 1394, Intel settled on a final specification in April for USB 2.0. The spec calls for increasing the bandwidth from USB 1.1's 12 Mbps to 480 Mbps or 60 MBps. The new spec calls for slight modification to a handful of stacks within the operating software, including the host stack.
While the speed is not earth-shattering - especially for high bandwidth applications such as high resolution imaging - it does bring the benefits of a strong endorsement from Intel. According to Jason Ziller, Intel technology initiatives manager and leader of the Intel USB 2.0 development effort, Intel plans to incorporate USB 2.0 chipsets and protocol stacks on their motherboards starting sometime in 2001. That time table could be shortened if the company pursues tentative plans to incorporate NEC's recently announced USB 2.0 discreet host controller, which provides for full bandwidth transmission.
Vision companies looking to explore USB 2.0 interface opportunities will be interested in Intel's announcement at the Japan USB 2.0 Developer's Forum. On June 15th, Intel announced that it will make a USB 2.0 Peripheral Developer Kit (PDK) available in early July. According to Ziller, the tool will allow developers to debug their prototype peripherals through transactions in a single step.
The PDK consists of three key features: a USB 2.0 host controller add-in card developed by Intel and containing the NEC host controller component; a USB 2.0 software driver stack developed by Microsoft Corporation that runs on Windows 2000; and a single-step debug software tool developed by Intel.
While USB 2.0 will not meet all industrial vision needs, Mike Kelley, sales manager for embedded vision system developer ESI (Portland, OR), which focuses on the semiconductor manufacturing industry, feels that USB 2.0 could off significant benefits for his customers. 'We have a customer that is insistent on taking the 1394 path. But we're cautioning them that if the rest of world - mainly Intel - is saying USB 2.0 and not 1394, then we want our customers to proceed very carefully,' Kelley said.
Channel Link standard is here
Perhaps the best compromise for high speed vision applications lies with an interface designed by a consortium of frame grabber and camera companies tentatively called Camera Link. Camera Link is based on National Semiconductors Channel Link chip set. Camera Link, comprised of 11-twisted pair wires terminated with an MDR 26 connector, provides a combination of the traditional low-voltage differential signal (lvds) with serial digital data flow. The result is a smaller cable that is capable of transporting up to 255 MBps of sustained data while maintaining the benefits of a parallel cable with the distance capabilities and environmental ruggedness accorded to serial digital lines, according to Dalsa's Miethig.
A dozen companies organized by Pulnix settled on a vision-based Channel Link standard in April. A finalized version somewhere between 20 and 30 pages is expected by the end of summer. Among its benefits are the ability to increase the bandwidth three-fold by adding a second Channel Link cable and the ease of incorporating asynchronous communications between the frame grabber and camera, along with four high speed control inputs for external synchronization, and other camera functions.
'Having a standardized connector and cable, we're reducing system costs and simplifying the user interface,' Miethig expalined. 'Future protocol enhancements to the standard can lead to the user plugging in one end of the cable into the camera and the other end into the frame grabber, and he's up and running.'
According to Miethig, 3M took part in the cabling discussions during the creation of what will be an open standard. 3M will also supply off the shelf cables designed to meet the requirements of Camera Link. Dalsa intends to have prototype cameras incorporating the Channel Link chip sets from National Semiconductor within the next few months, joining Pulnix and Basler who already offers some Channel Link-equipped models.
'One benefit is that the standard was developed with future growth in mind. There's enough spare control inputs into the camera and data bandwidth coming back that this should hold up for many years to come,' Miethig said. To put that in perspective, Dalsa's fastest commercially available camera is at 200 MB/s using 8 taps. The first single tap offering of DALSA's new generation camera, the Eclipse runs at 40 MHz.
The standard identifies National Semiconductor as the first chip set supplier, but Miethig added that Texas Instrument and Silicon Image make compatible chipsets. The National Semiconductor chips were chosen because the phase lock loop range was wider than the others. 'Or, if you wanted to spend big bucks, you can put in an FPGA from Altera,' Miethig added.
HOTLink legacy continues
One of the earliest digital transmission techniques used in machine vision is still seen as a viable technology by its proponents, including Wintriss Engineering (San Diego, CA) and PPT Vision Inc. (Eden Prairie, MN). At the time the 1394 standard was emerging, the Cypress HOTLink chipset was offering dual twisted pair transmission of up to 33 MBps. In addition to commensurate bandwidths, HOTLink offers considerably longer transmission lengths.
'. . . At the time the [Wintriss PCIHOTLink] board was built, the 1394 standard was 100 mbps and the device to device cable length was 4.5 m. The HOTLink standard which we use has an optimum of 330 mbps (~ 33 MBps) with the board to device cabling of over 100m,' explained Marc Angel, applications engineer at Wintriss.
Like Wintriss' inclusion of power cabling into their HOTLink products, PPT Vision has improved on the original standard as part of their product offerings. According to PPT Vision's Chuck Bourn, their Digital Serial Link (DSL) HOTLink system includes an equalizer, power supply and a specialized protocol designed to meet high-speed vision applications. '1394 has some latency for interrupts that can adversely affect high speed applications. PPT's protocol allows for immediate interrupts, and that's a big difference,' Bourn said. PPT officials say that eventually 75 percent of their camera offerings will be HOTLink compatible. 'We still see it as a viable technology,' Bourn added.
Incremental not paradigm changes
As the industry moves toward all-digital systems, some have questioned what such a move will ultimately mean. Many hope that digital delivery will eliminate the need for frame grabbers all together, reducing the system cost and making vision applications more user friendly. To those, Imaging Technology's (Bedford, MA) vice president of sales and marketing Brad Finney says, 'dream on.'
'The answer is simple, there are many features on a machine vision frame grabber that a 1394 interface card can't deliver,' he said. 'There are features on both the front and back end, such as camera interface, asynchronous, external synch, ability to buffer the image and free up the host from babysitting the video card - features that people use that aren't suplied by RS 170 and won't be supplied by a super cheap 1394 solution.'
Finney continues. 'If you take a normal frame grabber from Coreco Imaging and change it from an analog or digital parallel frame grabber to a 1394 frame grabber, you can run longer distance, but we didn't reduce the costs. In fact, we've increased the costs because the RS-170 interface chip set is less than the [others] . . . In most of our applications, the computer's close to the camera anyway.'
About the Author: Prior to his recent appointment as managing editor of VerticalNet's Wireless Design Online, Winn Hardin wrote for a variety of vision-focused trade publications, including Vision Systems Design, OE Reports and Photonics Spectra.
There are currently no comments for this article.
Leave a Comment:
All fields are required, but only your name and comment will be visible (email addresses are kept confidential). Comments are moderated and will not appear immediately. Please no link dropping, no keywords or domains as names; do not spam, and please do not advertise.