Serving Up Ethernet as I/O Fieldbus

Nov. 8, 2006
Users like different flavors of the technology.Retrofitting an old gantry requires a lot more than simply replacing the manual, pull-along crane with a robot—especially when the plan is for people and a high-speed robot to share the 100-foot long workspace in front of a bank of machines. Just ask Steve Dickerson, chief technology officer at CAMotion Inc., an Altanta-based systems integrator. “You have to do a very fast, repeated scan to make sure that none of the safety conditions have been violated,” he says. Fast and reliable communications are a must.
And not just any old communications scheme would work. “It had to be Ethernet-based because of the high data transfer rates and the ability to go to a number of destinations without special wiring,” says Dickerson. “It also had to be safety rated.” In other words, the job called for a very particular flavor of Ethernet. Consequently, CAMotion and its customer joined the growing ranks of manufacturers that are reading the menu of available flavors and choosing one that will work well as a fieldbus for controller-to-input/output (I/O) communications.The robot added all of the spice to the palletizing application that CAMotion just finished. While it is loading payloads as heavy as 150 pounds in one of the three work zones along the front of the machines, the operators are busy working at the machines in the other two zones. There can be no mechanical barriers to impede the free flow of people and materials. Instead, light curtains around each zone and various sensors on the bridge keep the robot away from the people and other objects.Two other factors complicated the mix. First was that the robot uses a vision system in its work. Second was that it travels overhead in a 100-by-25-foot workspace and the sole programmable logic controller (PLC) for the palletizing operation is on the ground. Not only would stringing 100 feet of cable be costly, but the user did not want any wires hanging from the robot. So the distributed I/O scheme had to include a wireless link between the robot and the PLC, one that would not let the vision system bog it down. Dickerson selected the Profinet local area network (LAN) from Siemens Energy & Automation Inc., Alpharetta, Ga., because this Ethernet protocol can process both safety-rated and regular I/O in a wireless manner. “We are not separating safety and standard I/O data from each other,” says Ralph Buesgen, Siemens marketing manager for Profinet in discrete automation. The safety-related data is embedded with the regular I/O data in a standard Institute of Electrical and Electronic Engineers IEEE 802.3 Ethernet telegram.“The Profinet safety solution works without redundant data and time stamps in the telegram, which simplifies the protocol structure and speeds the flow of data,” says Buesgen. “The maximum size of user data in a Profinet package is 1,440 bytes of user data; the maximum Ethernet-frame size is 1,520 bytes.” Consequently, the PLC can poll every I/O node, run a standard and safety program, and react in less than 50 milliseconds, even when transmitting vision data. Both the PLC and bridge have managed switches at the wireless access point to handle the communications. The managed switch allows the Ethernet communications to be given a different priority on different packages of information. The safety data have the highest priority. The regular I/O has the second highest priority, and the data from sensors such as the vision system have the lowest.Although the vision system can transmit images to the ground without overloading the system, the operator typically looks at the vision images during setups and debugging. The rest of the time, the data usually remains on the gantry. Likewise, most motion-control data also stay on the gantry because the drive and other necessary electronics travel with the robot along the bridge. “Real-time motion-control data are not going through the wireless portion of the Ethernet, except for some safety functions,” says Dickerson. “Motion control is all up in the moving bridge robot, where wired Ethernet updates information 1,000 times a second.”Dickerson likes the taste of fieldbus for controller-to-I/O communications because integrating safety and other I/O saved at least 30 percent of the cost of building the system. So he plans to make Ethernet a bigger portion of the job the next time around. “We’ll probably go to a serial interface based on Ethernet on the ground too,” he says.Injecting SimplicityThe desire for simplicity also was a driver at Ferromatik Milacron, of Malterdingen, Germany, about five years ago when the injection-molding machine builder began using Powerlink from Roswell, Ga.-based B&R Industrial Automation Corp. as the communications bus on its all-electric models. “Our goal was to reduce the number of variants that we have to support in the field,” says Thorsten Hoes, manager of research and development at Ferromatik Milacron.That meant using only one network to link the controller with the transducers that report screw position, clamping pressure and other variables for motion and process control. The network also had to carry signals for controlling the enclosure doors to keep people safe, and to ensure that events occur in the correct order to prevent damage to the machine and the mold.When the company began converting from hydraulic motors to electric ones, its engineering staff could see that the various kinds of analog inputs that they had been using before had to go. Not only did the variety of input techniques create complexity that was no longer wanted, but the analog communications also would not be able to handle the much larger volumes of information that the electric drives and motors would need. Although making the connections and getting data to flow were simple, making some important decisions was much harder. “We had to agree on what should be the content transported across the fieldbus,” says Hoes. “You have to agree upon what information you will be sending and what it will look like.”Hoes is glad that he and his group invested the time in making good decisions. “This is a very reliable connection to external devices,” he says. He reports that his company has not seen any problems with any of the 2,000 to 3,000 nodes on PowerLink buses that the company has in the field. Using Ethernet as an I/O bus solves the reliability problems that the designers had to spend time solving for the old connections in the past. Magnetic fields, for example, no longer interfere with data transmission.Another reason for the simplicity of the installation and the lack of complaints in the field is the engineering that B&R Automation put into its product. “Users want an approach to communications based on standard Ethernet nodes, rather than on proprietary ASIC (application specific integrated circuit) hardware,” explains Markus Sandhoefner, at B&R. “If you have a network that relies on proprietary hardware, then you depend on one supplier. This bears risk.”Standardization notwithstanding, vendors still make important contributions, thereby generating the various flavors of Ethernet that are out there. The most important ingredient of these flavors is the technique used to reduce collisions that cause data jams and slow communications. B&R introduced time slicing for its Powerlink. “It’s like a regulation of the traffic on the network so only one station is able to talk at a time,” says Sandhoefner. “It lets you communicate very fast, and, more importantly, in deterministic real time, which means the arrival of a telegram is guaranteed to the microsecond.“Some have gained determinism and speed for their networks through custom ASICs,” he continues. “We wanted a system that is as open as possible on the hardware side, so we do that time slicing in software, not hardware.” Therefore, users can build their networks on standard Ethernet nodes that are available anywhere, even those in typical notebook and desktop personal computers (PCs).For Accurpress America Inc., of Rapid City, S.D., response time was the initial reason that the builder of press brakes began using an Ethernet-based fieldbus in its Accell machines. A custom job had forced the company’s hand. “It required faster response times than conventional fieldbus technology could handle,” explains Alex Kvyatkovski, team leader for research and development in the company’s Surrey, British Columbia, Canada, facility.Until that job came along, Kvyatkovski and his design team had been using Lightbus from Beckhoff Automation LLC, Burnsville, Minn., as their fieldbus network. “Because it uses fiber optic cabling, Lightbus is very noise-tolerant and provides fast I/O processing,” he explains. The compact telegram structure of its protocol permits a transfer rate as high as 2.5 megabits per second (Mbps) of user data. To get the greater performance demanded by the custom job, Accurpress’ engineers decided to try Beckhoff’s Ethernet-based fieldbus, EtherCAT. The bus can process 1,000 distributed I/O in 30 microseconds (µs) or 100 axes in 100 µs. “Up to 1,486 bytes of process data can be exchanged with a single Ethernet frame,” says Skip Hansen, Beckhoff’s I/O systems product manager. “The transfer of this data quantity takes only 300 µs.“Thanks to hardware integration in the slave, and direct memory access to the network controller in the master, the complete protocol processing takes place within hardware and is independent of the run time of protocol stacks, CPU (central processing unit) performance, and software implementation,” he adds. “The speed of the EtherCAT I/O and [Beckhoff’s] TwinCAT Real Time software, for example, gives Accurpress extremely fast analog-digital I/O data and deterministic motion control.”The faster response times reduced cycle times significantly, making it much easier for the machine to hold its accuracy. As Accurpress’ most advanced press brake, the Accell machine uses closed-loop servo hydraulics and PC-based controls to generate parallelism and repeatability +/- 0.0004 inch. For more information, search keyword “Ethernet” at a Determined communication technologyEthernet has come a long way since its early days in the late 1970s. Back then, it used a shared medium, one in which all of the nodes use the same physical wires for communications. As one might imagine, packets of data from devices might bump into each other whenever they attempted to travel the data highway at the same time. Whenever that happened, the devices would wait for a random amount of time before trying to transmit their data again. If enough devices were competing, the volume generated by the continual re-sending of data could have overloaded the data highway.About 10 years ago, researchers solved the problem for most applications by using switches. Instead of connecting devices to a common medium, most vendors now connect them to packet switches, which are specialized, high-speed processors that regulate the flow of data and determine when they will travel the highway. “A device on a switch thinks it’s sharing a wire but finds it always available,” explains Harry Forbes, senior analyst at the ARC Advisory Group Inc., in Dedham, Mass. “So, there are no longer any collisions between devices competing for access to the medium.”The only applications for which the speed of data transfer might still be an issue are those involving the precise synchronization of motion. When it comes to using Ethernet as a fieldbus for
controller-to-input/output (I/O) communications, today’s switching technology is more than adequate.
In other words, many industrial applications do not need some of the higher performance technologies that exist today. “A lot of variables in an oil refinery, for example, aren’t going to change extremely rapidly,” offers Forbes. A run-of-the-mill Ethernet technology might work fine there. “If you’re just monitoring tank levels, you can use any number of protocols, including IP (Internet protocol) communications, for your I/O.”Invensys Process Systems, of Foxboro, Mass., has a different approach. “We address the major concern of Ethernet, the lack of determinism (specifying when data will arrive at its destination), through the use of a protocol that makes the controller the only master, the communications initiator, on the network,” says Alex Johnson, Invensys’ director of systems architecture.A range of Ethernet technologies exists. Switching is the norm, but vendors have a variety of variations on that theme. Architectures such as Profinet IRT (Siemens) and EtherCAT (Beckhoff Automation), for example, have no switches in the conventional sense. Rather, “they have put the switching into chips that implement their interfaces,” says Forbes. “So they have essentially built two-point switches into their devices.”Another architecture, called the Ethernet hub, seems to be one that can work in the most demanding motion-control applications. PowerLink, from B&R Automation, is an example of an Ethernet technology that uses this architecture instead of switches. “It’s really a machine control architecture for very precise synchronization,” says Forbes. “It uses a hub and an enhanced medium access protocol for allocating access to a shared medium.”