Networked I/O Rides the Ethernet Wave

Jan. 25, 2013
Users are taking advantage of the different ways suppliers are exploiting Ethernet and the various communications protocols available to distribute I/O devices.

There may be more than one way to skin a cat, but there is just one way of processing turkeys at the Cargill plant in Waco, Texas. And that way now relies on networked input-output (I/O) devices to oversee the ammonia-based refrigeration system at the heart of production. Like many users seeking to keep their processes as simple as possible, Minneapolis-based Cargill is linking the devices over an Ethernet-based network.

Keeping tabs on the refrigeration is crucial. “If the system were to go down, they would have to shut the plant down,” explains Barrett Davis, owner of Automate Co. LLC (, a systems integrator based in Pacific, Mo. For this reason, Cargill asked Davis to upgrade the system with controls and I/O from Opto 22 (, an automation supplier based in Temecula, Calif.

The upgrade at Cargill is indicative of a trend toward greater proliferation of Ethernet in industrial automation. “The use of Ethernet as a physical layer has been the biggest advancement in fieldbus I/O over the past decade,” observes Joey Stubbs, P.E., North American Representative for the Nuremberg-based EtherCAT Technology Group ( He attributes the trend to the ability of the automation industry to industrialize consumer-driven commodity products for fieldbus implementation, thereby driving development costs down.

“In the past, the world’s fieldbus organizations had to create their own standards for everything from the connector to the protocol to the stack,” Stubbs says. “Now, the physical layer can be handled by broader-based international standards organizations that drive these technologies for all areas of use—industrial, office, and consumer.”

Consequently, networked I/O has continued to advance, resulting in better performance, flexible topology, and simpler configuration. “With today’s networked I/O, system updates can be processed real-time in the sub-millisecond to microsecond range,” reports Kurt Wadowick, I/O product specialist at Beckhoff Automation LLC ( in Burnsville, Minn. These higher processing speeds continue to generate efficiencies in data collection and analysis.

Just as important has been the simultaneous evolution of open communications protocols. “Ethernet’s physical wiring combined with protocols such as EtherCAT, Profinet, EtherNet/IP, Powerlink and others were each originally developed individually by controls manufacturers,” explains Wadowick. “They, however, have re-emerged as open standards that can easily interface into other companies’ devices, such as drives, encoders and various kinds of industrial controls on the I/O network.”

As fieldbuses have become more open and dependent upon known physical technologies like Ethernet, I/O network update rates have increased to the point where they can keep pace with real-time events in the control system. Another important advantage has been less copper. “In the good old days, a great deal of wiring was required for industrial devices, says Wadowick. Today, he says, connecting these devices to the controller requires only the I/O network connection, control power and, in the case of high-energy devices, three-phase power.

Simplicity with TCP/IP
At Automate Co., Davis specializes in the Opto 22 products because they use standard transmission control protocol and Internet protocol (TCP/IP) on Ethernet-based networks. He likes the protocol’s simplicity. “It’s been well-thought-out from the get-go,” he says. “Serial has always been kind of a hassle to work with because you can have mismatches between two systems from different manufacturers.” Although he acknowledges that you can always get serial networks to work, he notes that it usually requires a lot of effort.

So, Davis usually argues for TCP/IP over some of the newer deterministic fieldbus networks wherever data collisions do not affect the reliability of TCP/IP communications in control applications. “These fieldbus solutions add complexity to Ethernet and TCP/IP,” he says. “They try to turn something nondeterministic by nature into a deterministic network. These attempts often over complicate the entire process and make interoperability between different systems difficult.” It also drives up costs, because installing such things as IGMP (Internet group management protocol) snooping switches is necessary.

Davis has found that the added complexity is unnecessary for most applications that he typically sees. “UDP [user datagram protocol] packets go across the network in a matter of a few microseconds,” he says. “Even if you get a collision, it has a whole scheme for handling that, and it retries very quickly.” Based on his analysis of Opto22’s hardware to respond, he is confident that he can configure most proportional-integral-derivative (PID) loops to scan rates of less than five milliseconds.

Although Davis believes that TCP/IP communications are fast enough for many I/O networks, an important exception is machinery requiring sub-millisecond timing. For now, most of these applications will rely on specialty bus networks, he says.

Delivers flexibility
Builders of industrial automation for heavy manufacturing industries like automotive are also seeing the benefits of networked I/O. Not only must the automated machinery there coordinate the motion of its own components, but it often must also synchronize its tasks with other equipment and people. And these lines must have the flexibility to accommodate a mix of products and design changes.

>> Don't Forget the I/O Networking Details: Noise, Power Supplies, Current. Click here for more information.

An example is vehicle production. “With high production rates and a wide variety of model mixes within today’s automotive factories, manufacturing systems are becoming much more complex,” says Andy Jones, global segment director for commercial vehicles for Comau—Adaptive Solutions of Southfield, Mich. “Even within commercial-vehicle manufacturing facilities, where the production rates can be more than 10 times lower than the typical passenger vehicle factory, networked I/O is still extremely important.”

Simplicity is also important to this business unit of Comau S.p.A., the Turin-based builder of assembly lines for a variety of industries. One reason is that the business unit serves industries, such as commercial vehicle manufacturing, that are relatively new to complex automation. Another reason is time to market. The time for bringing a vehicle from concept to market is as crucial to market share in commercial vehicles as it is for passenger vehicles.

A flexible and robust communications architecture, therefore, is very important for eliminating costly delays in designing, integrating and delivering the automation built by Comau’s Adaptive Solutions business. “Our customers expect reliable, cost-effective systems with the flexibility of changing to alternate component suppliers when necessary,” says Jones.

Although the existing infrastructure—as well as the needs and abilities of the user—will dictate the networking scheme chosen in the end, Camau engineers often specify EtherNet/IP, a standard protocol promulgated by ODVA (, with products from Milwaukee-based Rockwell Automation ( and others.

The EtherNet/IP protocol “enables all network devices to speak the same language, making network communications much more robust,” says Jones. This ability allows putting a variety of floor-level components, as well as integrated technology, including motion control, on the same network.

Right now, Comau is looking into using EtherNet/IP for changing the tools used by its robotic technology. “Automatic tool-changing is very common in our assembly systems, in order to make best use of the robots, precious floor space, and available cycle time by performing multiple operations within the same workspace,” explains Jones. Because quick-connect technology has not always been available for making these changes, the automation builder’s engineers would often have to design solutions that required communicating over DeviceNet, which added an extra protocol. Putting the robots on the EtherNet/IP network should eliminate this complication.

Another important feature of EtherNet/IP is that it supports the development of modular solutions that can be used again and adapted for future projects. Not only does reusing previously proven modules save time in design and commissioning, but it also streamlines equipment validation.

On the Safety Side
Networked I/O can be an improvement over safety relays as well, as the Lawrence Berkeley National Laboratory (LBNL) learned in Berkeley, Calif. The lesson came when it built Bella, the world’s most powerful laser, a petawatt laser that generates pulses that last for 40 fetoseconds. To protect workers and visitors from harmful radiation and exposure to laser light, LBNL installed safety system with a distributed I/O and human-machine-interface (HMI) architecture.

Bella occupies four moderate-sized rooms. The safety system monitors the 14 access doors and the shutters covering the large picture windows that observers use to view the laser. Some of the doors prohibit access to hazardous areas while the laser is on. If, however, other doors or the shutters are open, the programmable logic controller (PLC) will safely shut the laser down by taking control of the 18 shutters, three laser beam dumps, and 17 power-supply circuits.

“It wasn’t possible to do this project with safety relays. There is just way too much I/O for a programmable safety relay,” notes David Di Giorgio, lead engineer for the project to design and install the laser’s safety system. Di Giorgio is director of computer engineering at Deterministic Systems Inc. (DSi) (, a controls engineering company in Walnut Creek, Calif.

The heart of the system is an S7-319F safety PLC, the fastest model made by Siemens Industry Inc. ( of Alpharetta, Ga. Using ProfiSafe open communications, an Ethernet-based ProfitNet network links the PLC to four remote safety I/O racks and four HMIs, one each for each room. Distributing the racks to the rooms simplified installation. “The cabinet would have been quite long if we would have wired them all to one rack,” observes Di Giorgio. “Then, we would have had to figure out how to route all these cables and pay for all the copper.”

The ability to send safety information and control “from one room to another using only Ethernet and a set of power wires made the design clean and easily modifiable during the design phase,” Di Giorgio adds. “In the future, this will also allow LBNL to modify the system as the needs of the experiments change.”

DSi engineers got an idea of just how flexible the system is near the end of the project, during the final checking phase. The inspection uncovered that a group of four alarms failed to turn off a device as they should have. Rectifying this oversight was a matter of changing one instruction because the alarms were already grouped. Di Giogrio reports that actually reconfiguring the instructions and compiling and downloading the change took roughly 15 minutes, but that the whole event took about an hour when you include the discussions surrounding it.

Besides this ability to make changes on the fly, another benefit of the distributed I/O architecture is that it assists troubleshooting. “When you’re troubleshooting a problem, it is more convenient to have the rack in the same room,” notes Di Giorgio. More importantly, the network offers more diagnostics than relays do. Although relay-based systems can offer redundancy, they cannot detect a short circuit to a sensor coming from another system. Failsafe inputs and outputs on the safety PLCs, on the other hand, can detect short circuits.