Ethernet Takes Over

The ubiquitous network is moving down to the I/O level.

The huge pricing benefits of commodity technologies continue to expand their usage, often taking them into areas far distant from those they were built for. Ethernet is no exception. The ubiquitous network—which started out in the office—is today extending deeper into manufacturing environments, connecting sensors and other input/output (I/O) points to more sophisticated modules.

Ethernet was derided as too slow and too fragile just a few years ago, but today it’s going into applications that could be described as real-time environments. Some of them are harsh enough to make the factory floor seem like the front office.

Advanced Measurement Inc. uses Ethernet to link sensors in remote oil fields in northern Canada to ruggedized systems that send data to distant offices. The application has millisecond communication requirements as well as demanding environmental specifications. But those hurdles weren’t enough to offset the benefits of Ethernet.

“We like the ubiquitous nature [of Ethernet], and we’ve got tons of experience with it,” says Steve Conquergood, chief executive officer of the system integrator, in Calgary, Alberta, Canada. But it’s pricing that drives him away from proprietary fieldbuses. “Ethernet’s so dirt cheap, we use it everywhere it’s practical.”

That reasoning is shared by a growing number of developers who focus on more traditional industrial applications. While Ethernet may still be rare in the wilds of Canada, it’s becoming the norm in many factories around the globe. “Material handling is one of many fields that are pushing very hard to standardize on this,” says Helge Hornis, intelligent systems manager at automation components vendor Pepperl+Fuchs Inc., of Twinsburg, Ohio.

Managers who use Ethernet for higher level communications are extending the technology down to simple devices such as sensors. All types of dedicated buses are moving to Ethernet. “We’re seeing more activity with safety, putting it
into the Ethernet I/O rack instead of laying out a separate safety network. The real value in that is that safety becomes an I/O function. Instead of wiring special safety relays and safety networks, you wire everything as an I/O point, the same as if it were a motor,” says Jeremy Bryant, networking technology specialist at Siemens Energy & Automation Inc., the Alpharetta, Ga.-based automation supplier.

Many vendors note that the move away from proprietary fieldbuses has grown beyond a few brave souls to become a groundswell. “We’ve sold products that weren’t networked, but now almost no one buys anything that isn’t Ethernet-enabled. Our ARM-based modules all have either Ethernet or a wireless Internet Protocol connection,” says Joel Young, R&D vice president at vendor Digi International Inc., of Minnetonka, Minn., in reference to the company’s products based on ARM reduced instruction set computer, or RISC, processors.


The price of Ethernet chips, or more often, of the central processing units (CPUs) that include Ethernet “for free,” is a critical factor behind the adoption of Ethernet I/O. Prices for proprietary fieldbus chips are relatively fixed, but the price of an Ethernet peripheral on a CPU continues to fall from its already low price point.

But the ubiquity of Ethernet brings other benefits. Equipment providers cite a number of additional reasons for the broader adoption of the network. “We’re seeing a desire for a single, flat network, with one media structure,” Bryant says.

“People move to Ethernet for the commonality or because they want additional functionality. There are basic restraints in the proprietary fieldbuses that Ethernet solves, like node count and data packet size,” he adds.

The pricing benefits of Ethernet extend beyond the cost of purchasing controllers. Eliminating proprietary fieldbuses simplifies maintenance, even though the communication protocols for these lower-level links may well be different than the Internet Protocols used at higher levels. Everyone understands the basics of the network. “The reason for the increasing acceptance is, amongst others, fast start-up because everybody is familiar with Ethernet,” notes Hornis, of Pepperl+Fuchs.

This familiarity, coupled with success stories throughout the industry, is eliminating concerns that once prohibited the use of Ethernet in demanding applications. “Customers aren’t afraid of Internet Protocol networks any more. They may not understand them, but now they trust their suppliers,” says Digi International’s Young.

They can also put their faith in the Institute of Electrical and Electronics Engineers (IEEE) to continue pushing performance upward. Gigabit Ethernet chips and boards are inching into the ruggedized industrial world, providing a path beyond the 100 megabits per second (Mbps) speeds that are in broader use. The IEEE’s High Speed Study Group recently set the stage to push transfer rates up to 100 Gbits per second (Gbps) over fiber, predicting that this version could be completed by 2009.


Until that lofty speed and the transition to optical cables occur, speed remains the biggest concern for those who want to move Ethernet down to the I/O level. Ethernet wasn’t designed to be deterministic, though its increased speed makes it possible to do many tasks that were impossible when the network’s peak rate was 10 Mbits/second.

Now that 100 Mbps and faster products are well proven, many industrial networks can operate without concern that routine messages won’t arrive on time. That’s particularly true in applications with equipment that only sends a few small packets once in a while.

“With high-speed switching, people have gotten over their fear about collisions and late arriving packets. They realize that they don’t see a huge fluctuation in data rates unless they are moving huge files,” Young says. Now, many designers say that Ethernet can typically get down to delay times at the millisecond level. But just how far down into the millisecond level it goes, and how much certainty it can provide, remain open to interpretation.

For Advanced Measurement, a basic network didn’t offer enough assurance that messages would always arrive within the allotted time. The company came up with a solution that brought a lot in performance, but added only a little to cost. “When you’re pushing it down to the millisecond level, it’s not a generalist’s field. We use two physical buses, putting two Ethernet ports at each end, and using a switch so messages are not shared,” Conquergood says.

One of these parallel networks carries general purpose messages such as start or stop. The other handles messages that need to get through in a limited time. The number of nodes and the number of messages are both kept to a minimum, so there’s little chance of collisions or other delays. “The impact of this is that we create determinism at the appropriate level,” Conquergood says.


These techniques work for many tasks, but observers note that many applications require more determinism. Many applications go down to the microsecond level. To get to these levels of determinism, engineers must employ different protocols. A handful of real-time Ethernet protocols, including Cip Synch, EtherCat, EtherNet/IP, Ethernet-PowerLink, Profinet and Sercos III, have emerged for these time-sensitive applications.

Each of them brings different characteristics, ranging from compatibility, in the case of Profibus-based Profinet, to long
distance for EtherCat. “We propagate clocks to take the delay out of long runs,” says Skip Hansen, I/O system product
manager at Beckhoff New Automation Technology, in Burnsville, Minn. “In typical motion control loops, closing a loop in two milliseconds is usually OK. With EtherCat, we can close a loop in 400 microseconds.”

While determinism is usually considered for jobs that require very fast performance, suppliers note that some applications with comparatively slow requirements still require that messages arrive at precisely the right time. “Determinism isn’t necessarily low latency. It’s for users who want to know when something will happen,” Young says.

While implementing these high performance protocols adds some complexity and expense, observers note that this
doesn’t erase Ethernet’s overall pricing edge. “People realize that they have to think about protocols needed for their levels of determinism, and about their performance requirements. The cost of ownership for using one network is really beneficial, even if you’re using different protocols,” Bryant says.

ENTER 1588

These real-time protocols are also getting competition from the new kid on the block, IEEE 1588. This standard, now nearing completion of Version 2.0, makes it possible to synchronize clocks down to the sub-microsecond level. Time-stamped messages can be transmitted with assurance that they will arrive within a pre-determined time period.
Though clocks are synchronized to the sub-microsecond level, messages are rarely sent with timing requirements tighter than a few microseconds.

The standard, which is not limited to usage on Ethernet networks, is seeing growing application, which some cite as a
factor for the growth of Ethernet at this level. “New standards like IEEE 1588 help drive usage by providing timing of messages,” Young says. He adds, “There’s a lot of 1588 deployment starting, with more in Europe than in the U.S.”

U.S. suppliers are expressing more interest, as are other standards groups. The LXI Consortium, which represents the
test and measurement industry, has adopted 1588, pledging to implement Version 2.0 once it’s ready. Vendors are planning to follow suit. “1588 is a very compelling technology,” says Alex McCarthy, product manager at National Instruments, the Austin, Texas-based automation supplier.

For more information, search keyword “Ethernet” at










More in Control