News  Briefs  Features  Calendar  Opinions  References  Products  Subscriptions  Editorial FAQ  Media Kit

01 April 1998
Ethernet Bursts the Field-Bus War Bubble
by Steve Schoenberg

As the field-bus wars rage, a resolution is quietly making its way onto the factory floor—Ethernet.

Who won the field-bus war? The users, of course. They all stopped attending networking seminars and asking the editors of Industrial Computing to sort out the myriad of "standard" bus choices. Everyone just continued to install Ethernet. It started in the office, and, like most great ideas, it spread. It advanced from project to project until one day everyone realized that they were just a few connections away from complete manufacturing execution system integration.

Okay, so maybe not everyone has jumped on the Ethernet bandwagon, but, if not, they soon will. If you listen to the Ethernet critics, you'll hear all about its limitations: It's too expensive for device-level networking or it's not deterministic. In the meantime, your plant must have protected functional network layers. While the bus standard debates continue, all these issues have quietly been resolved. Let me share some overnight paradigm shifts that have changed the way engineers are thinking about industrial networking.

Have you looked lately at the cost of an Ethernet card? Recently, our team of development engineers showed me their plans for a RISC-processor-based Ethernet chip set for a new device-level product. I was astonished! Ethernet solutions are shrinking in size and cost and the new wiring schemes are a snap (literally). Just click a 10Base-T telephone-style plug into each device. (I, too, was a coax cable fan, but, like most engineers, I evolved.) Ethernet is getting less expensive by the minute, and it's one of the few network solutions that offers wiring choices.

Another paradigm shift: Fortunately, Ethernet is not deterministic. Remember that your goal is real-time performance. Determinism is just a technique to get you there. Simply put, you need data on time, every time-meaning you need it in real time. To achieve this feat, deterministic systems leave adequate time to perform all actions (necessary or not) during each and every scan. Determinism provides real-time results by guaranteeing the slowest performance all the time.

In contrast, modern computer-based systems offer real-time solutions by first throwing abundant processing power at the problem. (Does anybody want to compare the processor in a PLC with a 300-MHz Pentium II?) Add to that even a small amount of intelligent software and real-time systems start to fly. Just ask a technophile about preemptive multitasking and report on exception. The point? Ethernet has been a catalyst to the paradigm shift away from deterministic solutions to improve real-time results.

All this still leaves the need for protected functional network layers. We have always had dedicated I/O links from PLCs to the plant-floor computers using ModBus, Profibus, RS-232, and so on. Why not replace this potpourri of mediocre solutions with a dedicated Ethernet I/O network that is fast, reliable, inexpensive, and real time? Yes, I said real time! Remember that I/O data packets are tiny. An analog I/O point is only 2 bytes, and 200 discrete I/O points total only 25 bytes. At 10 Mbps, Ethernet I/O networks are so lightly loaded that occasional collisions become insignificant. Keep this I/O network separate from the plantwide Ethernet network. Windows 95 and NT let you install up to four Ethernet cards in the same computer. Now, you still have separate functional network layers, but you've eliminated the need to maintain multiple wiring schemes.

What about interoperability? After all, the goal is still to connect it all together. The answer starts with TCP/IP-clearly, the uncontested answer to global connectivity. But again, Ethernet gives you choices. Okay, say you already have DECnet, NetBEUI, Novell, or something else. It turns out that all these protocols sit inside the same Ethernet packets and travel the same highway. Yes, you can send ModBus or "proprietaryNet" messages over the same Ethernet as well. All protocols seamlessly share one Ethernet highway. Thus, in the quest for truly open systems, you can't get any better than this.

Recently, I had Road Runner Internet service installed in my home. Through a special modem, the cable TV was linked to my Ethernet 10Base-T hub. In just 15 minutes, I had a TCP/IP connection to the world. I was especially delighted to see that my existing Windows NetBEUI network was undisturbed. (Also, in my effort to be a better parent and communicate with my kids, I can now network with them.) The start-up cost for this multimegahertz Internet access, including high-tech equipment, installation, and training, was only $49! Ethernet is part of the new computer paradigm that tells us that "the less we pay, the more we get." There is a tremendous economy of scale when we all agree to do the same great things.

I wish I could say that I'm offering new ideas here, but I can't take the credit. Years ago, ABB, Foxboro, Siemens, and the rest of the big boys adopted Ethernet as their top-end real-time solutions. They typically made small improvements in the interests of performance and applied proprietary names to their networks as marketing people often do-just ask someone from Siemens if its expensive H1 network is Ethernet based. All of a sudden, they are all changing their tune and proclaiming their dedication to open standards. It was reported in a feature article entitled "Networking steps to the front of the class" (January 1998 InTech, page 36) that Ethernet is now the standard network in 72% of all plants that have adopted a standard.

I will offer one piece of advice. Don't throw out your Profibus, DeviceNet, or other proprietary network. There is great value in preserving your investment and maintaining the systems that work well for you. However, the next time a salesperson tells you the benefits of his or her proprietary solution, just ask if it comes with an Ethernet gateway.

To all the people in our industry who have spent countless, and often thankless, hours working to create, promote, and implement a viable field-bus standard, I offer my thanks. You have focused our attention on the need to standardize and defined valuable benchmarks that will make our jobs easier for decades to come. All we need to do now is acknowledge that Ethernet is the bus that meets our important standardization and performance objectives and that we are all choosing to use it in the field. When all is said and done, there's no doubt that Ethernet will be the winner of the field-bus war. IC

Additional Information

Author Information

Steve Schoenberg, president of SixNet (Clifton Park, N.Y.), has directed development of real-time network products since 1976. He holds an M.E.E.E. degree in electrical engineering and is a licensed professional engineer in New York. Schoenberg designed the first real-time remote I/O network for a PLC in 1979, and in 1984 he designed the first real-time I/O network coprocessor for the emerging PC bus. Presently, he is involved in the SixNet I/O for Windows products, which integrate distributed intelligent I/O into Windows-based systems using Ethernet. You can contact him via e-mail through the link above or c/o editors at IC@isa.org.