Is Ethernet the solution for industrial networking, or will it create additional problems?

According to a white paper published in April 1998 by Rockwell International Corp., "Ethernet's worldwide acceptance in industrial and office environments has created eagerness to expand its responsibilities on the plant floor. The network's performance capabilities make it ideal for tasks such as data monitoring and program maintenance. However, many predict recent advancements in Ethernet and the emerging Fast Ethernet technology will also enable it to handle mission-critical control responsibilities currently being managed by existing industrial automation networks. Meanwhile, others contend that Ethernet has a long way to go before it can assume an expanded role in the manufacturing environment."

Well before Ethernet was considered as "fair game" for industrial automation, control engineers had been watching (as well as participating in) the development of a myriad of bus systems. Over the last 30 years, however, individual buses have always evolved to serve niche markets. Development of appropriate silicon, network configurations, cabling and related connection hardware, and system software have proved to be complex, time consuming, and expensive that a "one size fits all" bus system has not emerged—yet.

To provide a complete "backbone" across an enterprise, a bus application must function across the bit-level sensor, device, control, and business levels. Developments in Ethernet technology, coupled with the build-up of corporate intranet/internet infrastructures, have given companies good reason to bring Ethernet down to the factory-floor level to leverage their network investments. The perceived ability of this technology to cross connect corporate information systems has driven companies like General Motors and Boeing to investigate its use.

Ethernet is already hard at work in process plants that use commercially available control networks. According to Dick Caro, vice president at Automation Research Corp. (ARC, Dedham, Mass), various process control systems are based on Ethernet, but use their own electronics, including The Foxboro Co.'s (Foxboro, Mass.) I/A Series Nodebus, ABB Industrial Automation's (Rochester, N.Y.) Advant, and Elsag Bailey's (Wickliffe, O.) Symphony products. In fact, Nodebus is actually redundant Ethernet with mechanisms added to ensure no single point of failure.

Mr. Caro says Ethernet is a well-developed, high-speed protocol. It is available as the standard 10 megabit/sec and as Fast Ethernet which runs at 100 megabit/sec. Because the basic "silicon" has been long developed, there is no premium for the faster version, PC cards average $50.

Another recent development includes IEEE specification 802.3z, or Gigabit Ethernet. This protocol was developed for use as an enterprise-wide backbone. Although its high speed (1,000 Mbits/sec) offers industrial users a definite speed advantage, it is expensive and makes its use at the workstation, printer, and node devices prohibitive, at least for the near future.

Ethernet is criticized as being nondeterministic. For most critical processes, even in a lightly loaded Ethernet, there is a chance a data "collision" may occur and a critical message may not get delivered. However, additions to the protocol can be made to reduce this chance. Many of Ethernet's recent proposed and approved changes, while designed for nonindustrial applications, have also had the effect of reducing the likelihood for collisions.

Ethernet was originally conceived as a multidropped bus network using insulation-piercing (or vampire) connectors. Although cheap and easy to install, this system proved to be unworkable. Engineering evolution and cost considerations teamed up to change it into a twisted pair-based star topology. Recent hardware developments increased throughput and effectively eliminated "nondeterministic" concerns.

Use of "intelligent" hubs, with message switching capability and full duplex wiring (the coaxial cable used in most industrial applications is half duplex), makes Ethernet fully deterministic and eliminates the problem of indeterminate delays in information reaching its destination. Switch hardware is down in price; $100 each for standard and $300 each for Fast Ethernet networks. Increased speed and carefully chosen topology reduce the chances of data collision to "infinitesimal," according to Ram Ramachandran, senior group manager, industrial communications R&D, at National Instruments (Austin, Tex.).

Ethernet's overhead is also a concern to potential industrial users. The protocol requires a full 32-bit address, a 32-bit cyclic redundancy code frame check sequence, and 46 bytes minimum message length. Ethernet requires this minimum message length for detection of collisions from two devices located at opposite ends of the network. Because specialized industrial networks (Profibus, FOUNDATION fieldbus, etc.) have fewer nodes and do not require long addresses, industrial users seem to have a groundless aversion to using long messages (46 bytes) to pass small data sets. According to ARC's Mr. Caro, "Ethernet is almost no protocol at all. It is true sharing of the wire. Ethernet does not get in the way of other protocols."

As Ethernet works its way into the process industries, it will have to overcome the problem of installation in hazardous environments. For areas requiring intrinsically safe operation, Ethernet is not recommended yet. Although not intrinsically safe as it stands, Ethernet can be routed through hazardous areas using explosion-proof conduit. Presently, there are no plans to certify it for explosion-proof applications.

Process industries in general, but especially those that handle explosive, flammable, or toxic substances, require the utmost in control system safety. The benchmark for safety systems is generally accepted as triple redundancy. While double redundancy is readily available for most Ethernet applications, triple redundancy is not. Safety systems commercially available today are highly specialized and usually proprietary in nature. Unless Ethernet is designed in by the manufacturer of a safety system, its use in a safety system is limited.

"Redundancy is inherent, utterly dominant in process control. Distributed control systems are redundant with nodes everywhere, but they haven't had a dominant fieldbus, until Ethernet," explains Steve Schoenberg, president of Sixnet (Clifton Park, N.Y.). Sixnet's EtherTrak modular Ethernet I/O blocks (CE, November product) aims to eliminate the need for most specialized industrial networks. The modules, under $400, benefit from Microsoft's (Redmond, Wa.) push into industrial applications, Mr. Schoenberg suggests.

Control engineers anxious to use Ethernet realize that "global acceptance does not an industrial network make." Despite the abundance of compatible products, impending solutions to both determinism and redundancy issues, and its ability to handle large amounts of data at relatively low cost, possible Ethernet migration to the plant floor raises questions that require hard answers.

• Can the amount and type of devices, size of data packets required, and exchange frequency be managed to deliver an acceptable level of system determinism and repeatability?
• Will the anticipated mix of information and control messages affect the applications timing requirements?
• Are the control devices interoperable, or does the application-level protocol vary by vendor?
• Will the required control devices and network components withstand plant-floor environment?
• Do the necessary components meet required agency approvals (FM, CE, UL, etc.) and/or ratings (NEMA, etc.)?

Finally, keep in mind that when a network is truly enterprise wide, it cuts across departmental boundaries. Management must clearly establish who will be responsible for installation, maintenance, and management of the network before expansion begins.

As Ethernet continues to make inroads in control applications both on its own and as part of proprietary systems, its success has also attracted other bus systems in progress. Development of an Ethernet-based FOUNDATION fieldbus is now in the works.

Although the original Fieldbus Foundation 31.25 kbit/sec H1 protocol is adequate for many process applications, incorporation of Ethernet into the H2 specification will extend functionality into systems that require higher speed. Ethernet-based versions of ControlNet, Profibus, Modbus, and Java technologies also are available or under development.

Although high-speed requirements are thought to be the domain of industrial automation, process systems also will benefit from accelerated data acquisition capabilities and an improved integration between fieldbus-compatible and installed control systems. As use of Ethernet grows in control networks on all levels, fieldbus (and other Ethernet-based), products will find their way into an increasing number of retrofit applications.

Even though Ethernet still "shares a ride" with other protocols in the industrial arena, its time to work there alone is approaching. As determinism and redundancy issues fade in the light of speed and hardware improvements, Ethernet could "stand alone" in some industrial control applications. When asked about the possibility, Chris LeBlanc, industrial communications production manager for National Instruments (NI), commented, "With the interest NI sees in Ethernet for control and the rate of technology development, that could be only one to two years from now."