Which bus, when?

July 12, 2001
It's now easier to see which kind of industrial bus best serves particular industries and applications.

Perry Sink
Synergetic Micro Systems Inc.
Downers Grove, Ill.

Edited by Miles Budmir

Connectors from Lumberg Inc., Midlothian, Va., are tailored for industrial Ethernet. The hybrid design combines an RJ45 electrical connection with the common M12/M18 shell, and protects against moisture buildup, dust, corrosion, EMI/RFI, vibration, and shock.

Snap Ultimate I/O from Opto 22, Temecula, Calif., supports 10 and 100 Mbits/sec Ethernet, as well as Modbus/TCP, XML, SMTP, and FireWire. It handles I/O, networking, and provides connectivity to other levels of the business.

Few topics in industry have been as vigorously debated as industrial networking. The so-called fieldbus wars raged throughout the 1990s with as many as two dozen contenders.

After all is said and done, five industrial networks are now widely deployed throughout the world. They are, in no particular order, DeviceNet, Ethernet, Foundation Fieldbus, Modbus, and Profibus.

Several networks have had varying degrees of success. SDS from Honeywell competed with DeviceNet, but never gained wide acceptance. Interbus from Phoenix Contact has done well in Europe. And ASInterface has found a limited niche of smaller applications that require just digital I/O.

In many respects, the fieldbus debates mirror the old Beta versus VHS conflict; the formats that prevailed are not necessarily technically superior. In the end, the success of a network depends on broad industry support and successful marketing. No surprise that the most successful networks are backed by large companies: Siemens with Profibus, Rockwell with DeviceNet, Schneider with Modbus, and Fisher-Rosemount with Foundation Fieldbus. Industrial Ethernet is somewhat of an exception, but the specific application protocols are still backed by large companies.

Fieldbus roundup
Profibus is the world's most widely used industrial network. Recognized by its characteristic purple cable and based on RS-485, it is associated with Siemens and is dominant in Europe and South America. Although recent revisions to the protocol have added configuration, diagnostic, and peer-to-peer capabilities, the most widely deployed version is simple Profibus DP that employs polled master/slave I/O.

On the downside, the high overhead-to-message ratio makes it less attractive for handling small amounts of data. There is also no power available on the cable.

DeviceNet features 24-V power on the cable and robust communication based on CAN technology, which has been used in automobiles for over 20 years. Originally designed as a cell-level device network for the automotive industry, DeviceNet has since expanded into assembly, welding, and material-handling machines. Also, most new semiconductor fabrication machines are wired with DeviceNet.

The protocol is optimized for small amounts of data (bits to dozens of bytes) and supports several message traffic options, as well as a nearly unlimited number of configuration and diagnostic variables. Its low cost is also a big plus as is widespread acceptance and efficient use of network bandwidth. However, message size is limited to 8 bytes/node/message.

Modbus is the oldest fieldbus, having been around for more than 20 years. It is used for everything from short serial linkage of smart devices to wide area networks.

A flexible, simple protocol with many capabilities, it is used most often with RS232 (serial, point-to-point) or RS-485 (serial with up to 31 devices). Gateways translate Modbus to almost any other protocol, and nearly every industrial device is available with a Modbus option. Drawbacks include a slow bit rate (38.4 kbits/sec max) on standard serial media and no peer-to-peer capabilities.

The new kid on the block is Foundation Fieldbus and it is quickly becoming the defacto standard in process control and hazardous environments. It works best where large amounts of data must transmit at low speed in an intrinsically safe manner.

Foundation Fieldbus H1 uses the 4-to20-mA standard with a max bit rate of 31.25 kbits/sec. This is a way of representing an analog value by a current draw that's between 4 and 20 mA, with 4 mA being zero and 20 mA full scale. Current is used instead of voltage because it avoids the problem of accounting for voltage drop in the wires. Thus, H1 wiring can run hundreds of feet without affecting the measurement. The 4-mA power source also energizes sensors. And in explosive environments, the low level of current won't produce a spark, providing intrinsic safety in oil, gas, and chemical applications.

Foundation Fieldbus HSE runs on Ethernet at 100 Mbits/sec, and the protocol is basically the same. Foundation Fieldbus protocol has special allowances for critical alarm functions and sophisticated data objects.

Fieldbus roundup








German government initiative and Siemens, 1985

Allen Bradley, 1994

Modicon, 1978

ISA, 1998

DEC, Intel, Xerox, 1976


9-pin D shell (impedence terminated) or 12-mm IP67 quick disconnect

12 and 18-mm waterproof plugs and receptacles, and 5-pin terminal block

DB-9 or application dependent

Application dependent

RJ45 or coaxial

Max # of nodes




240/segment, 65,000 possible segments

1,024 expandable with routers


100 m (copper, noo repeaters) to 24 km (fiber optic with repeaters)

100 to 500 m

350 m

1,900 m for H1

100 m (10BaseT) up to 50 km (mono mode, fiber with switches)

Bit rate

9,600 to 12 Mbits/sec

125, 250, and 500 kbits/sec

9,600 to 38.4 kbits/sec

31.25 kbits/sec and 100 Mbits/sec

10 Mbits/sec to 1 Gbits/sec

Message size

Up to 244 bytes of data/node/message

8 byes of data/node/message

Up to 254 bytes

128 bytes

46 to 1,500 bytes

Messaging formats

Polling (DP/PA) and peer-to-peer (FMS)

Polling, strobing, change of state, cyclic

Master/slave, discrete and analog I/O and parameters

Client/server, publisher/subscriber, event notification


Industrial Ethernet
During the last few years, Ethernet has received considerable attention as an alternative to dedicated industrial buses. It is a hot topic lately because it's well understood, relatively inexpensive, and not vendor-centric. For now, it is the dominant technology for office Local Area Networks (LANs). It is closely associated with the TCP/IP protocol, which is also used for arbitrating messages on the Internet, and it runs on copper, fiber, and wireless from 10 Mbits/sec up to 1 Gbit/sec.

Ethernet is not a protocol but a physical layer defined by IEEE 802.3, and TCP/IP is an arbitration mechanism. So without an agreed upon data format for process, automation, or embedded devices, Ethernet is chaos. This is being solved by porting existing automation protocols such as Modbus/TCP, Ethernet/IP, Foundation Fieldbus High Speed Ethernet, and ProfiNet, to Ethernet. With the exception of Modbus/TCP, all of these protocols have provisions for additional Web-oriented services to enable integration of devices with the enterprise and the Internet.

Ethernet is used extensively to bring data from control systems to business servers and the company LAN. It is only beginning to be used as a sensor-level network. But the industrial environment poses special challenges.

For one thing, Ethernet TCP/IP has a lot of overhead (message bytes dedicated to protocol arbitration, in addition to the data itself) and is designed to handle large amounts of data. If a device only sends 8 bytes of data, then the protocol is only working at 10% efficiency — the TCP/IP packet has about 68 bytes of overhead for every message. This is not necessarily a problem. But an Ethernet network with a large number of simple devices, all sending small amounts of data, may be slower than DeviceNet or Profibus, even though the Ethernet bit rate is higher.

Another concern is determinism. This is the degree of certainty with which a system will respond within a designated period of time. TCP/IP is a collision-based protocol: if two messages collide they both stop and retransmit later. This means that determinism is not possible with multiple devices. A piece of hardware called an Ethernet switch solves this problem. It provides dedicated bandwidth for each device without collisions.

Selecting a network
The choice of a network is as much political as it is technological. It has much to do with vendors, application requirements, and preferences. It also depends as much on whether the application needs vision, HMI, or just simple I/O, as it does on bit rates, arbitration schemes, and packet sizes.

The first decision is whether to go with an open or a proprietary network. Although the meaning of "open" is debatable, the key issue is whether a single vendor controls the network definition. In general, "open" means the specification is available for purchase by anyone and there are few legal restrictions on making compatible devices. Open is usually synonymous with widely available, although technically speaking, it just means that the technology is public knowledge.

On the other hand, proprietary networks like AB Remote I/O, Data Highway Plus, Modicon Modbus Plus, and GE Genius I/O are tightly controlled by their developers and cannot be supported without licensing agreements. In most cases, the product selection is much narrower and the technology is older.

Although users have been demanding interoperability for years, large vendors have avoided it. Why? Because interoperability gives more choices. So even in an open system, suppliers are always tempted to introduce proprietary elements (usually in the form of added features) which limit interoperability.

But the Internet has shaped expectations. People naturally ask, "If my son can play video games on the Web with some kid in Singapore, why won't your PLC talk to my encoder?" In such a climate, it is unrealistic for vendors to field proprietary communications. This is why all of the networks widely used today are supported by hundreds of vendors with unrestricted participation.

Selecting a network also depends on application-specific requirements and preferences. It depends on the PLC used, whether I/O is mostly discrete or analog, and the presence of any motion control. For instance, machines usually have a lot of fast, discrete I/O. Process applications have mostly slower analog I/O.

Analog-intensive, process-oriented applications will benefit most from Profibus and Foundation Fieldbus which is optimized for big chunks of data moving at slow speeds, rather than small bits of data moving quickly. DeviceNet tends to be used for more discrete applications.

For motion control, a specialized network like Sercos or CAN Open is generally the best fit. Even 100 Mbit/sec Ethernet can be used with the speed compensating for the lack of determinism.

Also important is that different protocols are preferred in different parts of the world. For example, U.S. customers may demand DeviceNet while European customers want Profibus. Careful planning may make it possible to handle both scenarios with common hardware footprints, software interfaces, and configuration tools. Deployments of the two networks can be 80% the same from the perspective of writing software and building equipment.

Fieldbus sites on the Web
The Web site of the Profibus trade organization contains news and technical support.
www.odva.orgThe Open DeviceNet Vendor Association home page.
www.modbus.orgBilled as a site by and for Modbus users, you'll find technical overviews as well as an active discussion group.
www.fieldbus.orgThe Fieldbus Foundation home page contains news, technical information, as well as a schedule of upcoming training seminars.
www.iaopennetworking.comAt the home page of the Industrial Automation Open Networking Association, you'll find news releases, discussions of issues surrounding Ethernet on the factory floor, and links to many other fieldbus and industrial networking sites.
www.synergetic.comHere you'll find a detailed fieldbus comparison chart which compares nearly two dozen fieldbuses. While here, you might also want to download an MP3 of the "Fieldbus Blues."

Common messaging formats

  • Master/slave — Master device asks for information; slave device cannot initiate message.
  • Peer-to-peer — Any node can initiate a message and communicate with another.
  • Broadcast (or strobe)— One node sends a message that all other nodes can receive.
  • Poll — Master communicates with slaves on an individual basis.
  • Change of state (or event driven)— Nodes report data only when there is a change.
  • Cyclic — Nodes report data periodically.
  • Client/server — Server responds to a query when the client requests data.
  • Producer/consumer (or publisher/subscriber) — Client notifies server that it wants to receive a certain type of message, and subsequently receives messages when they are published.

Bus basics

On the left is a control panel for a point-to-point wired automation system. On the right is a system wired with DeviceNet. Network-based systems dramatically reduce the number of wires and the associated complexity. Although they have a higher component cost, networked systems save wiring and labor costs and have the potential to report diagnostic data that is unavailable with point to point wiring.

A fieldbus is basically an industrial communication network. It lets remote I/O devices such as sensors, actuators, and controllers talk to central controlling devices such as PLCs or PCs. The main benefit is less wiring. The alternative to using an industrial network is to hardwire each data point — connecting every discrete and analog I/O point such as proximity switches, encoders, temperature controllers, and transducers, back to a central location. Wiring each physical connection leads to an unwieldy bundle of wires. It's also a labor-intensive task that may involve pulling wires through conduit and attaching them to terminal and junction blocks.

Process control plants with thousands of I/O points can look like a spaghetti city. This is why large factory automation and process control applications use fieldbuses extensively. With a network, one wire from each device is connected to a single network cable having four or five wires.

Devices that connect to a network can also report more information back to a central controller than their nonnetworked brethren. So in addition to current physical values, setpoints can also be monitored and changed over a network. A temperature controller, instead of being manually configured, can be set up over the network to report the temperature at the current time, set point, and some PID parameters. Analog temperature controllers may have to be programmed manually with buttons, knobs, etc. The digital version provides access to every variable inside the device.

One example is a mass flow controller used in semiconductor fabrication. The analog version may contain seven parameters. The DeviceNet version has more than 200. As always, the one chosen really depends on the application, what is being manufactured, and the degree of needed control.

As systems grow past the threshold of 100 I/O points, the added expense of network hardware is offset by savings in wiring time. These benefits grow exponentially with the size of the system.

Overall, the fieldbus area is growing quickly. Industry experts predict that in five years it will grow to nearly twice its current size. Driving this expansion will be building automation applications along with more traditional areas such as the automotive industry, and process and factory automation.

Voice your opinion!

To join the conversation, and become an exclusive member of Machine Design, create an account today!