By Dick Caro

In 1999, the question was: "Can  Ethernet be used in an industrial automation application?" Today, the question is: "Which version of Ethernet should I use for my industrial automation application?"

There is so much confusion over Ethernet in the manufacturing world that making the decision about which one to use can be very difficult. Most of the time, the decision is made when you pick your automation equipment supplier, since few suppliers in the industrial automation world support more than one version of high-speed, real-time, industrially hardened Ethernet.

Before we look at the many incarnations of Ethernet, let's get our technology straight. In all cases we are referring to full duplex switched Ethernet, not the old collision domain variety that was responsible for the belief that Ethernet was not deterministic. Todayís switched Ethernet is fully deterministic and highly suited for all types of real-time applications. The basic Ethernet protocol has not changed since Bob Metcalf created Ethernet for the Xerox Palo Alto Research Center in 1973, except for the minor adjustments made when Ethernet became the IEEE 802.3 standard. With the insertion of full duplex interfaces and a full duplex switch, the old collisions are gone, along with the old four-shield coaxial cable and its highly unreliable vampire clamp tap. Replacing expensive coax cable is inexpensive twisted pair unshielded cable, although many industrial applications do better with shielded cable.

The remaining piece of Ethernet that seems to be less than industrial quality is the connector - the eight-conductor RJ-45 plug and jack. Several manufacturers have addressed the plug and jack problem and now offer a rugged version of the RJ-45 designed for IP-67 use and a high vibration environment. Another approach is to use the DIN standard M12 connector in either the four- or eight-pin version that also provides an IP-67 rating and is essentially immune to vibration. All of these are more expensive than the RJ-45 connector, but not enough to eliminate Ethernet.

Industrial Ethernets
Ethernet has been adapted for industrial applications in process control, materials handling, discrete automation and motion control. Most of the adaptations consist of adding new application layer functions on top of the UDP/IP or TCP/IP network, transport and session layer protocols most commonly used with Ethernet. However, some of the Ethernet adaptations used for very high-speed multi-axis motion control applications make a few adjustments to the physical and data link layer protocols as well.

Table 1 shows the most popular Ethernet-based protocols used in industrial automation, and the automation applications they support. Process control concerns the regulatory feedback control of scalar or analogue variables usually involved in the manufacture of chemicals, petroleum products, pulp, paper, plastics and metals. Factory automation concerns the logical control of discrete, mostly two-position, items, some of which require interlocking and sequencing usually involved in the assembly or automated transport of an item. Motion control concerns the speed of a two- or three-dimensional movement to machine, assemble or place items with great positional accuracy.

Table 1

Protocol                                            Application
                                                            Process Control    Factory Automation    Motion Control
Foundation Fieldbus HSE                          *                          
EtherNet/IP                                                                                        *  
Modbus/TCP                                                 *                                  *
PROFInet IO                                                                                      *  
Sercos III                                                                                                                                 *
EtherCAT                                                                                                                                *
Ethernet Powerlink                                                                                                               * 


Foundation Fieldbus HSE
Foundation Fieldbus HSE was created to satisfy the original requirement for an upper-level network technology to interconnect the H1 fieldbus segments. The original fieldbus standard specified several new network technologies, all of which posed some risk, were not very fast, were expensive, and were relatively untested. By basing the upper-level network on Ethernet, HSE is fast, inexpensive, well tested and, therefore, risk-free.

HSE is implemented as a special application layer using unmodified full duplex switched Ethernet and UDP/IP protocols. All functionality of Foundation Fieldbus H1 is implemented in HSE, allowing this protocol to be used at the field level when necessary, and for real-time control loops. HSE adds a well-integrated bus redundancy capability to improve availability and reliability. Any speed or cabling technology suitable for Ethernet may be used for HSE. Wireless Ethernet technology, typically using Wi-Fi, has not been tested or approved, but many users are planning this typical Ethernet extension.

EtherNet/IP
EtherNet/IP is a specification of Open DeviceNet Vendors Association (ODVA) in co-operation with ControlNet International. These two "Open Specification" organizations produce the publicly available specifications for DeviceNet and ControlNet, respectively. Both ControlNet and DeviceNet use a common application layer called CIP (originally Control and Information Protocol, now Common Industrial Protocol), and so does EtherNet/IP.

EtherNet/IP uses unmodified standard switched Ethernet and a combination of UDP/IP protocols for real-time or high-speed real-time and control messaging, and TCP/IP for all connection and setup messages. Message data frames are identical with both DeviceNet and ControlNet when CIP is used, allowing EtherNet/IP to act as a common upper-level network for field networks using a number of devices with the lower level CIP messaging. This compensates for a common Ethernet problem in factory wiring where it is a star topology, and sensors and actuators are often in a linear topology. Typically, DeviceNet is used to interconnect sensors and actuators with a gateway to EtherNet/IP.

Modbus/TCP (also called Modbus TCP/IP)
Modbus was the original industrial data network, and Modbus/TCP is the Ethernet application of it. The purpose of Modbus/TCP is to connect host computers to controllers, and therefore it is not intended for fast, real-time control. Any of the traditional Modbus function codes (also called commands) may be used, where the user or the controller vendor defines the set of "registers" appropriate for the controller.

The original Modbus message frame was intended for transmission over a simple serial line using RS-232, 422/423, 485. That older specification contained a frame check sequence (FCS), a code that was used to validate that no data had been lost in the transmission. Modbus/TCP does not require the FCS since Ethernetís own cyclic redundancy code and the end-to-end messaging protocol of TCP provide greater message integrity. All messages use TCP/IP protocol. Most devices that support Modbus RTU also support Modbus/TCP.

PROFINet
PROFInet is the latest of a series of specifications originating from the Profibus organization. There are four different classes of PROFInet implementations, each for different applications:

• PROFInet IO - intended for high-speed data exchange between a PLC and its I/O unit, or between PLCs; uses TCP/IP transmission.

• PROFInet CBA - intended for data exchange in a distributed processing network; uses TCP/IP and DCOM data exchange.

• PROFInet RT - often called "soft real-time," RT is intended for real-time control data exchange; uses UDP/IP transmission.

• PROFInet iRT - isochronous real-time is intended for multi-axis motion control applications that require very high-speed synchronized data communications.

All forms of PROFInet except iRT use standard full duplex switched Ethernet. iRT uses a special chip to achieve its synchronization.

Sercos III
Sercos was the original digital network established for high-speed multi-axis motion control. Sercos III is based on the use of Ethernet components, but the protocol establishes a hardware implemented real-time channel in order to meet the synchronization needs for a large number of multi-axis motion controllers or digital drives. In addition, Sercos III provides dual ring topology for redundancy when needed for high reliability. Sercos III use of Ethernet is related to the cabling and switches used to build the network. Additionally, standard Ethernet TCP/IP messages can be carried on a non-real-time channel.

By using a counter-rotating ring network topology, Sercos III provides the capability for device-to-device communications not available in previous versions of Sercos. Use of the redundant form of Sercos requires unique cabling and does not take advantage of Ethernet's low-cost connectors, but does provide additional network reliability.

EtherCAT
EtherCAT is also intended for use in motion control applications where very high-speed communications are necessary. EtherCAT achieves high-speed multi-point communications by sending the entire database of values at one time. Each value is mapped to the database at a configured location. Each single database multi-cast therefore sends the entire database with extremely high efficiency. In this way, each axis of a multi-axis motion control application receives its data at the same time, since it knows where to access its setpoint or position data from the database. All transmissions use conventional Ethernet UDP/IP protocol without any custom hardware.

EtherCAT uses standard Ethernet except for the final connection to the field that uses LVDS (EIA-644) cable and connectors that are designed for high-speed, low-voltage data transmission and the industrial environment.

Ethernet Powerlink
Ethernet Powerlink is based on standard unmodified Ethernet with both TCP/IP and UDP/IP data frames as necessary for the application. A managing node must be specified to control communications with all real-time devices. The managing node regulates all communications dividing between real-time and ordinary communications. During the configured time interval for real-time communications, the managing node will allow communications with only one real-time node at a time, therefore avoiding any network collisions to assure both deterministic and high-speed data transfer. During the time period remaining, conventional Ethernet TCP/IP communications may occur.

Cabling for Ethernet Powerlink is conventional Ethernet cable and industrial RJ-45 or M12 connectors.
There is a version of industrial Ethernet available for your application. They are all better choices than proprietary networks. Use the one supported by your primary equipment supplier.


Dick Caro, CEO of CMC Associates in Acton, Mass., is a respected author and expert in industrial networking topics. His three books - Automation Network Selection, Wireless Networks for Industrial Automation and The Consumer's Guide to Fieldbus Network Equipment for Process Control - can be purchased by visiting www.isa.org/books or www.spitzerandboyes.com/Product/fbus.htm.