By Leslie Langnau, Managing Editor
Because top level executives want to know what’s going on in all departments of their company, they often implement the Ethernet protocol. It’s the backbone of more than 85 percent of the world’s LAN-connected PCs and workstations. Information technology (IT) departments use it for many applications.
Engineers in industrial automation departments, however, are often less receptive. Of the available networking choices, many view Ethernet as well as the International Standards Organization (ISO) seven-layer Open System Interconnection (OSI) model implementations as too slow and unpredictable. These engineers often prefer a shrunken, three layer communication system commonly known as a “fieldbus” network. Examples, such as DeviceNet, Profibus, and Modbus, comprise protocols conforming to the smaller three layer version of the OSI model: the physical layer, the transport layer, and the application layer.
“Some think that if a solution wasn’t developed by the industrial automation community, it’s not appropriate,” says Benson Hougland, vp marketing, Opto22 Inc, Temecula, California. “But why wouldn’t someone want the large R&D budget benefits of a Cisco System or other ‘IT’ type company?” Conversely, IT engineers aren’t keen on having automation applications tie into their Ethernet system. “
IT engineers’ resistance comes from their concern over possibly opening their network to loopholes to accommodate automation,” says Jason Haldeman, lead product specialist, Phoenix Contact, Middletown, Pa. “IT people control their Ethernet system. When automation engineers tie a cable into their network, the IT people want to have control of that network too, a scenario industrial engineers resist.” Ethernet now meets most industrial needs, including multi-axis synchronized motion control. The Ethernet specification has evolved over the years and lists many versions under IEEE 802.3_ that address earlier perceived shortcomings. It is a standard protocol that defines how data move through a medium (Category 5 wire and fiber optic cable) from one device to one or more devices in a predictable manner. Ethernet specifies only the mechanical and electrical connections of a network and how devices will access and share them.
It needs other protocols to create a full communications network.The OSI model recognizes that communication requires specific tasks be executed in specific order to ensure data integrity and reception. The model divides those tasks into seven layers, the fewest number deemed necessary to facilitate communications among disparate devices while enabling the greatest amount of flexibility for a huge range of applications. Ethernet meets the OSI specification for the first two layers of the model.The model was developed so that vendors would compete through layer seven, the application layer. The model was expected to promote generic hardware, and that upgrades, adaptations, and changes would be done through software, saving cost and time. Instead, industrial device vendors chose to compete on all levels, delaying the benefits of the OSI model.
Pick your option
Ethernet is incredibly flexible. It can be a deterministic system that fits most applications, transmission speed, or bandwidth.
The SFN unmanaged switches, from Phoenix Contact, use third-generation switch technology. They offer entry-level switch functions with five or eight ports (10/100 Mbps) in a narrow housing width. Also included is a range of 100Mbps glass fiber configurations, which support SC or ST style connectors. Each switch allows optional security frames to block unused RJ45 ports and lock in existing cables. This “physical layer 1” security restricts network access by unauthorized personnel, and yet is easy for plant floor personnel to implement.
Layout: Ethernet can be laid out in hubs and spokes favored by IT departments, multi-drops preferred by industrial engineers, or both ways in different parts of the same company. Ethernet can be connected peer-to-peer or master/slave. Although industrial engineers prefer master/slave, they may want to try peer-to-peer and learn its benefits.
Transmission/Bandwidth: Ethernet runs at 10Mbps, 100Mbps, and one Gigabit, fast enough for most time-sensitive applications. Multiple speed versions can be combined in one network. For example, one portion of an Ethernet configuration can operate at 10Mbps and another at 100Mbps. Most industrial automation applications are satisfied with 100Mbps. Gigabit Ethernet, known as GigE, is often a backbone network in industrial and information applications, but will soon be used in more task specific applications.The Ethernet protocol calls for data to be transmitted in “packets.” A minimum size packet is 84 bytes, with 46 bytes of actual data, not layer specific overhead data. Variations on this requirement can suit most applications.
Determinism: Switch devices and network segmentation guarantee deterministic, point-to-point networks that send messages to their destinations in microseconds. Large Ethernet networks take as little as 52µs to transmit a data packet. With 100Mbps versions, update time of I/O points can take less than 10 ms. “An Ethernet system can provide faster updates that a DeviceNet or Profibus system,” adds Haldeman. “But engineers need to be concerned about others communicating on that network. For example, speed can be increased if I/O is segmented from operator machine interface functions.
Switches are managed or unmanaged. Managed switches include processor intelligence that automatically logs device addresses and examines message contents for efficient forwarding. Unmanaged switches are essentially network routers, passing through any message that comes to them. Unmanaged switches will not prevent a phenomenon known as a broadcast storm. Most networks can broadcast messages to all attached devices, but if not handled properly, that messages can be resent by multiple devices multiple times, creating anoverload that clogs the network, reducing bandwidth. “A fieldbus network uses one master node to communicate to I/O points,” continues Haldeman. “Ethernet is not a one master system, any device can access it at any time.”
But Ethernet doesn’t need to follow the master/slave format to guarantee point-to-point communications. “The issues about deterministic bandwidth and time were based on Ethernet installations before the advent of the switch,” says Tom Edwards, senior technical advisor, Opto22. “Today, a switch costs what a hub cost five years ago and no one installs an Ethernet network that’s not a switched network. This means, the network is point-to-point, ensuring two devices communicate to each other without interference from other nodes. Therefore, the collision domain no longer exists. Plus today’s Ethernet is so fast and trouble free that determinism is no longer an issue either.” But one concern about switching is access. “There’s a chance that anyone can connect to that Ethernet system,” says Haldeman. “I can walk up to an Ethernet port, plug in a cable and hook up my laptop, and find a path to the Internet and connect it through that port. The problem is that this could bog down the baud rate. This is why security and managed switched components are becoming hot items.”
Distance: The wiring distance of Ethernet is not much of a concern anymore. The 10Mbps Ethernet networks typically run on unshielded twisted pair (UTP) cable. The 100Mbps version, known as Fast Ethernet, comes in three cable versions: 100BASE-TX media is usually Category 5 UTP cable; 100BASE-FX media is fiber-optic cable; and 100BASE-T4 uses two extra wires with level 3 UTP cable. The media for the 100BASE-TX standard is compatible with the 10BASE-T Ethernet standard.
Fiber optic cable is also used for Gigabit Ethernet. It offers benefits for applications that must deal with electronic transmissions or where environmental hazards are a concern. Plus, the standard allows fiber cabling to be to two km long. But most applications don’t need network distances of more than 100 meters. “Serial networks, for example, do have a slight advantage over lower speed Ethernet versions (10Mbps and lower) when it comes to distance,” notes Edwards. “Ethernet’s copper wire is limited to a 300 foot segment, while RS485 can go to 4,000 feet. You can boost Ethernet’s distance once, and then you are at the limit of what copper wire can handle.”
A solution is to use a media converter. This device supports copper wire to its maximum distance, connects to a fiber optic cable to transmit to thousands of meters, then connects back to copper wire at the destination end. “The converter and the cabling are not expensive anymore,” adds Edwards, “ and many engineers are doing this.”
First in, then out
Ethernet ports and connections have been showing up in all types of industrial devices, including I/O modules and sensors. Recent changes enable features such as remote monitoring and management, IT integration and voice over video. Some I/O modules, for example, incorporate switches into their design. In the newer modules, the switch may not be on the I/O module but on a control backplane. A processor on the I/O module talks to the backplane, and performs a task usually done by another controller or processor. One advantage is reducing the number of hardware devices needed in a communication system.
“Another advantage is the ability to take multiple I/O points for sensors, for example, and MUX them together to one IP address,” notes Hougland. If you have Ethernet there, you also probably have some processing power, so you can do a little preprocessing, perhaps give a datum an engineering unit or linearize a nonlinear signal from a thermocouple; all can happen at the I/O level. By the time you throw the data onto the Ethernet, they are in a usable, consumable form by a computer or controller.”
After an EthernetI/O product becomes a host on a peer-to-peer network, it is autonomous. Internet Protocol (IP) running on top of Ethernet in these arrangements adds more benefits to a communication system. With IP as the third layer, an industrial communication system can offer features similar to the Internet. “To browse a website,” says Hougland, “I use the http protocol. To send email, I use the SNMP protocol. To transfer a file, FTP, and so on. The same can happen in industrial communication. For closed loop control, I might use QoS or some other closed loop synchronization program. If I just need to gather data, I could use FTP, or maybe ModbusTCP and run it through Wonderware or another OI with that capability. Or perhaps I’ll use OPC. The bottom line is, once I’ve standardized on IP I have so much to choose from and can be very versatile and smart in terms of how I obtain or control the information on the network itself. Ultimately it’s all about routing IP because everyone is running on the Internet, which often runs over Ethernet.”
R-Series EtherCAT sensor is a high-speed networking device based on industrial Ethernet technology. From MTS Sensors, it handles the fastest rates of data transmission and communication in industrial settings. It is available for use with one to five position magnets depending on the particular network and can be set to document position, velocity, acceleration, as well as customer-specific “smart” functionality. Position resolution is accurate to 1µm, speed to 0.2 mm/s (0.008 inch/sec) and a sensor update time of 100µs (independent of stroke).
Every device on the network would have an IP address, letting users access that device through the Internet for remote operations. “Users simply go to an interface web page, like Internet Explorer,” says Dave Edeal, industrial marketing manager, MTS, Cary, N.C., “and a page opens for diagnostic or configuration information. They use this to query a sensor or device, and program or reconfigure on the fly. Also, they track when to replace a connected device, know whether it’s getting the message it expects, and so on. The device or sensor begins to look like a PC peripheral. Many designers want Ethernet for this level of communication because other standard buses tend to be more rigid.”
In addition to eliminating switch devices from an Ethernet design, newer I/O modules facilitate other configuration needs of Ethernet. For example, engineers no longer need to negotiation between 10Mbps and higher speed versions of Ethernet. Newer I/O devices will handle that task automatically through a feature known as auto-negotiation.
The SNAP-HD-G4F6 header cable, from Opto 22, lets users connect SNAP PAC programmable automation controllers and SNAP Ethernet I/O systems to plug-in, single-channel G4 digital I/O racks. The cable measures 6 ft (1.8m) in length and includes header connectors at each end.
Each layer of the model handles specific communication tasks.
Layer 1: The physical layer defines the electrical and mechanical connections to a network. A PC NIC card, for example, interfaces to this level, as does any device that will pass data through a physical medium.
Layer 2: The data link layer defines how devices access and share the physical medium and splits data into frames for transmission. It is usually split into two sub-layers: Logical Link Control (LLC) and Media Access Control (MAC). Switches often use this layer as they learn the MAC addresses of attached devices. Most versions of the Ethernet protocol do not go beyond this point.
Layer 3: The network layer defines how devices will establish, maintain and terminate network connections as well as route data to the appropriate portion of a network. The IP protocol, for example, resides here.
Layer 4: The transport layer works with the next layer up, ensuring the reliability and integrity of transmitted data. TCP usually resides here.
Layer 5: The session layer defines how two presentation entities exchange data. E-commerce applications extensively use this layer.
Layer 6: The presentation layer defines how application data are packed or unpacked and made ready for use by the next layer up, the application layer. Protocol conversions, encryption and decryption, as well as graphics expansion occur here.
Layer 7: The application layer works with end user and end application protocols, such as File Transfer Protocol (FTP) and various mail protocols.
A model network
The International Standards Organization (ISO) helped develop the seven-layer OSI model for network communications. The goal was to help users assemble “best-in-class” type systems using devices from different vendors. Even though TCP/IP and Ethernet dominate, the OSI model is generic and functions with all network types and media.
In brief, data are formed into a packet for transmission from device A to device B by an application at the top layer of the model, layer 7. The packet descends through each subsequent layer, acquiring various header and trailer data, as required by the protocols until it reaches Layer 1, where it is transmitted as a frame across the cable, wire, or fiber optic medium in use. When the frame reaches device B, each layer, beginning at layer 2, removes the header and trailer data as the packet ascends, delivering just the application data to the application.
Filed Under: Factory automation, Motion control • motor controls, Networks • connectivity • fieldbuses