Industrial networks serve as the interconnections between sensors, actuators, and the computers that monitor and control them. As the complexity of the automation rises within plants and factories, these connections are graduating from the more simplistic (e.g., several RS-232 serial data links) to more formal networks populated with greater numbers of sensors and controlled devices. Specifically, many of these proprietary networks installed by equipment manufacturers are giving way to more modern open networking technology, namely Ethernet and wireless.
Ethernet is the de facto standard for data-communications networks in almost all organizations. Because of its ubiquity, product costs are low and it’s relatively easy to build a network with off-the-shelf components.
However, there is one caveat: Ethernet was created mainly for an office environment. All of the equipment is usually indoors in a clean, climate-controlled atmosphere that’s generally free from electrical noise. A typical industrial environment is the exact opposite.
So can Ethernet really be a contender for the industrial world? The answer is yes, but with reservations. To be successful, an Ethernet local-area network (LAN) must handle and endure:
• Harsh environments: Harsh may not be a harsh enough word to describe industrial network environments. Temperatures can range from –40°F to above 140°F. Surroundings may include dust and smoke, oil and other lubricants, hazardous gases and chemicals, or other kinds of evil substances. Some of the equipment may be outdoors, such as on a pipeline or drilling rig. And don’t forget factors like vibration.
• Electrical noise: The industrial setting is also a dirty electrical environment. High-voltage ac power lines, motors and variable frequency drives, relays, solenoids, pumps, reactive loads, and all manner of devices turn off and on, generating spikes and interference. There’s also the ever increasing use of wireless, creating an ugly RF environment.
• Critical applications: The basic function of an industrial factory, plant, or operation is usually to manufacture some product, whether it’s automobiles, gasoline, or cookies, on a grand scale. Failure of the process due to machine or network downtime increases overall costs and can back up one process at the expense of another. So, it must be fixed fast. Otherwise, lost productivity can reach into the hundreds of thousands if not millions of dollars.
• Security: Office LANs certainly need security, but industrial networks also must be hack-proof and safe. Encryption, authentication, and other methods are needed to protect the equipment, the personnel, and the facility itself from unauthorized access, accidentally by employees or on purpose by the bad guys.
• Adaptability: An industrial network must accommodate both old and new equipment and processes. Most factories and plants are a mix of older legacy equipment and new hardware. Interfaces, protocols, and other sub-networks have to communicate and interoperate.
• Outside links: Any industrial network should be able to seamlessly link to a company office LAN, the Internet, or other network. The office LAN should never be used for industrial purposes, but invariably the industrial net needs the ability to talk to the office LAN via a bridge. Remember, different personnel will be installing and maintaining the office LAN (an IT person) and the industrial network (engineer, tech, or electrician).
Given all of these conditions, Ethernet can and has been used successfully with hardened equipment, environmental protection, ruggedized cabling, and other precautions.
ADAPTING ETHERNET TO INDUSTRY
A steady diet of updates and enhancements to the Ethernet IEEE 802.3 standard now make it virtually trouble-free to set up and use. Ethernet is a layer 1 and layer 2 network that defines only the physical layer (PHY) and the media access control (MAC). The most widely used PHY is unshielded twisted pair, such as CAT5e, which is also used in some industrial installations. However, shielded twisted pair is more common in industrial settings because the shielding provides extra physical protection as well as improved noise immunity. Fiber is also popular as the ultimate solution to the noise problem.
Most industrial data communications occur at rates from a few kilobits per second to about 115.2 kbits/s (typical rates are 9.6 and 19.2 kbits/s). Thus, the basic Ethernet speeds of 10-Mbit/s 10BaseT and 100-Mbit/s 100BaseTX may be overkill, but are nonetheless still viable.
Connectors tend to be one of the biggest problems with industrial Ethernet. Although used, common RJ-45 connectors generally aren’t acceptable since they get dirty and can corrode in some environments. Special rugged and sealed RJ-45 replacements have been developed (IP67), though, to keep out dust, dirt, and chemicals, as well as provide extra strength and strain relief.
Another major issue involves the length of cables. Ethernet specifications state a maximum range of 100 m with twisted pair. While that’s more than sufficient in typical office installations, it’s not enough for many large industrial applications. Because cables often are strung overhead, cable runs must factor in the vertical rises and horizontal runs. Products such as repeaters help extend the range of common cables. Fiber should be considered too, since it can extend the range to thousands of feet and even miles if needed. Standard Ethernet fiber options are available, as are copper-to-fiber converters.
It’s important to consider interconnection equipment for extending the network. Hubs are okay, but the more preferable option is the Ethernet switch (Fig. 1). Ethernet switches isolate different segments of the network from one another and generally speed up operations. In many cases, switches solve the latency problem inherent in Ethernet.
Latency is the delay between the initial signal and the time it’s received at the desired load. With long cables, latency is a problem in some situations. Latency also derives from the basic access method used in Ethernet—carrier sense multiple access with collision detection (CSMA/CD).
All nodes on an Ethernet bus share a common medium. Only one signal can flow at a time, so nodes must contend for access. First, nodes listen on the bus for activity. If no one else is transmitting, the bus sends a signal. If two nodes try to transmit simultaneously, a collision occurs. Both nodes back off, wait a short random period, and then try again. This process adds extra latency depending on how much traffic is on the networks, potentially causing problems in the process.
Many processes must operate in real time, meaning that instantaneous response or very low latency is mandatory. Such deterministic needs aren’t always met by Ethernet, but switches can often solve this problem. In some cases, special industrial protocols may be required to solve the latency problem.
When timing and synchronization is an issue, employ the IEEE 1588 standard called the Precision Time Protocol (PTP). This software option runs on any Ethernet system and provides time stamping, synchronization, and other timing functions that can mitigate problems in selected control applications.
One particularly beneficial feature of Ethernet is how well it works with TCP/IP. These are the main Internet protocols, but they’re also used in industry along with special protocols called fieldbuses (see “Fieldbuses Fit Into Industrial Networks”).
THE OSI MODEL
The Open Systems Interconnection (OSI) model (Fig. 2) is the International Standards Organization’s way of formally describing how data communications products and systems should be built and how they should work. There are seven formal layers, but most systems only use a few of them.
The physical layer 1 describes the signals, voltages, and interconnection medium. The data link layer 2 frames the data and defines the source and destination. The network layer 3 determines the data path in a network (e.g., routing). The transport layer 4 supplies a message structure and error detection. The session layer 5 provides opening and closing of connections and provides security. The presentation layer 6 delivers data conversion if needed. The application layer 7 is the actual program to be used.
Ethernet uses only the first two layers, the (PHY) and the data link or MAC layer. Most industrial networks that employ fieldbuses only use layers 1, 2, and 7. Industrial Ethernet uses layers 1 through 4. Layer 3 is provided by the Internet Protocol (IP) and layer 4 is provided by Transport Control Protocol (TCP) or User Datagram Protocol (UDP). The specific fieldbus is wrapped in TCP/IP for transmission.
Industrial Ethernet is, of course, standard Ethernet adapted to handle one or more of the popular fieldbuses. The PHY is standard but the MAC layer is generally modified to match up with the fieldbus as well as improve on the real-time performance of Ethernet. Furthermore, TCP/IP or UDP/IP protocols carry the fieldbus protocol at the network and transport layers.
Ethernet protocols are typically rated by their real-time performance. For example, class A is non-real-time (NRT) where the response time can be greater than 100 ms. Class B is soft-real-time (SRT) where the response time is in the 10- to 100-ms range. Class C is instantaneous real-time (IRT) with a response time of less than 1 ms for very critical applications. Some of the more widely used Ethernet protocols and their related fieldbuses include:
• Modbus TCP: Modbus packetsare encapsulated in TCP/IP and run on Ethernet.
• EtherNet/IP: This standard was developed and supported by Allen Bradley/Rockwell and puts DeviceNet/ControlNet on Ethernet. It uses the Control and Information Protocol (CIP) that’s mapped to TCP/IP or UDP.
• Ethernet Power Link: This fieldbus makes it possible to run DeviceNet protocol on Ethernet.
• EtherCAT: With EtherCAT, CAN and CANopen can be put on Ethernet.
• SERCOS: This fieldbus puts SERCOS I and II on Ethernet.
• Profinet: ProfiNet is an extension of the Profibus that uses TCP/IP so it can work on Ethernet. It’s available in class A, B, and C versions, depending on the application.
Designing industrial Ethernet is a real challenge. In the past, industrial Ethernet equipment has been implemented with ASICs or ASSPs, or with programming MCUs. While that solves the initial product design problems, it doesn’t protect the product from obsolescence, which can easily occur if a standard is changed, updated, or superseded. The goal of most industrial equipment manufacturers is to get as much as 15 years of life out of a product. With the pace of change today, that’s become more difficult than ever.
To that end, Jason Chiang of Altera has come up with another design solution. It integrates and updates processor functions with multiple industrial Ethernet protocols (Fig. 3a) and other custom logic on a reconfigurable FPGA. Thanks to the FPGA, you can manage the product platforms with different buses in multiple product lines with minimal engineering support. The FPGA gives the product the flexibility to adapt to changing and new standards in a timely manner. Figure 3b shows the Ethernet structure on the FPGAs.
At one time, wireless was strictly taboo in industry. It was thought to be too unreliable because of the many possibilities of link breakdown and potential problems regarding security. Today, wireless is broadly accepted across many industrial applications. It can still suffer from the aforementioned issues, but modern protocols and equipment have turned wireless into a far better alternative than wiring.
Wireless offers several stellar benefits. First, there’s no wiring. Cable, including installation and maintenance, is expensive and the wire and connectors are vulnerable to damage. Many installations require cables to be run in conduit for safety. However, a wireless link to a sensor or actuator can save thousands of dollars, and new equipment and links can be added more quickly.
One sticking point with wireless concerns its reliance on battery power for remote sensors or controllers. These devices may be hard to access. And, of course, short battery life means regular battery replacement to ensure uptime. Recently developed low-power wireless solutions help mitigate this problem, but it is still a major factor to consider when evaluating a wireless solution.
Many wireless technologies can be used to implement industrial monitoring and control, though only a few have seen large-scale adoption. Among these is wireless Ethernet or the IEEE 802.11 standard commonly known as Wi-Fi, IEEE 802.15.4/ZigBee, and proprietary methods using the popular industrial, scientific, and medical (ISM) frequencies (902 to 928 MHz, 2.4 GHz). Cellular technologies like GSM/GPRS/EDGE or CDMA are also finding their niche in longer-range applications.
Due to its complexity and power consumption, Wi-Fi isn’t the first choice for sensor and actuator links. It’s simply overkill for all but a few industrial applications. However, Wi-Fi does make a good connection to an office LAN via a bridge or gateway, and it can be a good option for some longer-range needs and those apps requiring high data speeds.
Perhaps the most accepted wireless technology in industry is based on the IEEE 802.15.4 standard. It uses the unlicensed ISM bands—868 MHz in Europe, 902 to 928 MHz in the U.S., and 2.4 GHz worldwide. The most popular of these, the 2.4-GHz version, achieves 250-kbit/s data rates. It features very low power consumption plus built-in security.
Several variations (enhancements) to the 802.15.4 standard help boost performance and/or reliability. For instance, the ZigBee standard builds a stack on top of the 802.15.4 stack to perform mesh networking or customize the technology to the application. The ZigBee Alliance offers multiple “profiles” for different applications in industry.
The Instrumentation Society for Automation (ISA), an industrial professional association and standards organization, is completing work on a variation of the standard called ISA100a. It employs a frequency-hopping technique to minimize interference in the crowded 2.4-GHz band and offset the effects of multipath fading. Nodes transmit successive packets on one of 16 different frequencies in the 2.4- to 2.485-GHz band using a pseudorandom hopping pattern. Another wireless technology called Wireless HART exploits a similar frequency-hopping scheme to carry the popular HART fieldbus protocol.
Sevral proprietary wireless products are designed specifically with industrial applications in mind. They opt for either 2.4 GHz or 900 MHz, depending on the desired range. The 1-W, 900-MHz devices feature a range from about 1500 feet to several miles (factoring in the environment), at a maximum 115-kbit/s data rate. The 63-mW, 2.4-GHz units will typically range up to 300 feet (again depending on the territory), at data rates up to 250 kbits/s. The Zlinx units from B&B Electronics are widely used to implement wireless Modbus networks (Fig. 4).
For more details on wireless, see “Cut The Links To Your Sensor/Actuator Networks.”