High Performance Industrial Networks
Demanding applications require high performance network protocols that are synchronized, with millisecond updates and jitter at a microsecond or less. Topology and configuration ease also are important, respondents to a Control Engineering survey said.
Mark T. Hoske
Control Engineering polled subscribers of its System Integration and Machine Control emailed newsletters to collect information on use of high performance industrial networks, those synchronized (time stamped) with updates in the millisecond range with microsecond jitter or better. While many communications networks serve the industrial space, few protocols effectively handle high-performance discrete applications, such as CNC, robotics, high-end packaging, and other high-speed motion control applications.
Of those responding, 65% use a high-performance network and of the 35% who do not, 53% plan to purchase one in the next 12 months. About 70 respondents currently use or plan to use a high speed network. They were asked to continue with the survey.
Of those who use a high-performance network, EtherCAT was the most popular, followed by “other,” including proprietary networks), SERCOS, Ethernet Powerlink, and Profinet IRT. Combining responses for the SERCOS and SERCOS III protocols edges that family of networks into the most-used category. Range of responses was diverse; other and proprietary ranked high. (See graphic.)
Investments in high-speed networks, and the related infrastructure, wire and cable, I/O devices, and compliant sensors, logic, and actuators accounted for approximately $1.9 million in spending, among 40 respondents to this question, more than $47,000 each (using the midpoint of each range). Most respondents also are spending on other types of industrial networks; for nearly 30% of respondents, this spending accounted for 75% of related budgets.
As for applications, nearly 45% of respondents use high-performance networks in material handling equipment (conveyors, transfer lines, elevators, and cranes). Robotics, computer numerical controls, and packaging machine applications also widely use such networks.
Why? Speed, topology, determinism were among top reasons the high-performance network protocol was chosen. Not surprisingly, performance issues accounted for 3 of the top 4 reasons and 5 of the top 10. Throughput, ease of configuration, cost of implementation, limited jitter, and availability of masters and/or slave devices (I/O, drives) were other attributes chosen by more than 25% of respondents.
Greatest influences on network protocol selection included internal staff, 33%, and the supplier of related hardware and software, 23%, followed by customers at 12%.
Survey respondents offered write-in advice about high-performance network protocols:
- Control system PLCs on which machines are running need to support high performance networks, and that’s the only way we are going to plan to implement new networks.
- International standards provide open, non-proprietary protocol.
- Look for speed, robustness, low hardware costs, ruggedness, and reliability, and ensure it’s well supported by vendors.
- Lots of marketing fluff clouds the picture on protocol choices. Do your homework; find the right one for your needs.
- Move away from proprietary and other interfaces that make it difficult to integrate with MES and ERP systems
- Synchronize communication among devices and schedule/synchronize with the control logic execution to avoid asynchronous elements in the loop. For process control we require intrinsic safety, which Foundation fieldbus provides.
- Out of the 24 years I have been applying and using industrial networks, EtherCAT is by far the highest performance and most cost effective network, (realtime or otherwise) that I have ever had the pleasure to apply.
- We use Rockwell Automation to supply all automation to ease training for our manufacturing engineers worldwide. Perhaps not best in every segment, support is from one supplier.
- Many networks create unexplained problems with noise (electromagnetic interference, EMI).
Joey Stubbs, PE, PMP, EtherCAT Technology Group: Ideally, the “high performance network” will be one network that represents a complete solution for I/O, drives, encoders, and data acquisition functions. This is an alternative to the traditional, multiple network solution – one network for motion, one for I/O, and most likely another completely different system for data acquisition. EtherCAT, the next generation Industrial Ethernet system, combines the best attributes of all of these dedicated networks and more into one all-purpose, high performance network. Three of the key ingredients to the success of EtherCAT are distributed clocks, time stamping and oversampling:
1) Distributed clocks deliver 64 bit clock synchronization of the EtherCAT network, without need for a special master card with a synchronization pulse.
2) Time stamping: The message receives time stamp information based on when a transition occurred in the field, without extra hardware or software on the slave device. The EtherCAT Slave Controller (ESC) has latching functionality in combination with its distributed clock (DC) to register and report back to the master when an event occurred in the field down to a 10 nanosecond (ns) resolution. For an output, the command to transition the output can be made with the same 10 ns resolution. This facilitates registering events that would happen between the cycle scans of traditional PLCs without the need for any additional processing requirements in the slave device.
3) Oversampling: Low-cost ESC supports subordinated interrupts as a function of the distributed clocks, allowing oversampling of input signals by hundreds or thousands of times within the master cycle scan. www.ethercat.org
Ronald Larsen, managing director, SERCOS North America: SERCOS I and II were the first highspeed real-time buses, with a rich command set, jitter less than 1 microsecond, mechanisms to ensure determinism and fiber optic transmission. Over 2.5 million nodes have been shipped. SERCOS III uses industrial Ethernet as a transmission medium, retaining the low jitter, high determinism and rich command set of SERCOS I and II.
SERCOS III operates at 100 MB/s using software loaded onto a standard FPGA or multi-protocol controller with master-slave communications over a ring or line topology and up to 511 devices per segment. No switches or hubs are required. Rather than using one Ethernet frame per command/ data telegram like some competing buses, SERCOS III efficiently packs multiple real-time telegrams into a single Ethernet frame, maximizing bandwidth usage for higher data throughput. SERCOS III includes a non-realtime channel for standard Ethernet devices to communicate with SERCOS masters and slaves for set-up, diagnostics and production reporting.
Unlike some competitive buses, SERCOS III allows this communication any time, not just during real-time operation. SERCOS III provides controller-to-controller cross communications. In addition, slave devices can communicate directly without master intervention, even over multiple rings. This system also facilitates bumpless recovery in 25 microseconds or less in case of a cable break or device failure. allowing real-time operation to continue. A SERCOS III product guide is available. www.sercos.com
Ethernet Powerlink Standardization Group: Ethernet Powerlink integrates features and abilities from Ethernet, CANopen, and a stack developed for real-time data communication. Powerlink keeps very close to the Ethernet standard, unlike some competing protocols. By retaining original Ethernet features, it reduces the cost of industrial deployment.
CANopen, a robust and proven protocol widely used throughout automation, simplifies setting up networks because of its wide use. The Ethernet Powerlink Standardization Group (EPSG) developed the Powerlink stack, which adds real-time capabilities to the protocol. www.ethernet-powerlink.org
Open PLC Network
Japan Electrical Manufacturers’ Association (JEMA): FL-net (OPCN-2) comprises the standard specifications of the controller level FA network developed by the Manufacturing Science and Technology Center, which realizes a communication network among programmable controllers, numerical control (NC), and robot controllers of multi-vendors.
Carl Henning, deputy director, PTO, Profibus and Profinet North America, says there are a number of factors to consider in selecting a high-precision network: speed, determinism, availability, support, and learning curve. All networks listed here are adequate for speed and determinism. Profinet has been used for motion control for over five years and is well-proven and therefore, widely available. Because Profinet is supported by a global organization, there are certified competence centers training centers, and test labs on every continent.
No other industrial Ethernet have the concept of competence centers, much less certify the centers. Because the high-performance aspects of Profinet are part of Profinet, the same network can be used for regular I/O, motion control, machine-to-machine integration, vertical integration, and functional safety. Even the high-performance section of the network is open to standard traffic like TCP/ IP. Wireless connections are possible and bumpless redundancy is supported.
CC-Link IE Field network
Chuck Lukasik, director, CC-Link Partner Association: CC-Link IE (Industrial Ethernet) Field network uses one gigabit-per second transmission and real-time protocol for control of remote I/O field devices with essentially no transmission delay. It delivers deterministic data exchange without the need for Ethernet switches. Determinism is guaranteed by using a token-passing technique. Cat5E cable and RJ45 connectors are the physical media. CC-Link IE Field offers both “Cyclic” and “Transient” methods of data exchange. Cyclic transmissions provide real-time, transparent data delivery services to all stations. A “common memory” model configured by a few simple parameter settings establishes deterministic data exchange for the network.
Specific data delivery timing can be calculated accurately before a network is commissioned. Transient messages are initiated “on-demand” from a specific station and they can be sent to one or more network stations and these stations can respond to transient requests for data. CC-Link IE Field allocates transient mode bandwidth so that cyclic communication remains deterministic; 100 MB Ethernet devices can be connected to this gigabit network via an Ethernet adapter. Seamless communication from field devices to controllers and among controllers provides an integrated control network.
Plug into Industrial Ethernet Protocols
CC-Lin IE extends 1 Gigabit industrial Ethernet to field devices, with copper
- Mark T. Hoske is Control Engineering content manager, www.controleng.com