Finding a Faster Fieldbus

Summary

Ethernet-based fieldbuses have their challenges, but speed, flexibility, and simplicity make EtherCAT a robust contender in industrial automation environments. The EtherCAT (Ethernet for control automation technology) fieldbus system was invented by Beckhoff Automation to apply Ethernet to automation applications requiring short data update times (cycle times) of less than 100 microseconds with low communication jitter of less than 1 microsecond. The protocol is standardized in IEC 61158 and is suitable for both hard and soft real-time computing requirements in automation technology.

Why fieldbuses are important

While the controller is obviously important to the controlled architecture, the bus system is the core. “The bus system defines system performance much more than the controller,” said Martin Rostan, executive director of EtherCAT Technology Group. “It’s the bus system that defines the choice of suppliers, especially if [users] pick a bus system that is supported on the controller side by only one supplier. Therefore, it’s also the bus system that defines the cost. It’s a bus system that defines if [users] have the choice, if [they] can go with centralized controls—closed, fast control loops—over the bus system or not.”

People used to think decentralized controls was a solution for difficult situations because neither their “centralized” controllers nor their bus systems had the performance to close fast-acting loops on a centralized CPU. They had to live with many programming languages and/or controllers to close those fast control loops in a decentralized way. However, with EtherCAT, users can do both. “Users can do either centralized or decentralized,” Rostan said.

“The control- and the bus-cycle times have a big impact on the reaction time,” Rostan explained. “It has an even bigger impact, not on the control cycle time, but whether you can make use of the control cycle time. We see people who use PC-based controls, which can execute PLC [programmable logic controller] code quickly and do cycle times in the sub-millisecond range, combine that with a legacy bus technology, or with one of our competitor’s bus technologies, which is more in the several millisecond range. Then, they’re disappointed by the performance of the overall architecture.”

Tackling the challenges of Ethernet fieldbuses

Three main challenges associated with Ethernet and fieldbus control performance are bandwidth utilization, stack delay issues, and switch delays.

Bandwidth utilization. The Ethernet frame is a minimum of 84 bytes (Figure 1). There is no smaller Ethernet frame than that. The interpacket gap is considered since it must be considered in a bandwidthcount situation. “If such a large container is used to transport a few bits or bytes, there will be a poor application data ratio,” Rostan said. “With four bytes of process data, for example, in such a frame, there would be less than 5% application data ratio—the rest is just air. It’s not the most efficient usage of bus bandwidth.”

Stack delays. Most people neglect stack delays based on the assumption that by the time the frame has arrived in the node, it’s available. “On the contrary, it must go through the local stack, with 1 MB of code or more if EtherNet/IP or Profinet is used,” Rostan said. “This takes time because these industrial Ethernet protocol stacks are big and time consuming to get through all those intermediate layers and get through to the application. In a simple case, the application is the output, or it’s the control algorithm of your device.”

Measurements show that EtherNet/IP takes up to 3 milliseconds to get through the protocol stack. According to Rostan, EtherCAT is lean in contrast. The EtherCAT stack is about 70 kB of code. “With digital input/ output (I/O) and EtherCAT, the stack delay time is zero because the EtherCAT chips have I/O onboard. In the hardware, there’s no software interaction whatsoever.”

High performance Ethernet on the fly

Rostan claims that EtherCAT is the fastest industrial Ethernet technology. Speed claims include:

• 1,000 distributed digital I/O in 30 microseconds
• 100 servo axis updates every 100 microseconds
• EtherCAT goes directly to the I/O slice, no sub-bus
• Optimal usage of the standard Ethernet port in the controls with no extra hardware

This performance is attributed to the EtherCAT functional principle: Ethernet on the fly (Figure 2). Characteristics include:

• Efficient: typically, only one Ethernet frame per cycle
• Ideal bandwidth utilization for maximum performance
• Resolves the three main performance problems (bandwidth utilization, stack delay issues and switch delays)
• Operates in real time and can reach down to the I/O level
• No underlying subsystems, thus no delays in bus coupler gateways
• Connects everything in one system: I/O, sensors, actuators, drives and displays

Rostan explained Ethernet on the fly further. “Instead of sending one frame to each node in each cycle in each direction, EtherCAT sends one frame through the node, and each device extracts data from that frame and inserts its data into the very same frame. In doing so, all sub-devices share one frame, and the bandwidth is even doubled—in ideal cases— because EtherCAT uses the same frame for input and for output data, so we get 200 megabits up front if we have a symmetrical process image.”

• Read, write or read-write
• Access to a specific sub-device through direct addressing, or access to multiple sub-devices through logical addressing (implicit addressing)

Logical addressing is used for the cyclical exchange of process data. Each datagram addresses a specific part of the process image in the EtherCAT segment, for which 4 GB of address space is available. During network startup, each sub-device is assigned one or more addresses in this global address space. If multiple sub-devices are assigned addresses in the same area, they all can be addressed with a single datagram.

Since the datagrams completely contain all the data access-related information, the master device can decide when and which data to access. For example, the master device can use short cycle times to refresh data on the drives, while using a longer cycle time to sample the I/O. A topology-related fixed process data structure is not necessary. This also relieves the master device in comparison to conventional fieldbus systems in which the data from each node had to be read individually, sorted with the help of the process controller and copied into memory.

With EtherCAT, the master device only needs to fill a single EtherCAT frame with new output data and send the frame via automatic direct memory access (DMA) to the MAC controller. When a frame with new input data is received via the MAC controller, the master device can copy the frame again via DMA into the computer’s memory—all without the CPU having to actively copy any data. In addition to cyclical data, further datagrams can be used for asynchronous or event-driven communication.

Flexible topology. EtherCAT supports all topologies—line, tree, star, or any combination of these (Figure 4). EtherCAT makes a pure bus or line topology with hundreds of nodes possible without the limitations that normally arise from cascading switches or hubs.

Final thoughts

Up to 65,535 nodes can be connected to one EtherCAT segment, so network expansion is virtually unlimited. Because of the practically unlimited number of nodes, modular devices such as “sliced” I/O stations can be designed in such a way that each module is an EtherCAT node of its own. Therefore, the local extension bus is eliminated. The performance of EtherCAT reaches each module directly and without any delays since there is no gateway in the bus coupler or head station.

_{This feature originally appeared in AUTOMATION 2023 Volume 2: Connectivity & Cybersecurity.}