Over the past decade, with the rapid development of internet technology, Ethernet has become the dominant network technology in commercial communication. Ethernet's communication speed is much higher than any current industrial fieldbus. As it is a standard network technology in the IT industry, thousands of companies participate in the development and production of related products, making it inexpensive and offering a wide range of choices. Therefore, people expect Ethernet to be applied to the industrial control field as well. With its low cost, extremely high communication speed, and globally widespread standard, it is expected to gradually replace the numerous bus systems in the existing industrial control industry, using Ethernet to achieve consistent communication from the management layer to the industrial field layer. The industrial control field and the IT industry have vastly different needs for network systems. To effectively apply Ethernet, it must be adapted to the specific requirements of the industrial environment. This article takes the real-time industrial Ethernet standard Ethernet Powerlink as an example to introduce the implementation scheme and practical application of industrial Ethernet. —Author's Preface I. Real-time Limitations of Standard Ethernet Currently, standard Ethernet can achieve transmission speeds of 100Mb/s or even 1000Mb/s, far faster than any fieldbus system. However, for industrial control, real-time performance is more important than transmission speed. A key characteristic of real-time performance is the determinism of time. Data transmission time during communication is not random but can be accurately predicted in advance. Although Ethernet has a high transmission rate, it cannot guarantee real-time communication between control devices. The communication mechanism of standard Ethernet IEEE 802.3 allows data transmission time to be arbitrarily delayed, thus negating the concept of real-time performance. However, in the field of industrial control, especially in the control of highly dynamic processes, real-time performance is indispensable. The reason for this communication time uncertainty in ordinary Ethernet is its access mechanism to the physical medium, CSMA/CD. CSMA/CD is the core of the Ethernet standard IEEE 802.3. If we want to fully utilize the advantages of Ethernet in the field of industrial control without changing the existing standard as much as possible, we must find a way to guarantee the determinism of data transmission time in Ethernet and enable real-time communication. 1. Requirements for Real-Time Performance in Industrial Control (1) Real-Time Performance In industrial control systems, real-time can be defined as the predictability of the system's response time to an event. After an event occurs, the system must react within a predictable time frame. How fast the reaction time needs to be depends on the controlled process. In chemical thermal process control, a reaction time of seconds is sufficient, while in high dynamic transmission control, the system reaction time must reach the microsecond level. In addition, the real-time performance in industrial control can be divided into two different categories: hard real-time and soft real-time (there is no clear boundary between them). Hard real-time: If the system response time requirement is not met, it will lead to fatal consequences (such as automotive ABS, aircraft, machine tools, etc.). Soft real-time: If the system response time requirement is not met, it will only affect the system control quality and will not cause serious consequences (such as building systems, elevators, warehouse management, etc.). (2) Jitter Jitter refers to the deviation in the completion or response time of the same process each time, that is, the time accuracy. The magnitude of jitter is very critical for some process controls such as motion control and some high-precision closed-loop controls. Take a shaftless printing machine as an example: Assume the printing speed is 25m/s, that is, every 40mm/µs. If the inter-axis communication jitter is greater than 40µs, there will be a deviation of more than 1mm, and the printing quality will definitely not meet the requirements, as shown in Figure 1. (3) The program in the communication cycle time control system runs in a periodic loop. All inputs are refreshed within one cycle, and the output is written after the calculation task is completed. The cycle time is determined by the controlled object. The cycle of high dynamic transmission control often needs to reach the millisecond level. After the system is networked, the network data exchange speed should correspond to the system calculation cycle time. In high-precision motion control of position control, electronic gears, and multi-axis linkage, the shorter the refresh time, the better. The shorter the time, the higher the control accuracy and the higher the dynamic performance that can be achieved. In multi-axis linkage, if the servo system performs position control with a cycle of 400µs, the information exchange between the axes should also be optimal with a cycle of 400µs to achieve the most accurate synchronization between axes. 2. Real-time level classification According to the different real-time requirements of different processes, the real-time performance can be divided into 4 levels (as shown in Figure 2). Among them, real-time level 4 is the most demanding in industrial control, mainly for the real-time requirements of mechanical transmission and motion control. Different fieldbus systems can be selected for these real-time requirements. If industrial Ethernet is to become the standard in the entire industrial control field, it must cover all these requirements for real-time performance and communication cycle, that is, it must meet the most stringent real-time requirements. II. Traditional methods to solve the real-time limitations of Ethernet Currently, there are several solutions to solve the problem of uncertain Ethernet data transmission time. They all have the following in common: none of them change the existing Ethernet communication mechanism, and the protocol is directly TCP/IP, which has many limitations. Representative methods include: (1) Low collision probability If there is not much data in the network, the collision probability will decrease. It increases exponentially with the increase of data communication. When the network load is less than or equal to 10%, it can be assumed that the collision can be avoided. The limitations of this method are: it cannot make full use of network bandwidth, wasting bandwidth; and it cannot guarantee 100% that the collision will not occur. (2) Using network switches to segment in the collision domain As shown in Figure 3, using network switches to segment is a completely different solution that can completely avoid the occurrence of collisions. Its principle is to separate the network domains that may have collisions using network switches. It is somewhat similar to a set of point-to-point connections. The limitations of this method are: data communication is affected by the delay caused by the allocation and buffering process of the network switch, and the transmission time characteristics are subject to the configuration of the network switch and will have some deviation. In high dynamic drive control, such deviation is not allowed; and setting up the network switch requires the staff to have a good understanding of network technology; in addition, its cost is relatively high. (3) IEEE 1588 time synchronization mechanism This method can better overcome the lack of real-time performance of Ethernet. The main principle is to synchronize the time of all stations in the network. A synchronization signal is used to periodically correct and synchronize the clocks of all stations in the network. Each frame of data sent by the station has its own time stamp, which tells the receiver the exact time when the task must be performed. Depending on the time accuracy requirement, a software clock or a hardware system clock can be used. If the accuracy requirement is very high, a hardware device must be added to the network to measure the time required for signal communication itself. Advantages of this method: it can achieve a high transmission time determinism, that is, real-time performance; it can directly use the TCP/IP protocol. The limitation is: all stations must have their own clocks, which is costly. At present, there is no hardware device to measure the time required for signal communication itself to meet higher accuracy requirements. The biggest problem is that this communication mechanism significantly impacts system programming. Since control tasks must be initiated via time-triggered methods, it increases programming difficulty and doesn't align with the programming habits of industrial control engineers. True Real-Time Ethernet – Ethernet Powerlink: Developed by B&R in Austria, the Ethernet Powerlink (EPL) standard is an industrial Ethernet that meets the most demanding real-time requirements (Level 4) and is already in practical application. B&R's initial approach to developing Ethernet Powerlink was to build a fieldbus system on top of standard Ethernet to meet the most demanding real-time requirements in control, while overcoming the limitations of the traditional solutions described above. Ethernet Powerlink's main technical specifications include: using standard Ethernet IEEE 802.3u (Fast Ethernet) as the transmission medium; a transmission rate of 100Mb/s; using standard hubs and standard wiring; a minimum real-time data transmission cycle of 200µs; jitter less than 1µs; simultaneous transmission of real-time and non-real-time data; simultaneous transmission of IP protocol; and use of standard Ethernet hardware. 1. Working Principle - Time Slot Communication Management Mechanism To avoid conflicts and make the most of bandwidth, EPL reorganizes the inter-station information exchange mechanism in the network in terms of time, and introduces a time slot management mechanism on the basis of CSMA. One of the stations in the network acts as the management station to manage network communication, gives synchronization beats to all other stations, and assigns publishing permissions to each station. Each station can only publish information after obtaining publishing permission. An EPL communication cycle can be divided into 4 stages (as shown in Figure 4): (1) Start stage: The administrator publishes the "Communication Cycle Start (SoC)" signal, which is broadcast to all stations. After this signal is issued, all stations synchronize. (2) Synchronization stage: In this stage, all stations exchange synchronization information. The management station sends a PRq frame to a station in a predefined order, requesting the station to publish information. After obtaining publishing permission, the station sends a PRs response frame in the form of a broadcast. All stations can receive this frame of information, including those stations that should receive this frame of information. The direct horizontal communication method between stations is very similar to that of the CAN bus. (3) Asynchronous phase: This phase is reserved for information that does not require real-time updates. The management station sends an "invitation" frame to a station, which can then publish asynchronous information, such as a frame of IP information. (4) Idle phase: The waiting time before the next cycle. 2. Topology: Any network topology can be achieved by using Hubs. Since only one station in the network can send information at the same time, there will be no conflict, so the number of Hubs used is unlimited. The system can use dual-port Hubs to achieve a single-line serial topology. Network switches are not recommended for EPL networks because they can cause higher system latency and jitter. External standard Ethernet (such as LAN) can directly access the EPL network through a gateway using the IP protocol. The application layer interface of Ethernet Powerlink V2 is based on the mechanism defined in the DS301 communication specification of CANOpen (defined by the CAN in Automotive organization). In this way, EPL can directly use a large number of device component features already defined in CANopen, realizing the consistency of CANopen and EPL network communication, and also simplifying the software layer transition from CANopen to EthernetPowerlink. 3. Ethernet Powerlink and TCP/IP Compatibility A key purpose of using Ethernet in industrial control is to achieve communication across the management layer, field layer, and even sensor and actuator layers. There are no location or system-related boundaries between layers; communication uses the IP protocol. Ethernet Powerlink achieves this through two mechanisms: simultaneous transmission of real-time and non-real-time (asynchronous) information; and transparent sending and receiving of IP protocol information in asynchronous time slots (as shown in Figure 5). The Ethernet Powerlink API is fully compatible with standard Ethernet driver functions. At the application layer, IP-based protocols or software can be used directly without modification. When real-time communication is not critical, such as during program downloads, system programming diagnostics, or parameter configuration, Ethernet Powerlink's open mode allows all stations to be used as ordinary Ethernet stations (non-real-time mode). In this mode, EPL stations are completely transparent to ordinary Ethernet. 4. Practical Applications and Standardization Process of Ethernet Powerlink Since its practical application three years ago, Ethernet Powerlink has received positive user feedback, with over 20,000 nodes deployed worldwide, rapidly expanding its market share. This demonstrates that Ethernet can meet the most demanding real-time requirements of control systems. From three-axis synchronous fully electronic injection molding machines (communication cycle 400µs, jitter <1µs) to 25-axis shaftless printing machines (communication cycle 800µs, jitter <1µs), and even large packaging machinery with 50 axes and 50 I/O stations (communication cycle 2.4ms, jitter <1ms), its performance has been tested in the field. Of course, besides technological advantages, the widespread adoption of a new communication standard is the most crucial factor in whether it can be accepted by a broad user base. To accelerate the adoption of the EPL standard, B&R released the technology immediately after developing it and entrusted a neutral institution (ETH Zurich) to manage the standard. Subsequently, several well-known European industrial control companies such as Lenze, Kuka Roboter, and Hirschmann joined the effort, establishing the Ethernet Powerlink Standardization Group (EPSG) to jointly develop related hardware and software products, enabling Ethernet Powerlink to be widely adopted in the industrial control industry as quickly as possible. To date, over 30 well-known industrial control companies worldwide, including end-users, have joined the EPSG (Electronic Power Link Group) to participate in the improvement of the Ethernet PowerLink standard and simultaneously develop and produce related hardware and software products. This makes the standard independent of any single company, a global, open industrial Ethernet standard. At the same time, EPSG is discussing with the IEEE standardization organization the possibility of making Ethernet PowerLink part of the IEEE's next-generation industrial Ethernet standard, thus making it a truly international standard.