Abstract: First, it is explained that there is no unified standard for real-time performance evaluation of industrial Ethernet. By comparing the real-time requirements and characteristics of industrial Ethernet used in process control and motion control, the differences and similarities of real-time performance evaluation of different industrial Ethernet are specifically reflected. Keywords: Industrial Ethernet; real-time communication; evaluation 1 There is no unified standard for real-time performance evaluation of industrial Ethernet According to reference [1], the real-time performance of a certain message is satisfied when its message response time is less than the specified time limit. The real-time performance of a certain node is satisfactory when all messages sent by the node can be responded to within the specified time limit. The real-time performance of the entire control network is satisfactory when the real-time performance of each message of each node distributed on the network is guaranteed. The real-time performance of the entire network must meet the following three time constraints: 1) There must be an upper limit to the time when each node obtains the right to communicate. If this value is exceeded, the right to communicate should be released immediately regardless of whether the current communication task is completed. This time constraint can prevent a certain node from occupying the bus for a long time, which would lead to the deterioration of the real-time performance of other nodes. 2) It should be ensured that every node on the network has the opportunity to obtain communication rights within a fixed time period to prevent individual nodes from experiencing poor or even lost real-time performance due to prolonged lack of communication rights. If even one node experiences this, the real-time performance of the entire network cannot be guaranteed. The length of this fixed time period is a metric for controlling the real-time performance of the network. 3) For urgent tasks, when their real-time requirements temporarily become very high, they should be given priority. Nodes with high real-time requirements should also have a greater chance of obtaining communication rights than other nodes. Therefore, assigning higher priority to certain nodes using a static (fixed) method and to certain communication tasks using a dynamic (temporary) method will ensure that the real-time performance of urgent tasks and important nodes is met. If we simply use these three time constraints to evaluate the access control methods commonly used in current control networks, we will find that some access control methods do not meet any of the constraints, such as the CSMA/CD method of Lonwork fieldbus. According to this principle, it seems that only token protocols can meet the actual requirements. However, in practical system applications, other access control methods are also used in systems with extremely high real-time requirements. For example, the Swiftnet protocol developed by Boeing is widely used in the aviation and aerospace fields, and the CAN bus protocol based on CSMA/CA is the standard protocol specification for the internal control circuits of high-end cars. There are many reasons for this established fact. On the one hand, with the advancement of network technology, various protocols are striving to improve their performance and adopt various methods to enhance real-time performance. On the other hand, real-time requirements are met through overall optimization by adjusting network configuration and load. Therefore, there is currently no specific universal standard for judging the real-time service of industrial Ethernet; real-time performance evaluation needs to be conducted based on actual applications and emerging new technologies. 2. Differences and Similarities in Real-Time Performance Evaluation of Different Industrial Ethernet Systems Industrial Ethernet can be divided into two main categories: industrial Ethernet for process control, such as HSE, and industrial Ethernet for discrete control, such as EPL. The following comparison of the real-time requirements and characteristics of industrial Ethernet for process control and motion control specifically reflects the similarities and differences in real-time performance evaluation between the two. Commonalities: The network response time has a unified model, and network latency is an important component. The real-time performance of the entire industrial Ethernet system is reflected by the network response time. Factors affecting network response time mainly come from three parts: the local system (source node processing), the industrial Ethernet network (transmission), and the destination node system (destination node processing). Figure 1 is a schematic diagram illustrating the time taken to send information from the source node to the destination node, which is the network response time T_delay. The total time delay can be divided into the following parts: the source node's time delay, the network channel's time delay, and the destination node's time delay. The source node's time delay includes preprocessing time T_pre, which is the sum of computation time T_scomp and encoding time T_scode; and a portion of the waiting time T_wait, which is the node queue time T_n-queue, depending on the total amount of data the source node needs to transmit and the network's transmission status. Network time delay includes: transmission time T_tx, which is the sum of frame transmission time T_frame and network physical propagation delay T_prop, depending on the size of the information, data transmission rate, and network cable length; and another part of waiting time T_wait, network blocking time T_block; the destination node time delay T_post is the post-processing time of the data, which is the sum of the destination node decoding time T_dcode and the destination node computation time T_dcomp. [align=center] Figure 1 Schematic diagram of industrial Ethernet response time[/align] Therefore, the total time delay can be expressed as: Figure 1 clearly shows the location of the industrial Ethernet network transmission part and its time delay (represented by T_ethernet): The difference between the two: The real-time type of industrial Ethernet used for process control is shown in the left diagram of Figure 2. Its real-time requirements can be called deterministic communication requirements. The time required to transmit the data message and the generation of response data are both time-deterministic. In industrial control systems, the time determinism of communication networks means that data transmitted through the network must be transmitted from the source to the destination within a predetermined time. That is, the data message arrives at the receiver within the deadline. [align=center]Figure 2 Different Real-Time Requirements of Industrial Ethernet in Different Control Fields[/align] The real-time type of industrial Ethernet used for motion control and precision manufacturing is shown on the right side of Figure 2. Delay and synchronization are the decisive factors for this type of real-time data communication. This type of real-time data communication has a precise and predictable timing, that is, when the data message arrives at the receiver, the generation of response data, and the time required to transmit the data message are all predetermined. Synchronization determines the accuracy of timing events that a distributed system can identify, usually referring to all components performing a recurring action simultaneously. Synchronization deviation can be constant or variable; the latter is generally called jitter. Constant deviation is not critical and is easily compensated, but jitter cannot be compensated. The magnitude of this quantity is critical for some controls, such as motion control or some high-precision closed-loop controls. Taking a shaftless printing press as an example: Assuming a printing speed of 25 m/s, this means printing 1 mm every 40 µs. If there is a jitter greater than 40 µs in the inter-axis communication, there will be a deviation of more than 1 mm, and the printing quality will certainly not meet the requirements. In high-precision motion control such as 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 400 µs, the information exchange between the axes should also ideally be at a cycle of 400 µs to achieve the most accurate possible synchronization between the axes. However, in relatively slow processes (such as thermalization in chemical engineering), refreshing the communication data every 400 µs is unnecessary. For process control systems with strong real-time requirements, the response time of real-time communication should be within the range of 5–10 ms. Therefore, the network time delay from one node to another is usually required to be less than 2–4 ms. For general real-time control requirements, the network time delay can be larger. 3 Conclusion From the above analysis, it can be seen that the real-time requirements of industrial Ethernet for discrete control, especially motion control, precision manufacturing and other fields are demanding. In addition to delay control, jitter control is also the main difficulty. At present, components based on commercially available (COTS) technology cannot meet the demanding jitter accuracy requirements of motion control, while industrial Ethernet for process control does not need to control jitter. The real-time requirements are not as demanding as those for industrial Ethernet for discrete control. At the same time, with the continuous development of commercial network hardware and software technology, it is possible to consider using the new COTS technology for application research. References [1] Zheng Wenbo. Control Network Technology. Beijing: Tsinghua University Press, Sprün Green Press, 2001. [2] Wang Fujun, Wang Chunping, Liu Huai. Analysis and Optimization of Information in Real-Time Control Network. Microcomputer Information, 2005, 21(1):37-38