Compared to civilian Ethernet switches, industrial Ethernet switches meet the needs of industrial sites in terms of design, component selection, strength, and applicability.
1. Topology
Topology refers to the arrangement of cables in a network. It is well known that EIA-485 or CAN uses a bus topology. However, in industrial Ethernet, due to the widespread use of hubs or switches, the topology is typically star or distributed star.
2. Wiring
Industrial Ethernet uses cables such as shielded twisted pair (STP), unshielded twisted pair (UTP), multimode or single-mode fiber optic cables. For speeds of 10Mbps, the requirements for twisted pair cables are not excessive, while for speeds of 100Mbps, Category 5 or Category 5e cables are recommended.
Fiber optic connections require a pair of fibers. Commonly used multimode fibers have wavelengths of 62.5/125μm or 50/125μm. Compared to the core of multimode fiber, the core of single-mode fiber is very thin, only about 10μm. Typically, multimode fiber is used for 10Mbps, while both single-mode and multimode fibers are suitable for 100Mbps.
3. Connectors and Connections
RJ-45 is the most common type of twisted-pair connector. It has two pairs of wires, one for transmitting and the other for receiving. In the definition of Media Dependent Interface (MDI), these four signals are identified as RD+, RD-, TD+, and TD-, respectively.
A communication link consists of a DTE (Data Terminal Equipment, such as a workstation) and a DCE (Data Communication Equipment, such as a repeater or switch). A hub port labeled MDI-X indicates that the DTE and DCE can be connected using a straight-through cable. If two DTEs or two DCEs are connected, a crossover cable method can be used, or the hub's uplink port can be used directly (without crossing the cables).
There are two types of fiber optic connectors: ST connectors are used for 10Mbps or 100Mbps; SC connectors are specifically for 100Mbps. Single-mode fibers typically use SC connectors. Connections between DTE and DCE can be made simply by following the TX and RX markings on the ports.
4. Industrial Ethernet vs. General Commercial Ethernet Products
What is Industrial Ethernet? Technically, it's compatible with IEEE 802.3, but its design and packaging are tailored to the requirements of both industrial and commercial applications. Designers in industrial settings want to utilize commercially available Ethernet chips and media while considering the specific requirements of industrial environments. The first considerations are high temperature, humidity, and vibration. The second is ease of installation within industrial control cabinets. The third is power requirements. Many control cabinets provide low-voltage AC or DC power. Wall-mounted power supplies are sometimes inadequate. Electromagnetic compatibility (EMC) requirements vary depending on the industrial environment's EMI (industrial immunity) and ESD (industrial vibration) requirements. Field safety standards differ significantly from office standards. Sometimes, harsh environment ratings are required. Factories may use industrial control cabinet standards, while building systems often use smoke standards. Clearly, low-cost commercial Ethernet hubs and switches cannot meet these requirements.
5. Speed and distance
When discussing distances in shared Ethernet networks, the concept of collision domains cannot be ignored.
Media access in shared Ethernet or half-duplex Ethernet is determined by Carrier Sense Multiple Access/Collision Detection (CSMA/CD). In half-duplex communication, sending and receiving cannot occur simultaneously, otherwise data collisions will happen. Before sending, a station first checks if there is an idle channel. During transmission, the station also listens for a period of time to ensure that no other station is transmitting synchronously during this time, and ultimately, the station's transmission is successful. Conversely, if a collision occurs, the source station sends a blocking signal to amplify the collision. Contesting stations retry after a delay (the delay time is determined by the algorithm and is random). Under this mechanism, all stations and all hubs must be within the same collision domain.
For industrial Ethernet, 10Mbps and 100Mbps are the most common. In a 10Mbps Ethernet network using only twisted-pair cabling, there are two distance-related concepts: segment and network diameter. The former refers to the distance between two devices (hubs, switches, or hosts), while the latter refers to the distance between the two furthest devices in the network. Regardless of whether it's a 10Mbps or 100Mbps network, the maximum distance between segments cannot exceed 100 meters. Considering network extension, the most useful rule is the 5-4-3 rule (only for 10Mbps repeaters). The rule states that a network can have a maximum of five segments, four repeaters, and no more than three mixed segments. Mixed segments refer to coaxial bus segments (obsolete). Since the maximum distance between twisted-pair segments is 100 meters, the maximum network diameter is 500 meters. The maximum distance for a fiber optic network segment can reach 2 kilometers, but the IEEE 802.3 standard stipulates that when using fiber optics, the maximum number of cascaded nodes cannot exceed three, and twisted-pair cabling must be used at the network endpoints. The two intermediate nodes must be fiber optic segments, and each segment must not exceed 1 kilometer. Thus, the entire fiber optic network segment length is limited to 2 kilometers. The 5-4-3 rule does not apply to 100Mbps networks. It is recommended to use a 100Mbps switch.
6. Hubs and switches
Repeater hubs (or simply hubs) are fundamental devices for building Ethernet topologies. They are multi-port devices with four, eight, or twelve ports and can be cascaded to form a distributed star topology. Hubs all comply with IEEE 802.3 repeater unit requirements. These requirements include preamble generation, symmetry, and amplitude compensation. Repeaters must retime the signal so that jitter caused by transceivers and cables does not accumulate as it propagates across multiple network segments. These devices can detect incomplete packets and collisions and generate a blocking signal. They also automatically isolate problematic ports to maintain normal network operation.
Another product line of interface adapters is the interface converter, sometimes called a transceiver. They convert one medium to another. The most important conversion is from twisted-pair to fiber optic. Since many hubs lack fiber optic ports, interface converters are used to support fiber optic applications in the network. These devices are transparent within the network. The ports do not store frames or detect collisions; they simply convert one medium into a signal compatible with the other.
Switching hubs (switches) can replace trunk hubs and improve network performance. Unlike physical layer devices—trunk hubs—switching hubs are actually bridges connecting two data links, meaning collision domains are terminated at each switch port. Therefore, adding switches expands the geographical reach of the network; cascading switches allows for large-scale network expansion. Switches are more complex than trunk hubs. Twisted-pair ports automatically negotiate rates with auxiliary ports (10Mbps or 100Mbps). Flow control is also negotiated. Full-duplex segments use the PAUSE scheme, while half-duplex segments typically use the backpressure scheme. A switch reads a complete frame and examines its source address to determine the port location of the connected Ethernet device. The switch then generates and maintains a port address table. From this point on, network communication is limited to the ports relevant to the current transmission. Because synchronous transmissions can be performed on these ports without any operation, network throughput is increased. The table is automatically updated based on changes in connection information.
If a port receives information that needs to be broadcast, grouped, or sent to an unknown address, the switch will automatically forward the information to all ports. Unlike trunk hubs, multiple collision domains exist here, and each collision domain must adhere to the rules described above.
Repeater hubs can be connected to switch ports. If the network consists entirely of switches, the twisted-pair cable segment should be kept to 100 meters, but there is no limit to cascading. Before using fiber optic cables, it must be specified whether it is half-duplex or full-duplex. The comparison between repeater hubs and switching hubs is clear: switches offer slightly better performance than hubs, but hubs are easier to understand; data communication can be observed using a network analyzer from any port. Switches, on the other hand, require broadcast transmission on a specific port for measurement. As a bridge, a switch stores and forwards entire data frames, introducing data latency. Hubs receive network signals without latency. Cascading switches further increases latency; therefore, hubs and switches each have their own applications in industrial Ethernet.
7. Half-duplex, Full-duplex
Half-duplex means that sending and receiving the same medium are asynchronous. Full-duplex, on the other hand, has separate send and receive paths. Full-duplex links are crucial for extending Fast Ethernet (100Mbps). A full-duplex link segment cannot exceed two devices, which can be network interface cards (NICs) or switch ports. Note: This does not apply to trunk hub ports; hubs do not have a full-duplex mode. This is because hubs are part of a collision domain, which amplifies collisions received by other ports. Full-duplex communication can be implemented with only two NICs; for more than two NICs, a switch must be considered for full-duplex operation.
10BASE-T and 10BASE-FL have separate transmit and receive paths, and can operate in full-duplex mode depending on the complexity of the network interface card or switch port. If these interfaces are configured in half-duplex mode, synchronization detection of receive and transmit will trigger collision detection. When the same interface is set to full-duplex, collision detection will be disabled because full-duplex does not comply with shared CSMA/CD rules.
The full-duplex link must be configured correctly. When a site is configured in full-duplex mode, the site or hub port sends frames while ignoring the CSMA/CD protocol. If the other end is configured in half-duplex mode, it will detect collisions and cause other problems, such as CRC errors, reduced network speed, and the loss of the advantages of Fast Ethernet.
As mentioned earlier, the network range at 100Mbps is reduced due to collisions. For twisted-pair segments and switch ports, the maximum segment distance is 100 meters (within the collision domain). The problem lies with fiber optic ports; for multimode fiber, the segment length is 2 kilometers; for single-mode fiber, it's 15 kilometers. In half-duplex mode, limited by the collision domain, the segment distance is 412 meters. Therefore, only in full-duplex mode (CSMA/CA ignored) can the fiber optic segment reach its limit. For Fast Ethernet, switching technology is recommended. For fiber optic ports in Fast Ethernet, full-duplex is recommended.
8. Automatic negotiation
With the widespread adoption of Fast Ethernet and its similar wiring rules to Traditional Ethernet, IEEE 802.3u recommends automatic configuration of Fast Ethernet, enabling Traditional Ethernet ports to work with other Fast Ethernet ports. This configuration protocol is based on National Semiconductor's NWay standard. Twisted-pair links automatically perform speed matching to facilitate data communication. This scheme is suitable for twisted-pair links. The situation is different with fiber optics. Although fiber optics have played a very important role in the history of Ethernet development, the speeds of two fiber optic devices cannot be automatically negotiated because 10BASE-FL devices operate at 850nm, and 100BASE-FX devices operate at 1300nm. They are not interoperable. However, for auto-negotiation protocols, self-negotiation between two fiber optic devices is possible (if communication is not an issue). Recognizing this, the newly introduced 100BASE-SX standard allows 850nm fiber to operate at 10Mbps or 100Mbps. At 100Mbps, the segment distance is 300 meters. Therefore, this should be considered during installation. Fiber optic speeds are usually fixed and do not undergo negotiation. Auto-negotiation protocols are successful on twisted-pair links. The advantage of auto-negotiation is that it eliminates the need for manual settings, allowing the devices themselves to determine their respective technical levels. The levels, from highest to lowest, are as follows:
1000BASE-T full-duplex highest
1000BASE-T
100BASE-T2 Full-Duplex
100BASE-TX Full-Duplex
100BASE-T2
100BASE-T4
100BASE-TX
10BASE-T Full-Duplex
10BASE-T minimum
The lowest priority is 10BASE-T (half-duplex, shared Ethernet), and the highest is 1000BASE-T full-duplex. This is a complete priority scheme, but it doesn't mean that a particular network interface card (NIC) can handle all of these technologies. In fact, some technologies are not commercially implemented, but they are all consistent with the IEEE 802.3 standard. Each port checks its own technical capabilities and determines the final speed (lower speed). For example, if a NIC supports 10BASE-T and the switch port is capable of 10BASE-T or 10BASE-TX, then 10BASE-T is ultimately chosen. If one NIC is 10BASE-T and another is 100BASE-TX, they cannot communicate because they are incompatible.
9. Transmission Protocol
The initial design did not address reliable end-to-end message delivery. The responsibility for network interconnection (communication between two networks) lies with Layer 3 – the network layer. Transport and interconnection become part of the protocol stack, with TCP/IP and SPX/IPX being two commonly used protocols. These two protocols are not interoperable, so Ethernet nodes must use compatible protocols. Due to its application on the Internet, TCP/IP has become the primary protocol, and this is also true in industrial networks. In fact, TCP/IP is a set of protocols defined by RFCs (Request for Comments) that have been around for many years. Besides Ethernet, TCP/IP also works with other data link technologies, residing above the physical/data link layer. At the transport layer, there are two important protocols: TCP and UDP. The former acknowledges received information. Both are useful. Above the protocol stack, several useful application layer protocols are used in industrial Ethernet. Addressing is an important topic for users. The IP protocol is responsible for routing data packets between stations that may be located in different networks. Each station has a unique 32-bit address (representing both the network address and host address). The address is represented in four bytes of dotted decimal notation. 128.8.120.5 is a valid address, but it's impossible to determine which part represents the host and which represents the network. Addresses are divided into five classes, A through E. The classification can be determined by observing the first byte.
IP address allocation is not simple and is usually done by the network administrator. Once allocated, it must be used on various stations in the network. IP addresses are allocated in two ways: static and dynamic. Dynamic allocation is done by the server, while static allocation is done by configuration. The following addresses are private addresses and cannot be assigned on routers. Therefore, they are not used on the Internet.
10.0.0.0~10.255.255.255
172.16.0.0~172.31.255.255
192.168.0.0~192.168.255.255
IP addresses and Ethernet MAC addresses are different and should not be confused. MAC addresses are assigned by the device manufacturer and are therefore globally unique. IP addresses are assigned during installation and can be reassigned as needed.
10. Application Layer Protocols
Once you determine the connectors and cables to use, whether to use a hub or a switch, and assign IP addresses, communication between sites is possible. Now you need to consider OSI higher-level compatibility. Recommended industrial automation protocols include Ethernet/IP, iDA, PROFInet, and Modbus/TCP. This does not include traditional internet applications—FTP, SNMP, SMTP, and TELNET. Users' devices may not support these protocols, so it's necessary to understand the compatibility of your own system.
Disclaimer: This article is a reprint. If it involves copyright issues, please contact us promptly for deletion (QQ: 2737591964 ) . We apologize for any inconvenience.
