Ethernet for Real Time Control?

Real-Time = Predictability

There are many reasons why Ethernet would be a great technology for real-time control networks. After all, Ethernet is familiar, it is ubiquitous and it is cheap. But Ethernet has a serious drawback: there is no way to know exactly when an instruction will be delivered. Real-time networks have always been based on the certainty that when instructions are sent they will be received and understood in about one millisecond.

The reason for Ethernet’s unpredictability is the CSMA/CD bus access procedure (Carrier Sense Multiple Access with Collision Detection). CSMA/CD has never been able to offer a guarantee of timely delivery. In the real-time world of machine control, this kind of chaos is simply unacceptable. The appeal of Ethernet for real-time networking is certainly strong, but without absolute timing certainty, there would be a huge risk to life and limb, not to mention the cost of ruined machinery and products.

If Ethernet cannot guarantee message delivery timing, further discussion is pointless. So the question is, “Can Ethernet be configured in such a way that it offers real-time control?” Fortunately, the answer is “Yes,” thanks to hardware and network configuration methods which are able to circumvent the chaotic CSMA/CD bus access procedure.

CSMA/CD bus access

Ethernet networks use the CSMA/CD bus access procedure which determines when individual computers can send data. If a participating device wants to send, it first tests if the bus has been pre-engaged by another participant (Carrier Sense). If that is the case, the first participant delays its data transmission to a later time, a procedure which in itself rules out real-time operation.

If the network is “idle”, meaning it is recognized as free, the device waits for a precise time interval called the “interframe gap”, during which the device can send the data in a serial bit data stream. Initially the preamble (seven bytes 10101010) is sent, then the start frame delimiter (bit combination: 10101011), followed by the 6-byte-long target address (MAC address of the target equipment), then the 6-byte source address (MAC address of the transmitting equipment). Next comes the 2-byte-long length/type field, followed by the data field (the actual operational information with a length from 50 to 1500 bytes), and finally a 4-bytelong frame check sequence.

Using this system, it is possible for two or more participants to transmit data at the same time, in which case the electrical signals are superimposed and the transmitted information is lost. For this reason, every transmitter tests for signals already present on the bus (collision detection). If it recognizes a collision, it sends a 32-bit long jamming signal and breaks off its transmission.

All participants then wait for an interval determined by a random number, after which they may attempt access again. Of course, this new transmission attempt may again result in a collision and the whole procedure begins again up to a maximum total of 16 attempts. In short, there is no way to determine when individual participants may transmit. For real-time systems this is completely unacceptable. This is why Ethernet switches are necessary.

Only switches offer real-time operation

An Ethernet switch is like an internet router or a telephone exchange in that it decides which data packet is sent where. An Ethernet switch chooses among many possible paths through the network, and therefore offers the potential for some control. The secret to real-time Ethernet is to define an individual ‘network’ for every individual piece of terminal equipment, on which only that terminal equipment communicates.

By using two cable pairs for each connection, data packets can be sent and received simultaneously (full-duplex operation). In this configuration, using full-duplex and switching, two pieces of terminal equipment cannot get in the way of one another, and collisions involving two or more data transmissions are impossible. The chaotic CSMA/CD bus access procedure is no longer a factor, and Ethernet offers a real-time capability, which makes it suitable for industrial and defense applications.

Layer 2 vs. Layer 3

The ISO/OSI model subdivides the complex process of computer communications into seven layers. Ethernet switches operate on layer 2 of the OSI model, while internet routers operate on layer 3 (Figure 1). Routers use IP addresses to send data packets from one part of the network to another and depend on comprehensive routing tables to find the best suitable path. Internet routers are usually so complex that most of their work can only be accomplished in software running on a processor. Ethernet switches, on the other hand, can be designed in hardware, creating an enormous speed advantage.

Switches need no support from an external host system, nor do they need a driver to be installed, and it makes no difference to the switch which operating system is running on the host system. Switches are autonomous and installation is simple: insert the board, connect the cable, and the switch works.

Even the configuration of the switching tables can be completed without an administrator because the switches are so intelligent they automatically learn which terminal equipment is connected to which port. Every incoming data packet contains both the target address (the 48-bit MAC address of the receiver) and the MAC address of the sender. The switch registers the port where a packet arrives along with the MAC address of the sender, and enters both pieces of information into the forwarding database.

This forwarding database is a table that lists a port for every MAC address. If a packet is sent to a specific MAC address, the switch refers to the forwarding table for the corresponding port number and sends the data packet through that port.

If a MAC address is not yet listed, the switch sends the packet to all ports. Thus it ensures that at all events the packet is delivered. When the addressee answers, it becomes known to which port the addressee is connected. Subsequently the switch routes data packets for this addressee via this port only.

To ensure the table is always current, the switch couples every entry to a timer, which is restarted every time data packets are received from the corresponding MAC address. If the equipment is removed from the switch, the timer runs out and the switch removes this entry from the forwarding database. This procedure (automatic address learning and ageing) ensures that the forwarding database contains no obsolete entries.

GE Fanuc Intelligent Platforms’ switches can automatically determine the type of Ethernet cable connected, so it is no longer necessary to use a crossover cable when connecting two Ethernet switches (automatic MDI/MDIX crossover). This cable sensing feature facilitates planning, installation and maintenance of industrial and office networks.

Figure 1

Ethernet Hubs & Ethernet Switches

The original bus structure of Ethernet networks is rare today even in the business office environment because it makes network planning, installation and fault detection expensive and time consuming. There has been a trend toward the star network topology in office applications where printers, servers and workstations are no longer connected to a common conductor. It is common today for network devices to be linked by coupling elements known as hubs. Hubs are devices which regenerate data signals, amplify and retransmit them.

Hubs are transceivers which receive signals, restore the eroded signal flanks, adjust amplitudes to designated values, remove jitter and then retransmit signals throughout the network. This makes larger networks possible. Additionally it is much easier to identify the faulty components and network sections if problems occur. Hubs work exclusively on the physical layer (layer 1); therefore they cannot interpret data packets.

Ethernet switches operate both on the physical layer and on layer 2. (Figure 1) They regenerate the signals and interpret them. An Ethernet switch interprets the target address and sends the data packet onwards exclusively to this address. Ethernet switches can thus deliver data packets with deterministic certainty and without collisions. The chaotic Ethernet CSMA/CD bus access procedure can be avoided with their help, allowing the implementation of industrial networks based on Ethernet which fully match the demanding real-time requirements of process control.

Ethernet in Harsh Environments

Yet real-time capability alone is not enough. Using Ethernet in harsh environments (for example in industry, on drilling platforms or in defense), the components must be sufficiently stable to withstand conditions that may be extreme. Operating in industrial and military applications is no piece of cake: dirt, dust, spray, intense heat and cold, vibration, shocks and jolts make life hell for Ethernet switches.

The solution consists in providing boards with large conduction cooling heat sinks, so that the heat generated is reliably dissipated through the board edges to the wall of the hermetically sealed enclosure. The enclosure finally radiates the heat away, via base plate cooling or via forced air, into the environment.

The GE Fanuc Intelligent Platforms CP3-GESW8N can work reliably in a range of -40°C to +85°C. The CP3-GESW8, the Ethernet switch for normal industrial environments, has an operational temperature range from 0°C to +70°C. The humidity can range from 5% to 95%. The ruggedized version has also been particularly well armored against strong vibration and shocks.

Conclusion

Because Ethernet is familiar, universal, and affordable, it would be a good networking technology for both the office and the production floor. However, because Ethernet is inherently unreliable when not used in a proper way, it has never been used in real-time applications on the production floor. This limitation can now be overcome through the use of switches and intelligent network design. With the proper hardware and network configuration, real-time Ethernet networks are now feasible.

Jürgen Eder, GE Fanuc Intelligent Platforms