A number of different industrial network topologies such as star or chicken foot are in use today for the various levels of the enterprise. The field level has different environmental and bandwidth requirements than the control system backbone. These requirements significantly influence the decision-making process of today's automation professional.
Today's industrial networks extend to all levels of the enterprise with field-level protocols such as Foundation fieldbus, Profibus, Modbus and others running on twisted-pair, to Ethernet typically from the control level, including remote I/O, and through the enterprise. Of course, now there are wireless permutations over these same domains, but we will leave wireless alone for now.
Ethernet is by default a star topology with the router or switch having individual spurs to each end device. In some cases this end device is another router or switch, and, as a result, a network is born. The majority of industrial Ethernet suppliers offer some form of a ring topology network as well to provide an economic and rapidly healing (typically 20–50 ms) method to increase the reliability of a network used to transmit control messages with minimal—hopefully no — interruption in the control of the process.
At present these ring topologies are proprietary, meaning all the switches in the ring must be from the same supplier. Now that's not really a major obstacle since most facilities use a single supplier for their infrastructures.
Rapid spanning tree protocol (RSTP) is an IEEE standard, and there is discussion underway to develop a standard for a ring topology, as well.
At the field level, more options are available. Daisy chain, star, spur, chickenfoot and tree are the most commonly considered. It's important to remember that the predominant process automation networks, Foundation fieldbus and Profibus-PA are based on the same physical layer and, most importantly, are wired in parallel. Also don't forget that all networks need terminators to match the system's characteristic impedance to avoid noise problems.
Daisy chain is widely used in RS-485 networks, including Profibus-DP and Modbus. The devices are linked by connecting each device or network node in series, one after another. It is the computer equivalent of a series electrical circuit with the same limitation: if one device in the network should fail, all devices downstream of that device are no longer visible. Those of us old enough recall old Christmas light strings in which if one bulb failed, the tree went dark. Fortunately, many protocols using RS-485 include connectors that are a combination daisy chain/spur connector with circuitry that isolates a failed device from the network, so the chain is not interrupted.
The spur network is very similar to daisy chain; the devices are strung along a trunk with each device connected to it by a short length of cable — minimum of 1 m.
The chickenfoot is the most commonly used configuration for process automation, in large part because it is most similar to a traditional analog instrument installation with a homerun cable — multiconductor — and individual cable spurs cables to the end devices. Many people call a chickenfoot installation a tree; however, the only tree it could represent is a palm tree with all the leaves at the top.
A tree configuration should have many branches, and the same is true for industrial networks. The devices are the leaves, and the trunk is the cable with terminators at either end. Normally one of those ends is connected to the control system I/O card, and then you can have a combination of spurs and chickenfoot termination assemblies along the trunk. This provides the end user with the optimal use of the installed field cable and takes full advantage of the capabilities of field-level industrial networks. Ethernet networks are built the same way with various routers and switches connected together, typically on a higher-speed trunk to connect a variety of end devices such as computers and printers.
The above does not include consideration of limitations imposed by the area classification of the installation, which obviously will impact the hardware selected to build the network.
No one network is the right answer in every situation, which means that engineering and thought is required to determine the optimal solution.