On Choosing a Network Topology

I know of no better route to understanding a topic than starting with the big picture before narrowing in on its micro details, and history is generally the best introduction to that big-picture overview. I was reminded of this while recently studying Advantech’s white paper, Understanding Device Level Connection Topologies, an excellent resource that covers the history that frames the automation-connectivity big picture before diving into the nitty-gritty of the current state of the connection-topology art.

To summarize, at a point in the not-so-distant past, data acquisition systems (DAQs) consisted of sensors hardwired directly to PLCs. Sensors were analog devices that simply reported on-versus-off status via the presence, or lack thereof, of an electrical signal. In contrast, today’s “devices” – switches, values, transmitters, motors, and such – typically involve smart sensors, sensors equipped with microprocessors that allow them to communicate with other devices and with host computers. Once devices were provided this ability to “talk,” the importance and complexity of interconnections between devices and controllers – the network – increased. Engineers must now not only decide which sensors and controllers to deploy where, but also by what network topology to connect those devices.

The white paper discusses the form and merits of seven basic topology models:

  1. Line: Also known as point-to-point topology, this is the simplest model wherein two points on the network are permanently connected by a single channel. Connection of a SCADA RTU to a central head via a leased pair line, is a common example of a line topology.
  2. Ring: A network wherein the last node is connected back to the first, forming a ring. Data is transmitted in a single direction along the ring and the signal is repeated at each node via a transmitter and receiver. If any node link is broken, data flow is interrupted.
  3. Mesh: Most international standard wireless protocols use mesh topology because of its high level of redundancy. Data can travel from node to node via any path in the mesh. Mesh is rarely used for complex wired systems because of the high cost of wired implementation.
  4. Star: A central node (hub or switch) is connected to each endpoint node. Star topology is among the easiest and least expensive to install and maintain. Stars can even be stacked via line connections between their central nodes. Of course, failure of the central node will cause failure of the entire system.
  5. Tree: A hierarchical arrangement comprised of at least three levels – a root node connected to one or more nodes lower in the hierarchy with line connections comprising the tree limbs. Tree topology is particularly appropriate for large wired networks.
  6. Bus: All devices are connected via a single bus cable, with each signal traveling the entire length of the bus until reaching the intended device. Bus topology is commonly used at the device level. Bus configurations are inexpensive to install, but can be expensive to manage over extended networks.
  7. Hybrid: “Ganging” or daisy-chaining mixed topologies is feasible as long as all nodes run the same protocol, or protocol is converted at critical junctures. Hybrid topologies are now very common.

The optimum topology is often dictated by the type of connection favored by the specific devices installed, as well as factors such as cost, durability, complexity, etc. Wired systems usually use an Ethernet protocol – Modbus TCP, Profibus, or Foundation fieldbus. Advantech’s ADAM-2000 wireless modules facilitate star, tree and mesh topologies without further configuration.

For more information about Advantech, please visit our website.