Assuring Packet Delivery

Classical Ethernet was partly based on probability theory.

Aw 1035 Gmintchell0803

In fact, Ethernet inventor, Bob Metcalfe, once wrote a column in an Information Technology magazine that often featured probability questions. One device that wanted to send information to another device would break the string of data into smaller “packets” and send them all off into the network. If it received a communication of a network failure, it would retry. Meanwhile, the recipient would re-assemble the packets into the message with a good probability that they all made it.

Network speeds got so fast that data subscribers seldom realized that there were re-tries in the system. Speed could trump probability. This was good enough that when engineers decided to use Ethernet for more than simply connecting to the enterprise system to send production information, they still didn’t worry. Even for most input/output messaging, the speed was great enough that messages got through.

Then engineers wanted to extend the benefits of Ethernet to the realm of very high-speed control, such as motion control. When you are moving axes and tools on a machine with response times in the sub-millisecond range, then probability doesn’t work. You want to know that the packets got there when they were supposed to get there.

Several solutions have been developed to accomplish this. Most use chips specially designed to synchronize the timing and delivery of the packets. This month, series Contributing Editor Terry Costlow looks at the technologies and applications that have extended Ethernet into the realm of deterministic networking.

Check out an on-demand Webcast on Deterministic Ethernet for Distributed Motion Control at

Subscribe to Automation World's RSS Feeds for Columns & Departments

More in Control