Using Ethernet for Data Acquisition

Ethernet note

InTech Magazine-01 June 2003
By Nicholas Sheble - Fieldbus News.
Write him at nsheble@isa.org.

Networking engineers continue to have questions regarding the efficacy of Ethernet and where it comes down regarding determinism. Is it or isn't it deterministic?

Networks that can reliably deliver messages on time are deterministic. However, the industry has defined that networks in which the time to deliver messages is unpredictable under the worst-case conditions are nondeterministic.

The property of the original 10Base2 Ethernet that makes it nondeterministic is the protocol for recovery from collision detection. The original ethernet used statistical setback which means that once the system detects a collision, both transmitting nodes stop their transmissions and wait a random interval before trying again.

This random number interval will be different in each node, so that one will retry before the other. It is the random number and statistical nature of resolving Ethernet's collisions that make it nondeterministic and unsuitable for most automation applications.

However, all modern 10/100/1000BaseT Ethernet-based automation networks totally avoid collisions and are therefore entirely deterministic and quite suitable for any automation purpose.

The technology that replaces collisions uses segmented networks based on active switching. With an Ethernet switch, each Ethernet segment consists of only one device and its own port on the switch.

The switch buffers all traffic and resolves all potential collisions while serving many nodes at the same time. Furthermore, most modern switches are full duplex, allowing nodes to receive at the same time as they send.