Networking

It's not Ethernet vs. fieldbus: Asking the right question about industrial communications

Reader feedback: Ethernet technology can solve fieldbus problems.

By Dick Caro

All of the Ethernet-based fieldbuses avoid the use of transmission control protocol (TCP) because it is nondeterministic.

I know you are not a networking expert, but I am. I found it difficult to read your cover article for the November 2016 edition. Let me explain.

Ethernet is a communications technology widely used to solve problems in industrial Fieldbus communications. Let me illustrate—EtherNet/IP, Profinet, PowerLink, EtherCAT, Sercos III, Foundation Fieldbus HSE and CC-Link IE are all fieldbuses and use Ethernet technology at layers 1 and 2 (physical and data link layer). Why do I say they are fieldbuses? They are all in the International Electrotechnical Committee (IEC) Standard 61158 that defines industrial fieldbuses. Each has different ways to use Ethernet, but none of them tolerate any collisions because it is collisions that caused old-fashioned coaxial cable Ethernet to be nondeterministic. Therefore all of these forms of Ethernet are fieldbuses and are deterministic.

Twenty years ago, offices and industry stopped using 10Base-2 coaxial-cable-based Ethernet because it just didn’t work well in the office and not at all in the factory. Since then, all Ethernet networks have used full duplex switches to buffer all messages and prevent collisions. While many industrial networks have used ruggedized Ethernet switches, several very high-speed machine control fieldbuses such as EtherCAT, PowerLink and Sercos III use Ethernet switch chips at each node and have the capability to form lightning-fast deterministic networks based on unmodified Ethernet protocol, and they call themselves fieldbuses.

 

By the way, anyone who refers to full duplex switched Ethernet in any form as nondeterministic is just behind times. We are so far advanced from those days, but there are some who are stuck in the old folklore that they cannot accept modern technical facts.

All of the Ethernet-based fieldbuses avoid the use of transmission control protocol (TCP) because it is nondeterministic. Instead, for real-time communications they use different applications of user datagram protocol (UDP) with Internet-protocol (IP) addressing. Some, such as Foundation Fieldbus HSE, do not allow any TCP-encoded messages, while others allocate fixed time slots for such TCP messages. This is only one type of time-sensitive networking (TSN) mentioned but not defined in your article.

Ebook: What manufacturers really think about the Industrial Internet of Things

So, it is not Ethernet vs. fieldbus. The question is: Will old non-IP fieldbus protocols go away? That would mean implementing IP at the edge device itself. This has already begun in both factory automation and process control, but there are many millions, or perhaps billions, of existing not-very-smart devices in the field that are not going away overnight.