Into the ETHER
15 Jan 2000
There has been no let-up in the fieldbus battle, with the main players, Foundation Fieldbus and Profibus, still struggling for supremacy. Yet in recent months another challenger has arisen: Ethernet, the technology which links office networks together.
Ethernet is, ironically, the oldest of the communications systems: it's been around for a quarter of a century. All the equipment needed to make an Ethernet network is readily available at a fraction of the cost of a proprietary control network. It has proved to be an ideal way to solve office network problems, especially since the rise of the internet; the TCP/IP protocol - the method the internet uses to communicate - is an internationally-accepted standard and is very efficient for office-type use.
And that's always been the sticking point. `Ethernet has always been criticised, especially by manufacturers of proprietary networks, as not deterministic and not robust enough to meet industrial needs,' said Tony Ciardiello, product manager of Schneider Electric.
The problems with Ethernet are certainly not trivial. As John French, managing director of NETdot3, explained, the original reason for using Ethernet in office networks was its then-huge bandwidth capability - 10Mb/s - and the relatively low speed of PCs. But the advent of client-server architecture, e-mail and high-resolution graphics put such a strain on the systems that by the early 1990s, when fieldbus development took off, Ethernet had acquired a reputation as unreliable and slow. `Ethernet was considered,' he said, `but for good reason at the time was thought useful only at the plant or information layer where determinism and resilience requirements were less stringent. Even today, there is a perception that Ethernet still suffers from the same constraints.'
This is not the case, Ciardiello argued. `With a properly structured network, Ethernet can offer equivalent, if not better performance than any proprietary network,' he claimed. `Remember, that's based on 10Mb/s, and as bandwidth increases then so will performance.'
So what's changed? French pointed to the development of full-duplex Ethernet switching as the main engine towards making the system deterministic. This is a vital property for a communication system controlling an industrial process. It means that the controlling device - whether a large server or PC - can scan a pre-determined number of `slave' devices, such as sensors, transmitters and drives, and recieve responses from them, within a set time. Traditionally, Ethernet could not do this. Only one device could use the `wire' to send or receive messages at any one time: the equivalent of a walkie-talkie system rather than a telephone. When there is a lot of traffic on such a system, messages don't reach the devices on the network and everything slows to a crawl. Clearly, when the network is entrusted with duties such as opening valves in response to sensor readings - some of which might be safety critical - this communication system is not suitable.
The full-duplex development removes this problem, French said. Every device on the network now has its own dedicated switch port, through which it can transmit or receive data, at wire speed, without any loss of information. The switches, which were first introduced in 1995, also ensure that messages only go to the relevant part of the network. Standard message prioritisation techniques ensure that `mission critical' information - such as a signal to open or close a valve, or change the speed of a drive - always jump to the front of a queue.
Resiliency is also less of a concern now than in the past, French claimed. This is the ability of a network to recover from the failure of any of its nodes or devices. A new development known as the redundant ethernet pseudo ring will allow networks to recover from failure within 20ms, said French. This works by using the switches to create a logical, rather than physical, break in the network if a component fails.
There are still some major limitations: power cannot be transmitted via an Ethernet; the cost of adding Ethernet connectivity to a device is around £10-20, although French expects this to halve in future; and there is, as yet, no standard for Ethernet connectors. However, as Ciardiello noted, more money is being spent on developing Ethernet than on all the other networks put together. `We are confident that Ethernet will win the fieldbus war in the next couple of years,' he said. PE