The next frontier for the established fieldbus organizations is to produce Ethernet TCP/IP application layers of their protocols.

By Perry Sink Marshall
Originally appeared in "The Industrial Ethernet Book" (C)2001 GGH Marketing Communications.

There are three major contenders: Modbus/TCP (Modbus protocol on TCP/IP), EtherNet/IP (the ControlNet/DeviceNet objects on TCP/IP) and ProfiNet, or PROFIBUS on Ethernet. A fourth, FOUNDATION Fieldbus High Speed Ethernet, may be a contender but is not designed to be a device level connection in the same way its counterpart, H1, is.

One could propose an infinite number of potential application layer protocols, and in fact right now there are, in addition to the above, a myriad of proprietary standards. But there are several significant advantages to employing existing bus architectures

  • Profiles for many devices have already been defined, and can be transferred to Ethernet with relatively little effort.
  • In systems which use, for example, PROFIBUS as an I/O level network, and PROFIBUS on Ethernet at the supervisory level, the relationship between the two networks is relatively transparent. Data can be transferred between the upper and lower network fairly easily.
  • Many developers and users are familiar with existing protocols, and this speeds product development and adoption.

The question is, what standards will eventually be adopted? Actually you could have all four protocols on the same wire, running simultaneously! Whatever you choose, when you take the web into consideration, the possibilities are tremendous for Ethernet-enabled automation systems.

But who will win in this wild west scenario? Contrary to perception, it will be difficult and expensive to get Ethernet to the sensor level. Using Ethernet to turn a valve on or off, or to connect a transmitter or other field device, is like filling a bathtub with a fire hose. However, it will do very well for `racks’ and clusters of I/O tying into a single node. Ethernet will not necessarily be ‘dirt cheap’ in industrial applications, and overall may prove to be more expensive.

Ethernet PC cards cost 1/10th as much as, for example, PROFIBUS cards for good reasons. They’re made by the millions; they’re passive, meaning that the PC does most of the work. And a PROFIBUS card usually has a processor handling communications, which brings substantial performance advantages. Then again, long term availability is a potential problem with consumer products, where life cycles can be measured in weeks! Finally, the quality of cards from an office supply store are certainly not for professional, industrial use.

In embedded applications Ethernet chips cost much more than say CAN chips. So devices themselves with Ethernet built in will definitely have a cost factor. I expect Ethernet to have costs more akin to PROFIBUS devices, which incorporate ASICs. Industrial grade cables and connectors will increase costs as well.

Not only are Ethernet ASICs more expensive: running a TCP/IP stack takes a bigger engine than a typical 8051 processor. So the processors will cost more, too. A TCP/IP packet has 68 bytes of overhead. For short messages, corresponding to typical industrial I/O, that’s a lot of overhead and 10Mbits may not be as fast as it sounds. In applications with high electrical noise, Ethernet may simply be impractical although the promise of fiber optic 100M and 1000Mbit networks probably overcomes most speed issues – at a cost!

Currently, you can purchase both ‘proprietary’ Ethernet products as well as ones that use open application layers like Modbus/TCP. And in some cases, you can buy controllers and software packages with drivers, and have a complete system. Is there a guarantee that what you buy now will still be an accepted standard in five years? Well, that’s a little iffy, given our `wild west’ proposition. If you want to go West, then place your bets! I am absolutely sure that you’ll be able to buy DeviceNet products five years from now. I think it’s pretty reasonable to guess that you’ll also be able to buy Ethernet/IP (DeviceNet on Ethernet) and PROFINet (PROFIBUS on Ethernet) products in five years as well. But at this moment, you’ll be hard pressed to assemble an automation system with Ethernet/IP or PROFINet products today.

So what can you do? One option is to choose a popular, open, non-Ethernet fieldbus. You’ll be more progressive than the majority and there’s little risk – if you get help from an experienced integrator and make sure your people are trained. And I think there’s reasonable assurance that, in the future, it will be compatible with an IP version of that same fieldbus.