Legacy Moves to Ethernet

Ethernet has become the de facto standard for business and industrial communication for several reasons: ubiquity, low cost, the ability to carry multiple protocols, and access to the Internet. According to the study "Industrial Ethernet Market Outlook Study" from ARC Advisory Group, worldwide sales of industrial Ethernet is expected to grow from 840 thousand units in 2004 to just over 6.

By Peter Cleaveland October 1, 2005
  • Industrial Ethernet growth

  • Serial device servers

  • Keep existing I/O points

  • Layered approach

What about 4-20 mA?
Moisture sensors move from serial to Ethernet

Ethernet has become the de facto standard for business and industrial communication for several reasons: ubiquity, low cost, the ability to carry multiple protocols, and access to the Internet. According to the study ‘Industrial Ethernet Market Outlook Study’ from ARC Advisory Group, worldwide sales of industrial Ethernet is expected to grow from 840 thousand units in 2004 to just over 6.7 million units in 2009, at a compounded annual growth rate (CAGR) of 51.4%.

‘The cost of Ethernet has become affordable to use all the way down to the micro PLC,’ says Bill Black, controllers product manager, GE Fanuc. A 250-ft roll of commercial-grade Cat 5 cable can be purchased at retail for $60, and that’s without shopping for the best price. A 1,000-ft spool of top-quality Cat 5 cable for severe industrial use goes for $270.

Perhaps the biggest advantage of the Ethernet physical layer is it can carry multiple Ethernet protocols and some others, for various purposes. ‘I can use Modbus TCP in a master/slave orientation to capture data for HMI/SCADA or maybe some data acquisition,’ says Benson Hougland, Vice President, Opto 22, ‘but if I need to transfer a file to that device, I can use FTP. If I want to have the device send an e-mail, I can simply use another well-know protocol called SMTP to send that e-mail and do it concurrently or simultaneously, whichever I choose.’

Connecting legacy devices

Probably the most direct way to connect a legacy serial device to Ethernet is with a serial device server (SDS). This device translates data back and forth between serial and TCP/IP, and provides a physical connection to Ethernet. ‘A common application of such a device is to allow point-to-point communication of two existing serial devices over a network,’ says Greg Dixson, product marketing manager for automation systems at Phoenix Contact. ‘Data is tunneled through the network and any range limitation concerns are eliminated.’

Some serial device servers also come with software that, when installed on a PC, makes the serial device appear to the control computer as if it were connected to a serial port on the PC. This can be a big help if a PC is replaced with a newer model that lacks serial ports. In addition, it provides access to the remote device from anywhere the plant Ethernet LAN is available. Many serial device servers also contain built-in Web servers, allowing access to the connected device via the Internet.

SDSs are available from multiple vendors, with prices typically ranging from a few hundred dollars to less than $100.

Some vendors suggest that an alternative to the SDS may be appropriate in situations where distances are short and there’s a need to increase data rates. ‘For those looking to add serial ports without opening the PC and adding a card to it,’ says David Johnson, product marketing manager at Quatech, ‘device servers have certainly worked well. But serial-to-USB adapters are also handy for doing that job. Many machine tools use one or the other, depending on how far away the PC running DNC software is from the CNC controllers.’

New for old

Architecting Ethernet and ControlNet as a backbone to a series of DeviceNets or other device-level networks.

While single devices may have dedicated serial links, a great deal of available industrial inputs and outputs connect via remote I/O units that multiplex many points to one cable. Many of these devices consist of a base that holds individual I/O modules and a plug-in network interface board. Tying existing I/O connections to Ethernet simply by replacing the network interface board can save a great deal of money.

When users want to connect these existing I/O points to Ethernet, says Hougland, they don’t want to replace everything. Opto 22 is addressing this customer need with its E1 and E2 brain boards. ‘They leave all their I/O in place—the racks, the I/O modules, the field lines, the power supplies—and they take out their serial brain board and put in our Ethernet brain-board,’ he says. ‘That means for a very low cost they’re able to upgrade a 20-year-old I/O system to today’s Ethernet networking.’

Opto 22 is not the only company to take this path. Mark DeCramer, product manager, advanced electronics, Wago Corp., says ‘[Wago’s] I/O modules are independent of the communications module. When upgrading to a different control network, all the I/O modules and wiring to field devices remains intact. Only the communications module needs to be exchanged.’ By changing communication modules, the Wago I/O System 750 can connect to Ethernet TCP/IP at 10 Mbps or 10/100 Mbps, as well as to multiple other buses.

Wireless Ethernet

Along with the growth in Ethernet has come an increase in the acceptance of wireless Ethernet (802.11b, also known as Wi-Fi), and many of the companies that provide serial device servers also have wireless device servers. ‘What’s impeded even more explosive growth in wireless,’ says Johnson, ‘are security concerns and the logistical issues of deploying a wireless network in a hot, noisy, metallic environment.’ Despite this, Quatech has made a big bet on wireless by merging with DPAC Technologies, which specializes in wireless connectivity,’ because we know the future is headed there,’ says Johnson.

Security is vital, but many security concerns can be addressed with basic tactics such as: not using the default password that came with the equipment, not using the default network name, blocking the service set identifier from being broadcast (closing the network), using medium-access control tables to specify that only certain users will have access the network, and making sure all access points are IEEE 802.11i-enabled.

Multi-layered architecture

While some advocate use of Ethernet down to the device level, it is common to see a more layered approach. This is seen in the architectures built on CIP (Common Industrial Protocol), such as Ethernet/IP, DeviceNet, and ControlNet—governed by independent network organizations, but with origins at Rockwell Automation.

Ethernet/IP has the advantage that, as an application-layer protocol, it is carried on Ethernet the same as TCP, SMTP, and IP. This makes expansion of an existing DeviceNet installation easy, says Brian Oulton, marketing manager for networking, Rockwell Automation. Rather than change everything, he says, it’s not difficult or expensive to ‘add a different network, ControlNet or Ethernet/IP, and add new devices to the new network and connect it to the old network. Oulton suggests using ‘Ethernet and ControlNet as a backbone to a bunch of little DeviceNets,’ (see ‘Layered architecture’ graphic). This architecture accommodates existing RS-232 and -422 systems, he adds. ‘It simplifies the topology and work involved because it gives you nice long distances.’

One criticism that has been leveled at Ethernet for many years is that it is indeterminate: Ethernet’s CSMA/CD media access control cannot guarantee a specific response time—an issue that helps sustain those in favor of the layered approach. The idea behind the layered architecture is that it allows lower levels to continue to use real-time protocols and lets the Ethernet layer handle large data files and other non time-critical material. And deterministic Ethernet protocols are available.

What about 4-20 mA?

For analog instruments connected by 4-20 mA loops, the lowest-cost way to get increased functionality continues to be HART. ‘The advantage for the user is the wealth of diagnostic and service data available over HART,’ says Bill Black of GE Fanuc. The other advantage of HART is that it uses the wiring originally installed for the 4-20 mA loop, so there’s no need to install a fieldbus.

Moisture sensors move from serial to Ethernet

Applied Instrumentation (Concord, CA) makes and installs specialty electronics for agriculture. The company’s primary business is in developing controls for crop dryers, including a line of proprietary low-frequency RF moisture meters that monitor the moisture content of products in bins, trucks, or storage silos. In California’s Central Valley, the controls are used primarily for monitoring the drying of walnuts, pistachios, filberts, and almonds.

According to Donald Osias, who owns and operates the firm, many smaller facilities use a handheld meter to monitor the drying process. This requires an operator to walk from bin to bin and plug into a sensor in each one. Larger facilities tend to have more sophisticated drying equipment, ‘and for those we have been building systems that monitor the moisture in all the bins in the plant and display all that information on a CRT display or an old PC in the dryer and send the data off,’ says Osias.

Most facilities that Applied Instruments works with have a fairly small number of I/O connections, with a large plant having perhaps 50 points, distributed throughout a large building. The company primarily used Optomux I/O units from Opto 22 to connect to the sensors for its clients. ‘It provided a simple ASCII text readable format, the costs were reasonable, and the relays were bulletproof,’ explains Osias. But in recent years, a problem developed: Newer computers no longer have easy support for serial data, and it became increasingly difficult to get the manufacturer-supplied C libraries and compilers to support it.

Osias plans to stop using RS-485 or 422 links and switch to Ethernet with any new installations of control equipment. Hoping to continue to use the installed base of Optomux I/O, a year ago Osias asked Opto 22 for a way to connect Optomux to Ethernet. He was told that a product to do that was under development: the E1 and E2 line of Ethernet brain boards. Introduced in June of this year, these boards plug into an Optomux base in place of the original B1 boards that communicate via an RS-485 serial link. E1, for digital data, and E2, for analog, provide both 10/100 Mbps Ethernet and RS-422/485 serial networking, and support multiple protocols simultaneously.

Osias is gratified that Opto 22 included this dual capability, as it allows his clients to retain their installed base of Optomux I/O and associated wiring. Recently, he installed an E1 on an Optomux base closest to the CPU at a client facility and used that same E1 as an external serial device server to connect the other Optomux units in the facility to the CPU. ‘I put a new controller in there with a new CPU for better monitoring capability; it’s all Ethernet, but we’re able to keep the old B1s that are out there running on serial.’