What’s Left to Say About Industrial Ethernet?

KEY WORDS Networks and communications Ethernet Device-level networks Fieldbus Everybody talks about the weather and industrial Ethernet, but nobody does anything about them. Well, not anymore. While Charles D. Warner and Mark Twain's famous quote still applies to meteorology, industrial Ethernet is moving from idea to reality on the plant-floor.

By Jim Montague, CONTROL ENGINEERING May 1, 2001

KEY WORDS

Networks and communications

Ethernet

Device-level networks

Fieldbus

Ethernet signaling is reportedly better than RS-232because it’s less susceptible to electrical noise.

Uniform, 8-bit frame size in Ehternet prevents undetected collisions.

Everybody talks about the weather and industrial Ethernet, but nobody does anything about them. Well, not anymore. While Charles D. Warner and Mark Twain’s famous quote still applies to meteorology, industrial Ethernet is moving from idea to reality on the plant-floor.

Multiplying hardware, software, and related solutions are giving Ethernet the momentum to reach manufacturing networks, while new cabling, connectors, and related products are outfitting historically office-based Ethernet for service in industrial environments. Likewise, previously disparate standards efforts are coalescing around common methods of implementing Ethernet, a trend likely to deliver much of the network interoperability-at least below Layer 7 or the application layer-that years of former fieldbus squabbles failed to produce.

Despite this good news, however, significant obstacles remain. Besides user reluctance and inexperience-not to mention historically, unhardened, office-based equipment-running Ethernet at increasingly higher speeds makes it more vulnerable to data traffic bottlenecks, collisions, and other problems. This is partially because Ethernet’s guts are based on Carrier Sense Multiple Access/Collision Detection (CSMA/CD), a contention-based access method that requires Ethernet nodes to listen before sending messages on the network.

And, despite recent standardizing efforts, users must still beware and shoulder most of the responsibility for researching, specifying, installing, configuring, and maintaining their industrial Ethernet networks. ‘Everyone talks about putting Ethernet on the factory floor, but most individuals really have not thought about how to actually do it, or they’ve simply ignored it,’ says Bob Lounsbury, project engineer for physical layers of core networks, Rockwell Automation (Cleveland, O.). ‘There was a big push because of Ethernet’s off-the-shelf cost advantages, but it’s an uphill battle to convince engineers to take this office network and put it in a high-noise, high-vibration, and high-temperature setting, and then expect it to work perfectly.

‘Ethernet has been going into industrial information and communication networks at 10 Mbps, but these applications could afford to be a little sloppy because they usually weren’t time critical. For example, if it takes 500 msec to update a data collection, monitoring, or other data network, it wouldn’t even be noticeable. However, a hiccup like this in a control network could shut down machines and critical applications, such as the motor control loops that maintain proper tension in papermaking machinery. There’s a huge desire to implement Ethernet, and it will happen, but it’s going to take some time.’

Goals, priorities, education

After assessing existing networks and deciding if industrial Ethernet might be able to help meet present and future business goals, users must evaluate and find the most appropriate type, amount, and method of using Ethernet in their specific application.

‘Users must prioritize their goals for each Ethernet project. They have to weigh how much web or Internet connectivity they want versus how much throughput they need,’ says Larry Komarek, automation systems business development manager, Phoenix Contact (Harrisburg, Pa.). ‘Infrastructure components for Ethernet are coming to market, so users must determine how much hardening, diagnostics, and speed they require. They also have consider the cost of support equipment.’

To teach engineers about applying industrial Ethernet, several organizations, companies, and integrators have launched training programs.

The Instrumentation, Systems and Automation (ISA) Society (Research Triangle Park, N.C.) is holding several courses nationwide through its ISA Training Institute, which have been presented by Eric Byres, P.E., research team leader, Advanced Information Technologies Group, British Columbia Institute of Technology (BCIT, Vancouver, B.C., Canada).

The Fieldbus Foundation (Austin, Tex.) offers training and components testing at its headquarters and at a new lab at the Southern Alberta Institute of Technology (SAIT, Calgary, Alberta, Canada). Also, Schneider Electric Automation (North Andover, Mass.) recently opened an Ethernet conformance lab at the University of Michigan (Ann Arbor, Mich.).

Physical layout of industrial Ethernet helpsreduce noise and helps network survive voltage transients

Appropriate design

Similar to any network, appropriate design is also essential for Ethernet. For example, attempting to control a large process or plant via a small network will likely create numerous bottlenecks and potential security problems. Mr. Lounsbury says a better design would be to assign most control work to an independent work cell, and only pass needed data back to the control room. Likewise, he adds sensible industrial Ethernet design should include the following elements:

Switches, instead of hubs. Ethernet switches provide a virtual connection, aid determinism, and keep network traffic to a minimum, which is preferable to hubs that repeat data to all network points;

Distributed control. This type of switch-based system uses a cellular architecture to build networks to perform different tasks;

Secure isolation between plant-floor and office networks. Using firewalls helps keep general traffic off control networks; and

Cable, connectors and equipment appropriate for the environment. Thorough research and testing by the industry is needed before deploying the typical RJ45 connectors on the factory floor. All Ethernet RJ45 connectors are not created alike. Some won’t survive vibration, and most aren’t sealed for industrial settings.

Standards, consistency sought

Ethernet’s main advantage, its historical popularity, may be an obstacle as it migrates to the plant-floor. There are so many Ethernet versions, flavors, specifications, and different solutions that interoperability may be nearly impossible.

‘What’s missing in industrial Ethernet today is consistent design standards or strategies. Many developers build and know their own solutions, but we need published standards on how to design proper industrial Ethernet networks. These are completely lacking now,’ says Mr. Byres. ‘An industrial Ethernet design standard should include the best of the Telecommunication Industry Association/Electrical Industry Association’s (TIA/EIA) existing 568 and 569 cabling standards for optimum Ethernet performance, maintenance, upgrading, and security. In fact, TIA/EIA’s 569 Committee’s TR 42.9 Working Group is seeking to modify 568 to meet industrial requirements.’

Good design, standards

Several standards projects are underway, including a snowballing effort by the Industrial Automation Open Networking Alliance (IAONA). [See this issue’s Market Update article.] Though it’s unlikely these efforts will produce agreement at Ethernet’s application layer, Mr. Byres says standardization will be possible at other layers.

‘It won’t be as bad as the fieldbus wars we were having. Because if everyone can agree on Ethernet, it will allow consistent design and cabling standards for running on the same cable at the same time,’ adds Mr. Byres. ‘We may not win the whole war for complete interoperability, but we will have won most of the battles, and be able to have a happy coexistence.’ He adds that common cabling, connectors, and addressing schemes will lead to consistent redundancy, and perhaps a common programming tool set in two or three years.

Speed = vulnerability

To provide 100-Mbps Ethernet, developers had to lower the amplitude and encode the data using a tertiary signaling method called MLT3 bit encoding, explains Mr. Lounsbury. The already low-amplitude voltage is split into three distinct levels, which allows it to run faster, while reducing radiated emissions. ‘However, this also means electrical noise bursts only have to be one-third of their previous level in relation to full voltage to corrupt data,’ says Mr. Lounsbury. ‘Control users need 100-Mbps Ethernet’s low turnaround times, but bitwidth at that speed is only 8 nsec. So, it only takes a short noise burst to damage an entire data packet.’

Users must also be cautious because, while 100 Mbps allows determinism, some Ethernet-enabling Transmission Control Protocol/Internet Protocol (TCP/IP) stacks in products are not optimized to allow that determinism. ‘Typical Ethernet switches and routers are optimized for delivering large data packets in an irregular pattern, but not in a constant, deterministic pattern,’ says Nick Jones, staff researcher, SST (Waterloo, Ontario, Canada). ‘Just because a switch supports 100 Mbps doesn’t mean it supports the real-time throughput that the data rate implies. In fact, some switches can only handle as little as 3% of the theoretical maximum throughput before experiencing significant determinism problems.’

Electrical noise

Because the capabilities of Ethernet components vary widely, users must closely examine specifications, especially for immunity to electrical noise. These include a component’s ability to reject electrical noise-usually defined as conducted interference from power supplies, electrical bursts and electromagnetic fields-without disrupting data packets and communications.

To be considered electrically hardened, network components must have a high common mode of rejection, according to Mr. Lounsbury. Ethernet is typically a differential communication system that uses two wires to transmit and two wires to receive data. These are configured as a twisted pair, either unshielded or shielded. Differential signaling helps increase immunity to electrical noise. ‘We find that high mode rejection on cabling and magnetics on the network interface card (NIC) can further enhance performance and reduce noise,’ he says.

Hardware is the path

To succeed in high-vibration, high-electrical noise, high-temperature, and frequently corrosive settings, Ethernet-based networks must also use components suitable for these environments.

‘Any imbalance due to poor cable or network construction in a common-mode noise environment will be translated into differential noise and result in higher error rates,’ says Mr. Lounsbury.

Manufacturers from several fields are seeking to solve these problems. Many are sealing traditional Ethernet cabling and connectors; others are building gateway solutions to traditional networking applications; some are devising PC-based solutions and on-chip devices; and a few are adapting less closely related technologies:

Rockwell Automation’s (Mayfield Heights, O.) third suite of Allen-Bradley products using the EtherNet/IP specification consist of 100-Mbps offerings in the ControlLogix, ProcessLogix, Flex I/O and PanelView lines. They’ll give users increased performance and bandwidth for Ethernet-based control applications, including time-critical applications, such as I/O devices and peer-to-peer interlocking.

JetLab from Jetter USA Inc. (Broadview Heights, O.) is a toolbox-based on its JetWeb control network components-that allows users to build software objects for passing data between machine and controllers via Ethernet and Jetter’s single, logical database.

Patent-pending RJ-Lnxx IP 67 industrial Ethernet cabling and connectors from Woodhead Connectivity (Northbrook, Ill.) combine a mini form factor with standard RJ-45 technology to seal out contaminants and protect against vibration and disconnects. RJ-Lnxx connectors are compatible with 10Base-T and 100Base-TX systems and can be used with commercial RJ-45 or industrial Ethernet components. Cordsets consist of Category 5 cable.

Pin-based, screw-on-coupling M12 connectors, traditionally used for sensors, were recently adapted by Lumberg Inc. (Midlothian, Va.) to aid Ethernet on the plant-floor. Lumberg’s ‘etherMate Interconnection System’ can communicate on shielded and unshielded cable according to 10Base-T and 100Base-T specifications, and its main components include RJ45-M12 bulkhead receptacle; ICAT5e braided and shielded cables; and PUR over-molded M12 connector cordsets for UTP and ScTP connection.

SST’s NetGateway is a headless-no moving parts-computer using a 200 MHz StrongARM processor with two Ethernet and four serial ports and support for two PC/104 interface cards. NetGateway helps users access plant-floor processes from office settings, while still keeping these networks separate.

Fast Ethernet Switch from InterlinkBT (Minneapolis, Minn.) provides eighteurofast 10/100 Mbps Ethernet ports; manages up to 1,000 media access control (MAC) addresses in its routing table; supports store-and-for- ward switching architecture; and has IP 67-rated connectors.

Users seeking to conserve slots in their PXI chassis can use National Instruments’ (Austin, Tex.) PXI-8212 GPIB and fast Ethernet interface for PXI/CompactPCI. Its Intel-based Ethernet controller chip is compatible with 10Base-T and 100Base-TX networks.

Fieldbus Coupler for Ethernet from Wago (Germantown, Wis.) uses Schneider Electric Automation’s Modbus on Ethernet protocol; uses hypertext transfer protocol (HTTP) to access the node and ‘boot-p’ to configure it; and connects to the Wago I/O System.

Opto22’s (Temecula, Calif.) Snap Ethernet I/O system is able to communicate with multiple devices simultaneously with different protocols for each device by using the multiple sessions allowed by its TCP/IP suite over its Ethernet physical layer.

Used to connect distant devices and an Ethernet network, RT1-TP/FL (10Base-X) and RT2-TX/FX (100Base-X) rail transceivers from Hirschmann’s Automated Network Solutions Group (Pine Brook, N.J.) have one twisted pair and one fiberoptic port for connected devices up to 40 km away.

Reaching down

In the future, more Ethernet switches and chip-based solutions will likely become fast and cheap enough for widespread Ethernet access to the device level, instead of classic I/O clusters reporting to an Ethernet network via a gateway.

Also, though Ethernet isn’t intrinsically safe yet, ISA’s MLT committee is examining how to develop intrinsically safe fiber-optic cable for use in Ethernet networks.

‘We’re probably two to five years away from a mainstream move by Ethernet TCP/IP to the fieldbus level, and about five years away from overcoming the connector and pricing issues for Ethernet to reach the device level,’ says Mark Fondl, president, Integrated Communication technologies (Newburyport, Mass.), and IAONA board member.

For more suppliers, go to www.controleng. com/buyersguide ; for more information, use the following circle numbers, online at www.controleng.com/freeinfo . Automationdirect.com wwwautomationdirecto.com 237

British Columbia Institute of Technology www.bcit.ca 238

Fieldbus Foundation. www.fieldbus.org 239

FieldServer Technologies. www.fieldserver.com 240

Hirschmann. www.hirschmann-usa.com 242

InterlinkBT. www.interlinkbt.com 243

Jetter USA. www.jetweb-technology.com 244

Lumberg. www.lumbergusa.com 245

National Instruments. www.ni.com 246

Opto22. www.opto22.com 247

Phoenix Contact. www.phoenixcon.com 248

QSI. www.qsicorp.com 249

Rockwell Automation www.automation.rockwell.com 250

Schneider Electric Automation www.schneiderautomation.com 251

Siemens Energy & Automation www.sea.siemens.com 252

Southern Alberta Institute of Technology www.sait.ab.ca 253

SST. www.mysst.com 254

VMIC. www.vmic.com 255

Wago. www.wago.com 256

Woodhead Connectivity www.woodheadconnectivity.com 257

Ethernet, OSM keep polyester, polypropylene process running

When transceivers failed and shut down Reemay Inc.’s (Old Hickory, Tenn.) former network, its machines kept running, but operators couldn’t make any changes or make other products. Reemay makes non-woven, spunbonded, polyester and polypropylene products, such as housewrap, carpet backing, fabric softener sheets, furniture and aerospace construction components, apparel interlinings, crop covers, air filters, and other products.

To prevent network crashes, resulting $15,000 per hour downtime, and 3 a.m. calls summoning him to the plant, engineer-in-charge George Reed opted for an industrial Ethernet-enabled Optical Switch Module (OSM) from Siemens Energy & Automation (Alpharetta, Ga.). OSM provides network redundancy, and notifies all network components when the switch is communicating.

Now, when a network line break occurs, OSM’s redundancy manager continues to operate the network OSM also e-mails Mr. Reed that the manager was activated, while allowing Reemay’s operators to continue running its process.