Ethernet focuses on floor, control
Harsh industrial noise levels and need for mission critical operations spur creation of infrastructure standards for performance and installation
By Robert Lounsbury
Ethernet emerged more than 20 years ago as a commercial network providing office connectivity to desktop computers and printers.
Office environments are relatively benign with respect to electrical noises. Applying the same network into the industrial environment presents the designer and installer with some challenges in obtaining the same or better performance.
Machine control de-mands reliable data transport to insure maximum network avail- ability and uptime.
By applying component enhancements, isolation, and separation correctly, top performance is possible and at a minimal cost.
Different Ethernet networks
There are three Ethernet media: co-axial, twisted pair, and fiber. The most popular is twisted pair (TP). The coaxial-based networks supported a maximum of 10MB/s and were limited to 200 meters (10Base2) and 500 meters (10Base5).
Today’s generic networks support 10 MB/s, 100 MB/s, and 1 GB/s and soon will support 10 GB/s. These new high-speed networks will primarily use the twisted pair (UTP and STP) and fiber as their physical layer mediums.
The emerging industrial Ethernet networks are following this concept of TP and fiber. They will primarily use 10MB/s and 100MB/s. The use of coaxial networks has declined due to the connection costs and their installation complexities.
Today’s industrial networks focus on using Ethernet as a control network. However, these new networks will not supplant the existing fieldbus networks in the industrial space.
There are several Industrial Ethernet networks defined by various consortia’s including EtherNet/IP, Modbus-IDA (Interface for Distributed Automation), and ProfiNet.
Not your office network
IT personnel have been providing network services for computing connectivity for many years in the front office. The need for simple reliable and inexpensive connectivity has created an interest in moving these services into the industrial areas.
However, there is no comparison of the environment in the front office to the factory floor. In fact, the noise levels in the harsh industrial environment are at least a magnitude (10X) greater than that of the office environment.
Further, networks deployed for factory automation applications are “mission critical” requiring high reliability infrastructures. All of this has spurred the recent creation of the industrial infrastructure standards for performance and installation in both ANSI/TIA and ISO/IEC.
In addition, many industrial controls companies have been providing Ethernet enabled controls equipment. This new breed of PLCs and IO is incorporating Ethernet as a control and information network.
Machine control traffic is dependent on unrestricted access to the network. Most of these products have web interfaces to help with configuration and maintenance.
The reliability of the network on the factory floor is mission critical. To make my point, I frequently make this statement: “If my computer cannot connect to the network in the office, I may not be able to print or access my e-mail. The cost to the company is minor, and there is no product waste.”
If the control network fails in the factory, many workers are idled, machines cannot run, and production stops. The costs can be in the hundreds of thousands of dollars per hour in lost productivity and revenues not counting scrap.
It does not take many network failures before a company begins to feel the financial losses, potentially leading to bankruptcy. A designer should keep this in mind when purchasing components and products for industrial applications.
Ask your self this question, “Would you bet your company on a $9 NIC card from your local computer store?”
The noises in the industrial environment are at least a magnitude greater than the office environment. Industrial controls engineers are expecting the same performance or better than that of the performance in the front office.
For example, IEEE 802.3 requires devices to meet a certain level of performance with 1V of noise, whereas industrial standards typically require the same performance in the presence of 10V of noise.
Given this, the network components must have greater noise immunity and/or be immune to high EMI within the environment. There is a further level of noise mitigation for Mechanical, Ingress, Chemical/ Climatic, and Electro-magnetic (MICE).
Most, industrial Ethernet networks use a combination of communication stack optimizations and band- width to achieve pseudo deterministic performance. Some customize the hardware to achieve deterministic behavior. Custom hardware raises questions regarding open standards, availability, and breadth of selection.
Deciding which network
Industrial control applications require topologies that generic standards do not support.
For example, a simple point-to-point connection between a switch and a PLC may be required, whereas the generic standards require the channel to contain a fixed horizontal cable with a work area cord.
These requirements are too complex and in many cases needless. Therefore, as part of the decision process, one must consider the needs of their application and which standard they should support.
In general, I recommend, whenever possible, one should follow the generic standards. These standards will provide the best success of meeting the requirements for standard IT applications such as voice, video, and data.
Here are two drawings that describe the cabling infrastructure in TIA terms. The hierarchical drawing begins at the campus distributor and ends at the intermediate distributor. It shows how a system can grow from two buildings (BD) horizontally to cover many buildings. Additionally, each floor (FD) will have at least one FD and possibly many intermediate distributors (ID), depending on the square footage of the floor coverage area. The hierarchical configuration provides redundancy at each level through the doted line connections.
Another way to look at this is to flatten out the view. This drawing shows how the automation island connects to the generic telecommunications infrastructure.
In the previous hierarchical scheme, the system ends at the telecommunications outlet (TO). The generic infrastructure ends at TO/AO (automation outlet).
If the TO is replaced with an AO, then the network may no longer be compliant to the generic standards.
This is true for other parts of the network where deviations exist and is permissible as long as performance is the intent and the designer is aware of the ramifications.
In most applications, more than one connection is necessary. In this case, the network will connect to the ID without a TO or AO or the ID will be move inside the automation island.
Connectivity from one automation island to another will be direct or go through the ID. In some cases, the ID may move into the automation island or there may be a need to support many different applications in the automation island such as voice data and video (VDV).
When the machine distributor is in a harsher environment like a MICE 2 or 3, its construction may be different. To distinguish between a distributor for normal MICE 1 environments and one of the industrial environments, there can also be a machine distributor. Technically the generic cabling ends at the AO if it is replacing the TO.
Inside automation isles
Here are a number of Ethernet based control and information network connectivity examples happening within automation islands.
At times, there is an implementation of a generic cabling system inside the automation island. There are restrictions placed on the cabling system so the cabling will support all of the “supported” applications (VDV).
In a compliant generic system, devices must connect through a work area cord from a TO. Between the cross connect and the TO/AO, there must be horizontal cabling that forms the fixed part of the cabling system. The TO/AO must be placed within 5 meters reach of the work area cord.
The distance between the cross-connect and the TO/AO is limited to 90 meters maximum. In a generic cabling system, the horizontal cabling must be of solid copper conductors.
In an industrial cabling system, the horizontal cabling may be of stranded cabling. This is largely due to the expected movement and flexing of the cabling components.
If your cabling infrastructure or machine cabling is only supporting control applications, then generic cabling methods may be too restrictive.
The international standards for industrial Ethernet are coming together so as to provide definitions and guidance up to the machine area or automation island.
If the network is providing connectivity to one of the fieldbuses within the automation island, then an AO should replace the TO. In this case, the incoming AO will have a specific definition per the flavor of the fieldbus.
Here are some point-to-point examples in a simple configuration. There are several tradeoffs in point-to-point topologies, which may not make any difference for industrial networks.
Those tradeoffs would be:
Future network re-configurations may be more difficult.
Network may not support all applications such as 1Gb/s, POE, or video. This is especially true if using two-pair cabling.
The benefits may far outweigh the loss in flexibility. Machine wiring is usually more permanent than office connections (less moves, add-ons, and changes).
Generally, the more connections that are in the channel, the greater the chance of a failure.
In short, flexibility comes at a cost that may be unnecessary to your application.
For the sake of differentiation, here is a name for each cord. The following attributes will help to give a better understanding. These are only examples of typical attributes, and the actual cables may need to be different.
Work area cord is usually in an enclosure and only requires protection to IP 20. In most cases, a standard off-the-shelf cord will work.
Industrial path cord represents the horizontal segment of a generic system and need to work in harsh environments. The cord will have plugs at each end, usually sealed. The cable is stranded conductors and will have balanced characteristics needed for high noise immunity.
Industrial work area cord is most likely on machines themselves or in the direct vicinity of machines. This cord is of stranded conductors. The jacket will provide protection against weld splatter, oil, and the like. The pairs will have the balanced characteristics needed for high noise immunity.
Keep in mind some industrial control network suppliers may require adherence to their standard to guarantee performance. Conveniently, at a minimum, most industrial control network standards are extensions to the existing generic and industrial standards.
ABOUT THE AUTHOR
Robert Lounsbury (flyingin98@road runner.com) is a member of the USTAG (US Technical Advisory Group) for IEC SC65C, Digital Communications with ISA acting as the group’s administrator. He is a technical researcher, standards writer, and speaker who has been working in the industrial networking field for the past 22 years. His book, Industrial Ethernet on the Plant Floor: A Planning and Installation Guide, is coming out early 2008.
NIC card is a network interface “Ethernet adapter” that is inside the back panel of the computer and connects to the cable or DSL modem.
ANSI: The American National Standards Institute is a non-governmental organization responsible for the development of voluntary manufacturing standards.
TIA: Telecommunications Industries Association
ISO/IEC: International Standard Organization/International Electric/Electrotechnical Commission
IP 20: The protection classification offered by an enclosure is shown by the letter IP (Ingress Protection) and two digits. The first digit indicates two factors: 1 means protection for persons, and 2 means protection for equipment. The second digit indicates the protection against water with zero meaning no protection.