Leveraging Internet protocol without breaking the bank
BY MICHAEL TENNEFOSS
When users talk industrial Ethernet, hit the key points: economy, value, and necessity.
In the beginning, there was cabling—lots and lots of cabling—and there was centralized control. Every sensor, every actuator, every display was connected by a separate cable that grew like ivy from a programmable logic controller (PLC), the brains of a traditional control system.
If a solenoid needed to be triggered in response to the activation of a limit switch, then the signal traveled from the limit switch through cabling to the PLC, which processed the information and sent a command to the solenoid over yet another cable. After testing the cabling, programming the PLC code, and commissioning the system, one was up and running—until something broke or a new function was required, that is.
If the PLC controller failed, if we needed a new actuator, or if the system needed to interface to a different control system, then problems arose.
Because we were dealing with a centralized control system, nothing could operate autonomously—when the PLC died, the entire system died.
If the system needed a new actuator, new cabling had to be installed and the PLC reprogrammed. If two different control systems had to be interconnected, then a custom serial gateway or intermediate interface logic was required.
There had to be a better way to build a control system.
ANSI 709.1 is an intelligent, distributed networking infrastructure. ANSI 709.1 is not a remote I/O scheme intended to extend the range of a PLC but rather an approach to network infrastructure that overcomes the fundamental limitations of PLCs and control networking.
This control network allows devices to communicate directly with one another on a peer-to-peer basis, without the need for a PLC or central controller. Each device is locally intelligent and able to communicate on a shared communication medium with any other device on the network. New devices need only connect to the shared medium, minimizing cabling changes.
A software download via network cabling reconfigures system functionality. Peer-to-peer communications allow the direct exchange of information among any or all of the devices without intervention by any central device, eliminating the single point of failure issue endemic to PLC-based systems.
If devices need central monitoring, a master/slave scheme can deterministically poll them.
ANSI 709.1 networks came into being as application control networks, not data networks. They are low cost, robust, and field proven. More than 16 million ANSI 709.1-compliant nodes are in service for pump and valve control, robotics, mass flow control, vacuum deposition, clean room air handlers/scrubbers, gas delivery and monitoring systems, welding, paint and adhesive application, measurement and instrumentation, SCADA systems, and related industrial applications.
With the spread of IP connectivity came the opportunity to carry ANSI 709.1 packets over TCP/IP-based networks. TCP/IP networks already installed for data and enterprise resource applications are ideal for tunneling control network traffic, with the proviso that the tunneling router adhere to the limits and limitations of the data network.
ANSI 709, combined with TCP/IP, provides an end-to-end, open, interoperable infrastructure for networking control devices over a variety of media, using robust physical layers and a comprehensive network management scheme, and meeting industrial EMC and environmental specifications.
These capabilities are responsible in part for the worldwide acceptance of ANSI 709.1 in a wide variety of applications and industries. The low cost of this technology—less than $10 for a twisted-pair node in reasonable volume-has helped, too.
ALLURING CONCEPT IS TCPAs TCP/IP-based networks have proliferated in the data networking world, the control industry has eyed with interest the high-speed, routable data technology used in TCP/IP local- and wide-area networks. This has begged the question: Why not adapt IP networking to sensor and actuator control?
The concept is alluring for many reasons. Ethernet has pervaded the data world and made its way into a wide variety of devices, many of which are low cost yet offer reliable performance in their intended applications.
ANOTHER ANGLEEqually important, TCP/IP has become a worldwide standard in the data world, just as Ethernet has become a ubiquitous medium. By leveraging these two standards and integrating them into control networking applications, it might be possible to leverage existing products, tools, and technology while proclaiming the use of an open standard.
Why not, indeed, jump on this bandwagon?
There are serious issues associated with the value, necessity, and economy of an all-Ethernet control system-issues generally glossed over in the enthusiasm that ensues whenever the subject of "industrial Ethernet" arises.
While TCP/IP is a standard in the data world, it is not a control protocol. If TCP/IP operates as open transport, one must still provide or develop an open control protocol. The temptation is for vendors to define a new control protocol riding on the coattails of the open TCP/IP standard, proclaiming the overall approach "open" because TCP/IP is open when in fact the control protocol itself is closed and proprietary.
The goals of vendors taking such an approach are threefold: first, to maintain a closed sole-source solution under the guise of an otherwise open standard; second, to extend the life of proprietary PLC or software-based PLC systems by using Ethernet-based devices as remote I/O for existing products; and third, to cause customers to migrate from low-cost sensors and actuators to higher-priced devices based on a more expensive, but no more capable, Ethernet channel.
Whatever the motivation, creating a new control protocol and implementing it in an open standard requires substantial and unnecessary development work, given that ANSI 709.1 is already an open standard and can already be used on a peer-to-peer and master/slave basis over TCP/IP networks.
Ethernet lacks the robust physical layers offered by ANSI 709.2 and ANSI 709.3. 10Base-T and 100Base-T systems will not meet European susceptibility limits except at high cost, while still being vulnerable to single points of failure in the point-to-point wiring and hubs.
While much work is under way to create industrial RJ-45 connectors, the resulting products are very expensive and cumbersome to install, relative to an ANSI 709.3 unshielded twisted pair.
Moreover, Ethernet does not support multidrop signaling or powered twisted-pair networks, placing severe restrictions on cable topologies and requiring the use of separate power wiring that ANSI 709.3 does not mandate.
Ethernet's original intended use-corporate data networks-represents relatively benign electrical environments. Electrostatic discharge (ESD), burst noise, electrical surges, radiated noise, magnetic fields, and so on manifest themselves at fairly low levels. Contrast this with a typical factory floor, which is a virtual firestorm of electrical activity.
Making an industrial Ethernet network work in such an environment requires the addition of surge suppressors, ESD protection devices, and a host of other protective mechanisms that ANSI 709.2 and ANSI 709.3 simply do not require.
MANAGE THE LION'S SHAREAs to necessity, the lion's share of industrial controls can be managed with 50+ millisecond (msec) response times typical of ANSI 709.1, while most <10 msec applications cannot accept the nondeterministic latency of an Ethernet connection. That makes the sweet spot of the industrial Ethernet market nodes requiring response times between 10–50 msec.
This is a niche market only.
Consider, finally, the matter of economy. A node employing the ANSI 709.1 protocol and ANSI 709.3 physical layer can typically build for less than $10 in reasonable volumes. Signal connections are simple: The node requires only an unshielded twisted pair. Electromagnetic compatibility (EMC) and regulatory compliance is easier because the physical layer inherently handles severe industrial environments.
A comparable node using industrial Ethernet will require Category 5 or better cabling, an industrial connector capable of terminating four twisted pairs, common mode and EMC protection compliant with high industrial EN 1000-4-x levels, a microprocessor capable of running both the TCP/IP and control network protocol stacks, and a power supply and separate power cabling capable of delivering the required voltage and current.
All this for a limit switch? A photoelectric detector? A conveyor speed sensor? A level sensor? The economics of using industrial Ethernet are simply unattractive for the vast majority of sensors, actuators, displays, and controllers one finds in an industrial plant. ANSI 709 provides an exceptionally cost-effective, robust, high-performance networking infrastructure for these devices, yet it also allows the use of standard TCP/IP networks—no new protocols required—as data tunnels when such a mode of signaling is required.
Together ANSI 709, Ethernet, and TCP/IP offer the most economical solution available for industrial control and represent the best industrial Ethernet solution available without requiring new development and invention. IT
Behind the byline
Michael Tennefoss has degrees from the University of California and the University of Minnesota. He is a vice president of Echelon Corp., a control networking company in California.