Control system technology is migrating toward an open concept where networks accommodate factory floor devices from different manufacturers, and control intelligence is distributed among these devices. Two new open networks — the Smart Distributed System (SDS) from Honeywell’s Micro Switch Div. and DeviceNet from Allen-Bradley Co. — offer a giant step forward in this move toward distributed machine control.
From a broad technological standpoint, these communication networks or buses share several common features. They:
• Communicate over a single cable into which all devices connect, Figure 1. Plug-in connections reduce wiring and make it easier to expand or modify the system.
• Let factory floor devices communicate with each other even though they come from various manufacturers.
• Encourage other vendors to develop devices that are compatible with the network, such as intelligent sensors and actuators.
• Are both based on a communication technology called Controller Area Network (CAN).
The two companies took different paths, however, in developing these networks. Honeywell launched a threephase program about 21/2 years ago. In Phase I, they established a network concept for connecting sensors, limit switches, and actuators. This was followed by development of PLC interfaces, sensors with CAN microprocessor chips to provide communication capabilities, and interfaces for sensors and other devices without CAN chips. Then, they initiated Beta sites to field test and refine the system.
Phase II, started within the past year, added smart sensors and actuators with advanced functions and diagnostic capabilities. Phase III, scheduled for next year, will expand system control and diagnostics. Throughout this program, Honeywell developed most of the hardware in-house, though partner companies also participated in the development.
Allen-Bradley, on the other hand, began about 1½ years ago to develop a network that links a wide variety of factory floor devices — including drives and starters as well as sensors and actuators — with each other and with PCs and PLCs. The company is developing some compatible sensors and drives in-house. But, their major thrust is to encourage other vendors to develop compatible devices within their respective areas of expertise. Numerous vendors have started this development effort.
Both of these systems have the potential to become a de facto device network standard. End users need to carefully evaluate each system and determine which is best for their applications. In the meantime, this overview will help you understand how the two systems work.
SMART DISTRIBUTED SYSTEM (SDS)
The SDS network, or bus, lets intelligent factory floor devices communicate with each other, and with PLCs or PCs, regardless of who manufactured the devices. Up to 64 sensors and other devices are linked over a single cable that carries both data and control power.
Unlike conventional 2-wire systems, this digital network provides system and device diagnostics through CAN chips embedded in PCs, PLCs, and other control components. Diagnostic capabilities cut installation and repair time, and aid troubleshooting.
Honeywell selected CAN because of its cost, robustness, and speed. “CAN is a proven communications technology” says Jeff Beal, SDS Marketing Manager. “Other systems are too slow for discrete manufacturing, where real time is measured in msec.”
To date, Honeywell has developed nearly all of the intelligent devices that communicate over the network. But, the system allows other vendors to develop and manufacture compatible devices, a feature that will expand the end user’s choice of I/O devices and controllers (PCs and PLCs). Accordingly, partner companies are developing devices such as valves, motor starters, and analog I/O.
Continue on page 2
The SDS network consists of a control interface, smart sensors, output devices, wiring, and an operator interface, Figure 2.
A control interface, called an Interface Terminal Strip (ITS), connects PCs or PLCs to the network. It converts serial data from the 4-wire bus into parallel inputs and outputs that controllers can process. Some companies are developing built-in interfaces for their own PLCs (see Partners section).
Smart sensors communicate with the network in one of two ways. First, conventional sensors and switches can tap in to the network through a Multiport Input Device, Figure 3.
Second, sensors can be built with CAN chips that let them communicate over the network. Honeywell has developed photoelectric sensors, proximity sensors, and limit switches with chips that provide diagnostic information and perform other functions such as on-delays and batch counting.
Adding intelligence to these sensors lets them send more detailed information, such as status, setup parameters, and operating condition, to a controller. Thus, a photoelectric sensor may report when its gain drops off — because of a dirty lens or being knocked out of alignment with its reflector. A mechanical limit switch can be programmed to report when it has been actuated a specified number of times as an aid to predictive maintenance.
Output devices (actuators) initiate action in response to instructions from a PC or PLC. A Multiport Output Device from Honeywell lets conventional actuators, such as contactors, solenoid valves, and simple motor starters, communicate over the network even though they come from different manufacturers.
Wiring connects components to the network. A single cable replaces up to 64 wires, reducing cost and installation time. Quick-disconnect couplings simplify adding or deleting devices and making repairs.
A hand-held operator interface (termed Activator) accesses any device on the network, so a user can program its setup parameters (normally open or closed), plus time delays and number of operating cycles. The Activator also interrogates each device to check its identity, status, and age.
Several SDS partners are incorporating network interfaces in their devices. For example, GE Fanuc developed an interface card that lets a Series 90-30 PLC connect to the network without a separate interface. FACTS Engineering is developing a similar interface for Siemens Simatic TI405 PLCs.
Any company that wants to develop compatible devices for the network can obtain a specification, which includes SDS application and physical layers for the CAN chip, at no charge and without license. The company then installs the two layers on a chip purchased from a vendor such as Motorola, Philips, Intel, or NEC, and imbeds the chip into its device. Optionally, a company can buy chips with the application and physical layers already installed.
Beta site testing
SDS has been installed at several Beta sites, both at Honeywell and in external manufacturing plants, for bottling, canning, conveying, packaging, and assembly line applications. External companies include some in the beverage, package distribution, and automotive industries. Here are some examples:
• Honeywell’s Freeport, Ill. facility includes four networked sensor assembly lines and a shipping conveyor, installed in mid-1992. Each assembly line has 160 inputs — proximity sensors and limit switches with CAN chips — and 160 outputs, all controlled by PLCs (TI Siemens and GE Fanuc) or IBM-compatible PCs. By eliminating I/O connections and wiring to individual devices, the network saved 150 hr in design and 284 hr in installation, and it reduced wiring cost by 15% compared to a conventional system.
The conveyor handles parts of various weights, ranging from ounces up to 20-lb, which requires sensors every 2 ft along an 80-ft span. With this many sensors, conventional wiring would not have been cost-effective. But the network’s single cable and quick-disconnect couplings saved 840 hr in installation and 17% in wiring cost.
Continue on page 3
• One external site is the General Motors Holden Engine Co. in Victoria, Australia. In October 1993, they applied SDS to a PLC-controlled overhead conveyor with 20 proximity sensors and bar code readers tied to the network through ITS units, plus 20 pneumatic valve actuators.
Sensors read a bar code on each motor to determine the type of vehicle in which it should be installed. The network eliminates the need to climb into dangerous areas to troubleshoot the sensors. System diagnostics tell operators when and where a problem exists, thereby reducing downtime caused by engines being sent to the wrong area.
This communication network is designed to connect various factory floor devices, such as photoelectric and proximity sensors, limit switches, bar code scanners, valve manifolds, motor starters, smart motor controllers, and adjustable- speed drives, to a control system without interfacing through I/O modules. These devices can communicate with each other (peer-to-peer) and with higher-level controllers (master-to-slave).
A single Belden cable connects plant floor devices to controllers, replacing individual wires from each device to the I/O. This means less wire and reduced cost. The twisted-pair cable, containing both signal and power lines, is arranged in a trunk-line, drop-line configuration that supports 64 nodes. Though nodes are typically sensors, they can also be intelligent blocks, each supporting 32 devices for a total of up to 2,048 devices on the network.
Quick-disconnect couplings simplify tear down and set up. And a user can remove one node without interrupting others.
The network lets operators obtain status information from smart factory floor devices so that deteriorating conditions are quickly detected and remedied. Networking makes it easy to set device parameters such as light or dark sensing in a photoelectric sensor or motor speed in a drive.
According to Bill Little, Allen-Bradley senior vice president and chief technical officer, DeviceNet “helps provide more flexibility, agility, and productivity by increasing the amount and rate of information flowing from plant floor devices to control systems.”
Allen-Bradley developed the network to meet the needs of both OEMs, who want reduced wiring costs, and end users, who want long life and predictive maintenance capabilities. Moreover, the network allows users to choose devices from a variety of vendors.
Partners develop components
To give the network a good start, Allen- Bradley is developing compatible photoelectric sensors, Figure 4, plus motor starters, drives, operator interfaces, bar code readers, and programmable controller interfaces.
But the company realizes it can’t develop every type of device in-house. So they encourage other companies (including competitors) to develop products in their own areas of expertise. To date, 50 companies have agreed to develop products for use with DeviceNet. Among them are market leaders in proximity sensors, controls, drives, valves, and motor starters. Here’s a sample of those companies and the devices being developed:
• Furnas Electric — ac drives and solid state starters.
• Modicon Inc.— programmable controllers.
• Moog (Germany) — servo controllers and drives.
• Namco Controls Corp. — sensors and sensor manifolds.
• Pro-Log Corp. — multi-axis motion controllers.
• Reliance Electric — ac vector drives and dc digital drives.
Sensor manufacturers, including Banner Engineering Corp., Pepperl+Fuchs Inc., and Turck Inc., will develop compatible proximity, photoelectric, and ultrasonic sensors. Several pneumatic valve manufacturers will also participate.
Other companies will develop software for device configuration, analysis of data traffic, diagnostics, and man-machineinterface. Horner Electric will develop an interface for GE Fanuc PLCs.
The network specification is open and available, without license, to any company interested in developing compatible products for the network. Interested companies need only to purchase Allen- Bradley’s network specification ($95 for specification plus updates), which includes application and physical layers for the chip, plus the CAN chips from one of the chip suppliers. They can also attend network technical sessions and obtain development help from companies such as the Dearborn Group, Huron Networks Inc., Online Development Inc., ProSoft, and S-S Technologies Inc.
Both SDS and DeviceNet are based on Control Area Network (CAN), a communications technology developed by Robert Bosch Gmbh of Germany in the mid 1980’s for smart systems in automobiles. CAN is also used as a factory-floor bus in Europe.
Located on a microprocessor chip, CAN is a communications protocol that defines rules for exchanging messages between devices on a network. It is organized similar to the ISO layer model (which defines the structure of communications networks) in that each layer performs a specific function. As used with SDS and DeviceNet, the CAN structure essentially consists of three layers: application, data link, and physical.
The basic chip contains the data link layer, which controls access to, and transmission over the network. It features variable- length messages, nondestructive arbitration, automatic error detection, and a data identifier.
A user controls message length by selecting the types of information to be communicated: error messages, batch counts, timer functions, or preventive maintenance data (number of operating cycles). In this way, the user can trade off transmission time for message length, which ranges up to 8 bytes.
Nondestructive arbitration occurs when two devices try to communicate over the network simultaneously. Normally, this causes a “collision” and both devices must try again. With the CAN signaling method, however, the message with the highest priority is allowed to transmit first while the other waits.
Error detection lets any device on the network detect incoming transmission errors and signal the transmitting device to try again.
A CAN identifier lets a device identify data in messages and accept only those mesages in its area of interest.
Other capabilities include peer-to-peer and multi-cast reception, in which one device broadcasts a message to several others simultaneously. For example, a controller could tell all four brakes on a vehicle to apply at the same time.
Not all CAN chips are the same. Available from Motorola, Philips, Intel, and NEC, they vary in capabilities depending on what features are needed. Also, they come in both a standalone version, which requires a separate microprocessor chip to interpret and process data, and an integrated version, where both CAN and microprocessor are embedded in one chip. The smallest embedded chip — about ½-in. wide, has been installed in an 18-mm. diam proximity sensor.
Honeywell and Allen-Bradley have both developed custom application and physical layers for the SDS and DeviceNet networks, respectively.
In general, the application layer defines information such as services available to the user, types of data that can be transmitted, and methods for reporting process variables. It also provides for both master- toslave and peer-topeer messages.
The physical layer specifies the communication method, which calls for a single cable for data communications and delivering power. It also specifies communication speeds. For SDS, messages are transmitted at rates ranging from 125 Kbit/sec. to 1 Mbit/sec. for distances to 1,500 ft, depending on the number of devices on the bus. DeviceNet features data rates of 125, 250, and 500 Kbit/sec. at distances to 1,640 ft, with no restrictions on the number of logical nodes.
Filling a niche in the network spectrum
Both SDS and DeviceNet are intended to fill a void in the spectrum of networks between simple low-level sensor networks and complex high-level fieldbus networks.
Low-level networks, such as ASI and Seriplex, connect simple On/Off devices (sensors) primarily for discrete applications. Though economical, they are limited to transmitting small amounts of data at a time, 1 byte or less.
The high end of the spectrum includes fieldbus networks, such as ISP, SP50, Profibus, and WorldFIP. Focusing on process applications, they connect complex devices such as flowmeters, pressure sensors, temperature controllers, and servo valves. These networks operate at baud rates of 1 to 2.5 Mbits/sec, and transmit up to 256 bytes of data at a time. Fieldbus networks are more expensive, as illustrated by a typical chip cost of about $75 compared with $3 to $35 for a CAN chip.
By comparison, CAN-based networks can transmit up to 8 bytes of data at a time.
Though not yet tested in factory applications, DeviceNet is designed to fill the need between these two levels for an open network that connects various discrete devices for real-time operation.
On the other hand, SDS is already established as a viable network through Beta-site testing and is proceeding toward a network that accommodates a larger variety of devices.