Ethernet, a digital network, is spreading from PLC and PC applications down to the motion level
Nearly every automation or motion device is available in a digital version. And nearly all devices that can connect to a digital network do so. Yet, most of today's motion controllers still rely on what is arguably an out-of-date technology (analog) to link to servo or step drives. As late as 1997, proprietary and analog interfaces dominated in the drive market with a 90% share, according to a report from Automation Research Corp. But things are changing.
Driving the change
The usual design approach for a multi-axis motion system is a PCbased controller connected through discrete wiring to a number of drives and inputs and outputs. The drives' command interfaces are either analog signals for servos or pulse-type signals (indexing) for steppers. Many times the drives will also have a serial interface, usually RS-232-based, to connect to a terminal device for programming or diagnostics.
But as intelligence distributes farther into manufacturing and devices of all kinds migrate onto networks, this tried-and-true approach will fall behind users' technical requirements. No where will this be more apparent than in industrial automation applications.
As a result, motion control must evolve from proprietary command and control schemes to an open networking environment where intelligence moves downstream to linked devices. There are several forces converging to push motion control along this path. The most important of these are:
Intelligent drives. Drives are no longer subservient to motion controllers. They are evolving from simple power devices with basic logic into mostly digital, "smart" devices. Advanced microprocessors and digital signal processors (DSPs) enable them to perform many functions previously executed on higher-end PLCs, CNCs, and general motion controllers - functions such as elaborate trajectory generation.
Just as significant, these new drives can operate independently in applications that don't require close coordination among multiple axes, which is the majority of motion applications.
Emerging motion networks. Except for Sercos and a few other protocols, there are no industrial networks that serve all motion requirements. The available choices, such as DeviceNet, Profibus-DP, and Interbus- S, can handle some motion control tasks. But these protocols originated in the PLC world to link inputs and outputs with sensors. Here, speed of response does not begin to match that needed for motion control.
Despite the limitations, though, engineers are successfully integrating device networks into systems to control motion. They're just not used to control coordinated multi-axis applications.
Such implementations do gain benefits, however. Wiring is simpler and less costly, installation and set-up are quicker, and integrators can query linked devices for data they can then use to improve operation or diagnose problems.
Information glut. Along with these forces, the one with perhaps the most profound influence has been the Internet. Because of its explosive growth, information technology is undergoing unprecedented change.
The foundations for this were put in place two decades ago, starting with the invention of Ethernet in the 1970's. The PC arrived in the early 1980's and by 1983, TCP/IP was incorporated into version 4.2 of the freely-distributed BSD Unix kernel. These milestones, though seemingly not related at the time, combined to set the stage for the "network age."
Ethernet and TCP/IP are already commonly used for communication among PLCs and other controllers. Now, this network system is linking the factory to management levels and the rest of the enterprise. Given the ever-lower costs of PC networking technology, engineers would be foolish to ignore commercial network solutions in motion control.
The Ethernet connection
So then, will every drive be networked? Will motion control be an open-system technology? If you view motion control devices and systems as embedded devices in the total fabric of industrial automation, then it follows that the future of motion control will be open and networked.
However, just as the protocols for the Internet and LANs converged for open networking of PCs, so too will the protocols for the larger market of embedded devices, including motion control systems. Already, Ethernetbased I/O is appearing on the market, and several motion controllers now come equipped with this protocol for connection to a host.
Ethernet may provide the migration path to open networking in industrial automation, and even motion control. Engineers already ask about motion solutions based on it, though they have concerns about the ability of a front-office LAN technology to give them the performance they seek in demanding automation applications. Nonetheless, it can no longer be dismissed as an inadequate industrial network because it lacks determinism or has communication collisions. Ethernet will be important for several reasons. Among them:
Low Cost. Even though some designers feel it is too expensive for device- level networking, the current cost of its chips is on par or lower than that of other protocols' chips. In addition, new chip sources continue to emerge. The NET+ARM chips from NETsilicon contain perhaps the first networking system-on-silicon. These chips have all the critical hardware and software necessary to make any embedded device network ready. And because of multi-vendor participation and the volume of the PC market, the costs will be driven to the point of inconsequence.
High Speed. There are no Ethernet cards rated at less than 10 Mbaud, a level that many industrial networks can't yet match. Moreover, the PC world is rapidly making 10/100base versions with 100 Mbaud capability the network standard. And although not needed at this point, Gigabit Ethernet is waiting in the wings.
Robust. To resolve conflicts among devices attempting to access the network, Ethernet uses carrier sense multiple access with collision detection (CSMA/CD) with the transmission and error checking features of TCP/IP to ensure error-free messaging. Also it's virtually immune to EMI noise. This was demonstrated years ago with tests conducted by Digital Equipment at several manufacturing sites. Therefore, harsh industrial environments will not bother it.
Determinism. The most frequently voiced concern about Ethernet is that it's not deterministic, that it does not guarantee data delivery in a specified time. One solution is to use Fast Ethernet (100 Mbps) on dedicated network links. This effectively lightens the message load on a system. The occasional packet collisions then become insignificant, yielding a virtually deterministic network.
Still, a few applications will simply require guaranteed real-time performance, such as some implementations of coordinated motion control. In these cases, engineers can isolate the host control and drives from other network traffic on one side of a network bridge. Devices on the other side can still access the system if they need to, but the bulk of the traffic will be intra-system and real-time.
With the advent of 100 Mbps to 1 Gbps Ethernet, coupled with the message prioritization (IEEE 802.1p) and redundancy (IEEE 802.12d) standards, segmenting the network can eliminate collisions and guarantee determinism in single master, multiple-slave arrangements.
Windows CE. This is the embeddable rendition of Windows. Version 3.0 adds real-time capabilities, opening the door for deterministic applications including those in motion control. The first real-time CEbased products are already available in PLCs. It won't be long before you'll find CE-based motion controllers performing complex contouring moves with networked drives and I/O.
Brian Kapitan is chief technologist and Ken Wyman is vice president of marketing and sales at API Motion Inc., in Amherst, N.Y.