By John Guite
Product Planning Manager
Parker Hannifin Corp.
Rohnert Park, Calif.
EDITED BY John R. Gyorki
Hundreds of motion-control users have chosen Ethernet over fieldbus and PCI. But, as far as industrial motion control is concerned, all are capable of carrying the same information between a highlevel machine controller and a motion controller. Ethernet is a networking protocol, PCI is a PC bus architecture, and fieldbus is used in industrial control. These three technologies are now competing for the same users on a level playing field. However, each comes with its own set of advantages and disadvantages.
Ethernet: A Worldwide Standard
Developed more than 20 years ago as a high-speed serial data-transfer network, Ethernet has become a worldwide standard and is now the most widely used Local Area Network (LAN) in existence. More than 85% of all installed network connections in the world are Ethernet. The most common Ethernet data transfer rate is 10 Mbits/sec, although most installations are rapidly converting to Fast Ethernet at 100 Mbits/sec. The next generation of Ethernet is called Gigabit Ethernet capable of 1 Gbits/sec (1,000 Mbits/sec). Ethernet is generally incorporated into a motion-control system through a stand-alone controller connected to the PC or network through a standard Ethernet cable.
Fieldbus: A Popular Protocol
The term fieldbus covers many different industrial network protocols. Two widely used protocols are DeviceNet and Profibus. Most fieldbus protocols have been developed and supported by specific PLC manufacturers such as Allen-Bradley and Siemens. However, a dominant fieldbus standard does not yet exist. Network performance and hardware interfaces differ depending on the fieldbus chosen, although data transfer rates from 500 kbits/sec to 12 Mbits/sec are typical. A PLC usually serves as the fieldbus master, which then communicates with distributed follower devices such as industrial I/O or motion-control systems. A motion controller is typically a stand-alone or busbased follower device on the fieldbus.
PCI: High-Speed Data Transfer
Peripheral Component Interconnect (PCI) architecture is currently the industry standard for desktop PCs. It's the latest in a line of ever-improving bus architectures offering faster speeds and greater throughput. PCI was intended to handle large amounts of data transferred by modern high-bandwidth applications such as audio and video peripherals. PCI is far superior to the bus architectures it replaces, such as ISA, EISA, and VESA because it moves the peripherals closer to the CPU, and eliminates processing latencies. It's a parallel interface, and the most common version has 32 data lines running at 133 Mbytes/sec. Most new PCI chipsets are capable of 64 data lines with throughputs of 266 Mbytes/sec. The next generation of PCI or PCI-X will run 1 Gbit/sec. A PCI motion-control solution comes in the form of a traditional busbased card that fits into an expansion slot of the PC.
Since PCI data-transfer rates are about 20 times greater than that of Ethernet and fieldbus, it is tempting to assume that PCI is the most obvious choice for a control system.
But, although speed is important, it's only one of the many criteria that must be considered to determine the optimal control structure. Other important issues in control-system specifications are flexibility, package size, PLC compatibility, installation and maintenance, and long-term viability. The following sections examine each of these areas relative to Ethernet, fieldbus, and PCI solutions.
While PCI may clearly win the datatransfer-rate race, the important issue is how that throughput is being utilized. It is unlikely that a motion-control application will use the entire bandwidth of the PCI bus. Usually the computer shares the PCI bus with several other devices such as an Ethernet card, a sound card, or a video application. In fact, being able to handle Fast Ethernet and other high-bandwidth functions is one of the main reasons that PCI has been upgraded from 32 to 64 bits. PCI is necessary as a backbone to carry all the information a network and other devices on the PCI bus transmit. Even if the entire PCI bandwidth were available, most PC-based motion controllers are simply not fast enough to send and receive large amounts of data as quickly as the PCI bus. In other words, the bottleneck is not the data-transfer protocol but the device sending and receiving the data itself.
Another speed-related concern is whether the majority of motion-control applications require the PCI-bus throughput. Look at the type of information typically sent and received by a motion controller (such as position, velocity, acceleration, or error) — the actual amount of data transferred every cycle is relatively small, perhaps 512 bytes. With an Ethernet motion controller, these 512 bytes equal only 4% of the overall bandwidth in a 10-Mbits/sec network.
Most motion-control applications are carried out by a variety of controllers. For example, a bus-based, stand-alone, or network motion controller may be required. Unfortunately, a PCI solution allows for only one form factor. In applications where the motion controller carries out functions independently of a PC or in situations where there is no PC control on the machine, a PCI-based solution is simply not feasible. On the other hand, most Ethernet or fieldbus controllers are also capable of operating as stand-alone devices.
As for remote-access networking, a PCI is not a viable solution. Implementing a motion controller on a factory network requires a controller with a network interface. Since most factory networks are Ethernet, the choice is even easier.
Often, when deciding between a standalone or network-capable controller and a PC-based controller, an engineer mistakenly assumes that a PC-based controller saves space since it resides in the PC. While this is true of the controller itself, there is also an external box required to connect servomotors, stepmotors, encoders, and I/O. In many cases this external box is as large or larger than a comparable standalone controller.
In many applications, a PLC is the primary controller that determines how far and fast a load should move. The communication between the two devices can range from discrete I/O to a network connection. Discrete I/O is not a favored method due to the time and effort required in wiring and installing the system. An Ethernet or a fieldbus solution is preferred in this case. Most PLCs now offer Ethernet as a standard networking option in addition to their fieldbus of choice.
Installation and Maintenance
Most firms cannot afford to have a DeviceNet or Profibus specialist on staff who thoroughly understands the network protocol. Even if a company could afford such a person, it is unlikely fieldbus would be their specialty. However, almost every company has a network administrator who is well versed and specializes in Ethernet protocol, making Ethernet all the more attractive for industrial control.
Connecting Ethernet to a PC is simpler than connecting bus-based controller cards to a computer. The only connection between the PC and an external Ethernet motion controller is a category-5 Ethernet cable. An Ethernet hub or switch that can add additional axes of control is readily available. Replacing a controller simply means configuring a new motion controller through software and making the same easy Ethernet connection. Bus-based controller-card installations, on the other hand, require that the computers be shut down or taken offline. Then having to remove the PC cover is an especially difficult task if the PC chassis is buried inside a machine. Finally, most bus-based motion controllers require two or more PCI slots for installation. Additional axes of control require additional slots and unwieldy cables snaked through yet another open expansion slot. If the controller fails for any reason, the whole process must be repeated to replace the controller.
When choosing control architecture, ask if this architecture will be available and viable for the life of the machine. Although it seems likely that PCI will be available for many years to come, it might eventually go the way of the ISA. How much longer before a faster and more efficient bus structure is developed for the PC? Also, how far can the fieldbus technologies be pushed, and who will motivate such action? Will companies like Oracle, IBM, or Coca-Cola care if a particular fieldbus becomes obsolete? Probably not, but they are likely concerned about the future of Ethernet because their company infrastructure is probably built upon it.
Ethernet has shown remarkable staying power. With over 85% of network connections in the world using it, there is widespread industry acceptance and thus little chance the technology is going to disappear soon. Since the 1970s Ethernet has seen many PC-bus architectures come and go. And with every new bus, many companies have developed Ethernet-compatible components.
The dominant technology — whether it ends up being Ethernet, fieldbus, or PCI — will need to be flexible, easy to maintain, and will require industry support and confidence in long-term viability. Ethernet clearly leads the group in these concerns, while the other two technologies provide only partial solutions. Ethernet-based motion controllers are flexible enough to operate in stand-alone applications as well as those requiring an easy and reliable interface to a PC or a PLC. Couple this with local or remote networking, and it's easy to see that Ethernet lets a manufacturer standardize on one motion controller regardless of the application. Overwhelming industry acceptance of the Ethernet ensures that speed and performance will continue to improve, as evidenced by the significant progress already made in Gigabit Ethernet and leading endusers. Time and experience will tell which technology will dominate the factory-automation industry. But networkable, remotely accessed industrial motion control is the key to the future, and Ethernet is most clearly the architecture of choice.