Motion controllers are available as cards that can be plugged into PCs, or as standalone units that may or may not include an amplifier. "The trend is away from PC cards and to boxed units," says John Mazurkiewicz, product line manager with Baldor Electric Co. (Fort Smith, AR). "It's just more convenient to make the connections."

Just as some people are smarter than others, some motion controllers are more intelligent than others. If the system will always be performing the same motions over and over again-a pick-and-place device, for example-a preprogrammed, general purpose controller may be easier to use and less expensive. These devices are already programmed with basic motion commands, and the engineer simply needs to enter the values. "All you have to do is tell it what positions you want to go to and how fast you want to get there," says Mazurkiewicz. "It's quick and easy, and you don't need a programmer to do it. You just fill in the blanks."

For more complex applications, a fully programmable controller is necessary. As their name implies, these controllers have to be programmed. "Some motion controllers have proprietary programming languages, so you have to learn that language," says Mazurkiewicz. "They are powerful, but if you lose that programmer, you have to get someone who knows the language or train someone to use it."

When specifying a motion controller, engineers should match the controller with the number and types of motors in the system. Some controllers are just for servomotors; some are only for step motors; and others can control both servos and steppers. Similarly, some controllers are just for rotary motors; some are for linear motors; and others can control both.

Next, engineers will need to specify the number and types of I/O they need. Some devices can be wired in series, so one I/O can serve multiple devices. "I/O cost money and real estate, so you should carefully gauge how much you need," says Mazurkiewicz.

The controller also has to be matched with the fieldbus connecting the various components of the motion control system. There are many to choose from: Ethernet, DeviceNet, Profibus, Sercos, CANopen, Firewire and CynqNet. Which to use depends on the speed and accuracy requirements of the application, wiring issues, and the types and brands of devices that will be connected to the bus, such as the human-machine interface and the host computer or PLC. Often, the choice of bus is a matter of industry, or even personal, preference.

Most controllers use proportional-integral-derivative (PID) algorithms to generate command signals for the motors. In proportional control, the command signal is generated from the difference between the programmed position and the actual position. The magnitude of the signal is proportional to the size of the error.

"Proportional control cannot achieve a zero steady-state error," says Mazurkiewicz. "Increasing the gain in the system can decrease steady-state error somewhat. However, when the gain is too high, the system overcorrects and may cause instability."

Integral control adds time into the mix. The command signal is developed from the product of the position error and the time that the error has persisted. Thus, over time, the deviation from the desired setpoint is minimized. Derivative control produces a corrective signal based on the rate of change of the signal. The faster the change from the set point, the larger the signal. This produces a faster response than the proportional-integral signal alone.