"This concept has been around for a while, about 10 years or so," said Roger Bullock, drive support manager at Baldor Electric in Fort Smith, Ark. "It hasn't been commonplace, because no manufacturer had perfected it and taken it into full production. That has changed in the past few years."

The most basic type of motor control—short of having no control at all—is an open-loop scalar inverter drive. With this configuration, speed regulation can be no better than the slip of the motor. For example, if a motor has 3-percent slip, speed regulation can be no more than 3 percent. Although this type of control is the most economical, it is not precise enough for many applications.

For closed-loop control, motors generally must have some sort of feedback sensor. The most accurate way of doing this is with an encoder. Such a device's input, along with the drive's algorithm, allow the drive to commutate the motor. It is also used to retransmit the signal, buffer it, and supply it to the user for positioning applications.

The most common type of motor feedback sensor is the Hall sensor. Another is a resolver, which involves two windings mounted on the shaft and two windings mounted to the body of the motor. One fixed winding excites the two that are rotating, and the other senses the voltage induced in the two rotating windings. By measuring the voltage, which varies sinusoidally, resolvers provide a very accurate reading of rotor position. This information enables the controller to fire the rotor so that it rotates properly and produces the maximum amount of torque possible.

Although feedback sensors such as Hall sensors, encoders, and resolvers are mature, reliable technologies, they do have their drawbacks. One is the cost of the sensor itself. A two-phase motor requires two Hall sensors, and three- and four-phase motors require four. That adds at least $1—the cost of the sensors plus the required plastic-bonded magnet and hub—to the total cost of the motor. This expense would be trivial in expensive, one-of-a-kind uses like a custom-built conveyor for a factory assembly line, but it is very significant in high-volume applications such as computer disk drives, which are manufactured in quantities of hundreds of thousands.

Another drawback of feedback sensors is that they require additional wiring, which can be prohibitive in applications such as hermetically sealed compressors. The wiring can also be vulnerable to electromagnetic interference—common in many industrial environments—which can hurt performance. In addition, even though feedback sensors are generally reliable, some experts claim that eliminating them can enhance reliability because this leaves fewer components that can break down. Also, since Hall sensors are semiconductors and magnets are subject to thermal stability problems, the Hall-sensor configuration can become less reliable at high speeds.

Removing the sensor from the motor gives closed-loop performance without closing the loop. While no sensor is available to provide feedback, there has to be some way of determining the motor's status. Instead of sensors, the algorithms in the drive monitor one or more key motor parameters and convert those data into information about motor speed. "It is only in recent years that the digital signal processors [DSPs] needed have become fast enough and inexpensive enough to make sensorless control feasible," said George Holling, president of Advanced Motion Controls Inc. (AMC) in Princeton, Wis.

State-of-the-art DSPs—referred to as "vecom" chips in Europe—have faster processing speeds than ever and 16-bit processing power, which has led to finer control algorithms. The controller cannot work with an analog signal; instead, the analog signal is converted into a digital signal that is broken down into elements fine enough for accurate control. Heavy processing capability is needed because measurements must be taken continuously. However, to reduce some of the processing requirements, most manufactures monitor the change of current rather than the current itself, so a new set of calculations is only needed when the change exceeds a preset value.

There are several ways to perform sensorless control. The most basic is to construct a mathematical model of the motor and verify that it reflects the motor's actual performance. This method requires no extra hardware, because the needed DSP is usually present in the controller. The disadvantage is that motor behavior usually varies with current and speed. Behavior also can be different among specific motors of a given model. As a result, this technique is relatively unreliable and used infrequently.

At this sawmill, a sensorless motor from Baldor Electric (right) is coupled via belts located under the shrouding to the chipper head.

Another method of sensorless technology can be likened to back EMF motor control. If the motor-circuit inductances and resistances are known, the resultant motor terminal voltage can be calculated where the motor terminal voltage is equal to motor speed. "Baldor uses this technique with its model reference adaptive control algorithm," Bullock said. "A dedicated DSP performing over 16,000 calculations per second is used. With user-entered motor data, the algorithm decouples the magnetizing- and torque-producing currents to determine precisely the motor load and speed."

The third main form of sensorless control, which can work with all types of motors, is indirect measurement through inductance sensing. Inductance is measured using one of several basic mechanisms, and phase and rotor position are derived from this value. This is accomplished by injecting a high-frequency signal and measuring the response. The method does require additional hardware, and it may be sensitive to switching noise.

Inductance sensing has several different approaches. The first uses current waveform. A typical variable-switch-reluctance motor, for example, has a hump in the current waveform. Inductance can then be derived from the slope of the current and the magnitude of the peak. This approach, however, cannot be used at motor start-up, and waveforms vary with motors and loads.

The second approach uses current level, where motor current is compared with an absolute level. This might not work for all motors, and again it does not work at low speeds.

The third approach is to measure current rise time, which requires that a pulse-width-modulated (PWM) signal be present. It does not work with firing-angle control.

Finally, inductance can be sensed by comparing the derivative of the motor current with a preset level. This does not depend on any particular current waveform, and works with PWM and firing-angle control. It will also work during start-up as long as a small current is injected into all phases. "We measure the derivative of current," said Holling. "After compensating for some imperfections, you can get highly accurate position sensing. Furthermore, this is a continuous process that works well even at high speeds. Relatively low-cost feedback does this, and this equipment is already there to monitor overcurrent conditions."

Another advantage to this fourth approach is that it uses a signal already present in the motor. (At start-up, a small amount of current must be injected in motor windings.) It can even work when net torque is zero, by comparing the net torque of motor windings and determining which one is most appropriate to energize. There must be comparable feedback among four phases.

Sensorless motors are proving to be most appropriate for applications where the position of the rotor does not have to be pinpointed with great accuracy, and all that is needed is speed and torque control. Prime candidates for sensorless motors are applications where open-loop scalar control had been used but better speed regulation and more torque at low speed are needed. Sensorless closed-loop control allows more direct control of motor torque than the open-loop scalar method, resulting in less motor heating and thus improved motor life. Typical applications include velocity control on conveyors, elevators, some pick-and-place applications, overhead gantries, appliance motors, and centrifuges.

"With a sensorless motor, you don't get as tight a loop closure," said Charles Geraldi, vice president of Global Servo Drives, a division of Lenze based in Montvale, N.J., "but it's decent. It's probably not the best solution if you need very accurate position control, but if you don't need it, you might as well not have to pay for it."

home | features | weekly news | marketplace | departments | about ME | back issues | ASME | site search

© 1998 by The American Society of Mechanical Engineers