Motor Drive Control Issues

Many adjustable-speed drive (ASD) applications require medium- or high-bandwidth torque control in order to achieve adequate control performance.  What this means is that if a drive system can function as a controllable source of torque, with fast dynamic response, it should be able to implement the desired motion control operation.

Prior to 1980, ASDs implemented with dc motors had significant advantages over other types.  The primary advantage of dc ASDs over ac ASDs, for example, was that dc drives could readily provide torque control through direct control of the armature current.  In contrast, ac drives up to that time usually used square-wave converters and were more suitable for simple speed control. Without torque control, the dynamic performance of a drive is at best slow and oscillatory, and in the worst case, unstable. High-performance applications such as industrial machine tools, spindle drives, cutters and machine tools require position control -- best implemented with precise torque control. The torque oscillations and sluggish response inherent in the square-wave converter are unacceptable for these applications. Instantaneous torque control is also highly desirable for many medium-performance applications such as conveyor belt drives, rollers, and electric vehicles which require fast torque response and precise speed control. Finally, while the transient response of low-performance applications such as fans and pumps is usually not important, instantaneous torque control has the advantage of eliminating the otherwise necessary current limiters. Conservative limits on the peak current, which also limit the peak torque, are required by square-wave drives to prevent "nuisance trips" caused by sudden surges in the torque demand. If a drive has instantaneous torque control capability, on the other hand, the maximum allowable torque setting serves to limit current, and can be provided quickly with low risk of spurious trips. True bipolar torque control allows four-quadrant operation of the drive, in which both speed and torque can be positive or negative.

A dc motor controller provides rapid torque control because the commutator maintains a fixed (and nearly ideal) torque angle at all times.  The instantaneous torque is proportional to the product of the armature current and the field current, and these are dynamically decoupled.  Dynamic decoupling means that field current and armature current can be adjusted independently without interference.  In a typical application, the field current (which changes slowly because of high winding inductance) is set and held at a target constant value.  The armature current is then adjusted to provide control of the shaft torque.  The dynamic response can be very fast:  even large dc machines have armature L/R time constants in the range of about 20 ms, so the torque can be adjusted at bandwidths of a few tens of hertz.

The challenge of a dc ASD is that it must be able to adjust the armature voltage and control armature current of a machine with minimal power loss. An early dc ASD, the Ward-Leonard system, controlled the dc voltages to the armature and the field with a simple algorithm, in which the desired torque and the magnetizing field strength served as command inputs.  In this system, a separate dc generator provided the adjustable dc source for the armature.  More recent dc drives use power electronics -- either a phase-controlled rectifier or a dc-dc converter from a separate dc supply -- to provide the armature control.  An ac motor, on the other hand, involves complex, nonlinear relationships between voltages, currents, fluxes, and torque. High-fidelity control of ac currents in these machines generally require fast power electronic components, which were not readily available until the 1970s. For these reasons, virtually all medium- or high-performance ASD's installed before 1980 were dc drives.  Many of these are still in operation today, but they are being supplanted rapidly by the new generations of ac drives.

The early generation of ac ASDs, which appeared mainly in the 1980s, used pulse-width modulation (PWM) to provide adjustable-frequency sinusoidal currents to ac machine stators.  The best of these drives provide excellent speed control but do not have direct torque capability.  Typically, the ac motor is treated as a "speed source" in which the actual speed is determined primarily by the stator input frequency.  The voltage is adjusted approximately so as to keep the no-load current fixed as the frequency changes.  Since this results in a frequency-proportional voltage, the method is often called "constant volts per hertz" control.  Commercial constant volts per hertz drives have become very sophisticated.  They permit adjustment of acceleration and deceleration rates, support wide speed ranges, and adjust for resistive drops to support tight regulation of speed.  Since drives of this type do not require any special sensing, and need only speed information for high performance, they are generally termed scalar control drive systems.  Scalar drives are popular in a wide range of ASD applications, and represent the most common type of ac drive.  They have the advantage that little information about the motor is needed to make them operate properly.

Field-Oriented Control

Speed control, while it supports a wide range of applications, does not cover the true high-end applications that require a torque source.  Several control algorithms were devised in the late 1960s by which ac induction motors could nearly achieve torque control. These methods met with only limited success, however, since they required cumbersome calculations which involved many unknown model parameters and numerous approximations. Additionally, they were limited by the slow switching speed and difficult commutation requirements of early thyristors. With the development of faster-switching and more easily controlled power electronic components such as the gate turn-off thyristor (GTO), the power bipolar junction transistor (BJT), and most recently the insulated-gate bipolar transistor (IGBT), researchers realized that new control schemes could be implemented which would provide high-bandwidth current control.

In 1972, F. Blashke published an approach to ac induction motor control known as field orientation.  In field orientation, the motor input currents are adjusted to set a specific angle between fluxes produced in the rotor and stator windings -- in a manner that follows from the operation of a dc machine.  More formally, the approach applies direct-quadrature (D-Q) two-axis analysis methods directly to the torque control problem.  When the dynamic equations for an induction motor is transformed by means of well-known rotating transformation methods into a reference frame that coincides with rotor flux, the results become similar to the dynamic behavior of a dc machine.  This allows the ac motor's stator current to be separated into a flux-producing component and an orthogonal torque-producing component, analogous to a dc machine field current and armature current. The key to field-oriented control is knowledge of the rotor flux position angle with respect to the stator. In Blashke, the rotor flux was measured by sensing coils.  Approaches in which rotor flux is sensed are now generally termed "direct field orientation" (DFO) methods.  It is also possible to compute the angle from shaft position information, provided other motor parameters are known.  This approach is now generally termed "indirect field orientation" (IFO), and is the basis of most commercializations of the field orientation concept.  The position sensor is often considered an undesired element, and various ways to develop "sensorless" field-oriented drives are active research topics today.  Whatever the field-orientation approach, once the flux angle is known, an algorithm performs the transformation from three-phase stator currents into the orthogonal torque- and flux-producing components.  Control is then performed in these components, and an inverse transformation is used to determine the necessary three-phase currents or voltages.

In principle, field orientation provides similar decoupled control of torque and flux which is inherently possible in the dc machine.  There are certain fundamental differences, however:  field-orientation provides what can be termed "asymptotic decoupling" -- the torque and field producing elements are decoupled if the field current is fixed.  This is a minor concern in most applications, since the field current is not usually adjusted rapidly in a drive application.  A more significant difference is the need for precise motor information.  The transformations and control algorithms presume ideal knowledge of various resistances and inductances within the machine.  In a dc machine, small parameter errors will alter the output torque but not the decoupling.  In an ac machine, parameter errors alter the transformation and can cause torque ripple and other problems.  A motor must be carefully characterized and measured for use with a field-oriented drive -- a process called commissioning that is often automated in commercial units.

Field-oriented torque control opened up many high-end applications for ac ASDs beyond those that can be served by speed control. In addition to the many benefits of the ac induction motor over dc machines (notably lower cost, better reliability, and higher efficiency), ac ASDs were found to offer faster torque response compared to dc drives.  This is achieved through the lower inertia of an ac motor compared to a dc motor of similar rating, and through the faster L/R time constants in many ac machines compared to dc machines. Recent improvements in IGBTs and in the latest controlled thyristors has now boosted sustainable switching frequencies into the multi-kilohertz range.  High-performance field-oriented drives now reach ratings well above 100 kW.  Pulse-width modulation and cost reductions in digital signal processors (DSPs) have made it much easier to compute transformations rapidly and enforce the desired current waveforms. These developments are leading to widespread industrial application of  field-oriented ac induction motor drives in high-end applications.  Drives of this type are generally termed vector drives to distinguish them from adjustable-speed scalar drive methods.

Types of Field Orientation

The basis of the field orientation algorithm is to use the flux angle, usually the rotor flux angle, to decouple the torque and flux components of the stator current. The most challenging, and ultimately the limiting feature of field orientation is the method whereby the flux angle is measured or estimated. The DFO method, in which direct measurement of flux is performed using flux sensing coils or Hall-effect devices, proved to be impractical for general use. The IFO method has become much more common.  In this case, the flux angle is not measured directly but is estimated from the equivalent circuit model and from measurements of the rotor speed, the stator current and the voltage. One common technique for estimating the rotor flux is based on the slip relation, which requires measurement of the rotor position and the stator current. Given sufficiently accurate current and position sensors, this method performs reasonably well over the entire speed range. Most high-performance ac ASDs in operation today employ indirect field orientation based on the slip relation.

A significant disadvantage of the slip relation is that it requires rotor position feedback from a shaft-mounted sensor, typically an optical encoder.  As mentioned above, "sensorless" (no shaft sensor) field-orientated control algorithms are a subject of active research. Most of these schemes depend on integrating the back emf voltage, which by Faraday's Law produces the stator flux. At high speeds, the back emf is approximately equal to the terminal voltage since the stator voltage drop is negligible, and integration of the stator voltage accurately estimates the field orientation angle. At low speeds, however, back emf-based schemes fail for two reasons. First, Faraday's Law implies that for a given distribution of linked flux, a coil's induced voltage amplitude is proportional to the fundamental frequency. Thus at low speeds, the back emf is small and voltage measurement is degraded by the relative increase in noise.  A second problem of emf methods is that while the voltage magnitude decreases with frequency, the load current for a given load torque does not change significantly with frequency. Thus at low speeds, the stator winding resistive drop is relatively much greater and the terminal voltage is no longer comparable to the back emf. In this case, estimating the back emf requires accurate knowledge of the stator resistance, but this varies with temperature. Therefore, a back emf-based field orientation scheme works well except when the drive frequency is reduced to the low speed range, where the dependence on the stator voltage measurement becomes increasingly unreliable due to noise and parameter detuning. Back emf-based schemes cannot provide field-oriented control over the entire speed range; at very low speeds, only the slip relation with rotor position information can provide sustained field orientated torque control. Additionally, even though back emf-based field orientation does not use the rotor speed directly, actual speed feedback is often important for steady-state speed regulation, and these schemes generally do not provide accurate rotor speed estimation because of detuning.

The literature mentions both DFO and IFO methods.  The original approach was developed based on rotor flux, but in the literature vector drive methods based on stator flux and air-gap flux can also be found.

Direct Torque Control

Direct torque control (DTO) is a technique similar in concept to field orientation, in which induction motor voltages or currents are set to values as close as possible to the ones needed to generate a desired torque.  Typically, the operation is performed with less intensive computation that for various field-oriented methods.  DTO is often implemented through space-vector modulation (SVM), in which the drive switches are set to alternate among just two of the possible output voltages so as to approximate closely the desired voltage vector.

Contents copyright (c) 1996-2002, IEEE.  All rights reserved.