Vector Control and Direct Torque Control

Introduction and Context

AC motors are mechanically robust, efficient, and relatively inexpensive, but their electromagnetic torque and flux are coupled through sinusoidally varying stator quantities. This coupling made early induction-motor drives difficult to control precisely. Scalar methods, such as constant-voltage-per-frequency control, provide satisfactory performance for pumps and fans but offer limited dynamic torque response and weak low-speed regulation.

Vector control and direct torque control (DTC) overcome these limitations by treating the motor variables as spatial vectors. Vector control transforms three-phase quantities into a rotating reference frame in which flux- and torque-producing components can be regulated independently. DTC directly selects inverter voltage vectors to control stator flux and electromagnetic torque without an inner current-control loop in its classical form.

These methods are central to industrial motion control, electric vehicles, machine tools, robotics, elevators, compressors, and renewable-energy conversion systems.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Explain stationary and rotating reference frames used in AC-machine control.
2. Apply Clarke and Park transformations to three-phase variables.
3. Develop flux and torque models for induction and permanent-magnet machines.
4. Design the principal structure of rotor-field-oriented and stator-field-oriented control.
5. Explain decoupling, current regulation, voltage limitation, and field weakening.
6. Compare classical DTC with field-oriented control.
7. Describe flux estimation, sensorless operation, and their practical limitations.
8. Select suitable control strategies for representative industrial applications.

1. Spatial Vectors and Reference-Frame Transformations

A balanced three-phase set can be represented by two orthogonal components rather than three independent phase variables. Let the phase quantities be x_a, x_b, and x_c, where x may denote voltage, current, flux linkage, or a related quantity.

The Clarke transformation maps the three-phase system into stationary orthogonal axes α and β. A power-invariant form is

For a balanced system, x_a + x_b + x_c = 0, so the zero-sequence component is zero. If zero-sequence currents are possible, an additional component must be retained:

The Park transformation rotates the stationary αβ axes through electrical angle θ_e:

The inverse transformation is

Here, d and q are rotating axes. The angle must correspond to the selected reference frame. For a rotor-flux-oriented induction-motor controller, θ_e is the electrical rotor-flux angle. For a surface permanent-magnet synchronous motor (PMSM), the d-axis is normally aligned with the rotor permanent-magnet flux.

The electrical and mechanical rotor speeds are related by

where ω_e is electrical angular speed in rad/s, ω_m is mechanical angular speed in rad/s, and p is the number of pole pairs. Mechanical speed in revolutions per minute is related by ω_m = 2πn_m/60.

Physical meaning of the d and q axes

In a properly selected rotating frame, the d-axis component primarily controls flux and the q-axis component primarily controls torque. This separation is not automatic; it depends on the machine model, the chosen angle, accurate parameters, and effective compensation of cross-coupling terms.

For a balanced sinusoidal set, a constant vector in the synchronous frame appears as a sinusoid in the stationary frame. This is the principal reason rotating-frame control converts an AC regulation problem into a DC-like regulation problem.

2. Induction-Machine Model

For an ideal symmetrical induction machine expressed in an arbitrary rotating d–q frame, the stator voltage equations are

The rotor equations, with rotor windings short-circuited, are

Here, v, i, and ψ are voltage in volts, current in amperes, and flux linkage in weber-turns; subscripts s and r denote stator and rotor; d and q denote axes; R_s and R_r are resistances in ohms; ω_k is reference-frame electrical speed; and ω_r is rotor electrical speed. The equations assume sinusoidally distributed windings, magnetic symmetry, and a linear magnetic circuit unless saturation is explicitly included.

The flux-current relationships are

where L_s and L_r are stator and rotor inductances and L_m is mutual inductance. The electromagnetic torque is

This general expression shows that torque is proportional to the cross product of the stator-flux vector and stator-current vector.

Rotor-flux orientation

Rotor-flux-oriented control selects the d-axis so that

The rotor-flux dynamics then become approximately

where rotor time constant T_r = L_r/R_r. At steady state,

for the common normalized model in which the rotor flux is represented consistently with the chosen inductance definitions. The torque becomes

under rotor-flux orientation. Thus i_ds controls rotor flux and i_qs controls torque. This is the central decoupling result for indirect or direct field-oriented control.

The slip electrical frequency required to maintain rotor-flux orientation is

and the synchronous electrical speed is

The formula assumes rotor-flux orientation and a nonzero estimated rotor flux. At very low flux or during transients, parameter errors can make this calculation inaccurate.

3. Permanent-Magnet Synchronous-Machine Model

For a PMSM in the rotor reference frame, neglecting zero-sequence effects, the voltage equations are

where L_d and L_q are d- and q-axis inductances and ψ_f is permanent-magnet flux linkage. The electromagnetic torque is

The first term is magnet torque. The second is reluctance torque and exists when the machine is salient, meaning L_d ≠ L_q. For a surface-mounted PMSM, L_d ≈ L_q, so

Consequently, below base speed, a typical surface-PMSM controller sets i_d^* = 0 and regulates i_q according to the torque demand. Interior PMSMs may use negative d-axis current to obtain maximum torque per ampere (MTPA), because reluctance torque can augment magnet torque.

The stator-flux linkage is

This equation is useful for flux estimation and voltage-limit analysis. It assumes the d-axis is aligned with the permanent-magnet flux and ignores saturation and cross-saturation.

4. Field-Oriented Control Architecture

A practical field-oriented controller normally contains the following signal path:

1. Measure two phase currents, DC-link voltage, rotor position or speed, and sometimes phase voltage.
2. Transform measured currents from abc to αβ and then to dq.
3. Compare measured d- and q-axis currents with references.
4. Use two current controllers to generate d- and q-axis voltage commands.
5. Compensate cross-coupling and back electromotive-force terms.
6. Transform voltage commands back to αβ.
7. Apply space-vector pulse-width modulation (SVPWM) or another modulation method.
8. Drive the inverter switches while enforcing current, voltage, thermal, and DC-link limits.

Current regulators

For a PMSM, ignoring coupling initially, a PI current controller may be written as

where K_p and K_i are proportional and integral gains. The units depend on the implementation and scaling of the error and integrator.

The PMSM voltage equations show that the axes are coupled by the terms −ω_e L_q i_q and ω_e L_d i_d. Feedforward decoupling can therefore use

The signs follow the voltage convention stated earlier. Controllers must use consistent transformation and voltage signs; otherwise, decoupling can destabilize rather than improve performance.

For an induction motor, the current references are usually produced by an outer speed controller. A torque reference is converted into i_q^* using the torque model, while i_d^* establishes the desired rotor flux. A speed PI controller may be represented as

with appropriate saturation and anti-windup. The speed-loop bandwidth is normally much lower than the current-loop bandwidth so that the inner current loops appear fast to the mechanical controller.

Voltage and current limits

The inverter cannot produce arbitrary voltage. For a two-level inverter with DC-link voltage V_dc, the available voltage vector is bounded by the modulation method. In a simplified circular approximation,

where V_max depends on the PWM convention and allowable modulation index. When the demanded voltage exceeds this limit, the controller must reduce flux, speed, or torque demand. Integral anti-windup is essential.

At high speed, the back-emf increases with speed. Field weakening reduces d-axis flux, commonly by applying negative i_d in a PMSM, so that the voltage remains within the inverter limit. Field weakening is constrained by demagnetization risk, current rating, thermal limits, and machine saliency.

Worked Example: PMSM Torque Current

An eight-pole-pair surface PMSM has permanent-magnet flux linkage ψ_f = 0.08 Wb. Neglect reluctance torque. Find the q-axis current needed for 12 N·m torque.

For eight pole pairs, p = 4 because an eight-pole machine has four pole pairs. The torque equation is

Therefore,

Thus the required q-axis current is 25 A, assuming ideal parameter knowledge, no losses, and a sinusoidal machine model. The d-axis reference for a surface-PMSM below base speed would commonly be zero.

5. Direct Torque Control

Classical DTC estimates stator flux and electromagnetic torque directly, compares them with reference values, and selects an inverter switching state from a lookup table. It does not require transformation into a rotating frame for its basic operation and does not use a conventional inner PI current loop.

The stator-flux model in the stationary frame is

where v_s and i_s are stationary-frame stator-voltage and current vectors. The torque estimate is

The estimated flux magnitude is

The inverter has six active voltage vectors and two zero vectors. During a short sampling interval, the approximate flux change is

Thus an active voltage vector generally advances or changes the stator-flux vector, while its tangential component strongly influences torque. A hysteresis comparator determines whether flux and torque should increase or decrease. The estimated flux sector and comparator outputs address a switching table.

DTC characteristics

DTC offers very fast torque response because the inverter voltage vector is selected directly in response to torque error. It has a relatively simple conceptual structure and does not require a rotor position sensor for the basic control law. Its disadvantages include variable switching frequency, torque and flux ripple, acoustic noise, sensitivity to stator resistance at low speed, and difficulties near zero speed. Modern variants use space-vector modulation, predictive control, fuzzy selection, or constant-switching-frequency regulators to reduce these effects.

6. Flux Estimation and Sensorless Operation

Flux estimation is central to both DTC and sensorless field-oriented control. The voltage model integrates the stator back-emf:

At low speed, the back-emf is small and the resistive voltage drop is relatively important. Small errors in R_s or current measurement can then create a large relative flux-estimation error. Integrator drift and inverter dead-time also matter.

The current model is based on machine flux-current relationships and rotor dynamics. It performs better at low speed in some induction-motor applications but depends strongly on rotor resistance, which varies with temperature. Practical observers combine voltage and current models. Examples include the model-reference adaptive system (MRAS), Luenberger observers, sliding-mode observers, extended Kalman filters, and nonlinear state observers.

A sensorless drive estimates rotor position or speed from measured currents and voltages. Benefits include lower cost, improved mechanical reliability, and elimination of encoder alignment problems. Limitations include poor observability at zero speed, parameter sensitivity, noise, computational burden, and possible instability during rapid regeneration or load transients.

For PMSMs, back-emf-based estimation becomes weak at low speed. High-frequency signal injection can exploit rotor saliency to estimate position near zero speed, especially in interior-PMSMs. It introduces additional acoustic noise, losses, and implementation complexity.

7. Comparison of FOC and DTC

Neither method is universally superior. FOC is often preferred where low ripple, predictable current limitation, and fixed switching frequency are important. DTC is attractive where rapid torque response and a direct electromagnetic control structure are priorities.

8. Illustrative Industrial Case: Sensorless Electric-Vehicle Traction Drive

Background

Consider an illustrative 60 kW interior PMSM traction motor supplied by a three-phase voltage-source inverter. The vehicle requires quiet operation, high torque at low speed, and extended constant-power operation above base speed. Eliminating the position sensor can reduce wiring and improve environmental robustness.

Problem

The controller must produce high starting torque, transition smoothly through base speed, limit battery current, and maintain stable operation during regenerative braking. A conventional back-emf observer is unreliable at standstill because the back-emf approaches zero.

Analysis

Below base speed, MTPA control is used. The controller computes d- and q-axis current references from a motor map that includes saliency and current limits. At higher speed, the voltage constraint becomes active, so negative d-axis current weakens the effective air-gap flux. During regeneration, the sign of q-axis current reverses while current and DC-link overvoltage limits remain enforced.

At standstill and very low speed, high-frequency voltage injection identifies the rotor saliency direction. As speed rises, the controller blends the injection-based estimate with a back-emf or observer-based estimate. A phase-locked loop tracks the estimated electrical angle, and plausibility checks detect disagreement between measured and estimated states.

Approach and solution

The drive uses:

• d–q current control with decoupling feedforward;
• MTPA below base speed;
• voltage-based field weakening above base speed;
• high-frequency signal injection at very low speed;
• an observer and phase-locked loop at medium and high speed;
• torque, current, thermal, and DC-link protection;
• anti-windup and controlled transitions between operating regions.

Outcome

The combined strategy provides sensorless starting capability, low torque ripple in normal operation, and constant-power performance over a wide speed range. The principal engineering cost is software complexity and the need to validate observer transitions under temperature, parameter variation, and regenerative operation.

Learning points

1. A sensorless strategy is usually a combination of operating-region-specific estimators rather than one universal algorithm.
2. Field weakening is constrained by voltage, current, thermal, and magnet-protection limits.
3. Control quality depends on measurement and protection architecture as much as on the control law.

9. Implementation Practice, Safety, and Contemporary Developments

Accurate current sensors must be synchronized with PWM timing. Offset calibration, gain matching, sampling delay compensation, and dead-time compensation can materially improve performance. Motor parameters should be identified over the expected temperature and frequency range; treating resistance and inductance as fixed can produce unacceptable errors.

Software should include startup sequencing, precharge and DC-link checks, gate-driver interlocks, overcurrent shutdown, overspeed protection, loss-of-position detection, and safe torque-off where required by the application. Functional safety design must follow the applicable machinery, drive, and electrical-installation standards in the jurisdiction. A control algorithm should never be evaluated only by simulation when a hardware fault can create hazardous torque.

Contemporary drives increasingly use model-predictive control, disturbance observers, embedded parameter adaptation, wide-bandgap semiconductor inverters, and data-driven condition monitoring. Artificial intelligence can assist with fault diagnosis, parameter estimation, and remaining-useful-life prediction, but it should not replace deterministic protection or be trusted without validation over abnormal operating conditions. Explainability, cybersecurity, model integrity, and fail-safe behavior are important governance concerns for networked drives.

Chapter Summary

Vector control represents three-phase machine variables in stationary and rotating coordinates so that flux- and torque-producing components can be regulated separately. Clarke transformation produces αβ variables, while Park transformation produces d–q variables rotating at a selected electrical angle. Correct angle definition is essential.

The induction-machine model shows that rotor-flux orientation makes d-axis current primarily responsible for flux and q-axis current primarily responsible for torque. PMSM torque contains magnet and, for salient machines, reluctance components. Field-oriented control uses current regulators, decoupling, modulation, outer speed control, and explicit management of voltage and current limits.

DTC estimates stator flux and torque and directly selects inverter voltage vectors. It offers very rapid torque response but classical implementations exhibit variable switching frequency and greater ripple. Flux estimation is difficult at low speed because back-emf is small and resistance, measurement, integration, and inverter nonidealities become significant. Sensorless control therefore commonly combines observers or signal-injection methods across operating regions.

Successful drive design requires coordinated modeling, measurement, tuning, protection, thermal analysis, parameter identification, and validation. The best control method depends on the required dynamic response, ripple, speed range, cost, sensor strategy, and safety obligations.

Key Terms

Reference frame: Coordinate system used to express machine variables.

Clarke transformation: Mapping from three-phase quantities to stationary αβ components.

Park transformation: Rotation of αβ quantities into d–q coordinates.

Space vector: Two-dimensional representation of a balanced three-phase quantity.

Rotor-flux orientation: Induction-motor orientation in which the d-axis aligns with rotor flux.

Field-oriented control: Control method that regulates flux and torque components independently.

Decoupling: Compensation of cross-coupling terms between d- and q-axis dynamics.

PMSM: Permanent-magnet synchronous motor.

Saliency: Difference between d-axis and q-axis inductance.

MTPA: Maximum-torque-per-ampere operation.

Field weakening: Flux reduction used to remain within inverter voltage limits at high speed.

DTC: Direct torque control based on direct flux and torque regulation.

Hysteresis comparator: Regulator that switches when an error crosses prescribed bands.

Flux observer: Algorithm that estimates flux from measured and modeled variables.

Sensorless control: Operation without a mechanical rotor-position sensor.

MRAS: Model-reference adaptive system for state or parameter estimation.

SVPWM: Space-vector pulse-width modulation.

Anti-windup: Method that prevents integrator accumulation during actuator saturation.

Electrical speed: Mechanical rotor speed multiplied by the number of pole pairs.

Zero-sequence component: Common-mode component not represented by αβ variables alone.

Review Questions

Short Answer Questions

1. What is the physical purpose of the Park transformation in AC-machine control?
2. State the torque equation for a surface-mounted PMSM under the sinusoidal machine model.
3. Why does rotor-flux orientation simplify induction-motor control?
4. What are the principal disadvantages of classical DTC?
5. Why is sensorless estimation difficult near zero speed?

Long Answer Questions

1. Derive and explain the induction-motor voltage equations in an arbitrary rotating reference frame.
2. Explain the complete signal flow of a PMSM field-oriented controller, including current regulation and PWM.
3. Compare voltage-model and current-model flux estimation, including their parameter sensitivities.
4. Explain field weakening in a PMSM and identify its electrical, thermal, and mechanical constraints.
5. Compare FOC and DTC for a machine-tool spindle and justify a suitable choice.

Analytical / Application Questions

1. A four-pole-pair surface PMSM has ψ_f = 0.10 Wb and must produce 18 N·m while reluctance torque is neglected. Calculate the required q-axis current.
2. An induction motor operates with rotor-flux orientation. Explain how an increase in q-axis current affects torque, slip frequency, and the risk of current-limit violation.
3. Design an operating-region strategy for a sensorless PMSM traction drive from standstill through field weakening. Specify the estimator, current references, limiting actions, and transition checks for each region.

Multiple Choice Questions with Answers

Question 1

Which transformation converts three-phase quantities into stationary orthogonal αβ components?

A. Clarke transformation

B. Park transformation

C. Fourier transformation

D. Laplace transformation

Correct Answer: A. Clarke transformation

Question 2

In rotor-flux-oriented induction-motor control, which current component primarily controls torque?

A. Zero-sequence current

B. d-axis current

C. q-axis current

D. Rotor resistance current

Correct Answer: C. q-axis current

Question 3

For a surface-mounted PMSM below base speed, a common d-axis current reference is:

A. Equal to rated q-axis current

B. Approximately zero

C. Always equal to negative rated current

D. Infinite

Correct Answer: B. Approximately zero

Question 4

Classical DTC directly regulates:

A. Only stator resistance

B. Rotor temperature and inductance

C. Stator flux and electromagnetic torque

D. Mechanical friction only

Correct Answer: C. Stator flux and electromagnetic torque

Question 5

A PMSM has p = 4, ψ_f = 0.08 Wb, and neglects reluctance torque. What q-axis current produces 12 N·m?

A. 12 A

B. 20 A

C. 25 A

D. 40 A

Correct Answer: C. 25 A

Question 6

Which condition most strongly limits basic back-emf sensorless estimation at zero speed?

A. Excessive mechanical inertia

B. Very small back-emf

C. Large pole-pair number alone

D. High acoustic frequency

Correct Answer: B. Very small back-emf

Question 7

The PMSM reluctance-torque term is proportional to:

A. ψ_f i_d only

B. (L_d − L_q)i_d i_q

C. R_s i_d only

D. DC-link capacitance

Correct Answer: B. (L_d − L_q)i_d i_q

Question 8

Why is anti-windup needed in a current controller?

A. To increase sensor noise

B. To prevent integrator accumulation during voltage saturation

C. To eliminate the rotor flux

D. To reverse the phase sequence

Correct Answer: B. To prevent integrator accumulation during voltage saturation

Question 9

In field weakening, negative d-axis current in a PMSM is mainly used to:

A. Increase the inverter resistance

B. Reduce effective flux and satisfy the voltage limit

C. Eliminate all torque production

D. Increase zero-sequence voltage

Correct Answer: B. Reduce effective flux and satisfy the voltage limit

Question 10

A major classical-DTC disadvantage is:

A. No electromagnetic torque control

B. Mandatory mechanical position sensor

C. Variable switching frequency and torque ripple

D. Inability to operate an inverter

Correct Answer: C. Variable switching frequency and torque ripple