Permanent-magnet synchronous motors (PMSM) provide a viable alternative to ac induction motors in many variable-speed applications. Advertised as “ac servo” or “dc brushless,” PMSM motors are traditionally found in positioning applications like robotics, machine tools, and high-speed packaging equipment. Given the steady advancements in drive technology over the past 20 years — integrated ASIC, DSP, and power transistor technology — PMSM is experiencing new success in variable-speed applications. The main reason PMSM is so attractive is due to its physical construction, which consists of permanent magnets mounted onto the rotor. This arrangement has several advantages as shown in the following table.
For example, PMSM motors can accommodate large airgaps between the rotor and stator. In some chemical pump and vacuum feed-through applications, this capability is exploited by adding non-magnetic material into the airgap to create a pressure, chemical, or environmental seal. As a result, magnetic couplings and fluidic seals are eliminated and the size and cost of the end-product reduced.
Another example can be found in laboratory environments, where PMSM technology is ideal for high-speed centrifuge operations, particularly bench-top products requiring small motors and high-speed operation. Not only do PMSM motors fit the bill, their ability to rapidly accelerate and decelerate high-inertial loads minimizes processing time.
A typical PMSM drive circuit contains an input diode bridge that rectifies ac-line voltage and a bank of capacitors, which then filter it. Together, these passive components form a simple ac-to-dc converter.
The right side of the conversion circuit is a “full three-phase bridge” and converts dc-to-ac. High-side transistors apply positive voltage to the motor phases, while low-side transistors apply negative voltage. By controlling bridge transistors' on and off states, the drive causes current to flow in or out any of the three motor phases.
Digital PMSM drives control all transistors through pulse width modulation (PWM). Analog feedback, whether voltage or current, is immediately digitized before being processed. Software processes all motor control functions, including modulation, field orientation, current, and velocity loop control. Digital control improves drive reliability by reducing the number of discrete components found in early designs and facilitates advanced algorithms for optimal motor performance.
Current and modulation
In a PMSM drive, PWM voltage regulates phase current, which in turn produces motor torque. Here's an explanation for one common PWM method: Applying voltage by turning on the high-side transistor produces the current I1(t). The phase-to-phase inductance, Lm, acts like a current filter, storing energy while I1(t) is increasing. When the applied voltage is removed, the energy stored in Lm then produes a fly-back voltage that generates a second current component, I2(t). As I2(t) decreases, energy is released from Lm.
The average level of I1(t) and I2(t) combines to form torque-producing current. In addition, the level of ripple current present produces unwanted I2R loss in the stator, and induces magnetic loss in the rotor. Many different PWM methods exist to address this problem. In high-performance PMSM applications, specialized drives minimize ripple current using advanced PWM techniques.
The magnetic circuit
The magnetic circuit of the PMSM motor is similar to an ac induction motor. Both motors utilize a stator assembly with specially distributed phase windings connected in either a wye or delta fashion. Stator laminations minimize airgap reluctance, facilitating a high level of flux coupling between the rotor and stator.
The fundamental difference between PMSM and ac induction is how magnetic poles are produced on the rotor. An ac induction motor induces magnetic poles that travel along the rotor's surface, a process that requires a small airgap and consumes a component of applied motor power. Conversely, PMSM motors create stationary poles on the rotor using fixed high-energy magnets. Permanent magnet rotor construction supports larger airgaps, reduces the rotor's inertia, and increases motor efficiency by eliminating power consumption associated with ac induction.
Field oriented control and torque
PMSM motors are three-phase machines that generate torque as a function of phase-current amplitude and stator geometry. Modern digital drives use Field Oriented Control (FOC) to separate the problem of torque production into two parts: aligning the applied stator field with the PM rotor and controlling the intensity of the applied field to regulate torque.
The process of field alignment begins by measuring phase current in the stator's U-V-W coordinate system. A vector transformation is used to calculate the amount of stator current that directly aligns with the rotor and the amount at a right angle to it. These two transformed measurements are commonly referred to as Id and Iq. Both are vector components of stator current that exist in the rotor's D-Q coordinate system.
PMSM torque production is optimal when no component of current is present along the D axis. To impose this condition, a current-control loop sets Id to zero. This control loop outputs a vector component of voltage that compensates for stator reactance. In turn, the Iq control loop then regulates torque-producing dc current along the rotor's Q axis. The Iq control loop outputs a vector component of voltage that compensates for the motor's back EMF and IR drop. FOC ends when both vector components of voltage are transformed from the rotor's D-Q coordinate system back into the stator's U-V-W coordinate system. U-V-W voltages are then applied to the phase leads using PWM.
Velocity feedback and control
In most brushless drives, velocity feedback v(t) is derived from position feedback P(t). The most common methods are differentiating P(t) or measuring time t to traverse the incremental distance dP. In either case, v(t) is compared to a desired velocity, the difference of which is velocity error.
A high-order control filter then forces the velocity error to zero. The filter's output feeds into the Iq loop and becomes a torque-command signal.
Any control filter calculating this signal introduces a time-delayed response as it interacts with the load and tries to eliminate velocity error. This means that velocity error is never zero when either the desired velocity or load torque constantly changes in time. For this reason a torque feedforward signal is commonly added to the filter's output to minimize velocity tracking error under dynamic conditions.
Improving PMSM operation
Drive selection affects PMSM motor performance. If not carefully considered, drive selection can limit precision, bandwidth, and efficiency in demanding applications. For instance, most high-speed PMSM motors run at elevated voltages and exhibit low inductance for maximum power output. To prevent motor overheating, the selected PMSM drive must regulate average current into a low-inductance load and minimize ripple current for efficient operation.
When precision is required, the selected drive must also exhibit sufficient numerical resolution to avoid signal quantization error. Another requirement is drive capability to appropriately scale control variables for smooth, quiet, and high-gain motor operation. In addition, feedback signals must be optimally processed to achieve high accuracy and repeatability.
Another area of concern is in applications with rapid acceleration and deceleration, which require high-bandwidth control. Here, the selected drive must operate control loops at high frequency for stability. A specialized control filter or feedforward technique may be necessary to control a highly dynamic load.
Mapped encoder-feedback correction
One example of a specialized PMSM drive technique is known as mapped encoder-feedback correction. Small-diameter motors are required in the optics industry to deliver precise velocity control and constant angular accuracy. However, as motor diameter decreases, the affects of encoder gradient (edge-distance) error and motor shaft run-out are amplified. Mapped encoder-feedback correction can be considered in this case.
A map is a table of coefficients permanently stored in the drive. As the motor rotates, position feedback selects coefficients from the table that help correct measured velocity. For example, if edge distance at an encoder position is known to be long, the calculated speed at that edge is multiplied by a coefficient greater than one. Conversely, if edge distance is short, the calculated speed is multiplied by a coefficient less than one.
Mapped encoder feedback correction can improve feedback accuracy by more than one order of magnitude. Also, this technique significantly reduces harmonic noise, enabling high-precision control.
Adaptive torque feedforward
Velocity tracking error is minimized when a torque feedforward signal is added to a velocity control filter's output. Many different methods for calculating a torque feedforward signal are possible. For example, during rapid acceleration, differentiating the “desired velocity” input signal produces a torque feedforward signal. This works well when the load is inertial and rigidly coupled to the motor.
In some applications, however, velocity remains constant while the load torque changes rapidly as a function of position. Two examples are peristaltic pumps for medical applications and CAM-driven mechanisms for packaging equipment. In such cases, a specialized technique known as adaptive torque feedforward may be considered. Here is one way to implement it.
As the load mechanism operates through one machine cycle, the output of the velocity control filter (the torque command signal) is recorded in a table as a function of position. On the next machine cycle, the previously recorded torque function is added to the control filter's output. The recorded torque function from the first cycle helps eliminate any systematic or repeating velocity error in the second cycle. This method is adaptive when recording and playback repeat continuously. Each time the motor travels through a new cycle, the control filter's output is averaged with previously recorded torque functions. In this way, the torque feedforward signal slowly adapts to changing load conditions over time and temperature.
Low package weight
Less heat generation
High torque-to-diameter ratio
Large airgap possible
Accuracy: speed or torque
Repeatability: speed or torque
Stiffness of regulation
Smoothness of operation
PMSM drive circuitry
Permanent-magnet synchronous motors convert mechanical energy returning from the load to electrical energy stored within the drive, a process called regeneration. During regeneration, charge (or excess energy) accumulates in the drive's dc-supply capacitors; input diode rectifiers prevent current from returning to the ac line. A shunt regulator circuit dissipates energy during regeneration for most PMSM drives in the 0.25 to 10-hp range.
Output bridge transistors apply voltage to a PMSM motor. PWM (or “on-off” control) drives individual high and low-side transistors. Motor inductance, Lm, acts as a current filter. The controller adjusts PWM on-time to regulate an average level of current flow.
PWM current waveform
The average current in a PWM waveform translates to torque-producing current. Ripple current, on the other hand, generates additional copper loss in the stator and induces magnetic loss in the rotor. Specialized PWM techniques can minimize both effects.
Four-pole PMSM motor
The magnetic structure of a four-pole PMSM motor consists of phase windings that are distributed spatially in lamination slots. Laminations reduce reluctance making it easier for flux to conduct between the rotor and stator.
Vector diagram for a two-pole PMSM motor
A vector diagram of a PMSM motor is key to understanding FOC. Stator voltage (PWM) and current are real quantities that exist in the U-V-W coordinate system. FOC transforms these quantities into equivalent values in the rotor's D-Q coordinate system, where all control is performed.
Field oriented control
This signal flow diagram shows how FOC works when applied to a PMSM motor. Torque production is optimal when no vector component of current exists along the rotor's D axis. Setting Id to zero forces this condition and produces a vector component of voltage to overcome the motor's reactance.
Velocity control loop
Position or time differentiation results in velocity feedback. The difference between velocity feedback and desired velocity produces velocity error, which is forced to zero by a PID filter. This output feeds into the current loop and becomes a torque signal.
Mapped encoder feedbackcorrection
A modified version of the velocity control diagram, this feedback method improves precision in laser scanning and similar optical applications.
When load profiles repeat as a function of position, this adaptive feedforward technique can be considered. Here, a table is established in the drive's memory to record the velocity control filter's torque command signal. As the load cycles, this signal plays back to minimize velocity-tracking error.