Tightly Integrated Devices Yield More Efficient Motor Control

Motor controls are comprised of the power electronics that control the application of voltage and current to the motor. Digital motor controllers manage the power electronics to achieve the target motion results such as speed or torque at a system level. The control algorithms implemented within the motor controller significantly influence the overall motor system efficiency.  

The most basic way to control a motor is by controlling voltage and current supplied. The most simple and primitive motor drivers use a linear amplifier which regulates the voltage supplied to the motor in a linear fashion. For example, if it is desired to deliver 10 amperes of current across a motor at zero speed that has only 1 ohm of phase-to-phase resistance, then only 10 volts is needed from the motor controller. If the linear motor controller is supplied with 50 volts, for example, then 40 volts multiplied by 10 amperes, or 400 watts is dissipated in the controller. Very few applications implement linear controllers because they are terribly inefficient.

PWM or Pulse Width Modulation schemes, on the other hand, depend on motor inductance and overall reactance to pulse the entire supply voltage across the motor at high speeds with duty cycles proportional to the average voltage they wish to produce (Figure 1). While PWM controllers introduce eddy-current losses in the motor, those losses are significantly lower than the improvement in efficiency gained within the controller using PWM techniques. The PWM controller has essentially three power states: a) Off, where the output power electronic switch is not conducting any current; b) On, where the output power electronic switch is conducting current, however, with very little voltage across it (1 to 3 volts); or c) switching from either off to on or on to off, where the power electronic switches incur switching losses proportional to how fast they can switch the current and how fast the voltage falls and rises across them. All of these states are relatively very low loss states compared to the linear controller.

With today’s processing power, it should be clear that digital control is really the only choice. Today’s microcontrollers are more capable than the digital signal processors of just 10 years ago. These devices are designed for real-time embedded applications and include peripherals and mechanisms to help ease the engineer’s job at applying them to control motors.

Unlike AC induction motors, where loss is created due to slip, permanent magnet synchronous motors (PMSM) are far more efficient, as they are synchronous machines. PMSM fall into several categories that impact performance, value, speed, motor constant and power density. The leading motor types are surface magnet (SM), interior permanent magnet (IPM) and axial flux, which can be either SM or IPM. The SM motor provides a robust platform that is relatively easy to control.

The IPM motor has the permanent magnets buried within rotor structure, which is made up of iron laminations stacked along the rotor. The introduction of the iron laminations adds rotor inductance that can be utilized to more effectively control the motor and effectively manipulate its characteristics and parameters. This added measure of control, however, comes with a price of control complexity when compared to SM motor control algorithms.

The control algorithm of choice for most permanent magnet motors today is the field-oriented control (FOC) algorithm. This algorithm “componentizes” the id and iq current and voltage vectors and closes the control loop around them after they have been subjected to coordinate transforms that reference them to the rotor’s angular position. Permanent magnet motors coupled with effective control algorithms are the most efficient combinations available today.

Figure 2 shows the block diagram of Field Oriented Control for a PMSM. Three measurements are taken as feedback from the motor: phase currents and position of the rotor. There are many ways of obtaining feedback to calculate position, such as a Hall effect sensor, encoder or sensorless using back emf. By measuring only two phase currents (ia and ib), the stator current vector which is comprised of 3 components ia, ib and ic can be constructed, this is because ia + ib + ic = 0.  The Clarke transform is used to convert these three current components into orthogonal, 2 axis coordinates, iα and iβ. Based on the position of the rotor, the angle θ can be calculated and is used to bias iα and iβ within the Park transform to essentially null the effect of rotor angular change, making the output of the Park transform a DC component relative to shaft angle.

The resultant current vector is mapped into a d-q axis as shown in Figure 2. These values are fed back into a PI (Proportional + Integral) controller as current error. The output of the PI regulator is a command for voltage. However, it is referenced to the rotor’s angular position. Before we can output voltage to the stator, we must now reference the voltage back to the stator, which is stationary. The Inverse Park transform accomplishes this by doing the inverse of the Park transform, just as the name suggests. The transform takes the rotor’s angle and offsets the id and iq voltage vectors into iα and iβ voltage vectors that now include amplitude components for the rotor’s angle. Finally, a Space Vector Pulse Width Modulator or SVPWM converts iα and iβ into three voltage vectors, Va, Vb and Vc.
As shown in Figure 3, an FOC control system is designed to keep id and iq orthogonal, which helps to ensure optimal angle control and better current utilization in producing torque. In order to ensure perpendicularity between these vectors, the algorithms must be computed quickly to minimize computational delays. Delays essentially obsolete the voltage vectors and decrease the effective efficiency because the ideal 90 degree angle between id and iq is not maintained.
In the past, because of the unavailability of fast controllers and digital devices, implementing FOC algorithms was only possible through the use of expensive, high-end DSPs or with assembly language. This constraint and the added cost of sensors have deterred designers from adoption in many applications that could benefit from the gains in torque management, control and efficiency. However, with the introduction of many complex and integrated electronic devices such as MCUs, lower cost DSPs and FPGAs, computational power and cost have come a long way in providing the power to make these methods available to even simple control applications. Implementing many of the control structures in hardware description language (HDL) in tandem with lower cost processors provides an excellent opportunity to segment the computing tasks into pieces appropriate for the subsystems in today’s SoCs. Figure 4 shows how this might be implemented.
For example, placing the Clarke, Park, PI controller and SVPWM in HDL would allow an MCU to pass off these calculations to hardware algorithms that can quickly crunch the numbers and spew back results, minimizing computational delay and reducing processor costs and computational requirements. In addition, hardware based actions that are clock driven provide faster system level protection for high-speed fault monitoring and protection schemes, making products more robust in even the harshest of environments.
Implementing FOC on Actel’s SmartFusion Intelligent Mixed Signal FPGA

SmartFusion devices integrate the three components essential to implementing motor control on a single device: an FPGA, ARM Cortex-M3 microcontroller and programmable analog subsystem. An analog-to-digital converter is used to capture motor feedback, such as phase current or back EMF voltage, while the Cortex-M3 microcontroller is used to manage various tasks such as sampling and communications. The FPGA enables implementation of algorithms in hardware, especially those that are system- and time-critical.  When implementing for motor efficiency, a user can partition the FOC algorithm between hardware and software. For example, system-critical components that require instantaneous response, such as critical fault detection and other supervisory functions, should be placed within the FPGA fabric.  Any tasks that are computationally intensive and iterative, such as PWM generation, should also be placed in the FPGA fabric.   Because these three components work across a high-speed 32-bit bus, off chip-delays are eliminated, yielding fast and flexible motor control implementation.