issue: November 2006 APPLIANCE Magazine
Using Sensorless Control to Catch a Spinning Fan
Email this Article
by Aengus Murray, International Rectifier
Over the years, engineers have turned to brushless DC drives with Hall sensor feedback due to their simple control circuitry. However, advancements in sensorless control offer designers alternative solutions.
A permanent magnet (PM) motor is the motor of choice for variable speed fans in air-conditioning applications. A traditional single-phase induction motor can operate with an efficiency as low as 20 percent when run at variable speed. In contrast, a permanent magnet AC motor can run with an efficiency of at least 85 percent. The high efficiency not only saves energy, but also enables a smaller motor design since the PM motor dissipates a lot less heat. The power rating of the fan is relatively small compared to the compressor power rating, but its energy consumption is still very important since it runs continuously.
U.S. air-conditioning manufacturers need to maximize the system efficiency over the full cooling range since the calculation of the Seasonal Energy Efficiency Ratio (SEER) rating is a weighted average of the efficiency over a range of cooling powers. Japanese manufacturers are now paying more attention to the fan efficiency since recently introduced government standards measure seasonal efficiency rather than the efficiency at rated cooling.
The traditional permanent magnet fan motor uses Hall Effect sensors for position feedback. A brushless DC motor generates trapezoidal shaped back EMF waveforms, as shown in Figure 1. Over a 60-degree electrical period, the voltage between one pair of windings is substantially constant, and the motor behaves just like a DC motor. As the rotor magnet turns, different pairs of windings will enter the flat-topped portion of their back EMF waveform. The Hall Effect sensors detect the position of the rotor magnet and provide the signals to the commutation circuit to select the appropriate pair of windings. Varying the voltage across the windings using pulse width modulation (PWM) controls the speed. The commutation and PWM control circuits require just a few logic gates, which makes the control hardware very inexpensive. In small fan motors, the power, control and Hall sensor components are typically all on a small circuit board mounted on the back of the motor.
Figure 1. Trapezoidal back EMF motor commutation sequence
Sensor Versus Sensorless
The simplicity of the control circuits has resulted in the widespread use of brushless DC drives with Hall sensor feedback. However, a number of performance and reliability issues have caused manufacturers to look for alternative solutions. One problem is the torque glitch caused by the non-ideal behavior of the brushless DC motor during commutation.
In Figure 1, the motor winding currents shown are rectangular with an almost immediate transition of current from one winding to the next at commutation. However, the real motor windings have inductance; therefore, there is delay in the transition of the current between windings. This delay becomes more significant at higher speeds as the back EMF in the incoming winding opposes the increasing current while the back EMF in the outgoing winding supports the decay of the current. The resultant dip in motor current generates a torque glitch at every commutation sequence. The torque glitch generates unwanted acoustic noise that has become more of an issue over the years as designers have continuously improved fan designs to minimize the wind noise. The problem is worse because inaccuracies in the alignment of the Hall sensors mean that the torque glitch magnitude varies between commutation points. The reliability of the Hall sensor assembly is another problem, especially in larger fans where the Hall sensor has to be mounted on a separate circuit board, close to the rotor.
Figure 2. Sinusoidal sensorless control algorithm
Advances in electronic circuit integration have enabled the introduction of more sophisticated control algorithms to solve the aforementioned issues. One innovation is the introduction of sinusoidal voltage control to eliminate the torque glitch at commutation. This requires a change in the motor design so that it generates sinusoidal back EMF waveforms. The application of sinusoidal voltages to the motor windings produces sinusoidal currents and smooth torque. This approach is very common in industrial servo drive systems, but only using a high-resolution position sensor such as a resolver. However, interpolating between Hall sensor switching points using a phase-locked loop produces the high resolution angle information needed to synthesize a smooth sine wave shape. A number of fans are available today with an embedded low-noise control circuit.
Another innovation is the elimination of the Hall sensor by using the motor windings voltage feedback signal, commonly called sensorless operation. When driving a brushless DC motor with a six-step commutation sequence, one winding is unconnected, so the zero crossing of the back EMF can provide a commutation signal. However, while this improves reliability, it does not solve the acoustic noise problem associated with six-step commutation. An indirect measurement of the back EMF is possible with sinusoidal current control based on the motor circuit model. The two-phase circuit model in the equation below simplifies the calculations since the rotor flux functions include simple sine and cosine terms. The stator winding voltage equations consist of the voltage across the inductance and resistance and the back EMF generated due to change in rotor flux coupled by the windings.
The voltage terms in the equation derive from the applied stator voltages, while the current terms derive from the measured winding currents. The application of integration along with algebraic manipulation yields the sine and cosine flux terms. The phase-locked loop decodes these trigonometric functions into rotor angle and rotor angular velocity.
The complete sinusoidal current control algorithm shown in Figure 2 has the well-known field-oriented control architecture often found in industrial drives. The vector rotation and Clarke blocks (ejq) transform the three-phase winding currents into two components producing torque (IQ) and flux (ID). The outer speed loop calculates the torque current needed to maintain the target speed while two current loops calculate the stator voltages needed to maintain the target current components. A feature of this control structure is that setting the flux current reference to zero maximizes the efficiency of motor operation. This is an added advantage of this control system over the sinusoidal voltage drive described previously.
Figure 3. Start up while turning at 300 rpm and ramp up to 1,800 rpm
The aforementioned algorithm operates over a wide speed range, but cannot operate down to zero speed. The problem is that at near-zero speed, the back EMF signal is much too low to detect and is non existent at start up. A special start-up sequence slowly ramps up the frequency of the winding currents to some minimum speed using frequency integration to estimate the rotor angle. Appropriate selection of the starting current generates the correct torque to accelerate the rotor at the same rate as the frequency ramp. However, even if the initial accelerating torque is too high, the phase advance of the rotor angle will reduce the torque until they exactly track.
The starting sequence works well when starting from zero speed; however, when starting a fan load, you cannot guarantee that the fan will not already be rotating. It is quite likely that the wind will drive the outdoor fan while it is unconnected. The speed can be very high when there is a typhoon or hurricane, and the application of the normal starting sequence would create a very large torque disturbance. In this case, the speed and direction of the outdoor fan must be determined before startup is attempted. If the fan is turning in the reverse direction, the controller applies braking torque before starting the fan in the correct direction.
The controller detects the initial speed and direction by closing the current control loops with zero torque and flux reference. The feedback loops force the applied stator voltages to track the winding back EMF so that no current flows. The rotor position estimator thus acquires the back EMF signals and derives rotor position and angle. This allows the controller to acquire the fan speed and direction without introducing torque disturbance. If the fan is already turning in the correct direction, as shown in Figure 3, the current controller starts operating in normal mode, and the motor accelerates to the target speed. If the fan is turning in the reverse direction, then the controller turns on the three lower inverter transistors, thus braking the motor. Once the motor stops, the controller starts the motor using the normal starting sequence and accelerates the motor to the target speed.
About the Author
Aengus Murray joined International Rectifier in 2005 and is involved in product marketing and motor control development efforts. He has a bachelor’s degree in Electrical Engineering and a Ph.D. in motor control. Murray has also worked at Kollmorgen Industrial Drives, Dublin City University and Analog Devices Inc. If you wish to contact Murray, e-mail: firstname.lastname@example.org