Direct Back EMF Detection for Sensorless BLDC Drives

Get Complete Project Material File(s) Now! »

Chapter III Improved Circuits for Direct Back EMF Detection

The back EMF detection method described in chapter II has to have minimum PWM off time. Therefore, it cannot go to 100% duty cycle. For some applications, the dc bus is low, like automotive applications. The bus voltage utilization is very important. An alternate back EMF detection is proposed in the chapter to solve the duty cycle limit problem.
For some applications, if the motor speed goes very low, the amplitude of the back EMF can be very low. In chapter II, Fig.2.18 shows a waveform when the amplitude of the back EMF is low. We find that the zero crossing is not evenly distributed. Theoretically, the zero crossing will happen every 60 electrical degree, so they should be evenly distributed. The inaccurate zero detection can cause wrong commutation, which probably could stall the motor at heavy load. Also bad zero crossing detection will cause bad speed regulation for close loop control because the speed feedback is based on the zero crossing.
The major reason is voltage drop across the diode in the inverter, biasing the back EMF signals. Also, the threshold voltage of the zero crossing comparison contributes to the unsymmetrical phenomenon to some extend.
In this chapter, the reason causing the unsymmetrical zero crossing is analyzed and an improved circuit is presented.
In high voltage applications, an unexpected signal delay issue for this sensorless scheme is identified. An improved circuit for high voltage applications is presented in this chapter as well.

Back EMF Detection During PWM On Time

In chapter II, we describe the direct back EMF detection during PWM off time. A true back EMF signal can be captured during that interval. However, one limitation for that method is that it cannot go to 100% duty cycle since we need minimum off time to have the time window to detect the back EMF. For some applications, it is not desirable. Can we detect the back EMF during PWM on time such that we don’t need that minimum off time?
We need to find out the terminal voltage signal during PWM on time for the floating phase. If phase A and B are conducting current, phase C is floating. The terminal voltage Vc is detected.
From equation 3.6, the terminal voltage equals to back EMF plus the half of dc bus voltage. If we compare Vc with Vdc/2, then we can get the zero crossing of the back EMF. To reduce the common mode voltage, we need to attenuate the signal. Since we only need to do this at high speed, the back EMF has high amplitude. The attenuation will not affect the sensitivity of the signal.
Therefore, if 100% duty cycle is necessary for some application, the back EMF detection can be done during PWM on time. At lower speed, the back EMF can be detected during off time. After the duty cycle is higher than certain level, the detection can be done during on time. The zero crossing reference changes to half of dc bus voltage, instead of ground.
This feature has been implemented into ST7MC, new version of ST72141 [25].

Improved Circuit for Low Speed/Low Voltage Applications

Biased Back EMF Signal

In the previous chapter, when we do the circuit analysis, we ignore the voltage drop in the diode and the switch. When the amplitude of the back EMF is very low, the voltage drop on these devices will affect the accuracy of the zero crossing detection.
Let’s do the circuit analysis again considering the voltage drop across the diode and the switch. For low voltage application, the power switch is low on-resistance MOSFET and the diode is the body diode of the MOSFET.
Same as in chapter II, the terminal voltage of the floating phase C is calculated during PWM off period.
If we ignore the second term of (3.13), the result is same as (2.6) in chapter II. However, at low speed and low voltage, the Back EMF itself is very small, the second term will play a significant role here. For low voltage MOSFET, Rdson is very low, Vmos can be ignored, so (3.11) can be rewritten as, Therefore, the voltage drop on the diode will bias the terminal voltage. When the back EMF ec is high enough at high speed, the impact of the second term of (3.12) is minimum, and can be neglect. At low speed, we need to consider the effect.
To verify the analysis, we use Mathcad to simulate the circuit considering voltage drop on diodes when the amplitude of the back EMF is low.
Fig.3.4 and Fig.3.3 show the simulation and the test results respectively. Fig3.3 shows that when sine waves are biased by a offset voltage, the zero crossings are shifted, unevenly distributed. The simulation and test results validate the analysis.
Meanwhile, at low speed, the slope of the back EMF near the zero crossing is very flat. The offset and the hysteresis loop of the comparator can shift the exact zero crossing point as well.

Improved Back EMF Detection Circuit for Low speed Applications

Complementary PWM

The first method to correct the offset voltage of the back EMF signal is to use complementary PWM. In previous chapter, the PWM is applied to high side only, shown in Fig2.5. When PWM is off, the current will freewheel through the body diode of bottom switch. The voltage drop across the diode causes biasing back EMF signal.
If the complementary PWM is applied, the freewheeling current will flow through the MOSFET instead of the body diode. Both motor terminal windings are tied to ground by MOSFETs, therefore, the offset voltage caused by body diode is eliminated. The PWM algorithm is shown in Fig3.5.
Another benefit of this complementary PWM, also referred to synchronous rectification, reduces the conduction loss during the current freewheeling period.
To implement the complementary PWM algorithm, dead time is necessary to prevent cross conduction between high side and low side MOSFET in the bridge. For ST72141, it has to implement this by external components. The new version ST7MC [25] will have built-in dead time generation.

Pre-conditioning Circuit to Correct the Biased Back EMF Signal

To eliminate the effect of diode voltage drop, we can also add a compensating t voltage to compensate the effect of the diode before the back EMF signal is sent to the comparator [15].
To alleviate problem caused by offset and hysterisis voltage of the comparator, we need to sharpen the slope of the Back EMF during the zero crossing period. We can use an amplifier to amplify the Back EMF signal.
A pre-conditioning circuit for back EMF zero crossing detection is developed to compensate the diode voltage drop and sharpen the slope of the back EMF signal near the zero crossing. The pre-conditioning circuit is shown in Fig.3.7.
TS274 is a high speed Op-amp. Chose R1 and R2 such that Vcon*R2/(R1+R2)=Vd/2. The positive input of the Op-amp is clamp at 0.7v by a diode because only signal close to zero crossing is of interest. So we only sharpen the slope of the back EMF near the zero crossing.
Therefore, this pre-conditioning circuit not only eliminates the offset caused by voltage on diode, but also sharpens the slope of the back EMF signal by amplifying the signal.
In Fig.3.8, the amplitude of the back EMF is very low. If there is no pre-conditioning circuit, the zero crossing point is A. the pre-conditioning circuit not only compensates the offset, but also sharpens the slope of the signal near the zero crossing. The zero crossing point is detected at B with the pre-conditioning circuit. Pay attention that amplitude voltage scaling for two signals is 10 times difference.
This improved circuit further extends the application of the sensorless BLDC drives. For conventional back EMF detection scheme, the speed range of the motor (maximum speed verse minimum speed) is typically about 5:1. For the proposed back EMF detection scheme without the pre-conditioning circuit, the speed range can be up to 10:1. The speed range can be extended to at least 50:1, if the pre-conditioning circuit is used. Fig3.10 shows the three-phase pre-conditioning circuit.

Improved Circuit for High Voltage Applications

In chapter II, it is mentioned that three phase terminal voltages are directly fed into the microcontroller through resistors, which limit the injected current. For different voltage applications, we need to choose different resistance accordingly. The injected current should be limited around 2 mA. For some household appliance applications, the dc bus voltage can be 300 v. The resistor value is chosen 160kΩ.
We find that zero crossing detection is not correct sometime when PWM duty cycle is high. The problem is caused by the large time constant of the current limit resistors. Inside the microcontroller, there is some parasitic capacitance. Since the outside resistance is high enough, even though the capacitance is low, the effect of RC time constant will show up. Fig.3.7 shows the winding terminal voltage Vmc and the voltage VC at the input pin of the microcontroller.
It is clear that VC has different delay time at the rising edge and the falling edge. The falling edge can last about 10us and the rising edge is very fast. The reason that causes the different slope is different exciting voltage source for the RC circuit. At the rising edge, the exciting source is motor terminal voltage. Since the terminal voltage is high (~300 v), the injecting current can charge the capacitor quickly. So the rising edge is short. At the falling edge, discharging source voltage is only 5v, thus, the discharging current is very small. Therefore, the falling edge is far longer than rising edge.
The back EMF signal is sampled at the end of PWM off time. If the PWM duty cycle is high enough such that the off time is less than 10us, the sampling result is not correct because the discharging period hasn’t finished yet.
To shorten the discharging time, we should reduce RC time constant. We can’t change the capacitance, but we can change the resistor value. Fig.3.13 shows a circuit that different time constants for rising edge and falling edge. At the rising edge of Vmc, Vmc will charge the capacitor through the 160k resistor. At the falling edge of Vmc, the discharging period, and the diode will forward conduct and the capacitor will be discharged through 20k resistor. Fig.3.14 shows the test result of circuit with variable RC time constant. The discharging time is about only 1us.
Fig.3.15 shows the schematic of the improved circuit of back EMF detection for high voltage applications.

Summary

In this chapter, improved circuitry for low speed /low voltage and high voltage applications are described.
The limitation of the back EMF detection during off time is that there is a requirement to have a minimum off time. Sometime this is undesirable for low voltage applications. An alternate back EMF detection during on time is proposed to make the duty cycle up to 100%, thus the bus voltage can be fully utilized.
The complementary PWM algorithm is proposed to eliminate the offset voltage caused by the body diode voltage when current is freewheeling through the diode. The offset voltage can cause untruthful zero crossing detection, especially at low speed.
The pre-conditioning circuit for low speed applications not only compensates the offset voltage caused by diodes, but also sharpens the slope of the back EMF signal near the zero crossing.
The variable time constant of the back EMF sensing circuit avoids the long falling edge of signals caused by large RC time constant.
The improved circuitry greatly expands the motor speed range. For example, for a 48 V motor, the speed operation range can be from 50rpm to 2500rpm. For high voltage operation, the sensorless system has been successfully used for 300v/30,000 rpm high-speed air blower applications.
Starting the Motor with the Sensorless Scheme

Introduction

The sensorless schemes are not self-starting. In order to sense the back EMF, the motor must be first started and brought up to a certain speed where the back EMF voltage can be detected. In practice, open-loop [16] starting the motor is accomplished by providing a rotating stator field with a certain increasing frequency profile. Once the rotor field begins to become attracted to the stator field enough to overcome friction and inertia, the rotor will turn. After the speed reaches a threshold voltage value, the back EMF can be detected, providing the position information, the system switches to synchronous commutation mode and the motor acts as a permanent magnet synchronous machine. If there is no specific requirement for start-up, like fans application, this open loop start-up can be satisfactory. However, for some applications, i.e., automotive fuel pump, the start-up has to be finished with 200ms to build up the pressure. It is very difficult to tune the start-up using the open-loop starting algorithm. On the other hand, if the starting torque is medium or high, usually it is difficult to start the motor with the open-loop algorithm.
In this section, a practical start-up tuning procedure with the help of a dc tachometer will be descried. The commutation during start-up is tuned in this way such that the motor speed has the maximum acceleration. This procedure is generally applicable for all different sensorless BLDC systems.

Test set-up

Connect motor with the load/dyno and a dc tachometer along the shaft through joint couplers. The output of the tachometer is a dc voltage that is directly proportional to the speed. From the tachometer voltage, we will know the instant speed of the motor. Fig. 4.1 shows the set-up.

Start-up Tuning Procedure

When the motor stops, the controller doesn’t know the rotor initial position. The first step is to align the motor to a known position by exciting two phases of the motor. For instance, we can choose phase A and phase B to be excited to set the initial position. After the rotor is in the initial position, a preset exciting pattern will be sent out. If three phases are alternately excited, the motor will start to accelerate. Table 4.1 and 4.2 show the exciting pattern for forward /backward rotation. The motor is driven with 6-step mode, and the exciting phase just repeats the same pattern after one cycle (6 steps).
The question is how to set the right time for each step. If we know the instant motor speed, we set the time for each step so that the motor speed keeps accelerating. We can get the instant motor speed from the dc tachometer.
We take forward rotation as an example for the start-up tuning.
Before starting, the position of the rotor is unknown. A pre-positioning is needed to place the motor in a known position. The pre-positioning is also called alignment. After phase A and phase B is excited, the rotor will align with the flux direction generated by phase A and phase B.
When the rotor approaching the alignment position, it will oscillate. The output of the tachometer will tell how long the oscillation lasts. Fig.4.2 shows current of phase A and the oscillation waveforms during the pre-positioning period. In order to reduce the oscillation, a progressive ramp-up current can be set to bring the rotor into the desired position. Applying a strong current level directly to the windings will make the rotor move more quickly and in turn this will make it oscillate more severe around the final position.
The pre-positioning period has to be long enough that the oscillation stops. Otherwise, the rotor will be at unknown position if it is still oscillating. In Fig.4.2, the time period, from T0 to T1, is the pre-positioning period. The oscillation stops before the end of pre-positioning. The tachometer signal shows the oscillation of the rotor.
After the pre-positioning phase is finished, the motor can be commutated to first step, phase A and C conducting current. At beginning of this commutation, the motor will start to turn because accelerating torque is produced. However, If this step lasts too long, the motor speed will first increase and then decrease. Fig.4.3 shows the waveform.
From the waveform, the speed of the motor rises from time T1, until time T2, since this step is set too long. So we should set time for this step as T2-T1. At time T2, the motor should commutate forward to next step.
The second step is when phase B and phase C are excited according to Table 4.1.
Similarly, we set the step time very long first and watch the output of the tachometer.
From Fig.4.4, we can find the right time, T3-T2, for the second step.
Continue to do the tuning, we can get the right time for following steps until the microcontroller can detects the back EMF and switches to synchronous commutation mode. With the help of the tachometer, we can have the best acceleration during the start-up.

Table of Content
Chapter I Introduction
1.1 Background
1.2 Brushless DC (BLDC) Motors and Sensorless Drives
Chapter II Direct Back EMF Detection for Sensorless BLDC Drives
2.1 Conventional Back EMF Detection Schemes
2.2 Proposed Direct Back EMF Detection Scheme
2.3 Hardware Implementation of the Proposed Back EMF Detection Scheme
2.4 Key Experiment Waveforms
2.5 An application Example: Automotive Fuel Pump
2.6 Summary
Chapter III Improved Circuits for Direct Back EMF Detection
3.1 Back EMF Detection During PWM On Time
3.2 Improved Circuit for Low Speed/Low Voltage Applications
3.3 Improved Circuit for High Voltage Applications
3.4. Summary
Chapter IV Starting the Motor with the Sensorless Scheme
4.1 Introduction
4.2 Test set-up
4.3 Start-up Tuning Procedure
Chapter V Conclusions and Future Research
5.1 Conclusions
5.2 Future Research
Reference
GET THE COMPLETE PROJECT