FAIRCHILD SEMICONDUCTOR POWER SEMINAR 2008 - 2009 BLDC Ripple Torque Reduction via Modified Sinusoidal PWM Shucheng Wang Abstract — This paper introduces a BLDC fan motor driver system using modified sinusoidal PWM (SPWM). The driver board is assembled in the motor to minimize system size. Based on the signal from the hall position sensor on the board, the controller generates the SPWM to the power module to drive the BLDC motor. Because it's SPWM instead of square PWM, there is no pulsation torque. The simulation and experimental results are presented to verify the stability of the driver system. I. INTRODUCTION The permanent magnet (PM) motor has many advantages. Compared to DC motors, they require lower maintenance due to the elimination of the mechanical commutator and they have a high-power density, making them ideal for high torque-toweight-ratio applications. Compared to induction machines, they have lower inertia allowing faster dynamic response to reference commands. They are more efficient due to the permanent magnets, which result in virtually zero rotor losses.[1] There are two kinds of PM motors, the permanent magnet synchronous motor (PMSM) and the brushless DC (BLDC) motor. They both have a permanent magnet on the rotor and require alternating stator currents to produce developed torque. The difference between the two is that the PMSM and the BLDC have sinusoidal and trapezoidal back-EMFs, respectively. The BLDC motor has the advantage of being a simple machine with higher power density, simple discrete position sensors, and simple control compared to a sinusoidal machine[2]. However, its disadvantage is the pulsating torque problem[3]. In theory, these two machines can be driven by either sinusoidal or rectangle state current. Practically, the PMSM requires sinusoidal stator currents to produce constant torque. For a BLDC motor, due to finite phase induction, the sum of the commutating currents is never constant and this is the reason for the generation of pulsation torque. The increment of current in one phase as a result of the other twophase commutation can sometimes be fatal. An improved implementation of direct torque control (DTC) to a permanent-magnet, brushless DC (BLDC) drive is introduced in reference [4]. The commutation torque ripple is minimized by combining the conventional two-phase switching mode with a controllable three-phase switching mode during periods when the phase currents are being commutated. A strategy for reducing commutation torque ripple in a position sensorless BLDC motor drive is proposed in reference [5]. The proposed method directly measures the commutation interval from the motor terminal voltage waveforms and does not require a current sensor and current control loop. However, when the duty cycle of the PWM is high, the compensation effect is not good, since the duty for commutation is limited to the maximal PWM duty. To compensate the duty of commutation, the PWM cycle is modified to synchronize the end of the point of commutation. This complex computation requires advanced MCU/DSP to achieve. A low-cost method to minimize the pulsating torque is introduced in this paper. In this method, the sinusoidal terminal voltage is employed. This induces a sinusoidal current with harmonics because of non-sinusoidal back-EMF. The harmonics and optimal angle for the voltage PWM calculation has been deduced in this paper. The torque ripple comparison between the proposed method and traditional six-step control for BLDC is shown by simulation. Compared to others, this method doesn’t require a current-sampling circuit, complex computation, or precision encoding. However, this method can’t be used where good dynamic performance is required because there is no current 1 FAIRCHILD SEMICONDUCTOR POWER SEMINAR 2008 - 2009 control loop or precision encoding. The testing was conducted on a fan motor used on an air-conditioner that requires high efficiency, quiet running, and low cost. The advantage of this method is to meet the requirement. The BLDC motor has been more and more popular for the outdoor fan application field because of low cost and high efficiency. K e ( −θ r ) = −K e (θ r ) where and dλe (θ r ) dθ r K e (θ r ) ≡ ωr ≡ dθ r dt (6) is the back-EMF coefficient is the rotor electrical speed (rad/s). Figure 1a shows the ideal Ke(θ r ) of the BLDC motor. II. BLDC MOTOR D,Q MODELING Because the BLDC motor is fed with a sinusoidal voltage, the harmonics can be analyzed in a d,q model[6][7]. Since there is different back-EMF between the BLDC motor and the PMSM, the d,q model of BLDC is unlike that of the PMSM. The harmonics of the back-EMF are analyzed in this section. The harmonics of the current and the harmonic torque induced by the harmonics in current and back-EMF are also presented. Finally, the optimal input voltage angle is introduced. A. Modeling of Back-EMF The BLDC motor is made with trapezoidal backEMF and the model is different from the PMSM. The BLDC motor is generally assumed to have three balanced phases connected in a Υ configuration; the N and the P pole of the PM is symmetrical in electrical position; and each pole pair is symmetrical in mechanical position. Based on the previous assumption, the flux linkage characters can be summarized as: λam = λ (θ r ) λbm = λ (θ r − α ) λ = λ (θ + α ) r cm (1) λ (θ r + π ) = −λ (θ r ) (2) λ ( −θ r ) = λ (θ r ) (3) where λam , λbm , and λcm are the phase a, b, and c flux-linkage due to permanent magnet, respectively; θ r is the electrical position of rotor; and α ≡ 2π / 3 . The back-EMF character of each phase can be derived from Equations 1, 2, and 3 as: ea = ωr K e (θ r ) eb = ωr K e (θ r − α ) e = ω K (θ + α ) r e r c (4) K e (θ r + π ) = −K e (θ r ) (5) Fig. 1. Back-EMF in the a b c frame and d,q frame. The voltage equations for the PM motor in the synchronous reference frame (SRF) are: r v ds rs + pLs r = v qs ωr Ls ea r − ωr Ls i ds r + T (θ r )eb rs + pLs i qs ec (7) where Ls and rs are the stator equivalent inductance and resistance, respectively, and T (θ r ) is the transformation matrix from three-phase to SFR. The definition of T (θ r ) ≡ 2 cos θ r 3 − sinθ r T (θ r ) is: cos(θ r − α ) cos(θ r + α ) − sin(θ r − α ) − sin(θ r − 2α ) (8) Substituting Equations 8 and 4 in Equations 7, the new voltage equation of BLDC motor is: r v ds r + pLs r =s v qs ωr Ls where K ed (θ r ) = r K ed (θ r ) − ωr Ls i ds r + ωr K (θ ) rs + pLs i qs eq r (9) 2 2 ∑ cos(θ r − iα )K e (θ r − iα ) 3 i =0 (10) 2 FAIRCHILD SEMICONDUCTOR POWER SEMINAR 2008 - 2009 K eq (θ r ) = − 2 2 ∑ sin(θ r − iα )K e (θ r − iα ) 3 i =0 (11) K e (θ r ) = ∑ − ai i =1,3,5,7,9... sin(iθ r ) (17) Considering Equations 5 and 6, the characters of K ed (θ r ) and K eq (θ r ) can be obtained as: Table I shows the harmonic relationship between the K e (θ r ) in three-phase and K ed (θ r ) & K eq (θ r ) in K ed (θ r + α / 2) = K ed (θ r ) K eq (θ r + α / 2) = K eq (θ r ) K ed ( −θ r ) = K ed (θ r ) K ( −θ ) = −K (θ ) r eq r eq SFR. Fig. 1(b) shows the A0 , A6 , and B6 can be calculated as: (12) K ed (θ r ) and K eq (θ r ) of the ideal A0 = a1 A6 = a7 − a5 B = −a − a 7 5 6 back-EMF calculation of each point. The result is coincident to Equation 12. (18) TABLE I HARMONICS RELATIONSHIP K ed (θ r ) K eq (θ r ) − a1 sin(θ r ) − a3 sin(3θ r ) 0 0 a1 0 − a5 sin(5θ r ) − a5 sin(6θ r ) − a5 cos(6θ r ) − a7 sin(7θ r ) − a9 sin(9θ r ) − a7 sin(6θ r ) 0 a7 cos(6θ r ) 0 − a11 sin(11θ r ) − a13 sin(13θ r ) − a11 sin(12θ r ) − a13 sin(12θ r ) − a11 cos(12θ r ) a13 cos(12θ r ) K e (θ r ) B. Modeling of the Current and the Torque As for the sixth harmonic in Equation 9, the pLs are dominant and the ωr Ls , −ωr Ls , and rs can be ignored to simplify the analysis process. The sixth harmonic of the current in SRF can be expressed as: Fig. 2. Ked and Keq harmonics amplitude. Based on Equation 12 the Ked(Θr) and Keq(Θr) can be written as: K ed (θ r ) = ∑ Bi sin( iθ r i = 6,12,18... K eq (θ r ) = ∑ Ai cos( iθ r i = 0,6,12,18... (13) ) ) (14) Figure 2 shows the harmonics of the Ked(Θr) and Keq(Θr) of the ideal trapezoidal back-EMF waveform. In Figure 2, the sixth harmonic is dominant. If the twelfth and higher harmonics are ignored, the back-EMF function can be written as: K ed (θ r ) = B6 sin 6θ r (15) K eq (θ r ) = A0 + A6 cos 6θ r (16) Based on Equations 5 and 6, the back-EMF coefficient can be expressed in Fourier form as: r i ds (6) = −B6 cos 6θ r 6ωr Ls (19) r i qs (6) = A6 sin 6θ r 6ωr Ls (20) The DC value of the current in the SRF can be expressed as: r r i ds ωr Ls v ds rs 1 r (0) = r 2 2 2 rs v qs − ωr Ke i qs(0) rs + ωr Ls − ωr Ls (21) The torque expression is: Te = ( 3P r r i qs K eq + i ds K ed 4 ) (22) Ignoring the higher harmonics, the detailed torque expression can be expressed as: Te = AA 3P i qs0 A0 + 6 0 + i ds0B6 sin6θr + A0i qs0 cos6θr 4 6 L ω r s (23) = Te0 + Te6 sin(6θr + γ ) 3 FAIRCHILD SEMICONDUCTOR POWER SEMINAR 2008 - 2009 where δ is the angle between the voltage vector and back-EMF vector. Figure 4 shows the stable voltage relationship. In MTPA control, the terminal voltage is: where Te0 = 3P i qs 0 A0 , 4 Te 6 = 2 A6 A0 2 2 6ω L + i ds 0B6 + A0 i qs 0 r s 3P 4 , and 6 A0 i qs 0ωr Ls . A A + 6 i B ω L ds 0 6 r s 6 0 γ = arctan The sixth harmonic component ratio is: η 6 ≡ Te 6 (24) Te0 For a typical BLDC motor with the load torque linear to running frequency, the sixth harmonic torque ratio is shown in Figure 3. When the operation frequency is higher than 6Hz, the sixth harmonic torque ratio is stable at A0/A6=5%. When it’s lower than 1Hz, the ratio is about 50%. Vdc ua = − 2 MI sin(θ r + δ ) Vdc MI sin(θ r − α + δ ) u b = − 2 u = − Vdc MI sin(θ + α + δ ) r c 2 (27) where Vdc is the DC link voltage and MI is the modulation index ( MI , defined as the peak phase voltage divided by half of the DC link voltage). qr ωr r r v qs = r si q s r Ke r ω + K eq ωr δ vs dr r r v ds Fig. 3. The sixth harmonic torque ratio in a typical motor. = −ω L i qs s r Fig. 4. The voltage vector for MTPA C. Optimal Angle of the Input Voltage When the motor works at the mode of maximum torque per ampere (MTPA), the copper losses are the smallest. It can be achieved by adjusting the input voltage of the BLDC. Normally, BLDC motor permanent magnet is surface mounted and there is no reluctant torque in the developed torque. Thus, the MTPA can be achieved by irds=0. Considering the DC value of the BLDC model, the stable voltage equation can be written as: r r v ds = −ωr Ls i qs r r v qs = rs i qs + ωr K eq r ω r Ls i qs r rs i qs + ω r K eq A. Simulation Model The BLDC motor model has been presented in reference [8].The dynamic system equations are: Ls Ls L s di a = uan − rs i a − ea dt di b = u bn − rs i b − eb dt di c = u cn − rs i c − ec dt (28) (25) where uan, ubn, ucn, ia, ib, ic, eb, eb, and ec are the phase voltages, currents, and back-EMF of phase a, b, and c, respectively. The developed torque and mechanical equation are: (26) P e i + eb i b + ec i c Te = a a ωr 2 The optimal angle of the voltage vector is: δ = arctan III. SYSTEM SIMULATION (29) 4 FAIRCHILD SEMICONDUCTOR POWER SEMINAR 2008 - 2009 2 dω r Jm = Te − TL P dt (30) where TL , P , and J m are load torque, pole number, and rotor inertia. Considering ia+ib+ic=0, the following can be can be derived from Equation 28: uan + u bn + ucn = ea + eb + ec (31) The neutral point voltage is: un = 1 (ua + ub + uc − ea − eb − ec ) 3 (32) Combining Equations 28 and 30, the system operation status equations, can be expressed as: di a Ls dt = ua − rs i a Ls di b = ub − rs i b dt di Ls c = uc − rs i c dt 2 dω r J m P dt = Te − ea − u n − eb − un − ec − u n (33) − TL B. SPWM Method In this system, the carrier waveform is changed for lower loss. Normally, the SPWM is generalized as in Figure 5. The sinusoid modulating signals are expressed in Equation 27. The three phases modulating signals are shown in Figure 5(a). By comparing the modulating signal and the carrier signal in Figure 5(b), the PWM is shown in Figure 5(c). Fig. 6. Modified modulation of a three-phase sinusoidal waveform. To save the loss, a modified SPWM is used, as shown Figure 6. The maximum MI value can be increased from 1 to 1.154 (= 2 / 3 ). The modified voltages are expressed as: u a' = u a − min(u a , u b , u c ) ' (34) u b = u b − min(u a , u b , u c ) ' u = u − min(u , u , u ) c a b c c C. System Configuration Figure 7 shows the schematic of the simulation. The hall sensor generates the absolute position ( θ n ) at time ( t n ) to the speed calculator. In the speed calculator, the low pass filter (LPF) generates estimated rotor speed by the input of the transient speed. The estimated rotor position is adjusted by adding the incremental angle ( ∆θ r ) in ( t − t n ) to θ n . Finally, the three phases of PWM are generated by the inverter. The inverter mechanism is based on Equations 29 and 33, and the motor model uses the Equation 32. There is no current loop and voltage loop in the inverter block. Fig. 5. General three-phase PWM. 5 FAIRCHILD SEMICONDUCTOR POWER SEMINAR 2008 - 2009 Vdc θ n − θ n −1 Fig. 7. Simulation schematic. D. Simulation Results and Comparison Figure 8 shows simulation results from the start. When the speed is low, the current is not stable due to the low resolution position sensor. The torque and current become stable after the MI increases to the maximal value. Figure 9 shows the stable segment of the simulation. There are 6n harmonics in the developed torque due to the non-sinusoidal back-EMF. The currents are synchronized with the back-EMF by fine tuning the MTPA angle. Figure 10 shows the simulation result for the traditional six-step control result. The simulation uses the same motor parameters and load characters. There are six developed torque pulses in a electrical cycle. To compare the torque ripple, the torque pulse has been minimized by optimizing the commutation angle. In Figure 9 and 10, the operation frequencies are almost the same. That means these two kinds of methods work at the same load point. In Figure 10, the magnitude of the torque ripple is 0.15Nm. In Figure 9, this value is 0.08Nm, which is half the traditional control result in Figure 10. Fig. 8. Modified SPWM control simulation results from 0 to 0.6s. 6 FAIRCHILD SEMICONDUCTOR POWER SEMINAR 2008 - 2009 Fig. 9. Modified SPWM control simulation results from 5(s) to 5.2(s). Fig. 10. Six-step control simulation results from 5(s) to 5.2(s). 7 FAIRCHILD SEMICONDUCTOR POWER SEMINAR 2008 - 2009 IV.EXPERIMENT RESULTS A. Experimental Setup The system is configured in Figure 11. The system includes the BLDC motor, hall sensor, controller, and Fairchild Smart Power Module (SPM®), and can be assembled in the motor. In this experiment, the selected SPM is the FSB50450, a 500V three-phase FRFET® inverter that includes a high-voltage integrated circuit (HVIC). The controller is the TB6551, a sinusoidal PWM controller for motor control. The three hall sensors are EW632’s. The motor parameters are shown in Table II. Figure 12 shows the back-EMF of this BLDC fan motor. It’s a typical 100w BLDC and the fifth and seventh harmonic back-EMF are about 5% and 0.5%, respectively. Fig .11. System configuration. TABLE II PARAMETER OF THE EXPERIMENTAL MOTOR Pole Number P 8 Stator Resistance rs 33 Ω Equivalent Self-Inductance L s 0.1 H Maxim Back-EMF Coefficient Terminal Voltage Angle Inertia Jm K e max 0.18 δ V·s/rad 0.08 rad 0.00012 Kg·m2 Fig. 12. One-phase back-EMF of the fan motor. Figure 13 shows the air-conditioner outdoor unit in which the tested fan is assembled. Figure 14 shows the tested fan motor in detail, while Figure 14(a) shows the fan assembled in the air- conditioner. The case of the outdoor unit has been dissembled. Figure 14(b) shows the fan separately, while Figure 14(c) and (d) show the built-in PCB board and the fan motor, respectively. The back case of the motor is the heat sink of the SPM after assembly. The SPM has been marked. SPM simplifies the built-in PCB design because of high-density integration and good thermal design. The motor winding has concentrated full-pitch distribution. The hall sensor is positioned on the bottom side of the PCB to detect the direction of the magnetic field. B. Experimental Results Figure 15 shows the starting waveform. Channel 1 and channel 3 are the PWM and the current of phase a, respectively. Figure 16 shows the stable working waveform. Channels 1-4 are the PWM in phase a, PWM in phase b, current in phase a, and PWM in phase b, respectively. 8 FAIRCHILD SEMICONDUCTOR POWER SEMINAR 2008 - 2009 There is no current spike and this shows the motor operation is stable and that the position calculation works well. and dynamic performance are not required. Fig. 13. The tested fan assembled in the air-conditioner. Fig .15. Operation waveform from start. (a) (b) (c) (d) Fig. 14. The fan motor for test. Fig .16 Operation waveform when the motor is stable. V. CONCLUSION There is a 6n harmonic torque instead of pulsating torque in the motor developed torque when the BLDC motor is driven by sinusoidal PWM. In the simulation, compared to the traditional six-step control, this method can reduce ripple about 50%. The test result shows this method can be used when the speed precision REFERENCES [1] [2] [3] Fernando Rodriguez and Ali Emadi, “A Novel Digital Control Technique for Brushless DC Motor Drives,” IEEE Trans. Ind. Electron., vol. 54, no. 3, pp. 2365-2373, Oct. 2007. P. Pillay and R. Krishnan, “Application Characteristics of Permanent Magnet Synchronous and Brushless DC Motors for Servo Drives,” IEEE Trans. Ind. Appl., vol. 27, no. 5, pp. 986–996, Sep./Oct. 1991. Bimal K.Bose, Modern Power Electronics and AC Drivers. Prentice Hall, 2002. 9 FAIRCHILD SEMICONDUCTOR POWER SEMINAR 2008 - 2009 [4] [5] [6] [7] [8] Y. Liu, Z. Q. Zhu, and D. Howe, “Direct Torque Control of Brushless DC Drives with Reduced Torque Ripple,” IEEE Trans. Ind. Appl., vol. 41, no. 2,pp. 599–608, Mar./Apr. 2005. D. K. Kim, K. W. Lee, and B.-I. Kwon, “Commutation torque ripple reduction in a position sensorless brushless DC motor drive,” IEEE Trans.Power Electron., vol. 21, no. 6, pp. 1762–1768, Nov. 2006. H.M.Ryu,” Synchronous Reference Frame d-q Modeling of Interior Permanent Magnet Synchronous Motor (IPMSM),” unpublished. C. W. Lu, “Torque Controller for Brushless DC Motors,” IEEE Trans. Ind. Electron., vol. 46, no. 2, pp. 471–473, Apr. 1999. P. Pillay and R. Krishnan, “Modeling of Permanent Magnet Motor Drives,” IEEE Trans. Ind. Electron., vol. 35, no. 4, pp. 537–541, Nov. 1988. Shucheng Wang was born in Jinzhou, China in 1979. He received a B.S. degree from Haerbin University of Science & Technology, Haerbin, China in 2001, and an M.S. degree in electronic engineering from Huazhong University of Science & Technology, Wuhan, China in 2004. Mr. Wang works for Fairchild Semiconductor in Shenzhen, China, where he is a technical and marketing engineer for the application support center team. His research interest is motor control. 10