Identifying, Analyzing and Mitigating First-, Second- and Third-Order Effects on Motor Control Performance in Vector Control of PMSM Motor Applications Renesas Electronics America Inc. Jeff Shoemaker, Sr. FAE Date: March 13, 2012 © 2011 Renesas Electronics America Inc. All rights reserved. 00000-A Agenda PMSM vector control First-Order Effects z Motor parameters z Back EMF, z Reluctance and Cogging torque Second-Order Effects z z z z PWM control frequency selection Dead time impact Feedback resolution and sensing Modulation techniques Third-Order Effects z Fixed point vs. floating point calculations z MCU/DSP architecture Q&A 2 © 2011 Renesas Electronics America Inc. All rights reserved. PMSM Vector Control Designers use vector control in PMSM applications to achieve efficiency and performance z Requires them to trade off many factors and effects z A balance is required between system level functionality and performance and efficiency Renesas has many years of experience working with motor control, in particular with vector control z Many control algorithms have been developed and tested z Showed how this can influence the MCU/DSC selection criteria Our experience has shown us that there are many factors to consider when developing and designing for optimum performance and high efficiency z We have attempted to stratify these considerations into three levels of effects from overall system level down to the MCU/DSC level 3 © 2011 Renesas Electronics America Inc. All rights reserved. Permanent Magnet Synchronous Motor Also known as Brushless DC (BLDC) motor Stator Windings Stator Rotor 4 © 2011 Renesas Electronics America Inc. All rights reserved. Parameter Value Motor Pole 4 Phase 3 Voltage 130 V Current 1.5 A Power 1/5 hp Speed 2500 rpm Inductance 27 mh Stator Resistor 5.1 ohm Hall sensors 3 Vector Control Current loop is inside, speed loop is outside Sensorless control has flux and position observer ω r* Commanded speed * iq Δωr ωr id = 0 * Observed actual speed Inverse Park transform * Uq iq Ud * α, β T −1 (θ ) Inverse Clarke transform Uα * Uβ * α, β id θ Park transform iq Clarke transform ia α, β iα id T (θ ) iβ α, β Current loop ωr Speed loop 5 © 2011 Renesas Electronics America Inc. All rights reserved. θ ib First-Order Effects 6 © 2011 Renesas Electronics America Inc. All rights reserved. First-Order Effects: Parameter Mismatch Manifests itself as a gross error in performance z E.g., Torque or speed off by 50%, instability, etc. A good place to start is the manufacturer’s datasheet or the ratings listed on the nameplate of the motor Table I. Motor specifications for Bodine 34B series BLDC motor. Is there a mismatch between what is in the datasheet versus the parameters that are used in the control algorithm? Are the order of magnitude and units correct? What is the meaning of specifications and ratings? 7 © 2011 Renesas Electronics America Inc. All rights reserved. 34B Series BLDC Motor Model 3306 Specifications Model Number 3306 [CAD Drawings] Category 34B Brushless DC Motor Speed (rpm) 2500 Rated Torque (oz-in) 81 Rated Voltage 130V Motor HP 1/5 Torque Constant (oz-in/A) 51 Voltage Constant (V/krpm) 38 Winding Resistance (ohms) 9.2 Winding Inductance (mH) 24 Rotor Inertia (oz-in-sec²) 0.0115 Radial Load (lbs) 42 Length XH (inch) 4.06 Weight (lbs) 6 Product Type 34B3BEBL Accessory Shaft NO Connection Diagram BLDC Connection First-Order Effects: Back EMF Back EMF is present in all PM Motors When the motor rotates, the magnetic field changes, and this induces a voltage that opposes the voltage applied to the motor z Va = ea +Ra*Ia + La*dia/dt z where Ra and La are the winding resistance and winding inductance respectively The term, ea, is the Back EMF denoted by z ea = k E * ωm z where kE is in units of V/(rad/sec) z Typically specified as V/kRPM by manufacturers 8 © 2011 Renesas Electronics America Inc. All rights reserved. First-Order Effects: Back EMF Waveform Found in manufacturer’s datasheet First-order effect is that it limits the top speed of the motor per a given amount of applied voltage because the higher speeds of the motor a larger Back EMF At some point, the amount of Back EMF equals the bus voltage, and there is not sufficient voltage that can be applied to the motor to maintain a certain speed Can be mitigated to some extent by employing techniques such as space vector modulation 9 © 2011 Renesas Electronics America Inc. All rights reserved. U Ω V W First-Order Effects: Reluctance & Cogging Torques The reluctance torque is the result of differences between the q and d axis inductances in a permanent magnet synchronous motor It manifests itself in the torque performance of the motor Mitigated by employing phase advancing techniques The cogging torque is created when the magnets in the rotor interact with the slots in the stator Undesirable first-order effect that is apparent when the motor is operating at low speed Mitigated by tuning or employing adaptive feed-forward or current injection techniques 10 © 2011 Renesas Electronics America Inc. All rights reserved. First-Order Effects: Others To Consider The motor torque constant, Kt, relates the torque produced by the motor with the current through the motor z Effective torque constant can be up to 20% lower at peak current than at lower current z Mitigate this effect is by properly characterizing Kt as a function of current and temperature. A lookup table can be used identify Kt and that can be used for the internal model of the motor and control algorithm 11 © 2011 Renesas Electronics America Inc. All rights reserved. Second-Order Effects 12 © 2011 Renesas Electronics America Inc. All rights reserved. PWM Control Frequency Selection Second-order effects are more pronounced in high performance algorithms, such as vector control, and dealt with assuming first-order effects have been satisfied Frequency selection is the result of a tradeoff between performance, efficiency and available CPU bandwidth The test data shown here is from a sensorless vector control algorithm implemented on an RX62N CPU board with a power stage driving a BLDC motor Results show much better control response while maintaining smooth current at the higher PWM freq at the cost of double the CPU bandwidth 13 © 2011 Renesas Electronics America Inc. All rights reserved. Speed Control @10kHz and 16kHz Motor Speed Using 10kHz Carrier Motor Speed Using 16kHz Carrier 4010 4010 4008 Accuracy is +/- 5 RPM 4008 Accuracy is +/- 3 RPM 4006 4006 4004 Speed in RPM Speed in RPM 4004 4002 4000 3998 3996 4000 3998 3996 3994 3994 3992 3992 3990 3990 0 10 20 30 40 50 60 70 Time (units of 10 msec) 26.5 % Bandwidth @10kHz 14 4002 © 2011 Renesas Electronics America Inc. All rights reserved. 80 0 20 40 60 80 Time (Units fo 10 msec) 44.8% Bandwidth @16kHz 100 120 140 Current Control @10kHz and 16kHz Phase current @10K 26.5 % Bandwidth @10kHz 15 © 2011 Renesas Electronics America Inc. All rights reserved. Phase current @16K 44.8% Bandwidth @16kHz Dead-Time Insertion and Mitigation The time inserted between turning OFF the high side and turning ON the low side, or vice versa z No power provided to motor during dead time z Keeps power module from blowing up! As a second-order effect, dead time creates a harmonic distortion in the current wave form and injects the disturbance in the current loop affecting the control performance Mitigated by simple to complex dead-time compensation techniques to recover performance Cost is CPU utilization but tradeoff may be worth it A+ Current ON A+ A+ Current ON A+ Current A- ON A- Current pass through blows up power module 16 © 2011 Renesas Electronics America Inc. All rights reserved. A- ON Δt Dead time prevents the power module blow up A- Dead Time Compensation @10Hz Without compensation, current profile has distinct harmonics With compensation, current profile is smooth Without dead time compensation 17 © 2011 Renesas Electronics America Inc. All rights reserved. With dead time compensation Dead Time Compensation @30Hz Without compensation, current profile has distinct harmonics With compensation, current profile is smooth Without dead time compensation 18 © 2011 Renesas Electronics America Inc. All rights reserved. With dead time compensation Feedback Mechanism Resolution Digital encoders are commonly used in high-end control where the speed loop is tightly coupled with the current loop Designers must consider the dynamic range, accuracy and speed loop rate that is required when selecting an encoder z Examples given in paper Some designers have a dynamic speed loop where the speed loop rate can change at different speeds z Must consider how loop dynamics and gains change in this case 19 © 2011 Renesas Electronics America Inc. All rights reserved. Feedback Mechanism Resolution (2) The ADC is used for current measurement and is available as a peripheral on an MCU/DSC with 10- or 12-bits resolution 12-bit resolution is better at resolving the zero crossing region and results in better performance z e.g, 10 amp max current per phase at 10-bits resolves 20 mA has much larger step size than 12-bits which resolves down to 5 mA Can also employ over sampling and filtering to increase resolution of current measurements Be wary of encoder quantization and resolution noise that can back feed into the speed loop z Manifests itself as a torque ripple Can be mitigated by employing fixed speed loop if small counts over speed range are possible or use dynamic loop rate at speed ranges where performance is critical 20 © 2011 Renesas Electronics America Inc. All rights reserved. Second-Order Effects: Modulation Techniques High-end control algorithms typically use sinusoidal methods and their enhancements to continually provide power to all three phases of the motor 10000 8000 6000 4000 Sine value 2000 0 1 40 79 118 157 196 235 274 313 352 -2000 Integer value for u -4000 Integer value for v Integer value for w -6000 -8000 angle -10000 Three sine waves one for each phase Input voltage to motor is up to max bus voltage per phase Phase to phase voltage is less due to third harmonic being cancelled as a second-order effect due to the phase voltages shifted 120 degrees apart PWM @20kHz, F = 50Hz 2 Vuv < Vbus U V U-V current W 1 U Magnitude Bus Voltage V Maximum Line-to-line Voltage is 86% of Vbus 0 -1 W -2 0 10 20 30 40 50 time (ms) Current should be here 21 © 2011 Renesas Electronics America Inc. All rights reserved. Current not at peak value Second-Order Effects: Modulation Techniques This second-order effect can be mitigated by injecting the third harmonic back into the phase voltages before it is applied to the motor z Known as third harmonic injection or space vector modulation z Additional scaling is required Additional scaling Pure sine wave and 3 rd harmonic Sine wave plus 3 rd harmonic The modified sine modulation has new PWM values which recover the lost phase to phase voltage The current value per phase can now achieve the desired maximum value sin θ +1/6sin3 θ PWM @20kHz, F = 50Hz 2 U V W U-V current Magnitude 1 0 -1 -2 0 10 20 30 time (ms) Current at peak value 22 © 2011 Renesas Electronics America Inc. All rights reserved. 40 50 Third-Order Effects 23 © 2011 Renesas Electronics America Inc. All rights reserved. Third-Order Effects: Fixed-Point vs Floating-Point Vector control algorithm requires implementing complex transformations iu a-axis ω iα F iw Id ω b-axis F ω F Iq d-axis iβ iv θ a-axis b-axis Clarke Transformation 1 ⎡ ⎡iα ⎤ ⎢1 − 2 ⎢i ⎥ = ⎢ 3 ⎣ β ⎦ ⎢0 ⎢⎣ 2 q-axis 1 ⎤ ⎡i ⎤ a 2 ⎥ ⎢i ⎥ ⎥ 3 ⎥⎢ b ⎥ − ⎢i ⎥ 2 ⎥⎦ ⎣ c ⎦ − Park Transformation ⎡ I d ⎤ ⎡ cos θ ⎢I ⎥ = ⎢ ⎣ q ⎦ ⎣− sin θ sin θ ⎤ ⎡iα ⎤ ⎢ ⎥ cosθ ⎥⎦ ⎣iβ ⎦ Vector control is typically implemented in a fixed-point DSP or MCU Fixed-point implementation has several issues such as saturation, continuous normalization and proper accuracy of the variable being used, especially for the 16-bit DSP devices 24 © 2011 Renesas Electronics America Inc. All rights reserved. Third-Order Effects: Fixed-Point vs Floating-Point Renesas implemented both fixed-point and floating-point sensorless vector control algorithms for BLDC motor control Results of comparison were presented at the Motor & Drive 2011 conference Implementation with floating-point uses significantly less code and CPU bandwidth ω r* Commanded speed * iq Δωr Speed Regulator ωr id = 0 * Observed actual speed iq id Inverse Inverse Park transform Clarke transform * Uq * iq PI Uα d,q α, β Regulator to id PI Regulator Ud * α, β T −1 (θ ) θ Uβ * to a, b, c Voltage Source 3-phase Inverter SIN PWM Clarke transform ia α, β to d,q id PWM1~6 Motor Model Based Flux and Position Observer Park transform iq DC Bus iα T (θ ) a,b,c to iβ α, β ib Current loop ωr θ 3-phase PMSM Speed Estimation Speed loop Parameter Fixed-Point Floating point Ratio Comments CPU bandwidth (μsec) 40 26 0.65 FPU 35% better Code size (bytes) 13816 7597 0.55 Code reduced by 45% 25 © 2011 Renesas Electronics America Inc. All rights reserved. Third-Order Effects: Processor Core Architecture Processor utilization and computational performance are third-order effects that can impact control loop response/performance and ability to perform other tasks z 26 Directly influenced by processor architecture and MCU/DSP/DSC selection Motor control testing by Renesas is done using a RX62x DSC High performance core with enhanced Harvard architecture, 5-stage pipeline, 64-bit wide pre-fetch queue minimizes bottlenecks between CPU and memories © 2011 Renesas Electronics America Inc. All rights reserved. Third-Order Effects: Processor Core Architecture 27 Wait state penalty due to slow flash is another third-order effect that impedes processor performance This can be mitigated by selecting a processor that has zero wait states up to the max CPU clock To alleviate CPU burden, take advantage of DMA to automate data transfers such as current measurement RX also has DTC (Data Transfer Controller) peripheral which allows for many more virtual DMA channels © 2011 Renesas Electronics America Inc. All rights reserved. Third-Order Effects: Synchronized ADC/PWM Triggering 28 The designer typically wants to have an automatic triggering of A/D conversion start relative to some point within the PWM timing cycle Having an advanced timer set with double buffering allows for asymmetric PWM and flexibility to set A/D conversion start trigger anywhere on the PWM timing cycle This is especially advantageous when implementing single-shunt sensorless vector control Interrupt skipping capability can allow designer to set up current loop interrupt at a fixed multiple of the PWM frequency to help reduce CPU utilization © 2011 Renesas Electronics America Inc. All rights reserved. Summary Renesas motor control team has shared our experiences within the engineering community A summary of first-, second- and third-order effects has been presented z First-order effects are system-level effects all designers must satisfy in order to have properly functioning motor control system z Second-order effects are pronounced in high-performance motor control implementations such as vector control z Identifying and mitigating third-order effects results in a more optimized and robust design The RX62T and RX62N DSC are very well suited for dealing with second- and third-order effects 29 © 2011 Renesas Electronics America Inc. All rights reserved. Jeff Shoemaker Jeffrey.Shoemaker@Renesas.com Thank you 30 © 2011 Renesas Electronics America Inc. All rights reserved. Renesas Electronics America Inc. © 2011 Renesas Electronics America Inc. All rights reserved.