IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 3, AUGUST Zhixun Ma, Jianbo Gao, and Ralph Kennel, Senior Member, IEEE

IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 3, AUGUST 2013 1253 FPGA Implementation of a Hybrid Sensorless Control of SMPMSM in the Whole Speed Range Zhixun Ma, Jianbo Gao, and Ralph Kennel, Senior Member, IEEE Abstract This paper presents an FPGA-based (Field Programmable Gate Array) sensorless controller for Surface Mounted Permanent Magnet Synchronous Machines (SMPMSM). A hybrid sensorless controller combining the signal injection technique and a linearly compensated flux observer is proposed. Using a Delta-Sigma A/D converter and FPGA oversampling technique, this work realizes a high performance high frequency (HF) injection sensorless control method which needs lower HF current response and introduces lower acoustic noises. The linearly compensated flux observer, based on back electromotive force (EMF) is used for sensorless control in the high speed range. The flux observer exhibits high dynamic and steady-state performance and is robust to parameter variation. Using model-based design, with the tools of MATLAB/Simulink and Simulink HDL (hardware description language) Coder, the whole control system is designed and implemented in a single chip. Experimental results demonstrate that the developed sensorless controller has high performance in the whole speed range. Index Terms Field programmable gate array (FPGA), flux observer, high frequency (HF) injection, model-based design (MBD), sensorless control, synchronous machine (SM). I. INTRODUCTION SURFACE mounted permanent magnet synchronous machines (SMPMSMs) have received increasing utilization in recent years because of their high energy efficiency and high torque density. In the high performance SMPMSM drive systems, position sensors such as resolvers or encoders are required to determine the rotor position. However, position sensors present several drawbacks, such as increased cost and reduced system reliability. Therefore, eliminating the position sensors is desirable. Sensorless control techniques presented in the literature can be mainly divided into two categories. The first, the fundamental model-based back electromotive force (EMF) tracking estimator [1] [6] has advantages of simplicity and straightforwardness. However, the performance deteriorates drastically at low speeds due to the low magnitude of EMF. Furthermore, most of the back EMF based observers require accurate and Manuscript received December 01, 2011; revised March 16, 2012 and July 20, 2012; accepted September 17, 2012. Date of publication October 02, 2012; date of current version August 16, 2013. Paper no. TII-11-961. The authors are with the Institute for Electrical Drive Systems and Power Electronics, Technical University of Munich, Munich 80333, Germany (e-mail: zhixun.ma@tum.de; jianbo.gao@tum.de; ralph.kennel@tum.de). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TII.2012.2221132 precise measurements of the machine parameters. This paper proposes a linearly compensated flux observer which is very robust to parameter variation. The second category exploits the anisotropic properties of PMSMs caused by the saliency of an interior magnet rotor and/or by the saturation of the stator iron. Most of these methods inject voltage signals and extract the position dependent inductance information from the measured stator currents. In the lower speed range and at a standstill, signal injection techniques are required and widespread [7], [8]. However, signal injection causes extra losses, torque ripple and transient disturbances. To reduce the interferences of the signal injection, it is advantageous to decrease the amplitude of the injected voltage, which, however, introduces a lower current response. High resolution current measurement is thus necessary to extract the position information with good signal-to-noise ratio. Using a Delta-Sigma modulator, and with proper decimation filter design, this paper gives a promising solution for current measurement. The former work about high resolution position estimation with Delta-Sigma modulated current measurement was presented in [9]. The combination of signal injection and back EMF tracking estimators is a good solution for the PMSM sensorless control in the whole speed range. Several hybrid sensorless control methods have been proposed in the literature for PMSMs [10] [13]. Nowadays, Field ProgrammableGateArrays(FPGAs)are good candidates to achieve high control performances in industrial control applications [14]. The motivation for model-based design (MBD) for FPGAs arises from the necessity of the design of complex control systems that require dedicated multipliers and computing units. Simulink HDL (Hardware Description Language) Coder is the software to generate HDL codes (VHDL or Verilog HDL)fromfixed point Simulink models. In addition, a few software products also convert Simulink models into HDL descriptions: System Generator for digital signal processing (DSP) from Xilinx, DSP Builder from Altera and Synplify DSP from Synopsys [15]. Each of them provides users with its own model blocks. Helped by these tools, users have to build the Simulink model with provided blocks to generate HDL codes. Compared with the three abovementioned software tools, Simulink HDL Coder, by which the generated HDL codes can be transplanted to any FPGA, is much more flexible. The interest in FPGA-based sensorless controllers for AC drives has been constantly rising in recent years due to their attractive cost and reliability [16] [19]. Paper [19] presents an FPGA-based sensorless controller, which combines the rotating high frequency (HF) injection method and the extended Kalman filter (EKF) algorithm. The sensorless controller was implemented in two FPGA chips. The HF injection method was tested 1551-3203 2012 IEEE

1254 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 3, AUGUST 2013 using an FPGA Actel A3P 1000, and the EKF algorithm was implemented with FPGA Xilinx Virtex-2P. This paper proposes an FPGA-based hybrid sensorless controller for SMPMSM combining the alternating voltage HF injection method and a linearly compensated flux observer. The sensorless controller, which is implemented in a single chip, is effective in the whole speed range including the standstill mode. The contributions of this paper are as follows: Using a Delta-Sigma A/D converter and the FPGA oversampling technique, this paper realizes a high performance HF injection sensorless control method which needs lower HF current response and introduces lower acoustic noises. A linearly compensated sensorless flux observer based on back EMF is proposed. The flux observer exhibits excellent dynamic and steady-state performance and is robust to parameter variation. It is effective with 0 3 times stator inductance variation at above 0.03 pu (per unit) speed. With increased speed the sensitivity of the flux observer to stator resistance variation decreases. At 0.03 pu speed, the flux observer works well with 0 2 times resistance variation with and without load. Finally, the implementation of the FPGA-based hybrid sensorless controller is presented. Experimentally, it is shown that the FPGA-based sensorless controller has high performance in the whole speed range. The injected carrier voltage in the stationary coordinate is mathematically represented in (4). A transformation of (4) to the multiplying (4) by. Together with the The solution is voltage model needs to be trans- In order to detect the carrier current, formed into the stationary coordinate (4) coordinate is achieved by (5) (6) (7) II. HF INJECTION METHOD A. HFModelofPMSM The fundamental written as voltage equations of PMSM can be where is the current positive sequence component and is the current negative sequence component. (8) Where,,,,, are the axis voltages, currents and inductances, is the stator resistance, is the magnetic flux, is the rotor speed. Since the frequency of the injected HF voltage is much higher than the fundamental frequency, the fundamental frequency is neglected in (1) in the HF model of PMSM. In the low speed range, the resistance voltage drop is very small and the time variation of inductances can also be neglected. Therefore, the HF voltage model can be derived from (1) as The index denotes the quantities related to the carrier frequency. B. Alternating Voltage HF Injection The alternating carrier voltage is injected into the estimated -axis, which is at a spatial displacement angle with respect to the real -axis [8]. The relationship can be written as (1) (2) (3) C. Closed-Loop Demodulation Position error signal can be obtained by demodulation of. It is transformed to a reference frame in a negative rotating direction at the approximated carrier frequency. This is achieved by through a low pass filter. The result is (10) Equation (10) contains the useful position error signal, where Equation (10) can be further written as (9) (11) (12) When is fairly small, (12) shows that the real part of the current response is proportional to the error angle.thisis used to track the rotor angle by a closed-loop tracking system. The closed-loop demodulation process is shown in Fig. 8.

MA et al.: FPGA IMPLEMENTATION OF A HYBRID 1255 and outputs directly the flux vector speed and the angle for sensorless control. Fig. 1. Flux and back EMF vectors in and frame. B. Stability Analysis of Linearly Compensated Flux Observer The small signal analysis model method is used to analyze the stability of the observer. The input is the phase disturbance error of the back EMF, and the output is the introduced phase variation of the flux. It is supposed that 1) The input phase step variation of the back EMF is small, such that and ; 2) the observer works in steady state, and and are positive; (3) the variables which are higher than the second order are neglected. As shown in Fig. 2, the relationship between the estimated and the real is (17) Based on the precondition, (17) is simplified to (18) Fig. 2. Diagram of the linearly compensated flux observer. III. FUNDAMENTAL MODEL-BASED METHOD Here, defining Since, (18) can be written as,thus (19) (20) A. Linearly Compensated Flux Observer The flux and back EMF vector diagram is shown in Fig. 1, where the frame is the stationary frame, the frame is the synchronous rotating frame, the flux vector is in the -axis, and is the angle between and the -axis. In the following equations, The relationship between, and is given as (13) (14) (15) (16) Therefore,,,,where is the amplitude of the flux, is the rotating speed of the flux vector, and is the back EMF vector [2], [3]. Through the above analysis, the diagram of the linearly compensated flux observer for the sensorless control is shown in Fig. 2. In Fig. 2,,,,, are the observed values of their corresponding theoretical actual values. is the compensator, where should have the same sign as. makes the flux always positive, so that the flux angle cannot converge to the opposite direction. This observer solves the drift problem When the system works in steady state,,,, the estimated speed is given as After neglecting the values of the second and higher orders That is (21) (22) (23) Due to,itis given. Through the above analysis, a small signal analysis diagram can be obtained as shown in Fig. 3. In order to find the influence of the compensator k to the observer, Fig. 3(a) can be simplified as (b). If there were no compensator, the closed-loop transfer function of the observer would be (24) It is easy to see that this characteristic equation would have a pair of imaginary roots. Thus the observer would

1256 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 3, AUGUST 2013 Fig. 3. Diagram of small signal analysis model (a) Small signal analysis of the linearly compensated flux observer; (b) Simplified small signal analysis model. be in a critical stable state. Therefore, the compensator is necessary. The proper compensator parameter will be given next. From Fig. 3, the open-loop and closed-loop transfer function are as follows: (25) (26) The characteristic equation of the closed-loop transfer function is. The relationship between the performance of the observer and the compensator is If, all the closed-loop characteristic roots are in the right side of the domain. The observer is unstable. If, the characteristic equation has a pair of imaginary roots. The observer is zero damping. If, all the closed-loop characteristic roots are in the left side of the domain. The observer is always stable. If, the characteristic equation has a pair of complex-conjugate roots.the observer is underdamping. If, the characteristic equation has a pair of equal minus real roots. The observer is critical damping. If, the characteristic equation has two minus real roots. The observer is overdamping. If, based on the second order system optimal parameter method, the roots of the characteristic equation are. The system can obtain the fastest response and least tuning time. Therefore, in the control system, the value of the compensator is. C. Parameter Sensitivity and Dynamic Test In the frame, back EMF is given by (27) where and are stator resistance and inductance respectively. The estimation of the back EMF is mainly based on and. For the robustness analysis, the parameters with identification errors are used in the observer. The estimated angle is Fig. 4. Sensorless performance based on the flux observer with parameter variation at 100 rpm (0.03 pu) with a 4 Nm load. (a).(b).(c).(d). compared with the real angle. Sensorless control performance basedontheflux observer is shown in Figs. 4 and 5 with different parameter variations. As a result, the proposed flux observer is very robust against parametric uncertainties. Through experiments, it has been shown that the sensitivity of parameter depends on the speed. The observer will be more sensitive to at a lower speed range. The flux observer can work well with at a speed of 100 rpm (0.03 pu) with a 4 Nm load (0.5 pu). At a speed of 900 rpm (0.3 pu), it can work, with the same load, at least with.fromtheexperimental results, for parameter, the parameter sensitivity is not aligned with speed. The flux observer works well with at the speed above 100 rpm. The high speed dynamic response of the flux observer is shown in Fig. 6. The machine is decelerated from 900 to 900 rpm. The estimated position follows the actual position very closely in the transient interval, with fast dynamics. The three curves in each block of Figs. 4 6, from the top down, are the U and V phase current, estimated error, and estimated rotor position, respectively. IV. COMBINATION OF SIGNAL INJECTION AND LINEARLY COMPENSATED FLUX OBSERVER At low speed, model-based observers cannot work well. Therefore, the signal injection method introduced in Section II is used. To perform a smooth transition between both strategies, a hybrid observer is used, in which the estimated flux angle is

MA et al.: FPGA IMPLEMENTATION OF A HYBRID 1257 Fig. 5. Sensorless performance based on the flux observer with parameter variation at 900 rpm (0.3pu) with a 4 Nm load. (a).(b). Fig. 7. Tools and Design flow for FPGA development. and verified in MATLAB/Simulink. At this time, the algorithm is validated. Then, the model is discretized to build the corresponding discrete model with the fixed point data type, which can be implemented in FPGA. Some special algorithms which are qualified for hardware implementation should be used in this phase. HDL code is then generated automatically from the discrete fixed point model by the Simulink HDL Coder. Second, the corresponding HDL code is synthesized, placed and routed with Altera Quartus II software. At the same time, the NIOS II soft core is designed by SOPC builder. Through Avalon bus, in NIOS II soft core, data can be read and written from the sensorless control Intellectual Property (IP) core. The target device is Altera Cyclone III FPGA EP3C40F484C7. B. FPGA-Based Controller Design for the Control Algorithm Fig. 6. Dynamic sensorless performance based on the flux observer from 900 rpm (0.3 pu) to 900 rpm without load. obtained with the linear combination of both observers. The switching strategy can be written as follows: (28) (29) where and are the up and low speed limitation of the transition process, respectively. The injected HF voltage is turned on and off using a hysteresis band at the speed slightly higher than the upper transition speed. The speed is obtained by switching both methods with a hysteresis band in the same manner. V. FPGA IMPLEMENTATION OF THE HYBRID SENSORLESS CONTROL A. Model-Based Design Flow for an FPGA-Based Controller Fig. 7 presents the design process and tools used in the control system. First, the continuous control model is constructed The proposed control algorithm has been implemented on an Altera cyclone III FPGA-based control board. The system clock frequency is 50 MHz. Fig. 8 demonstrates the system structure. The system contains: I/O interfaces, a NIOS II processor, an Avalon bus, a field oriented control model, and both signal injection and flux observer based sensorless control models. Because almost all FPGAs lack an analog-to-digital converter (ADC) appropriate for industrial applications, an additional ADC should be included in the system. The analog current input signal is converted into an oversampling digital signal by a four channel Delta-Sigma A/D converter ADS1204. The output signal of the system is a 6-route sinusoidal pulse-width modulation (SPWM) signal which is used for the inverter control. The NIOS II processor is a soft processor which is embedded in the FPGA. The field oriented control model and the sensorless control models are all connected with the Avalon bus. The NIOS II processor can exchange data with the models through the Avalon bus. The whole control algorithm is realized by hardware. The NIOS II processor is only used for debugging. The execution time of the sensorless controller is about 6.62. Its implementation result is shown in Table I. For the signal injection method of sensorless control, the resolution of current measurement is important, since the HF current response signal is quite low. Typical A/D converters in industrial inverters have only 12-bit resolution which introduces significant noises in the HF position information. This paper adopts a Delta-Sigma A/D converter.

1258 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 3, AUGUST 2013 Fig. 8. Block diagram of the sensorless field oriented controller. TABLE I FPGA UTILIZATION OF THE SENSORLESS CONTROLLER Fig. 9. Block diagram of Delta-Sigma A/D converter. The basic block diagram of a Delta-Sigma A/D converter is shown in Fig. 9. It consists of two parts. The firstpartis the second order Delta-Sigma modulator (ADS1204) that converts an analog input signal into a 1-bit digital data stream with a high data rate. The structure is shown in Fig. 10. With the appropriate digital filter and modulation rate, the A/D converter can be used to achieve 16-bit analog-to-digital conversion with no missing code. By analyzing the structure of the second order Delta-Sigma modulator, the output signal can be expressed as (30) where. So the output is one sample time delay of the input signal and the second order difference value of the quantizing noise. In the bandwidth of the signal, the signal gain is great and the noise is depressed to small values. Outside the bandwidth area, that is, in the high frequency area, the noise becomes stronger. Through noise shaping of the Delta- Fig. 10. Structure of the second order Delta-Sigma modulator. Sigma modulator, the noise is moved outside of the signal bandwidth. The second part is a low pass digital filter. The primary purpose of the digital filter is to remove the noises from the signal. The secondary purpose is to convert the 1-bit data stream at high sampling rates into a higher resolution data stream at a lower rate (decimation). The low pass digital filter is designed in FPGA. In this system, a filter [20] is used, which is represented in the z-domain as (31) The filter provides the best output performance at the lowest hardware size. It can be implemented as a simple architecture composed of adders and registers. Here, to fit the proportional

MA et al.: FPGA IMPLEMENTATION OF A HYBRID 1259 Fig. 11. filter implementation using MATLAB/Simulink. component of the PI (proportional plus integral) current regulator, the oversampling ratio M is set to 64. The output word width is 18-bit. The general filter implementation using MATLAB/SimulinkisshowninFig.11. With respect to the field oriented control, the most important constructions are the vector transformation parts Clarke and Park transformations. Clarke transformation can be implemented easily with an adder and a multiplier. For the rotating transformation, the CORDIC (Coordinate Rotation Digital Computer) algorithm [21], which is very qualified for hardware implementation, is adopted. Using only shift and add operations, this algorithm realizes vector rotating transformation with high speed and high resolution. The principle of the CORDIC algorithm is explicitly presented in [22]. For the HF injection sensorless control model, the input is the -axis current. After Park transformation, the HF response current is obtained through two band pass filters (BPF), which are realized by an infinite impulse response filter (IIR). By the phase locked loop (PLL) observer, the rotor position and speed are estimated for the field oriented closed-loop control. Aided by the MATLAB filter design tool, two forth-order BPFs and one second-order low pass filter (LPF) are designed and implemented. In the middle and high speed ranges, sensorless control is realized by the linearly compensated flux observer. HF voltage injection is turned off to avoid its interruption to the system. In the observer, there is a divider. It is a disadvantage for standard software solution. However, it is proper for FPGA, since FPGA consists of the implementation of a hardware divider. VI. EXPERIMENTAL RESULTS The sensorless control system represented in Fig. 8 was experimentally tested on a commercial 3-pole-pair SMPMSM servo motor made by Kollmorgen Seidel Corporation, fed by a two-level voltage source inverter (VSI), with a 6 khz PWM switching frequency. The motor parameters are given in Table II. The DC-link voltage was 300 V, the rated output voltage rms value of the inverter was 230 V. The shaft of the SMPMSM was mechanically connected to an induction motor load. An incremental encoder was fixed in the shaft to obtain the actual rotor speed and position. The frequency of the injected HF voltage signal was 1.2 khz. The measured maximum value of the d-axis HF current was about 180 ma, which ensured enough signal to noise ratio and lower acoustic noises. The saliency ratio in this machine was only 15%. (This value is calculated by, assuming.) According to (12), for the alternating HF injection method, the low saliency introduces a small real part of the current response TABLE II TECHNICAL DATA OF SMPMSM Fig. 12. HF injection sensorless control performance comparison between the PC and the FPGA-based system at 10 rpm. (a) Pentium system; (b) FPGA system. and makes it difficult to implement a traditional sensorless control algorithm. For comparison, the HF injection based sensorless control method was also implemented on a PC based real time system, which was equipped with a 600 MHz Pentium CPU (PC104) and several plug-in cards (12-bit A/D converter card, PWM generator card and so on). The sampling frequency was 16 KHz. Fig. 12 shows the performance comparison between the two systems. For the Pentium system, the maximum d-axis HF current was about 400 ma. Compared with FPGA-based system, the estimated position of the Pentium system had more ripples and the position estimated error was higher. Fig. 13 shows the sensorless speed and position performance from standstill to 800 rpm, with and without load. It is shown that sensorless control performance with load is better than that without load because the speed becomes much more stable when there is a load in the system. In Fig. 14, the speed reverse test shows high dynamic performance of the sensorless controller. The three curves in each block of Figs. 12 and 13, from the top down, are the U and V phase current, estimated error,

1260 IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS, VOL. 9, NO. 3, AUGUST 2013 Fig. 14. Speed reverse test from 60 to 60 rpm without load. the Delta-Sigma modulated current measurement and digital filters. Due to this high resolution current measurement and the FPGA oversampling technique based on Delta-Sigma A/D conversion, a much lower HF current response is sufficient to implement the sensorless control. From the experimental results, the magnitude of the HF current response based on the FPGA system was less than half of the PC-based system. This introduces lower acoustic noises, less losses and disturbances of the system. With a simple smooth switching strategy, the proposed FPGA-based hybrid sensorless controller has high performance over the whole speed range including the standstill mode. Fig. 13. Speed step from 0 rpm with a 10 rpm speed step every 0.4 second to 200 rpm, then 800 rpm (a) without load; (b) with 4 Nm load. and estimated rotor position, respectively. The three curves in Fig. 14, from the top down, are the U and V phase current, estimated rotor position, and estimated speed, respectively. VII. CONCLUSIONS This paper proposes an FPGA-based hybrid sensorless controller for SMPMSM using the alternating HF voltage injection technique and flux observer. Using model-based design, with tools such as MATLAB/Simulink and Simulink HDL Coder, the hybrid sensorless controller was designed and implemented on an FPGA board. The linearly compensated sensorless flux observer, which is robust to parameter variation and suitable for implementation in an FPGA system, is proposed. The stability analysis shows good performance of this flux observer despite its simple structure. High performance signal injection sensorless control is obtained with appropriate designs of REFERENCES [1] M. Jansson, L. Harnefors, and O. Wallmark, Synchronization at startup and stable rotation reversal of sensorless nonsalient PMSM drives, IEEE Trans. Ind. Electron., vol. 53, no. 2, pp. 379 387, Apr. 2006. [2] X. Ma and X. Wei, Voltage model of digital vector control and direct torque control systems, Trans. China Electrotech. Soc.,vol.19,no.3, pp. 65 69, Apr. 2004. [3] J. Gao, X. Wen, J. Chen, and F. Zhao, Novel motor stator flux observer basedonpll, Proc. CSEE, vol. 27, no. 18, pp. 41 47, Jun. 2007. [4] J. Hu and B. Wu, New integration algorithms for estimating motor flux over a wide speed range, IEEE Trans. Ind. Electron., vol. 13, no. 5, pp. 969 977, Sep. 1998. [5] F. Genduso, R. Miceli, C. Rando, and G. Galluzzo, Back EMF sensorless-control algorithm for high dynamic performance PMSM, IEEE Trans. Ind. Electron., vol. 57, no. 6, pp. 2092 2100, Jun. 2010. [6] G. Foo and M. Rahman, Sensorless sliding-mode MTPA control of an IPM synchronous motor drive using a sliding-mode observer and HF signal injection, IEEE Trans. Ind. Electron., vol. 57, no. 4, pp. 1270 1278, Apr. 2010. [7] J. Holtz, Acquisition of position error and magnet polarity for sensorless control of PM synchronous machines, IEEE Trans. Ind. Appl., vol. 44, no. 4, pp. 1172 1180, Jul. 2008. [8] M. Linke, R. Kennel, and J. Holtz, Sensorless position control of permanent magnet synchronous machines without limitation at zero speed, in Proc. IECON, 2002, pp. 674 679. [9] W. Hammel and R. Kennel, High-resolution sensorless position estimation using delta-sigma-modulated current measurement, in Proc. 2011 IEEE ECCE, 2011, pp. 2717 2724. [10] H. Iura, M. Inazumi, T. Kamei, and K. Die, Hybrid sensorless control of IPMSM for direct drive applications, in Proc. Int. Power Electron. Conf., 2010, pp. 2761 2767. [11] G. Wang, R. Yang, and D. Xu, DSP-based control of sensorless IPMSM drives for wide-speed-range operation, IEEE Trans. Ind. Electron., vol. 60, no. 2, pp. 720 727, Feb. 2013.

MA et al.: FPGA IMPLEMENTATION OF A HYBRID 1261 [12] E. Robeischl, M. Schroedl, and M. Krammer, Position-sensorless biaxial position control with industrial PM motor drives based INFORM and back EMF model, in Proc. IEEE IECON 02, 2002, pp. 668 673. [13] C. Silva, G. Asher, and M. Sumner, Hybrid rotor position observer for wide speed-range sensorless PM motor drives including zero speed, IEEE Trans. Ind. Electron., vol. 53, no. 2, pp. 373 378, Apr. 2006. [14] E. Monmasson, L. Idkhajine, M. N. Cirstea, I. Bahri, A. Tisan, and M. W. Naouar, FPGAs in industrial control applications, IEEE Trans. Ind. Inf., vol. 7, no. 2, pp. 224 243, May. 2011. [15] Y. Tezuka, S. Ichikawa, and Y. Noda, Design and implementation of hard-wired tracking control system, in Proc. TENCON 2010-2010 IEEE Region 10 Conf., 2010, pp. 299 304. [16] V. Delli Colli, R. Di Stefano, and F. Marignetti, A system-on-chip sensorless control for a permanent-magnet synchronous motor, IEEE Trans. Ind. Electron., vol. 57, no. 11, pp. 3822 3829, Nov. 2010. [17] G. Maragliano, M. Marchesoni, and L. Vaccaro, FPGA implementation of a sensorless PMSM drive control algorithm based on algebraic method, in Proc. 2010 IEEE Int. Symp. Ind. Electron., 2010, pp. 3083 3088. [18] L. Idkhajine, E. Monmasson, and A. Maalouf, Fully FPGA-based sensorless control for synchronous AC drive using an extended kalman filter, IEEE Trans. Ind. Electron., vol. 59, no. 10, pp. 3908 3918, Nov. 2012. [19] I. Bahri, A. Maalouf, L. Idkhajine, and E. Monmasson, FPGA-based implementation of sensorless AC drive controllers for embedded electrical systems, in Proc. Symp. Sensorless Contr. Electrical Drives, 2011, pp. 13 18. [20] J. Krah and C. Klarenbach, FPGA based field oriented current controller for high performance servo drives, in Proc. PCIM Power Conversion Intelligent Motion, Nürnberg, May. 2008. [21] Z. Zhou, T. Li, T. Takahashi, and E. Ho, FPGA realization of a highperformance servo controller for PMSM, in Proc. Appl. Power Electron. Conf. Expo., 2004, pp. 1604 1609. [22] Z. Ma, T. Friederich, J. Gao, and R. Kennel, Model based design for system-on-chip sensorless control of synchronous machine, in Proc. Symp. Sensorless Contr. Electr. Drives, 2011, pp. 85 89. Jianbo Gao wasborninchina.hereceivedtheb.s., the M.S. and the Ph.D. degrees from Tsinghua University, Beijing, China, in 1994, 1997, and 2001, respectively. In 2001 he joined the Fraunhofer Institute for Biomedical Engineering in Germany as a researcher. Since 2002 he became the Technical Director of the Chinese Branch of the Institute. After a short stay in the automotive industry from 2008 to 2009, he has been the researcher in the Institute for Electrical Drive Systems and Power Electronics at the Technical University of Munich since 2010. His interests include sensorless control of electric drives and signal processing technologies. Ralph M. Kennel (M 89 SM 96) received the Dr.- Ing. (Ph.D.) degree from the University of Kaiserslautern, Germany, in 1984. From 1983 to 1999, he was holding several positions with Robert BOSCH GmbH, Germany. Until 1997, he was engaged in the development of servo drives. Between 1997 and 1999, he was responsible for Advanced and Product Development of Fractional Horsepower Motors in automotive applications. From 1994 to 1999, he was appointed Visiting Professor at the University of Newcastle-upon-Tyne, U.K. Since 1999, he has been a Professor for electrical machines and drives at Wuppertal University, Germany. Since 2008, he is a Professor for electrical drive systems and power electronics at the Technische Universität München, Munich, Germany. His current research interests include sensorless control of ac drives, predictive control of power electronics, and hardware-in-the-loop systems. Prof. Kennel is a Fellow of Institution of Electrical Engineers, U.K., and a Chartered Engineer in the U.K. Zhixun Ma was born in Jiangsu, China, in 1984. He received the B.S. and M.Sc. degrees in electrical engineering in 2006 and 2009, respectively, from China University of Mining and Technology, Xuzhou, China. Since 2009, he has been with the Institute for Electric Drive Systems and Power Electronics, Technical University of Munich, where he is currently pursuing the Ph.D. degree. His main research interests include sensorless control of electrical drives and FPGA-based control of power electronics and drive systems.