Rapid Prototyping Applications on Three-Phase PWM Rectifier and Shunt

Active Power Filter

Kadir Vardar, Tolga Sürgevil, and Eyüp Akp nar

Dokuz Eylul University, Department of Electrical and Electronics Engineering

Kaynaklar Kampusu, 35160 Buca, Izmir, TURKEY
e-mail: [email protected] , [email protected] , [email protected]

Abstract electronics systems contains a background of inter-disciplinary

knowledgebase [13].
In this paper, the design procedures of a unity power factor The block scheme of a three-phase unity power factor PWM
pulse width modulated (PWM) rectifier and shunt active rectifier is given in Fig.1a. The major parts of this topology are a
power filter (APF) using MATLAB/Simulink blocksets and power converter that couples the AC supply voltages and DC
code generation tools for TMS320F2812 fixed-point digital link capacitor voltage through a boost inductance. In this
signal processor (DSP) is presented. A systematic design scheme, the rectifier circuit provides a regulated dc link voltage
approach is given to develop a rapid prototype system in a to the load providing the ac line currents almost sinusoidal and
reliable procedure. The design procedure uses Simulink at unity power factor. The control logic relies on measurement
model of the rectifier and shunt APF systems, control of dc link voltage, supply voltages and supply currents. The
circuitry, real-time DSP program code generation. The block scheme of a three-phase shunt APF compensating the
generated and self-developed codes on DSP are tested with current harmonics of a non-linear load is shown in Fig.1b. The
Simulink models; therefore, any possible error in the APF power circuit topology is the same as that of PWM rectifier
controller implemented within software is eliminated before in parallel with the load. The load current waveforms are
connecting the power converter to the supply. The code measured through the current sensors and their harmonic
generation tools of Simulink provide an efficient way in the contents are extracted through instantaneous reactive power
transition from the analysis stage to real system theory (IRPT) method. The inverse of the load current
implementation. The effectiveness of this procedure for harmonics are injected to the electrical network through APF.
safely starting up the system is experimentally verified. This is achieved by an inner current control loop to force the
APF currents track the reference current waveforms generated in
1. Introduction DSP in a similar fashion developed in PWM rectifier. The PI
controller regulates the DC bus voltage at the desired level.
The power quality issues emerging due to increasing usage of In the design procedure of such devices, it is often necessary
non-linear loads have been the major concern in last two to construct the complete system model to investigate its
decades and these problems have also created a specialized performance through computer simulations before
subject in power engineering. Employing active power filters implementing. In [14] it was noted that the use of hardware in
(APF), Static VAR Compensators (STATCOM), universal simulation of the power electronics converter is essential. The
power quality conditioners (UPQC) are some of the inclusion of DSP based design into computer simulations can be
technologies developed as solutions to power quality issues [1- carried out by using the program of DSP and co-operation with
4]. In industrial applications, it is convenient to employ APFs to PC. In recent years, the use of automatic code generation tools
reduce the harmonic content of supply current in motor drives. has become widespread. These tools can be used effectively to
APFs are also known to be the alternative tool for other power make a rapid prototyping and hardware-in-the-loop testing of
quality problems such as reactive power compensation, load controllers possible [15]. From this point of view, the Embedded
balancing, voltage regulation, and voltage flicker compensation Target tools that integrate the MATLAB/Simulink and Texas
[5]. From another point of view, the PWM rectifiers also Instruments TMS320C2000 DSP platforms are useful to develop
provide an indirect solution to power quality related problems in and validate the designs rapidly. These tools are used to generate
terms of offering salient features of operation at unity power a program in C-language for real-time implementation of the
factor and minimized harmonic current injection to the power Simulink model and also provide a flexibility to test the
system unlike conventional diode or thyristor controlled generated code with other Simulink models [16].
rectifiers. Due to their several advantages, the PWM rectifiers In this paper, a design procedure for three-phase unity power
are also found place in applications such as ac motor drives, factor PWM rectifier and shunt APF using MATLAB/Simulink
uninterruptible power supplies, etc. [6-9]. and TMS320F2812 DSP is presented. The procedure contains
Nowadays, many power electronics devices are integrated the modeling of the system in Simulink, real-time code
with digital signals processors (DSP) that offer a generic generation, testing the developed code on DSP with Simulink
structure and implementation of complex calculations required model, and experimental tests carried on the prototype system.
in a wide range of applications [10-12]. While the DSPs offer In Section 2, the Simulink modeling for computer simulations,
better solution to the control of many complex systems, real-time code generation for DSP controller and simultaneous
developing a DSP code and testing it on the real system requires operation of DSP with Simulink for PWM rectifier are
additional effort and attention since the design of power presented. In Section 3, the same procedure is applied for shunt
APF and the self-developed DSP code for this application is

I-258
tested with DSP in Simulink simulations. The measured results voltages with a gain. The switching pulses are generated in
and conclusions are given in sections 4 and 5, respectively. hysteresis current control block, which ensures the actual
currents are forced to track reference current waveforms in a
2. Modelling and Code Generation for PWM Rectifier specified hysteresis band. The output of the DC link voltage PI
controller is calculated as follows [17]:
The circuit model of the PWM rectifier is constructed in
MATLAB/Simulink, which contains power system and power I cm ( k ) = I cm ( k − 1 ) + K Δe [ e( k ) − e( k − 1 )] + K ee( k ) (1)
electronics tools to analyse the systems and specific blocks used
in code generation for C2000 DSPs. This model contains the where Icm is the ac line current magnitude, e is the error
measurement of voltage and current signals in the power circuit between the reference and the actual values of the dc link
via analog-to-digital conversion and digital processes that voltage, and KΔe and Ke are the proportional and integral gains
should be developed on DSP as shown in Fig.2. Here, the DSP for a given sampling rate. The indexes (k) and (k-1) indicates
section is performing calculations for the control circuitry and the sampled values at the present and the previous sampling
generating the gate pulses of IGBTs. This block is designed as a instants, respectively. The sampling rate for the digital PI
subsystem and contains the part from which the DSP code will controller is defined as the sampling rate of the ZOH blocks in
be generated. the model. The system parameters are given in Appendix.
Once the stable operation of the system is ensured in
simulations, the C-code of the DSP block is generated using
MATLAB C2000 code generation tools. The voltage and current
waveforms generated in simulation model are fed to DSP in
order to emulate the signals from the sensors of the physical
system. The DSP processes these waveforms and generates the
switching patterns at each simulation step. Here, the generated
code by Simulink contains the part after the measured signals
are read from analog-to-digital converter (ADC) registers and
before the gate signals are written into PWM registers only. So,
the initialization of these registers that should be done in the
complete DSP code developed for the designed system
operation is not included in the Simulink DSP subsystem. In that
way, the generated DSP code for current control loop and dc
link PI controller can be tested without using the real system in
order to avoid any failure and physical damage to IGBTs and
other power devices.

2.1. Simulink Model of PWM Rectifier

In the MATLAB/Simulink model of PWM rectifier, the

voltage and current signals are conditioned to 0-3V ADC levels
of F2812 DSP. The gains of the sensors, signal conditioning
circuits and ADC are defined in the main model as shown in
Fig.2. The system is simulated using discrete-time simulation
option and the integration step size of the solver is chosen to be
4 microseconds. The rectifier parameters are given in Appendix.
Fig. 1. Block schemes of the two designed systems a) PWM
Zero order hold (ZOH) blocks in the main model and the DSP
rectifier circuit b) shunt APF circuit
subsystem represents the sampling rate of the DSP including the
analog-to-digital conversion and execution time that is to
complete one cycle of calculations in the main program. Once 2.2. Code Generation and Simulation of PWM
the execution time on the DSP is measured, the sampling rate of Rectifier with DSP
the ZOH blocks in the model is set to an appropriate value larger
than the execution time considering the analog-to-digital The Simulink block scheme in Fig.3 is used for testing the
conversion brings about an extra conversion time. Here it must switching signals of IGBTs before connecting the DSP to IGBT
be noted that the integration step size should be kept as low as block in the system. This model acquires the switching signals,
possible with respect to the defined sampling rate (Tsmp) of ZOH which are generated by DSP, and applies them to IGBT bridge
blocks. That is because of the data transfer rate’s being limited block in Simulink. The memory locations are defined on DSP
to the sampling rate while the electrical quantities in the model for voltages, currents, and gate pulses. Simulink accesses these
is changing during this time interval and the states of the model memory locations when simulating the system with DSP and
should be solved during this interval with smaller step sizes. writes the variables into the specified memory locations at each
In the model of the PWM rectifier, the DSP Section numerical solver integration step. In order to ensure the
subsystem includes the program part that is loaded to DSP after operation of DSP at the sampling rate defined in Simulink, an
code generation. The DSP section is constructed for closed-loop additional memory location, which is accessed both Simulink
control incorporating the DC link voltage PI controller. In this and DSP is also defined. The status of this memory location is
block, the gains of measurement device and ADC are toggled at the beginning of each sampling interval defined in
transformed back to the real voltage and current values. The Simulink. When the Simulink model allows the DSP to execute
reference current waveforms are generated from the supply its code to obtain the switching states for the next sampling

I-259
interval, the calculations are performed on the DSP and
switching signals are written into the memory locations. Then,
the Simulink is enabled to read these values to perform the next
step of simulation. Hence, the calculations for the complete
system operation are performed in Simulink and DSP
sequentially: This kind of operation is essential especially in PI
controlled scheme, because the controller needs the previous
values of the sampled variables as shown in (1).
After generating the DSP code by using Simulink, the DSP
Section in the Simulink model is replaced by a developed
Simulink-Code Composer Studio (CCS) interface block. This
interface block in the model is an s-function [18] that provides
access to the memory locations reserved for the signals on DSP.
This s-function is created for the system designed. Hence, before
running the simulations with DSP, the generated code is loaded
to DSP and the communication between the MATLAB and CCS
was established.
Simulations are performed at fixed integration step size,
which is set forth in the Simulation Parameters in Simulink. The
proper operation of DSP code is tested before connecting it to
the real system. The DSP execution time is measured to be 50
microseconds for closed-loop rectifier operation in the generated Fig. 2. The Simulink model of PWM rectifier
code. Hence, the sampling rates are set to the values slightly
larger than these measured values. The results obtained from
Simulink alone and Simulink connected to DSP are obtained to
be very similar to each other. In Fig.4, the results for closed-
loop operation with DC link voltage PI controller are given.
Since the results of both simulation and DSP in simulation are
identical, the results of PI controlled rectifier are given for DSP
in simulation only. The DC link voltage reference was set to
300V and actual DC link voltage settles to its steady-state value
in 0.15s as shown in Fig.4.

3. Code Development for Shunt APF and its Fig. 3. DSP code generation block in Simulink
Simulation with DSP

When the model in the Simulink program is converted to DSP

code using Matlab DSP toolbox as being described in Section 2,
the sampling frequency of ADC is the main cycling frequency
and this becomes the switching frequency of currents compared
in hysteresis band. This frequency is getting lower as the
number of inputs read by ADC. In shunt APF application, the
mathematical burden of the DSP is much more than that of the Fig. 4. Simulation result of simultaneous operation of Simulink
case in PWM rectifier, because the DSP should also calculate with DSP for PWM rectifier a) AC line currents b) DC Link
the load harmonic content, produce filter reference currents voltage
through IRPT algorithm, and dc link PI controller. Also, in the
APF application, there are six currents and four voltages read by In order to increase the comparison frequency between actual
DSP. The shunt APF application requires an input boost and reference currents in hysteresis band, the reference current
inductance chosen at a considerably lower value in order to generation and hysteresis current controller are operated as two
achieve higher switching frequencies to eliminate a large portion independent tasks in the DSP code of shunt APF. The
of the load harmonic current content. This necessitates faster instantaneous reactive power calculation and extraction of
switching of IGBTs and current tracking. From this point of harmonics are performed in other dedicated routine called by
view, the code generation through Simulink DSP tools may not main program. While the reference current generation is
be effective. For instance, the execution time of the DSP is performed in the main program cycle at a rate of 10kHz, the
obtained approximately 50 microseconds in closed-loop hysteresis current controller compares the actual and reference
controlled PWM rectifier given in the previous section and this current values and updates the switching signals at a 100kHz
corresponds to a sampling rate of 20kHz, which is the rate of rate in an interrupt service routine (ISR).
updating the states of the switching signals. Considering the The operation of DSP with Simulink model of shunt APF is
IRPT algorithm implemented in the APF application, it is an also carried out in a similar way as mentioned in PWM rectifier
obvious fact that this rate will decrease to a lower value since section. The Simulink model reads the switching signals
the algorithm requires complex calculations. Because of this generated by DSP to perform numerical solver operation in
reason, the DSP code is developed by user in C-language in simulation and writes the results to the memory locations, which
order to keep the execution time as low as possible. are defined on DSP for line voltages, filter currents, load

I-260
currents, and dc link voltage at each integration step of the
numerical solver. While testing the DSP code with Simulink
model, the sampling rate in the model is set to 100kHz, which is
the rate of ADC and hysteresis current control (HCC) switching.
So, the interrupt service routine for HCC is removed into main
program cycle, which is not the case in the real system
operation. Finally, the rate for updating reference current value
was measured to be approximately 10kHz in the developed
program, these values are updated in the simulations over 10
main program cycle. The Simulink model of the system is
shown in Fig.5. A switching ripple filter is also connected to the
input terminals of APF to trap the high frequency components
that would appear on the filter currents and parameters of this
circuit are given in Appendix.
The communication between the MATLAB and CCS is
established using the s-function mentioned in Section 2.2 with a
few modifications to define additional memory locations. The
proper operation of DSP code for shunt APF is assured before
connecting it to the gate drive circuit of IGBTs in the prototype Fig. 5. The model of simultaneous operation of shunt APF
system. The entire circuit and control parameters of shunt APF model in MATLAB/Simulink and TMS320F2812 DSP
are given in Appendix. Fig.6 shows the results of obtained from
simultaneous operation of DSP with Simulink.

4. Result of Implementation

In the power converter part of the hardware, SEMIKRON

SEMITEACH three-phase IGBT module was used. This module
is equipped with built-in gate drive circuits and protection
circuits for short-circuit, overheating, and supply under-voltage
protection. The three-phase voltage and current waveforms are
measured via Hall-effect voltage transducers (LV-25P) and
current transducers (HXS20). The measured voltage and current
waveforms are input to ADC ports of the TMS320F2812 DSP
by conditioning these signals to 0-3V ADC voltage levels. Fig. 6. Simulation results obtained from simultaneous operation
of Simulink with DSP for shunt APF a) Diode rectifier currents
4.1. PWM Rectifier b) APF currents c) supply currents d) APF dc link voltage

Once the control algorithm is tested using the DSP and 5. Conclusions
Matlab-Simulink in a simultaneous operation, the DSP can be
connected to the power block having IGBTs, gate drives and The DSP program of a 3-phase unity power factor PWM
signal conditioning circuits. In the closed-loop control, the DSP rectifier is prepared by using MATLAB package program. This
can perform calculations and generate switching pulses in 50 is implemented using the facilities of MATLAB Simulink and
microseconds with standard floating-point operations, which is TMS320F2812 DSP tools. These embedded code generation
satisfactory for the control of the complete system. The results facilities provide a safe transition from the analysis stage to
of closed-loop operation of the real system at the start-up implementation one in power electronics applications. The
transient and steady-state are shown in Fig.7. Since the approaches given here can be implemented for rapid prototyping
simulations are performed under ideal conditions neglecting of the system in industrial applications. The proper operation of
noise effect on signal conditioning, the noise and switching the DSP is guaranteed through simulations before testing the
ripples due to the turn-on and turn-off characteristics of IGBTs code on physical system. The simultaneous operation of the
do not appear in the simulation results unlike the real system. DSP with the Simulink models was also established by the
The reference currents are updated in the main program cycle at developed model structure. This approach has been further
the rate of 20kHz, which is a satisfactory for rectifier operation. extended to shunt APF application in this work. MATLAB code
generation tools for the applications requiring a large amount of
4.2. Shunt APF mathematical computations such as IRPT in shunt APF maybe
inefficient; therefore, the structure of the DSP program can be
The experimental results obtained when shunt APF reorganized and developed by the user and tested in simulation
compensating the current harmonics of a diode rectifier load are model. Code generations tools of MATLAB can be considered
shown in Fig.8. During experiments, the dc link voltage is to be useful in providing rapid prototyping applications that
boosted to 650V and the APF reference currents for harmonic requires less mathematical computations. Even though the
compensation are generated approximately 4 cycles after the dc structure of the system needs more time consuming in
link voltage is settled to its reference value. The APF reference execution, the procedure described here can still provide an
currents are updated in the main program cycle at a rate of efficient and safe way in testing the DSP codes prepared for real
10kHz, which is a satisfactory frequency range for the system implementation.
compensation of load harmonics.

I-261
 Acknowledgment
This work was carried out as a part of project, “Power
Quality National Projects”, sponsored by Turkish Scientific and
Research Council and Turkish Electrical Power Transmission
Co. (TEA) under contract 106G012.

Appendix
PARAMERTS OF PWM RECTIFIER:
Parameter Symbol Value
Supply voltage peak value Em 100V Fig. 7. The dc link voltage and line current of PWM rectifier a)
AC side filter inductance Lf 13 mH
steady-state b) startup transient
Resistance of filter inductance Rf 0.1 ohms
DC link capacitor C 1.1 mF
Load resistance R 100 ohms
DC link voltage initial value Vdc(0) 160 V
Hysteresis band width εh 0.2 A
PI controller gains Ke 0.001
KΔe 0.1
PARAMETERS OF SHUNT APF
Parameter Symbol Value
Supply voltage peak value Em 310V
AC side filter inductance Lf 1.8 mH
Resistance of filter inductance Rf 0.1 ohms
AC side supply inductance Ls 100 μH
DC link capacitor C 2.35 mF
Diode rectifier load resistance RL 30 ohms
DC link voltage initial value Vdc(0) 550 V
Hysteresis band width εh 0.2 A
PI controller gains Ke 0.001 Fig. 8. Experimental results of the shunt APF a) load currents b)
KΔe 0.1
supply currents c) APF dc link voltage.
Switching ripple filter components L 0.25mH,
R 3 ohms, [8] M.P. Kazmierkowski, ‘Control strategies for PWM
C 30 μF rectifier/inverter-fed induction motors’. Proc. IEEE Int. Symp. on
Industrial Electronics, Puebla, Mexico, December 2000, pp.15-23.
6. References [9] M. Cichowlas, M. Malinowski, M. Jasinski, and M.P.
Kazmierkowski, ‘DSP based direct power control for three-phase
[1] H. Akagi, ‘Trends in active power line conditioners’, IEEE Trans. PWM rectifier with active filtering function’. Proc. IEEE Int.
Power Electronics, 1994, vol.9, no.3, pp. 263-268. Symp. on Ind. Elect.., Brasil, June 2003, pp.831-835
[2] M.K. Mishra, K. Karthikeyan, and P.K. Linash, ‘A development [10] M. Aredes, J.A.M. Neto, J.C.C. Ferreira, L.F.C. Monteiro, R.M.
and implementation of DSP based DSTATCOM to compensate Fernandes, and M.J.V. Siqueira, ‘A simplified control strategy for
unbalanced nonlinear loads’. Proc. IEEE Power India Conf., New a unified power quality conditioner prototype’. Proc. IEEE Power
Delhi, India, April 2006, 8 pp. Electronics Specialist Conf., Brasil, June 2005, pp.2592-2597.
[3] Akagi, H.: ‘New trends in active filters for power conditioning’, [11] E. Ozdemir, M. Ucar, M. Kesler, and M. Kale, ‘The design and
IEEE Trans. Ind. Applications, 1996, vol.32, no.6, pp. 1312-1322. implementation of a shunt active power filter based on source
[4] B. Singh, K. Al-Haddad, and A. Chandra, ‘A review of active current measurement’. Proc. IEEE Int. Electric Machines & Drives
power filters for power quality improvement’, IEEE Trans. Conf., Antalya, Turkey, May 2007, pp.608-613.
Industrial Electronics, 1999, vol.46, no.5, pp. 960-971. [12] C. Yanbo, Z. Fudan, and K.W.E. Cheng, ‘Shunt active power
[5] A.Emadi, A. Nasiri, and S.B. Bekiarov, ‘Uninterruptible Power filter-SIMULINK simulation and DSP-based hardware
Supplies and Active Filters’ (CRC Press, 2005). realization’. Proc. Int. Conf. on Power Electronics Systems and
[6] E.B. Barcenas, V. Cardenas, and J.A. Roffiel, ‘Shunt Active Power Applications, Hong Kong, November 2006, pp. 120-125.
Filters and PWM Rectifiers in Three-Phase Three Wire Systems: A [13] R. S. Balog, Z. Sorchini, J. W. Kimball, P. L. Chapman, and P. T.
Survey’. Applied Power Electronics Conference and Exposition, Krein, ‘Modern Laboratory-based education for power electronics
Dallas, Texas, March 2006, pp.1307-1313. and electric machines’, IEEE Trans. Power Systems, 2005, vol.20,
[7] M.P. Kazmierkowski, and L. Malesani, ‘Current control no.2, pp. 538-547.
techniques for three-phase voltage-source PWM converters: a [14] M. Dawande, V. Donescu, Z. Yao, and V. Rajagopalan, ‘ Recent
survey’, IEEE Trans. Industrial Electronics, 1998, vol.45, no.5, advances in simulation of power electronics converter systems’,
pp. 691-703. Springer India, 1997, vol.22, no.6, pp. 689-704.
[15] C. Defour, S. Abourida, and J. Belanger, ‘Hardware-in-the-loop
simulation of power drives with RT-LAB’. Proc. Int. Conf Power
Electr. and Drives Systems, Canada, 2005, pp. 1646–1651.
[16] The Mathworks, Inc.: ‘Embedded Target for the TI
TMS320C2000TM DSP Platform for use with Simulink-User’s
Guide Version 1’ (The Mathworks, Inc., 2003).
[17] T. Surgevil, and E. Akpinar, ‘Modelling of a 5-kW wind energy
conversion system with induction generator and comparison with
experimental results’, Renewable Energy, 2005, 30, pp. 913-29.
[18] ‘Writing s-functions’ (The Mathwoks, Inc., 2003).

I-262