You are on page 1of 9

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 25, NO.

1, JANUARY 2010 219

Solving EMI-Related Problems for Reliable


High-Power Converters Design With Precomputed
Electromagnetic Models
Handy Fortin Blanchette, Student Member, IEEE, and Kamal Al-Haddad, Fellow, IEEE

Abstract—Interactions between low-power control circuits such considerable time to construct the model. Thus, EMI analysis is
as DSP or field-programmable gate array (FPGA), and high-power often neglected. Consequently, converter designers must solve
switching components that generate high ∂i/∂t greatly decrease EMI problems at the prototyping stage, resulting in increased
the reliability of power converters. Moreover, they increase sub-
stantially the product’s time to market when introducing a new costs and engineering risks. The precomputed electromagnetic
design. The analysis of these couplings is traditionally done by models (PEMs) presented in this paper are the key elements
using the finite elements method, which requires high technical of the solution to this problem. The paper is organized as fol-
expertise for complex 3-D circuit geometries. To overcome this re- lows. In Section II, important electromagnetic issues are dis-
quirement, the authors propose an approach based on the use of cussed. In Section III, a step-by-step PEM design is shown. In
a precomputed electromagnetic model (PEM) library, which in-
cludes several metallic conducting structures present in modern Section IV, an application of these models is proposed, which
power converters. Transistor cases, as well as printed circuit board consists of validating the presence of an oriented field plane in
(PCB) copper traces, are included in this library. By using these a matrix converter cell. Experimental results are shown for this
library models, high-coupling areas with higher flux density can be application.
quickly identified in the converter 3-D layout. This helps to develop
a remedial strategy for avoiding higher flux density areas. The pro-
posed approach is used to analyze the electromagnetic behavior of II. MAGNETIC DIFFUSION AND FIELD PROPAGATION
a bidirectional matrix converter cell. Experimental results are also Designing a library containing PEMs of common metallic
presented to further demonstrate the viability of the new proposed
approach.
structures used in power electronics is based on the same funda-
mental principle as other existing libraries: the values included in
Index Terms—Electromagnetic coupling, electromagnetic inter- the library must be static. In electromagnetism, two fundamen-
ference (EMI), matrix converter, printed circuit layout, reliability.
tal phenomena go against this principle: the magnetic diffusion
observed by the conductor skin effect and the electromagnetic
I. INTRODUCTION wave propagation, which introduces a delay between the elec-
tromagnetic field and the source of this field. In order to work
UILDING A power converter printed circuit board (PCB)
B is a challenging task for designers because of electromag-
netic interference (EMI) problems and the requirement to work
in the static domain and create the proposed library, these two
important phenomena must be neglected. However, this cannot
be done without a rigorous analysis of the errors introduced by
with multiple material technologies. The flexible-foil PCB pre-
neglecting both phenomena. Let us consider first the magnetic
sented in [1] is a good example of new material used in power
diffusion. The equation managing this complex phenomenon is
electronics. EMI problems must be also solved by PCB design-
given by
ers. Two types of EMI can occur in power converters, namely
conducted EMI and radiated EMI. Conducted EMI study re- 
∇×H  = ∂H . (1)
lates to power converter capacitive paths, which are a source ∂t
of differential- and common-mode noise. Conducted EMI is
The relation (1) is difficult to solve for complex geometries
more extensively covered in publications [2]–[6] than radiated
such as those frequently used in power printed circuits. Since
EMI [7]. This is due to the fact that since radiated EMI requires
the goal of this study is to determine the effect of the magnetic
a 3-D model of the circuit geometry to compute the magnetic
diffusion on the external magnetic field, the selected structure,
field distribution, it is a very complex task. Obviously, finite-
shown in Fig. 1, is well adapted because a complex trace can be
element-analysis-based (FEA) softwares can be used to do this
decomposed into rectangular current tubes [8]. This is consid-
task. However, the simulation of 3-D parts, such as those used
ered an infinite rectangular printed circuit trace oriented along
in power converters, is quite cumbersome. Moreover, it takes
the x-axis.
The analysis is achieved in two steps. The first step is to
Manuscript received March 25, 2009; revised June 5, 2009. Current version compute the current density distribution in the conductor as
published January 29, 2010. This work was supported by the Natural Sciences
and Engineering Research Council, Canada, Research Chair and OPAL-RT a function of time. The next step is to compute the external
Technologies. Recommended for publication by Associate Editor P. Tenti. magnetic field by using current density distribution at a given
The authors are with the Department of Genie Electrique, Ecole de time. Obviously, the static case that is used to compute the
Technologie Supérieure (ETS), Montreal H3C1K3, QC, Canada (e-mail:
handy.blanchette@opal-rt.com; kamal.al-haddad@etsmtl.ca). library models is obtained when t → ∞. Fig. 2 shows the current
Digital Object Identifier 10.1109/TPEL.2009.2027708 densities magnetic diffusion in the metallic trace shown in Fig. 1.
0885-8993/$26.00 © 2010 IEEE
220 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 25, NO. 1, JANUARY 2010

Fig. 1. Typical PCB copper trace.

Fig. 3. External magnetic field distribution during the magnetic diffusion.

where J is a current density in a metallic conductor, c is the


speed of light, and ω is the current density angular frequency.
The switching currents flowing in the power transistors and
PCB traces contain many harmonics. Obviously, the phase shift
introduced by the complex numerator in (2) changes with the
frequency to preserve the waveform integrity. Consequently,
the propagation delay can be studied only for the fundamental
switching frequency. Modern power converters use switching
frequencies up to 200 kHz to reduce the size of passive elements.
Fig. 2. Current densities diffusion in a PCB copper trace. At this frequency, the propagation delay for a distance of 10 m
is 33 ns. This delay is sufficiently small to be neglected.
The two fundamental studies presented in this section are the
At t = 0, a current step is applied to the conductor. Immedi- basis for the construction of a precomputed library of electro-
ately, all the current flows at the conductor surface explaining magnetic models. With the magnetic diffusion being sufficiently
the high current densities present at the conductor periphery. small to be neglected, it is possible to compute the current densi-
As time progresses, the current densities diffusion toward the ties under static conditions by using the Laplace equation. Also,
conductor center is accentuated. As shown in Fig. 2, the current with the propagation delay of these signals being very small, it
densities in the conductor become uniform at t = 1 µs. is possible to link the magnetic field inside the converter to the
The external magnetic field generated by these current den- current densities flowing in the converter at the same time. This
sities is computed at y = 1.5 mm. The trace thickness (Ep ) completes the prerequisites to establish a PEM library.
being 0.5 mm, the distance between the trace and the point
where the magnetic field is computed is 1 mm along the y -axis. III. PRECOMPUTED ELECTROMAGNETIC MODEL DESIGN
This distance corresponds to an insulation of 1 kV, which is The precomputed models are generated in five steps. This
a minimum value even in low-voltage power converters, due section describes each step in detail.
to creepage requirements. The field modulus is computed at 1) The first step is to define the metallic structure geometry, in
t = 1, 12.5, and 875 ns, and the results are shown in Fig. 3. The particular, the structure dimensions and orientation in space as
maximum error between the field computed at the beginning illustrated in the 90◦ metallic elbow structure depicted in Fig. 4.
of the magnetic diffusion (t = 1 ns) and the field computed at A referential axis system must be also assigned to the structure
the end of the magnetic diffusion (t = 875 ns) is 12.8%. This to fix its origin. The referential axis system for the structure
error is sufficiently small to neglect the magnetic diffusion in shown in Fig. 4 is identified by the coordinate system (xp , yp )
the study of the magnetic field generated by power converters. fixed at the origin (0, 0, 0).
The next point to study is the impact of electromagnetic prop- 2) Since there is no closed form solution to compute the mag-
agation. Magnetic field propagation can be computed by using netic field generated by most of the metallic structures used in
the retarded magnetic potential formulation, given by (2) [9] power converters, such as the one shown in Fig. 4, a numerical
approach is required. In this study, the FEA is selected to com-
  −jωr
 = µ0 Je c plete this task. Consequently, the second step is to determine
A dV (2) the boundary conditions that must be applied to the metallic
4π V r
BLANCHETTE AND AL-HADDAD: SOLVING EMI-RELATED PROBLEMS FOR RELIABLE HIGH-POWER CONVERTERS DESIGN 221

Fig. 4. 90 ◦ metallic elbow structure.

Fig. 5. Interconnection between two models.

structure to compute the potential distribution over the trace.


With this distribution, one can obtain the current densities re- This constraint comes from the fact that current densities are
quired to compute the magnetic field. There are two types of perpendicular to the isopotential lines [11].
boundary conditions for metallic structure, namely the flux that 3) The third step is to compute the current densities according
is flowing outside the structure (Neumann condition) and the to the boundary conditions applied at the previous step. This task
fixed potentials inside the structure (Dirichlet condition) [10]. is handled by the FEA software.
All settings of the flux flow out of the structure are set to zero 4) The magnetic field is computed at this step. This is the most
because the current densities are not allowed to cross the ex- critical part of the PEM design. Most metallic structures, such as
ternal boundaries of the metallic structure, except where there MOSFET cases, are 3-D in nature and the current densities flowing
are connection points (called ports) with other structures. This in these structures are also 3-D. Consequently, to account for
condition can be easily handled by FEA software. this 3-D current densities distribution, the full magnetic vector
The next step is to apply the fixed potentials on the structure. potential formulation given by (3) must be solved according to
The magnitudes of these potentials are arbitrary because the boundary conditions
current flowing in the structure is normalized to one Ampere  = −µJ.
∇2 A (3)
before its inclusion in the library. How to apply these potentials
is a more complex issue. To keep things simple, the structure These boundary conditions are quite complex to compute be-
is divided into two parts, namely the input/output and the com- cause current densities are different everywhere inside the metal-
plex geometry. These parts are illustrated in Fig. 4 for the 90◦ lic structure. As previously discussed, these current densities
metallic elbow structure. The major concern faced during this are computed by solving the Laplace equation using the finite-
step is the length of the minimum input and output ports of the element method (FEM). The use of the FEM permits the compu-
structure. Eventually, many metallic structures are connected to tation of the electrical potential within the metallic structure. By
form a complete power PCB. However, current densities mis- using this electrical potential distribution, the current densities
alignment between two structures can arise if port lengths are are computed anywhere within in each element by interpolation.
too short. This misalignment reduces model precision because If the magnetic field is computed through the use of a second
the magnetic field orientation is directly related to the current FEA using the current densities provided by the first FEA, then
densities orientation [8]. This is illustrated in Fig. 5. At the top each element must be divided to compute the current densi-
of the figure, the output port of model 1 is too short and the ties into subelements. This results in a fine meshing. Averaging
current densities are not parallel to the structure. If model 2 is these current densities decreases the precision of the magnetic
connected to model 1, there will be a nonnull α angle between field computation, and is therefore, unsuitable for identifying
the current densities of both models showing a current densities converter leakage inductances [12], a key parameter for evalu-
misalignment. At the bottom of Fig. 5, the model 1 output port is ating the power semiconductors switching losses. To solve this
extended to force a parallel distribution of the current densities issue, the magnetic field is computed by two different methods.
at the end of the port. By extending the output port, the angle α Outside the metallic structure, the magnetic field is computed
will equal zero, eliminating discontinuities between the models. by the FEM. Inside the structure, the magnetic field is computed
222 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 25, NO. 1, JANUARY 2010

Fig. 7. PEM software implementation.

5) The fifth and final step is the storage of the PEMs. To avoid
storage of large matrices, the field is computed at a larger step,
Fig. 6. Gauss points for order 4 and 5 formulas.
as distance from the metallic structure increases. This is possible
because the field decreases by the inverse squared distance from
the current densities. Inside and very close to the structure, the
directly by using the Biot and Savart integral law. This integral field is computed every 0.25 mm. This zone is called the near-
law is outlined in the following equation: field area. At greater distances from the metallic structure, the
field variation is weak, and can be computed with a very rough
 step of 5 mm. This is the far-field area. Between the near- and
 = µ0 J × ar
B dV (4) the far-field areas, there is a transition area called the middle-
4π V r2
field area, where the field is computed every 1 mm. The total
where ar is the unit vector, directed from the current density storage space for each model is approximately 2.36 MB, which
J toward the point where the magnetic field is computed, and is very small in comparison with today’s commercially available
r is the distance between them. The main advantage of using storage technologies. This completes the description of PEMs.
this integral form is the resulting ease in accounting for the The next section describes how to use the PEM to perform
variable current densities in the conductor. The numerical inte- analysis of a PCB.
gration used to solve the Biot and Savart integral must be fast,
precise, and fully compatible with the FEM for postprocessing. A. PCB Design With PEM
The Gaussian quadrature integral method respects all these cri-
Five steps are required to build a PCB using the PEM library.
teria. The other advantage to using this integration method is the
The first step is to fix the circuit topology. The second step is
ability to easily change the integral formula order. Since the in-
to choose the converter components as a function of the power
tegrand must be evaluated at multiple locations in each element,
required. The third step is to place the parts in space. In fact,
it is possible that a Gauss point may match with the point where
these three first steps are identical to the usual approach taken
the magnetic field is computed, thereby causing a singularity
to build PCBs for power converters. The fourth step is to route
during the integration process [r becomes 0 in (4)]. By changing
the PCB by using PEMs included in the library. If a structure is
the Gauss formula order, the Gauss points are modified, thereby
not in the library, the procedure described in Section III must be
solving this problem. This is shown graphically in Fig. 6 for
used to include it. The fifth step is to join the PEM with straight
orders 4 and 5. With the fourth-order formula, a Gauss point
rectangular structures. The magnetic field distribution generated
matches with the point where the magnetic field is computed
by these straight structures can be computed by a close form for-
and the numerical integration diverges. In the case of a fifth-
mula given in [18]. The close form formula is convenient for
order formula, the numerical integral can be achieved without
these straight structures because they vary in height, width, and
significant difficulty. Converted to its isoparametric form, the
thickness at multiple places in the converter. Once the routing
Gauss integral is given by
is completed, any magnetic field analysis can be performed on
 
n 
n 
n the PCB. The basic software structure for using the models is
f (x, y, z)dV ≈ wi wj wk f (ξi , ηj , ψk ) |Jm | shown in Fig. 7. All the PEM models are included in the library.
V i=1 j =1 k =1 The model definition block contains the description of the PEMs
(5) used in the PCB and their orientations in space using homoge-
where f (ξi , ηj , ψk ) is the isoparametric transformation of (4), nous coordinate transformation [19]. This block also includes
wi,j,k are the weighting values and |Jm | is the Jacobian matrix straight structure descriptions and orientations. For example, the
determinant [10]. This isoparametric integration is very useful model illustrated in Fig. 8(c) contains seven PEMs and seven
for solving (4) because the element geometry does not affect straight structures. To complete the modeling of the PCB, one
the integration bounds. Moreover, the current densities for each must feed Tables I and II by extracting the coordinates with
element are directly extracted from the node potentials produced the routing tool. A script then calls the definition block with
by the first FEA. the structure’s currents (i1 , i2 , . . . , in ) and the point where the
BLANCHETTE AND AL-HADDAD: SOLVING EMI-RELATED PROBLEMS FOR RELIABLE HIGH-POWER CONVERTERS DESIGN 223

Fig. 8. (a) Three-phase bidirectional switching cell. (b) Printed circuit layout. (c) Decomposition in PEMs.

TABLE I IV. APPLICATION OF PEM TO A BIDIRECTIONAL MATRIX


EXAMPLE OF PEM TABLE DESCRIBING A PCB MODEL
CONVERTER CELL
Precomputed electromagnetic models can be used to design
both complex and simple converters. Of course, this powerful
method is best illustrated by designing a complex converter. The
selected application is the electromagnetic behavior analysis of
TABLE II a bidirectional three-phase switching cell used in both direct and
EXAMPLE OF STRAIGHT STRUCTURES TABLE DESCRIBING A PCB MODEL
indirect topologies of a matrix converter [13]–[17]. The electri-
cal diagram of this cell is shown in Fig. 8(a). The cell contains
12 power semiconductors: six power MOSFET (Q7 − Q12 ) and
six power SiC diodes (D1 − D6 ). In this application, the cell
currents are measured by three precision sensing resistors of
10 mΩ (R19 − R21 ), as shown in Fig. 8(a). The capacitors Cab ,
Cbc , and Cca are used to minimize the leakage inductances be-
magnetic field must be computed (px , py , pz ). The model defi- tween the semiconductors and the converter input filter. Since
nition block then sends the field components (Bx , By , Bz ) back the voltage across the sensing resistor is measured by a high-
to the script. The application of this approach is illustrated with speed differential amplifier, there is no common mass for all
an example in the next section. measurements. In many applications, a linear optically isolated
224 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 25, NO. 1, JANUARY 2010

Fig. 9. Fully isolated high-speed ADC block diagram.

amplifier is used to sense this differential voltage and to con-


nect all measured voltages to a common reference voltage. An
analog-to-digital converter (ADC) is connected to this common
reference voltage to read the amplifier outputs. The main prob-
lems with this approach are the amplifier’s limited bandwidth, Fig. 10. Covered surface used to compute the mutual inductance (in dark).
and the phase shift introduced by this isolated amplifier. These
problems preclude the use of this approach for high-speed appli-
cations such as power converters. In the proposed application,
to overcome the problems introduced by the isolated amplifier,
the ADC is directly integrated into the converter. Instead of
isolating the measured signals, the communication signals be-
tween the ADC and the main controller are isolated, as shown
Fig. 11. One-turn loop used to collect the magnetic field.
in Fig. 9.
Therefore, it is possible to acquire all the cell currents at the
same time without any phase shift. The main problem with this cuit must be studied. This exercise must then be repeated with
approach is communication reliability. Since the communication another mesh in the low-power circuit, and so on. For the cell
signals handled by the FPGA operate at 3.3 V logic with very shown in Fig. 8(c), there are two different meshes shown in
high impedance, and travel through the converter close-by to Fig 10 (meshes 1 and 2). To determine if the field orientation
high ∂i/∂t power trace signals, the reliability of this approach technique described previously produces expected results, the
is low if the converter magnetic field is not well controlled. It is mutual inductance between the test loop and both meshes must
possible to overcome this problem by using the magnetic field be studied. This is outlined in the next section.
orientation technique described in [18]. By using simple design
rules, this technique enables efficient control of the magnetic B. Field Analysis
field orientation in the converter for creating an oriented field
plane where the coupling between high- and low-power circuits As discussed in the previous section, mutual inductances be-
is very low. In the presented application, the oriented magnetic tween high- and low-power circuits are the source of radiated
field plane is used to insert the ADC board directly into the power EMI problems. In order to prove the usefulness of PEMs, mu-
converter, in close proximity to high time-varying magnetic tual inductances have been computed along the surface shown in
fields. The PEMs are used to study the results acquired through black in Fig. 10 and verified experimentally. The probe used to
the use of this approach. The experimental physical assembly collect the magnetic flux is depicted in Fig. 11. The experimen-
of the cell used in this application is shown in Fig. 8(b). The tal mutual inductance is found by dividing the voltage induced
magnetic-field-oriented planes are also identified in Fig. 8(b). in the probe by the time-varying current generated by MOS
switching. Fig. 12 illustrates the computed and experimental
mutual inductances between the probe, and meshes 1 and 2 (il-
A. Verification of Field-Oriented Plane Effectiveness by Mutual lustrated in Fig. 10) indicating good agreement between the two.
Inductance Computation
In Fig. 12, the mutual inductance is computed by a fifth-order
Mutual inductances computation is the natural approach used Gaussian quadrature integral [10] at 43 different positions be-
to study the interactions between two electrical circuits. Small tween −0.011 and 0.011 m. This computation is conducted in
mutual inductances between low- and high-power circuits in- 4.8 s on an AMD Phenom quad-core CPU clocked at 2.29 GHz.
dicate that good electromagnetic compatibility is achieved. In Very large mutual inductance asymmetry is illustrated in Fig. 12.
order to ensure that this condition exists for any converter op- This is due to sensing resistor structures that are slimmer than
erating point, the mutual inductances between a specific mesh PCB traces (producing higher current densities) and vertical
in the low-power circuit and all meshes in the high-power cir- positioning (0.0019 m above the PCB surface). This structure
BLANCHETTE AND AL-HADDAD: SOLVING EMI-RELATED PROBLEMS FOR RELIABLE HIGH-POWER CONVERTERS DESIGN 225

Fig. 13. Magnetic field distribution at z = 0.002 m using mesh 1 shown in


Fig. 10. The field is normalized from 0 to 100 (100 = 104.2 µT).

Fig. 12. Experimental and computed mutual inductances for meshes 1 and 2
illustrated in Fig. 10.

produces a flux misalignment varying with the height relative to


the PCB surface. This misalignment can be understood by plot-
ting the magnetic field intensity at different heights. In Fig. 13,
the field is plotted at z = 0.002 m. A zero field line, centered
at x = 0 m, is illustrated showing a good field symmetry be-
tween each side of the cell. A weak misalignment is visible at
z = 0.01 m, as shown in Fig. 14. The zero field line is squeezed
out of the center of the cell due to the magnetic field generated Fig. 14. Magnetic field distribution at z = 0.01 m using mesh 1 shown in
by the sensing resistor. Finally, the field computed at z = 0.02 Fig. 10. The field is normalized from 0 to 100 (100 = 20.83 µT).
m, shown in Fig. 15, presents strong asymmetry relative to the
center of the cell. This proves that the field-oriented plane ap-
proach may not be working well. However, it is important to
take into account that the maximal field intensity at z = 0.002
m is 5 times greater than the field at z = 0.01 m and 18.7 times
greater than the field at z = 0.02 m. Consequently, the field col-
lected by a circuit inserted in the field-oriented plane remains
very weak. However, Figs. 13–15 explain the mutual inductance
asymmetry shown in Fig. 12.

C. Fully Integrated Cell Performance


The presence of the field-oriented plane is verified by insert-
ing the high-speed, fully isolated ADC into the power bidirec-
tional cell, as shown in Fig. 16. The 10-bit ADC is clocked at
12.5 MHz, for a total conversion time of 1.5 µs. The first six
positive edges of the clock are used to sample the voltage given
by the differential amplifier. The next ten positive edges of the
clock are used to transmit the data. The theoretical waveforms
are shown in Fig. 17. In this example, the maximal current slope Fig. 15. Magnetic field distribution at z = 0.02 m using mesh 1 shown in
at the power MOSFET turn-OFF is 350 A/µs. This slope generates Fig. 10. The field is normalized from 0 to 100 (100 = 5.57 µT).
226 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 25, NO. 1, JANUARY 2010

signal must be kept low during the conversion. The EMI signal
is the voltage induced in the one-turn probe, which is inserted
in the power converter during the power MOSFET switching. The
DATA signal is the analog-to-digital conversion result. By com-
paring samples 1, 2, and 3 in Fig. 18, it can be seen that the
turn-OFF of the MOSFET does not affect the communication link
between the FPGA and the ADC. Neither signal (CS and DATA)
is affected by the power MOSFET switching.

V. CONCLUSION
Using PEMs is a quick and precise approach for predict-
ing magnetic field distribution in power converters. This new
drag-and-drop technology does not require a special expertise
in electromagnetics or finite elements, enabling more design-
ers to perform EMI analysis. One application was presented
in this paper; the introduction of a low-voltage high-speed iso-
Fig. 16. ADC inserted in the central field-oriented area.
lated ADC directly into a bidirectional power three-phase cell
(used in matrix converter applications) in close proximity to
high time-varying currents. Through the use of a proper field
orientation, the coupling between high- and low-power control
circuits is avoided, thereby increasing converter reliability. In
this application, the precomputed models were used to compute
the field distribution and confirm the expected field orientation.

REFERENCES
[1] E. C. W. de Jong, J. A. Ferreira, and P. Bauer, “Toward the next level of
Fig. 17. Theoretical analog-to-digital waveforms. PCB usage in power electronic converters,” IEEE Trans. Power Electron.,
vol. 23, no. 6, pp. 3151–2163, Nov. 2008.
[2] P. Kong, S. Wang, and F. C. Lee, “Common-mode EMI study and reduction
technique for the interleaved multichannel PFC converter,” IEEE Trans.
Power Electron., vol. 23, no. 5, pp. 2576–2584, Sep. 2008.
[3] H. Akagi and T. Shimizu, “Attenuation of conducted EMI emissions from
an inverter-driven motor,” IEEE Trans. Power Electron., vol. 23, no. 1,
pp. 282–290, Jan. 2008.
[4] D. Gonzalez, J. Balcells, A. Santolaria, J.-C. L. Bunetel, J. Cago,
D. Magnon, and S. Brehaut, “Conducted EMI reduction in power con-
verters by means of periodic switching frequency modulation,” IEEE
Trans. Power Electron., vol. 22, no. 6, pp. 2271–2281, Nov. 2007.
[5] F. Mihalic and D. Kos, “Reduced conductive EMI in switched-mode DC–
DC power converters without EMI filters: PWM versus randomize PWM,”
IEEE Trans. Power Electron., vol. 21, no. 6, pp. 1783–1794, Nov. 2006.
[6] M. Jin, M. Weiming, P. Qijun, K. Jun, Z. Lei, and Z. Zhihua, “Identifi-
cation of essential coupling path models for conducted EMI prediction in
switching power converters,” IEEE Trans. Power Electron., vol. 21, no. 6,
pp. 1795–1803, Nov. 2006.
[7] F. Costa and D. Magnon, “Graphical analysis of the spectra of EMI
sources in power electronics,” IEEE Trans. Power Electron., vol. 20,
no. 6, pp. 1491–1498, Nov. 2005.
[8] E. Boridy, Électromagnétisme : Théorie et Application. Montreal, QC,
Canada: Les Presses de l’Université du Québec, 1990.
[9] N. N. Rao, Elements of Engineering Electromagnetics. Englewood
Cliffs, NJ: Prentice-Hall, 2000.
[10] T. R. Chandrupatla and A. D. Belegundu, Introduction to Finite Elements
in Engineering. Englewood Cliffs, NJ: Prentice-Hall, 1997.
Fig. 18. Experimental high-speed ADC waveform. [11] J. Bastos and N. Sadowski, Electromagnetic Modeling by Finite Element
Methods. New York: Marcel Dekker, 2003.
[12] H. F. Blanchette and K. Al Haddad, “Leakage inductance calculation for
high power density converters applications,” in Proc. 33rd Annu. Conf.
a very large time-varying magnetic field in the converter, while IEEE Ind. Electron. Soc. (IECON 2007), Nov., pp. 1402–1407.
the induced voltage in high-impedance control circuits can eas- [13] R. Vargas, U. Ammann, and J. Rodriguez, “Predictive approach to increase
ily climb into the range of volts, if they are not well protected. efficiency and reduce switching losses on matrix converters,” IEEE Trans.
Power Electron., vol. 24, no. 4, pp. 894–902, Apr. 2009.
The experimental waveforms given by the isolated ADC are [14] M. Jussila and H. Tuusa, “Comparison of simple control strategies of
shown in Fig. 18. The chip select waveform (CS) indicates the space-vector modulated indirect matrix converter under distorted,” IEEE
beginning of conversion when it changes from high to low. This Trans. Power Electron., vol. 22, no. 1, pp. 139–148, Jan. 2007.
BLANCHETTE AND AL-HADDAD: SOLVING EMI-RELATED PROBLEMS FOR RELIABLE HIGH-POWER CONVERTERS DESIGN 227

[15] I. Sato, J.-I. Itoh, H. Ohguchi, A. Odaka, and H. Min, “An improvement Kamal Al-Haddad (S’82–M’88–SM’92–FM’07)
method of matrix converter drives under input voltage disturbances,” IEEE received the B.Sc.A. and the M.Sc.A. degrees from
Trans. Power Electron., vol. 22, no. 1, pp. 132–138, Jan. 2007. the University of Québec à Trois-Rivières, Trois-
[16] Y.-D. Yoon and S.-K. Su, “Carrier-based modulation technique for matrix Rivières, QC, Canada, and the Ph.D. degree from the
converter,” IEEE Trans. Power Electron., vol. 21, no. 6, pp. 1691–1703, Institut National Polytechnique, Toulouse, France, in
Nov. 2006. 1982, 1984, and 1988, respectively.
[17] P. W. Wheeler, J. Rodriguez, J. C. Clare, and L. Weinstein, “Matrix con- From June 1987 to June 1990, he was a Professor
verters: A technology review,” IEEE Trans. Ind. Electron., vol. 49, no. 2, at the Engineering Department, Université du Québec
pp. 276–288, Apr. 2002. à Trois-Rivières. In June 1990, he joined the teaching
[18] H. F. Blanchette and K. Al Haddad, “An efficient approach to design staff as a Professor of the Electrical Engineering De-
discrete packaging of bidirectional resonant power switch for matrix partment, École de Technologie Supérieure (ETS),
converter applications,” IEEE Trans. Power Electron., vol. 23, no. 4, Montreal, QC, Canada. Since 2002, he has been the holder of Canada Research
pp. 2195–2200, Jul. 2008. Chair in Electric Energy Conversion and Power Electronics. From 1992 until
[19] J. J. Craig, Introduction to Robotics: Mechanics and Control. 2003, he has been the Director of graduate study programs at the ETS. His cur-
Englewood Cliffs, NJ: Prentice-Hall, 2005. rent research interest include highly efficient static power converters, harmonics
and reactive power control using advanced active/hybrid filters, switch mode and
resonant converters, including the modeling, control, and development of pro-
totypes, for various industrial applications in electric traction, power supply for
drives, telecommunication etc. He is a Consultant and has established very solid
link with many Canadian industries where he was engaged in the field of power
electronics, electric transportation, aeronautics, and telecommunications. He is
Handy Fortin Blanchette (S’07) received the B.Eng. the Chief of ETS–Bombardier Transportation North America Division, a joint
and the M.Eng. degrees in electrical engineering, in industrial research laboratory on electric traction system and power electronics.
2001 and 2003, respectively, from the École de Tech- He has coauthored more than 300 TRANSACTIONS and conference papers.
nologie Supérieure (ETS), Montreal, QC, Canada, Dr. Al-Haddad is a Fellow of the Canadian Academy of Engineering and
where he is currently working toward the Ph.D. de- a Life Member of the Circle of Excellence of the University of Quebec. He
gree. received the outstanding researcher award and the medal of excellence from
From 1994 to 1997, he was engaged in indus- ETS in 2000 and 2009, respectively. He is very active in the IEEE Industrial
trial automation. From 1998 to 2000, he was with Electronics Society, where he is the Vice President Publications, and an Asso-
the Bombardier Transport–ETS Research Labora- ciate Editor of the IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS.
tory, Montreal, where he was involved in a high-
power traction system. From 2001 to 2003, he was
involved in the development of an electrical drive library in Simulink (MAT-
LAB) environment. In 2007, he was with OPAL-RT group, where he was in-
volved in power electronics real-time simulation. His current research interests
include power semiconductor modeling and thermal analysis.