Content uploaded by Maxime Dubois
Author content
All content in this area was uploaded by Maxime Dubois on Apr 20, 2018
Content may be subject to copyright.
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
Adrien Letellier, Student Member, IEEE, Maxime Dubois, Member, IEEE, João Pedro F. Trovão, Senior
Member, IEEE, Hassan Maher, Member, IEEE,
Abstract—This paper is concerned with the
determination of parasitic inductance values in very fast
switching power devices. To keep improving today’s power
converters, new technologies are studied which exhibit very
low switching times. The wide band gap semi-conductors
are among the key aspects of these improvements. Thanks
to their internal properties, they allow very fast di/dt and
dv/dt with very small footprint. Stray loop inductance needs
to be kept low, as it creates high peak voltage upon
switching of a transistor with fast di/dt. In particular, the
stray inductance value with respect to the loop size and
geometry needs to be calculated accurately at the design
stage of the power converters. This paper analyzes three
loop geometries and studies one with minimized stray
inductance and optimal current distribution. An analytical
method is proposed, which uses Biot-Savart law for an
accurate analytical estimation of the magnetic field intensity
in the selected geometry, leading to inductance calculation.
A comparison between the classical two-plate inductance
estimation formula and the proposed stray inductance
estimation is presented, proving more accurate value with
the method proposed in the paper. Finally, an experiment
has validated the new inductance estimation formula.
Index Terms—Power Electronics, Fast Switching,
Modeling, GaN, Design of Power Converters.
I. INTRODUCTION
he evolution of power electronics is tightly correlated to
the improvements in the semiconductor technology, the
use of new material and fabrication process, which all
look for higher compacity, higher efficiency and lower cost [1].
Wide Band Gap (WBG) components like GaN semiconductors
offer the possibility of switching currents in the range of 10 A
and voltages up to 650 V with 1 to 2 ns rise times while keeping
the cost low. These very short switching times offer the
possibility of decreased switching losses for MHz-range
switching frequencies [2].
In many applications, the switching topology makes proper
use of a low-side/high-side half-bridge configuration as the one
This work was supported by the Natural Sciences and Engineering Research
Council of Canada.
Adrien Letellier, Maxime Dubois and João Pedro F. Trovão are with the
Department of Electrical Engineering and Computer Engineering, Université
de Sherbrooke, Sherbrooke, QC, J1K 2R1 Canada (e-mails:
depicted in Fig. 1. As the switching current must flow through
copper material on a geometry (printed circuit board (PCB)
tracks or wires) characterized by a loop area A and a loop length
l, joule losses through the PCB tracks and inductance due to the
magnetic field created by the loop are inevitable. These two
behaviors can be modeled by lump resistance and inductance.
Thus, the inductance Lpcb and resistance Rpcb forming the power
loop, which also includes the half-bridge and the DC voltage
source as shown Fig. 1.
More specifically, Lpcb, Rpcb, and the switches internal
capacitors and packaging inductance LQ1,Q2 are the parameters
causing unwanted behaviors like overvoltage, ringing [3], [4],
parasitic losses and ultimately breakdown [5]. This
phenomenon is illustrated in Fig. 2 where a LTSpice®
simulation is conducted with a half-bridge circuit made of two
very fast GaN transistors. Two values of PCB inductance
(9.42 nH and 1.35 nH) have been simulated and reported. The
first one shows a peak voltage/bus voltage ratio of 2.2, where
the component maximum voltage rating of 100 V would be
exceeded with an important overvoltage and ringing which
would lead to the component breakdown. The second one
Adrien.Letellier@USherbrooke.ca, Maxime.Dubois@USherbrooke.ca;
Joao.Trovao@USherbrooke.ca)
Hassan Maher is with the Laboratoire Nanotechnologies Nanosystèmes,
CNRS UMI-3463, Institut Interdisciplinaire d’Innovation Technologique,
Université de Sherbrooke, Sherbrooke, QC J1K OA5, Canada (e-mail:
Hassan.Maher@USherbrooke.ca).
Calculation of PCB Power Loop Stray
Inductance in GaN or High di/dt Applications
T
Fig. 1. Power arm with an inductive load, Lpcb and Rpcb
are the conductor parasitic elements.
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
displays a peak voltage/bus voltage ratio of 1.3, with lower
overvoltage and ringing, keeping the switch under its maximum
voltage value thanks to a lower stray inductance.
In past literature, a few methods have been developed to
control and mitigate these parasitic elements [6], [7]. Some of
them aim at minimizing the current loop area and the current
path length. Other designs intend to minimize the generated
flux by using nullifying methods. Even though results were
obtained for reducing the overvoltage and losses [8], difficulty
remains for evaluating the power loop stray inductance without
numerical simulations [9], [10], [11]. The present work aims to
be complementary with the previous works by providing a way
to easily calculate the PCB stray inductance for a defined loop
shape without using Finite Element Analysis (FEA) or
experimental tests.
Going from the current loop physical dimensions to the
values of Lpcb and Rpcb, which in turn will lead to the overvoltage
across Q1 and Q2 without the use of a numerical simulation is a
difficult task.
The lump inductance of a loop formed by two parallel plates
(e.g. the top and bottom copper plates of a PCB) can be obtained
by using (1) which serves as a useful approximation in many
cases [12].
0pcb e
Ll
w
(1)
where e is the distance between the upper and lower conducting
plates, w is the width of these conducting plates and l is the
length of the conducting planes.
As will be shown in the paper, (1) can be used under certain
assumptions. However, the latter becomes inaccurate for low
values of the w/e ratios, which are characteristic of the
geometrical dimensions of very small switches used with GaN
semiconducting material.
The present paper will discuss the PCB positioning of GaN
transistors with a closer investigation at three possible layouts
for the copper tracks forming the current loop from the
decoupling capacitor to the half-bridge. The best suited
geometry will be identified and a mathematical development
using Biot-Savart law will be proposed for estimating the loop
inductance Lpcb with accuracy. The result is validated with
experimental and numerical simulation tools. A simplified
expression, giving a similar result within a known error range
is given and compared to the classical approximation equation.
The paper shows a comparison between different topologies.
The one most adapted to the current problem of fast switching
is then analytically studied. During a converter design
procedure, choices must be made for the power components and
thus for the power loop dimensions. Many software like Q3D
Extractor or Magnet can be used to model the conductors and
numerically find the inductor value. The current work aims to
provide a solution giving the inductance depending on the
power switch size without the need for finite-element
simulation. It then becomes possible to test many
configurations without the need for 3D modelling and
simulation. To obtain high accuracy, a FEA software needs to
work by using a fine mesh. For very high frequencies, the skin
effect is very thin, implying a very fine mesh. This is time
consuming and adds up to the total time if you want to test many
configurations.
A rough estimation on the time needed to test one
configuration using 3D FEA is of one hour. Of course, it
depends on the user’s skills and computer speed. In any case,
the FEA simulation is much longer than using analytical
expressions where the result is immediate. Moreover, the
simplified equation allows it to be integrated in scripts while the
co-simulation can be complicated and time consuming when
used.
The paper is organized as follows: section II presents the
three PCB layouts investigated for reducing Lpcb. Section III
presents the analytical derivation of Lpcb for the case of the two-
plate current loop. FEA validation of the obtained expression is
presented in section III. Section IV proposes a novel expression
for estimating the loop inductance. Section V presents a
practical demonstration of the switching behavior with the loop
geometry selected. Finally, section VI gives a conclusion to the
paper.
II. ANALYSIS OF THREE POWER LOOP CONFIGURATIONS
GaN-based power transistors exhibit very high-power
density. Thanks to the material properties, high voltages and
currents can be switched in the ns range, while keeping the chip
(a)
(b)
Fig. 2. Switching waveform of the component EPC2022
GaN transistor simulated with LTSpice®, (a) is for
Lpcb = 9.42 nH, (b) is for Lpcb = 1.35 nH. Input DC voltage
VDC = 50 V, load current Iload = 25 A.
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
extremely small. Fig. 3 shows the size difference between two
transistors package for a given voltage rating. A standard
silicon (Si) D²PAK MOSFET package is shown, with a much
larger footprint than the GaN HEMT (High Electron Mobility
Transistor), which enables low parasitic inductance.
Apart from the power loop PCB inductance, the transistor
package leads are themselves a source of parasitic inductance,
adding all of them together is the so called stray inductance.
When comparing the packages, numerous differences appear,
which have an impact on the circuit stray inductance. The
standard D²PAK exhibits a package inductance varying
between 5 nH and 7 nH [13], [14] while the package proposed
by the GaN transistor manufacturer EPC is given in the 0.1 nH
range for a size of 3.5x1.4 mm. This low package stray
inductance is key in minimizing the overall power loop
inductance.
The direct consequence of an important reduction in size is
the reduction of the width of the copper conductor and the
related magnetic flux produced by the power loop when
carrying the rated current. Hence, the underlying assumption
behind (1) is no longer valid. In this section, the three loops
studied with the assumption of very small footprint will use the
results from FEA to properly estimate the corresponding loop
stray inductance.
The following subsections will study the various loop
geometries proposed in past literature, such as the U-shape, the
Top-Bottom and the Top-Inner power loops, with the objective
of reducing the magnetic flux area created by the loop. Three
main topologies exist and exhibit different results [15].
Fig. 3. Comparison of a D²PAK standard packaging to
EPC2010C, a 200 V, 22 A transistor. (Source:
www.http://epc-co.com).
A. U-shape Power Path
The U-shape power path has been discussed in a number of
publications [6], [16] and is illustrated by Fig. 4(a). This
concept aims at making the current loop area as small as
possible and at the same time keeping it on one side of the PCB.
This solution is easy to implement and has the cost and
fabrication advantages of a single-sided PCB. The resistive path
exhibits low resistance value Rpcb. However, the loop parasitic
inductance Lpcb will be a concern in the case of high di/dt
switching. This is mainly due to the center magnetic flux
created upon the circulation of a current in the power loop. The
current density distribution J and field H are computed with
FEA and displayed in Fig. 4 as well.
Fig. 4 shows another aspect that needs to be carefully
investigated when dealing with Rpcb and Lpcb: the skin effect
behaviour observed in the power loop. Fast switching
transistors like GaN HEMT or Ultra-Fast MOSFETs are
intended to be used in Switch-Mode Power Supplies (SMPS)
with 1 MHz-range switching frequency. Moreover, the ringing
phenomenon observed across the transistor at turn-off involves
voltage and current oscillations in the 100 MHz range, as it was
shown in Fig. 2.
As the effect of the stray inductance and Lpcb is to affect the
circuit during such very fast transition periods, our analysis
must consider the skin effect behavior.
As shown in Fig. 4 for a 1 MHz frequency, the current
density is higher in the inner part of the U, where the magnetic
field is also stronger. The current is unevenly distributed at
1 MHz, making a poor use of the copper material when carrying
high current values. This current distribution will also lead to
the switches heating unevenly. The resulting values are a loop
inductance LPCB = 9.4 nH and a loop resistance of RPCB =
2.2 mΩ.
(a)
(b)
(c)
Fig. 4. (a) U-Shaped topology (b) |H| field produced by
the circulating current (c) current density J. Current
amplitude 10 A, frequency 1 MHz. The loop is 19x23 mm.
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
B. Top-Bottom Power Path
This method consists in mirroring the current path located on
the top layer of a PCB by implementing a symmetrical current
path on the PCB bottom layer. This forms a symmetry plan
orthogonal to the switches leading to a reduced area for the
circulation of the magnetic flux. This topology is illustrated in
Fig. 5(a). Since two layers are involved, the inter-layer
connections must be considered with extra care. The standard
manufacturing process does not allow for an exact control of
the vias manufacturing. A different method is chosen for this
study; a slot is designed at each end of the traces, instead of vias
and then filled with soldering tin.
This way it can be considered as a metal conductor where all
the parameters are known. The flux surface area is diminished,
leading to a lower inductance. The FEA results are shown Fig.
5(b), (c) where the current density appears well distributed,
lowering its maximal value compared to the U-shaped
topology.
In the geometry of Fig. 5, with a PCB thickness of
e = 1.6 mm the obtained inductance is Lpcb = 3.56 nH and the
1 MHz resistance is Rpcb = 1.6 mΩ. This result shows a great
improvement compared to the U-shaped method, with an
inductance reduction of 62% and a resistance reduction of 27%.
(a)
(b)
(c)
Fig. 5, (a) Top-bottom topology with a 1.6 mm PCB (b)
|H| field produced by the circulating current (c) current
density J. Current amplitude 10 A, frequency 1MHz. The
loop is 7.5x14.5 mm
C. Internal Layer Power Path
This method is similar to the Top-Bottom method, with the
exception of using a multi-layer PCB, in which an internal layer
is used for the current return instead of the PCB bottom layer.
This allows further reducing the loop area. The topology is
shown in Fig. 6(a) and the FEA results in Fig. 6(b), (c). As for
the Top-Bottom topology, the current density is well distributed
along the copper width with an even smaller flux loop area. The
results presented here are for a 1.6 mm PCB thickness with 2
internal layers equally spaced. Thus, the spacing between the
two flat conductors is one third of the PBC thickness, that is
e = 0.53 mm.
The FEA calculation gives an inductance of Lpcb = 1.35 nH
and a 1 MHz resistance of Rpcb = 0.95 mΩ. The stray
inductance is reduced by 62% and the resistance by 40%
compared to the previous method (section II.B.). The result can
be further improved by using a thinner PCB with more layers.
This configuration shows the lowest inductance value and an
evenly distributed current, allowing equal heat distribution.
These two parameters make it ideal for the present study.
(a)
(b)
(c)
Fig. 6. (a) Optimized topology using an internal layer,
(b) |H| field produced by the circulating current, (c) current
density J. Current amplitude 10 A, frequency 1 MHz. The
copper conductor is 7.5 mm wide and 19 mm long.
Spacing between the two conductor plates is 0.59 mm.
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
III. INTERNAL LAYER POWER PATH INDUCTANCE
CALCULATION
A. Modeling Methods
To determine the values of Lpcb, one of the most popular
method is to use FEA software where the PCB layout is
imported from the PCB design software. This is the method that
was used in section II. A faster method is to approximate it by
a broad side tracks inductance with the limits that the width
must be greatly higher than the distance between the
conductors. Then (1) is applied to the internal layer topology
displayed in Fig. 6.
All these methods are valid, but the cons are that either the
obtained value is approximated with an unknown error or a long
and fastidious simulation must be set and solved. To obtain a
better approximation that does not suffer the limits given
previously, a new analytical development for modeling the
value of Lpcb is presented in this paper. The analytical
development is based on the Maxwell’s equations and the Biot-
Savart law. Then a novel estimation formula is derived and
compared to the classical one.
B. Application of Maxwell and Biot-Savart Laws
Considering the skin and proximity effect, the considered
conductor height should be no more than the skin depth defined
by (2).
0
1
Cu
f
(2)
Then, the current density J is considered constant inside the
conductor height
and we will make the simplifying
assumption of h =
for the presented mathematical
development. Fig. 7 shows two cases: 1) copper thickness
superior to the skin depth and 2) for a thickness equal to the skin
depth. The second case allows us to consider J constant and will
be used for the mathematical development.
Where f is the frequency of the signal considered (typically
300 MHz during the ringing phenomenon), µ0 is the void
permittivity and
Cu is the conductivity of copper.
It is then possible to use Biot-Savart law [17] to retrieve the
H field and associated magnetic flux
for one conductor and
use (3) to obtain the value of the stray inductance Lpcb, based on
the current Ipcb and the number of turns n. In the present case,
Ipcb is assumed as a constant value, flowing in the thickness
determined by the skin depth, as detailed above, n = 1 and the
flux
is calculated in between the two conductors of Fig. 7.
The coordinates referential and axis used in this analysis are laid
out as described in Fig. 7.
pcb pcb
n
LI
(3)
The Biot-Savart law is used, in order to provide the H field
for any specified point depending on the current density and
geometry. Applied to our problem geometry we need to solve
(4) giving the field at any point (x0, z0) around the conductor.
0
0
03
0
4s
Vconductor
J r r r
B r dxdydz
rr
(4)
The coordinates (x0, z0) are associated to vector r0, whereas
coordinates (x, z) are associated to vector r, which defines any
given point inside the conductor. The Biot-Savart law
formulated by (4) considers a surface current density vector js,
which has a constant value J and orientation over the entire
conductor volume Vconductor, that is, inside the conductor
between the boundaries -w/2 < x < w/2 and –h/2 < z < h/2.
Vector js is oriented parallel to the y-axis in the upper conductor
and anti-parallel to the y-axis in the bottom conductor. Given js
orientation, the magnetic field at any point will have an x-axis
component Hx and a z-axis component Hz, with no y-axis
component. As the parallel conductors extend in the y-axis from
y = -∞ to y = +∞, we can demonstrate (5):
3
2 2 2
0 0 0
22
00
1
2
dy
x x y y z z
x x z z
(5)
Inserting (5) into (4), the upper conductor will create a field
Hupper:
22 0
00 22
00
22
00
22 0
00 22
00
22
,2
,0
,2
wh
upperX wh
upperY
hw
upperZ hw
zz
J
H x z dzdx
z z x x
H x z
xx
J
H x z dxdz
z z x x
(6)
where the current density in the top and lower conductor J is
given by (7).
I
Jwh
(7)
(a)
(b)
Fig. 7. Skin depth effect and its consideration for the
analytical development. (a) is when the considered thickness
h > δ and (b) is for h ≤ δ.
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
The solution for the H field generated at each point (x0, z0) in
the space around the conductors is detailed in Appendix I,
giving (8) as a solution.
00
0
1
0222
00
0
1
0222
00
22
00
022
00
22
00
022
00
00
0
1
022
0
,
22
2 tan
424
22
2 tan 24
22
/ 2 ln 22
22
/ 2 ln 22
,
22
2 tan
42
upperX
upperZ
H x z
w h z
Jzh h z w x
w h z
zh h z w x
w x h z
wx w x h z
w x h z
wx w x h z
H x z
h w x
Jxw w x h
2
0
0
1
0222
00
22
00
022
00
22
00
022
00
4
22
2 tan 24
22
/ 2 ln 22
22
/ 2 ln 22
z
h w x
xw w x h z
w x h z
hz w x h z
w x h z
hz w x h z
(8)
Similarly, the lower conductor will create a field Hlower
22 0
00 22
00
22
00
22 0
00 22
00
22
,2
,0
,2
wh
e
lowerX wh
e
lowerY
hw
e
lowerZ hw
e
zz
J
H x z dzdx
z z x x
H x z
xx
J
H x z dxdz
z z x x
(9)
Which accepts (10) as a solution. Naturally, the total field is
the sum of (8) and (10) at any point.
The field generated by two conductors can now be calculated
analytically for each desired coordinate (x0, z0) with the
resulting H field displayed in Fig. 8, with the geometrical
parameters conductor width w = 7.5 mm, thickness h = 35 m,
spacing e = 0.42 mm. The calculation is done for a current I =
10 A. As a comparison, a 2-D FEA was also performed on the
exact same geometry with the results shown in Fig. 9. The
obtained field is the same in both distribution and intensity in
both analytical computation and FEA.
00
0
1
0222
00
0
1
0222
00
2
2
00
02
2
00
2
2
00
02
2
00
,
22
2 tan
424
2 3 2
2 3 tan 3 2 4
22
/ 2 ln 2 3 2
22
/ 2 ln 2 3 2
lowerX
H x z
w h z e
Jz e h h z e w x
w h z e
z e h h z e w x
w x h z e
wx w x h z e
w x h z e
wx w x h z e
00
0
1
022
2
00
0
1
022
2
00
2
2
00
02
2
00
2
2
00
02
2
00
,
22
2 tan
424
22
2 tan 24
22
/ 2 ln 22
2 3 2
3 / 2 ln 2 3 2
lowerZ
H x z
h w x
Jxw w x h z e h
h w x
xw w x h z e h
w x h z e
h z e w x h z e
w x h z e
h z e w x h z e
(10)
As expected, this result shows that the field is concentrated
between the two conductors, due to the currents going in
opposite ways on the top and bottom conductors. The field
becomes nearly null outside the inner space and concentrated
inside. The analytical formulations of (8) and (10) will be of
great value for considering the end effect, which show some
variation in the field intensity and will require careful
consideration for low geometrical ratios w/e. In Fig. 10, the H
field is plotted along the z axis. The difference between FEA
and analytical calculation is inferior to 1%.
Fig. 8. Analytical result for the H field computed with
(8) and (10). The dimensions are e = 0.42 mm,
w = 7.5 mm, h = 35 µm, I = 10 A.
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
When w >> e, the curve obtained in Fig. 10 can be
approximated by an external H field of 0 and an internal one of
I/w, which is the basis for the conventional inductance
expression of (1).
Within the cope of this paper, our aim is to obtain better
accuracy on the PCB inductance value, which will be useful in
the accurate estimation of the overvoltage on a transistor at turn-
off. To determine the inductance, the flux
between the two
plates must be calculated. This is done by integrating the
magnetic field over a surface S parallel to the plane (y, z) inside
the loop formed by the top layer, bottom layer and the two slots
shown Fig. 6.
The flux can be expressed by (11):
2
0 0 0 0
2
0, 0,
h
X upperX lowerX
h
Se
B dS l H z H z dz
(11)
Inserting (11) into (3) gives (12):
2
00 0 0
2
0, 0,
h
pcb upperX lowerX
he
l
L H z H z dz
I
(12)
Then the exact value of the inductance is obtained for the two
conductors perfectly aligned, with the same dimensions by
inserting (8) and (10) into (12), giving (13). The detail of the
inductance calculation is included in the Appendix.
2
21
0
22
2
2 1 1
22
22
22
22
2
tan
2
22
tan tan
22
22
log log
222
pcb lw
L he e e
wh w
ww
h h e h e h
ww
ww
e h e h
weh
ww
eh
(13)
Of course, the resulting inductance expression is a bit
cumbersome compared to the conventional (1). This issue is
addressed in the next section, with a more usable
approximation, and where accuracy is enhanced compared to
(1).
IV. FORMULA SIMPLIFICATION
A. Standard approximation limits
For small track width w, the H field along the z axis is
displayed in Fig. 11 (for w/e = 1 and h = 35 µm). The mean
value of |H| calculated with (8) and (10) between the two traces
is 12183 A/m while the traditional approximation using I/w will
give 23810 A/m, with an error of 95% on the field estimated
and inductance calculation.
In the application where GaN transistors are used with the
loop and dimensions highlighted in Fig. 6, the width is close to
the space between the conductors, (1) cannot be used. Fig. 12
shows the difference in inductance value calculated with (13)
compared to the value obtained with (1) for typical GaN
transistors footprint. Differences over 10% are obtained. The
conductor width associated to some GaN components and the
driver loop size are also indicated on Fig. 12. As presented in
Fig. 12, even if the approximation is suitable for wide copper
traces, it is not suited for the fast switching GaN technology. A
new estimation formula with a lower error and higher range of
use is then needed.
Fig. 9. FEA simulation of the H field distribution in 2-D
with Magnet®. The dimensions are e = 0.42 mm,
w = 7.5 mm, h = 35 µm, I = 10 A.
(a)
(b)
Fig. 10, H comparison between (a) analytical and (b)
FEA calculation along the z axis (for x=0). The dimensions
are e = 0.42 mm, w = 7.5 mm, h = 35 µm, I = 10 A.
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
Fig. 12. Difference between Lpcb/l calculated with (1) and
Lpcb/l calculated with (13). The error becomes significantly
higher than 10% for a ratio of w/e < 7. The indicated GaN
components width shows that their inductance cannot be
well estimated with (1). EPC2022, EPC2014 and GS61008
are GaN transistors available on the market.
B. Proposed Approximation
To overcome this limitation, (13) can be approximated with
(14), using the assumption that 50 mm > w > 0.25 mm and
2 mm > e > 100 µm. The approximation was conducted using
curve fitting numerical tools. The inductance per meter Lpcb/l
was first calculated for the previously given dimensions,
considering the standard value of h = 35µm. Then, using (1) as
a starting point, the custom equation (14) coefficients were
found thanks to the least square algorithm of Levenberg-
Marquardt [8].
024.0
1
1
0w
e
l
w
e
Lpcb
(14)
The resulting approximation shows a R² coefficient of 0.999,
indicating an extremely close result between the exact formula
and the estimation considering that a R² of 1 is a perfect
approximation. (14) introduces a correction factor to the
conventional expression (1) with the term within parentheses
with a maximum error of 6% observed in the studied range. Fig.
13. shows the difference between (14) and (13), the ranges for
the loop width w and spacing e are adapted to allow the
dimensions of different GaN components to be seen. The result
is greatly improved, and the resulting formula remains easy to
use. With this new approximation, the power loop inductance
and the gate loop inductance can be found with a known
maximal error allowing to evaluate the loop characteristics at
the design stage.
The comparison is conducted for the power loop of
components EPC2014, GS61008 and also to calculate the gate-
source driver loop. The result in Table 1 shows that the new
proposed equation allows for a better estimation of the
inductance for various size of tracks. In particular, for the gate-
source stray inductance, it becomes possible to estimate
standard (0.254 mm) tracks inductance. The error is reduced
from more than 300% with (1) to less than 4% with (14). The
deviation between the FEA 3D simulation and the
approximated value of (14) is within 10%. A similar
comparison is made with the top-bottom loop topology detailed
in II.B. The result is given Fig. 14. The flux is well concentrated
between the conductors, (1), (13) and (14) can then be applied
to this topology and the result is given in the table inside Fig.
14.
Table 1. Comparison between the FEA result and the
analytical formulas for different GaN HEMT components,
power loop and driver loop.
Component
EPC2014
power loop
GS61008
power loop
EPC2014
Gate drive
loop
Loop dimensions
(w x l x e)
1.7 x 9.5 x 1
(mm)
7 x 15.5 x 1
(mm)
0.254 x 7.2
x 1 (mm)
Lpcb with (13)
(exact expression)
4.67 nH
2.52 nH
7.82 nH
Lpcb with (14) (new
approximation), error
4.60 nH,
1.5%
2.50 nH,
0.8%
8.11 nH,
3.7%
Lpcb with (1) (classical
approximation), error
7.02 nH,
50%
2.78 nH,
10%
35.6 nH,
355%
Lpcb FEA result with
Magnet®,error
4.31 nH,
7.7 %
2.28 nH,
9.5%
8.81 nH,
13%
Fig. 11. |H| along the z axis, for x = 0 with a reduced size
of w = 0.42 mm and e = 0.42 mm, the traditional
approximation cannot be used in this case.
Fig. 13. Difference between Lpcb/l calculated with (14)
and Lpcb/l calculated with (13). The error is limited to 6%.
The indicated GaN components width shows that their
inductance can be well estimated with (14).
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
(a)
Component
EPC2014
Loop dimensions (w, l, e)
1.7 mm x 9.5 mm x 1.6 mm
FEA result with Magnet®
5.72 nH
Result of (1) and error
11.2 nH, 96%
Result of (13) and error
6.20 nH, 8.4%
Result of (14) and error
6.07 nH, 6.1%
(b)
Fig. 14. Simulation of the Top-Bottom for the component
EPC2014, (a) the FEA simulation result, (b) inductance
result and calculation for this topology.
It appears that the proposed formula (14) allows estimation
better estimation (6.1 % difference with FEA) if compared to
the classical approximation (1), which exhibits an error of 96%.
It shows the adequacy of the proposed solution to two
geometries (internal layer and Top-Bottom).
V. PRACTICAL VALIDATION
To validate both the analytical expression (13) of Lpcb and the
approximated (14), a laboratory test is conducted, using a GaN
HEMT with the prototype visible Fig. 15. Each important part
is shown and the inductances that appear are calculated with
(14). This prototype allowed for validation of the numerical
approach and to perform power measurement. The dimensions
and the small e value of 175µm have been chosen to allow the
smallest inductance, hence the smallest overshoot possible.
This prototype can switch currents on a ns scale and thus needs
a very low inductance in its power loop. We use a 1 mm-thick
PCB with 2 internal layers (e = 175 µm). The component used
is the GS61008p from GaNSystems (w = 7 mm). The power
path architecture uses the optimized topology presented in Fig.
6.
Fig. 15. Simulation of the Top-Bottom for the component
EPC2014, (a) the FEA simulation result, (b) inductance
result and calculation for this topology.
When switching the result of the low side rising voltage is
the one presented in Fig. 16.
A common way to measure the stray loop inductance is to
measure the ringing frequency using (15).
2
1
2
stray
oss ring
LCf
(15)
The PCB loop inductance is then obtained by subtracting the
two inductances of the transistor package LQ1 and LQ2. Using
the component technical document, Coss(50 V) = 280 pF. Then
the measured inductance is Lstray = 1.45 nH. Considering a
package inductance of 0.1 nH per transistor, similar to the one
given by the HEMT manufacturer EPC [13], and removing the
capacitors total inductance of 0.87 nH we can deduce the
Fig. 16. Low side turn-off voltage, an overvoltage and
ringing occur, due to the PCB power path inductance and the
internal component capacitance Coss
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
experimental value of the PCB loop inductance to be
Lpcb(measured) = 0.38 nH.
Using (14), the approximated theoretical value is 0.53 nH,
giving a difference of 150 pH between measurement and
theory. The obtained error comes from the lack of detailed
packaging inductance analysis from the manufacturer and some
3-D effects, which are not considered in the field calculation.
Fig. 17 shows the result of the obtained inductance
depending on the e/w ratio. It appears clear that the traditional
approximation introduces a high error with a ration over 0.2.
The gate loop values taken for the graph are the ones calculated
with the FEA method and shows a very close match with the
presented work, while the traditional calculation exhibits a very
high error.
Fig. 17. e/w ratio and the location of the experiment and a
gate loop inductance calculation
VI. CONCLUSION
This paper presented the most optimized power loop for high
speed switching devices like GaN power transistors. Many
studies use this topology and the value of the PCB inductance
is known through PCB numerical simulation or formula
approximation. These two tools have advantages and
drawbacks, such as time consumption or unknown error. The
paper presented a complete 2-D analysis of the magnetic field
obtained between two conducting plates, by using Biot-Savart
law and allowed to demonstrate that the conventional
inductance formulation is an approximation that can no longer
be used in certain form factors, especially with small
conductors. It was found that for PCB traces of width/spacing
inferior to 7, traditional expression (1) will overestimate the
inductance by more than 10%. With a ratio of 1, this
overestimation increases to 95%.
To improve the inductance estimation, a new formula is
proposed by (14). It remains easy to use and was tested for
numerous width and conductor spacing with a FEA software.
The validation through a practical experience shows the
correspondence between the real measurement and the new
approximation. This work intends to help the design of high
speed switching loops where all the stray inductances need to
be known at the design stage of the power converters.
APPENDIX I
The detailed calculation steps to solve (4) and obtain Hupper
and Hlower are presented here. The H field calculated is only for
one conductor, to obtain the total field at each point, the
solutions of both the top and the inner conductors must be
added.
22 0
00 22
00
22
00
22 0
00 22
00
22
,2
,0
,2
wh
upperX wh
upperY
hw
upperZ hw
zz
J
H x z dzdx
z z x x
H x z
xx
J
H x z dxdz
z z x x
(16)
Knowing that
ln u
uu
(17)
2222
0 0 0 0
2
2
2222
0 0 0 0
2
2
, ln
4
, ln
4
wh
upperX h
w
hw
upperZ w
h
J
H x z z z x x dx
J
H x z z z x x dz
(18)
222
0 0 0 0
2
22
00
222
0 0 0 0
2
22
00
, ln 4 2
4
ln 4 2
, ln 4 2
4
ln 4 2
w
upperX w
h
upperZ h
J
H x z x x h z
x x h z dx
J
H x z z z w x
z z w x dz
(19)
To calculate the complete integral, solving the following
equation is needed.
2
ln c x a b dx
(20)
2
2
2
2
ln
()
ln 2 ()
c x a b dx
c x a
a x c x a b dx
c x a b
(21)
2
2
2
ln
1
ln 2 2 ()
1
c x a b dx
a x c x a b dx dx
c x a
b
(22)
2
21
ln
ln 2 2 tan
c x a b dx
bc
a x c x a b x a x
cb
(23)
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
Applied to Hx with
0
xa
,
2
0
2zhb
,
4c
for the first logarithm term and
0
xa
,
2
0
2zhb
,
4c
for the second
logarithm term, it comes:
22 2
0 0 0 0
11
0 0 0 0 0
22 00
00 2
4 2 2 2
, ln 2 tan 2 tan
4 2 2
42
w
upperX
w
x x h z x x x x
J
H x z x x h z h z
h z h z
x x h z
(24)
00
11
0 0 0 0
22
2 2 2 2
0 0 0 0
2 2 2 2
0 0 0 0
00
2 2 2 2
0 0 0 0
2 2 2 2
, 2. 2 tan 2. 2 tan
82 4 2 4
2 2 2 2
2 ln 2 ln
2 2 2 2
upperX
w h z w h z
J
H x z z h z h
h z w x h z w x
w x h z w x h z
w x w x
w x h z w x h z
(25)
By applying the same methodology to HupperZ, it comes:
00
11
0 0 0 0
22
2 2 2 2
0 0 0 0
2 2 2 2
0 0 0 0
00
2 2 2 2
0 0 0 0
2 2 2 2
, 2. 2 tan 2. 2 tan
82 4 2 4
2 2 2 2
2 ln 2 ln
2 2 2 2
upperZ
h w x h w x
J
H x z x w x w
w x h z w x h z
w x h z w x h z
h z h z
w x h z w x h z
(26)
Then, the final field, including both conductors of the same dimensions with a spacing of e is:
00
11
0 0 0 0
22
2 2 2 2
0 0 0 0
00
11
00
22
2 2 2 2
0 0 0 0
2 2 2 2
, 2 2 tan 2 2 tan
82 4 2 4
2 2 2 2 2 3 2
2 2 2 tan 2 2 3 2 tan
2 2 4 2 3 2 4
X
w h z w h z
J
H x z h z h z
h z w x h z w x
w e h z w e h z
e h z e h z
e h z w x e h z w x
2 2 2 2
0 0 0 0
02 2 2 2
0 0 0 0
2 2 2 2
0 0 0 0
02 2 2 2
0 0 0 0
00
(w 2x ) +(2e h 2z ) (w-2x ) +(h 2z )
2 ln (w 2x ) +(2e 3h 2z ) (w 2x ) +(h 2z )
(w 2x ) +(2e h 2z ) (w 2x ) +(h 2z )
2 ln (w 2x ) +(2e 3h 2z ) (w 2x ) +(h 2z )
,8
Z
wx
wx
J
H x z
00
11
022
22
0 0 0 0 0
00
11
022
22
0 0 0 0 0
0
2 2 2 2
2 2 tan tan
2 4 2 3 2 2
2 2 2 2
2 2 tan tan
2 4 2 3 2 2
(w-
2 ln
w h z h w x
wx h z w x w x h z e h z e
h w x h w x
wx w x h z w x h z e h z e
hz
2 2 2 2
0 0 0 0
0
2 2 2 2
0 0 0 0
2 2 2 2
0 0 0 0
00
2 2 2
0 0 0
2 x ) +(h-2 z ) (w 2 x ) +(h 2 z )
2 ln
(w 2 x ) +(h 2 z ) (w 2 x ) +(h 2 z )
(w-2 x ) +(2e h 2 z ) (w 2 x ) +(2e 3h 2 z )
2 2 ln 3 2 2 ln
(w 2 x ) +(2e h 2 z ) (w 2 x ) +
hz
h e z h ew z
2
0
(2e 3h 2 z )
(27)
To calculate the inductance, the flux between the two plates must be calculated. The surface is considered parallel to the plane
(y, z) and inside the loop formed by the top layer, bottom layer and the two slots as shown Fig. 6.
-h 2
00
h
-2
. 0,
BX
Se
H dS l H z dz
(28)
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
h2
00
h
-2
0,
PCB X
e
l
L H z dz
I
(29)
Then the exact value of the inductance for the two conductors perfectly aligned, with the same dimensions is:
22
2 1 2 1
0
22
22
2
2122
22
22
tan tan
22
222
tan log log
2222
pcb lww
L he e e h h
wh w w
ww
e h e h
w
w
e h e h e h
www
eh
(30)
REFERENCES
[1] J. Millan, P. Godignon, X. Perpina, A. Perez-Tomas, and J. Rebollo,
“A Survey of Wide Bandgap Power Semiconductor Devices,” IEEE
Transactions on Power Electronics, vol. 29, no. 5, pp. 2155–2163,
May 2014.
[2] A. Letellier, M. R. M. R. M. R. Dubois, J. P. J. P. J. P. Trovao, and
H. Maher, “Gallium Nitride Semiconductors in Power Electronics for
Electric Vehicles: Advantages and Challenges,” in 2015 IEEE
Vehicle Power and Propulsion Conference (VPPC), 2015, pp. 1–6.
[3] J. Noppakunkajorn, D. Han, and B. Sarlioglu, “Analysis of High-
Speed PCB With SiC Devices by Investigating Turn-Off Overvoltage
and Interconnection Inductance Influence,” IEEE Transactions on
Transportation Electrification, vol. 1, no. 2, pp. 118–125, Aug. 2015.
[4] A. Letellier, M. R. Dubois, J. P. Trovao, and H. Maher, “Overvoltage
Reduction with Enhanced Snubber Design for GaN-Based Electric
Vehicle Drive,” in 2017 IEEE Vehicle Power and Propulsion
Conference (VPPC), 2017, pp. 1–6.
[5] D. Reusch and J. Strydom, “Understanding the Effect of PCB Layout
on Circuit Performance in a High-Frequency Gallium-Nitride-Based
Point of Load Converter,” IEEE Transactions on Power Electronics,
vol. 29, no. 4, pp. 2008–2015, Apr. 2014.
[6] D. Reusch, “Optimizing PCB layout,” Optimizing PCB Layout. EPC-
CO, pp. 1–6, 2014.
[7] W. Kangping et al., “An optimized layout with low parasitic
inductances for GaN HEMTs based DC-DC converter,” in 2015 IEEE
Applied Power Electronics Conference and Exposition (APEC),
2015, pp. 948–951.
[8] K. Wang, X. Yang, L. Wang, and P. Jain, “Instability Analysis and
Oscillation Suppression of Enhancement-Mode GaN Devices in
Half-Bridge Circuits,” IEEE Transactions on Power Electronics, vol.
8993, no. c, pp. 1–1, 2017.
[9] C. Ni, Z. Zhao, and X. Cui, “Inductance Calculation Method Based
on Induced Voltage,” IEEE Transactions on Magnetics, vol. 53, no.
6, pp. 1–4, Jun. 2017.
[10] J. Wang, H. S. Chung, and R. T. Li, “Characterization and
Experimental Assessment of the Effects of Parasitic Elements on the
MOSFET Switching Performance,” IEEE Transactions on Power
Electronics, vol. 28, no. 1, pp. 573–590, Jan. 2013.
[11] J.-L. Schanen, C. Guerin, J. Roudet, and G. Meunier, “Modeling of
Printed Circuit Board loop inductance,” IEEE Transactions on
Magnetics, vol. 30, no. 5, pp. 3590–3593, Sep. 1994.
[12] A. Lidow, J. Strydom, M. de Rooij, and D. Reusch, GaN Transistors
for Efficient Power Conversion, Second Edi. Wiley, 2014.
[13] A. Lidow and M. De Rooij, “eGaN ® FET Electrical Characteristics.”
EPC-co, pp. 1–9, 2012.
[14] H. J. Won-suk Choi, Dongkook Son, “AN-4178 New High-Voltage
SMD Package, Power88 for High- Efficiency, and Low-Profile
Power Systems.” Fairchild, p. 7, 2015.
[15] K. Wang, L. Wang, X. Yang, X. Zeng, W. Chen, and H. Li, “A
Multiloop Method for Minimization of Parasitic Inductance in GaN-
Based High-Frequency DC–DC Converter,” IEEE Transactions on
Power Electronics, vol. 32, no. 6, pp. 4728–4740, Jun. 2017.
[16] H. J. Zhang, “PCB layout considerations for non-isolated switching
power supplies,” Application Note, Linear Technology, vol. 57, no.
15, pp. 41–59, 2012.
[17] J. D. Jackson, Classical Electrodynamics, Third edit. Wiley, 1998.
Adrien Letellier obtained his Master degree
in power electronics and automatics in
ESME Sudria in 2011. He started his PhD in
2015 and focuses on understanding how to
achieve high frequency power conversion
using GaN semiconductors. From 2011 to
2013 he joined Valeo group as a system
engineer to design motor algorithms. From
2014 to 2015 he went as a laboratory
engineer at the engineer school ESME Sudria to improve the
design of various converters and supervise students’ graduation
projects focused on power conversion and control algorithms.
M. Letellier received the creativity price in 2011 for its project
on the induction motor sensorless vector control.
Maxime R. Dubois (M’99) obtained his
B.Sc. in Electrical Engineering from the
Université Laval, Québec, Canada in 1991.
He received a Ph.D cum laude from Delft
University of Technology in The Netherlands
in 2004. Between 2004 and 2011, he has been
with the Université Laval. Since 2011, Prof.
Dubois has been Associate Professor at
University of Sherbrooke, Canada. He is the
founder of Eocycle Technologies Inc., a company specialized
in the development of TFPM. He is also the founding professor
of the company AddEnergie Technologies. His fields of interest
are electrical machines and power electronics applied to the
field of wind energy, energy storage and electric vehicles. He
was the Technical Program Committee Chair of the 2015 IEEE
VPPC and a Guest Editor for the Special Issue of IET Electrical
Systems in Transportation on Design, Modeling and Control of
electric Vehicles.
João Pedro F. Trovão
(S’08 - M’13 - SM’17) was born in
Coimbra, Portugal, in 1975. He received the
M.Sc. degree and the Ph.D. degree in
Electrical Engineering from the University
of Coimbra, Coimbra, Portugal, in 2004 and
2013, respectively. From 2000 to 2014, he
was a Teaching Assistant and an Assistant
Professor with the Polytechnic Institute of Coimbra–Coimbra
Institute of Engineering (IPC–ISEC), Portugal. Since 2014, he
0885-8993 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPEL.2018.2826920, IEEE
Transactions on Power Electronics
IEEE TRANSACTIONS ON POWER ELECTRONICS
has been a Professor with the Department of Electrical
Engineering and Computer Engineering, University of
Sherbrooke, Sherbrooke, QC, Canada, where he holds the
Canadian Research Chair position in Efficient Electric Vehicles
with Hybridized Energy Storage Systems. He is an
author/coauthor of over 75 journal and conference papers. His
research interests cover the areas of electric vehicles,
hybridized energy storage systems, energy management and
rotating electrical machines. J. P. F. Trovão is the General Chair
of the 2018 IEEE Vehicle Power and Propulsion Conference,
Chicago, US. He was the Technical Program Committee Co-
Chair of the 2017 IEEE Vehicle Power and Propulsion
Conference, Belfort, France, the General Co-Chair and the
Technical Program Committee Co-Chair of the 2014 IEEE
Vehicle Power and Propulsion Conference, Coimbra, Portugal,
as well as the Award Committee Member for the 2015
(Montreal, Canada) and 2016 (Hangzhou, China) IEEE Vehicle
Power and Propulsion Conferences. He was a Guest Editor for
the Special Issue of IET ELECTRICAL SYSTEMS IN
TRANSPORTATION ON ENERGY STORAGE AND ELECTRIC POWER
SUB-SYSTEMS FOR ADVANCED VEHICLES. He was a Guest
Editor for the Special Issue of IEEE TRANSACTIONS ON
VEHICULAR TECHNOLOGY ON ELECTRIC POWERTRAINS FOR
FUTURE VEHICLES.
Prof. Hassan Maher received the Ph.D.
degree in integrated micro-and opto-
electronics and sensors from University of
Paris XI and the HDR in Physics from
University of Lille1. In 1996 he joined the
CNET, Bagneux, France, working on the
growth of InP HEMT by MOCVD and
design, implementation and characterization
of InP composite channel HFETs for PIN-HEMT circuits. In
2000 he joined the CSDL at Simon Fraser University, BC,
Canada, working on AlGaN/GaN microwave field-effect power
transistors. In 2001 he joined PerkinElmer-Optoelectronics,
QC, Canada, working on the development of the PIN-HBT
circuits. In 2003 he joined OMMIC, Limeil-Brévannes (Paris),
France as project leader. He was promoted in 2011 director of
R&D division. He was working as a prime coordinator of
several FP7, ESA, ANR and other national projects dealing
with RF MMICs based on HEMTs (InP, GaAs, Metamorphic,
Pseudomorphic and GaN), RITD (diode) and HBTs. As of 2012
he is a professor at the Université de Sherbrooke, QC, Canada
and a member of the Laboratoire Nanotechnologies
Nanosystèmes (LN2)-CNRS-UMI-3463, Institut
Interdisciplinaire d’Innovation Technologique (3IT). His
research is focused on advanced fabrication processes of III-V
(GaAs, InP, GaN) devices and circuits