2354 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO.

3, MARCH 2018

Bus Bar Design for High-Power Inverters

Alan Dorneles Callegaro , Student Member, IEEE, Jing Guo , Student Member, IEEE,
Michael Eull, Student Member, IEEE, Benjamin Danen, Student Member, IEEE, Jason Gibson,
Matthias Preindl , Member, IEEE, Berker Bilgin, Senior Member, IEEE, and Ali Emadi, Fellow, IEEE

Abstract—This paper presents a comprehensive analysis about quality of the work; indeed, many interesting results are pre-
bus bar design procedure. Some applications in terms of rated sented that, when put together, help to guide the design process.
power and shape are investigated regarding their particular re- In [1], the basics for low-inductance bus bar design are laid out
quirements and challenges. The dc-link capacitor selection is one
of the first and most important steps. It not only dictates the bus and the benefits of such an approach are exhibited. At around
bar complexity but also is the key to accomplish a high-power den- the same time, other researchers presented a different approach
sity prototype. Current density and distribution is discussed in this to bus bar design for a single-phase inverter that removed one
paper based on simulation results. Moreover, the effects of stray layer in the stackup and showed a benefit in the inductance, al-
inductance and capacitance are explained along with the dc-link
though it did cost them in complexity [4]. In [3], several theories
capacitors and power semiconductor devices. Simulated results are
compared with measurements by a high precision impedance ana- are tested and their respective benefits and drawbacks shown.
lyzer, which shows the reliability of 3-D modeling-based designs. A bus bar for Silicon Carbide MOSFET modules was designed
and tested in [5], albeit at low frequencies, adding to the general
Index Terms—Bus bar, high-power inverter, power electronics,
SRM inverter, stray capacitance, stray inductance, three-phase
body of work.
inverter. Some unique approaches were taken as more researchers pub-
lished on the subject. One of the most interesting is presented
in [8], where a multilayer design, with several forward, return,
I. INTRODUCTION and grounding planes are present. The result is a lower induc-
US bars have been present in power distribution systems tance and reduced electromagnetic emissions. Others built on
B for many years. In their most basic form, bus bars are
large conductors used to transmit significant quantities of cur-
this and previous ideas by analyzing special structures for mul-
tilevel converters. In [2] and [6], the respective authors built
rent where a wiring scheme is infeasible. With power transis- multilevel converters with bus bars and, in the latter, a listing of
tors continuing to move upward in current levels and switching bus bar topologies from previous studies for different converter
frequency, laminated bus bars have been attracting increasing applications was presented.
interest from both industry and academia for the system ben- Other researchers have considered thermal effects driven by
efits they exhibit. These include a low impedance via tightly the generated power losses [9], [10]. In [9], a model is cre-
coupled conducting planes, a simplification in system assembly ated to analyze the steady state and transient performance of a
and reliability enhancements. single bus bar, independent of material and geometry. In [10],
Many studies have been undertaken that involve the design a scalable lumped parameter thermal model is developed for a
and use of a bus bar for some applications [1]–[8]. Often, the laminated bus bar, with estimated and actual temperatures being
design of the bus bar and necessary considerations are not dis- reasonably close.
cussed in great detail, with most of the attention being paid to Few sources discuss design practices explicitly. A good re-
minimizing the stray inductance. This does not detract from the source in the literature is [7], where several key aspects are
considered: integration of the capacitors and switches, holes,
Manuscript received November 3, 2016; revised February 6, 2017; accepted
insulation material, and overall dimensioning. Bus bar manu-
March 23, 2017. Date of publication April 6, 2017; date of current version facturers normally provide some advice and support on their
December 1, 2017. This work was supported in part by the Canada Excellence websites as well, with one such example being found on [11].
Research Chairs Program and in part by the Natural Sciences and Engineering
Research Council of Canada. Recommended for publication by Associate Editor
The provided information is often not exhaustive and does not
H. Wang. (Corresponding author: Alan Dorneles Callegaro.) cover complex geometries, leaving designers to estimate perfor-
A. D. Callegaro, J. Guo, B. Danen, B. Bilgin, and A. Emadi are with mance under such conditions, opening up the opportunity for
the McMaster Automotive Resource Centre, McMaster University, Hamilton,
ON L8P 0A6, Canada (e-mail: [email protected]; [email protected];
suboptimal designs to be manufactured.
[email protected]; [email protected]; [email protected]). With respect to multilayer designs, they afford the opportu-
M. Eull and M. Preindl are with the Department of Electrical Engineering, nity to decrease losses and, potentially, the parasitic inductance
Columbia University in the City of New York, New York, NY 10027 USA
(e-mail: [email protected]; [email protected]).
of the interconnected system. Some research has been done on
J. Gibson is with the MERSEN Electrical Power, New York, NY 14623 this subject, with a proposed reduced layer bus bar proposed in
USA (e-mail: [email protected]). [4], where the dc-link bus bars are parallel to one another on
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
the same plane and the phase output is separated from the two
Digital Object Identifier 10.1109/TPEL.2017.2691668 by a layer of insulation. Two instances of intricate multilayer

0885-8993 © 2017 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.
CALLEGARO et al.: BUS BAR DESIGN FOR HIGH-POWER INVERTERS 2355

Fig. 1. Bus bar design procedure. Components selection, ac and dc current density analysis, 3-D modeling and finite element analysis (FEA) simulations, and
insulation material selection are involved in this process.

arrangements for multilevel inverters are presented in [6] and [2] terms of mechanical construction, the number of inputs and
and, in [8], a truly multilayer bus bar design is shown. In all three outputs can dictate the design complexity. Also, on the electrical
papers, moving to multilayer designs has benefited the induc- point of view, the average and rms current amplitude as well as
tance. There are certainly added costs and complexities moving the low- and high-frequency components are responsible for the
from a single conductor bus bar to a multilayer one. One of the bus bar thickness and number of connections in order to improve
main design considerations that has to be designed around is the the current distribution. However, the most crucial point for a
hi-pot test, which is a high-potential test to verify the electrical good bus bar design is the dc-link capacitance requirement.
insulation between the bus bar conductors. However, coupling As illustrated by Fig. 1, a bus bar design is composed of
multilayer designs with aluminum as the conductor could yield several steps. Power semiconductors and dc-link capacitor ge-
benefits in terms of mass reduction, but at the expense of higher ometry are chosen to optimize the power density as well as to
losses. minimize the bus bar complexity. Some examples are presented
This paper endeavors to outline a design process for a bus at the first step including air and water cooled configurations.
bar. Experimental results for five physically realized designs The next step comprehends terminal connection size in regards
are presented to show the validity of the approaches used. In to current density and skin effect. It is important to note that,
Section II, general design considerations are discussed where from the results obtained on bus bars A, B, and D, terminal con-
a step-by-step procedure for a bus bar design and optimization nections have a significant impact on stray inductance. Once bus
is proposed. Section III provides a detailed analysis of current bar thickness and terminals are defined, dc and ac connections
ripple generated by the switching operations of a three-phase are defined on a 3-D CAD model. For a better current distri-
inverter using sinusoidal pulse width modulation (PWM). A bution, dc input connections must be symmetrically positioned
number of key subjects are covered in Section IV—including relative to the power modules. For the ac current distribution, the
the current density, skin and proximity effects, and parasitic symmetry between dc-link capacitor and power semiconductor
parameters—and simulations are provided to show how they modules is also necessary. Sharp corners and bends can cause
may be obtained. Section V provides experimental results to eddy currents and consequently voltage drops, which results
substantiate the design principles and simulations discussed in in losses and heat generation, “the surface impedance has the
preceding sections. Last, conclusions are drawn on the efficacy maximum value at the conductor edge. The sharper the edge,
of the presented paper. the greater the value of Zs,max ” [12]. To avoid these effects,
corners and bends should have a sufficient radius, as illustrated
II. DESIGN CONSIDERATIONS in Fig. 1. Simulation analysis is an important step in order to
guarantee the bus bar performance. This is can be an iterative
A. Bus Bar Shape Definition and Manufacturing Feasibility process where the terminal connection, dimensions, or geome-
Distinct converter topologies and applications require special try of the bus bar can be updated to match requirements such
characteristics for the shape and dimensions of a bus bar. In as current density, distribution, and stray inductance. Finally,
2356 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 3, MARCH 2018

Fig. 2. DC and ac current paths for different bus bar applications. The dc
current represents the input average current and the ac component is circulating
between the IGBT/MOSFET terminals and dc-link capacitor.

according to the application requirements, the insulation mate-

rial and configuration are defined and the manufacturing process
begins.
Fig. 2 shows the five applications that are discussed in the
paper and Fig. 3 shows the prototypes for those applications.
Fig. 3(a) and (b) illustrates two solutions where both the IGBT
module and film capacitor are coupled to a water cooled ther-
mal plate. Although both bus bar designs look similar, there is
a major difference between them. In the first case, the dc in-
put is connected to the left side which makes both dc and ac
current components circulate through the bus bar increasing its
current density and losses. In the second design, as shown in
Fig. 4, the dc input connection is closer to the IGBT module and
Fig. 3. High-power inverter prototypes with five different bus bar designs.
consequently the majority of the current on the bus bar is the (a) Bus bar type A, three-phase inverter. (b) Bus bar type B, four-phase asym-
high-frequency component; as a result the ohmic losses can be metric bridge switched reluctance machine (SRM) inverter. (c) Bus bar type C,
reduced. During the component selection process of bus bar B, three-phase SRM inverter. (d) Bus bar type D, dual three-phase inverter. (e) Bus
bar type E, three-phase inverter.
the SBE power ring capacitor was chosen due its higher current
rating and capacitance value. Another option for this case was
the LH3 series film capacitor from electronic concepts, with
slightly lower rated current and capacitance at the same voltage
rating.
Simpler bus bar configurations are shown in Fig. 3(c)
and (e). Type C consists of a flat bus bar connecting the in-
put and output with the dc-link capacitor, while type E is shaped
around the capacitors. In most cases, the ability to share the heat
sink between the power semiconductors and dc-link capacitor is Fig. 4. DC and ac current flow in bus bar type B.
lost in this configuration. For applications such as a dual-inverter
topology, where two inverters are sharing the same bus bar and greatly increase the cost and complexity of the design. In sum,
dc-link capacitors, bus bar type D has its dc input connection in the bus bar design starts along with the power electronics con-
the middle of the bus bar, as illustrated in Fig. 3(d). verter design.
Cylindrical and rectangular capacitors are most commonly Although the general shape and configuration of the bus bar is
employed, and in terms of technology, the options are ceramic, defined by the designer and the component layout of the system,
electrolytic, or film capacitors. For the sake of safety and robust- the manufacturer will make suggestions for complexity and cost
ness, film capacitor are widely implemented nowadays. The de- reduction. Since all bends or forms in the metal increase cost,
sign and number of capacitors also affects the number of holes the majority of shapes that deviates the bus bar from the flat
and connection points. This increases the possibility of an elec- plate stock shape will be done using a separate manufacturing
trical short area (hi-pot failure) and complicates the design. operation. Each individual manufacturing operation can have its
The most common and easiest connection method for a ca- own costs associated with it, such as:
pacitor onto a bus bar is a screw or bolt on connection. Soldering 1) General run-time cost. This is the general labor, machine
or spot welding connection methods can also be used, but they rate, and overhead cost applied to an operation.
CALLEGARO et al.: BUS BAR DESIGN FOR HIGH-POWER INVERTERS 2357

2) Setup cost. This is the time to set up and calibrate the ma- two potentials) has moisture or has dirt or debris on it, the
chine to perform the manufacturing operation. The setup creepage distance may be significantly reduced and the current
time is spread across the lot size being run. can track along the surface of the insulation easier to create a
3) Certain operations, such as bending the metal, may require short. The severity of the contamination is typically defined by
a unique tool, fixture, or die to perform the operation. This the pollution degree rating [15].
unique tool will cost to that job as a one-time charge. Pollution degree affects the creepage and clearance distances
In general, the process of reviewing or creating a design re- required to ensure the safety of a product and it is classified
quires an in-depth knowledge of the manufacturing process. In according to the amount of dry pollution and condensation in
terms of bending, a general guideline is to make the inside ra- the environment. Safety standards bodies such as Underwriters
dius of a bend equal to or greater than the overall thickness of Laboratories (UL) and IEC categorize them as follows:
the bar or sheet to bend. For bending raw metal sheets or bars, Pollution Degree 1: No pollution or only dry, nonconduc-
it is possible to make the inside bend radius smaller than the tive pollution occurs and has no effect. Examples are, sealed
metal thickness, but it can introduce problems or make the op- components and within air/water tight enclosures.
eration more difficult. Inside and outside corners of a stamped Pollution Degree 2: Normally only nonconductive pollution
plate should be radiused for several reasons. By radiusing the occurs. Temporary conductivity caused by condensation is to be
inside and outside corners of a stamped plate, the lamination expected, commonly present in offices and laboratories.
will stretch and form better over the corner, there will be fewer Pollution Degree 3: Conductive pollution or dry nonconduc-
eddy currents and resulting hot spots and the sharp corner will tive pollution that becomes conductive due to condensation. Of-
require less time and care to reduce its burr. ten the case in industrial environments or manufacturing areas
When designing a higher level assembly, whether it is assem- (harsh environments).
bled by hand with operators or by automated machinery, the Pollution Degree 4: The pollution generates persistent con-
physical space and limitations must be considered. The order ductivity caused by conductive dust, rain, or snow.
of assembly is important. Once the first component is attached, The pollution degree can also impact the selection of the in-
we must make sure that there is room to attach the second. Not sulation type. Certain materials will simply not hold up well
only the actual space it will occupy, but also the path to get the under certain conditions. Nomex for example is a popular insu-
second component into position should be taken into account. lation material that has been in use for decades. It is however, an
Once all of the assembly is verified, there has to be room to aramid paper. This means it is hydroscopic and has the potential
maneuver the required tools. If a screw is needed to assemble a to absorb moisture. Once this happens, the dielectric rating is
component, then there has to be room for the screw driver and severely degraded and may also cause delamination.
an operator’s hand to work the screw driver. Finally, possible Since each insulation material has an upper and a lower tem-
servicing such as periodical inspection, maintenance, cleaning, perature limit for functionality, this must also be taken into
or replacement must be considered. account. The upper limit is typically the parameter that causes
design issues. One must consider both the ambient temperature
and the self-heating temperature rise of the bus bar. The lower
B. Conductor Insulation temperature limit is rarely a problem, but must be considered
The selection of the electrical insulation is driven by the op- if there are any shock or vibration conditions that can occur at
erating voltage, the operating temperature, and the environment those lower temperatures.
in which it has to function. The operating voltage dictates the re- How the bus bar is to be used and what it comes into contact
quired dielectric strength of the insulation which in turn depends with must be considered, when selecting the insulation material.
on the material used. Most laminating films are relatively durable, but can be cut
The most common materials are: Nomex, Tedlar, Mylar, Kap- creating an insulation fault. It must also be considered if another
ton, Ultem, Valox, epoxy-glass, heat shrink tubing, and epoxy component may impact or rub against the laminated surface,
powder coating, or some combination of them [13]. For high- which can result in a hi-pot failure.
frequency ac applications, the material Nomex type 410 from With any of these potential “severe physical use” conditions,
DuPont has a dielectric strength of 34 kV/mm for a thickness an alternate material should be considered, such as powder coat-
of 0.18 mm and 32 kV/mm for 0.51 mm [14]. As the insula- ing or a secondary material on top of the laminated surface.
tion thickness increases, the dielectric-strength decreases, and Powder coating is a hard, durable, and chemical resistant dielec-
consequently, as estimated by (6) in Section III-B, the bus bar ca- tric coating that can be used instead of or in conjunction with
pacitance is also reduced. A higher capacitance value decreases lamination. Secondary barrier materials such as FR-4 (Flame
the overall bus bar impedance and consequently reduces the Retardant) or GPO3 (Glass Polyester Laminate, NEMA LI-1)
noise produced by parasitic inductances at high frequencies. are also common in these conditions.
This explains the importance of high-dielectric-strength mate- The choice of the thickness is typically driven by the dielectric
rial combined with high-technology manufacturing. requirement, but can also be driven by the physical design of the
The creepage distance, which can cause short circuits, should conductors themselves. The flexibility of the insulating material
be considered as well and is dictated by the exterior material must also be considered. Under severe vibration or flexation
of the insulation film (bare film or coated with resin) and the conditions, powder coating can crack and fail. The design may
environment it is in. If the surface of the bus bar (between also require the conductor to be formed after being laminated.
2358 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 3, MARCH 2018

threshold would be the inverse: more material, in either axis,

is needed to reduce the current density and, consequently, the
losses. More material, in turn, leads to a more expensive and
Fig. 5. Conductors insulation: (a) Simplest and least expensive. (b) Insulation heavier bus bar. For complex bus bar shapes, the mass can be
is pinched closed and sealed on itself. (c) Epoxy filled edges.
approximated by breaking it up into discrete segments that are
easy to calculate. For a more precise calculation, the mass of the
dielectric material and the adhesive holding the pieces together
can be calculated; however, these are relatively small values
when compared to the conductor and, unless the application is
severely weight restricted, need not be considered.

D. Noncopper Conductors

Fig. 6. Bus bar thickness design considerations based on maximum current Additional subjects of interest, though not detailed in depth
density J [A/mm2 ]. in this paper, are the transition to noncopper materials (e.g.,
aluminum). Working with aluminum has advantages and draw-
This requires a film that can stretch and bend with the conductor. backs. Aluminum, relative to copper, has a density that is 33%
Most films can do this, Nomex is one that cannot. lower than that of copper; however, both its electrical and ther-
Finally, as illustrated in Fig. 5, the edge conditioning of the bus mal conductivity are 60% lower in comparison, depending on
bar defines the conductor edges protection from contamination the alloy used. Therefore, an aluminum design should weigh
and shorting. Open edge is the most susceptible to contamina- 66% less than a copper one of equivalent thickness. But, be-
tion and hi-pot failure and is typically used in clean and dry cause of the conductivity difference explained above, the weight
environments. The sealed edge gives some protection against reduction from a copper bus bar is about 33%. The cost saving
mild contamination and is the most common and typically not of a cheaper per-pound price of aluminum is also eroded by this
more expensive than the open edge. However, the tooling cost as well.
can be higher for complicated laminating fixtures. In the third Aluminum is also particularly good for mass production. It
method, the channels formed in the open edge technique are now can be made as a mould that offers more consistent results.
filled with an epoxy. This gives a strong and durable barrier to However, the manufacturability impact of doubling the conduc-
contamination and moisture. tor thickness must be considered. Going from a 1-mm thick
copper conductor to a 2-mm thick aluminum one is typically
C. Conductor Thickness and Current Density not a problem. But if the copper conductor is already 3 mm
thick, doubling it to 6 mm will greatly impact many aspects of
When mechanical strength is not the main requirement, the
the manufacturing process.
copper thickness is defined by the input and output terminal
Other lossy conductors, such as nickel, can be added to the
connectors. As illustrated by Fig. 6, given the maximum current
surface of the copper conductors to reduce high-frequency har-
density J [A/mm2 ] and the length l available, the thickness w
monics [17]. This combination will attenuate the high-frequency
is calculated to keep the current density within its specifica-
noise, which is pushed to the surface to travel through the lossy
tion. For instance, in high current applications, the dc input is
nickel due to the skin effect. The low-frequency current will be
split in two or more connectors in order to improve the current
distributed throughout the less-resistive copper.
distribution with acceptable copper thickness. As illustrated by
Fig. 9, dc current distribution is improved by splitting the posi-
tive and negative terminals in three. This reduces ohmic losses III. SIMULATIONS
and evenly spread the heat across the bus bar, which reduces In this section, simulations that are used to predict the bus bar
the hot spots. Typically, the bus bar conductors are sized for a performance are introduced. Following the steps described in
30 ◦ C self-heating temperature. Section II, the bus bar can be preliminarily designed. Then the
The lower boundaries in bus bar design require: a minimum next step is the bus bar performance evaluation. If the predicted
conductor thickness to prevent it from melting when the nominal bus bar performance meets the requirements, the bus bar can be
current is applied and a minimum insulation thickness to sustain sent for manufacturing; otherwise, modifications in the initial
the intended operating voltage. An upper boundary does not design might be demanded based on the simulation results.
explicitly exist and, thus, is decided by the designer as a result of Bus bars can be designed according to the packages and loca-
system constraints and desires: size, weight, cost, and electrical tions of the selected switching devices and dc-link capacitors. To
and thermal performance. Depending on the thickness of the evaluate the performance of a bus bar before it is manufactured,
copper, it may be prudent to thicken the dielectric material to analysis could be done by simulation using an electromagnetic
minimize the impact of swarf when pressing the sheets together. simulation software, such as ANSYS Maxwell. In the simula-
A rule of thumb for bus bar design is to not allow the current tions, the current density and distribution can be estimated based
density to exceed 5A/mm2 [16]. A higher threshold would mean on different inverter operating conditions. Additionally, the cur-
that a smaller cross sectional area is needed, allowing for a rent excitations should be defined by (1), where Iave,in , Irm s,in ,
size reduction in the vertical or horizontal directions. A lower Irms,ripple , M , and cos φ indicate the average and rms value of
CALLEGARO et al.: BUS BAR DESIGN FOR HIGH-POWER INVERTERS 2359

TABLE I
SYSTEM PARAMETERS OF APPLICATIONS

Bus bar parameters A B C D E

V dc [V] 300 600 600 425 300

I r m s, o u t [A] 212 400 600 365 450
I a v e, i n [A] 195 300 250 400 410 Fig. 7. Bus bar simulation: dc analysis equivalent circuit.
I r m s, i n [A] 230 350 425 435 490
I r m s -r i p p l e [A] 123.5 - - 237 270
requirement; thus, the overheat on entire bus bar or in some of
its local areas can be prevented. In addition, some applications
the dc-link current, rms ripple current, modulation index, and contain more than one dc-link capacitor and, to ensure that each
power factor, respectively, [18]–[21]: capacitor operates under similar conditions, the ac current dis-
 tribution and the current sharing among the capacitors should
Irm s,ripple = I 2 rm s,in − I 2 ave,in be balanced. The phase currents, and consequently the current
  √  through the power devices, are defined by the load demand and
√
 3M 2 3M 9 2 individually controlled by the control algorithm. Therefore, for
= Irm s ·  + − M · cos2 φ.
2π π 8 a three-phase balanced load, currents through the power semi-
conductors are also balanced.
(1)
In an inverter, the dc bus current can be presented by the sum
As it is mentioned above, the design should meet the require-
of dc and ac current components. The dc current component is
ment of current density limitation (a typical value for passive
Iave,in , while the ac current component is the dc-link capacitor
cooled bus bars is 5 A/mm2 ) to prevent overheating in any part
current, and its rms value is Irm s,in . As a result, the current
of the bus bar. Meanwhile, to achieve balanced current sharing
distribution of dc component and ac component can be evaluated
at the output terminals and among the dc-link capacitors, the
by dc analysis and ac analysis, respectively. The bus bar is
current distribution on bus bars needs to be evaluated. [22]
modeled with a single average R-L model. In the design process,
Furthermore, 3-D analysis is necessary to calculate the para-
asymmetries are typically analyzed with FEA using a 3-D model
sitic parameters of the bus bar, so resistance and bus bar power
of the bus bar.
loss can be obtained. Resistance varies depending on the fre-
In dc analysis, a dc current is injected from the dc input
quency of the ac current. The relationship between the frequency
terminals into the bus bar, as shown in Fig. 7. Rp , Lp , Rn ,
and the resistance can be obtained through simulation as well.
and Ln stand for the resistance and stray inductance of bus
However, the resistance of the bus bar is typically small and the
bar positive plate (p) and negative plate (n), respectively. At the
amount of power loss is usually negligible compared to the total
same time, the positive plate and the negative plate of the bus bar
power loss of the entire inverter. Moreover, the value of bus bar
are shorted at the power switching device terminals [4]. Thus,
stray inductance can be estimated. To prevent switching devices
the current distribution on the bus bar between input terminals
being damaged by large voltage spikes during turn-OFF tran-
and switching device is analyzed. To implement the simulation,
sients and to reduce switching power loss of the device and the
excitation current is defined as the rms value of inverter input
stray impedance of the bus bar, the total inductance in the circuit
current.
has to be minimized. Generally, the total circuit inductance in-
A preliminary design of an inverter dc bus bar with one set
cludes dc-link capacitor equivalent series inductance, switching
of dc input is shown in Fig. 9(a). It is obvious that the current
device equivalent inductance, and the bus bar stray inductance.
distribution is unbalanced and higher current density is acquired
Even though it is possible to select switching devices and ca-
around the area near dc input terminals on both plates.
pacitors with relatively small inductance, there is only a limited
To improve the current distribution, a bus bar with three sets
improvement on the total circuit inductance. Nevertheless, by
of dc inputs is designed. It can be seen in Fig. 9(b) that dc inputs
laminating two conductors (positive and negative plate) the mu-
are distributed along one side of the bus bar and the current flow
tual inductance is maximized and significant reduction in the
on the entire bus bar is balanced. Therefore, the current density
bus bar stray inductance can be achieved. As a result, the to-
analysis shown in Fig. 6 holds.
tal inductance in the circuit is also minimized. Therefore, the
Then the ac analysis can be implemented to evaluate the bus
stray inductance analysis is very important in dc bus bar design.
bar current distribution between capacitors and switching de-
Another parameter that needs to be analyzed is the stray ca-
vices. In other words, the current sharing among dc-link capac-
pacitance, which should be maximized in the design to reduce
itors is analyzed. The equivalent circuit of ac analysis is shown
the stray impedance and the electromagnetic interference (EMI)
in Fig. 8. Here, an ac excitation current is injected into the bus
[1], [23]. Table I shows the parameters corresponding to each
bar from switching device terminals. In addition, the positive
bus bar design given in Fig. 3.
plate and the negative plate of the bus bar are shorted at dc-link
capacitor terminals.
A. Current Density
The current distribution in ac analysis in the preliminary de-
The analysis of current density and current distribution is in- sign in Fig. 9(c) is similar to the revised design in Fig. 9(d).
dispensable. A bus bar design should satisfy the current density This validates that the ac current distribution is defined by the
2360 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 3, MARCH 2018

Fig. 10. Skin effect simulation analysis for a rectangular conductor showing
Fig. 8. Bus bar simulation: ac analysis equivalent circuit. the comparison between analytic and FEA results.

FEA results is shown by Fig. 10:

1 √
δ=√ (1 − e−t π f σ μ ). (2)
πf σμ

An imbalance in the current density distribution is possi-

ble when considering complex geometries and, in particular,
those that are asymmetric. In these cases, it may be necessary to
guide currents—both dc and ac—to their respective locations by
adding features to reroute current. For example, in the ac ripple
current case, currents flow from the power module through the
dc-link capacitors and back. If a significant imbalance is present,
then it could result in accelerated failure of the capacitors and,
consequently, for the inverter as a whole. Equal current sharing
Fig. 9. DC and ac analysis for inverter dc-link current density and distribution between devices is critical to the long-term reliability of neg-
with three sets of dc input (the negative plate is shown on the left and the ative temperature coefficient components, such as electrolytic
positive plate is shown on the right, maximum current density of 5 A/mm2 ). (a)
DC analysis for inverter dc-link current density and distribution with one set of
capacitors, IGBTs, and diodes.
dc input. (b) DC analysis for inverter dc-link current density and distribution When assessing the rerouting of currents, the focus is
with three sets of dc input. (c) AC analysis for inverter dc-link current density normally placed on minimizing the current density through a
and distribution with one set of dc input. (d) AC analysis for inverter dc-link
current density and distribution with three sets of dc input.
vulnerable path or the current balancing of active and passive
components. The most reliable method by which these issues
can be avoided is to ensure that the paths are symmetric and
location of the capacitors and switches rather than those of dc of the same length. This, however, is not always possible and,
input tabs. indeed, for many designs, is infeasible. Hence, in some cases,
Both the skin and proximity effects contribute to design chal- it becomes necessary to consider adding intentional obstacles
lenges by increasing the losses and reducing the region of effec- (holes) to the current flow path to make it an unattractive route.
tively utilized conductor. If the goal is to stay below a prescribed Fig. 11 illustrates such an idea. The point of injection is one
current density, then these effects need to be carefully studied. of three IGBT module terminals and, in this case, it is in a
In the dc case, the thickness of the conductor is chosen to meet location where it is impossible to guarantee equal current shar-
a chosen maximum current density. In the ac case, when skin ing between capacitors, with the left-most one experiencing
and proximity effects are present, current flows closer to the the highest current and, therefore, losses. Fig. 11(a) shows the
surface of the material, reducing the effective conducting area. bus bar with no alterations, whereas Fig. 11(b) and (c) shows
This requires a reconsideration of a bus bar’s layout. A bus it with symmetrical cuts, triangular and round added, respec-
bar designed with dc current density in mind would result in tively. The current density, while not drastically different, has
a significant amount of wasted conductor with a sufficiently been reduced on the left-hand side with current being pushed
high-frequency ac current. rightward, thereby reducing the stresses on the left-most ca-
The skin effect can be considered using the skin depth, which pacitor. A sweep of incision dimensions can be made and the
is the distance from the surface that the ac current flows. This current sunk by each capacitor evaluated until a suitable result
depth depends on the conductor itself as well as the frequency of is achieved.
the current. Generally, the skin depth can be approximated while In order to address the current distribution problem, Fig. 12
neglecting the thickness of the conductor. When the material illustrates two different design for dc input connections for the
thickness is comparable to the skin depth, it should be taken into same application. On first design, the input connectors are con-
account [24], [25]. This is often the case with bus bars currents centrated at the edge of the bus bar. This configuration may cause
at intermediate frequencies, when the skin depth is similar to the a hot spot due to the higher current density on the top and bot-
thickness of the bus bar. The equation to approximate skin depth tom layer at the same bus bar region. This overheat may damage
while accounting for the thickness is given in (2). An illustration the insulation material and consequently reduce the converter
of the skin effect and the comparison between analytical and robustness.
CALLEGARO et al.: BUS BAR DESIGN FOR HIGH-POWER INVERTERS 2361

Fig. 11. AC current guiding by using incisions in the bus bar. (a) Bus bar with no cut. (b) Bus bar with triangular cut. (c) Bus bar with round cut.

Fig. 13. Example of bus bar structure.

However, the total ohmic loss on the bus bar is only several
Fig. 12. Current distribution analysis for two configurations of input dc current watts that is negligible compared to the inverter power loss.
connections, maximum current density of 5 A/mm2 . 2) Stray Inductance: The stray inductance defines the volt-
age spike during turn-OFF transient of the switching device. A
TABLE II high voltage spike, which may damage the semiconductors, is
STRAY RESISTANCE OBTAINED FROM SIMULATIONS caused by a large parasitic inductance. Furthermore, it results in
higher switching power loss and EMI, and it also restricts the
Frequency [kHz] Resistance [mΩ] Bus bar power loss [W] switching frequency of the inverter due to the thermal limita-
Negative plate Positive plate Total tions [28]. If the connection terminals for capacitors and power
0 0.158 1.41 1.53 2.94 modules are asymmetrically located, the parasitic impedance
10 0.219 1.49 1.42 2.91 difference in each current path may affect current sharing [29].
20 0.251 1.87 1.77 3.64 To understand and reduce this voltage spike and its resulting
30 0.270 2.07 1.96 4.03
40 0.284 2.22 2.14 4.36 effects, the stray inductance of the bus bar should be estimated
50 0.295 2.33 2.21 4.54 and minimized. A bus bar is a collection of parallel plates, and
60 0.304 2.41 2.28 4.69 an example is given in Fig. 13, where w, t, l, and d indicate
70 0.311 2.47 2.35 4.82
80 0.317 2.54 2.41 4.95 width, thickness, and length of each conductor, and the distance
90 0.322 2.57 2.44 5.01 between two conductors, respectively. The inductance of flat
100 0.327 2.62 2.48 5.10 plates (3) can be used to approximate the self-inductance of the
entire conductor [30]–[32], which is valid only for the dc current
analysis:
B. Parasitics 
2l w+t
Lself = 2l log( ) + 0.5 + 0.2235( ) × 10−7 .
1) Resistance: The evaluation of parasitics in a bus bar, in- w+t l
cluding resistance, stray inductance, and capacitance, is also (3)
very important. First, the resistance affects power loss on a bus From (3), it can be concluded that an increase in either the
bar. Thus, low bus bar resistance is preferred. Generally, the re- width or thickness reduces the inductance. However, thick bus
sistance is determined by materials and shape of the conductor. bar design renders high material cost. To achieve the minimized
Bus bar E is taken as an example in resistance calculation. inductance and material cost, the bus bar is designed with large
Table II presents the resistance and ohmic loss acquired by dc- width and small thickness, which is commonly utilized by con-
and ac-analysis. It demonstrates that as the current frequency ventional flat plate design.
increases, the value of resistance grows and it results in higher Furthermore, the flux linkages of two conductors will cancel
ohmic loss on both negative and positive plates [26], [27]. each other due to their opposite current directions shown in
2362 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 3, MARCH 2018

TABLE III
STRAY INDUCTANCE [NH] OBTAINED FROM SIMULATIONS

Frequency [kHz] Bus bar type

A B C D E

10 21.38 24.01 11.16 13.90 9.82

20 20.46 23.33 10.68 13.14 9.36
30 20.05 23.04 10.46 12.80 9.19
40 19.79 22.87 10.34 12.59 9.10
50 19.62 22.77 10.27 12.45 9.05
60 19.49 22.7 10.22 12.35 9.00
70 19.39 22.65 10.18 12.26 8.98
80 19.31 22.62 10.16 12.20 8.96
90 19.24 22.59 10.14 12.14 8.94
100 19.19 22.57 10.12 12.10 8.93

Fig. 14. Current flow in equivalent circuit during the turn-OFF of phase-A
Fig. 13, and the total stray inductance of the dc bus bar can be upper switch.
expressed in the following equation [33]–[35]: TABLE IV
SYSTEM PARAMETERS AND ESTIMATED VOLTAGE SPIKE
Ltotal = 2 · (Lself − LM ) . (4)
Total inductance and Bus bar type
It can be seen in (4) that the way to design a bus bar with estimated voltage spike

a total inductance as small as possible is to design it with the A B C D E

maximized mutual inductance and is to maximize the over-
L c [nH] 5 5 5 15 40
lap area of two conductors. Therefore, the laminated bus bar No. of caps 1 2 2 5 5
structure is preferred in the design [33]. Total L c [nH] 5 2.5 2.5 3 8
The stray inductance of the bus bar will change with fre- L IGBT [nH] 20 23 18 14 18
L bus−bar [nH] 21.38 24.01 11.16 13.90 9.82
quency as well, due to the skin effect and unintended parasitics. Δ I [A] 212 400 600 516 640
It has been shown that the inductance can be split into two, the Δ t [ns] 75 105 160 60 160
internal inductance that changes with frequency, and the exter- V dc [V] 300 600 600 425 300
Δ V [V] 130 188 118 265 143.28
nal inductance that will be constant [36]. Results from [37] back V peak [V] 430 788 718 690 443.28
up this claim, showing the measured inductance up to 5 MHz
that asymptotes at high frequencies. The stray inductance ac-
quired in simulations is indicated in Table III. This shows similar possible while the stray inductance is minimized [1], [41]:
behavior, and asymptotes in a similar way.

Now that the bus bar stray inductance is obtained, the max- L
Z= . (5)
imum voltage spike can be calculated by analyzing the current C
flow during the turn-OFF transient of the switching devices. Take The stray capacitance of a laminated bus bar is defined by
the inverter topology as an example, when phase A upper switch the geometry of conductors and the thickness of dielectric ma-
is being turned OFF, the current flow is analyzed when currents terial between positive and negative conductors, which can be
in Phase A and C are positive, whereas it is negative in Phase B. obtained in (6) [42],
The equivalent circuit is given in Fig. 14. In the figure, it can be
ω·l
seen when T1 is being turned OFF, the current (ia1 ) in blue path C = ε0 · ε r (6)
decreases from ia to zero while the current (ia2 ) in green path d
increases from zero to ia . Thus there are two current changes where, εr is relative permittivity.
causing voltage spike across T1 . Based on the expression above, the maximum capacitance
And then, according to [36], [38]–[40], the voltage spike and can be achieved by enlarging the area of one conductor overlap-
the voltage across T1 during turn-OFF transient are calculated ping the other and reducing d; nevertheless, the self-inductance
in Table IV. The parameters used in calculations are selected is proportional to l. Thus, to maximize the capacitance without
based on the peak value of inverter output current, IGBT current increasing the inductance, thickness of the dielectric material d
fall time during turn-OFF, and the dc-link voltage. should be minimized under the condition of satisfying insulation
3) Stray Capacitance: The stray capacitance benefits the requirement [23], [31]. As per the previous discussion, simula-
system by reducing the total impedance of the bus bar and pro- tions can be used for a more accurate capacitance estimation for
viding a filter for high-frequency noise. Generally, the resistance both dc and ac analysis.
of a bus bar is small and negligible, thus the total impedance of An additional benefit of the capacitance is its ability to filter
the bus bar can be expressed by (5). Obviously, to achieve low high-frequency noise. EMI is an issue in power electronics due
impedance, the stray capacitance should be designed as large as to the several kilo hertz switching frequency of the inverter.
CALLEGARO et al.: BUS BAR DESIGN FOR HIGH-POWER INVERTERS 2363

This switching introduces high-frequency noise into the system.

The laminated structure of the bus bar creates a high-frequency
capacitor that helps mitigate the noise propagation [8], [43],
though this unintended filter is likely not enough to completely
remove the issue.

C. High-Frequency Bus Bar Modeling & Analysis

An unavoidable result of fast switching devices is the high-
frequency harmonics, termed EMI [44]. The magnitude and
propagation of these harmonics depends on the switching de- Fig. 15. Actual bus bar setup and 3-D model for measurement and simulation,
vice, the switching speed, and the components of the system. respectively.
The bus bar has a significant effect on this by introducing induc-
tance and capacitance to the main paths of the converter. EMI magnetic core around the bus bar, which was shown to further
can be detrimental to various parts of the system, and should be reduce the EMI, but the size of the core may interfere with the
considered as a metic when planning to switch at high frequen- design layout and reduce the power density. Adding additional
cies. layers will also increase the complexity and cost to the bus bar
The noise produced by the devices can propagate through the construction, so should only be employed when necessary.
electrical system (conducted EMI) via the stray capacitance of The amount of EMI can be estimated and used to classify
the system to grounded components, such as the capacitance be- the bus bar performance based on regulation values. It can also
tween the bus bar and the cold plate. The harmonics can transfer be useful to simulate the effect of adding an EMI filter, and
through the ground to susceptible devices imposing a number to evaluate its effectiveness. Simulations are typically done by
of negative effects. Typical problems caused are measurement finding an equivalent circuit for the bus bar and running a cir-
errors in sensors, false control signals, and interference in the cuit simulation [47], [48]. Modern electromagnetic software,
gate voltages of switching devices [45]. Excessive noise can such as ANSYS Q3D, gives the ability to find an equivalent
deteriorate control, reduce efficiency and cause the system to circuit from FEA, but other techniques including PEEC [49]
malfunction. Depending on the use, there may also be regula- and Cauer Networks [50] can be used as well. If the physical
tions [46]. bus bar is available, the model can be validated or improved
At high frequencies, the harmonics will begin to radiate as by measuring the impedance spectrum. This can be done by
well, using the propagation path as an antenna. Reducing EMI using an impedance analyzer or by employing time domain
in general will reduce the radiated interference, but steps can reflectometry [51].
be taken to further reduce it by adding a shielding layer to the Modeling these effects can give insight into future problems
design [44]. It can be used to shield the entire bus bar, or to as well as validate the design. Of course this is most useful
separate different positive and negative conducting plates. while designing the bus bar to allow changes if the EMI is too
The source of EMI is the fast current changes causing over- large, or to help design a filter to mitigate it. If the magnitude of
voltages and the resulting oscillations. Through the reduction conducted EMI is too large, an external filter may be required.
of the bus bar’s inductance, the overvoltage is reduced as well However, this will lead to increase in cost and reduction in
as the EMI [1]. Further, having a low impedance generally re- power density. By knowing and designing for reduced EMI, the
duces the EMI of the system [7]. Minimizing the inductance and filter size can be reduced, or removed entirely. Keeping the EMI
increasing the capacitance are some of the main goals of a bus small ensures the system is more robust, accurate, and safe.
bar, and has been discussed above. Another benefit of the ca-
pacitance is the implicit differential mode filter, helping reduce
the harmonics between the positive and negative plates. IV. TESTING
If necessary, this filter can be expanded on by adding more In order to validate the bus bar designs and analysis, stray
layers to the bus bar, for example adding a grounded layer inductance and capacitance were measured using a precision
between the positive and negative plates to create a common impedance analyzer, e.g., Keysight 4294A. As previously stated,
mode filter as well. This has been demonstrated in [8] showing the capacitor connections must be short circuited to evaluate
reduced EMI, with only a small increase in size. A benefit of the bus bar stray inductance. As shown in Fig. 15, the capacitor
using the bus bar as a filter is the minimal equivalent stray terminals were shorted using stainless steel washers. These con-
inductance. Any external filter will have significantly more due nections add parasitic inductances to the circuit and by adding
to the required connection. the exact same connections and material to the 3-D model the
Another method to filter the EMI is by adding a lossy material results can be confronted side by side.
to the surfaces of the conducting plates. This has been shown The measured and simulated results for stray inductance are
in [17], where nickel is added to the surface of the copper presented in Table V for bus bars A, B, and C. Moreover, for
conductor. High-frequency harmonics, which are pushed to the the stray capacitance the dc-link capacitor connections were left
surface due to the skin effect, will travel through the lossy nickel, open while the capacitance was analyzed at the IGBT terminal,
attenuating the noise. This can be taken further by adding a and the results are presented on Table VI. Both results present
2364 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 3, MARCH 2018

TABLE V
MEASURED STRAY INDUCTANCES [NH] OF DESIGNED BUS-BARS (L s :
SIMULATION RESULTS; L m : MEASURED RESULTS)

Frequency [kHz] Bus bar type

A B C

Ls Lm Ls Lm Ls Lm

10 21.38 30.7 24.01 25.4 11.16 46.2

20 20.46 29.9 23.33 23.55 10.68 18.6
30 20.05 26.3 23.04 23.39 10.46 14.3
40 19.79 25.6 22.87 23.25 10.34 11.6
50 19.62 25.3 22.77 22.15 10.27 9.8
60 19.49 25.05 22.7 22.26 10.22 9.6
70 19.39 24.95 22.65 21.5 10.18 9.3 Fig. 16. Bus bar E testing setup.
80 19.31 24.8 22.62 22.22 10.16 10.1
90 19.24 24.15 22.59 21.85 10.14 10.5
100 19.19 23.8 22.57 21.92 10.12 11.2

TABLE VI
MEASURED STRAY CAPACITANCES OF DESIGNED BUS-BARS (C s : SIMULATION
RESULTS; C m : MEASURED RESULTS)

Frequency [kHz] Bus bar stray capacitance [nF]

Fig. 17. Current in one dc-link capacitor. (a) Overview. (b) Zoomed view.
A B C

Cs Cm Cs Cm Cs Cm

10 4.26 4.443 11.059 10.98 24.005 24.9

20 4.259 4.422 11.056 10.92 23.999 24.78
30 4.257 4.409 11.053 10.89 23.992 24.7
40 4.256 4.399 11.049 10.86 23.984 24.64
50 4.255 4.39 11.047 10.83 23.977 24.6
60 4.254 4.383 11.044 10.82 23.97 24.55
70 4.253 4.377 11.041 10.80 23.964 24.52
80 4.252 4.371 11.048 10.79 23.958 24.48
90 4.252 4.366 11.036 10.77 23.954 24.46
100 4.251 4.363 11.034 10.75 23.95 24.44 Fig. 18. DC input connection types used in experimental verification for bus
bar E. (a) DC connection type I. (b) DC connection type II. (c) DC connection
type III.

relatively good accuracy, which proves the reliability of 3-D

modeling and simulations techniques.
In order to test the current sharing among the dc-link capaci-
tors, bus bar E is selected, which is designed for an inverter with
five dc-link capacitors and three IGBT modules. To prevent any
of the capacitors being overloaded, the connection holes on the
bus bar E are located symmetrically; therefore, the ac current
distribution should be balanced and it is validated by experi-
mental results. The experimental setup is shown in Fig. 16. Fig. 19. DC-link capacitor current with connection type I. (a) Currents in
The ac current on the bus bar circulates between five dc-link capacitor 1, 3, 5. (b) Currents in capacitor 2, 3, 4.
capacitors and three IGBT modules, as a result, the experimental
verification for ac current distribution can be implemented by
examining the currents in each dc-link capacitors. The current
in one of the capacitors is shown in Fig. 17(a), while a zoomed
in view is shown in Fig. 17(b).
To validate the balanced distribution of ac current on the bus
bar E, the capacitor currents are measured under three different
connections, seen in Fig. 18, where the dc input terminals were
chosen as current injection points.
Using three Rogowski ac current probes, five capacitor Fig. 20. DC-link capacitor current with connection type II. (a) Currents in
currents are measured and presented in Figs. 19–21. The capacitor 1, 3, 5. (b) Currents in capacitor 2, 3, 4.
CALLEGARO et al.: BUS BAR DESIGN FOR HIGH-POWER INVERTERS 2365

V. CONCLUSION
A comprehensive bus bar design analysis is presented in this
paper. Based on different application requirements, packages of
power modules, and dc-link capacitors, five bus bar layouts are
designed. The current density, current distribution, and parasitic
parameters of each bus bar are evaluated by simulations and
experiments.
Fig. 21. DC-link capacitor current with connection type III. (a) Currents in 1) The current density of each bus bar is under the specifica-
capacitor 1, 3, 5. (b) Currents in capacitor 2, 3, 4. tion.
2) Based on the given power module locations, the current
distribution of dc component is defined by the positions
of dc input tabs, while that of ac component is related
to where the dc-link capacitors are located. In addition,
some intentional obstacles can be created on the bus bar
to achieve balanced current sharing.
3) Bus bar resistance and ohmic power loss are obtained.
It is proved that this resistance is usually in the order of
milliohms and the bus bar total ohmic loss is less than
Fig. 22. Experimental data plotted IGBT voltage and current curves during
turn-OFF transient. (a) When Ic = 101 A. (b) When Ic = 448.8 A. 10 W. Thus, comparing to the converter power loss, the
ohmic loss of bus bar is negligible.
TABLE VII 4) The bus bar inductance and capacitance are investigated.
ESTIMATED AND EXPERIMENTAL MEASURED VOLTAGE SPIKE According to the simulation results and measurements, the
predicted values of inductance and capacitance are close to
Parameters Case 1 Case 2 the measurements. The voltage spike caused by the stray
Calculated Experimental Calculated Experimental
inductance is estimated and experimentally validated.
In sum, a bus bar assembly (laminated) will typically have
I c (A) 208 208 306 306 a lower profile and use less space than wires or cables. The
V c e (V) 300 300 300 300
Δ t (ns) 115 115 140 140
lamination process utilizes a thin dielectric film to separate the
L stray (nH) 44.34 42.02 44.34 43 conductors. This thin and consistent gap results in lower and
V spike (V) 80.2 76 96.9 94 more consistent stray inductance. Along with inductance, the
V peak (V) 380.2 376 396.9 394
capacitance of the bus bar is also improved (increased). Espe-
cially in inverters, this lower inductance and high capacitance
might allow eliminating the snubber capacitors. In terms of ther-
experimental results show that a bus bar is an effective way mal management, the increased surface area of flat conductors
to distribute current between passive and active components. over wires means improved and more rapid cooling. Power den-
Considering the inverter implemented by bus-bar E, the sity is also a major requirement where the physical space taken
symmetry between capacitors and IGBT modules allows a up by a laminated bus bar is typically smaller than wires. In addi-
balanced operation in a three-phase balanced system; therefore, tion, the shape allows for tighter and improved use of the space.
the high-frequency current will also be balanced. Overall, this makes bus bars a clear choice when designing a
Finally, to verify the evaluated stray inductance in the circuit high power converter.
loop, double pulse test [52], [53] is implemented for bus bar E to
capture the IGBT switching transients. The data of voltage spike ACKNOWLEDGMENT
during turn-OFF transient is obtained experimentally, and it is
plotted in Fig. 22. Utilizing the calculation method illustrated in This research was undertaken thanks to funding from the
previous section, the estimated voltage spike and experimental Canada Excellence Research Chairs (CERC) Program. The au-
measured voltage spike are demonstrated in Table VII. To ensure thors would like to thank MERSEN for their technical contribu-
these results comparable, the estimated result are obtained based tion to this publication. The authors also gratefully acknowledge
on the experimentally measured current and transient time. ANSYS for their support with Maxwell and Q3D software, and
As presented by on Table V, the stray inductance of bus bar C CMC Microsystems for their support with Solidworks software.
is lower than the values of A and B. Bus bar C is expected to have REFERENCES
a higher stray inductance because its total length is bigger than
the other designs. However, bus bars A and B were designed [1] M. C. Caponet, F. Profumo, R. W. De Doncker, and A. Tenconi, “Low stray
inductance bus bar design and construction for good EMC performance
with rectangular terminals, resulting in a higher stray inductance in power electronic circuits,” IEEE Trans. Power Electron., vol. 17, no. 2,
value. Also, bus bar E has the same IGBT connection and also pp. 225–231, Mar. 2002.
presents small values for this quantity, as shown in Table III. [2] J. Wang, B. Yang, J. Zhao, Y. Deng, X. He, and Z. Xu, “Development of a
compact 750KVA three-phase NPC three-level universal inverter module
This represents an important point to be considered during the with specifically designed busbar,” in Proc. 25th Annu. IEEE Appl. Power
design process. Electron. Conf. Expo., 2010, no. 50737002, pp. 1266–1271.
2366 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 33, NO. 3, MARCH 2018

[3] J. M. Guichon et al., “Busbar design: How to spare nanohenries?” in [27] A. T. Bryant, K. K. Vadlapati, J. P. Starkey, A. P. Goldney, S. Y. Kandilidis,
Proc. Conf. Rec.—41st IAS Annu. Meeting (IEEE Ind. Appl. Soc.), 2006, and D. A. Hinchley, “Current distribution in high power laminated bus-
vol. 4, pp. 1865–1869. bars,” in Proc. 14th Eur. Conf. Power Electron. Appl., 2011, pp. 1–10.
[4] F. Zare and G. F. Ledwich, “Reduced layer planar busbar for voltage source [28] F. B. Leferink, “Inductance calculations; methods and equations,” in Proc.
inverters,” IEEE Trans. Power Electron., vol. 17, no. 4, pp. 508–516, Int. Symp. Electromagn. Compat., 1995, pp. 16–22.
Jul. 2002. [29] R. Bayerer, “Parasitic inductance hindering utilization of power devices,”
[5] F. Alkayal and J. B. Saada, “Compact three phase inverter in silicon carbide in Proc. 9th Int. Conf. Integr. Power Electron. Syst., Mar. 2016, pp. 1–8.
technology for auxiliary converter used in railway applications keywords [30] G. L. Skibinski and D. M. Divan, “Design methodology and modeling
advantages of sic power switches comparing to silicon ones for a power of low inductance planar bus structures,” in Proc. 5th Eur. Conf. Power
converter application,” in Proc. 15th Eur. Conf. Power Electron. Appl., Electron. Appl., 1993, pp. 98–105.
Lille, France, 2013, pp. 1–10. [31] X. Zhu, D. Su, Y. Zhang, and L. Wei, “Bus bar design for EMC perfor-
[6] C. Chen, X. Pei, Y. Chen, and Y. Kang, “Investigation, evaluation, and mance of power converters in fuel cell electric vehicles,” in Proc. IEEE
optimization of stray inductance in laminated busbar,” IEEE Trans. Power Int. Conf. Veh. Electron. Safety, 2006, pp. 144–147.
Electron., vol. 29, no. 7, pp. 3679–3693, Jul. 2014. [32] H. J. Beukes, J. H. R. Enslin, and R. Spee, “Busbar design considera-
[7] M. Khan, P. Magne, B. Bilgin, S. Wirasingha, and A. Emadi, “Lami- tions for high power IGBT converters,” in Proc. 28th Annu. IEEE Power
nated Busbar design criteria in power converters for electrified powertrain Electron. Spec. Conf., 1997, vol. 2, pp. 847–853.
applications,” in Proc. 2014 IEEE Transp. Electrific. Conf. Expo., 2014, [33] M. Ando, K. Wada, K. Takao, T. Kanai, S. Nishizawa, and H. Ohashi,
pp. 1–6. “Design and analysis of a bus bar structure for a medium voltage inverter,”
[8] D. O. Thomas, M. Sylvain, G. J. Michel, J. L. Schanen, and A. Perregaux, in Proc. 14th Eur. Conf. Power Electron. Appl., 2011, pp. 1–10.
“Reduction of conducted EMC using busbar stray elements,” in Proc. 24th [34] X. Deng et al., “Development of a compact 60 kw three-phase asymmetry
Annu. IEEE Appl. Power Electron. Conf. Expo., 2009, pp. 2028–2033. half-bridge power convertor with specifically designed busbar for switched
[9] R. T. Coneybeer, W. Z. Black, and R. A. Bush, “Steady-state and tran- reluctance machines,” in Proc. IEEE NW Russia Young Researchers Elect.
sient ampacity of bus bar,” IEEE Trans. Power Del., vol. 9, no. 4, Electron. Eng. Conf., 2015, pp. 311–316.
pp. 1822–1829, Oct. 1994. [35] H. Yu, Z. Zhao, T. Lu, L. Yuan, and S. Ji, “Laminated busbar design and
[10] L. Smirnova, R. Juntunen, K. Murashko, T. Musikka, and J. Pyrhonen, stray parameter analysis of three-level converter based on HVIGBT series
“Thermal analysis of the laminated busbar system of a multilevel con- connection,” in Proc. 2015 IEEE Appl. Power Electron. Conf. Expo., 2015,
verter,” IEEE Trans. Power Electron., vol. 31, no. 2, pp. 1479–1488, pp. 3201–3207.
Feb. 2016. [36] X. Zhang, H. Zhang, R.-W. Yu, and G.-J. Tan, “Planar bus bar optimum
[11] Design guide formulas. Mersen. [Online]. Available: http://www. design in high-power converters based on FEM analysis,” in Proc. 2nd
busbar.com/. Accessed on: Aug. 2, 2016. Int. Symp. Power Electron. Distrib. Gener. Syst., 2010, pp. 167–170.
[12] J. Wang, J. D. Lavers, and P. Zhou, “Modified surface impedance boundary [37] L. Liu, X. Cui, and L. Qi, “Simulation of electromagnetic transients of the
condition applied to eddy current problems,” IEEE Trans. Magn., vol. 28, bus bar in substation by the time-domain finite-element method,” IEEE
no. 2, pp. 1197–1200, Mar. 1992. Trans. Electromagn. Compat., vol. 51, no. 4, pp. 1017–1025, Nov. 2009.
[13] Bus bar solutions—Electrical power distribution. MERSEN. [Online]. [38] H. Wen and W. Xiao, “Design and optimization of laminated busbar
Available: http://ep-us.mersen.com/. Accessed on: Jul. 15, 2016. to reduce transient voltage spike,” in Proc. 2012 IEEE Int. Symp. Ind.
[14] Ddupont nomex paper type 410. Dupont. [Online]. Available: http:// Electron., 2012, pp. 1478–1483.
www.dupont.ca/. Accessed on: Jul. 2, 2016. [39] A. Abrishamifar, R. Lourakzadegan, R. Esmaili, and M. Arefian, “Design
[15] Pollution degree rating for electrical equipment. National instrumrnts. and construction of a busbar for spike reduction in an industrial inverter,”
[Online]. Available: http://www.ni.com/. Accessed on: Oct. 18, 2016. in Proc. 1st Power Electron. Drive Syst. Technol. Conf., 2010, pp. 13–17.
[16] J. Allocco, “Laminated bus bars for power system interconnects,” in Proc. [40] L. Popova et al., “Modelling of low inductive busbars for medium voltage
Appl. Power Electron. Conf., 1997, vol. 2, pp. 585–589. three-level NPC inverter,” in Proc. 2012 IEEE Power Electron. Mach.
[17] F. Luo, A. C. Baisden, and D. Boroyevich, “Design of a hybrid busbar Wind Appl., 2012, pp. 1–7.
filter combining a transmission-line busbar filter and a one-turn inductor [41] Z.-N. Ariga and K. Wada, “Laminated bus bar structure for low induced
for dc-fed three-phase motor drive systems,” IEEE Trans. Power Electron., noise,” in Proc. 7th Int. Conf. Integr. Power Electron. Syst., 2012, pp. 1–6.
vol. 28, no. 12, pp. 5588–5602, Dec. 2013. [42] A. Hino and K. Wada, “Resonance analysis for DC-side laminated bus-bar
[18] P. A. Dahono, Y. Sato, and T. Kataoka, “Analysis and minimization of of a high speed switching circuit,” in Proc. 2013 1st Int. Future Energy
ripple components of input current and voltage of PWM inverters,” IEEE Electron. Conf., 2013, pp. 751–756.
Trans. Ind. Appl., vol. 32, no. 4, pp. 945–950, Jul./Aug. 1996. [43] W. Xuesong, Z. Zhengming, and Y. Liqiang, “Conducted EMI reduc-
[19] J. Kolar and S. Round, “Analytical calculation of the RMS current stress tion in IGBT-based converters,” in Proc. 2010 Asia-Pacific Electromagn.
on the DC-link capacitor of voltage-PWM converter systems,” IEE Proc.— Compat., 2010, pp. 230–234.
Elect. Power Appl., vol. 153, no. 4, pp. 535–543, Jul. 2006. [44] G. Skibinski, J. Pankau, R. Sladky, and J. Campbell, “Generation, control
[20] X. Pei, W. Zhou, and Y. Kang, “Analysis and calculation of dc-link current and regulation of EMI from AC drives,” in Proc. 32nd IEEE-IAS Annu.
and voltage ripples for three-phase inverter with unbalanced load,” IEEE Meeeting Conf., 1997, pp. 1571–1583.
Trans. Power Electron., vol. 30, no. 10, pp. 5401–5412, Oct. 2015. [45] W. Xuesong, Z. Zhengming, and Y. Liqiang, “Conducted EMI reduc-
[21] X. Lu, H. Cha, and F. Z. Peng, “Optimizing capacitance in SPWM con- tion in IGBT-based converters,” Asia-Pacific International Symposium on
verter/inverter for series hybrid electric bus systems,” in Proc. 2009 IEEE Electromagnetic Compatibility, Beijing, 2010, pp. 230–234.
Veh. Power Propulsion Conf., 2009, pp. 852–858. [46] “Cispr11: Industrial, scientific and medical (ISM) radio-frequency equip-
[22] R. J. Pasterczyk, C. Martin, J.-M. Guichon, and J. L. Schanen, “Pla- ment electromagnetic disturbance characteristics limits and methods of
nar busbar optimization regarding current sharing and stray inductance measurement,” Int. Special Committee Radio Interference, 2004.
minimization,” in Proc. 2005 Eur. Conf. Power Electron. Appl., 2005, [47] S. Chen et al., “Towards EMI prediction of a pm motor drive for automotive
pp. 1–9. applications,” in Proc. 18th Annu. IEEE Appl. Power Electron. Conf.
[23] Z. Wang and G. Chen, “Study on planar busbar regarding stray inductance Expo., 2003, pp. 14–22.
minimization and oscillation suppression for high power converter,” in [48] H. Shim et al., “Inverter modeling including non-ideal IGBT characteris-
Proc. Int. Conf. Sustain. Power Gener. Supply, 2009, pp. 1–7. tics in hybrid electric vehicle for accurate EMI noise prediction,” in Proc.
[24] S. M. Wentworth, M. E. Baginski, D. L. Faircloth, S. M. Rao, and L. IEEE Int. Symp. Electromagn. Compat., 2015, pp. 691–695.
S. Riggs, “Calculating effective skin depth for thin conductive sheets,” [49] D. Cottet, I. Stevanovi, B. Wunsch, D. Daroui, J. Ekman, and G. Antonini,
in Proc. 2006 IEEE Antennas Propag. Soc. Int. Symp., Jul. 2006, “EM simulation of planar bus bars in multi-level power converters,” in
pp. 4845–4848. Proc. Int. Symp. Electromagn. Compat., 2012, pp. 1–6.
[25] T. Hiraoka, T. Tokumitsu, and M. Aikawa, “Very small wide-band MMIC [50] M. Jones and A. W. Kelley, “Wideband circuit model for busbar
magic T’s using microstrip lines on a thin dielectric film,” IEEE Trans. impedance,” in Proc. 15th Annu. IEEE Appl. Power Electron. Conf. Expo.,
Microw. Theory Techn., vol. 37, no. 10, pp. 1569–1575, Oct. 1989. 2000, pp. 839–845.
[26] F. Zare and G. Ledwich, “A plane busbar impedance calculation using [51] H. Zhu, A. R. Hefner, and J. S. Lai, “Characterization of power electronics
Maxwell’s equations,” in Proc. 11th Int. Symp. High Voltage Eng., 1999, system interconnect parasitics using time domain reflectometry,” IEEE
vol. 2, pp. 148–151. Trans. Power Electron., vol. 14, no. 4, pp. 622–628, Jul. 1999.
CALLEGARO et al.: BUS BAR DESIGN FOR HIGH-POWER INVERTERS 2367

[52] J. Guo, H. Ge, J. Ye, and A. Emadi, “Improved method for MOSFET Matthias Preindl (S’12–M’15) received the B.Sc.
voltage rise-time and fall-time estimation in inverter switching loss calcu- degree in electrical engineering (summa cum laude)
lation,” in Proc. Transp. Electrific. Conf. Expo., Jun. 2015, pp. 1–6. from the University of Padua, Padua, Italy, the M.Sc.
[53] R. Bayerer and D. Domes, “Power circuit design for clean switching,” in degree in electrical engineering and information tech-
Proc. 2010 6th Int. Conf. Integr. Power Electron. Syst., Mar. 2010, pp. 1–6. nology from ETH Zurich, Zurich, Switzerland, and
the Ph.D. degree in energy engineering from the Uni-
Alan Dorneles Callegaro (S’15) received the B.Sc. versity of Padua, in 2008, 2010, and 2014, respec-
and M.Sc. degrees in electrical engineering from Fed- tively.
eral University of Santa Catarina (UFSC), Florianop- He was an R&D Engineer of power electronics and
olis, Brazil, in 2011 and 2013, respectively. In 2015, drives at Leitwind AG, Sterzing, Italy (2010–2012),
he joined the McMaster Automotive Resource Cen- a Postdoctoral Research Associate with the McMas-
tre, McMaster University, Hamilton, Canada, where ter Institute for Automotive Research and Technology, McMaster University,
he is currently working toward the Ph.D. degree in Hamilton, ON, Canada (2014–2015), and a Sessional Professor in the Depart-
electrical engineering at the Canada Excellence Re- ment of Electrical and Computer Engineering, McMaster University (2015). He
search Chair in Hybrid Power train Program. is currently an Assistant Professor in the Department of Electrical Engineering,
He was with the Power Electronics Institute, Flo- Columbia University in the City of New York, NY, USA.
rianopolis, Brazil, from 2013 to 2014. His research Dr. Preindl received the Career Award of the Futura Foundation in South
interests include high-power inverters, switched reluctance machines, noise and Tyrol, Italy, and the CAREER Award of the US National Science Foundation in
vibration analysis of traction motors, and motor control. 2016 and 2017, respectively.

Berker Bilgin (S’09–M’12–SM’16) received the

Ph.D. degree in electrical engineering from Illinois
Institute of Technology, Chicago, IL, USA. He is
currently working toward the MBA degree at the De-
Groote School of Business, McMaster University.
Jing Guo (S’15) received the B.S., M.S., and Ph.D. He is the Research Program Manager in Canada
degrees in electrical engineering from Beijing Jiao
Excellence Research Chair in Hybrid Powertrain Pro-
Tong University, Beijing, China, in 2009, Illinois
gram in McMaster Institute for Automotive Research
Institute of Technology (IIT), Chicago, IL, USA,
and Technology, McMaster University, Hamilton,
in 2012 and McMaster University, Hamilton, ON, ON, Canada. He is managing many multidisciplinary
Canada, in 2017, respectively.
projects on the design of electric machines, power
She was a Research Assistant at the Electrical and
electronics, electric motor drives, and electrified powertrains.
Computer Engineering Department, IIT from Octo-
Dr. Bilgin was the General Chair of the 2016 IEEE Transportation Electrifi-
ber 2012 to December 2012. Currently, she is a post cation Conference and Expo.
doctoral fellow in the Program of the Canada Ex-
cellence Research Chair in Hybrid Powertrain at Mc-
Master Automotive Resource Centre, Hamilton, ON, Canada. Her main research
interests include power switch modeling and inverter design. Ali Emadi (S’98–M’00–SM’03–F’13) received the
B.S. and M.S. degrees in electrical engineering (high-
est distinction) from Sharif University of Technology,
Michael Eull (S’12) photographs and biographies not available at the time of
publication. Tehran, Iran, in 1995 and 1997, respectively, and the
Ph.D. degree in electrical engineering from Texas
A&M University, College Station, TX, USA, in 2000.
He is the Canada Excellence Research Chair in
Benjamin Danen (S’17) received the Bachelor’s of Hybrid Powertrain, McMaster University, Hamilton,
Engineering and Master’s of Engineering in electri- ON, Canada. Before joining McMaster University, he
cal engineering from McMaster University, Hamil- was the Harris Perlstein Endowed Chair Professor of
ton, ON, Canada, in 2014 and 2016, respectively. engineering and the Director of the Electric Power
He joined MARC in September 2014 as a Re- and Power Electronics Center and Grainger Laboratories, Illinois Institute of
search Assistant for the Canada Excellence Research Technology in Chicago, IL, USA, where he established research and teaching
Chair in Hybrid Powertrain Program, where he cur- facilities as well as courses in power electronics, motor drives, and vehicular
rently serves as a Research Engineer. His areas of power systems. He was the Founder, Chairman, and President of Hybrid Electric
research include motor control, system modeling and Vehicle Technologies, Inc., a university spin-off company of Illinois Tech. He
power electronics. is the principal author/coauthor of more than 400 journal and conference papers
as well as several books including Vehicular Electric Power Systems (2003),
Energy Efficient Electric Motors (2004), Uninterruptible Power Supplies and
Active Filters (2004), Modern Electric, Hybrid Electric, and Fuel Cell Vehicles
Jason Gibson received the Associates degrees in en-
(2nd ed, 2009), and Integrated Power Electronic Converters and Digital Control
gineering, the Bachelor’s degree in business admin-
(2009). He is also the editor of the Handbook of Automotive Power Electronics
istration, and the Executive Master’s degree in busi-
and Motor Drives (2005) and Advanced Electric Drive Vehicles (2014).
ness administration. He recently received the Exec-
Dr. Emadi was the Inaugural General Chair of the 2012 IEEE Transportation
utive MBA degree in Business Administration and
Electrification Conference and Expo and has chaired several IEEE and SAE
Management from Rochester Institute of Technology conferences in the areas of vehicle power and propulsion. He is the founding
(RIT).
Editor-in-Chief of the IEEE TRANSACTIONS ON TRANSPORTATION ELECTRIFI-
He is shifting his professional focus to Product
CATION. He received numerous awards and recognitions. He was the advisor
Management and executive leadership. He has 20
for the Formula Hybrid Teams at Illinois Tech and McMaster University, which
years of experience in custom engineered product received the GM Best Engineered Hybrid System Award at the 2010, 2013, and
design in technologies ranging from bus bars, forg-
2015 competitions.
ings, plastic injection molding, solenoids, and valves.