Development of a Zero-Voltage-Transition Bidirectional DC-DC Converter for a
Brushless DC Machine EV Propulsion System
Xinxiang Yan and Allan Seckold
Tactical Technologies Pty Ltd.
PO Box 6547 Baulkham Hill BC
NSW 2153, Australia
Tel: 61-2-9680 9088, Fax: 61-2-9680 9099
Email: xinxiang.yan@ieee.org
Abstract - This paper presents the development of a zero-
voltage-transition (ZVT) bidirectional dc-dc converter for a
brushless dc machine (BLDCM) electric vehicle (EV)propulsion
system. The dc-dc converter serves two main purposes. Firstly, it
is used as the power electronics interface required for load
leveling EV batteries. By employing high power ultracapacitors,
the battery energy density and peak power density requirements
are decoupled, offering opportunities to design and use EV
batteries that are optimized for energy density, life and low cost,
with less attention being given to the peak power. Secondly, the
dc-dc converter is used to achieve speed extension and constant
power operation of the BLDCM. To determine how well the dc-
dc converter works together with the inverter l motor, and to
develop a proper control method for the whole propulsion
system, mathematical analysis and MatlablSimulink simulations
were carried out based on control theories. The zero-voltage-
transition (ZVT) technique has been adopted in the bidirectional
dc-dc converter to solve the problems associated with high
frequency operation of PWM power converters, such as high
switching loss resulting in low efficiency thus requiring larger
heat sinks, and severe EMI noise. The final ZVT bidirectional
dc-de converter had an efficiency of greater than 96%.
- POWER TOPOLOGY AND OVERALL POWER
I. BACKGROUND
CONTROL OF AN EV PROPULSION SYSTEM
Of all the sub-systems used in an electric vehicle (EV), the
battery remains the main barrier to success. The challenge is
that design choices for energy density, power density, lifetime,
weight, volume, and cost of the battery are all compromises.
The three leading performance criteria of a battery, energy,
power and life, are inextricably linked within an “eternal
triangle”. Improvement in any one comes at the expense of
one or both of the others.
Employing high power density ultracapacitors (UC) in EV
drivelines can significantly reduce the peak power
requirement for the battery. The peak power during
acceleration is supplied by the ultracapacitors while the
batteries supply the continuous power for normal driving.
This offers opportunities to design and use EV batteries that
are optimized for energy density, life and low cost with less
attention being given to the peak power. The example shown
in [l] uses Zinc Bromine batteries, which are optimized for
energy density, lifetime and low cost. A power electronics
interface is required for the battery load leveling purpose.
0-7803-7262-X/02/$10.00Q 2002 EEE.
Authorized licensed use limited to: Tec de Monterrey. Downloaded on July 25, 2009 at 14:57 from IEEE Xplore. Restrictions apply.
Dean Patterson
Department of Electrical Engineering
University of South Carolina
Columbia SC 29208, USA
Tel: (803) 777-7362, Fax: (803) 777-8045
Email: patterson@ieee.org
An axial flux permanent magnet brushless dc machine
(BLDCM) is employed in the EV in [l] owing to its very high
efficiency, high reliability and easy control. However, the
natural “constant torque” torque-speed characteristic of a
BLDCM and its speed limitation determined by battery voltage
do not suit usual traction applications. The desired torque-
speed characteristic for traction applications is constant torque
below the base speed of the motor and constant power above
the base speed, which implies a reduction in the achievable
torque as speed increases. This torque-speed characteristic
minimizes the power rating of the motor and controller and
consequently, the cost, weight, and volume of the whole
propulsion system. To achieve extension of speed range and
constant power operation of a BLDCM, flux-weakening
control is one method. However flux-weakening methods for a
BLDCM result in poor efficiency and require complicated rotor
design and fabrication. By boosting the battery voltage to a
higher value through a dc-dc converter, the speed range of a
BLDCM can be extended. While the motor voltage increases,
the motor current decreases. Therefore, constant power
operation above the base speed can be achieved. As
regeneration operation is also required, the dc-dc converter
must be bidirectional.
It can been seen that both the EV battery load-leveling
scheme and the constant power operation of BLDCM at speeds
above the base speed require a bidirectional dc-dc converter.
However, the EV battery load-leveling scheme employing an
ultracapacitor bank requires a bidirectional dc-dc converter
with the ultracapacitors connected to its lower-voltage side and
the battery to its higher-voltage side, while the constant power
operation of the BLDCM at speeds above the base speed
requires another bidirectional dc-dc converter with the battery
connected to the lower-voltage side and the motor / inverter to
the higher-voltage side.
To avoid employing two dc-dc converters, a novel
topology of a dc-dc converter was proposed by the authors in
[l]. Fig. 1 shows its power stage schematic. It consists of a
simple bidirectional dc-dc converter, a Schottky diode D, an
SCR and a changeover contactor, three of which are used to
move the battery from the higher-voltage side to the lower-
voltage side of the converter for constant power operation of
the BLDCM. Since the dc-dc converter and ultracapacitors
1661
 L
Fig. I Topology of the multifunctionaldc-dc converter used in [I]
t J 'Maximum/Torque ' Fig. 3 Control diagram of the bi-directional dc-dc converterfor the purpose of
speed
Fig. 2 The concept of the overall power management
work only for a short term, they together are called a Short
Term Power Unit (STPU).
The overall power control scheme for all speeds is
illustrated in Fig. 2. At very low speeds, all the power comes
from the battery. In this mode, high torque can be obtained,
since the required power, which is proportional to motor
speed, is within the output power capability of the battery.
Above a threshold motor speed, nl, the STPU is activated to
provide additional power when high acceleration is required.
While the shaded triangle shows the extra power required for
high acceleration, the STPU actually provides the power
shown in the hatched rectangle and the battery provides the
remainder of the power, in order to simplify the control. The
operation of the STPU is disabled at speed n2, due to the upper
power limit of the STPU. At motor speeds above the base
speed, nb, the battery voltage is boosted to a higher value for
speed extension and constant power operation of the motor.
In the whole speed range, from zero speed to maximum
speed n, the maximum torque envelope is shown as a dashed
line. This maximum torque-speed envelope is similar to the
conventional torque-speed characteristic of an internal
combustion engine (ICE) with a transmission. That is, the
motor provides high torque at low speeds and lower torque at
high speeds, which is familiar to ICE vehicle drivers.
11. SIMULATION RESULTS OF THE VOLTAGE AND CURRENT
TRANSIENTS IN THE SHORT TERM POWER UNIT WHILST LOAD
LEVELING THE BA'RERIES
In establishing the effectiveness of the STPU, simulations
on MatlabBimulink were carried out to observe the voltage
1662
Authorized licensed use limited to: Tec de Monterrey. Downloaded on July 25, 2009 at 14:57 from IEEE Xplore. Restrictions apply.
1 ..^*
load leveling the batteries
and current transients of the STPU. Fig. 3 shows the control
diagram of the bi-directional dc-dc converter for the purpose
of load leveling the batteries. While the dc-dc converter is
working, the ultracapacitor current Idc1 (the inductor current)
is kept constant. Therefore, a negative feedback loop is closed
on the inductor current. The current source IJOad represents
the inverter motor combination which is current-controlled. A
PI compensation network has been designed for the current
control. The bus capacitor of the inverter, cbm, is considered
as part of the STPU. IdCz is defined as the output current of
the STPU in boost mode when it is positive and as the input
current of the STPU in buck mode when it is negative.
A. STPU Voltage and Current Transients in Boost Mode
Fig. 4 shows the simulated battery current, bus voltage,
and output current of the STPU in boost mode. When the
inverter current is +40 A, only the batteries provide the power
and the STPU is disabled. At an inverter current demand of
+120 A, the STPU is activated, and both the STPU and the
batteries provide power to the motor I inverter. When the
ultracapacitor voltage is 90% of full voltage, and the
ultracapacitor current is set to 150 A, the STPU provides 80.5
A and the batteries 39.5 A, as shown in Fig. 4(a). With the
same ultracapacitor current setting but at 70% of full voltage,
the STPU provides less current, being 61 A, and the batteries
provide more, being 59 A, as shown in Fig. 4(b). In both cases,
transients take about 2 ms and the variation and ripple of the
bus voltage are very small.
B. STPU Voltage and Current Transients in Buck Mode
The voltage and current transients of the STPU in buck
mode are similar to those in boost mode, except that all the
currents in buck mode are in the opposite directions and
become negative in value.
Fig. 5 shows the battery current, bus voltage, and input
current of the STPU in buck mode. When the ultracapacitor
voltage is 70% of full voltage, and the ultracapacitorcurrent
 (b)
Fig. 4 Battery current, bus voltage and currents of the STPU in boost mode
Step load initial value -40A, final value - 120A
Current command in the ultracapacitors (the inductor): 150A
Trace 1: bus voltage
Trace 2: current command in the motor I inverter
Trace 3: output current of the STPU
Trace 4:battery current
(a) Ultracapacitor voltage is 82.8V, 90% of full voltage
(b) Ultracapacitor voltage is 64.4V, 70% of full voltage
is set to -150 A, the STPU absorbs 73 A of 120 A regenerative
current and the batteries take the rest (47 A), as shown in Fig.
5(a). With the same ultracapacitor current setting, but at 90%
of full voltage, the STPU absorbs more regenerative current,
95 A, and the batteries take 25 A, as shown in Fig. 5@). In
both cases, transients take about 2 ms and the bus voltage is
very smooth.
The simulation results of voltage and current transients in
the STPU in both boost mode and buck mode have shown the
excellent dynamic performance of the STPU. These, along
with the high efficiency of the STPU which will be presented
later, have demonstrated the effectiveness of the STPU.
111. DISCUSSION
OF THE CONTROL OF THE COMPLETE DRIVE
SYSTEM
This section studies two different current control methods
of the BLDCM at speeds above the base speed when using a
boost converter for speed extension. The study focuses on
1663
Authorized licensed use limited to: Tec de Monterrey. Downloaded on July 25, 2009 at 14:57 from IEEE Xplore. Restrictions apply.
(b)
Fig. 5 Battery current, bus voltage and currents of the STPU in buck mode
Regenerative current: -120A
Current command in the ultracapacitors (the inductor): -150A
Trace 1: bus voltage
Trace 2: battery current
Trace 3: current taken by the STPU
Trace 4: current command in the motor I inverter
(a) Ultracapacitorvoltage is 64.4V, 70% of full voltage
(b) Ultracapacitorvoltage is 82.8V, 90% of fullvoltage
control system stability analysis in the s-plane when using the
two dfferent methods.
A. Method I: The Converter - Current controlled; The Inverter
- Commutation Only
Since a boost converter is employed, the question arises,
"can we use the boost converter to regulate the motor current
while we let the inverter perform commutation only in order
to minimize the switching loss?" Allured by the minimization
of switching loss, the author in [2] tried to implement this
control scheme but did not succeed. This control issue will be
investigated below.
As the inverter only performs commutation, it can be
neglected in modeling the control system. In a 60" interval,
the three-phase BLDCM is equivalent to a dc motor, as shown
in Fig. 6. The system consists of two current loops. The
current reference of the inner loop is obtained from the output
of the PI controller of the outer loop. In the inner current
loop, since there is only one energy storage component L1 that Considering
implies a first order pole in the left half plane (LHP),very
good control performance can be readily achieved. Then the AVbuso= d(La -
diu +iaRa) = (sL, + Ra)Aia (3 -4)
transfer function of the inner closed current loop, Gil(s), can dt
be optimized as a small inertia element. For fuaher and simplifying Equation (3.3) give
simplicity, we assume that the inductor current follows its
reference perfectly. Thus we have 2
SLlZinobin = Aia [Lac~svbusoS
Gil(~) 1. (3.1) + (RaCbusVbus + LazaO)s + 2Raza0 + EaOl a (3.5)

The motor current control loop is then simplified as shown in Therefore, we have
Fig. 7 in which &i(s) is the transfer function of the converter.
Now we go back to Fig. 6 to derive &i(s). IC&)=-- Aia - -Ll'inOS-24nZinO +Vino

Aiin Lacburvb-Sz +(RacbmvbbUso +L,Ja,)s+2Ra~,ofEao

Assume that the switch and diode are ideal (no loss and no
energy storage). The instantaneous power equation relating = as+b (3.6)
cs2 +ds+e
input and output quantities of the system is obtained by
differentiating the energy quantities where
-i:Rin --(-Llik)
d 1 = -((-Chvbus)
d 1 2 a = -Lllino
dt 2 dt 2
d 1
+- (- L,i; ) + i; R, + iaEa *
(3.2) ~ b = -2RinZin0 + Vino
dt 2
C = LaCbusVbusO
Linearizing the power equation around the operation point
(Vi,o, L o , Ea09 and Iao) yields = RaCbusVbusO + LazaO
5, Aiin - 2RinZin0Aiin- sLIIinoAiin = s ~ b u s v b u s o ~ v b w e = 2RaZa0 + Ea0. (3.11)
+ sLaZaoAia + 2RaZa0Aia + EaOAia. (3.3)
At the rated operation point,

Vinolino -1zLRin = z a o ~ a o+ ~ 2* a (3.12)

From Equation (3.12), we can find I ~ O With

. the false solution
rejected, we have

(3.13)

In addition,
BLDCM
Modulation Equivalent vbusO = EaO -k IaORa * (3.14)
c i t

Using the values of the system, Vbo = 120V, Eao = 170V, Ia0 =
80A, Ri, = O.O60hm, R, = O.O260hm, Cbus = 470oUF, La =
182uH, L1 = 4 W ,we have Iho = 122.18A, vb,&= 172.6V,
and
' L d
Kci (SI = -0.002224~+105.34 . (3.15)
0 . 0 0 0 1 4 7 ~+0.03559s+174.16
~
Fig 6 Motor currentcontrol Method I
Ki(s)contains a zero and two poles:
Inductor Current z1 = 4 . 7 3 6 5 ~ 1 0 ~
Template
IQ-ref p1 = (-O.1143+1.0825i)x1O3
+
p2 = (-0.1143 -1.0825i)x103

Fig. 7 Motor current control loop with unspecifiedcontrol plant using Fig. 8 shows the root locus of Ki(s). As can be seen, the
Method I two poles are a complex conjugate pair that are very close to

1664

Authorized licensed use limited to: Tec de Monterrey. Downloaded on July 25, 2009 at 14:57 from IEEE Xplore. Restrictions apply.
 0000 , makes the control much easier. Excellent performance of the
I current control of the BLDCM can be easily achieved with a
4000. current-regulated voltage source inverter [3]. Based on this
fact, for the purpose of the analysis and design of the boost
2000 -
converter, the current-regulated inverterlmotor can be modeled
B 0 .
as a dc current source Io. Then the power system is simplified
E as illustrated in Fig. 9.
-2000 - Similar analysis based on energy balance of the power
system, as performed in Section A, has shown that proper
-4000 - design of the PI controller results in that the closed-loop
voltage transfer function only or nearly has a first-order pole in
the LHP, suggesting a stable system. Therefore, this control
method has been used in this EV propulsion system.
Fig. 8 Root locus of K&)
Iv.ZERO-VOLTAGE-TRANSITIONIMPLEMENTATION OF THE
the imaginary axis and far from the real axis. The root locus BIDIRECTIONAL DC-DC CONVERTER
strongly suggests that little margin is left for designing a
stable and high-performance closed-loop system for the Soft-switching technologies have been good solutions for
purpose of motor current control. No matter whether a the problems associated with high frequency operation of
proportional-integral (PI) or a proportional-intepl- PWM power converters, such as high switching loss resulting
differential (PID) controller is used, it is very difficult to effect in low efficiency thus requiring larger heat sinks, and severe
satisfactory compensation. EM1 noise.
The control difficulty mainly originates from the existence Among various soft-switching techniques, the zero-
of the big capacitor CbW. If Cbus is decreased, the poles would voltage-transition (ZVT) [4-61 technique is favorable for this
move away from the imaginary axis and towards the real axis. bidirectional dc-dc application and has been adopted, based on
Unfortunately, a big Cbus is preferable to absorb ripple current two main reasons. Firstly, ZVT realizes soft switching in a
and make the bus voltage stable. PWM converter without imposing additional switching voltage
and current stress and conduction loss, allowing the use of
B. Method ZZ: The Converter - Voltage Source Output; The lower voltage power devices, which usually have better
Inverter - Current Mode Control conduction and switching performance than higher voltage
power devices. The use of the lower voltage power devices
An alternative method is to control the inverter and the dc- contributes to the high efficiency. Secondly, because the
dc converter separately, that is, the boost converter is auxiliary ZVT circuits are not in the main power path and are
controlled as a voltage source output and the inverter is actively controlled, the capability of bidirectional operation of
current regulated with a voltage source input. the basic dc-dc converter remains, which is essential for this
The big capacitor in between decouples the two stages and application.
Fig. 10 shows the circuit diagram of the ZVT bidirectional
dc-dc converter. Symbols with lower case subscript ‘x’
represent the additional components for ZVT boost operation,
and symbols with lower case subscript ‘y’ represent the
additional components for ZVT buck operation, while symbols
with lower case subscript ‘xy’ represent the additional common
IO j components for both ZVT boost operation and ZVT buck
operation. The fundamental principles of ZVT converters are

1
Equivalent
circuit of
Modulation Motor/
& Driver inverter

vin c

11.
‘Vbueref 5
Fig. 9 Equivalent system diagram using Method I1 of the motor current
I I
control at the motor speeds above the base speed Fig. 10 Diagram of the ZVT bidirectional dc-dc converter

1665

Authorized licensed use limited to: Tec de Monterrey. Downloaded on July 25, 2009 at 14:57 from IEEE Xplore. Restrictions apply.
 r
shown in [4-61.
The ZVT bidirectional dc-dc converter has been designed,
constructed and tested. The experimental results are
satisfactory. ZVT operation in both boost mode and buck
mode has been achieved and the efficiencies are very high.
As an example of the experimental results, Fig. 11 and Fig. 12
show the oscilloscope waveforms on the main power switch i 500 1500 2500 3500 I
!
ootpla Power gu)
and auxiliary switch in boost mode. The current waveforms
of the power switches and diode were not measured as it was
not convenient to measure them due to the compact layout of Fig. 14 Measured efficiency vs. output power in buck mode
the prototype. However zero-voltage transition for both turn
on and turn off of the main boost switch SI and the auxiliary achieved in both the operation modes.
boost switch S, can still be clearly observed from Fig. 11 and
Fig. 12. V. CONCLUSIONS
Fig. 13 and Fig. 14 show the measured efficiencies vs.
output power in boost mode and buck mode respectively. To achieve a desirable torque-speed characteristic in an
96% efficiency over a wide output power range has been electric vehicle system employing a BLDCM, a bidirectional
dc-dc converter is used for load leveling EV batteries, and
=& JZlrOTrlp:d_ M Par: 9.000#r MEASURE speed extension and constant power operation of the BLDCM.
Matlab/Simulink simulations and mathematical stability
analysis are carried out and have proven the effectiveness of
.. .. .. .. employing the bidirectional dc-dc converter in the EV
propulsion system. The ZVT technique has been adopted in
the bidirectional dc-dc converter to solve the problems
associated with high frequency operation of PWM power
.. .. . . . converters, such as high switching loss resulting in low
L....:.... ....:... .........
. . . . efficiency thus requiring larger heat sinks, and severe EMI
M1 20.0V
_.I

CH2 5 M V
_yL ..............
M5 . 0 0 ~ ~
YLL.-YL!.LY.)* .
CM 1BAUV noise. An efficiency of 96% has been achieved on the
constructed ZVT bidirectional dc-dc converter.
Fig. 11 Waveforms of V,, and Vbl
showing soft switching operation of the main boost switch REFERENCES

-
[ 11 X.Yan and D. Patterson, “Improvement of drive range, acceleration and

I-]
deceleration performance in an electric vehicle propulsion system,” The
IEEE Power Electronics Specialist Conference (PESC’99), South
1 Carolina, USA, 1999.
. . :/.:.v:.,:.;,
. r-; f [2] J. Ofori-Tenkorong, “Permanent-magnet synchronous motors and
associated power electronics for direct-drive vehicle propulsion,” Ph.D.
.......... .........................
. .
m thesis, Department of Electrical Engineering, Massachusetts Institute of
. . . . .
:
I,,L
.. ..
i . :. . .. .A. v, . .
; 1
... ... . . . . ;
:
..*
,”q [3]
Technology, 1996
D. Patterson, X.Yan,and S. Camillen, “A very high efficiency controller
-a for an axial flux permanent magnet wheel drive in a solar powered
[ vehicle,” The IEEE International Conference on Power Electronics
Drives and Energy Systems for Industrial Growth (PEDES98), Perth,
Australia, 1998.
[4] R. Streit and Daniel Tolhk, “High efficiency telecom rectifier using a
Fig. 12 waveforms of V, and V d novel soft-switched boost-based input current shaper,” The 13th

;’::TI
showing soft switching operation of the auxiliary boost switch International Telecommunication Energy Conference (INTEEC’91),
1991.
[5] S. Ben-Yaakov, Gregory Ivensky, Oleg Levitin, and Alexander Treiner,
“Optimization of the auxiliary switch components in a flying capacitor
ZVS converter,”APEC‘95, Dallas, TX, USA, 1995.
[6] G. Hua and F. C. Lee, “Soft-switching techniques in PWM converters,”
IEEE Transactions on Industrial Electronics, vol. 42, 1995.
;. 94
E
92

90
LO00 2000 3000 4000 5000

ouiput Power gv)

Fig. 13 Measured efficiency vs. output power in boost mode

1666

Authorized licensed use limited to: Tec de Monterrey. Downloaded on July 25, 2009 at 14:57 from IEEE Xplore. Restrictions apply.