A Unity Power Factor Converter Using Half
A Unity Power Factor Converter Using Half
3, MAY 1998
487
Abstract A single-phase high-efficiency near-unity powerfactor (PF) half-bridge boost converter circuit, which has been
proposed earlier by other researchers, is presented with detailed
analysis. This converter is capable of operating under variable
PF. However, the focus of this paper is in achieving unity PF
operation only. The efficiency of this circuit is high because there
is only one series semiconductor on-state voltage drop at any
instant. The existence of an imbalance in the voltages of the
two dc-link capacitors, which was noted before, is confirmed
here. The cause for the imbalance is analyzed using appropriate
models, and a control method to eliminate it is discussed in detail.
Analysis and design considerations for the power circuit using
the fixed-band hysteresis current control (HCC) technique are
provided. The analytical results are verified through simulation
using switched and averaged circuit models of the scheme and
also through experimental work. At 90-V ac input and 300W 300-V output, the experimental prototype demonstrates an
efficiency of 96.23% and a PF of 0.998. This converter, with
its relatively high dc-output voltage, is well suited for the 110-V
utility supply system. A circuit modification for universal input
voltage range operation is also suggested.
Index TermsHalf-bridge boost topology, IEC 1000-3-2, power
factor correction, switch-mode rectifier.
I. INTRODUCTION
488
(a)
(b)
(c)
Fig. 1. Circuit variations of the boost PFC topology.
SRINIVASAN AND ORUGANTI: UNITY POWER FACTOR CONVERTER USING BOOST TOPOLOGY
(a)
489
(b)
Fig. 4. Current waveforms in HCC technique. (a) Over one line cycle. (b) Over one switching cycle.
(a)
(b)
Fig. 5. The two modes of operation of the half-bridge circuit. (a) Mode 1: positive inductor-current-slope mode. (b) Mode 2: negative inductor-current-slope mode.
490
time and the memory storage requirement for the output are
significantly less than in the simulation using actual switches.
Due to these, unlike with a switched model, the averaged
model simulation can be easily run several times to optimize
converter and controller performance.
The averaged model approach has been verified to give correct results for numerous systems including the one discussed
in this paper (see Section V-B2). It is found to be very useful
in closed-loop compensator design.
Despite the drawbacks in the simulation of the switched
model, it is still useful and necessary in the study of switching
transients and in the determination of quantities such as
peak voltage and current stress of the device, duty ratio,
and switching frequency variations. Thus, the two models
complement each other in their usefulness. The system under
study was simulated using both the averaged and switched
models.
2) Computation of Switching Frequency and Duty Cycle
(Using Switched Model): Referring to Figs. 4(b) and 5, the
expressions for
and
are given by
(3)
(4)
The expression for the switching frequency is therefore given
by
where
(5)
and
maximum switching frequency
It can be shown that
(6)
occurs at an angle
(7)
for
where
etc.
(8)
or
for
where
etc.
(9)
SRINIVASAN AND ORUGANTI: UNITY POWER FACTOR CONVERTER USING BOOST TOPOLOGY
491
(a)
(b)
Fig. 8. Variation of switching frequency and duty cycle over a line cycle ( = 0:5). (a) Switching frequency. (b) Duty cycle.
(a)
(b)
Fig. 9. Ripple voltage waveforms across capacitors ( = 0:5) across (a) C1 and (b) C2.
For a special case of equal to 0.5, both (8) and (9) will result
in the same value as may be expected. Fig. 8(a) shows the
variation of the normalized switching frequency
over a line cycle for equal to 0.5. From (3) and (4), the duty
cycle of switch S1 is given by
(10)
for
over a line
Fig. 8(b) shows the variation of
cycle. The duty cycle is maximum at the negative peak of the
line voltage and minimum at the positive peak. The maximum
and minimum duty cycles are given by
(17)
The low-frequency capacitor ripple voltages
and
are
(18)
(11)
(19)
(12)
The duty cycle
(13)
From (10) and (13), the average duty ratios of S1 and S2 are
and , respectively.
3) Circuit Analysis Using Averaged Model:
(20)
492
Fig. 10.
(b)
Fig. 12. Variations of the diode and switch average currents with
= 0:5).
(for
Fig. 11.
= 0:48.
(21)
From (18) and (19), the overall output voltage ripple
given by
is
(22)
In Fig. 10, the normalized ripple voltage waveforms are
shown for
equal to 3.23 and for
equal to 0.5. The
output voltage ripple is small compared to the ripples across
the individual capacitors. It may be noted that for values of
different from 0.5, the output voltage ripple is not a pure
second-harmonic as in conventional boost PFCs. This is seen
from the waveform in Fig. 11, which is plotted based on (22)
for equal to 0.48. The line-frequency oscillations in Fig. 11
is also confirmed through simulations and experiments (Section IV). In closed-loop control, the line-frequency oscillations
at the output can increase the total harmonic distortion (THD)
of the input current.
b) Range of values for : As mentioned earlier, the voltages across C1 and C2 must each be greater than the input
voltage over the line cycle. Neglecting all ripple voltages for
C1
(23)
SRINIVASAN AND ORUGANTI: UNITY POWER FACTOR CONVERTER USING BOOST TOPOLOGY
493
TABLE II
POWER CIRCUIT PARAMETERS AND COMPONENT VALUES
DESIGN SPECIFICATIONS
TABLE I
HALF-BRIDGE BOOST PFC CIRCUIT
FOR THE
that the overall conduction loss can be quite small since the
conduction voltage drop of the diode is usually much smaller
than that of the active switch.
III. CIRCUIT IMPLEMENTATION
The implemented single-phase half-bridge converter is
shown in Fig. 13. The design specifications are given in
Table I. Aluminum electrolytic capacitors of value 1000 F
(400 V) with low equivalent-series resistance (ESR) of 80
m each were used for the capacitors C1 and C2. The boost
inductor was constructed from two core sets of Philips EC-70
core (material grade 3C8 ferrite), and the air-gap width was
experimentally adjusted to obtain an inductance of 1.27 mH.
An IGBT allows a fast recovery diode to be externally paralleled across it. Also, IGBTs on-voltage drop is low. Hence,
the IGBT (IRGBC40U) and diode (MUR1560) mounted on
suitable heat sinks were used.
The Unitrode UC3706 driver integrated circuit (IC) fed
by an isolated 15-V supply was used for the gate drive.
The output voltage feedback is obtained through a resistive
divider, while the line current was sensed using a Hall-effect
sensor (LEM module LA 50-P/SP 1). The sinusoidal signal for
shaping the input current was obtained from the line through
a small step-down transformer. The hysteresis band
was
set at 1.7 A. This results in about 20% ripple in the line
current at full-load power. The measured value of
in the
implementation was 1.67 A. The summary of the important
power circuit parameters and component values are presented
in Table II. It must be noted that the experimental circuit is
494
(a)
Fig. 14.
(a)
Fig. 15.
(b)
Input current waveform: (a) simulated (switched model) and (b) experimental.
(b)
Voltage ripple across individual capacitors [(i) across C1 and (ii) across C2]: (a) simulated (switched model) and (b) experimental.
Fig. 16. Imbalance in the capacitor voltage waveforms [(i) voltage across
C1 and (ii) voltage across C2] (experimental).
out of control for the duration for which the input voltage
is greater than the capacitor voltagethe input current then
takes a shape similar to the input current waveform of a typical
rectifier-capacitor-load front-end circuit, as also pointed out in
Section II-A3b.
In Fig. 18, the variation of switching frequency with time
based on simulation (using switched model) is depicted. Com, the effect of
paring this with Fig. 8(a), drawn for
imbalance on the minimum switching frequency may be seen.
A comparison of the analytical, simulation, and experimental
results for the maximum and minimum switching frequencies
measured at 300-W and 300-V output is presented in Table IV.
The results agree quite well. Similar comparisons have been
made for the variations of duty cycle with time as well. These
are not presented here due to lack of space.
The variations of efficiency (measured with Voltechs
PM3000 A equipment) with respect to rms input voltage and
load are shown in Fig. 19. The efficiency is generally high
reaching a maximum of around 96.23% at an input voltage of
90 V and a load power of 225 W.
V. IMBALANCE
IN
CAPACITOR VOLTAGES
SRINIVASAN AND ORUGANTI: UNITY POWER FACTOR CONVERTER USING BOOST TOPOLOGY
COMPARISON
OF THE
HARMONIC COMPONENTS
TABLE III
CURRENT AGAINST
OF THE INPUT
(a)
Fig. 17.
THE
495
(b)
Output voltage ripple: (a) simulated (switched model) and (b) experimental.
496
(a)
Fig. 19.
(b)
Experimental efficiency results. (a) Variation with line voltage. (b) Variation with load power.
(31)
(32)
Even though (30) and (31) are nonlinear, (32) is a firstorder linear differential equation. The solution of (32), after
substituting for
, is given by
(33)
From (33), it is seen that the imbalance caused by the initial
conditions of the system does not decay with time. The
dependence of the imbalance on the initial conditions, and
also the time at which the input is applied, explains why the
magnitude of the imbalance is arbitrary as observed both in
the experiment and simulation.
B. Elimination of Imbalance
The closed-loop control system with the imbalance control
is shown in Fig. 21. Here, the shaded blocks are extra blocks
added to the system in Fig. 6. The signal
is obtained
after multiplying
(output voltage of the controller
)
with the sinusoidal signal
in phase with the line voltage
waveform. Because of the fast inner current control loop, any
dynamics associated with it may be neglected while studying
the voltage-loop performance. Therefore, in the original system, the input line
is the same as
. The operation
Fig. 20. Model of the half-bridge circuit used to analyze the imbalance in
capacitor voltages.
where
and
are the dc values of the currents through
the controlled sources.
(i.e.,
). This introduces a positive
Now, let
offset
in
. From (13), the average value
of
is equal to . Therefore, applying Kirchoffs current
for the dc currents, we note that
law (KCL) at junction
will increase by
, whereas
will increase by
(
magnitude will decrease). The dc-current
component
causes C2 to charge upthe component
causes C1 to discharge. As a consequence,
reduces. In steady state, the dc value of
is made
zero through this closed-loop action. In Section V-B1, we
SRINIVASAN AND ORUGANTI: UNITY POWER FACTOR CONVERTER USING BOOST TOPOLOGY
Fig. 21.
497
COMPARISON
OF
RESULTS
FOR
MAXIMUM
AND
TABLE IV
MINIMUM SWITCHING FREQUENCY
OF
AND
Vo = 300 V)
where
is the time constant and is the variable
of integration. The constant in (36) is obtained by setting
time equal to zero in (38).
As
(Fig. 21) is nearly constant over a line cycle,
will be sinusoidal with no dc offset. The term with the
integral sign in (36) will therefore have only ac components
as seen later in (38). Hence, from (36) it is seen that the dc
imbalance caused by the initial conditions approaches zero
asymptotically in steady state. The settling time
for the
dc imbalance to decay to zero can be estimated to be equal
to
approximately. Making equal to
infinity (i.e.,
) in (36) once again verifies the existence
of the imbalance in the original system.
is given by
Under steady state,
(37)
Substituting (37) in (36) and performing the integration,
under steady state is given by
(38)
and
(35)
(36)
498
(a)
(b)
(c)
Fig. 22. Averaged model simulation results after imbalance correction: (a) input current waveform, (b) voltage across individual capacitors [(i) across
C1 and (ii) across C2], and (c) output voltage ripple.
TABLE V
POWER FACTOR VERSUS u : A COMPARISON OF SIMULATED AND
ANALYTICAL RESULTS (@ 90-V INPUT AND 300-W OUTPUT)
real power
apparent power
(39)
From Table V
(40)
may be considered as a good choice. This will result in a PF
equal to 0.995 and in a settling time equal to 127 ms, which
is about six and a half line cycles for a 50-Hz power system.
SRINIVASAN AND ORUGANTI: UNITY POWER FACTOR CONVERTER USING BOOST TOPOLOGY
(a)
Fig. 23.
499
(b)
Experimental results after imbalance correction: (a) voltage across individual capacitors [(i) across C1 and (ii) across C2] and (b) output voltage ripple.
VII. CONCLUSION
Fig. 24.
500
Ramesh Srinivasan (S93) received the B.Sc. degree in physics from University of Madras, Madras,
India, in 1987 and the M.E. degree in electrical
engineering from the Indian Institute of Science,
Bangalore, India, in August 1991. He has been
working toward the Ph.D. degree in electrical engineering at the National University of Singapore,
Singapore.
From August 1991 to November 1993, he was
working in industry in research and development.
He was involved in the design and development of
a variety of high-frequency high-density switch-mode power supplies and
uninterruptible power supply systems. His research interests are converter
topologies, modeling, simulation, and control, especially as applied to power
electronics. He is currently involved with an electric vehicle project at the
Singapore Polytechnic, Singapore.