1888 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 23, NO. 4, JULY 2008
Space Vector Modulator for Vienna-Type Rectifiers
Based on the Equivalence Between
Two- and Three-Level Converters:
A Carrier-Based Implementation
Rolando Burgos, Member, IEEE, Rixin Lai, Student Member, IEEE, Yunqing Pei, Member, IEEE,
Fei (Fred) Wang, Senior Member, IEEE, Dushan Boroyevich, Fellow, IEEE, and Josep Pou, Member, IEEE
Abstract—This paper presents the equivalence between two-
and three-level converters for Vienna-type rectifiers, proposing a
simple and fast space vector modulator built on this principle. The
use of this duality permits the simple compliance of all topological
constraints of this type of nonregenerative three-level rectifier,
enabling as well the extension of its operating range by the use
of simpler two-level overmodulation schemes. The proposed algo-
rithm is further simplified by deriving its carrier-based equivalent
implementation, exploiting the direct correspondence existent
between the zero-sequence vectors of Vienna-type rectifiers and
the zero state vectors of two-level converters. As a result, the
proposed algorithm is also capable of controlling the rectifier
neutral point voltage. This feature makes it attractive as well
for neutral-point-clamped inverters, complementing previous
carrier-based space vector modulators developed for these con-
verters. A complete experimental evaluation using a 2 kW digital
signal processor–field programmable gate array controlled Vi-
enna-type rectifier is presented for verification purposes, asserting
the excellent performance attained by the proposed carrier-based
space vector modulator.
Index Terms—Carrier, pulsewidth modulation (PWM), space
vector modulation (SVM), three-level converter, Vienna rectifier.
I. INTRODUCTION
A. Vienna Rectifier, Equivalent Topologies, and Space Vector
Modulation
T HE Vienna rectifier was developed and proposed with thepurpose of maximizing the power density of three-phase
power supplies for telecommunication applications [1]–[3].
Providing power factor correction (PFC) for diode rectifiers,
Manuscript received July 18, 2007; revised October 14, 2007. Published July
7, 2008 (projected). This work was supported by the Engineering Research
Center Shared Facilities, the National Science Foundation under NSF Award
Number EEC-9731677, and the CPES Industry Partnership Program. Recom-
mended for publication by Associate Editor P. Barbosa.
R. Burgos, R. Lai, F. Wang, and D. Boroyevich are with the Center for Power
Electronic Systems (CPES), Virginia Polytechnic Institute and State University,
Blacksburg, VA 24061 USA (e-mail: rolando@vt.edu; rp.burgos@ieee.org).
Y. Pei is with the School of Electrical Engineering, Xi’an Jiaotong University,
Xi’an 710049, China.
J. Pou is with the Terrassa Industrial Electronics Group (TIEG), Universitat
Politecnica de Catalunya, Catalunya 08222, Spain.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TPEL.2008.925180
its specific circuit topology evolved from nonregenerative
two-level and three-level rectifiers [4]–[9], where together with
the latter ones forms a family of functionally equivalent circuit
topologies [8]–[11]. More than a decade later and thanks to their
notable performance and the advancements on semiconductors
devices, magnetic materials, capacitor and cooling technolo-
gies, these rectifiers remain as one of the preferred choices
when power density is a design objective [12], [13]. As such,
aircraft applications where both power density and specific
weight are of utmost importance have become a potential user
of this type of nonregenerative three-level rectifier [14], [15],
which has also been recently considered for industrial motor
drives, power supplies and active filters [13], [16]–[18].
For power supplies in the several kilowatts range, Vi-
enna-type rectifiers can employ switching frequencies well
above 100 kHz depending on specific power quality and elec-
tromagnetic compatibility (EMC) requirements [13]–[15]. For
this reason hysteresis or sinusoidal-pulsewidth modulation
(PWM) modulators are used since they can be implemented by
simple analog circuitry [3], [19], [20]. Along these lines several
alternative methods have been proposed seeking to improve the
converter performance without using digital processors, among
others one-cycle control and multiplier-less analog controls
[21], [22]. And yet, though limited in capabilities, advanced
analog control schemes for specific operating conditions have
been successfully implemented [23], [24].
For applications optimized for lower switching frequencies,
controls by means of digital signal processors (DSP) and field
programmable gate arrays (FPGA) may be readily employed for
Vienna-type rectifiers [9], [15]–[18]. Motor drives are a good
example of this since a single DSP can be used to control both
rectifier and inverter stages. To take advantage of digital con-
trols, several space vector modulation (SVM) algorithms have
been proposed, exploring the boundaries and limitations of con-
tinuous and discontinuous schemes [16], [25], [26]. However,
the apparent complexity of its implementation has somewhat
thwarted the usage of SVM, a consequence of the nonregen-
erative nature of this type of three-level rectifier which requires
corresponding line currents and phase voltages to have the same
instantaneous polarity [20]. The neutral-point-clamped (NPC)
three-level inverter on the contrary has none of the above con-
straints [27].
Regarding the NPC inverter, significant work has been done
on its modulation scheme since it has been an intrinsic lever in
0885-8993/$25.00 © 2008 IEEE
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
BURGOS et al.: SPACE VECTOR MODULATOR FOR VIENNA-TYPE RECTIFIERS 1889
balancing the neutral point voltage of this converter. Both car-
rier-based PWM [28]–[31] and SVM techniques have been used
[32]–[35], although the latter has usually been finally imple-
mented exploiting the equivalence with the former [34]–[41].
This equivalence however is not as straightforward and direct
as in two-level converters where the relationship between car-
rier-based PWM and SVM has been well established [42]–[44].
Specifically, some initial study showing this equivalence for
NPC inverters was presented in [36], [37]-valid only at a single
operating point. A more complete analysis and carrier-based
SVM scheme was later presented for the three-level NPC in-
verter in [38], and for four- and -level inverters in [39] and
[40]. However, the analyses presented in [36] and [40] only con-
sidered the case of equal time distribution between redundant
vectors, and hence did not take into account the neutral point
voltage balancing problem, essential for this type of topology.
References [39], [40] correctly identified though the constraints
added by the stacked carrier signals in the generation of the
zero-sequence component on which the PWM and SVM equiv-
alence is built. Only [41] has presented of late a theoretical
analysis for -level inverters including the neutral point voltage
balance, nonetheless veiled behind an intricate geometrical de-
scription.
Another relevant simplification of SVM in NPC converters
is attained by exploiting their equivalence with two-level
converters, as was shown in [45] and recently for th-order
multilevel inverters in [46], [47]. This analogy is extended in
this paper to Vienna-type rectifiers, and is used to identify the
subset of space vectors active during each switching cycle, thus
easily complying with all topological constraints associated
to enforcing the equal polarity between corresponding line
currents and voltages [48], [49]. The converter operating region
is then extended by using a simple two-level overmodulation
algorithm [50]. The equivalence with two-level converters is
also used to show the functional correspondence between the
redundant zero-sequence vectors of Vienna-type rectifiers and
the zero state vectors of two-level converters. This duality is
the basis of the proposed carrier-based SVM algorithm, and as
such is used to derive its zero-sequence duty cycle generator.
This is done by equating the relative conduction times of the
zero-sequence vectors thus mirroring the zero vector control
in two-level converters. As a result, the proposed algorithm is
capable of controlling the neutral point voltage of the rectifier.
This paper presents the complete, detailed derivation of the
proposed algorithm together with a thorough experimental eval-
uation using a 2 kW DSP–FPGA controlled Vienna-type recti-
fier prototype. The algorithm is first developed in space vector
terms exploiting the two-level and three-level converter equiv-
alency, and then in terms of its carrier-based implementation,
providing significant insight into this realization and its addi-
tional usage for three-level NPC inverters. The results obtained
verify the excellent capabilities of the proposed carrier-based
SVM algorithm.
II. ELECTRICAL STATES OF VIENNA-TYPE RECTIFIERS
Fig. 1 shows the circuit schematic of the Vienna-type rectifier
considered in this paper, comprised of a main diode bridge and
three ac switches connecting the input phases to the dc-bus neu-
tral point. The rectifier prototype built and used for evaluation
of the proposed algorithm is rated at 2 kW, 60 V rms, 60 Hz,
Fig. 1. Circuit schematic of Vienna-type rectifier topology considered in this
paper.
200 V dc, and uses SiC Schottky diodes for the main bridge
diodes and Si MOSFETs with their body diodes to implement
the bidirectional switches. A digital controller built on a DSP
(AD21160M SHARC) and FPGA (Xilinx XCV-400) architec-
ture is used to implement the modulator and controls using a
sampling frequency of 40 kHz. The boost inductors are 160 H
rated at 2 kW, and the dc bus capacitors are 20 F rated at 450 V.
This rectifier as a three-level topology has 25 valid electrical
states of the following type:
(1)
In this paper, switching functions (for ) take
values in representing the per unit voltage of the
phase-leg with respect to the neutral point, and not the state
of the corresponding bidirectional switch. These 25 states are
a subset of the 27 states of the NPC inverter, since these topolo-
gies have only one null or zero state. Per this switching function
definition and Fig. 1, the rectifier phase to neutral point voltages
under balanced steady state conditions are given by
(2)
These 25 states may be converted into the plane using
the following space vector transformation
(3)
(4)
so that the resultant space vectors have magnitude zero,
one, , or two. Correspondingly, the maximum magnitude for
a per unit sinusoidal voltage reference vector normalized to
is , where is the modulation index 1 0 and
amplitude of the phase duty cycles in abc-coordinates. The 25
space vectors labeled with their corresponding abc-coordinate
state realization are shown in Fig. 2(a).
A major constraint of this topology is that the phase to neutral
point voltage can only be switched between the neutral point
and the dc bus rail having polarity equal to that of the corre-
sponding line current, that is when 0 or when 0
as shown in Fig. 1. This limits the number of active states that
can be applied at any given time to a subset of eight vectors
depending on the instantaneous polarity of the converter line
currents, or equivalently depending on the sector where the line
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
1890 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 23, NO. 4, JULY 2008
Fig. 2. Space vectors of Vienna-type rectifier: (a) 25 vectors labeled with their
corresponding electrical state realization S = [S S S ]. (b) Space
vector diagram illustrating the six Sectors {I, II,. . . ;VI} created by the loca-
tion of the input current vector i , depicting as well the active voltage-vector
hexagons and corresponding vectors for two different locations of i , Sectors
I and IV, respectively.
current vector lies in the plane. A total of six possible Sec-
tors exist, {I, II, VI}, which activate six different set of space
vectors forming a hexagonal region. Fig. 2(b) shows as example
the cases when the current space vector lies in Sectors I and
IV, respectively. This active hexagon demarcates the valid and
realizable location for the reference voltage vector. As shown
in this figure, the active vectors in every case correspond to the
two realizations of the redundant vector connecting the origin of
the plane to the center of the hexagonal region, and to the
six vectors forming this hexagonal region (including the zero
vector). Notice that due to the current polarity restrictions, only
one of the two realizations for the other two redundant space
vectors forming the hexagon are active during this time; namely
and in Sector I, and and
in Sector IV as shown in Fig. 2(b).
III. EQUIVALENCE BETWEEN TWO-LEVEL AND
THREE-LEVEL SPACE VECTOR MODULATION
Fig. 3 shows the space vectors of a two-level voltage source
converter in the plane consisting of six active vectors
and two zero vectors . The active vectors di-
vide this hexagon into six Sectors, {I, II, VI}, which de-
limit the possible location of the reference voltage vector
without requiring overmodulation. For either SVM or its car-
rier-based implementation the sequence of states applied during
a switching cycle is
(5)
for odd sectors, or with reversed active vector transitions for
even sectors, where and correspond to the vectors leading
and lagging [42], [43]. The generic volt-second balance
during a switching cycle with controlled relative conduction
times of the zero vectors and is then given by
(6)
where corresponds to the switching period,
, , and , to the conduction times of the respective space
vectors, and to the controlled time ratio between vectors
and [44].
For Vienna-type rectifiers as well as for the NPC converter
topology, SVM is implemented by synthesizing the reference
vector using the three nearest vectors forming a triangle around
Fig. 3. Space vector diagram of two-level voltage-source converter showing:
six Sectors {I, II,. . . ;VI}, six active vectors fv ; v ; . . . ; v g, two zero vectors
fv ; v g, and the synthesis of the voltage reference vector v in Sector I, in-
dicating the corresponding lagging and leading vectors v and v ; v and v ,
respectively.
Fig. 4. Two- and three-level space vector equivalency: (a) synthesis of v in
three-level SVM diagram using three adjacent space vectors forming a triangle
around it (v , v and v ) and (b) synthesis of equivalent reference voltage vector
v in two-level equivalent space vector diagram for Vienna-type rectifiers.
it. For Vienna-type rectifiers, restrained to the eight active vec-
tors depending on the current vector location, one of the three
nearest vectors forming the triangle in question is always the re-
dundant vector pointing to the center of the active hexagon. If
this redundant vector is named , the volt-second balance for
this type of rectifier can be expressed as
(7)
where and correspond to the vectors lagging and leading
as shown in Fig. 4(a). Since has two switching state real-
izations as described in Section II, (7) may be rewritten by split-
ting into its two realizations, namely and , and using a
constant as the time ratio between them as
(8)
These two vectors are named and given that injects
current into the dc-link neutral point, that is in Fig. 1,
while draws current from it , resulting in the re-
spective charge and discharge of the dc-link midpoint voltage.
The volt-second balance (8) is by inspection equivalent to the
expression given in (6). It is apparent then that the role of the
zero vectors in two-level converters-which shape the zero-se-
quence voltage, and that of the zero-sequence vectors in Vi-
enna-type rectifiers is identical. Now, since
(9)
and
(10)
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
高亮
Luli
注释
为什么取1/2
BURGOS et al.: SPACE VECTOR MODULATOR FOR VIENNA-TYPE RECTIFIERS 1891
when 2, i.e., the sought operating conditions,
(8) may be rewritten as
(11)
Rearranging terms in this expression yields
(12)
where may be understood as an origin translation in the
plane. This defines a new set of equivalent two-level vectors
denoted by the superscript in the following, with which the
volt-second balance (12) may be rewritten as
(13)
This shows that the volt-second balance of Vienna-type rec-
tifiers is identical to that of two-level converters after applying
an origin translation of . This is verified by replacing
into the two-level volt-second balance expression
(6) as
(14)
Fig. 4(b) illustrates this concept, depicting the two-level syn-
thesis of the equivalent reference voltage vector . The equiv-
alence is finally complete by stating that
(15)
As mentioned in Section I, a similar duality has been estab-
lished for the NPC converter [45]–[47]; however, in the case of
this converter the relationship is more involved since depending
on the magnitude and location of the reference voltage vector
more than one set of redundant vectors of the converter might
be active (up to three for low modulation indexes), in which case
the correspondence with two-level converters is not one-to-one
as it is for Vienna-type rectifiers regardless of the reference
vector location or magnitude.
IV. SPACE VECTOR MODULATION SCHEME
The space vector modulator for Vienna-type rectifiers may
be easily implemented by exploiting the equivalence with
two-level converters. First, the sector where the current space
vector lies is determined using a simple algorithm developed
for the space vector modulation of two-level current-source
converters that enables the straightforward sector identification
without using trigonometric relationships [48]. Once the cur-
rent vector is located, the rectifier active hexagon is uniquely
determined together with all its associated voltage vectors,
including the three vectors closest to , namely and and
the redundant zero-sequence vector . The origin translation
can then be applied generating the two-level equivalent
reference voltage vector and the equivalent vectors and
. The location of the new reference vector in the two-level
space vector diagram is determined similarly to the location of
the current vector but by applying the voltage-source converter
equivalent algorithm [49]. The respective conduction times
and for the equivalent vectors and are then
determined by
(16)
Once the conduction times are known the corresponding
three-level vectors , , and can be applied for ,
and using (15) and the sequence described by (8).
An advantage of this two-level equivalent approach is that the
main requirement for the modulation of Vienna-type rectifiers
may be easily addressed, i.e., that the voltage reference vector
lie within the active hexagon. This can be achieved by simply
limiting to lie within the sinusoidal region of the two-level
space vector diagram, or by using any overmodulation technique
for two-level converters which has the advantage of extending
the converter operating region by 10%. In this case, a fast algo-
rithm based on if–then rules covering both overmodulation and
bang-bang regions as presented in [50] was used. This algorithm
modifies the equivalent two-level reference keeping it re-
strained to the voltage hexagon defined by the location of the
current vector while following the actual reference as close as
possible. Fig. 5 shows the active voltage hexagon and the linear
and overmodulation regions as well as the algorithm in question.
Naturally, this approach also covers the actual overmodulation
region of Vienna-type rectifiers, that is when lies outside
both the active hexagon and the three-level space vector region.
V. PROPOSED CARRIER-BASED SPACE VECTOR MODULATOR
A. Basis of P