A THREE-PHASE SOFT-SWITCHED HIGH POWER DENSITY DC/DC

CONVERTER FOR HIGH POWER APPLICATIONS.

R.W. De Doncker D.M. Divan M.H. Kheraluwala

University of Wisconsin-Madison
Dept. of Electrical & Computer Engineering
1415 Johnson Drive
Madison, WI 53706-1691
Tel. (608) - 262 5702

Abstract
This paper presents three dcldc converter topologies suitable for This paper proposes a new soft-switching dcldc converter
high power density high-power applications. All three circuits topology suitable for high power applications. Soft switched
operate in a soft switched manner making possible a reduction in converters are characterized by intrinsic modes of operation which
device switching losses and an increase in switching frequency. allow an automatic and lossless resetting of the snubber elements
The three-phase dual-bridge converter proposed is seen to have the through an appropriate recirculation of trapped energy. The
most favorable characteristics. This converter consists of two three capability to eliminate losses associated with the snubber now permit
phase inverter stages operating in a high frequency six-step mode. the use of oversized snubbers resulting in dramatically lower device
In contrast to existing single phase ac link dc/dc converters, low switching losses, even at substantially higher frequencies.
r.m.s. current ratings are obtained for both the input and output filter Examples of soft switched dcldc converters are the parallel output
capacitors. This is in addition to smaller filter element values due to SRC operated above resonance [9], the resonant pole, the pseudo-
the higher frequency content of the input and output waveforms. resonant converter [2,3] and all quasi-resonant converters [4,5].
Furthermore, the use of a three-phase symmetrical transformer For multi-quadrant operation and for dclac inverter applications,
instead of single phase transformers and a better utilization of the typical examples of soft-switchedtopologies are the resonant dc link
available apparent power of the transformer (as a consequence of the inverter and the quasi-resonant current mode or resonant pole
controlled output inverter) significantly increase the power density inverter [5]. The proposed circuit utililizes the resonant pole as the
attainable. basic switching element for both input and output devices and yields
substantial benefits in power density and operating characteristics.
Introduction
Soft Switched DCDC Converters
The area of high power density dc /dc converters has been an
important research topic, especially for switched mode power The preferred dc/dc converter topology for high power
supply applications rated at up to 500 watts. The needs of the next applications has been the full bridge circuit operated at constant
generation of aerospace applications require extremely high power frequency under a pulse width control strategy. The topology
densities at power levels in the multi-kilowatt to megawatt range. features minimal voltage and current stresses in the devices,
The implications of realizing high power density and low weight minimum VA rating of the high frequency transformer, as well as
systems at these power levels have rarely been addressed. This low ripple current levels in the output filter capacitor. The power
paper examines considerations for the selection of topologies which density levels that can be reached are limited by peak and average
can realize the low weight constraints that are of primary device switching losses, transformer leakage inductances and output
importance. rectifier reverse recovery. The factors above constrain the maximum
Recognizing that higher switching frequencies are the key to frequency attainable, and thus the smallest size possible, given the
reducing the size of the transformer and filter elements, it is apparent state of the art in component technology. Most of the soft switching
that some form of soft switching converter with zero switching loss converters reported in literature attempt to tackle one or more of the
is required, if system efficiencies and heat sink size are to be problems listed above, typically at the expense of substantially
maintained at a reasonable level. By far the most attractive circuit higher component stresses [4,5]. For high power operation, that is
so far has been the series resonant converter (SRC) [l]. Using unacceptable. Soft switching variations of the full bridge converter
thyristors with a single LC circuit for device commutation and are thus the most favoured topologies.
energy transfer, the topology is extremely simple in realization and
offers the possibility of power densities in the 0.9-1.0 Kg/KW
range at power levels up to 100 KW.
The following problems can be identified with the SRC.
Thyristor commutation requirements demand higher current ratings
from the devices and higher VA ratings from the LC components.
Thyristor recovery times significantly slow down the maximum
switching frequencies attainable. Snubber inductors and RC
networks are needed to effect current transfer without encounteringa
diode recovery problem. Capacitive input and output filters have to
handle ripple currents at least as large as the load current. Although
switching frequencies in the 10 KHz range yield dramatic reduction
in converter size when compared to conventional hard switching
circuits, it is clear that systems operating at similar frequencies and
with lower component ratings are potentially capable of even higher
li
power densities.

Figurel: Pseudo resonant full bridge dc/dc converter [21.

88CH2565-0/88/0000-07%$01.OO 0 1988 IEEE

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on June 06,2020 at 13:39:12 UTC from IEEE Xplore. Restrictions apply.
 The pseudo-resonant dcldc converter proposed in reference 2
is shown in Figure 1. It uses capacitive snubbers and can be
designed with device stresses approaching that of the conventional
full bridge. However, the circuit uses the transformer as a voltage
transfer element and the interactions of its leakage inductance (Li)
and the output rectifer are unresolved. The maximum switching
frequency limit is reached when the energy lost due to Ll and the
peak diode reverse recovery current become unacceptable. 3
Transferring the output filter inductance to the ac side
completely changes the operating characteristics of the circuit [2].
As shown in Figure 2a, the inductance L i s now in series with the
transformer leakage inductance, Ll. Further, energy trapped in
(LI+Lr) during reverse recovery is now transferred to the load in a
lossless manner. This implies that a further increase in switching
frequency may be possible. It can also be shown that the additional
inductor L required in the previous circuit (Figure 1) is no longer
absolutely essential, as adequate control range can be obtained at
constant frequency while maintaining soft-switching for all devices.
Operating waveforms for this single-phase phase shifted dc/dc
converter are shown in Figure 2b. It can be seen that soft switching
is obtained by ensuring that device turn-on only occurs when its
anti-parallel diode is conducting. The biggest problem with this
circuit seems to be the high ripple current in the output filter
capacitor. For low voltage power supplies, this is a very important
consideration. For high power high voltage aerospace applications,
it is not seen to be a major problem because power densities
obtainable with multi-layer ceramic capacitors are at least as good as
with inductors.
It has been proposed in Reference [7] that the diode recovery
process is akin to the existence of an active device in anti-parallel
with it. Observing that the circuit in Figure 2a naturally handles the
diode reverse recovery process, it is proposed that the diodes be
replaced by active devices as shown in Figure 3a. Many high
frequency converters already use synchronous rectifiers in
essentially the same location. The converter can now be operated
Figure 3: Topology B
a) single phase dual bridge dcldc converter
b) waveforms for topology B

with a simpler control strategy in which the input and output bridges
generate square waves which are phase shifted from each other. In
keeping with our philosophy, regions of operation can be identified
which permit soft-switching of & devices in bridges. Figure
3b shows operating waveforms for the single phase dual bridge
dc/dc converter topology.
The circuit in Figure 3a can be extended to a polyphase
version. The three-phase dual bridge circuit is shown in Figure 4a.
Control parameters :a) l3 : phase shift between VA VB Again, examining the modes of operation for the converter, it is
b) d : Vo/ Vs possible to identify regions where both sets of switches experience
soft-switching. As in the case of the single phase dual bridge
converter, both bridges generate quasi-square waves phase shifted
from each other. It should be noted that the soft-switchingtransition
is actually resonant in nature [2,6] but is assumed to be almost
instantaneous for the derivation of first-order operating
characteristics. The three-phase dual bridge converter has
substantially lower filter ratings when compared to its single phase
counterpart. Consequently, it has the potential of realizing the
highest power density.
It should be noted that all three converters, denoted converter
A, B, and C for circuits in Figure 2, 3, and 4 respectively, exhibit
desirable properties with regard to parasitics such as device storage
time, transformer leakage inductance and diode reverse recovery. It
is shown in the paper that transformers which use the leakage
impedance as an energy transfer element have the potential of
reaching higher power densities. While this technique has been
extensively used at lower power levels, it has been felt that the
higher VA rating of the composite transformer was an unacceptable
penalty at higher power levels. It will be shown that the resulting
increase in switching frequency possible, more than compensates
Figure 2: a) Topology A, single phase shifted dcldc for the increased VA rating, allowing a substantial reduction in the
converter overall size of the converter. The use of dual bridges also yields
b) Waveforms for topology A unexpected gains in power density, as will be shown, and permits
bidirectional power flow. Analysis in paper is restricted to
unidirectional power flow only.

797

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on June 06,2020 at 13:39:12 UTC from IEEE Xplore. Restrictions apply.
 Analvsis Of Single Phase Converters
Converter A:
In order to derive the operating charateristics of the three
dcldc converters, it is assumed that the transfer of current from
device to diode on turn-off is instantaneous. The actual switching
locus depends on the value of capacitance, C, used and the current
level. For a typical device such as a BIT with a current fall time,
tf,and a turn-off current of IP, the device energy loss per switching
cycle can be found approximately to be [81

As there are no turn-on or snubber dump losses, C can be made

fairly small while retaining a fast switching characteristic and low
device losses, simultaneously. This justifies the assumption of a
fast, almost instantaneous switching transtion for analysis over a full
cycle.

T *IO-4

0
0
N

N
0
'0.00 0.33 0.66
T *lo-4
N
0

Figure5: Equivalent circuits of three dcldc converter topologies:

. q , ,
0
, , i ,
a) Single-phase phase-shifted converter, topology A,
b) Single-phase dual bridge converter, topology B,
c) Three phase dual bridge converter, topology C.
0.00 0.33 0.66 1.00 1.33 1.67 2.00
T *IO+
The equivalent circuits for the three converters are shown in
Figure 5. Replacing the transformer with an equivalent inductance,
L, simplifies circuit analysis. For converter A, three operating
Figure4: Topology C, a) Circuit schematicof the proposed 3 phase modes can be identified as shown in Figure 2b. The phase shift
dcldc converter, between the two resonant poles is t$ = atp, where 61is the switching
b) Operating waveforms: frequency. The equivalent voltage applied across the load is V&
V p Primary line-neutral voltage,
VAS- Secondary line- neutral voltage, and has pulse width of p = 6xb. The inductor current i as a function
I A primary
~ line current, of 8 = cot is given below. In Mode 1,
IAS- Secondary line current,
Is- Source side dc link current,
I(j- Load side dc link current.

798

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on June 06,2020 at 13:39:12 UTC from IEEE Xplore. Restrictions apply.
 where Vs, Vo are input and output dc voltages and i(0) is the initial Operating characteristicsfor converter A are calculated based
on the principles listed above and are shown in Figure 7. Figure 7a
current at 0 = 0. Mode 1 ends at 0 = $. Given the output rectifier it
depicts the range of control possible with variation of p while
is clear that OS@$. In Mode 2, maintaining soft-switching operation for all four devices. It is seen
that maximum power transfer occurs at p = 180" and d = 0.58. The
range of control obtainable at various values of d is limited by the
(3) requirement of soft-switching. Figure 7b shows a variation of the
transformer KVA versus the power output Po It is seen that at the
design point, the KW/&VA)~ratiois approximately 0.3 pu.
similarly, the current in Mode 3 can be found to be
PHRSE-SHIFT DC-DC CONVERTER
i(0) =
(21 p) +
-(0 - i(p) (4) OUTPUT DIODE B R I D G E

At the end of the half cycle i(x) = -i(O). Solving for i(O), we can
obtain the complete current waveform.
The soft-switching constraints require that the device be
conducting at turn-off. From Figure 2b this implies that i(rc) >0,
Further, given the diode bridge on the output, i($)=O. Using the
above relationships,

I$ = 1(p - drc)
2
where

From i(8) and the converter switching functions, the supply and
output average currents, (I, and Io) can be found to yield power
%:OO 3b.00 6b.00 Sb.00 liO.00 lkO.00 l'so.00
transfer at a given p, d and U. This is found to be BETR IDEG)

(7) PHASE-SHIFT DC-DC CONVERTER

OUTPUT D I O D E BRIDGE
Input and output filter capacitor rms current ratings can also be In
0

calculated from i(0). Peak device voltage and current ratings are
also easily found. 1'
Estimating the size/power density of the transformer is not
as straightforward as it might seem. As the leakage reactance is
used for the main energy transfer element, the size associated with it
is fairly significant. It is presumed that the VA rating of the
'composite' transformer is smaller than that obtained for the
combined VA ratings of an ideal transformer and series inductances
corresponding to the leakage inductance. Figure 6 shows the
equivalent circuit model of a single phase transformer. The KVA
rating of the transformer is then calculated to be (neglecting
magnetizing current),

O<VA)T12(Vm. I) + 2 12. - ( 21 (8)

Ideal Leakage
Transformer

where Vm and I are rms quantities and XI = uL. For given values
of V, and Po, it is also assumed that transformer size is proportional
to its KVA rating and inversely proportional to the frequency.
x,/ 2 x,/ 2
Figure7: Topology A
a) Variation of output power versus p for various values
of V f l s = d,. boundary shown corresponds to (I = 0,
vs b) O<VA)Tversus the output power Po as a function of d,
Outer boundary corresponds to p = NOo,inner boundary
- -
to I$ = 0.

Figure6: Equivalent circuit model of a single phase transformer.

799

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on June 06,2020 at 13:39:12 UTC from IEEE Xplore. Restrictions apply.
 Converter B: The constraints which define soft-switching boundaries can now be
specified for the input and output bridges to be i(0) I O and i($) 2 0
A similar analysis can be carried out for converter B. This respectively. These constraints enclose the desired operating region
converter has only two modes of operation. The two bridges are for the converter. Exceeding the Fist constraint results in natural
presumed to operate with phase shift $. The cuirent i(e) is once commutation of the input bridge devices and gives snubber dump.
again given by equations 2 and 3 for modes 1 and 2 respectively. For the output converter, the constraint equation corresponds to
diode bridge operation.
The boundary conditions now dictate that i(0) = -i(a) at the end of
Figure 8a shows the variation of Po as a function of $ for
Mode 2. Solving for i(e), the output power Po, is: different values of d. The upper boundary corresponds to the input
converter, while the lower curve represents diode operation. For
(9) d=l it can be seen that $ can vary over the entire range of 0 to a12
giving control from zero to full power. Figure 8b plots the
transformer KVA against the output power for various values of d.
The boundary, corresponding to an output diode bridge, is identical
1-PHASE DC-DC CONVERTER to that in Figure 7b with p = R.
Examination of Figure 8b shows that a transformer KVA of
1.356 pu is required with output diodes if d is to be varied over the
0
range 0 to 1. With dual bridges, for the same transformer KVA, it
N-
is now possible to transfer Po=O.59 pu, as against a previous
maximum of 0.3 pu. Taking into consideration the phase shift
control possible in converter A, the KW/(KVA)T ratio increases
O
U)

d
from 0.3 to 0.435 an astounding improvement of 45% in the power
density.
Compared to normal hard switched converters with
N
0
KW/(KVA)Tratios approching unity, this may seem to be very poor
transformer utilization. However, if the switching frequency for the
proposed converter can be made substantially higher, actual
size/weight could be much lower. Transformer sizing will be
examined in greater detail later in the paper. It is apparent that along
with further gains in transformer power density, a significant
reduction in input/output filter size and ripple current rating will
result from selecting the three phase dc/dc dual bridge converter.
Analvsis Of Three Phase DC/DC COnvertq
The circuit schematic of the new three phase dual bridge
soft-switching ac link dc/dc converter is shown in Figure 4. The
proposed converter consists of two three-phase inverter stages, each
00
P H I 0 (DEGI
operating in a six-step mode with controlled phase shift. Using two
active bridges not only permits bidirectional power flow, but also
allows control at a fixed frequency. The ac link transformer is Y-Y
connected and is three phase symmetric with the leakage inductances
1-PHASE DC-DC CONVERTER
used as energy transfer elements.

iip
i is

lkp

ks

Figure9: Schematic of a symmetric 3 phase transformer.

In the following analysis, it is assumed that the primary and

secondary resistances of the transformer can be neglected and the
%% OI.15 0130 0:VS 0160 0:75 d.90
turns ratio is 1:l. Figure 9 shows the schematic of the 3 phase
KWCl s)m”mtric transformer. Using the relationships C ii, = 0 and iiS

Figure8: Topology B, = 0 for Y connected transformers, the transformer equations can be

derived to be
a) Variation of output power versus $ for various values
of d, upper boudary - natural commutation locus for
input bridge, lower boundary - diode operation for
output bridge,
b) (KVA)Tversus Po as a function of d. Boundary
corresponds to diode owration for output bridge.

800

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on June 06,2020 at 13:39:12 UTC from IEEE Xplore. Restrictions apply.
 and
for Cp I 0 I x13

where V. and Vis are the primary and secondary voltages for the i h for n13 I 0 ICp + x13
phase, E,, and Lsl are the primary and secondary leakage
inductances, Lpp and Lss are self inductances and Lps,Lsp are the
mutual inductances between the appropriate phase windings given
by the superscript notation used.
Using properties of a symmetric transformer and defining
for 4 + x13 I 0 I 2x13
Lm = Lf, + LY, = LFP + L"PP
one can derive

for 2x13 I 0 I 2x13 + 4

di
o(L,+L ) J = V . (t)-- Lm Vis(') (14)
PI dt IP L, + Lsl
for Cp + 2x13 I 0 I x
where CT is a leakage factor given by

Equating i(0) = -i(x) and solving, we obtain

i(0) = -vs [ Y - d C p - F ]
The value of (T is typically a small number around the ratio of the 3WL,
leakage to the magnetizing inductance.
Equations 13 and 14 are the basic equations which govern Using i(0), the supply side dc link current can be reconstructed
the current in the circuit. Further assuming that hi= (< L,,, (for using the input bridge switching function, and is shown in Figure
1 : l turns ratio), then equations 13 and 14 reduce to 4b. From this, the average output power is calculated to be
di .
L 1s= Vi$') - v. (t)
cs dt 1P

di The above results apply for 0 I $ 5 x13.

L A
dt
=v.1P(t) - Vis(')
For x13 I $ 5 2x13, a similar set of equations can be
derived. The average output power for $ in this range can then be
where found to be

La = a(L, + L,,) = a(L, + LPl) E (Lsl + LPl) (18)

The simplified single equivalent circuit reduces to that shown in

Figure 6 with Lo = L. It can be seen that with the assumptions Based on i(0), the important parameters such as the transformer
above, ii, = iip = i(0). KVA, input and output capacitor ripple current and peak device
In order to calculate the three line currents, the classic six stresses can be found.
step line to neutral voltage waveform is assumed for both the Figure 10a is similar to Figure 8a and shows the variation of
primary and secondary windings. The difference between the two Po as a function of (I for different values of d Once again, the curve
voltages is applied across b. Six modes, corresponding to
different driving voltages, can be identified over a 180' conduction for d=l shows wide range of control, i.e. from zero power for $ = 0
cycle. Using the property of a balanced 3 phase set and E iip = 0, to maximum power for $ = W2. The lower boundary corresponds
one can obtain full information by calculating two currents over 113 to the soft-switching locus for a diode output bridge. This locus is
of a period.
derived by finding the value of $ = & such that i((Ic) = 0. For 0 I@
Solving for i(0) over a half period O+x, I x/3 this yields the lower boundary dl,
forOI0SQ
( $ = I - - 3Cp
(1 +d) 0
Vs -
i(0) = i(0) + - 2x
3 WL,
The upper boundary governs the transition for the input bridge
between natural commutation and soft-switching. This corresponds
to the relationship i(0) = 0, which yields dU,
80 1

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on June 06,2020 at 13:39:12 UTC from IEEE Xplore. Restrictions apply.
 3-PHASE DC-DC CONVERTER Ln
0

0
Ln

. 00
P H I 0 (DEGI

* 00
Figurelo: Topology C,
a) Variation of output power versus (I for d=O.25 to
d=2.5 in steps of 0.25. Upper boundary - natural
0
commutation for input bridge, lower boundary -
diode operation for output bridge,
b) (KVA)T versus Po for d=O.1 to 1.0, in steps of 0.1,
boundary indicates diode operation,
c) R.M.S current in input capacitor filter.
d) R.M.S current in output capacitor filter.
du=- 1
(29)
1 - -3@
2Jc

I 2 ~ 1 3the
For xI3 I @ , lower and upper boundaries are obtained to
be

d = -2- - 3@
' 3 J c (30)

d,=- 1
(31)
3 3@
2 J c

KWO Figure 10b shows the KVA rating of the transformer for 0 I
d I 1 as a function of the output power. The locus corresponding to
output diode bridge operation is also plotted. Once again, it can be
seen that maximum output power with the dual bridge is 0.46 pu
with a KW/(KVA)T ratio of 0.48. For the diode bridge, Po30.265
0
m pu and KW/(KVA)-r=0.38, a substantial difference. Figures 1Oc
and 10d show the ripple current in the input and output capacitor
filters as a fucnction of d and @. It will be seen that for d=0.5 and 9
= 60°, the output current ripple goes to zero. At the maximum
power transfer point of Po=0.46 pu, @ = 50.1" and the output
current ripple is 0.095 pu while the input current ripple is 0.0925
pu. For lower values of d, the output current ripple increases.
However, under all conditions, the ripple is substantially smaller
than for either of the single phase converters. Clearly, given the
operating range of the converter, an optimization is possible which
yields the smallest filter size.
The analysis of the two dual bridge converter topologies has
yielded interesting and fairly counter-intuitive results in terms of
overall system power density. It is not obvious why, given a
transformer, one is able to obtain more power with a dual bridge
than with a single bridge. Also, it is not clear whether the resulting
system, using transformer leakage inductances gives power density
than a conventional hard switched dcldc converter in which the
leakage elements are parasitics. In order to examine these issues
better, a fundamental component model of the system is invoked and
PH I D (DEGI * O0 analyzed next.

802

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on June 06,2020 at 13:39:12 UTC from IEEE Xplore. Restrictions apply.
 Fundamental Model Analvsis
(36)
XL
This is approximately equal to 2Po as $h is assumed small. Thus the
transformer size for hard-switching will be proportional to s h ,
where

(37)
Similarly, examining the softiswitched converter, we can see that
Figure 11: Fundamental model of the dual maximum power transfer occurs at Q = qS = xi2 and d = 1. Under
bridge dcldc converter these conditions and at a frequency q,we find for the same ratings

Once again ignoring winding resistances and core losses, the o , L = X =- e

transformer model in Figure 6 can be invoked for both the single
phase and three phase converters. Figure 11 shows the fundamental
' Po
equivalent circuit for the two converters. VSof angle of 0"and Vo of
(KVA)T = 4P0 (39)
angle of are the two voltage sources applied across the effective
leakage impedence term XI = jwL. This model is identified to the
synchronous machine equivalent circuit and may be expected to The size is now obtained by dividing (KVA)T by a.
demonstrate similar properties. The current vector i can be Comparing the sizes of the two transformers, we can see that
calculated to be

where Ssis the sizing constant for the soft switched converter.
The power output can then be calculated to be As typical values for Oh are in the 2- 10' range, it can be seen
that significant reduction in transformer size are possible by
switching to a scheme where the leakage inductances are the current
$
P =-dsin$
(33) transfer elements. This comparison is funher strengthened when the
losses resulting from interaction of diode reverse recovery and
O XI leakage inductance are considered for a current source output dcldc
converter.
which is identical to that for a synchronous machine. Using
equation 8, the KVA of the composite transformer can be found to ComDarison of pro ed Convert
be
The three converters proposed have been presented in detail
v2 If2 including sufficient infomation for the development of operating
( K v A ) ~1 d 1(-2d2COS2$+d4) +( I - ~ ~ c o s Q + ~ ~(34)
]] characteristics. In the interests of maintaining reasonable paper
length, only a small fraction of the converter curves are presented.
Ideal Transformer Leakage In order to better compare the three topologies, Table 1 presents the
detailed specifications for various components based on the
Once again, this includes leakage and ideal transformer terms with equations and curves presented in the paper. The design is
the assumption that the equivalent composite transformer will be at denormalized so as to conform to a 100 KW specification at 500
least as small as the two components taken separately. Volts 1200 Amps dc.
As expected, maximum power is transfered at Q = ~ 1 2a, Examining peak device stresses, the 3 phase dual bridge
result corroborated by Figures 8a and loa. Once again, if the offers the lowest (Vce LJ stress at (1.265 load KW) as opposed
KW/(KVA)T is plotted against Q for different d, d=l shows the to a factor of 3.45 for converter A. Converter B is reasonably good
possibility of maximizing that ratio. Plotting Po versus (KVA)Talso at 1.35. Similar conclusions are seen from the transformer (KVA)T
shows similar shapes with less than 10% error due to harmonic ratings. Converter C needs a transformer which is 30% smaller than
components being neglected. Consequentlly, the fundamental for converter A and 10% smaller than for converter B. Both dual
model is assumed to be sound and gives a good first estimate on bridge converters compare reasonably well until we come to input
system design. and output filters. The total capacitive KVA needed for converter C
Considering the issue of transformer size may also be easier is 40.7 KVA as opposed to 141.6 KVA for converter B and 226
in the fundamental model. Assume that for a given (KVA)T, KVA for converter C, a truly substantial reduction. The r.m.s
transformer size is inversely proportional to the frequency, given the current ratings calculated for capacitors are well within the capability
core material. Given the task of designing a transformer for a of multilayer ceramic capacitors available commercially, provided
conventional hard switched dcldc converter, one chooses d = 1 and proper packaging is done.
L small for maximizing transformer utilization. This implies a small A value of OL has been specified in order to attain the
value for $I = Oh. If this transformer is designed for frequency 0 = desired specifications. The actual choice of o and L will depend on
Wh, then from equation 33 and 34, we get the core material and detailed transformer design. A dual bridge 3
phase converter is presently being fabricated at the 50 KW level and
detailed experimental results will be presented in a later paper. It is
anticipated that limitations on transformer size minimization will be
(35) imposed by the weak scaling factors that govern how the leakage
inductances reduce with size as operating frequency is increased.
Although it has been shown that the dual bridge converters are
inherently capable of higher power density, converter A has many
803

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on June 06,2020 at 13:39:12 UTC from IEEE Xplore. Restrictions apply.
 TABLE 1

Converter A Converter B Converter C

single phase single phase three phase
phase shift dual bridge dual bridge

1. Converter- (Vs) 862.07V 500v 500v

a. Power rating (Po) l00KW l00KW l00KW
b. Output voltage (Vo) 500V (d=0.58) 500V (d=l) 500V (d=l)
c. Output current (Io) 200A 200A 200A
d. @ 37.8' 45.75' 50.1'
e. 13 0 - 180'
2. Device S-
a. Type of device MCT MCT MCT
b. No of devices 4+4 (diodes) 4+4 = 8 6+6 = 12
c. Peak voltage Vs = 862.07V vs = 500v v s = 500V
d. Peakcurrent 401.02A 270.68A 253.45A
e. (Vpk * Ipk)/KW 3.45 1.35 1.265

3 . T r a n s f o r m x S ~ aOL
~ 2.24R 1.475fi 1.15R
a. Max. primary volts 862.07V 500v 2Vsf3 = 333V
b. Peak primary current 401.02A 270.68A 253.45A
c. Max. sec. volts 500v = v o 500v 2Vof3 = 333V
d. Max. sec. current 40 1.02A 270.68A 253.45A
e. KVA 298.59 229.83 207.83
f. RMScument 2 17.06A 244.07A 162.78A
4.
a. Input filter
i. Capacitor voltage Vs = 862.07V 500v 500v
ii. Capacitor rms current 196.27A 141.02A 40.17A
iii. Capacitor KVA 169.2 70.5 1 20.1
b. Output filter
i. Capacitor voltage v o = 500v 500v 500v
ii. Capacitor rms current 113.53A 142.37A 41.3A
iii. Capacitor KVA 56.77 71.19 20.65

desirable characteristics. For higher output voltages, where active The three phase dual bridge converter proposed has the most
devices may be unable to operate, diode rectifiers may be the only favorable characteristics including:
alternative along with converter A. Further, for applications where
the reliability or cost penalties of additional gate turn-off devices is - small number of components
unacceptable, again converter A may be the only viable option. - low device and component stresses
- zero (or low) switching losses for all devices
Conclusions - small filter components
- high efficiency (no trapped energy)
Three new dc/dc converter topologies suitable for high - high power density using a symmetrical 3 phase
power-density high power applications have been presented in this transformer
paper. All three circuits operate in a soft switched manner making - bidirectionalpower flow
possible a reduction of device switching losses and an increase in - buck-boost operation possible
switching frequency. Along with soft-switching, all proposed - low sensitvity to system parasitics
circuits utilize the leakage reactance of the ac link transformers as - parallel operation possible as a result of current transfer.
active current transfer elements and eliminate problems of interaction
between these leakage inductances and diode reverse recovery. The Acknowledgements
dual bridge topologies are also capable of buck-boost operation and
bi-directional power flow, although that aspect has not been This project has been supported through a grant from the
analyzed in detail in this paper. National Aeronautics and Space Administration, NASA Lewis
The current transfer mode of operation makes it easier to Research Center, Cleveland, OH.
parallel multiple modules for extending the power capacity of the
system. The use of a three phase ac link system dramatically
reduces the capacitor ripple curzents making it possible to use high References.
power density multi-layer ceramic capacitors. The dual bridge
converters are also seen to offer an unexpected gain in the power [l]. F.C. Schwarz, J.B. Klaassens, "A Controllable 45-kw Current
density attainable as a result of the controlled action of the two Source for DC Machines", IEEE, Transactions IA, Vol. IA-15,
bridges. As the snubbers used'are purely capacitive, these would NO. 4, JulyJAugust 1979, pp.437-444.
supplement the internal device capacitance, giving a clean power
structure. The total number of system components is also seen to be [2]. D.M. Divan, 0. Patterson, "A Pseudo Resonant Full Bridge
minimal - the input and output filter capacitors, two bridges and one DC/DC Converter", IEEE-PESC 1987, Conf. Rec. pp. 424-
transformer. All device and component parasitics are seen to be 430.
used favorably.

804

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on June 06,2020 at 13:39:12 UTC from IEEE Xplore. Restrictions apply.
 [3]. A.S.Kislovski, "Half Bridge Power Processing Cell Utilizing a
Linear Variable Inductor and Thyristor Dual Switches", IEEE-
PESC 1988 Conf. Rec., pp. 284-289.
[4]. K.H. Liu & F.C. Lee, "Zero-Voltage Switching Technique in
DC/DC Converters", IEEG-PESC Conf. Rec., pp. 58-70, June
1986.
[5]. W.A. Tabisz and F.C. Lee, "Zero Voltage Switching Multi-
Resonant Technique - A Novel Approach to Improve
Performance of High Frequency Quasi-Resonant Converters,
"IEEE-PESC 1988 Conf. Rec., pp. 9-17.
[6]. D.M. Divan and G. Skibinski, "Zero Voltage Switching Loss
Inverters for High Power Applications", IEEE-IAS Conf.Rec.,
1987, pp. 627-634.
[7]. D.M. Divan, "Diodes as Pseudo Active Elements in High
Frequency DCDC Converters", IEEE-PESC 1988 Conf. Rec.,
pp. 1024-1030,
[8]. D.M. Divan, G. Venkataramanan and R. DeDoncker, "Design
Methodologies for Soft Switched Inverters", to be published,
IEEE-IAS 1988 Conf. Rec.
[9]. R.L. Steigerwald, "High-Frequency Resonant Transistor
DC/DC Converters", IEEE Transactions on Industrial
Electronics, Vol. IE-31, N0.2, May 1984, pp. 181-191.

805

Authorized licensed use limited to: NATIONAL INSTITUTE OF TECHNOLOGY CALICUT. Downloaded on June 06,2020 at 13:39:12 UTC from IEEE Xplore. Restrictions apply.