Modular Power Electronic Converters (PDFDrive)

Aalborg Universitet
Modular Power Electronic Converters in the Power Range 1 to 10 kW

Klimczak, Pawel
Publication date:
2009
Document Version
Publisher's PDF, also known as Version of record
Link to publication from Aalborg University
Citation for published version (APA):

Klimczak, P. (2009). Modular Power Electronic Converters in the Power Range 1 to 10 kW. Aalborg: Institut for
Energiteknik, Aalborg Universitet.
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
? Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
? You may not further distribute the material or use it for any profit-making activity or commercial gain
? You may freely distribute the URL identifying the publication in the public portal ?
Take down policy
If you believe that this document breaches copyright please contact us at vbn@aub.aau.dk providing details, and we will remove access to
the work immediately and investigate your claim.
Downloaded from vbn.aau.dk on: januar 23, 2018

Modular Power Electronic Converters
in the Power Range 1 to 10 kW
by
Pawel Klimczak
Aalborg University, Denmark

Institute of Energy Technology
November 2009
i
Acknowledgements
First of all I’d like to thank my family, especially my wife, for all the help they
gave me and for their patience during the last period I wrote the thesis.
I’d like to thank my supervisor Associate Professor Stig Munk-Nielsen for his
advices and guides he gave me during the project period. I’d like to thank
involved in the project: Uffe Borup (Danfoss Solar Inverters A/S), Paul Thøgersen
(KK-Electronic A/S), Henrik Kragh (Grundfos Management A/S), Christian Wolf
(Grundfos Management A/S), Michael Andersen (Danmarks Tekniske Universitet),
Mads Lundstrøm (IRD A/S) and Klaus Moth (APC Denmark) for their input and
their response to my work.
I owe a special thank to my fellow student Morten Nymand for fruitful
discussions we had.
I appreciate financial support from Dansk Energi net under PSO project no.
562/06-14-26808.
During my study I’ve stayed at Kassel University, Germany for three months in
2008. I’d like to thank Prof. Peter Zacharias for his supervision during my study
abroad. Also I owe a special thank to Benjamin Sahan and Samuel Vasconcelos
Araujo for their help in Kassel.
Finally, I’d like to thank my colleagues and fellow students at Aalborg
University – we had many interesting discussions and some fun too.
Thank everyone
Pawel Klimczak
iii
Abstract
Thanks to CO2 emission reduction policies and increasing prices of fossil fuels a
significant growth in field of sustainable energy sources (SES) is being observed
during last decade. A government support and take-off projects in Europe and US
shall ensure an increasing trend in future too. Some of SES based plants , like
hydro-, geothermal-, biofuel-plants, use synchronous generators directly connected
to the grid. But some other SES technologies, like fuel cell or photovoltaic, require
a power electronic converter between the energy source and the load or the grid.
Work presented in this thesis concentrates on dc-dc non-isolated converters suitable
for high voltage gain applications, like uninterruptible power supply (UPS) and
some of sustainable energy sources. A special attention is on reduction of power
losses and efficiency improvements in non-isolated dc-dc step-up converters.
During literature study many different non-isolated dc-dc step-up topologies
were found, however not all of them are desired for high voltage gain applications.
It’s found that converters based on an inductor and a coupled-inductor principle (a
boost and a center tapped boost converters) as well as converters derived from
isolated converters (a non-isolated flyback-boost, a non-isolated push-pull-boost and
a non-isolated two-inductor-boost converters) are good candidates for future
investigation. Analysis and comparison of selected, most promising topologies
indicated that a non-isolated push-pull-boost and a non-isolated two-inductor-boost
converters are the best candidates for applications requiring a high voltage gain.
Design of a high efficiency converter requires a detailed knowledge and accurate
prediction of power losses. For this purpose average steady-state models of
selected topologies and component loss models are developed and implemented in
MATLAB. Converter models base on analysis of ideal waveforms and are built-up
from set of equations describing values essential for power loss calculation, e.g.
average or rms current values. These data are used by component models to
calculate losses in particular components. It’s important that component models
use parameters from datasheets in most cases. It enables performance comparison
of different topologies as well as comparison of different components. The
proposed modeling approach was verified using a basic boost converter breadboard.
With small modification these models may be used for design purposes, like search
for optimum output power and optimum switching frequency for given topology and
given MOSFETs.
Using developed tools and models the converter breadboard was designed. The
breadboard demonstrated a very high efficiency, comparable with present state-of-
the-art isolated converters.
A modular converter concept and its influence on a fuel cell converter overall
efficiency were investigated too. Based on simulation and measurement results it
iv
was demonstrated that the parallel modular converter used in a fuel cell application
achieves a high efficiency over wide range of the output power. Moreover, the
efficiency increases while the output power decreases, which is opposite to a solid
converter solution.
v
Contents
Acknowledgements ........................................................................................................... i
Abstract............................................................................................................................ iii
Contents ............................................................................................................................ v
Abbreviations and Symbols............................................................................................. ix
Abbreviations .............................................................................................................. ix
Style of writing ............................................................................................................ ix
Symbols ....................................................................................................................... ix
Chapter 1 Introduction .................................................................................................. 1
1.1 Background and motivation ............................................................................. 1
1.2 Potential applications........................................................................................ 2
1.2.1 Fuel cell .................................................................................................... 2
1.2.2 Photovoltaic .............................................................................................. 3
1.2.3 Uninterruptible power supply................................................................... 3
1.3 Overview on a non-isolated system.................................................................. 3
1.4 Problem definition ............................................................................................ 5
1.4.1 Project limitations..................................................................................... 6
1.4.2 Used tools ................................................................................................. 6
1.5 Thesis outline.................................................................................................... 6
1.6 List of publications ........................................................................................... 8
Chapter 2 Non-Isolated Dc-Dc Converters ................................................................... 9
2.1 General overview on non-isolated dc-dc step-up converters........................... 9
2.2 Review of topologies ...................................................................................... 11
2.2.1 Basic boost converter ............................................................................. 11
2.2.2 Multi-phase boost converter ................................................................... 13
2.2.3 Cascaded boost converter ....................................................................... 17
2.2.4 Center tapped boost converter ................................................................ 19
2.2.5 Non-isolated flyback-boost converter .................................................... 22
2.2.6 Non-isolated push-pull-boost converter ................................................. 24
2.2.7 Non-isolated two-inductor-boost converter............................................ 26
2.3 Summary......................................................................................................... 29
Chapter 3 Power Losses in Dc-Dc Converters............................................................ 33
3.1 Overview on model levels .............................................................................. 33
3.2 Modeling approach ......................................................................................... 34
3.2.1 Currents in the circuit ............................................................................. 36
3.3 Transistor ........................................................................................................ 37
3.3.1 Conduction loss ...................................................................................... 37
3.3.2 Switching loss......................................................................................... 37
3.3.3 Gate loss ................................................................................................. 40
3.4 Diode .............................................................................................................. 40
3.4.1 Conduction loss ...................................................................................... 41
vi
3.4.2 Reverse recovery loss ............................................................................. 41

3.4.3 Forward recovery.................................................................................... 43
3.4.4 Capacitive loss........................................................................................ 43
3.5 Magnetic device.............................................................................................. 44
3.5.1 Copper loss ............................................................................................. 44
3.5.2 Core loss ................................................................................................. 48
3.5.3 Types of magnetic devices and their losses............................................ 50
3.6 Capacitor......................................................................................................... 51
3.7 Other losses .................................................................................................... 53
3.8 Model verification .......................................................................................... 53
Chapter 4 Design and Optimization of the Converter................................................. 57
4.1 Converter analysis .......................................................................................... 57
4.1.1 Voltage gain............................................................................................ 59
4.1.2 Duty cycle below 50%............................................................................ 60
4.1.3 Currents .................................................................................................. 62
4.1.4 Inductor and transformer ........................................................................ 64
4.1.5 Summary................................................................................................. 66
4.2 Design............................................................................................................. 66
4.2.1 Power MOSFETs.................................................................................... 67
4.2.2 Power diodes .......................................................................................... 70
4.2.3 Magnetic components............................................................................. 72
4.2.4 Transformer design................................................................................. 73
4.2.5 Inductor design ....................................................................................... 77
4.2.6 Integrated inductor-transformer.............................................................. 84
4.2.7 Capacitors ............................................................................................... 92
4.3 Summary......................................................................................................... 93
Chapter 5 Modular Converter ..................................................................................... 95
5.1 Scaling of converters ...................................................................................... 95
5.1.1 Transistor ................................................................................................ 95
5.1.2 Inductor................................................................................................... 97
5.1.3 Transformer ............................................................................................ 99
5.1.4 Capacitor............................................................................................... 101
5.2 Overview on modules interconnection ......................................................... 102
5.3 Fuel cell converter ........................................................................................ 103
5.4 Modular converter ........................................................................................ 105
5.4.1 Interleaved PWM generation................................................................ 108
5.5 Summary....................................................................................................... 110
Chapter 6 Experimental results ................................................................................. 111
6.1 Experimental setup and breadboards............................................................ 111
6.2 Magnetic components measurement ............................................................ 114
6.3 Efficiency measurement and estimation....................................................... 116
6.4 Summary....................................................................................................... 119
Chapter 7 Conclusion................................................................................................ 121
7.1 Conclusions and new aspects ....................................................................... 122
Appendix A Publications .......................................................................................... 125
A.1 A Single Switch Dual Output Non-Isolated Boost Converter (APEC 2008)126
A.2 Comparative Study on Paralleled vs. Scaled Dc-dc Converters in High
Voltage Gain Applications (EPE-PEMC 2008) ....................................................... 131
vii
A.3 Boost Converter with Three-Stage Switching Cell and Integrated Magnetics
(APEC 2009) ............................................................................................................ 137
A.4 High Efficiency Boost Converter with Three State Switching Cell (PCIM
2009) 143
A.5 Integration of Magnetic Components in a Step-Up Converter for Fuel Cell
(EPE 2009) ............................................................................................................... 149
Appendix B Models .................................................................................................. 159
B.1 Transistor loss model (lib_trans.m) .............................................................. 159
B.2 Transistor data library (data_trans.m) .......................................................... 160
B.3 Diode loss model (lib_diode.m) ................................................................... 164
B.4 Diode data library (data_diode.m)................................................................ 164
B.5 Magnetic device general loss model (lib_mag.m)........................................ 166
B.6 Magnetic core library (data_magcore.m) ..................................................... 167
B.7 Magnetic material library (data_magmat.m) ................................................ 167
B.8 Boost converter main file (main_boost.m) ................................................... 168
B.9 Boost converter model (conv_boost.m)........................................................ 169
B.10 Non-isolated push-pull-boost converter main file (main_pushpullboost.m) 170
B.11 Non-isolated push-pull-boost converter model (conv_pushpullboost.m) .... 171
B.12 C-code used in the microcontroller .............................................................. 175
Bibliography ................................................................................................................. 181
ix
Abbreviations and Symbols
Abbreviations
CCM continuous conduction mode
DCM discontinuous conduction mode
ESR equivalent series resistance of a capacitor
FC fuel cell
FEM finite element method
FOM figure of merit
HV high voltage (about the output voltage)
IC integrated circuit
IGBT insulated gate bipolar transistor
LV low voltage (significantly lower than the output voltage)
MOSFET metal-oxide semiconductor field-effect transistor
PV photovoltaic source
SES sustainable energy source(s)
SVS switched voltage source inverter
UPS uninterruptible power source
VHV very high voltage (significantly higher than the output voltage)
Style of writing
i or i(t), etc. instantaneous values
I(rms), etc. rms values of ac and dc components
I, etc. average, dc values
I(ac), etc. rms values of ac component
I(pp), etc. peak-to-peak values
∆i, etc. amplitude values
Symbols
Symbol Description Unit
A area m2
AP area product values of magnetic core m4
x
B flux density, flux density T

C capacitance F
D transistor duty cycle %
δ penetration depth m
feq equivalent frequency Hz
fr repetition frequency Hz
fs switching frequency Hz
FR ac to dc resistance factor -
Φ magnetic flux Wb
gfm transistor forward transconductance S
h layer thickness m
i, I current A
IC capacitor current A
ID diode current A
Iin input current A
IL inductor current A
Iout output current A
IT transistor current A
l length m
L inductance H
LM magnetizing inductance H
λ volt-seconds applied to the winding V·s
M voltage gain or number of layers -
µ permeability H/m
n turns ratio, number of turns (inductor) -
n 1 , n2 number of turns (transformer) -
N number of phases -
η efficiency or porosity factor %
pFe specific core loss (per volume unit) W/m3
P power or power loss W
PC capacitor loss W
PCuac ac copper loss W
PCudc dc copper loss W
PDcap diode capacitive loss W
PDcond diode conduction loss W
PDoff diode turn-off loss W
PDon diode turn-on loss W
PFe core loss W
xi
Pin input power W

Pout output power W
PTcond transistor conduction loss W
PTgate transistor gate drive loss W
PTrr loss due to diode reverse recovery W
PTsw transistor switching loss W
Q charge C
QG gate charge C
Qrr reverse recovery charge C
R resistance Ω
rDS(on)T on-state resistance temperature coefficient Ω/ºC
rFT forward resistance temperature coefficient of diode Ω/ºC
RDS(on) transistor on-state resistance Ω
RDS(on)25 transistor on-state resistance at 25ºC Ω
RF diode forward resistance Ω
RF25 diode forward resistance at 25ºC Ω
RL1 inductor winding resistance Ω
Rn transformer winding resistance Ω
ρ resistivity Ω·m
t time s
tif current fall time s
tir current rise time s
tvf voltage fall time s
tvr voltage rise time s
T temperature or time period ºC, s
Tj junction temperature ºC
W energy or energy loss J
v, V voltage V
vDT forward voltage temperature coefficient of diode V/ºC
VGS gate-source voltage V
VD diode anode-cathode voltage V
VD(F) diode forward voltage (for loss calculation) V
VD(R) diode reverse voltage V
Vin input voltage V
Vn winding voltage V
Vout output voltage V
VT transistor drain-source voltage V
VT(off) transistor blocking voltage
xii
1
Chapter 1
Introduction
Work presented in this thesis concentrates on dc-dc non-isolated converters
suitable for high voltage gain applications, like uninterruptible power supply (UPS)
or some of sustainable energy sources (SES). Three main issues are presented.
First, state-of-the-art in dc-dc non-isolated step-up converters is presented. Several
topologies are presented, its features are discussed in general and the most
promising solutions are selected for further investigation. An overview on power
loss distribution in relation to presented topologies it given too. Second, the most
promising topology is chosen. The converter breadboard is being designed,
optimized and built. Laboratory tests are performed to verify its performance and
efficiency. Finally, parallel operation of multiple dc-dc modules is investigated.
It includes interaction between modules and a controller, optimum utilization of a
single module and final tests.
First, this chapter will discuss background for the research work made in this
thesis. It includes introduction to and discussion on possible applications and
issues related to a non-isolated grid connected system. Next, the initial problem is
defined and preliminary limitations are specified. The outline of the thesis and the
list of publications are presented at the end of this chapter.
1.1 Background and motivation

Thanks to CO2 emission reduction policies and increasing prices of fossil fuels a
significant growth in field of sustainable energy is being observed during last
decade. A government support and take-off projects in Europe and US shall ensure
an increasing trend in future too [1, 2]. SES is a wide term and it covers all
renewable energy sources such as biofuels, solar power, wind power, wave power
and geothermal power. Hydrogen power and sometimes nuclear power are
considered as SES too. Some of SES based power plants, including hydro-,
geothermal-, biofuel-plants, use synchronous generators directly connected to the
grid. But some other sources like fuel cells (FC) and photovoltaic (PV) basically
produce a dc voltage which has to be converted to an ac voltage suitable for the grid.
So, an interface between the source and the grid is a need and the power electronic
converter is the interface.
This project, entitled “Modular power electronic converters in the power range 1
to 10 kW”, is being realized in cooperation between Aalborg University, Danmarks
Tekniske Universitet and companies: Grundfos Management A/S, KK-Electronic
2
A/S, IRD Fuel Cell Technology A/S and Danfoss Solar Inverters A/S (former
PowerLynx A/S). Each of these companies has interest in development and
implementation of highly efficient power electronic converter, which converts a low
dc voltage to 50/60 Hz ac voltage required by utility grid or commercial loads. The
converter like this will be suitable for sustainable energy sources, like fuel cells and
photovoltaic and for UPS systems in the power range from few kilowatts to tens of
kilowatts.
1.2 Potential applications

This project focuses on high voltage gain applications in which a step-up dc-dc
converter must be used. It includes sustainable energy, especially fuel cells and
photovoltaic as well as battery based UPS systems. This section will present
general aspects of these applications.
1.2.1 Fuel cell

In the simplest words a fuel cell is an electrochemical device which converts
chemical energy of a fuel and an oxidant directly into electrical energy and heat
available for users. Basically all fuel cells operate on the same principle, however
they use different electrolytes materials. Most recognized kinds of fuel cells are:
proton-exchange membrane fuel cell (PEMFC), alkaline fuel cell (AFC), phosphoric
acid fuel cell (PAFC), molten carbonate fuel cell (MCFC), solid oxide fuel cell
(SOFC) and direct methanol fuel cell (DMFC). Another classification bases on the
operating temperature of a fuel cell. Low temperature fuel cells operate below
200˚C and are suitable for portable and automotive applications and for low power
co-generation. Medium temperature fuel cells operate between 200˚C and 600˚C
and can be used for automotive applications and co-generation including combined
heat and power (CHP). Finally high temperature fuel cells operate above 600˚C
and the main application is CHP plants in the power range up to hundreds of
kilowatts. The no-load voltage of a single cell is limited to 1.23 V in an ideal case,
but in practice it’s about 1 V for hydrogen fuel. During normal operation the
voltage drops down to 0.7-0.5 V. To increase the output voltage cells are
connected in series and forms fuel cell stack. Due to practical problems related to
the cell voltage balancing and fuel/oxidant distribution limit number of series
connected cells and the stack output voltage [3, 4].
The fuel cell market is just growing. The major barriers for wide spread use of
fuel cells are a high cost, a short lifetime of a fuel cell and lack of a pure hydrogen
fuel. However a significant growth is predicted in next years. In Europe it’s
expected to install fuel cell plants with total power of 1 GW before 2015 [5, 6].
Nowadays there are several commercially available fuel cells. Most of them are
a low temperature PEMFC with the output power from few hundred watts up to few
kilowatts. The output voltage of such fuel cell usually is somewhere between 12-
60 V.
3
1.2.2 Photovoltaic
Photovoltaic cells are usually known as 'solar cells'. Photovoltaic cells work by
transforming the photon energy from solar radiation directly into electrical energy
without an intermediate mechanical process. There are many inorganic and
organic materials used to manufacture photovoltaic cells. The most spread
however is silicon [7]. Based on silicon crystal structure there are mono-crystalline
cells, poly-crystalline cells and thin-film (amorphous) cells. Among them mono-
crystalline PVs provide the highest efficiency, but they are the most expensive. On
the other end there are thin-film PVs which are fairly cheap, but have worse
performance. A single PV cell delivers the voltage up to 0.6 V at no-load
conditions. To increase the output voltage single cells are connected in series and
forms PV panels. Often panels are connected in series and forms PV strings [8].
Contrary to the fuel cell market, the PV market is well developed and it’s
growing fast. In 2008 there was about 15 GW of photovoltaic power installed
worldwide [9]. The same analysis predicts new PV plants with total power of 22
GW in 2013. Current research projects are primarily focused on cost reduction [1].
There are many PV panels manufacturers around the world. Most of
commercial PV panels have the output power in the range of 100-400 W and the
output voltage about 20-45 V (per single panel). To increase the power or the
voltage level panels are connected and they form strings and arrays.
1.2.3 Uninterruptible power supply

Uninterruptible power supply (UPS) is a device providing an emergency power
when grid is not available. UPS are primarily used for power back-up of:
computers, servers, telecom equipment and medical equipment. The output power
of UPS varies from few hundred volt-amps to megavolt-amps and the back-up time
vary from seconds to hours. UPS based on lead-acid batteries can deliver power
up to tens of kilovolt-amps and back-up time from few minutes up to few hours. A
high power battery based UPS usually incorporates so called double conversion.
Because of a high voltage battery banks used in such UPS there is no need for high
step-up voltage conversion. This solution provides the best protection but it’s the
most expensive one. Many of commercial low power UPS products are so called
off-line or standby UPS. These products usually have a single 12 V or 24 V lead-
acid battery, thus there is an obvious need for a high step-up conversion. UPS
application however creates one more challenge for a power converter. Contrary to
a fuel cell or a photovoltaic cell, a battery is a rechargeable source, which requires
bidirectional power flow. Bidirectional power flow in low power off-line UPS is
ensured in two ways – by use of a bidirectional converters or by use of a dedicated
battery charger next to the main converter.
1.3 Overview on a non-isolated system

Lack of an isolation between an input and an output of the non-isolated power
conversion system creates certain issues. One of the most important questions is
about a source grounding?
4
Figure 1.1 A non-isolated power conversion system with a full-bridge inverter
Figure 1.2 A non-isolated power conversion system with a half-bridge inverter
Figure 1.3 A non-isolated power conversion system with a SVS inverter
Figure 1.1 presents a non-isolated converter with a full-bridge grid connected

inverter. The source cannot be grounded in this case and it floats. Terminals of
the source may jump between positive and negative potential of hundred volts with a
high frequency. Usually it’s not a problem if the source is a battery pack in an
isolated enclosure. In case of floating PV panel one has to take care about ground
leakage current, especially in case of a large PV arrays . Proper grounding of a PV
panel terminal may reduce leakage ground current [10, 11]. Moreover such
grounding is required in some countries (e.g. in USA) [11-13].
If grounding of a source terminal is required usually the first choice would be a
system similar to the one presented on Figure 1.2. A half-bridge inverter is used
and the source has a solid connection with neutral point of the grid, which usually is
grounded in a load center [12, 13]. The main drawback of this solution is need for
twice dc-link voltage compared with the full-bridge inverter based configuration.
One way is to use of two series connected sources Vin1 and Vin2 together with two
independent dc-dc converters for each dc-link voltage. If only a single source Vin1
is available then an inverting converter (buck-boost) or a dual-output converter is
5
required [14-16]. A simple single-input dual-output step-up converter is presented

and discussed in Appendix A, paragraph A.1. Use of two sources will increase
cost of the whole system on one hand. On the other use of a single source requires
a higher voltage gain which turns into efficiency degradation and more complex
converter.
A power conditioning system with the switched voltage source (SVS) inverter
presented on Figure 1.3 may solve problems addressed above [17]. Such
configuration provides a solid connection between the source terminal and the grid
neutral wire. In the same time the SVS inverter requires the same dc-link voltage
like the full-bridge inverter. Reference [11] presents another single-stage non-
isolated power converters for systems with a grounded source.
1.4 Problem definition

Nowadays there are many different converter topologies and selection of the
most suitable one is not straightforward choice. Such selection is a difficult trade
off between converter’s complexity, estimated cost, size and weight, power level
and expected efficiency.
The goal of this project is to investigate features of different non-isolated dc-dc
step-up converters dedicated for fuel cell application and suitable for other high
voltage gain applications.
The aims of this project can be covered by the following statements:
• Can a non-isolated dc-dc converter work in a high voltage gain
application as well as an isolated transformer based converter do?
• Which non-isolated dc-dc topology is the best candidate for fuel cell
application?
• How a modular converter can improve performance of the fuel cell
converter?
In order to achieve these aims the work is performed in several areas.
Following main areas are covered in this thesis:
• Modeling – averaged models of various dc-dc step-up converters are
developed. Also particular components and parts of dc-dc converters are
modeled in relation to power loss and heat dissipation in these
components. Combining averaged circuit model and component model
will result in robust modeling tool to estimate power losses and efficiency
of the converter circuit under specific operational conditions.
• Analysis – different power electronic circuits are analyzed and their
features and performance are evaluated on a common basis. Efficiency
is one of the most important factors.
• Optimization – selected converters are being designed and optimized
according to project limitations.
6
• Modularization – parallel operation of several modules and its impact on

overall system performance is investigated. Special attention is on
efficiency improvement by intelligent usage of modules.
1.4.1 Project limitations

The work performed in this thesis is done under following conditions:
• The focus is on dc-dc non-isolated step-up converters only
• The focus is on the converter’s power stage only
• Input voltage range is 30 V to 60 V dc
• Output voltage is 400 V dc (for the SVS inverter) or ±400 V dc (for the
half-bridge inverter)
• Power level range is from few hundred watts up to few kilowatts per
module
• Converter design incorporates current state of the art components and
materials
• Best case efficiency target is 98%
• Input current ripples should be reduced
1.4.2 Used tools

The general purpose simulation program MATLAB/Simulink® is chosen as the
main simulation and calculation tool in this project. It supports equation based
calculations prepared as script files as well as graphical signal flow simulations
prepared as Simulink models.
Power electronic circuit simulations are supported by PLECS® [18] which is a
toolbox for MATLAB/Simulink®. It allows control of circuit simulations by
simple, often iterative scripts thus many time-consuming simulations are done
automatically.
For more advanced design and optimization of magnetic components a simulator
called FEMM 4.0.1 [19] is used. This software incorporates finite element
methods (FEM) for electromagnetic fields simulations. It’s a simple and robust
tool providing reliable results within a short time. More advanced users may use
FEMM 4.0.1 software in connection with MATLAB® and prepare scripts i.e. for an
inductor winding or an air gap optimization.
1.5 Thesis outline

The thesis is divided into seven chapters and an appendix. It begins with broad
overview on applications for dc-dc step-up converters and important aspects of non-
isolated ones. Various converter topologies are introduced too. Next, different
power loss mechanisms present in dc-dc converters are described and reader is
guided thru optimization process. Then a modular converter concept is introduced
and it’s demonstrated how the modular converter can improve a power conversion
efficiency.
7
• Chapter 1 - Introduction
This chapter. Includes background, motivation, application description,
problem definition and project limitations. Also it describes general tools
used.
• Chapter 2 - Non-Isolated Dc-Dc Converters
An overview on different approaches for boosting of a dc voltage and current
state-of-the-art non-isolated step-up dc-dc converters is given. Next, more
detailed description and analysis of selected converters is done. Then
converters are compared against basic boost converter and against each other.
Required passive components, number and utilization of semiconductors, etc.
are taken into account. These factors approximate future converter size and
cost in compare with other solutions.
• Chapter 3 - Power Losses in Dc-Dc Converters
In this chapter issues related to power losses in dc-dc converters are
discussed. First modeling approach is presented. Then sources of major
losses are pointed and appropriate models are introduced. These models
bases on references, but some of them require minor adjustments. Finally
suitable converter models are developed and simulation results are presented.
Results shall indicate potential peak efficiency of each topology, thus it’s
possible to justify them in terms usability in a high voltage gain and high
efficiency applications.
• Chapter 4 - Design and Optimization of the Converter
Optimization is a critical part of the converter design process. It results may
lead to good or bad design. This chapter describes optimization of the
selected converter with special focus on magnetic components.
• Chapter 5 - Modular Converter
Modularity of a power converter is not a new approach. First it was used in
telecom, space or military applications. Recently it can be found in
commercial products too. It’s well known that modular architecture reduces
costs of production and maintenance, improves reliability, enables system
scalability etc. In this chapter impact on system power conversion
efficiency is being investigated. An example application will be fuel cell
based UPS, which operates under variable load and variable input voltage, so
optimization of a solid converter become difficult.
• Chapter 6 - Experimental results
Report from the converter build-up process and laboratory work is presented
in this Chapter. First laboratory equipment and measurement setup is being
presented. Next, measured electrical values of magnetic components are
provided and compared with calculated values. Finally, efficiency of three
breadboards is measured and presented.
• Chapter 7 - Conclusion
This chapter summarize and conclude the whole project, points out findings
and suggests future work.
8
• Appendix A - Publications
Conference papers related to and published during the project period by the
author of this thesis are provided to leader in this appendix.
• Appendix B - Models
In this appendix source code of converters and components models used are
provided. It also contains C-code of the multiphase PWM generator using
microcontroller.
1.6 List of publications

• P. Klimczak and S. Munk-Nielsen, "A single switch dual output non-isolated
boost converter," in Applied Power Electronics Conference and Exposition,
2008. APEC 2008. Twenty-Third Annual IEEE, 2008, pp. 43-47.
• P. Klimczak and S. Munk-Nielsen, "Comparative study on paralleled vs.
scaled dc-dc converters in high voltage gain applications," in Power
Electronics and Motion Control Conference, 2008. EPE-PEMC 2008. 13th,
2008, pp. 108-113.
• P. Klimczak and S. Munk-Nielsen, "Boost Converter with Three-State
Switching Cell and Integrated Magnetics," in Applied Power Electronics
Conference and Exposition, 2009. APEC 2009. Twenty-Fourth Annual IEEE,
2009, pp. 1378-1383.
• P. Klimczak and S. Munk-Nielsen, "High Efficiency Boost Converter with
Three State Switching Cell," in International Exhibition & Conference for
Power Electronics, Intelligent Motion, Power Quality PCIM Europe 2009
Nurnberg, Germany, 2009.
• P. Klimczak and S. Munk-Nielsen, "Integration of Magnetic Components in a
Step-Up Converter for Fuel Cell," in European Conference on Power
Electronics and Applications, 2009. EPE 2009. 13th Barcelona, Spain, 2009.
9
Chapter 2
Non-Isolated Dc-Dc Converters
There are many different applications for dc-dc boost converters. One of them
is a low/medium power fuel cell based UPS systems. In this application boost
converter is used to boost the low variable voltage from the fuel cell (or the battery)
and provide the high quality, regulated dc voltage to the inverter. For many years
isolated topologies with high frequency step-up transformer have been used
commonly. However, if there is no need for galvanic isolation between the input
and the output of the converter a non-isolated step-up converter might be an
interesting and beneficial solution.
Many non-isolated step-up converters were found during a literature study. In
following section a general overview on different voltage amplification techniques is
given first. Next, several step-up topologies are presented and described in details
– starting from the simplest boost converter and ending up with so called non-
isolated two-inductor boost converter.
2.1 General overview on non-isolated dc-dc

step-up converters
One approach for step-up converter bases on the switched-capacitor principle –
charging a number of capacitors in parallel and discharging them in series [20, 21].
However “pure” switched capacitor circuit require significant number of stages to
achieve high voltage gain. It means large number of elements (capacitors, diodes
and transistors). Moreover control of the output voltage is limited and voltage
conversion ratio depends strongly on the circuit configuration. The advantage of
this solution is absence of any magnetic devices, so the converter can be very
compact – could be even assembled as an integrated circuit (IC). In [22] switched
capacitor step-up converter is integrated with boost converter in order to have high
voltage gain with reasonable duty ratio and good efficiency. Also boost converter
has a regulation purpose, so output voltage is controlled well. Another switched
capacitor converter is proposed in [23] – in this paper basic concept bases on
resonant energy transfer among the switched capacitors. The converter voltage
transfer ratio is determined by the number of stages and fixed.
Another approach is presented in [5, 18, 24-28], where a family of converters
with a coupled-inductor (a tapped-inductor, an autotransformer) is introduced.
Using coupled inductors these converters can provide a high voltage gain without
10
extreme duty cycle. However the leakage inductance of the coupled inductor may
cause an additional voltage stress on the active switch, so the switching loss
increases and the converter’s efficiency decreases. Employing a snubber circuit
voltage stress (voltage spikes) on the active switch can be attenuated, thus a low
voltage rated transistor with a lower on-state resistance can be utilized. In the
references [5, 28] a boost converter with a high voltage gain (up to 20) and a very
good efficiency (up to 97%) is presented. This converter bases on the coupled-
inductor principle, but it incorporates a series capacitor in order to increase the
voltage gain. Also, a passive regenerative snubber recovers energy stored in the
leakage inductance, ensuring a good performance of this topology. Reference [29]
presents a boost converter with a coupled inductor and a voltage multiplier (a
voltage doubler). In this converter the output voltage is a sum of the output voltage
from the boost converter stage and the voltage multiplier stage.
A similar idea (a boost converter and a voltage multiplier) is presented in [30].
In this paper there is a number of parallel and interleaved strings. Each string
contains the boost converter and a number of voltage multiplier stages. However
to ensure a high voltage gain at a low duty cycle a number of multiplier stages is
required. So, for a higher power and a high voltage gain a large number of parallel
and series connected stages is used, so the structure become complex and potentially
expensive. Also paralleled strings have a lot of interconnections one to the others,
thus reliability of the whole system decreases.
A converter based on a three-state switching cell and a voltage doubler is
presented in [31] and it presents a very good efficiency. This converter provides a
high voltage gain, a reduced voltage stress on transistors and ensures reduced input
current ripples. This converter can be considered as a non-isolated push-pull-boost
converter. This topology is presented in details in section 2.2.6.
References [32-34] present a family of Luo converters with a positive and a
negative output voltage. Luo converters family comes from SEPIC and ZETA
(dual-SEPIC) converters. A voltage gain of the elementary circuit can be increased
by using a voltage lift technique and an auxiliary circuit. This auxiliary circuit
contains capacitors, diodes and inductors. There are self-lift, re-lift and multiple
lift circuits presented in the literature. Self-lift Luo converter in continuous
conduction mode (CCM) has the same voltage transfer function like the basic boost
converter in CCM, but Luo converter contains significantly more elements and it has
a large ripple in the input current. I.e. quadruple-lift Luo converter in CCM has
voltage gain 4 times higher than the basic boost converter in CCM, but Lou
converter is very complex now and it contains a lot of components (2 active
switches, 7 diodes, 6 capacitors and 5 inductors [34]). One can find more about
Lou converters and this voltage boost (voltage lift) technique in [35].
All of converter groups presented above are dc-dc converters which regulate
voltage in a dc-link. It means that in all cases power conversion has at least two
stages. A Z-source inverter is introduced in [36, 37]. It combines a boost
converter and an inverter in one stage converter. The reference [36] presents
comparison of the Z-source inverter and the boost-buck inverter. In this
comparison the boost-buck inverter achieves slightly better efficiency in whole
11
range of the output power. Number of components is similar in both converters,

but the Z-source inverter exerts serious voltage and current stress on the
components.
At the moment topology based on a coupled inductor or a three-state switching
cell seem to be good candidate for future investigation, because of its high
conversion ratio, proved high efficiency, low voltage transistors and overall low
component count. However converters with a coupled inductor reveal a common
drawback – large ripples in the input current.
2.2 Review of topologies

This section presents several basic non-isolated dc-dc step-up topologies. It
includes a converter diagram, main waveforms and short operation description.
Key parameters and equations are presented for each topology. Design equations
developed in following paragraphs are used at the end of the chapter to compare
presented topologies one against the other. Based on the comparison the best
canditade topology is selected for future investigation.
Because of requirement of a small input current ripple it is assumed that a
converter operates in CCM most of the time. Also it’s assumed that all capacitors
are large, so capacitor voltages have major dc component with a very small and
negligible ripple. Finally, for sake of simplicity it’s assumed that all circuits are
lossless, thus (2.1) and (2.2) are fulfilled.
Vout = M ⋅ Vin (2.1)
I in = M ⋅ I out (2.2)
2.2.1 Basic boost converter

Figure 2.1 presents a boost converter, which is a basic and well-known circuit
with voltage step-up capability. In CCM there are only two stages over a single
switching period and main waveforms are presented on Figure 2.2. During stage 1
the transistor T1 is turned-on, the input voltage Vin is being applied to the inductor
L1, the input current iin(t) increases and the inductor L1 stores energy. The diode D1
is reverse biased and the load Rout is supplied from the capacitor C1. During stage 2
the transistor T1 is turned-off and the diode D1 conducts and the inductor L1 passes
stored energy to the output capacitor C1. The input current iin decreases.
In CCM the converter’s voltage gain M is given by (2.3). The transistor T1
Figure 2.1 Diagram of a basic boost converter

12
Figure 2.2 Waveforms of a basic boost converter
blocking voltageVT1 and diode D1 reverse voltage VD1 are equal to the output voltage
Vout (under assumption of small output voltage ripple). The transistor rms current
IT1(rms) is given by (2.4). The diode average forward current is equal to the average
output current. Required power rating of the transistor and the diode in terms of
the output power Pout are given by (2.5) and (2.6) respectively. For a small input
current ripple and a very large duty cycle the rms current is approximated by the dc
value. Required inductance L1 depends on parameters like input voltage Vin,
switching frequency fs, duty cycle D and allowed input current ripple ∆iin and it’s
given by (2.7). Output capacitor C1 size is given by (2.8), where ∆vout is allowed
amplitude of the output voltage ripple.
1
M= (2.3)
1− D
2
1  ∆i 
I T1(rms) = I in ⋅ 1 + ⋅  in  ⋅ D ≈ M ⋅ I out (2.4)
3  I in 
volt×ampT1 = VT1(off) ⋅ I T1(rms) ≈ Vout ⋅ M ⋅ I out = M ⋅ Pout (2.5)
volt×amp D1 = Vout ⋅ I out ≈ Pout (2.6)

Vin ⋅ D
L1 = (2.7)
2 ⋅ ∆iL1 ⋅ f s
I out ⋅ D
C1 = (2.8)
2 ⋅ ∆vout ⋅ f s
This topology has an advantage of robustness and simplicity. However fairly
large inductor and very poor utilization of the transistor are main drawbacks of this
topology operating in a high voltage gain application.
13
Figure 2.3 Diagram of a 2-phase boost converter
2.2.2 Multi-phase boost converter

Use of several boost converters in parallel expands the output power of the whole
system, while the input current is shared between two or more converter modules.
Operation with interleaved switching scheme gives an advantage of reduction of the
input current ripple and it leads to inductor size reduction. Additionally the output
voltage ripple is reduced too, so the output capacitor C1 can be reduced. Figure 2.3
presents the two phase boost converter. It’s composed of two paralleled boost
converters which inputs and outputs are connected in parallel. If both converters
are identical and transistors are operated with the same duty cycles then the input
current is shared equally among paralleled converters.
The multi-phase boost converter operates in the same way like the basic boost
converter does. Figure 2.4 presents key waveforms of the two phase boost
converter. Gating signals for both transistors have the same duty cycle and they
are phase shifted by 180º (interleaving). One may observe that the input current
ripple is smaller then ripples observed in particular inductor currents. Main
challenge in case of paralleled boost converters is to provide exactly the same duty
cycle for all transistors. Even a small mismatch in duty cycles may lead to a
significant current sharing unbalance and reduce reliability of the system.
Different active and passive current sharing methods are discussed in [38].
In case of paralleled converters the voltage gain M does not change and it’s given
by (2.9). Also transistors’ blocking voltage and diodes’ reverse voltage doesn’t
change and are about the output voltage Vout. The transistor rms current is given by
(2.10). A significant inductor current ripple cannot be neglected this time and it
has an influence on the transistor power rating according to (2.11). Term N means
number of paralleled modules, sharing the input current. Required power rating of
the diode D1 is given by (2.12). The required inductance L1 is given by (2.13).
14
Figure 2.4 Waveforms of a 2-phase boost
It’s important to note that allowed inductor current ripple may become
significantly larger than allowed input current ripple (∆iL1>∆iin), thus required
inductance L1 can be reduced. The current ripple reduction for depends on number
of interleaved phases and duty cycle. Simulation results for 2, 4, and 8 phase
interleaved boost converter are presented on Figure 2.5. However one should be
aware that increased inductor current ripple may lead to increased ac copper losses
in the inductor winding and increased conduction losses in the transistor.
1
M= (2.9)
1− D
1
1 phase
0.8 2 phases
4 phases
8 phases
∆iin/Iin dc
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
D
Figure 2.5 Reduction of the relative input current ripple
in a multi phase boost converter
15
2 2
I 1  ∆i  M 1  ∆i 
I T1(rms) = in ⋅ 1 + ⋅  L1  ⋅ D ≈ ⋅ I in ⋅ 1 + ⋅  L1  (2.10)
N 3  I L1  N 3  I L1 
2
M 1  ∆i 
volt×ampT1 = VT1(off) ⋅ I T1(rms) ≈ Vout ⋅ ⋅ I in ⋅ 1 + ⋅  L1  =
N 3  I L1 
(2.11)
2
M 1  ∆i 
= ⋅ Pout ⋅ 1 + ⋅  L1 
N 3  I L1 
I out Pout
volt×amp D1 = Vout ⋅ ≈ (2.12)
N N
Vin ⋅ D
L1 = (2.13)
2 ⋅ ∆iL1 ⋅ f s
I out ⋅ D
C1 = (2.14)
2 ⋅ ∆vout ⋅ f s ⋅ N
Furthermore two inductors, L1 and L2, can be coupled together on a single
magnetic core resulting in a two phase boost converter with a coupled inductor, like
the one presented on Figure 2.6. The coupled inductor is presented as an ideal
transformer with two identical windings n1=n2, and a storage inductance LM in
parallel to the winding n1. Key waveforms of this converter are presented on
Figure 2.7. It is important to note that this converter operates with the transistor
duty cycle below 50%, i.e. without overlapping. In CCM there are four basic
operation stages over a single transistor switching period. During stage 1 both
transistors are turned-off and energy stored in the inductance LM is being released
thru diodes D1 and D2 to the output capacitor C1. The input current splits into two
identical halves and flows thru both windings n1 and n2. At the beginning of stage
2 the transistor T1 is turned-on. Input voltage Vin is applied to the winding n1.
The same voltage induces in the winding n2, but it’s too low to forward bias the
body diode of the transistor T2, so there is no current flow in the winding n2. During
this stage the whole input current iin(t) flows thru the winding n1. Energy is stored
in the storage inductance LM. Stage 3 is identical to the stage 1 while both
transistors are turned-off. Stage 4 is symmetrical to the stage 2, but the transistors
T2 is being turned-on and T1 remains turned-off.
16
Figure 2.6 Diagram of a two phase boost converter with a coupled inductor
Figure 2.7 Waveforms of a 2-phase boost converter with a coupled inductor
In CCM the voltage gain M is given by (2.15). The transistors’ blocking

voltage and diodes’ reverse voltage doesn’t change and are about the output voltage
Vout. The transistor rms current is given by (2.16) and the transistor power rating is
given by (2.17). Both diodes share the output current equally and the diode power
rating is given by (2.18). The effective frequency “seen” by the coupled inductor
LM and the output capacitor C1 is double of the switching frequency fs. It enables
use of a smaller inductor (2.19) and a smaller capacitor (2.20), in compare to the
basic boost converter.
1
M= (2.15)
1− 2 ⋅ D
2
1  ∆i 
I T1(rms) = I in ⋅ 1 + ⋅  in  ⋅ D ≈ M ⋅ D ⋅ I out (2.16)
3  I in 
volt×ampT1 = VT1(off) ⋅ I T1(rms) ≈ Vout ⋅ M ⋅ D ⋅ I out = M ⋅ D ⋅ Pout (2.17)
I out Pout
volt×amp D1 = Vout ⋅ ≈ (2.18)
2 2
17
Vin ⋅ D
L1 = (2.19)
2 ⋅ ∆iLM ⋅ fs
I out ⋅ D
C1 = (2.20)
2 ⋅ ∆vout ⋅ f s
The boost converter with the coupled inductor overcomes current sharing
problem [39] and with proper winding arrangement results in compact and efficient
design of the coupled inductor, even in spite of significant ac component in winding
currents [40].
2.2.3 Cascaded boost converter

Theoretically a basic boost converter is able to provide infinite voltage gain
while duty cycle approaches 100% according to (2.3). However the basic boost
converter has the voltage gain from 2-5 in products [41] up to 15-20 in breadboards
[42].
The limitation of the voltage gain comes from two major directions. On the one
hand parasitics present in the power circuit should be taken into account, while
resistive losses in the inductor, in the transistor and in the diode effectively limit the
maximum voltage gain [43]. Also the converter with a very large voltage gain
requires oversized transistor according to (2.5). Such high voltage transistor
usually exhibits a high on-state resistance. In connection with a large input current
results in a high conduction loss, which limits both, efficiency and rated power of
the converter module. On the other hand the voltage gain curve become steeper
while duty cycle increases. At one point it’s fairly difficult to control the
converter, while a very small variation in duty cycle D results in a large change in
the voltage gain M. This small variation in the duty cycle D may be desired, as a
response to change of operating conditions. But it can be undesired, e.g. caused by
delays in the driver circuit or different switching of the transistor due to the
temperature or ageing.
If a very high voltage gain is required it may be more beneficial to use of two or
more series connected (cascaded) boost converters, like presented on Figure 2.8.
This approach gives some advantages, but it creates new challenges in the same
time. Main advantages include a high voltage gain, a good power decoupling
between the output and the input, better utilization of semiconductors, presence of
an intermediate dc bus. Major drawbacks are more complex circuit, more complex
controls and a potential stability problem.
Figure 2.8 Diagram of a cascaded boost converter

18
The voltage gain of the cascaded boost converter operating in CCM is the
product of the voltage gain of each stage (2.21). The transistor T1 and the diode D1
have to handle the intermediate voltage VC1, while the transistor T2 and the diode D2
have to handle the output voltage Vout. Transistors’ rms currents are given by
(2.22) and (2.23), under assumption of a small input current ripple and a large
intermediate current ripple. Required power rating of both transistors is the sum of
respective power ratings and it’s given by (2.26). Similarly the requires power
rating of both diodes is given by (2.29). For a large voltage gain M cascading of
two or more boost converters lead to a significant reduction of the required
transistors power rating, but in the same time it increases required diodes power
rating by number of cascaded converter stages.
It’s important to note that “smooth” input current is ensured by the inductor L1
only (2.30). The capacitor C2 determines the output voltage ripple (2.33). The
intermediate voltage vC1(t) and the intermediate current iL2(t) may contain larger
ripples, so the capacitor C1 and the inductor L2 can be reduced – equations (2.32)
and (2.31) respectively. Moreover the transistor T1 can operate with higher
switching frequency then the transistor T2 can, i.e. fs1>fs2. It allows to reduce the
inductance L1 and capacitance C1 further.
1 1
M = M1 ⋅ M 2 = ⋅ (2.21)
1 − D1 1 − D2
2
1  ∆i 
I T1(rms) = I in ⋅ 1 + ⋅  in  ⋅ D1 ≈ M 1 ⋅ M 2 ⋅ D1 ⋅ I out (2.22)
3  I in 
2
1  ∆i 
I T2(rms) = I L 2 ⋅ 1 + ⋅  L 2  ⋅ D2 ≈
3  I L2 
(2.23)
2
1  ∆i 
≈ M 2 ⋅ D2 ⋅ 1 + ⋅  L 2  ⋅ I out
3  I L2 
Vout
volt×ampT1 = VT1(off) ⋅ I T1(rms) ≈ ⋅ M 1 ⋅ M 2 ⋅ D1 ⋅ I out = M 1 ⋅ D1 ⋅ Pout (2.24)
M2
2
1  ∆i 
volt×ampT2 = VT2(off) ⋅ I T2(rms) ≈ Vout ⋅ M 2 ⋅ D2 ⋅ 1 + ⋅  L 2  ⋅ I out =
3  I L2 
(2.25)
2
1  ∆i 
= M 2 ⋅ D2 ⋅ 1 + ⋅  L 2  ⋅ Pout
3  I L2 
volt×ampT = volt×ampT1 + volt×ampT2 =

 2 
 1  ∆iL 2   (2.26)
= M 1 ⋅ D1 + M 2 ⋅ D2 ⋅ 1 + ⋅   ⋅ Pout
 3  I L2  
 
19
Vout
volt×amp D1 = ⋅ I out ⋅ M 2 ≈ Pout (2.27)
M2
volt×amp D2 = Vout ⋅ I out ≈ Pout (2.28)
volt×amp D = volt×amp D1 + volt×amp D2 = 2 ⋅ Pout (2.29)
Vin ⋅ D1
L1 = (2.30)
2 ⋅ ∆iin ⋅ f s1
VC1 ⋅ D2
L2 = (2.31)
2 ⋅ ∆iL 2 ⋅ f s2
I L 2 ⋅ D1
C1 = (2.32)
2 ⋅ ∆vC1 ⋅ f s1
I out ⋅ D2
C2 = (2.33)
2 ⋅ ∆vout ⋅ fs2
In order to extend power level of the cascaded boost converter it’s possible to
parallel one or more cascaded stages.
2.2.4 Center tapped boost converter

A basic configuration of a center-tapped boost converter [25, 27] is presented on
Figure 2.9. The converter is similar to the two stage cascaded boost converter
(Figure 2.8) but there is no second transistor T2. Instead, two inductors L1 and L2
are magnetically coupled now and they form the coupled inductor, also called an
autotransformer or a center tapped inductor. In fact this coupled inductor works in
the way similar to a flyback transformer. The coupled inductor is modeled as a two
winding ideal transformer n1:n2 and a storage inductance LM connected in parallel
with the primary winding of the transformer. For sake of simplicity the leakage
inductance is neglected. The converter operates in CCM if the magnetizing current
iLM(t) (representation of the magnetic flux) is continuous. For simplified analysis
it’s assumed that there are only two major operating stages in CCM.
The converter main waveforms are presented on Figure 2.10. At the beginning
of stage 1 the transistor T1 is turned on and the input voltage Vin is being applied
across the winding n1 and the inductor LM. Rising input current flows thru storing
inductance LM and the transistor T1 only. Positive voltage vn2(t) induces in the
winding n2, the diode D2 is reverse biased, so there is no current flow in the winding
Figure 2.9 Diagram of a center-tapped boost converter

20
n2. Depends on operating point and the turns ratio n is may happen that induced
during this stage voltage vn2(t) is larger than the voltage vC1(t). In such case the
diode D2 experience the reverse voltage even larger than the output voltage Vout
(2.35). During stage 2 the transistor T1 is turned-off and energy stored in the
storage inductance LM is being released. Voltages across both windings n1 and n2
are negative now, so diodes D1 and D2 may become forward biased. In an ideal
case diode currents are almost identical and their average value is the same.
However an exact shape of these currents depend on several factors like:
capacitances, diodes forward voltage, winding resistance and leakage inductance,
etc.
The voltage gain of the converter is given by (2.36). Required transistor and
diode D1 voltage rating is reduced by the coupled inductor turns ratio n and it’s
given by (2.37). In the same time the diode D2 has to handle larger voltage which
is given by (2.38). In order to find transistor rms current, first the magnetizing dc
current ILM is found with (2.39). Assuming that the magnetizing current has a
small ripple then the transistor rms current is found with (2.40). Both diodes
conduct the same average current which is equal to the output dc current Iout.
Power rating of the transistor is given by (2.41) and the power rating of diodes is
given by (2.44). Required storage inductance LM is given by (2.45). The output
voltage ripple is limited by the capacitor C2 which capacitance is given by (2.46).
The capacitor C1 acts as a regenerative snubber. Its capacitance should be large
enough to store energy coming from the leakage inductance of the coupled inductor
and clamp the voltage vC1(t) at the certain level.
n2
n= (2.34)
n1
Figure 2.10 Waveforms of the center tapped boost converter

21
1
vC1 < vn 2 
→ <n (2.35)
1− D
1+ n ⋅ D
M= (2.36)
1− D
1
VT1(off) = VD1(R) = ⋅ Vout (2.37)
n ⋅ D +1
n
VD2(R) = ⋅ Vout (2.38)
n ⋅ D +1
n +1 n ⋅ D +1 n +1
I LM = I in ⋅ = I out ⋅ ⋅ (2.39)
n ⋅ D +1 1− D n ⋅ D +1
2
1  ∆i  n +1
I T1(rms) = I LM ⋅ D ⋅ 1 + ⋅  LM  ≈ ⋅ D ⋅ I out (2.40)
3  I LM  1 − D
volt×ampT1 = VT1(off) ⋅ I T1(rms) ≈

1 n +1 n +1 (2.41)
≈ ⋅ Vout ⋅ ⋅ D ⋅ I out = M ⋅ ⋅ D ⋅ Pout
n ⋅ D +1 1− D ( n ⋅ D + 1)
2
1 1
volt×amp D1 = ⋅ Vout ⋅ I out = ⋅ Pout (2.42)
n ⋅ D +1 n ⋅ D +1
n n
n ⋅ D +1 n ⋅ D +1
1+ n
volt×amp D = volt×amp D1 + volt×amp D2 = ⋅ Pout (2.44)
n ⋅ D +1
Vin ⋅ D
LM = (2.45)
2 ⋅ ∆iLM ⋅ f s
I out ⋅ D
C2 = (2.46)
2 ⋅ ∆vout ⋅ fs
Major advantages of this topology are a high voltage gain under reasonable duty
cycle and reduced voltage stress on the transistor. Unfortunately this topology has
several drawbacks. The main one is a large input current ripple determined by turns
ratio n. Also the diode D2 may exhibit increased voltage stress if (2.35) is fulfilled.
Finally, the winding arrangement of the coupled inductor may become difficult.
First, the storage inductance LM has to be fairly large, which in connection with a dc
bias result in a significant number of turns in the winding n1. The winding n2 has n
times more turns. Both windings have to have a low dc and ac resistance in the
same time, which is a challenge for a significant number of turns.
More detailed analysis of the converter, including presence of the leakage
inductance is presented in [25, 27]. Moreover similar solutions based on a coupled
inductors are presented in [5, 18, 24, 26, 28].
22
Figure 2.12 Diagram of a flyback-boost converter
2.2.5 Non-isolated flyback-boost converter

A very interesting modification of a flyback converter is presented on Figure
2.12 [19]. In comparison to the original flyback converter this one has two extra
components – the diode D1 and the capacitor C1. These additional components act
as a regenerative snubber for the transistor T1. It overcomes a problem related to a
leakage inductance of a flyback transformer. Series connection of capacitors C1
and C2 improves the voltage gain M in compare to the original flyback converter too.
A flyback transformer n1:n2 is modeled as an ideal transformer with a storage
inductance LM connected in parallel to the winding n1. For sake of simplicity the
leakage inductance is neglected. The converter operates in CCM if the
magnetizing current iLM(t) (representation of the magnetic flux) is continuous. This
simplified analysis includes only two major operating stages in CCM.
Key waveforms of the converter are presented on Figure 2.11. At the beginning
Figure 2.11 Main waveforms of the flyback-boost converter

23
of stage 1 the transistor T1 is turned on. The input voltage Vin is applied to the
winding n1 and the storage inductor LM. The diode D1 is reverse biased and the
input current flows thru the storage inductor LM and the transistor T1. Also a
positive voltage vn2 induced in the winding n2 and the diode D2 is reverse biased, so
the current in2 is zero during this stage. The load Rout is being supplied from output
capacitors C1 and C2 which are connected in series. During stage 2 the transistor T1
is off. Energy stored in the inductance LM is passed thru diodes to capacitors.
Both diodes conduct the same average current, but exact shape of each diode current
depends on several factors like: output capacitance, diode forward voltage, winding
resistance, leakage inductance and so on. In an ideal case both diode currents are
almost identical.
The voltage gain of the converter is given by (2.47). The transistor T1 and the
diode D1 blocking voltages are given by (2.48). The diode D2 blocking voltage is
given by (2.49) and this voltage may exceed the output voltage Vout in many cases.
The transistor T1 rms current is found with (2.51), where ILM is the magnetizing dc
current given by (2.50). Both diodes conduct the same average current equal to the
output dc current Iout. With these information required power ratings of all
semiconductors are found with equations (2.52) to (2.55). The converter requires
fairly large output capacitors since they are connected in series and the effective
output capacitance is given by (2.57). Finally (2.56) gives required value of the
storage inductance LM.
1+ n ⋅ D
M= (2.47)
1− D
1
VT1(off) = VD1(R) = ⋅ Vout (2.48)
1+ n ⋅ D
n
VD2(R) = ⋅ Vout (2.49)
1+ n ⋅ D
n +1
I LM = I out ⋅ (2.50)
1− D
2
1  ∆i  n +1
I T1(rms) = I LM ⋅ D ⋅ 1 + ⋅  LM  ≈ ⋅ D ⋅ I out (2.51)
3  I LM  1 − D
volt×amp T1 = VT1(off) ⋅ I T1(rms) ≈

1 n +1 n +1 (2.52)
≈ ⋅ Vout ⋅ ⋅ D ⋅ I out = M ⋅ ⋅ D ⋅ Pout
n ⋅ D +1 1− D ( n ⋅ D + 1)
2
1 1
n ⋅ D +1 n ⋅ D +1
n n
n ⋅ D +1 n ⋅ D +1
24
1+ n
volt×amp D = volt×amp D1 + volt×amp D2 = ⋅ Pout (2.55)
n ⋅ D +1
Vin ⋅ D
LM = (2.56)
2 ⋅ ∆iLM ⋅ f s
I out ⋅ D
Cout = (2.57)
This converter has the same voltage gain like the center tapped boost converter
presented in section 2.2.4. Also the flyback-boost converter has the same voltage
stress on the diode D2 but it requires larger output capacitors then the center-tapped
boost. The flyback-boost converter gives an idea how to transform some of
isolated topologies into non-isolated ones by adding the regenerative snubber
composed of the diode D1 and the capacitor C1. This approach overcomes
problems related to a leakage inductance of a transformer and improves voltage gain
in the same time.
2.2.6 Non-isolated push-pull-boost converter

A non-isolated push-pull-boost converter is presented on Figure 2.13. This
converter can be derived from a current-fed push-pull [43] converter using the same
approach like described in section 2.2.5. Additional components - diodes D1, D11
and the capacitor C1 - act as a regenerative snubber, recycling energy stored in a
leakage inductance of the push-pull transformer. Also series connection of all
capacitors improves the voltage gain M. In literature this topology is also called a
boost converter with a three state switching cell and a voltage multiplier [6, 31, 44].
Key waveforms of the converter are presented on Figure 2.14. Basically the
converter operates in CCM and with duty cycle above 50% (overlapping mode).
Under such conditions there are four major operating stages. Stages 1 and 3 are
identical. Both transistors are turned-on and share the input current equally, since
both primary windings are effectively connected anti-parallel and n1=n11. During
this period the input inductor L1 is being charged. No voltage induces in the
Figure 2.13 Diagram of a non-isolated push-pull-boost converter [6, 31, 44]

25
Figure 2.14 Main waveforms of the non-isolated push-pull-boost converter
winging n2 and there is no current flow in this winding too. Next, at the beginning
of stage 2 the transistor T1 is turned-off. The current in1 continuous to flow thru the
diode D1 to the capacitor C1. During this stage windings n1 and n11 are effectively
connected in series and the voltage vC1(t) is applied to them. Positive voltage vn2(t)
induces in the winding n2, the diode D3 is forward biased and the capacitor C3 is
recharged during this stage. The last stage 4 is similar to stage 2, but this time the
transistor T1 remains on and the transistor T11 is turned-off. During this stage
diodes D11 and D2 conduct and capacitors C1 and C2 are recharged.
The voltage gain of this converter is given by (2.59). Voltage stress of
transistors and diodes D1 and D11 is reduced and given by (2.60). Diodes D2 and
D3 have to handle reverse voltage given by (2.61). Assuming that the input current
had a very small ripple and the current in2(t) is almost rectangular, the transistor rms
current is given by (2.62). The average forward current of diodes D2 and D3 is
equal to the output dc current Iout. The average forward current of diodes D1 and
D11 equals half of the output dc current, since these diodes are connected in parallel.
Required power rating of transistors and diodes is given by equations (2.63) to
(2.67). Required storage inductance L1 is found with (2.68), while (2.69) gives the
effective output capacitance.
n2 n2
n= = (2.58)
n1 n 11
1+ n
M= (2.59)
1− D
1
VT1(off) = VT11(off) = VD1(R) = VD11(R) = ⋅ Vout (2.60)
1+ n
26
n
VD2(R) = VD3(R) = ⋅ Vout (2.61)
n +1
2 2
M ⋅ I out n I out   M ⋅ I out 
I T1(rms) ≈ (1 − D ) ⋅  + ⋅  + ( 2 ⋅ D − 1) ⋅   (2.62)
 2 2 1− D   2 
2
1 M n  M 
2 (2.63)
≈ ⋅ (1 − D ) 
⋅ +  (
+ 2 ⋅ D − 1)   ⋅ Pout
⋅
n +1  2 2 ⋅ (1 − D )   2 
volt×ampT = 2 ⋅ volt×ampT1 (2.64)

1 I 1
volt×amp D1 = ⋅ Vout ⋅ out = ⋅ Pout (2.65)
n +1 2 2 ⋅ ( n + 1)
n n
n +1 n +1
1+ 2 ⋅ n
volt×amp D = 2 ⋅ ( volt×amp D1 + volt×amp D2 ) = ⋅ Pout (2.67)
1+ n
Vin ⋅ ( D − 0.5 )
L1 = (2.68)
2 ⋅ ∆iin ⋅ fs
I out ⋅ D
Cout = (2.69)
Major advantages of this topology are: high voltage gain, limited input current
ripple and reduced voltage stress on transistors. In the same time the converter
reveals important drawbacks including: complicated design of the three-winding
transformer, three series connected output capacitors and number of diodes
effectively connected in series. However even in spite of these drawbacks this
topology seems to be a good candidate for a step-up converter for a fuel cell.
2.2.7 Non-isolated two-inductor-boost converter

A non-isolated two-inductor boost converter is presented on Figure 2.15. This
converter is derived from a two-inductor boost converter using approach presented
in section 2.2.5. Diodes D1, D11 and the capacitor C1 act as a regenerative snubber
for transistors T1 and T11.
Key waveforms are presented on Figure 2.16. Basically the converter operates
in CCM with duty cycle above 50%. There are four basic operational stages.
Stages 1 and 3 are identical. Both transistors are on, the input current iin(t)
increases and energy is stored in inductors L1 and L2. The winding n1 is essentially
shorted and no voltage induces in the winding n2. All diodes are reverse biased, so
the load is supplied from output capacitors. At the beginning of stage 2 the
transistor T1 is turned-off and the diode D1 starts to conduct because of continuous
inductor current iL1(t). The capacitor voltage vC1(t) is being applied to the winding
27
Figure 2.15 Diagram of a non-isolated two-inductor-boost converter

n1 and voltage n·vC1(t) induces in the winding n2. If this voltage is larger than the
capacitor voltage vC3(t) then the diode D3 become forward biased and conducts the
current in2(t) recharging the capacitor C3. Stage 4 is similar to stage 2 but the
transistor T11 is turned-off and T1 remains on. Diodes D11 and D2 conduct and
capacitors C1 and C2 are recharged during stage 4.
The voltage gain of this converter is given by (2.71). Voltage stress of
transistors and diodes D1 and D11 is given by (2.72). Diodes D2 and D3 have to
handle reverse voltage given by (2.73). Assuming that the inductor currents have
only a very small ripple and the current in2(t) is almost rectangular, the transistor rms
current is found with (2.74). The average forward current of diodes D2 and D3 is
equal to the output dc current Iout. The average forward current of diodes D1 and
D11 equals half of the output dc current, since these diodes are connected in parallel.
Required power rating of transistors and diodes is found with equations (2.75) to
(2.79). Required storage inductances L1 and L2 are found with (2.80), while (2.81)
gives the effective output capacitance.
n2
n= (2.70)
n1
1+ 2 ⋅ n
M= (2.71)
1− D
1
VT1(off) = VT11(off) = VD1(R) = VD11(R) = ⋅ Vout (2.72)
1+ 2 ⋅ n
2⋅n
VD2(R) = VD3(R) = ⋅ Vout (2.73)
2 ⋅ n +1
2 2
 M ⋅ I out I   M ⋅ I out 
I T1(rms) ≈ (1 − D ) ⋅  + n ⋅ out  + ( 2 ⋅ D − 1) ⋅   (2.74)
 2 1− D   2 
28
Figure 2.16 Waveforms of the non-isolated two-inductor boost converter

2
1 M n  M 
2 (2.75)
≈ ⋅ (1 − D ) ⋅  +  + ( 2 ⋅ D − 1) ⋅   ⋅ Pout
2 ⋅ n +1  2 (1 − D )   2 
volt×ampT = 2 ⋅ volt×ampT1 (2.76)

1 I 1
volt×amp D1 = ⋅ Vout ⋅ out = ⋅ Pout (2.77)
2 ⋅ n +1 2 2 ⋅ ( 2 ⋅ n + 1)
2⋅n 2⋅n
2 ⋅ n +1 2 ⋅ n +1
1+ 4 ⋅ n
volt×amp D = 2 ⋅ ( volt×amp D1 + volt×amp D2 ) = ⋅ Pout (2.79)
1+ 2 ⋅ n
Vin ⋅ D
L1 = L2 = (2.80)
2 ⋅ ∆iL1 ⋅ fs
I out ⋅ D
Cout = (2.81)
This topology has similar advantages like the non-isolated push-pull-boost
converter presented in section 2.2.6. It includes a high voltage gain, reduced
29
Table 2.1 Component requirements comparison for different dc-dc topologies

Voltage gain Transistors Diodes Magnetic Capacitors
M components
1
Basic boost 1 HV 1 HV 1 inductor (CCM) 1 HV output
1− D
N-phase 1 HV output
1 N inductors (DCM
interleaved N HV N HV for N times higher
1− D or CCM)
boost switching freq.
2-phase 1 HV output
boost with 1 1 coupled inductor for double
2 HV 2 HV
coupled 1− D with n2=n1 switching
inductor frequency
1 1 1 inductor (CCM)
Cascaded ⋅ 1 LV 1 LV 1 LV intermediate
1 inductor (CCM or
boost 1 − D1 1 − D2 1 HV 1 HV 1 HV output
DCM)
1 coupled inductor /
Center- 1+ n ⋅ D 1 LV 1 LV snubber
1 LV flyback transformer
tapped boost 1− D 1 VHV 1 HV output
with n2>n1
1 coupled inductor / 2 LV output

Flyback- 1+ n ⋅ D 1 LV
1 LV flyback transformer all series
boost 1− D 1 VHV
with n2>n1 connected
1 inductor (CCM) 3 LV output

Push-pull- 1+ n 2 LV
2 LV 1 push-pull all series
boost 1− D 2 HV
transformer connected
2 inductors (DCM
Two- 3 LV output
1+ 2 ⋅ n 2 LV or CCM)
inductor- 2 LV all series
1− D 2 HV 1 two-winding
boost connected
transformer
voltage stress on transistors and smooth input current. Moreover this topology
utilizes a simple two-winding transformer instead of a more complex three-winding
one. Main drawbacks of this topology are several diodes effectively connected in
series and three series connected output capacitors.
2.3 Summary
In this chapter the state-of-the-art in non-isolated dc-dc step-up converters has
been presented. At the beginning of the chapter different approaches for a dc
voltage boosting were introduced. Based on results presented by authors, as well
as considering topologies complexity and features it’s found that converters with an
inductive energy storage are the most suitable ones. They have a fairly low
component count and a very good performance among different converter kinds
(switched capacitor circuits, resonant, Luo, etc.).
Next, in section 2.2 several different non-isolated step-up topologies were
described in details, including diagrams, waveform sketches and key equations.
30
The survey starts with a well known basic boost converter and guides the reader thru
more and more advanced topologies, including coupled inductor topologies.
Finally, three topologies derived from isolated dc-dc step-up converters are
presented. These are the flyback-boost (known), the push-pull-boost (known) and
the two-inductor-boost (new).
Table 2.1 summarizes the voltage gain of considered converters and their
component count – separately transistors, diodes, magnetics and capacitors. The
knowledge about number of components alone is not enough, so a short description
of components is provided too. In the table following abbreviations are used: LV
states for a low voltage, i.e. a voltage significantly lower than the output voltage;
HV states for a high voltage, i.e. about the output voltage; VHV states for a very
high voltage, i.e. a voltage significantly higher than the output voltage. Analyzing
the table one can find that the simple topology, like the basic boost converter
provides the lowest voltage gain, but it has the fewest parts. Improvement in the
voltage gain is done by the price of the component count – it applies to both, passive
and active components. The choice of the most suitable topology is not clear yet,
especially that only an approximate voltage rating of components is know so far.
Analysis and comparison of equations describing converters is not convenient since
some of equations are quite complex. Instead, a numerical example is used for a
quantitive comparison of components’ size.
The example design bases on the specification presented in section 1.4.1. The
input voltage is Vin=30 V, the output voltage is Vout=400 V and the output power
Pout=1000 W. It’s assumed that the efficiency of all converters is 100%. Also no
safety margins are taken into account at this point, so absolute minimum values are
calculated. The size of the components is expressed in terms of the apparent power
or in terms of the average stored energy. Low voltage semiconductors are
considered to be below 150 V devices. Moreover, the input current ripple is
limited to 20% (peak-peak) of the rated current. The output voltage ripple is 5%
(peak-peak) of the dc output voltage. The switching frequency is assumed to be
fs=50 kHz basically. Some converters however have more then one transistor
operating in phase shift manner. In such case two scenarios are presented for
passive components – a) the switching frequency is reduced, so input / output ripples
remains at 50 kHz; b) the switching frequency is kept at 50 kHz, so input /output
ripple base frequency increases.
The basic boost converter is considered to be a reference topology, since it’s the
simplest one. This topology has a very poor utilization of the transistor.
Paralleling and interleaving of boost converters does not improve it, but interleaving
technique enables the reduction of passive components’ size, even if the switching
frequency is reduced (scenario b). Cascading (series connection) of two or more
boost converters improves utilization of transistors. Now, two much smaller
transistors are required instead of a large one. Unfortunately, utilization of diodes
is degraded and the size of passive components may increase significantly. The
center-tapped boost converter and the non-isolated flyback-boost seem to overcome
these problems. They ensure a good utilization of the transistor and the diode.
Also size of passive components seems to be similar to the basic boost converter.
However, these converters have two inherent drawbacks. First, the winding n2 (see
31
Table 2.2 Size of components in a 1 kW converter

Transistors Diodes apparent Energy stored in Energy stored in
apparent power power inductors capacitors
[VA] [VA] [mJ] [mJ]
Basic boost 12840 1000 45.6 184
2-phase a) 42.4 a) 184

12840 1000
interleaved
(2×8420) (2×500) b) 21.2 b) 92
boost
4-phase a) 34.7 a) 184

12848 1000
interleaved
(4×3212) (4×250) b) 8.7 b) 46
boost
2-phase
boost with 18148 1000 a) 46 a) 184
coupled (2×9074) (2×500) b) 23 b) 92
inductor
Cascaded
6268 2000 72 290
boost
(3468+2800) (1000+1000) (37.3+34.7) (150+140)
(120 V bus)
Center-
1225 197.4
tapped boost 4345 46.2
(306+919) (45.4+152)
(n=3)
Flyback- 1225 151.2
4345 46.2
boost (n=3) (306+919) (45.4+105.8)
a) 27.3
Push-pull- 4816 1668
b) 13.6 a) 243.8
boost (n=2) (2×2408) (2×167+2×667)
1607 VA transf. (37.4+2×103.2)
a) 54.8 (2×27.4) b) 121.9
Two-
4816 1668 (18.8+2×51.6)
inductor- b) 27.4 (2×13.7)
(2×2408) (2×167+2×667)
boost (n=1) 1333 VA transf.
Figure 2.9 and Figure 2.12) is not clamped and it resonates with the diode D2
parasitic capacitance. It results in a huge voltage overshoot across the diode.
Second, a huge input current ripple, thus they require an additional input filter.
The non-isolated push-pull-boost and the non-isolated two-inductor-boost converters
are current fed ones, so input current ripple is limited. Both topologies have a lot
in common – they incorporate two low voltage transistors and the voltage doubler
rectifiers. Also they offers a good utilization of transistors and stored energy is
comparable with the basic boost converter. In the same time these converters
provides a very high voltage gain. The non-isolated two-inductor-boost converter
offers the highest theoretical voltage gain in this comparison. However one should
remember that the turns ratio n definition is slightly different for the push-pull
transformer and for the two-winding transformer (equations (2.58) and (2.70)). So,
the effective voltage gain in terms for number of turns (not turns ratio) will be
similar in both cases.
32
At this point the non-isolated push-pull-boost seems to be a good candidate for

future investigation. First, its good performance is proven [6], while the converter
complexity is moderate. Second, this topology has a relatively low requirements in
terms of the components rating and it may turn into a compact and cost effective
design. Finally, the push-pull transformer can be integrated with the input inductor
on a single core, thus the size of the converter can be reduced further (see Chapter
4).
33
Chapter 3
Power Losses in Dc-Dc Converters
At the beginning of Chapter 2 the assumption of lossless circuits was made. It
meant that the output power of the converter equals the input power and no losses
are dissipated in the converter. This assumption simplified analysis of converters
and helped to develop converter’s equations. Unfortunately, in the real world
converters are not lossless. The output power of the converter is always lower then
the input power and the difference is the power losses dissipated as heat in the
converter (3.1). Ratio between the output power and the input power is the power
conversion efficiency or just the efficiency (3.2). Knowledge about power losses
and heat dissipated in the converter operating under different conditions is critical
for a good design and a reliable product. One way to gain this knowledge is to
assemble many breadboards and test them intensively. This approach however is
very expensive and requires a lot of time. Another way is to use modeling and
simulation in the design process.
Ploss = Pin − Pout (3.1)
Pout P
η= = 1 − loss (3.2)
Pin Pin
This chapter deals with estimation of power losses in dc-dc non-isolated
converters. First a brief overview on different model levels is given and modeling
approach is explained. Next, major loss sources are pointed out and suitable loss
models are introduced. Finally the presented modeling approach is being applied
to the boost converter. Validation of the modeling approach is presented at the end
of this chapter.
3.1 Overview on model levels

Nowadays simulation is a widespread and important tool for design and
optimization. It is possible to simulate behavior of almost any natural system and
power electronics is just one of many applications. To achieve reliable and
accurate results it is important to use good models. The question now is what is a
good model? The answer is not straightforward at all. Usually the more accurate
model is more complex one and requires more computation time. Sometimes
however there is no need for a very complex model while a simpler one gives
satisfactory results.
34
Figure 3.1 Different switching network models - a) a component model; b) an

ideal switch model; c) an average model [45]
Figure 3.1 [45] presents three models of the same switching network of a boost
converter. Model a) is so called a component model (levels 1 to 5 in [46]). It
takes into account many details of the MOSFET and the diode including its
switching behavior, thermal characteristics, ageing and so on. Models of this kind
are usually very accurate, but they require a lot of computation time. Also these
models need many parameters which shall be calculated from a silicon geometry or
extracted from measurements.
Model b) is so called an ideal switch model (level 0 in [46]). The MOSFET and
the diode are replaced by ideal switches. Sometimes the model has implemented a
very basic static characteristic of the component, like on-state resistance or forward
voltage drop. Although a dynamic characteristic and other complex phenomena
are omitted, it’s still possible to observe current and voltage waveforms present in
the converter circuit. Thanks to the simpler model and larger time step the
simulation speed increases. It’s possible to simulate more switching cycles and
observe a dynamic behavior of the converter as a whole, which is useful in a control
loop design or an operational check of a topology.
Model c) is an average model, which can be considered as a highest level model.
Instead of a component models or ideal switches the model is described by averaged
functions [47]. All signals are averaged over a switching period and no high
frequency variations are taken into account. It doesn’t provide a deep insight into
the converter’s waveforms, but dynamic behavior of the whole converter is being
modeled accurately. Computation time is being reduced further. An average
model is suitable for a control loop design, simulation of long time periods and
analysis of the circuit on a system level.
3.2 Modeling approach

Figure 3.2 presents the diagram of the incorporated model. The model is built
up from several interconnected modules, which are organized in a hierarchical
structure. The main script runs and supervise the whole loss estimation process.
This module can run a simple task, like the power loss calculation in a single
operating point. It can perform more complex tasks, like a minimum power loss
35
Figure 3.2 The block diagram of the model
search too. The main provides input data about the operating point (the input
voltage, the output voltage, the output power etc.) and about used components to the
lower level module – the converter model module. The main script collects the
loss calculation results and prepares them for the print out.
The converter model’s main task is calculation of currents and voltages essential
for the power loss calculation by the component models. It includes dc, ac and rms
values of currents flowing thru the component. Current and voltage values just
before and just after switching are calculated too. Making of a converter model
inside a proper circuit simulation software would be the first choice for power loss
estimation. Such converter model will base on ideal or detailed component models
(level 1 or 2 in [46]), which provide detailed waveforms and accurate results on one
hand. On the other hand simulations of several topologies operating under
different conditions may take a lot of time. Since power losses are calculated in a
steady state it’s possible to incorporate a simpler, average model is incorporated.
Such a model will provide comparable results in a significantly shorter time. The
converter model is built up from the set of equations describing particular currents
and voltages, based on the operating point parameters delivered by the main script.
In paragraphs 3.2.1 the method for calculation of respective current values is
presented.
The component models are responsible for a direct power loss calculation based
on values provided by higher level models and component parameters. All
considered losses together with relevant equations are described in paragraphs from
3.3 to 3.6. It’s important that presented models base on datasheet values in most
cases. So, component parameters extraction is vastly reduced or eliminated at all.
36
Figure 3.3 Trapezoidal segments
3.2.1 Currents in the circuit

Dc, ac and rms current values are important for estimation of conduction power
losses inside components. For estimation of the switching loss knowledge about
currents just before and just after switching instant is required. These values can
be found easily with a circuit simulator, but it requires a significant amount of
computation time. In this paragraph a simple analytical solution is provided.
One can observe that current waveforms of all converters presented in Chapter 2
are composed from trapezoidal segments. Assumption of piecewise linear currents
is true for ideal converters only, but it’s enough for power losses estimation.
Figure 3.3 presents a simple trapezoidal waveform, similar to a transistor current
in a boost converter (Figure 2.2). Finding instantaneous values at the beginning
and at the end of each trapezoidal segment allows calculation of dc and rms values
according to (3.3) and (3.4). These equations can be generalized ((3.5) and (3.6)
)[47]. Then the ac value can be found with (3.8).
 t − t   i ( t ) + i ( t1+ ) 
I =  2 1  ⋅  2−  (3.3)
 t3 − t 0   2 
 t −t  1
 t3 − t 0  3
(
I ( rms ) =  2 1  ⋅ ⋅ i ( t1+ ) + i ( t1+ ) ⋅ i ( t2− ) + i ( t2− )
2 2
) (3.4)
I = ∑
n (
 tk − tk −1   i ( tk − ) + i t( k −1) + ) 
⋅ (3.5)
k =1  t n − t0   2 
 
n
 tk − tk −1 
I ( rms ) = ∑
k =1  t n − t0 
 ⋅ uk (3.6)
1
(
uk = ⋅  i t( k −1)+ ) + i ( t( ) ) ⋅ i ( t 
2
) + i ( tk − )
2
k −1 + k−  (3.7)
3  
I ( ac ) = I (2rms ) − I 2 (3.8)
In the same way all current waveforms are analyzed and respective equations in
MATLAB® format are presented in Appendix B.
37
3.3 Transistor
Majority of power transistors used in power electronic converters are IGBTs and Power
MOSFETs (referred here as MOSFETs). IGBTs are desired for high power
applications. They handle voltages exceeding 600-1000 V and currents much greater
than 100 A. The price for it is a poor switching performance, thus IGBTs’ switching
frequency is limited. On the other hand in applications where voltages do not exceed
600 V MOSFETs are preferred because of its fast switching and relatively low
conduction losses. This section deals with power losses present in a MOSFET
transistors. Two major sources are recognized: a conduction loss and switching losses.
Further total switching losses break up into turn-on, turn-off, gate and reverse recovery
losses. However the term switching loss used in following sections refers to turn-on
and turn-off losses only. The reverse recovery loss is caused by the diode reverse
recovery phenomena and it is discussed in section 3.4.2.
3.3.1 Conduction loss

MOSFET conduction loss, also called on-state loss, is caused by the current
flowing thru the transistor’s internal resistance (channel resistance) RDS(on) and it’s
simply described by (3.9). The transistor rms current IT(rms) is found by analysis of
converter current waveforms. The transistor rms current of selected topologies are
presented in Chapter 2. On-state resistance RDS(on) is published in datasheets while
it’s a basic parameter of any MOSFET. In the transistor operating in the ohmic
region (also called the on-state region) the resistance RDS(on) practically depends only
on the junction temperature Tj. Actual on-state resistance RDS(on) for given junction
temperature Tj are read from datasheet plots. For simulation purposes resistance-
temperature relationship is approximated by (3.10). Term RDS(on)25 is the on-state
resistance at the junction temperature of 25ºC and rDS(on)T is the temperature
coefficient of the resistance.
PTcond = RDS( on ) ⋅ I T(rms) 2 (3.9)
RDS( on ) = RDS( on )25 + (Tj − 25D C ) ⋅ rDS( on )T (3.10)
3.3.2 Switching loss

Estimation of turn-on and turn-off losses is not that simple. Switching of a
MOSFET is a complex process, which depends on operating conditions and the
converter topology. In the literature [43, 47-50] one can find more or less detailed
descriptions of the MOSFET switching. However mathematical description of the
switching loss is often limited to a very general equation (3.11). Gate controlled
switching of an inductive clamped load is assumed here. This kind of switching is
true for all converters presented in Chapter 2. Figure 3.4 shows the MOSFET with
its parasitic capacitances placed in the equivalent switching circuit. Equations
(3.12), (3.13) and (3.14) define the input capacitance Ciss, the output capacitance Coss
and the reverse transfer capacitance Crss. Capacitances Coss and Crss are strongly
variable with the drain-source voltage (here referred as vT(t)). The current source Il
represents the inductor current, which is considered to be constant during switching.
38
Figure 3.4 A clamped inductive load switching circuit

and parasitic capacitances of the MOSFET
The voltage source Vc represents the capacitor voltage (the clamping voltage) and
it’s also constant during switching. The diode D is assumed to be an ideal one, so
no reverse recovery and no forward recovery is taken into account at this point.
The MOSFET is controlled by the square-wave voltage vG, which is VG either zero.
The gate current iG(t) is limited by the gate resistor RG.
PTsw = fs ⋅ ∫ vT ( t ) ⋅ iT ( t ) dt (3.11)
switching
transitions
Ciss = CGS + CGD (3.12)

Coss = CDS + CGD (3.13)
Crss = CGD (3.14)
The gate controlled switching has several stages like presented on Figure 3.5.
From the switching power loss point of view only four stages are important. Those
are:
• t1-t2 – called current rise time tir, during this period the drain current iT(t)
linearly increases from zero to the load current Il, the drain-source voltage
vT(t) remains almost constant at the clamping voltage level Vc
• t2-t3 – called voltage fall time tvf, during this period the drain-source voltage
vT(t) falls from the clamping level Vc to the on-state voltage RDS(on)·Il, this
voltage transition is strongly non-linear
• t6-t7 – called voltage rise time tvr, during this period the drain-source voltage
vT(t) rises from the on-state voltage to the clamping voltage Vc, also this
voltage transition is non-linear
• t7-t8 – called current fall time tif, during this period the drain current iT(t)
linearly falls from the load current level Il to zero
39
Figure 3.5 MOSFET switching waveforms
The total energy dissipated during turn-on and turn-off is expressed as the sum of
energy portions (3.15) dissipated during four stages mentioned above. Energy
dissipated during current rise time and current fall time is given by (3.16) and (3.17).
It’s simply calculated as the area of gray triangles from Figure 3.5 since the drain
current rises and falls in almost linear manner and the drain-source voltage remains
constant. Current rise time tir is found by analysis of the gate circuit and the gate-
source voltage vGS(t) [48]. After the voltage VG is applied to the gate circuit the
gate-source voltage vGS(t) increases. Once the voltage vGS(t) reaches the threshold
level VGS(th) the drain current iD(t) starts to rise. The transistor is in its active
region, so the gate-source voltage vGS(t) and the drain current iT(t) are coupled
together by the transconductance gfm. The gate-source voltage rises to the plateau
voltage Vplateau given by (3.20). In the same time the drain current reaches the load
current level Il and the current transition is done. The gate circuit is basically a RC
circuit driven by the square-wave voltage vG(t), so the current rise time tir is found as
time required by the gate-source voltage vGS(t) to increase from the threshold level
VGS(th) to the plateau voltage Vplateau (3.18). Current fall time tif is found in similar
way. The gate circuit is discharged and the gate-source vGS(t) decreases from the
plateau voltage Vplateau to the threshold voltage VGS(th) (3.19).
PTsw = fs ⋅ (Wir + Wvf + Wvr + Wif ) (3.15)
Wir = tir ⋅ I l ⋅ Vc (3.16)

Wif = tif ⋅ I l ⋅ Vc (3.17)
 V G −VGS( th ) 
tir = RG ⋅ Ciss ⋅ ln 
 V G −Vplateau 
(3.18)
 
 Vplateau 
tif = RG ⋅ Ciss ⋅ ln   (3.19)
 VGS( th ) 
 
Il
Vplateau = VGS( th ) + (3.20)
g fm
40
Estimation of energy dissipation during voltage rise and voltage fall requires
more sophisticated approach because of non-linear voltage transitions. During
these phases the gate-source voltage vGS is constant as stated before. This fact has
two important implications. First, the gate current iG is constant and it’s easy to
find with (3.21), assuming that vGS(t)=VG during turn-on transition and vGS(t)=0
during turn-off transition. Second, the gate current iG(t) charges or discharges the
Crss capacitance only. Since Vplateau<<Vc the slope of vGD(t) is about the same like
the slope of vT(t) (3.22), which makes analysis of vT(t) easier [50]. The range of
vT(t) and vGD(t) variation is divided into n discreet steps. Next, (3.23) is used to
find the time tk required to increase or decrease the voltage vT(t) from one discreet
value vT(k) to the other discreet value vT(k+1). For each step value of Crss is
updated in respect to the actual vT(k) voltage. Finally energy dissipated during
voltage fall time is found as a sum of energy portions dissipated in each step (3.24).
In the same way energy dissipated during voltage rise time is found.
vGS ( t ) − Vplateau
iG ( t ) = (3.21)
RG
dvT ( t ) dvGD ( t )
≈ (3.22)
dt dt
vT ( k + 1) − vT ( k )
tk = Crss ( vT ( k ) ) ⋅ (3.23)
iG ( t )
n
Wvf = ∑ tk ⋅ vT ( k ) ⋅ I l (3.24)
k =0
3.3.3 Gate loss

Gate loss comes from charging and discharging the MOSFET input capacitance
Ciss thru the gate resistor RG and it’s given by (3.25). Term QG is the gate total
charge which is available in datasheets. Most of this energy is dissipated in the
external gate resistor and the driver. Portion of this energy might be dissipated
inside the MOSFET if there is any significant internal gate resistance.
PTgate = f s ⋅ QG ⋅ VG (3.25)
3.4 Diode
Next to a transistor, a diode is a key semiconductor component. Fast switching
of the diode has the same importance like a low conduction loss. Schottky diodes
provide outstanding switching performance and a low conduction loss in the same
time, but the blocking voltage is limited to 150-200 V. Fast silicon (pn-junction)
diodes have been used if higher blocking voltage is required. Despite a great
technology improvement done during years the reverse recovery behavior remained
the main drawback of silicon diodes. Recent development in wide bandgap
semiconductor materials (SiC, GaN and similar) enabled mass production of next
41
generation high voltage Schottky diodes, which overcome reverse recovery problem.
Following sections describe estimation of different losses present in power diodes.
3.4.1 Conduction loss

Forward characteristic of a power diode is non-linear and it’s presented on Figure
3.6. For simulation purposes a linear approximation is used. It includes the
forward voltage drop VD(F) and the forward resistance RF. This simple model is
sufficient for the conduction loss calculation according to (3.26). Sometimes the
diode rms current ID(rms) is replaced by the diode average current ID for sake of
simplicity. However if the ac component of the diode current is significant then
the diode rms current shall be used. Also one shall remember that both, the
forward voltage drop VD(F) and the forward resistance RF are temperature dependent
((3.27) and (3.28)). All required diode parameters are extracted from forward
characteristic plots available in datasheets.
PDcond = VD( F) ⋅ I D + RF ⋅ I F( rms ) 2 (3.26)
RF = RF25 + (Tj − 25D C ) ⋅ rFT (3.27)
VD( F) = VD( 25) + (Tj − 25D C ) ⋅ vD(T ) (3.28)
3.4.2 Reverse recovery loss

The reverse recovery (or diode turn-off) behavior of a silicon power diode is
considered as the most important for the circuit performance, while it may lead to
severe losses in the transistor and the diode itself. Switching of an inductive
clamped load is assumed. The switching circuit and related switching waveforms
are presented on Figure 3.9 and Figure 3.7. When the transistor is turned-on the
load current Il is being commutated from the diode to the transistor. The diode
current fall rate di(t)/dt may exceed 1000 A/µs in many cases, but the pn-junction
diode will not move from conduction to blocking state immediately. The stored
Figure 3.6 Forward characteristic of a diode and its linearization

42
charge has to be removed from the junction vicinity before the diode can block the
reverse voltage. It results in the reverse current iD(t) flowing thru the diode D, the
transistor T and the clamping voltage source Vc. Exact shape of the diode reverse
current depends on many factors, where the most important are: the diode material
and manufacturing technology, the junction temperature, the diode current fall rate
di(t)/dt. In the literature one can find several physical diode models suitable for the
reverse recovery loss calculation [51, 52].
Majority of the reverse recovery power loss is dissipated in the transistor
commutating with the diode. A simple and robust method for calculation of this
loss is described in [47]. The method is valid if the time tf is significantly less then
the time ts (see Figure 3.7), which is true for a low softness factor s (3.29) or so
called “snappy” diodes. Additional power PTrr dissipated in the transistor is simply
described by (3.30) [47]. Values of the reverse recovery time trr and the reverse
recovery charge Qrr are found in datasheet plots as a function of the diode current
slope di(t)/dt. The diode current slope di(t)/dt is determined either by the transistor
gate driver (the transistor current rise time tir), or by the clamping voltage and stray
inductances present in the circuit – whichever gives lower di(t)/dt.
tf
s= (3.29)
ts
PTrr = fs ⋅ (Vc ⋅ I l ⋅ trr + Vc ⋅ Qrr ) (3.30)

Some energy is dissipated in the diode during its turn-off too. For “snappy”
diodes this is a very small loss and usually is neglected. However in case of soft
recovery diodes (a large softness factor s) it should be taken into account [53].
Assuming linear voltage and current transitions in the diode the power dissipated in
Figure 3.7 Reverse recovery of a pn-junction silicon diode

diode waveforms (top) and transistor waveforms (bottom)
43
Figure 3.8 Forward recovery of a diode
the diode is given by (3.31). The time tf is found with (3.32) by rearranging (3.29),
while the softness factor s and the reverse recovery time trr are available in
datasheets.
PDoff = fs ⋅ ( tf ⋅ I RM ⋅ Vc ) (3.31)
s ⋅ trr
tf = (3.32)
s +1
3.4.3 Forward recovery

The diode waveforms during its turn-on are presented on Figure 3.8. When the
diode passes from the blocking state to the conduction state the forward voltage
temporary increases to the peak forward voltage VFRM before it drops to the steady-
state forward voltage VD(F). During fast switching the peak forward voltage VFRM
may exceed 100 V in some cases. This voltage adds to the transistor turn-off
voltage and it has two important results. First, it creates voltage overshoot across
the transistor and may lead to the device failure. Second, the increased turn-off
voltage of the transistor may cause increased switching loss during its turn-off.
Forward recovery and the peak forward voltage VFRM causes power loss PDon in the
diode. Using linear approximation of waveforms, like presented in [53], this power
is found with (3.33). This simplified method assumes that the diode current rise
time is very short in compare to the forward recovery time tfr of the diode, so the
current iA is almost rectangular. This assumption often leads to overestimation of
actual turn-on energy in the diode. Since this energy loss is fairly small in compare
to conduction loss and turn-off loss it’s often neglected.
PDon = f s ⋅ ( tfr ⋅ I l ⋅ VFRM ) (3.33)
3.4.4 Capacitive loss

Figure 3.9 presents a clamped inductive load switching circuit and the diode
capacitance CD is marked. During the converter operation this capacitance is
charged and discharged periodically. During diode turn-on energy stored in the
diode capacitance is lost [47]. In the simplest form the capacitive loss PDcap is
described by (3.34). The estimation is not trivial since the diode capacitance CD is
a function of the reverse voltage vD(R)(t). The range of the reverse voltage variation
(from –Vc to 0) is divided into n discreet steps. Then the actual value of the diode
44
Figure 3.9 A clamped inductive load switching circuit

and parasitic capacitance of the diode
capacitance CD is found from the datasheet plot for each voltage step. The total
charge stored in the diode capacitance is calculated as a sum of small charges for
each voltage step (3.35).
1 
PDcap = fs ⋅  ⋅ QD ⋅ Vc  (3.34)
2 
( )( )
n
QD = ∑ CD vD( R ) ( k ) ⋅ vD( R ) ( k + 1) − vD( R ) ( k ) (3.35)
k =1
When pn-junction silicon diode is used the capacitive loss is often neglected
since it’s small in compare to other loss components, especially in compare to
reverse recovery loss. However if a Schottky diode is used then the capacitive loss
should be taken into account. It’s because a Schottky diode has larger capacitance
and it has no reverse recovery loss. So, the capacitive loss become larger and
visible among other loss components.
3.5 Magnetic device

Magnetic components are very important parts of any power converter. In
many cases they are designed especially for the particular converter, rather then
selected among available off-the-shelf parts. Core loss and copper loss are two
major loss sources in magnetic devices regardless it’s type and design details.
Following sections describe these losses, modeling of them and their impact on
particular magnetic devices like an inductor, a transformer or a flyback transformer.
3.5.1 Copper loss

The copper loss is one of two loss mechanisms present in most of magnetic
devices. Basically it’s caused by the current flowing thru the winding, whose
resistance is non-zero. So called the low frequency copper loss is expressed by
well known equation (3.36), where I(rms) is the rms value of the current and R is the
low frequency resistance given by (3.37). The low frequency resistance bases only
on the wire cross section area ACu, the wire length lCu and its resistivity ρ. It
45
Figure 3.10 Skin effect
doesn’t include any high frequency effects. In a switch mode converter operating
in kilohertz or megahertz range the low frequency approach limits to the dc current
components only. The equation (3.36) become (3.38) in such case.
PCu = R ⋅ I (2rms ) (3.36)
ρ ⋅ lCu
Rdc = (3.37)
ACu
PCudc = Rdc ⋅ I 2 (3.38)

Estimation of the ac copper loss requires a more sophisticated approach.
Basically the ac copper loss is expressed by (3.39), similarly to the dc copper loss.
However at high frequency the winding resistance increase is observed (Rac>Rdc).
The resistance increase is basically caused by eddy currents inside the winding
conductor – the most pronounced phenomenas responsible for eddy currents are the
skin effect and the proximity effect. Both phenomenas are widely described in the
literature [43, 47, 54-62], so here only a brief explanation is given together with a
simple method for estimation of the ac copper loss.
PCuac = Rac ⋅ I ac2 (3.39)
The skin effect is explained using Figure 3.10 [47]. There is the non-magnetic
wire conducting the high frequency current i(t). The current i(t) induces the
magnetic flux Φ(t). According to Lenz’s law the ac flux Φ(t) induced current
(eddy currents) inside the wire. These eddy currents tend to oppose the ac flux and
flow in the manner presented on Figure 3.10. Eddy currents add to the main
current i(t) and it results in a non-uniform current density inside the wire. The high
frequency current i(t) do not penetrate to the center of the wire, but flows on the
surface of the wire. Because the center of the wire is not utilized, the effective wire
cross section are is reduced thus the ac resistance increases over the dc resistance
given by (3.37). The ac resistance Rac of a single wire is given by (3.40), where δ
is so called penetration depth and h is the wire thickness. The penetration depth is
given by (3.41), where f is the current frequency, ρ is the wire resistivity and µ is the
magnetic permeability of the wire (in most cases µ=µ0).
h
Rac = ⋅ Rdc (3.40)
δ
46
Figure 3.11 Proximity effect
ρ
δ= (3.41)
π ⋅µ ⋅ f
The proximity effect is shortly explained using Figure 3.11 [47]. There are two
parallel wires in a close proximity. The wire 1 carries the high frequency current
i(t), while the wire 2 is open circuit and its net current is zero. The penetration
depth δ is significantly smaller then the wires thickness h. Now, the current i(t)
generates the flux Φ(t) around the wire 1. The flux attempts to penetrate the wire 2
and by Lenz’s law current is induced in the wire 2. Because the wire 2 is open
circuit the induced current path has to close inside the wire 2 – the current flows in
one direction on the left side of the wire 2 and in the opposite direction on the right
side, like presented on Figure 3.11. In a multilayer designs the proximity effect
may lead to a severe ac copper loss.
In the literature one may find different methods for estimation of the ac
resistance and the related ac copper loss [43, 47, 54-62]. Some of these methods
base on Dowell’s work and one dimensional (1D) approach, while some other
incorporate more complex two dimensional (2D) approach. Here the simple
method for estimation of the ac losses is presented [47]. The method bases on 1D
approach and it is intended and suitable for analysis of windings made out of an
uniform foil of the thickness h. It’s possible to extend the method to windings
made out of a square wire and a round wire. In this method the resistance factor FR
binding the dc resistance Rdc and the ac resistance Rac is used. The resistance factor
FR is found based on the winding geometry (number of layers m, the layer thickness
h, penetration depth δ etc.) with (3.42), where ϕ is the relative penetration depth
given by (3.43) (for foil windings). Functions G1(ϕ) and G2(ϕ) are given by (3.44)
and (3.45) respectively. Solving (3.42) for M-layer design leads to (3.46), which is
plotted on Figure 3.12.
M
Rac 1
FR = =
Rdc M
∑ ϕ ⋅( 2 ⋅ m
m =1
2
− 2 ⋅ m + 1) ⋅ G1 ( ϕ ) − 4 ⋅ m ⋅ ( m − 1) ⋅ G2 ( ϕ ) (3.42)
h
ϕ= (3.43)
δ
47
sinh ( 2 ⋅ ϕ ) + sin ( 2 ⋅ ϕ )
G1 ( ϕ ) = (3.44)
cosh ( 2 ⋅ ϕ ) − cos ( 2 ⋅ ϕ )
sinh ( ϕ ) ⋅ cos ( ϕ ) + cosh ( ϕ ) ⋅ sin ( ϕ )

G2 ( ϕ ) = (3.45)
cosh ( 2 ⋅ ϕ ) − cos ( 2 ⋅ ϕ )
 2 
FR = ϕ G1 ( ϕ ) + ⋅ ( M 2 − 1) ⋅ ( G1 ( ϕ ) − 2 ⋅ G2 ( ϕ ) )  (3.46)
 3 
As mentioned above, the method can be extended for a windings made out of a
round wire filling the whole core window width. For that purpose the relative
penetration depth ϕ given by (3.43) is replaced by the one given by (3.47). The
coefficient η is so called porosity factor. It’s defined as the ratio between the
actual layer copper area to the area of the effective foil conductor (3.48). For round
wires that span the bobbin the typical value of η is 0.8. As presented on Figure
3.12 a large ac copper loss occurs for a large number of layers m and a large relative
penetration depth ϕ. So, the general guideline is to avoid multilayer designs, to use
the interleaved windings whenever possible [43, 47, 59] and to keep the layer
thickness below the penetration depth (h<δ). The method does not include
presence of the air gap, the edge effects and it’s not valid for complex or non-
uniform winding configurations (e.g. arranged on a toroidal core).
π d
ϕ = η⋅ ⋅ (3.47)
4 δ
ACu round
η= (3.48)
ACu foil
The analysis above is valid for sinusoidal currents, which are rare in switch mode
3
10
M=1
M=2
2 M=3
10
M=5
M=10
FR
1
10
0
10 −1 0 1
10 10 10
Relative penetration depth
Figure 3.12 The resistance factor FR as a function of the relative penetration depth
ϕ and the number of layers M
48
converters. Typical current waveform is far from being sinusoidal and it contains a
significant harmonic content. Such current waveform is expressed using Fourier
series and the ac copper loss is calculated for each harmonic separately [47, 59].
Nowadays, thanks to the increasing processing power of desktop computers 2D
and 3D finite element methods (FEM) are used for design and optimization of
magnetic devices. FEM approach provides a very good accuracy and enables
analysis of any complex winding and core configuration. It enables automated
harmonic analysis of arbitrary current waveform too.
3.5.2 Core loss

The core loss is the second loss mechanism present in most of magnetic devices.
It’s commonly accepted that the core loss is caused by two physical effects – the
hysteresis of the material and eddy currents inside the core [47]. Not all energy
required to change the magnetization of the core can be recovered. Part of this
energy is lost inside the core as a heat and electrically observed as a hysteresis loop.
Because the core is a better or worse conductor placed in a variable magnetic field
eddy currents flow inside the core. Both phenomenas contribute to the total core
loss.
Over the years several different methods and modeling approaches ware
presented. Reference [63] points out most common approaches for modeling of the
core loss: the hysteresis models (Jiles-Atherton model and Preisach model), the loss
separation model and the empirical model (Steinmetz equation). Reference [64]
gives a good insight into magnetization process of ferrites bases on domains
(domain walls) hypothesis. Also it states that the only origin of losses are eddy
currents around moving domain walls. However even detailed knowledge about
physical origin of the core doesn’t provide necessary practical method for
calculation of the core loss. So far the method based on the work done by
Steinmetz is widely used. The foundation of the method is so called Steinmetz
equation (3.49).
pFe = k ⋅ f α ⋅ ∆Bβ (3.49)
This empirical equation states that the specific core loss (loss per volume unit) is
a function of the magnetization frequency f and the flux density amplitude ∆B.
Parameters k, α and β are determined experimentally and they are commonly found
in core datasheets. Simple and robust Steinmetz equation together with parameters
available for different materials (ferrites, powders) makes this method very popular
and ready to use. However three main drawbacks of this method are recognized.
First, parameters (k, α, β) published in datasheets are valid only for sinusoidal
excitation. It makes the method inaccurate for switch mode converters, while the
magnetic flux in the core is not sinusoidal. Second, the basic Steinmetz equation
doesn’t include any dc bias influence. It may lead to miscalculation of the core
loss in case of dc inductors or flyback transformers. Finally, the method doesn’t
include temperature changes, so it’s valid and accurate for one temperature only.
Usually parameters k, α and β are given for 100ºC, which is close to the operating
temperature of many magnetic components.
49
To overcome non-sinusoidal flux problem several methods were introduced [63,

65-68]. So called modified Steinmetz equation (MSE) seems to be robust and
efficient method [63]. This method doesn’t require any additional parameters
beyond Steinmetz equation coefficients k, α and β. The method bases on the fact,
that the core loss is a function of the magnetization rate rather than the switching
frequency alone. MSE is given by (3.50), where fr is the repetition frequency, feq is
so called equivalent frequency. The repetition frequency fr in a switch mode
converter equals the switching frequency. The equivalent frequency feq is defined
as the frequency of the sinusoidal flux density, whose has the same flux density
amplitude ∆B and the same average magnetization rate like the considered arbitrary
flux density waveform. For a piecewise linear flux density waveform the
equivalent frequency is given by (3.51).
pFe = ( k ⋅ f eqα -1 ⋅ ∆Bβ ) ⋅ f r (3.50)
2
2 K  Bk − Bk −1  1
2 ∑
f eq = ⋅  ⋅ (3.51)
π k = 2  2 ⋅ ∆B  tk − tk −1
Influence of the dc bias (premagnetization) on the core loss is discussed in
literature and several authors reported an increased core loss under dc bias
conditions [63, 69, 70]. Estimation of such a loss is difficult, mainly due to lack of
data from manufacturers and lack of a robust, widely accepted model. In [63] a
simple extension of Steinmetz equation is proposed. In this method the parameter
k from Steinmetz equation (3.49) is replaced by new parameter k’ given by (3.52).
The new parameter depends on both dc and ac components of the flux density. It
also utilizes two new parameters K1 and K2 which are extracted for each ferrite
material, since they are not provided by manufacturers so far.
 
 ∆B 
−
k' = k ⋅ 1 + K1 ⋅ Bdc ⋅ e K2   (3.52)
 
 
Analysis of a ferrite datasheet shows that the core loss is a strong function of the
core temperature. I.e. the popular power ferrite material 3F3 operating at 100 kHz
and 100 mT has the specific core loss 170 mW/cm3 in room temperature (20 ºC).
Under the conditions (frequency and flux density), but in operating temperature of
100 ºC the specific core loss goes down to about 70 mW/cm3 [71]. In the same
datasheet Steinmetz equation parameters are given for operating temperature of 100
ºC only. So, estimation of the core loss based on Steinmetz equation only may lead
to significant errors if the operating temperature differs from 100 ºC (e.g. during the
converter warm-up period). To overcome this problem a simple extension of
Steinmetz equation is proposed in [65] and it’s given by (3.53). It simply includes
parabolic correlation between the core loss and the temperature T. Parameters ct2,
ct1 and ct have to be extracted experimentally or from the datasheet for the core
material. Moreover these parameters are the flux density and the frequency
dependent in most cases. So, similarly to Steinmetz equation parameters they are
valid only for the certain frequency/flux density range.
50
Figure 3.13 The flyback converter and the flyback transformer currents
pFe = k ⋅ f α ⋅ ∆Bβ ⋅ ( ct2 ⋅ T 2 − ct1 ⋅ T + ct ) (3.53)

Even in spite of these limitation Steinmetz equation proves to be a very robust
and fairly sufficient method for the core loss preliminary estimation. More
accurate results, by the cost of computation time, are obtained with a combined
hysteresis-FEM models [72].
3.5.3 Types of magnetic devices and their losses

Various kinds of magnetic devices are used in power electronics. Depending on
the kind of the magnetic device core and copper losses might be balanced or one of
them can be dominant. In this section losses in magnetic components of non-
isolated dc-dc converters are discussed and the focus is on converters presented in
Chapter 2. It includes a dc inductor, a flyback transformer and a high frequency
transformer.
A dc inductor is a storage element, so it utilizes an air gap to store energy. For
a constant permeability µ the flux density B(t) is directly proportional to the current
iL(t) (3.54), so a small current ripple ∆iL indicates a small ac flux density component
∆B. In case of a gapped ferrite core it means a very low core loss, so it’s most
likely so called saturation limited design (powder cores however may still exhibit a
considerable core loss). Also a very small ac current ensures a fairly low ac copper
loss, even if the ac resistance is much larger than the dc resistance of the winding.
In a dc inductor design dc copper loss is dominant. In case of an inductor current
with a larger ac component (e.g. close to DCM) the ac copper loss may become
significant very quickly, so the ac resistance become an important constraint.
µ ⋅ n ⋅ iL ( t )
B (t ) = (3.54)
lg
A flyback transformer , similarly to the dc inductor, stores energy in an air gap.
The difference is that the flyback transformer uses one winding to store energy (e.g.
winding n1) and another to release energy (e.g. winding n2). The equivalent circuit
and relevant waveforms are presented on Figure 3.13. The inductor LM represents
the magnetizing inductance referred to the primary winding n1. Assuming a
51
constant permeability the flux density B(t) is directly proportional to the

magnetizing current iLM(t) – by analogy to (3.54). Now, the ac flux density ∆B
may remain small, so the core loss is small too. But both winding currents in1(t)
and in2(t) contain a large dc and ac components, since they are discontinuous. In a
flyback transformer design the focus is on reduction of both, dc and ac winding
resistances.
A transformer contrary to the flyback transformer doesn’t store energy and it
utilizes an ungapped ferrite core having a very high permeability µ. Now, the flux
density B(t) is not a direct function of winding currents, but it’s proportional to the
volt-seconds λ1 applied to the winding (3.55). In a typical high frequency design
the peak-peak flux density swing is large and limited by the core loss, not by the
saturation level. Usually both core and copper loss have to be taken into
consideration.
λ1
∆B = (3.55)
n1 ⋅ Ac
More detail about losses in other magnetic components one can find in [47].
3.6 Capacitor
In many cases a capacitor is treated as a lossless component. In fact it’s not
lossless. The complete equivalent circuit of a real capacitor is presented on Figure
3.14 a). It contains the ideal capacitor C, the parallel resistance Rp, the series
resistance Rs and the stray inductance Ls. If the capacitor operates well below it’s
resonance frequency then the stray inductance Ls is neglected and the complete
equivalent circuit reduces. Figure 3.14 b) presents the series equivalent circuit,
which contains the ideal capacitor C and so called equivalent series resistance ESR.
From the power loss point of view ESR is the key component and the capacitor
power loss is given by (3.56) in a general case. ESR value is sometimes available
in the capacitor datasheet.
PC = ESR ⋅ I C2 ( rms ) (3.56)
However many manufacturers publish so called dissipation factor (or loss factor)
tanδ instead of ESR. The dissipation factor is given by (3.57) and relates to the
parallel equivalent circuit presented on Figure 3.14 c). The circuit contains the
ideal capacitor C and the parallel resistance Rp, which originally represented the
finite resistance of the capacitor dielectric material. Unfortunately the parallel
circuit is not convenient for the analysis, since the capacitor current iC(t) splits into
two currents flowing thru the capacitor C and the parallel resistor Rp. Using (3.58)
and (3.59) one can find ESR based on the dissipation factor tanδ and transform the
parallel circuit into the series circuit. The capacitance C remains not affected.
1
tan δ = (3.57)
2 ⋅ π ⋅ f ⋅ C ⋅ Rp
52
Figure 3.14 Capacitor equivalent circuits - a) complete; b) series; c) parallel
Figure 3.15 Current flow in a boost converter's node A
1
Rp = (3.58)
2 ⋅ π ⋅ f ⋅ C ⋅ tan δ
 1 
 Rp ⋅ ( j ⋅ 2 ⋅ π ⋅ f ⋅ C ) 
ESR = Re   (3.59)
R + 1 
 p ( j ⋅ 2 ⋅π ⋅ f ⋅ C ) 
 
Analysis of the series equivalent circuit is fairly easy since there is only one path
for the capacitor current iC(t). To find the capacitor current iC(t) let’s consider the
node A of the boost converter presented on Figure 3.15. The diode D conducts
unidirectional pulse width modulated current, which contains ac and dc components.
Assuming that the output capacitor C is large and the output voltage ripple is
negligible, the output current is pure dc current Iout. So, the ac component of the
diode current can flow only thru the capacitor C. Moreover, it’s well known that in
steady state there is no dc current in any capacitor. So, the rms capacitor current is
given by (3.60). In many cases it’s enough to estimate the power loss in the
capacitor, but for some types of capacitors ESR (or tanδ) is a strong function of the
frequency. In such case it’s possible to use harmonic analysis of the capacitor
current and find the power loss caused for each harmonic.
I C ( rms ) = I D2 ( rms ) − I out
2
(3.60)
53
3.7 Other losses

Losses described in sections from 3.3 to 3.6 above are associated with particular
components in the power circuit of the converter and they are major part of total
losses inside the converter. In a low power converters (let’s say below 100 W)
losses coming from the control circuitry (analog, microcontroller, DSP or similar)
and from the measurement circuitry and sensors may become a significant part of
total losses. In medium / high power converters (let’s say above 1000 W) these
losses are usually fairly small in compare to the power circuit losses.
Any commercial converter has to fulfill certain requirements regarding
harmonics and EMI pollution [73, 74]. It means that the converter has to equipped
with input and output filters or EMI filters. Such a filters are built from passive
components (magnetics and capacitors), so they suffer the same kind of losses like
passive components in the power circuit.
In a low voltage / high current application a substantial conduction loss may
occur in a high current path (PCB track or busbar) and interconnections. So, in
such case it’s necessary to ensure a large enough copper cross section area.
Moreover, one shall be aware of the skin effect in case of an ac current path. Such
an ac wiring should be as short as possible too in order to reduce a stray inductance
and improve EMI performance. More details and guidelines about a layout design
on can find in [75].
For the power loss estimation purposes non of these losses is taken into account.
It simplifies calculations on the one hand. On the other hand gives a good basis for
a fair comparison between different topologies.
3.8 Model verification

The modeling approach and models presented in Chapter 3 have to be verified
and 200 W boost converter is built for this purpose. All kinds of losses discussed
in previous paragraphs exist in this simple topology, so they can be calculated,
measured and verified. Verification process is done in following steps:
• validation of the converter model
• components parameters extraction
• calculation of losses
• measurement of losses
First, the averaged model of the boost converter is compared against the circuit
simulator based model. The focus is on current values crucial for the power loss
estimation. It includes: the input current, the transistor rms current, the diode
current, the output current and currents values just before and just after switching.
Results of the comparison are summarized in Table 3.1. A good agreement
confirms usefulness of the presented modeling approach.
54
Table 3.1 Currents calculated by the PLECS model and the averaged model
PLECS Averaged model
Input dc current [A] 4.51 4.50
Input peak-peak current [A] 7.70 7.69
Transistor rms current [A] 4.10 4.10
Transistor turn-on / turn-off currents [A] 0.64 / 8.35 0.65 / 8.35
Diode rms current [A] 2.91 2.90
Output dc current [A] 1.50 1.50
Table 3.2 Operating points used for the model verification

Reference 1 2 3 4
Input voltage [V] 30 20 30 30 30
Output voltage [V] 90
Output power [W] 135 135 200 135 135
Gate resistor [Ω] 18 18 18 5 18
Inductor (low / high core loss) low low low low high
Next, components parameters are extracted and losses are calculated. Most of
required parameters are read from components’ datasheets and only very few are
extracted from measurements (e.g. inductor winding resistance). Power losses are
calculated for the reference operating point first. Then, a single operational
parameter is disturbed in this way, that only a single kind of loss varies significantly,
while other remain about the same. Tested operating points are presented in Table
3.2 while calculated power losses are summarized in Table 3.3.
Finally, power losses in the breadboard are measured. The total converter loss
is found with used of precise multimeters, as a difference between the input power
and the output power (3.61). This measurement is usually enough to justify
efficiency of a breadboard. Unfortunately data collected in this way are
insufficient for the model verification. Knowledge about loss distribution in the
converter is a must. In an ideal case each kind of loss in each component shall be
known. In practice it’s easy to find some kind of losses (e.g. conduction losses),
while a direct measurement of some others is hard (e.g. the switching loss or the
core loss).
Ploss = Pin − Pout (3.61)
The inductor copper loss is fairly easy to find, since it’s easy to measure the
inductor current – the dc component IL dc is measured with a precise ampmeter while
the peak-peak value iL(pp) is measured with a current probe. The inductor dc and ac
resistances are measured with a high precision RLC-meter.
Estimation of the transistor conduction loss bases on the on-state resistance read
from the datasheet and the measured transistor rms current. A direct measurement
of the current is difficult, because it’s a high frequency current path and use of a
55
precise ampmeter is not possible, while accuracy of current probes is insufficient.

However, the transistor rms current may be calculated from the inductor current and
the duty cycle (3.62). Now, the transistor conduction loss is found too.
2
1  ∆i 
I T( rms ) = IL ⋅ 1+ ⋅  L  ⋅ D (3.62)
3  IL 
A direct measurement of the transistor switching losses is straightforward in the
theory, but in practice it is not easy. It requires a very fast high voltage probe and a
special current sensor, which can be placed directly on the transistor lead (e.g.
Rogowski coil). Next, the time delays of the voltage sensing path and the current
sensing path have to match exactly. In case of a very fast transistor switching a
limited bandwidth may lead to a significant measurement error. Here, a method
based on the transistor temperature measurement is proposed. First, the transistor
is placed on a separate heatsink and the turned-on transistor is connected to a dc
voltage source. The transistor drain current, the drain-source voltage drop and the
temperature are measured, so the relation between the transistor loss and the
temperature are found. This method has an additional benefit of measuring the
actual on-state resistance of the transistor, so datasheet values are verified too.
Then the transistor with the heatsink are placed in the converter. During normal
operation of the boost converter the following losses dissipate in it: the transistor
conduction loss, the transistor switching loss, the diode reverse recovery loss and the
loss related to the diode capacitance. Since the transistor conduction loss is known,
the remaining part relates to the switching loss.
The diode losses are estimated in the same way like the transistor loss was.
First, the diode is placed on a separate heatsink and the heatsink temperature versus
conduction loss is measured in dc conditions. Then, the diode together with the
heatsink are placed in the converter, so the temperature of the heatsink indicates the
total diode loss.
The capacitor loss due to ESR may be found by a direct measurement of the
capacitor current iC(t). In many cases however the capacitor leads are very short
and it’s not possible to connect any current sensor or ampmeter. In such a case
analysis of the diode current and the output current is useful, like described in
section 3.6. ESR of the capacitor may be read from the datasheet, but for better
accuracy it’s measured with a precise RLC-meter.
Now the only remaining loss in the power circuit of the boost converter is the
inductor core loss. Measurement of the core loss usually bases on observation of
the hysteresis loop [70, 76]. This method requires however an additional
equipment, like a hysteresis loop tracer. In the breadboard two inductors are used.
The first is designed in the way that the flux density amplitude is very small and the
core loss is very small too. The other inductor is designed to have a significantly
higher flux density amplitude, so the core loss become a visible part of the total
converter loss. Both inductors have a similar inductance value, so the input current
waveform shall not be affected, thus other losses remain about the same. The core
loss is considered as a remaining part of the converter total losses according to
(3.63).
56
Table 3.3 Power loss simulation (calculation) results

Reference 1 2 3 4
Transistor conduction loss 0.29 0.74 0.64 0.29 0.29
Transistor switching loss 0.62 0.82 0.85 0.25 0.62
Diode conduction loss 0.93 1.01 1.49 0.93 0.93
Core loss 0.24 0.13 0.24 0.24 1.73
Ac copper loss 0.41 0.25 0.42 0.41 0.14
Dc copper loss 0.11 0.24 0.23 0.10 0.05
Capacitor loss 0.03 0.04 0.06 0.03 0.03
Total calculated loss 2.63 3.23 3.93 2.25 3.79
Table 3.4 Power loss measurement results

Reference 1 2 3 4
Transistor total loss 0.94 1.56 1.57 0.55 0.94
Diode total loss 0.99 1.16 1.62 0.97 1.01
Estimated core loss 0.26 0.15 0.26 0.26 1.87
Ac copper loss 0.41 0.25 0.41 0.41 0.14
Dc copper loss 0.11 0.24 0.24 0.10 0.05
Capacitor loss 0.03 0.04 0.06 0.03 0.03
Measured total loss 2.74 3.40 4.16 2.32 4.04
PFe = Ploss − ( PTcond + PTsw ) − PDcond − PTrr − PCuac − PCudc − PC (3.63)

The converter losses are measured in several operating points, like it was
simulated before. Table 3.3 and Table 3.4 contain results of simulations and
measurements.
57
Chapter 4
Design and Optimization of the Converter
The converter comparison made in Chapter 2 clearly indicated the non-isolated
push-pull-boost converter as a good candidate for a fuel cell converter. In this
chapter design and optimization of this topology is presented. It starts with a
detailed analysis of the converter, with a special focus on magnetic components
operating under different conditions. Then a preliminary calculations are done and
power semiconductors are selected among available state-of-the-art components.
The last part of this chapter deals with optimization of magnetic components. It’s
presented that integration of the inductor and the push-pull transformer leads to a
significant size reduction compared with the push-pull-boost converter with
separated inductor and the transformer.
4.1 Converter analysis

In this section the non-isolated push-pull-boost converter, presented on Figure
4.1, is discussed. For reader convenience the basic operating stages, already
presented in paragraph 2.2.6, are described again. Then more detailed discussion
of the converter operation is provided in following paragraphs.
This topology originally was presented in [44] and it’s called the boost converter
based on three-state switching cell by the authors. In [31] a detailed description of
the converter, together with key equations and waveforms, is provided. It makes a
Figure 4.1 The non-isolated push-pull-boost converter diagram

58
Figure 4.2 Key waveforms of the converter operating in overlapping mode
base for the converter analysis presented in following text.

The converter may be split into two parts, like it’s shown by the dashed line on
Figure 4.1. The high current part of the converter consists of the inductor, primary
windings of the push-pull transformer and both transistors. The low current part
consists of all diodes, the secondary side of the transformer and capacitors.
Basically the converter operates in CCM with the duty cycle above 50% (conditions
for operation with the duty cycle below 50% are introduced in section 4.1.2). Gate
signals are 180º phase shifted, like presented on Figure 4.2. Under this conditions
in a steady state the converter has four operating stages. Two of them (1 and 3) are
identical, while two remaining stages (2 and 4) are symmetrical.
Stage 1 lasts from time t0 to t1. During this stage both transistors are turned-on
and diodes D1 and D11 are reverse biased. Primary windings of the transformer are
shorted and the primary voltage is zero during this period. So, the secondary
voltage is zero too and diodes D2 and D3 are reverse biased. The load is supplied
from the output capacitors during this period. Neglecting the primary winging
resistance and the transistor on-state resistance the whole input voltage Vin is applied
to the inductor L1. The inductor current ripple amplitude is given by (4.1). By
rearranging (4.1) the inductance L1 is found with (4.2).
Vin ⋅ ( D − 0.5 )
∆iL1 = (4.1)
2 ⋅ L1 ⋅ fs
Vin ⋅ ( D − 0.5 )
L1 = (4.2)
2 ⋅ ∆iL1 ⋅ f s
Stage 2 lasts from t1 to t2. At the beginning of this stage the transistor T1 is
turned-off and it remains off until time t2. Because there is a non-zero leakage
59
inductance of the winding (not shown on Figure 4.1) the current in1(t) is commutated
from the transistor T1 to the diode D1. The diode is forward biased now, so the
capacitor voltage vC1(t) is applied to series connected primary windings. A positive
voltage vn2(t) induces in the secondary winding and if the induced voltage is higher
than the capacitor voltage (vn2(t)>vC3(t)), then the diode D3 become forward biased.
Energy stored in the inductor L1 during stage 1 is now transferred to output
capacitors C1 and C3.
Stage 3 lasts from t2 to t3 and it’s identical to stage 1. Stage 4 lasts from t3 to t4
and it’s symmetrical to stage 2, while the transistor T11 is turned-off. During stage
4 capacitors C1 and C2 are recharged.
4.1.1 Voltage gain

The static voltage gain of the converter is found by analysis of the inductor volt-
second balance [47] and assuming that capacitor voltages have negligible ripples.
From the inductor L1 point of view there are only two operational stages in CCM –
charging periods (stages 1 and 3) and discharging periods (stages 2 and 4). During
the charging period the input voltage is applied to the inductor and the volt-second
λL1+ is given by (4.3). During the discharge period the voltage across the inductor
equals the difference between the input voltage and the half of the capacitor voltage
VC1, since n1=n11. The volt-second λL1- applied to the inductor during this period is
given by (4.4). Now, both volt-seconds during charging and discharging periods
have to stay in balance. Now, equations (4.3) and (4.4) are combined in such a
way that they describe the capacitor voltage VC1 (4.5). Capacitor voltages VC2 and
VC3 are found with (4.6). The output voltage Vout is a sum of capacitor voltages
given by (4.7), so the voltage gain is given by (4.8).
Vin
λ L1+ = ⋅ ( D − 0.5 ) (4.3)
fs
 VC1 
 2 − Vin 
λ L1- =  ⋅ 1− D
( ) (4.4)
fs
1
VC1 = Vin (4.5)
1− D
n
VC 2 = VC 3 = ⋅ VC1 (4.6)
2
n +1
Vout = VC1 + VC 2 + VC 3 = ⋅ Vin (4.7)
1− D
Vout 1 + n
M= = (4.8)
Vin 1 − D
60
4.1.2 Duty cycle below 50%

An isolated current-fed push-pull converter cannot operate with duty cycle below
50% in CCM. It’s simply because a continuous current path for the inductor
current has to be ensured, so at least one of transistors has to be turned-on. The
non-isolated version of this converter overcomes this limitation. Diodes D1 and
D11 provide a continuous path for the inductor current, when both transistors are off.
Before converter’s operation with the duty cycle below 50% is discussed let’s
take a look on what happen at 50% duty cycle. In this operating point there are
only two operating stages – 2 and 4, when one transistor is turned-on and the other is
turned-off. Stages 1 and 3 don’t exist since there is no overlapping here. There
are no inductor charging or discharging periods, the input current is ideally smooth
and the inductor voltage vL1(t) is zero all the time. The input voltage Vin is applied
directly to one of primary windings and since n1=n11 the capacitor voltage VC1 has to
be twice the input voltage (4.9). The output voltage is given by (4.10). If the
duty cycle D=50% is substituted into (4.7) one can see that (4.7) and (4.10) equals
each other – so the voltage gain is continuous on this boundary.
VC1 n1 + n11
= =2 (4.9)
Vin n1
Vout = 2 ⋅ (1 + n ) ⋅ Vin (4.10)

Now, the duty cycle decreases below 50% slightly and there are four major
operating stages again. Figure 4.4 presents related waveforms. Stages 1 and 3 are
symmetrical – one transistor is turned-on and the other is turned-off. Stages 2 and
4 are identical – both transistors are turned-off.
During stage 1 the transistor T1 is turned-off and the diode D1 conducts. The
capacitor voltage VC1 is applied to series connected primary windings of the
Figure 4.3The converter waveforms at duty cycle lower than 50%

61
Figure 4.4 The converter observed waveforms while operating with duty cycle of
50% (top-left), 40% (top-right), 39.5% (bottom-left) and 25% (bottom-right),
Ch1 – the gating signal for T1; Ch2 – the drain-source voltage of T1;
Ch3 – the secondary winding voltage; Ch4 – the inductor current
transformer. Also the positive voltage induces in the secondary winding and the
capacitor C3 is charged thru the diode D3.
During stage 2 both transistors are turned-off and diodes D1 and D11 conducts.
Primary windings are effectively shorted and there is no voltage across the
secondary winding.
Stage 3 is symmetrical to stage 1 and stage 4 is identical to stage 2.
The capacitor voltage VC1 is higher then the input voltage Vin. Stages 2 and 4
are considered as the inductor discharging periods, while stages 1 and 3 are charging
ones. The voltage gain is found by analysis of the inductor volt-second balance
and it’s given by (4.11).
Vout 1 + n
M= = (4.11)
Vin 1 − D
However it’s found empirically that at lower duty cycles the voltage gain doesn’t
follow (4.11). Figure 4.4 presents observed waveforms for duty cycles below 50%.
It’s clear that at one point the secondary voltage (the voltage across capacitors C2
and C3) collapses and only the capacitor C1 contributes to the output voltage.
Observed waveforms look like in a two phase boost converter with a coupled
inductor (see section 2.2.2). The voltage gain of the converter is measured and
62
Voltage gain
4
3 Push−pull−boost conv.
2−ph. boost with coupled inductor
Measured gain
2
0.3 0.35 0.4 0.45
Duty cycle
Figure 4.5 Measured voltage gain of the converter
vs. theoretical gain given by (2.15) and (4.11)
compared with prediction given by (4.11) for the non-isolated push-pull converter
and (2.15) for the two phase boost with the coupled inductor. Figure 4.5 shows the
result. Above the crossing point the voltage gain follows (4.11) and below the
crossing point it follows (2.15). The critical duty cycle (at the crossing point) is
given by (4.12).
n
Dcr = (4.12)
2 ⋅ n +1
It’s very important to note that the converter may operate safely with duty cycle
between 50% and the critical duty cycle. However operation below the critical
duty cycle is not recommended since the whole output voltage appears across the
capacitor C1 only. So, both transistors and diodes D1, D11 may experience an
excessive and undesired voltage stress.
4.1.3 Currents
Respective dc, rms and ac currents are calculated using the method briefly
introduced in paragraph 3.2.1 and using general equations (4.13), (4.14) and (4.16).
n
 t − t   i ( t ) + i ( tk −1 ) 
I = ∑  k k −1  ⋅  k  (4.13)
k =1  t n − t0   2 
n
 tk − tk −1 
I ( rms ) = ∑k =1  t n − t0 
 ⋅ uk (4.14)
1
uk = ⋅  i t( k −1)+( ) + i ( t( ) ) ⋅ i ( t 
2
) + i ( tk − )
2
k −1 + k−  (4.15)
3  
63
Figure 4.6 The primary winding current waveform and

its values at the beginning and at the end of each stage
I ( ac ) = I (2rms ) + I 2 (4.16)
In the non-isolated push-pull-boost converter transistor and diode currents are
parts of the primary winding current. To find the primary winding current, and
thus the diode and the transistor currents, it’s necessary to find the primary current
values at the beginning and at the end of each stage (trapezoidal segment), i.e. just
before and just after transistor switching, like presented on Figure 4.6. For this
purpose following assumptions are made:
• the input current is fairly smooth
• currents iD1(t) and iD11(t) have the same shape like currents iD3(t) and iD2(t)
respectively, but they are lower by half (4.17).
1
iD1 (t ) = ⋅ iD3 (t ) (4.17)
2
During stages 1 and 3 the inductor current splits into two exact halves and flows
thru both primary windings. The winding currents are given by (4.18) during these
periods.
1
in1 ( tk ) = in11 ( tk ) =
⋅ iL1 ( tk ) k =0 + ,1− ,2 + ,3− (4.18)
2
During stage 2 the winding current in1(t) flows thru the diode D1 only. Also the
secondary winding current in2(t) flows thru the diode D3 only. At any time the sum
of the transformer winging currents has to be zero (4.19). Also the inductor current
may flow only into primary windings (4.20).
in11 (t ) − in1 (t ) − n ⋅ in 2 (t ) = 0 (4.19)
iL1 (t ) = in1 (t ) + in11 (t ) (4.20)
Manipulation of these equations leads to the solution for the winding current
in1(t) during stage 2 (4.21). The other winding current in11(t) is found with (4.22).
64
iL1 (tk )
in1 (tk ) = (4.21)
2 ⋅ (1 + n )
k =1+ ,2 −
iL1 (tk )
in11 (tk ) = iL1 (tk ) − (4.22)
2 ⋅ (1 + n )
k =1+ ,2 −
Stage 4 s symmetrical to stage 2 and primary currents in1 and in11 transpose only
((4.23) and (4.24)).
iL1 (tk )
in1 (tk ) = iL1 (tk ) − (4.23)
2 ⋅ (1 + n )
k = 3 + ,4 −
iL1 (tk )
in11 (tk ) = (4.24)
2 ⋅ (1 + n )
k = 3 + ,4 −
Rearrangement of (4.19) results in the equation for the secondary current in2(t)
(4.25) and it’s valid for all stages.
in11 ( t ) − in1 ( t )
in 2 ( t ) = (4.25)
n
Now, the inductor current iL1(t) in time instants t0, t1, t2 and so on is found as the
minimum or the maximum instantaneous inductor current (4.26). It’s assumed that
the inductor current is constant during a short switching period – so it’s the same
just before and just after switching (iin(tk-)=iin(tk+)).
 I L1 − ∆iL1 k = 0,2,4
iL1 ( tk ) =  (4.26)
 I L1 + ∆iL1 k =1,3
In the considered converter respective time instants tk are found based on the
switching frequency fs and the duty cycle D. Details are presented in Appendix B
in form of MATLAB® equations.
4.1.4 Inductor and transformer

The inductor current ripple is proportional to the applied volt-second which is
given by (4.27) for the duty cycle above 50%. Figure 4.7 presents the normalized
inductor current ripple as a function of the duty cycle. Theoretically, there inductor
current is smooth at 50% and 100% duty cycle, while the maximum current ripple
occurs at 75% duty cycle. The inductor current contains a large dc component with
a limited ac ripple. So, in the inductor design the focus is on a low dc resistance.
Vout ⋅ (1 − D ) ⋅ ( D − 0.5 )
λ L1 = (4.27)
fs
The transformer flux density amplitude ∆B is found by analysis of the volt-
second applied to the transformer and it’s given by (4.28). The equation is truth for
duty cycle above 50%. Figure 4.8 presents the normalized flux density amplitude
as a function of duty cycle. The largest flux density exists at 50% duty cycle and it
decreases while the duty cycle increases.
65
Normalized current ripple

0.8
0.6
0.4
0.2
0
0.5 0.6 0.7 0.8 0.9 1
Duty cycle
Figure 4.7 The normalized inductor current ripple
as a function of the duty cycle
1
Normalized flux density
0.8
0.6
0.4
0.2
0
0.5 0.6 0.7 0.8 0.9 1
Duty cycle
Figure 4.8 The normalized transformer flux density
as a function of the duty cycle
Vout ⋅ (1 − D )
λ n1 = (4.28)
f s ⋅ (1 + n )
At this point it’s important to note that each of primary currents contains three
components: the half of the input dc current, the half of the input current ac ripple,
and the reflected secondary current. All three components are presented on Figure
4.9 and their paths are presented on Figure 4.10. The input current (dc and ac
ripple) flows thru the inductor winding and thru primary windings which are
effectively in parallel and the resistance seen by this current component is expressed
as RL1+Rn1||Rn11. The secondary winding current flows in the secondary winding,
66
Figure 4.9 The primary current components
Figure 4.10 Flow of current components in the push-pull transformer
but it’s also reflected to the primary windings, which are effectively series connected
for this current component and the resistance seen by this current component is
expressed as Rn2+Rn1+Rn11. So, it’s very important to arrange transformer windings
in such way, that resistance seen by all current components is low.
4.1.5 Summary
In this section the non-isolated push-pull-boost converter has been analyzed in
details. It’s operation under different conditions has been explained. Also set of
equations describing the converter has been developed and presented to the reader.
These equations are included into the averaged converter model, which is used for
the converter design process described in following paragraphs.
4.2 Design
In the previous section the non-isolated push-pull-boost converter has been
analyzed in details and set of equations have been developed. In this paragraph the
preliminary design considerations and calculations are presented.
The specification of the breadboard originate from the project limitations in
section 1.4.1 and it’s presented in Table 4.1. The primary application is a modular
converter for a fuel cell. It results in variable input voltage and gives some
freedom in terms of the converter’s output power (modularity). So, some
provisions are done. The converter’s power losses are optimized for the lowest
input voltage, which results in the highest input current. However all components
have to sustain the highest input voltage. The peak efficiency of 98% leaves a
room for only 2% of losses. So, the projected power loss distribution at about half
of the rated output power (efficiency peak point) is as follow:
67
Table 4.1 The specification of the breadboard

Input voltage range Vin(min)-Vin(max) 30-60 V
Output voltage Vout 400 V
Rated output power Pout 100-1000 W
Input current ripple at rated current Iin(pp) 20%
Output voltage ripple Vout(pp) 5%
Target efficiency η 98%
• 0.5% for MOSFETs

• 0.5% for diodes
• 0.5% for the transformer
• 0.3% for the inductor
• 0.2% for capacitors and other losses.
The first step is to find circuit parameters like the turns ratio, the duty cycle
range. As mentioned before the converter may operate below 50% duty cycle, but
in order to avoid an excessive voltage stress it’s desired to operate in overlapping
mode and the minimum duty cycle is assumed to be 50%. The maximum allowed
turns ratio is found with (4.29) at maximum input voltage Vin(max). However use of
a foil winding in the transformer requires an integer turns ration, thus n=2 is used
(see section 4.2.4). The minimum and maximum duty cycle is found at the highest
and at the lowest input voltage ((4.30) and (4.31)).
n=
Vout
Vin ( max )
( )
⋅ 1 − D( min ) − 1 =
400
60
⋅ (1 − 0.5 ) − 1 = 2.33 ≈ 2 (4.29)
Vin ( max ) 60
D( min ) = 1 − (1 + n ) ⋅ = 1 − (1 + 2 ) ⋅ = 0.55 (4.30)
Vout 400
Vin ( min ) 30
D( max ) = 1 − (1 + n ) ⋅ = 1 − (1 + 2 ) ⋅ = 0.78 (4.31)
Vout 400
At this point two key parameters are still not known – the rated output power and
the switching frequency. How to fix them in this case? One can pick the output
power and the switching frequency arbitrary and then try to find suitable
components – especially transistors. The other way is exactly opposite – find a
state-of-the-art transistor first and then find the optimum operating point for it. The
second way is shortly described in paragraph 4.2.1. Once the output power and the
switching frequency are known, one may calculate currents in the circuit and design
magnetic components.
4.2.1 Power MOSFETs

Selection of the most suitable MOSFET is a very important step in any converter
design and the choice shall be careful. There are many available MOSFETs, so
68
Table 4.2 Basic data of IFRP4321PbF MOSFET

Breakdown voltage 150 V
Rated current 78 A
On-state resistance (at 25 ºC) 12 mΩ
Total gate charge 71 nC
Input capacitance 4460 pF
how to find the one? How to make sure, that the chosen MOSFET will perform
well in the converter?
The first step is very general and simply rejects all parts with inappropriate
voltage and current ratings. Only devices which have the current and the voltage
rating inside a certain range goes thru to the next step.
The second step bases on so called figure of merit (FOM). In this method the
performance of transistor is evaluated based on product of two transistor’s
parameters. It’s important that each of these parameters is directly linked with one
of two major loss mechanisms in the transistor – the conduction loss and the
switching loss. The most common parameters used are the on-state resistance vs.
the total gate charge (4.32) or the on-state resistance vs. the effective output
capacitance of the transistor (4.33). It’s expected that the device with the minimum
FOM will provide the best overall performance, while it minimizes total losses. In
fact, the method points out two or three components which are evaluated in the third
stage.
FOM QG = RDS( on ) ⋅ QGtot (4.32)
FOM Coss = RDS( on ) ⋅ Coss (4.33)

Once the transistor is selected the optimum operating point is found. The
250
200
[nC]
IRFP4568PbF
Gtot
150
Q
100 IRFP4115PbF
IRFP4321PbF
5 10 15 20
R [mohm]
DS(on)
Figure 4.11 On-state resistance vs. total gate charge of several 150 V MOSFETs
69
limit PTcond
1.2
P
Tcond
1
[W]
Tcond
0.8
0.6
P
0.4
0.2
0
200 400 600 800 1000
P [W]
out
Figure 4.12 The conduction loss limit and the actual conduction loss
vs. the output rated power of the converter
averaged converter model and the transistor model (both introduced in Chapter 3)
are used to calculate actual power losses in the transistor operating under different
conditions.
Now, the numerical example it provided. The wanted transistor is the 150 V
rated MOSFET (4.34). After looking at the high power end (1000 W) the
transistor’s rms current is about 18.1 A (according to (2.62) and the method
presented in section 4.1.3). Including the current derating factor between 2-10, the
wanted transistor’s rated current is between 36.2-181 A. Using a search engine
[77] 54 different MOSFETs are found. They are sorted by the on-state resistance
and top 10 components go to the second stage. Figure 4.11 shows the on-state
resistance vs. the total gate charge of considered MOSFETs. The top three
MOSFETs are marked with different colors. At the time of the converter design
only IRFP4321PbF [78] was available and since it was superior in compare with
other available devices it’s selected for further work. Basic parameters of the
MOSFET are summarized in Table 4.2.
1 1
VT1( off ) = VT11( off ) = VD1( R ) = VD11( R ) =
⋅ Vout = ⋅ 400 = 133.3 V (4.34)
n +1 1+ 2
Once the transistor is selected the next step is to find the optimum operating
point for it. This point shall cover the converter peak efficiency point, thus the
transistor loss limit is known. It equals 0.5% of the output power at the peak
efficiency point. The transistor operates in its optimum when the conduction loss
equals the switching loss, so the allowed conduction loss in a single transistor is
0.125% of the converter output power at the peak efficiency point. Figure 4.12
shows the conduction loss limit for a given rated power and the actual conduction
loss calculated by the model. The crossing point indicated the optimum rated
power, which in this case is about 500 W. So, the expected conduction loss is
about 0.31 W per MOSFET at 250 W output power (efficiency peak point). It’s
70
0.5
limit PTsw
0.4 P
Tsw
[W]
Tsw 0.3
0.2
P
0.1
0
2 4 6 8 10
f [Hz] 4
s x 10
Figure 4.13 The switching loss limit and the actual switching loss
vs. switching frequency
important to note that calculations are sensitive to the MOSFET junction

temperature, which is 40 ºC in this case.
The allowed switching loss is easily found since it equals the conduction loss in
the optimum point. Knowledge about the allowed switching loss is used to find the
optimum switching frequency. Still, the calculation is not straightforward since the
gate drive circuit has a great influence on switching times and the switching loss.
For the first approximation the gate circuit parameters are taken from the MOSFET
datasheet. Figure 4.13 shows the allowed switching loss and the actual switching
loss of the transistor. The crossing point indicated the optimum frequency and it’s
located close to 65 kHz. In the breadboard the switching loss will increase,
because of losses associated with the stray inductance and the diode. So, the
switching loss is reduced on account of these additional losses and it’s fixed at 50
kHz. In future it’s possible to work with the gate drive circuit and optimize
switching of the MOSFET.
Now the output rated power and the switching frequency are fixed and they equal
500 W and 50 kHz respectively.
4.2.2 Power diodes

According to the assumed power distribution the allowed total diode loss is 1.25
W at the lowest input voltage and 250 W output power (efficiency peak point). A
power diode usually is preselected based on its maximum reverse voltage and
forward current values (average, rms and peak) [79]. The choice of the optimum
device bases on the forward characteristic (conduction loss) and the switching
performance (reverse recovery and junction capacitance).
The reverse recovery loss is recognized as a severe loss mechanism in hard
switching converters. This issue is addressed first and the focus is on Schottky
diodes only, since they exhibit almost no reverse recovery loss. Moreover,
71
Table 4.3 Basic data of 30CPQ150PbF Schottky diode

Rated average current (per leg, at 25 ºC) 15 A
Non-repetitive peak current (5 µs sine) 1000 A
Forward voltage drop (at 25 ºC, estimated) 0.5 V
Forward resistance (per leg, at 25 ºC, estimated) 29 mΩ
Table 4.4 Basic data of IDT10S60C SiC Schottky diode

Rated average current (at 25 ºC) 10 A
Non-repetitive peak current (10 µs) 350 A
Forward voltage drop (at 25 ºC, estimated) 0.9 V
Forward resistance (at 25 ºC, estimated) 59 mΩ
Schottky diodes have a lower forward voltage in compare with silicon pn-junction
diodes having the same breakdown voltage. The price for it would be: a larger
reverse leakage current, a larger diode capacitance and a low maximum breakdown
voltage (up to 150-200 V) [43], than silicon pn-junction diodes. Since diodes D1
and D11 are 150 V rated (4.35) they are selected among available Schottky diodes.
Based on the average forward current requirement (4.29) there are 43 different
diodes preselected [77]. Even in spite of the increased diode capacitance the
conduction loss is expected to be the dominant loss mechanism. Thus, the diode
selection bases on its forward characteristic only and the device with the lowest
forward voltage drop is selected – it’s 30CPQ150PbF [78] dual in pack diode and
the basic parameters are summarized in Table 4.3.
1 1
VD2( R ) = VD3( R ) =
⋅ Vout = ⋅ 400 = 133.3 V (4.35)
n +1 1+ 2
For the given turns ratio diodes D2 and D3 require at least 350 V reverse voltage
(4.36). There are many 400 V ultra fast silicon diodes available, however their
reverse recovery behavior will cause significant losses since the reverse recovery
current flows thru several components, like presented on Figure 4.14. Next choice
is to use a novel silicon-carbide (SiC) diode instead. Currently SiC diodes are
available in 300 V, 600 V and 1200 V classes [80]. Use of a 600 V rated diodes
seems to be a safe choice and they are available from few manufacturers. Again,
the diodes are preselected based on its average forward current rating. Many
devices have a very similar forward voltage drop, however analysis of the forward
characteristic clearly indicated that the device with a higher current rating shall
provide a lower conduction loss. The diode IDT10S60C [80] is selected and it’s
basic parameters are summarized in Table 4.4.
n 2
VD2( R ) = VD3( R ) = ⋅ Vout = ⋅ 400 = 266.7 V (4.36)
n +1 1+ 2
72
Figure 4.14 The diode D3 reverse recovery current path
The estimated conduction loss in 150 V diodes is 0.16 W per diode, and in 600 V
SiC diodes it’s 0.72 W per diode. According to components datasheets there is no
reverse recovery loss. The estimated capacitive diode loss in 150 V diodes and 600
V diodes is 0.05 W and 0.07 W per diode. The total estimated power loss in all
diodes is about 2 W which exceeds allowed diode loss of 1.25 W. Selected diodes
are current state-of-the-art devices and paralleling of them will not reduce the power
loss. The only way to keep losses within limits is to reduce losses in other
components, like the inductor or the transformer.
4.2.3 Magnetic components

Magnetic components are a very special parts of any power converter. It’s
because they are designed and optimized for a particular converter, while
semiconductors and capacitors are selected among several off-the-shelf parts.
Number of available and appropriate parts is limited, so it’s relatively easy to find
the most suitable one. The choice may base on the component parameters, price or
availability. Of course one can find some off-the-shelf inductors or transformers.
In many cases these parts may satisfy some of the requirements, but it’s most likely
that they will not fulfill all of them. The only way is to design and optimize
magnetic components specially for the particular converter.
The first step is to choose a core shape and a suitable material. Core
manufacturers usually provide core selection guidelines and application notes, where
recommendations and practical clues are given. References [41, 81] provide a very
good discussion on properties of particular core shapes in different applications.
Based on them it’s found that EE shaped core is a good candidate to work with. It
provides a low winding cost, a simple assembly and a good flexibility. EE cores
are available in many different sizes and they are made out of different materials –
both ferrite and powder. Also it’s fairly easy to place different kind of wires on a
EE core, including a Litz wire, a round or a rectangular wire or a foil.
Once the core shape and material are determined, the magnetic device is about to
be designed. In the literature one may find several design procedures – some of
73
Table 4.5 The transformer design input data

Rated output power Pout 500 W
Switching frequency fs 50 kHz
Turns ratio 1:1:2
Primary currents In1 / In1(ac) / In1(rms) (at 250 W and 30 V) 4.25 / 1.97 / 4.66 A
Secondary current In2(ac) (at 250 W and 30 V) 1.91 A
Primary currents In1 / In1(ac) / In1(rms) (at 500 W and 30 V) 8.68 / 3.92 / 9.52 A
Secondary current In2(ac) (at 500 W and 30 V) 3.89 A
them are so called single pass procedures, while some others are iterative [43, 47,
82, 83]. In following paragraphs a detailed description of magnetic components
design is presented. A single pass area product AP method [43, 83] is used for a
core size approximation. Then the core size and basic winding parameters are
specified more precisely. Finally FEMM 4.0.1 software is used to minimize
windings resistance.
4.2.4 Transformer design

The transformer design input parameters are summarized in Table 4.6. For the
transformer core the power ferrite material type P is used [84]. The goal is to keep
power losses in the transformer below the limit stated in section 4.2, i.e. 1.25 W at
250 W output power and the lowest input voltage. First, the flux density amplitude
is calculated by rearranging natural Steinmetz equation (4.37). The calculation
bases on the allowed core loss which shall not exceed 100-150 mW/cm3 in the worst
case (the highest winding current).
1 1
 p β  100  2.86
∆B = 0.1⋅  Fe α  = 0.1⋅  1.36 
= 0.15 T (4.37)
 k ⋅ fs   0.158 ⋅ 50 
An appropriate core size is found using the area product method (4.38). Term
Pt in the equation is so called apparent power of the transformer and it’s defined as
the sum of power handled by all windings [83] and it’s given by (4.39). For given
input parameters the core size equivalent to E42/21/20 [84, 85] is sufficiently large.
It’s dimensions are summarized in Table 4.6.
Pt 1260
AP = = = 3.94 cm 4 (4.38)
kCu ⋅ k f ⋅∆B ⋅ J ⋅ f s 0.4 ⋅ 4 ⋅ 0.1⋅ 4e6 ⋅ 50e3
Pt = ∑ I nk ( rms ) ⋅ Vnk ( rms ) = 1260 VA (4.39)

k
Next step is to find number of primary winding turns. Two limitation are taken
into account. First, as mentioned before the worst case core loss per volume unit
shall not exceed 100-150 mW/cm3 and the related flux density amplitude of about
74
Table 4.6 Dimensions of EE42/21/20 core

Area product (maximum / effective) AP 5.97 / 4.03 cm4
Core cross section Ac 2.44 cm2
Window dimensions (maximum) 29.6×8.65 mm
Window dimensions (effective) 25.9×6.68 mm
Core volume 22.7 cm3
Table 4.7 The primary winding number of turns vs. core loss at 30 V input voltage
Number of turns Flux density amplitude Core loss per volume Total core loss
3
6 0.100 T 32 mW/cm 0.73 W
7 0.086 T 21 mW/cm3 0.48 W
3
8 0.075 T 14 mW/cm 0.32 W
9 0.067 T 10 mW/cm3 0.23 W
0.15 T. So, the minimum number of turns required to satisfy this limitation is
found with (4.40).
n1 =
(
Vout ⋅ 1 − D( min ) ) =
2 ⋅ (1 + n ) ⋅ fs ⋅ Ac ⋅ 2 ⋅ ∆B
(4.40)
400 ⋅ (1 − 0.78 )
= = 4 turns
2 ⋅ (1 + 2 ) ⋅ 50e3 ⋅ 2.44e-4 ⋅ 2 ⋅ 0.15
The second limitation is the maximum allowed power loss in the transformer.
Assuming that the core loss equals the copper loss in the maximum efficiency point
it allows only 0.625 W of the total core loss, i.e. the maximum core loss per volume
unit is 28 mW/cm3 (at the minimum input voltage of 30 V). For different number
of turns in primary winding the flux density amplitude and the core loss are
calculated. In the first approximation Steinmetz equation is used. Results are
summarized in Table 4.7. It’s found that 7, 8 and 9 turns fulfill the allowed core
loss limitation. However lower number of turns will benefit in lower winding
resistance and thus lower copper losses at high power / low input voltage end. For
future calculations number of the primary winding turns is 6.
The actual core loss is corrected using modified Steinmetz equation. First, the
lowest input voltage case is analyzed. The flux density amplitude is given by
(4.41) and its waveform is presented on Figure 4.15. The equivalent frequency and
the resulting core loss are found ((4.42) and (4.43)). In the same way the highest
input voltage case is analyzed and the calculated core loss per volume unit is about
150 mW/cm3.
∆B =
(
Vout ⋅ 1 − D( max ) ) = 0.1 T (4.41)
2 ⋅ (1 + n ) ⋅ fs ⋅ Ac ⋅ 2 ⋅ n1
75
Figure 4.15 The transformer flux density waveform
2
2 K  B − Bk −1  1
f eq = 2 ⋅ ∑  k  ⋅ = 90 kHz (4.42)
π k = 2  2 ⋅ ∆B  tk − tk −1
(
pFe = k ⋅ f eqα -1 ⋅ ( ∆B ⋅10 ) ⋅ f r =
β
) (4.43)
(
= 0.158 ⋅ 901.36−1 ⋅ ( 0.1 ⋅10 )
2.86
) ⋅ 50 = 39 mW/cm3
With turns ratio 1:1:2 it gives 12 turns in the secondary winding. For the given
rated winding currents (Table 4.5) the minimum cross section areas of winding
conductors are found ((4.44) and (4.45)). The minimum window area is calculated
with (4.46) and the considered core has the window sufficiently large to
accommodate windings.
I n1( rms ) 9.52
An1 = = = 2.38 mm 2 (4.44)
J 4
I n 2( rms ) 3.89
An 2 = = = 0.97 mm 2 (4.45)
J 4
nk ⋅ Ank
Wa ( min ) = ∑ = 1.34 cm 2 (4.46)
k kCu
The next step is to arrange winding on the core. The primary currents are
composed from dc and ac components (see paragraph 4.1.4), which makes the
windings design a challenge. Use of a single solid wire may result in a low dc
resistance, but because of ac effects (see paragraph 3.5.1) the ac resistance will be
large and the ac copper loss may become unacceptably high. In high frequency
transformers Litz wires are used to reduce ac resistance and thus related losses.
However Litz wires have a poor copper fill factor and the winding dc resistance
increase in compare with a solid wire. Keeping a low dc resistance in a push-pull
transformer is important too. A foil winding may be a good alternative. It may
provide a good copper fill factor (depends on foil and isolation thickness) and the dc
resistance is low. Moreover, if the foil winding is built-up from thin layers (in
compare with the penetration depth) then the ac loss may stay low. However foil
windings may have a significant interwinding capacitance because of a winding
large area facing the other winding. To keep this capacitance low an isolation
material with a low permittivity. Also a distance between windings is important.
76
Figure 4.16 Possible winding arrangements of a push-pull transformer

on a EE core using solid wires (a and b) and foil (c and d)
Figure 4.16 presents different winding arrangements of a push-pull transformer.

The arrangement a) has both primary winding are stacked one on the top of the
other. The secondary winding in placed in the top of primary windings. Since the
transformer current component see primary windings as series connected this simple
winding arrangement has a large ac loss, because of primary windings multilayer
placement. Using interleaving technique the secondary winding may be located
between primary windings (Figure 4.16 b). It reduces number of primary winding
layers, thus the ac copper loss is reduced. Such a winding arrangement can be
implemented using a copper foil. Depends on a layer thickness two solutions for
the secondary winding are possible (Figure 4.16 c and d). In both cases the turns
ration is the integer number. The winding arrangement d) has an advantage of less
isolation layers in compare with the arrangement c). The winding arrangement d)
is used in the design.
The effective window width and height are taken from Table 4.6. The primary
winding is made out of 24 mm width foil. To satisfy the required wire cross
section area (4.44) the minimum foil thickness is 0.1 mm (0.15 mm is available and
used). The secondary winding is made out of 11 mm wide foil and it’s minimum
required thickness is 0.09 mm (0.15 mm is available and used). The transformer
operates at 50 kHz and the penetration depth at this frequency is about 0.34 mm.
With the layer thickness of 0.15 mm and interleaving a very low ac resistance seen
by the transformer current component is expected. Including the isolation of 0.1
mm the total winding height is about 6 mm and it fits the available window size.
Figure 4.17 presents the simulation result of the transformer 2D FEMM model.
The current with known amplitude and frequency is applied to windings and copper
losses are calculated in each layer. Next, the resistance (dc or ac) is calculated and
extrapolated from 2D model to 3D.
77
Figure 4.17 The push-pull transformer FEMM model

Table 4.8 The transformer resistances and power losses
Calculated primary winding resistance (dc / ac) 2.7 / 2.9 mΩ
Calculated secondary winding resistance (dc / ac) 11 / 12.5 mΩ
Calculated primary winding copper loss at 250 W (dc / ac) 0.05 / 0.02 W
Calculated secondary winding copper loss at 250 W (ac) 0.05 W
Calculated core loss 0.7W
Table 4.8 summarized the calculated winding resistances and power losses in the
transformer operating at the output power of 250 W and the input voltage of 30 V.
The estimated winding losses are about 0.21 W and the core loss is about 0.7 W.
The total calculated transformer loss is below assumed limit. However, the core
loss calculation is valid for the core temperature of 80 ºC and since the transformer
will operate well below this temperature the actual core loss may increase
(according to datasheet plots). Since temperature coefficients ct, ct1 and ct2 (see
paragraph 3.5.2) are not known the temperature influence is not calculated. Also,
since there is window are available it’s possible to increase number of turns (e.g. up
to 8 primary winding turns) in order to reduce the core loss by the price of a larger
winding resistance.
4.2.5 Inductor design

The design input parameters are summarized in Table 4.9. The core material is
selected first. For dc inductors powder cores are used often. However they
exhibit a non-negligible core loss at higher frequencies, i.e. KoolMu material at 100
kHz and 50 mT has the core loss of about 200 mW/cm3. Ferrite core made out of P
78
3
10
2
10
R
F
1
10
0
10
2 4 6 8 10
R [ohm] −3
dc x 10
Figure 4.18 Allowed ac to dc resistance factor FR as a function
of the dc resistance of the inductor winding
material [84] is used instead of a powder core. The dissipated power is one of the
most important constraints in the design. To reach the efficiency goal the allowed
power dissipation in the inductor is about 0.75 W at 250 W output power and the
lowest input voltage. To fulfill this requirement both – dc and ac resistance have to
stay below the certain limit. Figure 4.18 presents the maximum allowed ac to dc
resistance factor FR as a function of the actual dc resistance of the inductor winding.
The required inductance L1 is found for the minimum input voltage (i.e.
maximum duty cycle) and the rated input current (4.47). The design bases on a
gapped ferrite core and it’s assumed that the inductor operates in its linear region.
So, the flux density is proportional to the inductor current, thus the dc and ac flux
density components are found ((4.48) and (4.49)). The area product method links
the core dimensions (window area and core cross section area) with the inductor
electric and magnetic parameters (like the energy stored in the air gap, the current
density or the peak flux density). For given input parameters the required area
product is calculated (4.50) and it’s found that core size equivalent to E42/21/15 [84,
85] is large enough. It’s dimensions and parameters are summarized in Table 4.10.
L1 =
(
Vin ( min ) ⋅ D( max ) − 0.5 ) = 30 ⋅ ( 0.78 − 0.5) = 48.3 µΗ (4.47)
2 ⋅ ∆iL1 ⋅ fs 2 ⋅1.74 ⋅ 50e3
I L1 17.4
B = B( max ) ⋅ = 0.35 ⋅ = 0.32 T (4.48)
I L1( max ) 19.1
∆B = B( max ) − B = 0.35 − 0.32 = 0.03 T (4.49)
L1 ⋅ I ⋅ I ( max ) 48.3e-6 ⋅17.4 ⋅19.1

AP = = = 2.87 cm 4 (4.50)
kCu ⋅ B( max ) ⋅ J 0.4 ⋅ 0.35 ⋅ 4e6
79
Table 4.9 The inductor design input data

Rated output power Pout 500 W
Switching frequency fs 50 kHz
Turns ratio 1:1:2
Input current ripple at rated current Iin(pp) 20%
Input current IL1 (at 250 W and 30 V) 8.5 A
Input currents IL1 (at 250 W and 30 V) 17.4 A
Table 4.10 Dimensions of EE42/21/15 core

Area product (maximum / effective) AP 4.74 / 3.17 cm4
Core cross section Ac 1.85 cm2
Window dimensions (maximum) 29.6×8.65 mm
Window dimensions (effective) 26.2×6.79 mm
Core volume 17.3 cm3
Center leg dimensions a×b 12.2×15.2 mm
Then the number of turns and the air gap length are found ((4.51) and (4.52)).
In the air gap the flux is spread over a larger area, thus the effective air gap area is
larger than the core cross section area and it is given by (4.53). Once the effective
air gap area is known, the number of turns is corrected (4.54).
L1 ⋅ ∆iL1 48.3e-6 ⋅1.74
n= = = 15.1 ≈ 15 turns (4.51)
∆B ⋅ Ag 0.03 ⋅1.85e-4
µ ⋅ n 2 ⋅ Ag 4e-7 ⋅ π ⋅152 ⋅1.85e-4

lg = = = 1.1 mm (4.52)
L1 48.3e-6
Ag ' = ( a + lg ) ⋅ ( b + lg ) = 2.03 cm 2 (4.53)
L1 ⋅ lg 48.3e-6 ⋅1.1e-3
n' = = = 14.4 ≈ 14 turns (4.54)
µ ⋅ Ag ' 4e-7 ⋅ π ⋅ 2.03e-4
The number of turns and the length of the air gap are known now. In following
steps the minimum wire cross section area is found (4.55) based on the inductor rms
current. Then the minimum required window area is calculated (4.56). The
selected core provides a significantly larger window area.
I L1( rms ) 17.43
AL1 = = = 4.36 mm 2 (4.55)
J 4e6
80
Figure 4.19 FEMM model of the inductor,

the winding made out of ∅2.4mm wire
n ⋅ AL1 14 ⋅ 4.36e-6
Wa ( min ) = = = 1.53 cm 2 (4.56)
kCu 0.4
Using modified Steinmetz equation and the method presented in section 3.5.2 the
core loss is estimated for the minimum input voltage and the maximum duty cycle.
The repetition frequency of the ripple is twice the switching frequency, i.e. 100 kHz.
The equivalent frequency is found with (4.57), and it’s 82 kHz. The estimated core
loss per volume unit is about 0.5 mW/cm3 (4.58). This calculation doesn’t include
the core temperature and the dc premagnetization effects, thus the actual core loss
may increase. However even if the core loss increases several times of the base
value the total core loss will stay low and most likely negligible.
2
2 K  B − Bk −1  1
f eq = 2 ⋅ ∑  k  ⋅ = 82 kHz (4.57)
π k = 2  2 ⋅ ∆B  tk − tk −1
(
pFe = k ⋅ f eqα -1 ⋅ ( ∆B ⋅10 ) ⋅ f r =
β
) (4.58)
(
= 0.158 ⋅ 82 1.36 −1
⋅ ( 0.015 ⋅10 )
2.86
) ⋅100 ≈ 0.5 mW/cm 3
Now, the winding arrangement is designed. Since the input current has a great
dc component and fairly small ac ripple, the focus is on a low dc resistance. The
first choice is a solid round wire and the required wire diameter is ∅2.4 mm.
Figure 4.19 presents the winding made out of a single wire arranged into two
identical layers. The major drawbacks of this solution includes the difficult
winding due to the wire size, partial utilization of the window width, significant ac
81

the winding made out of two paralleled ∅1.7mm wires

the winding made out of 11×0.4 mm foil
loss due to the layer thickness (using Dowell’s method the ac to dc resistance factor
is about 30).
Use of a several parallel wires, improves the window width utilization and
reduces height of the layer in the same time. As a result the ac copper loss may
decrease (the ac to dc resistance factor is about 21). Finally, it’s easier to wind few
82

the winding made out of 24×0.2 mm foil
Table 4.11 Calculated copper losses and winding resistances of the inductor
2× 24×0.2 mm
Wire size / type ∅2.4 mm 11×0.4 mm 24×0.2 mm
∅1.7mm (notch)
Dc resistance 4.7 mΩ 4.3 mΩ 4.4 mΩ 4.2 mΩ 4.3 mΩ
Ac resistance 0.43 Ω 0.64 Ω 0.3 Ω 0.25 Ω 0.19 Ω
Ac to dc resistance factor 92 149 68 60 44
Dc copper loss (at 250 W) 0.34 W 0.31 W 0.32 W 0.30 W 0.31 W
Dc copper loss (at 500 W) 1.43 W 1.30 W 1.33 W 1.26 W 1.31 W
Ac copper loss (input voltage 0.43 W 0.65 W 0.30 W 0.25 W 0.19 W
dependent)
thinner wires then a single thick one. Example of such a winding arrangement
using two parallel ∅1.7 mm wires is presented on Figure 4.20.
Add in of parallel wires and reduction of the wire diameter will eventually lead
to a foil-like winding, composed of many thin wires placed side by side. Use of a
foil winding for a dc inductor may provide a good copper fill factor (depends on a
foil wire / isolation thickness ratio) and thus dc copper loss will stay low. Two
alternative designs using a foil winding are considered. The first design has 7
layers and 2 turns per layer. Its FEMM model is presented on Figure 4.21. The
foil size is 11×0.4 mm and it fits the copper cross section are required by (4.55).
Using Dowell’s method it’s found that the ac to dc resistance factor for this winding
configuration is about 37, so the ac loss are relatively high.
83
Figure 4.23 Simple rectangular winding notch reducing ac copper loss
The other design has 14 layers, 1 turn per layer and its FEMM model is
presented on Figure 4.22. The foil size is 24×0.2 mm and the copper cross section
area is larger then required by (4.55). The penetration depth at 100 kHz is about
0.24 mm and the relative penetration depth is 0.8. Using Dowell’s method it’s
found that ac to dc resistance factor FR for such configuration is about 10.
Dowell’s method used for a preliminary estimation of ac copper loss doesn’t take
into account any effect of an air gap or an edge effect and the ac copper loss may be
greatly underestimated. FEMM models (shown on figures above) can predict
copper losses accurately and results are summarized in Table 4.11. Both winding
arrangements incorporating solid round wire doesn’t exceed the allowed loss limit of
0.75 W. The copper loss of foil windings are below the limit and the arrangement
presented on Figure 4.22 provides the lowest total copper loss. Further
improvement and the loss reduction is possible. Since the winding doesn’t occupy
whole window it’s possible to increase the foil thickness, e.g. up to 0.3 mm and thus
reduce the dc copper loss. On Figure 4.22 one may observer a red area in the
winding placed close to the air gap. In this part of the winding significant eddy
currents are induced by the fringing field around the gap. The method proposed in
[56, 61] enables reduction of eddy currents induced by the fringing field simply by
removal part of the winding close to the gap. The authors discuss different shapes
and sizes of notches and their impact on ac and dc copper losses. However, even a
simple, rectangular winding notch presented on Figure 4.23 reduces ac loss by a
small cost of increased dc resistance due to a reduced copper cross section area.
It’s particularly helpful in case of larger air gaps or foil windings placed very close
to the gap. In case of the winding from Figure 4.22 the winding is placed on a coil
former and it’s already relatively distant from the gap. The rectangular notch in
fact reduces ac resistance, but in the same time the dc resistance increases. The
total copper loss however is reduced by about 10% in compare with a full window
width foil winding.
84
Figure 4.24 Magnetic circuits of gapped inductor (a) and push pull transformer (b)
4.2.6 Integrated inductor-transformer

The concept of integration of the inductor and the transformer on a single core
was previously presented in [86] for an isolated Cuk converter and in [87] for an
isolated push-pull converter with ripple-free input current.
The input voltage varies in a wide range and the ratio between the minimum and
the maximum voltage is 1:2. At the rated power and the lowest input voltage the
inductor current reaches its maximum value. The flux density in the inductor core
is current dependant and the inductor is designed to sustain the current without core
saturation. Contrary to this the flux density amplitude in the transformer core
depends on the input voltage and reaches it’s maximum at the highest input voltage
(and the lowest duty cycle, see Figure 4.8). Moreover, many high frequency
transformers are so called thermal limited designs, which means that the maximum
flux density amplitude is limited by allowed core loss rather than by saturation level.
It leaves some space for dc flux. In this conditions, at the highest input voltage the
inductor current may reach about half of the maximum current. So, in terms of
flux density only one core at the time is fully utilized, while the other is utilized
partially only. The same is true for other current fed dc-dc converters, like full
bridge or two inductor boost converter. Moreover, the push-pull boost converter
(isolated and non-isolated) suffers an additional copper loss due to the input current
flowing thru primary windings. A properly designed integrated inductor-
transformer may improve utilization of the magnetic core and windings.
The challenge is to place all windings on a single core in such way that they will
not disturb each other, i.e. the ac flux generated by the transformer winding shall not
induce any voltage in the inductor winding and vice versa.
Figure 4.24 presents the equivalent circuit of the inductor magnetic circuit.
Magneto motive force (MMF) is represented by the voltage source nL1·iL1(t). The
reluctance of the center leg including the air gap reluctance is represented by the
linear resistor Rg. Resistor Ro indicates the outer leg reluctance, but since it’s very
small in compare with the air gap reluctance it can be neglected.
Figure 4.24 presents the equivalent circuit of the push-pull transformer magnetic
circuit arranged on an EE core. All windings are placed on the center leg of the
core. MMFs generated by windings are represented by three voltage sources which
85
Figure 4.25 Magnetic circuit of integrated inductor-transformer
Figure 4.26 Reduced magnetic circuit

of integrated inductor-transformer
directions refer to the converter diagram shown on Figure 4.1. Resistors Ro

represent reluctance of outer legs while the resistor Rc represents reluctance of the
center leg without any air gap.
These two magnetic circuits are arranged on a single gapped EE core and Figure
4.25 presents the electrical equivalent circuit. The inductor winding is placed on
the center gapped leg. Transformer windings are split into two identical halves and
placed on outer legs. In this way there is no undesired interaction between the
transformer and the inductor. Fluxes related by the inductor and the transformer
are present in outer legs, which in spite of discussion above improves overall
utilization of the magnetic core. Unfortunately such winding arrangement is
complex. It has a large total number of turns, so copper losses may be high.
Another solution is needed.
The inductor current splits into two halves when it flows into transformer
primary windings. The flux generated by the inductor winding also splits into two
halves once it goes into outer legs. So, there is a good agreement between the
inductor current, primary current components and related fluxes and it’s possible to
use transformer primary windings as a part of the inductor winding. For proper
operation of the transformer the secondary winding is split into two symmetrical
halves (n2a and n2b) and arranged on both outer legs, so it’s not affected by inductor
flux. The equivalent circuit of the proposed winding arrangement is presented on
Figure 4.26. The proposed integrated inductor-transformer is described by (4.59),
where Lk is self inductance of nk winding, Mji is mutual inductance between ni and nj
86
windings and so on. Self inductances are defined by (4.60) while mutual
inductances are given by (4.61) in general case.
M L1n 2b   dt 
diL1
 vL1   L1 M L1n1 M L1n11 − M L1n 2a
v    di 
 n1   M n1L1 Ln1 − M n1n11 − M n1n 2a − M n1n 2b   dtn1 
 vn11  =  M n11L1 − M n11n1 Ln11 M n11n 2a M n11n 2b  ⋅  didtn11  (4.59)
   
 vn 2a   − M n 2aL1 − M n 2an1 M n 2an11 Ln 2a M n 2an 2b   din 2 
 v   M n 2bL1  dt 
 n 2b  − M n 2bn1 M n 2bn11 M n 2bn 2a Ln 2b   din 2 
 dt 
nk2
Lk = (4.60)
Rk
ni ⋅ n j
M ij = kij (4.61)
Rij
Since outer legs and windings arranged on the core are symmetrical its possible
to simplify the matrix in (4.59) because some of mutual inductances are identical
(assuming no leakage flux). Reluctance of the center leg, including the air gap, is
given by (4.62) and (4.63) gives reluctance of the outer leg. Now respective self
and mutual inductances are given by (4.64), (4.65) and (4.66).
lc l
Rg = + g (4.62)
µ0 ⋅µ r ⋅ Ac µ 0 ⋅ Ag
lo
Ro = (4.63)
µ 0 ⋅µ r ⋅ Ao
n2
L1 =
Rc + R2o
n12
Ln1 = Ln11 = (4.64)
Ro + Rg Ro
2
n2a
Ln 2a = Ln 2b =
Ro + Rg Ro
87
n ⋅ n1
M L1n1 = M L1n11 = M n1L1 = M n11L1 = k L1n1 ⋅
Rg + Ro Ro
n ⋅ n2a
M L1n 2a = M L1n 2b = M n 2aL1 = M n 2bL1 = k L1n 2a ⋅
Rg + Ro Ro
n1 ⋅ n11
M n1n11 = M n11n1 = kn1n11 ⋅
Ro + Ro Rg
(4.65)
n1 ⋅ n2a
M n1n 2a = M n11n 2b = M n 2an1 = M n 2bn11 = kn1n 2a ⋅
Ro + Ro Rg
n1 ⋅ n2b
M n1n 2b = M n11n 2a = M n 2bn1 = M n 2an11 = kn1n 2b ⋅
Ro + Ro Rg
n2a ⋅ n2b
M n 2an 2b = M n 2bn 2a = kn 2an 2b ⋅
Ro + Ro Rg
Ro Ro
k L1n1 =
Ro
Ro Ro
k L1n 2a =
Ro
Ro Rg
kn1n11 =
Ro (4.66)
kn1n 2a ≈ 1
Ro Rg
kn1n 2b =
Ro
Ro Rg
kn 2an 2b =
Ro
Now, so called equivalent inductance L1eq has to be found. During the inductor
charging periods (stages 1 and 3, see paragraph 4.1) the input voltage is applied to
the inductor winding and parallel connected primary windings of the transformer.
In a traditional push-pull transformer primary windings are shorted during this
periods and there is no voltage across them (except voltage drop due to wire
resistance and leakage inductance), so the whole input voltage is applied to the input
inductor. However, in the integrated magnetic device presented on Figure 4.26
primary windings contribute to energy storage and the voltage across them is non-
zero. So, (4.67) is truth and L1eq is unknown equivalent inductance. (4.68) and
(4.69) are substituted into (4.67). During charging periods there is no current in the
secondary winding so (4.70) is truth. Also (4.71) and (4.72) are truth during these
periods, so (4.73) gives a solution for the equivalent inductance. It’s important to
note that the physical inductance of the inductor L1 may be significantly lower then
the equivalent inductance L1eq seen by the converter.
88
10
AP [cm4] 6
4 APint
AP
int ind
2
AP
int tr
0
0.1 0.15 0.2 0.25 0.3 0.35 0.4
B [T]
ind
Figure 4.27 Integrated inductor-transformer area product

vs. inductor flux density
diin
Vin = vL1 + vn1 = L1eq ⋅ (4.67)
dt
diin di di di di
vL1 = L1 ⋅ + M L1n1 ⋅ n1 + M L1n11 ⋅ n11 − M L1n 2a ⋅ n 2 + M L1n 2b ⋅ n 2 (4.68)
dt dt dt dt dt
diin di di di di
vn1 = M n1L1 ⋅ + Ln1 ⋅ n1 − M n1n11 ⋅ n11 − M n1n 2a ⋅ n 2 − M n1n 2b ⋅ n 2 (4.69)
dt dt dt dt dt
din 2
=0 (4.70)
dt
din1 din11
= (4.71)
dt dt
diin din1 din11
= + (4.72)
dt dt dt
L1eq = L1 + M L1n1 + M n1L1 (4.73)
The area product method is used to determine the core size required by the
integrated inductor-transformer. For this calculation values from paragraphs 4.2.4
and 4.2.5 are taken. Because two magnetic devices share the same core calculation
of the core size is more complex. The simplest way is to add area products
required by the transformer and by the inductor (4.74). Since the inductor and the
transformer fluxes share the same core the total flux density shall stay below
saturation (4.75).
Pt L1 ⋅ I ⋅ I ( max )
AP = + (4.74)
kCu ⋅ k f ⋅∆Btr ⋅ J ⋅ fs kCu ⋅ ( Bind + ∆Bind ) ⋅ J
89
10
AP [cm ] 6
4
4 AP
int
APint ind
2
AP
int tr
0
0.1 0.15 0.2 0.25 0.3 0.35 0.4
B [T]
ind
Figure 4.28 Integrated inductor-transformer area product

vs. inductor flux density, Bsat=0.4 T
∆Btr + ( Bind + ∆Bind ) < Bsat (4.75)

By adjusting ratio between the transformer flux density ∆Btr and the inductor flux
density (Bind+∆Bind) it’s possible to find the minimum required area product of the
core, like presented on Figure 4.27. The minimum required area product found in
this way equals 7.95 cm4 and it’s significantly larger then the total area product
required by separated inductor and transformer (6.80 cm4). Such a high value is a
result of a lower copper fill factor of particular windings and a lower maximum flux
density of the inductor.
This simple calculation presented above deals with the transformer and the
inductor area products separately and doesn’t include some important aspects.
Since the transformer flux density amplitude is well below saturation it’s possible to
inject some significant dc flux without increasing the core size (some additional
window area is required however). First the area product required by the
transformer only is calculated (4.76). Next, rearrangement of (4.50) enables
calculation of how much energy it’s possible to store in a given core using only this
‘free’ flux density space (4.77).
Pt
APtr = (4.76)
kCu ⋅ k f ⋅∆Btr ⋅ J ⋅ fs
APtr
L1 ⋅ I ⋅ I ( max ) = (4.77)
kCu ⋅ ( Bsat − ∆Btr ) ⋅ J
The required area product of the core as a function of the inductor flux density is
presented on Figure 4.28. For the transformer flux density amplitude below 0.09 T
the used core is sufficiently large to handle the inductor. Once the transformer flux
density amplitude increases there is not enough ‘free’ room for the inductor and a
bigger core is required. According to this calculation the minimum area product
90
Figure 4.29 The integrated magnetic number of turns and air gap
length calculation flowchart
equals 4.33 cm4. The core E42/21/20, the one used for the push-pull transformer
(see paragraph 4.2.4) is large enough. The optimum inductor flux is about 0.31 T.
With assumed saturation level of 0.4 T it leaves about 0.09 T for the maximum
amplitude of the transformer flux density. According to calculation from
paragraph 4.2.4 the worst case (at the highest input voltage) transformer flux density
amplitude is about 0.15 T. However, one shall remember that the dc
premagnetization may increase core loss significantly (see paragraph 3.5.2).
Once the core size and core type are know number of turns in particular winding
as well the air gap length are calculated. First, the transformer number of turns is
found using (4.78). It’s important that transformer windings are placed on outer
legs, so the effective core cross section area is different from the value given in a
standard datasheet.
Vout ⋅ (1 − D )
n1 = (4.78)
2 ⋅ (1 + n ) ⋅ f s ⋅ Ac ⋅ 2 ⋅ ∆B
Next, the inductor number of turns and the length of the air gap are calculated in
the iterative process, which is shown on Figure 4.29. Starting from a single turn in
91
Figure 4.30 FEMM model of the integrated inductor-transformer
the inductor winding the required equivalent inductance is calculated. Successive,

ascending air gap length values are substituted into (4.62) and resulting equivalent
inductance is calculated with (4.73). This loop is continued until the required
equivalent inductance is found. Then the flux densities related to the inductor and
the transformer are calculated for different operating conditions. The transformer
flux density amplitude is calculated by rearranging (4.78) for a given number of
turns. The inductor related flux is found by analysis of the equivalent circuit from
Figure 4.26 and it’s given by (4.79). From this one can find flux density increase
in the center and outer legs due to the inductor.
nL1 + 12 ⋅ n1
Φ c = iin ⋅ (4.79)
R g + 12 ⋅ Ro
If the total flux density exceeds the saturation level at any operating point one
additional turn is added to the inductor winding and next iteration starts.
Using this method it’s found that the transformer primary windings have 13 turns
each, the secondary winding has 26 turns (split into two symmetrical halves) and the
inductor winding has 10 turns. The air gap length is 1.8 mm. Required wire cross
section area doesn’t change in compare with previous calculations (4.44), (4.45) and
(4.55) that is 4.36 mm2 for the inductor winding, 2.38 mm2 for the primary winding
and 0.97 mm2 for the secondary winding. Use of a foil winding become somehow
difficult, because of increased number of turns and thus increased interwinding
capacitance.
Instead of a foil paralleled round wires are used as follow: inductor winding 3×
∅1.5 mm, primary windings 3× ∅1.0 mm and the secondary winding 2× ∅0.8 mm.
The proposed winding arrangement is presented on Figure 4.30.
92
Table 4.12 The integrated inductor-transformer resistances and losses

Calculated inductor winding resistance (dc / ac ripple) 2.8 / 179 mΩ
Calculated primary winding resistance (dc / ac / ac ripple) 6.9 / 17.2 / 100 mΩ
Calculated secondary winding resistance (dc / ac) 32.7 / 37.5 mΩ
Calculated inductor winding copper loss at 250W (dc / ac ripple) 0.20 / 0.18 W
Calculated primary winding copper loss at 250 W (dc / ac / ac ripple) 0.13 / 0.06 / 0.01 W
Calculated secondary winding copper loss at 250 W (ac) 0.14 W
Calculated core loss 0.29 W
It’s important to note that the transformer related flux, the major source of the
core loss, flows only in outer legs, so only outer legs contribute to the total core loss.
The core loss is calculated using modified Steinmetz equation (see paragraph 3.5.2).
Winding resistances are found using FEMM software and they are summarized
in Table 4.12. Also, losses present in the integrated magnetic device are shown in
the same table. The total calculated loss of the device operating at 250 W is about
1.2 W. For comparison, the total loss of the push-pull transformer and the dc
inductor operating in the same conditions is about 1.14 W.
4.2.7 Capacitors
Output capacitors of the converter are designed to filter high ripples rather then
low frequency ripples coming from an inverter. The allowed output voltage peak-
to-peak ripple is 5% of the output voltage, i.e. 20 V peak-to-peak. It’s assumed
that the voltage ripple are distributed among output capacitors according to EQ.
VCj ( pp ) VCj
= j =1,2,3 (4.80)
Vout ( pp ) Vout
According to the converter analysis from paragraph 4.1 the capacitor C1 is
recharged during stages 2 and 4, while it discharges during stages 1 and 3, so
effectively it’s recharged twice per the switching period. The required capacitance
is given by (4.81).
I out ⋅ ( D − 0.5 ) 1.25 ⋅ ( 0.78 − 0.5 )
C1 = = = 1.05 µF (4.81)
Vout ( pp ) ⋅ VC1 ⋅ fs 5% ⋅133.3 ⋅ 50e3
The capacitor C2 is recharged during stage 4 only and discharges during stages 1,
2 and 3. The capacitor C3 operates symmetrically to the capacitor C2 and it’s
charged during stage 2 and it discharges during stages 1, 3 and 4. The required
capacitance is given by (4.82).
I out ⋅ D 1.25 ⋅ 0.78
C2 = C3 = = = 2.93 µF (4.82)
Vout ( pp ) ⋅ VC 2 ⋅ fs 5% ⋅133.3 ⋅ 50e3
93
Table 4.13 The loss budget of the 500W converter,

output power 250 W, input voltage 30 V
Transistor losses (conduction / switching / gate) 2× 0.32 / 0.32 / 0.07W
Diode losses (low voltage diode / high voltage diode) 2× 0.21 / 0.79W
Transformer losses (copper / core) 0.21 / 0.7 W
Inductor losses (copper dc / ac) 0.31 / 0.19 W
Integrated inductor-transformer (copper inductor / transformer / core) 0.38 / 0.54 / 0.29 W
Other losses (interconnections, capacitors, etc.) 0.5 W
Total loss (with separated / integrated magnetics) 5.11 / 5.13 W
Capacitors are preselected based on their capacitance and their rated dc voltage.
Then parts with the lowest ESR are selected – C1 is 2.2 µF 160 V MKP1839, C2 and
C3 are 4.7 µF 160 V MKP1839. The capacitor current is found as a difference
between the neighboring diode rms current the output dc current. Estimated
capacitor loss due to ESR is negligible.
4.3 Summary
In Chapter 4 the non-isolated push-pull-boost converter detailed analysis has
been made fist and important equations describing the converter has been developed.
Next, a numerical example of the converter design has been presented step by step.
First, based on the assumed peak efficiency relative loss limits are set. Then,
transistors are selected among available state-of-the-art devices and converter
parameters (the output power and the switching frequency) are adjusted in the way
that transistors are used in optimum way. In following paragraphs other
components are selected (diodes, capacitors) or designed (magnetics). The design
of the transformer and the inductor bases on a single pass calculations and the area
product method for the core size estimation. Winding resistances of a final design
are calculated with FEMM software for better accuracy. The final part of the
chapter deals with the design of the integrated inductor-transformer for the
converter. It’s demonstrated that the properly designed integrated magnetic device
improves utilization of the magnetic core and winding. As a result the size of
magnetic components is significantly reduced, while the calculated losses remains
on a comparable level (see Table 4.8, Table 4.11and Table 4.12).
Finally, the estimated losses of the 500 W converter operating at 250 W output
power and 30 V input voltage are summarized in Table 4.13. In both cases (with
separated magnetic devices and the integrated one) estimated losses are just about
the assumed limit of 5 W. This calculation indicates that the target efficiency is
achievable for this converter topology. The calculation however includes only
losses in the power circuit of the converter and gate drive loss. It doesn’t take into
account losses associated with controller and sensor circuitry.
95
Chapter 5
Modular Converter
A modular converter is not a new concept. This kind of converters are used in
telecomm, database and server centers, aerospace and space applications.
Modularity provides several well known advantages for both, a system designer and
a system user. It includes a system flexibility and scalability, possible cost
reduction of installation and maintenance, design standardization and improved
reliability due to redundancy. However a simultaneous operation of many
converter modules and their interactions creates challenges. The most recognized
and pronounced are current/voltage sharing among modules [38, 88-91] and
potential system instability due to multiple control loops [92-94].
In this Chapter a general overview on a converter scaling and paralleling is
given. Then, different interconnections of converter modules are presented.
Finally, benefit of a modular step-up converter in a fuel cell application is
demonstrated. It includes a theoretical discussion and practical verification.
Also, the idea can be extended for other applications.
5.1 Scaling of converters

The rated output power of a converter is Pout(max) and it’s peak efficiency ηmax
occurs at about half of the rated output power Pout(max)/2. The converter is scaled up
in power and now the rated power is twice larger Pout(max)’=2·Pout(max). Efficiency
curve shall has the same shape and it peaks at Pout(max)’/2=Pout(max). Since the peak
efficiency doesn’t change the allowed losses are twice larger now, but the loss
distribution shall not change in the scaled converter. How does the scaling process
influences particular components?
5.1.1 Transistor
First, transistor losses are considered. At the efficiency peak point the
conduction loss equals the switching loss, as explained in paragraph 4.2.1. Once
the converter is scaled in power the transistor current and allowed losses increase
proportionally (PTcond’=2·PTcond, PTsw’=2·PTsw). Since the conduction loss is
proportional to the second power of the current the on-state resistance has to
decrease by about half. It's simply done by selection of a larger transistor or by
paralleling several devices. Paralleling two or three transistors may be relatively
easy, but putting more of them in parallel become a challenge. One of difficulties
is to ensure proper layout and equal stray inductances in all current paths. Small
96
0.5
limit PTcond
1.2
P 0.4
1 Tcond
PTcond [W]
[W]
0.8 0.3
Tsw
0.6 0.2
P
limit PTsw
0.4
0.1 PTsw
0.2
0 0
200 400 600 800 1000 2 4 6 8 10
P [W] f [Hz] 4
out s x 10
Figure 5.1 Conduction (left) and switching (right) losses of IRFP4321PbF

2.5 3
limit P limit P
Tcond 2.5 Tsw
2 P P
Tcond Tsw
2
PTcond [W]
[W]
1.5
1.5
Tsw
1
P
1
0.5 0.5
0 0
0 500 1000 1500 2000 2 4 6 8 10
P [W] f [Hz] 4
out s x 10
Figure 5.2 Conduction (left) and switching (right) losses of IRFP4568PbF
differences in stray inductances may lead to unexpected voltage oscillations across

paralleled MOSFETs [95]. Also dynamic current sharing may become a problem
[71]. It may be necessary to reduce switching speed and switching loss may
increase. To keep switching loss at acceptable level the switching frequency is
reduced. Sometimes however a device having larger current handling capability is
available, so there is no need for devices paralleling. Unfortunately in most cases a
larger transistor has a worse switching performance.
Here two transistors are compared – IRFP4321PbF and IRFP4568PbF. Both
transistors have a similar value of FOMQG but different on-state resistance and
switching performance. Results of the comparison are presented on Figure 5.1 and
Figure 5.2. In the comparison the current rise/fall rate diT/dt is assumed to be
similar in both cases and it results in different gate resistances. The larger
transistor (IRFP4568PbF) is able to process more then twice larger power with the
same conduction loss in percentage terms. However keeping the switching loss on
the certain level requires a significant reduction of the switching frequency, even in
spite a lower gate resistance.
If the converter is being scaled up further it may be necessary to change used
technology – it applies to both, layout and transistors. At certain power level it’s
97
use of PCB is not possible anymore, thus busbars are used instead. It potentially
results in longer interconnections and it eventually may lead to the switching
frequency reduction. In the same time it may be necessary to change transistor
technology, e.g. IGBT instead of MOSFET. Certainly such change has an impact
on the optimum power level and the optimum switching frequency.
From this point two cases shall be considered – first, the switching frequency of
the scaled converter remains the same (fs’=fs) or second, the switching frequency
decreases (for sake of simplicity, by half (fs’=fs/2).
5.1.2 Inductor
Now, the boost inductor is being scaled up. Since the allowed current ripple
increases the required inductance changes according to (5.1).
V (
 in ( min )
)
⋅ D( max ) − 0.5 1
L1 ' =
( )
Vin ( min ) ⋅ D( max ) − 0.5 
=
2 ⋅ 2 ⋅ ∆iL1 ⋅ f s
= ⋅ L1
2
(5.1)
2 ⋅ ∆iL1 '⋅ fs ' (
 Vin ( min ) ⋅ D( max ) − 0.5

) = L1
 2 ⋅ 2 ⋅ ∆iL1 ⋅ f2s
The required core size is found with area product method. Assuming the same
maximum flux density, current density and copper fill factor it’s given by (5.2).
 12 L1 ⋅ 2 ⋅ I ⋅ 2 ⋅ I ( max )
 = 2 ⋅ AP
L1 '⋅ I '⋅ I ( max ) '  kCu ⋅ B( max ) ⋅ J
AP' = = (5.2)
kCu ⋅ B( max ) ⋅ J  L1 ⋅ 2 ⋅ I ⋅ 2 ⋅ I ( max )
 k ⋅B = 4 ⋅ AP
 Cu ( max ) ⋅ J
References [83, 96] provide core dimensions growth guidelines – for linear
dimensions (5.3), areas (5.4) and volumes (5.5). So, the core volume will increase
by factor of 1.68 if the switching frequency remains constant or by factor of 2.83 if
the switching frequency is reduced by half. Moreover it’s stated that the current
density shall decrease (5.6) because of a relatively smaller increase in the cooling
surface.
1
length = klength ⋅ AP 4 (5.3)
2
area = karea ⋅ AP 4 (5.4)
3
volume = k vol ⋅ AP 4 (5.5)
1
J = k J ⋅ AP 8 (5.6)
The scaled inductor, similar to the transistor, conducts twice larger current and
allowed copper loss is twice larger too. So the dc resistance of the inductor
winding, shall decrease by half. When the inductor is being scaled three
parameters change – the number of turns n (5.7), the mean length per turn MLT
(5.3) and the allowed copper cross section area ACu. To keep the current density on
98
the same level the copper cross section area is double and the resulting dc resistance
is given by (5.8). In the first case, the dc resistance of the scaled inductor decreases
by more than half, so expected dc copper loss increases slower than assumed. In
the second case the dc resistance decreases slower then expected, so it may be
necessary to increase the wire cross section area and thus reduce the current density.
As the result the required area product may increase further due to a lower current
density (5.2). Also, in the second case, the total copper area (wire cross section
multiplied by number of turns) growths faster then the window area does. It should
be take into account too.
 12 L1 ⋅ 2 ⋅ ∆iL1 −1
 2 = 2 2 ⋅n
L '⋅ ∆iL1 '  ∆B '⋅ 2 ⋅ Ag
4
n' = 1 = (5.7)
∆B '⋅ Ag '  L1 ⋅ 2 ⋅ ∆iL1
= 1⋅ n
 ∆B '⋅ 4 24 ⋅ Ag

 2− 2 ⋅ n ⋅ ρ ⋅ 2 4 ⋅ MLT Rdc
1 1
 = 11
n '⋅ ρ ⋅ MLT'  2 ⋅ ACu 24
Rdc ' = = (5.8)
ACu ' 1
 n ⋅ ρ ⋅ 4 4 ⋅ MLT Rdc
 = 1
 2 ⋅ ACu 22
Alternatively the required wire cross section area may be estimated for given dc
resistance, number of turns and MLT (5.9).
 2− 2 n ⋅ρ ⋅ 2 4 ⋅ MLT
1 1
3
 = 2 4
⋅ ACu
n '⋅ρ ⋅ MLT '  1
2 ⋅ R dc
ACu ' = = (5.9)
Rdc ' 1
 n ⋅ρ ⋅ 4 ⋅ MLT
4
3
 = 2 2 ⋅ ACu
2 ⋅ Rdc
1

Such increase of the wire size leads to reduction of relative penetration depth ϕ
(5.10) and to possible increase of the ac resistance. For a dc inductor conducting a
smooth current ac resistance may not be so important, but it has a great impact on
performance of a coupled inductor or a flyback transformer. So, for notation
simplicity a square wire is considered, but it’s valid for round wires as well. In the
first case the relative penetration depth increases, thus the skin effect become more
painful. On the other hand less turns may result in fewer layers and thus the higher
relative penetration depth could be compensated. In the second case reduction of
the switching frequency counteract the relative penetration depth which remains
constant. With the same number of turns and possibly the same number of layers
the ac to dc resistance factor will not change most likely.
 2 ⋅h 1
 = 22 ⋅ ϕ
h' ACu '  δ
ϕ' = = = (5.10)
δ' δ'  2 ⋅h = ϕ
 2 12 ⋅ δ
99
5.1.3 Transformer
Similar considerations are done for scaling of a high frequency transformer.
The allowed core and copper loss of the scaled transformer are assumed to be twice
larger – PFe’=2·PFe and PCu’=2·PCu. The area product formulation for a transformer
is given by (5.11). Contrary to the dc inductor the transformer design is thermal
limited most likely, thus reduction of the switching frequency enables increase of
the flux density amplitude.
Pt '
AP ' = =
kCu ⋅ k f ⋅∆B '⋅ J ⋅ fs '
 2 ⋅ Pt
 k ⋅ k ⋅∆B ⋅ J ⋅ f = 2 ⋅ AP (5.11)
 Cu f s
= − αβ
 2 ⋅ P '  f 
t
= 4 ⋅   ⋅ AP=2.83 ⋅ AP α = 1
s
( )
 k ⋅ k ⋅ fs β ⋅ ∆B ⋅ J ⋅ 1 ⋅ f
α
 fs '  β 2
 Cu f f s ' 2 s
Of course there is a certain limit on the total core loss. If the switching
frequency and the flux density are constant the core loss per volume unit remains the
same and the required area product is twice larger (5.11). So, the core volume
increases by factor of 1.68 according to (5.5) and the total core loss increases by the
same factor. Growth of the area product in the second scenario depends on
material properties. For α=1 and β=2 the area product will increase by factor of
2.83 and the core volume will increase by factor of 2.18. In the second case
reduction of the calculated flux density amplitude may be required in order to keep
the total core loss on the allowed level. The flux density amplitude, the core
volume and the total core loss are calculated iteratively.
However, the core loss isn’t the only one loss mechanism in a transformer.
Also copper losses, especially ac copper loss, is very important. Like it was
previously with the transistor and the inductor scaling the resistance of the
transformer winding has to decrease by half. Three variables decide about the dc
resistance – the number of turns, the wire cross section area and MLT. MLT
growth is given by (5.3) again. The number of turns (of the primary winding) is
given by (5.12). In the first case only the core cross section area increases, while
the allowed flux density and the applied volt-second remain constant. In the
second case the applied volt-second value increases due to the switching frequency
reduction. However, in the same time the allowed flux density amplitude increases
(under assumption of constant specific core loss). Both variables have influence on
the calculated number of primary winding turns and they counteract the applied volt-
second increase. To keep the current density constant the wire cross section area
increases twice. Resulting primary winding dc resistance is given by (5.13). In
the first case the total copper area (the wire cross section area multiplied by the
number of turns) increases with the same rate like the window area does, thus the
copper fill factor in (5.11) doesn’t change. In the first case the dc resistance
decreases by more than half. In the second case the dc resistance change depends
strongly on the required number of turns and thus it indirectly depends on the core
100
material properties. For assumed values of α and β the dc resistance increases

faster then assumed thus a potential dc copper loss may become higher then
expected.
 λ −1
 12 = 2 2 ⋅ n1
λ'  2 ⋅ Ac ⋅ 2 ⋅ ∆B
n1 ' = = (5.12)
Ac '⋅ 2 ⋅ ∆B '  2⋅λ
= n1
( )
α
 2 2 ⋅ Ac '⋅ 2 ⋅ ffs ' β ⋅ ∆B '
1
 s
 2− 2 ⋅ n ⋅ ρ ⋅ 2 4 ⋅ MLT Rdc
1 1
 = 11
n '⋅ ρ ⋅ MLT'  2 ⋅ ACu 24
Rn1dc ' = = (5.13)
ACu ' 1
 n ⋅ ρ ⋅ 2.83 4 ⋅ MLT
 = 0.65 ⋅ Rdc
 2 ⋅ ACu
In a transformer however the ac resistance has a great importance. To keep the

ac resistance low it’s necessary to ensure a low dc resistance and a low ac to dc
resistance factor FR. For simplicity let’s assume that a square wire is used and
windings are interleaved, so only single layer solution is considered. Moreover,
it’s assumed that the original design had the relative penetration depth ϕ=1.
Similarly to the inductor case discussed before, the relative penetration depth
increases if the switching frequency remains constant. In such case the ac to dc
resistance factor increases too. If the switching frequency decreases by half, the
relative penetration depth remains constant, so the ac to dc resistance factor doesn’t
change.
 2 ⋅h 1
 = 22 ⋅ ϕ
h' ACu '  δ
ϕ' = = = (5.14)
δ' δ'  2 ⋅h = ϕ
 2 12 ⋅ δ
Using Dowell’s method and curves presented on Figure 3.12 are used for
estimation of the ac to dc resistance factor and the resulting ac resistance. So, in
the first case the resulting ac resistance is 0.51 of the original one. In the second
case the resulting ac resistance is 0.65 of the original one. However, if the number
of layers in both designs increases to two the resulting ac resistance of the primary
winding is 0.73 and 0.65 of the original ac resistance in the first and the second case
respectively.
101
−5
x 10
capacitance per volume unit [F/m3]

3 1
2.5
0.8
volume [m3]
1.5 0.6
1
0.4
0.5
0 0.2
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
capacitance [F] −5 capacitance [F] −5
x 10 x 10
Figure 5.3 Volume and capacitance density vs. capacitance
of 160 VDC film capacitors
−4
x 10 capacitance per volume unit [F/m3] 8
1.4
1.2 7
1 6
volume [m3]
0.8
5
0.6
4
0.4
0.2 3
0 2
0 0.5 1 0 0.2 0.4 0.6 0.8 1
capacitance [F] −3
x 10 capacitance [F] x 10
−3
Figure 5.4 Volume and capacitance density vs. capacitance

of 400 VDC aluminum electrolytic capacitors
5.1.4 Capacitor
Since the converter is scaled up in power only the voltage rating of used
capacitors doesn’t change. To keep the output voltage ripple on acceptable level
the required capacitance is given by (5.15) in general case. In the first case, the
required capacitance increases twice due to twice larger discharge current Iout. In
the second case the required capacitance is four times larger than the original one
due to a larger discharge current Iout and a longer discharge time (lower switching
frequency fs). Figure 5.3 presents relation between the volume (size) and the
capacitance of several 160 VDC capacitors. Figure 5.4 presents the same
relationship for 400 VDC capacitors.
I out ⋅ D
C1 = (5.15)
2 ⋅ ∆vout ⋅ f s ⋅ N
102
Figure 5.5 Possible modules interconnections: a) series-input series-output;

b) series-input parallel-output; c) parallel-input series-output;
d) parallel-input parallel-output; e) series connection of modules
5.2 Overview on modules interconnection

Figure 5.5 presents possible connections of isolated single-input single-output
converter modules [97]. Series-input series-output connection isn’t a very common
one. This kind of connection can be useful in high voltage converters, where
converter modules share input and output voltages. It allows use of simple
switches, rather then complex ones made out of series connected devices. Series-
input parallel-output connection is desired for a step-down application, where a
conversion ratio is high, i.e. single stage power supply for a microprocessor.
Modules input terminals are series connected and share the input voltage, so low
voltage switches may be used. Output terminals are parallel connected increasing
the output current handling capability. Moreover, the voltage conversion ratio of a
single module is significantly lower then the conversion ration of the whole
converter, so the overall performance may increase. Parallel-input series-output
connection is suited for a step-up applications. A high input current splits and it’s
shared among several modules, while the output voltage is built-up from output
voltages of modules. Again, the voltage gain of a single module is lower than the
gain of a whole converter and it may turn into a better performance. Series
connection of input or output terminals requires a careful control of voltage sharing
to avoid failure. Redundancy implementation is possible but it requires smart
103
0.98
0.96
efficiency
0.94
0.92 Vin=30 V
V =40 V
in
0.9
V =50 V
in
0.88
50 100 150 200 250
P [W]
out
Figure 5.6 Estimated efficiency of a boost converter operating

at different input voltages
protections and bypasses. However series connection of the input or the output
requires a galvanic isolation, thus it’s not feasible for a non-isolated converter.
Parallel-input parallel-output connection can be used for isolated and non-
isolated converters. This kind of connection has no impact on voltage conversion
ratio of a single module, but it enables easy extension of the power level.
Moreover, using interleaving technique size of the input and the output filter may be
reduced significantly, as described in paragraph 2.2.2.
Figure 5.5 e) presents series connected converters and it may be considered as a
modular converter. This kind of converter is used in a multibus converter and in a
converter with a very high conversion ratio. Use of this connection reduces voltage
conversion ratio of each module and improves utilization of switches, as it was
explained in paragraph 2.2.3. Also, presence of an intermediate bus improves
power decoupling between the input and the output of the converter.
5.3 Fuel cell converter

Apart from loss mechanisms described in Chapter 3 and the converter design
calculations presented in Chapter 4 total losses of a boost converter can be described
by a function similar to the one given by (5.16). Exact values of particular
coefficients depend on topology and components used. However, most likely the
loss curve and the efficiency curve of a step-up converter will be similar to the one
presented on Figure 5.6. Now, if the converter is optimized for the low input
voltage, like it was presented in Chapter 4, the efficiency peaks at about half of the
rated power. However, once the input voltage increases the efficiency peak moves
toward higher power. Also, the peak efficiency will increase slightly in most cases.
It’s true for many isolated and non-isolated step-up converters and it has a great
impact on the system performance.
104
0.8
FC voltage
0.6
0.4
0.2
0 0.2 0.4 0.6 0.8 1 1.2
FC current
Figure 5.7 The fuel cell output voltage as a function
of the fuel cell output current (example curve)
0.98
0.96
efficiency
0.94
0.92
0.9
0.88
50 100 150 200 250
Pout [W]
Figure 5.8 Estimated efficiency of a boost converter operating

at variable input voltage (50 V at 50 W down to 30 V at 250 W)
Ploss = f (Vin , I in , Vout , I out ) =

(5.16)
= kVin ⋅ Vin jVin + k Iin ⋅I in jIin + kVout ⋅ Vout jVout + k Iout ⋅ I out jIout
Figure 5.7 presents the fuel cell voltage as a function of the fuel cell output
current. It’s just an example curve, but it indicates important feature of current fuel
cells. At no load condition the fuel cell output voltage is the highest and it
decreases wile the fuel cell is being loaded. Now, the voltage / power profile is
being combined with the step-up converter model and its efficiency curves from
Figure 5.6. Efficiency of the converter is calculated for variable input voltage and
the result is presented on Figure 5.8. At the rated output power and the minimum
105
0.96
0.94
efficiency
0.92
250 W converter
0.9 2.5 kW converter
0 500 1000 1500 2000 2500

P [W]
out
Figure 5.9 Estimated efficiency of the 250 W boost converter

and the scaled 2.5 kW boost converter (ideal scaling) at 30 V input voltage
input voltage the power losses are dominated by conduction losses due to a high
input current. At the other end, at light load the voltage dependant losses
dominate, while the conduction losses are fairly low. An increased input voltage at
light load leads to further rise of voltage dependant losses and efficiency degradation
– compare curve on Figure 5.8 and curve related to 30 V input voltage on Figure
5.6.
5.4 Modular converter

As presented above a fuel cell output characteristic and a step-up converter
characteristic doesn’t match. It may result in worse poor utilization of converter
performance in terms of the conversion efficiency, especially at light loads.
Use of a parallel-input parallel-output modular converter may help to overcome
this drawback. Now, let’s assume that the converter, which efficiency curves are
presented on Figure 5.6 and Figure 5.8 is being scaled up in power – let’s assume
that the rated power of the new converter is 2.5 kW and the efficiency curve is
“stretched” in power (ideal scaling), like presented on Figure 5.9. The peak
efficiency in both cases is the same and it occurs at about half of the rated power (at
the lowest input voltage). Also efficiency at rated power in both cases is the same.
Instead of scaling of the converter one may use ten smaller converters in parallel
and achieve the same power level and the same efficiency profile like a single
smaller converter has. Use of several paralleled converter modules requires careful
current sharing – it includes both, control algorithm and current sensors in each
module. On the other hand, using interleaving modulation scheme it’s possible to
reduce size of passive components, like the output capacitor or boost inductors.
Moreover it’s possible to improve converter efficiency for partial and light loads in
compare with the scaled converter. Figure 5.10 presents estimated efficiency
106
0.96
0.95
efficiency
1 module
0.94 2 modules
4 modules
6 moduls
0.93
8 modules
10 modules
0.92 scaled converter
0 500 1000 1500 2000 2500
P [W]
out
Figure 5.10 Estimated efficiency of the modular boost converter

and the scaled converter operating with 30 V input voltage
0.96
0.94
efficiency
0.92 1 module
2 modules
0.9 4 modules
6 modules
0.88 8 modules
10 modules
0.86 scaled converter
0 500 1000 1500 2000 2500
Pout [W]
Figure 5.11 Estimated efficiency of the modular boost converter

and the scaled converter operating with variable input voltage
curves for 2, 4, 6, 8, and 10 modules operating in parallel. In compare with the

scaled converter (dashed curve on Figure 5.10) the modular converter performs
better at partial and light loads. At the rated power efficiency of both converters is
about the same.
If the input voltage is current dependent, like presented on Figure 5.7 use of the
modular converter may provide even more benefits. Figure 5.11 presents estimated
efficiency of the scaled converter and the modular converter operating at variable
input voltage (different number of modules is turned-on).
107
0.96
0.95
efficiency
0.94
1 module
2 modules
0.93 3 modules
4 modules
0.92
0 200 400 600 800
P [W]
out
Figure 5.12 Measured efficiency of the modular boost converter

operating at 30 V input voltage
0.96
0.95
efficiency
0.94
1 module
2 modules
0.93 3 modules
4 modules
0.92
0 200 400 600 800
Pout [W]
Figure 5.13 Measured efficiency of the modular boost converter

operating at variable input voltage
To verify these considerations a 4-phase 800 W modular boost converter is built

and tested. Figure 5.12 presents measured efficiency of the breadboard operating
at 30 V input voltage and 400 V output voltage. It’s demonstrated that the
efficiency profile and the peak efficiency remain the same, regardless number of
operating modules. Also, turning-off some of modules while the converter
operates at light load actually improves efficiency, so it’s possible to maintain a high
efficiency over a wide load ranges.
Figure 5.13 presents measured efficiency of the breadboard operating at variable
voltage – 50 V at 100 W output power drops down to 30 V at 800 W in linear
manner. Now, efficiency of the modular converter actually increases when the
108
Figure 5.14 Block diagram of the proposed PWM generator
Figure 5.15 Key waveforms in the proposed PWM generator
output power decreases. It’s opposite tendency in compare with a scaled boost
converter.
5.4.1 Interleaved PWM generation

Described above method for efficiency improvement is valid aside from used
modulation scheme – interleaved or synchronous. However, interleaving is
preferred since it provide benefit of potentially smaller passive components. The
challenge now is to get high quality multiphase gating signals which are interleaved
equally, even if number of phases changes. So far no dedicated solution (like
integrated circuit) was found on the market.
The solution proposed here bases on two sawtooth generators, the delay unit and
set of comparators (two comparator per PWM channel), like presented on Figure
5.16. Figure 5.15 presents key waveforms. The delay unit delays the sawtooth
signal by the programmable time td, which equals the transistor turn-on time. Each
PWM channel has two comparators. One associated with original sawtooth signal
and it’s responsible for turn-on signal generation (rising edge of the comparator
output). The other comparator is associated with the delayed sawtooth signal and it
generates turn-off signal (rising edge of the comparator output). Basically both
109
Figure 5.16 Diagram of external logic circuit
Figure 5.17 Picture of the XC167 board with external logic connected
comparators (in the same PWM channel) are connected to the same reference
voltage VChi(ref). Value of the reference voltage of the last PWM channel is just
below the sawtooth peak voltage. Reference voltages of all remaining channels are
proportionally lower, like presented on Figure 5.15. If number of active channels
are about to change (increase of decrease) reference voltages are recalculated in the
way that they are distributed uniformly from zero to the peak of the sawtooth signal.
In this way all outputs are interleaved equally and have the same duty cycle,
regardless the number of active channels. Additionally reference voltage of one
comparator in a single channel may be changed and it results in duty cycle change in
this channel. It can be used in an active current sharing scheme.
Since comparators provide only turn-on and turn-off signals it’s necessary to
process them further. Figure 5.16 presents a simple logic, based on two D-type
flip-flops with asynchronous reset input, which detects rising edges of comparator
outputs and process them into PWM waveforms.
The solution proposed here is implemented in microcontroller XC167CI-16F40F
from Infineon [80]. This microcontroller provides 3+1 PWM channels, which
unfortunately are desired to be part of 3-phase inverter (either grid connected or for
electric drive). The microcontroller consists up to 32 capture-compare (CAPCOM)
channels which are used for turn-on and turn-off signals generation as explained
above. External D-type flip-flops (7474, dual in package) are used to get PWM
waveforms. Figure 5.17 presents 4-phase implementation. C-code used in the
microcontroller is available in Appendix B, paragraph B.12.
110
5.5 Summary
This chapter deals with scaling and paralleling of dc-dc step-up converters.
First scaling in power of the converter was discussed. It included
considerations about use of higher power semiconductors and growth of passive
components. Two cases were discussed in parallel – first, where the switching
frequency didn’t change after scaling of the converter, and the second, where use of
a higher current rated transistors lead to the switching frequency reduction.
Next, impact of the variable input voltage on the converter efficiency was
discussed based on the fuel cell converter example. It’s found that at the light load
(and high input voltage) the efficiency of the converter drops due to higher voltage
dependent losses in the converter. On the other end, at the rated output power (and
the lowest input voltage) conduction losses dominate in the converter. So, variable
input voltage makes optimization of the converter more difficult.
Finally, the parallel modular converter concept was presented as an alternative
for converter scaling. Intelligent use of the modular converter in fuel cell or similar
applications may benefit in a better efficiency at medium and light loads in compare
with the scaled converter.
111
Chapter 6
Experimental results
Some experimental results have been presented already. At the end of Chapter
3, in paragraph 3.8 the power loss model was verified using the boost converter
breadboard (rated power 200 W, input voltage 20-30 V, output voltage 90 V). In
paragraph 5.4 performance of a modular converter was demonstrated using 4-phase
800 W boost converter, which input voltage is 30-50 V and the output voltage is 400
V and it fits the specification from paragraph 1.4. Efficiency achieved by this
converter was in the range of 95-96%.
To demonstrate performance of the push-pull-boost converter three breadboard
based on this topology are built and tested. In this chapter experimental setup is
presented first. It includes description of the equipment used, as well as
presentation of three breadboards. Also electrical measurements of custom made
magnetic components are presented. Then observed key waveforms and efficiency
measurement results of all breadboards are presented and discussed briefly.
6.1 Experimental setup and breadboards

The diagram of the experimental setup used for efficiency measurement is
presented on Figure 6.1 and the picture of the setup is presented on Figure 6.2. The
converter under test (CUT) is supplied from the regulated dc power supply Toellner
TOE-8872 (60 V, 65 A, 1500 W) [98]. The electronic load Zentro-Elektrik
EL1000/800/20 (800 V, 20 A, 1000 W) [99] operating in current control mode is
used. Auxiliary power supply GWINSTEK GPS-4303 (30 V, 3 A) [100] is used to
power up gate drivers and microcontroller board (Infineon XC167 [80]), which is
used as PWM gating signal generator. Use of the electronic load instead of
variable resistors ensures more accurate control of the output power and improves
system stability. Four precise 6 ½ digit multimeters Fluke 8845A [101] are used
Figure 6.1 Diagram of the experimental setup

112
Figure 6.2 Picture of the experimental setup used
Figure 6.3 Picture of the 500 W push-pull-boost converter
for the input and output voltages and current measurements. All values, except the
input current are measured directly. To measure the input current the precise 40 A
/ 60 mV shunt resistor is used. To improve thermal stability the shunt resistor is
placed on the heatsink, so its temperature variations are greatly reduced. All
measurements are done when the converter under test is warmed up. During
measurement all multimeters are triggered synchronously, so influence of a low
frequency changes is minimized.
Additionally, the digital scope Tektronix TDS-3014C [102] is used to observe
and record waveforms. In connection with the digital scope high voltage
differential probes Tektronix P5205 [102] and current probes Tektronix TCP-202
(15 A peak) [102] and LEM PR-50 (30 A peak) [103] are used.
Using this equipment three breadboards are tested. The first one is 500 W
push-pull-boost converter made according to the design presented in Chapter 4.
This converter has the dc inductor and the push-pull transformer arranged on two
113
Figure 6.4 Picture of the 500 W push-pull-boost converter

with integrated inductor-transformer
Figure 6.5 Picture of the 1 kW push-pull-boost converter

with integrated inductor transformer
separate cores and the assembled converter is presented on Figure 6.3. More
details about this breadboard is presented in Appendix A, paragraph A.4.
The second prototype is also 500 W push-pull-boost converter made according to
the design calculations from Chapter 4. The difference is in magnetic components
since this converter has integrated inductor-transformer. The assembled converter
is presented on Figure 6.4. More details about this breadboard is presented in
Appendix A, paragraph A.5.
The third breadboard is a scaled-up version of the second one. The scaling
process is simplified and there are only two major changes in compare with the
second breadboard – original transistors IRFP4321PbF [78] are replaced by newer
ones IRFP4568PbF [78], which have lower on-state resistance, and integrated
magnetic device is redesigned to handle higher power. Beside it, the switching
114
Figure 6.6 Picture of the assembled dc inductor and push-pull transformer

for the 500 W push-pull-boost converter, cores size EE42/21/20
Figure 6.7 Picture of the assembled integrated inductor-transformer

for the 500 W push-pull-boost converter, core size EE42/21/20
frequency remains the same, the diodes remain the same and the general converter
layout remain the same too.
In addition to three breadboards presented above the 300 W push-pull-boost
converter was made. This converter was the first breadboard with integrated
inductor-transformer (EE42/21/20 core). More details about this breadboard is
presented in Appendix A, paragraph A.3.
6.2 Magnetic components measurement

In Chapter 4 magnetic components for the push-pull-boost converter were
designed. Their electrical parameters were found using FEMM simulation
software. In this paragraph electrical parameters of these custom made magnetic
components are measured using precise RLC-meter HP/Agilent 4284A. To make
the comparison easier calculated and measured resistance values are presented as
115
Table 6.1 Calculated winding resistances of the dc inductor (core EE42/21/15) and
the push-pull transformer
Calculated Measured
Resistance seen by the input dc current (RL1+Rn1||Rn11) 5.7 mΩ -
Resistance seen by the input current ripple (RL1+Rn1||Rn11) 200 mΩ -
Resistance seen by the secondary current (Rn2+Rn1+Rn11) 18.3 mΩ -
Table 6.2 Calculated and measured winding resistances of the dc inductor (core
EE42/21/20) and the push-pull transformer
Calculated Measured
Resistance seen by the input dc current (RL1+Rn1||Rn11) 3.8 mΩ 4.2 mΩ
Resistance seen by the input current ripple (RL1+Rn1||Rn11) 105 mΩ 120 mΩ
Resistance seen by the secondary current (Rn2+Rn1+Rn11) 18.3 mΩ 21 mΩ
Table 6.3 Calculated and measured winding resistances of the integrated inductor-
transformer for the 500 W converter (EE42/21/20 core size)
Calculated Measured
Resistance seen by the input dc current (RL1+Rn1||Rn11) 6.3 mΩ 8 mΩ
Table 6.4 Calculated and measured winding resistances of the integrated inductor-
transformer for the 1 kW converter (EE55/28/21 core size)
Calculated Measured
Resistance seen by the input dc current (RL1+Rn1||Rn11) 2.3 mΩ 2.8 mΩ
resistances seen by the particular current component, like explained in paragraph

4.1.3. The input current (dc and ac ripple) flows thru the inductor winding and thru
primary windings which are effectively in parallel and the resistance seen by this
current component is expressed as RL1+Rn1||Rn11. The secondary winding current
flows in the secondary winding, but it’s also reflected to the primary windings,
which are effectively series connected for this current component and the resistance
seen by this current component is expressed as Rn2+Rn1+Rn11.
The dc inductor used in the first breadboard is different form the one designed in
paragraph 4.2.5 since it bases on a larger core – EE42/21/20 instead of EE42/21/15.
It results in lower number of turns and thus a lower winding dc resistance in
compare with the design calculations in paragraph 4.2.5. The push-pull
transformer made according to the design calculations from paragraph 4.2.4. Dc
and ac resistances of both inductor designs are summarized in Table 6.1 and Table
6.2.
116
Figure 6.8 Picture of the assembled integrated inductor-transformer

for the 1 kW push-pull-boost converter, core size EE55/28/21
Figure 6.7 presented assembled integrated inductor-transformer for the 500 W

push-pull-boost converter made according to design calculations from paragraph
4.2.6. Table 6.3summarizes calculated and measured winding resistances.
In addition, the integrated inductor-transformer has been scaled-up in the way
that it’s suitable for the 1 kW push-pull-boost converter. The assembled device is
presented on Figure 6.8. It bases on EE55/28/21 core, ferrite material type P.
Using the same guidelines like presented in paragraph 4.2.6 it’s found that the
inductor winding has 6 turns, the primary winding of the transformer has 8 turns and
the secondary winding of the transformer has 18 turns. Calculated and measured
winding resistances are summarized in Table 6.4.
6.3 Efficiency measurement and estimation

In this paragraph efficiency estimation and measurement results are presented.
Figure 6.9 and Figure 6.10 present measured and estimated efficiency of the 500 W
converter with separated inductor and transformer.
Figure 6.11 and Figure 6.12 shows measured and estimated efficiency of the 500
W converter with integrated inductor-transformer.
Figure 6.13 and Figure 6.14 shows measured and estimated efficiency of the 1
kW converter with integrated inductor-transformer.
All simulations were done under following conditions:
• temperature of all components was constant during all simulations
• the dc-bias and the temperature influence on the core loss was not
included (certain coefficients in the model were set to zero)
• additional conduction losses caused by wiring and interconnections were
not included
• calculated (not measured) values of winding resistances were used
• effect of the interwinding capacitance in case of the foil winding is not
inclueded
117
0.98
0.97
efficiency 0.96
0.95
30V
0.94
50V
0.93
0.92
0 200 400 600
P [W]
out
Figure 6.9 Measured efficiency of the 500 W push-pull-boost converter with

separated inductor and transformer at 30 V and 50 V input voltage
0.98
0.97
0.96
efficiency
0.95
0.94
30 V
0.93
50 V
0.92
0 200 400 600
Pout [W]
Figure 6.10 Estimated efficiency of the 500 W push-pull-boost converter with

separated inductor and transformer at 30 V and 50 V input voltage
118
0.98
0.97
efficiency 0.96
0.95
30V
0.94
45V
0.93
0.92
0 200 400 600
P [W]
out
Figure 6.11 Measured efficiency of the 500 W push-pull-boost converter with

integrated inductor-transformer at 30 V and 45 V input voltage
0.99
0.98
0.97
efficiency
0.96
0.95
30 V
0.94 45 V
0.93
0 200 400 600
Pout [W]
Figure 6.12 Estimated efficiency of the 500 W push-pull-boost converter with

integrated inductor-transformer at 30 V and 45 V input voltage
119
0.98
0.97
efficiency 0.96
0.95
0.94
30V
0.93
0.92
0 200 400 600 800 1000 1200
P [W]
out
Figure 6.13 Measured efficiency of the 1 kW push-pull-boost converter with

integrated inductor-transformer at 30 V input voltage
0.98
0.97
0.96
efficiency
0.95
0.94
0.93 30V
0.92
0 200 400 600 800 1000 1200
Pout [W]
Figure 6.14 Estimated efficiency of the 1 kW push-pull-boost converter with

integrated inductor-transformer at 30 V input voltage
6.4 Summary
In this chapter the experimental setup – equipments and breadboards – was
presented and briefly described. Then estimated and measured values of winding
resistance of magnetic components were provided. Finally, measured and
estimated efficiency curves of three breadboards were presented.
121
Chapter 7
Conclusion
In this thesis non-isolated dc-dc step-up converters suitable for high voltage gain
applications were investigated. At the beginning the literature study provided a
good overview on different step-up topologies. It’s found that converters based on
an inductor and a coupled-inductor principle (a boost and a center tapped boost
converters) as well as converters derived from isolated converters (a non-isolated
flyback-boost, a non-isolated push-pull-boost and a non-isolated two-inductor-boost
converters) are good candidates for future investigation. Description of selected
topologies includes the circuit scheme, the diagram of key waveforms, the key
equations as well as operation explanation and features of the converter. At the end
of Chapter 2 all selected topologies are compared against each other in terms of
required number of components and required ratings (current, voltage, etc.) of
components. This quantitive comparison together with description indicated that
the non-isolated push-pull-boost converter is the best candidate for the non-isolated
fuel cell converter. Also the non-isolated two-inductor-boost converter is a very
good candidate since it requires almost identical semiconductors and capacitors like
the non-isolated push-pull-boost converter does. The main difference between
these topologies is in magnetic components – the non-isolated push-pull-boost
converter requires a smaller inductor, while the non-isolated two-inductor-boost
converter requires a smaller and simpler transformer. Since a high efficiency is a
main goal its important to predict losses in a converter in advance. So, modeling
and calculation of losses is a very important part of the project. Used modeling
approach and loss models are introduced and described in Chapter 3. Validation of
these models is provided at the end of the Chapter based on the boost converter
breadboard. In Chapter 4 the non-isolated push-pull-boost converter is being
analyzed in details. Then the design example is provided to the reader. Also
integration of the inductor and the push-pull transformer is presented in this Chapter.
It’s demonstrated that such integration of magnetic components on a single core may
reduce size of the converter, while losses remain on the same level. In Chapter 5
converter scaling problem is presented. It includes discussion on growth of
components and the impact on the converter operation and performance. Next
modular converter approach is presented as an alternative for scaling. Based on
simulations and measurements it’s demonstrated that the efficiency of the modular
converter remains high over wide range of loads. In case of the fuel cell
application the efficiency of the modular converter increases while the output power
decreases, which is exactly opposite in compare with a solid step-up converter.
Finally, in Chapter 6 laboratory setup and equipment are described and experimental
122
results are presented. Three breadboards based on the non-isolated push-pull-boost

converter topology are realized and tested demonstrating a very high efficiency,
exceeding 97% in case of both 500 W breadboards.
7.1 Conclusions and new aspects

Below specific conclusions and novelties are summarized:
• It’s found that a non-isolated push-pull-boost and a non-isolated two-
inductor-boost topologies are the best candidates for a high voltage gain
converters, including a fuel cell application. These topologies are more
complex then a basic boost converter and they require more components,
but they provide better utilization of switches and they require smaller
passive components. Also these two topologies have a higher voltage
gain, so they operate at a lower duty cycle than the basic boost converter
does.
• It’s demonstrated that a modular step-up converter may improve
efficiency of the fuel cell converter. First, the converter design is not
limited by the required output power, thus it’s possible to utilize available
components (especially MOSFETs) in optimum way ensuring low losses.
Second, using a modular converter it’s possible to improve efficiency at
light and partial loads by turning-off one or more modules.
• It’s demonstrated that a basic boost converter is able to achieve peak
efficiency of 95.5%, which is well below level set by current state-of-the-
art isolated converters (up to 98%) [104, 105]. However, using a proper
non-isolated topology it’s possible to achieve a similar performance in
terms of the peak efficiency and the size of magnetic components.
Integration of an inductor and a transformer on a single core enables
further size reduction of magnetic components. It’s believed that the
limit of performance with available components was reached.
• Averaged steady-state models of converters together with component loss
models is a useful tool in prediction of converter power losses.
Component loss models base on parameters available in component
datasheet in most cases, so parameters extraction is minimized or even
avoided at all. With small changes in the main script file these models
were used to estimate an optimum power level and an optimum switching
frequency for a given MOSFETs. It’s possible to develop similar
averaged models for other topologies, including isolated dc-dc converter
and inverters.
• It’s believed that two dc-dc topologies developed during this project are
novel. The first topology is so called a single switch dual output non-
isolated boost converter and it was presented on 23rd Annual Applied
Power Electronics Conference and Exposition APEC 2009, Austin, USA
(see Appendix A.1). The second topology is a non-isolated two-
inductor-boost converter (see paragraph 2.2.7) derived from an isolated
two-inductor step-up topology.
123
125
Appendix A
Publications
126
A.1 A Single Switch Dual Output Non-Isolated

Boost Converter (APEC 2008)
A Single Switch Dual Output

Non-Isolated Boost Converter
Pawel Klimczak Stig Munk-Nielsen
Institute of Energy Technology Institute of Energy Technology
Aalborg University Aalborg University
Pontoppidanstraede 101 Pontoppidanstraede 101
9220 Aalborg, DENMARK 9220 Aalborg, DENMARK
pak@iet.aau.dk smn@iet.aau.dk
Abstract- There are many applications for high gain dc-dc

converters. In several of these applications galvanic isolation is
not required, but there are some safety issues regarding missing
isolation and leakage current. Usage of a half-bridge inverter
and a dual dc-link may solve this issues. In this paper very
simple dual output non-isolated boost converter is presented.
Single active switch is used to control both, positive and negative
output voltages. The converter is desired to boost unregulated
low input voltage 25-50 Vdc to regulated high voltage ±400 Vdc
in dual dc-link. In this paper proposed topology is analyzed and
explained. Obtained experimental results prove good
performance of the converter over wide range of input voltage
and output power.
I. INTRODUCTION
Fig. 1 - Proposed new single switch dual output boost
Demand for high gain and high efficiency dc-dc converters converter
is always present. Most recognized application for this kind
of converters were UPS systems for network servers and
telecom equipment. Nowadays new applications emerge: fuel II. ANALYSIS OF THE PROPOSED CONVERTER
cell (FC) and photovoltaic (PV) based distributed generators The proposed single switch dual output boost converter is
up to few kilowatts. Usually, in such cases, the low voltage presented on Fig. 1. In the simplest word, it is a basic boost
from the dc source (25-50 Vdc) is converted to the higher dc- converter [8] with very few additional components: a
link voltage (350-400 Vdc) [1]. Nowadays many applications secondary winding n2 (magnetically coupled with the primary
incorporates many smaller, paralleled and interleaved dc-dc winding n1) and a voltage doubler (D2, D3, C2, C3). In a steady
converters instead of one large converter [2, 3]. The modular state voltage transfer functions of an ideal converter are given
architecture provides better reliability, scalability, possibly by (1). For n1=n2 and continuous conduction mode (CCM) the
higher efficiency and lower costs of production and converter is able to generate symmetrical positive and
maintenance. Also in some applications transformer-less negative voltages.
converters are found as more efficient, smaller and cheaper 1
solution. According to [4] usage of 3-level neutral-point- Vout_pos Vin
1 D
clamped (NPC) inverter is good choice in order to decrease (1)
n2 n D
leakage current to the ground and fluctuations of the PV panel Vout_neg VC 2 VC 3 Vin 2 Vin
n1 n1 1 D
in the non-isolated converter. In such solution one terminal of
the source is connected to the grid neutral wire. On the other
hand this solution requires positive and negative dc-link For a little more detailed analysis it is assumed that both
voltages. In [5, 6] there are presented non-isolated converters (positive and negative) loads are resistive and balanced. The
with dual (±) output. coupled inductor is modeled as a two-windings ideal
In this paper a simple and efficient dual output non-isolated transformer with unity turns ratio (n1=n2) and a magnetizing
dc-dc boost converter is presented (Fig. 1). The proposed inductor LM in parallel to the primary winding. Leakage
topology bases on the one presented in [7], but different inductances Ls1 and Ls2 are very small and it’s influence is
connection of the secondary winding n2 and voltage doubler negligible, however winding resistances Rs1 and Rs2 are taken
(D2, D3, C2, C3) enables to generate positive and negative into account. Diodes are modeled as an ideal diode in series
output voltage using only one active switch. This converter is with voltage source (forward voltage VDi, i=1, 2, 3) and
a candidate for module in a parallel modular boost converter. resistor (forward resistance RDi, i=1, 2, 3). As long as SiC
127
Fig. 2 - Operational waveforms
diodes are used no reverse recovery phenomena is considered.

Moreover it is assumed that capacitors C1 and C3 have equal
capacitances. Capacitor C2 has a capacitance significantly
larger than C3 has. The converter operates in the continuous
conduction mode (CCM). Under above assumptions there are
four operational states. Basic waveforms are presented on
Fig. 2.
State 1 (t0-t1): At the beginning of this state the transistor T is
being turned-on. During this period the input voltage Vin is
applied to the primary winding, the magnetizing current iLM(t)
is rising linearly and the energy is stored in the magnetizing
inductance LM. A voltage slightly lower than the input voltage
induces in the secondary winding n2 and the current in2 ramp-
up to zero amps. Diode D3 become reverse biased. If the
voltage across capacitor C2 is lower than the induced voltage
vn2(t), then diode D2 become forward biased and the capacitor
C2 being charged. The converter configuration is presented on
Fig. 3a. If leakage inductances are very small and negligible,
then in2 current in time period t0-t1 is given by (2). In order to
make the equation easier to read all resistances are grouped as
a primary side resistance Rpri=Rs1+RDSon; and a secondary side
resistance Rsec=Rs2+RD2. It is important to note that the
magnetizing current iLM(t) is time dependent value and it has
influence on charging the capacitor C2. Because vC2(t) voltage
Fig. 3 - Basic operational states of the converter
ripples are much smaller than the average output voltage
Vout_neg it is assumed that output current Ineg is constant in time
period t0-t1.
128
 Vin iLM t Rpri vC 2 (t0 ) VD 2 

iin 2 t  
 Rpri Rsec 
 
(2)
 t 
exp   I neg
 R R C 
 pri sec 2 
State 2 (t1-t2): In this state the transistor remains turned-on.

Capacitor C2 become fully charged and secondary winding
current in2 falls to a small value, about Ineg. In the same time
the voltage drop on primary side resistances (Rs1, RDSon)
become larger due to the rising magnetizing current iLM(t) (3).
It means that slightly lower voltage is applied to the ideal
transformer’s primary winding n1, lower voltage induces in
the secondary winding n2 and lower voltage is applied to the
capacitor C2. If in2(t) current become smaller than Ineg load Fig. 4 - Observed waveforms
current then the capacitor C2 discharges through the load
resistor Rload_neg and voltage vC2(t) falls too. Falling vC2(t) is lower than it was at the end of previous period. It means
voltage drives current in2(t) to rise. In many cases it may that difference between induced voltage vn2(t2+T) and
happen that voltage vn1(t) drops faster than voltage vC2(t) and capacitor voltage vC3(t2+T) become larger. The larger voltage
current in2(t) falls to zero. Both diodes D2 and D3 are reverse difference will cause higher charging current. New
biased. The converter is in the state presented on Fig. 3b. equilibrium will be established, but by a cost of a lower
average VC3 voltage.
vn1 t | Vin Rs1 RDSon iLM t (3)
III. EXPERIMENTAL SETUP AND RESULTS
I neg in 2 t
vC 2 (t ) | vC 2 (t1 ) t (4) In order to verify proposed idea the 200 W demonstrator
C2 converter has been built and tested. Desired input voltage
range is 25-50 Vdc and it fits output voltage of the fuel cell
State 3 (t2-t3): At the beginning of this state the transistor T is stack up to few kW, single PV panel or many of lead acid
being turned-off. At the beginning the whole input current batteries. The output voltage is ±400 Vdc (dual dc-link) and it
flows through D1 diode to the capacitor C1. The voltage vn1(t) is suitable for half bridge 1-phase inverters and 3-phase 4-
in t2 moment is given by (5). Current flow in the converter is wires inverters. The converter was designed to operate at
presented on Fig. 3c. switching frequency fs=30 kHz, but is was tested at switching
frequencies from 25 kHz up to 35 kHz. A single 600 V
vn1 (t2 ) Vout_pos Vin iin t2 Rs1 RD1 (5) MOSFET (Fairchild, FCH47N60) and two 600 V SiC diodes
(Infineon, SDT05S60) are used (D1, D3). Diode D2 is silicon
ultra fast diode 15ETL06 from International Rectifier. Due to
If the voltage given by (5) is higher than C3 capacitor
the converter inherent properties this diode has no reverse
voltage then the diode D3 become forward biased and starts to
recovery problem and it was chosen because of its low
conduct (Fig. 3d). The current in secondary winding is given
forward voltage drop. The coupled inductor bases on the
by (6). Energy stored in the magnetizing inductor is
toroid powder core (Magnetics, KoolMu 77930). It has 51
transferred to output capacitors C1 and C3. An amount of
turns primary and 51 turns secondary winding.
energy transferred to each capacitor depends on the load
Fig. 4 shows observed waveforms of the operating
connected to each output and it’s represented by a shaded
converter for minimum desired input voltage 25 Vdc (duty
areas on Fig. 2.
ratio equal to 93.8% at fs=25 kHz) and 100 W output power.
vn1 t2 vC 3 (t2 ) VD 3
iin 2 t One may compare these waveforms with theoretical ones
Rs 2 RD 3 presented on Fig. 2. First of all it is important that in2 (Ch3)
(6)
 t  current slopes are very steep during transistor transients. It
exp   I neg means that leakage inductances are fairly small, like assumed
 R R C 
 s2 D3 3 
before. Also a positive part of the observed current in2 falls
Now, lets assume that the converter operates with a very exponentially almost to zero value. Capacitor C2 is fully
high duty cycle. It may happen that the transistor turn-off charged. However negative part of the current in2 does not fall
time become so short, capacitor C3 cannot be full recharged to zero – as explained before the turn-off time is too short and
and it’s voltage vC3(t) at the beginning of the next switching capacitor C3 cannot be fully recharged.
period is lower. Consequently vC3(t) at the end of next period
129
Fig. 5 - Measured output voltages (fs=25 kHz) Fig. 7 - Measured efficiency of the converter (fs=25 kHz)
Fig. 6 - Measured output voltages (fs=30 kHz) Fig. 8 - Measured efficiency of the converter (fs=30 kHz)
In fact, for lower input voltage Vin, negative output voltage simple topology with low overall component count. The
Vout_neg ‘drops’ (Fig. 5). It is more visible at higher power and negative voltage is generated using the coupled inductor and
higher current. One explanation for this may be a greater the voltage doubler.
influence of the parasitic components. Another one is a fact Analysis of the ideal converter was presented and principle
that a larger amount of energy has to be transferred to of operation was explained. This analysis proved, that it is
capacitor C3 in each cycle. Comparing Fig. 5 and Fig. 6 one possible to create symmetrical output voltage over a wide
may notice that for higher switching frequency voltage range of the input voltage.
asymmetry become larger, especially for lower input voltage Finally experimental data were presented. Recorded
and higher output power. On the other hand, by comparing waveforms confirmed principle of operation of the converter,
Fig. 7 and Fig. 8 one may notice that for higher switching but also indicated influence of parasitics. Measured output
frequency an efficiency of the converter is higher. It’s also voltages pointed a potential limitation of the voltage gain.
important that for lower voltage (and higher input current)
ACKNOWLEDGMENT
efficiency drops due to the conduction losses. For 100 W
output power and high voltage efficiency drops a little also. This project is financed by Dansk Energi net under PSO
It’s mainly due to the core losses in the coupled inductor. project no. 562/06-14-26808.
IV. CONCLUSION REFERENCES
In this paper new solution for a single-switch dual-output [1] Z. Qun, T. Fengfeng, and F. C. Lee, "A front-end DC/DC
non-isolated boost converter is presented. This is a very converter for network server applications," in Power Electronics
130
Specialists Conference, 2001. PESC. 2001 IEEE 32nd Annual, [7] B. Ju-Won, R. Myung-Hyo, K. Tae-Jin, Y. Dong-Wook, and K.
2001, pp. 1535-1539 vol. 3. Jong-Soo, "High boost converter using voltage multiplier," in
[2] V. Dang Bang, Y. Lembeye, J. P. Ferrieux, J. Barbaroux, and Y. Industrial Electronics Society, 2005. IECON 2005. 31st Annual
Avenas, "New high power - high ratio non isolated DC-DC boost Conference of IEEE, 2005, p. 6 pp.
converter for Fuel cell applications," in Power Electronics [8] N. Mohan, T. M. Undeland, and W. P. Robbins, Power
Specialists Conference, 2006. PESC '06. 37th IEEE, 2006, pp. 1-7. electronics : converters, applications, and design, 3rd ed.: John
[3] S. B. Kjaer, J. K. Pedersen, and F. Blaabjerg, "A review of single- Wiley & Sons, 2003.
phase grid-connected inverters for photovoltaic modules," Industry [9] J.-H. Kim and S.-K. Sul, "PWM Method of a Three-Level Neutral-
Applications, IEEE Transactions on, vol. 41, pp. 1292-1306, 2005. Point-Clamped Converter under Neutral Point Imbalance
[4] T. Kerekes, R. Teodorescu, and U. Borup, "Transformerless Condition," in Applied Power Electronics Conference, APEC 2007
Photovoltaic Inverters Connected to the Grid," in Applied Power - Twenty Second Annual IEEE, 2007, pp. 1089-1094.
Electronics Conference, APEC 2007 - Twenty Second Annual [10] A. von Jouanne, D. Shaoan, and Z. Haoran, "A simple method for
IEEE, 2007, pp. 1733-1737. balancing the DC-link voltage of three-level inverters," in Power
[5] Y. Kanthaphayao and C. Boonmee, "Dual-output DC-DC power Electronics Specialists Conference, 2001. PESC. 2001 IEEE 32nd
supply without transformer," in TENCON 2004. 2004 IEEE Annual, 2001, pp. 1341-1345 vol. 3.
Region 10 Conference, 2004, pp. 41-44 Vol. 4.
[6] F. L. Luo, "Double-output Luo converters, an advanced voltage-
lift technique," Electric Power Applications, IEE Proceedings -,
vol. 147, pp. 469-485, 2000.
131
A.2 Comparative Study on Paralleled vs. Scaled

Dc-dc Converters in High Voltage Gain
Applications (EPE-PEMC 2008)
Comparative Study on
Paralleled vs. Scaled Dc-dc Converters
in High Voltage Gain Applications
Pawel Klimczak*, Stig Munk-Nielsen*
* Aalborg University / Institute of Energy Technology, Aalborg, Denmark, e-mail: pak@iet.aau.dk
Abstract—Today power converters are present in many

commercial, medical and industrial applications. A lot of
them are high power and high current applications. In
order to increase power handling capability several
transistors or diodes are paralleled often. However such
paralleling may lead to converter’s performance
degradation or switches quick failure. A parallel modular
converter built of many paralleled modules may be an
interesting alternative, while a modular converter provides
well known advantages like scalability, improved reliability
and lower cost. This paper investigates possibility of
improving an efficiency by intelligent usage of a modular
boost converter in a high voltage gain application.
Keywords— DC power supply, Interleaved converters,

Parallel operation, MOSFET
I. INTRODUCTION
If the power level is low it’s possible to use only a
single power semiconductor as a switch. For higher
power levels a single transistor or a single diode may not
be enough to handle high current and it is required to use
bigger component or few ones in parallel. Usually it’s
easy to parallel two or three components, but for larger
Fig. 1 - Parallel modular boost converter (a) and a boost converter
number of paralleled components interconnections with paralleled/scaled devices (b)
become more complex. It’s difficult to ensure equal stray
inductances in each current path and some components now design and optimize a converter for a power level
may experience larger stress, especially during fast which ensures the highest efficiency or the best
switching transients [1, 2]. In the same time due to on- components utilization. Further on, the system can be
state resistance mismatch and temperature rise current easily expanded by connecting more modules in parallel.
sharing problem between transistors may arise. Also there It gives obvious benefits for a customer – instead of
are some applications, e.g. UPS, solar inverters or drive replacing whole system, it’s necessary to add more
inverters, in which basically the same converter topology modules. However paralleling of many modules creates
is used for different power levels. Often in such case a new challenges. It may lead to unequal load sharing [3-7]
basic converter is scaled up or down. This approach or system instability [8-10].
requires additional design work, more complex This paper investigates possibility of improving the
production, costly testing facilities and so on. One of efficiency by intelligent usage of a modular boost
disadvantages of a single converter is rather narrow range converter in a high voltage gain application.
where high efficiency is achieved – a converter can be
optimized for one operating point only. Optimization II. SCALING AND PARALLELING OF A BOOST CONVERTER
process is more complex if input voltage varies in a wide
range, e.g. fuel cell applications. Fig. 1 presents a parallel modular converter (a) and a
boost converter with paralleled semiconductors (b). Input
A paralleling of whole dc-dc converters may be an and output voltages of both converters are the same and
alternative for paralleling or scaling of components inside the difference is only in rated power of each one. The
a converter. This approach provides several, well known output power of a single module is limited by a current
advantages. It includes expandability, higher reliability, rating of used transistor – ideally a single module should
easy maintenance and cheap production due to design consist only of one transistor and one diode – a device
standardization and mass production. Additionally, paralleling should be avoided. To increase power range
parallel modular converter provides more flexibility for a several modules are paralleled and share the load.
system designer and a customer. A system designer can
132
Contrary to that approach Fig. 1 b presents a boost References [11, 12] give a guidance about core scaling.
converter with paralleled and scaled devices. This Now, let’s assume that the scaled boost converter has to
converter consists of several paralleled transistors and process twice more power for the same input and output
diodes. Higher power level is achieved by redesigning the voltages and the same switching frequency. Thus the
converter. inductor dc current Idc has to be twice larger. Assuming
the same relative ripples Ipp% the peak current Imax will be
A. Power Semiconductors twice larger too (3), but required inductance will be only
In high power and high current applications the main half according to (4).
switch is chosen to have sufficiently large current rating.
However if such switch is not available or it’s § I pp% ·
performance is not satisfactory then often practice is to I max = I dc ⋅ ¨ 1 + ¸ (3)
© 2 ¹
parallel few transistors, split current between them and
reduce total on-state losses. Paralleling of two or three
transistors usually is relatively easy, but putting more Vin ⋅ D
devices in parallel may become a challenge – especially in L= (4)
f s ⋅ I dc ⋅ I pp%
medium/high frequency applications. One, but not the
only one, difficulty is to ensure proper layout which By substituting new Idc, Imax and L values to (1) one will
provides equal current paths for all transistors – the same find required area product Ap which is exactly twice larger
stray path length, the same stray inductance etc. During a than original one. Extending previous assumptions with
transistor turn-on period energy is stored in the parasitic constant peak induction Bmax and constant current density
stray inductance (leads, connections etc.). When the J it’s possible to find required number of turns and dc
transistor is being turn-off this energy is transferred copper losses of the scaled inductor using (5) and (6)
between transistor’s capacitance and the stray inductance respectively. Mean length per turn growth rule is given by
causing oscillations. If there is an exact symmetry in the (7).
circuit each transistor absorbs the same amount of energy
and experience the same oscillations. But in practice there L ⋅ I max
are small differences e.g. due to component value n= (5)
tolerance or temperature difference. As it is stated in [2] Bmax ⋅ Ac
such parameter differences may lead to unexpected
oscillations in paralleled MOSFETs. In order to reduce n ⋅ MLT ⋅ ρ
oscillation different snubber circuits are used, but it Pdc = I dc2 ⋅ (6)
ACu
increases component count and overall converter
complexity. Handling of a high current by a switch
MLT = K MLT ⋅ Ap 4
1
composed of many devices can push the switching (7)
frequency down and result in more bulky passive
components. It’s found that a scaled inductor should have 0.707
Also equal current sharing among several transistors is number of turns of the original one and dc copper loss will
an important issue. As long as MOSFET transistors have increase by 68% only (under the assumption of optimum
a positive temperature coefficient they have an inherent design and Pfe=Pcu [11]). However under the same
self-balancing capability – up to some extend. Reference assumption of optimum design the current density has to
[1] analyzes these problems in details and introduces a decrease when the inductor is scaled up according to (9).
dynamic current sharing from a gate side as a solution for
surface = K s ⋅ Ap 2
1
a current sharing problem during on-state and switching (8)
periods. However this method bases on gate current
control and may require additional components for a Kj
control circuit. J= 1
(9)
Ap 8
B. Growth of Passive Components
Not only transistor have to be rated for high current. In the real world such ideal scaling is very rare, because
Also all passive components have to be sufficiently large. of limited number of different core sizes. In many cases it
In the literature one can find at least two approach for results in a scaled component which is too bulky or not
power inductor core size selection – one bases on core optimum.
geometry factor Kg [11] and the other approach uses so Finally the output capacitor C1 (Fig. 1) has to be
called area product Ap [11, 12]. Area product for an sufficiently larger if the higher power has to be processed.
inductor core is given by (1). Area product of a particular Eq.(10) gives a required capacitance. It can be seen that
core can be found by multiplication core cross section are for constant output voltage Vout, voltage peak-peak ripples
a and window area (2). Vpp%, duty cycle D and switching frequency fs the output
filter capacitance C1 is a directly proportional to the
L ⋅ I dc ⋅ I max average output current Iout.
Ap = [cm 4 ] (1)
Bmax ⋅ J ⋅ K u
I out ⋅ D
C1 = (10)
Vout ⋅ Vpp% ⋅ f s
Ap = Ac ⋅ Wa (2)
Above assumptions will result in sufficiently larger
output capacitor and greater capacitor rms current as well.
133
TABLE I – MODULE INPUT CURRENT DEVIATION UNDER NON-IDENTICAL

MODULES OPERATION
2 modules 4 modules 8 modules

Rs -20% +4.7% +7.1% +8.4%
LM -20% 0.0% 0.0% 0.0%
RDS -20% +5.2% +7.9% +9.3%
Vf -20% +1.5% +2.2% +2.6%
Rf -20% +0.8% +1.1% +1.3%
Worst case +13.0% +20.4% +24.4%
Duty +0.1% +32.1% +40.4% +53.5%
TABLE II – TRANSISTOR RMS CURRENT DEVIATION UNDER NON-

IDENTICAL MODULES OPERATION
2 modules 4 modules 8 modules

Rs -20% +4.4% +6.7% +7.9%
LM -20% +1.7% +1.7% +1.7% Fig. 2 - 4-phase modular boost converter
RDS -20% +4.8% +7.4% +8.8%
Vf -20% +1.4% +2.1% +2.4%
Rf -20% +0.7% +1.1% +1.2%
Worst case +12.2% +19.2% +23.0%
Duty +0.1% +30.4% +44.0% +50.9%
III. MODULAR POWER CONVERTER APPROACH

Paralleling of dc-dc converters is a well know technique
for extending an output power and may be an interesting
alternative for a large or scaled up converters with
multiple paralleled transistors. This section gives an
overview on few common issues and challenges related to
paralleling of dc-dc converters.
A. Current Sharing
The most recognized is current sharing problem. If the
single dc-dc module has relatively low power it is possible Fig. 3 - Scaled boost converter
to relay on a simple droop method. Also most of components values are selected to fit a breadboard
paralleled converters has an inherent self-balancing specification described in Section V.
functionality – up to some extent. If one of modules
First, simulation of 2, 4 and 8 paralleled identical
conducts higher current it heats up, so it’s resistances
increase and unbalance current is limited in this way. But modules (Fig. 1 a) operating in open loop configuration
were performed. Results confirm stable operation and
if safety limits are lower or a single module has higher
power then active current sharing method is required. equal current sharing among all modules.
Different paralleling and current sharing methods are Second step is to simulate behavior of non-identical
presented in [3]. This paper describes several passive and modules. It’s done by changing component values in one
active methods, including Master-Slave and Average of modules. Only one parameter at the time was changed
Current Programming Methods. More details about active in order to find out which parameter deviation introduces
current sharing methods one may find in [4-7]. the largest current unbalance. Following parameters
(component values) were decreased by 20%: inductor
B. Instability of Paralleled Dc-Dc Converters winding resistance Rs, inductance LM, transistor on-state
Another problem caused by paralleled dc-dc converters resistance RDS, diode forward voltage Vf and diode
is potential system instability. The one reason for this is forward resistance Rf.
that dynamic of the parallel system is different that Next, based on previous observations the ‘worst case’
dynamic characteristic of a single converter. It means that scenario was simulated – all components values are
a controller tuned for a single module may not work increased or decreased by 20% to create the largest
properly with paralleled converters leading to undesired unbalance.
system oscillations. The other reason for instability is Finally, duty cycle value was increased by +0.1% in
presence of multiple current control loops which are one of modules to create unbalanced conditions.
introduced by some of current control sharing methods. Simulation results are summarized in Table I and Table
This problem is widely discussed in [8-10]. II, where module input current deviation and transistor
rms current deviation are presented. It has been found that
IV. SIMULATION OF NON-IDENTICAL MODULES PARALLEL the inductor series resistance and transistor on-state
OPERATION resistance have big influence on current unbalance. Both
Due to potential current unbalance problem several resistances are located on low voltage and high current
simulations of non-identical modules are done using side of the converter. Also, the same disturbance results
Matlab®/PLECS® software. Model parameters and in larger unbalance when more modules are paralleled.
However, the modular converter like this seems to be very
134
by MOSFET’s conduction loss at the lowest input voltage.

To achieve higher power level more identical modules are
connected in parallel to the first one. Each transistor has
its own gate driver and it’s controlled by an individual
signal. A single Infineon XC167 microcontroller is used
for PWM generation. All PWM signals have the same
duty cycle and they are equally interleaved – phase shift
between PWM signals depends on number of operating
modules and it’s adjusted continuously.
B. Description of the Scaled Converter
The second converter is scaled up to 400 W version of a
single 200 W boost module (Fig. 1 b, Fig. 3). It is
designed to process twice larger power like the single
module and achieve the same efficiency. This converter
consist two paralleled MOSFETs and two SiC diodes –
the same type like in the modular converter described
above. Each transistor has its own gate driver, but PWM
signal is common for both transistors. Both MOSFETs
operate synchronously, so effective on-state resistance
RDSon is half of a single MOSFET resistance. Also it
enables to process twice larger current. Due to higher
current the inductor has to be redesign and scaled up –
Fig. 4 - Efficiency of a single boost module (m1), two modules in substituting new current and inductance values to (1) it’s
parallel (m2) and the scaled converter (LB) at 30 V input voltage found that new core should provide twice larger area
product value. There are at least three ways to achieve
sensitive to duty cycle unbalance – even a very small higher power handling capability of the inductor: 1) put
deviation in duty cycle may result in huge current two smaller inductors in parallel; 2) redesign an inductor
unbalance. So, the special attention should be given to using two stacked smaller cores; 3) redesign an inductor
PWM gating signals and gate drivers’ design. using one larger core. This converter has the storage
inductor built on two stacked toroid cores (Magnetics,
High Flux 58071). Based on area product approach it’s
V. EXPERIMENTAL SETUP
found that such stacked cores have twice larger power
Two demonstrator converters are build and tested in handling capability (the same window area Wa, twice
order to compare their performance and efficiency. larger core cross section area Ac) and it fit’s to
requirements. Number of turns is decreased by half (34
A. Description of the Modular Converter turns) and, due to doubled core cross section area AC,
The first one is a modular boost converter built-up from inductance of the inductor is decreased by half too.
four paralleled boost converter modules (Fig. 1 a, Fig. 2). Specific core losses given by Steinmetz equation shall
Desired input voltage range is 30-50 Vdc and output remain on the same level, but total core loss will be twice
voltage is regulated at 350 Vdc. A single 600 V MOSFET larger due to larger core volume. The winding wire size is
(Fairchild, FCH47N60) and a single 600 V SiC diode selected in respect to required winding dc resistance,
(Infineon, SDT05S60) are used. The storage inductor which should be half of dc resistance of the inductor in
bases on a toroid powder core (Magnetics, High Flux modular converter. For twice larger current it gives twice
58071) and it has 68 turns winding. Maximum output larger copper loss.
power of a single module is 200 W and in fact it’s limited Using this design approach, the converter is expected to
Fig. 5 - Input current (Ch1) and Fig. 6 - Module input currents (Ch1 and Ch2), converter input
transistor drain-source voltage (Ch3, Ch4) current (Math) and transistor drain-source voltages (Ch3 and Ch4)
135
Fig. 7 - Efficiency of a modular converter at 30 V input voltage
have twice larger maximum power and similar efficiency

to the single converter module described previously.
VI. EXPERIMENTAL RESULTS Fig. 8 - Efficiency of a modular converter at variable input voltage
At the beginning three converter configurations were efficiency of the whole system. One may observe that
compared in terms of achieved efficiency. At an input efficiency of such modular system increases while output
voltage of 30 V the single module (curve m1) efficiency power decreases – this behavior is opposite to most of
was measured in the power range from 50 to 200 W. Next traditional converters, where efficiency drops at low load.
efficiency of a scaled 400 W boost converter (curve LB)
was measured and compared with two paralleled modules VII. CONCLUSION
(curve m2) under the same conditions (input voltage 30 V,
output power 100-400 W). Results are presented on Fig. This paper gives an overview on scaling and paralleling
4. One may observe that efficiency curves m1, m2 and of boost converters. Two demonstrator converters are
LB are very similar –it means that the scaling process is built, tested and results are presented. Based on these
done correctly. Also, it confirms that paralleling of two results it's stated that modular converter gives more
switches shall not affect the efficiency significantly. flexibility and provides additional degree of freedom for a
system designer. Adjusting number of operating modules
Fig. 5 and Fig. 6 present input currents and transistor
leads to optimum utilization of the converter. In case of a
voltages of a scaled 400 W converter and two interleaved
fuel cell application, with variable input voltage, the
modules respectively. One may find that input current
parallel converter provides high efficiency over a wide
ripples are significantly lower for interleaved modular
range of output power levels.
system. Moreover, output voltage ripples have a twice
larger frequency, and it results in smaller output capacitor.
Also it’s should be emphasized that the parallel converter
operates in open loop configuration and current sharing is ACKNOWLEDGMENT
very good in spite of absence of current controller. This project is financed by Dansk Energi net under PSO
Next the input voltage is set to 30 V and efficiency was project no. 562/06-14-26808.
measured for different number of modules (up to four
modules in parallel) for different power levels. Results REFERENCES
are presented on Fig. 7. By adjusting number of operating
[1] W. Hongfang and F. Wang, "Power MOSFETs Paralleling
modules it’s possible to keep high efficiency over a wide
range of output power. By adding more modules it’s easy Operation for High Power High Density Converters," in Industry
to extend power level of the system. Applications Conference, 2006. 41st IAS Annual Meeting.
Finally, measurements under variable input voltage Conference Record of the 2006 IEEE, 2006, pp. 2284-2289.
were done. Variable input voltage simulates behavior of a [2] J. G. Kassakian and D. Lau, "An analysis and experimental
PEM fuel cell – output voltage of the stack is highest at verification of parasitic oscillations in parralleled power
no-load condition and decreases while output power MOSFET's," Electron Devices, IEEE Transactions on, vol. 31, pp.
increases. In the lab setup input voltage was set to 50 V 959-963, 1984.
for 100 W and decreased linearly to 30 V at 800 W output [3] L. Shiguo, Y. Zhihong, L. Ray-Lee, and F. C. Lee, "A
power. Under such conditions the efficiency was
classification and evaluation of paralleling methods for power
measured for different number of modules. Results are
supply modules," in Power Electronics Specialists Conference,
presented on Fig. 8 together with input voltage value.
One may observe that efficiency of four modules in 1999. PESC 99. 30th Annual IEEE, 1999, pp. 901-908 vol.2.
parallel (curve m4) drops quickly while output power [4] T. Kohama, T. Ninomiya, M. Shoyama, and F. Ihara, "Dynamic
decreases, e.g. down to 94% at 340 W. In this case it’s analysis of parallel-module converter system with current balance
beneficial to turn-off one or two modules and increase
136
controllers," in Telecommunications Energy Conference, 1994.

INTELEC '94., 16th International, 1994, pp. 190-195.
[5] D. J. Perreault, R. L. Selders, and J. G. Kassakian, "Frequency-
based current-sharing techniques for paralleled power converters,"
in Power Electronics Specialists Conference, 1996. PESC '96
Record., 27th Annual IEEE, 1996, pp. 1073-1079 vol.2.
[6] N. Hur and N. Kwanghee, "A robust load-sharing control scheme
for parallel-connected multisystems," Industrial Electronics, IEEE
Transactions on, vol. 47, pp. 871-879, 2000.
[7] D. J. Perreault, K. Sato, R. L. Selders, Jr., and J. G. Kassakian,
"Switching-ripple-based current sharing for paralleled power
converters," Circuits and Systems I: Fundamental Theory and
Applications, IEEE Transactions on [see also Circuits and Systems
I: Regular Papers, IEEE Transactions on], vol. 46, pp. 1264-1274,
1999.
[8] H. H. C. Iu and V. Pjevalica, "Experimental study of instabilities
in two parallel-connected boost converters under current mode
control," in Power Electronics and Applications, 2005 European
Conference on, 2005, p. 8 pp.
[9] S. K. Mazumder, "Stability analysis of parallel DC-DC
converters," Aerospace and Electronic Systems, IEEE
[10] J. M. Zhang, X. G. Xie, X. K. Wu, and Q. Zhaoming, "Stability
study for paralleled DC/DC converters," in Power Electronics
Specialists Conference, 2004. PESC 04. 2004 IEEE 35th Annual,
2004, pp. 1569-1575 Vol.2.
[11] C. Wm.T.Mclyman, Transformer and Inductor Design Handbook,
Third Edition. New York: Routledge, 2004.
[12] P. Wallmeier, "Pre-optimization of linear and nonlinear inductors
using area-product formulation," in Industry Applications
Conference, 2002. 37th IAS Annual Meeting. Conference Record
of the, 2002, pp. 2445-2450 vol.4.
137
A.3 Boost Converter with Three-Stage Switching

Cell and Integrated Magnetics (APEC 2009)
Boost Converter with Three-State Switching Cell

and Integrated Magnetics
Pontoppidanstraede 101 Pontoppidanstraede 101
9220 Aalborg DENMARK 9220 Aalborg DENMARK
Abstract – Fuel cell systems often require high voltage gain

and dc-dc step-up converter is a critical part. Scope of this
paper is integration of inductor and transformer on a single is D3 C3
iout
core. Usage of integrated magnetics improves utilization of
magnetic core and thus size and weight of the converter may be ns D2 C2 +
iin=iM LM ip11 np11
D11
reduced. R0 vout
-
I. INTRODUCTION + Vin ip1 np1
- D1 C1
An increasing interest in alternative and renewable energy T11 T1
creates a demand for novel power conversion systems, which
interfaces different energy source to a load or to a utility grid. Fig. 1 - Diagram of a boost converter
Very often it’s necessary to boost a low voltage from a source with a three state switching cell and a voltage multiplier
(small fuel cells, most of photovoltaic panels) to a high
voltage suitable for an inverter. The required voltage gain II. ANALYSIS OF THE CONVERTER
may exceed value of 20 in many cases. Such gain is
attainable for boost converter with SiC diodes and CoolMOS The selected topology is presented on Fig. 1. It
transistors, but by a cost of poor semiconductors utilization incorporates so called 3-state switching cell (T1, T11, D1, D11
and limited efficiency [1]. In the literature [2-6] one may find and the transformer) [7, 10]. Explanation of the converter’s
several topologies with a coupled inductor, which offer a high operation maybe more intuitive if one considers this topology
voltage gain. The main drawback among these topologies is as a modified current fed push-pull converter – with diodes
large input current ripple and such ripples may lead to PV D1, D11 and capacitor C1 as a lossless clamp for transistors.
maximum power point tracking mismatch or to quick Key waveforms are presented on Fig. 2. Basically the
degradation of a fuel cell. For these reasons an additional converter operates with transistor duty cycle larger than 50 %,
input filter is required, but it increases size, cost and i.e. overlapping mode. Assuming continuous conduction
complexity of the converter. The topology introduced in [7] mode (CCM) there are four operating modes. Mode 1 and 3
and presented on Fig. 1 offers a high voltage gain, reduced are identical while both transistors are turned-on and share the
voltage rating of MOSFETs and input current ripple is limited current equally (np1=np11). During this period input inductor
by inductor LM. Reference [8] reports a very good VGS
performance of this topology in a high voltage gain t
application, like a fuel cell. However in such application the
Iin, IM
converter has to operate under variable, load dependent input
voltage. Typically for fuel cell minimum to maximum input
voltage ratio is about 1:2 [9]. While the converter operates
t
under different conditions it might be difficult to optimize
magnetic components. On one hand they are sized for high Ip1
input current. One the other hand they are designed for the
highest input voltage too. But in fuel cell application high t
D1 T1
voltage and high current doesn’t occur at the same time. It Ip11
may result in poor utilization of magnetic cores of the
inductor and the transformer. T11 D11 t
The main focus in this paper is to design and implement of
Is t
an integrated magnetic, which should reduce size and weight
of the converter.
Fig. 2 - Main waveforms of the converter
138
LM is being charged. Because both primary windings are dc current with relatively small ac component (ripples).
shorted there is no voltage across secondary winging, so there Majority of dc current content allows using of a round wire
is no current flow in the secondary winding too. Next, in for inductor winding. It characterizes relatively low cost and
mode 2 transistor T1 is being turned-off and energy stored in good copper fill factor Ku. Since dc current is dominant a low
LM is being released thru primary winding (np1) and diode D1 dc resistance is a key. In order to improve copper fill factor
to capacitor C1, and thru secondary winding and diode D3 to one may use two or three thinner wires instead of thick single
capacitor C3. The last mode 4 is symmetrical to mode 2 – wire. But this approach makes the winding more complex
while transistor T1 remains turned-on and T11 is turned-off. and expensive. If required number of turns is relatively small
During this period diode D11 and diode D2 conduct. one may incorporate a foil winding, which provides excellent
Capacitors C1 and C2 are recharged. copper fill factor. There are two main drawbacks of an
It’s important that capacitor C1 is recharged twice during inductor foil windings, especially when ac current component
the switching period, while capacitors C2 and C3 are increases. First, it’s a multilayer design and one turn means
recharged only once. Static voltage gain of the converter is one layer so proximity losses become significant very
given by (1), where n is primary to secondary turns ratio. quickly. The solution is a design with as few turns as
Voltage stress is given by (2) for primary side semiconductors possible. Second significant drawback relates to an air gap in
and by (3) for secondary side diodes D2 and D3. a ferrite core. Because of a fringing field huge eddy currents
1 n induces in inner layers of the winding. It creates significant
Vout Vin (1) losses and risk of hot spot. Using of a distributed air gap or a
1 D
powder core helps to solve this problem. If a gapped ferrite
1 core is the only option then special winding arrangement is
VT1 VT11 VD1 VD11 Vout (2)
1 n necessary. It’s simply done by notching inner layers around
n the air gap, so fringing field doesn’t penetrate the winding
VD2 VD3 Vout (3) [11, 12]. Example of such winding is shown on Fig. 3.
n 1 A design procedure based on area-product approach is well
III. MAGNETIC COMPONENTS described in [13], so here only few most important steps are
presented. First, required inductance LM is to be determined
Magnetic components are essential for any dc-dc converter
and for considered converter it’s given by (4), where ¨IM is
design. Their performance reflects directly on converter’s
amplitude of current ripples. By rearranging (4) it’s found
performance. So, typical design of magnetics focuses on
that the largest current ripples occur at duty cycle D=0.75.
reduction of losses inside magnetics and on reduction of
parasitics, so external circuit is not affected. But on the other Vout D 0.5 1 D
LM (4)
hand the magnetics are significant part of the converter size f s 2 'I M 1 n
and weight, so size reduction should be taken into
consideration too. In many cases integration of magnetics Assuming that the inductor operates in linear region, below
may improve utilization of a core and thus size may be saturation it’s easy to find dc and ac induction components
reduced. (Bind and 'Bind) while they are directly proportional to
inductor current dc and ac components (IM(dc) and 'IM).
A. Dc inductor Core size is characterized by area product Ap, which is
In considered topology dc inductor LM conducts essentially product of core cross section area Ac and window area Wa
[13]. Required area product value is given by (5). Current
IM(dc) is maximum inductor dc current, i.e. at minimum input
voltage and rated power. When the dimensions of the core
are know one may find required number of turns (6) and
length of the air gap (7).
Ap

LM I M dc I M dc 'I M (5)
Bind 'Bind J K u
LM 'I M
nM (6)
'Bind Ac
P0 nM2 Ac
lg (7)
LM
For small air gaps effective area of the gap Ag is close to
Fig. 3 - Notched foil winding with low dc and ac
resistance arranged on EE ferrite core core cross section area Ac, but larger gaps effective area of the
gap is larger and have to be corrected [13].
139
Ip1 IM (dc) IM (ac) Reflected Is a) b)
np1·Ip1 +
-
Rg -
t Rc + Rc
np11·Ip11
Ip11
nM·IM +
-
ns·Is +
-
t
Fig. 5 - Magnetic circuits of an inductor (a)
Fig. 4 - Primary currents and its components and a push-pull transformer (b)
B. Transformer
Fig. 4 presents primary currents and its components: input Rc Rc -
np1a·Ip1 + + np1b·Ip1
dc current, input current ripples and reflected secondary -
- Rg
current. This combination of ac and dc components creates + + np11b·Ip11
np11a·Ip11 -
challenge for design of transformer windings. Reduction of
nM·IM +
- -
both, dc and ac resistances of primary windings, is crucial.
nsa·Is +
-
+ nsb·Is
Usage of Litz wire helps to reduce skin and proximity effects.
Unfortunately in the same time Litz wire has poor copper fill
factor Ku, so dc resistance of such winding is high. Also Fig. 6 - Magnetic circuit of the integrated inductor-transformer
leakage inductance of such winding is high affecting the
external circuit. Round wire has better copper fill factor and
dc resistance is reduced in compare to Litz wire. But due to 1 Vout 1 D
np1 np11 (11)
skin effect ac loss is fairly high. 2 'Btr Ac 2 1 n fs
Foil winding seems to be a good solution. It has a very
good copper fill factor and resulting dc resistance is low. It’s found that the largest induction amplitude occurs at
Small thickness of the foil reduces skin effect, while duty cycle D=0.5, i.e. for the highest input voltage.
interleaving of primary and secondary windings attenuates C. Integrated magnetics
proximity effect [14]. Fig. 5a presents magnetic circuits of an inductor. Magneto-
Design procedure based on area product Ap is described in motive force (MMF) is represented as voltage source nM·IM,
details in [13]. First, size of the core is calculated from (8). while the air gap reluctance is represented by linear resistor
Variable Pt is transformer apparent power, i.e. sum of input Rg. Resistor Rc represents reluctance of the ferrite core and
and output power of the transformer. For push-pull it’s very small in compare to reluctance of the air gap
transformer it’s given by (9). (Rc<<Rg).
Pt Fig. 5b shows magnetic circuit of a push-pull transformer.
Ap (8) There are three voltage sources representing MMF of
2 K u K f 'Btr J f s
windings. MMFs directions refer to the converter diagram on
Pt 1.41 Pin Pout (9) Fig. 1. Reluctance of the core is represented by resistor Rc.
Induction amplitude ¨Btr may not exceed saturation level of These two circuits can be combined together on a single EE
the core. However in many cases peak induction in the core. The simplest way is to place the inductor winding on
transformer is well below saturation because of maximum the center gapped leg. Resulting magnetic circuit is shown on
allowed core losses. Specific core loss Pfe as a function of Fig. 6. Transformer windings are split into two identical
induction amplitude and switching frequency is given by well halves and placed on outer legs. In this way flux generated
known Steinmetz equation (10). In considered converter by inductor winding does not affect transformer windings.
primary winding number of turns is given by (11), which is Also flux generated by the transformer has no influence on
true for D=(0.5, 1). the inductor. Unfortunately it results in complex winding
arrangement, large overall number of turns and large winding
Pfe k f sD 'BtrE (10)
resistance.
140
10
- Rc Rc -
np1·Ip1 + + np11·Ip11 A
p int
Rg 8 A
p int ind
Ap int tr
nM·IM +
- - 6
Ap [cm ]
4
nsa·Is +
-
+ nsb·Is
4
Fig. 7 - Reduced magnetic circuit of the integrated inductor-
transformer 2
Analyzing primary side current waveforms presented on

0
Fig. 4 one may find that input current flows thru transformer 0.1 0.15 0.2 0.25 0.3 0.35 0.4
primary windings. Since there is agreement between current Bind [T]
waveforms it’s possible to use transformer primary windings Fig. 8 - Integrated inductor-transformer required area product versus
as a part of the inductor winding. For proper operation of the inductor flux
transformer the secondary winding is split and arranged on
both outer legs, so it’s not affected by inductor flux. transformer and by the inductor (17). One can change ratio
Proposed magnetic circuit is presented on Fig. 7 and currents between the inductor and the transformer flux and find
directions refers to Fig. 1. The magnetic circuit from Fig. 7 is minimum required core size. Fig. 8 presents how required
described by (12), where LM is self inductance of nM winding, area product depends on inductor flux. This method gives
MMp1 is mutual inductance between nM and np1 windings and only approximated core size, but it’s a good starting point for
so on. further optimization.
Design of the integrated magnetic is an iterative process. Pt
When both transistors are turned-on the input voltage is Ap
2 K u K f 'Btr J fs
applied to inductor and transformer’s primary windings, so (17)
(13) is truth. Assuming that secondary current is is zero and

LMeq I M dc I M dc 'I M
both primary currents are equal ip1=ip11 one can write (14) and
(15). Combining these equations with (13) gives equivalent
Bind 'Bind J K u
inductance LMeq (16). With good coupling between both Once the core size is determined inductor and transformer
primary windings mutual inductance Mp1p11 value is close to windings are designed according to guidelines presented in
self inductance Lp1 value, so they will cancel each other. previous sections. First, number of turns in transformer
Equivalent inductance LMeq is larger then physical inductance winding is find with (11). Next, in an iterative process one
LM, so in this way required inductance LM is reduced. can find self inductances and mutual inductances (12), find
diM required number of turns nM (6) and size of the air gap lg (7).
ª vM º ª LM M Mp1 M Mp11 M Msa M Msb º ª dt º Contrary to typical transformer, the integrated magnetic
« » « dip1 »
« »
M p1sb » « dt »
« vp1 » « M p1M Lp1 M p1p11 M p1sa usually operates under significant dc flux conditions. It has
«v » « » two major effects. First, because of dc flux in the core
« M p11M M p11p1 Lp11 M p11sa M p11sb » « dip11 »
« p11 » « » dt increased core losses may occur. Unfortunately core
« vsa » « M saM M sap1 M sap11 Lsa M sasb » « dis »
« dt » datasheets contain power loss data only for pure sinusoidal
« » « M sbM M sbp1 M sbp11 M sbsa Lsb »¼ « dis »
¬ vsb ¼ ¬
¬ dt ¼
excitation, so it’s difficult to calculate core losses in advance.
Second, the peak induction is higher in compare to traditional
(12) transformer. Thus there is no space for any undesired dc flux.
diM It requires careful control of dc current flowing in primary
Vin vM vp1 LMeq (13)
dt windings np1 and np11.
diM dip1 dip11 IV. EXPERIMENTAL SETUP AND RESULTS
v M LM M Mp1 M Mp11 (14)
dt dt dt In order to verify theoretical analysis and calculations a 300
dip1 dip11 W breadboard is built and tested. Input voltage varies from
diM
vp1 M p1M Lp1 M p1p11 (15) 30 to 50 Vdc and output voltage is regulated at 400 Vdc.
dt dt dt Switching frequency is fs=50 kHz and the transformer turns
1 1 ratio is n=2. According to (2) it allows using 150 V
LMeq LM M
M p1M Lp1 M p1p11 (16)
2 2
Mp1 MOSFETs (IRFP4321PbF) and 150 V Schottky diodes
(30CPQ150PbF) on primary side. On the secondary side of
Approximated core size is found using area product
the transformer there are 600 V SiC diodes (IDT10S60C).
approach simply by summarizing area product required by the
141
Fig. 9 - Assembled converter with integrated inductor-transformer Fig. 11 - Input current Iin (bottom)
and its spectrum (top) in balanced case
Fig. 10 - Main waveforms of the converter Fig. 12 - Input current Iin (bottom)
Ch1 - gating signal; Ch2 - VDS; Ch3 - Vn2; Ch4 - Iin and its spectrum (top) in unbalanced, saturated case
It’s found that E42/21 core (0P44022EC) is sufficiently

98
large. With nM=7 turns and 1 mm air gap in the center leg it
gives inductance LM=16µH. Transformer induction 97.5
amplitude is limited to 'Btr=0.15 T at the highest input 97
voltage and it’s limited by core losses. Transformer primary
Efficiency [%]
96.5
winding has np1=np11=12 turns, which results in equivalent
inductance LMeq=46µH. The secondary winding has ns=24 96 V =30 V
in
turns and it’s divided into to halves, nsa=nsb=12 turns on each 95.5
of outer legs. Primary and secondary windings are
95 Vin=45 V
interleaved to reduce proximity effect and ac losses.
Assembled 300 W breadboard is presented on Fig. 9. Fig. 10 94.5
presents basic operational waveforms of the converter at 30 V
94
input voltage. 0 50 100 150 200 250 300 350
Control of primary windings dc currents is critical to avoid Pout [W]
outer legs saturation. In case of integrated inductor- Fig. 13 - Measured efficiency of the converter
transformer it’s possible to detect such saturation by
observation of the input current only. Fig. 11 shows the input input current changes. Now, 50 kHz component appears in
current and its spectrum in an unsaturated and balanced case. the current spectrum presented on Fig. 12.
The input current has only a large dc component and 100 kHz Fig. 13 presents measured efficiency of the converter at 30
component, which refers to ripples at double switching V and 45 V input voltage. For lower input voltage efficiency
frequency. In case of small unbalance of transformer primary
currents one of outer legs become saturated and shape of the
142
peaks at 97% at 175 W and remains above 96.5% up to 300 Power Electronics Specialists Conference, 2000. PESC 00. 2000 IEEE
W. For higher input voltage efficiency exceeds 97%. 31st Annual, 2000, pp. 858-863 vol.2.
[11] J. D. Pollock and C. R. Sullivan, "Gapped-inductor foil windings with
V. CONCLUSION
low AC and DC resistance," in Industry Applications Conference, 2004.
A boost converter with a three state switching cell and 39th IAS Annual Meeting. Conference Record of the 2004 IEEE, 2004,
integrated magnetic has been presented in this paper. The p. 557.
converter has the inductor and the push-pull transformer [12] J. D. Pollock and C. R. Sullivan, "Modelling Foil Winding
arranged on a standard gapped EE core. Such integration Configurations with Low AC and DC Resistance," in Power
allows using of transformer primary windings as a part of the Electronics Specialists Conference, 2005. PESC '05. IEEE 36th, 2005,
inductor winding, thus smaller physical inductance is pp. 1507-1512.
required. It results in smaller and lighter core. To verify the [13] C. W. T. McLyman, Transformer and inductor design handbook:
idea 300 W breadboard is built and tested. The breadboard Marcel Dekker, 2004.
exhibits good performance. Also size of the magnetic [14] M. Nymand and M. A. E. Andersen, "A new approach to high
components is reduced. efficiency in isolated boost converters for high-power low-voltage fuel
ACKNOWLEDGEMENT cell applications," in Power Electronics and Motion Control
Conference, 2008. EPE-PEMC 2008. 13th, 2008, pp. 127-131.
This project is supported by Dansk Energi net under PSO [15] S. Cuk, "New magnetic structures for switching converters," Magnetics,
project no. 562/06-14-26808. IEEE Transactions on, vol. 19, pp. 75-83, 1983.
REFERENCES
[1] P. Klimczak and S. Munk-Nielsen, "Comparative study on paralleled
vs. scaled dc-dc converters in high voltage gain applications," in Power
Electronics and Motion Control Conference, 2008. EPE-PEMC 2008.
13th, 2008, pp. 108-113.
[2] B. Ju-Won, R. Myung-Hyo, K. Tae-Jin, Y. Dong-Wook, and K. Jong-
Soo, "High boost converter using voltage multiplier," in Industrial
Electronics Society, 2005. IECON 2005. 31st Annual Conference of
IEEE, 2005, p. 6 pp.
[3] Z. Qun and F. C. Lee, "High-efficiency, high step-up DC-DC
converters," Power Electronics, IEEE Transactions on, vol. 18, pp. 65-
73, 2003.
[4] W. Rong-Jong and D. Rou-Yong, "High step-up converter with
coupled-inductor," Power Electronics, IEEE Transactions on, vol. 20,
pp. 1025-1035, 2005.
[5] K. C. Tseng and T. J. Liang, "Novel high-efficiency step-up converter,"
Electric Power Applications, IEE Proceedings -, vol. 151, pp. 182-190,
2004.
[6] R. J. Wai and R. Y. Duan, "High-efficiency DC/DC converter with high
voltage gain," Electric Power Applications, IEE Proceedings -, vol.
152, pp. 793-802, 2005.
[7] G. V. Torrico-Bascope, R. P. Torrico-Bascope, D. S. Oliviera, F. L. M.
Antunes, S. V. Araujo, and C. G. C. Branco, "A Generalized High
Voltage Gain Boost Converter Based on Three-State Switching Cell,"
in The 32nd Annual Conference of the IEEE Industrial Electronics
Society IECON'06, Paris, France, 2006, pp. 1927-1932.
[8] P. Zacharias, B. Sahan, S. V. Araújo, F. L. M. Antunes, and R. T.
Bascopé, "Analysis and Proposition of a PV Module Integrated
Converter with High Voltage Gain Capability in a non-isolated
Topology," in The 7th International Conference on Power Electronics
Daegu, Korea: IEEE, 2007.
[9] W. Vielstich, Handbook of fuel cells fundamentals, technology, and
applications editors, Wolf Vielstich, Arnold Lamm, Hubert A.
Gasteiger: Wiley, 2003.
[10] G. V. T. Bascope and I. Barbi, "Generation of a family of non-isolated
DC-DC PWM converters using new three-state switching cells," in
143
A.4 High Efficiency Boost Converter with Three

State Switching Cell (PCIM 2009)
High Efficiency Boost Converter with Three State Switching Cell

Pawel Klimczak, Aalborg University, Denmark
Stig Munk-Nielsen, Aalborg University, Denmark
Abstract
The boost converter with the three-state switching cell seems to be a good candidate for a dc-dc stage for
non-isolated generators based on alternative energy sources. It provides a high voltage gain, a reduced
voltage stress on transistors and limited input current ripples. In this paper the focus is on performance
improvement of this type of the converter. Use of foil windings helps to reduce conduction losses in
magnetic components and to reduce size of these components. Also it has been demonstrated that the
regulation range of this type of converter can be increased by operation with duty cycle lower than 50%.
1. Introduction
Increasing interest in alternative and renewable
energy creates a demand for novel power
conversion systems, which interfaces different
energy sources to a load or to an utility grid. Very
often it’s necessary to boost a low voltage from a
source (commercial fuel cells, most of photovoltaic
panels) to a high dc voltage suitable for an Fig. 1. Diagram of the three state switching
inverter. In many cases safety isolation is not cell boost converter
requires, thus it’s possible to use a non-isolated
topology. The required voltage gain may exceed good performance of this topology in high voltage
value of 20 in some cases. Such gain is attainable gain application (Vin=35 V, Vout=400V). The main
for a basic boost converter with SiC diodes and focus of this paper is to improve a power density
novel MOSFET transistors (SuperFET, CoolMOS and an efficiency of this type of converter.
etc.), but by a cost of poor semiconductors
utilization and the efficiency limited by MOSFET
conduction losses [1]. Further efficiency 2. Description of the converter
improvement requires low voltage switches which The selected topology is presented on Fig. 1. It
ensure lower conduction losses and improved incorporates so called three-state switching cell
switching performance. In the literature [2-6] one
may find several topologies based on a coupled vGS
inductor principle, which offer a high voltage gain
and reduced voltage stress on the active switch. iin=iM t
The main drawback among these topologies is
large input current ripples. On one hand significant
input current ripples may lead to faster degradation
of a fuel cell stack or disturb maximum power point t
tracking in case of PV application. An additional ip1
input filter can solve these problems, but it
increases size, complexity and cost of the D1 T1
ip11 t
converter. Also large ripples result in a large peak
current, which leads to increased conduction
losses and requires oversized components. The T11 D11
topology introduced in [7] and presented on Fig. 1 is t
offers a high voltage gain and reduced voltage t
stress on transistors. Reference [8] reports a very Fig. 2. Basic waveforms of the converter
144
(switches T1, T11, D1, D11 and the transformer) [7, end, at the very low duty cycles the converter
9]. The secondary side of the transformer with behaves like a two-phase interleaved boost with
diodes D2, D3 and capacitors C2, C3 compose the coupled inductor [11]. In this case voltage gain is
voltage doubler rectifier. Basically the converter given by (4).
operates in so called overlapping mode, i.e. with 1
the duty cycle larger than 50% and in continuous Vout Vin (4)
1 2 D
conduction mode (CCM). Under such conditions diodes D2 and D3 are
forward biased all the time so there is no voltage
2.1. Operation with duty cycle across capacitors C2 and C3. The whole the output
larger than 50% voltage appears across the capacitor C1. It may
Key waveforms for are presented on Fig. 2. There create an additional voltage stress on transistors
are four basic operating modes. Mode 1 and 3 are and diodes D1 and D11.
identical while both transistors are turned-on and The boundary between one and the other mode is
share the input current equally (np1=np11). During defined by the crossing of voltage gain functions
this period the input inductor LM is being charged. (1) and (4). By combining these two equations one
Both primary windings are effectively connected may find the critical duty cycle Dcr at the crossing
anti-parallel. In this case no voltage induces in the point (5).
secondary winging and there is no current flow in n
Dcr (5)
the secondary winding too. Next, in mode 2 the 2n 1
transistor T1 is being turned-off and energy stored
in LM is being released thru the primary winding
(np1) and the diode D1 to the capacitor C1, and thru 3. Magnetic components
the secondary winding and the diode D3 to the Magnetic components are essential for any dc-dc
capacitor C3. In this operating mode primary converter design. Their performance reflects
windings are effectively connected in series and directly on converter’s performance. Typically
the voltage VC1 is applied to them. The voltage design and optimization process focuses on loss
equal to VC1·n induces in the secondary winding. reduction inside magnetics as well as on reduction
Term n means the transformer primary-to- of parasitics, so external circuit is not affected.
secondary turns ratio. The last mode 4 is similar to Volume of magnetics is an important issue too.
mode 2 – while transistor T1 remains turned-on
and T11 is turned-off. In this mode diodes D11 and
D2 conduct. It’s important that the capacitor C1 is
3.1. Dc inductor
charged twice over one cycle, while capacitors C2 In the considered topology the dc indutor LM
and C3 are charged only once. The converter conducts essentially the dc current with limited ac
voltage gain in CCM is given by (1). The maximum component (ripples). It means that the flux ac
voltage across switches is given by (2) and (3). component is fairly low too, thus core losses are
1 n small and the design is so called saturation limited
Vout Vin (1) design [12]. To improve performance of such
1 D
design it is important to reduce dc and ac
1
VT1max VD1max VC1 Vout (2)
1 n
n
VD2max VD3max V out (3)
1 n
2.2. Operation with duty cycle

lower than 50%
Unlike a conventional current-fed push-pull
converter [10] the topology presented on Fig. 1 is
able to operate with the duty cycle lower than 50%.
However the voltage gain function becomes
strongly non-linear in this case. For duty cycles
just below 50% converter operates normally and
it’s voltage gain follows (1). Also the voltage stress
on components fits to (2) and (3). On the other Fig. 3. Foil winding arrangement on a gapped
core
145
conduction losses, but low dc resistance is crucial. Fig. 4 presents transformer primary currents and
Basically one may choose between a solid round its components: the input dc current iin(dc), the input
wire, a Litz wire and a foil winding. Use of a single current ripple iin(ac) and the reflected secondary
round wire provides a good window fill factor, a low current is’. Fig. 5 shows how these components
dc resistance and a low manufacturing cost. To flow in particular windings of the transformer. The
improve window utilization one may use few input current flows into the beginning of the
thinner wires or a rectangular wire if available. winding np11 and into the end of the winding np1 like
Another choice is a Litz wire which provides a good presented on Fig. 1 and Fig. 5. For this current
ac/dc resistance ratio, but it requires a larger component primary windings are effectively
window area because of a poor copper fill factor. connected anti-parallel. In an ideal case half of the
Finally a winding made of a copper foil may input current flows into one primary winding while
provide a very good copper fill factor (depends on the other half goes into the other. In this way
copper and insulation thickness) and a very low dc fluxes generated by these currents cancel each
resistance. However two main drawbacks are other, so there is no effect on the secondary side
recognized. First, it's a multilayer design since one of the transformer. Secondary current is flows in
turn means one layer. In this case proximity losses the secondary winding and the reflected secondary
may become large quickly. The solution is to use current is’ flows in primaries as presented on Fig.
as few turns as possible and the copper thickness 5. For this current component primary windings
smaller than the penetration depth for given are connected effectively in series.
frequency. Second drawback relates to use of a Such mixture of ac and dc components creates a
ferrite core with a discrete air gap. Because of a challenge for the design of transformer windings,
fringing field around the air gap huge eddy currents especially primaries. Reduction of dc and ac
induce in inner layers of the winding. It creates resistances of primary windings is crucial. Use of a
large losses and risk of hot spot in proximity of the Litz wire helps to reduce skin and proximity effects
gap. This issue is solved by use of a core with and provides a good ac/dc resistance ratio.
distributed air gap, e.g. a powder core. If a core Unfortunately in the same time a Litz wire has a
with a discrete air gap is an only option then the poor copper fill factor, and thus requires a core
special winding arrangement is necessary. It is with a large window. Also the leakage inductance
simply done by notching of inner layers of the of such winding is high. A round wire has a better
winding, so the fringing field doesn't penetrate thru copper fill factor and a dc resistance is reduced in
the winding [13, 14]. An example of such winding compare to the Litz wire. But due to the skin effect
is presented on Fig. 3. the ac loss is fairly high. By taking several thinner
round wires one may reduce layer thickness and
3.2. Three winding transformer the ac resistance, but it requires fairly complicated
winding arrangement. A foil winding seems to be a
ip1 iin(dc) iin(ac) is` good solution once again. It has a very good
copper fill factor and the dc resistance is low.
Small thickness of the foil reduces the skin effect,
t while interleaving of primary and secondary
ip11
t
Fig. 4. Transformer primary currents and their
components
np1
ns
iin
is
is'
np11
Fig. 5. Primary currents flow in transformer Fig. 6. A three winding transformer foil
windings windings arrangement
146
windings attenuates the proximity effect [15]. of the switching frequency). The transformer
The proposed winding arrangement of the three bases on E42/21 core too. Because of core losses
winding transformer is presented on Fig. 6. its peak induction is limited to 0.15 T at high input
Primary and secondary windings are interleaved. voltage and it requires at least 6 turns in each of
Primary windings np1 and np11 are made of a thin primary windings. Primary windings are made of
foil occupying whole available window width. Each 24x0.15 mm copper foil. The transformer turns
of primary windings has just a single turn per layer ratio is np1:np11:ns=1:1:2, thus the secondary
and the secondary winding is sandwiched between winding has 12 turns and it’s made of 10x0.15 mm
primaries. The secondary winding has few turns copper foil. Measured dc resistance is
arranged in a single layer like presented on Fig. 6. Rp1dc=Rp11dc=3 mȍ for the primary winding and
Such winding arrangement provides low dc and ac Rsdc=12 mȍ for the secondary. The ac resistance
resistances for all mentioned above current of the transformer is measured in respect to
components and a good coupling between primary particular current components flowing in its
and secondary windings. This solution has two windings and described in section 3.2. The ac
main drawbacks. First, because of intensive resistance seen from the secondary winding while
interleaving there is a large surface area of primary primaries are connected in series and shorted is
and secondary winding facing each other. Rsac+Rp1ac+Rp11ac=21 mȍ. This resistance is related
Depends on the insulation material and the to the secondary current and the reflected
insulation thickness it may result in a significant secondary current. Next the resistance for input
interwinding capacitance which decreases current dc and ac components is evaluated with
performance. The second drawback is reduced primary windings connected anti-parallel.
flexibility in terms of turns ratio. It’s fairly easy to
achieve integer turns ratio, like 1:2 or 1:3. But it
may be difficult to achieve non-integer turns ratio,
like 1:2.5.
4. Experimental results
In order to verify theoretical analysis and
simulations the 500 W breadboard has been built
and tested. The converter input voltage is 30-50 V
and the output voltage is regulated at 400 V. The
switching frequency is fs=50 kHz. The assembled
converter is presented on Fig. 7. The inductor LM
bases on E42/21 ferrite core with 1 mm air gap.
The winding is made of 24x0.3 mm copper foil and
there are 12 turns, which gives the inductance Fig. 8. Waveforms of the converter operating
LM=47 PH. It limits input current ripples to 3.5 A with Vin=10V and 50% duty cycle
peak to peak, which is about 20% of the maximum
input dc current. Measured dc resistance of the
inductor winding is RMdc=2.7 mȍ while respective
ac resistance is RMac=120 mȍ for 100 kHz (double
Fig. 9. Waveforms of the converter operating

Fig. 7. Assembled 500 W converter with Vin =10 V and 40% duty cycle
147
5.5
Voltage gain
4.5
3.5
Gain given by (1)
3
Gain given by (4)
2.5 Measured gain
2
0.3 0.35 0.4 0.45
Duty cycle
Fig. 10. Waveforms of the converter operating Fig. 12. Voltage gain of the converter operating
with Vin =10V and 39.5% duty cycle at low duty cycle
98
97
96
Efficiency [%]
95
30V
94
50V
93
92
0 100 200 300 400 500 600
P [W]
out
Fig. 11. Waveforms of the converter operating Fig. 13. Measured efficiency of the converter
with Vin =10 V and 25% duty cycle for 30 V and 50 V input voltage
Measured dc resistance is about Rp1dc||Rp11dc=1.5 while voltages VC2 and VC3 are given by the
mȍ and the ac resistance is about Rp1ac||Rp11ac=1.8 transformer secondary voltage (Ch3). For duty
mȍ. The transformer turns ratio is 1:1:2 and it cycles 50% and 40% output capacitors share the
enables to use of 150 V MOSFETs (IRF4321PbF). output voltage equally and the voltage gain follows
Diodes D1 and D11 are 150 V Schottky’s (1). When the duty cycle decreases just below Dcr
(30CPQ150PbF) while D2 and D3 are 600 V SiC (Fig. 10) the voltage VC1 quickly increases (the
diodes (IDT10S60C). peak of Ch2) and voltages VC2 and VC3 decrease
Next, the test at low duty cycles and reduced input (the amplitude of Ch3). When the duty cycle
voltage is being performed. The input voltage is decreases farther then the transformer secondary
set to 10 V and the duty cycle is changed from voltage collapses down to zero. In this case only
50% down to 25%. From (5) it’s found that the the capacitor C1 contributes to the output voltage.
critical duty cycle Dcr=40%. Fig. 8, Fig. 9, Fig. 10 Fig. 12 compares theoretical voltage gains (given
and Fig. 11 present observed waveforms at 50%, by (1) and (4)) against measured voltage gain.
40%, 39.5% and 25% duty cycle respectively. On Finally the efficiency of the converter has been
these figures Ch1 presents the gating signal for the measured using four precision digital multimeters
transistor, Ch2 presents the drain-source voltage Fluke 8845A. Since the input current exceeds the
of the respective transistor, Ch3 presents the multimeter current range an external precise shunt
transformer secondary voltage and Ch4 presents resistor has been used. Measured efficiency
the input current. The voltage VC1 is determined by curves are presented on Fig. 13.
the transistor drain-source peak voltage (Ch2),
148
5. Conclusion Voltage Gain Boost Converter Based on

Three-State Switching Cell," in The 32nd
This paper describes the boost converter with the Annual Conference of the IEEE Industrial
three-state switching cell. To improve the design Electronics Society IECON'06, Paris, France,
and reduce size of the converter foil windings are 2006, pp. 1927-1932.
used instead of round wires or Litz wires. The dc [8] P. Zacharias, B. Sahan, S. V. Araújo, F. L. M.
inductor with foil winding is characterized by a very Antunes, and R. T. Bascopé, "Analysis and
low dc resistance. For the three winding Proposition of a PV Module Integrated
transformer (the push-pull transformer) proper Converter with High Voltage Gain Capability in
arrangement of foil windings ensures a very low a non-isolated Topology," in The 7th
resistance for all current components present in its International Conference on Power
windings. Measured peak efficiency exceeds 97% Electronics Daegu, Korea: IEEE, 2007.
for low and high input voltage. Also it has been [9] G. V. T. Bascope and I. Barbi, "Generation of
demonstrated that the regulation range of this type a family of non-isolated DC-DC PWM
of converter can be increased by operation with converters using new three-state switching
duty cycle between 50% and Dcr. cells," in Power Electronics Specialists
Conference, 2000. PESC 00. 2000 IEEE 31st
Annual, 2000, pp. 858-863 vol.2.
Acknowledge [10] E. H. Kim and B. H. Kwon, "High step-up
This project is supported by Dansk Energi net resonant push-pull converter with high
under PSO project no. 562/06-14-26808. efficiency," Power Electronics, IET, vol. 2, pp.
79-89, 2009.
[11] L. Po-Wa, L. Yim-Shu, D. K. W. Cheng, and L.
References Xiu-Cheng, "Steady-state analysis of an
interleaved boost converter with coupled
[1] P. Klimczak and S. Munk-Nielsen,
inductors," Industrial Electronics, IEEE
"Comparative study on paralleled vs. scaled
dc-dc converters in high voltage gain
[12] C. W. T. McLyman, Transformer and inductor
applications," in Power Electronics and Motion
design handbook: Marcel Dekker, 2004.
Control Conference, 2008. EPE-PEMC 2008.
[13] J. D. Pollock and C. R. Sullivan, "Gapped-
13th, 2008, pp. 108-113.
inductor foil windings with low AC and DC
[2] B. Ju-Won, R. Myung-Hyo, K. Tae-Jin, Y.
resistance," in Industry Applications
Dong-Wook, and K. Jong-Soo, "High boost
Conference, 2004. 39th IAS Annual Meeting.
converter using voltage multiplier," in
Conference Record of the 2004 IEEE, 2004,
Industrial Electronics Society, 2005. IECON
p. 557.
2005. 31st Annual Conference of IEEE, 2005,
[14] J. D. Pollock and C. R. Sullivan, "Modelling
p. 6 pp.
Foil Winding Configurations with Low AC and
[3] Z. Qun and F. C. Lee, "High-efficiency, high
DC Resistance," in Power Electronics
step-up DC-DC converters," Power
Specialists Conference, 2005. PESC '05.
Electronics, IEEE Transactions on, vol. 18, pp.
IEEE 36th, 2005, pp. 1507-1512.
65-73, 2003.
[15] M. Nymand and M. A. E. Andersen, "A new
[4] W. Rong-Jong and D. Rou-Yong, "High step-
approach to high efficiency in isolated boost
up converter with coupled-inductor," Power
converters for high-power low-voltage fuel cell
Electronics, IEEE Transactions on, vol. 20, pp.
applications," in Power Electronics and Motion
1025-1035, 2005.
Control Conference, 2008. EPE-PEMC 2008.
[5] K. C. Tseng and T. J. Liang, "Novel high-
13th, 2008, pp. 127-131.
efficiency step-up converter," Electric Power
Applications, IEE Proceedings -, vol. 151, pp.
182-190, 2004.
[6] R. J. Wai and R. Y. Duan, "High-efficiency
DC/DC converter with high voltage gain,"
Electric Power Applications, IEE Proceedings
-, vol. 152, pp. 793-802, 2005.
[7] G. V. Torrico-Bascope, R. P. Torrico-Bascope,
D. S. Oliviera, F. L. M. Antunes, S. V. Araujo,
and C. G. C. Branco, "A Generalized High
149
A.5 Integration of Magnetic Components in a Step-

Up Converter for Fuel Cell (EPE 2009)
Integration of Magnetic Components in a Step-Up Converter for Fuel Cell

Pontoppidanstræde 101 Pontoppidanstræde 101
Aalborg, Denmark Aalborg, Denmark
Tel.: +45 9940 7331 Tel. +45 9940 9242
www.iet.aau.dk www.iet.aau.dk
Acknowledgements
This project is supported by Dansk Energi net under PSO project no. 562/06-14-26808.
Keywords
«DC power supply», «photovoltaic», «renewable energy systems», «switched-mode power supply»,
«uninterruptible power supply (UPS)»
Abstract
Fuel cell generator is a good example of a high voltage gain application in which dc-dc step-up
converter is a critical part. The input voltage of the converter decreases while the output power
increases. It creates challenges in design of the converter’s magnetic components. Scope of this
paper is integration of the dc inductor and the transformer on a single core. Such integration
improves utilization of the core and windings. It leads to size reduction of the converter.
Introduction
An increasing interest in alternative and renewable energy creates a demand for novel power
conversion systems, which interfaces different energy source to a load or to a utility grid. Very often
it’s necessary to boost a low voltage from a source (small fuel cells, most of photovoltaic panels) to a
high voltage suitable for an inverter. The required voltage gain may exceed value of 20 in many
cases. Such gain is attainable for boost converter with SiC diodes and CoolMOS transistors, but by a
cost of poor semiconductors utilization and limited efficiency [1]. In the literature [2-6] one may
find several topologies with a coupled inductor, which offer a high voltage gain. The main drawback
among these topologies is large input current ripple and such ripples may lead to PV maximum power
point tracking mismatch or to quick degradation of a fuel cell. For these reasons an additional input
filter is required, but it increases size, cost and complexity of the converter. The topology introduced
in [7] and presented on Fig. 1 offers a high voltage gain, reduced voltage rating of MOSFETs and
input current ripple is limited by inductor LM. Reference [8, 9] reports a very good performance of
this topology in a high voltage gain application, like a fuel cell. However in such application the
converter has to operate under variable, load dependent input voltage. Typically for fuel cell
minimum to maximum input voltage ratio is about 1:2 [10]. While the converter operates under
different conditions it might be difficult to optimize magnetic components. On one hand they are
sized for high input current. One the other hand they are designed for the highest input voltage too.
But in fuel cell application high voltage and high current doesn’t occur at the same time. It may
result in poor utilization of magnetic cores of the inductor and the transformer.
The main focus in this paper is to design and implement of an integrated magnetic, which should
reduce size and weight of the converter.
150
Description of the converter
Fig. 1 Diagram of the converter
The selected topology is presented on Fig. 1. It incorporates so called 3-state switching cell (T1, T11,
D1, D11 and the transformer) [7, 11]. Explanation of the converter’s operation become more intuitive
if one considers this topology as a modified current fed push-pull converter – with diodes D1, D11 and
capacitor C1 as a lossless clamp for transistors. Key waveforms are presented on Fig. 2. Basically
the converter operates with transistor duty cycle larger than 50 %, i.e. overlapping mode. Assuming
continuous conduction mode (CCM) there are four operating modes. Mode 1 and 3 are identical,
while both transistors are turned-on and share the input current equally (np1=np11). During this period
the input inductor LM is being charged and the input current iin rises. Because both primary windings
are shorted there is no voltage across the secondary winging, so there is no current flow in the
secondary winding too. Next, in mode 2 the transistor T1 is being turned-off and energy stored in LM
is being released thru the primary winding (np1) and the diode D1 to the capacitor C1, and thru the
secondary winding ns and the diode D3 to the capacitor C3. The last mode 4 is symmetrical to mode 2
– while the transistor T1 remains turned-on and T11 is turned-off. During this period the diode D11
and the diode D2 conduct. Capacitors C1 and C2 are recharged. Static voltage gain of the converter
is given by (1), where n is primary to secondary turns ratio. Voltage stress is given by (2) for the
primary side semiconductors and by (3) for the secondary side diodes D2 and D3.
1 n
Vout Vin (1)
1 D
1
VT1 VT11 VD1 VD11 Vout (2)
n 1
n
VD2 VD3 Vout (3)
n 1
Fig. 2 Main waveforms of the converter

151
Magnetic components
Magnetic components are essential for any dc-dc converter design. Their performance reflects
directly on converter’s performance. On the one hand a design of magnetic components focuses on
reduction of losses inside them and on reduction of parasitics, so external circuit is not affected. On
the other hand magnetic components are significant part of the converter size and weight, so size
reduction should be taken into consideration too. In many cases integration of magnetics may
improve utilization of a core or windings and thus size may be reduced.
The dc inductor
In the considered topology the dc inductor LM conducts essentially the dc current with relatively small
ac component (ripples). Majority of dc current content allows using of a round wire for the inductor
winding. It characterizes relatively low cost and good copper fill factor Ku. Since the dc current is
dominant a low dc resistance is a key. In order to improve copper fill factor one may use two or
three thinner wires in parallel or a rectangular wire instead of a single thick round wire. If required
number of turns is low one may use a foil winding, which may provide a very good copper fill factor
and thus a very low dc resistance [9].
A design procedure based on area-product approach is well described in [12], so here only major steps
are presented. First, the required inductance LM is to be determined. For the considered converter
it’s given by (4), where IM is the amplitude of current ripples. By rearranging (4) it’s found that the
largest current ripples occur at the duty cycle D=0.75.
Vout D 0.5 1 D
LM (4)
f s 2 'I M 1 n
Assuming that the inductor operates in its linear region, below saturation it’s easy to find dc and ac
induction components (Bind and Bind) since they are directly proportional to the inductor dc and ac
current components (IM(dc) and IM).
The core size is characterized by the area product Ap, which is the product of the core cross section
area Ac and the window area Wa [13]. Required area product Ap value is given by (5). The current
IM(dc) is the inductor maximum dc current, i.e. at the minimum input voltage Vin and rated power.
When the dimensions of the core are know one may find required number of turns (6) and length of
the air gap (7).
Ap

LM I M dc I M dc 'I M (5)
Bind 'Bind J K u
LM 'I M
nM (6)
'Bind Ac
P0 n Ac
2
M
lg (7)
LM
For small air gaps effective area of the gap Ag is close to core cross section area Ac (8). For larger
gaps the effective area of the gap increases and have to be corrected according to (9) [12].
Ag Ac a b (8)
Fig. 3 Magnetic circuits of an inductor (a) and a push-pull transformer (b)

152
Fig. 4 The transformer primary currents and its components
Ag a lg b lg (9)
Equivalent magnetic circuit of the inductor is presented on Fig. 3a. Magneto-motive force (MMF) is
represented as voltage source nM·IM, while the air gap reluctance is represented by linear resistor Rg.
Resistor Rc represents reluctance of the ferrite core and it’s very small in compare to reluctance of the
air gap (Rc<<Rg).
The push-pull transformer
Fig. 4 presents the transformer primary currents and its components: input dc current IM(dc), input
current ripples IM and reflected secondary current is’. This combination of ac and dc components
creates challenge for design of the transformer windings. Reduction of both, dc and ac resistances of
primary windings is crucial. Use of a Litz wire helps to reduce skin and proximity effects.
Unfortunately in the same time the Litz wire has poor copper fill factor Ku, so one end up with a bulky
transformer or a large dc resistance. Also the leakage inductance of such a winding is relatively high
affecting the external circuit. Round wire has a better copper fill factor and dc resistance is reduced
in compare to Litz wire. However inappropriate design of such windings may result in significant ac
losses due to the skin and proximity effects. Interleaving of the primary and the secondary windings
helps to reduce ac losses. A foil winding may provide a very good copper fill factor and a low dc
resistance. Interleaved winding arrangement a large area of the primary and the secondary winding
facing each other, thus the ac resistance is reduced. Unfortunately, in the same time interwinding
capacitance increases. For a few turns and thick isolation between primary and the secondary it may
not be a severe problem and such design will have a very good performance [9]. But reduction of the
isolation thickness and thus the distance between the primary and the secondary winding may result in
significant increases of the interwinding. A large interwinding capacitance results in a significant
degradation of the converter’s performance.
Design procedure based on area product Ap is described in details in [12]. First, size of the core is
calculated from (10). Variable Pt is transformer apparent power, i.e. sum of input and output power
of the transformer.
Pt
Ap (10)
2 K u K f 'Btr J f s
Induction amplitude Btr may not exceed saturation level of the core. However in many cases the
peak induction is well below the saturation level because of the maximum allowed core loss.
Specific core loss Pfe as a function of induction amplitude and switching frequency is given by well
known Steinmetz equation [13]. Once the induction amplitude Btr and the core size are known the
number of primary turns can be found with (11). It’s found that the largest induction amplitude
occurs at duty cycle D=0.5, i.e. for the highest input voltage.
1 Vout 1 D
np1 np11 (11)
2 'Btr Ac 2 1 n f s
Fig. 3b shows the magnetic circuit of a push-pull transformer. There are three voltage sources
representing MMFs of windings. MMFs directions refer to the converter diagram on Fig. 1.
Reluctance of the core is represented by the resistor Rc. By analyzing primary current’s (Fig. 4) one
can find that MMFs caused by the input current (dc and ac components) cancel each other. There is
153
no effect in the secondary winding due to the input current components flowing in the primary
windings.
The integrated magnetic
There are two major reasons for integration of the inductor and the transformer into a single device.
First, the high frequency transformer is a thermal limited design, so the induction amplitude Btr might
be well below the core saturation level. It leaves some room for dc flux which may come from the
inductor winding nM. Second, the input current flows thru the primary windings of the push-pull
transformer but creates no flux since MMFs related to the primary windings are opposite. The only
result of the input current flow in the primary windings is a copper loss. However with a proper
winging arrangement it’s possible to use primary windings as a part of the inductor windings.
Fig. 3 presents magnetic circuits of the inductor and the push-pull transformer. These two circuits
can be combined together on a single EE core. The simplest way is to place the inductor winding nM
on the center gapped leg. Transformer windings are split into two identical halves and placed on
outer legs. Resulting magnetic circuit is shown on Fig. 5. In this way the inductor and the
transformer fluxes are present in the outer legs of the core, which improves its utilization. However
the flux generated by inductor winding does not affect transformer windings. Also flux generated by
the transformer has no influence on the inductor. This solution results in complex winding
arrangement, large overall number of turns and possible large winding resistances.
The inductor current splits into two halves in the transformer primary windings. The flux related to
the inductor winding also splits into two halves once it goes into the core outer legs. So, there is a
great agreement between inductor current, the transformer primary current components (IM(dc) and IM)
and related fluxes. It’s possible to use the transformer primary windings as a part of the inductor
winding. For proper operation of the transformer the secondary winding is split and arranged on
both outer legs, so it’s not affected by inductor flux. Proposed magnetic circuit is presented on Fig. 6
where currents directions refer to Fig. 1. The magnetic circuit from Fig. 6 is described by (12),
where LM is self inductance of nM winding, MMp1 is mutual inductance between nM and np1 windings
and so on.
Design of the integrated magnetic is an iterative process. When both transistors are turned-on the
input voltage is applied to inductor and transformer’s primary windings, so (13) is truth. Assuming
that secondary current is zero and both primary currents are equal ip1=ip11 one can write (14) and (15).
Combining these equations with (13) gives the equivalent inductance LMeq (16). With a good
Fig. 5 Magnetic circuit of the integrated inductor-transformer
Fig. 6 Reduced magnetic circuit of the integrated inductor-transformer

154
coupling between both primary windings the mutual inductance Mp1p11 value is close to the self
inductance Lp1 value, so they will cancel each other. It’s important to note that the equivalent
inductance LMeq is larger then the physical inductance LM, so in this way the inductance LM may be
reduced. It turns into fewer inductor turns nM and possibly lower conduction losses.
Approximated core size is found using area product approach simply by summarizing area products
required by the transformer and by the inductor (17). One can change ratio between the inductor
induction Bind and the transformer induction ǻBtr and find the minimum required core size. This
method gives only an approximated core size, but it’s a good starting point for further optimization
and the size reduction.
Once the core size is determined the inductor and the transformer windings are designed according to
guidelines presented in previous sections. First, number of turns in the transformer primary winding
is found with (11). Next, in an iterative process one can calculate self inductances and mutual
inductances (12), find required number of turns nM (6) and size of the air gap lg (7).
Contrary to a typical transformer, the integrated magnetic usually operates under significant dc flux
conditions. It has two major effects. First, because of dc flux in the core increased core losses may
occur. Unfortunately core datasheets contain power loss data only for pure sinusoidal excitation, so
it’s difficult to calculate core losses in advance. Second, the peak induction is higher in compare to a
traditional transformer. Thus there is a very little space for any undesired dc flux. It requires
careful control of dc currents flowing thru primary windings np1 and np11.
diM
ª vM º ª LM M Mp1 M Mp11 M Msa M Msb º ª dt º
« » « « »
M p1p11 M p1sa M p1sb »» « dt »
dip1
« vp1 » « M p1M Lp1
« v » « M p11M M p11p1 « »
Lp11 M p11sa M p11sb » « dip11 » (12)
« p11 » « » « dt »
« vsa » « M saM M sap1 M sap11 Lsa M sasb » dis
« » «M « dt »
M sbp1 M sbp11 M sbsa Lsb »¼ « dis »
¬ vsb ¼ ¬ sbM ¬ dt ¼
diM
Vin vM vp1 LMeq (13)
dt
di dip1 dip11
v M LM M M Mp1 M Mp11 (14)
dt dt dt
di dip1 dip11
vp1 M p1M M Lp1 M p1p11 (15)
dt dt dt
1 1
LMeq LM M Mp1 M p1M Lp1 M p1p11 (16)
2 2
Ap
Pt

LMeq I M dc I M dc 'I M (17)
2 K u K f 'Btr J f s Bind 'Bind J K u
Experimental setup and results

In order to verify theoretical analysis and calculations the 500 W breadboard is built and tested. The
input voltage varies from 30 to 50 Vdc and the output voltage is regulated at 400 Vdc. The switching
frequency is fs=50 kHz and the transformer turns ratio is n=2. According to (2) it allows using 150 V
MOSFETs (IRFP4321PbF) and 150 V Schottky diodes (30CPQ150PbF) on primary side. On the
secondary side of the transformer there are 600 V SiC diodes (IDT10S60C).
It’s found that the E42/21 core (0P44022EC) is sufficiently large. With nM=10 turns and 1.8 mm air
gap in the center leg it gives the inductance LM=18 µH. The transformer induction amplitude is
limited to ǻBtr=0.15 T at the highest input voltage and it’s limited by core losses. The transformer
primary winding has np1=np11=13 turns, which results in the equivalent inductance LMeq=48 µH. The
secondary winding has ns=26 turns and it’s divided into to halves, nsa=nsb=13 turns on each of outer
legs. The primary and the secondary windings are interleaved to reduce proximity effect and ac
losses. The assembled integrated inductor transformer is presented on Fig. 7. Fig. 8 presents the
155
Fig. 7 The integrated inductor-transformer (core type EE42/21)
Fig. 8 The assembled 500 W step-up converter
a) b)
Fig. 9 The observed input current and its spectrum in balances case (left) and in case of saturation of
the other leg (right)
assembled 500 W converter. Below the integrated transformer-inductor two MOSFETs (T1, T11) and
diodes (D1, D11 in a single pack) are visible. Control of dc currents in primary windings is critical to
avoid outer legs saturation. In case of integrated inductor-transformer it’s possible to detect such
saturation by observation of the input current only. Fig. 9a shows the input current and its spectrum
in an unsaturated and balanced case . The input current has only a large dc component and 100 kHz
component, which refers to input current ripples at double switching frequency. In case of small
unbalance of the transformer primary currents one of outer legs become saturated and shape of the
input current changes. Now, 50 kHz component appears in the current spectrum, like presented on
Fig. 9b.
Fig. 10 presents measured efficiency of the converter at 30 V and 45 V input voltage. For the lower
input voltage the efficiency peaks at 97.4% at 250 W and remains above 97% up to 450 W. For the
higher input voltage the efficiency exceeds 97.5%.
156
98
97.5
97
Efficiency [%]
96.5
96 30V
95.5 45V
95
0 100 200 300 400 500 600
Pout [W]
Fig. 10 Measured efficiency of the converter
Conclusion
A boost converter with a three state switching cell and integrated magnetic has been presented in this
paper. The converter has the inductor and the push-pull transformer arranged on a standard gapped
EE core. Such integration allows using of transformer primary windings as a part of the inductor
winding, thus smaller physical inductance is required. It results in smaller and lighter core. To
verify the idea the 500 W breadboard is built and tested. The breadboard exhibits a good
performance and a compact assembly.
References
[1] P. Klimczak and S. Munk-Nielsen, "Comparative study on paralleled vs. scaled dc-dc converters in high
voltage gain applications," in Power Electronics and Motion Control Conference, 2008. EPE-PEMC
2008. 13th, 2008, pp. 108-113.
[2] B. Ju-Won, R. Myung-Hyo, K. Tae-Jin, Y. Dong-Wook, and K. Jong-Soo, "High boost converter using
voltage multiplier," in Industrial Electronics Society, 2005. IECON 2005. 31st Annual Conference of
IEEE, 2005, p. 6 pp.
[3] Z. Qun and F. C. Lee, "High-efficiency, high step-up DC-DC converters," Power Electronics, IEEE
[4] W. Rong-Jong and D. Rou-Yong, "High step-up converter with coupled-inductor," Power Electronics,
IEEE Transactions on, vol. 20, pp. 1025-1035, 2005.
[5] K. C. Tseng and T. J. Liang, "Novel high-efficiency step-up converter," Electric Power Applications,
IEE Proceedings -, vol. 151, pp. 182-190, 2004.
[6] R. J. Wai and R. Y. Duan, "High-efficiency DC/DC converter with high voltage gain," Electric Power
Applications, IEE Proceedings -, vol. 152, pp. 793-802, 2005.
[7] G. V. Torrico-Bascope, R. P. Torrico-Bascope, D. S. Oliviera, F. L. M. Antunes, S. V. Araujo, and C.
G. C. Branco, "A Generalized High Voltage Gain Boost Converter Based on Three-State Switching
Cell," in The 32nd Annual Conference of the IEEE Industrial Electronics Society IECON'06, Paris,
France, 2006, pp. 1927-1932.
[8] P. Zacharias, B. Sahan, S. V. Araújo, F. L. M. Antunes, and R. T. Bascopé, "Analysis and Proposition
of a PV Module Integrated Converter with High Voltage Gain Capability in a non-isolated Topology,"
in The 7th International Conference on Power Electronics Daegu, Korea: IEEE, 2007.
[9] P. Klimczak and S. Munk-Nielsen, "High Efficiency Boost Converter with Three State Switching Cell,"
in International Exhibition & Conference for Power Electronics, Intelligent Motion, Power Quality
PCIM Europe 2009 Nuremberg, Germany, 2009.
[10] W. Vielstich, Handbook of fuel cells fundamentals, technology, and applications editors, Wolf Vielstich,
Arnold Lamm, Hubert A. Gasteiger: Wiley, 2003.
157
[11] G. V. T. Bascope and I. Barbi, "Generation of a family of non-isolated DC-DC PWM converters using
new three-state switching cells," in Power Electronics Specialists Conference, 2000. PESC 00. 2000
IEEE 31st Annual, 2000, pp. 858-863 vol.2.
[12] C. W. T. McLyman, Transformer and inductor design handbook: Marcel Dekker, 2004.
[13] J. Reinert, A. Brockmeyer, and R. W. A. A. De Doncker, "Calculation of losses in ferro- and
ferrimagnetic materials based on the modified Steinmetz equation," Industry Applications, IEEE
159
Appendix B
Models
B.1 Transistor loss model (lib_trans.m)

1 %Calculation of transistor conduction, switching and gate losses
2 %INPUT DATA FROM CONVERTER FILE
3 trans_model;
4 trans_Tj;
5 trans_fs;
6 trans_Von;
7 trans_Voff;
8 trans_Ion;
9 trans_Ioff;
10 trans_VGS;
11 trans_Rgon;
12 trans_Rgoff;
13 trans_Irms;
14
15 %INPUT DATA FROM DATA FILE
16 trans_RDSon=trans.RDSon(trans_model)+(trans_Tj-
25)*trans.Rtemp(trans_model);
17 trans_Cgs=trans.Cgs(trans_model);
18 trans_Cosse=trans.Cosse(trans_model);
19 trans_Coss=[trans.Coss0(trans_model), trans.Coss1(trans_model),
trans.Coss2(trans_model) trans.Coss3(trans_model)];
20 trans_Crss=[trans.Crss0(trans_model), trans.Crss1(trans_model),
trans.Crss2(trans_model) trans.Crss3(trans_model)];
21 trans_VDS=[trans.VDS0(trans_model), trans.VDS1(trans_model),
trans.VDS2(trans_model) trans.VDS3(trans_model)];
22 trans_Rgontot=trans.Rgint(trans_model)+trans_Rgon;
23 trans_Rgofftot=trans.Rgint(trans_model)+trans_Rgoff;
24 trans_Qg=trans.Qg(trans_model);
25 trans_Qmill=trans.Qmill(trans_model);
26 trans_VGSth=trans.VGSth(trans_model);
27 trans_gfs=trans.gfs(trans_model);
28
29 %Switching times calculation
30 trans_Ciss=trans_Cgs+interp1(trans_VDS,trans_Crss,trans_Von,'pchip
');
31 %Turn-on delay time
32 trans_tdon=trans_Rgontot*trans_Ciss*log((trans_VGS)/(trans_VGS-
trans_VGSth));
33 %Current rise time
34 trans_tir=trans_Rgontot*trans_Ciss*log((trans_VGS-
trans_VGSth)/(trans_VGS-trans_VGSth-trans_Ion/trans_gfs));
35
160
36 %Energy dissipated during voltage fall time

37 trans_Vvf_mx=linspace(trans_Von,trans_Ion*trans_RDSon,100);
38 trans_tvf_mx=0;
39 trans_Evf=0;
40 for trans_i=2:size(trans_Vvf_mx,2)
41 %Volt fall time estimation due to Cgd (gate side) behavior
42 trans_Cgd=interp1(trans_VDS,trans_Crss,trans_Vvf_mx(trans_i),'pchi
p');
43 trans_tvf_mx(trans_i)=trans_tvf_mx(trans_i-
1)+(trans_Vvf_mx(trans_i-1)-
trans_Vvf_mx(trans_i))*trans_Rgontot*trans_Cgd/(trans_VGS-
trans_VGSth-trans_Ion/trans_gfs);
44 trans_Evf=trans_Evf+(trans_tvf_mx(trans_i)-trans_tvf_mx(trans_i-
1))*trans_Ion*trans_Vvf_mx(trans_i);
45 end
46 trans_PTon=trans_fs*trans_Ion*trans_Von*trans_tir/2+trans_Evf*tran
s_fs;
47
48 %turn-off transient
49 trans_Ciss=trans_Cgs+interp1(trans_VDS,trans_Crss,trans_Voff,'pchi
p');
50 %Turn-off delay time
51 trans_tdoff=trans_Rgofftot*trans_Ciss*log((trans_VGS)/(trans_VGSth
+trans_Ioff/trans_gfs));
52 %Current fall time
53 trans_tif=trans_Rgofftot*trans_Ciss*log((trans_VGSth+trans_Ioff/tr
ans_gfs)/(trans_VGSth));
54
55 %Energy dissipated during voltage rise time
56 trans_Vvr_mx=linspace(trans_Ioff*trans_RDSon,trans_Voff,100);
57 trans_tvr_mx=0;
58 trans_Evr=0;
59 for trans_i=2:size(trans_Vvr_mx,2)
60 %Volt rise time estimation due to the Cgd (gate side) behavior
61 trans_Cgd=interp1(trans_VDS,trans_Crss,trans_Vvr_mx(trans_i),'pchi
p');
62 trans_tvr_mx(trans_i)=trans_tvr_mx(trans_i-
1)+(trans_Vvr_mx(trans_i)-trans_Vvr_mx(trans_i-
1))*trans_Rgofftot*trans_Cgd/(trans_VGSth+trans_Ion/trans_gfs);
63 trans_Evr=trans_Evr+(trans_tvr_mx(trans_i)-trans_tvr_mx(trans_i-
1))*trans_Ioff*trans_Vvr_mx(trans_i);
64 end
65 trans_PToff=trans_fs*trans_Ioff*trans_Voff*trans_tif/2+trans_Evr*t
rans_fs;
66
67 %on-state losses
68 trans_PTcond=trans_Irms^2*trans_RDSon;
69
70 %gate drive losses
71 trans_PTgate=trans_VGS*trans_Qg*trans_fs;
B.2 Transistor data library (data_trans.m)

1 %MOSFET database
2 trans_i=1;
161
3 trans.type(trans_i)=['01.ideal MOSFET'];
4 trans.Vbrr(trans_i)=inf; %V
5 trans.Id(trans_i)=inf; %A
6 trans.RDSon(trans_i)=0; %ohm
7 trans.Rtemp(trans_i)=0; %ohm/K
8 trans.VDS0(trans_i)=1; %V
9 trans.VDS1(trans_i)=10; %V
10 trans.VDS2(trans_i)=1e3; %V
11 trans.VDS3(trans_i)=1e6; %V
12 trans.Coss0(trans_i)=0; %F
16 trans.Crss0(trans_i)=0; %F
20 trans.Cosse(trans_i)=0; %F
21 trans.Cgs(trans_i)=0; %F
22 trans.Rgint(trans_i)=0; %ohm
23 trans.Qg(trans_i)=0e-9; %C
24 trans.Qmill(trans_i)=0e-9; %C
25 trans.VGSth(trans_i)=0; %V
26 trans.gfs(trans_i)=1e3; %S
27
28 trans_i=trans_i+1;
29 trans.type(trans_i)=['02.ipw60r045cp'];
30 trans.Vbrr(trans_i)=650;
31 trans.Id(trans_i)=60;
32 trans.RDSon(trans_i)=40e-3;
33 trans.Rtemp(trans_i)=.56e-3;
34 trans.VDS0(trans_i)=5;
38 trans.Coss0(trans_i)=17000e-12;
42 trans.Crss0(trans_i)=1000e-12;
46 trans.Cosse(trans_i)=320e-12;
47 trans.Cgs(trans_i)=6800e-12;
48 trans.Rgint(trans_i)=1;
49 trans.Qg(trans_i)=150e-9;
50 trans.Qmill(trans_i)=50e-9;
51 trans.VGSth(trans_i)=3;
52 trans.gfs(trans_i)=60;
53
55 trans.type(trans_i)=['03.irfb4228pbf'];
162
79
81 trans.type(trans_i)=['04.fch47n60'];
106
107 trans.type(trans_i)=['05.fdb2532'];
163
131
133 trans.type(trans_i)=['06.irfp4321pbf'];
155 trans.VGSth(trans_i)=4.5;
157
159 trans.type(trans_i)=['07.irf4568pbf'];
162 trans.RDSon(trans_i)=4.8e-3;
164
181 trans.VGSth(trans_i)=3.5;
B.3 Diode loss model (lib_diode.m)

1 %Calculation of diode losses (conduction, rev.recovery and
capacitive)
2 %INPUT DATA FROM CONVERTER FILE
3 diode_model;
4 diode_Tj;
5 diode_I;
6 diode_Irms;
7 diode_Ioff;
8 diode_Voff;
9 diode_fs;
10
11 %INPUT DATA FROM DATA FILE
12 diode_Vf=diode.Vf(diode_model)+(diode_Tj-
25)*diode.Vftemp(diode_model); %for temp<>25C
13 diode_Rf=diode.Rf(diode_model)+(diode_Tj-
25)*diode.Rftemp(diode_model); %for temp<>25C
14 diode_Qrr=diode.Qrr(diode_model);
15 diode_trr=diode.trr(diode_model);
16 diode_C=[diode.C0(diode_model), diode.C1(diode_model),
diode.C2(diode_model), diode.C3(diode_model)];
17 diode_VR=[diode.VR0(diode_model), diode.VR1(diode_model),
diode.VR2(diode_model), diode.VR3(diode_model)];
18
19 %Conduction power losses
20 diode_PDcond=diode_I*diode_Vf+diode_Irms^2*diode_Rf;
21
22 %Reverse recovery losses
23 diode_PDrr=diode_fs*(diode_Voff*diode_Ioff*diode_trr+diode_Voff*di
ode_Qrr);
24
25 %Diode capacitive loss
26 diode_VR_mx=linspace(0,diode_Voff);
27 diode_QC=0;
28 for diode_i=1:(size(diode_VR_mx,2)-1)
29 diode_QC=diode_QC+interp1(diode_VR, diode_C,
diode_VR_mx(diode_i),'pchip')*(diode_VR_mx(diode_i+1)-
diode_VR_mx(diode_i));
30 end
31 diode_PDC=diode_QC*diode_Voff/2*diode_fs;
B.4 Diode data library (data_diode.m)

1 %Diode database
2 diode_i=1;
165
3 diode.type(diode_i)=(['01.ideal_D']);
4 diode.Vrrm(diode_i)=inf; %V
5 diode.If(diode_i)=inf; %A
6 diode.Vf(diode_i)=0; %V
7 diode.Vftemp(diode_i)=0; %V/K
8 diode.Rf(diode_i)=0; %ohm
9 diode.Rftemp(diode_i)=0; %ohm/K
10 diode.Qrr(diode_i)=0; %C
11 diode.trr(diode_i)=0; %s
12 diode.VR0(diode_i)=0; %V
15 diode.VR3(diode_i)=1e6; %V
16 diode.C0(diode_i)=0; %F
20
21 diode_i=diode_i+1;
22 diode.type(diode_i)=(['02.sdb06s60']);
23 diode.Vrrm(diode_i)=600;
24 diode.If(diode_i)=6;
25 diode.Vf(diode_i)=.9;
26 diode.Vftemp(diode_i)=-.88e-3;
27 diode.Rf(diode_i)=92e-3;
28 diode.Rftemp(diode_i)=.368e-3;
29 diode.Qrr(diode_i)=21e-9;
30 diode.trr(diode_i)=0;
31 diode.VR0(diode_i)=1;
35 diode.C0(diode_i)=220e-12;
39
41 diode.type(diode_i)=(['03.idt10s60']);
45 diode.Vftemp(diode_i)=-1e-3;
47 diode.Rftemp(diode_i)=.21e-3;
48 diode.Qrr(diode_i)=0;
58
60 diode.type(diode_i)=(['04.30cpq150']);
166
64 diode.Vftemp(diode_i)=-1.5e-3;
66 diode.Rftemp(diode_i)=-.08e-3;
67 diode.Qrr(diode_i)=0;
B.5 Magnetic device general loss model

(lib_mag.m)
1 %Calculation of core and copper losses in mag devices
2 %Core loss MSE + temp. + dc-bias
3 %INPUT PARAMETERS FROM CONVERTER FILE
4 mag_fs; %Hz
5 mag_feq; %Hz
6 mag_deltaB; %T
7 mag_Bdc; %T
8 mag_mat; %-
9 mag_core; %-
10 mag_Rdc; %ohm
11 mag_FR; %-
12 mag_I; %A
13 mag_Iac; %A
14 mag_T; %deg C
15
16 %INPUT PARAMETERS FROM DATA FILE
17 mag_kfs=1;
18 %Frequency range selection
19 while mag_fs>magmat.fs(mag_kfs,mag_mat)
20 mag_kfs=mag_kfs+1;
21 end
22 mag_k=magmat.k(mag_kfs,mag_mat);
23 mag_alpha=magmat.alpha(mag_kfs,mag_mat);
24 mag_beta=magmat.beta(mag_kfs,mag_mat);
25
26 mag_K1=magmat.K1(mag_kfs,mag_mat);
27 mag_K2=magmat.K2(mag_kfs,mag_mat);
28
29 mag_ct=magmat.ct(mag_kfs,mag_mat);
30 mag_ct1=magmat.ct1(mag_kfs,mag_mat);
31 mag_ct2=magmat.ct2(mag_kfs,mag_mat);
32
33 mag_volume=magcore.volume(mag_core); %cm3
34
167
35 %Core loss calculation

36 mag_kprim=mag_k*(1+mag_K1*mag_Bdc*exp(-mag_deltaB/mag_K2)); %dc
bias coefficient
37 mag_ktemp=mag_ct2*mag_T-mag_ct1*mag_T+mag_ct; %temperature
coefficient
38 mag_Pfevol=mag_kprim*mag_ktemp*(mag_fs/1e3)*(mag_feq/1e3)^(mag_alp
ha-1)*(mag_deltaB*10)^mag_beta/1e3; %loss per volume unit
39 mag_Pfe=mag_Pfevol*mag_volume;
40
41 %Copper loss
42 for mag_i=1:size(mag_Rdc,2)
43 mag_Pcudc(mag_i)=mag_Rdc(mag_i)*mag_I(mag_i)^2;
44 mag_Pcuac(mag_i)=mag_FR(mag_i)*mag_Rdc(mag_i)*mag_Iac(mag_i)^2;
45 end
B.6 Magnetic core library (data_magcore.m)

1 %Core database
2 magcore_i=1;
3 magcore.model(magcore_i)=(['01.E42/21/20']);
4 magcore.volume(magcore_i)=22.7; %cm3
5 magcore.Ag(magcore_i)=2.44; %cm2
6 magcore.lm(magcore_i)=9.7; %cm
7 magcore.Wamax(magcore_i)=2.55; %cm2
8 magcore.Wa(magcore_i)=1.78; %cm2
9
10 magcore_i=magcore_i+1;
11 magcore.model(magcore_i)=(['01.E42/21/15']);
12 magcore.volume(magcore_i)=17.3; %cm3
13 magcore.Ag(magcore_i)=1.85; %cm2
14 magcore.lm(magcore_i)=9.7; %cm
15 magcore.Wamax(magcore_i)=2.55; %cm2
16 magcore.Wa(magcore_i)=1.78; %cm2
B.7 Magnetic material library (data_magmat.m)

1 %Ferrite - powder materials database
2 magmat_i=1;
3 magmat.mat(magmat_i)=(['01.ideal ferrite']);
4 magmat.fs(:,magmat_i)= [100e3 500e3 inf inf inf];
5 magmat.k(:,magmat_i)= [0 0 0 0 0];
6 magmat.alpha(:,magmat_i)= [1 1 1 1 1];
7 magmat.beta(:,magmat_i)= [1 1 1 1 1];
8 magmat.K1(:,magmat_i)= [0 0 0 0 0];
9 magmat.K2(:,magmat_i)= [1 1 1 1 1];
10 magmat.ct(:,magmat_i)= [1 1 1 1 1];
11 magmat.ct1(:,magmat_i)= [0 0 0 0 0];
13
14 magmat_i=2;
15 magmat.mat(magmat_i)=(['02.P material Magnetics']);
16 magmat.fs(:,magmat_i)= [100e3 500e3 inf inf inf];
17 magmat.k(:,magmat_i)= [.158 .0434 7.36e-7 0 0];
168
18 magmat.alpha(:,magmat_i)= [1.36 1.63 3.47 1 1];

19 magmat.beta(:,magmat_i)= [2.86 2.62 2.54 1 1];
20 magmat.K1(:,magmat_i)= [0 0 0 0 0];
21 magmat.K2(:,magmat_i)= [1 1 1 1 1];
22 magmat.ct(:,magmat_i)= [1 1 1 1 1];
B.8 Boost converter main file (main_boost.m)

1 close all;
2 clear all;
3
4 run ../lib/data_magmat.m;
5 run ../lib/data_magcore.m
6 run ../lib/data_diode.m;
7 run ../lib/data_trans.m;
8
9 %CONVERTER INPUT DATA
10 conv_Vin=30;
11 conv_Vout=90;
12 conv_fs=50e3;
13 conv_Pout=135;
14 conv_eff=1;
15
16 %TRANSISTOR INPUT DATA
17 transT1_model=1+5;
18 transT1_Tj=50;
19 transT1_VGS=14;
20 transT1_Rgon=18+2;
21 transT1_Rgoff=18+2;
22
23 %DIODE INPUT DATA
24 diodeD1_model=1+3;
25 diodeD1_Tj=50;
26
27 %INDUCTOR INPUT DATA
28 magL1_L1=52e-6;
29 magL1_deltaB=0.065;
30 magL1_Bdc=0;
31 magL1_mat=1+1;
32 magL1_core=1;
33 magL1_T=30;
34 magL1_Rdc=4.9e-3;
35 magL1_FR=85/4.9;
36
37 %Run converter model
38 run ./conv_boost.m
39
40 ('done')
169
B.9 Boost converter model (conv_boost.m)

1 clear mag_*;
2 clear trans_*;
3 clear diode_*;
4 clear cap_*;
5
6 %CONVERTER INPUT DATA FROM MAIN FILE
7 conv_Vin;
8 conv_Vout;
9 conv_fs;
10 conv_Pout;
11 conv_eff;
12 magL1_L1;
13
14 %Input power and duty cycle
15 conv_Pin=conv_Pout/conv_eff;
16 conv_D=1-conv_Vin/conv_Vout;
17
18 %Currents in the circuit
19 conv_Iout=conv_Pout/conv_Vout;
20 conv_Iin=conv_Pin/conv_Vin;
21 conv_Iinpp=conv_Vin*conv_D/conv_fs/magL1_L1;
22
23 conv_Iint0p=conv_Iin-conv_Iinpp/2;
24 conv_Iint1m=conv_Iin+conv_Iinpp/2;
25 conv_Iint1p=conv_Iint1m;
26 conv_Iint2m=conv_Iint0p;
27
28 %Time instants
29 conv_t0=0;
30 conv_t1=conv_D/conv_fs;
31 conv_t2=1/conv_fs;
32
33
34 %TRANSISTOR
35 %INPUT DATA FROM MAIN FILE
36 trans_model=transT1_model; %NOTE: trans_model=1 is an ideal
transistor
37 trans_Tj=transT1_Tj;
38 trans_VGS=transT1_VGS;
39 trans_Rgon=transT1_Rgon;
40 trans_Rgoff=transT1_Rgoff;
41 %CALCULATED TRANSISTOR INPUT DATA
42 trans_fs=conv_fs;
43 trans_Irms=sqrt(conv_D*1/3*(conv_Iint0p^2+conv_Iint0p*conv_Iint1m+
conv_Iint1m^2));
44 trans_Ion=conv_Iint0p;
45 trans_Ioff=conv_Iint1m;
46 trans_Von=conv_Vout;
47 trans_Voff=conv_Vout;
48 %Library call
49 run ../lib/lib_trans.m
50 %Output data from loss model
51 transT1_PTcond=trans_PTcond;
52 transT1_PTon=trans_PTon;
53 transT1_PToff=trans_PToff;
170
54 transT1_PTgate=trans_PTgate;
55
56 %DIODE
58 diode_model=diodeD1_model; %NOTE: diode_model=1 is ideal diode
59 diode_Tj=diodeD1_Tj;
60 %CALCULATED DIODE INPUT DATA
61 diode_I=conv_Iout;
62 diode_Irms=sqrt((1-
conv_D)*1/3*(conv_Iint1p^2+conv_Iint1p*conv_Iint2m+conv_Iint2m^2))
;
63 diode_Ioff=conv_Iint2m;
64 diode_Voff=conv_Vout;
65 diode_fs=conv_fs;
66 %Library call
67 run ../lib/lib_diode.m
69 diodeD1_PDcond=diode_PDcond;
70 diodeD1_PDrr=diode_PDrr;
71 diodeD1_PDC=diode_PDC;
72
73 %INDUCTOR
75 mag_deltaB=magL1_deltaB;
76 mag_Bdc=magL1_Bdc;
77 mag_mat=magL1_mat;
78 mag_core=magL1_core;
79 mag_T=magL1_T;
80 mag_Rdc=magL1_Rdc;
81 mag_FR=magL1_FR;
82 %CALCULATED INDUCTOR DATA
83 mag_I=conv_Iin;
84 mag_Iac=conv_Iinpp/2/sqrt(3);
85 mag_fs=conv_fs;
86 mag_feq=2/pi^2*((2*mag_deltaB/2/mag_deltaB)^2/(conv_t1-
conv_t0)+(2*mag_deltaB/2/mag_deltaB)^2/(conv_t2-conv_t1));
87 %Library call
88 run ../lib/lib_mag.m
90 magL1_Pcuac=mag_Pcuac;
91 magL1_Pcudc=mag_Pcudc;
92 magL1_Pfe=mag_Pfe;
B.10 Non-isolated push-pull-boost converter main

file (main_pushpullboost.m)
1 close all;
2 clear all;
3
4 run ../lib/data_magmat.m;
5 run ../lib/data_magcore.m;
6 run ../lib/data_diode.m;
7 run ../lib/data_trans.m;
8
171
9 %CONVERTER INPUT DATA

10 conv_Vin=30;
11 conv_Vout=400;
12 conv_fs=50e3;
13 conv_Pout=250;
14 conv_eff=.98;
15
16 %TRANSISTOR INPUT DATA
17 transT1_model=6;
18 transT1_Tj=40;
19 transT1_VGS=14;
20 transT1_Rgon=5+2;
21 transT1_Rgoff=5+2;
22
23 %DIODE INPUT DATA LOW VOLTAGE
24 diodeD1_model=4;
25 diodeD1_Tj=40;
26
27 %DIODE INPUT DATA HIGH VOLTAGE
28 diodeD2_model=3;
29 diodeD2_Tj=40;
30
31 %INDUCTOR INPUT DATA
32 magL1_L1=48.3e-6;
33 magL1_deltaB=0.03;
34 magL1_Bdc=0.27;
35 magL1_mat=1+1;
36 magL1_core=1;
37 magL1_T=30;
38 magL1_Rdc=4.2e-3;
39 magL1_FR=247e-3/magL1_Rdc;
40
41 %TRANSFORMER INPUT DATA
42 magTR_deltaB=0.075;
43 magTR_Bdc=0;
44 magTR_mat=2;
45 magTR_core=1;
46 magTR_T=40;
47 magTR_n=2;
48 magTR_Rdc=[4.5e-3 18e-3];
49 magTR_FR=[4.13e-3 16.8e-3]./magTR_Rdc;
50
51 %RUN CONVERTER MODEL
52 run ./conv_pushpullboost.m;
53
54 ('done')
B.11 Non-isolated push-pull-boost converter model

(conv_pushpullboost.m)
1 clear mag_*;
2 clear trans_*;
3 clear diode_*;
4 clear cap_*;
172
5
6 %CONVERTER INPUT DATA FROM MAIN FILE
7 conv_Vin;
8 conv_Vout;
9 conv_fs;
10 conv_Pout;
11 conv_eff;
12 magL1_L1;
13 magTR_n;
14
15 %INPUT POWER AND DUTY CYCLE
16 conv_Pin=conv_Pout/conv_eff;
17 conv_D=1-conv_Vin*(1+magTR_n)/conv_Vout;
18
19 %CURRENTS IN CIRCUIT
20 covn_Iout=conv_Pout/conv_Vout;
21 conv_Iin=conv_Pin/conv_Vin;
22 conv_Iinpp=conv_Vin*(conv_D-0.5)/conv_fs/magL1_L1;
23
24 conv_Iint0=conv_Iin-conv_Iinpp/2;
25 conv_Iint1=conv_Iin+conv_Iinpp/2;
26 conv_Iint2=conv_Iint0;
29
30 conv_In1t0p=conv_Iint0/2;
31 conv_In1t1m=conv_Iint1/2;
32 conv_In1t1p=conv_Iint1/2/(1+magTR_n);
33 conv_In1t2m=conv_Iint2/2/(1+magTR_n);
34 conv_In1t2p=conv_Iint2/2;
35 conv_In1t3m=conv_Iint3/2;
36 conv_In1t3p=conv_Iint3-conv_Iint3/2/(1+magTR_n);
37 conv_In1t4m=conv_Iint4-conv_Iint4/2/(1+magTR_n);
38
39 conv_In2t1p=(conv_In1t3p-conv_In1t1p)/(magTR_n);
40 conv_In2t2m=(conv_In1t4m-conv_In1t2m)/(magTR_n);
41
42 conv_un1t01=1/3*(conv_In1t0p^2+conv_In1t0p*conv_In1t1m+conv_In1t1m
^2);
^2);
^2);
^2);
46
^2);
48
49 conv_an1t01=(conv_In1t0p+conv_In1t1m)/2;
53
55
56 %TIME INSTANTS
173
57 conv_t0=0;
58 conv_t1=(conv_D-0.5)/conv_fs;
59 conv_t2=0.5/conv_fs;
60 conv_t3=(conv_D)/conv_fs;
61 conv_t4=1/conv_fs;
62
63 conv_d01=(conv_t1-conv_t0)/(conv_t4-conv_t0);
67
68 %TRANSISTOR
70 trans_model=transT1_model;
71 trans_Tj=transT1_Tj;
72 trans_VGS=transT1_VGS;
73 trans_Rgon=transT1_Rgon;
74 trans_Rgoff=transT1_Rgoff;
75 %CALCULATED TRANSISTOR INPUT DATA
76 trans_fs=conv_fs;
77 trans_Irms=sqrt(conv_d01*conv_un1t01+conv_d23*conv_un1t23+conv_d34
*conv_un1t34);
78 trans_Ion=conv_In1t2p;
79 trans_Ioff=conv_In1t1m;
80 trans_Von=conv_Vout/(1+magTR_n);
81 trans_Voff=conv_Vout/(1+magTR_n);
82 %LIBRARY CALL
83 run ../lib/lib_trans.m;
84 %OUTPUT DATA FROM LOSS MODEL
85 transT1_PTcond=trans_PTcond;
86 transT1_PTon=trans_PTon;
87 transT1_PToff=trans_PToff;
88 transT1_PTgate=trans_PTgate;
89
90 %DIODE LOW VOLTAGE
92 diode_model=diodeD1_model;
95 diode_I=conv_d12*conv_an1t12;
96 diode_Irms=sqrt(conv_d12*conv_un1t12);
97 diode_Ioff=conv_In1t2m;
98 diode_Voff=conv_Vout/(1+magTR_n);
100 %LIBRARY CALL
101 run ../lib/lib_diode.m;
106
107 %DIODE HIGH VOLTAGE
109 diode_model=diodeD2_model;
112 diode_I=conv_d12*conv_an2t12;
113 diode_Irms=sqrt(conv_d12*conv_un2t12);
174
114 diode_Ioff=conv_In1t2m;
115 diode_Voff=conv_Vout*2/(1+magTR_n);
117 %LIBRARY CALL
118 run ../lib/lib_diode.m;
123
124 %INDUCTOR
126 mag_deltaB=magL1_deltaB;
127 mag_Bdc=magL1_Bdc;
128 mag_mat=magL1_mat;
129 mag_core=magL1_core;
130 mag_T=magL1_T;
131 mag_Rdc=magL1_Rdc;
132 mag_FR=magL1_FR;
133 %CALCULATED INDUCTOR INPUT DATA
134 mag_I=conv_Iin;
135 mag_Iac=conv_Iinpp/2/sqrt(3);
136 mag_fs=conv_fs*2;
137 mag_feq=2/pi^2*(1/(conv_t1-conv_t0)+1/(conv_t2-conv_t1));
138 %LIBRARY CALL
139 run ../lib/lib_mag.m;
141 magL1_Pcuac=mag_Pcuac;
142 magL1_Pcudc=mag_Pcudc;
143 magL1_Pfe=mag_Pfe;
144
145 %TRANSFORMER
147 mag_deltaB=magTR_deltaB;
148 mag_Bdc=magTR_Bdc;
149 mag_mat=magTR_mat;
150 mag_core=magTR_core;
151 mag_T=magTR_T;
152 mag_Rdc=magTR_Rdc;
153 mag_FR=magTR_FR;
154 %CALCULATED INPUT DATA
155 mag_I=[conv_Iin/2 0];
156 mag_Iac=[sqrt(conv_d01*conv_un1t01+conv_d12*conv_un1t12+conv_d23*c
onv_un1t23+conv_d34*conv_un1t34-(conv_Iin/2)^2)
sqrt(2*conv_d12*conv_un2t12)];
157 mag_fs=conv_fs;
158 mag_feq=2/pi^2*(1/(conv_t2-conv_t1)+1/(conv_t4-conv_t3));
159 %LIBRARY CALL
160 run ../lib/lib_mag.m;
162 magTR_Pcuac=mag_Pcuac;
163 magTR_Pcudc=mag_Pcudc;
164 magTR_Pfe=mag_Pfe;
175
B.12 C-code used in the microcontroller

1 /* Modulator - phase shifted PWM signal generator by
pak@iet.aau.dk */
2 #define USE_PEC
3
4 #include <xc161.h>
5 #include <stdio.h>
6 #include <intrins.h>
7 //#include <math.h>
8
9 #define CPU_CLK ((unsigned long) 20000000)
10
11 #pragma PECDEF(7)
16
17 // Macros useful for dealing with Interrupt and PEC registers
18 #define IC_IE(x) (((x) == 0) ? 0x0000 : 0x0040)
19 #define IC_ILVL(x) (((x) << 2) & 0x003C)
20 #define IC_GLVL(x) ((x) & 0x0003)
21 #define IC_PEC(x) (((x) & 0x0007) | 0x0038)
22
23 unsigned long f;
24 unsigned int i, n, step, ticks, duty_tmp, rel_tmp;
25 unsigned int bdata reload[1];
26 unsigned int bdata duty[1];
27 unsigned int bdata ooff[1];
28 unsigned int chrefon[9], chrefoff[9];
29
30
31 void channel_inc (void) interrupt CC1_CC4=20 {
32 if (n<8) {
33 n=n+1;
34 rel_tmp = 0xFFFF - ((unsigned long) (CPU_CLK/f));
35 ticks=(0xFFFF-rel_tmp);
36 reload[0]=rel_tmp;
37 step=ticks/n;
38 i=0;
39 while (i<n){
40 chrefon[i]=0xFFFF-i*step-step/2;
41 chrefoff[i]=chrefon[i];
42 i++;
43 }
44 i=n;
45 while (i<9){
46 chrefon[i]=0x0;
47 chrefoff[i]=rel_tmp+ticks/2; // turn-off impulses
are always available
48 i++;
49 }
50
51 CC1_CC11=chrefon[0]; //P2.11
52 CC1_CC10=chrefon[1]; //P2.10
53 CC1_CC9=chrefon[2]; //P2.9
176
55
56 CC1_CC12=chrefon[4]; //P2.12
57 CC1_CC13=chrefon[5]; //P2.13
58 CC1_CC14=chrefon[6]; //P2.14
59 CC1_CC15=chrefon[7]; //P2.15
60
61 CC2_CC18=chrefoff[0]; //P9.2
65
70 }
71 }
72
73
74 void channel_dec (void) interrupt CC1_CC5=21 {
75 if (n>0) {
76 n=n-1;
77 rel_tmp = 0xFFFF - ((unsigned long) (CPU_CLK/f));
78 ticks=(0xFFFF-rel_tmp);
79 reload[0]=rel_tmp;
80 step=ticks/n;
81 i=0;
82 while (i<n){
83 chrefon[i]=0xFFFF-i*step-step/2;
84 chrefoff[i]=chrefon[i];
85 i++;
86 }
87 i=n;
88 while (i<9){
89 chrefon[i]=0x0;
90 chrefoff[i]=rel_tmp+ticks/2; // turn-off impulses
are always available
91 i++;
92 }
93
94 CC1_CC11=chrefon[0]; //P2.11
95 CC1_CC10=chrefon[1]; //P2.10
98
99 CC1_CC12=chrefon[4]; //P2.12
100 CC1_CC13=chrefon[5]; //P2.13
101 CC1_CC14=chrefon[6]; //P2.14
102 CC1_CC15=chrefon[7]; //P2.15
103
108
177
113 }
114 if (n==0) { // decrease gain when all modules are down -
avoid inrush current when modules turn-on again
115 //gain=11;
116 duty[0]=reload[0]+1;
117 }
118 }
119
120 void pwm_inc (void) interrupt CC2_CC30=69 {
121 if (n>0){
122 if (duty[0]<0xFFFA) {
123 if (duty[0]<(reload[0]+ticks/2))
124 duty_tmp=duty[0]+3 ;
125 else
126 duty_tmp=duty[0]+1;
127 duty[0]=duty_tmp;
128 }
129 }
130 else if (n==0 & f<60000){
131 f = f + 1000;
132 reload[0]=0xFFFF-((unsigned long) (CPU_CLK/f));
133 ticks=(0xFFFF-reload[0]);
134 CC1_T0REL=reload[0];
135 CC2_T7REL=reload[0]+36;
136 CC2_CC18=reload[0]+ticks/2; //P9.2
144 }
145
146 }
147
148 void pwm_dec (void) interrupt CC2_CC31=70 {
149 if (n>0){
150 if (duty[0]>reload[0]+1) {
151 if (duty[0]<(reload[0]+ticks/2))
152 duty_tmp=duty[0]-3 ;
153 else
154 duty_tmp=duty[0]-1;
155 duty[0]=duty_tmp;
156 }
157 }
158 else if (n==0 & f>20000){
159 f = f - 1000;
160 reload[0]=0xFFFF-((unsigned long) (CPU_CLK/f));
161 ticks=(0xFFFF-reload[0]);
178

172 }
173 }
174
175 //void adc_to_duty (void) interrupt ADC_IC=40 {
176 // duty_tmp=reload[0] + ((double) ((ADC_DAT & 0x00FF) +
256)/512)*ticks;
177 // duty[0]=duty_tmp;
178 //}
179
180 //void frequency_setup (unsigned int rel){
181 //}
182
183 //void serial_setup (void){
184 // DP3 |= 0x400;
185 // DP3 &= ~0x0800;
186 // P3 |= 0x0400;
187
188 // S0BG = 0x0040;
189 // S0CON = 0x8011;
190
191 // S0TBUF = 0x0040;
192 // S0RBUF = 0x000F;
193 //}
194 //****************************************************************
******************************
195 void main (void) {
196 // Port direction and output type setup
197 DP2=0xFF00;
198 ODP2=0x0000; //push pull outputs
199 ALTSEL0P2=0xFF00; // alternate function of P2 - required in
non-staggered mode od CAPCOM1
200 POCON2=0x4400; // output driver control strong driver soft
edge
201
202 DP9=0x00FF;
203 ALTSEL1P9=0x00FF;
204 //POCON9=0x0044; // output driver control strong driver soft
edge
205
206 DP7=0x0030;
207 ALTSEL1P7=0x0030;
208 //POCON7=0x0040; // output driver control strong driver soft
edge
209
210 //PICON=0x0060; // 0x0060 - use cmos treshold on ports P6 &
P7
211
212 // Frequency f=(305Hz-20MHz) @ 20MHz
213 f=30000;
214 reload[0]=0xFFFF-((unsigned long) (CPU_CLK/f)); //
initial value
215 ticks=(0xFFFF-reload[0]); //initial value
216
217 // A-D conversion setup
218 //P5DIDIS_P0=1; // pin as AD converter input, alternate
function
219 //ADC_CON=0xF210; //0xB210
220 //ADC_CON1=0x1000; // 1000 - 8-bit resolution
179
221
222 // CAPCOM setup
223
226
227 CC1_T01CON=0x0000; //0x0707;
228 CC1_IOC=0x0004; //0x0004 non-staggered mode -
CC_clock=CPU_CLK
229
230 CC1_M0=0x0004; // 0x7 - T0/T7 Compare Mode 3 (one event /
period, IRQ + pin)
231 CC1_M1=0x0011; // 0x4 - T0/T7 Compare Mode 0 (many events /
period, IRQ only)
232 CC1_M2=0x7777; // 0x1 - T0/T7 Capture Mode, Rising Edge; 0x2
- T0/T7 Capture Mode, Falling Edge;
233 CC1_M3=0x7777; //
234
235 CC2_T78CON=CC1_T01CON; // the same settings like T0 & T1
236 CC2_IOC=CC1_IOC; // the same settings like T0 & T1
237
238 CC2_M4=0x7777;
239 CC2_M5=0x0077;
240 CC2_M6=0x0000;
241 CC2_M7=0x1177;
242
247
252
253 // Interrupts and PEC setup
254 CC1_CC0IC=IC_IE(1)|IC_PEC(7); // Reload T7 on "duty" compare
event in CC0
255 PECC7 = 0x00FF;
256 DSTP7 = (unsigned int) &CC2_T7;
257 SRCP7 = _sof_ (reload);
258
259 CC1_T0IC=IC_IE(1)|IC_PEC(5); // Reload duty ratio CC0 on T0
overflow
260 PECC5 = 0x00FF;
261 DSTP5 = (unsigned int) &CC1_CC0;
262 SRCP5 = _sof_ (duty);
263
264 CC1_CC4IC=IC_IE(1)|IC_ILVL(12)|IC_GLVL(0); // Active
channels increase P6.4
265 CC1_CC5IC=IC_IE(1)|IC_ILVL(12)|IC_GLVL(1); // Active
channels decrease P6.5
266
267 CC2_CC30IC=IC_IE(1)|IC_ILVL(12)|IC_GLVL(2); // PWM increase
P7.6
268 CC2_CC31IC=IC_IE(1)|IC_ILVL(12)|IC_GLVL(3); // PWM decrease
P7.7
269
180
270 // Global interrupt enable

271 PSW_IEN=1;
272
273 // Peripherials run
274 CC2_T78CON_T7R=1;
275 CC1_T01CON_T0R=1;
276
277 // Initial variable values
278 n=0; //initial number of active channels
279 duty[0]=reload[0]+1;
280 while(1){
281
282 };
283 }
181
Bibliography
[1] Z. Fu and W. Qu, "Implementation of a current share method for paralleling of
DC/DC converter without control interconnection," in Electrical Machines and
Systems, 2003. ICEMS 2003. Sixth International Conference on, 2003, pp. 394-
397 vol.1.
[2] P. Klimczak and P. Munk-Nielsen, "A single switch dual output non-isolated
[3] G. Hoogers, Fuel cell technology handbook CRC Press, 2003.
[4] W. Vielstich, A. Lamm, and H. A. Gasteiger, Handbook of fuel cells
fundamentals, technology, and applications: Wiley, 2003.
[5] W. Rong-Jong and D. Rou-Yong, "High step-up converter with coupled-
inductor," Power Electronics, IEEE Transactions on, vol. 20, pp. 1025-1035,
2005.
[6] P. Zacharias, B. Sahan, S. V. Araújo, F. L. M. Antunes, and R. T. Bascopé,
"Analysis and Proposition of a PV Module Integrated Converter with High
Voltage Gain Capability in a non-isolated Topology," in The 7th International
Conference on Power Electronics Daegu, Korea: IEEE, 2007.
[7] O. B. Cruz and L. D. Olavarrieta, "A Bird's Eye View of Materials and
Manufacturing Processes for Photovoltaic Cells," in Electronics,
Communications and Computers, 2005. CONIELECOMP 2005. Proceedings.
15th International Conference on, 2005, pp. 251-256.
[8] A. Goetzberger and V. U. Hoffmann, Photovoltaic solar energy generation:
Springer, 2005.
[9] "Global Market Outlook Until 2013," in
http://www.epia.org/fileadmin/EPIA_docs/publications/epia/Global_Market_Ou
tlook_Until_2013.pdf: European Photovoltaic Industry Association, 2009.
[10] T. Kerekes, R. Teodorescu, and U. Borup, "Transformerless Photovoltaic
Inverters Connected to the Grid," in Applied Power Electronics Conference,
APEC 2007 - Twenty Second Annual IEEE, 2007, pp. 1733-1737.
[11] H. Patel and V. Agarwal, "A Single-Stage Single-Phase Transformer-Less
Doubly Grounded Grid-Connected PV Interface," Energy conversion, ieee
transactions on, vol. 24, pp. 93-101, 2009.
[12] "IEEE guide for terrestrial photovoltaic power system safety," IEEE Std 1374-
1998, 1998.
[13] W. Vielstich, Handbook of fuel cells fundamentals, technology, and applications
editors, Wolf Vielstich, Arnold Lamm, Hubert A. Gasteiger: Wiley, 2003.
182
[14] Y. Kanthaphayao and C. Boonmee, "Dual-output DC-DC power supply without

transformer," in TENCON 2004. 2004 IEEE Region 10 Conference, 2004, pp.
41-44 Vol. 4.
[15] P. Klimczak and S. Munk-Nielsen, "A single switch dual output non-isolated
[16] F. L. Luo, "Double-output Luo converters, an advanced voltage-lift technique,"
Electric Power Applications, IEE Proceedings -, vol. 147, pp. 469-485, 2000.
[17] O. Won-Sik, H. Sang-Kyoo, C. Seong-Wook, and M. Gun-Woo, "Three phase
three-level PWM switched voltage source inverter with zero neutral point
potential," in Power Electronics Specialists Conference, 2004. PESC 04. 2004
IEEE 35th Annual, 2004, pp. 4405-4410 Vol.6.
[18] Z. Qun and F. C. Lee, "High-efficiency, high step-up DC-DC converters,"
Power Electronics, IEEE Transactions on, vol. 18, pp. 65-73, 2003.
[19] K. C. Tseng and T. J. Liang, "Novel high-efficiency step-up converter," Electric
Power Applications, IEE Proceedings -, vol. 151, pp. 182-190, 2004.
[20] H. Shu-Hung Chung, "Design and analysis of a switched-capacitor-based step-
up DC/DC converter with continuous input current," Circuits and Systems I:
Fundamental Theory and Applications, IEEE Transactions on [see also Circuits
and Systems I: Regular Papers, IEEE Transactions on], vol. 46, pp. 722-730,
1999.
[21] M. On-Cheong, W. Yue-Chung, and A. Ioinovici, "Step-up DC power supply
based on a switched-capacitor circuit," Industrial Electronics, IEEE
[22] O. Abutbul, A. Gherlitz, Y. Berkovich, and A. Ioinovici, "Step-up switching-
mode converter with high voltage gain using a switched-capacitor circuit,"
Circuits and Systems I: Fundamental Theory and Applications, IEEE
Transactions on [see also Circuits and Systems I: Regular Papers, IEEE
Transactions on], vol. 50, pp. 1098-1102, 2003.
[23] K. K. Law, K. W. E. Cheng, and Y. P. B. Yeung, "Design and analysis of
switched-capacitor-based step-up resonant converters," Circuits and Systems I:
Regular Papers, IEEE Transactions on [Circuits and Systems I: Fundamental
Theory and Applications, IEEE Transactions on], vol. 52, pp. 943-948, 2005.
[24] D. A. Grant and Y. Darroman, "Extending the tapped-inductor DC-to-DC
converter family," Electronics Letters, vol. 37, pp. 145-146, 2001.
[25] Z. Qun, T. Fengfeng, and F. C. Lee, "A front-end DC/DC converter for network
server applications," in Power Electronics Specialists Conference, 2001. PESC.
2001 IEEE 32nd Annual, 2001, pp. 1535-1539 vol. 3.
[26] C. E. A. Silva, R. P. T. Bascope, and D. S. Oliveira, "Proposal of a New High
Step-Up Converter for UPS Applications," in Industrial Electronics, 2006 IEEE
International Symposium on, 2006, pp. 1288-1292.
[27] D. M. Van de Sype, K. De Gusseme, B. Renders, A. R. Van den Bossche, and J.
A. Melkebeek, "A single switch boost converter with a high conversion ratio," in
Applied Power Electronics Conference and Exposition, 2005. APEC 2005.
Twentieth Annual IEEE, 2005, pp. 1581-1587 Vol. 3.
183
[28] R. J. Wai and R. Y. Duan, "High-efficiency DC/DC converter with high voltage
gain," Electric Power Applications, IEE Proceedings -, vol. 152, pp. 793-802,
2005.
[29] B. Ju-Won, R. Myung-Hyo, K. Tae-Jin, Y. Dong-Wook, and K. Jong-Soo,
"High boost converter using voltage multiplier," in Industrial Electronics
Society, 2005. IECON 2005. 31st Annual Conference of IEEE, 2005, p. 6 pp.
[30] L. C. Franco, L. L. Pfitscher, and R. Gules, "A new high static gain nonisolated
DC-DC converter," in Power Electronics Specialist Conference, 2003. PESC
'03. 2003 IEEE 34th Annual, 2003, pp. 1367-1372 vol.3.
[31] G. V. Torrico-Bascope, R. P. Torrico-Bascope, D. S. Oliviera, F. L. M. Antunes,
S. V. Araujo, and C. G. C. Branco, "A Generalized High Voltage Gain Boost
Converter Based on Three-State Switching Cell," in The 32nd Annual
Conference of the IEEE Industrial Electronics Society IECON'06, Paris, France,
2006, pp. 1927-1932.
[32] L. Fang Lin, "Luo-Converters, a series of new DC-DC step-up (boost)
conversion circuits," in Power Electronics and Drive Systems, 1997.
Proceedings., 1997 International Conference on, 1997, pp. 882-888 vol.2.
[33] F. L. Luo, "Negative output Luo converters: voltage lift technique," Electric
[34] F. L. Luo, "Positive output Luo converters: voltage lift technique," Electric
[35] F. L. Y. H. Luo, Advanced DC/DC converters. Boca Raton: CRC Press, 2004.
[36] B. Farhangi and S. Farhangi, "Comparison of z-source and boost-buck inverter
topologies as a single phase transformer-less photovoltaic grid-connected power
conditioner," in Power Electronics Specialists Conference, 2006. PESC '06. 37th
IEEE, 2006, pp. 1-6.
[37] L. Poh Chiang, D. M. Vilathgamuwa, Y. S. Lai, C. Geok Tin, and Y. Li, "Pulse-
width modulation of Z-source inverters," Power Electronics, IEEE Transactions
on, vol. 20, pp. 1346-1355, 2005.
[38] L. Shiguo, Y. Zhihong, L. Ray-Lee, and F. C. Lee, "A classification and
evaluation of paralleling methods for power supply modules," in Power
Electronics Specialists Conference, 1999. PESC 99. 30th Annual IEEE, 1999,
pp. 901-908 vol.2.
[39] L. Po-Wa, L. Yim-Shu, D. K. W. Cheng, and L. Xiu-Cheng, "Steady-state
analysis of an interleaved boost converter with coupled inductors," Industrial
Electronics, IEEE Transactions on, vol. 47, pp. 787-795, 2000.
[40] W. Wei and L. Yim-Shu, "A two-channel interleaved boost converter with
reduced core loss and copper loss," in Power Electronics Specialists Conference,
2004. PESC 04. 2004 IEEE 35th Annual, 2004, pp. 1003-1009 Vol.2.
[41] "Ferroxcube Product Selection Guide," in
http://www.ferroxcube.com/appl/info/PSG2003.pdf, 2003.
[42] P. Klimczak and S. Munk-Nielsen, "Comparative study on paralleled vs. scaled
dc-dc converters in high voltage gain applications," in Power Electronics and
Motion Control Conference, 2008. EPE-PEMC 2008. 13th, 2008, pp. 108-113.
184
[43] N. Mohan, T. M. Undeland, and W. P. Robbins, Power electronics : converters,

applications, and design, 3rd ed.: John Wiley & Sons, 2003.
[44] G. V. T. Bascope and I. Barbi, "Generation of a family of non-isolated DC-DC
PWM converters using new three-state switching cells," in Power Electronics
Specialists Conference, 2000. PESC 00. 2000 IEEE 31st Annual, 2000, pp. 858-
863 vol.2.
[45] H. Kragh, Modeling, Analysis and Optimisation of Power Electronic Circuits for
Low-Cost Drives, PhD thesis, Aalborg University, 2000
[46] I. Budihardjo, P. O. Lauritzen, K. Y. Wong, R. B. Darling, and H. A. Mantooth,
"Defining standard performance levels for power semiconductor devices," in
Industry Applications Conference, 1995. Thirtieth IAS Annual Meeting, IAS '95.,
Conference Record of the 1995 IEEE, 1995, pp. 1084-1090 vol.2.
[47] R. W. Erickson and D. Maksimovic, Fundamentals of power electronics, 2nd ed.
Norwell, Mass.: Kluwer Academic, 2001.
[48] J. Brown, "Modeling the switching performance of a MOSFET in the high side
of a non-isolated buck converter," Power Electronics, IEEE Transactions on,
vol. 21, pp. 3-10, 2006.
[49] Z. J. Shen, X. Yali, C. Xu, F. Yue, and P. Kumar, "Power MOSFET Switching
Loss Analysis: A New Insight," in Industry Applications Conference, 2006. 41st
IAS Annual Meeting. Conference Record of the 2006 IEEE, 2006, pp. 1438-
1442.
[50] J. Waite and T. G. Wilson, Jr., "Use of simulation to understand and predict
switching losses in a two-stage power factor corrected AC-to-DC converter," in
Applied Power Electronics Conference and Exposition, 1996. APEC '96.
Conference Proceedings 1996., Eleventh Annual, 1996, pp. 88-95 vol.1.
[51] C. L. Ma and P. O. Lauritzen, "A simple power diode model with forward and
reverse recovery," Power Electronics, IEEE Transactions on, vol. 8, pp. 342-
346, 1993.
[52] S. P. Pendharkar, M. Trivedi, and K. Shenai, "Dynamics of reverse recovery of
high-power P-i-N diodes," Electron Devices, IEEE Transactions on, vol. 43, pp.
142-149, 1996.
[53] R. Burkel and T. Schneider, "Fast Recovery Epitaxial Diodes (FRED)
Characteristics - Applications - Examples," in
http://www.ixyspower.com/images/technical_support/Application%20Notes%20
By%20Topic/FREDs,%20Schottky%20and%20GaAS%20Diodes/IXAN0044.pdf,
I. P. A. Note, Ed., 1999.
[54] S. L. M. Berleze and R. Robert, "Skin and proximity effects in nonmagnetic
conductors," Education, IEEE Transactions on, vol. 46, pp. 368-372, 2003.
[55] K. W. Klontz, "Skin and proximity effects in multi-layer transformer windings
of finite thickness," in Industry Applications Conference, 1995. Thirtieth IAS
Annual Meeting, IAS '95., Conference Record of the 1995 IEEE, 1995, pp. 851-
858 vol.1.
[56] J. D. Pollock and C. R. Sullivan, "Modelling Foil Winding Configurations with
Low AC and DC Resistance," in Power Electronics Specialists Conference,
2005. PESC '05. IEEE 36th, 2005, pp. 1507-1512.
185
[57] A. Reatti and M. K. Kazimierczuk, "Comparison of various methods for

calculating the AC resistance of inductors," Magnetics, IEEE Transactions on,
vol. 38, pp. 1512-1518, 2002.
[58] A. M. Urling, V. A. Niemela, G. R. Skutt, and T. G. Wilson, "Characterizing
high-frequency effects in transformer windings-a guide to several significant
articles," in Applied Power Electronics Conference and Exposition, 1989.
APEC' 89. Conference Proceedings 1989., Fourth Annual IEEE, 1989, pp. 373-
385.
[59] J. P. Vandelac and P. D. Ziogas, "A novel approach for minimizing high-
frequency transformer copper losses," Power Electronics, IEEE Transactions
on, vol. 3, pp. 266-277, 1988.
[60] N. Xi and C. R. Sullivan, "An improved calculation of proximity-effect loss in
high-frequency windings of round conductors," in Power Electronics Specialist
Conference, 2003. PESC '03. 2003 IEEE 34th Annual, 2003, pp. 853-860 vol.2.
[61] J. D. Pollock and C. R. Sullivan, "Gapped-inductor foil windings with low AC
and DC resistance," in Industry Applications Conference, 2004. 39th IAS Annual
Meeting. Conference Record of the 2004 IEEE, 2004, p. 557.
[62] F. Robert, P. Mathys, and J. P. Schauwers, "Ohmic losses calculation in SMPS
transformers: numerical study of Dowell's approach accuracy," Magnetics, IEEE
[63] J. Reinert, A. Brockmeyer, and R. W. A. A. De Doncker, "Calculation of losses
in ferro- and ferrimagnetic materials based on the modified Steinmetz equation,"
Industry Applications, IEEE Transactions on, vol. 37, pp. 1055-1061, 2001.
[64] C. D. Graham, "Physical origin of losses in conducting ferromagnetic materials
(invited)." vol. 53: AIP, 1982, pp. 8276-8280.
[65] M. Albach, T. Durbaum, and A. Brockmeyer, "Calculating core losses in
transformers for arbitrary magnetizing currents a comparison of different
approaches," in Power Electronics Specialists Conference, 1996. PESC '96
Record., 27th Annual IEEE, 1996, pp. 1463-1468 vol.2.
[66] L. Jieli, T. Abdallah, and C. R. Sullivan, "Improved calculation of core loss with
nonsinusoidal waveforms," in Industry Applications Conference, 2001. Thirty-
Sixth IAS Annual Meeting. Conference Record of the 2001 IEEE, 2001, pp.
2203-2210 vol.4.
[67] A. Van den Bossche, V. C. Valchev, and G. B. Georgiev, "Measurement and
loss model of ferrites with non-sinusoidal waveforms," in Power Electronics
Specialists Conference, 2004. PESC 04. 2004 IEEE 35th Annual, 2004, pp.
4814-4818 Vol.6.
[68] K. Venkatachalam, C. R. Sullivan, T. Abdallah, and H. A. T. H. Tacca,
"Accurate prediction of ferrite core loss with nonsinusoidal waveforms using
only Steinmetz parameters," in Computers in Power Electronics, 2002.
Proceedings. 2002 IEEE Workshop on, 2002, pp. 36-41.
[69] C. A. Baguley, B. Carsten, and U. K. Madawala, "The Effect of DC Bias
Conditions on Ferrite Core Losses," Magnetics, IEEE Transactions on, vol. 44,
pp. 246-252, 2008.
186
[70] A. Brockmeyer, "Experimental evaluation of the influence of DC-

premagnetization on the properties of power electronic ferrites," in Applied
Power Electronics Conference and Exposition, 1996. APEC '96. Conference
Proceedings 1996., Eleventh Annual, 1996, pp. 454-460 vol.1.
[71] W. Hongfang and F. Wang, "Power MOSFETs Paralleling Operation for High
Power High Density Converters," in Industry Applications Conference, 2006.
41st IAS Annual Meeting. Conference Record of the 2006 IEEE, 2006, pp. 2284-
2289.
[72] Y. Zhai and L. Vu-Quoc, "Analysis of Power Magnetic Components With
Nonlinear Static Hysteresis: Proper Orthogonal Decomposition and Model
Reduction," Magnetics, IEEE Transactions on, vol. 43, pp. 1888-1897, 2007.
[73] "American National Standard guide on the application and evaluation of EMI
power-line filters for commercial use," ANSI C63.13-1991, 1991.
[74] "IEEE standard procedure for measuring conducted emissions in the range of
300 kHz to 25 MHz from television and FM broadcast receivers to power lines,"
ANSI/IEEE Std 213-1987, 1988.
[75] D. L. Jones, "PCB Design Tutorual," in
http://alternatezone.com/electronics/files/PCBDesignTutorialRevA.pdf, 2004.
[76] F. Dong Tan, J. L. Vollin, and S. M. Cuk, "A practical approach for magnetic
core-loss characterization," Power Electronics, IEEE Transactions on, vol. 10,
pp. 124-130, 1995.
[77] "Farnell," in http://www.farnell.com.
[78] "International Rectifier," in http://www.irf.com.
[79] M. H. Rashid, Power electronics handbook Academic Press, 2001.
[80] "Infineon Technologies," in http://www.infineon.com.
[81] "Magnetics Cores for SMPS," in http://www.mag-inc.com/pdf/ps-01.pdf.
[82] A. V. d. Bossche and V. C. Valchev, Inductors and tranformers for power
electronics CRC Press, 2005
[83] C. W. T. McLyman, Transformer and inductor design handbook: Marcel
Dekker, 2004.
[84] "Magnetics," in http://www.mag-inc.com.
[85] "Ferroxcube," in http://www.ferroxcube.com.
[86] S. Cuk, "New magnetic structures for switching converters," Magnetics, IEEE
[87] C. Rong-Tai and C. Yung-Yaw, "A novel single stage push pull converter with
integrated magnetics and ripple-free input current," in Power Electronics
Specialists Conference, 2004. PESC 04. 2004 IEEE 35th Annual, 2004, pp.
3848-3853 Vol.5.
[88] N. Hur and N. Kwanghee, "A robust load-sharing control scheme for parallel-
connected multisystems," Industrial Electronics, IEEE Transactions on, vol. 47,
pp. 871-879, 2000.
[89] T. Kohama, T. Ninomiya, M. Shoyama, and F. Ihara, "Dynamic analysis of
parallel-module converter system with current balance controllers," in
187
Telecommunications Energy Conference, 1994. INTELEC '94., 16th

International, 1994, pp. 190-195.
[90] D. J. Perreault, K. Sato, R. L. Selders, Jr., and J. G. Kassakian, "Switching-
ripple-based current sharing for paralleled power converters," Circuits and
Systems I: Fundamental Theory and Applications, IEEE Transactions on [see
also Circuits and Systems I: Regular Papers, IEEE Transactions on], vol. 46,
pp. 1264-1274, 1999.
[91] D. J. Perreault, R. L. Selders, and J. G. Kassakian, "Frequency-based current-
sharing techniques for paralleled power converters," in Power Electronics
Specialists Conference, 1996. PESC '96 Record., 27th Annual IEEE, 1996, pp.
1073-1079 vol.2.
[92] H. H. C. Iu and V. Pjevalica, "Experimental study of instabilities in two parallel-
connected boost converters under current mode control," in Power Electronics
and Applications, 2005 European Conference on, 2005, p. 8 pp.
[93] S. K. Mazumder, "Stability analysis of parallel DC-DC converters," Aerospace
and Electronic Systems, IEEE Transactions on, vol. 42, pp. 50-69, 2006.
[94] J. M. Zhang, X. G. Xie, X. K. Wu, and Q. Zhaoming, "Stability study for
paralleled DC/DC converters," in Power Electronics Specialists Conference,
2004. PESC 04. 2004 IEEE 35th Annual, 2004, pp. 1569-1575 Vol.2.
[95] J. G. Kassakian and D. Lau, "An analysis and experimental verification of
parasitic oscillations in parralleled power MOSFET's," Electron Devices, IEEE
[96] P. Wallmeier, "Pre-optimization of linear and nonlinear inductors using area-
product formulation," in Industry Applications Conference, 2002. 37th IAS
Annual Meeting. Conference Record of the, 2002, pp. 2445-2450 vol.4.
[97] R. Ayyanar, R. Giri, and N. Mohan, "Active input-voltage and load-current
sharing in input-series and output-parallel connected modular DC-DC converters
using dynamic input-voltage reference scheme," Power Electronics, IEEE
[98] "TOELLNER Electronic Instrumente GmbH," in
http://www.toellner.de/pdf_gb/8871.pdf.
[99] "zentro-elektrik GmbH KG " in http://www.zentro-
elektrik.de/inhalte/Elektronische_DC_Lasten_2007_12_11_20_09_17/ELA_500
_EL1000_series.pdf.
[100] "GOOD WILL INSTRUMENT CO., LTD. - GW Instek," in
http://www.gwinstek.com/html/en/showproductdetal-e.asp?pd1_sn=120.
[101] "Fluke Deutschland GmbH," in
http://www.fluke.de/comx/show_product.aspx?locale=dede&pid=35911.
[102] "Tektronix, Inc.," in http://www.tek.com.
[103] "LEM Probes," in http://www.lem.com
http://www.techedu.com/pdf/LEM/PR50.pdf.
[104] M. Nymand and M. A. E. Andersen, "A new approach to high efficiency in
isolated boost converters for high-power low-voltage fuel cell applications," in
Power Electronics and Motion Control Conference, 2008. EPE-PEMC 2008.
13th, 2008, pp. 127-131.
188
[105] M. Nymand and M. A. E. Andersen, "New Primary-Parallel Boost Converter for

High-Power High-Gain Applications," in Applied Power Electronics Conference
and Exposition, 2009. APEC 2009. Twenty-Fourth Annual IEEE, 2009, pp. 35-
39.

Modular Power Electronic Converters (PDFDrive)

Uploaded by

Copyright:

Available Formats

Modular Power Electronic Converters (PDFDrive)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Modular Power Electronic Converters (PDFDrive)

Uploaded by

Copyright:

Available Formats

Aalborg Universitet

Modular Power Electronic Converters in the Power Range 1 to 10 kW

Link to publication from Aalborg University

Citation for published version (APA):

Downloaded from vbn.aau.dk on: januar 23, 2018

Aalborg University, Denmark

3.4.2 Reverse recovery loss ............................................................................. 41

Abbreviations and Symbols

B flux density, flux density T

Pin input power W

1.1 Background and motivation

1.2 Potential applications

1.2.1 Fuel cell

1.2.3 Uninterruptible power supply

1.3 Overview on a non-isolated system

Figure 1.1 A non-isolated power conversion system with a full-bridge inverter

Figure 1.2 A non-isolated power conversion system with a half-bridge inverter

Figure 1.3 A non-isolated power conversion system with a SVS inverter

Figure 1.1 presents a non-isolated converter with a full-bridge grid connected

required [14-16]. A simple single-input dual-output step-up converter is presented

1.4 Problem definition

• Modularization – parallel operation of several modules and its impact on

1.4.1 Project limitations

1.4.2 Used tools

1.5 Thesis outline

1.6 List of publications

2.1 General overview on non-isolated dc-dc

range of the output power. Number of components is similar in both converters,

2.2 Review of topologies

2.2.1 Basic boost converter

Figure 2.1 Diagram of a basic boost converter

Figure 2.2 Waveforms of a basic boost converter

volt×ampT1 = VT1(off) ⋅ I T1(rms) ≈ Vout ⋅ M ⋅ I out = M ⋅ Pout (2.5)

volt×amp D1 = Vout ⋅ I out ≈ Pout (2.6)

Figure 2.3 Diagram of a 2-phase boost converter

2.2.2 Multi-phase boost converter

Figure 2.4 Waveforms of a 2-phase boost

Figure 2.7 Waveforms of a 2-phase boost converter with a coupled inductor

In CCM the voltage gain M is given by (2.15). The transistors’ blocking

volt×ampT1 = VT1(off) ⋅ I T1(rms) ≈ Vout ⋅ M ⋅ D ⋅ I out = M ⋅ D ⋅ Pout (2.17)

2.2.3 Cascaded boost converter

Figure 2.8 Diagram of a cascaded boost converter

volt×ampT = volt×ampT1 + volt×ampT2 =

2.2.4 Center tapped boost converter

Figure 2.9 Diagram of a center-tapped boost converter

Figure 2.10 Waveforms of the center tapped boost converter

volt×ampT1 = VT1(off) ⋅ I T1(rms) ≈

Figure 2.12 Diagram of a flyback-boost converter

2.2.5 Non-isolated flyback-boost converter

Figure 2.11 Main waveforms of the flyback-boost converter

volt×amp T1 = VT1(off) ⋅ I T1(rms) ≈

2.2.6 Non-isolated push-pull-boost converter

Figure 2.13 Diagram of a non-isolated push-pull-boost converter [6, 31, 44]

Figure 2.14 Main waveforms of the non-isolated push-pull-boost converter

volt×ampT = 2 ⋅ volt×ampT1 (2.64)

2.2.7 Non-isolated two-inductor-boost converter

Figure 2.15 Diagram of a non-isolated two-inductor-boost converter

Figure 2.16 Waveforms of the non-isolated two-inductor boost converter

volt×ampT1 = VT1(off) ⋅ I T1(rms) ≈