Академический Документы
Профессиональный Документы
Культура Документы
by
Djilali Hamza
Queens University
Kingston, Ontario, Canada
(November, 2011)
Abstract
Increasing power density of switch mode power supplies, by increasing their switching frequency
has becoming a challenging obstacle for EMI mitigation. The passive EMI suppression technique
has always been the primary solution to fulfill the EMC requirement in terms of conducted
emission limits. However, the call for stringent power supplies specifications renders the passive
techniques less desirable, due to their increasing size and power losses. In other words, the greater
the power density of the converter, the bigger the passive filter. Other suppression techniques
such as the spread spectrum frequency modulation (SSFM), and soft switching, prove to have less
performance and much complex to implement. The active analog EMI filters provide the basic
noise suppression technique; however, their performance is dramatically impeded at higher
frequency. This solution requires an additional small size passive filter to complete the EMC
spectrum for conducted emissions.
Digital active filtering techniques offer advantages of flexibility, fewer external components and
reduced overall size and power losses as compared to conventional passive filtering techniques.
In this thesis DSP-based and FPGA-based EMI control techniques to mitigate the conducted
emissions of switch mode power converters are proposed. These techniques are implemented inlieu of the passive filtering techniques, by keeping equal or better performance. Moreover, these
solutions can be configured as a stand-alone or integrated into the converter digital controller
algorithm.
Finally, the proposed solutions are implemented into three types of power converters, namely, a
AC-DC power factor corrected converter, DC-AC micro-inverter for Photovoltaic application,
and DC-DC for Electric Vehicle (EV) battery charger. Analytical, simulation and experimental
results are provided to verify the proposed solutions.
ii
Acknowledgements
I would like to express my sincere gratitude and many thanks to Dr. Praveen Jain, the director of
ePOWER centre, for his support and confidence during the course of this work. Dr. jain's
extensive research vision has been a source of inspiration and an example for me to overcome all
the hurdles encountered in this research work.
I also thank my committee members, in particular, Dr. John E. Quaicoe of Memorial University
of Newfoundland, for their insights and valuable observations. In addition, I would like to extend
my thanks to my past and present colleagues, as well as, all the graduate students of Queen's
ePOWER centre for their cooperation and understanding.
Financial support in the form of tuition waver award under Queen's University Employee Tuition
Assistant Program, is gratefully acknowledged.
Finally, my heartfelt appreciation goes to my parents, my wife, my daughter and my son for their
love, support and patience.
iii
To my Parents,
My wife Mei
My daughter Myriam
& My son Yusuf
iv
Statement of Originality
(Required only for Division IV Ph.D.)
I hereby certify that all of the work described within this thesis is the original work of the author.
Any published (or unpublished) ideas and/or techniques from the work of others are fully
acknowledged in accordance with the standard referencing practices.
(Djilali Hamza)
(November, 2011)
Table of Contents
Abstract ............................................................................................................................................ ii
Acknowledgements ......................................................................................................................... iii
Statement of Originality................................................................................................................... v
Table of Contents ............................................................................................................................ vi
List of Figures .................................................................................................................................. x
Acronyms: ................................................................................................................................ xvii
Symbols: ..................................................................................................................................... xx
Chapter 1 Introduction ..................................................................................................................... 1
1.1
Background ...................................................................................................................... 1
1.2
1.2.1
1.2.2
1.3
1.4
1.5
B.2
D.2
ix
List of Figures
Fig. 1.1 Cost of EMC solution during the Design Lifecycle [2] ...................................................... 2
Fig. 1.2 EMI coupling mechanism block diagram ........................................................................... 3
Fig. 1.3 Inductive Coupling ............................................................................................................. 4
Fig. 1.4 Capacitive Coupling ........................................................................................................... 5
Fig. 1.5 CM and DM paths in power converter ............................................................................... 6
Fig. 1.6 Differential Mode currents is generated by the normal switching action of the MOSFET 7
Fig. 1.7Common Mode currents coupled to chassis through stray capacitance .............................. 8
Fig. 1.8 CM noise due to the Drain heat-sink stray capacitance ................................................... 9
Fig. 1.9 Possible loop areas for radiated emission in SMPS .......................................................... 10
Fig. 1.10 Fields of EMC ................................................................................................................ 12
Fig. 2.1 MOSFET Drain to Heat-sink stray capacitance, the CM current returns through the LISN
impedance (large return path) ........................................................................................................ 19
Fig. 2.2 Reducing the CM current by floating the heat-sink.......................................................... 20
Fig. 2.3 Reducing the CM current by screening/shielding the MOSFET ...................................... 20
Fig. 2.4 Reducing the CM noise current by using LC circuit ........................................................ 21
Fig. 2.5 CM current generation through core and primary-to-secondary stray capacitance .......... 22
Fig. 2.6 Transformer core is referenced to the DC-link, shorter current path................................ 22
Fig. 2.7 shield connection between the primary and the transformer core .................................... 23
Fig. 2.8 shielding primary and secondary winding ........................................................................ 23
Fig. 2.9 Transformer windings configuration; (a) single windings, fringing fields; (b) split
windings creates leakage fields that tend to cancel........................................................................ 25
Fig. 2.10 Schematic diagram of an input EMI filter for SMPS ..................................................... 26
Fig. 2.11 Basic passive EMI filter circuit configurations: (a) single stage LC-circuit, (b) -circuit,
(c) T-circuit, (d) Multistage LC-circuit
...................................................................................... 28
Fig. 2.12 Basic EMI filter Configuration for CM and DM attenuation ........................................ 28
Fig. 2.13 Insertion loss (IL) measurement: (a) Reference circuit, (b) Filter under test inserted ... 30
Fig. 2.14 Single ended CM equivalent circuit diagram ................................................................ 31
Fig. 2.15 Interface impedances of the passive EMI filter .............................................................. 33
Fig. 2.16 zero voltage transition (ZVS) circuit implemented in a DC/DC converter .................... 36
Fig. 2.17 ZVS Transition in main switch S1.................................................................................. 37
Fig. 2.18 HS Transition in main switch S1 .................................................................................... 37
Fig. 2.19 EMI profile with Soft Switching circuit ......................................................................... 37
Fig. 2.20 EMI Profile with Hard Switching ................................................................................... 38
Fig. 2.21 EMI Profile with ZVS and HS [38] ................................................................................ 39
Fig. 2.22 Effect of SSFM technique on an nth harmonic of a clock signal; (a) un-modulated clock;
(b) modulated clock ....................................................................................................................... 40
Fig. 2.23 Resulting spectrum of different modulating signal waveforms; (a) sine wave; (b) square
wave; (c) ramp ............................................................................................................................... 41
Fig. 2.24 Spectral content of the modulated signal ........................................................................ 43
Fig. 2.25 The configuration of the active input EMI filter............................................................. 46
Fig. 2.26: Equivalent harmonic circuit of the converter: (a) Equivalent circuit with passive filter
only, (b) Equivalent circuit with hybrid active and passive filters ................................................ 47
Fig. 2.27: Noise attenuation of the active circuit (Zin=50//50H, Rs=50, Cs=5F, Cin=20F,
k1=6106, k2 =100, N=15) .............................................................................................................. 49
Fig. 2.28: Injection transformer ..................................................................................................... 51
Fig. 2.29: Experimental circuit diagram ........................................................................................ 52
Fig. 2.30: Conducted EMI noise spectrum result with the input passive filter only ...................... 53
xi
Fig. 2.31: Conducted EMI noise spectrum result with the combination of passive and active input
EMI filters. ..................................................................................................................................... 54
Fig. 3.1 Application of DAEF versus PEF in power converters ................................................... 57
Fig. 3.2 General Scheme of the Digital Input EMI filter ............................................................... 59
Fig. 3.3 Impulse Sampling Model.................................................................................................. 60
Fig. 3.4 Circuit example of a Sampling time-domain analog signal .............................................. 62
Fig. 3.5 Simulation waveforms showing the sampling process using impulse function and the
recovered signal ............................................................................................................................. 62
Fig. 3.6 Block Diagram of an Analog-to-Digital Converter (ADC) .............................................. 63
Fig. 3.7 Block Diagram of a Digital-to-Analog Converter (DAC) ................................................ 64
Fig. 3.8 Feedback Diagram of the DSP-Based Digital EMI filter ................................................. 68
Fig. 3.9 Frequency response - magnitude of the DSP-based EMI filter ........................................ 70
Fig. 3.10 Frequency response - phase of the DSP-based EMI filter ............................................. 70
Fig. 3.11 Schematic block diagram of the simulation circuit ....................................................... 71
Fig. 3.12 Conducted Noise measurement without Input EMI Filter ............................................. 74
Fig. 3.13 Conducted Noise measurement with Passive Input EMI Filter ...................................... 74
Fig. 3.14 Conducted Noise measurement with Proposed Digital Input EMI Filter ....................... 75
Fig. 3.15 Output of the ADC and the DAC in time domain ......................................................... 76
Fig. 3.16 Conducted emissions testing experimental setup ........................................................... 77
Fig. 3.17 Input and output voltage signal of the proposed DAE Filter .......................................... 78
Fig. 3.18 Conducted emission spectrum of EUT without filters.................................................... 79
Fig. 3.19 Conducted emission spectrum of EUT with PEF ........................................................... 79
Fig. 3.20 Conducted emission spectrum of EUT with DAEF........................................................ 80
Fig. 3.21 Comparative attenuation between PEF and DAEF........................................................ 81
Fig. 4.1 Proposed FPGABEST block diagram connected to a buck converter.............................. 84
xii
Fig. A.1 OrCAD Simulation Schematic of the DC-DC power converter .................................... 160
Fig. A. 2 OrCAD Simulation Schematic of the DAEF Module.................................................. 161
xv
List of Tables
Table 2-1: Performance comparisons of the passive and the hybrid filters ................................... 54
Table 3-1 Simulated Converter Parameters for DSP-based EMI filter .......................................... 72
Table 3-2 Comparison of attenuation performance between PEF and DAEF using simulation
results ............................................................................................................................................. 75
Table 3-3 Comparison between experimental and simulation results ........................................... 80
Table 5-1 Characteristics Evaluation of different PV systems architectures ............................... 100
Table 6-1 Converter parameters.................................................................................................. 133
xvi
Acronyms:
AC
Alternative Current
ADC
Analog-to-Digital Converter
AEM
AM
Amplitude Modulation
ASIC
CFL
CISPR
ITC
CM
Common Mode
DAC
Digital-to-Analog Converter
DAEF
DC
Direct Current
DM
Differential Mode
DOD
Department of Defense
D-Q
Direct Quadrature
DSP
DSPBEST
EEC
EM
Electromagnetic
EMC
Electromagnetic Compatibility
EMI
Electromagnetic Interferences
xvii
ESR
EST
EV
Electrical Vehicle
FCC
FCHEV
FCV
FM
Frequency Modulation
FPGA
FPGABEST
HPF
High-Pass Filter
HS
Hard Switching
HV
HV
ICEV
IEC
IL
Insertion Loss
IPEM
LISN
LSB
LV
MOSFET
MPP
MPPT
MSB
PCB
PFC
PHEV
PI
PSD
PV
Photovoltaic
PWM
Pulse-Width-Modulation
RHP
Right Half-Plane
SMPS
SSCG
SSFMT
TF
Transfer Function
VHDL
ZOH
Zero-Order-Hold
ZCS
Zero-Current Switching
ZVS
xix
Symbols:
U0
Um
R1
R2
Rin
Lf and Rf
Cf and ESRcf
Rd and Cd
Zeq
ZIF
Input Impedance of the passive EMI filter loaded with the DC/DC converter with
the output port open
ZOF
Output Impedance of the passive EMI filter connected to an input DC bus with
the input port shorted
ZDC
Zin
Modulation index
Rate of modulation, %
Cs and Rs
Rb, Rb1 and Cb Bias circuit (resistance and capacitance) of the OPAmp
xx
Vref
in
iinj
Iin
Current fed back into the utility mains with the passive filter only
I'in
Current fed back into the utility mains with the hybrid filter
Av
Closed-loop voltage gain for the current feedback amplifier with unity gain
A(s)
k1 and k2
DC gain (typically 105 to 107) and the cutoff frequency of the OPAmp
vop
sw
Switching frequency
m(t)
s(t)
Sampled signal
Maximum frequency of the band-limited signal
Sensed analog signal
Sampled signal of the
Piece-wise linear function of the sampled signal
EMI source function (noise current) at the quite port, utility side
X(s)
EMI source function (noise current) at noisy port, the converter side
X(s)
K1
Injector gain
xxi
K2
G(s)
(f1)
(f2)
XD
XQ
XR
XI
Xm
PM
Phase margin
Km
Modulator gain
Vs
Kb
Plant DC gain
Gp(s)
Plant TF
Gd(s)
D-Q transformation TF
Gc(s)
Type-III compensator TF
Tcl_i
Number of bits
Output voltage sensor transfer function
Ksense
Sensing gain
Gain of the PWM
Gi(s)
Gv(s)
xxiii
Chapter 1
Introduction
1.1
Background
From the very beginning, there was an obvious trade-off between efficiency or functional
characteristics of the power electronic system and the EMC performances. One of the major
challenges is to predict, with reasonable accuracy, the noise levels for a given device early in the
design phase. In order to accomplish this task, the designer must deal with more or less empirical
models for the component behavior in various frequency ranges and handle different and specific
simulation methods. EMI should be considered early in design cycle of a product, in order to
decrease expensive post-development modifications for EMI compliance. On the other hand, as
shown in Fig. 1.1, the more the EMI problems are left at the end of the design cycle of the
product, the more the cost and time to market increases. Hence, it is necessary to resolve all the
issues pertaining to EMI at the design stage of the product.
Degree of Freedom
Design
Solution Cost
Prototyping
Production
Fig. 1.1 Cost of EMC solution during the Design Lifecycle [2]
1.2
A receptor circuit whose operation is negatively impacted by the transmitted energy (the
victim circuit).
All three elements must be present for EMI to take place remove any one and there can be no
interference. This is illustrated in Fig. 1.2 below.
Fundamentally, EMI requires recognition of the fields caused by rapidly changing currents and
voltages. While these characteristics are quantitatively described by Maxwells equations, we
need only to know that electronic noise may be induced by coupling between circuit elements
through the action of either a magnetic or an electric field. A magnetic field will cause a changing
current in a conductor to induce a voltage in another according to:
Where, M is the mutual inductance between the source and the victim circuit. Fig. 1.3 illustrates
the concept of inductive coupling.
Radiated EM
Near-Field
Magnetic Energy
Far-Field
Magnetic Energy
Similarly, an electric field will cause a changing voltage on a surface to induce a current to flow
in another conductor according to:
Where, C is the capacitance coupling the source to the victim. Fig. 1.4 illustrates the concept of
capacitive coupling.
E-field
Near-field
Conducted Energy
Far-field
Radiated Energy
These equations tell us that where we have rapidly changing currents (high di/dt) as in the
conductors in series with power switching devices we can expect to see an induced voltage
across other conductors coupled by a mutual inductance. And where there is a high dv/dt as on
the drain contacts of the power switching FETs any parasitic capacitance can couple an induced
current into another path in the vicinity circuits.
More specifically, the switch mode power converter is based on a non-linear action of the
switching devices, such as MOSFETs which are controlled by the pulse modulated signal with
variable or fixed frequency, to step-down or to step-up the input voltage to a desired level. These
switching waveforms contain significant energy levels at the fundamental switching frequency
and its multiple harmonics and therefore generate EMI problems. The EMI is transmitted in two
forms: radiated and conducted. Usually conducted noise is several orders of magnitude higher
than the radiated noise into free space.
CM
1 to 10uF
Decoupling Cap
0.1uF
50
DM
EUT
(SMPS, DC/DC
Converter)
GND
50
1 to 10uF
Decoupling Cap
0.1uF
Neutral
CM
CM
50
CM and DM add vectorially
EMI (Line) = CM + DM
EMI (Neutral) = CM - DM
Spectrum Analyzer
R_load
The contributions of these two modes are inherent to the basic operation of the switching power
supply. The action of the internal power switches causes fast di/dt rates in the differential current
at both the input and the output of the power supply. As illustrated in Fig. 1.6, the input filter
ideally would decouple any high frequency noise which is external to the power supply.
However, residual ripples and switching spikes exist as a differential mode noise source with
bidirectional current flowing into input port and exiting from the output port. On a Printed Circuit
Board (PCB) the DM noise resides in the trace routing, parts placement, current paths that can be
translated into radiated emissions in free space once the current paths meets the criteria of a loop
antenna.
R_load
I_DM
I_DM
I_DM
DC
I_DM
Main Switch
I_DM
I_DM
Pulse Control
signal
Fig. 1.6 Differential Mode currents is generated by the normal switching action of the MOSFET
There are also sources of rapidly changing voltage within the power supply which can couple
noise through parasitic capacitance to earth ground, some of which are shown in Fig. 1.7. This
type of noise in the ground path, which can be seen as common mode noise on all power supply
terminals, is measured with respect to ground.
DM_filter
CR1
T1
C_out
R_load
D1
C_bulk
DC
Main Switch
Pulse Control
signal
Chassis Ground
I_CM
I_CM
Chassis Ground
So far we have mentioned the sources of EMI in the SMPS that can be generated through
capacitive and inductive coupling to the ground chassis, in the form of differential and common
mode. However, this is not the only locations of the EMI source where high dv/dt signals might
introduce ground noise. Heat sinks are also potential problem area, as safety requirements
typically do not allow heat-sinks to be charged to a high voltage potential when fault condition
occur. Hence, they must be grounded to chassis.
Another potential induced noise source is the transformer primary to secondary coupling. This is
a CM noise that can be returned back to the line impedance stabilization network (LISN) through
the chassis, creating a large current loop. The additional sources of CM noise that can be found in
DC/DC converters are noise coupled through the stray capacitance of the output rectifier to the
heat-sink and the output rectifier to the transformer core.
A noise path through the heat-sink stray capacitance is illustrated in Fig. 1.8.
50 LISN
CM noise path
CR1
T1
C_out
R_load
D1
CD_HS
C_bulk
DC
Main Switch
Pulse Control
signal
Heat Sink
Chassis Ground
Chassis Ground
1.2.2
In an electronic apparatus or circuit such as found in switch mode power supplies, EMI energy
can be mutually interchanged between radiated and conducted form. For example, EMI is initially
generated in a conducted form through an interconnecting wire or PCB trace. This conducted
energy creates an electromagnetic field around the conductor according to Faradays law. If there
is then mutual inductance or capacitive coupling to another conductor, then the radiated energy is
transformed back to conducted noise in the adjacent conductor. However, the RF energy which
propagates along the trace at specific length could become a radiating loop antenna. Thus, the
circuit becomes susceptible to external EMI, according to the antenna reciprocity theorem.
Typical radiated emission scenario is illustrated in
Fig. 1.9.
As for the radiated emission in the power converters, it depends on the switching frequency
employed in the circuit. Power converters, operating below 500 KHz switching frequency, tend to
have their noise bandwidth up to 16 MHz which is not a concern in terms of EMC standards
compliance (30 MHz to 1GHz). However, it is a problem for the internal circuit operation of the
power converter [8]-[10].
T1
Clamp/Reset
CR1
C_out
R_load
D1
C_bulk
Main Switch
DC
Naux
Rg
in
Fr
gi
ng
fie
To PWM VCC
Fr
in
gi
ng
fie
ld
ld
10
systems. All these are manifestations of EMI and demonstrate the need to design systems which
are compatible with their electromagnetic environment. There are two aspects to EMC. First,
systems must be designed so that they do not emit significant amounts of unintended
electromagnetic (EM) radiation into their environment. This aspect is described as emission and
may be divided in turn into conducted and radiated emission. Second, systems must be capable of
operating without malfunction in their intended environment. This aspect is described as
immunity, or alternatively, as susceptibility. Hence, all EMC analyses and design techniques aim
to address these two aspects using circuit-based and field-based experimental, analytical, and
numerical techniques.
It is important to realize why EMC has become so important in recent years. As is usual in such
cases, there are several reasons: Modern design relies increasingly on the processing of digital
signals, i.e., signals of a trapezoidal shape with very short rise and fall times. This gives them a
very broad frequency spectrum and thus they are more likely to interfere with other systems.
Most modern designs rely on clocked circuits with clock frequencies exceeding 2 GHz. This
implies very short transition times and also the presence of several harmonics well into the
microwave region. Such a broad spectrum makes it inevitable that some system resonances will
be excited forming efficient antennas for radiating EM energy into the environment and coupling
to other systems. Voltage levels for switching operations have steadily decreased over the years
from hundreds of volts (vacuum tubes) to a few volts in modern solid-state devices. This makes
systems more susceptible to even small levels of interference. We make a much greater use of the
EM spectrum as, for instance, with mobile phones and other communication services. Equipment
is increasingly constructed using small cabinets made out of various plastics and composites in
contrast to traditional design, which used metal (a good conductor) as the primary constructional
material. This trend meets the need for lighter, cheaper, and more aesthetically pleasing products.
11
However, poor conductors are not good shields for EM signals, thus exacerbating emission and
susceptibility problems. Miniaturization has become a trade option, as smaller and lighter mobile
systems are required. This means close proximity between circuits and thus greater risk of intrasystem interference (cross talk). We rely increasingly on electronics to implement safety critical
functions. Examples are airbags systems for cars, automatic flight controls in aircraft, etc... It is,
therefore, imperative that such circuits be substantially immune to EMI. In addition, military
electronic systems are continuously exposed to very hostile EM environments.
These points illustrate the engineering need to design electromagnetically compatible systems. To
ensure that this compatibility exists, a relatively new engineering discipline, namely
electromagnetic compatibility (EMC), has evolved. EMC is defined as the study and analysis to
resolve electromagnetic interaction problems in the field of electrical engineering. EMC is
branched into two distinct categories: EM emissions and EM susceptibility with alternating
medium of propagation. The two categories are fundamentally reciprocal.
Fig. 1.10 illustrates the areas of EMC fields.
12
It is important to mention that EMC is a system level consideration. However, all electronic subcomponents must undergo compliance testing. This does not imply that compliant subcomponents mean that the system is EMC compliance. So it is common to test for
electromagnetic noise generation from a power supply as a stand-alone box, and also the ultimate
standards that have to be met apply to the system as a whole with the power supply as an internal
component or subsystem.
International standardization bodies have recognized for many years the need to define standards
and procedures for the certification of systems meeting EMC requirements. These are the
responsibility of various national standard bodies and are overseen by the International Electrotechnical Commission (IEC).
Furthermore, various regulations have been imposed by government authorities to limit the level
of the EMI emissions. These regulations have been developed as EMC standards and are a
condition to product marketing and circulation. A product is considered to be compliant when the
limits applying to that product are not exceeded. There are many ways of achieving EMC
compliance, using different suppression techniques. These techniques have their own advantages
and limitations. These advantages are measured based on their EMI attenuation and their
drawbacks are evaluated based on their impact on the power converter overall size , efficiency,
and cost. Conducted Interference filtering technique, either passive or active is the most used
technique due to its attenuation performance, regardless of the size and the cost factors. However,
stringent power converters specification and cost requirements, make these EMI solutions less
desirable or the performance is compromised with the cost and size. A combined active and
passive filtering technique (hybrid) seems to be the solution for size and performance to a certain
extent. The main purpose of the hybrid EMI filter is to reduce the size of the passive filter, by
splitting the EMC frequency spectrum for the conducted emissions in two allocations. The low
13
frequency allocation is targeted by the active filter whereas the high frequency allocation is
assigned for the passive filter. The combination of these filters covers the whole EMC spectrum,
from 150 KHz to 30 MHz, achieving a significant overall attenuation in the conducted emissions,
while reducing the passive filter size and increasing the performance.
2. Development of a digital active EMI filter using concurrent FPGA based implementation.
3. Integration of the proposed EMI solution in a digital controller of a grid-tied PV inverter as
an industrial application.
i. Analysis and modeling of the micro-inverter
ii. Verification of the micro-inverter control loop stability
4. Development of a DC-DC converter prototype for electric vehicle battery charger that
consists of the proposed EMI solution integrated into the digital controller of the power
converter.
i. Design of a digital control compensator including the proposed EMI solution
ii. Verification of the control loop stability of the converter
EMI suppression technology, based on Digital Signal Processing (DSP) approach is introduced.
This method extracts the interference noise from the input leads of the power converter using a
sensing circuit, the noise is then fed to an Analog-to-Digital Converter (ADC) to digitize the
noise signal. The digital form of the noise signal is then fed to the DSP core for further
processing. The output of the Digital-to-Analog Converter (DAC) is an image representation of
the original noise signal which is used to suppress the incident noise in the power converter. This
method of suppression does not employ any passive components as compared to the passive
filtering method which requires both bulky capacitors and choke inductors. In chapter 4, a new
FPGA-based approach to suppress conducted EMI emission has been presented. This method
uses a Very high speed integrated circuit Hardware Description Language (VHDL) algorithm to
process the noise signal with a negligible delay in the phase. Both approaches have been
compared in terms of their EMI attenuation performance. The analysis and modeling of the
proposed methods are presented. The simulation and experimental verification to validate the
proof-of-concept have been laid down. In addition, two industrial applications have been selected
as a test bed. These are namely, the solar micro-inverter application in Chapter 5 and Electrical
vehicle (EV) DC/DC power converter in Chapter 6. Finally, Chapter 7 summarizes the
contribution of the achieved milestones and provides guidance for future research directions.
16
Chapter 2
Review of EMI Control techniques in Power converters
2.1 Introduction
For cost-effective design approach, EMC should be considered at early stage of the power
converter design. Hence, designing to achieve EMC involves a series of measures to reduce
emissions at the source. This can be done by identifying and minimizing the coupling paths and
diverting CM noise away from ground. As product development progresses from the design stage
to testing the prototype and mass production, the range of available noise suppression techniques
decreases steadily. As the first step, the power converter should be analyzed for its EMI
generation to determine the required measures of suppressions that can be implemented at the
design stage without the need for passive EMI filters. The later, are not cheap and have a direct
impact on the stability characteristics of the power converter, adding the cost and the PCB real
estate.
ground. These components should be placed as close to each other as possible. The PCB should
be laid out in such a way to minimize loop area of return current paths. A DM coupling via a loop
antenna leads to unwanted noise radiation and will jeopardize the circuit immunity to external
noise. Hence, grid grounds and ground planes should be always added as additional layers to the
PCB design. With the ground plane the return current flows directly underneath the PCB tracks at
higher frequencies. Power converters are not sealed box products. They need to be supplied by a
source power so that they can convert power to the specific load. This is done through a power
cable. This cable at specific length becomes an antenna radiating CM noise. This CM noise can
be prevented by cable filtering and by additional ground planes during PCB design. Shielding is
another technique to suppress radiating magnetic fields. This technique can be applied to
magnetic components such as transformers and power inductors to contain the fringing fields.
However, this method is very expensive when used for this purpose and magnetic fields are very
difficult to shield. Thus it is better to combat the problem at its source by minimizing current
loops and containing magnetic fields. This leads to the discussion of the two major circuit
elements that are responsible for most CM EMI generation namely the heat-sink and the
transformer.
18
). This CM noise
current can be converted into CM voltage noise in the artificial line impedance (LISN) and part of
this current can produce a radiating E-field due to the large return path that acts as a loop antenna.
Cr1
LISN
Cout
Rload
50
GND
MOSFET
N
50
Icm
Ground
Fig. 2.1 MOSFET Drain to Heat-sink stray capacitance, the CM current returns through the LISN
impedance (large return path)
One of the most efficient solutions for suppressing the CM noise voltage is achieved by reducing
the value of the stray capacitance built between the MOSFET and the heat-sink. Three methods
are known; the first one is to float the heat-sink from ground and alternatively to tied it to source
pin of the MOSFET as shown in
Fig. 2.2. This ensures that the CM noise current flowing through the stray capacitance C d_HS, is
kept confined within the primary circuit rather than returning through the LISN. However, this
type of referencing is not acceptable by safety standards, should the heat-sink become live after a
short circuit condition.
19
Cr1
LISN
Icm
Cout
Rload
50
GND
CD_HS
MOSFET
50
Fig. 2.2 except that in this method the heat-sink is grounded as per safety standards requirements.
The CM current return path is confined in the primary circuit, because the impedance between the
shield and the source of the MOSFET is very low compared to the impedance between the heatsink and the ground.
Cr1
LISN
Icm
Cout
50
GND
Cstr1
Cstr2
MOSFET
N
50
Zs <<
GND
20
Rload
The third method of reducing the CM current path and bypassing the LISN is using LC circuit.
This method uses an inductor between the heat-sink and ground and connects a capacitor between
the heat-sink and the source pin of the power MOSFET. This circuit operates as follows: at high
frequency the inductor acts as an open circuit and the capacitor as a short circuit. At DC level, the
inductor acts as a short circuit and the capacitor as an open circuit. In this way the high frequency
CM noise current takes the path of least impedance while the heat-sink is kept referenced to
ground.
Fig. 2.4 illustrates the concept.
Cr1
LISN
Icm
Cout
Rload
50
GND
Cstr1
MOSFET
50
ZL=2fL
Zc=1/2fC
GND
21
Cr1
LISN
Cout
Rload
50
GND
Ccore
MOSFET
N
Cp-s
50
ICM
ICM
GND
Fig. 2.5 CM current generation through core and primary-to-secondary stray capacitance
To reduce the inter-winding and the transformer core stray capacitances, two methods can be
applied; first referencing the core to the DC-link of the power circuit, will provide a shorter path
to the CM current and keep it confined in the primary circuit. The layout of this configuration is
shown in Fig. 2.6.
Cr1
LISN
Cout
Rload
ICM
50
GND
Ccore
MOSFET
50
GND
Fig. 2.6 Transformer core is referenced to the DC-link, shorter current path
As for reducing the unwanted inter-winding stray capacitance, an electrostatic shield can be
placed between the primary and the secondary windings and referenced to the DC-link, as
illustrated in Fig. 2.7. This approach will divert the CM noise away from the ground and provide a
shorter return path. This is adequate solution for low output voltage converters.
22
Electrostatic shield
Cr1
LISN
Cout
Rload
50
GND
MOSFET
N
50
Stray capacitance
GND
Fig. 2.7 shield connection between the primary and the transformer core
For higher output voltages, the stray capacitance between the shield and the secondary winding
becomes significant and may couple CM noise currents into the secondary side of the power
converter. To circumvent this current path, a secondary shield is required. This configuration is
shown in Fig. 2.8.
Electrostatic shield
Cr1
LISN
Cout
Rload
50
GND
MOSFET
N
50
Stray capacitance
GND
the leakage inductance generates a transverse magnetic field between windings. Depending on the
coupling factor of the magnetic transformer, some of this field may be coupled to the transformer
core, and the rest acts as a magnetic dipole antenna fringing out into surrounding space with an
intensity which decays as the cube of the distance as derived from Maxells field equations. To
suppress the fringing field, a change in the winding procedure to interleave the primary as shown
in Fig. 2.9, will produce two leakage fields with opposite polarities which ultimately cancel each
other. Due to the skin effect at high frequency, it is preferable to use different types of magnet
wire such as Litz wire, or bifilar windings. This will result in significant attenuation of radiated
emissions.
Another method of reducing the fringing magnetic fields from a transformer is the use of a
conductive copper foil strapped around the transformer. The copper foil provides a path for the
eddy currents that result from the leakage inductance magnetic dipole. The current flowing in the
flux strap creates an opposing magnetic dipole which tends to cancel the original field in the
vicinity of the power transformer.
Inductors are also potential generators of stray magnetic fields caused by the side gaps of the core
which are outside the coil winding. By changing the core geometry so that the gap is in the center
leg, the flux will be fully contained within the winding, hence, reducing the source of EMI
radiation.
24
Fringing Fields
Primary
Secondary
Core
Primary
Coupled Fields
Primary
Secondary
Primary
Secondar
y
Primary
Primary
Secondar
y
Core
Coupled Fields
(a)
(b)
Fig. 2.9 Transformer windings configuration; (a) single windings, fringing fields; (b) split windings
creates leakage fields that tend to cancel
In this section, the conventional passive EMI filters are introduced. The circuit configuration and
design considerations in terms of mismatch impedances are discussed.
EMI FIlter
Power Converter
L
Zs
T1
CR1
R_load
C_out
C_bulk
Cx
D1
AC
Main Switch
Zg
Pulse Control
signal
L
Cy
Cy
Chassis Ground
26
X L 2fL
XC
(2.1)
1
2fC
(2.2)
It is important to note that the basic approach to EMI filtering is to use a set of series inductors
and parallel capacitors to divert the flow of EMI currents away from the auxiliary circuit. This is
done by using the high inductor impedance to prevent the flow of the EMI noise from flowing in
the circuit and the low capacitor impedance to divert the EMI noise towards the ground. The EMI
current flows through the path of the least impedance.
Hence, filtering common-mode EMI requires capacitors connected to ground. These capacitors
are referred as Y capacitors and safety regulations limit these capacitors to relatively low values.
Consequently, high values of inductance are essential for effective filtering. However,
differential-mode filtering requires capacitors across the input lines. These capacitors are referred
as X capacitors.
L1
C1
L1
C2
L2
L2
L3
C2
C1
C3
Fig. 2.11 Basic passive EMI filter circuit configurations: (a) single stage LC-circuit, (b) circuit, (c) T-circuit, (d) Multistage LC-circuit
Phase
Y
LOAD
Neutral
Earth GND
28
The passive EMI filter can be evaluated by its insertion loss which can be defined as the ratio of
the generator voltage and the measured voltage when the filter is inserted between the generator
and the measuring instrument as shown in Fig. 2.13.
The insertion loss (IL) can be generally expressed in terms of voltages as:
U
R2
IL 20 log 0
U m R1 R2
(2.3)
Where: U0 is the generator voltage and Um is the measuring instrument voltage when the filter
under test is inserted in the circuit. R1 is the generator source impedance, and R2 is the measuring
instrument input impedance.
In practice, R1 and R2 should be equal; in this case equation (2.3) can be simplified as follows:
U0
IL 20 log
2 U m
(2.4)
The attenuation is a transfer function that reflects the performance of the filter at each frequency
with the real circuit impedance, rather than assumed impedance as in the case of the insertion loss
measurements. To respond to the challenges of low profile DC/DC converters and their fast
growing switching frequency (>1MHz), it is important to consider the following parameters: the
size (small footprint), performance (at least 30 dB attenuation across the frequency spectrum) and
cost. The first two parameters are circuit related and provide a lot of room for research on the
combination of the passive and the active EMI filters as an integral part in an open frame
configuration of DC/DC type converter.
29
Measuring Instrument
(R1=R2=R)
Signal Generator
R1
R2
U0
U0
2
(a)
Measuring Instrument
Signal Generator
R1
R2
Um
(b)
Fig. 2.13 Insertion loss (IL) measurement: (a) Reference circuit, (b) Filter under test
inserted
2.3.3 Source and Load Impedance Variations
One of the main problems in designing passive EMI filters for power converters is caused by the
arbitrary selection of the source and the load impedance values. This means that there is no
guarantee that the arbitrary selected values that are supposed to reflect the source and the load
impedances are valid values, given the fact that a typical EMI filter is to be installed in power
converters which can be used in variety of applications and supply networks. The high frequency
impedance of the DC/DC converter which represents the noise source varies widely, depending
on the converter circuit topology, circuit layout, types of MOSFET and the switching frequency.
Furthermore, the load impedance which is the supply impedance of the passive EMI filter is even
less predictable than the source impedance, because the load impedance depends on the facility
network from which the power converter would be supplied and varies with the number of
30
electrical equipment connected to the supply network [13]-[19]. Nonetheless, modeling the power
converter in terms of its EMI noise, as shown in [20], is an essential step in designing the proper
EMI filter.
Rf
Z2
Lf
Z1
Cf
Input
DC
Voltage
Z3
Rd
ESRcf
Cd
ESRcd
Rin
(DC/DC
converter
input side)
Z1 R f sL f
(2.5)
Z 2 ESRC f
1
sC f
(2.6)
Where: ESRc is the equivalent internal series resistance of the filter capacitor
Z3
1
ESRCd Rd
sCd
(2.7)
Then the equivalent impedance of the EMI filter at the input of DC/DC converter is given by:
Z eq Z1
Z 3 Z 2 Rin
Rin Z 3 Rin Z 2 Z 2 Z 3
(2.8)
The values of damping resistor and energy-storage capacitor can be selected according to:
Rd
Lf
(2.9)
Cf
Cd 2 C f
(2.10)
The input and the output impedances of the passive EMI filter are important parameters that must
be known when connecting the filter to the input side of the DC/DC converter. This condition sets
a boundary requirement between the EMI filter and the DC/DC converter in terms of input/output
impedances [25]. The input EMI filter can adversely affect the performance of the DC/DC
converter which can be driven into continuous oscillations and instability. Thus, the input passive
EMI filter should be designed with the issue of the impedance compatibility in mind. Fig. 2.15
32
shows the interface impedance of the input passive EMI filter with the input DC voltage bus and
the DC/DC converter.
Input Interface
Input DC Bus
Output Interface
ZDC
ZIF
DC/DC Converter
ZOF
Zin
Z DC Z ' IF
(2.11)
Where, ZDC is the output impedance of the input DC source, and ZIF is the input impedance of the
filter with the output port open
2. At the interface between the input filter and the DC/DC converter:
Z 'OF Z in
(2.12)
Where, ZOF is the output impedance of the filter with the input port shorted, and Zin is the input
impedance of the DC/DC converter.
33
Since impedances ZDC and Zin are system dependant and can not be known, the two impedances
ZIF and ZOF can be considered as design parameters and can be calculated or measured.
The input impedance can be selected as high as possible to minimize interactions with the input
DC bus, and the output impedance can be selected as low as possible to minimize interactions
with the loop gain of DC/DC converter.
34
35
CS1
D1
S1
vS1 -
Ls
Cs
iS2
gS1
Vin
SR1
ir
iS1
S2
gS2
Tx
gSR1
iSR1
Co
CS2 +
vS2
-
+
R L Vo
-
D2
SR2
gSR2
iSR2
Fig. 2.16 zero voltage transition (ZVS) circuit implemented in a DC/DC converter
36
60
40
Vds1
20
Vgs1
0
-20
475 . 6us
475 .8 us
476 .0 us
476 .2 us
476 .4 us
476 .6 us
476 .8 us
477 .0us
Tim e
54.7
40.0
Vds1
20.0
Vgs1
0
- 18.5
475.900us
476.000us
476.100us
476.200us
476.300us
476.400us
476.500us
476.600us
476.700us
476.800us
Time
100mV
10mV
100uV
1.0uV
150KHz
300KHz
1.0MHz
3.0MHz
10MHz
Frequency
30MHz
100mV
10mV
100uV
150KHz
300KHz
1.0MHz
3.0MHz
10MHz
30MHz
Frequency
The simulation results reveal the difference between the HS and ZVS converters with respect to
their EMI conducted profiles. It is found that the advantage of using the ZVS circuit as an EMI
solution is not significant. It slightly contributes to the EMI attenuation especially at higher
frequency where the CM noise dominates. However, at lower frequency, it exhibits an increase in
the fundamental peak and the first few harmonics, up to 3MHz.
To support what has been claimed in the section above, a comparison between two boost
converters, one employing ZVS technique whereas the other one uses hard switching method, is
provided. It is concluded that the figure of merit in terms of EMI conducted emissions is
insignificant and can be quantified to few dB only and can be even negligible at low frequencies.
This is illustrated in the experimental waveform shown in Fig. 2.21 below. Hence, the ZVS
method cannot be regarded as an EMI solution but rather as an option to slightly improve the
conducted emissions, if and only if the resonant converter topology is sought to be suitable for a
specific application. In fact, under EMC standards, 2 or 3dB is considered an uncertainty margin
for measurements; hence it is not worth it to implement a ZVS circuit just to accomplish the EMI
attenuation of a couple of dB, especially at lower frequencies.
38
spectrum clock generation (SSCG) method [40]-[43] to reduce the radiated emission. This
method is based on frequency modulating (FM) the system clock signal which results in power
spectrum density (PSD) with uniform sideband harmonic amplitudes. The energy of each discrete
frequency harmonic is spread over a wider bandwidth, thus reducing the amplitude of the
harmonic contents of the pulse signal.
Fig. 2.22 illustrates the effect of the SSCG technique on a single harmonic of a clock signal.
Amplitude
Frequency
Frequency
(a)
(b)
Fig. 2.22 Effect of SSFM technique on an nth harmonic of a clock signal; (a) un-modulated
clock; (b) modulated clock
As an example of signal modulation, consider a modulated signal
Where:
angle which is varied as a result of the modulation. In terms of frequency, this can be expressed
as:
(2.14)
Where:
(2.15)
The key to maximizing the attenuation of a clock signal fundamental and its harmonics is the type
of the modulating waveform, either being a sine wave, triangular or saw tooth as illustrated in
Fig. 2.23. Optimum attenuation can be achieved using triangular modulating waveform
Fig.2.23(c), compared to sinusoidal waveform Fig.2.23(a).
(a)
(b)
(c)
Fig. 2.23 Resulting spectrum of different modulating signal waveforms; (a) sine wave; (b)
square wave; (c) ramp
Two useful parameters to describe the characteristics of the modulated signal are the modulation
index K and the rate of modulation, % given by:
41
(2.16)
(2.17)
The parameter , gives an idea on how wide the energy of a single harmonic will be spread
relative to
the modulated signal is spread within the bandwidth. This can be expressed as:
(2.18)
Since the modulated clock signal is periodic in nature with a period of
and a phase of
(2.19)
is either 5V or 0V
(2.22)
The frequency spectrum of the modulated clock signal is represented by delta functions at integer
multiples of
number of sidebands increases, resulting in signal energy that is more evenly distributed in the
bandwidth. An even distribution of signal energy provides a greater overall attenuation of the
fundamental harmonic amplitude of the modulated signal. This can be seen in Fig. 2.24.
42
It is demonstrated in [52], [53] that using SSCG to modulate the system clock with triangular
waveform at 1% frequency deviation, have resulted maximum of 16.7dB in the radiated emission
attenuation. Also it is found that the level of reduction in the radiated emissions depends on the
amount of the frequency deviation or the modulation index.
Similarly, it is shown in [54],[55] that the application of the SSFM techniques in a resonant
inverter based compact fluorescent lamp (CFL) can help to reduce the output lamp current power
spectral density (PSD) to an EMC compliant level. An optimum attenuation of 12db was
achieved using multi-slope ramp or triangular modulation while controlling the effect of
amplitude modulation (AM) affect to keep the output current ripples within the desired level.
This technique has been proven to have low attenuation at lower frequency harmonics
with a maximum attenuation around 10dB at optimum modulation index and triangular
waveform. However good performance is observed at higher harmonics.
The technique is useful in power converters only when the switching frequency falls
within the EMC spectrum of the conducted emissions.
Increasing the modulation index beyond a certain value may impact the output current
ripple.
The modulating frequency may interact with the feedback loop bandwidth of the power
converter and may results in continuous oscillations.
Higher frequency performance of this technique makes it more suitable in digital circuit
design application, by modulating the system clock to mitigate the radiated emissions.
44
to avoid saturation, in particular when it is placed at the output of the power converter. The ripple
voltage which is in the phase of the input ripple current is fed into an inverting feedback amplifier
to inverse the phase of the sensed noise signal. The output signal of the amplifier is injected back
into the converter through a transformer with the desired gain. This injected voltage is converted
into current through the shunt inductor which is part of the passive filter elements.
(2.27)
Where:
(2.28)
Assume that the current that is fed into the utility mains in Fig. 2.26(a) and Fig. 2.26(b) is Iin and
I'in respectively, the performance of the active filter can be evaluated by its insertion loss defined
as:
IL
I in
I in
(2.29)
In the frequency domain, the current Iin and I'in can be derived according to Fig. 2.26:
Iinj
Fig. 2.26: Equivalent harmonic circuit of the converter: (a) Equivalent circuit with passive filter only,
(b) Equivalent circuit with hybrid active and passive filters
47
1
Z in sLin R ESR _ Lin
(2.30)
1
R ESR _ Cin
sC in
The noise current in circuit Fig. 2.26 (b) can be expressed as:
I in I n
Z1
B Z 2 E Z3 Z 4
(2.31)
Where, Lin and Cin are inductor and capacitor components of the input passive filter. Resr_Lin and
Resr_Cin are the equivalent series resistances for the inductor and the capacitor components
respectively. Rs and Cs are the resistive and capacitive components of the sense circuit. N is the
gain of the injection transformer.
B 1
E 1
Z1
Z in
(2.32)
Rs
1
sCs 1 Av N Rs
Z in
1 Av N Rs
Rs
sC s
(2.33)
1
R ESR _ Cin
sCin
(2.34)
1
sC s
(2.35)
Z3
1
RESR _ Cin Rs
sCin
(2.36)
Z4
1
Rs
sC s
(2.37)
Av is the closed-loop voltage gain for the current feedback amplifier with unity gain that includes
the frequency-dependent open-loop trans-conductance gain function A(s):
48
Av
A( s)
1 A( s)
A( s)
(2.38)
k1
1 s k2
(2.39)
Where k1 is the DC open-loop voltage gain (typically 105 to 107), and k2 reflects the cutoff
frequency. These two parameters describe the frequency-dependent open loop characteristics of
the chosen OPAmp.
Therefore, the noise attenuation of the hybrid circuit can be found by substituting Iin and I'in in
equation (2.29) by equations (2.30) and (2.31) respectively, and Fig. 2.27 shows the noise
attenuation in frequency domain.
Fig. 2.27: Noise attenuation of the active circuit (Zin=50//50H, Rs=50, Cs=5F, Cin=20F,
k1=6106, k2 =100, N=15)
It can be seen that in the frequency range of interest (150 kHz-3 MHz), the active filter has
substantial noise attenuation, also referred as insertion loss of 30 dB at the fundamental switching
frequency of the DC converter, and around 20 dB at the first several harmonics. It is also found
that the noise attenuation is not only dependent on the characteristics of the OPAmp (k1 and k2),
49
but also highly dependent on the value of the shunt inductor Lin. In Fig. 2.27, the smaller value of
Lin results in better noise attenuation. However the minimum value of Lin should be limited by the
cut-off frequency of the input passive filter, which should be set at a decade lower than the
resonant frequency of the output filter of the DC/DC converter to avoid interaction.
50
_1
and vit
_2
are the voltages at the primary and secondary windings of the injection
transformer; vop is the output voltage of the OPAmp; Rs is the resistance of the sensing element;
s is the switching frequency; in is the input ripple current; iinj is the injected current noise.
Ideally, the injected current noise should cancel the input ripple current:
(2.44)
From (2.43), the ratio of the injection current and the ripple current should be close to 1;
(2.45)
Or
(2.46)
51
From (2.46) above, the turns ratio of the injection transformer must match the impedance of the
bypass inductor in order to achieve maximum noise attenuation.
52
Fig. 2.30: Conducted EMI noise spectrum result with the input passive filter only
Another set of measurements were taken with the active EMI filter inserted between the LISN
and the passive filter. The corresponding result of the EMI noise spectrum, as per EN55022
standards, is shown in Fig. 2.31. As can be seen, significant noise attenuation is achieved at the
fundamental switching frequency and at the first harmonic. This is due to the contribution of the
active filter. Table 2-1, compares the peak attenuation accomplished by the passive and the hybrid
filters.
53
Fig. 2.31: Conducted EMI noise spectrum result with the combination of passive and active
input EMI filters.
An attenuation exceeding 30dBV is observed at the switching frequency and at the first higher
harmonic. These attenuation magnitudes make a significant difference in the EMC compliance in
terms of conducted emissions.
Table 2-1: Performance comparisons of the passive and the hybrid filters
Peak attenuation
Peak attenuation
Delta
attenuation
(dBV)
(dBV)
(dBV)
0.7
88.9
57.9
31
1.4
79.9
43.4
36.5
2.1
57.2
27.4
29.8
2.8
70.6
35.3
35.3
3.6
67.7
30.3
37.4
Harmonic Frequency
(MHz)
54
2.7 Summary
In this chapter, a review of the existing EMI suppression techniques in power converters, have
been presented. There are several techniques, namely the basic technique, the zero-voltage
crossing technique, the SSFM technique, the passive filtering technique and the active analog
filtering technique. Each of these techniques are evaluated according to their attenuation
performance. It has been shown that the passive and the active analog filtering techniques are the
most desirable EMI solutions. Furthermore, the limitations of each technique have been
discussed.
Finally, the simulation results have been shown and have demonstrated that the passive filtering
technique outperformed the other techniques, in particular at high frequency, in terms of noise
attenuation, design simplicity and cost. However, the size is a major drawback in the passive
filtering technique.
55
Chapter 3
Proposed DSP-Based EMI Suppression Technique
3.1 Introduction
The active analog EMI filters provide the basic noise suppression technique and their main
advantages are low cost and ease of use. However, their limitations call for a requirement of
additional passive elements to complete the EMC spectrum, in terms of noise attenuation. Also,
the issue of the negative impedance seen by the converter can have a great impact on its stability.
This is mainly due to the components selection of the passive elements of the EMI filter and the
final installation of the converter. In addition, the size of the passive filter is product specific and
varies with the input parameters of the converter, such as rated current and voltage.
The performance versus cost reduction trends of digital circuits has made possible their
application for power converters digital controller techniques [76]-[83]. They are usually based
on digital signal processor (DSP) that exploits their mathematical oriented resources. The main
limitation of using DSPs in high switching power converters is their sequential operation, that is,
instructions are executed one after the other resulting in a delayed signal. However, this issue can
be neglected if the number of instruction that needs to be executed is very small. Other limitation
in the digital control, is the issue of the limit-cycle oscillation. This is due to the small change in
the output voltage that can cause a further change in the duty cycle.
In this chapter DSP-based EMI suppression technique (DSPBEST) is proposed. It can also be
referred to as digital active EMI filter (DAEF). The proposed DSP-based filtering method
overcomes the drawbacks of the analog suppression techniques. Meanwhile, the DSP execution
time delay does not impact the DAEF performance, since it requires only a few instructions to
56
inverse the phase of the sensed signal. The following are two examples of the comparison
between active/passive analog EMI filter and the proposed DAEF.
Fig. 3.1 shows a simplified diagram comparing the DAEF and the passive EMI filter (PEF) in
terms of their size variation versus converter power rating. The DAEF presents stronger
competitive application in medium to high power converters. In low power applications, the
DAEF presents a less desirable option. This is due to the constant size of the DAEF as compared
to the PEF. However, this statement holds only when the DAEF is configured as a stand-alone
solution. In an integrated version, the DAEF is still the best solution in terms of size and
performance.
Filter size
PEF
DAEF
Co
nv
e
po rgenc
int
e
Low Power
50w
High power
This chapter is partitioned as follows. In section 2, the principle of operation of the proposed
DSPBEST which is also referred to as DAEF, is described. The sampling theory on which the
proposed technique is based on is briefly revisited and the building blocks of the proposed
technique are illustrated. Section 3 presents the circuit analysis and the derivation of the transfer
function of the proposed technique and the key design waveforms are given. Sections 4 and 5
validate the proposed technique through simulation and experimental results respectively. Finally,
section 6 summarizes the key elements of the proposed technique.
58
Z>>>
AC
Vin
Cinj
PFC Converter
Rinj
Rs
Cbulk
Vout
LOAD
Cs
LPF
HPF
Iin
ADC
DAC
PFC Controller
Control Algorithm
59
m(t)
S(t)
i(t)
(a)
Ts
(b)
is given by:
(3.1)
The frequency domain transformation of i(t) is also an impulse train given by:
(3.2)
Where:
60
(3.4)
Multiplication in the time domain corresponds to convolution in the frequency domain. Hence,
the frequency domain transformation of the sampled signal is given by:
(3.5)
To recover the continuous band-limited signal the sampled signal s(t) is passed through an ideal
low-pass filter having a frequency response:
(3.6)
Where,
of the band-limited signal must be less than or equal half the sampling frequency
i.e.:
(3.7)
Inequality (3.7) is known as the Nyquist condition for perfect reconstruction of a band-limited
signal. The minimum sampling frequency that satisfies inequality (3.7) is known as the Nyquist
rate. If the Nyquist condition of (3.7) is not satisfied, the spectra of the images overlap and will
cant be recovered from the sampled signal.
Fig. 3.4 shows a simulation circuit for sampling a time domain signal, with the delay introduced
by the low-pass filter.
61
mt
sampledsignal
IN2
V
V1
100
IN
OUT
V
IN1
OUT
recov eredsignal
0.1dB
60dB
Impulse
V
VOFF = 0
VAMPL = 1
FREQ = 1000
V2
V1 = 0
V2 = 1
2.0V
0V
-2.0V
V(mt)
2.0V
Impulse function
0V
V(Impulse)
1.0V
0V
-1.0V
V(sampledsignal)
1.0V
0V
-1.0V
0s
Recovered Signal
Filter delay
0.5ms 1.0ms
V(recoveredsignal)
1.5ms
2.0ms
2.5ms
3.0ms
3.5ms
4.0ms
4.5ms
5.0ms
5.5ms
6.0ms
6.5ms
7.0ms
7.5ms 8.0ms
Time
Fig. 3.5 Simulation waveforms showing the sampling process using impulse function and the
recovered signal
with precise tolerance. However, digital signal processing can tolerate less precise components
making digital signal processors less susceptible to temperature, aging and manufacturing
tolerances. Furthermore, digital systems offer more extensive programmability than analog
systems. However, all naturally occurring signals that are encountered in the real world are
analog signals. This requires the transformation of such signals from the analog domain to the
digital domain to make use of the powerful computational processing power of the digital signal
processors. The digital signal then has to be transformed back to the analog domain. This
transformation is done by using analog-to-digital and digital-to-analog converters. Fig. 3.6 shows
the block diagram of an analog-to-digital converter (ADC).
The sampler provide the discrete signal in the time domain then it is followed by the quantizer,
which is a many-to-one transformer that maps a range of the continuous signal into a discrete
level. The quantizer performs approximation to the analog signal by approximating it to one of a
finite number of discrete levels. After being quantized, the coder maps each quantized level into a
binary code-word.
Analog Input
(EMI source)
N-Bits Digital
Output
In
S/H
010...
X*(t)
x(t)
Sample
and Hold
Quantizer
Uniform
Encoder
In the digital-to-analog converter (DAC), the reverse operations to those of the analog-to-digital
converter occur as shown in Fig. 3.7. The decoder transforms the binary code into a quantized
signal level. Because the quantizer is a many to one transformer, i.e. it maps a range of the
continuous signal into a discrete level, hence, it has no inverse equivalent in the digital-to-analog
63
converter. Thus, any quantization noise added to the signal is stuck to it and cant be removed by
the digital-to-analog converter. Finally, a low-pass filter converts the time-discrete (sampled)
signal into a continuous analog signal.
Analog Output
(Inverted EMI signal)
N-Bits Digital
Input
010...
Y(t)
X*(t)
Uniform
Decoder
Lowpass RF Filter
in Fig. 3.7 is a
The output function, before the low-pass filter, is a piece-wise linear constant or ramp function in
the nth T period and can be represented by:
(3.8)
So that
64
(3.9)
where:
(3.10)
And,
(3.11)
Where a1,2 and b1,2 are constants determined by the particular sample-and-hold. In the case of the
zero-order-hold (ZOH), the output has a constant value equal to x(nT) throughout the interval .
The Laplace transform, of the function
is as follows:
(3.12)
Replacing
(3.13)
Where,
(3.14)
And,
(3.15)
Using the shifting properties of the transform for (3.13), S1 can be determined as,
65
(3.16)
Similarly, for S2, we have
(3.17)
Substituting (3.16), (3.17), (3.14) and (3.15) into (3.13) the general transfer function can be
written as follows.
(3.18)
, For
(3.19)
Thus, comparing (3.19) with (3.9), yields a1=a2=b1=0 and b2=1. Substituting these values into
(3.18), the general transfer function for the ZOH-based DAC can be simplified as follows:
(3.20)
The relation between the sensed analog signal x(t) to the sampled signal variable x*(t) in the Sdomain is:
(3.21)
The Input-to-Output transfer function for the ZOH DAC may then be written as:
(3.22)
While the transfer function from Input x(t) to Output y(t) is complicated by the repeating
spectrum, the effective frequency response is the continuous Laplace transform transfer function
of the impulse response which can be expressed as,
66
(3.23)
Thus, by substituting
(3.24)
Where
67
+
-
Y(s)
X(s)
Interference Noise
X(s)
Discrete
System
K1
ZOH+DAC
Injector
G(s)
D(s)
K2
Sensor
H(s)
(3.27)
Where,
is the EMI source function at the quite port, utility side;
is the EMI source function at noisy port, the converter side;
is the injected EMI noise function, after processing;
In theory,
nullification of the EMI noise. However, in reality, this cannot be realized due to the parasitic
capacitance inherent in the circuit. Therefore,
K1 is the injector gain;
K2 is the bits inversion algorithm implemented in the DSP device;
is the Laplace transform transfer function of the high-pass filter and it is given by:
(3.28)
Where,
is equal to
68
D(s) is the Laplace transform transfer function of the ZOH as derived in the previous section.
(3.29)
Where, T is the ADC clock/sampling period.
G(s) is the Laplace transform transfer function of the RC low-pass filter given by:
(3.30)
Where,
is equal to
Substituting H(s), D(s), and G(s) in (3.27), the closed-loop transfer function of the feedback
diagram of Fig. 3.8 can be expressed as:
(3.31)
The frequency response of (3.31) in terms of magnitude and phase are illustrated in Fig. 3.9 and
Fig. 3.10 respectively.
For higher noise attenuation, the gain of the feedback transfer function of (3.31) should be as
large as possible. This can be done by increasing the injector gain K 1. To achieve an attenuation
of at least 50dB within the bandwidth of 10 KHz to 30MHz, a gain of 100 is required.
In Fig. 3.10, the effect of the ZOH frequency properties on the overall attenuation transfer
function is apparent. It is a shift of 1800 where the sine function changes sign.
69
Magnitude (dB)
15
30
45
60
10
100
1 10
1 10
1 10
1 10
1 10
1 10
1 10
Frequency (Hz)
200
Phase (Deg.)
100
100
200
10
100
1 10
1 10
1 10
1 10
1 10
1 10
1 10
Frequency (Hz)
Lo
SW1
SW2
Resr
Co
Vdc
Ro
Resr
Rsense
CCVS
PWM Module
Rpar
Rseries
Controller
Cf
LISN
Sensed noise signal
resolution of these devices was selected to be 14 bits and the sampling frequency of the ADC was
set at 200 Mbps which reflects around 10 times the upper frequency of the EMC standards for
conducted emissions (30MHz). In this case Shannons theory for sampling, mentioned in the
previous chapter, is not sufficient, considering the sensed signal amplitude and frequency
variations. Thus, oversampling is required to achieve adequate and complete signal discretization.
Normally, for accurate comparisons between simulated and real plots, the simulation time should
be adjusted in order to match the normalized filter bandwidth at -6dB (200Hz and 9 kHz) as per
CISPR16 standards [86]. However, to simplify the various timing values and limit the number of
simulated data points, 500Hz and 10 kHz will be used as analysis bandwidths. The converter
parameters are given in Table 3.1.
Table 3.1 Simulated Converter Parameters for DSP-based EMI filter
Parameters
Values
Input voltage
12Vdc non-isolated
Output voltage
5Vdc
Output current
4 Amps
Switches S1 & S2
IRF640
Switching Frequency
500KHz
Output filter
Lf=1.5mH; Cf=0.2uF
14bits DAC, 14bits ADC, RC-low-pass
LISN components
72
First, the simulation run without the EMI filter connected to the power converter. Then the
passive analog EMI filter was added to the converter circuit in order to observe the overall
contribution in terms of input noise attenuation. The third case, the digital EMI filter was added to
the circuit but without the Passive EMI filter. In all simulation cases, Fast Fourier Transform
(FFT) was performed on the sensed voltage at the LISN sense port in order to represent the noise
voltage in terms of its harmonic contents in the frequency domain and log compress the Y axis to
represent the plots in dBV form to be compared to the compliance limits.
Waveforms showing the simulated conducted EMI noise spectrum without the EMI filter, with
the passive EMI filter, and with the proposed digital filter are depicted in Fig. 3.12, Fig. 3.13, Fig.
3.14 respectively. As can be seen from these figures, the EMI noise is attenuated by more than
20dBuV, when the digital EMI filter is introduced in the circuit. Note that the digital EMI filter
outperforms the passive EMI filter at the fundamental frequency (500 KHz) and the first few
harmonic peaks. These peaks are labeled on the plot along with their corresponding frequencies.
Table 3.2 presents selective peaks from both PEF and DAEF attenuation performance at different
frequencies.
73
120dBuV
(500KHz,90.3dBu
V)
Conducted Noise
100dBuV
(1MHz,88.3dBuV)
(1.5MHz,84.9dBu
V)
(2MHz,81.28dBu
V)
(2.5MHz,81.4dBuV)
80dBuV
Conducted Emissions CISPR, Class A
60dBuV
40dBuV
20dBuV
0dBuV
150KHz
1.0MH
z
10MHz
30MHz
Frequency
120dBuV
100dBuV
Conducted Noise
(500KHz,82.6dBu
V)
80dBuV
Conducted Emissions CISPR, Class A
60dBuV
(1.0MHz,67.08dBu
V)
(1.5MHz,56.68dBu
V)
(2.0MHz,84.06dBu
V) (2.5MHz,43.76dBu
V)
40dBuV
20dBuV
0dBuV
20dBuV150KHz
1.0MHz
10MHz
Frequency
Fig. 3.13 Conducted Noise measurement with Passive Input EMI Filter
74
30MHz
120dBuV
Conducted Noise
100dBuV
(500K,71.99dBuV)
(1.0M,71.93dBuV)
(1.5M,69.0dBuV)
80dBuV
(2.0M,65.57dBuV)
(2.5M,65.5dBuV)
60dBuV
40dBuV
20dBuV
0dBuV
-20dBuV
1.0MHz
150KHz
10MHz
Frequency
Fig. 3.14 Conducted Noise measurement with Proposed Digital Input EMI Filter
Table 3.2 Comparison of attenuation performance between PEF and DAEF using
simulation results
Peak attenuation
Peak attenuation
Peak attenuation
with PEF
with DAEF
(dBV)
(dBV)
(dBV)
0.5
82.6
71.99
90.3
67.8
71.93
88.93
1.5
56.68
69.0
84.9
50.06
65.57
81.28
2.5
43.76
65.50
81.4
Harmonic Freq.
Peaks (MHz)
75
30MHz
The binary plots of the most significant bit (MSB) and least significant bit (LSB) are shown in
Fig. 3.15 as well as the noise signal at the output of the DAC and the sensed noise signal. As
mentioned before, the injected noise signal theoretically should be the exact replica of the sensed
noise signal with 180o phase shift. However, at higher frequencies, above 30MHz, the injected
signal start to exhibit a slight delay in the phase shift as shown below.
U13:DB0
U13:DB1
U13:DB2
U13:DB3
U13:DB15
U13:DB14
U13:DB13
U33:DB12
2.0V
Sensed Noise
Freq= 33MHz to 500MHz
Injected Noise
0V
-2.0V
573.9900us
574.0000us
V(R87:1)
574.0100us
574.0200us
574.0300us
574.0400us
V(Vsense)
574.0500us
574.0600us
574.0700us
574.0800us
Time
Fig. 3.15 Output of the ADC and the DAC in time domain
76
passive EMI filter installed in the EUT. In the third case, the measurements were done with only
the proposed DSP-based digital active EMI filter. The test setup is shown in Fig. 3.16 below.
The experiment plots pertaining to the proof-of-concept are illustrated in the following figures.
Fig. 3.17 shows the sensed ripple voltage and the injected ripple voltage at the input and the
output of the DAEF. Note that the waveforms are 1800 out-of-phase for EMI noise cancellation
with the frequency identical to the switching frequency of the power converter or the EUT.
DAE Filter
AD/DC Power
Converter
(EUT)
77
Fig. 3.17 Input and output voltage signal of the proposed DAE Filter
The plot in Fig. 3.18 shows the spectrum of the conducted emissions of the EUT without any
filtering in place. High amplitude peaks up to 80dB in magnitude can be seen in the lower
frequency spectrum. The plot in Fig. 3.19 shows significant amplitude attenuation with the
introduction of the passive filter at the input of the EUT. Similarly, Fig. 3.20 shows the
attenuation performance when the PEF is replaced by the DAEF. Attenuation up to 30dB can be
achieved using this filtering method.
Table 3-3 presents selective peaks from both PEF and DAEF attenuation performance at different
frequencies. A comparative graph is depicted in Fig. 3.21 to support the validity of the method
and to confirm the claim of replacing the PEF by the DAEF while improving the attenuation
performance.
78
dBuV
100
90
80
70
EN 55022; Class B Conducted, Quasi-Peak
60
50
40
30
20
10
0
10
4/17/2011 4:04:24 PM
dBuV
100
90
80
70
60
50
40
30
20
10
0
1
10
(Start = 0.15, Stop = 30.00) MHz
79
dBuV
100
90
80
70
60
50
40
30
20
10
0
1
10
4/17/2011 4:46:12 PM
Peak attenuation
Peak attenuation
with PEF
with DAEF
(dBV)
(dBV)
(dBV)
0.5
47.46
36.49
53.12
31.43
34.13
43.78
1.5
25.20
29.69
50.00
40.0
26.34
51.74
2.5
40.0
22.69
51.86
Harmonic Freq.
Peaks (MHz)
80
3.8 Summary
In this chapter, a novel DSP-based technique to suppress conducted EMI emissions in power
converters has been proposed. This technique exploits the theory of sampling using data
acquisition devices such as ADC and DAC for discrete time conversion of the EMI noise source.
The impulse function to represent the sampling process has been used for the sensed signal
recovery. The system building blocks have been explained. The ZOH of the signal acquisition has
been pointed out of being the interface between the continuous and the discrete signal. The
analysis of the mixed system that is partially discrete and partially continuous has been
performed. The system transfer function has been derived using s-domain equivalent parameters.
The frequency response of the attenuation transfer function has led to the determination of the
magnitude and the phase. The figure-of-merit of the attenuation transfer function, in terms of
magnitude, has revealed that the gain of the injector has a direct impact on the increase of the
magnitude within the desired bandwidth. Finally, the simulation and the experimental results
prove the validity of replacing the PEF by the proposed DAEF in the power converter.
81
Chapter 4
Proposed FPGA-Based EMI Suppression Technique (FPGABEST)
4.1 Introduction
The issues with the existing methods previously mentioned in terms of their performance and
their limitations have paved the way for the proposed technique that surpasses these drawbacks.
The DSP based solution for EMI mitigation proposed earlier, has a slight disadvantage which
resides in the delay embedded in the execution of the algorithm instructions that is processed
sequentially. This execution delay can results in phase lag between the sensed signal and the
injected signal which has a direct impact on the attenuation performance. Therefore, in this
chapter, an FPGA based EMI suppression technique is proposed to alleviate the problem of the
phase delay while using the same interface circuit. This method uses a field programmable gate
array (FPGA) device, in order to exploit its concurrent operation [88]-[91]. All the internal logic
elements of the FPGA, and therefore the control algorithm, are executed continuously and
simultaneously. The control algorithm has been developed in Very high speed integrated circuit
Hardware Description Language (VHDL) [92]-[96]. This method is as flexible as any software
solution. The same algorithm can be synthesized into any FPGA device and even has a possible
direct path to a custom chip. In this way, the FPGA could be substituted by an Application
Specific Integrated Circuit (ASIC), opening interesting possibilities in EMI filtering techniques in
terms of performance, cost and flexibility. VHDL has been used for modeling the phase inversion
by using a logic inverter implemented in VHDL.
This chapter is organized as follows. In section 2, the principle of operation of the proposed
FPGABEST is described. In section 3, the circuit analysis and the expression for the z-domain
82
transfer function of the proposed technique is derived. The key design waveforms are presented.
Finally, section 4 summarizes the key elements of the proposed technique and provides a brief
comparison between the DSPBEST and FPGABEST in terms of their performance and
limitations. Note that both techniques can be referred to, for simplicity, as Digital EMI filter.
83
Lf
Z>>>
Iout
SW1
Rload
SW2
ESR
Vout
Vin
Cf
Sensor
Injector
Controller/
Compensator
Analog Interface
ADC
DAC
14
N-bits inverter
n
14
Y(s)
X(s)
Quiet
Port
Discrete System
(ZOH + DAC)
K1
G(s)
Injector
D(z)
K2
T
H(s)
Sensor
Similar to the approach used for the DSP-Based EST, the noise attenuation transfer function can
be derived using the equivalent two port network model as shown in Fig. 4.3
X(s)
Y(s)
Feedback Path
F(s)
Quiet Port
Noisy Port
Where, F(s) is the feedback components blocks of the proposed technique. Y(s) is the noise
current at the Quiet port and X(s) is the noise current at the Noisy port.
The insertion-loss approach could also be used to derive the attenuation transfer function.
However, this method is more complex.
As opposed to the DSP-Based EST, in which the transfer function was derived in continuoustime, the transfer function of the FPGA-Based EST is obtained using discrete-time model.
Furthermore, it is imperative to investigate the characteristics of the discrete-time model, simply
because, the sensor and the injector blocks exist in continuous-time, whereas the sampler and the
85
ZOH are discrete-time. This mixed continuous and discrete time models make the analysis of the
feedback loop more complicated. To overcome this problem, one would convert all the feedback
components to their discrete-time equivalent at specific sampling instances. In this case, the
complete feedback loop can be analyzed using straight-forward application of standard discretetime model. Thus, this leads to the discussion of the approximation methods which are used to
convert the continuous-time model (s-domain) to its equivalent discrete-time (z-domain).
There are three main methods of approximation, the numerical integration which is sub-divided
onto three methods, the forward rule, the backward rule and the trapezoidal rule also known as
the Tustins method. The second approximation is called the pole-zero mapping which maps the
zeros and the poles of the continuous-time into the poles and zeros of the discrete-time
equivalents, using the relation
The third approximation is the zero-order hold (ZOH) equivalent. This method is preferred
because of its simplicity and accuracy as compared to other approximations.
To obtain an expression for the noise attenuation transfer function, each block of the feedback
path of must be converted to its z-domain equivalent. The sensor, the injector and the DAC-ZOH
will be approximated in the next paragraphs and the final expression for the noise attenuation will
be derived thereof.
written as.
86
(4-3)
Hence, the ZOH equivalent approximation can be applied to obtain the discrete-time equivalent.
(4-4)
We have
(4-5)
Substituting equation (5) into (4), the discrete-time transfer function can be obtained as.
(4-6)
4.3.2 The Injector discretization
The s-domain expression of the RC Low-pass filter (LPF) is given by.
(4-7)
Where
written as.
(4-8)
Similarly, the ZOH equivalent approximation can be applied to obtain the discrete-time
equivalent.
(4-9)
After the partial expansion of
written as.
(4-10)
87
(4-12)
The term
reduces to
(4-13)
After substituting (4.13) into (4.12), and re-arranging, the discrete-time transfer function
of the DAC can be expressed as.
(4-14)
4.3.4 Discrete-time Closed Loop Transfer Function
The closed-loop transfer function of the system can be obtained as:
(4-15)
Where F(s) is the feedback path in the continuous time:
(4-16)
Where K1 and K2 are constant gain as defined in Chapter 3.
(4-17)
88
(4-18)
(4-19)
Let 1 = a and 1/2 = b, then
(4-20)
Since
(4-22)
Then
(4-23)
Simplifying and rearranging this equation, F(z) can be finally written as:
89
(4-24)
Finally,
(4-26)
Where,
and
are the corner frequency of the LPF and the HPF respectively equal to 10 KHz and 30
MHz; The choice of the K1 is based on the transmission line loss, in an ideal situation the loss is
negligible and the injection gain is unity. Hence, base on measurements, the value of K 1 is
optimized to be equal to 30.
Substituting the numerical values in (4.26), the final transfer function becomes,
(4-27)
The plot of the frequency response magnitude of the above discrete-time equation provides an
evaluation in terms of EMI attenuation of the FPGA-Based EST. This plot is in-line with the one
obtained in Chapter 3 using the continuous-time model. The plot in Fig. 4.4 exhibits an
attenuation performance of more than 70dB at 100 KHz with the loop gain K 1 maintained at 30;
however, the effectiveness of this method depends primarily on the open loop gain of the digital
active filter, i.e. increasing the value of K1 will increase the performance of the noise attenuation,
as shown in Fig. 4.4.
90
4.4 Summary
In this chapter, an FPGA-based technique to suppress conducted EMI emissions in power
converters has been introduced. The principle of operation of the proposed technique has been
explained. The sampled noise is processed using VHDL algorithm, with a set of instructions that
can be executed simultaneously, hence reducing the delay time. The analysis and design approach
were carried out in z-domain. Using the theory of ZOH approximation the closed loop transfer
function is derived in discrete-time. With the sample rate of 10 times the noise bandwidth, it was
found that the approximation was adequate in meeting the design requirements. Furthermore, the
91
frequency response magnitude of the noise attenuation was plotted and shown that increasing the
gain of the injector k1, will directly impact the performance of the noise attenuation. Finally,
comparing the two techniques presented in chapter 3 and chapter 4, one can say that the
difference resides in their performance and the cost of the implementation. In other words, the
FPGA-Based EST performs slightly better than the DSP-Based technique, since this later rely on
the sequential execution of the algorithm, whereas, the FPGA-Based technique, the algorithm is
executed simultaneously which can be translated into a better attenuation. The other factor is the
cost; in this case the cost for implementing the DSP-Based technique is much cheaper than the
FPGA-Based technique.
92
Chapter 5
Integration of the Proposed DAEF in a Digital Controller of a grid-tied
Photovoltaic Micro-inverter
5.1 Introduction
This chapter presents an industrial application case study, in which the proposed DAEF is
integrated into an FPGA-based digital controller of a grid-tied micro-inverter used in solar power
conversion. As opposed to the stand-alone version of the DAEF, the control algorithm is added to
the inverter main program to form a digital control system. Hence, further reduction of the size,
cost and space of the overall power inverter printed circuit board (PCB).
In section 5.2, the general description of the grid-tied PV inverters and their use in Photovoltaic
(PV) power generation are presented. Section 5.3, as a case study, the modeling of the microinverter controller including the DAEF has been investigated. The expression for the control-tooutput transfer function is required to verify the stability of the micro-inverter, using the gain and
phase margin criteria. Experimental results showing the performance of the integrated DAEF in
the micro-inverter are illustrated in section 5.4. Finally, a summary is given in section 5.5.
the solar energy. There are different configurations for connecting this equipment to the grid.
However, all these approaches are based on an inverter which converts the solar energy into
electrical energy that can be exploited by the consumer. The power inverter is the critical
component in the PV power conversion. A significant progress in the development of the
inverters dedicated to PV systems has been achieved in the last decade. These inverters are not
limited to DC/AC power conversion, but also they do track the maximum power delivered by the
PV panel to maximize the energy throughput. Moreover, they are equipped with monitoring
circuits to ensure a safe operation and to provide protection in the event of power failure, hence,
increasing the reliability of the system. These safety features are enforced by international and
local safety standards such as the UL1741 [97] and IEEE1547 [98]. Currently, there are three
main interface architectures of PV inverters: the centralized inverters, the strings technology and
inverters integrated into the panels or micro-inverters. These architectures are further reviewed in
the following subsections.
350V is required at the input of the inverter. This is not always the case, since a bad panel or
partial shading will prevent the PV system to generate an adequate DC voltage level at the input
of the inverter. There are other examples, such as high-voltage DC cables interconnecting the PV
modules and the power inverter, power losses due to a centralized maximum power point
tracking (MPPT), and losses in the string isolation diodes. The grid-connected interface stage
(the bridge inverter) is usually line commutated by means of thyristors, generating many current
harmonics and hence poor power quality. Moreover, the reliability of the system is compromised,
since it depends only on one centralized inverter. Thus, when a failure on the central inverter
occurs, it causes a complete system shutdown, consequently, causing production downtime. All
these issues make this architecture less desirable. Thus, a different architecture that overcomes
these problems, in particular the issue of power quality, is required.
AC
DC
AC
Grid
eliminating those caused by the isolation diodes. The technical properties of this architecture
increase the reliability of the PV system. However, the number of medium power inverters is
considerably increased which dramatically increases the cost of the PV system as compared to the
previous architecture.
The advantages and the benefits to use the string inverter architecture over the centralized
architecture are detailed in [102]. However, both architectures have the issue of shading which
has an impact on the PV system throughput. A better solution consists of integrating the low
power inverter known as micro-inverter in each solar panel. This will be briefly described in the
following section.
AC
DC
AC
DC
AC
DC
AC
Grid
96
Further system expansion can easily be achieved since a new string with DC/DC converter can be
plugged into the existing platform.
The multi-string architecture can be manipulated to produce a modular PV system to mitigate the
problem with MPPT error detection. This modification consists of integrating the DC/DC
converter with its own MPPT controller to the solar panel. The DC/DC converters are connected
in series to provide the power inverter with an adequate voltage and the strings are then connected
in parallel in order to achieve the desired power. This is illustrated in Fig. 5.4.
DC
String1
AC
DC
DC
DC
String2
DC
DC
String3
DC
97
AC
Grid
DC
DC
DC
DC
DC
DC
AC
DC
DC
DC
AC
Grid
String1
String2
DC
DC
DC
DC
98
AC
DC
AC
Grid
AC
DC
AC
DC
AC
DC
99
String Inverter
Architecture
Micro-Inverter
Architecture
Architecture
PV Input Voltage
DC Losses
(Cables)
340-800
150-800
20-80
~1%
Negligible
~2%
Inverter Efficiency
95-97%
92-96%
87-93%
Fault Detection
Simple
Complex
Complex
In any case, the selection of architecture depends on the required power delivery and the cost per
watt of the installed PV system. This is depicted in Fig. 5.6.
3$/W
String architecture
2.5$/W
Multi strings
architecture
2$/W
Central Architecture
1.5$/W
Micro-inverters
architecture
1$/W
0$/W
1kW
10kW
100kW
As an interface between the PV panels and the DC/AC inverter, the DC/DC converter performs a
dual function of a) boosting the PV DC voltage of 40V to a DC voltage of 340V which is
required at the input of the inverter in order to feed the grid with 240Vac; and b) tracking the
maximum power delivered by the PV panel to maximize the energy throughput. Then the second
stage DC/AC inverter converts the DC power into an instantaneous current and voltage suitable
for the grid injection.
The digital control system consists of three main control functions: the MPPT control, the output
inductor current ILf, and the EMI control. Therefore, in the case of the two stage micro-inverter, a
control algorithm is designed to keep the first stage of the inverter (DC/DC converter) operating
at maximum power. While, the second stage control algorithm is dedicated to regulate the
instantaneous current or voltage to be injected into the grid.
PV Panel
Dboost
Lboost
Dflyback
T1
S1
S3
Lf
Z>>> Grid
Cboost
SWboost
Cbus
SWflyback
Cf
S2
S4
To S2 & S3
Vbus
To S1 & S4
Iin
ILf
Vin
Control Algorithm
(MPPT+ Inverter
Control + EMI Control)
ADC
DAC
FPGA-Based Controller
Vgrid
Sensor
(HPF)
Injector
(LPF)
101
focus on the integration of the proposed DAEF into the second stage control of the microinverter.
To verify the seamless integration of the DAEF into the digital controller of the micro-inverter, it
is required to investigate the stability of the inverter. This can be achieved by deriving the openloop transfer function of the micro-inverter and use one of the stability criteria such as root-locus
or gain-phase margin techniques. Furthermore, a compensator or a controller design is often
required to satisfy the stability conditions under different load and line variations.
As opposed to the DC/DC converters, it is difficult to design a micro-inverter/inverters controller
with high control gain at the fundamental frequency. This is particularly due to their time varying
voltages and/or currents. However, there exist three types of controller design techniques which
are currently used in single-phase power inverters.
The first technique [117], [118] is similar to that of DC/DC converters, with the assumptions that
the voltages and currents are DC variables instead of time variant parameters. This linear timeinvariant controller design approach proves to be inadequate for this type of systems since it
causes a significant steady state error in both amplitude and phase. In order to damp the resonance
of the LC filter in a single-phase inverter and to improve the control bandwidth, various multiple
loop controllers were proposed in literature, including capacitor current feedback, inductor current
feedback and their variations [119]-[124].
The second design approach [125]-[127] is to use a carrier sinusoidal control signal which is
multiplied by the output of the compensator rectified signal to produce a sinusoidal wave shape.
This signal is fed to the pulse width modulator to provide the gating signal of the inverter bridge
MOSFETs. This approach has two negative aspects: first, the gain of the control loop is
continuously varied along the sinusoidal waveform. The minimum gain at the zero crossing point
may cause significant distortion of the output signal due to the variation in the loop gain; second,
102
the regulator performance is very poor under variable load which results in excessive harmonic
contents due to the modulation effect of the compensator output with the carrier reference signal.
The third approach [128] which is adopted in this chapter, presents an alternative way for
controlling the instantaneous inductor current of single-phase DC/AC inverters using Direct
Quadrature (D-Q) reference frame control technique. In a D-Q reference frame, the physical
(Real) Circuit, in conjunction with an "Imaginary Orthogonal Circuit'', is transformed from the
stationary frame to the DQ rotating frame so that the steady state voltage or current in DQ frame
becomes DC variables allowing the design of controller using design techniques developed for
DC/DC converters. This design approach can achieve infinite control gain which results in zero
steady-state error at the fundamental frequency and better dynamic performance of the system.
Vgrid
ILf
Driver
Vin
Pin
cycle
delay
ID
R-I
IIm
To
D-Q
Imaginary Orthogonal
Transform
Circuit Variable
IQ
Duty_D
Current
Compensator
in
D-Q
Frame
Duty_Q
D-Q
To
R-I
Transform
Duty_R
To Switch Drivers
Iin
LPF
Switching Pattern
MPPT
Controller
HPF
EMI
Controller
To injection point
S1 & S4
S2 & S3
PLL
ID_ref
IQ_ref
Vgrid
Fig. 5.8 Block diagram micro-inverter controller architecture in rotating D-Q frame
103
It can be seen from Fig. 5.8 that the real variables, which are the output inductor current ILf and
the grid voltage Vgrid, are used to construct the imaginary circuit variables. These later are
transformed into the D-Q rotating frame using the Real-Imaginary (R-I) stationary frame. The
Proportional and Integral (PI) compensator is designed in the rotating frame with constant current
reference ID_ref and IQ_ref. This generates the duty cycle control signals Duty_Q and Duty_D. The
next step is the inverse transform of the duty cycle signals from D-Q frame to R-I stationary
frame. Finally, the duty cycle signals of the imaginary circuit is decoupled from the R-I stationary
frame and only the duty ratio of the real circuit is applied to the DC/AC inverter stage.
Equations (5.1) and (5.2) [128] provide the definition of the rotating transformation matrix from
the stationary frame to D-Q rotating frame. Equation (5.3) and (5.4) [128] are the inverse
transformation from the D-Q rotating frame, back to R-I stationary frame.
(5.1)
(5.2)
Where
and
and
represent the inductor current and the capacitor voltage in the rotating frame,
(5.3)
(5.4)
The circuit model in R-I stationary frame is depicted in Fig. 5.9.The average model of the real
and imaginary circuits can be obtained using the inductor current and the capacitor voltage of the
LC output filter and can be expressed in (5.5) and (5.6).
104
I_R
L_f
R_ESR
DRVbus
C_f
V_R
Zgrid
Ip
Vbus
I_I
R
+
V_I
DIVbus
-
Zgrid
(5.5)
(5.6)
By applying the D-Q transformation stated in (5.1) and (5.2) to (5.5) and (5.6), yield a circuit
model in the rotating frame which is expressed as.
(5.7)
(5.8)
The circuit model reflecting this transformation is depicted in Fig. 5.10.
105
wL_fI_q
L_f
I_D
R_ESR
DdVbus
wC_eqV_q
C_f
V_D
Zgrid
Ip(D_Q)
wL_fI_d
L_f
I_Q
Vbus
R_ESR
DqVbus
wC_eqV_d
C_f
V_Q
Zgrid
Topology: first stage buck-boost DC/DC converter, second stage full bridge DC/AC
inverter
The block diagram of the inductor current control loop in the s-domain and its corresponding
digitized system are shown in Fig. 5.11and Fig. 5.12 respectively. Four transfer functions (TF)
are required to verify the system stability. Namely, the plant or the inverter Gp(s), the decoupling
TF resulting from the D-Q transformation Gd(s), the DAEF transfer function FDAEF(s) derived in
chapter 2, and the compensator Gc(s) which need to be designed to fulfill the stability
requirement.
Compensator(Controller)
ILf_ref(s)
Gd(s)
Gc(s)
Gp(s)
GDAEF(s)
ILf(s)
K(sensor)
ILf_ref(n)
+
U(n)
Gc(z)
Hc
ZOH
Gd(s)
Gp(s)
GDAEF(s)
K(sensor)
FPGA/DSP
ILf(t)
The open-loop transfer function (TF) for the system without the compensator, can be written as.
(5.9)
Where;
is the modulator gain;
108
Therefore, the uncompensated control-to-inductor current open loop transfer function of the
micro-inverter system, can be evaluated in Matlab and written as,
(5.17)
The bode plot of the open loop uncompensated system transfer function is shown in Fig. 5.13.
The plot exhibits an unstable system, with infinite phase margin and the -20dB slope is far from
crossing the unity gain line (0dB line). Hence, to satisfy the stability criteria, the system must be
compensated.
Fig. 5.13 Bode plot of the open loop un-compensated control system
5.4.1 Compensator design
As previously mentioned, the imaginary circuit model provides a way to transform the real
circuit into the D-Q rotating frame. This converts the sinusoidal steady state into a DC steady
109
state operating point. This conversion allows the micro-inverter compensator to be designed
according to DC/DC converters design method. In this case, the type III compensator is found to
be adequate for this 7th order control system (5.17).
The type III compensator is referred to as double-pole double-zero compensation network
because it introduces double zeros into the error amplifier compensation to reduce the steep gain
slope above the double pole caused by the L-C filter and its associated -1800 phase shift. Type III
compensation network can achieve fast transient response and may provide more than 70 phase
boost. The circuit diagram of the type III compensator is shown in Fig. 5.14.
C2
C3
R3
R2
Vin
C1
Vout
R1
+
Rbias
Vref
GC ( s )
1 sR2C1 1 sC3 R1 R3
CC
sR1 C1 C2 1 sR3C3 1 sR2 1 2
C1 C2
(5.18)
R1 is arbitrarily chosen to be 10 k, so that the parameters values of the circuit can be evaluated
using the K-factor method [130] as follows:
M P 90
K tan
45
4
(5.19)
110
Where M is the desired phase margin in degree, and P is the phase-shift of the converter at
crossover frequency.
C2
1
15.2
2Fc GEA R1
pF
(5.20)
C1 C2 K 1 20 pF
(5.21)
R2
K
1.2 M
2Fc C1
(5.22)
R3
R1
7.5 K
K 1
(5.23)
C3
1
2Fc K R3
1.4 nF
(5.24)
Replacing the parameters values in (5.18), the transfer function of the compensator can be rewritten as,
(5.25)
The type III compensator has a double zero located at a frequency fz below Fc, and a double pole
located at a frequency fp above Fc given as,
fz
Fc
6.5 kHz
K
(5.26)
(5.27)
For complete calculation details of the type III compensator used in this design, please refer to
Appendix D of this thesis.
The gain and phase of the compensator transfer function are shown in Fig. 5.15 and Fig. 5.16
respectively. The plots reveal the amount of the gain and phase boost required to achieve the
system stability.
111
110
Magnitude (dB)
80
50
20
10
1
10
100
1 10
1 10
1 10
1 10
1 10
Frequency (Hz)
Phase (deg)
50
90
130
1
10
100
1 10
1 10
1 10
1 10
Frequency (Hz)
112
1 10
The control-to-inductor current open loop transfer function of the micro-inverter system,
including the compensator, is evaluated in Matlab and written as,
(5.28)
The Bode plot of the micro-inverter compensated control system is shown in Fig. 5.17. It can be
seen that the control system exhibits a gain margin of 28dB and phase margin of 79 degrees.
These parameters are large enough in providing the desired stability of the closed loop control
system.
113
The Nyquist plot, including the time-delay produced by the ZOH function, is shown in Fig. 5.18.
According to the Nyquist theorem, the net number of the encirclement N is equal to the number
of zeros in the right half-plane (RHP) minus the number of open loop poles P in the right halfplane, in other words;
(5.29)
In the case of the micro-inverter, the number of poles of the open loop TF is P=0, and the number
of zeros in the RHP of the closed loop TF is Z=0, therefore, N=0, which implies that the system is
stable. This is shown in Fig. 5.18, where there is no encirclement of the -1 point. This also proves
that the ZOH delay function is insignificantly low to have an impact on the system stability.
Fig. 5.18 Nyquist Plot of the micro-inverter closed loop control system
114
Test Receiver
PV Simulator
LISN
The first test was performed with the passive EMI filter only. The results are presented in Fig.
5.20.
The second round of testing was done without any filters in the micro-inverter unit. i.e.: the
passive EMI filter components were removed from the converter unit. The results are shown in
Fig. 5.21.
dBuV
80
70
60
50
40
30
20
10
0
-10
-20
1
10
(Start = 0.15, Stop = 30.00) MHz
Fig. 5.20 Conducted emissions spectrum of the micro-inverter with passive EMI filter
The last measurement was conducted with the DAEF prototype connected to the AC-side of the
micro-inverter unit. The resulting EMC spectrum is depicted in Fig. 5.22.
As it can be observed, from Fig. 5.20 and Fig. 5.22, the EMI attenuation performance of the
DAEF prototype can match or outperform the one with the passive EMI filter. Furthermore, this
confirms the viability of the proposed DAEF to replace the conventional passive EMI filter.
116
Hence, significant reduction of the PCB space, regardless of the power rating of the microinverter unit, can be achieved.
dBuV
80
70
60
50
40
30
20
10
0
-10
-20
1
10
(Start = 0.15, Stop = 30.00) MHz
Fig. 5.21 Conducted emissions spectrum of the micro-inverter without EMI filters
dBuV
80
70
60
50
40
30
20
10
0
-10
-20
1
10
(Start = 0.15, Stop = 30.00) MHz
Fig. 5.22 Conducted emissions spectrum of the micro-inverter with DAEF installed
117
5.6 Summary
The main objective of this chapter is to prove the feasibility of replacing the conventional passive
EMI filter by the DAEF in a solar grid-tied micro-inverter. This is to say that the DAEF can be
integrated into the digital controller of the micro-inverter.
First, this chapter started by providing a brief introduction of the grid-tied photovoltaic inverter. It
described one of the critical components that made up the solar PV system, the micro-inverter. Its
principle of operation was briefly discussed. Then different PV architectures were presented. It is
found that the grid-tied micro-inverter configuration is the best suited to resolve the issue of
shading and minimize cable losses between PV panels. However, there is a trade-off between
efficiency and architecture. The centralized configuration tends to have better efficiency than the
micro-inverter string configuration.
Second, the design and modeling of the micro-inverter was thoroughly analyzed using the D-Q
modeling technique to transform the real circuit into an imaginary circuit, in order to find the DC
operating point. In this state the compensator can be designed using the DC/DC design method.
The closed loop and open loop transfer functions were derived. The stability of the micro-inverter
was verified.
Finally, the experiment results were presented to prove the validity of the DAEF to be integrated
in the micro-inverter circuit as an EMI suppression technique. Hence, significant reduction of the
PCB space regardless of the power rating can be achieved.
118
Chapter 6
Proposed DAEF Integration in a DSP-Based DC-DC Digital Controller
Used in Electric Vehicle (EV) Battery Charger
6.1 Introduction
This chapter presents an industrial application case study, in which the proposed DAEF is
integrated into a DSP-based digital controller of a DC-DC converter used in electrical vehicles
power conversion system. As opposed to the stand-alone version of the DAEF, the control
algorithm is added to the inverter main program to form a digital control system, leading to
further reduction of the size, cost and space of the overall power inverter printed circuit board
(PCB).
In section 6.2, general description of the EV power conversion system and the different
topologies that are used as a battery chargers, are presented. In section 6.3, the closed loop control
system of the full bridge DC-DC converter is studied. The design of the compensator, taking into
account the DAEF transfer function, is presented. The frequency response of the dual voltage and
current control loops is required to verify the stability of the DC-DC converter, using the gain and
phase margin criteria. Experimental results showing the performance of the integrated DAEF in
the digital controller are illustrated in section 6.4. Finally, a summary is given in section 6.5.
transportation and a reduction of the environmental impact caused by the vehicles. The
sociological approach requires a radical change in personal attitudes, and a significant overhaul of
the economical activities related to transportation. The attitude is somewhat impossible given our
dependence on cars. However, a technical solution to reduce the environmental impact is rather
collectively acceptable. This is embodied by the electric vehicle (EV). Moreover, the advent of
the all-electric vehicle faces two pitfalls: a lack of technical maturity and a difficulty of the public
acceptance. This is mainly due to its autonomy which is limited by the energy stored in the
battery and the re-charge time; whereas, the conventional internal combustion engine vehicle
(ICEV) has more mobility in terms of mileage and an insignificant time to re-fill the tank. Today,
most of the technical efforts focus on these two aspects namely autonomy and re-charge time.
Unfortunately, the results have not reached the market expectations. This leads to a compromise
solution which is the hybrid (gas/electrical) vehicle (HV). While, the all-electric vehicle is not
ready for large scale deployment, the hybrid vehicle (HV) and fuel cell vehicle (FCV) are a
reality with more than a million cars on the road.
Many research efforts have been focused on developing efficient, reliable, and low-cost power
conversion techniques for the future new energy vehicles [131]-[136]. Two distinct
configurations are available: the plug-in hybrid electric vehicle (PHEV); and the fuel cell hybrid
electric vehicle (FCHEV). This later has an advantage of using hydrogen based energy required
for its autonomy. However, the FCHEV is seen as a long term solution due to its actual cost and
its complex manufacturing. The FCHEV configuration is shown in Fig. 6.1.
120
Gas Tank
Fuel Cell
Combustion Engine/
Motor
Boost Conv.
Buck/Boost
Conv.
DC/AC
Inverter
Battery
Electric
Motor
Wheel
Clutch System
Wheel
Differentiator
As for the PHEV, it is a full hybrid vehicle with particular characteristics of having high voltage
batteries which can be charged using a conventional 110Vac outlet. The drive train consists of
two drive options, one is the full drive in which the vehicle is driven using full electrical energy
stored in the batteries, whereas, the other option is the mixed drive where the combustion engine
is used when necessary. These options make the PHEV one of the best vehicles in terms
performance, lower CO2 emissions and higher fuel economy. The first PHEV prototype was
121
designed in 2004 by California cars initiative. Other prototypes emerged then after. The block
diagram of the PHEV configuration is depicted in Fig. 6.2.
Gas Tank
HV
Batteries
AC/DC
Converter
Full Bridge
DC-DC
Converter
Combustion
Engine/Motor
DC/AC
Inverter
LV
Battery
Auxiliary
Circuits
Electric
Motor
Wheel
Clutch System
Wheel
Differentiator
In both configurations, the alternative energy source is used to assist the propulsion of the vehicle
during transient and to absorb the kinetic energy during regenerative braking. In the FCHEV
topology, the fuel cell pack is connected to the DC Bus via a boost converter and the energy
storage battery is connected to the DC Bus via a bidirectional DC-DC converter. In the PHEV,
122
the source of energy is directly drawn from the grid and it is used to charge the high voltage (HV)
battery pack through the AC-DC converter which includes power factor correction (PFC)
strategy. The HV battery provides the required energy to drive the electric AC motor through the
DC-AC inverter. The full bridge DC-DC converter is tapped from the HV battery source to
charge the low voltage (LV) battery to provide power to auxiliary circuits. The focus of this
chapter is to integrate the DAEF into the digital controller of the full-bridge DC-DC low voltage
battery charger. The stability assessment of the DC-DC converter including the DAEF will be
investigated, to ensure a safe operation of the converter while providing a significant EMI
suppression.
asymmetric auxiliary circuit is shown in Fig. 6.3. The steady-state analysis of this circuit is
explained in [148].
S1
S3
C1
C3
Laux1
Laux2
VAO
HV Battery
300Vdc
VBO
C4
S2
S4
C2
Ll
Lf1
Lf2
T1
LV Battery
12Vdc
Cf
SR1
SR2
losses. Hence, the overall efficiency of the power converter can be significantly improved. The
operating principles of the current doublers synchronous rectifier are fully addressed in [149].
125
better load transient response. The phase margin and bandwidth of the closed loop system are also
improved as a result.
point from the sensed point to prevent the noise signal from flowing towards the auxiliary circuits
interfacing the power converter.
S3
C1
C3
Laux1
Laux2
VAO
HV Battery
300Vdc
VBO
C4
S4
S2
Hsense1(s)
C2
Vout
Ll
Lf1
Lf2
T1
LV Battery
12Vdc
Cf
SR1
SR2
Hsense(s)
Hinj(s)
Sensor
Vs1
Kpwm
KADC
KDAC
Kinv(z)
KADC
Phase-shift Modulator
KADC
Iref
Z-Td/Ts
Delay
ZOH
Hinteg(z)
Integrator
Hinteg(z)
Hc(z)
Current loop
Compensator
Integrator
HV(z)
Voltage loop
Compensator
Vref
Fig. 6.4 closed loop block diagram of the EV auxiliary battery charger
In order to investigate the stability of the converter system including the DAEF, the frequency
response of all the blocks need to be characterized.
127
In order to have an infinite dc loop gain to ensure a zero steady-state error, the output of
compensator passes through an Euler integrator. The integrator is described by the following
transfer function.
(6.2)
A soft complex conjugate zero pair is chosen due to its 180 degrees phase boost and its gain
increase of 40dB/decade. This is illustrated in Fig. 6.5. The transfer function of the complex zero
pair is given by.
(6.3)
8
20log H comp f 2
9
20log H comp f 2
10
20log H comp f 2
12
20log H comp f 2
171.102
10
11
De
cr
ea
sin
arg H comp f 2
arg H comp f 2
Magnitude
20log H comp f 2
arg H comp f 2
50
arg H comp f 2
arg H comp f 2
100 100
100
100
1 10
1 10
Frequency
f
1 10
5
10
180
200
150
180
Phase
100
180
50
CD
180
ecr
eas
ing
180
5.324 50
100
100
1 10
1 10
f
Frequency
1 10
5
110
Fig. 6.5 Bode Plot of soft complex digital zero-pair used for the system compensation
Using the conventional average modeling techniques, the closed loop control-to-output voltage
transfer function and control-to-inductor current transfer function of the uncompensated system,
can be expressed respectively as.
(6.4)
(6.5)
128
Where;
(6.8)
is the output voltage sensor which consists of the sensing gain
RC filter whose time-constant is
and a first-order
The transfer function of the control to output current of the converter is derived as
(6.11)
Where
and
are the output filter capacitor, the output inductor and the load resistor
respectively.
Similarly, the control to output voltage of the converter is derived as.
(6.12)
129
is the transfer function of the digital active EMI filter and it is given by.
(6.13)
Using equations (6.2), (6.3), (6.4) and (6.5), the loop gains of the inner current loop and the outer
voltage loop of the compensated system can be respectively written as.
(6.14)
(6.15)
Fig. 6.6, Fig. 6.7, Fig. 6.8, and Fig. 6.9 show the frequency response of the converter system for
the inner current and the outer voltage loop gains with compensation, taking into account the
parameters presented in Table 6-1. These plots indicate that the system without the compensation
would result in an unstable system. Adding a soft-complex zero pair of b=256 produces a phase
boost of 180o at about 20 KHz in the inner current loop. The digital zero generates a significant
increase in the phase margin of the voltage loop. The compensated current loop has a cross-over
frequency of 200 KHz with a phase margin of 40o. Similarly, the outer voltage compensated loop
gain exhibits a 64o phase margin at 6 KHz cross-over frequency. Low control bandwidth is used
for the voltage loop, due to the slow response of the energy accumulator or the battery. Further
details can be found in Appendix D.
130
150
Magnitude (dB)
100
50
0
0
50
100
150
10
100
1 10
4
5
1 10
1 10
Frequency (Hz)
1 10
1 10
uncompensated system
compensated system
Fig. 6.6 Frequency response of the current loop gain (Magnitude)
270
Magnitude (dB)
180
90
0
90
180
180
270
10
100
1 10
1 10
1 10
Frequency (Hz)
4
1 10
uncompensated
compensated
131
1 10
100
65
Magnitude (dB)
30
5
40
75
110
145
180
10
100
3
4
1 10
1 10
Frequency (Hz)
1 10
1 10
Uncompensated
Compensated
Fig. 6.8 Frequency response of the outer voltage loop gain (Magnitude)
270
Magnitude (dB)
180
90
0
90
90
180
180
270
10
100
1 10
1 10
Frequency (Hz)
3
1 10
Uncompensated
Compensated
Fig. 6.9 Frequency response of the outer voltage loop gain (Phase)
132
1 10
300 Vdc
Output voltage
12 Vdc
Output current
90 Amps
Output inductor Lf
1.6 H
Output capacitor Cf
200 F
Load resistance Ro
0.133
Switching frequency fs
200 KHz
250 MHz
133
DSP Board
DC-DC
Converter
(EUT)
134
Fig. 6.12 Transient response to a step down change in the load current
135
The conducted emission measurements were carried out according to CISPR16-1 test method.
The first test was done with the passive EMI filter designed into the DC-DC converter and the
result is shown in Fig. 6.13. An average noise peak of 60dBuV across the spectrum (150 KHz
30 MHz) can be observed, with the highest peak at 74dBuV.
Fig. 6.13 Conducted emission spectrum with the passive EMI filter
The second test was conducted with no EMI filter installed in the DC-DC converter. The result is
reflected in Fig. 6.14. From this figure, an average noise peak of 80dBuV can be seen, with the
highest peak at 94dBuV.
136
137
From the above conducted EMI spectrum plots, it can be deduced that replacing the passive EMI
filter with the DAEF produces similar or better performance, across the frequency range of 150
KHz to 30MHz.
6.5 Summary
In this chapter, the seamless integration of the DAEF has been demonstrated as being a valid EMI
solution for an industrial application such as the Electric vehicle DC-DC battery charger. This
chapter briefly introduce the power conversion system of a battery powered Electric Vehicle. It
describe the different variations of the EV, namely the hybrid (HEV) and the PHEV. This later
proves to be the best drive system configuration compared to other topologies. The circuit
analysis for the full-bridge resonant DC-DC converter used as a battery charger in the PHEV was
presented. The digital controller design, including the DAEF was derived and the stability
assessment was theoretically verified. Finally, experimental results were illustrated to validate the
co-existence of the DAEF with the digital controller and to achieve a significant EMI attenuation
without the need for the passive EMI filter.
138
Chapter 7
Conclusions & Future Work
7.1 Conclusions
Industrial trends tend to converge towards power converters that can operate at switching
frequencies in the 1-10 MHz range, to achieve greater power density and improved transient
response. However, these technological advances come with a price tag. A tremendous EMI noise
is generated in the power unit, which consequently pollute the utility systems. An immediate
solution would be a passive EMI filter to suppress the conducted emissions and to fulfill the EMC
requirements set forth by government authorities and international standards. The actual
challenge is that the passive EMI filtering solutions are inadequate in following the power density
trend. In other words, the greater the power density of the converter, the bigger the passive filter.
In most cases, the passive EMI filter takes one quarter to a half of the PCB space of the power
converter. To meet the next generation requirements of these applications, in terms of EMI
suppression, four ideas have been proposed in this thesis. These ideas can be summarized as
follows:
1. A DSP-based technique to suppress conducted EMI emissions in power converters has
been proposed. This technique exploits the theory of sampling using data acquisition
devices such as ADC and DAC for discrete time conversion of the EMI noise source. The
impulse function to represent the sampling process has been used for the sensed signal
recovery. The figure-of-merit of the attenuation transfer function, in terms of magnitude,
has revealed that the gain of the injector has a direct impact on the increase of the
magnitude within the desired bandwidth. Finally, the simulation and the experimental
139
results prove the validity of the proposed EMI suppression technique, which can the
passive EMI filters in the power converter.
The content of this chapter has been submitted to the IEEE transactions on power
electronics.
2. An FPGA-based technique to suppress conducted EMI emissions in power converters has
been introduced. The principle of operation of the proposed technique has been
explained. The sampled noise is processed using VHDL algorithm, with a set of
instructions that can be executed simultaneously, hence reducing the delay time. The
analysis and design approach were carried out in z-domain. Using the theory of ZOH
approximation the closed loop transfer function is derived in discrete-time. With the
sample rate of 10 times the noise bandwidth, it was found that the approximation was
adequate in meeting the design requirements. Finally, comparing the two techniques
presented in chapter 3 and chapter 4, it can be concluded that the difference resides in
their performance and the cost of the implementation. In other words, the FPGA-Based
DAEF performs slightly better than the DSP-Based DAEF, since this later rely on the
sequential execution of the algorithm, whereas for the FPGA-Based technique, the
algorithm is executed simultaneously which results in improved attenuation. The other
factor is the cost; in this case the cost for implementing the DSP-Based technique is much
cheaper than the FPGA-Based technique.
The content of this chapter is subject to patent application in the US and Canada. It has
also been submitted for publication in the IEEE Transactions on Industrial Electronics.
3. Integration of the Proposed DAEF in a Digital Controller of a grid-tied Photovoltaic
Micro-inverter has been realized. The design and modeling of the micro-inverter was
140
thoroughly analyzed using the D-Q modeling technique to transform the real circuit into
an imaginary circuit, in order to find the DC operating point. In this state the compensator
can be designed using the DC/DC design method. The closed loop and open loop transfer
functions were derived. The stability of the micro-inverter was verified. Finally, the
experiment results were presented to prove the validity of the DAEF to be integrated in
the micro-inverter circuit as an EMI suppression technique. Hence, significant reduction
of the PCB space regardless of the converter power rating can be achieved.
4. The seamless integration of the DAEF has been demonstrated as being a valid EMI
solution for an industrial application such as the Electric vehicle DC-DC battery charger.
The power conversion system of a battery powered Electric Vehicle has been briefly
introduced. Different variations of the EV have been described, namely the Hybrid
Electric Vehicle (HEV) and the Plug-in Hybrid Electric Vehicle (PHEV). The circuit
analysis for the full-bridge resonant DC-DC converter used as a battery charger in the
PHEV has been presented. The digital controller design, including the DAEF has been
derived. The stability assessment has been theoretically verified. Finally, experimental
results were illustrated to validate the co-existence of the DAEF with the digital
controller and to achieve a significant EMI attenuation without the need for the passive
EMI filter.
be possible to explore an application specific integrated circuit (ASIC) to provide a smaller, more
reliable and cost effective solution.
The integrated solution proposed in chapter 5 and chapter 6 is based on control algorithm to
reverse the phase of the sensed interference signal. The DAC and the ADC converters used in this
application are discrete components, hence proper PCB grounding need to be addressed. Future
research might be possible to convert the DAC and the ADC functions into a VHDL code as well
as the sense and the injection circuits. The final product would be one single package digital EMI
filter that can be a plug-and-play or an add-on feature into the power converter controller.
142
References
[1]
IEC Draft 22G-WG4-11 Electromagnetic compatibility of power drive systems, Oct. 1994
[2]
[3]
M.C. Caponet, & F. Profumo, "Devices for the separation of the common and differential
mode noise: design and realization," Applied Power Electronics Conference and Exposition,
2002. APEC 2002. Seventeenth Annual IEEE , vol.1, no., pp.100-105 vol.1, 2002
[4]
A. A. Toppeto, Test method to differentiate common mode and differential mode noise, in
Proc. IEEE EMC Symp., Rotterdam, Netherlands, 1979, pp. 497-502
[5]
[6]
J. Roudet, R. Scheich, and A. Orlandi, Origin and propagation of common mode currents
occurring in a bridge rectifier in Proc., 15th Int. Telecommun. Energy Conf., Paris, France,
Sept. 27-30, 1993
[7]
C. R. Paul and K. B. Hardin, Diagnosis and reduction of conducted noise emission, IEEE
Trans. Electromagnetic Compatibility., vol. 30, no. 4, pp. 553-560. 1988. New York: Van
Nostrand Reinhold, 1991
[8]
[9]
J. Roudet, R. Scheich, and J. Mahdavi, Conducted RFI emission from A DC-DC converter
with sinusoidal line current, in Proc. IEEE 1993 IAS Conf, Toronto, Canada, 1993
143
[10]
[11]
[12]
[13]
B. Choi, and B. H. Cho, Intermediate Line Filter Design to Meet Both Impedance
Compatibility and EMI Specifications IEEE Transactions on Power Electronics, Vol.10,
No.5, pp. 583-588, Sept. 1995
[14]
[15]
H. Matsuda et al, "Analysis of common mode noise in switch mode power supplies", NEC
Technical Report. 1998, Vol. 51, pp. 60-65
[16]
S. Ye, W. Eberle, and Y. F. Liu, A Novel EMI Filter Design Method for Switching Power
Supplies IEEE Transactions on Power Electronics, Vol. 19, No. 6, November 2004
[17]
P. Li and B. Lehman, Accurate Loop Gain Prediction for DC-DC Converter Due to the
Impact of Source/Input Filter IEEE Transactions on Power Electronics, Vol.20, No.4, pp.
754- 761, July 2005
[18]
D. Zhang, D. Y. Chen, M. J. Nave, and D. Sable, Measurement of noise impedance of offline converters, IEEE Transactions on Power Electronics, vol. 15, pp. 820825, Sept. 2000
144
[19]
[20]
M. Jin, and M. Weiming, Power Converter EMI Analysis Including IGBT Nonlinear
Switching Transient Model IEEE Transactions on Industrial Electronics, Vol. 53, No. 5,
October 2006
[21]
C. Hua, C. Shen, and G. Peng, Design of an Input Filter to Suppress Conducted EMI of a
Single Phase UPS Inverter In Proc. IEEE 22nd International Conference IECON, vol.3,
pp.1699 1704, 5-10 Aug.1996
[22]
F.L. Martin, F.W. Arthur, Input filter Design for multiple-module DC power systems IEEE
Transactions on Power Electronics, Vol.11, No.3, pp. 472 - 488, May 1996
[23]
F. C. Lee, and Y. Yu, Input filter design for switching regulators, IEEE Transactions on
Aerospace and Electronic System, vol. AES-15, no. 5, pp. 627-634, 1979
[24]
[25]
M. J. Nave, Power Line Filter Design for Switched-Mode Power Supplies. New York: Van
Nostrand Reinhold, 1991
[26]
[27]
[28]
G. Hua, et al, Novel Zero Voltage Transition PWM Converters, Proceedings of the Power
Electronics Specialist Conference, Spain, June 29 - July 3, 1992.
145
[29]
M. M. Jovanovic, W. A. Tahisz, and F. C. Lee, High frequency off- line power conversion
using zero voltage switching quasiresonant and multi-resonant techniques, IEEE Trans.
Power Electronics, pp. 459-469. Oct. 1989
[30]
[31]
[32]
[33]
[34]
M. M. Jovanovic, D. C. Hopkins, and F. C. Lee, Evaluation and design of megahertzfrequency off-line zero current switched quasiresonant converters IEEE Trans. Power
Electron., pp. 136-146, Jan. 1989
[35]
D. C. Hopkins, M. M. Jovanovic, F. C. Lee, and F. W. Stephenson, Hybridized off-line 2MHz zero current switched quasiresonant converters IEEE Trans. Power Electronics, pp.
147-154, Jan. 1989
[36]
P. Jain. H. Soin, and M. Cardella, Constant frequency resonant dc/dc converter topologies
with zero switching losses IEEE Trans. Aerospace Electronic Syst., vol. 30, no. 2 pp. 534544, Apr. 1994
[37]
P. K. Jain, Asymmetrical pulse width modulated resonant dc/dc converter U.S. Patent
5159541, Oct. 27, 1992
146
[38]
[39]
[40]
[41]
[42]
K. B. Hardin, J. T. Fessler, and D. R. Bush, Spread spectrum clock generation for the
reduction of radiated emissions in Proc. IEEE Int. Symp. Electromagnetic Compatibility,
1994, pp. 227231
[43]
H. Li, Y. Cheng, and D. Puar, Dual-loop spread-spectrum clock generator, in Dig. Tech.
Papers, IEEE Int. Solid State Circuits Conf., 1999, pp. 184185
[44]
[45]
[46]
[47]
[48]
[49]
T. Ninomiya, M. Shoyama, C. F. Jin, and G. Li, EMI issues in switching power converters,
in Proc. IEEE Int. Symp. Power Semiconductor Devices I. C, (ISPSD), Jun. 47, pp. 8186,
2001
[50]
[51]
[52]
[53]
[54]
S. Johnson and R. Zane, Custom Spectral Shaping for EMI Reduction in High Frequency
Inverters and Ballast IEEE Transactions on Power Electronics, Vol. 20, No. 6, November
2005
[55]
L. Laskaii, P. N. Enjeti, and I. J. Pitel, White noise modulation of high frequency high
intensity discharge lamp ballasts, IEEE Trans. Ind. Appl., vol. 34, no. 3,pp. 597605,
May/Jun. 1998
148
[56]
[57]
Y. C. Son and S. K. Sul, A new active common mode EMI filter for PWM inverter, IEEE
Trans. Power Electron., vol. 18, no. 6, pp. 13091314, Nov. 2003
[58]
P. C. Murphy, An active ripple filtering technique for improving common mode inductor
performance, IEEE Power Electron. Letter., vol. 2, no. 2,pp. 4550, Jun. 2004
[59]
T. Farkas and M. F. Schlecht, Viability of active EMI filters for utility applications, IEEE
Trans. Power Electron., vol. 9, pp. 328337, May 1994
[60]
D. Y. Lee and B. H. Cho, Design of an input filter for power factor correction (PFC) AC to
DC converters employing an active ripple cancellation, in Proc. 3rd Intersoc. Energy
Conversion Eng. Conf., 1996, pp. 582586
[61]
D. C. Hamill, An efficient active ripple filter for use in DCDC conversion, IEEE Trans.
Aero. Electronic Syst., vol. 32, pp. 10771084, July 1996.
[62]
[63]
I. Takahashi, A. Ogata, H. Kanazawa, and A. Hiruma, Active EMI filter for switching noise
of high frequency inverters, in Proc. Power Conversion Conf., Nagaoka, Japan, 1997, pp.
331334
[64]
A. L. Julian, R. Cuzner, G. Oriti, and T. A. Lipo, Active filtering for common mode
conducted EMI reduction in voltage source inverters, in Proc. IEEE APEC, 1998, pp.934
939
[65]
S. Ogasawara, H. Ayano, and H. Akagi, An active circuit for cancellation of common mode
voltage generated by a PWM inverter, IEEE Trans. Power Electron., vol. 13, pp. 835841,
Sept. 1998
149
[66]
[11] D. C. Hamill and P. T. Krein, A zero ripple technique applicable to any DC converter,
in Proc. IEEE PESC Rec., pp.11651171, 1999,
[67]
L. Bruton, Active RC Circuits: Theory and Design. Englewood Cliffs, NJ: Prentice Hall,
1980.
[68]
J. Marrero and C. Pang, Ripple current reduction circuit, U. S. Patent 5038263, Aug. 1991
[69]
L. LaWhite and M. F. Schlecht, Design of active ripple filters for power circuits operating in
the 110MHz range, IEEE Trans. Power Electron., vol. 3,no. 3,pp. 310317, Jul. 1988.
[70]
A. C. Chow and D. J. Perreault, Design and evaluation of an active ripple filter using voltage
injection, in Proc. IEEE Power Electronics Specialist Conf.01, Jun. 1721, pp. 390397,
2001
[71]
Y. C. Son and S. K. Sul, Generalization of active filters for EMI reduction and harmonics
compensation, in Proc. 38th Industry Applications Conf., Oct. 1216, pp. 12091214, 2003
[72]
M. El-Habrouk, Active power filters: A review, Proc. Inst. Elect. Eng. Electric Power
Appl., vol. 147, pp.403413, Sep. 2000
[73]
W. Chen, X. Yang, and Z. Wang, An Active EMI Filtering Technique for Improving Passive
Filter Low Frequency Performance," IEEE Transactions on Electromagnetic Compatibility,
Vol. 48, No.1, Feb. 2006
[74]
Hamza, and P. K. Jain, Conducted EMI noise mitigation in DC-DC converters using active
filtering method, IEEE Power Electronics Specialists Conference, pp. 188 194, 2008
[75]
K.W. Yuk, and, L.Y. Shu, and K.C. David, Synthesis of a wide bandwidth current sensor
using a current transformer for active power filter applications, IEEE- PESC, Vol.1, pp. 270
273, 2000
[76]
Y. Duan, and H. Jin, Digital controller design for switch mode power converters, in Proc.
IEEE Appl. Power Electron. Conf. (APEC), vol. 2, pp. 967973, 1999
150
[77]
[78]
[79]
[80]
[81]
M. Fuand Q. Chen,A DSP based controller for Power Factor Correction (PFC) in a rectifier
circuit, in Proc. IEEE Appl. Power Electron. Conf. (APEC),vol.1, pp.144149, 2001
[82]
Y.T. Feng, G. L.Tsai, and Y. Y. Tzou, Digital control of a single stage single switch flyback
PFC AC/DC converter with fast dynamic response, in Proc. IEEE Power Electron. Spec.
Conf. (PESC), vol. 2, pp.12511256, 2001
[83]
A. Prodic, J. Chen, R.W. Erickson, and D. Maksimovic, Digitally controlled low harmonic
rectifier having fast dynamic response, in Proc. IEEE Appl. Power Electron. Conf. (APEC),
vol. 1, pp. 476782, 2002
[84]
L. Keh-La, K. Armin, and H. J. Bedrich " Modular Low-Power, High-Speed CMOS Analogto-Digital Converter for Embedded Systems", press, 2004
[85]
[86]
[87]
[88]
A. de Castro, P. Zumel, O. Garca, T. Riesgo, and J. Uceda, Concurrent and Simple Digital
Controller of an AC/DC Converter with Power Factor Correction Based on an FPGA, IEEE
Transactions on Power Electronics, Vol. 18, No. 1, January 2003
[89]
S. Hauk,The roles of FPGAs in reprogrammable systems, Proc. IEEE, vol. 86, pp. 615
638, Apr. 1998
[90]
W. Tang, Y. Jiang, G. C. Hua, F. C. Lee, and I. Conen, Power factor correction with flyback
converter employing charge control, in Proc. IEEE Appl. Power Electron. Conf. (APEC),
pp. 293298, 1993
[91]
[92]
[93]
A. Rushton, Ed., VHDL for Logic Synthesis. New York: Wiley, 1998.
[94]
[95]
[96]
IEEE Standard VHDL Analog and Mixed-Signal Extensions, IEEE Std 1076.1,1999
[97]
Equipment
for
Use
With
[98]
IEEE1547 IEEE Standard for Interconnecting Distributed Resources With Electric Power
Systems, IEEE Std. 1547, 2003.
[99]
[100]
H. Haeberlin, Evolution of inverters for grid connected PV-systems from 1989 to 2000, in
Proc. 17th Eur. Photovoltaic Solar Energy Conf., Munich, Germany, Oct. 2226, pp. 426
430, 2001
[101]
F. Chan, & H. Calleja , "Reliability Estimation of Three Single-Phase Topologies in GridConnected PV Systems," Industrial Electronics, IEEE Transactions on , vol.58, no.7,
pp.2683-2689, July 2011
[102]
[103]
N. A. Rahim, J. Selvaraj, "Multistring Five-Level Inverter With Novel PWM Control Scheme
for PV Application," Industrial Electronics, IEEE Transactions on , vol.57, no.6, pp.21112123, June 2010
[104]
A.M. Pavan, S. Castellan, S. Quaia, S. Roitti, and G. Sulligoi, " Power Electronic
Conditioning Systems for Industrial Photovoltaic Fields: Centralized or String Inverters? "
Clean Electrical Power, 2007. ICCEP '07. International Conference on, pp., 208 214, 21-23
May 2007
[105]
1999
[106]
T. Shimizu, K. Wada, and N. Nakamura, A flyback type single phase utility interactive
inverter with low-frequency ripple current reduction on the DC input for an AC photovoltaic
module system, in Proc. IEEE PESC02, vol. 3, pp. 14831488, 2002
[107]
S. B. Kjaer and F. Blaabjerg, Design optimization of a single phase inverter for photovoltaic
applications, in Proc. IEEE PESC03, vol. 3, pp. 11831190, 2003
[108]
M. Nagao and K. Harada, Power flow of photovoltaic system using buck-boost PWM power
inverter, in Proc. PEDS97, vol. 1, pp.144149, 1997
[109]
[110]
E. Achille, T. Martir, C. Glaize, and C. Joubert, Optimized DC-AC boost converters for
modular photovoltaic grid-connected generators, in Proc. IEEE ISIE04, pp. 10051010,
2004
[111]
D. C. Martins and R. Demonti, Grid connected PV system using two energy processing
stages, in Conf. Rec. 29th IEEE Photovoltaic Specialists Conf., pp. 16491652, 2002
[112]
A. Lohner, T. Meyer, and A. Nagel, A new panel integratable inverter concept for gridconnected photovoltaic systems, in Proc. IEEE ISIE96, vol. 2, pp. 827831, 1996
[113]
[114]
Q. Li, and P. Wolfs, A Review of the Single Phase Photovoltaic Module Integrated
Converter Topologies with Three Different DC Link Configurations IEEE Transactions on
Power Electronics, vol. 23, no. 3, May 2008
154
[115]
[116]
[117]
D.N. Zmood and D.G. Holmes, "Stationary frame current regulation of PWM inverters with
zero steady state error," Proceedings of IEEE Power Electronics Specialists Conference, pp.
1185-1190, 1999,
[118]
frame
generalized integrators for current control of active power filters with zero steady state error
for current harmonics of concern under unbalanced and distorted operation conditions,"
IEEE Transactions on Industry Applications, Vol. 38, No. 2, pp. 523 -532, Mar/Apr 2002,
[119]
E. Villanueva, P. Correa, J. Rodriguez, & M. Pacas , "Control of a Single-Phase Cascaded HBridge Multilevel Inverter for Grid-Connected Photovoltaic Systems," Industrial Electronics,
IEEE Transactions on , vol.56, no.11, pp.4399-4406, Nov. 2009
[120]
[121]
[122]
M. Castilla, J. Miret, J. Matas, L. Garcia de Vicuna, & J.M. Guerrero , "Control Design
155
R. Gonzalez, E. Gubia, J. Lopez, & L. Marroyo , "Transformerless Single-Phase MultilevelBased Photovoltaic Inverter," Industrial Electronics, IEEE Transactions on , vol.55, no.7,
pp.2694-2702, July 2008
[124]
R.A. Mastromauro, M. Liserre, T. Kerekes, & A. Dell'Aquila, "A Single-Phase VoltageControlled Grid-Connected Photovoltaic System With Power Quality Conditioner
Functionality," Industrial Electronics, IEEE Transactions on , vol.56, no.11, pp.4436-4444,
Nov. 2009
[125]
[126]
Q. Zhang, X. Sun, Y. Zhong, M. Matsui, & B. Ren, "Analysis and Design of a Digital PhaseLocked-Loop for Single-Phase Grid-Connected Power Conversion Systems," Industrial
Electronics, IEEE Transactions on , vol.58, no.8, pp.3581-3592, Aug. 2011
[127]
M. Castilla, J. Miret, J. Matas, L.G. de Vicua, & , J.M. Guerrero, "Linear Current Control
Scheme With Series Resonant Harmonic Compensator for Single-Phase Grid-Connected
Photovoltaic Inverters," Industrial Electronics, IEEE Transactions on , vol.55, no.7, pp.27242733, July 2008
[128]
L. Zhigang, L. Alesi, Z. Xiaohu, J. Li, and A. Huang, Digital Controller Design for PV
Inverter with Model Based Technique IEEE proceedings, Energy Congress and Exposition
(ECCE), pp.2950-2955, 2010
[129]
SPARQ systems inc. is a spin-off company under Queens university and its technology
156
[131]
N.D. Weise, K.K. Mohapatra, & N. Mohan, "Universal utility interface for Plug-in Hybrid
electric vehicles with vehicle-to-grid functionality," Power and Energy Society General
Meeting, 2010 IEEE , vol., no., pp.1-8, 25-29 July 2010
[132]
[133]
S.A. Rahman, N. Zhang, & Z. Jianguo, "Modeling and simulation of an energy management
system for Plug-in Hybrid Electric Vehicles," Power Engineering Conference, 2008. AUPEC
'08. Australasian Universities , vol., no., pp.1-6, 14-17 Dec. 2008
[134]
X. Xu, C. Wang, G. Liao, C.P. Yeh, W. Stark, "Development of a Plug-In Hybrid Electric
Vehicle educational demonstration unit," North American Power Symposium (NAPS), 2009 ,
vol., no., pp.1-7, 4-6 Oct. 2009
[135]
S.G. Wirasingha, & A. Emadi, "Classification and Review of Control Strategies for Plug-In
Hybrid Electric Vehicles," Vehicular Technology, IEEE Transactions on , vol.60, no.1,
pp.111-122, Jan. 2011
[136]
M. Ferdowsi, "Plug-in Hybrid Vehicles - A Vision for the Future," Vehicle Power and
Propulsion Conference, 2007. VPPC 2007. IEEE , vol., no., pp.457-462, 9-12 Sept. 2007
[137]
[138]
[139]
Y.C. Chuang, & Y.L. Ke, A novel high-efficiency battery charger with a buck zero-voltageswitching resonant converter, IEEE Trans. Energy Conversion., Vol 22, No 4, pp. 848854,
2007
[140]
Egan, M.G., OSullivan, D.L., Hayes, J.G., Willers, M.J., Henze, C.P.: Power-factorcorrected single-stage inductive charger for electric vehicle batteries, IEEE Trans. Ind.
Electron., Vol 54, No 2, pp. 12171226, 2007
[141]
With
Buck
H.L. Chan, K.W.E. Cheng, & D. Sutanto, A ZCSZVS bi-directional phase-shifted DCDC
converter with extended load range, IEE Proc. Electronic Power Applications, Vol 150, No
3, pp. 269277, 2003
[143]
B.P. McGrath, D.G. Holmes, P.J. Mc Goldrick, & A.D. McIver, Design of a soft-switched 6kW battery charger for traction applications, IEEE Trans. Power Electronics, Vol 22, No 4,
pp. 11361144, 2007
[144]
R. Ayyanar, & N. Mohan, A novel full-bridge DCDC converter for battery charging using
secondary-side control combines soft switching over the full load range and low magnetics
requirement, IEEE Trans. Industrial Applications, Vol 37, No 2, pp. 559565, 2001
[145]
Z.
directional battery charger for plug-in hybrid electric vehicle application," Energy
Conversion Congress and Exposition, 2009. ECCE 2009. IEEE , vol., no., pp.3930-3936, 2024 Sept. 2009
158
[146]
P.K. Jain, W. Kang; H. Soin, & X. Youhao, "Analysis and design considerations of a load
and line independent zero voltage switching full bridge DC/DC converter topology," Power
Electronics, IEEE Transactions on , vol.17, no.5, pp. 649- 657, Sep 2002
[147]
Pahlevaninezhad, M.; Drobnik, J.; Jain, P.; Bakhshai, A.; , "A Load Adaptive Control
Approach for a Zero Voltage Switching DC/DC Converter Used for Electric Vehicles,"
Industrial Electronics, IEEE Transactions on , Vol. 59, No.2, pp. 920-933, 2012
[148]
H. Al-Atrash, & I. Batarseh, "Digital Controller Design for a Practicing Power Electronics
Engineer," Applied Power Electronics Conference, APEC 2007 - Twenty Second Annual
IEEE , pp.34-41, Feb. 25 2007- March 1 2007
[149]
Y. Panov, and M.M. Jovanovic, "Design and performance evaluation of low-voltage/highcurrent DC/DC on-board modules," IEEE Transactions on Power Electronics, Vol 16, No 1,
pp. 26-33, 2001
159
Appendix A
Simulation Schematics
S1
M1
IRF640
L2
1uH
V1
R5
output
0.01
C2
100u
g1
g2
IRF640
M2
E1
R81
1
12Vdc
+
-
+
-
R6
0.002
S2
R82
0.00001
+
-
+
-
E
E411
+
-
F1
F
+
-
GAIN = 1
R83
100m
E412
E
C80
20p
R86
0.002
C70
R12
23.7k
L5
1.5mH
C81
536p
816
R80
E410
C82
+
-
pwm_out
+
-
Error Amplifier
stp(V(%IN))
0.2uF
L3
50uH
L4
50uH
C83
100nR85
R84
50
2
C84
100n
R87
C85
10
10n
10k
OUT
+
U7
SAW
V308
GAIN = 10
1
R11
244p
1.5Vdc
V2 = 2.5
PER = 2u
TR = 1.98u
Vinject
V407
50
Vsense
160
16 Binary Inverters
U32
1
2
INV
U33
2
INV
U34
2
INV
U35
2
INV
U36
2
INV
U37
2
INV
U38
V408
IN1
IN2
Vsense
OUT
2
INV
U39
1Vdc
U14
Vinject
INV
U40
U13
OFFTIME = .0025uS
DSTM1
ONTIME = .0025uS
CLK
DELAY =
STARTVAL = 0
OPPVAL = 1
IN
CNVRT
STAT
LO
reference
OVR
REF
LGND
V3
2V
DB15
DB14
DB13
DB12
DB11
DB10
DB9
DB8
DB7
DB6
DB5
DB4
DB3
DB2
DB1
DB0
2
INV
U41
2
INV
U42
2
INV
U43
2
INV
U44
16 Bits ADC
2
V
1
V
CNVWHI = 0.01n
CNVWLO = 0.02n
DB15
DB14
DB13
DB12
DB11
DB10
DB9
DB8 OUT
DB7
DB6
DB5
DB4
DB3 REF
DB2
DB1 LGND
DB0
reference
16 Bits DAC
INV
U45
2
INV
U46
2
INV
U47
1
V
2
INV
Title
<Title>
Size
Document Number
Custom<Doc>
Date:
161
R
Sheet
of
Appendix B
Circuit Layout & Selected Components List
`
Fig. B.2 PCB Bottom Layer of the DAEF
162
163
Component
Manufacturer/Distributer
KAD5514P-25
IntersiL
ISL5957-SOIC
IntersiL
TMS320F28335
Texas Instrument
Altera Corporation
Crystal Oscillator
Si590
Silicon Labs
RF Transformer
ADT1-1WT
Mini Circuits
RC-Sensor
0.1F/1K
Digikey
RC-injector
1nF/30
Digikey
Digikey
DC-DC Converter
Freescale
DC-AC Micro-inverter
200W PV Inverter
SPARQ systems
164
Appendix C
Matlab Analysis file
k1=100;
k2=-1;
T=0.5e-8;
% Sensor TF
% Injector TF
numc=[2
0];denc=[T
+8e+8*T
0];%
ZOH
transfer
function,
using
approximation
[num1,den1]=series(k1*k2*numa,dena,numb,denb);
[num2,den2]=series(num1,den1,numc,denc);% open loop TF of DAEF
[num,den]=cloop(num2,den2,-1);
printsys(num,den)
Km=0.4;
% modulator gain
Ka=-25;
% Controller gain
Kb=80;
% Plant DC gain
Lf=500e-6;
Cf=2.2e-6;
Z=0.3;
% Output impedance
165
Pade
[num3,den3]=series(nump,denp,numd,dend);
%TF of Gd(s)*Gp(s)
loop
TF
of
Gp(s)*F(s)and Kd
printsys(num4,den4)
% open loop TF
w=logspace(2,4,600);
[mag,phase,w]=bode(num4,den4,w);
margin(mag,phase,w);figure
% This is the type 3 compensator used for the inverter control system
num_comp=[0.58e-9 4.8e-5 1];
loop
with
compensator printsys(num6,den6)
[num7,den7]=feedback(num6,den6,numk,denk,-1);%closed
loop
compensator
printsys(num7,den7)
w=logspace(2,4,600);
[mag,phase,w]=bode(num6,den6,w);% Open loop Phase & Gain Margin
margin(mag,phase,w);figure
[mag1,phase1,w]=bode(num7,den7,w);% Closed loop PM & GM
margin(mag1,phase1,w);figure
[Gm,Pm,Wcg,Wcp]=margin(mag1,phase1,w);
H=tf(num7,den7);
nyquist(H); % This is the Nyquist plot of the Closed loop TF
title(['Gm=',num2str(Gm),'Pm=',num2str(Pm)]);figure
166
with
step(num7,den7);
End of program
Matlab output file;
(C. 1)
2)
(C. 2)
3)
(C. 3)
4) This is the overall open loop transfer function of the inverter
system;
(C. 4)
(C. 5)
167
Appendix D
MathCAD Analysis File
f1 10010
1 f1
6
f2 30 10
2 f2
6
f 100 200 50 10
( f) 2 f
j
k1 10 20 100
k2 1
1
6
25010
T F k1
j 1 j 2
j 2 1 2 j 1 2
k1 k2 2
T j
1 e
168
15
20log TF100 ( f)
30
45
60
100
1 10
1 10
1 10
f
1 10
1 10
1 10
100
arg TF 100 ( f)
180
100
200
10
100
1 10
1 10
1 10
1 10
1 10
169
1 10
1 10 1.2 10 400010
j
1
6
20010
j T
1e
j
100
150
20log ZOH
200
250
300
6
1 10
1 10
1 10
1 10
170
1 10
10
50
arg ZOH
180
100
150
200
6
1 10
1 10
1 10
1 10
Duty cycle:
D
Vo
Vin
D 0.8
Output current: Io 2 3 20
Switching frequency:
Output filter:
Lo 50010
Zo(Io) 0.3
171
1 10
10
Co 2.2 10
Resonant frequency:
fLC
1
2 LoCo
3
fLC 4.799 10
Frequency response:
6
f 10 50 2 10
( f) 2 f
j
Feedback gain:
KFB
PWM gain:
Vref
Vin
KPWR Vin
KMOD
1
Vs
K LC Io
Filter gain:
1
Lo
2
1 j
j LoCo
Zo( Io)
172
40
4
20log T ( f) 2
32
68
104
140
10
1 10
1 10
100
1 10
1 10
1 10
( f)
45
arg T ( f) 2
180
90
135
180
100
1 10
1 10
1 10
1 10
( f)
2
Fig. D.6 Phase plot of the plant transfer function
173
1 10
1
fs
5
Fc 1 10
180
90
Boost(2) 46.828
Choose Compensator type III when the boost is less than 180 deg.
The K factor:
Boost ( 2)
K tan
45
180
K 2.319
Double-zero location:
Fz
Fc
K
Fz 6.567 10
174
Double-pole location:
4
Fp 1.523 10
Fp Fc K
BW 5.228 10
BW Fp Fc
G
20
R1 10 10
C2
GEA 104.773
1
2 FcGEAR1
11
C2 1.519 10
11
C1 2.003 10
C1 C2(K 1)
R2
R3
C3
K
6
2 Fc C1
R2 1.21 10
R1
K1
R3 7.583 10
1
2 Fc K R3
C3 1.378 10
TEA
175
C1C2
C1 C2
110
Magnitude (dB)
80
50
20
10
1
10
100
1 10
1 10
1 10
1 10
1 10
Frequency (Hz)
Phase (deg)
50
90
130
1
10
100
1 10
1 10
1 10
1 10
Frequency (Hz)
176
1 10
Output current :
Io 90
Ro
Vo
Io
Ro 0.133
Filter Inductor :
Lo 1.6 10
Filter Capacitor:
Co 20010
Input Voltage :
Vin 300
Clock/Sampling frequency:
Fs 20010
1
5
f 10 150 10010
vo ( s)
d ( s)
177
i( s)
d ( s)
Laplace function:
s ( f) j2 f
Vin
Lo
2
1 s ( f)
s ( f) LoCo
Ro
Gv( f)
Gi( f)
Ro s ( f) Lo s ( f) LoCoRo
1
20
178
1 2 f1
2 2 f2
Hsen (f)
Ksense
1 s(f)
ADC Gain:
n
Kadc
12
Vadc
PWM Gain:
Kpwm( f) 1.2
n 10
j 2
Hcomp( f c) 1 2 e
b
f
Fs
j 2 f
1
Fs
1 e
179
10
11
20log Hcomp f 2
13
20log Hcomp f 2
10
20log Hcomp f 2
12
20log Hcomp f 2
20log Hcomp f 2
50
100
10
1 10
1 10
100
1 10
1 10
1 10
1 10
200
11
13
12
10
11
arg Hcomp f 2
arg Hcomp f 2
arg Hcomp f 2
arg Hcomp f 2
arg Hcomp f 2
180
150
180
100
180
180 50
180
50
100
1 10
1 10
1 10
1 10
Fig. D.10 Phase of the digital compensator for different values of b and c
180
1 10
Hinteg( f)
j 2
1e
f
Fs
Hdelay( f) e
f
Fs
ZOH( f)
1e
j 2
f
Fs
f
Fs
1
Fs
Gdaef ( f)
s ( f) 1 s ( f) 2
f
j 2
K1 K2 2
Fs
s ( f) 1 2 s ( f) 1 2
1 e
Open loop gain of the inner current loop can be expressed as:
Ti_comp( f c) Hdelay(f) ZOH ( f) Hsen ( f) Kadc Gi( f) Hcomp ( f c) Hinteg( f)
T v_compf
( c)
Gv( f) Gdaef ( f) Hsen ( f) Kadc Hinteg( f) Hcomp( f c) Hcomp( f c) Hinteg( f) Hdelay( f) ZOH( f) Kpwm( f)
1 Hdelay( f) ZOH( f) Hsen ( f) Kadc Gi( f) Hcomp( f c) Hinteg( f)
T v_unf
( )
150
20010
Magnitude (dB)
100
20log Ti_uncomp( f)
20log Ti_comp( f c)
50
0
0
50
100
150
10
100
1 10
1 10
f
1 10
1 10
Frequency (Hz)
uncompensated system
compensated system
Fig. D.11 Bode plot of the Gain Transfer function of the inner current loop
182
1 10
Magnitude (dB)
270
225
180
135
180 90
arg ( Ti_uncomp( f) )
45
0
180 45
arg ( Ti_comp( f c) )
90
135
180
225
270
10
20010
180
180
100
1 10
1 10
f
1 10
1 10
1 10
Frequency (Hz)
uncompensated
compensated
Fig. D.12 Phase response of the open loop transfer function for the current inner loop
100
610
Magnitude (dB)
65
30
20log Tv_un( f)
20log Tv_comp( f c)
40
75
110
145
180
10
1 10
1 10
100
1 10
1 10
f
Frequency (Hz)
Uncompensated
Compensated
Fig. D.13 Bode plot of the open loop gain transfer function for the outer voltage loop
183
Magnitude (dB)
270
231.43
192.86
154.29
115.71
360
77.14
arg( Tv_un( f) )
38.57
0
360 38.57
arg( Tv_comp( f c) )
77.14
115.71
154.29
192.86
231.43
270
10
610
90
180
100
1 10
1 10
1 10
f
Frequency (Hz)
Uncompensated
Compensated
Fig. D.14 Phase response of the open loop transfer function for the outer voltage loop
184
1 10
Appendix E
DSP Program
/****************************************************************
*****************
// This code is modified to reverse the phase of the acquired
noise signal // coming from the ADC outputs
//
// The task of this program is to run ADC0 in continuous mode.
// The results are used to adjust the duty cycle of PWM1
// The 12 bit digit is out on pins GPIO4-GPIO15
// The Strobe signal is given on GPIO27.
//
//
*****************************************************************
*****************/
#include
<math.h>
#include
"PS_bios.h"
#include "DSP2833x_Device.h"
// DSP2833x Headerfile Include
File
#include "DSP2833x_EPwm_defines.h"
typedef float DefaultType;
#define
GetCurTime() PS_GetSysTimer()
void Task();
DefaultType
fGblV2 = 0.0;
unsigned int a=0;
//-----------------------------------------------void InitEPwm1Gpio(void)
{
EALLOW;
GpioCtrlRegs.GPAPUD.bit.GPIO0 = 0;
// Enable pull-up on
GPIO0 (EPWM1A)
GpioCtrlRegs.GPAPUD.bit.GPIO1 = 0;
// Enable pull-up on
GPIO1 (EPWM1B)
GpioCtrlRegs.GPAMUX1.bit.GPIO0 = 1;
// Configure GPIO0 as
EPWM1A
GpioCtrlRegs.GPAMUX1.bit.GPIO1 = 1;
// Configure GPIO1 as
185
EPWM1B
EDIS;
}
//-----------------------------------------------void InitEPwm1()
{
// EPWM Module 1 config
EPwm1Regs.TBPRD = 0xFFFF;
EPwm1Regs.TBPHS.half.TBPHS = 0; // Set Phase register
EPwm1Regs.TBCTL.bit.CLKDIV = TB_DIV1;
EPwm1Regs.TBCTL.bit.HSPCLKDIV = TB_DIV1;
EPwm1Regs.TBCTL.bit.CTRMODE = TB_COUNT_UP;
EPwm1Regs.TBCTL.bit.PHSEN = TB_DISABLE;
// Master
module
EPwm1Regs.TBCTL.bit.PRDLD = TB_SHADOW;
EPwm1Regs.TBCTL.bit.SYNCOSEL = TB_CTR_CMPB;
// Sync
down-stream module
EPwm1Regs.CMPCTL.bit.SHDWAMODE = CC_SHADOW;
EPwm1Regs.CMPCTL.bit.SHDWBMODE = CC_SHADOW;
EPwm1Regs.CMPCTL.bit.LOADAMODE = CC_CTR_ZERO; // load on
CTR=Zero
EPwm1Regs.CMPCTL.bit.LOADBMODE = CC_CTR_ZERO; // load on
CTR=Zero
EPwm1Regs.AQCTLA.bit.ZRO
EPwm1A
EPwm1Regs.AQCTLA.bit.CAU
EPwm1Regs.AQCTLB.bit.ZRO
EPwm1A
EPwm1Regs.AQCTLB.bit.CAU
GpioDataRegs.GPASET.bit.GPIO27 = 1;
a = AdcRegs.ADCRESULT0;
GpioDataRegs.GPADAT.all =a;
EPwm1Regs.CMPA.half.CMPA = a;
GpioDataRegs.GPASET.bit.GPIO27 = 1;
GpioDataRegs.GPACLEAR.bit.GPIO27 = 1;
186
}
//-----------------------------------------------void Initialize(void)
{
PS_SysInit(30, 10);
PS_InitTimer(0, 0xffffffff);
PS_ResetAdcConvSeq();
PS_SetAdcConvSeq(eAdcCascade, 0, 1.0);
PS_AdcInit(0, !0);
// ----------------------------------EALLOW;
GpioCtrlRegs.GPAPUD.all = 0;
31
GpioCtrlRegs.GPADIR.all = 0xFFFF;
// GPIO0-31 = output
187