TP 22

P.A.
JANSE VAN RENSBURG

Effective Coupling for Power-Line Communications
A
dissertation submitted to the Faculty of Engineering and the Built
Environment, as partial fulfilment of the requirements for the degree
D.Ing (Electrical and Electronic)
University of Johannesburg
Supervisor: Prof. H.C. Ferreira

January 2008
A
Abstract
ABSTRACT
T
HE technique of using a live power cable to simultaneously transport a
communication signal, has been practiced since the early 1900’s. In most cases,
power-line communications has been implemented as a retrofit technology, with its
main benefit being the utilization of a ‘free’ existing network. This driving force of power-line
communications is typical for high-, medium-, and low-voltage distribution networks, as well
as intra-building networks currently targeted for home automation and home networking.
Researchers have thus focused on the optimum use of these existing power-line channels,
often accepting the inherent drawbacks of this hostile communication channel.
Apart from unpredictable noise sources, two main disadvantages of the low-voltage power-
line network as a communication channel, are i) the unknown power cable characteristics and
topology and ii) time-dependent fluctuation of the power-line impedance level as loads are
unpredictable switched into, and out of the network. These two factors have obscured the
requirements for proper coupling and impedance adaptation to the degree that most
researchers and manufacturers have merely accepted this typical ≈ 20-dB coupling loss as one
of the inherited disadvantages of the power-line channel. Most researchers and manufacturers
have thus defaulted to a guessed power-line impedance level, and have used one fixed coupler
winding ratio under all circumstances, regardless of power-line conditions.
This study has shown that proper coupling and impedance adaptation can yield significant
transmission gains even with limited (qualitative) knowledge of a power-line channel and its
topology. After formulating design steps for an impedance-adapting coupler that facilitates bi-
directional transmission, the impact of the fluctuating power-line impedance on coupler
bandwidth was investigated. Next, impedance adaptation strategies were considered and the
tradeoff between series cable requirements and parallel load requirements was explored. A
model of sufficient simplicity was developed to facilitate qualitative description and
classification of power outlets – functioning as communication nodes. Very interesting
simulation results were obtained and these were verified using a laboratory setup of
characterized power cables and calibrated loads. Next, these simulation results were employed
to improve power-line transmission over a live, uncharacterized 220-V residential network by
means of i) classifying typical residential rooms qualitatively in order to choose proper coupler
winding ratios and ii) using an innovative dual coupler for dedicated on-off switching with
harsh loads, thereby mitigating the fluctuating impact of said loads on low-voltage power-line
communications.
A
Abstract
OPSOMMING
Effektiewe Koppeling vir Kraglyn-Kommunikasie
S EDERT die vroeë 1900’s is stroomdraende kragkabels al benut om terselfdertyd

kommunikasieseine te dra. In die meeste gevalle word kraglyn-kommunikasiestelsels
eers later toegevoeg tot bestaande kragstelsels, met die hoof voordeel dat die bestaande
kraglyn-netwerke dus ‘verniet’ benut kan word as kommunikasienetwerke. Hierdie dryfveer
van kraglyn-kommunikasie is tipies vir hoogspanning-, mediumspanning- en laagspannings-
netwerke, sowel as gebou-netwerke wat tans vir gebou-outomatisering beoog word. Navorsers
fokus dus gewoonlik op die optimale benutting van bestaande kraglyn-netwerke, en aanvaar
maar grotendeels die inherente nadele van hierdie moeilike kommunikasiekanaal.
Behalwe vir onvoorspelbare ruisbronne, is die twee groot nadele van die laagspanningsnetwerk
as kommunikasiekanaal i) die onbekendheid van kragkabel-eienskappe en netwerk-topologie
en ii) die tyd-afhanklike wisseling van die kraglyn-impedansievlak soos laste onvoorspelbaar in
en uit die netwerk geskakel word. Hierdie twee faktore het die vereistes vir doeltreffende
koppeling en impedansie-aanpassing tot so ’n mate verberg, dat navorsers en modem
vervaardigers bloot koppelverliese van ongeveer 20 dB aanvaar het as een van die oorgeërfde
nadele van die kraglynkanaal. Meeste navorsers en vervaardigers val dus terug op geskatte
kraglyn-impedansievlakke, en gebruik een vaste koppel-windingsverhouding onder alle
omstandighede, ongeag die werklike kraglyn toestande.
Hierdie studie het getoon dat doeltreffende koppeling en impedansie-aanpassing
indrukwekkende transmissie-verbeteringe kan lewer, selfs met geringe (kwalitatiewe) kennis
van ’n kraglynkanaal en die topologie daarvan. Nadat ontwerpstappe geformuleer was vir ’n
impedansie-aanpassende koppelbaan wat tweerigting-transmissie fasiliteer, was die impak van
die wisselende kraglyn-impedansievlak op koppelbaan-bandwydte ondersoek. Volgende was
strategieë vir impedansie-aanpassing ondersoek, en die teenstrydige behoeftes van serie-kabels
teenoor parallelle laste, is geoptimeer. Hiervoor was ’n model ontwikkel wat eenvoudig genoeg
is om kwalitatiewe beskrywing en klassifikasie van kragpunte (wat as kommunikasie-nodes
dien) moontlik te maak. Baie interessante simulasie-resultate was verkry, en dié was bevestig
deur ’n laboratorium-opstelling van gekarakteriseerde kragkabels en gekalibreerde laste.
Vervolgens was hierdie simulasie-resultate benut om kraglyn-transmissie oor ’n lewendige 220-
V residensiële netwerk te verbeter deur i) die kwalitatiewe klassifikasie van tipiese kamers,
sodat doeltreffende koppel-windingsverhoudings gekies kan word, asook ii) die
implementering van ’n oorspronklike dubbel-impedansie koppelbaan wat outomaties skakel
saam met swaar laste, om sodoende the wisselende impak van dié laste op kraglyn-
kommunikasie teë te werk.
B
Briefing
BRIEFING
C OUPLING a sensitive modem or an expensive measuring instrument to a live power-

line cable, can cause even an electrical engineer to be apprehensive about the task. It
was therefore attempted to write this thesis in such a style that practical,
implementation issues are always explained simply, while emphasizing the importance of the
concept. Deeper theoretical and mathematical analyses that would enhance a reader’s insight
of the topic, are included in some of the chapters, but the outcome of said theory would
typically be discussed, guiding the reader in certain tradeoffs to be made. Unless specifically
indicated otherwise, all quantities in this thesis are in standard SI units.
Chapter 1 is intended to give the reader with little or no background on power-line

communications a quick overview of the history, promising applications as well as difficulties
of this field. Many important references are cited for the reader who plans to investigate some
of these matters further. Next, Chapter 2 is aimed to outline some of the most common high-
frequency and other parasitic side-effects of passive components, as these are used to
construct coupling circuits as well as modems.
In Chapter 3, the concept of filtering and coupling is carefully introduced, whereafter the
operation of a typical transformer-capacitor coupling circuit is investigated, emphasizing the
filter mechanisms of the coupler. Chapter 4 shows that an impedance-adapting coupler can be
used for bi-directional communication, and further shows the reader practical design steps for
such a bi-directional coupler. As the internal parasitic impedances of a coupling transformer
determine the filter characteristics of the coupler, and also impact on core saturation, these are
given special consideration in Chapter 5. Practical tuning of magnetizing and leakage
inductances is also discussed in this chapter.
Impedance adaptation, by adjusting winding ratios of the coupling transformer, is investigated

in Chapter 6. The trade-off between series impedances (cables) and parallel impedances
(loads) is especially considered. Simulation details are given, for the reader who would like to
implement a similar simulation on a spreadsheet, or develop the model further. In Chapter 7,
bi-directional couplers (designed according to Chapter 4) are used for impedance adaptation,
both in the laboratory as well as in a live 220-V residential power-line network. Finally,
Chapter 7 describes a novel dual-impedance coupler, showing the reader how to design such a
coupler using commercially available coupling transformers. Results confirm that the
suggested dual coupler can be used to improve transmission and reception from switching
loads, as well as reduce network impedance fluctuation.
C
Content
CONTENT
P. Publications i
F. Figures iii
T. Tables viii
1. PLC History, Applications and Hindrances 1
1.1 PLC History and Applications 2
1.2 Inherited Hindrances 5
1.3 Regulations 8
1.4 Conclusion 10
1.5 References 11
2. Component Selection and Circuit Layout 15
2.1 Introduction 16
2.2 Conductors and Circuit Layout 17
2.3 Resistors 19
2.4 Capacitors 21
2.5 Inductors 23
2.6 Transformers 25
2.7 Component Integration 27
2.8 Conclusion 28
2.9 References 29
3. Coupling Fundamentals 31
3.1 Filtering Concepts 32
3.2 Transformer Couplers 35
3.3 Conclusion 40
3.4 References 41
4. Design of a Bi-Directional Coupler 42
4.1 Introduction 43
4.2 Circuit 43
4.3 Design 45
4.4 Bi-Directional Measurements 48
C
Content
CONTENT…
4. Design of a Bi-Directional Coupler…
4.5 Influence of Fluctuating Power-Line Impedance 50
4.6 Conclusion 52
4.7 References 52
5. Coupling Transformer Impedances 53
5.1 Introduction 54
5.2 Summarized Theory of Inductance 54
5.3 Leakage Inductance Considerations 57
5.4 Magnetizing Inductance Considerations 61
5.5 Conclusion 64
5.6 References 64
6. Impedance Adaptation 65
6.1 Introduction 66
6.2 Model for Transmitter-Receiver Chain 67
6.3 Simulation Details 69
6.4 Simulation Results 70
6.5 Recommendations 76
6.6 References 78
7. Experimental Verification 79
7.1 Introduction 80
7.2 Experimental Setup 84
7.3 Laboratory Verification of Simulation Results 88
7.4 Classifying Outlets for Impedance Adaptation 92
7.5 Dual Impedance-Adapting Coupler 95
7.6 Conclusion 101
7.7 References 102
8. Conclusion 103
8.1 Summary and Core Results 104
8.2 Further Research Possibilities 109
8.3 Closure 110
8.4 References 111
P
Publications
PUBLICATIONS
Overlapping
1. Peer-reviewed conference papers directly related to this thesis
Chapter
[1] P. A. Janse van Rensburg, H. C. Ferreira, “Practical aspects of component 2
selection and circuit layout for modem and coupling circuitry,” Proceedings of
the 7th International Symposium on Power-Line Communications and its Applications
(ISPLC’03), Kyoto, Japan, 26-28 March 2003, pp. 197-203.
[2] P. A. Janse van Rensburg, H. C. Ferreira, “Coupling circuitry: understanding 3

the functions of different components,” Proceedings of the 7th International
Symposium on Power-Line Communications and its Applications (ISPLC’03), Kyoto,
Japan, 26-28 March 2003, pp. 204-209.
[3] P. A. Janse van Rensburg, H. C. Ferreira, “Step-by-step design of a coupling 4

circuit with bi-directional transmission capabilities,” Proceedings of the 8th
International Symposium on Power-Line Communications and its Applications
(ISPLC’04), Zaragoza, Spain, 31 March - 2 April 2004, pp. 238-243.
[4] P. A. Janse van Rensburg, H. C. Ferreira, “The role of magnetizing and 5

leakage inductance in transformer coupling circuitry,” Proceedings of the 8th
[5] P. A. Janse van Rensburg, H. C. Ferreira, A. J. Snyders, “Coupler winding 6

ratio selection for effective power transfer to a power-line communications
receiver,” Proceedings of the 10th International Symposium on Power-Line
Communications and its Applications (ISPLC’06), Orlando, USA, 26-29 March
2006, pp. 290-295.
[6] P. A. Janse van Rensburg, H. C. Ferreira, “Classifying typical residential rooms 7

for narrow-band impedance adaptation,” Proceedings of the 11th International
Symposium on Power-Line Communications and its Applications (ISPLC’07), Pisa,
Italy, 26-28 March 2007, pp. 47-52.
[7] P. A. Janse van Rensburg, H. C. Ferreira, “Dual Coupler for Dedicated 7

Switching with Loads,” Proceedings of the 11th International Symposium on Power-
Line Communications and its Applications (ISPLC’07), Pisa, Italy, 26-28 March
2007, pp. 53-58.
i
P
Publications
PUBLICATIONS…
Overlapping
2. Peer-reviewed journal papers directly related to this thesis
Chapter
[8] P. A. Janse van Rensburg, H. C. Ferreira, “Design of a bidirectional 4, 5
impedance-adapting transformer coupling circuit for low-voltage power-line
communications,” IEEE Transactions on Power Delivery, vol. 20(1), January
2005, pp. 64-70.
[9] P. A. Janse van Rensburg, H. C. Ferreira, “Coupler winding ratio selection for 6, 7
effective narrow-band power-line communications,” IEEE Transactions on
Power Delivery, vol. 23(1), January 2008, pp.140-149.
Overlapping
3. Peer-reviewed conference papers indirectly related to this thesis
Chapter
[10] A. J. Snyders, H. C. Ferreira, A. W. Ballot, P. A. Janse van Rensburg, “The 4, 5
impact of coupling/filtering on orthogonal M-FSK,” Proceedings of the 9th
(ISPLC’05), Vancouver, Canada, 6-8 April 2005, pp. 206-209.
[11] A. J. Snyders, H. C. Ferreira, P. A. Janse van Rensburg, “Carrier gain 4, 5

adjustment for improved power-line signal-to-noise ratios,” Proceedings of the
10th International Symposium on Power-Line Communications and its Applications
(ISPLC’06), Orlando, USA, 26-29 March 2006, pp. 39-43.
ii
F
Figures
FIGURES
Chapter 1 Page
Fig. 1.1 (a) High-voltage transmission line with simple topology and (b) a peek into 3
the complexity of the low-voltage network.
Fig. 1.2 (a) Global growth of the internet compared to fixed-line and mobile 4
phones and (b) penetration of the internet per region, showing the ‘digital
divide’ between Africa and other developed areas.
Fig. 1.3 Simplified representation of the cross-sectional area of a conductor that is 5
utilized by a certain (a) dc current, (b) low-frequency ac current, and (c)
high-frequency ac current.
Fig. 1.4 Estimated radiated emission for a 100-kHz PLC system compared to a 10- 7
MHz PLC system at 1-mA current levels.
Fig. 1.5 Summary of maximum radiated emissions allowed. 9
Chapter 2 Page
Fig. 2.1 (a) General equivalent parallel-resonant circuit for resistors, and (b) typical 20
⏐ZTOTAL⏐ vs. logarithmic frequency plot.
Fig. 2.2 (a) General equivalent RC circuit for carbon-composite and metal-film 21
resistors, and (b) typical normalized ⏐ZTOTAL⏐ vs. logarithmic frequency
plot.
Fig. 2.3 (a) General equivalent circuit for capacitors, and (b) typical ⏐ZTOTAL⏐ vs. 22
logarithmic frequency plot.
Fig. 2.4 (a) General parallel-resonant equivalent circuit for inductors, and (b) typical 23
⏐ZTOTAL⏐ vs. logarithmic frequency plot.
Fig. 2.5 (a) Comprehensive lumped equivalent circuit for transformers, and (b) 26
typical ⏐ZP⏐ (no-load impedance seen from primary) vs. logarithmic
frequency plot.
Chapter 3 Page
Fig. 3.1 HF-LF model for a (a) 1-µF capacitor and (b) 1-µH inductor in a 50-Hz 33
sinusoidal power and 500-kHz sinusoidal carrier scheme.
Fig. 3.2 The HF-LF model for a 1-µF capacitor terminating into a 50-Ω load 34
resistor (50-Hz sinusoidal power and 500-kHz sinusoidal carrier scheme).
iii
F
Figures
FIGURES…
Chapter 3… Page
Fig. 3.3 Measured Bode plots for a 1-µF capacitor terminating into a 50-Ω load 34
resistor: (a) voltage transfer function and (b) phase transfer function.
Fig. 3.4 (a) Simplified linear model for a transformer and (b) practical non-linear B- 35
H curves.
Fig. 3.5 Suggested coupling circuit. 37
Fig. 3.6 Measured amplitude response of (a) 1:1 transformer only and (b) 1:1 38
transformer with 0.22-µF capacitor in series with primary terminal.
Fig. 3.7 Measured amplitude response of a 1:1 transformer with 0.22-µF capacitor 39
in series with primary terminal but with 25-Ω secondary termination.
Fig. 3.8 (a) Measured amplitude response of a 1:1 transformer with 0.22-µF 40
capacitor in series with primary terminal, total series inductance of 31 µH
and terminating in 25-Ω load. b) Measured amplitude response of a 1:1
transformer with 1-µF capacitor in series with primary terminal, total series
inductance of 69 µH and terminating in 50-Ω load.
Chapter 4 Page
Fig. 4.1 (a) Suggested coupling circuit and (b) equivalent circuit for frequencies 44
close to the resonant point.
Fig. 4.2 Photographs of (a) the coupling transformer designed in this section and 45
(b) a coupling transformer designed for a 250-kHz-to-500-kHz carrier
frequency.
Fig. 4.3 Measured amplitude response of the coupling circuit designed for 49
CENELEC B, C, and D band when (a) receiving a signal from the power
line side and (b) transmitting a signal into a 1-Ω resistor.
Fig. 4.4 Measured amplitude response of the coupling circuit designed for 250 kHz 50
to 500 kHz when (a) receiving a signal from the power-line side and (b)
transmitting a signal to the power-line side (1-Ω resistor).
Fig. 4.5 Measured amplitude response for a 1:1, 80-kHz coupling circuit with (a) 51
50-Ω terminating impedance and (b) 25-Ω terminating impedance.
iv
F
Figures
FIGURES…
Chapter 5 Page
Fig. 5.1 Plot of λ vs. i to show non-linearity of inductive behavior. 55
Fig. 5.2 Simplified representation of a transformer’s flux paths and MMF 56
distribution across the core window.
Fig. 5.3 General transformer equivalent circuit, assuming linear core material, and 57
losses modeled with resistors.
Fig. 5.4 Cross-section through a planar transformer and MMF distribution across 58
window.
Fig. 5.5 Cross-section through transformer (top view) to indicate winding 60
protruding from core.
Fig. 5.6 (a) More comprehensive model for the coupling circuit designed in 61
Chapter 4 and (b) simulated amplitude response for this coupling circuit.
Fig. 5.7 Measured response of the coupling circuit designed in Chapter 4. 63
Chapter 6 Page
Fig. 6.1 Transmitter-receiver pair coupled through the power-line channel. 67
Fig. 6.2 Simulated circuit with all components referred to the power-line side. 69
Fig. 6.3 Thévenin equivalent circuit for Fig. 6.2. 70
Fig. 6.4 Performance (SRX vs. ZS) for certain winding ratios where (a) ZII = 0.4 Ω 71
(more typical of CENELEC frequencies) and (b) ZII = 25 Ω (more typical
of higher frequencies).
Fig. 6.5 (a) Optimum receiver winding ratio aRX and (b) corresponding 72
performance for certain fixed transmitter winding ratios with ZII = 0.4 Ω.
Fig. 6.6 (a) Optimum transmitter winding ratio aTX and (b) corresponding 73
performance for certain fixed receiver winding ratios with ZII = 25 Ω.
Fig. 6.7 Optimum symmetrical winding ratio, aTX = aRX = a for a symmetrical 74
power-line environment.
Fig. 6.8 Optimum asymmetrical winding ratio, aTX for asymmetrical power-line 75
channel with transmitter far from the load (ZL : ZR = 10:1).
Fig. 6.9 Optimum asymmetrical winding ratio, aTX for asymmetrical power-line 76
channel with transmitter close to the load (ZL : ZR = 1:10).
v
F
Figures
FIGURES…
Chapter 7 Page
Fig. 7.1 Transmitter-receiver pair coupled through the power-line T-model of 80
Chapter 6.
Fig. 7.2 Performance (SRX vs. ZS) for certain winding ratios where (a) ZII = 0.4 Ω 82
(severely loaded) and (b) ZII = 25 Ω (lightly loaded).
Fig. 7.3 Top view of Fig. 6.7 – a three-dimensional solution of (7.1) for a 83
symmetrical power-line topology.
Fig. 7.4 Three pairs of couplers built into 3-prong plugs, typical for South Africa 84
and United Kingdom.
Fig. 7.5 Utilizing external 50-Ω terminations in parallel with an instrument input 85
impedance to force the ‘receiver’ input impedance down to a wanted
‘power-line’ impedance level, in this case 12.5 Ω.
Fig. 7.6 Influence of terminating load on measured amplitude response of the 86
selected 1:4 transformer.
Fig. 7.7 Influence of coupling capacitor on measured amplitude response of the 87
selected 1:4 transformer, terminating into a worst-case 3-Ω load.
Fig. 7.8 Emulated 50-Ω transmitter (Goldstar FG2002C function generator) and 88
50-Ω receiver (IFR 2397 spectrum analyzer).
Fig. 7.9 (a) Extension cord and (b) calibrated 50-Ω loads in parallel, with 1:1 89
coupler.
Fig. 7.10 Performance of the three coupling pairs between 50 kHz and 550 kHz in 90
a disconnected laboratory setup.
Fig. 7.11 Measured performance of the three coupling pairs for power-line 90
conditions C and D.
Fig. 7.12 Measured performance for lightly-loaded power-line conditions E and F. 91
Fig. 7.13 Performance of the coupling pairs for power-line conditions G and H. 92
Fig. 7.14 Live 220-V transmission measurements from a residential lounge to 94
kitchen.
Fig. 7.15 Live transmission measurements between a kitchen and bedroom. 94
Fig. 7.16 Live transmission measurements from the border of a residence (gate) to 95
the laundry.
vi
F
Figures
FIGURES…
Chapter 7… Page
Fig. 7.17 Proposed dual coupler, incorporating a 4:1 winding ratio for permanent 96
connection between modem and main load Z1 as well as a 2:1 winding
ratio between modem and power line.
Fig. 7.18 Photo of the 4:1 vs. 2:1 dual coupler prototype. 97
Fig. 7.19 Laboratory measurements showing transmission fluctuation for different 98
coupler setups when switching a 3-Ω load.
Fig. 7.20 Live transmission measurements from the border of a residence (gate) to 99
the laundry, as influenced by the receiver-coupler.
Fig. 7.21 Dual coupler connected to the stove supply, with 4:1 portion connected 100
across oven (bake and grill) elements.
Fig. 7.22 Live transmission measurements between a kitchen outlet and stove. 100
Chapter 8 Page
None. –
vii
T
Tables
TABLES
Chapters 1 - 5 Page
None. –
Chapter 6 Page
Table 6.1 Classification of ZS and ZII Values. 68
Chapter 7 Page
Table 7.1 Classification of ZS and ZII Values. 81
Table 7.2 Winding Ratio Choice. 83
Table 7.3 Approximate |ZS| of Cable Pairs. 89
Table 7.4 Room Classification. 93
Chapter 8 Page
None. –
viii
1
Chapter
CHAPTER 1
PLC History, Applications and Hindrances
A
LTHOUGH the superimposing of a communication signal onto a power waveform
has been practiced for over a century, power-line communications only recently (in
the 1990’s) started gaining momentum. Traditional applications were limited to voice
and control signals over high-voltage and medium-voltage networks, as well as peak-demand
management (load shedding) and automatic meter reading over low-voltage networks.
However, the advances in semiconductor theory and production in the last half-century and
associated growth in the consumer electronics market, facilitated the computer- and internet
boom. Subsequently this has set the stage for home networks and home automation – ideal
candidates for power-line communications.
The utilization of the power-line channel for communications – an application it was never
designed for – makes it an interesting and challenging field. At low frequencies, closer to the
frequency the network was designed for, power-line cabling is not as hostile towards the
communication signal. Unfortunately, bandwidth is directly limited by frequency, and only
limited data rates can be achieved with low-frequency power-line communications. As the
communication frequency is pushed up to increase bandwidth, so do the complications
increase. Instead of a conducting copper network, the power-line channel acts more and more
like a complex antenna as frequency rises. Both from an engineering efficiency viewpoint, as
well as environmental and interference perspectives, this potential radiated emission of the
power-line communication signal has become an important consideration.
1
PLC HISTORY, APPLICATIONS AND HINDRANCES
1
1.1 PLC History and Applications
The concept of superimposing an information-carrying signal onto a power

waveform has been practiced for over a century. Currently, the term power-line
communications or PLC is commonly accepted as the generic name of this field of
Electrical Engineering.
T
HE first known patents for power-line communications (PLC) systems were registered in
Britain and Germany in 1897 and 1901 respectively. It is interesting that the British
patent was awarded to Routin and Brown from Switzerland [1], while the German
patent was registered by a French engineer, Loubery [2]. Therefore PLC technologies have
been available and have been utilized for over a century. Unbeknown to the general public,
utility companies have utilized PLC for many decades to implement their own telephone
systems over their distribution lines since the 1920’s [3], [4]. Another widespread use of this
technology behind the scenes, has been load shedding, e.g. switching off hot-water geysers
from a central monitoring location during peak hours [5]-[7]. However, only recently have
consumers become aware of the promises of this ‘free’ power-line network when American
and European countries started to allow production and sales of power-line modems for
home networks and home automation [8]-[10].
In his 1901 patent, Loubery suggested a multi-tone signaling scheme, i.e. different tones
(frequencies) represent different messages – similar to the more complex DTMF (dual-tone
multi-frequency) technology [11] that is well known today. Further, the patent suggests dedicated
tuned filters to receive and interpret the message – one filter for each tone. In hindsight, it is
now clear that this simple system can only communicate very crudely with the receiver – but it
remains impressive that the fundamental concept of 1901 is still used in the current modern
high-speed PLC technologies [10]. The OFDM (orthogonal frequency division multiplexing) [12] and
all FSK (frequency shift-keying) [13] techniques “pack the eggs in many baskets” by making use of
multiple frequencies or frequency bands to transmit data.
Later, in the 1920’s, transmission of voice over HV (high-voltage) lines became very popular, as
these power-line networks were widespread long before dedicated telephone networks were
established, and penetrated deeply into rural areas [3]. The utility companies themselves
urgently required bidirectional communication for the management of their power stations,
substations, distribution lines, as well as synchronization and energy distribution. In the
following decades, it became clear that HV and MV (medium-voltage) power-lines are reliable
candidates for data transmission, mainly because the topology of these lines are simple,
yielding a predictable transfer function. See Fig. 1(a). However, transformers spoil the
simplicity of these combined HV and MV networks and typically need to be bypassed for a
proper link to other lines [14].
2
(a) (b)
Fig. 1.1 (a) High-voltage transmission line with simple topology and (b) a peek
into the complexity of the low-voltage network. (From http://images.google.com July
2007)
Currently, in third-world countries where telephone networks are still restricted, these power-
line networks are widespread, and can facilitate the desperate need for communication – see
Fig. 1.2(b). It is hoped that internet access over power-lines will soon facilitate telephony,
education, banking and even remote medical consultations in these third-world countries.
Therefore some African and South American researchers have enthusiastically pursued this
PLC field [13], [15]-[17].
As the LV (low-voltage) power-line topology makes reliable communication very difficult, this
branch of PLC has historically lacked behind. Before the 1990’s, the only application that
could be justified, was load control to limit peak energy demand. This technique, used since
the 1930’s was known as ripple control or ripple carrier signaling [3], [18]. Because of the low carrier
frequency (< 3kHz) and the hostile LV topology, transmitter power had to be very high for
successful communication – even in the order of kilowatts [5].
Another informal and relatively unknown step for power-line communications during World
War II was prompted by government restrictions on amateur radio broadcasting. These
practically-minded radio amateurs opted for a different route by transmitting their signals
through power-line cables, thereby limiting possible interference with the military. A chapter
on “Carrier Current Communication” has subsequently been included in some early editions
of the yearly Radio Amateur’s Handbook [19].
The use of PLC technology to implement automatic meter reading (AMR) has been a drawing
card for municipalities ever since the inception of PLC [1], but this application only gained
momentum in the 1970’s [20]-[24]. Apart from data acquisition for the billing of electricity,
water, and gas consumption, researchers foresee even more advanced functions that might be
3
1
implemented in the near future. These include i) warning or even disconnection on non-
payment of bill, ii) broadcasting of simple news messages or disaster-management messages,
iii) load-flow analysis, and iv) even statistical gathering of the power-line network
characteristics, e.g. noise data.
Later, during the 1980’s and 1990’s, research into the use of PLC for home automation [25],
[26] began to accelerate. As the consumer electronics market grew exponentially, so did the
scope for intelligent devices and home networking and automation. Typical functions to be
automatically and/or remotely controlled, include lighting, heating, air conditioning,
multimedia, alarm systems and even cooking. A building might be programmed for instance,
to switch on a minimum number of lights at night, and further to automatically switch lights
on and off as persons enter and exit a room. Electricity expenditure on heating may be
reduced by fine-tuning on and off periods to the household routine. Multimedia applications
might include automatic recording of radio or television programs, and rebroadcasting them
through the home network.
Also, as personal computers (PCs) became more affordable in the 1990’s, local area networks
(LANs) over power-lines started receiving attention [27], [28]. In subsequent years, the world-
wide web or internet became a reality, and a growing number of families in developed countries
started utilizing the internet from their homes for instant access to information, media,
shopping and banking [29]. See Fig. 1.2. The number of personal computers per household
has also risen, and home networking is becoming an important target of the information
technology (IT) sector. Power-line communications is an ideal vehicle for intra-building
computer networks and home automation, and some companies have had low-speed narrow-
band home automation products on the market since the 1990’s e.g. [30], [31].
(a) (b)
Fig. 1.2 (a) Global growth of the internet compared to fixed-line and mobile
phones (in billions) and (b) % internet penetration per region (2004), showing the
‘digital divide’ between Africa and other developed areas. (From
http://www.itu.int/ITU-D/ict/publications/ July 2007)
4
1
As regulations in different countries were relaxed in the 2000’s, wide-band high-speed
technologies became viable [10], [32]-[35]. The HomePlug consortium [36] have specified a
number of standards as the technology developed, including HomePlug 1.0 (supporting 14
Mbps), HomePlug 1.1 (supporting 85 Mbps), and HomePlug AV (supporting 200 Mbps).
Take note that these raw bit rates are at the physical layer, and do not represent the bit-rates
that a user would experience from a software program communicating through the network –
which might be as little as 25 % of the specified raw bit rate. The data throughput further has
to be shared amongst all the devices connected to the PLC network, as is the case with the
bandwidth of any network.
1.2 Inherited Hindrances

Part and parcel of the power-line communications channel, is the fact that neither the physical
cabling nor the topology of the network was originally designed for communication
frequencies. Further, the very appliances serviced by the electricity supply, drain large portions
of the transmitted waveform and feed harmonics and other noise back into the power grid in
an unpredictable manner, thereby polluting the channel. These inherited disadvantages make
power-line communications a challenging field.
1.2.1 Attenuation and radiation
The frequency band(s) within which the system operates also has a direct influence on these
hindrances. For lower modulation frequencies (e.g. 100 kHz), physical attenuation in the
cabling is not as severe as for high frequencies (e.g. 10 MHz) where the skin effect causes a
factor ≈10 smaller effective conductor area (as the frequency is larger by factor 100) [37]. See
Fig. 1.3. for a simplified representation of the skin-effect. Also, self-inductance of a straight
conductor, causes an impedance proportional to frequency [38] – the higher the frequency, the
more a signal is hindered by the cabling. These concepts will be discussed in greater depth in
Chapter 2.
(a) (b) (c)
Fig. 1.3 Simplified representation of the cross-sectional area of a conductor that is

utilized by a certain (a) dc current, (b) low-frequency ac current, and (c) high-
frequency ac current. The higher the frequency, the higher the current density
(and effective resistance).
5
1
Also, power appliances and other loads impact strongly on low-frequency systems, as these
inductive and especially resistive loads conduct low frequencies well, thereby draining the
communication signal [39], [40]. Conversely, high-frequency systems are less affected by L-R
loads, as inductive loads admit very little of the high-frequency current to pass through. Even
resistive loads cause little attenuation compared to the cabling of the power network, as most
resistive loads behave inductively at higher frequencies, except if the self-resonant frequency is
exceeded [37], [38]. Capacitive loads and many protective devices such as varactors and zener
diodes are however detrimental to transmission of high-frequency signals [4].
Network topology fundamentally influences the success of power-line communications [41].

Medium- and high-voltage power-line networks as used by utility companies often show a
point-to-point topology, with a single line (and return path) between the two points [4]. A
simple topology, for instance a single line or ring, can facilitate impedance matching [42]
and/or make signal attenuation more predictable. Therefore such simple-topology systems can
be designed to transmit much higher data rates at lower transmitted power levels than for
complex topologies. As was mentioned in Section 1.1 though, interconnecting transformers
spoil the simplicity of combined HV and MV networks and typically need to be bypassed for
proper transmission to the next section [14].
In residential or low-voltage power-line communications, a complex star or branching

topology is enforced by the physical layout of electrical power cables [4], [5], [41]. Signal
attenuation is different in each branch and multiple mismatched branches cause multiple
reflections of the signal [41], [43]. Due to the nature of the low-voltage (LV) network, LV
power-line communications is subject to this open, unpredictable network. Electricity
consumers and their appliances switch loads in an uncontrolled manner, causing sudden and
harsh impedance fluctuations. Furthermore, the uncertainty of load fluctuations and
associated noise, increases with complexity. These fluctuations make impedance matching
impossible, and only a small portion of the already limited signal is available to the receiver
[44]. All the abovementioned factors hamper the success of low-voltage power-line
communications and forces the designer to communicate at lower data rates and / or higher
power levels than for a simple topology.
Topology also directly impacts on the potential radiated emission of a PLC network.
Common-mode currents that typically leak back to earth through parasitic capacitances, form
large antenna-like loops – radiating much of the common-mode energy. Reference [45] and
[46] describe the radiating mechanisms of differential mode and common-mode currents, and
[47] shows that guided-wave radiation dominates at 100 kHz for distances up to 170 m from
the source, while at 10 MHz the radiated wave dominates even in the near field (< 10-m
distances). A further conclusion made in [47] is that for the same transmitted current, a 10-
MHz PLC system inherently causes between 40 dB and 80 dB more radiated emission than a
100-kHz PLC system. See Fig. 1.4.
6
Fig. 1.4 Estimated radiated emission for a 100-kHz PLC system compared to a 10-
MHz PLC system at 1-mA current levels (from [47]).
Fortunately, the higher a PLC signal’s frequency, the more viable network conditioning
becomes as higher frequencies require smaller (cheaper) reactive components. One kind of
network conditioning, is to ‘block off’ certain branches of the network, preventing the
communication signal from entering these branches, and so only allowing the power
waveform to flow further. Unfortunately, for lower frequencies, this technique requires an
expensive series inductor capable of carrying the load current [48]-[50]. For high-frequency
PLC systems though, even ferrite cores clamped around the appropriate conductor, may boost
the self-inductance of said cable to sufficiently impede the PLC signal. Thus topology is
improved and the unnecessary dissipation of the communication signal in the down-stream
low-impedance loads is prevented to a certain degree. Furthermore, high-frequency noise
signals that originate at the loads are prevented from entering the conditioned network. If the
cost warrants the improvements in performance, this technique can therefore be used to
design a more favorable channel both in terms of noise and topology.
1.2.2 Noise
Apart from the unpredictable time-variant transfer function caused by all the influences
mentioned above, noise is the other major obstacle to successful power-line communications.
Noise originating from the host of loads as well as their switching, is unfortunately conducted
through the copper network, polluting the power-line communications channel. Once again,
fewer sources of noise impact on HV and MV networks than on LV networks. High-voltage
and medium-voltage power-line noise sources include lightning strikes and other atmospheric
discharges, low-level corona discharges as well as circuit-breaker transients [16]. Capacitor
banks being switched for power factor correction on MV lines cause further noise peaks.
7
1
Although some of the HV and MV noise may appear on the LV network, the interconnecting
transformers do filter these to a large degree. Apart from radiated interference originating
outside a building e.g. from lightning and vehicles driving past, noise on the LV network is
largely due to the switching and operation of intra-building appliances. Again, the fact that the
LV network is an open, uncontrolled network, allows this chaotic superimposing of various
disturbances onto the network.
Power waveform disturbances do impact on PLC, but these are more of a concern to power
utility companies and typically cannot be controlled or improved by a residential end-user.
Waveform disturbances may be classified as over-voltages, under-voltages, frequency
variations, harmonic distortions and even outages [51]. An outage would not interrupt a local
PLC network, however the PLC modems do require power to operate, and thus the local PLC
network would also be off, unless power is sourced from an uninterruptible power supply (UPS).
Superimposed disturbances are normally caused by the switching and operation of intra-
building appliances as was mentioned above. In [52], this kind of noise is classified into four
categories, namely i) noise synchronous to the power waveform, ii) periodic but asynchronous
noise, iii) smooth-spectrum noise and iv) single-event (a-periodic) impulse noise. See [4], [5] or
[16] for a discussion of these noise classes. The reason for classifying noise is to develop
strategies to mitigate the impact of said noise on data transfer. Periodic disturbances as
classified under i) and ii) are easier to model and alleviate than unpredictable disturbances such
as those under iv). Therefore the time-variant stochastic study of a-periodic noise has been the
bottleneck when modeling the noise characteristics of the power-line communications
channel. As these fluctuations occur randomly, researchers have tried to describe and predict
said noise patterns statistically [44], [53]-[58].
1.3 Regulations
When the first analog telephone call was made by Alexander Bell, no telecommunications
regulation existed as nobody knew or could foresee that such an invention would become a
large-scale reality [59]. Similarly, no particular standard existed when power-line
communications was first developed. But as the promise of large-scale economic
implementation grew, so did the need for governments to regulate the deployment and
operation thereof. To date, unfortunately no finality has been reached by major international
standard-setting stakeholders [60].
Background on the different stakeholders and their interdependency can be found in [61]. As
with many other standards, it seems as if the USA is content with having a separate standard –
probably motivated by the fact that the USA market is large enough to develop and sustain
the technology. In the European Union however, many committees and regulatory bodies
have to find common ground [61]. Fig. 1.5 illustrates some of the current differences between
EU regulators. Technical issues to be discussed and specified, include i) frequency bands, ii)
emission limits and background noise (both impacting on signal-to-noise ratios), iii)
8
1
measurement standards including coupling, and even iv) maximum modem output power
levels [61], [62].
Fig. 1.5 Summary of maximum radiated emissions allowed, from [63]. The (Pk)
label indicates peak detection, while (QPk) indicates quasi-peak detection. Take
note that different regulations specify measurements at different distances, and
formulas were used to normalize these to a 3-m distance.
In the quest for simple and fair regulations, an important separation of frequency bands has to
be made as was mentioned in Section 1.2. PLC systems operating under 1 MHz (strictly < 525
kHz) should be referred to as low-frequency or LF-PLC, while those above 1 MHz (strictly > 1.6
MHz) should be referred to as high-frequency or HF-PLC [45]. It is well known and documented
that LF-PLC systems can be modeled and regulated using simple electrical circuit theory and
instrumentation. However, HF-PLC systems, especially above 10 MHz, need to be
approached using electromagnetic theory. In essence, each PLC system can be viewed as a
complicated, unpredictable antenna. Antenna theory is an already complex field and a design is
typically limited to a tuned frequency. Therefore the accurate wideband antenna-modeling of a
complex time-variant PLC system is an almost impossible task. It is therefore the writer’s
opinion that HF-PLC modeling as well as regulation standards need to be based on
electromagnetic theory and measurements further linked to statistics.
For a thorough discussion of various regulators as well as details pertaining to standards, the
reader is referred to [62] where the maximum allowed field strength, and distance at which
said field strength is to be measured, are also discussed. Consensus has also not been reached
on the measurement strategy to obtain the average field strength [45]. Some countries use a
peak detector (e.g. UK), while others employ an average detector (e.g. USA), the latter
9
1
detector typically yielding ≈ 10 dB lower readings. Furthermore, the time duration of this
measurement has not been agreed upon and it is clear that a quasi-peak detector with specified
measurement duration would be a fair solution [45]. Fig. 1.5 from [63] highlights the
inconsistency between different maximum field strengths, by normalizing the measurement
distances to 3 meters, using formulas and other assumptions. Also see [45] for USA FCC part
15 limits, German NB30 limits and British MPT1570 limits.
The PLC-regulation disparity between different countries has an even deeper root, at the
judicial level. Some regulators do not require a-priori or pre-approval of a new technology, and
these regulators would only act upon complaints of interference [59]. Such a strategy is almost
unthinkable, and opens the door for large-scale uncontrolled deployment of culpable devices.
Followed by legitimate or even nasty, unfounded complaints by the public or competitors, this
would result in unnecessary confusion, impacting negatively on the already cautious consumer
market [62].
Although each country has the right to determine its own telecommunication regulations,
history has proven that the more countries subscribe to a certain standard, the more successful
its economic implementation will be [62]. A fragmented body of PLC regulators will further
hamper the interoperability of equipment and cause dissatisfaction amongst consumers. Many
other competing technologies such as ISDN (integrated services digital network), DSL (digital
subscriber line), Cable-TV and Wireless networks have been available for years [59], [62] – and as
with all consumer electronics products, these technologies are offered to consumers at ever-
decreasing installation and subscription costs. If the power-line communications opportunity
is to be seized, fair and practical international regulations would certainly have to be finalized
soon.
1.4 Conclusion
T
HIS chapter gave the reader a quick overview of the historical development of the
power-line communications field and mentioned many of the exciting applications
that have been implemented in the past, as well as some that may become viable
soon. The disadvantages of the power-line channel as a retrofit network were also described.
Most of these hindrances are high-frequency side-effects, as the frequency of a typical power-
line communications signal is between 3 and 6 orders higher than the frequency of its power
waveform counterpart – for which the power-line network was designed.
Further, the reader was given background on the all-important topic of regulations pertaining
to power-line communications. As an introductory chapter, important references are provided
throughout the text, for the reader who intends to investigate these issues further. The website
referenced in [64] is a very useful one, as it allows free downloads of all ISPLC (International
Symposium on Power-Line Communications) conference papers that appeared since the inception of
the symposium in 1997, until 2004. However, since 2005, the yearly ISPLC conference papers
have been archived by the IEEE [65].
10
1
1.5 References
[1] J. Routin, C. E. L. Brown, “Power line signaling electricity meters,” London: British Patent 24833, 1897.
[2] C. R. Loubery, “Einrichtung zur elektrischen Zeichengebung an die Theilnehmer eines Starkstromnetzes,”
Berlin: German Patent 118717, March 15, 1901.
[3] H.-K Podszeck, Carrier Communication over Power Lines, 4th Edition, New York: Springer-Verlag, 1972.
[4] K. Dostert, Powerline Communications, Englewood Cliffs, NJ: Prentice-Hall, 2001.
[5] O. Hooijen, Aspects of Residential Power Line Communications, Aachen: Shaker Verlag, 1998.
[6] A. J. Baggot, B. E. Eyre, G. Fielding, F. M. Gray, “Use of London’s electricity supply system for centralized
control,” Proceedings of the IEE, vol. 125(4), April 1978, pp. 311-327.
[7] A. Forrest, F. M. Gray, “Maximum demand limitation by the automatic and remote control of nonessential
loads using the supply network as the communications medium,” Proceedings of the 3rd International Conference on
Metering Apparatus and Tariffs for Electricity Supply, (MATES ’77), London, UK, 15-17 November 1977, pp.
112-116.
[8] D. Radford, “Spread-spectrum data leap through ac power wiring,” IEEE Spectrum, Nov. 1996, pp. 48-53.
[9] P. A. Brown, ‘Power line communications - past, present, and future,” Proceedings of the 3rd International
Symposium on Power-Line Communications and its Applications (ISPLC’99), Lancaster, UK, 30 March - 1 April
1999, pp. 1-8.
[10] K. H. Afkhamie, S. Katar, L. Yonge, and R. Newman, “An Overview of the upcoming HomePlug AV
Standard,” Proceedings of the 9th International Symposium on Power-Line Communications and its Applications
[11] M. J. Park, S. J. Lee, D. H. Yoon, “Signal detection and analysis of DTMF receiver with quick Fourier
transform,” Proceedings of the 30th Annual Conference of the IEEE Industrial Electronics Society (IECON’04), Busan,
Korea, 2-6 November 2004, vol. 3, pp. 2058- 2064.
[12] V. Chakravarthy, A. S. Nunez, J. P. Stephens, A. K. Shaw, M. A. Temple, “TDCS, OFDM, and MC-CDMA:
a brief tutorial,” IEEE Communications Magazine, vol. 43(9), September 2005, pp. s11-s16.
[13] A. J. Snyders, H. C. Ferreira, P. A. Janse van Rensburg, “Carrier gain adjustment for improved power-line
signal-to-noise ratios,” Proceedings of the 10th International Symposium on Power-Line Communications and its
Applications (ISPLC’06), Orlando, USA, 26-29 March 2006, pp. 39-43.
[14] IEEE Guide for Power-Line Carrier Applications, IEEE Standard 643-1980.
[15] J. Anatory, N. Theethayi, M. Kissaka, N. Mvungi, “Broadband power line communications: the factors
influencing wave propagations in the medium voltage lines,” Proceedings of the 11th International Symposium on
Power-Line Communications and its Applications (ISPLC’07), Pisa, Italy, 26-28 March 2007, pp. 127-132.
[16] H. C. Ferreira, H. M. Grové, O. Hooijen, and A. J. H. Vink, Power Line Communication (in Wiley
Encyclopaedia of Electrical and Electronics Engineering), New York: John Wiley & Sons, 1999, pp. 706-
716.
[17] H. C. Ferreira, “PLC issues in emerging countries: technical and economical factors for the developments of
potential PLC markets” ISPLC 07 Panel Session II in Proceedings of the 11th International Symposium on Power-
Line Communications and its Applications (ISPLC’07), Pisa, Italy, 26-28 March 2007.
[18] J. Newel, “History and theory of ripple communication systems,” Conference Record of the 1976 National
Telecommunications Conference (NTC’76), Dallas, USA, 29 November - 1 December 1976, pp. 2.3.1-2.3.6.
[19] A. F. Collins, The Radio Amateur’s Hand Book, LaVergne: Lightning Source Inc., 2004.
[20] G. Lokken, N. Jagoda, R. J. D’Autenil, “The proposed Winconsin Electrical Power Company load
management system using power line carrier over distribution lines,” Conference Record of the 1976 National
Telecommunications Conference (NTC’76), Dallas, USA, 29 November - 1 December 1976, pp. 2.2.1-2.2.3.
11
1
[21] G. Kaplan, “Two-way communication for load management,” IEEE Spectrum, vol. 14(8), August 1977, pp.
47-50.
[22] B. E. Eyre, “Results of a comprehensive field trial of a United Kingdom customer telemetry system using
mainsborne signaling,” Proceedings of the 5th International Conference on Metering Apparatus and Tariffs for Electricity
Supply (MATES ’87), London, UK, 13-16 April 1987, pp. 252-256.
[23] M. C. King, “Experimental systems for tele-reading over the low-voltage network,” Proceedings of the 6th
International Conference on Metering Apparatus and Tariffs for Electricity Supply, (MATES ’90), London, UK, 3-5
April 1990, pp. 154-157.
[24] C. Nunn, P. M. Moore, P. N. Williams, “Remote meter reading and control using high-performance PLC
communications over the low voltage and medium voltage distribution networks,” Proceedings of the 7th
International Conference on Metering Apparatus and Tariffs for Electricity Supply, (MATES ’92), London, UK, 17-19
November 1992, pp. 304-308.
[25] W. Downey, “Central control and monitoring in commercial buildings using power line communications,”
Proceedings of the 1st International Symposium on Power-Line Communications and its Applications (ISPLC’97), Essen,
Germany, 2-4 April 1997, pp.115-119.
[26] N.-H. Ahn, T.-G. Chang, H. Kim, “A systematic method of probing channel characteristics of home power
line communication network,” Digest of Technical Papers: 21st International Conference on Consumer Electronics
(ICCE’02), Atlanta, USA, 18-20 June 2002, pp. 378-379.
[27] J. O. Onunga, R. W. Donaldson, “Personal computer communications on intrabuilding power line LAN’s
using CSMA with priority acknowledgements,” IEEE Journal on Selected Areas of Communication, vol. 7(2),
February 1989, pp. 180-191.
[28] D. Tuite, “Power line spread-spectrum saves copper in LAN’s and control systems,” Computer Design,
November 1992, pp. 50-53.
[29] R. Dettmer, “The net effect,” IEE Review, March 1995, pp. 67-71.
[30] http://www.echelon.com (July 2007)
[31] http://www.yitran.com (July 2007)
[32] http://www.advanceddd.com (July 2007)
[33] http://www.cogency.com (July 2007)
[34] http://www.intellon.com (July 2007)
[35] http://www.powerlinecommunications.net (July 2007)
[36] http://www.homeplug.org (July 2007)
[37] J. Auvray, M. Fourrier, Problems in Electronics, including lumped constants, transmission lines and high
frequencies, ISBN 0-08-016982-1, Oxford: Pergamon Press, 1973.
[38] C. Bowick, RF Circuit Design, Carmel: Howard W. Sams & Co, 1991.
[39] P. Sutterlin, “A power line communication tutorial – challenges and technologies,” Proceedings of the 2nd
International Symposium on Power-Line Communications and its Applications (ISPLC’98), Tokyo, Japan, 24-26 March
1998, pp. 15-20.
[40] R. M. Vines, H. J. Trussel, K. C. Shuey, J. B. O’Neal, Jr., “Impedance of the residential power-distribution
circuit,” IEEE Transactions on Electromagnetic Compatibility, vol. EMC-27(1), February 1985, pp. 6-12.
[41] O.G. Hooijen, “On the relation between network-topology and power line signal attenuation,” Proceedings of
the 2nd International Symposium on Power-Line Communications and its Applications (ISPLC’98), Tokyo, Japan, 24-26
March 1998, pp. 45-56.
[42] C. R. Perez, “An automatic impedance adapter for medium-voltage communications equipment,” Proceedings
of the 4th International Symposium on Power-Line Communications and its Applications (ISPLC’00), Limerick, Ireland,
5-7 April 2000, pp. 218-224.
12
1
[43] D. Chaffanjon, “A real knowledge of propagation: the way of efficiency and reliability making PLC
generalization feasible,” Proceedings of the 2nd International Symposium on Power-Line Communications and its
Applications (ISPLC’98), Tokyo, Japan, 24-26 March 1998, pp. 57-66.
[44] M. H. L. Chan, R. W. Donaldson, “Attenuation of communication signals on residential and commercial
intrabuilding power-distribution circuits,” IEEE Transactions on Electromagnetic Compatibility, vol. EMC-28(4),
November 1986, pp. 220-230.
[45] H. Dalichau, “Evaluation of different frequency bands regarding their qualification for inhouse powerline
communication,” Proceedings of the 5th International Symposium on Power-Line Communications and its Applications
(ISPLC’01), Malmö, Sweden, 4 -6 April 2001, pp. 203-210.
[46] P. Favre et al, “Common mode current and radiations mechanisms in PLC networks,” Proceedings of the 11th
International Symposium on Power-Line Communications and its Applications (ISPLC’07), Pisa, Italy, 26-28 March
2007, pp. 348-354.
[47] H. Dalichau, “EMC aspects of inhome-PLC: crosstalk between neighbouring apartments…,” Proceedings of the
6th International Symposium on Power-Line Communications and its Applications (ISPLC’02), Athens, Greece, 27-29
March 2002, pp. 218-224.
[48] K. Dostert, “New PLC approaches for high speed indoor digital networks,” Proceedings of the 5th International
Symposium on Power-Line Communications and its Applications (ISPLC’01), Malmö, Sweden, 4 -6 April 2001, pp.
253-258.
[49] A. Klip, “A short overview of the possibilities in PLC, inside and outside the private house,” Proceedings of the
1st International Symposium on Power-Line Communications and its Applications (ISPLC’97), Essen, Germany, 2-4
April 1997, pp. 120-126.
[50] P. A. Brown, “Directional coupling of high frequency signals onto power networks,” Proceedings of the 1st
International Symposium on Power-Line Communications and its Applications (ISPLC’97), Essen, Germany, 2-4 April
1997, pp. 121-126.
[51] L. M. Milanta, M. M. Forti, “A classification of the power-line voltage disturbances for an exhaustive
description and measurement,” Symposium Record of the 1989 IEEE International Symposium on Electromagnetic
Compatibility (EMC’89), 8-10 September 1989, pp. 332-336.
[52] R. M. Vines, H. J. Trussel, L. J. Gales, J. B. O’Neal Jr., “Noise on residential power distribution circuits,”
IEEE Transactions on Electromagnetic Compatibility, vol. EMC-26(4), November 1984, pp. 161-168.
[53] S. N. Talukdar, J.C Dangelo, “Uncertainty in distribution PLC attenuation models,” IEEE Transactions on
Power Apparatus and Systems, vol. PAS-99(1), January / February 1980, pp. 328-334.
[54] A. Schiffer, “Statistical channel and noise modeling of vehicular dc-lines for data communication,” Proceedings
of the 51st IEEE Vehicular Technology Conference (VTC 2000 - Spring), Tokyo, Japan, 15-18 May 2000, pp. 158-
162.
[55] J. C. Dangelo, S. N. Talukdar, “A stochastic model for PLC systems,” IEEE Transactions on Power Apparatus
and Systems, vol. PAS-100(11), November 1981, pp. 4464-4472.
[56] A. Voglgsang, T. Langguth, G. Korner, H. Steckenbiller, R. Knorr, “Measurement, characterization and
simulation of noise on powerline channels,” Proceedings of the 4th International Symposium on Power-Line
Communications and its Applications (ISPLC’00), Limerick, Ireland, 5-7 April 2000, pp. 139-146.
[57] D. Benyoucef, “A New Statistical Model of the Noise Power Density Spectrum for Powerline
Communication,” Proceedings of the 7th International Symposium on Power-Line Communications and its Applications
[58] G. Marubayashi, “Noise measurements of the residential powerline,” Proceedings of the 1st International
Symposium on Power-Line Communications and its Applications (ISPLC’97), Essen, Germany, 2-4 April 1997, pp.
104-108.
[59] M. Harris, “Power-line communications – a regulatory perspective,” Proceedings of the 3rd International
Symposium on Power-Line Communications and its Applications (ISPLC’99), Lancaster, UK, 30 March - 1 April
1999, pp. 131-138
13
1
[60] J. Newbury, “Standardization issues for PLC & DPLC in Europe,” ISPLC 07 Panel Session I in Proceedings of
the 11th International Symposium on Power-Line Communications and its Applications (ISPLC’07), Pisa, Italy, 26-28
March 2007.
[61] J. Newbury, “Regulatory requirements for power-line communications systems operating in the high
frequency band,” Proceedings of the 5th International Symposium on Power-Line Communications and its Applications
(ISPLC’01), Malmö, Sweden, 4 -6 April 2001, pp. 305-310.
[62] P. Strong, “Regulatory & consumer acceptance of powerline products,” Proceedings of the 5th International
Symposium on Power-Line Communications and its Applications (ISPLC’01), Malmö, Sweden, 4 -6 April 2001, pp.
175-184.
[63] J. Newbury, J. Yazdani, “From narrow to broadband communications using the low voltage power
distribution network,” Proceedings of the 7th International Symposium on Power-Line Communications and its
Applications (ISPLC’03), Kyoto, Japan, 26-28 March 2003, pp. 120-124.
[64] http://www.isplc.org/Proceedings/ (July 2007)
[65] http://ieeexplore.ieee.org/ (July 2007)
*****
14
2
Chapter
CHAPTER 2
Component Selection and Circuit Layout
T
HE coupling interface between modem and power line is by nature a difficult one to
design and implement, as the two applicable waveforms represent two extremes –
high power at a low frequency versus low power at a high frequency. These extremes
often specify contradictory requirements for the various components involved, as is explained
in this chapter. Apart from selection- and design criteria for power-line communications, the
impact of parasitic effects are especially discussed.
Firstly, conductors are considered, as these form interconnections between other components.
Also, inductors and transformers are constructed using insulated conductors, and said
characteristics of conductors thus need to be well understood. As circuit layout has a direct
impact on parasitic effects, this topic is included in the discussion on conductors.
Next, resistors, capacitors, and inductors are discussed, emphasizing high-frequency and other
stray effects that need to be taken into consideration when designing a modem or coupling
circuit. Lastly, transformers are considered, as transformers are found in the majority of
coupling circuits, and provide impedance adaptation as well as help protect modem circuitry
from the power waveform.
15
COMPONENT SELECTION AND CIRCUIT LAYOUT
2
2.1 Introduction
When designing a prototype coupling circuit for power-line communications,

components need to be carefully chosen for accuracy and reliability under operating
conditions. Conductors and circuit layout are also crucial, and therefore included in
this chapter.
T
HE superimposing of a communication signal on a power waveform implies that
communication circuitry and power circuitry would have to be carefully designed
and/or interfaced for optimal compatibility between the two systems. Power systems
and communication systems operate at two different extremes – power systems at very low
frequencies and very high power, current and/or voltage levels and communication systems at
very high frequencies and very low power, current and voltage levels. To be able to design said
communication systems as well as a proper interface between power and communication
systems, components as well as circuitry must be fundamentally understood at these extremes.
This dilemma of designing for both the power waveform as well as the communication
waveform is illustrated below:
Blocking inductors: these have to be designed for the power frequency (to prevent saturation)
and for the power current (to prevent volt-drops). But blocking inductors have to function
properly especially at the modulation frequency, and therefore the self-resonant point needs to
be above said frequency. Air-core or gapped-core inductors are well suited to this application.
Coupling capacitors: these carry the communication current and thus have to be high-frequency
capacitors (self-resonant point has to be higher than the modulation frequency). Conversely,
they have to filter the power voltage (dropped across the component) as well as voltage surges
and therefore need to be high-voltage capacitors.
Coupling transformers: the main function is galvanic isolation and impedance adaptation, but the
coupling transformer has to freely pass the high-frequency communication signal and has to
be designed as such. Unfortunately the power waveform has a much lower frequency and
much higher voltage, and the power waveform would thus have a saturating influence of at
least 105 compared to the communication waveform – see Section 5.4. Therefore the power
waveform is typically first filtered by ≈ 100 dB before entering the coupling transformer.
In most designs, circuit diagrams are drawn using the lumped-parameter model. This model
assumes that a component is purely resistive, capacitive or inductive. Actually all components
have a mixture of these three attributes. The attributes not taken into account in the ideal
lumped-parameter model, are called stray or parasitic components. In the following sections,
the influences of these parasitic parameters as well as other obstacles will be discussed.
16
2
Apart from the existence of parasitic impedances, their magnitudes and influence on the
circuit are aggravated by voltage and current extremes. The effects of parasitic capacitances are
worse at high voltages whilst the effects of parasitic inductances are aggravated by high
currents. This implies that component behaviour in a power-line communications application
could be totally different to laboratory behaviour. Magnetic components (inductors and
transformers) are especially prone to unwanted side effects and have to be designed carefully.
A real-world component’s behavior depends mostly on frequency, current, voltage and
temperature of which frequency is the most common. Although temperature can have a
substantial influence on the characteristics of components and circuits, its influence will
generally only be mentioned in this chapter.
2.2 Conductors and Circuit Layout

Conductors are often overlooked during an electrical / electronic design, as these are always
assumed to be ideal in circuit diagrams. But the negligible resistance of a certain conductor
combined with the contact resistance at both ends, can cause a significant voltage drop if a
large current passes through it. Furthermore, at high frequencies, the skin effect and proximity
effect cause all conductors to have a higher ac resistance than its associated dc resistance.
Other influences that need to be considered, are the parasitic inductance of straight
conductors and loop-forming conductors, and parasitic capacitance between closely spaced
conductors.
The dc resistance of any uniform conductor can be calculated using the classical R = ρl/A
formula. It is not uncommon for long cables to have dc resistance figures of a few ohms.
When large currents flow through these, large voltages develop across them. A cable carrying
100 A of current that has a resistance of only 1 Ω causes a loss of 100 V across its terminals.
This negative effect can be reduced by increasing the cross-sectional area of the conductor.
Contact resistance falls beyond the scope of this discussion, but a few general guidelines can
be followed to minimize its influence. Contact surface areas should be large, especially solder
joints, as the resistivity of solder is much higher than that of copper. Any oxidation or dirt
should be thoroughly removed, as these obstruct the flow of current and effectively reduce
the contact area. In high-current applications, pressure is typically applied to the contact
surface in order to reduce contact resistance.
The self-inductance of a straight length of conductor, carrying alternating current, is [1]:
⎡ ⎛ 4l ⎞⎤
L ≈ 2l ⎢2.3 log 10 ⎜ − 0.75 ⎟⎥ [nH] (2.1)
⎣ ⎝ d ⎠⎦
where l and d represent the length and diameter of the conductor. The fundamental reason for
this phenomenon is that any change in current is opposed by the magnetic field surrounding
the current, creating a back-EMF type voltage across the terminals of the cable. The longer
the conductor, the larger its parasitic self-inductance becomes. The length / diameter ratio
17
2
also influences a conductor’s stray inductance – the thicker the conductor, the lower its
parasitic self-inductance. Although this inductance is theoreticaly independent of frequency, its
reactance is directly proportional to frequency, causing its impact on a circuit to increase as
frequency increases. Thus even component leads and interconnecting wires can cause
unwanted parasitic effects at high frequencies. To minimize this parasitic inductance in
practice, one would make connecting wires as short and thick as is practically convenient.
Using two (or more) spaced wires to carry the same current, has almost the same effect on
straight-line inductance as using one conductor with a diameter of the said spacing [4].
Another source of stray inductance is created when alternating current flows through any loop
formed by conductors and other components. This loop is effectively a one-turn air core
inductor. Apart from the parasitic inductance, the loop also acts as a transmitting and
receiving antenna, aggravating EMI in the circuit [2]. To minimize these negative effects,
circuit layout is extremely important – the cross-sectional area of every single loop in the
circuit needs to be made as small as practically possible. Also, pairs of interconnecting wires
are normally twisted for the same reason.
The term skin effect refers to the fact that high-frequency currents tend to flow on the surface
of a conductor. As discussed in the straight-line inductance section, any change in current is
opposed by the magnetic field surrounding the current, effectively manifesting as inductance.
These magnetic fields are stronger in the centre of a conductor, resisting the flow of current
more than on the surface. One could say that the impedance experienced by alternating
current in the centre of the conductor is higher than for ac currents flowing on the surface.
Because a large portion of the cross-sectional area is not utilized, the conductor’s ac resistance
for that specific frequency is higher than for any lower frequency or dc. The distance from
the surface at which the current density drops to 1/e or 0.368 of the maximum current density
on the surface, is called the skin depth, depth of penetration or critical depth. Theoretically, skin-
effect calculations for various waveforms can get very involved [3], but the skin depth δ for a
sine-wave with frequency f is given by [4]:
1
δ = [m] (2.2)
πσµ 0 µ R f
where µR and σ are the relative permeability and conductivity of the conductor material
respectively and µ0 is the permeability of free space. When designing a high-frequency
practical circuit, one would first calculate the skin depth of the specific metal at the operating
frequency. One could either use metal plate conductors or a twisted bundle of insulated
strands, called Litz wire [5]. A metal plate conductor would have a thickness between δ and 2δ,
and the required width to produce cross-sectional area for a certain current density. Billings [6]
gives guidelines for optimum (minimum ac resistance) plate thickness and strand diameter,
which is influenced by various factors. For Litz wire, a circular enamel-insulated strand of wire
with a diameter in the order of δ to 2δ would be chosen. The required cross-sectional area for
18
2
a certain current density would then be obtained in order to calculate the required number of
insulated strands.
The proximity effect is caused by the magnetic fields of other high-frequency alternating currents
in close proximity to the conductor in which the effect is observed. The stronger the magnetic
field, the stronger the opposition to changes in current. As the skin effect, the proximity effect
can also increase the effective ac resistance experienced by alternating currents. Fortunately
this effect only plays a substantial role in inductors and transformers, where conductors are in
close proximity, and especially where windings are tightly wound to fit into a certain core
window.
Although not as commonly observed as magnetic effects, parasitic capacitance can play an
important role where conductors are closely spaced, as predicted by the parallel plate
capacitance formula
ε 0ε R A
C= [F]. (2.3)
d
Symbols A and d represent shared cross-sectional area and distance between two parallel
conducting surfaces, ε0 the permittivity of free space and εR the relative permittivity of the
material between the conducting surfaces. A typical example of a parasitic capacitance would
be the inter-winding capacitance of a transformer, which manifests between the primary and
secondary windings, as the area between the two layers of conductors is big while the
separating distance between them is small [7]. Often insulating materials have a higher εR than
air, which increases this parasitic effect. Co-axial cables are also well known for this effect, as
the two co-axial conductors have a large common area between them, directly proportional to
the cable length.
2.3 Resistors
For power-line communications in general, one would try to avoid using resistors, as
resistance, in essence, implies a loss of power, either of the communication signal or the
power waveform. However, a resistor can be used for various purposes such as a linear
current-measuring device (low-resistance, series connected resistor strangely called a shunt [8]),
as a voltage sensing device (voltage divider circuit [9]), to bias transistor circuits or to shape
the quality factor (selectivity) of a filter or resonant circuit [1].
Although resistors theoretically have a constant impedance magnitude and zero phase angle,
they also show deviations from ideal when exposed to high frequencies. Some of the resistor-
types that will be discussed are wire-wound, carbon-composite, metal-film and thin-film chip
resistors.
The wire-wound resistor exhibits the worst performance when high-frequency currents are
applied. The manufacturing method involves the winding of a high-resistivity wire around a
19
2
cylindrical heat-resistant substrate, which typically has a relative permeability (µR) close to 1.
This means that the wire-wound resistor is essentially also a cylindrical air-core inductor. See
[2] for the necessary formulas to calculate this parasitic inductance. The distributed
capacitances between adjacent windings are typically lumped as a parasitic shunt capacitor. See
Fig. 2.1.
(a) (b)
Fig. 2.1 (a) General equivalent parallel-resonant circuit for resistors [1], and (b)
typical ⏐ZTOTAL⏐ vs. logarithmic frequency plot.
Furthermore, the straight-wire inductance of each terminal lead also adds to the total value of
the parasitic series inductance. Special high-frequency wire-wound resistors that are useful for
frequencies up to 200 kHz, are cross-wound or bifilarly wound in a special way to cancel flux-
linkages and so minimize the parasitic inductive effect [2], [10]. Also, the cross-sectional area
of the ceramic substrate (effectively an air-core), and thus the parasitic inductance, is
sometimes minimized by using a thin, flat substrate instead of a cylindrical one (see (2.7)).
Take note that the general equivalent circuit for resistors is a parallel-resonant one. Instead of
having a constant resistance (reactance) and zero phase shift, any resistor will first show
inductive, resonant and ultimately capacitive characteristics if the frequency is raised high
enough. See Fig. 2.1(b). Furthermore, the magnitude of the resistive part influences the quality
factor (relative width of resonant peak). The bigger the magnitude of a wire-wound resistor
compared to its parasitic reactances, the lower the resonant peak, and the less significant the
influence of the parasitic reactances. Thus low-value wire-wound resistors are more
susceptible to parasitic effects, and should be treated with greater care. For carbon composite
and metal film resistors, the contrary is true.
Carbon composite resistors are manufactured from densely packed carbon or dielectric
granules, and a very small capacitance exists between each granule. The aggregate of these
form the resistor’s parasitic shunt capacitance [11]. Because the lead inductance of this type of
resistor is negligible compared to its resistance and parasitic capacitance, the impedance of this
component is reasonably constant at low frequencies where the resistance dominates, but rolls
off at high frequencies where the capacitance dominates. This effect is known as the Boella
effect, named after its discoverer from Italy [11]. See Fig. 2.2.
20
(a) (b)
Fig. 2.2. (a) General equivalent RC circuit for carbon-composite and metal-film
resistors, and (b) typical normalized ⏐ZTOTAL⏐ vs. logarithmic frequency plot [1].
Carbon composite is indicated by a dotted line, and metal film by a solid line.
Metal-film resistors show excellent linear characteristics for frequencies as high as 10 MHz,
above which the shunt parasitic capacitance dominates, producing a roll-off effect similar to
carbon-composite resistors. For low-value resistors (< 10 kΩ) at very high frequencies,
resistance caused by the skin effect reduces the roll-of slope by neutralizing the parasitic
capacitive reactance to a certain degree. In resistors smaller than 100 Ω, resonance can also be
observed between the parasitic lead inductance and shunt capacitance [1].
Thin-film chip resistors have recently been developed to minimize stray effects. These are
manufactured on Alumina or Beryllia substrates, and show negligible parasitic reactance from
dc to frequencies in the order of 2 GHz [1]. Because surface mount technology is used, lead
inductance is virtually done away with. Also, in contrast to a granular material, the uniform
metal resistive layer has a negligible stray capacitance [2]. Unfortunately, these surface-mount
chip resistors can only carry small currents due to thermal constraints. For high-power
applications, band resistors can be used, made from a strip of boron-carbon encapsulated by a
heat-radiating material [2].
2.4 Capacitors
Capacitors are used extensively in power-line communications, most commonly to couple the
communication signal to the power line [12], but also as part of more sophisticated, higher-
order filters [13] and general communication / modem circuitry [14]. Just as any other passive
electronic component, a real-world capacitor has all three L, R and C parameters, but is called
a capacitor because the C component dominates. Fig. 2.3 shows the equivalent circuit that is
generally valid for capacitors [1], [15]. Often the series inductance is omitted [15], [16] to
simplify the model, but sometimes more complicated models are used [17]. Take note that Fig.
2.3(a) is a series-resonant one, incorporating losses associated with contacts and terminals (RS)
as well as the dielectric (RP). RS is typically small, in the order of 0.1 to 1 ohm, but RP is usually
in the mega-ohm to giga-ohm range [1].
21
(a) (b)
Fig. 2.3 (a) General equivalent circuit for capacitors [1], and (b) typical ⏐ZTOTAL⏐
vs. logarithmic frequency plot.
Generally, capacitors are quite linear up to a certain point in frequency, called the self-resonant
frequency. This frequency, fRES is determined by the product of the capacitance C with its
parasitic inductance L:
1
f RES = [Hz] (2.4)
2π LC
Above resonance though, the parasitic inductance dominates, making the capacitor behave
inductively. For an ideal capacitor, L would be zero, and the resonant response and associated
inductive response above resonance, would be shifted to an infinite frequency. Thus for very
high frequency applications, capacitance values together with their parasitic inductances, must
be minimized in order to shift the self-resonant frequency higher than the operating
frequency.
The quality factor Q of a capacitor is determined by its total internal parasitic resistance at a
certain frequency, called equivalent series resistance (abbreviated ESR), made up of the
combination of RS and RP. Q is the inverse of the capacitor’s power factor, and its relationship
with the ESR (symbol r) at a certain frequency f is as follows:
1 1
Q= = [–] (2.5)
cos φ 2πfrC
Apart from frequency, temperature is the second-most important factor that influences the
characteristics of capacitors negatively. Generally, the higher a capacitor’s εR, the smaller its
physical dimensions for a certain capacitance, and the more prone it is to have non-linear
temperature characteristics. Also, a high Q-value (low power factor) means that very little
power is dissipated in the capacitor, making it operate at a lower, more stable temperature
than a low-Q counterpart.
Many different materials and manufacturing processes are used to produce capacitors. Three
different groups of dielectrics exist, and can be used to classify capacitors as air-dielectric,
22
2
solid-dielectric or liquid-dielectric [5]. Air-dielectric capacitors are typically used as variable
capacitors for tuning applications. Low-permittivity (εR) capacitors are also used where good
temperature stability is required. An insulator with εR ≈ 1 can be used to raise the breakdown
voltage for high-voltage applications. Solid-dielectric capacitors include ceramic, mica, paper
and metalized plastic-film capacitors of which ceramic capacitors are generally suitable for
high frequencies [2]. Some ceramic capacitors are specially manufactured for high-frequency
applications, using ribbon connector leads, or as surface-mount chip capacitors [1]. Metalized
plastic-film capacitors exhibit very good tolerances (typically 2 %) for temperatures below 90
0
C and are generally used where space is not a constraint. Liquid-dielectric capacitors include
electrolytic capacitors and super capacitors [2]. These are used where large capacitances are
required such as for filtering or UPS applications.
2.5 Inductors
In power-line communications, inductors are commonly used to de-couple portions of the
power line from the power-line communications network, as inductors impede high-frequency
signals. This technique prevents the unnecessary dissipation of transmitted power in certain
sections of the power line and associated loads [12]. Inductors also often form part of
coupling circuits, filter circuits and tuning circuits [13]. The most common equivalent circuit
for inductors is shown in Fig. 2.4, and is very similar to a resistor’s equivalent circuit [1], [18],
[19]. In this parallel-resonant circuit though, the inductance dominates the other parameters at
low frequencies, and thus functions as an inductor.
(a) (b)
Fig. 2.4 (a) General parallel-resonant equivalent circuit for inductors [1], and (b)
typical ⏐ZTOTAL⏐ vs. logarithmic frequency plot.
The resistance in Fig. 2.4(a) models losses in the winding as well as magnetic core (and strictly
speaking dielectric losses associated with the stray capacitance as well as any radiated losses)
[2]. The capacitance C is the lumped value of the total stray capacitance, formed between
neighboring conductors that have a volt-drop between them. Thus for high-voltage
applications, the influence of C is much more significant, as higher volt-drops exist between
23
2
conductors. To minimize the parasitic capacitance one would space conductors as far apart as
possible, increasing the distance between any two miniature capacitive layers. This is achieved
in the single-layer solenoid often used for high frequencies [11]. Unfortunately, this spacing
degrades the inductive coupling behavior of the component, resulting in a smaller effective
inductance.
A fundamental equation defining inductance L carrying a current i, is given by Faraday’s law

and can be written in different forms:
dφ d ⎛ di dL ⎞
e = −n = − ( Li ) = −⎜ L + i ⎟ [V] (2.6)
dt dt ⎝ dt dt ⎠
Symbol e represents the induced voltage across the inductor (often called back-EMF) that
always opposes the cause, hence the negative sign. Symbols n and φ represent the number of
turns and flux linked by the turns. Take note that the last term of (2.6) is often neglected,
assuming L constant. Practically, this can often be very inaccurate. Various factors can cause L
to change abruptly [2], the most critical being 1) non-linearity of core µE at higher excitation
levels as the B-H curve flattens, 2) frequency dependency of µE indirectly caused by skin and
proximity effects, influencing effective dimensions of coil and 3) temperature dependency of
µE as core heats to reach operating temperature. (It is important to understand that the
effective permeability, µE, is not a predetermined value, but rather an observed ratio between
B and H for a certain core under certain operating conditions.)
Air-core inductors have definite advantages over magnetic core inductors: they are almost 100
% linear with respect to excitation, and saturation effects do not apply. If the conductors are
well designed for high frequencies (e.g. Litz wire), the air core inductor is also very linear with
respect to frequency and to a lesser degree, temperature changes [2]. The disadvantages of air-
core inductors coincide with the advantages of magnetic core inductors: air-core inductors are
generally large and require many turns compared to cored inductors. Because of the higher
number of turns, the parasitic resistance and capacitance is higher than for a cored inductor.
Also, the large cross-sectional area makes the inductor act as a radiating and receiving antenna,
aggravating EMI in the circuit and surroundings where it is used.
Magnetic cores having an effective permeability µE of 1000 is not uncommon, and therefore
would boost the inductance of an air-core counterpart by factor 1000. See (2.7), the general
design equation for an ideal homogeneous thyroidal inductor (having N turns wound on a
core with effective cross-sectional area A and flux path-length l ) [2]:
A
L = µ0 µ E N 2 ⋅ [H] (2.7)
l
Symbols µ0 and µE are permeability of free space and the effective relative permeability of the
core material respectively. If one compares a cored inductor with an air-core inductor at a
certain fixed inductance, the physical dimensions (factor A/l ) of the cored inductor is smaller
24
2
by factor µE if N is kept constant. Alternatively, the number of turns is reduced by a factor
equal to the square root of µE (keeping geometry fixed). Practically, a combination of both is
implemented, yielding a smaller inductor with a fewer number of turns. Fewer turns implies a
smaller parasitic capacitance as well as a smaller parasitic resistance (that can lead to an
increase in Q, depending on the introduced core losses). Another advantage is the adjustability
of a cored inductor, which can be realized by moving the core in and out of the winding.
Magnetic-core inductors also have disadvantages. All magnetic materials saturate at a certain
level of excitation, and this causes harsh non-linearity in a circuit. Saturation of inductors and
transformers has to be carefully considered for power-line communications applications, as
signals get distorted when large currents saturate magnetic cores. Also, different magnetic
materials have different losses, influencing the quality factor. The uniqueness of a certain
magnetic material though, is summarized by its permeability characteristics, which varies with
temperature, frequency and excitation. Consequently, the inductance changes from its original
intended value, causing impedance mismatches, ineffective filtering and mistuning, to name a
few negative effects.
2.6 Transformers
A transformer is a device with two or more inductors that are magnetically coupled. Typically
a high-permeability magnetic core is used in order to effectively couple the two (or more)
windings. At low frequencies (most power applications), it is sufficiently accurate to model a
transformer by a magnetizing and leakage inductance. The magnetizing inductance models the
energy storage in the magnetic core, whilst the leakage inductance models the energy stored
outside of the core (flux in the windings as well as air, oil or resin surrounding the
transformer). A well-designed transformer would have a large magnetizing inductance
compared to its leakage inductance, implying effective coupling.
At high frequencies though, and for very sensitive applications, the parasitic influences of a
transformer has to be modeled more accurately. Fig. 2.5 shows a more comprehensive model
[7] that can be used for any transformer. LM and RC represent the magnetizing impedance and
core losses (hysteresis and eddy current) respectively. Both the primary and secondary
windings have conduction losses represented by R as well as leakage inductance represented
by LL. The parasitic capacitances are the lumped values of capacitance between different turns
and layers of the same winding (CP and CS) and capacitance between the primary and
secondary windings (CPS). In order to design a transformer where parasitic influences are
minimal, some practical guidelines can be followed:
25
(a) (b)
Fig. 2.5 (a) Comprehensive lumped equivalent circuit for transformers [7], and (b)
typical ⏐ZP⏐ (no-load impedance seen from primary) vs. logarithmic frequency
plot.
Leakage inductances are largely dependent on transformer geometry. Langsdorf [20] shows
that the total leakage inductance is almost proportional to the height-to-width ratio of the
windings. Using a transformer core that facilitates a winding of double the original width,
would necessitate only half the winding height for the same number of turns, effectively
reducing the leakage inductance by factor 4.
Interleaving of layers of primary and secondary turns can have a significant impact on parasitic
reactances, as interleaving reduces the magneto-motive force (MMF), magnetic field strength H
and associated flux density B inside the transformer core and windings. Leakage inductance is
further reduced by a factor almost equal to the number of layers used, for instance by a factor
4 when using two primary and two secondary layers [20].
This technique can also reduce the effect of inter-winding capacitance (CPS), as the relative
voltage between layers is smaller. A larger capacitive area is introduced though, which could
aggravate CPS, depending on the specific geometry and design. Furthermore, CP and CS are
drastically reduced because the distance between capacitive layers is increased by a factor of 10
or even more. Core losses is also less (RC smaller) because of lower excitation of the core.
Winding losses represented by RP and RS can be minimized by using Litz wire or plate
conductors to minimize the skin effect. See Section 2.2. Also, the proximity effect can be
minimized by using an optimum number of layers when applying the interleaving technique
[6].
In general, thyroidal transformers exhibit good characteristics (low parasitic reactances)

compared to other cores. When few turns are required for a specific design, parasitic
capacitances can be minimized by spacing windings apart to minimize CPS and by spacing
turns equally to minimize CP and CS [1].
26
2
2.7 Component Integration
Conventional electric circuit theory is based on discrete, ideal L, R and C components that are
connected with ideal conductors, and therefore the majority of electronic designs are done
using this approach. As was made clear in the preceding sections, every electronic component
has a certain measure of inductance (L), resistance (R) and capacitance (C ). Depending on
operating frequency, current and voltage levels as well as temperature, a specific L, R or C will
typically dominate the behavior of the component. To be able to design and manufacture an
accurate, reliable electronic system, one has to take the necessary precautions against
unforeseen behavior when doing a design.
Components are purchased or manufactured to behave as ideal as possible under the

operating conditions, and modeled as such. Lead-inductance, for instance, is made
insignificant with the use of surface mount technology. All other parasitic influences are
minimized when designing and manufacturing the component. If certain parasitic influences
would change a component’s behavior substantially, it would be modeled by an equivalent
circuit that again consists of discrete (ideal) components. The latter is also not accurate, as
parasitic influences are prone to change with frequency, voltage, current and temperature.
The explosive growth of semiconductor integrated circuits since the 1950’s caused a general
belief that even passive components will eventually be replaced by monolithic semiconductor
integrated circuits. Contrary to this belief, passive component sales have doubled between
1990 and 1995, contributing 70 % to the global electronic component sales in 1995 [21]. In
the majority of today’s state-of-the-art electronic equipment, passive components dominate
integrated circuits, both in terms of component count and board space occupied.
A typical video camcorder has 450 passive components and 10 integrated circuits (ICs). This 45:1
ratio drops to 20:1 in digital cellular phones where one still finds 150 to 200 passive
components [21]. Although passive components are cheap compared to active components or
integrated circuits, the logistics associated with stocks, purchasing and insertion on a printed
circuit board, doubles the initial purchasing cost of a typical passive component. Furthermore,
the required board space hinders miniaturization of the system. Therefore engineers have been
researching various techniques to minimize passive component size and cost, while improving
performance and predictability.
Functional integration of discrete passive components involves the manufacturing of various

discrete L or R or C components into a common substrate or package. Although these are still
discrete devices in one chip, this technique facilitates miniaturization, enhances performance
by reducing circuit parasitic parameters, increases reliability and reduces manufacturing costs
as well as assembly costs [21].
Although one cannot get away from using discrete components, a different approach,
electromagnetic integration, seems to be the obvious solution in some cases. Here one would make
use of the fact that real-world components have a mixture of characteristics, and actually
27
2
design a single passive component to have two or more contrasting influences, for instance a
LC component or a RC component. Apart from facilitating miniaturization and reducing
manufacturing costs as well as assembly costs, reliability and circuit performance is enhanced:
parasitic parameters of the component is absorbed into its functionality and used as part of
the circuit.
As early as 1929, the first electromagnetic-integrated LC component was proposed [22] and in
1949 the first (British) patent was awarded for an electromagnetic-integrated LC component
[23]. Electromagnetic-integrated (distributed) RC components have also been investigated
since the 1960’s [24]-[26]. This research was partly stimulated by the development of
semiconductor integrated circuits, as the parasitic parameters in these semiconductor
integrated circuits had to be understood. It became obvious that parasitic parameters might as
well be utilized as part of a distributed structure (electromagnetic-integrated component) [27].
High current and voltage applications (at high frequencies) pose further difficulties, and only
since the late 1980’s did researchers manage to cross the 100-W barrier [28], [29].
Because of material and manufacturing constraints, neither functional- nor electromagnetic-

integrated passive components were initially commercially viable. Researchers are currently
still working actively to design and produce cost-effective passive integrated devices for micro-
electronics [21], [30], [31] as well as power electronics [32]-[34].
2.8 Conclusion
P OWER systems and communication systems operate at two different extremes –

power systems at very low frequencies and very high power, current and/or voltage
levels and communication systems at very high frequencies and very low power,
current and voltage levels. These extremes can cause unwanted behavior if components are
not selected carefully and / or circuit layout is planned properly.
Frequency behavior of passive components was discussed, and practical design guidelines
were given. Generally components are chosen to have a self-resonant point of at least an
octave above the operating frequency. The self-resonant point of a certain component can be
raised by either reducing the parasitic values (as discussed) or if not possible, designing the
circuit using smaller component values (not true for wire-wound resistors though). Parasitics
introduced by circuit layout should also be minimized and/or designed for.
Finally, it is suggested that components be characterized at low-frequency power-line voltages

and currents where applicable, but first and foremost over the range of modulation
frequencies, applying low power.
28
2
2.9 References
[1] C. Bowick, RF Circuit Design, Carmel: Howard W. Sams & Co, 1991.
[2] W. K. Chen (ed.), The Circuits and Filters Handbook, Boca Raton: CRC Press and IEEE Press, 1995.
[3] C.-S. Yen et al, “Time-domain skin-effect model for transient analysis of lossy transmission lines,” Proceedings
of the IEEE, vol. 7(7), July 1982, pp. 750-757.
[4] J. Auvray, M.Fourrier, Problems in Electronics, including lumped constants, transmission lines and high frequencies, ISBN
0-08-016982-1, Oxford: Pergamon Press, 1973.
[5] F. E. Terman, Electronic and Radio Engineering, 4th edition, New York: Mc Graw-Hill, 1955.
[6] K. H. Billings, Handbook of Switchmode Power Supplies, New York: Mc Graw-Hill, 1989.
[7] A. Baccigalupi et al, “On a circuit theory approach to evaluate the stray capacitances of two coupled
inductors,” IEEE Transactions on Instrumentation and Measurement, vol. 43(5), October 1994, pp. 774-776.
[8] D. W. Braudaway, “Behaviour of resistors and shunts: with today’s high-precision measurement capability
and a century of materials experience, what can go wrong?” IEEE Transactions on Instrumentation and
Measurement, vol. 48(5), October 1999, pp. 889-893.
[9] S. R. Naidu, A. F. C. Neto, “The stray-capacitance equivalent circuit for resistive voltage dividers,” IEEE
Transactions on Instrumentation and Measurement, vol. IM-34(3), September 1985, pp. 393-398.
[10] D. W. Braudaway, “Precision resistors: a review of material characteristics, resistor design, and construction
practices,” IEEE Transactions on Instrumentation and Measurement, vol. 48(5), October 1999, pp. 878-883.
[11] K. Henney, Radio Engineering Handbook, 5th edition, New York: Mc Graw-Hill, 1959.
[12] H.-K. Podszeck, Carrier Communication over Power Lines, 4th Edition, New York: Springer-Verlag, 1972.
[14] K. K. Clarke, D.T. Hess, Communication circuits: analysis and design, Reading: Addison-Wesley, 1971.
[15] J.-Y. Kim et al, “High-frequency response of amorphous tantalum oxide thin films,” IEEE Transactions on
Components and Packaging Technologies, vol. 24(3), September 2001, pp. 526-533.
[16] C.-H. Lai, T.-Y. Tseng, “Analysis of the ac electrical response for (Ba,Pb)TiO3 positive temperature
coefficient ceramics,” IEEE Transactions on Components and Packaging Technologies, Part A, vol. 17(2), June 1994,
pp. 309-315.
[17] Y. Sakabe et al, “High-frequency measurements of multilayer ceramic capacitors,” IEEE Transactions on
Components and Packaging Technologies, Part B, vol. 19(1), February 1996, pp. 7-14.
[18] J. Muciek, A. Muciek, “A comparison of low-frequency models of inductance standards,” IEEE Transactions
on Instrumentation and Measurement, vol. IM-36(3), September 1987, pp. 830-833.
[19] R. Hanke, K. Dröge, “Calculated frequency characteristic of GR1482 inductance standards between 100 Hz
and 100 kHz,” IEEE Transactions on Instrumentation and Measurement, vol. 40(6), December 1991, pp. 893-896.
[20] A. S. Langsdorf, Theory of alternating-current machinery, New York: McGraw-Hill, 1955.
[21] G. van de Walle, “Integration of passive components: an introduction,” Philips Journal of Research, vol. 51(3),
1998, pp. 353-361.
[22] H. Peierls, Elektrisher Wickelkondensator, German Patent, PS 609 980, D.R., 1929.
[23] E. K. Cole ltd., W.A. Stickley, British Patent, 623 501, 1949.
[24] W. M. Kaufman, “Theory of a monolithic null device and some novel circuits,” Proceedings of the IRE,
September 1960, pp. 1540-1545.
[25] R. W. Wyndrum Jr., “Distributed RC Notch Networks,” Proceedings of the IEEE, vol. 51(2), February 1963,
pp. 374-375.
29
2
[26] R. Levy, T. E. Rozzi, “Precise design of coaxial low-pass filters,” IEEE Transactions on Microwave Theory and
Techniques, vol. MTT-16(3), March 1968, pp. 142-147.
[27] H. J. Carlin, “Distributed circuit design with transmission line elements,” Proceedings of the IEEE, vol. 59(7),
July 1971, pp. 1059-1081.
[28] H. Van Der Broek et al, “A compact high performance 300W / 5V switched mode power supply,” Proceedings
of 3rd European Conference on Power Electronics and Applications (EPE ’89), Aachen, Germany, 9-12 October 1989,
pp. 1455-1459.
[29] I. W. Hofsajer et al, “Functional component integration in a multi-kilowatt resonant converter,” Proceedings of
5th European Conference on Power Electronics and Applications (EPE ’93), vol. 2, Brighton, UK, 13-16 September
1993, pp. 125-130.
[30] N. J. Pulsford, “Applications of passive integration,” Philips Journal of Research, vol. 51(3), 1998, pp. 411-428.
[31] W.-Y. Choi, S. Trolier-McKinstry, “A voltage-controlled tunable thin-film distributed RC notch filter,”
IEEE Transactions on Components and Packaging Technologies, vol. 24(1), March 2001, pp. 33-37.
[32] I. W. Hofsajer, J. A. Ferreira, J. D. Van Wyk, “Design and analysis of planar integrated L-C-T components
for converters,” IEEE Transactions on Power Electronics, vol. 15(6), November 2000, pp. 1221-1227.
[33] J. T. Strydom, J. D. Van Wyk, J. A. Ferreira, “Some limits of integrated LCT modules for resonant
converters at 1MHz,” Conference Record of the 34th IEEE Industry Applications Society Conference (IAS ’99), vol. 2,
Phoenix, USA, 3-7 October 1999, pp. 1411-1417.
[34] J. T. Strydom et al, “Power electronic subassemblies with increased functionality based on planar
subcomponents,” Conference Record of the 31st IEEE Power Electronics Specialist Conference (PESC’00), vol. 3,
Galway, Ireland, 18-23 June 2000, pp. 1273-1278.
*****
30
3
Chapter
CHAPTER 3
Coupling Fundamentals
I N essence, a coupling circuit is a special filter that allows a high-frequency

communication signal to pass through it (preferably in both directions) while ‘blocking’
or filtering the power waveform. Therefore this chapter introduces filtering concepts in a
very simple, yet practical way in order to clarify the uncertainty that exists over coupling
circuits for power-line communications. If no impedance adaptation is required, a standard
passive filter circuit may be used, as long as the applicable components are rated to block the
power waveform voltage. Extra protective circuitry is also advisable.
However, if impedance adaptation is required between the modem and the power line,
transformers are indispensable. This chapter therefore expands on the role of transformers in
coupling circuitry. As the internal series (‘leakage’) impedance of a transformer is inductive, all
transformers exhibit some kind of low-pass effect, filtering frequencies higher than this cut-
off point. However, if the coupling circuit is required to implement a band-pass filter, extra
components are required. Typically, a single series capacitor is sufficient, and this capacitor
effects a high-pass filter, attenuating frequencies below this cut-off point. This chapter further
illustrates that bandwidth of this coupling circuit is dependent on the terminating impedance.
Therefore the series-resonant point of the leakage inductance and coupling capacitor should
preferably be designed to match the transmission frequency of the modem.
31
COUPLING FUNDAMENTALS
3
3.1 Filtering Concepts
A power-line communications coupler needs to do two filtering tasks – blocking the

power waveform while simultaneously allowing the modulated communication signal
to pass.
F ILTERING (ability to discriminate between different frequencies) forms the basis of

most coupling and de-coupling circuits. De-coupling filters are typically used for
network conditioning whereas coupling filters feed the communication signal to and
from the power line. Other applications of filtering include the filtering of noise, filtering to
improve phase distortion and filtering to prevent incoming and outgoing EMI. It is obvious
from the above examples that proper filter design is i) essential for minimum signal
attenuation when coupling to the power line and ii) can drastically improve signal to noise
ratios.
As any repetitive waveform can be expressed as a mathematical sum of sinusoids (called its
Fourier expansion), a periodic waveform can be thought of as a unique mixture of sinusoids,
each with a certain frequency and amplitude, that form one superimposed signal. Every
possible sub-sinusoid of a signal is changed by the filtering process, depending on its
frequency [1], [2]. This change is two-fold: amplitude is attenuated by a certain factor, and
phase angle is shifted to a certain degree. After the filtering process, the resultant waveform is
once again the sum of the individual sub-sinusoids. Thus the filtered waveform is now
composed to a lesser degree of some frequency (sinusoidal) components because they have
been attenuated or filtered to a certain degree. As each sub-sinusoid is typically phase-shifted by
a different angle, the filtered waveform does not only show phase delay but also phase
distortion, a side-effect that has to be kept in mind [1], [2].
Single and paired capacitors are used extensively in power-line communications to couple the
communication signal to the power line while ‘blocking’ the low-frequency power signal [3],
[4]. This application is well known in transistor circuit theory, as a series capacitor is typically
used to disconnect or block dc biasing voltages but pass small-signal ac voltages. This is
possible due to the frequency-dependent impedance of a capacitor (valid for a specific
frequency sine-wave):
1 1
ZC = = ∠ − 90 0 [Ω] (3.1)
jωC 2πfC
From (3.1) it is obvious that any capacitor’s impedance will tend to infinity at dc (f = 0 Hz),
blocking any dc component of a signal. At low frequencies, such as power-line frequencies, its
impedance is high enough to revert almost 100 % of the signal to lower impedance paths of
the power line (current divider rule). At the communication signal frequency though, the
32
3
coupling capacitor would be designed to have a low enough impedance to admit a large
portion of the communication signal, making it available to the receiver.
If a power-line communication system uses a sinusoidal carrier signal of 500 kHz

superimposed on a 50-Hz sinusoidal power waveform, the following conclusions can be made
for a 1-µF series capacitor: the capacitor can be modeled as two parallel impedances, one
parallel path for low frequencies and another for high frequencies. See Fig. 3.1(a).
(a) (b)
Fig. 3.1 HF-LF model for a (a) 1-µF capacitor and (b) 1-µH inductor in a 50-Hz
sinusoidal power and 500-kHz sinusoidal carrier scheme.
The low-frequency (LF) path has a reactance of 3.18 kΩ whereas the high-frequency (HF)
path has a reactance of only 0.318 Ω (inversely proportional to frequency). Both the HF and
LF path impedances have the same phase angle though (-900).
Inductors function complementary to capacitors: high-frequency signals (or harmonics) are

blocked and low-frequency signals (or harmonics) passed, reason being that an ideal inductor
has a unique impedance for each sub-sinusoid of a waveform, proportional to the sub-
sinusoid’s frequency f:
Z L = jωL = 2πfL∠ + 900 [Ω] (3.2)
In a power-line network, any series inductor would impede the high-frequency

communication signal to a certain degree, from flowing through that specific path.
Unnecessary loss of transmitted power can be prevented by ‘blocking’ off branches of the
network that will not be utilized for power-line communications. If a 1-µH series-inductor is
considered with the same power-line communication scheme as discussed above (50 Hz with
500-kHz carrier), it can be modeled as a 0.314-Ω reactance for low frequencies in parallel with
a 3.14-kΩ reactance for high frequencies. See Fig. 3.1(b). Both the HF and LF path
impedances have the same phase angle of +900 as for an ideal inductor.
It must be understood that a series capacitor or inductor usually functions as a first-order

filter, as these often terminate into a resistive measuring load or resistive power load. In Fig.
3.2, the HF-LF model is used further to illustrate a 1-µF capacitor terminating into a 50-Ω
load resistor. The voltage-divider rule can be used to obtain the voltage transfer function as
33
3
VOUT 50∠0 0
= ≈ 1∠0.364 0 [–] (3.3)
V IN 50∠0 + 0.318∠ − 90 0
0
for the HF path, and
VOUT 50∠0 0
= ≈ 0.0157∠89.10 [–] (3.4)
V IN 50∠0 0 + 3.18E 3∠ − 90 0
for the LF path. VOUT is measured across the load resistor while VIN represents the voltage
applied across the R-C branch. For this specific example of Fig. 3.2, the high-frequency
sinusoid is hardly attenuated, and phase-shifted by only 0.3640 whereas the low-frequency
voltage signal is attenuated by 98.43 % or 36 dB (and phase-shifted by 89.10). Also refer to
Fig. 3.3 which affirms latter calculations.
Fig. 3.2 The HF-LF model for a 1-µF capacitor terminating into a 50-Ω load
resistor (50-Hz sinusoidal power and 500-kHz sinusoidal carrier scheme).
Although the discussed HF-LF model facilitates the understanding of a simple filter circuit,
the Bode plot presents a summary of attenuation and phase shift for a logarithmic range of
frequencies. When designing a FSK modulation scheme for instance, the attenuation and
phase shift of different modulation frequencies can be graphically considered. See Fig. 3.3.
(a) (b)
Fig. 3.3 Measured Bode plots for a 1-µF capacitor terminating into a 50-Ω load
resistor: (a) voltage transfer function and (b) phase transfer function. Note the
influence of stray effects above 1 MHz. Instrument: HP3577B Network Analyzer.
34
3
Inductor-capacitor combinations can also be utilized as second-order low-pass or high-pass
filters, the main difference being a roll-off figure of 12 dB/octave compared to 6 dB/octave
for first-order RC and RL filters. Band-pass and band-stop filters can also be realized with LC
combinations, but only have a 6 dB/octave roll-off figure [4].
Series resistance can also be introduced to dampen resonance and so reduce the Q-factor of
the second-order filter. As this technique involves an impedance mismatch, it is a non-
selective attenuation technique and is therefore not recommended. If the mismatch is not
severe, unattenuated frequencies closer to series-resonant points are affected more than other
attenuated frequencies.
The more severe the mismatch, the more prone it is to attenuate all frequencies. Typically,
filter circuits also terminate into resistive loads, thereby influencing the bandwidth of the
second-order system to be more selective (higher Q) or less selective (lower Q), depending on
the configuration. If a filter circuit terminates into a reactive load, its characteristics could
change to that of a higher-order circuit, complicating the response of the system.
3.2 Transformer Couplers

Transformers are often used as coupling devices as they provide galvanic isolation between
their primary and secondary windings (between the power circuitry and communication
circuitry) and act as limiters [5]. Other uses of transformers include voltage / current
transformation, adaptation of impedance levels and filtering.
Because transformer design involves so many parameters it is often a non-precise,

experimental procedure. Therefore transformers are typically not suitable for accurate
complex impedance matching but rather for crude equaling of resistive impedances or else
reactance magnitudes. Fig. 3.4 shows a simplified equivalent circuit for a transformer,
assuming an ideal linear B-H curve. Practical B-H curves are non-linear though, and depend on
excitation levels.
(a) (b)
Fig. 3.4 (a) Simplified linear model for a transformer and (b) practical non-linear
B-H curves.
35
3
When designing a transformer, it is important to limit the maximum flux density in the core
for various reasons:
• The higher the flux density, the more non-linear the B-H curve, and the more distortion is
introduced in the signal across the secondary winding.
• If the core is saturated, induction ceases (no signal reaches the secondary), and the
primary winding appears as a short circuit as it has a very low magnetizing impedance
(that of an air core, as µ(t) = dB/dH).
• The enclosed area of the B-H curve is proportional to the power losses in the core. The
core temperature can influence the transformer’s frequency response. Failure caused by
overheating is also possible.
Practically, B can be controlled by implementing one or more of the following:
• Choose a core with a large enough cross-sectional area.

• Limit the maximum applied voltage and minimum applied frequency (see (3.5)).
• Interleave layers of primary and secondary turns to reduce leakage flux.
• Increase the number of turns, keeping the same winding ratio.
The latter increases a transformer’s magnetizing inductance LM by factor N2, and so limits the
magnetizing current necessary to excite the transformer. Smaller magnetizing current implies
less dissipation in the core. In general, the more effective a transformer’s window surface is
filled with primary and secondary turns, the more efficient the transformer will be. Wire
diameter is typically optimized to produce equal losses in the windings and the core.
Interleaving layers of primary and secondary conductors will further reduce the MMF and so
reduce leakage inductance, core losses and copper losses.
In a power-line communication application, the B-H curve of a transformer would show small
high-frequency hysteresis curves superimposed on a large low-frequency (50-Hz) hysteresis
curve. The sum of the low-frequency and high-frequency flux densities is limited to BSAT and
therefore the two signals have to share the available maximum flux density. A transformer’s
maximum excitation limit is always determined by the highest voltage level and lowest
operating frequency [7] as (3.5) (valid for sine wave) clearly illustrates:
v MAX
B MAX = [T] (3.5)
2πNAf MIN
From (3.5), it can be shown that a 50-Hz, 311-V(PEAK) power signal would waste the majority
of the available BSAT compared to a 500-kHz, 10-V(PEAK) communication signal. The frequency
difference of factor 104 multiplied with the voltage difference of factor 31, shows that the
power signal would saturate a core of BSAT = 600 mT while the communication signal has not
even utilized 2 µT of the 600-mT available flux density.
36
3
For this very reason, the low-frequency power sinusoid needs to be filtered to reduce its
amplitude drastically before entering the coupling transformer. This is typically achieved by
connecting a series capacitor to one or both of the transformer primary terminals [5], [8].
Most transformers show natural band-pass filtering characteristics because of internal
impedances [9], [10]. Remember that any transformer will saturate at a frequency too low for
its design (see (3.5)). As the core goes into saturation, the induced secondary voltage drops to
zero. At high frequencies though, the small series leakage inductances dominate to form a
low-pass LR filter in association with parallel load resistance and other resistances [9], [10].
The series capacitor that is used to prevent saturation of the coupling transformer in
conjunction with the transformer’s leakage inductance creates a series-resonant coupling
circuit. If this series-resonant circuit is terminated into a resistive load, a second order band-
pass filter is realized. This band pass filter has a roll-off figure of 6 dB/octave or 20
dB/decade (see Fig. 3.6(b)). If a higher roll-off figure is required, parallel capacitors have to be
introduced to both the primary and secondary windings. See [11], [12] for details. The center
frequency of this band-pass filter is at the series-resonant point
1
fR = [Hz] (3.6)
2π LC
where L refers to the series inductance and C refers to the series capacitance. L typically
consists only of the leakage inductance referred to primary, but can be enlarged with a series
inductor. The bandwidth of the filter is determined by the respective low-frequency and high-
frequency –3dB cut-off points
1
f LF = [Hz] (3.7)
2πRC
and
R
f HF = [Hz] (3.8)
2πL
where R refers to the terminating resistance. The model of the suggested coupling circuit is
shown in Fig. 3.5 below:
Fig. 3.5 Suggested coupling circuit.
37
3
Fig. 3.6(a) shows the measured amplitude response (⏐H(jω)⏐ vs. frequency) of a RS 196-369
1:1 coupling transformer only, terminating in the 50-Ω instrument input impedance. Unless
stated otherwise, all measurements were done with a HP 3577B 200-MHz network analyzer
with 50-Ω output and input impedances.
The measured cut-off point in Fig. 3.6(a) of 382 kHz corresponds to the theoretical value (see
(3.8)) of 419 kHz, caused by leakage inductance (typical value 19 µH) terminating in the 50-Ω
instrument impedance.
The low-frequency cut-off point in Fig. 3.6(a) did not manifest during the measurement, as the
lowest frequency measured was 20 Hz, and the applied voltage was too low to cause saturation
at this frequency (see (3.5)). It was determined experimentally that the low-frequency cut-off
point is ≈ 900 Hz for a 20-VP-P sine wave and ≈ 650 Hz for a 2-VP-P sine wave. However, the
maximum test signal that could be applied without overloading the instrument was only -6
dBm, approximately 112 mV if a 50-Ω load is assumed.
Fig. 3.6(b) shows the transfer function of the 1:1 transformer with a 0.22-µF capacitor in
series with primary terminal (also 50-Ω secondary termination). The measured center
frequency and cut-off points correspond closely to the calculated values of fR ≈ 78 kHz, fLF ≈
14.5 kHz and fHF ≈ 419 kHz.
(a) (b)
Fig. 3.6 Measured amplitude response of (a) 1:1 transformer only and (b) 1:1
transformer with 0.22-µF capacitor in series with primary terminal (both 50-Ω
secondary termination).
The influence of both the terminating- and series resistances on the filter characteristics was
investigated. See Fig. 3.7 below. Fig. 3.7(a) is a follow-up of Fig. 3.6(b), but the terminating
resistance was decreased to 25 Ω. Here the center frequency is still 78 kHz, but the cut-off
points have moved closer to form a narrower pass band. The theoretical values of fR ≈ 78 kHz,
fLF ≈ 29 kHz and fHF ≈ 210 kHz correspond closely to the measured values.
38
3
In Fig. 3.7(b) the influence of series resistance (as discussed in a previous paragraph) is shown.
Fig. 3.7(b) is a follow-up of Fig. 3.7(a), but a 50-Ω resistor was inserted in series with the
transformer primary.
Although the impedance mismatch caused by the 50-Ω resistor does cause a broader pass-
band, the pass-band region is attenuated by approximately 10 dB, whereas already filtered
frequencies (compare with Fig. 3.6(a)) are not attenuated further. This may impact on signal-
to-noise ratios, and is therefore not recommended. The use of a series resistor is only
warranted when impedance matching is improved by its insertion into the circuit.
(a) (b)
Fig. 3.7 Measured amplitude response of a 1:1 transformer with 0.22-µF capacitor
in series with primary terminal but with 25-Ω secondary termination. Fig. 3.7(a)
illustrates that a decreased termination resistance decreases the bandwidth. Fig.
3.7(b) shows the effect of an impedance mismatch (50-Ω series resistor) on the
filter characteristics.
For confirmation purposes, both the series inductance and capacitance were made larger. Fig.
3.8(a) follows on Fig. 7(a) but a series inductor of 12 µH was added to the leakage inductance
of 19 µH. Fig. 3.8(a) confirms that the center frequency has moved down to ≈ 61 kHz and the
high-frequency cut-off point has dropped to ≈ 128 kHz. Fig. 3.8(b) shows the measured
amplitude response of a 1-µF capacitor in series with 69-µH of inductance terminating in a
50-Ω load, once again confirming the theoretical expectations.
As a final step in the measurement process, typical protective back-to-back zener diodes were
inserted in parallel with the secondary winding and caused no visible change in the amplitude
response. Furthermore a 10-µF electrolytic capacitor was placed in series with the final output
stage (often required at the input port of instruments) and this also had no visible effect on
the amplitude response.
39
(a) (b)
Fig. 3.8 (a) Measured amplitude response of a 1:1 transformer with 0.22-µF
capacitor in series with primary terminal, total series inductance of 31 µH and
terminating in 25-Ω load. b) Measured amplitude response of a 1:1 transformer
with 1-µF capacitor in series with primary terminal, total series inductance of 69
µH and terminating in 50-Ω load. Measured values correspond well with (3.6) to
(3.8).
If the only function that a transformer provides over and above filtering, is that of protection,
its necessity in the coupling circuit must be seriously considered, as transformers are bulky,
heavy and expensive compared to other passive components. If it is possible to design a
system such that impedance levels (at the carrier frequency) are nearly equal, a coupling
transformer becomes redundant. In such a case it is suggested that coupling equipment be
designed as symmetrical, bi-directional passive filters with additional protective circuitry [13].
A simple series-resonant LC-R band-pass filter (which behaves the same as the circuit in Fig.
3.5) can be designed with (3.6) to (3.8). This coupling circuit could include varistors on the
power-line side and zener diodes on the communication side for protection.
3.3 Conclusion
T
HE design of coupling and de-coupling circuitry is often neglected at the expense of
reduced signal levels and increased noise levels. This chapter summarized the fact that
coupling capacitors or de-coupling inductors don’t merely pass or block signals, but
that their filtering characteristics are quite dependent on the loads into which the waveforms
terminate. The functioning of a coupling transformer was also investigated, and it was shown
that the leakage inductance together with a series capacitor, form a series-resonant band-pass
filter that can be modeled as a simple LC-R circuit. Design equations were given and these
were verified by measuring the voltage transfer function of the coupling circuit for different
values of L, C and R.
40
3
3.4 References
[1] D.S. Humpherys, Analysis, Design and Synthesis of Electrical Filters, Englewood Cliffs: Prentice-Hall, 1970.
[2] W.K. Chen, Passive and Active Filters: Theory and Implementations, New York: John Wiley & Sons, 1986.
[3] H.-K Podszeck, Carrier Communication over Power Lines, 4th Edition, New York: Springer-Verlag, 1972.
[6] K.C. Abrahams, “A novel high-speed PLC communication modem,” IEEE Transactions on Power Delivery, vol.
7(4), October 1992, pp. 1760-1768.
[7] J. Millman, H. Taub, Pulse, Digital and Switching Waveforms, New York: McGraw-Hill, 1965.
[8] W. Downey, “Central control and monitoring in commercial buildings using power line communications,”
Proceedings of the 1st International Symposium on Power-Line Communications and its Applications (ISPLC’97), Essen,
Germany, 2-4 April 1997, pp.115-119.
[9] Philips Components, Soft Ferrites Data Handbook MA01, 1991.
[10] K.K. Clarke, D.T. Hess, Communication circuits: analysis and design, Reading: Addison-Wesley, 1971.
[11] F.E. Rogers, The Theory of Networks in Electrical Communications and Other Fields, London: Macdonald & Co,
1957.
[12] F.E. Terman, Electronic and Radio Engineering, 4th edition, New York: Mc Graw-Hill, 1955.
[13] G. Telkamp, “A low-cost power-line node for domestic applications,” Proceedings of the 1st International
Symposium on Power-Line Communications and its Applications (ISPLC’97), Essen, Germany, 2-4 April 1997, pp. 32-
35.
*****
41
4
Chapter
CHAPTER 4
Design of a Bi-Directional Coupler
A
LTHOUGH some authoritative texts on power-line communication suggest separate
coupling circuits for transmitting and receiving data, it was envisaged that a properly
designed coupling circuit would be able to adapt modem impedances and power-line
impedances to be roughly equal. Equality in terminating impedances further suggests that filter
characteristics would be similar in both the transmit- as well as receive directions, yielding a bi-
directional coupler. Chapter 4 shows that this is indeed the case, and that a single coupling
circuit can be used to transmit and receive data from a power-line modem.
After giving a concise introduction of the suggested coupling circuit and its operation, this
chapter focuses on practical design steps for an impedance-adapting transformer-capacitor
coupling circuit. The design was based on the European CENELEC standard which
prescribes maximum signal levels as well as allowed carrier frequency bands. Although the
design included an in-depth transformer design, the information is still applicable and useful if
a commercial coupling transformer is to be selected. The practical fine-tuning of transformer
magnetizing- and leakage inductances and its influence on the pass-band are also considered.
The designed coupling circuit was constructed and experimental measurements show good bi-
directional symmetry when transmitting and receiving data. Lastly, the impact of a fluctuating
power-line impedance level on coupler bandwidth is illustrated, and concluding design
tradeoffs are discussed.
42
DESIGN OF A BI-DIRECTIONAL COUPLER
4
4.1 Introduction
A properly designed coupler can adapt impedance levels to be roughly equal – thus
facilitating bi-directional band-pass transmission of power-line communication
signals. However, fluctuating power-line impedance levels cause the bandwidth of a
coupler to fluctuate, and therefore needs to be considered in such a design.
T
RANSFORMER-CAPACITOR coupling circuits are used extensively in low-voltage
power-line communications, mainly because i) the transformer provides galvanic
isolation from the power-line network and ii) the transformer acts as a limiter when
saturated by high-voltage transients. As power-line access impedances are generally very low,
poor power transfer is achieved because of the mismatch between modem impedance and
power-line impedance. However, in this chapter it is shown that a properly designed coupling
circuit can i) adapt the impedance level of a certain modem to a chosen typical impedance
level of the power line and ii) subsequently be used as a bi-directional coupler for two-way
communication.
The step-by-step design of a 1:7 coupling transformer is described, and amplitude response
measurements clearly indicate the bi-directional symmetry. Also, it is shown that fluctuating
power-line impedance values cause the bandwidth of the coupling filter to fluctuate when a
signal is transmitted. In the receiving direction though, the bandwidth of the coupling filter
stays constant as the modem input impedance stays constant. Lastly, some tradeoffs are
discussed in order to choose a typical minimum power-line impedance value for design
purposes.
4.2 Circuit
In Chapter 3 [1] a typical coupling circuit (shown in Fig. 4.1(a)) was analyzed and shown to
behave as a simple series-resonant band-pass filter at frequencies close to the series resonant
point. The equivalent circuit is shown in Fig. 4.1(b). The center frequency of this band-pass
filter is at the series-resonant point
1 [Hz] (4.1)
fR =
2π LC
where L refers to the series inductance and C refers to the series capacitance. L typically
consists only of the leakage inductance referred to primary, but can be enlarged with an
external series inductor. The bandwidth of the filter is determined by the respective low-
frequency and high-frequency –3-dB cut-off points
1
f LF = [Hz] (4.2)
2πRC
43
4
and
R
f HF = [Hz] (4.3)
2πL
where R refers to the terminating resistance (RP when transmitting and R’M when receiving).
(a) (b)
Fig. 4.1 (a) Suggested coupling circuit and (b) equivalent circuit for frequencies
close to the series-resonant point. ZP and ZM refer to the power-line and modem
terminating impedances respectively whereas R’M represents the referred modem
impedance (to the power-line side).
Take note that ZP, the equivalent impedance of the power-line network (see Fig. 4.1) was not
taken into account during measurements in [1]. Typical impedance values of 0.1 Ω to 2 Ω for
frequencies in the CENELEC (Comité Européen de Normalisation Electrotechnique) B, C, and D
bands have been reported [2], [3]. This low impedance value implies that a large portion of the
available signal power is dissipated unnecessarily. Even if it is possible to improve (raise) the
power-line impedance using network conditioning, it is still good practice to design the input
impedance of the modem to equal the power-line access impedance as seen from the modem.
This would mean a splitting of transmitter current (and power) at the series-resonant point
and thus only 3 dB of power is lost.
If the modem input impedance cannot be designed to equal the power-line impedance, it is
necessary to equalize the impedance levels by using a step-up transformer. A transformer is a
powerful impedance-transforming device, as any impedance is reflected by the square of the
winding ratio. If a transformer is used to equalize the terminating impedances on either side of
Fig. 4.1(b), another advantage results: the coupling network can now be used as a bi-
directional coupler with the same filter characteristics from both sides. The step-by-step
design of such a transformer coupling circuit for the CENELEC bands is discussed in the
following section. See Fig. 4.2(a) below, showing a photograph of this prototyped coupling
transformer.
44
4
4.3 Design
4.3.1 Frequency specifications
The bi-directional coupler will be designed for the CENELEC B, C, and D bands [3], [4], for
the frequency range ≈90 kHz to ≈150 kHz. Thus 90 kHz would be the worst-case frequency
for core considerations (see (4.6)) and 150 kHz would be the frequency where copper losses
would be a maximum.
4.3.2 Impedance levels / winding ratio
Although the power-line access impedance fluctuates between 0.2 Ω to 2 Ω and more [2], [5],
a 1-Ω power-line impedance level will be used for this design, as it satisfies various tradeoffs
(discussed in Section 4.5). It is also assumed that a 50-Ω modem impedance needs to be
adapted to the 1-Ω impedance of the power line. This can be done using a 1:7 transformer
winding ratio. The 50-Ω modem impedance appears as (1/7)2⋅50 ≈ 1 Ω on the power-line
side, whereas the 1-Ω power-line impedance appears as (7/1)2⋅1 ≈ 49 Ω on the modem side.
(a) (b)
Fig. 4.2 Photographs of (a) the coupling transformer designed in this section and
(b) a coupling transformer designed for a 250-kHz-to-500-kHz carrier frequency.
4.3.3 Maximum voltage levels
According to the CENELEC specification, a maximum signal level of 116 dBµV (≈ 0.63
VPEAK) is allowed for the B, C, and D bands. Although this figure represents a measured value,
the measurement setup consists of a balanced voltage-divider network, which implies that only
50 % of the true signal is measured [3], [4]. Thus a maximum voltage of 122 dBµV (which
represents ≈ 1.26 VPEAK or 0.89 VRMS) can be injected into the power-line network and this
represents a voltage of 8.81 VPEAK or 6.23 VRMS on the modem side of the 1:7 transformer.
4.3.4 Maximum power and current
In order to calculate the maximum power level, a minimum power-line impedance of 0.25 Ω
is assumed. This gives a maximum power throughput of 3.17 W and maximum currents of
3.56 ARMS and 0.51 ARMS for the power-line side and modem side respectively.
45
4
4.3.5 Core
An E20 core was chosen, manufactured from Mn-Zn ferrite material [6]. This core set has the
following properties: core cross-sectional area Ae = 31.2 mm2, flux path-length le = 42.8 mm,
intrinsic permeability µi = 3800, and saturation flux density BSAT = 400 mT.
4.3.6 Current density
Billings [7] shows that the current density for a certain rise in conductor temperature is given
by the empirical equation
J ≈ α⋅AP -0.125 [A/cm2] (4.4)
where α is a constant for a certain core shape and temperature rise, and AP refers to the area
product of the core (product of core and window areas). The core under consideration has an α
of 4.5 (for ∆T = 30 0C) and an area product AP = 0.08923 cm4. Equation (4.4) yields a current
density of 6.1 A/mm2 for a 30 0C rise in conductor temperature. Reference [8] lists values of α
for various other core shapes and temperatures.
4.3.7 Skin effect
The optimum strand diameter is typically chosen between δ and 2δ, depending on the
proximity effect and other design factors. For copper at a temperature of 50 0C, the
penetration depth δ at a certain frequency f is
1 0.0699 [m] (4.5)

δ= =
πσµ 0 µ R f S f
if the relative permeability µR of copper is assumed as 1. Symbol σ represents the conductivity

whereas µ0 is the permeability of free space. Equation (4.5) yields a δ of 0.18 mm for 150
kHz. The optimum strand diameter can also be obtained from graphs [7] and was 0.36 mm for
this design. The closest (lower) gauge available at the time of construction was a 0.25-mm
diameter copper wire with a cross-sectional area of 0.049 mm2.
4.3.8 Number of strands
The number of insulated strands necessary to obtain a Litz-wire bundle with sufficient cross-
sectional area (for the current density of 6.1 A/mm2) can now be calculated as approximately
twelve strands for the power-line side and two strands for the modem side.
4.3.9 Number of turns
A typical window fill-factor is 0.5, and it is thus expected that only 8 power-line side turns
together with 56 modem-side turns would fit onto the transformer’s bobbin.
4.3.10 Flux density
Although the power waveform can have an overwhelming influence on the flux density (see
Chapter 5 or [9]), the superimposing flux density contribution BM of the communication
waveform on the total flux density is
46
4
VM ≈ 9 mT (4.6)
BM =
2πfAN
which is valid for a sine wave, subscript M referring to peak values on the modem side.
Symbols f, A, and N refer to frequency, core cross-sectional area, and number of modem-side
turns respectively. As BSAT ≈ 400 mT (depending on temperature), less than 3 % of the
available flux density is required by the communication waveform.
4.3.11 Leakage inductance
The value of the transformer’s leakage inductance (which is dependent on the number of
turns and winding geometry) can now be calculated (see Chapter 5 or [9] for details):
µ0 T
LL = ⋅ ⋅ (a + b + 3c) ⋅ N 2 ≈ 294 nH (4.7)
3 w
This equation is valid for one primary and one secondary layer, i.e. no interleaving
(sandwiching). T and w represent the mean turn length and window width respectively and a, b
and c refer to the primary winding thickness, secondary winding thickness and insulation
thickness respectively. N refers to the number of turns of the side to which the total leakage
inductance needs to be referred to.
4.3.12 Enlarging of leakage inductance
According to (4.3), L = 294 nH would yield a high-frequency cut-off point of approximately

540 kHz, which should preferably be lower for the CENELEC B, C, and D frequency bands.
A cut-off frequency of 300 kHz, one octave above the upper limit, would be more
appropriate. This cut-off frequency requires a leakage inductance of 530 nH, which can be
accomplished by letting the single-layer primary winding on the power-line side protrude out
of the transformer window. See Section 5.3.
This enlarged leakage inductance was later measured and fine-tuned to approximately 530 nH
using a HP 4284 precision LCR meter. The primary winding layer had to protrude
approximately 3 mm from both sides of the transformer in order to obtain a 530-nH leakage
inductance. (Alternatively, an external 230-nH inductor can be inserted in series with the
transformer primary terminal.) Also see Chapter 5 or [9] for equations and alternative methods
to enlarge/reduce leakage inductance.
4.3.13 Series capacitor
The required series capacitor can be calculated by either choosing a center point for the band-
pass filter (see (4.1)) or by choosing a low-frequency cut-off point (see (4.2)). A low-frequency
cut-off point of 50 kHz was chosen, also roughly an octave below the lower limit of the
CENELEC bands. Equation (4.2) yields a series capacitor of 3.2 µF, and together with the
series leakage inductance has a series resonant point of 122 kHz. As this capacitor carries the
power waveform’s voltage, it needs to be rated at the peak voltage, including a safety factor of
50 % or more.
47
4
4.3.14 Magnetizing inductance
The value of the magnetizing inductance LM does not influence the pass-band of the coupling
circuit, but is required to determine the filtering of the low-frequency power waveform. See
Chapter 5 or [9]. The total inductance (magnetizing plus leakage inductance) of a transformer
is given by
µ 0 µ e Ae [H] (4.8)
LT = ⋅ N 2 ≡ AL ⋅ N 2
le
where subscript e refers to effective. Ae and le represent the core cross-sectional area and flux
path-length respectively and N represents the number of turns (of the side to which the
inductance needs to be referred to). AL is called the inductance factor of a certain core set, and
is supplied by manufacturers in order to easily calculate inductance values for different air gaps
and number of turns. To obtain a more accurate value of magnetizing inductance, a
transformer’s leakage inductance needs to be subtracted from the above value. For this design,
LM ≈ 100 µH.
4.3.15 Check combined flux density levels
Because of its low frequency, the power waveform can easily saturate the transformer,
depending on how well it is filtered. See Chapter 5 or [9] for details and equations. For this
design, it is expected that the power waveform would cause a flux density of ≈ 43 mT.
Combined with the communication waveform, a total flux density of 52 mT is expected – well
below the saturation flux density of 400 mT.
4.3.16 Reducing magnetizing inductance
If the combined flux density level of the transformer is not sufficiently low, the magnetizing
inductance of the transformer can be made smaller by introducing an air-gap to lower LM. A
smaller magnetizing inductance would help to filter low frequencies more effectively. Refer to
Chapter 5 or [9] for more details.
4.4 Bi-Directional Measurements

In order to verify that the constructed coupling circuit (see Fig. 4.2(a)) does function as
designed, the amplitude response (⏐H(jω)⏐ vs. frequency) was measured with an HP 3577B 5-
Hz-to-200-MHz network analyzer with 50-Ω output and input impedances. Fig 4.3 shows the
band-pass filter effect of the coupler prototype in the receiving as well as transmitting
direction for an emulated 1-Ω power-line impedance. Take note of the bi-directional
symmetry – the coupling circuit shows very similar filter characteristics in both receiving and
transmitting directions.
48
4
The measured center frequency and cut-off points correspond to the calculated values of fR ≈
122 kHz, fLF ≈ 50 kHz and fHF ≈ 300 kHz. Another interesting observation is the +15-dB and
–19-dB peaks when receiving and transmitting respectively. This is caused by the 1:7
transformer winding (voltage) ratio that implies approximately +17 dB when receiving and –
17 dB when transmitting. It must be emphasized that the amplitude response as was measured
by this instrument, represents voltage ratios expressed as decibels and not power levels. Thus
the +17 dB and –17 dB levels should be taken as reference levels, as indicated in Fig. 4.3(a)
and (b).
(a) (b)
Fig. 4.3 Measured amplitude response of the coupling circuit designed for
CENELEC B, C, and D band when (a) receiving a signal from the power-line side
and (b) transmitting a signal into a 1-Ω resistor (emulating a power-line access
impedance). Take note of the bi-directional symmetry. The difference in reference
levels is caused by the 1:7 transformer winding ratio.
Another coupling circuit for frequencies between 250 kHz and 500 kHz (typical in USA /
Japan), was designed and constructed. See Fig. 4.2(b). Because of the higher cut-off
frequencies, smaller reactive component values are required – see (4.2) and (4.3). Cut-off
frequencies of 167 kHz and 750 kHz were chosen, yielding C ≈ 0.95 µF and LL ≈ 210 nF and
a center point of ≈ 350 kHz. This coupling circuit also showed good symmetry and correlated
well with the model in Fig. 4.1(b). Measured amplitude responses for the receiving and
transmitting directions are shown in Fig. 4.4(a) and (b) respectively.
49
(a) (b)
Fig. 4.4 Measured amplitude response of the coupling circuit designed for 250
kHz to 500 kHz when (a) receiving a signal from the power-line side and (b)
transmitting a signal to the power-line side (1-Ω resistor). Take note of the bi-
directional symmetry and different reference levels caused by the 1:7 transformer
winding ratio.
4.5 Influence of Fluctuating Power-Line Impedance

Equation (4.1) shows that the terminating impedance does not influence the resonant point
(center point) of the band-pass filter. However, (4.2) and (4.3) show that a varying terminating
impedance does influence the respective cut-off points and therefore the bandwidth of the
filter. Fig. 4.5 shows the influence of terminating impedance on the amplitude response of a
1:1 coupling circuit designed around a 80-kHz center point. In Fig. 4.5(b), the narrower pass-
band caused by a lower terminating impedance is evident.
When receiving a signal, this problem is not encountered: as the modem input impedance and
transformer winding ratio is constant, the referred modem impedance stays constant. This
implies a constant amplitude response when receiving a signal through the coupling circuit.
Fig. 4.3(a) and Fig. 4.4(a) would therefore stay constant during power-line impedance
fluctuations because the modem terminating (input) impedance stays constant. However,
when transmitting a communication signal into the power line, the power-line terminating
impedance varies with time. As stated before, typical impedance values are 0.2 Ω to 2 Ω for
frequencies in the CENELEC B, C, and D bands [2], [3]. Impedance values as low as 0.1 Ω
and as high as 10 Ω are also not uncommon [2], [5].
50
(a) (b)
Fig. 4.5 Measured amplitude response for a 1:1, 80-kHz coupling circuit with (a)
50-Ω terminating impedance and (b) 25-Ω terminating impedance. Take note how
the bandwidth is reduced by a lower terminating impedance. For this circuit L ≈
19 µH and C ≈ 0.22 µF.
The lowest power-line impedance value would result in the narrowest pass-band whilst the
maximum power-line impedance value would result in the widest pass-band. It is therefore
best to design the filter for the lowest impedance values expected for a certain power-line
network. If the power-line impedance (and transmitting bandwidth) then increases, no harm is
done: a wider pass-band usually has no detrimental effect when transmitting a communication
signal. Rather, when a signal is received, it is of paramount importance that proper filtering
takes place to improve signal-to-noise ratios.
The minimum power-line impedance that is designed for, must be chosen wisely, taking
various interdependent factors into consideration. The lowest value of power-line impedance
will determine the maximum power level (and thus transformer core size) for a certain voltage
level. This power rating then influences the conductor cross-sectional area, and number of
turns that can physically fit into the transformer window. Consequently, the number of turns
influences magnetizing and leakage inductance as well as flux density. This inter-dependency
of parameters is typical of any transformer design, and usually a number of design iterations
have to be done before a satisfactory result is obtained.
51
4
4.6 Conclusion
T
HE first step when designing a coupling circuit would be to decide on a typical
minimum power-line impedance value. The coupling transformer’s turns ratio is then
chosen to adapt the modem impedance to be roughly equal to this chosen power-line
impedance. (This facilitates symmetrical filtering in both directions.) The maximum power
output of the modem would be matched to the coupling transformer’s maximum power in
order not to overload the coupling circuit. If the power-line impedance drops below the
chosen design value, the communication signal is still transmitted at the modem’s maximum
power level, but below the maximum voltage levels as determined by government regulations.
Thus transformer size and cost (for optimum performance at a chosen minimum power-line
impedance level) is traded against reduced performance at lower power-line impedance levels.
4.7 References
[1] P. A. Janse van Rensburg, H. C. Ferreira, “Coupling circuitry: understanding the functions of different
components,” Proceedings of the 7th International Symposium on Power-Line Communications and its Applications
[2] H. C. Ferreira, H. M. Grové, O. Hooijen, and A. J. H. Vink, Power Line Communication (in Wiley
Encyclopaedia of Electrical and Electronics Engineering), New York: John Wiley & Sons, 1999, pp. 706-
716.
[3] K. Dostert, Powerline Communications, Upper Saddle River: Prentice Hall PTR, 2001.
[4] “Signalling on Low-Voltage Electrical Installations in the Frequency Range 3 kHz to 148.5 kHz”,
CENELEC Standard EN 50065-1, 1991.
[5] O. Hooijen, Aspects of Residential Power Line Communications, Aachen: Shaker Verlag, 1998.
[6] Soft Ferrites Data Handbook MA01, Philips Components, 1991.
[7] K. H. Billings, Handbook of Switchmode Power Supplies, New York: Mc Graw-Hill, 1989.
[8] W. T. McLyman, Transformer and Inductor Design Handbook, New York: Marcel Dekker, 1978.
[9] P. A. Janse van Rensburg, H. C. Ferreira, “The role of magnetizing and leakage inductance in transformer
coupling circuitry,” Proceedings of the 8th International Symposium on Power-Line Communications and its Applications
*****
52
5
Chapter
CHAPTER 5
Coupling Transformer Impedances
A
S the magnetizing and leakage inductances of the coupling transformer impact on
the filter characteristics of the coupling circuit, these internal transformer
impedances are discussed in more depth in this chapter. Although Chapter 5 is
aimed towards the design of a coupling transformer, a reader intending to utilize a commercial
coupling transformer will still gain invaluable knowledge, helping with the selection of an
appropriate coupling transformer.
At the outset, the fundamental definition of inductance is discussed, after which a simplified
transformer equivalent circuit is derived. Next, theoretical design equations are derived for a
transformer’s leakage inductance. As the leakage inductance impacts directly on the upper cut-
off point of the coupling circuit’s pass-band, the designer is shown how to increase or
decrease leakage inductance for a certain coupling circuit design. Core saturation, a very
important topic, is discussed next, as transmitted data is lost during a core saturation event.
Simulations as well as practical measurements show that the magnetizing inductance (with
coupling capacitor) causes a 40-dB/decade filter slope at low frequencies. Therefore the
magnetizing inductance should be designed to adequately filter the power waveform – in order
to prevent core saturation. Magnetizing inductance design equations are presented, and
practical ways to increase or reduce this inductance are discussed.
53
COUPLING TRANSFORMER IMPEDANCES
5
5.1 Introduction
Together with the coupling capacitor, the internal impedances of a coupling

transformer determine the pass-band of the filter circuit. Leakage inductance
impacts on the upper cut-off point, while magnetizing inductance influences the
lower cut-off point.
I
N Chapters 3 and 4 (or see [1], [2]) it was shown that accurate modeling and design of
leakage inductance is necessary in order to design an effective transformer-capacitor
coupling circuit. Furthermore, transmission through this coupling circuit is ruined if the
transformer saturates, and for this reason accurate magnetizing inductance design is necessary.
Although theoretical understanding is required for such a coupling transformer design,
practical knowledge is often necessary to alter a design parameter. Therefore various hands-on
methods to increase or decrease these impedance values are discussed in the following
sections. Portions of Section 5.2 and 5.3 were also presented in [3].
5.2 Summarized Theory of Inductance

5.2.1 Definition of inductance
Fundamentally, inductance is defined by the well-known equation
diL
vL = L ⋅ [V] (5.1)
dt
and therefore inductance L is the constant of proportionality between instantaneous voltage
and the time derivative of instantaneous current for a certain inductor. Furthermore, Faraday’s
Law states that induced voltage v can be expressed as
dφ dλ
v=N⋅ ≡ [V] (5.2)
dt dt
where λ (number of flux-linkages) is defined as Nφ – the product of N (number of turns) and

φ (coupled flux). By equating (5.1) and (5.2), inductance can be expressed in terms of λ as
dφ dλ
L= N⋅ = [H], (5.3)
di di
a non-linear equation which is material dependent. If λ is plotted against i for a certain

inductor, the resulting curve exhibits a similar shape to the applicable B-H curve of the
magnetic material used (see Fig. 5.1 below).
54
Fig. 5.1 Plot of λ vs. i to show non-linearity of inductive behavior. Dotted line
indicates assumption that magnetic material is linear up to saturation level.
In the rest of this chapter, a linear λ-i relationship (with no risidual magnetic effects) will be
assumed in order to make the modeling and mathematics manageable. It is therefore assumed
that the inductor core has a constant µR. (This assumption is accurate only for air-core
inductors or cored inductors where excitation is minimal.) Based on these assumptions,
inductance can now be written as
φ λ
L= N⋅ = [H], (5.4)
i i
the ratio between the resultant flux-linkage and the current associated with this flux-linkage.
5.2.2 Inductance in transformers
In a transformer core, two main flux paths exist, and the associated fluxes are called mutual
(or magnetizing) flux and leakage flux. These two flux components are responsible for a
transformer’s magnetizing and leakage inductance respectively. For an inductor though, the
magnetizing and leakage flux components add to make up the total inductance. Refer to Fig.
5.2.
In Fig. 5.2, φN1 represents the total flux induced by MMF N1i1. A certain component of φN1
‘leaks’ back to the bottom leg of the core and is labeled φL1. This primary leakage flux φL1
therefore does not couple with the secondary winding N2. The remaining (mutual) primary
flux φM1 however does couple with N2 and induces opposing MMF N2i2 with flux φN2.
Similarly, only φM2 (the magnetizing- or mutual portion of φN2) couples with N1, to oppose φM1
whereas the uncoupled flux φL2 leaks back through the core window and surrounding space.
Using (5.4), the magnetizing inductance LM of a transformer can now be written as
φM φM1 φM 2 [H] (5.5)

LM = N 1 ⋅ = N1 ⋅ − N1 ⋅
i1 i1 i1
where φM = φM 1 − φM 2 represents the resultant mutual or magnetizing flux, as the two mutual flux
components oppose each other.
55
Fig. 5.2 Simplified representation of a transformer’s flux paths and MMF

distribution across the core window.
As the leakage flux components do not oppose each other, but rather cause a superimposing
effect, the resultant ‘leakage inductance’ LL can be written as
φL φ L1 φ L2 [H] (5.6)
LL = N 1 ⋅ = N1 ⋅ + N1 ⋅ ≡ LL1 + LL 2
i1 i1 i1
where LL1 and LL2 represent the primary and secondary ‘leakage inductances’ respectively.
Assuming negligible losses for now, and a linear λ-i relationship – implying that flux-linkage
Nφ = λ = Li (5.4), the voltages across the primary and secondary windings can now be
obtained by taking the time derivative of applicable flux-linkages for each winding – see (5.2).
The different (voltage-equivalent) terms may also be referred to the other side of the
transformer using the appropriate turns-ratio, yielding:
di1 ⎛ di N di ⎞ [V] (5.7)

v1 = LL1 + LM ⎜⎜ 1 − 1 ⋅ 2 ⎟⎟
dt ⎝ dt N 2 dt ⎠
di 2 ⎛ di N di ⎞ [V] (5.8)
v 2 = − LL 2 − LM ⎜⎜ 2 − 2 ⋅ 1 ⎟⎟
dt ⎝ dt N 1 dt ⎠
These two equations, with resistors added in order to model losses, give rise to the widely
used transformer equivalent circuit shown in Fig. 5.3 below. For an ideal transformer, the
series components need to be negligibly small, whereas the parallel components need to
exhibit very large values.
56
Fig. 5.3 General transformer equivalent circuit, assuming linear core material, and
losses modeled with resistors.
The secondary leakage inductance can be referred to the primary side, but this causes
inaccuracies as the magnetizing current does not flow through LL2. For a large magnetizing
inductance though, the secondary leakage inductance may be safely referred to the primary
side using
2
⎛N ⎞
LL 2 (referred ) = ⎜⎜ 1 ⎟⎟ ⋅ LL 2 [H]. (5.9)
⎝ N2 ⎠
5.3 Leakage Inductance Considerations

5.3.1 Influence on filtering
The leakage inductance of a coupling transformer forms a series-resonant band-pass filter with
the external (series) coupling capacitor. For frequencies close to resonance, the effect of the
magnetizing inductance is negligible. See Section 5.4 as well as [1] or [2].
5.3.2 Calculation of leakage inductance
An alternative approach to calculate inductance, is to use two energy equations, namely the
energy stored in an inductor (5.10) as well as the energy stored in a certain three-dimensional
space through volume-integration of the B-H product (5.11):
Li 2
W= [J] (5.10)
2
1 µ µ [J] (5.11)
W=
2v
∫ BHdv = 0 r ∫ H 2 dv
2 v
By equating (5.10) and (5.11), and solving for L, inductance can be expressed as
µ 0µ r [H]. (5.12)
L= ∫H
2
dv
i2 v
As an example, the leakage inductance of the following planar transformer in Fig. 5.4 will be
calculated. This transformer has separately wound primary and secondary windings (no
interleaving). Also see [4] for a different winding configuration.
57
5
Assuming no fringing of flux outside the windings, and also that µR = 1 for volumes v1, v2, and
v3 (volumes taken up by windings N1, N2, and isolation between windings respectively) the
leakage inductance can be calculated. The total ‘leakage’ energy stored in volume v1, is given by
µ0 2
W1 = ∫ dW1 where dW1 = H ⋅ dv1 [J]. (5.13)
v
2
1
As graphically depicted in Fig. 5.4 the MMF in the primary winding (volume v1) can be written
as
x N1i1 for 0 < x < a

H= ⋅ [At/m]. (5.14)
a w
Fig. 5.4 Cross-section through a planar transformer and MMF distribution across
window.
Take note that w, the width of a planar winding, represents the effective magnetic path-length
of the leakage flux. It is now also assumed that the winding structure, viewed from the top,
has a cylindrical shape with average turn-length T in order that dv = A⋅dx = Tw⋅dx. It follows
that
µ 0 ⎛ x N1i1 ⎞
a 2
W1 = ∫ ⋅⎜ ⋅ ⎟ ⋅ Tw ⋅ dx [J] (5.15)
0 2 ⎝a w ⎠
⇒ W = µ 0 ⋅ ⎛⎜ N1i1 ⎞⎟ ⋅ Tw ⋅ x 2 ⋅ dx
2 a
[J] (5.16)
1
2 ⎝ aw ⎠
∫
0
⇒ W = µ 0 ⋅ ⎛⎜ N1i1 ⎞⎟ ⋅ Tw ⋅ 1 ⋅ a 3
2
[J] (5.17)
2 ⎝ aw ⎠
1
3
⇒ W1 = µ 0 aT ⋅ ( N1i1 ) 2 [J]. (5.18)

6w
Similarly, it can be shown that the ‘leakage’ energy W2 stored in winding N2 is
58
5
µ 0 bT
⋅ ( N1i1 )
2
W2 = [J] (5.19)
6w
as N1i1 ≈ N2i2. Furthermore, the energy stored in volume v3 (isolation space between windings)
can be calculated by
µ 0 ⎛ N1i1 ⎞
c 2
W3 = ∫ ⋅⎜ ⎟ ⋅ Tw ⋅ dx [J], as (5.20)
0 2 ⎝ w ⎠
H=
N1i1 [A⋅t/m] for volume v3. Therefore (5.21)
w
µ 0 cT
⋅ ( N1i1 )
2
W3 = [J]. (5.22)
2w
The total ‘leakage’ energy stored in the transformer can be obtained by adding the individual
energy components, and therefore the total energy stored in the ‘leakage inductance’ LL of the
transformer is given by
1
WL = W1 + W2 + W3 = LLi 2 [J]. (5.23)
2
From (5.18), (5.19), (5.22) and (5.23) it follows that
µ 0 N 12 T
LL = ⋅ ⋅ ( a + b + 3c ) [H] (5.24)
3 w
where µ0 = 4π x 10-7 is the permeability of free space and N1 the number of turns on the
primary winding. T represents the average turn-length whereas w denotes core window width.
As shown in Fig. 5.4, symbols a, b and c represent the thickness of the primary, secondary and
insulation layers respectively.
5.3.3 Enlarging leakage inductance
If a transformer’s leakage inductance is to be utilized as a filter inductor (instead of using an

external, discrete inductor) its value often has to be increased. Depending on the design
constraints, various options may be followed, all yielding a more non-ideal transformer:
• Increase the insulation thickness c between the primary and secondary layers – refer to
(5.24).
• Increase the number of turns N1 of the primary (and secondary to keep winding ratio the
same). See (5.24). This implies higher current densities in the windings.
• Insert ferrite (or other material) with permeability µR between primary and secondary
windings to drastically boost the leakage flux. Equation (5.24) then becomes
µ0 N12T
LL = ⋅ ⋅ (a + b + 3µ R c) [H]. (5.25)
3 w
• Allow the external winding to protrude out of the core. See Fig. 5.5 below.
59
Fig. 5.5 Cross-section through transformer (top view) to indicate winding

protruding from core. External loop has a radius R.
If the original leakage inductance of the transformer without the protruding loop is negligible,
the leakage inductance with protruding loop is given by an empirical equation [5]
2
N1 R 2
LL = [µH] (5.26)
230 R + 250 H
where R and H represent the radius and height of the protruding loop respectively.
5.3.4 Reducing leakage inductance
For high-frequency power-line communication applications, the leakage inductance has to

resonate with the series coupling capacitor at this higher carrier frequency. This implies that
smaller reactive component values are required. Although it would be the easiest and most
economical to reduce the coupling capacitance value, the leakage inductance may have to be
reduced in certain cases. Some general guidelines are given below:
• Choose an appropriate core type that encloses the windings to minimize external leakage.
• Minimize thickness of insulating layer between windings.
• Interleave layers of primary and secondary windings to lower the effective window MMF.
Using the method described in Section 5.3, it can be proven that the theoretical leakage
inductance of an interleaved transformer (where the primary winding is divided into q layers
and secondary winding is divided into q+1 layers) is given by
µ 0 N12T
LL = ⋅ ⋅ ( a + b + 6c ) [H] (5.27)
6q w
Comparing (5.27) with (5.24) it is evident that leakage inductance can be reduced by almost 75
% by merely dividing the primary winding into q = 2 layers and the secondary winding into
three layers.
60
5
5.4 Magnetizing Inductance Considerations
5.4.1 Influence on saturation and filtering
As the total flux density in a coupling transformer core is made up of the communication
waveform’s flux density superimposed on the filtered power waveform’s flux density, the
influence of both have to be considered in order to prevent saturation. The influence of the
filtered power waveform on core saturation can be gauged by its v/f factor, also called the
volt-time product. Equation (5.28) refers to Fig. 5.6(a) and summarizes the superimposing
influence of the filtered power waveform (subscript P) and the referred communication / modem
waveform (subscript M) on the total flux density (subscript T), and is valid for sinusoidal
waveforms only.
1 ⎛V ' V' ⎞
BT = BP + BM = ⋅ ⎜⎜ P + M ⎟⎟ [T] (5.28)
2πAN ⎝ fP fM ⎠
The ratio of the filtered power waveform’s v/f value to the communication waveform’s v/f
value will determine how the two waveforms share (or exceed) the total flux density, as core
cross-sectional area (A) and number of primary turns (N) are fixed for a certain design. Take
note that in Fig. 5.6(a) and equation (5.28) the voltage V’P refers to the peak filtered power
voltage, whereas V’M refers to the peak referred communication voltage. Furthermore, the
minimum value of fM has to be used in order to calculate the maximum flux density.
(a) (b)
Fig. 5.6 (a) More comprehensive model for the coupling circuit designed in
Chapter 4 [2] and (b) simulated amplitude response for this coupling circuit. Here
the reference level is 0 dB as the 1:7 transformer winding ratio of the coupling
circuit of Chapter 4, was not included in the simulation.
As an example, the coupling circuit designed in Chapter 4 [2] will be discussed. This coupling
circuit was designed to transmit and receive any 122-dBµV or 1.26-VPEAK CENELEC B, C, or
D signal (95 kHz-148.5 kHz) to and from a 50-Hz, 220-VRMS, 1-Ω power line. Thus on the
61
5
power-line side of the coupling transformer, the received communication waveform will have
a maximum v/f factor of only 1.26 V / 95 kHz = 13 µV/Hz, yielding a flux density of BM ≈ 9
mT for the chosen core set. For the filtered power waveform to have a v/f figure (and flux
density contribution) of say twice this value, it needs to be filtered to a voltage of only 1.3
mVPEAK or by 107.6 dB. Alternatively, a power waveform attenuation level of –100 dB would
represent 62.2 µV/Hz (BP ≈ 43 mT) whereas –90 dB would represent 196.7 µV/Hz (BP ≈
136.2 mT), all still within acceptable limits as BSAT = 400 mT for this ferrite core.
At a first glance, it seems as if this band-pass coupling circuit designed in Chapter 4, would
only be able to filter the power waveform by a maximum of 60 dB, as only three decades is
available for filtering, and the roll-off slope is –20 dB/decade. However, it must be
emphasized that the model in Chapter 4 [2] is only valid for frequencies close to the resonant
point. A more comprehensive model of the coupling circuit that includes the magnetizing
inductance, is shown in Fig. 5.6(a). The voltage transfer function of this circuit (in the
receiving direction) is expressed by
V 'P s 2 RCLM [–] (5.29)

= 3
VP s CLM LL + s RC ( LM + LL ) + sLM + R
2
where R represents the referred modem impedance. From (5.29) it is clear that the power-line
impedance has no influence on the voltage transfer function in the receiving direction. The
amplitude response of (5.29) was simulated and is shown in Fig. 5.6(b). As can be seen in Fig.
5.6(b), the simulated amplitude response of the coupling circuit under consideration correlates
well with the simple band-pass model of Chapter 4 for frequencies above 1.6 kHz, but a roll-
off slope of –40 dB/decade is evident below 1.6 kHz. This is due to the influence of the
parallel magnetizing inductance at lower frequencies. At 1.6 kHz, the reactance of this
coupling transformer’s 100-µH magnetizing impedance nearly equals the 1-Ω (referred) load
resistor, and at frequencies below 1.6 kHz, its reactance dominates the parallel branch
(LM⎪⎢R), attenuating these frequencies more than for a pure resistor.
It is not necessary to simulate the circuit in Fig. 5.6(a) for every design, as the magnitude of
the filtered power waveform voltage for a certain frequency (voltage across the magnetizing
inductance) can be calculated using the voltage divider rule:
X LM X R [V] (5.30)
V 'P = ⋅ VP
X C + X LL + ( X LM X R )
Subscripts LL and LM refer to leakage and magnetizing inductance respectively, whereas

subscript R refers to the modem impedance referred to the power-line side. Once V’P is
known, its contribution to the flux density can be calculated with (5.28).
The simulated amplitude response (Fig. 6(b)) of the coupling circuit designed in Chapter 4 [2]
was also verified practically with an HP 3577B 5-Hz-to-200-MHz network analyzer with 50-Ω
output and input impedances – one measurement is shown in Fig. 5.7. Comparing Fig. 5.7
with Fig. 5.6(b), it can be seen that the prototyped coupling circuit (designed in Chapter 4),
62
5
performed as expected, but deviates from the simulated coupling circuit for low frequencies.
This is largely due to the limited power output of the measuring instrument – the coupling
transformer core under measurement could not be driven deeply into saturation for a reliable
low-frequency measurement.
From the 50-Hz cursor measurement in Fig. 5.7 though, it is evident that for the specific
coupling circuit designed in Chapter 4 [2], a total attenuation of 100 dB is attainable. Thus the
power waveform is expected to occupy 43 mT of flux density whilst the communication
waveform is predicted to utilize 9 mT. The sum of these two figures, 52 mT, will be the total
maximum flux density in the coupling transformer’s core. This value will be the same,
regardless of whether the coupling circuit is used to receive data or transmit data: although the
voltages on the modem side is seven-fold those on the power-line side, the number of turns is
also seven-fold and thus cancels the influence of the former.
Fig. 5.7 Measured response of the coupling circuit designed in Chapter 4. The
cursor was positioned at ≈ 50 Hz and measured –83 dB (+17 dB reference level).
5.4.2 Calculation of magnetizing inductance
It was shown in Section 5.2 that for a linear magnetic core,
φ [H], where φ = ℑ = Ni
L=N⋅ [Wb]. (5.31)
i ℜ ℜ
Furthermore reluctance ℜ is defined by
le [A⋅t/Wb] (5.32)
ℜ=
µAe
where µ represents permeability, le is the effective flux path length and Ae is the effective core
cross-sectional area. From (5.31) and (5.32) follows that total inductance
N 2 Ae
L = LM + LL = µ0 µ R ⋅ [H]. (5.33)
le
63
5
Equation (5.33) gives reasonably accurate values for magnetizing inductance as leakage
inductance LL is typically negligible. Leakage inductance can be subtracted from L to obtain
more accurate answers for LM.
5.4.3 Enlarging / reducing magnetizing inductance
From (5.33) it is clear that magnetizing inductance can be increased by either increasing µR,
increasing A or decreasing l (using a different transformer core), or alternatively increasing N
(thinner conductors or bigger core). It must be mentioned that µR is also dependent on the air
gap if a core set is used. In this case µR can be boosted by decreasing the air-gap, or even
clamping the core set together. Conversely, magnetizing inductance can be decreased by
inserting an air-gap between certain legs of a core. This technique of fine-tuning the air-gap
for a certain magnetizing inductance is widely used. Take note that the air-gap doesn’t alter
BSAT but only lowers µe to skew the B-H curve [6]. Although BSAT stays the same, a smaller
magnetizing inductance would help to filter low frequencies more effectively, as the –40
dB/decade slope starts at a higher frequency. Refer to Fig. 5.6(b) and Fig. 5.7.
5.5 Conclusion
I N order to design an effective coupling circuit, the internal impedances have to be

designed properly, as they can have a detrimental effect on coupling by either i) filtering
the communication signal or ii) allowing the power waveform to saturate the coupling
transformer. Therefore the fundamental definition of inductance, as well as leakage inductance
and magnetizing inductance were discussed. Design equations were derived and various
practical methods to increase and decrease these inductances were discussed.
5.6 References
[2] P. A. Janse van Rensburg, H. C. Ferreira, “Step-by-step design of a coupling circuit with bi-directional
transmission capabilities,” Proceedings of the 8th International Symposium on Power-Line Communications and its
Applications (ISPLC’04), Zaragoza, Spain, 31 March - 2 April 2004, pp. 238-243.
[3] P. A. Janse van Rensburg, “Tegnologie vir ’n geïntegreerde gedeeltelik-serieresonante mutator,” M.Ing
thesis, Department of Electrical and Electronic Engineering, Rand Afrikaans University, South Africa, 1997.
[4] A. S. Langsdorf, Theory of Alternating-Current Machinery, 2nd ed., New York: McGraw-Hill, 1955.
[5] P. C. Theron and J. A. Ferreira, “The zero voltage switching partial series resonant converter,” IEEE
Transactions on Industry Applications, vol. 31(4), July/August 1995, pp. 879-886.
[6] J. G. Kassakian, M. F. Schlecht, G. C. Verghese, Principles of Power Electronics, Reading, Massachusetts:
Addison-Wesley, 1991.
*****
64
6
Chapter
CHAPTER 6
Impedance Adaptation
I N Chapter 4, the design of a coupling circuit was based on a purely parallel power-line
impedance. It was assumed that the adaptation would be sufficient if the transformer
winding ratio equalized the modem impedance with the (theoretically) parallel power-line
impedance. However, in a real-world scenario, transmitters and receivers are always separated
by series cables, and also connected to parallel loads via series cables. Take note that series and
parallel impedances have conflicting adaptation requirements. A perfect channel would consist
of an infinitely small series impedance, and an infinitely large parallel impedance, leaving a
direct connection between matched transmitter and receiver impedances.
Thus the goal of Chapter 6 is to investigate this trade-off by means of a simple T-model for
the power line. The main reason for choosing a simple model, is to linguistically describe any
power-line condition, and utilize the results for improved transmission. Although the model
may seem very simple initially, it later proved that some results could only be obtained with
numerical analysis – even after reducing the number of parameters by placing certain
constraints on the circuit parameters.
Simulation results are summarized by presenting these as three-dimensional graphs, showing

optimum winding ratios for symmetrical as well as asymmetrical power-line conditions.
Finally, it is recommended that manufacturers separate couplers from modems and provide a
range of couplers to the consumer, with appropriate guidance in order to choose good
winding ratios.
65
IMPEDANCE ADAPTATION
6
6.1 Introduction
Power-line cables (series impedances) and loads (parallel impedances) impose

conflicting impedance adaptation requirements. This tradeoff is investigated in this
chapter.
B
EFORE this project, hardly any significant literature on the topic of low-voltage
power-line communications coupling existed. To this extent, the design of said
coupling circuitry was investigated and documented in [1] (alternatively, see Chapter
4). However, for simplification, only a parallel power-line (load) impedance was assumed in
Chapter 4 and [1]. In this chapter, the optimization of a crucial coupling circuit parameter is
investigated, namely the winding ratio of transmitter and receiver coupling transformers.
As a starting point, the power line is modeled by taking into consideration a two-fold lumped
series impedance as well as a lumped parallel impedance. Although the accuracy of such a
simple, lumped model might be questioned, it is shown in Chapter 7 that results are
sufficiently accurate to determine good winding ratios – yielding high gains. Bear in mind that
winding ratio is not continuously variable, yet influences impedance transformation by its
square. For instance, the difference between selecting a 2:1 transformer over a 1:1 transformer
implies an increase in impedance transformation from 1:1 to 4:1 – this is equivalent to a 300
% increase in impedance transformation, reflected to the primary and a decrease of 75 %,
reflected to the secondary. Hence even when model accuracy is poor, useful results can be
obtained.
Another reason for using only one lumped parallel impedance and a two-fold lumped series
impedance, is to qualitatively describe the power-line channel. The series impedance will be
classified using the terms short line, average line and long line. The parallel impedance will be
classified by the terms severely loaded, normally loaded and lightly loaded. See Section 6.2 for a
suggested classification, conveniently chosen for this study. These categories further facilitate
the interpretation of results (see Section 6.4), and will also aid with practical recommendations
(Section 6.5).
Given the constraints and assumptions mentioned in Section 6.2, this chapter attempts to
answer the questions stated below:
1) Is it worthwhile attempting impedance matching?
2) Of what order should practical winding ratios be?
3) Is there a good, versatile receiver winding ratio, aRX ?
4) Is there a good, versatile transmitter winding ratio, aTX ?
66
6
5) Can power transfer be optimized with aTX = aRX?
6) Can aTX and aRX be optimized independently?
7) How should these findings be interpreted and implemented?
6.2 Model for Transmitter-Receiver Chain

Fig. 6.1 shows a lumped-parameter model for the power-line channel, with a transmitter-
receiver pair coupled through the channel. For this study, the transmitter output impedance
ZTX as well as the receiver input impedance ZRX equal 50 Ω. The coupling transformers have a
winding ratio of aTX :1and 1:aRX for the transmitter- and receiver-side respectively.
Fig. 6.1 Transmitter-receiver pair coupled through the power-line channel. A

simplified model with two lumped series impedances and one lumped parallel
impedance, represents the power line for frequencies below 1 MHz.
It is assumed that both coupling circuits are properly designed (see Chapters 4 and 7), and that
frequencies of interest occupy the flat band of the coupling circuits. This implies that the
coupling capacitors as well as magnetizing and leakage inductances can be neglected, and that
both coupling circuits can be modeled as ideal transformers (see Chapter 5 or [2]). For this
study, the upper frequency limit was chosen as 1 MHz, thus including the ≈ 100-kHz
CENELEC (Comité Européen de Normalisation Electrotechnique) [3] bands as well as ≈ 500-kHz
systems, typical for Japan and USA. Consequently, the flat band of said coupling circuitry has
to stretch to ≈ 1 MHz in order for results and conclusions to be valid.
As the interconnecting network, the power-line channel is modeled by a series impedance ZS

consisting of a left- or transmitter-side component ZL as well as a right- or receiver-side
component, ZR. This series impedance represents the power-line cabling, and is largely related
to the characteristics and length of the power line, including the return line. It is assumed that
power-line cabling distances are short compared to the applied wavelength, and that a
lumped-parameter network would adequately model the cabling at these wavelengths.
Practically, these cable distances should be less than ≈ 400 m at 100 kHz, and less than ≈ 80
m at 500 kHz, for reasonable results. Another assumption, is that cable inductance dominates
behavior below 1 MHz, as power lines typically terminate in low-impedance loads [4], [5].
However, as explained in Section 6.1, even when these assumptions are violated to a certain
67
6
degree, results can in most cases still be used for improved power transfer, as the impedance
transformation discontinuities for different transformer winding ratios, typically far outweigh
the inaccuracies of the model.
The parallel power-line impedance in Fig. 6.1 (ZII), is used to model loads that are connected
to the grid. It is well known that the power-line access impedance as measured between live
and neutral at a specific point, can vary considerably with location and time. Surprisingly,
some reported power-line impedance measurements, e.g. [6] - [8] were done between live and
earth, or neutral and earth. Although potentially useful for power-line communications, the
earth wire forms a separate network not to be confused with the real power network.
Presumably, the main reason for not measuring between live and neutral in these cases, was
for the fear of damaging expensive measuring equipment.
Citing true access impedance measurements for the CENELEC bands, [9] and [10] list typical
values ranging between 0.2 Ω and 2 Ω. Reference [11] further reports outdoor measurements
ranging between 0.1 Ω and 18 Ω for A-band CENELEC frequencies, while [5] and [12] report
values between 1 Ω and 20 Ω for frequencies below 100 kHz.
As a starting point for this study, moderate values of ZS and ZII were chosen, ranging between
1 Ω and 10 Ω. However, as these impedances may represent inductances or capacitances, and
the 50-kHz to 500-kHz portion of the band already represents a decade change in frequency,
it was decided to investigate all values of ZS and ZII between 0.1 Ω and 100 Ω. Phase angle is
not taken into consideration for any of the impedances though, only impedance magnitudes
are utilized. Therefore, strictly speaking, apparent power is the optimizing criterion, and
complex impedance matching is not attempted.
To facilitate the discussion of the influence of the power-line characteristics on performance,

the various ranges of series and parallel power-line impedances are categorized in Table 6.1.
As the simulations span over three orders for both ZS and ZII and the results are further
presented in logarithmic scale, it was decided to specify three classes for each. See Table 6.1.
Table 6.1
Classification of ZS and ZΙΙ Values
ZS (equal to ZL + ZR)
‘Short Line’ ‘Average Line’ ‘Long Line’
0.1 to 1 Ω 1 to 10 Ω 10 to 100 Ω
ZII
‘Lightly Loaded’ ‘Normally Loaded’ ‘Severely Loaded’
100 to 10 Ω 10 to 1 Ω 1 to 0.1 Ω
68
6
Take note that these impedance values – and thus the classification thereof – is totally
dependent on frequency. For CENELEC frequencies, average line might refer to a 20-m cable,
whereas the same cable could act as a long line at 500 kHz. The same holds for parallel loads,
and for a certain project, a custom classification of series and parallel impedances should be
done, taking into consideration impedance range, typical impedance values, as well as the
influence of frequency on these.
6.3 Simulation Details

In order to simulate the behavior of the circuit in Fig. 6.1, and investigate the requirements for
impedance matching, all components were referred to the power-line side. See Fig. 6.2.
Fig. 6.2 Simulated circuit with all components referred to the power-line side. ZTX
and ZRX both equal 50 Ω, but their referred values are influenced by the choice of
winding ratio. ZS (sum of ZL and ZR) as well as ZII are varied between 0.1 Ω and
100 Ω.
A Thévenin equivalent circuit for Fig. 6.2, can be obtained by treating Z'RX as the load, and
considering the power-line impedances as part of the source. See Fig. 6.3. Disconnecting Z'RX,
and short-circuiting E'TX, it follows that ZTHEV seen from the receiver side, can be expressed as
Z THEV = Z R + ( Z II ( Z L + Z 'TX )) [Ω] (6.1)
or
⎛ ( Z )( Z L + Z 'TX ) ⎞
Z THEV = Z R + ⎜⎜ II ⎟⎟ [Ω]. (6.2)
⎝ Z II + ( Z L + Z 'TX ) ⎠
Also, ETHEV, the Thévenin equivalent no-load voltage can be found using the voltage-divider
rule and is effectively the voltage across ZII when Z'RX is disconnected:
⎛ Z II ⎞
ETHEV = E 'TX ⎜⎜ ⎟ [V] (6.3)
⎝ II
Z + ( Z L + Z ' TX ) ⎟⎠
69
Fig. 6.3 Thévenin equivalent circuit for Fig. 6.2. Expressions for ZTHEV and ETHEV
are given in (6.2) and (6.3).
ETHEV can now be used to determine the apparent power SRX delivered to the receiver for
matched or unmatched conditions:
2
V ' 2RX 1 ⎛ Z ' RX ETHEV ⎞
S RX = = ⎜⎜ ⎟⎟ [VA] (6.4)
Z ' RX Z ' RX ⎝ Z ' RX + Z THEV ⎠
The comparative performance of various combinations of winding ratios and power-line
impedances can now be plotted after choosing a reference input, say ETX = 1 VRMS for all
cases, and using the following formula to convert to the dBm scale:
⎛ S ⎞
S RX = 10 log10 ⎜ RX ⎟ [dBm] (6.5)
⎝ 0.001 ⎠
6.4 Simulation Results

6.4.1 Is it worthwhile attempting impedance matching?
In order to become familiar with the influence of winding ratio on power transfer under
different power-line conditions, various combinations of aTX = aRX, ZL = ZR, and ZII were first
investigated. Fig. 6.4 shows two sets of these simulation results for certain winding ratios.
The benefit of impedance transformation is already evident in this first glimpse of the
behavior of the circuit. By comparing in Fig. 6.4(a), the 4:1:4 pair with the default 1:1:1 pair, it
is evident that 20 dB of gain is possible over a large range of ZS values, yet a smaller loss (< 10
dB) is sacrificed for ZS values larger than 25 Ω. Similarly, in Fig 4(b), by choosing a 2:1:2 pair
over a default 1:1:1 pair, gains of ≈ 5 dB are possible for the biggest range of ZS values.
Using Table 6.1 to summarize the results in Fig. 6.4, it can be stated that, for 50-Ω modems,
impedance transformation is important for a severely- to normally-loaded power-line channel (low
to average values of ZII), but that it can be detrimental in some cases, e.g. a power-line channel
with long lines (large values of ZS), especially if the power line is lightly loaded.
70
6
6.4.2 Of what order should practical winding ratios be?
If one fixed (general purpose) winding ratio has to be implemented in hardware, it is clear
from Fig. 6.4 that practical winding ratios would typically be between the extremes of 1:1 and
say 10:1, depending on the expected range of power-line conditions. However, if winding
ratios are to be selected for very specific conditions, these limits change. The following
sections show that optimum winding ratios, given the limitations of this study, are as low as
0.6:1 and as high as 25:1.
(a) (b)
Fig. 6.4 Performance (SRX vs. ZS) for certain winding ratios where (a) ZII = 0.4 Ω
(more typical for CENELEC frequencies) and (b) ZII = 25 Ω (more typical for
higher frequencies). All results are normalized to a 1-VRMS transmitter output. Fig.
6.4(a) shows that impedance transformation is important for low values of ZII
(severely-loaded power lines), but both Fig. 6.4(a) and 4(b) indicate that
impedance transformation can be detrimental for larger values of ZS (long lines).
6.4.3 Is there a good, versatile receiver winding ratio?
From the Thévenin equivalent circuit in Fig. 6.3, for maximum apparent power transfer to the
receiver, Z'RX should equal ZTHEV, which yields the following equation after substituting
winding ratios into (6.2):
Z RX ⎛ ( Z II )( Z L + ( Z TX aTX
2
)) ⎞
Z ' RX = = Z + ⎜ ⎟ [Ω] (6.6)
⎜ ⎟
⎝ Z II + Z L + ( Z TX aTX ) ⎠
2 R 2
a RX
This Thévenin condition for maximum apparent power transfer as expressed in (6.6), has two
unknown variables, aTX and aRX, given a certain power-line condition (certain ZL, ZR and ZII).
Either aTX or aRX can be solved, if the other is known or kept fixed. Equation (6.7) expresses
the solution for a matched aRX in terms of a known aTX value and non-referred impedance
values:
⎛
a RX = Z RX ⋅ ⎜⎜
{
Z II + ( Z TX aTX2
) + ZL } ⎞
⎟ [–] (6.7)
{ } ⎟
⎝ {Z II + Z R } ⋅ ( Z TX aTX ) + Z L + {Z II ⋅ Z R } ⎠
2
71
6
This condition was simulated for the whole range of power-line conditions as explained in
Section 6.2. Fig. 6.5 summarizes one set of results showing the required matched aRX value,
together with its performance at a parallel power-line impedance of 0.4 Ω (more applicable to
CENELEC frequencies).
(a) (b)
Fig. 6.5 (a) Optimum receiver winding ratio aRX and (b) corresponding
performance for certain fixed transmitter winding ratios with ZII = 0.4 Ω.
Compare with Fig. 6.4(a) to see drastic improvement in performance.
Take note that in Fig. 6.5, the same four transmitter winding ratios were chosen as for Fig.
6.4, and represent non-optimum winding ratios. In Fig. 6.5(b), for each data point, aRX was set
to the optimum value though, as obtained from Fig. 6.5(a). Comparing Fig. 6.5(b) to Fig.
6.4(a), the improvement is evident. All four curves show improved performance, and the
penalty when using higher winding ratios, for a power-line channel with long lines is diminished.
In the search for a good, versatile receiver winding ratio, Fig. 6.5(a) does provide some clues.
Fig. 6.5(a) shows that a certain optimum aRX is compatible with a range of transmitter winding
ratios (aTX) as the four curves coincide for most values of ZS. This observation suggests that
optimized aTX and aRX values are largely independent of each other. Fig. 6.5(a) says nothing
about performance though, and Fig. 6.5(b) as well as Fig. 6.6(b) suggest that aRX = 2 or aRX =
4 are possible candidates for a good, versatile receiver winding ratio, depending on the
applicable power-line conditions.
6.4.4 Is there a good, versatile transmitter winding ratio?
Similarly, the optimum transmitter winding ratio aTX can be obtained by manipulating the
Thévenin condition for maximum apparent power transfer given in (6.6):
⎛
aTX = Z TX ⋅ ⎜⎜
{
Z II − ( Z RX a RX2
) − ZR } ⎞
⎟ [–] (6.8)
{ } ⎟
⎝ {Z II + Z L } ⋅ ( Z RX a RX ) − Z R − {Z II ⋅ Z L } ⎠
2
Equation (6.8) yielded complex results for most combinations of aRX, ZL, ZR, and ZII, as the
subject of the square root was negative for most of these combinations. Subsequently
numerical analysis had to be employed to maximize SRX while varying aTX, keeping other
72
6
parameters fixed. The Newton-Raphson method, incorporated into Microsoft ExcelTM was
used in all cases. For each optimum solution of aTX, it was confirmed that impedance
matching had been accomplished by comparing values of Z'RX and ZTHEV. A converging
solution was obtained for all combinations of aRX, ZL, ZR, and ZII, and one set of results is
presented in Fig. 6.6, for ZII = 25 Ω.
Take note that for this higher 25-Ω parallel power-line impedance (ZII) – more typical for
higher frequencies – optimum winding ratios are lower. Another interesting result in Fig.
6.6(a), is that an almost-constant aTX = 2 is required for maximum power transfer, regardless
of whether the signal is received through a 1:1 or a 2:1 transformer. Fig. 6.6(b) was obtained
by choosing the same four winding ratios as before (but in this case for the receiver) while
setting aTX to its optimum value for each data point. In order to compare performance, Fig.
6.6(b) has to be evaluated against Fig. 6.4(b) as both represent the same power-line conditions.
In the quest for a good, versatile transmitter winding ratio, Fig. 6.6(a) does show that aTX = 4
seems like middle ground, but again, this graph does not indicate performance, and
performance is not linearly related to winding ratio. By consulting Fig. 6.4, Fig. 6.5(b), and Fig.
6.6(b), it seems like aTX = 2 is also possibly a candidate for good, all-round transmitter
performance (subject to power-line conditions).
(a) (b)
Fig. 6.6 (a) Optimum transmitter winding ratio aTX and (b) corresponding
performance for certain fixed receiver winding ratios with ZII = 25 Ω. Numerical
analysis had to be employed to optimize aTX as (6.8) yielded complex results for
most data points. Compare with Fig. 6.4(b) to see improved performance.
6.4.5 Can power transfer be optimized with aTX = aRX?
Another goal of this study, is to determine the viability of optimized, equal winding ratios at
transmitter and receiver under different power-line conditions. If a constraint of equality is
placed on the two winding ratios (i.e. aTX = aRX = a), an analytical expression can be found
(using (6.6)) that describes the winding ratio of two identical coupling transformers that will
73
6
yield an optimized result, given this constraint. The fourth-order polynomial in (6.9) expresses
this requirement for optimized, equal winding ratios:
a 4 {Z II Z R + Z R Z L + Z L Z II } + a 2 {Z TX ( Z II + Z R ) − Z RX ( Z II + Z L )} − Z TX Z RX = 0 (6.9)
The roots of (6.9) also had to be found using numerical analysis, i.e. searching for the value of
a, that would make the left-hand side of (6.9) equal to zero (within a certain tolerance). This
procedure was repeated for each data point under consideration, until a converging result was
obtained. Next, matching was confirmed by comparing values of Z'RX and ZTHEV. In Fig. 6.7,
the complete set of optimized symmetrical results is displayed in a three-dimensional graph.
Fig. 6.7 Optimum symmetrical winding ratio, aTX = aRX = a for a symmetrical
power-line environment.
Take note that above set of solutions for symmetrical winding ratios only provide optimum
power transfer to the receiver under the constraint aTX = aRX = a. Further investigation
confirmed that these optimized results only correspond with matched results if the power line
is symmetrical, i.e. ZL = ZR.
6.4.6 Can aTX and aRX be optimized independently?
It was anticipated that for an asymmetrical power-line channel, i.e. ZL ≠ ZR, different values of
aTX and aRX would be required for optimum power transfer. In order to examine the
characteristics of an asymmetrical power-line channel, the total series power-line impedance ZS
was divided in a 10:1 ratio, i.e. ZL : ZR = 10:1 or ZL : ZR = 1:10. It proved that these
independently optimized, matched winding ratio values indeed were not identical to the
symmetrical ones obtained through (6.9).
Results from (6.9) were utilized though as the starting point for each numerical analysis, as
fewer iterations are required when the initial value is closer to the final optimized value. For
each data point, aTX was optimized first while keeping aRX fixed at the symmetrical optimum
74
6
from (6.9). Next, aRX was optimized while keeping aTX fixed at the new asymmetrical optimum.
For certain data points, the procedure had to be repeated twice before a matched, converging
solution was accomplished. Matching was confirmed by comparing ZTHEV with Z'TX for each
data point. Fig. 6.8 shows the optimum transmitter winding ratio aTX, where the load is far
from the transmitter (ZL : ZR = 10:1).
Fig. 6.8 Optimum asymmetrical winding ratio, aTX for asymmetrical power-line
channel with transmitter far from the load (ZL : ZR = 10:1).
Comparing Fig. 6.8 to Fig. 6.7, the shape of the two planes show much similarity, but for
smaller values of ZII (normally- to severely-loaded power line) it is evident that required winding
ratios are slightly lower. This is due to the larger (asymmetrical) distance from this ZII,
requiring lower winding ratios for impedance transformation. Further, Fig. 6.8 has a larger
plateau for long lines (large values of ZS) as the series impedance ZS now has a more dominant
influence on transmitter performance when ZL : ZR = 10:1.
Conversely, in Fig. 6.9, the optimum winding ratio is shown when the load is in close vicinity
to the transmitter. In this case, it is clear that for a normally- to severely-loaded power line, ZII
prescribes the need for more stringent impedance transformation.
Still, for a lightly-loaded power line (large values of ZII), this plane is almost identical to Fig. 6.7
and Fig. 6.8. This effect is largely due to the fact that a lightly-loaded power line is more prone to
be symmetrical than a heavily-loaded power line, as the series impedances now dominate.
75
Fig. 6.9 Optimum asymmetrical winding ratio, aTX for asymmetrical power-line
channel with transmitter close to the load (ZL : ZR = 1:10).
Since a ZL : ZR ratio of 10:1, seen from the transmitter side implies a ZL : ZR ratio of 1:10 seen
from the receiver side, it turned out that the optimum receiver winding ratio aRX far from the
load, is identical to the optimum transmitter winding ratio aTX far from the load. The same
holds for close vicinity to the load. This is a very important observation, as it implies that one
optimized coupling circuit can be used for both transmitting and receiving data, as was
suggested in [1].
6.5 Recommendations
H OW should these findings be interpreted and implemented? Fig. 6.2 clearly shows
that an ideal power-line channel would have an infinitely large parallel impedance
combined with zero series impedance, yielding a perfect match between transmitter
and receiver. Likewise, from the results shown in Fig. 6.4, right through to the results of Fig.
6.9, it is clear that a tradeoff exists between satisfying the requirements for enlarging the
relative value of the dominating load impedance ZII (seen from the Tx or Rx side) by using a
winding ratio larger than 1 – yet keeping the relative value of cable inductance ZS (seen from
the Tx or Rx side) small enough not to negate the first improvement.
Although the optimum winding ratio results of this chapter range from below 1:1 to over 20:1,
it is wise to be cautious against over-compensation in case the loading of the power line is
reduced. In summary, the following is recommended for 50-Ω modems with transmission
frequencies below 1 MHz:
76
6
1) Instead of a default 1:1 winding ratio, (no impedance transformation) implement a
conservative 2:1 winding ratio and gain up to 5 dB for most power-line conditions.
2) If close proximity to a normally- to severely-loaded power line is guaranteed, use larger winding
ratios, e.g. a = 4. Gains of up to 10 dB and more are possible.
3) Only if a modem is positioned at a lightly-loaded remote location, implement low winding

ratios, e.g. 1:1 or even 0.5:1 in extreme cases.
It is further proposed that manufacturers separate couplers from modems, and provide a
range of couplers so that the end-user may choose the correct coupler (winding ratio) for a
certain power-line condition. Limited information may be provided as a guideline to the
consumer. Alternatively, various rooms of a residential property may be classified, and
assigned good winding ratios as is done in the next chapter.
It must be mentioned that results must be interpreted for a certain modulation frequency, as
this impacts directly on ZS. Higher frequencies require more conservative (smaller) winding
ratios, as cable inductance has a stronger impact on transmission. Conversely, at lower
frequencies, load impedances can be transformed more aggressively by using larger winding
ratios.
Also take note that the power-line voltage has a direct impact on typical values of ZII.
Contrasting a 110-V network against a 220-V network, ohm-values for the same appliance
(power level) would be four times more severe (smaller). Therefore 110-V networks would
require more drastic impedance adaptation – i.e. higher winding ratios.
77
6
6.6 References
[1] P. A. Janse van Rensburg, H. C. Ferreira, “Design of a bidirectional impedance-adapting transformer
coupling circuit for low-voltage power-line communications,” IEEE Transactions on Power Delivery, vol. 20(1),
January 2005, pp. 64-70.
1998, pp. 15-20.
[6] J. R. Nicholson, J. A. Malack, “RF impedance of power lines and line impedance stabilization networks in
conducted interference measurements,” IEEE Transactions on Electromagnetic Compatibility, vol. EMC-15(2),
May 1973, pp. 84-86.
[7] J. A. Malack, J. R. Engstrom, “RF impedance of United States and European power lines,” IEEE
Transactions on Electromagnetic Compatibility, vol. EMC-18(1), February 1976, pp. 36-38.
[8] P. J. Kwasniok, M. D. Bui, A. J. Kozlowski, S. S. Stuchly, “Technique for measurement of powerline
impedances in the frequency range from 500 kHz to 500 MHz,” IEEE Transactions on Electromagnetic
Compatibility, vol. 35(1), Feb. 1993, pp. 87-90.
[9] H. C. Ferreira, H. M. Grové, O. Hooijen, and A. J. H. Vink, Power Line Communication (in Wiley Encyclopaedia of
Electrical and Electronics Engineering), New York: Wiley, 1999, pp. 706-716.
[11] O. Hooijen, Aspects of Residential Power Line Communications, Aachen, Germany: Shaker Verlag, 1998.
[12] M. Tanaka, “High frequency noise power spectrum, impedance and transmission loss of power line in Japan
on intrabuilding power line communications,” IEEE Transactions on Consumer Electronics, vol. CE-34(2), May
1988, pp. 321-326.
*****
78
7
Chapter
CHAPTER 7
Experimental Verification
T
HE power-line model of Chapter 6 was deliberately chosen as a simple model, to
facilitate qualitative interpretation of the impedance adaptation results, and therefore
the applicability of said model had to be tested experimentally. The T-model of
Chapter 6 is further based on the assumption that cable distance is proportional to cable
impedance – therefore certain frequency limits had to be set in order to ensure the assumed
inductive behavior of the cabling. In this chapter, bi-directional couplers (designed according
to Chapter 4) are therefore tested for frequencies below 1 MHz – both in the laboratory as
well as in a live 220-V residential power-line network. Test results for these couplers show
good back-to-back fidelity (< 1 dB), and the impedance adaptation strategy developed in
Chapter 6 proved very successful, yielding significant transmission gains.
Firstly, a controlled laboratory setup was used in order to directly verify the experimental
results of Chapter 6. This laboratory setup mimicked the T-model with two extension cords as
series impedances with a calibrated load between them as a parallel load impedance. Various
extension cord lengths were combined with two applicable load values, being 3 ohm and 25
ohm, in order to validate the simulation results of Chapter 6.
After successful validation with a laboratory setup, the three-dimensional set of (symmetrical)
optimum winding ratios of Chapter 6 was used to classify typical residential rooms such as a
kitchen, bedroom, etc. and a good winding ratio was specified for each room. Live 220-V
measurements in a residential power-line environment proved very successful, and step-wise
improvement can be seen as the transmitter- and receiver winding ratios approach the
suggested optimum value.
Finally, a novel dual-impedance coupler was prototyped and tested for switching loads in a
live 220-V residential power-line environment. Results show that such a dual coupler can aid
in the mitigation of switching power-line loads by means of providing two optimized winding
ratios for on and off conditions. Another inherited advantage results – power-line impedance
levels are more stable, facilitating more accurate impedance adaptation from other outlets.
79
EXPERIMENTAL VERIFICATION
7
7.1 Introduction
To substantiate the theoretical content of this thesis, this chapter describes the
deployment of a range of prototype couplers. Firstly, the theoretical simulation
results of Chapter 6 are verified in a controlled, disconnected power-line network.
Subsequently the successful utilization of impedance adaptation strategies is
demonstrated for a live 220-V residential network. Typical residential rooms are
classified for impedance adaptation and a novel dual-impedance coupler is tested for
harsh switching loads.
C
OUPLING losses, including impedance mismatches have been underestimated and
neglected by power-line communications researchers and manufacturers. Further, the
variations in network topology combined with unknown cable and load impedances,
obscure the requirements for impedance adaptation. These factors have caused manufacturers
and researchers to default to guessed impedance levels and fixed winding ratios under all
conditions.
After investigating coupling circuit design and impedance adaptation mechanisms, the final
step of this project was to implement properly designed couplers in conjunction with
impedance adaptation strategies in a real-world power-line network. However, the simulation
results of Chapter 6 was first verified in a controlled laboratory environment, before
attempting to utilize these results for an uncharacterized power-line environment. Some core
simulation results of Chapter 6 are presented below, as these will be verified in this chapter.
Also, important deductions are made in order to apply these in the live 220-V trials.
The lumped-parameter power-line T-model that was simulated in Chapter 6 [1] is shown in
the dotted box, Fig. 7.1, with a transmitter-receiver pair coupled through the channel. For
both the simulation results in Chapter 6 as well as the experimental measurements in this
chapter, the transmitter output impedance ZTX as well as the receiver input impedance ZRX
equal 50 Ω. The coupling transformers have a winding ratio of aTX:1 and 1:aRX for transmitter
and receiver side respectively.
Fig. 7.1 Transmitter-receiver pair coupled through the power-line T-model of

Chapter 6.
80
7
It is essential that both coupling circuits are properly designed (see Chapter 4 or [2]), to exhibit
a flat response for frequencies of interest. For all measurements in this chapter, the upper
frequency limit was chosen as 1 MHz, thus including the ≈ 100-kHz CENELEC bands [3] as
well as ≈ 500-kHz systems, typical for Japan and USA.
It is also assumed that power-line cabling distances are short compared to the applied
wavelength, and that a lumped-parameter network would adequately model the cabling at
these wavelengths. Practically, these cable distances should be less than ≈ 400 m at 100 kHz,
and less than ≈ 80 m at 500 kHz, for reasonable results. Another assumption, is that cable
inductance dominates behavior below 1 MHz, as power lines typically terminate in low-
impedance loads [4], [5]. However, as was explained in Chapter 6, even when these
assumptions are violated to a certain degree, results can in most cases still be used for
improved power transfer, as the impedance transformation discontinuities for different
transformer winding ratios, typically far outweigh the inaccuracies of the model.
One reason for choosing a simple power-line model in Chapter 6, was for qualitative
description of the appropriate impedances. This would further enable an end-user of a power-
line communications coupling device to linguistically choose good couplers for different
power-line conditions – see Section 7.4. In Table 7.1, a convenient classification is shown that
will also be used for practical implementation in this chapter.
Table 7.1
Classification of ZS and ZΙΙ Values
ZS (equal to ZL + ZR)
‘Short Line’ ‘Average Line’ ‘Long Line’
0.1 to 1 Ω 1 to 10 Ω 10 to 100 Ω
ZII
100 to 10 Ω 10 to 1 Ω 1 to 0.1 Ω
Fig. 7.2 shows the first set of simulation results to be verified, each for a certain power-line
condition, and four symmetrical pairs of couplers. In Section 7.3, these simulation results are
verified experimentally in a laboratory setup. Using Table 7.1 to summarize the results in Fig.
7.2, it can be stated that, for 50-Ω modems, impedance transformation is important for a
severely- to normally-loaded power-line channel (low to average values of ZII), but that it can be
detrimental in some cases, e.g. a power-line channel with long lines (large values of ZS),
especially if the power line is lightly loaded.
81
(a) (b)
Fig. 7.2 Performance (SRX vs. ZS) for certain winding ratios where (a) ZII = 0.4 Ω
(severely loaded) and (b) ZII = 25 Ω (lightly loaded). Results are normalized to a 1-
VRMS transmitter output. Impedance transformation is important for low values of
ZS (short lines), but can be detrimental for larger values of ZS (long lines). Points
E through H further refer to specific power-line conditions that are experimentally
verified in this chapter (Section 7.3).
Further, the central result of Chapter 6 [1] is expressed by the fourth-order polynomial in
(7.1). The roots of (7.1) had to be solved using numerical analysis (Newton-Raphson method).
Thus a certain optimum winding ratio a had to be found that would make (7.1) zero for each
power-line condition (parameters ZL, ZR and ZII) – ZTX and ZRX being 50 Ω. See Chapter 6 or
[1] for more details.
a 4 {Z II Z R + Z R Z L + Z L Z II } + a 2 {Z TX ( Z II + Z R ) − Z RX ( Z II + Z L )} − Z TX Z RX = 0 (7.1)
In Fig. 7.3, a top-view of the three-dimensional solution (Fig. 6.7) for polynomial (7.1) is
shown where ZL = ZR, i.e. a symmetrical power-line topology. The points labeled A through G
in Fig. 7.3, describe specific power-line conditions and are experimentally verified in this
chapter, Section 7.3.
Fig. 7.3 neatly summarizes the simulation results of Chapter 6 [1] – showing the required
winding ratios for optimum transmission between two 50-Ω modems. With the guidance of
the contour lines in Fig. 7.3, the suggested winding ratio choice for each of the nine power-
line conditions (of Table 7.1) is listed in Table 7.2.
82
Fig. 7.3 Top view of Fig. 6.7 – a three-dimensional solution of (7.1) for a
symmetrical power-line topology. Bold grid-lines indicate the classification of
power-line impedances into the nine categories of Table 7.1. Contour lines
facilitate the choice of optimum winding ratio for each category. Points A through
H further refer to specific power-line conditions that are experimentally verified in
this chapter (Section 7.3).
Table 7.2 may now be used to classify any power outlet in order to choose a good winding
ratio for said outlet. Also, different regions (or rooms) of a building may be classified, and
assigned good winding ratios as is done in Section 7.4. Although the suggested winding ratios
of Table 7.2 range from 1:1 to 12:1, it is wise to be cautious against over-compensation in case
the loading of the power line is reduced.
Table 7.2
Winding Ratio Choice
ZII
ZS
(ZL+ ZR)
100 to 10 Ω 10 to 1 Ω 1 to 0.1 Ω
‘Short Line’
4:1 7:1 12:1
0.1 to 1 Ω
‘Average Line’
2:1 4:1 5:1
1 to 10 Ω
‘Long Line’
1:1 2:1 2:1
10 to 100 Ω
83
7.2 Experimental Setup

7.2.1 Couplers
The couplers to be tested, and used for impedance adaptation, were designed according to
Chapter 4 [1], but commercially available high-frequency transformers were used. In order to
experiment with various symmetrical as well as asymmetrical pairs of couplers, two 4:1
coupling circuits, two 2:1 coupling circuits, as well as two 1:1 coupling circuits were designed
and built into three-prong plugs, typical for South Africa and United Kingdom. See Fig. 7.4.
Fig. 7.4 Three pairs of couplers built into 3-prong plugs, typical for South Africa
and United Kingdom.
As these coupling transformers were not designed, but purchased, some of the essential
parameters of these high-frequency transformers had to be tested, as limited information is
supplied with a typical datasheet. Bandwidth especially has to be verified at expected power-
line impedance levels, rather than erroneously assuming the published 50-Ω termination
bandwidth. In this section, coupling transformers are characterized in a transmitter-setup, i.e. a
50-Ω transmitter sending a signal into a low-impedance power-line network. If impedance
adaptation is sufficient, it can be assumed that the coupling transformer will behave similarly
in a receiver setup. See Chapter 4.
Although bandwidth fluctuation can be estimated using datasheets, it is still suggested that,
where possible, the voltage transfer function of the chosen transformer be measured at the
expected power-line impedance level. A network analyzer is ideal for this, as long as the
tracking generator output impedance equals the envisaged modem output impedance. Input
impedance is less of a problem, as dummy loads in parallel with the input impedance may be
used to force the instrument input impedance down to the correct envisaged ‘power-line’
impedance level. See Fig. 7.5.
84
Fig. 7.5 Utilizing external 50-Ω terminations in parallel with an instrument input
impedance to force the input impedance down to a wanted ‘power-line’
impedance level, in this case 12.5 Ω. For this measurement, the reference level
would be -6 dB, as only 1/4 of the received power is measured, and 3/4 dissipated.
Alternatively, a signal generator (with wanted output impedance) may be used in conjunction
with a spectrum analyzer. The maximum-hold function of a spectrum analyzer is ideal for this,
and gives the operator chance to sweep the signal generator by hand. (Most results in this
chapter were measured using this technique.) It is of critical importance though to calibrate
the measurement setup first, by measuring the back-to-back transfer function. Any deviation
from a flat response, may be used to correct further measurements.
7.2.2 Upper cut-off point
If the transformer bandwidth is not explicitly provided, a good indication of the upper-limit of
the pass-band may be obtained from the leakage inductance of the transformer (see Chapter
4). The upper cut-off frequency can be determined with
R
f HF = [Hz] (7.2)
2πL
using the published leakage inductance L and expected terminating (or power-line) impedance
R. Take note that impedance values may be transferred to any side of the transformer, as
reflected values will yield the same answers.
Even if the expected bandwidth of the coupling transformer is sufficient, the limits still need
to have a safety factor included, as fluctuation of the (terminating) power-line load impacts
proportionally on the upper and lower cut-off points. For instance, if the upper cut-off
frequency of a coupling transformer is published as 1 MHz (typically measured with 50-Ω
termination), this figure would drop to 200 kHz if the terminating, reflected power-line
impedance is envisaged to be 10 Ω. See (7.2). As an example, Fig. 7.6 shows the measured
amplitude response for the selected 1:4 transformer – measured for different terminating
impedances. For this measurement though, no series capacitor was used, and thus the
bandwidth fluctuation only affected the upper cut-off point.
85
Fig. 7.6 Influence of terminating load on measured amplitude response of the

selected 1:4 transformer. The upper cut-off point of the pass-band is directly
proportional to the terminating load reactance. Here the curves were not
normalized, and amplitudes cannot be compared.
7.2.3 Lower cut-off point
Conversely, the low-frequency cut-off point is controlled by the coupling capacitor C together
with terminating resistance R:
1
f LF = [Hz] (7.3)
2πRC
This lower cut-off point needs to be low enough to allow data transmission, yet high enough
for filtering purposes in order to prevent core saturation. Chapter 5 or [2] shows that the
power waveform needs to be filtered by roughly 100 dB for acceptable flux density levels.
If the terminating impedance is assumed to be 10 Ω, and perhaps rises to 50 Ω, the low-

frequency cut-off point would shift lower by factor 5, resulting in much less filtering of the
power-waveform. This increase in power-waveform magnitude would typically be between 20
dB and 40 dB per decade drop in cut-off frequency, and can easily cause saturation of the
coupling core. In turn, this would blank out the data stream for the entire duration of this
saturation event. Further, overheating and failure of the coupling transformer would be
imminent.
86
Fig. 7.7 Influence of coupling capacitor on measured amplitude response of the

selected 1:4 transformer, terminating into a worst-case 3-Ω load. The lower cut-off
point of the pass-band is indirectly proportional to the coupling capacitance as
well as terminating load impedance. Here the curves are normalized, and the
reference level is -12 dB as only 1/16 of the load power was measured by the
spectrum analyzer.
The required series capacitor can be calculated by choosing a low-frequency cut-off point (7.3)
in conjunction with the expected terminating load. As this capacitor ‘blocks’ the peak power
voltage, it needs to be rated as such, including a safety factor. Fig. 7.7 shows the measured
frequency (amplitude) response for the lower cut-off point of the selected 4:1 transformer for
different capacitors. For this coupler, a 0.47-µF capacitor was deemed sufficient (see Fig. 7.7)
for a 4:1 winding ratio which represents the worst case – and therefore the same would be
suitable for the 2:1 and 1:1 couplers.
7.2.4 Magnetizing inductance and saturation
The value of the magnetizing inductance LM does not necessarily influence the pass-band of
the coupling circuit, but is required to determine the filtering of the low-frequency power
waveform. For these commercially available transformers, LM is in the order of 400 µH.
Further, the point where magnetizing impedance equals terminating impedance, provides a
hint to obtain the frequency where the lower slope of the transfer function starts to drop
drastically by 40 dB/decade [1]:
R [Hz] (7.4)
f 40 dB ≈
2π ⋅ LM
which is roughly 20 kHz for this design. In Fig. 7.7, it is evident that below 20 kHz, the slope
drops by ≈ 12 dB/octave or ≈ 40 dB/decade. Therefore, the two decades and two octaves of
87
7
frequency ‘space’ between a 50-Hz power waveform and the 20-kHz point, already represents
104 dB of filtering – with no threat of saturating the core.
7.2.5 Emulated transmitter and receiver
For both the laboratory as well as live 220-V measurements, the experimental setup consisted
of an emulated transmitter and receiver with 50-Ω output and input impedances respectively,
connected through the power-line channel via the couplers to be tested. See Fig. 7.8. The
transmitter was emulated using a Goldstar FG2002C function generator, while the receiver
was emulated with an IFR 2397 spectrum analyzer.
Fig. 7.8 Emulated 50-Ω transmitter (Goldstar FG2002C function generator) and
50-Ω receiver (IFR 2397 spectrum analyzer).
For referencing purposes, the back-to-back transfer function of each coupling pair was first
measured, i.e. no power-line channel in-between. All three pairs (4:1:4, 2:1:2, 1:1:1) exhibited
excellent fidelity (< 1 dB) in the 50-kHz to 550-kHz range, and response of the worst case
(4:1:4 pair) is included in Fig. 7.10 through Fig. 7.13 as a benchmark.
7.3 Laboratory Verification of Simulation Results

For verification of the simulation results of Chapter 6, the transmitter and receiver were now
connected (via couplers) using disconnected power cables to form a symmetrical power-line
channel similar to the simulated T-model. Typical extension cord, consisting of three braided
1-mm2 copper cores, was used to form 2-m, 5-m, 10-m and 20-m pairs of power cabling. See
Fig. 7.9(a).
88
(a) (b)
Fig. 7.9 (a) Extension cord and (b) calibrated 50-Ω loads in parallel, with 1:1
coupler.
The inductance of this power cable was measured to be ≈ 400 nH/m. Neglecting resistance
and capacitance of the cabling as was motivated in Section 7.1 (also see [4], [5]), the
approximate series impedance (magnitude) of each cable pair was calculated at 100 kHz and
500 kHz, shown in Table 7.3.
Table 7.3
Approximate |ZS| of Cable Pairs
Total Series Power-Line Impedance |ZS|
Cable Length
@ 100 kHz @ 500 kHz
2m + 2m ≈1Ω ≈5Ω
5m + 5m ≈ 2.5 Ω ≈ 12.5 Ω
10m + 10m ≈5Ω ≈ 25 Ω
20m + 20m ≈ 10 Ω ≈ 50 Ω
Various parallel combinations of wide-band 50-Ω BNC terminations were used to load this
disconnected laboratory setup via a 1:1 coupler. See Fig. 7.9(b). Some possible load values are
50 Ω, 25 Ω, 12.5 Ω, 6.25 Ω, and 3.125 Ω, but two specific values were chosen to coincide
with the lightly loaded and normally loaded classes of Table 7.1, namely the 25-Ω combination and
the 3.125-Ω combination respectively. For clarity, results will be limited and each power-line
condition be given a label, coinciding with a certain point on the plane in Fig. 7.3. Measured
results for power-line conditions A and B are shown in Fig. 7.10 together with the back-to-
back 4:1:4 coupling pair reference line.
For measurements in Fig. 7.10, a ZII ≈ 3 Ω load was placed between two 10-m extension
cords. At 100 kHz, ZS ≈ 5 Ω (point A, Fig. 7.3), while at 500 kHz, ZS ≈ 25 Ω (point B, Fig.
7.3). From the simulation results in Fig. 7.3, it is expected that at point A, a 4:1:4 pair would
89
7
yield best results, followed by the 2:1:2 and 1:1:1 pairs – this is indeed the case in Fig. 7.10.
However, from point B in Fig. 7.3, it is expected that the 2:1:2 pair would perform best,
followed by the 1:1:1 pair and lastly the 4:1:4 pair. This is also experimentally confirmed by the
measurements in Fig. 7.10.
Fig. 7.10 Performance of the three coupling pairs between 50 kHz and 550 kHz in
a disconnected laboratory setup. Here, a ZII ≈ 3 Ω load was placed between two
10-m extension cords. For power-line conditions A and B, performance
corresponds with simulation results in Fig. 7.3.
Fig. 7.11 Measured performance of the three coupling pairs for power-line
conditions C and D. Here, a ZII ≈ 3 Ω load was placed between two 20-m
extension cords, i.e. ZS ≈ 10 Ω @ 100 kHz (point C) and ZS ≈ 50 Ω @ 500 kHz
(point D). The coupler pairs performed as expected for both points C and D in
Fig. 7.3.
90
7
Next, two 20-m cables were used, connecting the same ZII ≈ 3 Ω load between transmitter
and receiver. This setup is labeled power-line condition C and D, at 100 kHz and 500 kHz
respectively. Coupling pair performance (Fig. 7.11) was as expected from Fig. 7.3.
A third set of results is shown in Fig. 7.12. Points E and F describe a lightly-loaded power line of
ZII ≈ 25 Ω, connected between two 20-m cables being equivalent to ZS ≈ 10 Ω @ 100 kHz
(point E) and ZS ≈ 50 Ω @ 500 kHz (point F). Results in Fig. 7.12 and Fig. 7.13 can also be
compared with expected, simulated performance in Fig. 7.2(b), where ZII = 25 Ω. For power-
line condition E (ZS ≈ 10 Ω), Fig. 7.2(b) predicts that the 2:1:2 pair would outperform the
1:1:1 and 4:1:4 pair by ≈ 2 dB and ≈ 5 dB respectively. Experimental results in Fig. 7.12
indicate a similar trend with the 2:1:2 pair outperforming the 1:1:1 and 4:1:4 pair by ≈ 2.5 dB
and ≈ 6 dB respectively. Also, for ZS ≈ 50 Ω, Fig. 7.2(b) predicts that the 1:1:1 pair would
outperform the 2:1:2 pair and 4:1:4 by ≈ 4 dB and ≈ 8 dB respectively. Measured results for
point F in Fig. 7.12 indicate a similar trend, with the 1:1:1 pair outperforming the 2:1:2 and
4:1:4 pair by ≈ 4.5 dB and ≈ 10 dB respectively.
Fig. 7.12 Measured performance for lightly-loaded power-line conditions E and F.

In this laboratory setup, a ZII ≈ 25 Ω load was placed between two 20-m extension
cords, i.e. ZS ≈ 10 Ω @ 100 kHz (point E) and ZS ≈ 50 Ω @ 500 kHz (point F).
Performance at points E and F compare favorably to expected performance (Fig.
7.2(b) and Fig. 7.3).
Lastly, the same ZII ≈ 25 Ω load was connected between two 2-m extension cords. For power-
line condition G, ZS ≈ 1 Ω whereas for point H, ZS ≈ 5 Ω. The results in Fig. 7.13 also
correlate with simulation results of Fig. 7.3, and show much similarity, compared to the
expected performance in Fig. 7.2(b).
91
Fig. 7.13 Performance of the coupling pairs for power-line conditions G and H.
Here, a ZII ≈ 25 Ω load was placed between two 2-m extension cords, i.e. ZS ≈ 1 Ω
@ 100 kHz (point G) and ZS ≈ 5 Ω @ 500 kHz (point H). Compare with expected
performance in Fig. 7.2(b) and Fig. 7.3.
7.4 Classifying Outlets for Impedance Adaptation

In order to illustrate the improved performance that is possible by choosing good winding
ratios qualitatively without exact knowledge of the channel, measurements were done between
50 kHz and 550 kHz in a live, uncharacterized residential 220-V power-line environment.
These measurements however cannot be compared directly with the laboratory results of
Section 7.3, as the topology is unknown and differs considerably from the simulated T-model.
Further, all power-line cabling utilized, were solid 2.5 mm2 copper conductors (AWG14),
whereas the laboratory setup in Section 7.3 was made up of stranded 1 mm2 (AWG18) copper
wire. This implies that the inductance per unit length is different, and thus directly affects the
scaling of the frequency axis.
For these measurements, we merely used the linguistic terms of Table 7.4 to classify the
estimated distances between transmitter/receiver outlets and estimated position of the
dominating load. The severity of the dominating power-line load impedance was also
estimated and classified using Table 7.4. Although the severity of these loads are unknown at
communication frequencies, their 50-Hz ratings were used to gauge loading of the network.
As a first example, a residential kitchen is considered. Power outlets in the kitchen are deemed
to be central, i.e. close to other loads and the distribution box – this points to short distances
(short line), row 1 of Table 7.4. Further, normally-loaded conditions are possible (column 2 of
Table 7.4), as many appliances may be connected to the network, e.g. kettle, toaster,
dishwasher, deep fryer, microwave oven, etc. For this power-line condition, Table 7.4 suggests
that a 7:1 winding ratio would yield best transmission results. However, one needs to be
92
7
conservative to prevent over-compensation in case fewer loads are connected. Therefore a 4:1
winding ratio is suggested for kitchen outlets.
Table 7.4
Room Classification
ZII
ZS
(ZL+ ZR)
100 to 10 Ω 10 to 1 Ω 1 to 0.1 Ω
‘Short Line’ 4:1
7:1 12:1
0.1 to 1 Ω e.g. Kitchen
‘Average Line’ 2:1
4:1 5:1
1 to 10 Ω e.g. Bed, Lounge
‘Long Line’ 1:1 2:1
2:1
10 to 100 Ω e.g. Gate e.g. Laundry
Next, a bedroom outlet is considered. Although bedrooms are typically further from the
distribution box, a distance of 10 m or 20 m probably acts as an average line for CENELEC
frequencies. Loading is usually also less severe (lightly loaded), and might involve bed lamps,
small power supplies or perhaps the occasional hair drier being plugged in. Therefore Table
7.4 suggests a 2:1 winding ratio for bedrooms. Similar arguments also lead to a winding ratio
choice of 2:1 for a lounge, laundry, or dining room power outlet. Only lightly-loaded remote
power outlets (long lines) would benefit from a 1:1 winding ratio – for instance at the gate of a
residential property.
The results presented in Fig. 7.14 through 7.16 show that even without accurate knowledge of
the power-line environment, good winding ratio choices can be made by using Table 7.4,
resulting in beneficial transmission gains of up to 10 dB and more. As a first live experimental
trial, transmission was investigated between the lounge and kitchen. Table 7.4 recommends a
2:1 winding ratio for the lounge, and a 4:1 winding ratio for the kitchen. Fig. 7.14 shows the
improvement from a default 1:1:1 pair, progressing step-by-step as the winding ratio choice
approaches the suggested 2:1:4 pair. The proposed 2:1:4 pair (black line, Fig. 7.14) yielded
more than 10 dB compared to a default 1:1:1 setup (light-gray line, Fig. 7.14).
93
Fig. 7.14 Live 220-V transmission measurements from a residential lounge to

kitchen. The 2:1:4 pair suggested by Table 7.4 (black) yielded more than 10 dB
compared to a default 1:1:1 setup (light-gray). The 2:1:1 and 1:1:4 pairs indicate the
step-by-step improvement as the winding ratio choice improves.
Next, live transmission measurements were done between the kitchen and bedroom. Here a
4:1:2 pair (black line, Fig. 7.15) was expected to perform best (see Table 7.4), and yielded ≈ 5
dB at 100 kHz, and ≈ 12 dB at 500 kHz compared to a default 1:1:1 setup (light-gray line, Fig.
7.15). The other two lines in Fig. 7.15 confirm that transmission improves as the winding
ratios approach the optimum pair.
Fig. 7.15 Live transmission measurements between a kitchen and bedroom. Here
a 4:1:2 pair (black) was expected to perform best (see Table 7.4), and yielded ≈ 5
dB at 100 kHz, ≈ 12 dB at 500 kHz compared to a default 1:1:1 setup (light-gray).
The 2:1:1 and 4:1:1 pairs exhibit intermediate improvements.
94
7
The third set of live measurements in Fig. 7.16, shows transmission from the border of a
residence (gate) to the laundry, for different winding ratios. In this case, only marginal gains
are possible as distances are large, and loads relatively light. The 1:1:2 pair as proposed by
Table 7.4 (black line, Fig. 7.16) yielded only ≈ 2 dB compared to a default 1:1:1 setup (light-
gray line, Fig. 7.16). The 1:1:4 and 2:1:4 lines in Fig. 7.16 indicate that over-compensation can
be detrimental, especially for remote outlets (long lines).
Fig. 7.16 Live transmission measurements from the border of a residence (gate) to
the laundry. Here only marginal gains are possible as distances are large, and
loads relatively light. Table 7.4 proposed a 1:1:2 pair (black) which yielded only ≈ 2
dB compared to a default 1:1:1 setup (light-gray). The 1:1:4 and 2:1:4 curves warn
that one should guard against over-compensation (especially for long lines).
It must be mentioned that proper winding ratios can be even better gauged for a certain fixed
frequency, whereas the results shown in this section span over a decade of frequencies. Higher
frequencies require more conservative (smaller) winding ratios, as cable inductance has a
stronger impact on transmission. Conversely, at lower frequencies, load impedances can be
transformed more aggressively by using larger winding ratios.
7.5 Dual Impedance-Adapting Coupler

The extremes as well as unpredictability of the power-line access impedance has always been
one of the main enemies of low-voltage power-line communications. Much work has been
done on field measurements as well as statistical modeling of the power-line access
impedance. Other researchers have focused on modulation techniques and spread-spectrum
techniques in order to safeguard data against fluctuating channel conditions. Another
approach to mitigate fluctuating power-line impedance is network conditioning – ‘blocking
off’ harsh loads from the power-line network by means of series-connected filters e.g. [6].
The installation of these filters is labor-intensive and a side effect is the back-E.M.F. voltage
95
7
spike across the filter inductance, generated when currents are abruptly switched off. These
filters are expensive, especially for low-frequency applications, as they have to be rated at the
typical current that is drawn by the load (to be ‘blocked off’).
However, said loads are often ideal candidates for home automation, requiring the
communication signal to reach these loads. A hot water geyser for instance, may be controlled
directly through the power-line channel in order to save electricity. In future, a stove oven may
be switched on from a remote location via a cellular phone, connecting to the in-home power-
line network in order for a meal to be ready upon arrival at home. Therefore this section
investigates an alternative approach to try and mitigate harsh loads, and investigates the
possibility of utilizing a simple, dual coupler as an alternative to network conditioning.
Take note that the goal of network conditioning is two-fold, i.e. raising the general power-line
access impedance level, and reducing fluctuations thereof. This requirement for little
fluctuation is largely due to the general practice of only using one kind of coupling circuit
throughout a product range. A fixed coupler operates at its optimum only at a specific power-
line impedance level. It was shown in Section 7.4 [1] that the use of different couplers for
different power-line conditions (in the same building) can result in substantial transmission
gains. In this section, this concept is taken one step further – a dual coupler provides two
optimized couplers integrated into one circuit – ideal for switching loads.
The proposed dual-coupler circuit diagram is shown in Fig. 7.17. The built-in appliance switch
is utilized to drive a low-power switching circuit in order to switch between the suggested 2:1
and 4:1 winding ratios. (Throughout this section, the switching of the appliance is not
controlled, but assumed as an unpredictable parameter.) Further, a center-tapped transformer
is used to implement both winding ratios, resulting in no extra costs over a standard coupling
transformer. Another cost implication is the addition of C2, which typically requires a much
lower capacitance than for C1.
Fig. 7.17 Proposed dual coupler, incorporating a 4:1 winding ratio for permanent
connection between modem and main load Z1 as well as a 2:1 winding ratio
between modem and power line (in case the main load Z1 is disconnected from
the power).
96
7
Take note that the main load, Z1 is permanently connected to the modem via a 1:4 winding
ratio (seen from power-line side). For the case when Z1 is switched off (as shown in Fig. 7.17)
this permanent connection does unnecessarily drain a portion of the communication signal.
However, the magnitude of Z1 is enlarged by factor 16, seen from the modem side. Thus only
a small portion of the communication signal is sacrificed (e.g. less than 3 dB for Z1 > 3 Ω).
In order to test the performance of the proposed dual coupler, two prototypes were designed
and constructed. Fig. 7.18 shows a photograph of the first prototype. The first dual coupler
offers the choice of a 4:1 vs. 2:1 winding ratio (see Fig. 7.17), while a second dual coupler was
designed and constructed to provide a 2:1 vs. 1:1 choice for longer distances and higher
frequencies – as was motivated in Section 7.4.
Fig. 7.18 Photo of the 4:1 vs. 2:1 dual coupler prototype. A coaxial BNC connector
is used for the modem side, while crimped lugs are used for the power-line side, in
order to easily connect to typical heating element terminals. Also, a 1-MΩ
discharge resistor was connected across capacitor C2 in Fig. 7.17 for safety
reasons.
Before testing the dual coupler in a live power-line environment, its performance was first
investigated in a back-to-back setup, using artificial loads (parallel combinations of 50-Ω BNC
terminations). The experimental setup was identical to the one described in Section 7.2 and
consisted of an emulated transmitter and receiver with 50-Ω output and input impedances
respectively, connected through the power-line channel via the coupling circuitry under
consideration.
Fig. 7.19 shows some back-to-back (i.e. no power line) transmission results for a standard 2:1
transmitter-coupler, while varying the receiver-coupler, for frequencies below 1 MHz. The top
three curves in Fig. 7.19 correspond to a no-load condition (emulating the main appliance load
being off). As can be seen in Fig. 7.19 (top curve), a standard 2:1:2 pair shows good fidelity,
with ≈ 1 dB drop between 100 kHz and 1 MHz. The second 2:1:4 curve indicates the penalty
associated with the impedance mismatch between a standard 2:1 transmitter-coupler and a 4:1
receiver-coupler. The third curve (2:1:dual pair) indicates a marginal penalty of ≈ 3 dB
97
7
(compared to 2:1:2 pair), as the main 3-Ω load is permanently connected across the dual
coupler’s 4:1 winding even though the 2:1 winding receives the signal. See Fig. 7.17.
Conversely, the bottom two curves in Fig. 7.19, correspond to a full-load condition (emulating
the 3-Ω main appliance load being on). These two curves show that a gain of ≈ 2 dB is
possible by using a dual coupler (4:1 winding ratio) instead of a standard coupler (2:1 winding
ratio).
Another important goal of the dual coupler, is to reduce fluctuations in impedance levels –
and consequently reduce fluctuations in performance. The labeled brackets in Fig. 7.19
indicate the measured fluctuation for different coupler setups. The ‘2:1 only’ bracket shows ≈
10 dB fluctuation for a standard 2:1:2 coupler pair, when switching a 3-Ω load.
The ‘dual’ bracket also indicates the fluctuation in transmission levels when switching a 3-Ω
load between on and off, but here the dual coupler reduces fluctuation to ≈ 5 dB instead of ≈
10 dB. Thus the dual coupler allows a more stable, but lower network impedance, facilitating
more accurate impedance adaptation from other outlets.
Fig. 7.19 Laboratory measurements showing transmission fluctuation for different

coupler setups when switching a 3-Ω load.
An alternative setup to Fig. 7.17 is possible – where both the load as well as the correct
coupler winding are switched together (the main load is therefore not permanently connected).
This strategy would allow better performance during ‘off’ conditions at the expense of slightly
more complicated and costly switching circuitry. However, network impedance fluctuation
would be worse.
98
7
In order to investigate the performance of a dual coupler over a standard one-winding coupler
in a live, 220-V residential power-line network, measurements were also done between 50 kHz
and 550 kHz as in the preceding sections – thus including the CENELEC B, C, and D bands
(roughly from 95 kHz to 150 kHz), as well as ≈ 500-kHz systems, typical for Japan and USA.
Although the characteristics of this residential network is not known, coupler winding ratios
had previously been optimized for different transmission routes – see Section 7.4 [1]. In the
following sets of results, the two dual coupler prototypes were in both cases used as receiver
couplers. However, it must be emphasized that these dual couplers are equally effective when
used as couplers to transmit from harsh switching appliances into the power-line network.
As a first step, transmission between the gate and laundry was investigated. As distances are
large, a 2:1 vs. 1:1 dual coupler was chosen. For this set of measurements, a 3.6-kW resistive
load was used in the laundry, emulating an appliance, with the spectrum analyzer connecting
to said load via the dual vs. standard receiver-coupler. For both sets of results below, the dual
coupler was configured according to Fig. 7.17 (main load permanently connected). The
emulated transmitter was connected at the gate of the property through a standard 1:1
coupler. Transmission results are shown in Fig. 7.20 – the outer envelope shows on-off
fluctuation for a standard 1:1 receiver-coupler, while the two inner curves show little
fluctuation when using a dual coupler at the receiver. In addition ≈ 3 dB is gained when the
main load is on, as the 2:1 portion of the dual coupler is automatically activated for on-
conditions.
Fig. 7.20 Live transmission measurements from the border of a residence (gate) to
the laundry, as influenced by the receiver-coupler. Here the 1:1 vs. 1:2 dual coupler
was loaded with a 3.6-kW resistive load. The outer envelope shows on-off
fluctuation for a standard 1:1 coupler, while the two inner curves show little
fluctuation when using a dual coupler. Furthermore, ≈ 3 dB is gained when the
main load is on.
99
7
For a second example, the stove was chosen as a typical appliance that would benefit from a
dual coupler. See Fig. 7.21 and 7.22. The transmission route was chosen to be from a kitchen
outlet to the stove. Take note that the stove has dedicated power-line cabling to the
distribution board, and does not form part of the kitchen sub-network. A standard 2:1
transmitter-coupler was used, while the receiver-coupler was varied to obtain comparative
results.
Fig. 7.21 Dual coupler connected to the stove supply, with 4:1 portion connected
across oven (bake and grill) elements.
Fig. 7.22 Live transmission measurements between a kitchen outlet and stove.
The proposed 4:1 vs. 2:1 dual receiver-coupler accomplishes ≈ 2 dB of gain
compared to a default 2:1 receiver-coupler when the load is on. Fluctuation when
switching is better, but not drastically reduced as only ≈ 4 kW of the ≈ 8 kW was
permanently connected across the 4:1 winding.
100
7
As both the stove and kitchen networks are quite central, and thus in closer vicinity to low-
impedance loads, the 4:1 vs. 2:1 dual coupler was chosen for this setup. The transmission
results are summarized in Fig. 7.22, showing the fluctuation between a no-load and full-load
condition – comprising of three stove plates as well as two oven elements, totaling ≈ 8 kW.
Fluctuation when switching is better, but not drastically reduced as for practical reasons, only
≈ 4 kW of the ≈ 8 kW was permanently connected across the 4:1 winding (see Fig. 7.21). Fig.
7.22 also confirms that performance fluctuation is reduced by the dual coupler, and that ≈ 2
dB is gained by the automatic utilization of the built-in 4:1 winding ratio of the dual coupler
when the appliance is loaded.
7.6 Conclusion
T
HE title of this thesis, “Efficient Coupling for Power-Line Communications”
summarizes the aim of this project. Although coupling circuit design had been
researched and tested in a laboratory (Chapter 4), it was still to be proven in field
trials. Also, in Chapter 6, impedance adaptation mechanisms were investigated, yielding
promising simulation results. Thus these theoretical impedance adaptation strategies also still
had to be verified in a real-world power-line network.
As the highlight of this project, this chapter therefore investigated the deployment of a range
of in-house prototype couplers in a residential 220-V power-line network. Before applying
impedance adaptation strategies in a live network though, the performance of the couplers
were tested in a disconnected laboratory power-line network – see Section 7.3. Instead of
defaulting to one fixed winding ratio for all modems under all circumstances, Section 7.4
showed that a certain room or location may be classified (based on typical power-line
conditions) in order to select a good winding ratio per room (for these typical conditions).
The approach followed in Section 7.5 is even more specialized – here an integrated dual-
impedance coupler is deployed for certain switching loads, where this switching mechanism
impacts on the local power-line impedance level to a large extent. Live 220-V results show that
the proposed dual coupler can aid in the mitigation of switching power-line loads by means of
i) providing two dedicated winding ratios tuned for on and off conditions, and ii) a more
stable but lower power-line impedance level – this facilitates more accurate impedance
adaptation from other power outlets.
101
7
7.7 References
[1] P. A. Janse van Rensburg, H. C. Ferreira, “Classifying typical residential rooms for narrow-band impedance
adaptation,” Proceedings of the 11th International Symposium on Power-Line Communications and its Applications
(ISPLC’07), Pisa, Italy, 26-28 March 2007, pp. 47-52.
January 2005, pp. 64-70.
1998, pp. 15-20.
[6] P. A. Brown, “Directional coupling of high frequency signals onto power networks,” Proceedings of the 1st
International Symposium on Power-Line Communications and its Applications (ISPLC’97), Essen, Germany, 2-4 April
1997, pp. 121-126.
*****
102
8
Chapter
CHAPTER 8
Conclusion
A
S the concluding chapter of this thesis, Chapter 8 describes the overall contribution
of this project. First, a concise overview of the goals and accomplishments of this
project is given. Next, the content and findings of each chapter is summarized to
give the reader an overview of the specific work covered by each chapter. Also, it is intended
to show how the preceding chapters progressed – to gather, develop and present the necessary
theory for effective power-line communications coupling. Apart from theoretical content, the
reader is briefed on experimental measurements in most of the chapters. These include
laboratory testing of concepts, as well as live testing in a 220-V residential power-line
environment. Further, all the conference contributions and journal papers relevant to this
project, are cited. Before the closure, some future research possibilities are mentioned. Most
of these suggestions involve more specialized application of the results presented in Chapters
6 and 7, while some other suggestions entail further expansion of the simplified model of
Chapter 6 – in order to investigate impedance adaptation for higher modulation frequencies.
103
CONCLUSION
8
8.1 Summary and Core Results
This chapter summarizes the content of this thesis, discusses the most important
results and concludes with some possibilities regarding the furthering of this research
topic.
T
HE title of this thesis, “Efficient Coupling for Power-Line Communications”
describes the overall aim of this project. Practically, this study also aimed to clarify the
uncertainty and fears over connecting sensitive modems as well as expensive
measuring equipment (through couplers) to peak voltages of 500 V or more. After a concise
introduction to the power-line communications field (Chapter 1), conflicting requirements of
the power waveform versus the communication waveform, were discussed in Chapter 2.
Especially high-frequency parasitic side effects had to be taken into consideration, as the
frequency of the superimposed communication signal is 3 to 6 orders higher than the power
waveform frequency.
In Chapter 3, basic filter concepts were discussed, and the operation of a typical transformer-
capacitor coupling circuit was considered. Chapter 4 followed with a practical, step-by-step
design example for a bi-directional 100-kHz, 500-V coupler, also showing laboratory
measurements to verify correct operation. As the internal impedances of a coupling
transformer determines its filter characteristics, the design and fine-tuning of leakage and
magnetizing inductances were discussed in Chapter 5.
Impedance adaptation strategies were considered in Chapter 6, and a simple T-model was
used to investigate the tradeoff between cable- and load impedance adaptation for frequencies
below 1 MHz. Chapter 7 followed with practical testing of different couplers designed
according to Chapter 4. Utilizing the simulation results of Chapter 6, impedance adaptation
techniques were also implemented in a live 220-V residential power-line network. These
results showed significant transmission gains, and indicate clear step-wise improvement as
impedance adaptation is improved according to the simulation results of Chapter 6. Finally, a
dual-impedance coupler was designed and prototyped for the mitigation of harsh switching
loads. Results confirmed that transmission is improved, and that the local network impedance
level is more stable when a dual coupler is used instead of a standard, fixed-winding coupler.
8.1.1 Chapter 1 – PLC history, applications and hindrances
This chapter gave the reader a quick overview of the historical development of the power-line
communications field and mentioned many of the exciting applications that have been
implemented in the past, as well as some that may become viable soon. Although the
superimposing of a communication signal onto a power waveform has been practiced for over
a century, power-line communications only recently (in the 1990’s) started gaining
104
CONCLUSION
8
momentum. Traditional applications were limited to voice and control signals over high-
voltage and medium-voltage networks, as well as peak-demand management (load shedding)
and automatic meter reading over low-voltage networks. However, the advances in
semiconductor theory and production in the last half-century and associated growth in the
consumer electronics market, facilitated the computer- and internet boom. Subsequently this
has set the stage for home networks and home automation – ideal candidates to be
implemented through power-line communications.
Next, the disadvantages of the power-line channel as a retrofit network were also described.
The utilization of the power-line channel for communications – an application it was never
designed for – makes it an interesting and challenging field. Most of these hindrances are high-
frequency side-effects, as the frequency of a typical power-line communications signal is
between 3 and 6 orders higher than the 50-Hz or 60-Hz frequency of its power waveform
counterpart – for which the power-line network was designed. For low-frequency PLC (below
≈ 500 kHz), power-line cabling is not as hostile towards the communication signal. Rather,
loads draining the communication signal, are the main enemy.
Unfortunately, bandwidth is directly limited by frequency, and only limited data rates can be
achieved with low-frequency power-line communications. As the communication frequency is
pushed up to increase bandwidth, so do the complications increase. Instead of a conducting
copper network, the power-line channel acts more and more like a complex antenna as
frequency rises. Both from an engineering efficiency viewpoint, as well as environmental and
interference perspectives, this radiation of the power-line communication signal has become
an important consideration. Therefore, in Chapter 1, the reader was also given background on
the all-important topic of regulations pertaining to power-line communications. As an
introductory chapter, important references were provided throughout the text, for the reader
who intends to investigate this field further.
8.1.2 Chapter 2 – component selection and circuit layout
The coupling interface between modem and power line is by nature a difficult one to design
and implement, as the two applicable waveforms represent two extremes – high power at a
low frequency versus low power at a high frequency. These extremes often specify
contradictory requirements for the various components involved, as was explained in this
chapter.
Frequency behavior of passive components was discussed, and practical design guidelines
were given. Generally components are chosen to have a self-resonant point of at least an
octave above the operating frequency. The self-resonant point of a certain component can be
raised by either reducing the parasitic values (as discussed in Chapter 2) or if not possible,
designing the circuit using smaller component values.
As conductors are often overlooked, these were discussed first. Conductors typically have an
even bigger impact on a circuit’s parasitic behaviour than the components themselves. Also,
inductors and transformers are constructed using insulated conductors, and said characteristics
105
CONCLUSION
8
of conductors thus need to be well understood. As circuit layout also has a direct impact on
parasitic effects, this topic was included in the discussion on conductors.
Next, resistors, capacitors, and inductors were discussed, emphasizing high-frequency and
other stray effects that need to be taken into consideration when designing a modem or
coupling circuit. Lastly, transformers were considered, as transformers are found in the
majority of coupling circuits, and provide impedance adaptation as well as help protect
modem circuitry from the power waveform. The content of this chapter was presented as a
paper at the ISPLC’03 conference, Kyoto, Japan [1].
8.1.3 Chapter 3 – coupling fundamentals
The design of coupling and de-coupling circuitry is often neglected at the expense of reduced
signal levels and increased noise levels. In essence, a coupling circuit is a special filter that
allows a high-frequency communication signal to pass through it (preferably in both
directions) while ‘blocking’ or filtering the power waveform. Therefore Chapter 3 introduced
basic filtering concepts in a very simple, yet practical way in order to clarify the uncertainty
that exists over coupling circuits for power-line communications. It was further explained that
coupling capacitors or de-coupling inductors don’t merely pass or block signals, but that their
filtering characteristics are quite dependent on the loads into which the waveforms terminate.
If no impedance adaptation is required for a modem, standard passive filter circuits may be
used, as long as the components are rated to block the power waveform voltage. Extra
protective circuitry is also advisable.
However, if impedance adaptation is required between the modem and the power line,
transformers are indispensable. This chapter therefore expanded on the role of transformers
in coupling circuitry. As the internal series (‘leakage’) impedance of a transformer is inductive,
all transformers exhibit some kind of low-pass effect, filtering frequencies higher than this cut-
off point. However, if the coupling circuit is required to implement a band-pass filter, extra
components are required. Typically, a single series capacitor is sufficient, and this capacitor
together with the leakage inductance, form a series-resonant band-pass filter that can be
modeled as a simple LC-R circuit. Design equations were given and these were verified by
measuring the voltage transfer function of practical coupling circuits for different values of L,
C and R. The content of this chapter was also presented at the 2003 ISPLC conference in
Kyoto, Japan [2].
8.1.4 Chapter 4 – design of a bi-directional coupler
Although some authoritative texts on power-line communications suggest separate coupling

circuits for transmitting and receiving data, it was envisaged that a properly designed coupling
circuit would be able to adapt modem and power-line impedances to be roughly equal.
Equality in terminating impedances further implies that filter characteristics would be similar
in both the transmit- as well as receive directions, yielding a bi-directional coupler. Chapter 4
showed that this is indeed the case, and that a single coupling circuit can be used to transmit
and receive data from a power-line modem.
106
CONCLUSION
8
After giving a concise introduction of the suggested coupling circuit and its operation, this
chapter focused on practical design steps for an impedance-adapting transformer-capacitor
coupling circuit. The design was based on the European CENELEC standard which
prescribes maximum signal levels as well as allowed carrier frequency bands. Although the
design included an in-depth transformer design, the information is still applicable and useful if
a commercial coupling transformer is to be selected. The practical fine-tuning of transformer
magnetizing- and leakage inductances and their influence on the pass-band were also
mentioned.
The designed coupling circuit was constructed and experimental measurements in Chapter 4
show good bi-directional symmetry when transmitting and receiving data. Next, the impact of
a fluctuating power-line impedance level on coupler bandwidth was illustrated, and concluding
design tradeoffs were discussed. The work in this chapter also yielded a conference paper [3]
that was presented at the 2004 ISPLC conference, Zaragoza, Spain.
8.1.5 Chapter 5 – coupling transformer impedances
Chapter 5 showed that in order to design an effective coupling circuit, the internal impedances
have to be designed properly, as they can have a detrimental effect on coupling by either (i)
filtering the communication signal or (ii) allowing the power waveform to saturate the
coupling transformer. Although Chapter 5 is aimed towards the design of a coupling
transformer, a reader intending to utilize a commercial coupling transformer will still gain
invaluable knowledge, helping with the selection of an appropriate coupling transformer.
At the outset of Chapter 5, the fundamental definition of inductance was discussed, after
which a standard simplified transformer equivalent circuit was derived. Next, theoretical
design equations were derived for a transformer’s leakage inductance. As the leakage
inductance impacts directly on the upper cut-off point of the coupling circuit’s pass-band, the
designer was shown how to increase or decrease leakage inductance for a certain coupling
circuit design.
Core saturation, a very important topic, was discussed next, as transmitted data is lost during a
core saturation event. Simulations as well as practical measurements showed that the
magnetizing inductance (with coupling capacitor) causes a 40-dB/decade filter slope at low
frequencies. Therefore the magnetizing inductance should be designed to adequately filter the
power waveform – in order to prevent core saturation. Magnetizing inductance design
equations were presented, and practical ways to increase or reduce this inductance were
discussed.
The work in this chapter also yielded a conference paper [4] that was presented at the 2004
ISPLC conference, Zaragoza, Spain. Also, in conjunction with the results of Chapter 4, two
other conference contributions were made as co-author – investigating the impact of coupler
filtering on modulation performance [5], [6].
107
CONCLUSION
8
8.1.6 Chapter 6 – impedance adaptation
In Chapter 5, the design of a coupling circuit was based on a purely parallel power-line
impedance. It was assumed that the adaptation would be sufficient if the transformer winding
ratio equalized the modem impedance with the (theoretically) parallel power-line impedance.
However, Chapter 6 showed that in a real-world scenario, transmitters and receivers are
always separated by series cables, and also connected to parallel loads via series cables. It was
further shown that series and parallel impedances have conflicting adaptation requirements. A
perfect channel would consist of an infinitely small series (cable) impedance, and an infinitely
large parallel (load) impedance, leaving a direct connection between the matched transmitter
and receiver pair.
Thus the goal of Chapter 6 was to investigate this trade-off by means of a simple T-model for
the power-line channel. The main reason for choosing a simple model, was to linguistically
describe any power-line condition, and utilize the results for improved transmission. Although
the model may seem very simple initially, it later proved that some results could only be
obtained with numerical analysis – even after reducing the number of parameters by placing
certain constraints on the circuit parameters. Very interesting and useful simulation results
were obtained, and three-dimensional solutions for optimum winding ratios (given certain
constraints) were presented. Finally, interpretation of simulation results in order to implement
these practically, was discussed. Portions of this chapter was presented as a conference paper
at the 2006 ISPLC in Orlando, U.S.A [7].
8.1.7 Chapter 7 – experimental verification
Although coupling circuit design had been researched and tested in the laboratory (Chapter 4),
it still had to be proven in field trials. Also, in Chapter 6, impedance adaptation mechanisms
were investigated, yielding promising simulation results. Thus these theoretical impedance
adaptation strategies also still had to be verified in a real-world power-line network.
As the highlight of this project, Chapter 7 therefore investigated the deployment of a range of
prototype couplers in a residential 220-V power-line network. Before applying impedance
adaptation strategies in a live network though, the performance of the couplers were tested in
a disconnected laboratory power-line network. This laboratory setup mimicked the T-model
with two extension cords as series impedances with a calibrated load between them as a
parallel load impedance. Various extension cord lengths were combined with two applicable
load values, being 3 Ω and 25 Ω, in order to validate the simulation results of Chapter 6.
After successful validation with a laboratory setup, the three-dimensional set of (symmetrical)
optimum winding ratios of Chapter 6 was tested in a live residential power-line network.
Instead of defaulting to one fixed winding ratio for all modems under all circumstances,
Chapter 7 showed that a certain room or location may be classified (based on typical power-
line conditions) in order to select a good winding ratio per room (for these typical conditions).
Live measurements proved very successful, and step-wise improvement can be seen as the
transmitter- and receiver winding ratios approach the suggested optimum value.
108
CONCLUSION
8
Finally, an even more specialized approach was followed – an integrated dual-impedance
coupler was designed, prototyped and deployed for certain switching loads, where this
switching mechanism impacts on the local power-line impedance level to a large extent. Live
220-V results showed that such a dual coupler can aid in the mitigation of switching power-
line loads by means of providing two dedicated winding ratios, tuned for on and off
conditions. Another inherited advantage results – power-line impedance levels are more
stable, facilitating more accurate impedance adaptation from other outlets.
Portions of Chapter 7 were presented as conference papers [8], [9], at the 2007 ISPLC, Pisa,
Italy.
8.2 Further Research Possibilities

In order to fully exploit the findings presented in this thesis, follow-up research is warranted –
some possible projects are suggested below:
1) Statistical field data on fluctuating power-line conditions can aid in the choice of good (but
conservative) winding ratios as spelled out by Chapters 6 and 7.
2) Network conditioning, in order to keep ZII larger than ≈ 12.5 Ω can make winding ratio
choice much more predictable, especially for asymmetrical power-line conditions. See
three-dimensional simulation results, Chapter 6. Therefore the design of de-coupling
circuits for network conditioning, as well as cost-versus-performance tradeoffs are
outstanding issues that need to be researched.
3) Improved intra-building performance is possible by combining wiring lay-out (distances)

as well as load (appliance) severity and -position on a two- or three-dimensional drawing –
a computer model would be advisable for large buildings. Custom classification of
impedances such as in Chapter 7 would facilitate the choice of good coupler winding
ratios at any given point in the building.
4) The proposed impedance adaptation strategy of Chapter 7, and in particular the dual
coupler concept, is even more appropriate for closed systems such as automotive power-
line communications [10], [11]. If a vehicle can be designed to have all major loads
conditioned or adapted by such a dual coupler, a very stable (though low) network
impedance would be accomplished.
5) In future, intelligent couplers may utilize fuzzy logic to switch between windings of a
multi-tap coupling transformer, similar to the dual coupler discussed in Chapter 7. See
Table 7.2 for possible fuzzy sets describing the nine suggested power-line conditions.
6) For wide-band applications, a high-frequency lumped-parameter cable model of sufficient

accuracy for frequencies up to 10 MHz or perhaps even 30 MHz, can be substituted in
place of ZL and ZR in Chapter 7. Refer to Chapter 6 where it was argued that these models
109
CONCLUSION
8
do not have to be very accurate, as impedance discontinuities between different
transformer winding ratios are very abrupt. Also, load characterization (of ZII) would have
to be done for these higher frequencies. Substituting these high-frequency models in place
of the impedances of the suggested T-model of Chapters 6 and 7, will facilitate winding
ratio choice for wide-band applications.
7) Protective coupler circuitry, including varistors, zener diodes and also coupling
transformers, is also an outstanding topic that needs attention. Further, the mechanisms
involved when voltage and current transients are absorbed, have to be investigated by
power-line communications researchers. For instance, the effect of impulses on coupling
transformer operation (and saturation) and the subsequent influence on data transmission,
has to be clarified and mitigated. This seems like a logical and necessary topic to be
refined, in order to have robust and predictable couplers as a reliable link in the power-line
communications chain.
8.3 Closure
A
LTHOUGH the field of low-voltage power-line communications has progressed
rapidly over the past decade, a very important link in the communication chain has
been neglected, namely coupling. Before this project, hardly any published literature
existed on this topic, especially for low-voltage power-line network coupling. After
investigating high-frequency and other parasitic side-effects at component level, the
mechanisms of coupling circuitry were examined. Further, impedance adaptation and bi-
directional transmission were considered, leading to a successful bi-directional coupler of
which the practical step-by-step design was documented for future reference [12]. The
highlight of this project proved to be an investigation into the influence of coupler winding
ratio on impedance adaptation of the low-voltage power-line network [13]. Very interesting
simulation results were obtained and these were confirmed using a laboratory setup of
characterized power cables and calibrated loads. As a final step, these simulation results were
leveraged to improve power-line transmission over a live 220-V residential network by means
of i) classifying typical residential rooms to choose proper winding ratios and ii) using a dual
coupler for dedicated on-off switching with harsh loads, thereby mitigating the fluctuating
impact of said loads on low-voltage power-line communications.
110
CONCLUSION
8
8.4 References
[1] P. A. Janse van Rensburg, H. C. Ferreira, “Practical aspects of component selection and circuit layout for
modem and coupling circuitry,” Proceedings of the 7th International Symposium on Power-Line Communications and its
Applications (ISPLC’03), Kyoto, Japan, 26-28 March 2003, pp. 197-203.
[3] P. A. Janse van Rensburg, H. C. Ferreira, “Step-by-step design of a coupling circuit with bi-directional
transmission capabilities,” Proceedings of the 8th International Symposium on Power-Line Communications and its
Applications (ISPLC’04), Zaragoza, Spain, 31 March - 2 April 2004, pp. 238-243.
[5] A. J. Snyders, H. C. Ferreira, A. W. Ballot, P. A. Janse van Rensburg, “The impact of coupling/filtering on
orthogonal M-FSK,” Proceedings of the 9th International Symposium on Power-Line Communications and its Applications
[6] A. J. Snyders, H. C. Ferreira, P. A. Janse van Rensburg, “Carrier gain adjustment for improved power-line
signal-to-noise ratios,” Proceedings of the 10th International Symposium on Power-Line Communications and its
Applications (ISPLC’06), Orlando, USA, 26-29 March 2006, pp. 39-43.
[7] P. A. Janse van Rensburg, H. C. Ferreira, A. J. Snyders, “Coupler winding ratio selection for effective power
transfer to a power-line communications receiver,” Proceedings of the 10th International Symposium on Power-Line
Communications and its Applications (ISPLC’06), Orlando, USA, 26-29 March 2006, pp. 290-295.
[8] P. A. Janse van Rensburg, H. C. Ferreira, “Classifying typical residential rooms for narrow-band impedance
adaptation,” Proceedings of the 11th International Symposium on Power-Line Communications and its Applications
(ISPLC’07), Pisa, Italy, 26-28 March 2007, pp. 47-52.
[9] P. A. Janse van Rensburg, H. C. Ferreira, “Dual Coupler for Dedicated Switching with Loads,” Proceedings of
the 11th International Symposium on Power-Line Communications and its Applications (ISPLC’07), Pisa, Italy, 26-28
March 2007, pp. 53-58.
[10] P. A. Janse van Rensburg, H. C. Ferreira, “Automotive power-line communications: favourable topology for
future automotive electronic trends,” Proceedings of the 7th International Symposium on Power-Line Communications
and its Applications (ISPLC’03), Kyoto, Japan, 26-28 March 2003, pp. 103-108.
[11] P. A. Janse van Rensburg, H. C. Ferreira, A. J. Snyders, “An experimental setup for in-circuit optimization of
broadband automotive power-line communications,” Proceedings of the 9th International Symposium on Power-Line
Communications and its Applications (ISPLC’05), Vancouver, Canada, 6-8 April 2005, pp. 322-325.
January 2005, pp. 64-70.
[13] P. A. Janse van Rensburg, H. C. Ferreira, “Coupler winding ratio selection for effective narrow-band power-
line communications,” IEEE Transactions on Power Delivery, vol. 23(1), January 2008, pp.140-149.
*****
111

TP 22

Uploaded by

Copyright:

Available Formats

TP 22

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

TP 22

Uploaded by

Copyright:

Available Formats

P.A.

JANSE VAN RENSBURG

D.Ing (Electrical and Electronic)

Supervisor: Prof. H.C. Ferreira

S EDERT die vroeë 1900’s is stroomdraende kragkabels al benut om terselfdertyd

C OUPLING a sensitive modem or an expensive measuring instrument to a live power-

Chapter 1 is intended to give the reader with little or no background on power-line

Impedance adaptation, by adjusting winding ratios of the coupling transformer, is investigated

[2] P. A. Janse van Rensburg, H. C. Ferreira, “Coupling circuitry: understanding 3

[3] P. A. Janse van Rensburg, H. C. Ferreira, “Step-by-step design of a coupling 4

[4] P. A. Janse van Rensburg, H. C. Ferreira, “The role of magnetizing and 5

[5] P. A. Janse van Rensburg, H. C. Ferreira, A. J. Snyders, “Coupler winding 6

[6] P. A. Janse van Rensburg, H. C. Ferreira, “Classifying typical residential rooms 7

[7] P. A. Janse van Rensburg, H. C. Ferreira, “Dual Coupler for Dedicated 7

[11] A. J. Snyders, H. C. Ferreira, P. A. Janse van Rensburg, “Carrier gain 4, 5

PLC History, Applications and Hindrances

The concept of superimposing an information-carrying signal onto a power

1.2 Inherited Hindrances

1.2.1 Attenuation and radiation

(a) (b) (c)

Fig. 1.3 Simplified representation of the cross-sectional area of a conductor that is

Network topology fundamentally influences the success of power-line communications [41].

In residential or low-voltage power-line communications, a complex star or branching

Component Selection and Circuit Layout

When designing a prototype coupling circuit for power-line communications,

2.2 Conductors and Circuit Layout

The self-inductance of a straight length of conductor, carrying alternating current, is [1]:

A fundamental equation defining inductance L carrying a current i, is given by Faraday’s law

In general, thyroidal transformers exhibit good characteristics (low parasitic reactances)

Components are purchased or manufactured to behave as ideal as possible under the

Functional integration of discrete passive components involves the manufacturing of various

Because of material and manufacturing constraints, neither functional- nor electromagnetic-

P OWER systems and communication systems operate at two different extremes –

Finally, it is suggested that components be characterized at low-frequency power-line voltages

I N essence, a coupling circuit is a special filter that allows a high-frequency

A power-line communications coupler needs to do two filtering tasks – blocking the

F ILTERING (ability to discriminate between different frequencies) forms the basis of

If a power-line communication system uses a sinusoidal carrier signal of 500 kHz

Inductors function complementary to capacitors: high-frequency signals (or harmonics) are

Z L = jωL = 2πfL∠ + 900 [Ω] (3.2)

In a power-line network, any series inductor would impede the high-frequency

It must be understood that a series capacitor or inductor usually functions as a first-order

for the HF path, and

3.2 Transformer Couplers

Because transformer design involves so many parameters it is often a non-precise,

• Choose a core with a large enough cross-sectional area.

Fig. 3.5 Suggested coupling circuit.

Design of a Bi-Directional Coupler

4.3.2 Impedance levels / winding ratio

4.3.3 Maximum voltage levels

4.3.4 Maximum power and current

4.3.6 Current density

J ≈ α⋅AP -0.125 [A/cm2] (4.4)

4.3.7 Skin effect

1 0.0699 [m] (4.5)

if the relative permeability µR of copper is assumed as 1. Symbol σ represents the conductivity

4.3.8 Number of strands

4.3.9 Number of turns

4.3.10 Flux density