John Dongbin Suh
Center for Communications and Signal Processing Department of Computer Science
North Carolina State University
CCSP-TR-88/11~
Suh, John D. Multiconductor Signal Propagation in Distribution Line
Carrier Networks (under the direction of J.B. O'Neal, Jr.)
A mathematical model for predicting multiconductor signal
propagation in distribution line carrier networks is formulated and
tested with empirical data. The multiconductor model accounts for
general source and load termination conditions and can be extended for
analyses of multiconductor systems of order 'n'. Current and voltage
propagation measurements (a total of 14 sets) conducted on actual
three-phase distribution lines are presented. It is shown that
certain discontinuities set up different wave patterns on each line,
which introduce electromagnetic coupling effects between phase
conductors at carrier frequencies.
Several test cases are computer-simulated to assess the
validity of the mathematical model. It is found that, overall, the
predicted results based on the multiconductor model agree with that
of the measured data, and hence, the mathematical model is valid.
The derivation of the distributed parameters of the multiconductor
model is presented and implemented in a computer simulation. It
is concluded that although the theoretically derived parameters are
adequate in predicting signal profiles, a higher degree of accuracy
ACKNO~LEDGEMENTS
The author ~ishes to thank Dr. J.B. O'Neal, Dr. Sasan Ardalan, and
Dr. Steven Vright of North Carolina State University; Kay Clinard, Lou Gale,
and other members of Carolina Power and Light's Distribution Automation
Research Unit; Ken Shuey of Yestinghouse Electric Corporation, and Jamey
Phillips for their invaluable guidance, support, and assistance in this
TABLE OF CONTENTS
Page
LIST OF SYMBOLS v
1. INTRODUCTION... . . .. 1
2. FORMULATION OF MATHEMATICAL MODEL FOR MULTICONDUCTOR 3
SIGNAL PROPAGATION
2.1 General Single-Phase Transmission Line Theory 3
2.2 Multiconductor Transmission Line Equations 7
2.3 Multiconductor Transmission Line Model 12
for Three-Phase Systems
3. PER-UNIT LENGTH IMPEDANCE AND ADMITTANCE MATRICES 19
3.1 Impedance of Cylindrical Wire with Return Path 20
3.2 Self and Mutual Impedances of Parallel Wires with 24
Unequal Current Distribution and Ground Return
3.3 Application of Carson's Line to the Derivation 27
of the Per-Unit Length Impedance Matrix
3.4 Derivation of the Admittance Matrix 30
3.5 Effect of Earth on Capacitance 33
3.6 Experimental Determination of Multiconductor 37
Line Parameters
4. NETWORK AND TEST DESCRIPTION 40
4.1 Test Set-Up and Measurements 40
5.1 Experimental and Theoretical Results of Test #lb 53 (Bundled Conductor)
5.2 Experimental and Theoretical Results of Test #lc 61
5.3 Effect of Capacitive and Inductive Loading on 63
Propagation (Decoupled Case Studies)
5.4 Experimental and Theoretical Results of Test #ld 66
5.5 Experimental and Theoretical Results of Tests 72
#2c and #2d
6. CONCLUSIONS 0 • • • • • • • • • • • • • • • • • • • • • • • • • • 84
REFEREl'lCES 87
APPENDIX A - CALCULATION OF LINE PARAMETERS: PER-UNIT ... 92
LENGTH IMPEDANCE [Z] M~D ADMITTANCE [Y] MATRICES
A.I - Calculation of [Z] for Three-Phase System 94
with Neutral Yire (Vertical Geometry)
A.2 - Calculation of [Y] for Three-Phase System 105
~ith Neutral Yire (Vertical Geometry)
A.3 - Per-Unit Paramters for Delta Configuration ... 110
APPENDIX B - CALCULATION OF DISTRIBUTED PER-UNIT 112 LENGTH IMPEDANCE Zp AND ADMITTANCE Yp FOR A
BUNDLED CONDUCTOR
B.l - Calculation of Zp for Bundled Conductor 113
(Vertical Geometry)
B.2 - Calculation of Yp for Bundled Conductor 118
(Vertical Geometry)
B.3 - Propagation Constant and Characteristic 119
Impedance
LIST OF SYMBOLS
V Voltage
I Current
R Resistance
L Inductance
C Capacitance
G Conductance
f Frequency
Z Impedance
Y Admittance
[ ] Matrix representation
y propagation constant
a Attenuation constant
a
Phase constantZ Characteristic impedance
o
Yo Characteristic admittance
r
Source reflection coefficients
r
L Load reflection coefficientH Magnetic field intensity
J Current density
B Magnetic flux density
uo Permeability of free space
8 Permittivity constant
r Radius
~ Flux linkage
s
GMD Geometric mean distance
Dm Mutual geometric mean distance
X Reactance
M Mutual inductance
q Charge density
E Electric field intensity
[f]
Propagation matrixK Kilo
Q Ohm
Hz Hertz
A Yavelength
v Velocity of propagation
p
A great deal of progress has been made in the research and
development of distribution power line carrier (DLe) technology over
the last ten years. Distribution line carrier communications enable
utilities to implement distribution automation applications, which
include load control, remote meter reading, line sectionalization,
and fault monitoring. Distribution line carrier, which is very
different from power line carrier (PLC) over transmission lines,
utilizes two-way communications from a central point to many remote
locations in a radial tree-topology network. In such a complex feeder
network consisting of taps and multiple branches, it is difficult
to achieve uniform signal strengths.
Extensive research has been done toward the development of
distribution automation systems {1},{2} in actual feeder networks.
Over the past several years, a series of DLC studies have been made
in which testing and simulation were possible in a controlled,
mini-mally complex environment. These studies were based on experimental
measurements conducted at Carolina Power and Light Company's
distri-bution automation test facility. Hemminger {3} researched the effect
of distributed transformer loading on the propagation of DLC signals
in a single-phase network. Borowski {4} later extended the
single-phase network to include branching, and analyzed network response to
line parameter variations.
signal propagation in multiconductor networks. As the number of
conductors in the system increases, electromagnetic effects of a
system of parallel conductors increase the complexity of obtaining a
solution to the classical multiconductor transmission problem. In
Chapter 2, the mathematical model for multiconductor distribution
lines will be formulated for a general system of 'n' conductors.
Steady state closed-form solutions for voltage and current as a
func-tion of distance will also be presented. In Chapter 3, the per-unit
length equivalent circuit parameters of the multiconductor model will
be defined and theoretically derived for a typical distribution network
operating at carrier frequencies. A description of the propagation
measurements along several multiconductor networks will be presented
in Chapter 4, and will be analyzed in Chapter 5, where several test
cases will be compared to simulations based on the theoretical model.
Finally, an assessment of the mathematical model and program
implemen-tation will be presented based on the correlation between experimental
Chapter 2 - Formulation of Mathematical Model for Kulticonductor Signal Propagation
Theoretical investigations into the propagation of
electro-magnetic waves in multiconductor transmission systems have been
carried out by various authors {7,8,13,36}. Although the theory
is developed for many applications, such as microstrip directional
couplers, shielded pair cables, and power line carrier networks, the
mathematical basis for modelling is common to any multiconductor system.
In the following sections, the mathematical model for multiconductor
signal propagation in distribution line carrier networks will be
formulated.
Section 2.1 - General Single-Phase Transmission Line Theory
The propagation of electromagnetic waves in overhead conductors
can either be characterized from a distributed parameter circuit point
of view using Kirchoff's equations, or by a more rigorous approach in
which the electromagnetic field in a multiconductor system is
deter-mined from Maxwell's equations. The latter approach is discussed by
Kuznetsov {8}, who expresses the solution to the wave equations in
terms of contour integrals. The former approach will be investigated
in great detail, since voltage and current are considered to be easily
measurable quantities, constrained to follow fundamental differential
The general transmission line equations for voltage and
current are well-known for single-phase networks, as are the various
equivalent circuit networks (T-type, L-type, Pye, etc.). The
distri-buted per-unit length equivalent circuit for an incremental length ~
is shown in Figure 2-1. From Kirchoff's circuit laws, we obtain the
following partial differential equations:
dV(X,t)/dX
ai(x,t)/dX
-Ri(x,t) - Lai(x,t)/at
-Gv(x,t) - Cav(x,t)/at
<2.1>
<2.2>
In Equations 2.1 and 2.2, also known as the general equations of
telegraphy, voltage and current are functions of two independent
variables, time (t) and distance along the line (x). The familiar
distributed parameters R,L,C, and G represent the per-unit length
resistance, inductance, capacitance, and conductance, respectively.
These distributed parameters, which appear in the above equations as
"coefficients" for single-phase lines, comprise the elements of the
distributed admittance and impedance matrices for multiconductor
systems, as ~e shall further investigate in Chapter 3.
Transformation of Equations 2.1 and 2.2 into the frequency
domain reduces the partial differential equations for voltage and
current to a system of linear ordinary differential equations as a
+
l'
R6X
L6X
G6X
6X
--- C6X
! I
I
Figure 2-1 Distributed per-unit length circuit model for
dV(x)/dx
dI(x)/dx
-ZI(x)
-YV(x)
<2.3>
<2.4>
where Z
=
R + jooL and Y=
G + jwC. Differentiation and crosssubstitution of Equations 2.3 and 2.4 yield:
ZYV(x)
YZI(x)
<2.5>
<2.6>
The solutions to the above system of linear, homogeneous
differ-ential equations are veIl known:
V(x)
I(x) (liZ )(Ae-yX - BeyX)
o
<2.7>
<2.8>
where characteristic impedance 2
=
(2/y)1/2=
(R+jwL)/(G+jwC)}1/2o
and propagation constant y
=
(2y)1/2. The constants A and Baredetermined by the imposed boundary conditions at the source and load.
It can also be shown {28} that closed form steady-state solutions
for voltage and current exist in the form:
V(x)
v
ZS 0
z
+ Zo S
I(x)
v
sz
+ Zo S
<2.10>
where V
s represents the source voltage, Zs the source impedance,
r
s the source reflection coefficient, andr
L the load reflectioncoefficient. In the following section we shall see how Equations 2.1
to 2.10 for single-phase transmission lines are related to systems of
equations for multiconductor transmission lines of 'n' parallel
conductors.
Section 2.2 - Multiconductor Transmission Line Equations
Several researchers {17,24} have studied the theory of
uniform multiple-coupled transmission lines. The mathematical
model for these multiconductor systems is used extensively in the
modelling and prediction of crosstalk in various environments. The
same model will be used to predict voltage and current propagation
in distribution networks consisting of 'n' conductors. The system
under consideration consists of 'n' overhead conductors, numbered
from 1 to 'n'. The conductors are assumed to be parallel to the
surface of the earth. Vl' V2, · .. ,Vn will represent the voltages in
each of the phase conductors. Likewise, II' 1
2, ... , In will
current on an elemental section of line dx, can be written in matrix
form as:
d[V(x)]/dx
d[I(x)]/dx
-[Z][I(x)]
-[Y][V(x)]
<2.11>
<2.12>
vhere [V] and [I] are vectors of dimension 'n x l' and [Z] and [Y] are
'n x n' per-unit length impedance and admittance matrices
respect-ively. Symbolically, for a system of 'n' conductors,
[V]
[I]
<2.13>
<2.14>
where T denotes transpose. Ma~rices [Z] and [Y] are arbitrarily
represented as:
2
11 Z12 Z1n
I 221 222 Z2n
<2.15>
[2]
I
I
I
I
I
Zn1 Zn2 '3nnI
Y11 Y12 Yln
l
Y
21 Y22 Y2n
<2.16>
[Y]
I
I
iI
Y
n1 Yn2 Ynn
J
\
-The elements on the diagonal of [Z], namely Z11' Z22' ... , Znn
represent self- or internal impedance. They take the form Z.. =
JJ
R..+jooL .. , where R.. is the internal resistance of conductor "j",
JJ JJ JJ
and L
j j its self-inductance. The off-diagonal elements Zjk (j~k),
h di ·b d t l ' d t b the J'th and kt h
represent t e Istrl ute mu ua In uc ance et~een
wires. They are expressed in complex form as Rj k + jwLj k. Similarly,
the diagonal elements Y11' Y22' ... 'Ynn represent self-admittances in
the form G.. + jwC .. , where G.. is the self-conductance and C.. the
JJ JJ JJ JJ
the self-capacitance. The off-diagonal elements Y
j k represent "mutual"
admittances in the form Gj k + jwC
j k, where Gj k represents mutual conductance and C
j k the line to line, or mutual capacitance. In
distribution networks where spacings bet~een conductors are
signifi-cant, the conductance component of the admittance is assumed to be
negligible. The derivation of each element in [Z] and [Y] will be
presented in Chapter 3. Intuitively, the off-diagonal, or "mutual"
terms of [Z] and [Y] exist because the lines are geometrically
voltage and current in each conductor. Differentiating equations
2.11 and 2.12 and cross-substituting yield:
2 2
d [I(x)]/dx
[Z][Y][V(x)]
[Y][Z][I(x)]
<2.17>
<2.18>
Note that these multiconductor transmission line equations correspond
to Equations 2.5 and 2.6 for single-phase lines. At this time, it
is convenient to define the following 'n x n' matrices:
Propagation Matrix [f] ( [Z][y])ll2
Characteristic Admittance Matrix
Characteristic Impedance Matrix [Z ]
=
Inv{([Z][y])1/2}[Z]o
where "Inv" denotes the inverse of a matrix. The corresponding
multiconductor solutions to Equations 2.17 and 2.18 are analogous to
the single-phase solutions (Equations 2.7 and 2.8), and are obtained
by solving a system of '2n x 2n' linear, homogeneous second-order
differential equations:
[V(x) ]
[I(x)]
<2.19>
where 'n x n' matrices [A] and [B] are determined from a set of '2n'
boundary conditions. In solving the equations for a system of 'n'
parallel conductors, there are generally 'n' roots (eigenvalues) that
correspond to 'n' "modes" of propagation. This approach involves the
diagonalization of the matrix product [YZ] in order to isolate or
effectively decouple the voltage and current equations.
In {13}, Paul presents a method which utilizes chain
matrix parameters in constructing solutions to the classical
multi-wire transmission line equations. In {14}, several matrix identities
are given by Paul, along with a set of matrix equations which
incor-porate terminal constraints for the total solution of line currents.
Similarly, a highly systematic technique for solving systems of
multiconductor equations which utilizes Green's matrix is formulated
by Gruner {39}. Gruner's method is valid for arbitrarily
ter-minated networks and can be applied in various situations, which
include voltage or current excitation applied at any point along the
network. Alternatively, a derivation by Riddle {33} arrives at
multiconductor closed-form solutions for voltage and current as a
function of line length. These closed-form solutions, which exist in
matrix form for a multiconductor system, are very similar to Equations
2.9 and 2.10 for single-phase systems, and are convenient for
com-puter simulation and algorithm development. The aforementioned
methods to the solution of the multiconductor transmission line
simulations in {33}. The method developed by Riddle is shown to be
more computationally efficient and versatile. Consequently, the
measured results (to be discussed in Chapter 5) are simulated using
the closed-form multiconductor solutions developed by Riddle.
Section 2.3 - Hulticonductor Transmission Line Model for Three-Phase Systems
~e will no~ consider a multiconductor system composed of three
homogeneous lines above a neutral plane. The per-unit length
equiva-lent circuit model for an elementary length dx is sho~n in Figure 2-2;
The mutual coupling elements in [Y] are connected in shunt from phase
to phase. Similarly, the self-capacitance terms located on the
diagonal are represented by a phase to neutral shunt connection.
Distributed resistances per-unit length are connected in series with
the line. The arro~s between conductor self-impedances denote the
mutual inductive coupling between phases, which is analogous to ideal
transformer coupling. Assuming that all conductors are homogeneous
implies that [Y] and [Z] are symmetric matrices. Hence, Yjk=Ykj and
Zjk=Zkj. From Figure 2-2 we utilize Kirchoff's relations to
produce a system of six differential equations:
dV (x)/dx
1 <2.21>
Ij
(x~ Ij(~+h),.
~
I
ZjJ
dx.
+
+
\.
V/yJ Y,odxJJ
\
V/:t:+dx)
I
ref
-
",
~k(h
" " '" "-,
I
I I
V
k(x)I Ykkdx Vk(x+dx)I
I
I
Zkk
dx
+
+
I
l-"
4'(~)
I
k(x+d~)<2.23>
<2.24>
<2.25>
<2.26>
The above system of equations can be expressed in the matrix form
of Equations 2.11 and 2.12 as follows:
V1(X)l
(
,...
Z13l I 1(X)l
I 2
11 212
I
d/dx V
2( x ) 221 ?~22 223 12( x ) <2.27>
V
3( x ) 231 232 233 1 3( x )
J
"
iYll+Y12+Y13'"
("
I I(x )I -Y12 -Y13 \I Vl ( X) ,
I
I
d/dx 1
2( x ) ' = - I
I
-Y21 121+Y22+Y23 -Y23 V2( x )<2.28>
I
l
V3( x ) 1
3( =< ) 1/
I
-Y31 -Y32 Y31+Y32+Y33"
From the system of equations above, vhich ~ere determined by the
per-unit length multiconductor model of Figure 2-2, it is evident that as
the number of conductors increases, the general solutions become more
the incorporation of terminal constraints further complicate the
solution process. Thus, we assume that the system is composed of
uniform conductors, and that wave propagation is constrained to TEM,
or "quasi-TEM" mode. A fundamental property of TEM (transverse electro-magnetic) waves is that the components of the electric field intensity
vector E and magnetic intensity vector H only exist perpendicular
to the direction of propagation along the line. This implies that
E and H are zero, where x denotes the distance along the line.
x x
In a strict sense, since losses are eminent in any real transmission
network, the longitudinal components E and H are not zero, and
x x
hence, we cannot assume pure TEM wave propagation. The waves actually
exhibit a combination of TE and TM modes of propagation. These
"hybrid" waves are referred to as "quasi-TEM" waves. Since the
longitudinal electric and magnetic field components are assumed
to be considerably smaller than the transverse components, the
so-called "quasi-TEM" wave can be approximated by the TEM wave. For the remainder of the discussion, we will assume "quasi-TEM" wave
propagation.
For a system of 'n' equations, '2n' boundary conditions must
be incorporated to solve for the constants ('nxl' vectors) [A] and
[B) in equations 2.19 and 2.20. Note that these matrices are
equivalent to the constants associated with forvard and backvard
travelling waves for single-phase transmission lines. For our
specific system (as modeled in Figure 2-2), a total of six boundary
Ye will consider for simplicity, a linear, reciprocal network
in which a source voltage is applied at the sending end of only one
line, the other two lines terminated in an arbitrary impedance between
phase and neutral at the source, as shown in Figure 2-3. Similarly at
the load, arbitrary impedances are also connected between phase and
neutral. Consequently, the source and load impedan~es of the system
can be represented as 3x3 diagonal matrices. It is often convenient
to represent source and load impedances as respective admittances in
shunt with a Norton equivalent current source in order to accomodate
the open-circuit load, ~hich is quite ·common in distribution networks
operating under normal conditions. Thus, the source and load
termina-tions can be represented by diagonal admittance matrices as follows:
a
a
a
a
o
a
o
a
a
a
o
l
o
IYL3
J
<2.29>
<2.30>
source and load matrices. Additional line to line terminations would
create off-diagonal terms (The analysis for incorporating these
termination conditions is discussed in detail in {l} and (12) ).
Yith boundary conditions known at the source and load, along
with the per-unit length impedance and admittance parameters, the
closed-form expressions for voltage and current as a function of
dis-tance can be solved.
The solution to the multiconductor transmission problem is
important in characterizing forms of crosstalk, which exist in a
multitude of applications, including distribution line carrier. The
coupling effects, ~hich must be analyzed from an electromagnetic point
of view, inherently reside in the per-unit length circuit parameters,
Chapter 3 - Per-unit Length Impedance and Admittance Matrices
The distributed parameters of the mathematical model
present-ed in Chapter 2 will be definpresent-ed by considering individual elements
of the impedance matrix [Z] and the admittance matrix [Y]. The terms
of [Z], namely the self and mutual resistances and inductances, can
be derived by examining the flux linkages both internal and external
to the conductor. Equations describing self and mutual impedances
are based upon a modification of Carson's line equations for wave
propagation in parallel overhead wires with ground return. The- 1
actual calculations become complicated at higher frequencies when
currents tend to redistribute themselves toward the outer surface of
the conductor, thus resulting in an increased resistance per-unit
length and a decreased inductance per-unit length.
The admittance matrix is composed of self and mutual
con-ductance and capacitance terms. The capacitance terms can be
derived by considering the physical conductor geometry in reference
to an e~uipotential earth surface. The conductance per-unit length,
on the other hand, is affected by factors that may not be
control-lable or measurable.
The derivation of net~ork parameters [Z] and [Y] will be
outlined in this chapter. Other references {11,30,31,32) contain
more complete and rigorous analyses of these parameters.
several test cases, which will be discussed later in Chapter 5. An
experimental method for obtaining actual network parameters from
open-circuit and short-circuit input impedance measurements will
also be presented in this chapter.
Although the derivations of [Z] and [Y] are considered
separately by effectively isolating respective magnetic and electric
fields, it is evident that the equations for voltage and current
accommodate both field effects. Hence, the term transverse
electro-magnetic (TEM), or "quasi-TEM" (as discussed earlier) applies in the formulation of per-unit length distributed parameters.
Section 3.1 - Impedance of Cylindrical Vire with Return Path
From Ampere's law for static magnetic fields,
,. r
eH-dI = : J.ds I , <3.1>
which states that the line integral of a static magnetic field
intensity around a closed path must equal the total current enclosed
by the path. For a typical segment of cylindrical ~ire, as sho~n in
closed path
ds
;>
II
;'~
I
~
J
1
I
(
..
(
,
\
\ ~
\
\",
1 <3.2>
<3.3>
H<p 1/2nr <3.4>
The magnetic flux density B around this path is expressed as:
B u H
o <3.5>
where u is the permeability of free space (u
o 0
-7
4nxlO Him) •
Thus, B 2xlO-7I l r <3.6>
Assuming a medium of constant permeability, the equations for
inductance of a return circuit consisting of two parallel wires can be
derived. The total inductance of the circuit is found by dividing the
sum of the internal and external flux linkages by current I. Yoodruff
{S} expresses the total number of flux linkages per meter length about
one wire as:
~tot 10-71{2ln(D/r)+u/2u)o <3.7>
where D denotes the distance between phase and neutral wire and r the
radius of the conductors (assuming both are homogeneous). Thus, the
L lO-7(2ln(D/r) + u/2u )
o <3.8>
Woodruff extends the theory to a parallel system of 'n' homogeneous
conductors, in which mutual geometric mean distances (GMD) and
geo-metric mean radii (GMR) are utilized. As a result, the total number
of linkages about a conductor with self GMD D and mutual GMD D
s m
with respect to neutral current is:
~tot 2xlO-7I{ln(D /D)} linkages per meter
m s <3.9>
Dividing by I and converting to units of miles, we obtain the
following expressions for inductance and reactance:
O.3219(ln(D /D)} mH/mile
m s
2.020xl0-3j{ln(D /D)} ohms/mile
m s
<3.10>
<3.11>
From these equations and Carson's line equations, we can derive line
parameters (see Appendix B for derivation of Z and Y ) for a
p p
homogeneous three phase system in which phase conductors are injected
with the same current, and this effectively reduces to a single
Section 3.2 - Self and Mutual Impedances of Parallel Vires with Unequal Current Distribution and Ground Return
Ye will now consider a group of parallel, non-zero, current
carrying conductors, in which each wire experiences an induced voltage
due to flux linkages between current-carrying conductors.
Two segments of parallel wires (denoted by a-a' and b-b') are
shown in Figure 3-2. This particular circuit model will be used to
describe self and mutual inductance terms of the per-unit length
impedance matrix [2]. The circuit model is analogous to a one-turn
air core transformer equivalent. From field theory, if an applied
potential V creates a current I in the direction shown, a
mag-aa a
netic flux ~ba linking coil 'b' due to the current in 'aT will be
established. Lenz's law states that a counterflux ~ab will oppose
~ba' thus creating an induced current in the direction b-b'. Thus,
a mutual impedance term establishes the effective induced voltage in
the opposing wire, and the circuit equations may be written as:
V
a V 'a
v '
b<3.12>
<3.13>
Several
where Zaa
=
Ra a + jwLaa, and Zab = Ra b + jwLa b , etc.
references {31,32} show the computations for parallel cylindrical
wires. This involves the summation of partial self-inductance terms
~~
+'10(4.' - - - - ~ T:l~Cl I
-0.-
«
a
~+ 0 t
Va.
Vr;.,1
- .i-
o-.-L
to..b :
t
htL
I I
liLt-
---~I.
-- ---T-- tI
lH
I
b
Ib
~ j,'+
GI\IV' )
0 +Vb Vb'
---L.
.L
where length of line's' is much larger than radius r, the inductance per
unit length is:
Lis 10-7/2 + 2x10-7{In(2s/r)-1} <3.14>
Since the GMR for cylindrical wires is D
s
therefore express self-inductance as:
O.779r {5}, ~e can
L 2x10-7{In(2s/D )-l}
s <3.15>
Likewise, mutual inductance Mis determined from geometric mean
distance (D ) and is defined as: m
M 2xlO-7{In(2s/D )-l}
m <3.16>
Although the above formulas for self and mutual inductances imply that
they are functions of line length s, we shall see that these terms
"cancel out" when equivalent expressions for self and mutual impedances
Section 3.3 - Application of Carson's Line to the Derivation of the Per-Unit Length Impedance Matrix
The basis for describing wave propagation in overhead
conduc-tors with earth return was presented by J.R. Carson in 1926. Various
authors (11,30} have used Carson's line with earth return in
trans-mission line applications, such as zero-sequence impedance calculations
for fault analysis. Here, we will derive self and mutual impedance
terms using a rather heuristic approach in which earth return is used
in the circuit model. This involves the utilization of Carson's line
with earth return, as shown in Figure 3-3. An overhead wire of unit
length (denoted by length a-a') carries a conductor current I , and
a
returns through the earth through a ficticious "ground conductor"
beneath the surface of the earth (denoted by length g-g'). Similar to
the method of images, which is commonly used in the computation of
sequence capacitances, the earth is assumed to extend infinitely with
uniform resistivity. The distance between the overhead conductor and
the ficticious "ground conductor" is denoted by D . This distance
ag
is a function of earth resistivity p, and is adjusted so that the
calculated inductance is equal to that measured by test (30}. From
equations 3.12 and 3.13, we can represent Carson's line (Figure 3-3)
I z
a. a aa I
.-.,.. a
+
r
v
a
D REF
ag
)
7V
!
g
= -I a
J
+
---...
~I Fictitious earth3
return conductorI,.
1 UNIT·1· .
I
z ag
[Va - va'l= 'Zaa
lVg-Vg'j
: -, II
' -I
a J
<3.17>
Note that voltages V , V " V , and V ' are all referenced to ground.
a a g g
Thus, we know that Vg 0, and V ' - V 'a g
=
o.
Subtracting the twoequations enables us to solve for V :a
V
a (zaa + Zgg - 2zag)1a Zaa aI
where Zaa z + Z - 2z
aa gg ag Zaa denotes the "total" self
impedance of conductor "a" vith earth return accounted for, whereas ..
lower case z denotes the self-impedance of conductor "a" without aa
earth return. Zaa can be regarded as the total self impedance, since
it contains an earth resistance term r .g From equations 3.15 and
3.16, we can express each component of total self-impedance Z as:
aa
z aa
z gg
z ag
r
a + jOOk{ln(2s/Ds a)-1} Q/unit length
r + jwk{ln(2s/D )-1} Q/unit length
g sg
jwk{ln(2s/Dag)-l} Q/unit length
<3.18>
<3.19>
<3.20>
where Dsa and Dsg denote self GMD's of conductors "a" and "(J""o ,
respectively. Combining terms from Equations 3.18, 3.19. and 3.20 and
Z
aa (Ra g e+ R ) + jwkln(D /Dsa) <3.21>
where De is commonly defined as De = D 2/ D {1l}. The parameter
ag sg
De is dependent upon both earth resistivity p and frequency!, and
is defined by:
D
e 2160(p/j)1/2 <3.22>
An identical methodology is followed for deriving a system of three
phase conductors ~ith or vithout ground wires. The derivation is shown
in Appendix A. Given the physical geometries and conductor
specifi-cations, it is possible to theoretically calculate per-unit length
impedance parameters. Factors such as skin effect can be estimated by
a method also shown in Appendix A and also in {22}.
Section 3.4 - Derivation of the Admittance Matrix
Just as magnetic field effects are considered for studying
inductance, the distribution of the electric fields determine the
capacitance of a system of parallel conductors. The shunt admittance
matrix [Y], as mentioned earlier, consists of conductance and
capacitive reactance terms. However, the conductance term is usually
Unknown and often uncontrollable factors such as changes in
atmos-pheric conditions, dirt, and corona contribute to leakage current
between conductors, which often make conductance impossible to
measure. Thus, we assume negligible conductance contribution at
at distribution voltages and consider only capacitance terms.
Capacitance between conductors is defined as charge per unit
of potential difference. It is dependent upon the size and spacing
of the conductors relative to each other and to an earth conducting
plane. Intuitively, we can visualize current, or the movement of
charge, to increase and decrease with the instantaneous value of the
alternating voltage impressed on the system. This is evident in
standing waves, where charging current flows even in the presence
of an open-circuit load, as we shall later investigate.
Recall from field theory that the potential difference
between tva points P1 and P
z
external to a linear charge density q(see Figure 3-4) is equal to the integral of the potential gradient E:
J E-dx
J
(q/2n£x)-dx <3.23>Thus, by superposition, for an n-wire system carrying charge
densities qa' qb,···,qn' located above a ground plane, the
difference in potential between any two wires will be the sum of
+q
-
.../
f
I II
/
/
/
~
/
/
/
'"
,..("
/ '
v .
aJ (1/2n8)(qa1n{Daj/r} + qbln{Dbj/Dba} +
+ ••• + qnIn{D ./DnJ na})
<3.24>
Section 3.5 - Effect of Earth on Capacitance
The presence of earth as a conducting medium must be
account-ed for when calculating capacitance. The assumption that the earth
is a perfect conductor of infinite extent in a horizontal plane will
enable us to understand the effects of a conducting earth on
capacitance calculations.
Consider a parallel two conductor system with earth return
as shown in Figure 3-5. The physical location of the conductors is
defined with respect to a coordinate system in which the earth plane
is used as the horizontal reference axis and the axis of symmetry of
the pole structure as a vertical reference. In charging the conductor,
the earth surface and conductor plane can be regarded as equipotential
surfaces, since the earth has a charge equal in magnitude to that of
the conductor but opposite in sign. Assuming the earth is of uniform
resistivity and infinite, its surface can be replaced by a ficticious
conductor of the same size and shape as the overhead conductor at a
di~tance equal to that of the overhead conductor to earth. This
en-tails that if the earth is removed and a charge equal and opposite to
pole structure
conductor j
/
I
I
/!
!)
II
d ..
1J
conductor i
earth
I
I
I
I
I
I
~
image ofconductor i
\
D .. \
1J
\
\
\
I
I
I
I
I
I
I
I
1
image ofconductor j
1
~
j
then the plane midway between the two occupies the same position as
the equipotential surface. This ficticious conductor, having charge
equal and opposite to that of the overhead conductor is called the
image conductor.
Thus, since calculations involve only lengths between
con-ductors and their respective image conductors, the admittance matrix
is dependent only upon the physical geometry of the conductors
relative to earth.
From equation 3.22, a system consisting of four overhead
conductors can be equivalently expressed in matrix form as:
V B B
ab B B
\
;' qa
1
a aa ac an !
Vb 1/2Jt£ Bba Bbb B
bc Bbn
I
:: I
<3.25>
Vc Bca Bc b B Bcd
cc
V B B i
nb B B qn
J
n na nc nn
where the elements of matrix [B] are determined by the geometry of the
conductors as follows from Figure 3.5:
B ..
1J 1n(D . .IJ/d .. }1J <3.26>
dij diIStance between 1.th and ' thJ con uctor ford (i~j)
radius of it h conductor for (i:j)
D.. distance between the jth conductor and the image of the ith
conductor.
Equation 3.25 can also be expressed in a form similar to equation 3.23:
vhere
[V] = 1/2n£[B] [1']
~
=
[qa qb qc qn]T<3.27>
We can arbitrarily define a charge coefficient matrix [P] as:
[P]
=
(1/2Jt€)[B]where [P]
=
[C]-I, since q[V] = [P]['f]
cV and
<3.28>
<3.29>
To obtain the total per-unit length admittance matrix [Y], we apply the
following relations: From Ohm's law, the current vector [I] is:
[ I ] [Y] [V] <3.30>
Current is also defined as the derivative of charge vith respect to
time. Thus,
[ I ] d['f]/dt j
-r
'¥] <3.31>[ I ]
Thus, from 3.30:
jw[C] [V] <3.32>
[Y] [1] [V]-l JW. p-1 jw(2nEB-1) <3.33>
The complete calculation of the per-unit length admittance matrix for
the actual test network is given in Appendix A for a four conductor
system (Three phase wires, one neutral) with earth return.
Section 3.6 - Experimental Determination of Hulticonductor Line Parameters
A measurement technique for determining the per-unit length
parameters of a multiconductor network is presented in {21}. It is
formulated in terms of measurable short and open-circuit line
impedances at a particular frequency. From transmission line theory
for single-phase lines, the short-circuit input impedance Z and
sc
open-circuit input impedance Z can be expressed in terms of
atten-oc
uation constant a, phase constant
a,
characteristic impedance Z , ando
line length 1 as:
z
sc
z
ocZ tanht« + jS)l
o
Z cotht« + jf3)l o
<3.34>
Multiplying equations 3.34 and 3.35 and solving for Z we obtain:
a
Z
o {Zsc ocZ }1/2 <3.36>
The expression for propagation constant in terms of Z and Z is:
sc oc
y = ex + jS = (arctanh{Z IZ }1/2)/1
sc oc <3.37>
Similarly, for a multiconductor system consisting of{N' conductors
(excluding ground wire), the resulting expressions for the
short-circuit and open-short-circuit input impedance matrices are:
[tanh(fl)][Z ]
o
Inv[tanh(rl)]-[Z] o
Solving for Z we obtain:
o
<3.38>
<3.39>
[Z ]
o {[ Zsc]Inv[Zoc]}-1/2 [2sc] <3.40>
The propagation matrix [f] is expressed as:
[ f]
1/')
{arctanh([Z ]Inv[Z ]) ~}/l
Having obtained [Z ] and [f] from equations 3.40 and 3.41, we can
o
solve for [Z] and [Y] by the following relations {14):
[Z]
[ Y]
[f][Z] o
Inv[Z ]·[f]
a
<3.42>
<3.43>
Thus, from the knowledge of the input impedance matrices for
open-circuit and short-open-circuit load conditions, the multiconductor line
parameters [Z] and [Y] can be obtained.
A technique for measuring these input impedance matrices
is presented in {21}, where ratios of voltage to current are measured
by effectively "isolating" self and mutual impedance and admittance
terms at a chosen frequency. The reader is referred to {21} for a
detailed explanation of this measurement procedure and the results
for a four-conductor line.
This experimental method of calculating line parameters could
not be implemented in our tests because the network was not conducive
to the measurement of open-circuit and short-circuit input impedances.
Thus, the validity of the mathematical model can only be determined
Chapter 4 - Network and Test Description
In order to gain an understanding of carrier signal
propa-gation on multiconductor distribution lines, several tests ~ere
performed on actual de-energized distribution networks at Carolina
Power and Light's Distribution Automation Test Facility. The test
facility offers a controlled environment in which propagation
measurements of voltage and current as a function of distance can
be performed. The 23 kV test facility, which was constructed to
Carolina Power and Light's distribution engineering standards, is
composed of spans of single-phase and three-phase sections of line.
These spans can be configured into various lengths of three-phase
and single-phase "netTHorks" by controlling oil break switches located
at various "switching poles". A more detailed description of the
test facility is presented in {6}.
Section 4.1 - Test Set-Up and Measurements
Propagation measurements were performed by injecting a 25 kHz
carrier signal at the sending end of a multiconductor network and
measuring voltage and current magnitudes at approximately equidistant
intervals along the network. The 25 kHz frequency was also used in
previous OLe experiments {3,~} in order to "visualize" nodes and
antinodes in standing ~ave patterns. These signal nulls occur at
1.77 miles at 25 kHz. Due to the physical limitations of the network
(approximately four miles each of three-phase and single-phase sections
of line), about one-half of a wavelength can be plotted for either
the three-phase or single-phase spans, which should yield sufficient
information for standing vave analysis.
Although the test facility accommodates both single-phase
and three-phase sections of overhead and underground distribution
lines, only the overhead conductors were utilized. These conductors
are classified as #2 AVG aluminum, and are spaced according to
distri-bution standards (Refer to Figures A-2 and A-3 as shown in Appendix
A). The neutral conductor follows an intermittent grounding scheme,
where grounding occurs at each pole (located about 300 feet apart).
Figure 4-1 shows the experimental set-up for carrier injection
at the sending end of the network, where a function generator (HP
3311A) in series with a power amplifier (HP 467A) is used to generate
the 25 kHz sinusoidal carrier signal. The signal is coupled to the
phase (denoted by A,B,C) and neutral (N) conductors by twisted pair
16 gauge wires, which are clamped on to each of the overhead
conductors. A connection box (banana-type connectors) vas built to
select either single-phase injection on conductor A, or three-phase
injection on all phases A,B, and C. The peak-to-peak source voltage
used for the propagations tests was SO volts- At the receiving end,
16 guage twisted pair wire was also used for any loads which were to
HP 3311A Function Generator
HP 467A Pover
Amplifier
R
R lOkQ
Frequency
Counter (25 KHz)
/
TO PHASE B
Figure 4-1
Connection Box
to phase.
The actual voltage and current measurements were performed in
a bucket truck using a battery-operated portable dual-channel
oscillo-scope (Tektronix 305 DHM). The voltage probe was modified by attaching
large alligator-type clips to signal and ground leads of a coaxial
cable, enabling voltage measurement from phase to phase and phase to
neutral, simply by clamping on to the wires. The current measurements
were made using a combination of a Fluke current transformer (clamp-on
with 1000:1 turns ratio) and a Tektronix current probe (model P6021).
In order to compensate for the 1000:1 decrease in current, the output
current of the Fluke was increased by placing 500 turns (type 40 A~G
wire) of the Fluke's secondary to the primary side of the Tektronix
current probe. Thus, the calibrated net transformer ratio was
approx-imately 1.75:1 at 25 kHz. Currents were measured on phases A,B, and
C only, since Hemminger {3} showed that almost no current flowed on
the neutral conductor more than two pole spans (about 600 feet) from
the source.
Section 4.2 - Network Configurations and Boundary Conditions
Multiconductor propagation measurements of voltage and
cur-rent were performed on three diffecur-rent networks. These net~orks are
shovn in Figures 4-2, 4-3 and 4-4. "Network #1", as shown in Figure
4-2, consists of a homogeneous span of parallel conductors A,B,C,
~_._--- 3.59 miles ---~
I
a O~--~---..,O
b o---~
...--..---..-;---....
---=o
C O,..,~~-
...
I:.:III:I:lI----...
---_rj
n
(")c~---...
o
source
Figure 4-2
TABLE 4-1 Boundary conditions for tests conducted on Network #1
# OF PHASES SOURCE LOAD
TEST # INJECTED CONDITIONS CONDITIONS
ZSA SQ ZLA OPEN
1
la 2
SB OPEN ZLB OPEN
(PHASE A)
ZSC OPEN ZLC OPEN
ZSA
=
5Q ZLA OPENIb 3 2
SB 5Q ZLB OPEN
ZSC 5Q ZLC OPEN
ZSA SQ ZLA OPEN
1
lc ZSB lOKQ
ZLB +jZol
(PHASE A)
ZSC lOKQ ZLC -jZ02
ZSA 52 ZLA OPEN
Id 3 ZSB SQ ZLB +jZol
ZSC SQ ZLC -jZ02
2
corresponds to Figure 4-2, where the parallel conductors assume a
vertical geometry with phase A on top, B in the middle, and C on the
bottom, the closest to neutral conductor N. The total length of the
network is approximately 3.59 miles, or 18,962 feet. In a sense, the
network is symmetrical in that all conductors (A,B,C,N) span the same
total length, (i.e. there are no discontinuities that result from
unequal line lengths). Thus, this particular network is easily
modeled as one continuous, homogeneous section of line. In actuality,
this is not the case due to variations in conductor geometry
result-ing from transpositions, differresult-ing pole heights, etc. However, for
simplicity, we shall model this "symmetric" network as one continuous
section of line, characterized by the same set of distributed
per-unit values throughout.
A series of tests was made on Network #1 by applying
sev-eral combinations of boundary conditions at the source and load.
At the source, current was injected on either: (i) phase A only
(single phase injection), or (ii) on all phases A,B, and C
(three-phase injection). On the receiving end, load conditions were applied
as follows: (i) all phases A,B,C open, or (ii) phase A open, phase B
terminated in +jZ , and phase C terminated in -jZ (loads
connect-o 0
ed between phase and neutral). No phase to phase source or load
terminations were applied. The combinations of two source and two
load conditions resulted in a series of four independent tests
for example denotes the load impedance connected between phase A and
neutral. Likewise, ZSA denotes the source impedance connection between
A and N. The term "open" implies an open-circuited load.
Network #2, as shovn in Figure 4-3, is very similar to Network
#1 with the exception of phase A, which extends about 1.18 miles
further, spanning a total length of 4.77 miles. The lengths of Band
C are unchanged at 3.59 miles. The extension of phase A causes the
network to be asymmetric. Hence, a discontinuity exists in the
net-work, for which two sets of line parameters must be employed to the
multiconductor solution equations presented in Chapter 2. Some
interesting coupling phenomenon result from this discontinuity, as
we shall see in the next chapter. The source and load conditions
for this network and the corresponding test cases are summarized in
Table 4-2. Note that the load conditions change only for phase A.
Network #3, shown in Figure 4-4, adds another discontinuity
by effectively reducing the length of phase conductor C by one-half.
Here, we are interested in the effect of discontinuities (which result
from an open-circuit) on the propagation in each of the phase
conduc-tors. The three-phase netvork is reduced to a two-phase network after
propagating 1.82 miles, and is then further reduced to a single-phase
network after traveling 3.59 miles. Because the test facility is
composed entirely of single and three-phase distribution lines, a
"true" tl,rJo-phase system W'as not physically configurable. However, it
4.77 miles
!,
"
3.59 miles
b o~---o
I
I
c o~---o
,
n
0---...0
i .. source load
I
~I
TABLE 4-2 Boundary conditions for tests conducted on Network #2
# OF PHASES SOURCE LOAD
TEST # INJECTED CONDITIONS CONDITIONS
ZSA = SQ ZLA SHORT
1
2a ZSB OPEN ZLB OPEN
(PHASE A)
ZSC OPEN ZLC OPEN
ZSA = SQ ZLA OPEN
1
2b ZSB OPEN ZLB OPEN
(PHASE A)
ZSC OPEN ZLC OPEN
ZSA SQ ZLA SHORT
2c 3 ZSB 5Q ZLB OPEN
ZSC SQ ZLC OPEN
ZSA 5Q ZLA OPEN
2d 3 ZSB 52 ZLB OPEN
--- 4.77 miles
I ~ 3.59 miles
~._---_.
b O~---O
~ 1.82 miles ~
I
C
c...
---~o--n c,·~---~---~---.--.-(O
!
load--~
TABLE 4-3 Boundary conditions for tests conducted on Network #3
# OF PHASES SOURCE LOAD
TEST # INJECTED CONDITIONS CONDITIONS
ZSA SQ ZLA SHORT
1
3a ZSB lOKQ ZLB OPEN
(PHASE A)
ZSC lOKQ ZLC OPEN
ZSA
=
5Q ZLA=
OPEN1
3b ZSB lOKQ ZLB OPEN
(PHASE A)
ZSC lOKQ . ZLC OPEN
ZSA
=
SQ ZLA 4101
3c ZSB lOKQ ZLB OPEN
(PHASE A)
ZSC lOKQ ZLC OPEN
ZSA
=
5Q ZLA SHORT3d ')
ZSB SQ ZLB OPEN
..J
ZSC SQ ZLC OPEN
ZSA SQ ZLA OPEN
3e j
ZSB SQ '7 OPEN
...) L1
LB
ZSC SQ ZLC OPEN
ZSA SQ ZLA 410Q
3f 3 ZSB SQ
ZLB OPEN
wire physically exists with both ends open, and can therefore be
considered as an equipotential neutral conductor. Table 4-3
sum-marizes the various source and load boundary conditions that were
applied to Network #3. A total of six tests were conducted. Note
again that loads ZLB and ZLC at their receiving end remains open in
all test cases. The 10 kilohm source impedance was used to simulate
the impedance of a substation transformer {3}.
The results of all propagation tests are tabulated and
plotted in Appendix C. Because of the extensive amount of
experi-mental data, a few tests, particularly those which were conducive
to simulation, will be analyzed and commented upon in greater detail
Chapter 5 - Test Results and Analysis
The experimental and theoretical results of several test
cases are analyzed in the sections to follow. Because of the
ex-tensive amount of data from the various tests described in the last
chapter, only a few of the more significant results will be discussed.
Computer simulation results will be compared to empirical data to
check the validity of the mathematical model.
Because the solution process to the multiconductor problem
is so mathematically complex, it lends very little insight to the
actual physical occurrences along the distribution network.
Conse-quently, an analytical approach will be taken which involves the
application of transmission line fundamentals. The Smith chart, for
example, provides an accurate solution and a physical interpretation
of what happens on the line.
Section 5.1 - Experimental and Theoretical Results of Test lIb
(Bundled Conductor)
As described in Table 4-1, test case #lb was conducted on
a symmetric netvork, ~here each conductor spans a distance of
approx-imately 3.59 miles. For this particular test, three phases (A,B,C)
were injected with a 25 kHz sinusoid. At the load, all phases were
terminated in an open circuit. The standing wave patterns for voltage
A total of ten locations were chosen as measurement points. A spline function was used in plotting the subsequent standing wave patterns
for each phase. Note that the voltage minimum and current maximum
both occur approximately at the midpoint of the network, a
quarter-wavelength from the load. This is because the network length is very
close to an electrical half-wavelength at 25 kHz.
For a lossless line, a half-wavelength (\12) at 25 kHz
corresponds to a distance of 3.73 miles. From previous propagation
tests conducted by Hemminger {3}, the velocity of propagation of an
unloaded distribution line was measured to be 95% of the speed of
light in air, which indicates that the line is virtually loss less and
conducive to TEM, or quasi TEM modes of propagation. Thus, an
electri-cal half-vavelength adjusted for 95% speed of propagation is actually
about 3.54 miles, which is very close to the total netvork length of
3.59 miles. The symmetry of the standing wave patterns indicates that
the network is indeed very close to a half-wavelength.
Figure 5-1 shows the voltage standing wave pattern for phases
A, B, and C. Because of the uniformity of the network (all phases
being of equal length and all conductors uniform), we would
anti-cipate that all phases would exhibit identical standing ~ave patterns.
In essence, this particular three-phase system can be considered as a
single-phase bundled conductor, in which the three phases constitute
the bundle. As discussed in Chapter 3, a system of parallel wires
carrying unequal currents will mutually induce voltage on neighboring
PHASE TO fo'[UTRAl VOltAGES (T£ST,es) '0 to .0 , o L 10
,
S ao 10 ~~ ... I"...
It .• .,. ..... l." I." . . . . I.... ..M ,... 1.&1
o.04 0 40 0 ' .0 1. 20 I.60 2 . 00 J . I t I . " I. 19 J.Jt
DUrAlIIICI n . SOuUI (MIL'S)
Olst·JnC. fr'Olfo ~~ (",tlea)
PHASE CuRMNTS (TEST'I8)
'ea • -..----I
..
• ---- .,
u.
-....
r:
'" .' ; / .' ! ,-i \ \ \ \\.~\
\...
\ ... \ ~...-\ .110 .'70 .,Iot ..JtSIWULATED PHASE CURREKTS(TESTliB) C\S·('~t lA)
.'M
.,...
....
6 ...J
1. ...-,
...
-.
I~
.
I
" " - - l S '~... ·
I
-,
~ ' . - t - + ' 0 ..···0 I ~,_.-A
o " , -A- ' ...
""~-'''
-.
. , . ' / , ,. > :D-<,~_..,,/ / '0
,.'/,.,D
»<>. /
a " "
:~
y'
70 '0 60 10 40 20 '0 10 100.00 0.40 0.10 r.20 1.60 Z. 00 2.J, 2.7f J. l ' 3. , .
DISUNCI "'OM SOURel (HILlS)
.... .... ..at .... .... ... •.•• J.lI 1.W
[Iht·:...:~ fr.jfA ~.:. (",1l~&)
lJ1
V1
'~ via current in phase \a~ producing an induced voltage across the
mutual inductance term Zab. In the case of a bundled conductor, or
any balanced three-phase circuit ~hereIIJ can be considered as the
superposition of equal currents I , I
b, and I , the flux linkages
a c
between phases will essentially be zero because the magnetic field
intensity H is zero. Thus, no mutual induction takes place. This is
evident in Figure 5-1, where all phase voltages are equal. The phase
to phase voltages were also measured to be zero, which justifies the
absence of mutual coupling effects on a uniform network.
The phase currents, as expected, are not exactly equal at all
points along the network. This can be attributed to the physical
geometry of the conductors. Although the network is considered to
be symmetric, other factors, such as unequal spacings between phases,
variations in conductor heights above ground (due to terrain), and
skin effect can cause unequal current division between homogeneous
conductors. Skin effect results in a decrease in current density
toward the center of the conductor. This inequality in current
density is caused by a longitudinal element near the center of the
conductor being surrounded by more magnetic lines of force, hence
reducing the net driving emf at the center element. Thus, virtually
all of the current is concentrated near the surface of the conductor.
In addition, the netvork experiences changes in conductor geometry
in the form of "tvists". These tvists involve the transposition of
afore-mentioned nonuniformities have some effect on voltage and current
distribution, and is the probable cause for the slightly different
standing wave patterns between the phase conductors shown in Figure
5-1.
Since this particular network can be treated as a single
bundled conductor, certain transmission line parameters such as
characteristic impedance and propagation constant can be evaluated
from empirical data. For single phase lines, the characteristic
impedance is obtained by the following relation:
z
o (2 Z )1/2sc oc
<5.1>
where Z represents the short-circuit input impedance, and 2
sc oc
the open-circuit input impedance. From Figure 5-1, the impedances
for short-circuit and open-circuit load conditions can be obtained
by assuming that the netvork length is equal to an electrical
half-wavelength. This assumption is valid due to the symmetry of the
standing wave patterns, as discussed before. Since voltage and
cur-rent minimas and ~aximas repeat every half-wavelength, the input
impedance for this symmetrical network can be represented by:
I
v
III
oc oc
12
I
375 ohmso
55 volts/12.3 rna
7 volts/222.2 rna
4.47 kohms
The open circuit input impedance is obtained at the end of the line,
where V
=
V ,and current I = I .=
I + Ib + I .
Like-oc max oc mIn a c
wise, the input impedance looks like a short circuit at the midpoint
of the network where V = V . , and I = I
sc mIn sc max Note that
the currents for both short-circuit and open-circuit loads are the
superposition of individual currents flowing in each of the phase
conductors. Hence, we see that the input impedance is dependent upon
line length. Because this particular network is one-half wavelength
long, the input impedance always "looks" like the load impedance. If,
however, the network length were ~ quarter-wavelength, then the input
impedance for an open-circuit load would look like a short.
The empirical value for characteristic impedance 2 of
o
375 ohms is consistent with that derived theoretically for a bundled
conductor. The theoretical value for Z of 372 ohms is derived
o
in Appendix B, and is based on methods described in Chapter 3 for
computing per-unit length matrices [Z] and [Y]. From a previous
experiment conducted on an unloaded single phase network {3}, 2
0
was measured to be in the neighborhood of 450 ohms. The decrease in
Z for a bundled conductor is expected since the equivalent
geomet-o
ric mean radius is greater than that of a single conductor. Hence,
the inductance is decreased with the addition of conductors, while
the capacitance to ground is increased, which results in a net
de-crease in the magnitude of the characteristic impedance.
maxima for each phase conductor all occur at the same location
with respect to the load (midpoint of the network), we may deduce
that the propagation constant (y
=
a + jS) for each conductoris the same. Furthermore, the propagation matrix
r
is diagonal.Recall from Chapter 2 that the propagation matrix contains elements
(eigenvalues) that define the modes of propagation. It is evident
from Figure 5-1 that the imaginary components
<a ,
aa,
~ba )
c of thepropagation constant are equal.
A simulation program based on the mathematical model presented
in Chapter 2 was written by M. Riddle {33). Several test cases were
simulated for comparison vith empirical results. Figure 5-2 shows the
theoretical standing wave patterns generated for the symmetric network
denoted by test #lb. The per-unit length parameters, namely the [Z]
and [Y] matrices were derived by the methodology described in Chapter
3. These parameters are derived in Appendix A for the test network
used in making the propagation measurements. In comparing the measured
data (Figure 5-1) to that of the theoretical (Figure 5-2), we see
noticeable differences, the most prominent being the lack of symmetry
in Figure 5-2. Another discrepancy between the two plots exists in
the phase voltages being unequal in Figure 5-2, ~hereas the measured
voltage standing wave patterns in Figure 5-1 are identical.
There are several possible explanations for these apparent
differences. The first is that the theoretically derived per-unit
length parameters do not accurately represent the actual test network
reality, this is not the case. Many approximations were made due to
uncertainties in the network parameters. For example, the physical
geometry of the conductors changes throughout the netVlork from
"verti-cal" to "delta" configurations, as mentioned previously. These
conductor t~ists, which cannot be regarded as a true transposition
in a strict sense, have an effect similar to that of a complete
transposition cycle in that the rotating of conductors effectively
reduces, or "cancels out" mutual impedance effects. In the derivation
of the per-unit length impedance matrix, these twists were not taken
into account. Other uncertainties, such as conductor heights are
inherent in varying terrain levels. In addition, non-uniform conductor
spacings, and earth resistivity also have an effect on signal
propaga-tion. These "non-uniformities" in the network could affect the
calculated parameters [Z] and [Y], but their sensitivity to these
non-uniformities is not yet knoVln. However, based on several "trial
and error" variations of line length, the experimental standing
wave-forms could be matched. In particular. the network seemed especially
sensitive to line lengths corresponding to multiples of
quarter-wavelengths, vhich is the case in test #lb.
Another possible source of error could be attributed to lack
of precision in the numerical solution process. Since the computer
simulation involves matrix functions (multiplication, diagonalization,
inversions, etc.) vhich are used iteratively, a very slight error in