Você está na página 1de 9

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS, VOL. 37, NO.

8, AUGUST 1990 991

A Simple Introduction to the


Transmission-Line Modeling
MATTHEW N. 0. SADIKU. MEMBER,
IEEE,AND LAWRENCE C. AGBA

Abstract -The transmission-line modeling, a numerical method for Schwingers [7], [8] work, the power and flexibility of
field problems, is presented in simple terms that undergraduate stu-equivalent circuits become more obvious to engineers.
dents non-experts can understand. The method is shown to be generally
applicable to a wide range of problems. Numerical examples are pro-
The recent applications of the idea to scattering prob-
vided for waveguide problems in one and two dimensions. lems, originally due to Johns [91, has made the method
more popular and attractive.
I. INTRODUCTION The transmission-linemodeling (TLM), otherwise known
as the transmission-line matrix method, is a numerical
T HE LINK BETWEEN field theory and circuit the-
ory, the major theories on which electrical engineer-
ing is based, has been exploited in developing numerical
technique
lents. It is
for solving field problems using circuit equiva-
based on the equivalence between Maxwells
techniques to solve certain types of partial differential equations and the equations for voltages and currents on
equations arising in field problems with the aid of equiva- a mesh of continuous two-wire transmission lines. The
lent electrical networks [ll. There are three ranges in the main feature of this method is the simplicity of formula-
frequency spectrum for which numerical techniques for tion and programming for a wide range of applications
field problems in general have been developed. In terms [lo]. As compared with the lumped network model, the
of the wavelength A and the approximate dimension I of TLM is more general and performs better at high fre-
the apparatus, these ranges are [2] quencies where the transmission and reflection properties
of geometrical discontinuities cannot be regarded as
A >> L lumped [7].
h=L The TLM is not new to experts in numerical analysis of
h =-5< L. EM related problems. However, the method has been
presented in various contexts dictated by the applications.
In the first range, the special analysis techniques are This paper seeks a unified, general presentation of the
known as circuit theory; in the second, microwave theory; method in simple terms that students with basic knowl-
and in the third, geometric optics (frequency-independent). edge of differential equations and circuit theory can un-
Hence the fundamental laws of circuit theory can be derstand.
obtained from Maxwells equations by applying an ap- In Section 11, a unified treatment of the TLM method
proximation valid when A >> I. However, it should be
is presented whereby it is shown to be applicable in
noted that circuit theory was not developed by approxi- solving various problems involving the diffusion equation,
mating Maxwells equations, but rather was developed
Poissons equation, or the wave equation. A special appli-
independently from experimentally obtained laws. The
cation of the method is made to wave propagation prob-
connection between circuit theory and Maxwells equa-
lems in Section 111. Section IV presents examples of wave
tions (summarizing field theory) is important; it adds to
propagation problems in one- and two-dimensional space.
the comprehension of the fundamentals of electromagnet-
ics (EM). According to Silvester and Ferrari, circuits are
mathematical abstractions of physically real fields; never- 11. THEORETICAL BACKGROUND
theless, electrical engineers at times feel they understand The TLM method involves dividing the solution region
circuit theory more clearly than fields [3]. into a rectangular mesh of transmission lines. Junctions
The idea of replacing a complicated electrical system by are formed where the lines cross, creating impedance
a simple equivalent circuit goes back to Kirchhoff and discontinuities. A comparison. of the transmission-line
Helmholtz. As a result of Parks [41, Krons [5], [6], and equations and Maxwells equations allows equivalences to
be drawn between voltages and currents on the lines and
Manuscript received July 27, 1989. This paper was recommended by electromagnetic fields in the solution region. Before we
Associate Editor T. K. Ishli.
M. N. 0. Sadiku is with the Department of Electrical Engineering, apply the method, it seems fit to review briefly the basic
Temple University, Philadelphia, PA 19122. concepts of transmission lines and also show how the
L. C. Agba is with the Department of Electrical and Computer TLM method can be applied to a wide range of field
Engineering, Florida Atlantic University, Boca Raton, FL 33431.
IEEE Log Number 9036287. related problems.

0098-4094/90/0800-0991$01.00 01990 IEEE


992 IEEE TRANSACTIONS ON CIRCUITSAND SYSTEMS, VOL. 37, NO. 8, AUGUST 1990

Differentiating (3) with respect to z and (6) with respect


to t, the two equations become
a2v
__ -- R -
a i + L -a2z
az2 az a m (7)
a2z av a2v
I- ataz
-G-+C,-.
at at
(8)
Substituting (6) and (8) into (7) gives
~
-
z Z+Al a2v a2v av
Fig. 1. T-type equivalent circuit model of a different length of a -.
az
=L C 7
at
+ ( R C + G L )-
at
+ RGV. (9)
two-conductor transmission line.
Similarly, we obtain the equation for current I as
A. General, Basic T-type Equivalent Model a21 a2z ai
-= LC? + ( RC + G L ) - + RGI. (10)
Consider an elemental portion of length A1 of a two- az at at
conductor transmission line. We intend to find an equiva- Equations (9) and (10) have the same mathematical form,
lent circuit for this line and derive the line equations. The which in general may be written as
equivalent circuit of the portion of the line is shown in
Fig. 1, where the line parameters R , L, G , and C are a2 a2@ a@
-= LC7+(RC+GL)-+RG@ (11)
resistance per unit length, inductance per unit length, az at at
conductance per unit length, and capacitance per unit where @ ( z ,t ) is replaced by either V ( z ,t ) or Z(z, t).
length of the line, respectively. The model in Fig. 1 may
represent any two conductor lines. The model is called B. Special, Basic T-type Equivalent Models
the T-type equivalent circuit; other types of equivalent Ignoring certain transmission line parameters in ( 1 1 )
circuits are possible, but we end up with the same set of leads to the following special cases [ 111:
equations. In the model of Fig. 1,we assume without loss
of generality that wave propagates in the + z direction, a) L = C = 0 yields
from the generator to the load.
By applying Kirchhoff's voltage law to the left loop of
the circuit in Fig. 1, we obtain
where k, = RG. Equation (12) is the one-dimensional
AI AI aZ
2
+
V (z , t ) = R-Z( z , t ) L - -( z , t ) + V ( z + A 1 / 2 , t ) elliptic partial differential equation called Poisson's equa-
2 at tion.
(1) b) R = C = 0 or G = L = 0 yields
or
V(z+A1/2,t)-V(z,t) az
- = RI( 2 , t ) + L -( 2 , t ) .
A1/2 at
where k 2 = G L or RC. Equation (13) is the one-dimen-
(2) sional parabolic partial differential equation called the
Taking the limit of (2) as AI -+ 0 leads to diffusion equation.
c) R = G = 0 (lossless line) yields

Similarly, applying Kirchhoff's current law to the main


node of the circuit in Fig. 1 gives where k, = L,C. This is the one-dimensional hyperbolic
+
I( 2, t ) = I( z AI, t ) + AZ partial differential equation called Helmholtz equation or
= I( z + + +
A l , t ) GAW( z A1/2, t ) simply wave equation. Thus under certain conditions, the
one-dimensional transmission line can be used to model
av
+ CAl-(Z
at
+ A1/2,t) problems involving an elliptic, parabolic, or hyperbolic
partial differential equation (PDE). The transmission line
or of Fig. 1 reduces to those in Fig. 2 for these three special
I ( z + A1, t ) - I( Z,t ) cases, part (b) holding when G = L = 0. It is the model in
-
A1
= GV( z + A2/2, t ) Fig. 2(c) that is of interest in this paper since we intend to
av apply the TLM to wave propagation problems.
+ C - -z ( z + A1/2, t ) . (5)
C. Other Useful Line Characteristics
As AI + 0, (5) becomes
Apart from the equivalent models, other trans-
az av
--(z,t)
a2
= G V ( Z t, ) + C -at( z,t). (6)
missions-line parameters are of interest. A detailed expla-
nation of these parameters can be found in standard field
SADIKU AND AGBA: INTRODUCTION TO TRANSMISSION-LINE MODELING 993

(a)

RAL/2

I
1
T Fig. 3. Equivalent network of a two-dimensional TLM shunt mode.

Lbl/2

I
Lbl/2 2

+
~~

(C)
Lboundaly C
Fig. 2. Transmission-line equivalent models for (a) elliptic PDE,Pois-
son's equation; (b) parabolic PDE, diffusion equation; (c) hyperbolic Fig. 4. Transmission-line matrix and boundaries.
PDE, wave equation.

is made up of a large number of such building blocks as


theory texts, e.g. [12l, [131. We briefly present these im- depicted in Fig. 4. Notice that in Fig. 4, single lines are
portant parameters. For the lossless line in Fig. 2(c), the used to represent a transmission line pair. Also, a uniform
characteristic resistance internodal distance of A1 is assumed throughout the
Rc = im matrix.

the wave velocity A. Equivalence Between Network and Field Parameters


1 We refer to Fig. 3 and apply Kirchhoffs current law at
U=- node 0 to obtain
m + +
I,( x - A 1 / 2 ) - I,( x A 1 / 2 ) I,( z - A 1 / 2 )
and the reflection coefficient at the load
- Z , ( ~ + A l / 2 ) = 2 C b l -3V
.Y (16)
RL - Rc at
r= Dividing through by AI gives
RL + Rc
where R L is the load resistance.
Z,( x - AZ/2) - Z,( x + A Z / 2 )
AI
111. APPLICATION OF THE TLM TO WAVE I,( z - A 1 / 2 ) - Z,( z + A 1 / 2 ) avY (17)
PROPAGATION PROBLEMS + AI
=2C-.
at
The generality of the TLM method has been demon- Taking the limit as A i + 0 results in
strated in the previous section. In this section, the method
is applied specifically to the problems of wave propaga-
az, az, 3VY
-2 c - .
tion in lossless media. In order to show how Maxwell's az ax at
equations may be represented by the transmission-line Applying Kirchhoffs voltage law around the loop in the
equations, the differential length of the lossless transmis- x - y plane gives
sion line between two nodes of the mesh is represented by aZ,( x - AZ/2)
inductors and capacitors as shown in Fig. 3 for two- V,( x - A1 /2) - LA1 /2
dimensional wave propagation problems. At the nodes, at
pairs of transmission lines form impedance discontinu- aZ,( x A Z / 2 ) +
- LA1/2 - V,( x + A 1 / 2 ) = 0. (19a)
ities. The complete network of transmission-line matrices at
994 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS, VOL. 37, NO. 8, AUGUST 1990

Upon rearranging and dividing by Al, we have Consider the situation for which E, = E, = Hy = 0 and
Vv(x - A1/2) - Vv(x + A1/2) all derivatives with respect to y are zero. It is noticed at
once that this mode is a transverse electric (TE). Thus
A1 (23) reduces to
=-
2
L az,(
at
x - A1/2)
- + - L2 ai,( x +atA1/2) . (19b) aH,
----
aH, aE,
az ax
-E-
at (24a)
Again, taking the limit as A1 + 0 gives

ax
---
aV, 81,
L--.
at
Taking similar steps on the loop in the y - z plane yields
aVY
-- - -L-.
az at Taking similar steps in (24a)-(c) as were taken for
These equations will now be combined to give a wave (18a)-(c) results in another Helmholtz equation:
equation. Differentiating (18a) with respect to t , (18b)
with respect to x , and (18c) with respect to z , we have
a2r, a2z,
_-_-- a2v,
- 2c- Comparing (24) and (25) with (18) and (21) yields the
azat axat at2 following equivalence between the parameters:
a2vy
-=-
a2z,
E, = VY
L-
ax2 atax
a2vy a2zz H, E - I,
-=- L-.
az ataz H, = I,
Substituting (20b) and (20c) into (20a) leads to p=L
a2vy a2vy a2vy Er2C
- + y-- 2LC-.
ax2 az at2
Thus in the equivalent circuit
Equation '(21) is the Helmholtz wave equation in two-
dimensional space. voltage at a node is E,,
In order to show the field theory equivalence of (18) current in the z direction is - H,,
and (21), consider Maxwell's equations current in the x direction is H,,
inductance per unit length represents the permeabil-
aH ity of the medium,
VXE=-p-
at twice the capacitance per unit length represents the
and permittivity of the medium.
dE B. Dispersion Relation of Propagation Velocity
QXH=E--.
at For the basic transmission line in the TLM that has
Expansion of (22) in the rectangular coordinate system p, = E , = 1, the inductance and capacitance per unit length
yields are related by [81
1 1
{(m.4
m = =c
(27)

where c = 3 X lo8 m/s is the speed of light in vacuum.


Notice from (27) that for the equivalence made in (261, if
voltage and current waves on each transmission-line com-
ponent propagate at the speed of light c, the complete
network of intersecting transmission lines represents a
medium of relative permittivity twice that of free space.
This implies that as long as the equivalent circuit in Fig. 3
aH, aH, aEy is valid, the propagation velocity in the TLM mesh is 1/fi
---- -E-
az ax at of the velocity of light. It is shown in the Appendix that
the propagation constant 0, on the TLM mesh is
dHY aH, aE,
--E-.
ax ay at sin(pnAl/2) =fisin[oA1/(2c)]. (28)
SADlKU A N D AGBA: INTRODUCTION TO TRANSMISSION-LINE MODELING 995

0'7b
I.... ....

0.6

0.50 0.05 01 0.15 0.2 0.25

Normalized Frequency A ~ / A
Fig. 5. Dispersion of the velocity of waves in two-dimensional TLM
network.

If U, ( = w / p n ) is the velocity of waves on the mesh, (28)


can be written in terms of the ratio r = U, / c = o / p , c ,
i.e.,
I l-Il2V I

By selecting different values of Al/A, the corresponding


values of r = U,,/ c can be obtained numerically as shown
in Fig. 5. From Fig. 5, we observe that the TLM can only 2
, ...... j ........! 1

represent Maxwell's equations over the range of frequen-


cies from zero to the finest network cutoff frequency, ........ .........
1
which occurs at o A l / c = ~ / or2 AI/A = 1/4. Over this
range, the velocity of the waves behaves according to the (4
characteristic of Fig. 5. For frequencies much smaller
than the network cutoff frequency, the propagation veloc- Fig. 6. The impulse response of a node in a matrix.
ity approximates to l / & of the free-space velocity.
The more general case of four impulses being incident
C. Incident and Reflected Pulses on the four branches of a node can be obtained by
If kVi and kVL are the voltage impulses incident upon applying the superposition principle to the previous case
and reflected from terminal rz of a node at time t = of a single pulse. Hence, if at time t = kAl/c, voltage
kAl/c, we derive the relationship between the two quan- impulses kv;, kvi,kvl, and kv'
are incident on lines 1-4
tities as follows. Let us assume that a voltage impulse respectively, at any junction node as in Fig. 6(c), the
function of unit magnitude is launched into terminal 1 of combined voltage reflected along line 1 at time t =
a node, as shown in Fig. 6(a), and that the characteristic (k + 1)Al/c will be [9], [lo]
resistance of the line is normalized. A unit-magnitude
delta function of voltage and current will then travel
towards the junction with unit energy (S,). Since line 1
has three other lines joined to it, its effective terminal
resistance is 1/3. With the knowledge of its reflection In general, the total voltage impulse reflected along line n
coefficient, both the reflected and transmitted voltage at time t = (k + l)Al/c will be
impulses can be calculated. The reflection coefficient is
R,-R,
r = R,+ R,
-
_
1/3-1
_ _ _ E _ _

1/3+1
1
2 (30)
k+lK=-
2
[ m=l
k vi -
,771 V'
k n, a = 1 , 2 , 3 , 4 . (33)

so that the reflected and transmitted energy are This idea is conveniently described by a matrix equation
relating the reflected voltages at time (k + l)Al/c to the
s, = s,r2= 114 ( 3 1 ~ ) incident voltages at the previous time step kAl/c:

where
and transmitted
s,= s , ( i - rz) = 314
subscripts quantities,
i , r , and t respectively.
indicate incident,
(31b)

Thus reflected,
a voltage
impulse of - 1/2 is reflected back in terminal 1 while a
voltage impulse of 1/2 = [3/4 + 31'" will be launched
into each of the other three terminals as shown in
k+l

[;I=[ i -I
-1 1

Also, an impulse emerging from a node at position


1

-1
1)
k

[;I. i

(34)

Fig. 6(b). (z, x) in the mesh (reflected impulse) becomes automati-


996 IEEE TRANSACTIONS ON CIRCUITS A N D SYSTEMS, VOL. 37, NO. 8, AUGUST 1990

Any resistive load at boundary C (see Fig. 4) may be


simulated by introducing a reflection coefficient r
k 41 c(P , 4 ) = kVi( P + 1 , 4 ) = r [kV[( P 9 ) ] (36)
9

where

r=-R,-1 (37)
R,+l
EXCITATION
and R, is the surface resistance of the boundary normal-
ized by the line characteristic impedance. If, for example,
a perfectly conducting wall ( R , = 0) is to be simulated
along boundary C', (37) gives = - 1, which represents a
short circuif, and
k + l % ( P , q )= - k V 4 ( P 7 4 ) (38)
is used in the simulation.
FIRST ITERATION
E. Computation of Fields and Frequency Response
We continue with the TE mode of (24a)-(24c) as our
example, and calculate E,, H,, and H,. E, at any point
corresponds to the node voltage at the point, H, corre-
sponds to the net current entering the node in the x
direction (see (26)) while H, is the net current in the
negative z direction. For any point ( z = m , x = n ) on the
grid of Fig. 4, we have for each kth transient time.
SECOND ITERATION
1
Fig. 7. Scattering in a two-dimensional TLM network excited by a k ~ y ( m , n= )- [ ~ ; ( m , n+,V;(m,n)
) +,~;(m,n)
Dirac impulse. 2
+kv(m>n)] (39)
cally an incident impulse at the neighboring node. Hence
-k H , ( m > 1 =k V;( m n 1
> - kV4 ( m > n 1 (40)
k H , ( m , n )= k v ; ( m , n )-,V,'(m,n>. (41)
Thus a series of discrete delta-function of magnitudes
E,, H , , and H, corresponding to time intervals of A / / c
are obtained by the iteration of (34) and (35). (Notice that
reflections at the boundaries A and B in Fig. 4 will
cancel out, thus HL= 0.) Any point on the mesh can serve
Thus by applying (34) and (351, the magnitudes, posi- as an output or observation point. Equations (39)-(41)
+
tions, and directions of all impulses at time ( k l)Al/c provide the output-impulse functions for the point repre-
can be obtained at each node in the network provided senting the response of the system to an impulsive excita-
+hattheir corresponding values at time k A l / c are known. tion. These output functions may be used to obtain the
The impulse response may, therefore, be found by ini- output waveform. For example, the output waveform cor-
'ially fixing the magnitude, position and direction of travel responding to a pulse input may be obtained by convolv-
of impulse voltages at time t = 0, and then calculating the ing the output-impulse function with the shape of the
.:tate of the network at successive time intervals. input pulse.
The propagation of pulses in the TLM is illustrated in Sometimes we are interested in the response to a
Fig. 7, where the first two iterations following an initial sinusoidal excitation. This is obtained by taking the
cxcitation pulse in a two-dimensional shunt-connected Fourier transform of t h e impulse response. Since the
TLM are shown. We have assumed free-space propaga- response is a series of delta functions, the Fourier trans-
tion for the sake of simplicity. The scattering process form integral becomes a summation, and the real and
(incidence and reflection of pulses) forms the basic algo- imaginary parts of the output spectrum are given by [91,
rithm of the TLM method [lo],[141. [IO1
N
D. Boundary Representation
Re [ F (AI/A)] = ,Ices ~ (42a)
Boundaries are usually. placed
. halfway between two k=l
%odes,in order to ensure synchronism. In practice, this is
achieved by making the mesh parameter A1 an integer
fraction of the structure's dimensions.
Im[F(Al/A)] =
N

k=l
,Isin ( ~ (42b)
SADIKU AND AGBA: INTRODU(TI0N TO TRANSMISSION-LINE MODELlNG 997

where F(A1 / A ) is the frequency response, I is the value TABLE I


NORMALIZEDIMPEDANCE OF A TEM WAVE
of the output-impulse response at time t = k A l / c , and N DISCONTINUITY
WITH FREE-SPACE
is the total number of time intervals for which the calcula-
tion is made. Henceforth, N will be referred to as the
number of iterations.
Number of
IV. NUMERICAL
EXAMPLES iterations 100 150 200

A Fortran program was written for the numerical calcu- 0.002 0.9789 -0.1368 0.9730 -0.1396 0.9781 -0.1253 0.9747 -0.1282
lations for one-dimensional TEM waves. This example is 0.004 0.9028 -0.2432 0.8980 -0.2322 0.9072 -0.2400 0.9077 -0.2356
0.006 0.8114 -0.3068 0.8229 -0.2979 0.8170 -0.3046 0.8185 -0.3081
taken from [9]. The calculations were carried out on a
0.008 0.7238 -0.3307 0.7328 -0.3457 0.7287 -0.3404 0.7256 -0.3390
25 x 11 rectangular matrix. TEM field-continuation 0.010 0.6455 -0.3201 0.6367 -0.3350 0.6396 -0.3281 0.6414 -0.3263
boundaries were fixed along x = 2 and x = 10, producing 0.012 0.5783 -0.2730 0.5694 -0.2619 0.5742 -0.2680 0.5731 -0.2707
boundaries, in effect, along the lines x = 1.5 and x = 10.5. 0.014 0.5272 -0.1850 0.5313 -0.1712 0.5266 -0.1797 0.5255 -0.1765
The initial impulse excitation was on all points along the 0.016 0.4993 -0.0609 0.5043 -0.0657 0.5009 -0.0538 0.5018 -0.0545
line z = 4, and at all subsequent time intervals the field 0.018 0.5002 -0.0790 0.4987 -0.0748 0.5057 -0.0785 0.5057 0.0768

along this line was set to zero. In this way, interference


from boundaries to the left of the excitation line was TABLE I1
avoided. Calculations in the z-direction were terminated IMPEDANCE OF A RECTANGULAR
NORMALIZED
at z = 24, so that no reflections were received from points WAVEGUIDEWITH SIMPLELOAD
at z = 25 in the matrix, and the boundary C in Fig. 4, Ae/A TLM Results Exact Results
situated at z = 24.5, was, therefore, matched to free space.
IZI ArdZ) IZI Adz)
The output-impulse response for E, and H, was taken at
0.020 1.9391 0.8936 1.9325 0.9131
the point z =14, x = 6 , which is 10.5 mesh points away 0.021 2.0594 0.6175 2.0964 0.6415
from the boundary C , for 100, 150, and 200 iterations. 0.022 1.9697 0.3553 2.0250 0.3603
Since the velocity of waves on the matrix is less than 0.023 1.7556 0.1530 1.7800 0.1438
that in free space by a factor u n / c (see Fig. 5), the 0.024 1.5173 0.0189 1.5132 0.0163
0.025 1.3036 -0.0518 1.2989 -0.0388
effective intrinsic impedance presented by the network
0.026 1.1370 -0.0648 1.1471 -0.0457
matrix is less by the same factor. The magnitude of the 0.027 1.0297 -0.0350 1.0482 -0.0249
wave impedance on the matrix, normalized to the intrinsic 0.028 0.9776 0.0088 0.9900 0.0075
impedance of free space, is given by Z = lEyl/ IH,I and is 0.029 0.9620 0.0416 0.9622 0.0396
tabulated in Table I, together with Arg(z), for the various 0.030 0.9623 0.0554 0.9556 0.0632

numbers of iterations made. A comparison is made with


the exact impedance values [12]. formulation and programming. Another advantage of the
The second example taken from [15] is on a rectangular
TLM method resides in the large amount of information
waveguide with simple load. The Fortran program used generated in one single computation. Not only is the
for the numerical analysis is basically similar to that of
impulse response of a structure obtained, yielding in turn
one-dimensional simulations. A 25 X 11 matrix was used
its response to any excitation, but the characteristics of
for the numerical analysis of the waveguide. Short circuit
the dominant and higher order modes are also accessible
boundaries were placed at x = 2 and x = 10, the widthin the frequency domain through the Fourier transform.
between the waveguide walls thus being 9 mesh points.The method is limited only by the amount of memory
The system was excited at all points along the line z = 2,
storage required, which depends on the complexity of the
and an impulse function of the output was taken from the
structure. The number of iterations required varies from
point ( x = 6, z = 12). The C boundary at z = 24 repre-
the hundreds to the thousands, depending on the size and
sented an abrupt change to the intrinsic impedance ofcomplexity of the TLM mesh.
free space. The cutoff frequency for the waveguide occurs
Although the application of TLM in this paper has
[161 at A l / A , , = 1/18, and A n is the network-matrix wave-
been limited to the wave propagation problem, the method
length, which corresponds to A l / A = 42/18 since has a wide range of applications. Simple applications of
_
A n - un
_-- 6 JLC 1 the
found
technique to diffusion problems, for example, can be
in Lohse et al. [17] and Wong 1181. The method is
A c - = = = m = z .
recently extended to antenna problems [191, [20]. For
A comparison between the results for the normalized further development and applications of the TLM meth-
guide impedance using this method is made with exact ods, one should consult [lo], [21-251.
results in Table 11.
V.CONCLUSIONS APPENDIX
This paper has presented the transmission-line model- THEDISPERSION RELATION OF THE TLM MESH

ing in simple terms that nonexperts can understand. The In a two-port network, the ABCD transfer matrix is
main feature of the TLM method is the simplicity of commonly used to relate the input voltage V , and current
998 IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS, VOL. 37, NO. 8 , AUGUST 1990

1 have a propagation constant y,, = a,, + jp,,, then

cosh ( y , , A l ) = cos( e) -tan (8/2) sin ( e ) . (A.7)


This equation describes the range of frequencies over
which propagation can take place (passbands), i.e.,
4 --
L A
b
ne12 ne13 lcos(e)-tan(e/2)sin(e)1<1 (A.8)
Fig. 8. Lossless transmission line with shunt element in center. and the range Of frequencies Over which propagation
cannot occur (stopbands), i.e.,
I , to the output voltage V, and current Z,, i.e., /cos(e) -tan (8/2),sin (e)l> 1. (A.9)
[ S ] = rAc .I[
B -3 ('4.1)
For the lowest frequency propagation region,
Yn = jP,, (A.lO)
and
This matrix is convenient to use when several two-port
2rA1 w
networks are connected together in cascade so that the e=-- - -Al. (A.ll)
output from one forms the input to the next. To obtain A C
the ABCD matrix for a transmission line with shunt Introducing (A.10) and (A.11) in (A.71, we obtain
element in the center as in Fig. 8, we consider the line as
comprising of three networks connected in cascade. The
transfer matrix of a lossless transmission line of electrical
length e/2( = r A l / A ) and characteristic impedance Z,
COS ( p , , A l ) = COS ( w
- tl)-tan(
:)sin( F). (A.12)
is [26] Applying some trigonometric identities to (A.12) results
COS^/?. j~,sin0/21 in
final
(
sin -) = & s i n ( X )
wAl '

(A.13)

which is a transcendental equation. Furthermore, since


The matrix of a shunt admittance is w A l TAI
- (A.14)
2c A
and
Therefore, the overall ABCD matrix for the whole line in
Fig. 8 is

2rr
p =- (A.16)
Ar
where r = U,,/c. Substituting (A.14) and (A.16) into
(A.131, we have

J . (A.4) r A1 T A1
COS e /2 sin( ;-T) = a s i n . ( h). (A.17)
For the network in Fig. 3, the voltage and current at By Selecting different values of A l / A 7 the corresponding
node i are related to those at node i + 1 by the transfer- values of r = U,,/C can be obtained as in Fig. 5 for
matrix equation two-dimensional problems.
Following the same procedure, the dispersion relation
(cos e /2) ( j sin e 1 2 ) 1 for three-dimensional problems can be derived as
(jsin0/2) COS^/^)

('4.5)
sin .:( y) = 2sin ( r : ) . (A.18)

Thus for low frequencies in three-dimensional space, the


where it is assumed that the characteristic impedance of network propagation velocity may be considered constant
the line is unity. If the waves on the periodic structure and equal to c/2.
SADIKU AND AGBA: INTRODUCTION T O TRANSMISSION-LINE MODELING 999

ACKNOWLEDGMENT [20] I. Palocz and N. Marcovitz, A network-oriented approach in the


teaching of electromagnetics, ZEEE Trans. Education, vol. E-28,
The authors like to express their deep apprecia- pp. 150-154, Aug. 1985.
[21] Numerical Techniques for Microwave and Mullimeter-Wave Passive
tion to Verne11 Ross for typing the manuscript and for Structures. T. Itoh, Ed. New York Wiley, 1989, pp. 496-591.
being such a pleasure to work with. The comments of the [22] Y. C. Shih and W. J. R. Hoefer, Dominant and second-order
mode cutoff frequencies in fin lines calculated with a two-dimen-
-.
reviewers are appreciated. sional TLM momam. ZEEE Trans. Microwave Theorv Tech.. vol.
M-%T-28, pp.14z3-1448, Dec. 1980.
J. W. Bandler and P. B. Johns, Transmission-line modeling and
sensitivity evaluation for lumped network simulation and design in
the time domain, J . Franklin Inst., vol. 304, no. 1, pp. 15-23,
REFERENCES 1977.
S. Akhtarzad and P. B. Johns, Solution of Maxwells equations in
G. Kron, Numerical solution of ordinary and partial differential three spzi:e dimensions and time by the TLM method of numerical
equations by means of equivalent circuits, J . Appl. Phys., vol. 16, analysis, h o c . Inst. Elect. Eng., vol. 122, no. 12, pp. 1344-1348,
pp. 172-186, Mar. 1945. Dec. 1975.
C. H. Durney and C . C . Johnson, Introduction to Modem Electro- P. B. Johns and M. OBrien, Use of the transmission-line mod-
magnetics. New York: McGraw-Hill, pp. 286-287, 1969. elling (TLM) method to solve non-linear lumped networks,, Radio
P. P. Silvester and F. L. Ferrari, Finite Elements for Electrical Elect. Eng., vol. 50, no. 1/2, pp. 59-70, 1980.
Enaineers. Cambridee. MA: Cambridee Univ. Press. U. 24. 1983. Microwave Transmission Circuits. G. L. Ragan, Ed. New York
R.-H. Park, Defin5ion of an ideal-synchronous Sachine and McGraw-Hill, 1948, pp. 544-547.
formulaters for armature flux linkages, Gen. Elect. Rev., vol. 31,
pp. 332-334, 1928.
G. Kron, Equivalent circuit of the field equations of Maxwell,
Proc. IRE., pp. 289-299, May 1944.
G. Kron, Equivalent Circuits of Electrical Machinery. New York m
Wiley, 1951.-
N. Marcovitz and J. Schwineer. On the reuroduction of the
electric and magnetic fields pFoduced by currehts and discontinu-
ity in wave guides-I, J. Appl. Phys., vol. 22, no. 6, pp. 806-819,
June 1951. Matthew N. 0. Sadiku (S82-M82-M88) re-
J. Schwinger and D. S. Saxon, Discontinuities in Waveguides. New ceived the M.Sc. and Ph.D. degrees from
York: Gordon and Breach, 1968. Tennessee Tech-University, Cookeville, TN, and
P. B. Johns and R. L. Beurle, Numerical solution of 2-dimen- is currently an Associate Professor of electrical
sional scattering problems using a transmission line matrix, Proc. engineering at Temple University, Philadelphia,
ZEEE, vol. 118, pp. 1203-1208, Sept. 1971. PA.
W. J. R. Hoefer, The transmission-line matrix method-Theory His primary research interests are numerical
and applications, IEEE Trans. Microwave Theory Tech., vol. techniques in electromagnetics, local area com-
MTT-33, pp. 882-893, Oct. 1985.
B. J. Ley, Computer Aided Analysis and Design for Electrical Engi- puter networks, and superconductivity. He is the
neers. New York: Holt, Rinehart and Winston, pp. 815-817, author of Elements of Electromagnetics (HRW,
1970 19891.
M. N. 0. Sadiku, Elements of Electromagnetics. New York Saun- Dr. Sadiku is a Registered Professional Engineer in the state of
ders, 1989, pp. 494-565, chap. 11. Florida and a member of ASEE.
D. K. Cheng, Field and Wave Electromagnetics. Reading, MA.
Addison-Wesley, 1983, chap. 9, pp. 370-442.
G. E. Mariki and G . Yek, Dynamic three-dimensional TLM 0
analysis of microstrip lines on anisotropic substrate, ZEEE Trans.
Microwave Theory Tech., vol. MTT-33, pp. 789-799, Sept. 1985.
P. B. Johns, Applications of the transmission-line matrix method
to homogeneous waveguides of arbitrary cross-section, in Proc.
ZEEE, vol. 119, p 1086-191, Aug. 1972. Lawrence C. Agba received the M.S. degree
R. A. Waldron, fheory of Guided Electromagnetics Waves. Lon- from Florida Atlantic University, Boca Ra-
don, U K Van Nostrand Reinhold, 1969, pp. 157-172. ton, where he is currently working toward the
A. M. Lohse, ef al., Computer graphics for transient fields, Ph.D. degree in electrical engineering. His re-
ZEEE Trans. Education, vol. E-20, pp. 64-68, Feb. 1977.
C. C. Wong, Solution of the network analog of one-dimensional search interests include numerical modeling of
field e uations using the ladder method, ZEEE Trans. Education, electromagnetic problems, artificial neural net-
vol. E-18,pp. 176-179, Aug. 1985. works, and VLSI hardware design.
W. R. Zimmerman, Network analog of Maxwells field equations
in one and two dimensions, ZEEE Trans. Education, vol. E-25,
pp. 4-9, Feb. 1982.