24.15.

Problems 1221

a. For each of the values M = 20, 50, 100, 200, solve Hallén’s equation for a delta-gap
input with voltage V0 = 1 volt using both the exact and the approximate kernels. Plot
the real and imaginary parts of the current Im = I(zm ) versus zm over the right half
of the antenna, that is, 0 ≤ zm ≤ h, where h = l/2.
25
b. Fit to the computed current samples Im of the exact kernel to King’s three-term approx-
imation. Then, place the fitted points on the same graphs as in part (a). Discuss how Coupled Antennas
well or not the three-term approximation fits the exact-kernel and the approximate-
kernel current.
Repeat by using a two-term approximation. Discuss how well or not the two-term
approximation fits the exact-kernel and the approximate-kernel current.
c. To illustrate the logarithmic singularity near z = 0, evaluate the limiting expression at
the points zm , m = 1, 2, . . . , M, for M = 200 (the point z0 = 0 is to be skipped):

4kaV0  
Ilog (zm )= −j ln k|zm | + const.
η

Adjust the constant so that this expression agrees with the exact-kernel current at the
25.1 Near Fields of Linear Antennas
point z1 , that is, Ilog (z1 )= I1 . Then, plot the imaginary parts of Im and Ilog (zm ) versus
In calculating mutual coupling effects between closely-spaced linear antennas, we need
zm .
to know the fields produced by an antenna at near distances. The fields generated by a
d. Repeat parts (a–c) for the antenna radius a = 0.001 and then for a = 0.008. Discuss thin wire antenna with current I(z) were worked out in Sec. 15.4.
the effect of changing the radius on the quality of the solution, both for the exact and
We summarize these results here. All field components can be obtained from the
the approximate kernel cases.
knowledge of the z-component of the magnetic vector potential Az (z, ρ):
e. Repeat parts (a–d) for the antenna length l = 1.0λ. Comment on the success of the
exact versus approximate kernel calculations versus the parameters l, a, M. h 
μ e−jkR
f. For each value of M and current solution Im , −M ≤ m ≤ M, the input impedance of Az (z, ρ)= I(z ) dz , R = ρ2 + (z − z )2 (25.1.1)
4π −h R
the antenna can be calculated from the center sample I0 , that is, Z0 = V0 /I0 . Similarly,
the input admittance is:
1 I0 where h is the half-length of the antenna, h = l/2, and the geometry is shown in
Y0 = = = G0 + jB0 Fig. 25.1.1. We have used the approximate thin-wire kernel because it differs little from
Z0 V0
the exact kernel for distances ρ > a (typically, when ρ  5a.)
where G0 , B0 are its real and imaginary parts, that is, the input conductance and sus-
ceptance.
For each of the values M = 1, 2, . . . , 100, calculate the corresponding conductance and
susceptance, G0 (M), B0 (M), using the exact and the approximate kernels and plot
them versus M. Use the length and radius l = 0.5λ and a = 0.005λ.
This is a time-consuming question. It requires that you solve the Hallén equation
for each value of M for the exact and approximate kernels and pick the center value
I0 . Discuss the convergence properties of the exact versus the approximate kernel
calculation.

Fig. 25.1.1 Fields of a thin wire antenna.

25.1. Near Fields of Linear Antennas 1223 1224 25. Coupled Antennas

Then, the non-zero field components Ez , Eρ , Hφ can be constructed from the two where we used I (0±)= ∓kIm cos kh and I (±h)= ∓kIm . Inserting this result into
alternative sets of formulas: Eq. (25.1.4) and rearranging some constants, we find:

jωμEz = ∂2z Az + k2 Az jωμEz = ∂2z Az + k2 Az jηIm  

Ez (z, ρ)= − G(z − h, ρ)+G(z + h, ρ)−2 cos kh G(z, ρ) (25.1.7)
4π
jωμEρ = ∂ρ ∂z Az , ∂ρ (ρHφ ) = jω ρEz (25.1.2)
The quantities G(z − h, ρ), G(z + h, ρ), G(z, ρ) can be written conveniently as follows:
μHφ = −∂ρ Az jωEρ = −∂z Hφ
e−jkR0 
As a first approximation, we will assume that the current I(z) is sinusoidal. This G(z, ρ) = , R0 = ρ2 + z2
R0
is justified only when the antenna length is near half a wavelength λ/2. Most coupled
e−jkR1 
antenna arrays that are used in practice, such as Yagi-Uda, satisfy this condition.
G(z − h, ρ) = , R1 = ρ2 + (z − h)2 (25.1.8)
We note also that the near fields resulting from the sinusoidal current assumption R1
do not satisfy the correct boundary conditions on the surface of the antenna, that is
e−jkR2 
the condition Ez (z, ρ)= 0 at z = 0 and ρ = a. In Sec. 25.2, we consider an improved G(z + h, ρ) = , R2 = ρ2 + (z + h)2
R2
approximation of the near fields that addresses these issues. Thus for now, we will
assume that: where R0 , R1 , R2 are recognized to be the distances from the center and the two ends
 
sin k(h − |z|)   of the antenna to the observation point, as shown in Fig. 25.1.1. Thus, we can write:
I(z)= I0 = Im sin k(h − |z|) (25.1.3)
sin kh


where we distinguish between the current I0 at z = 0 and the maximum current Im = jηIm e−jkR1 e−jkR2 e−jkR0
Ez (z, ρ)= − + − 2 cos kh (25.1.9)
I0 / sin kh. For half-wavelength antennas, we have kh = π/2, I0 = Im , and the current 4π R1 R2 R0
becomes I(z)= I0 cos kz.
In principle, one could insert Eq. (25.1.3) into (25.1.1) and perform the required Next, we determine Hφ from Ampère’s law in (25.1.2) by noting that ρEz is a complete
integrations to get Az . However, for the purpose of determining the fields, this is not derivative with respect to ρ. Indeed, for any of the quantities R, we have:
necessary. Combining (25.1.1) and (25.1.2), we obtain:
h e−jkR e−jkR 1
2 μ 2  2 2   ∂ρ (e−jkR )= −jk(∂ρ R)e−jkR = −jkρ ⇒ =− ∂ρ (e−jkR )
jωμEz (z, ρ)= ∂z Az + k Az = I(z )(∂z + k )G(z − z , ρ) dz (25.1.4) R R jkρ
4π −h

where we denoted G(z − z , ρ)= e−jkR /R and replaced ∂2z by ∂2z . Next, we use the Applying this result to all three terms of Eq. (25.1.9), we have:
differential identity:  
jηIm 1
  ρEz (z, ρ)= − ∂ρ e−jkR1 + e−jkR2 − 2 cos kh e−jkR0
I(∂2z + k2 )G − G(∂2z + k2 )I = ∂z I∂z G − G∂z I (25.1.5) 4π −jk

Because of the assumed form (25.1.3), I(z ) satisfies the Helmholtz equation, (∂2z + Inserting this into Ampère’s law, ∂ρ (ρHφ )= jω ρEz , and rearranging some con-
k )I(z )= 0, and therefore, the integrand of (25.1.4) becomes a complete derivative:
2
stants, we find:
 
I(z )(∂2z + k2 )G(z − z , ρ)= ∂z I(z )∂z G(z − z , ρ)−G(z − z , ρ)∂z I(z ) (25.1.6)
jIm  −jkR1 
∂ρ (ρHφ )= ∂ρ e + e−jkR2 − 2 cos kh e−jkR0
Integrating the first term, we obtain: 4π
h
  which can be integrated trivially, giving:
∂z I(z )∂z G(z − z , ρ) dz = I(h)∂z G(z − h, ρ)−I(−h)∂z G(z + h, ρ)= 0
−h

where we used the end-conditions I(h)= I(−h)= 0. The second term in (25.1.6) is a jIm  
Hφ (z, ρ)= e−jkR1 + e−jkR2 − 2 cos kh e−jkR0 (25.1.10)
little trickier because ∂z I(z ) is discontinuous at z = 0. Splitting the integration range, 4πρ
we obtain:
h   h A possible integration constant in ρ is dropped because the field must vanish when
  0  
∂z G(z − z , ρ)∂ I(z ) dz =

z

+ ∂z G(z − z , ρ)∂z I(z ) dz its source vanishes, that is, when Im = 0. Finally, we obtain Eρ from Faraday’s law in
−h −h 0
(25.1.2). Noting the differentiation property:
   
= G(z, ρ)I (0−)−G(z + h, ρ)I (−h) + G(z − h, ρ)I (h)−G(z, ρ)I (0+)
 

z −jkR
  ∂z (e−jkR )= −jk e , R = ρ2 + z2
= kIm 2 cos kh G(z, ρ)−G(z − h, ρ)−G(z + h, ρ) R
25.2. Improved Near-Field Calculation 1225 1226 25. Coupled Antennas

we obtain from jωEρ = −∂z Hφ : King’s three-term approximation, or a three-term fitted to a numerical solution, pro-
vides a better approximation to the current, and one may expect that the fields generated

jηIm z − h −jkR1 z + h −jkR2 z −jkR0 by such current would more closely satisfy the boundary condition (25.2.2). This is what
Eρ (z, ρ)= e + e − 2 cos kh e (25.1.11)
4πρ R1 R2 R0 we discuss in this section.
Because the current need not satisfy the Helmholtz equation, I (z)+k2 I(z)= 0, we
The field expressions (25.1.9)–(25.1.11) have been used widely primarily for the pur- must revisit the calculations of the previous section. We begin by assuming that I(z) is
pose of calculating mutual impedances. They appear in many textbooks and some early symmetric in z and that it vanishes at the antenna end-points, that is, I(±h)= 0. The
references are [1800,1802,2,3]; see also [1784]. electric field Ez (z, ρ) at distance ρ is obtained from Eq. (25.1.4):
It is worth also to verify that the exact expressions for the fields give correctly the h
radiation fields that were derived in Sec. 17.3. At large distances, we can make the 4πjωEz (z, ρ) = I(z )(∂2z + k2 )G(z − z , ρ)dz
approximations: −h
h (25.2.3)
R0 = r, R1 = r − h cos θ, R2 = r + h cos θ e−jkR

= 2
I(z )(∂z + k ) dz
2

where r is the radial distance and θ the polar angle. Replacing ρ = r sin θ, the magnetic −h R
field (25.1.10) becomes approximately: 
where R = (z − z )2 +ρ2 . Applying the differential identity (25.1.5) and the end-point
jIm  −jk(r−h cos θ)  conditions I(±h)= 0, we obtain,
Hφ (r, θ)= e + e−jk(r+h cos θ) − 2 cos kh e−jkr
4πr sin θ h  
4πjωEz (z, ρ) = G(z − z , ρ) I (z )+k2 I(z ) dz −
which simplifies into: −h
 z =h (25.2.4)
jIm e−jkr cos(kh cos θ)− cos kh − G(z − z , ρ)I (z )
Hφ (r, θ)= (25.1.12)
2πr sin θ z =−h

This agrees with the results of Sec. 17.3. The assumed symmetry of I(z) implies a discontinuity of its derivative at z = 0. In-
deed, setting I(z)= F(|z|), for some continuous and continuously differentiable func-
tion F(·), we find,
25.2 Improved Near-Field Calculation
I (z) = sign(z)F (|z|) ⇒ I (0+)= −I (0−)= F (0)
The current on a thin linear antenna is determined from the solution of the Hallén or
I (z) = 2δ(z)F (0)+sign2 (z) F (|z|)
Pocklington integral equations; for example, the latter is,
h Using these into Eq. (25.2.4) and splitting the integration range [−h, h] into three
I(z )(∂2z + k2 )G(z − z , a)dz = −4πjω Ein (z) (25.2.1) parts, [−h, 0−], [0−, 0+], [0+, h], we obtain:
−h
h  h  0−  0+  h  0−  0+  h  0−  h  0−  h
− = + + − − − = + − −
For a center-fed antenna, the impressed field is related to the driving voltage V0 at the −h −h −h 0− 0+ −h 0− 0+ −h 0+ −h 0+
antenna terminals by Ein (z)= V0 δ(z). The boundary condition that the net tangential
where we have canceled the terms over [0−, 0+]; indeed, it is easily verified that:
E-field vanish on the antenna surface requires that,
 0+  
Ez (z, a)= −Ein (z)= −V0 δ(z) (25.2.2) G(z − z , ρ) I (z )+k2 I(z ) dz = 2G(z, ρ)F (0)
0−

where Ez (z, a) is the field on the antenna surface (i.e., at ρ = a) generated by the  0+
current. Thus, the net field is zero, Ez,tot (z, a)= Ez (z, a)+Ein (z)= 0. It follows then G(z − z , ρ)I (z ) = 2G(z, ρ)F (0)
0−
from Eq. (25.2.2) that Ez (z, a) must vanish along the antenna, except at z = 0.
As we saw in Sec. 24.4, the assumption of a sinusoidal current can be justified on the Using the following notation for the principal-value integral,
basis of Pocklington’s equation, but it represents at best a crude approximation. The h  0−  h
resulting electric field does not satisfy condition (25.2.2), as can be seen setting ρ = a − = +
into Eq. (25.1.9). −h −h 0+
25.2. Improved Near-Field Calculation 1227 1228 25. Coupled Antennas

it follows from Eq. (25.2.4) that, Our procedure for obtaining improved near fields is to first get an improved solution
h for the current I(z) and then use it in Eq. (25.2.6) to calculate the field Ez (z, ρ). We will
 
4πjωEz (z, ρ) = − G(z − z , ρ) I (z )+k2 I(z ) dz use the three-term approximation for the current:
−h     
 0−  h     kz kh
I(z)= A1 sin(k|z|)− sin(kh) + A2 cos(kz)− cos(kh) + A3 cos − cos
− G(z − z , ρ)I (z ) − G(z − z , ρ)I (z ) 2 2
−h 0+ (25.2.11)
and fix the coefficients A1 , A2 , A3 by fitting this expression to a numerical solution as
which gives,
discussed in Sec. 24.6, and then, use Eq. (25.2.11) into (25.2.6) with the integral term
h   evaluated numerically. Fig. 25.2.1 shows the results of such a calculation for a half-
4πjωEz (z, ρ) = − G(z − z , ρ) I (z )+k2 I(z ) dz + wave antenna l = 0.5λ with radius a = 0.005λ. Fig. 25.2.2 shows the results for a
−h
 (25.2.5) full-wave antenna l = 1.0λ with the same radius. The required quantities appearing in
  (25.2.6) are calculated as follows:
+ 2I (0+)G(z, ρ)−I (h) G(z − h, ρ)+G(z + h, ρ)
     
kh 3 kz
where we used I (h)= −I (−h). Finally, we can write, I (z )+k2 I(z )= −k2 A1 sin kh − k2 A2 cos kh − k2 A3 cos − cos
2 4 2
 h −jkR  
e
4πjωEz (z, ρ) = − I (z )+k2 I(z ) dz +
−h R I (0+)= −I (0−)= kA1
 −jkR1 (25.2.6)
e−jkR0 e e−jkR2  
1 kh
+ 2I (0+) − I (h) + I (h)= −I (−h)= kA1 cos kh − kA2 sin kh − kA3 sin
R0 R1 R2 2 2

The last three terms are the standard terms found in the previous section. The The numerical solutions were obtained by solving the Hallén equation with point-
principal-value integral term represents the correction that must be added to enable the matching, pulse basis functions, and the exact kernel using M = 100 upper-half current
boundary conditions. The other field components can now be obtained from Ez using samples In . These current samples were then used as in Eq. (24.6.15) to obtain the
similar procedures as in the previous section. For Hφ , we find: parameters A1 , A2 , A3 .
h The upper-left graphs show the current I(z) of Eq. (25.2.11) together with the sam-
 
−4πjkρHφ (z, ρ) = − e−jkR I (z )+k2 I(z ) dz + ples In to which it was fitted.
−h (25.2.7) The upper-right graphs show the magnitude of Ez (z, ρ) as a function of ρ for z fixed
  at z = 0.2h. The behavior of Ez (z, ρ) is consistent initially with a logarithmic depen-
+ 2I (0+)e−jkR0 − I (h) e−jkR1 + e−jkR2
dence on ρ as predicted by King and Wu [1730,1731] and discussed below, followed then
which may also be written in the form: by the expected 1/ρ decrease arising from the last three standard terms of Eq. (25.2.6),
h which are represented by the dashed curves.
The left middle-row graphs display the logarithmic dependence more clearly by plot-
− 4πjkρHφ (z, ρ)= I(z )(∂2z + k2 )e−jkR dz (25.2.8)
−h ting the real and imaginary parts of Ez (z, ρ) versus ln(ρ/a), including the King-Wu
approximation of Eq. (25.2.15).
obtained by reversing the above differential identity steps. Similarly, we have:
The right middle-row graphs show the magnitude of the field Ez (z, a) at the surface
h of the antenna as a function of z over the interval 0 ≤ z ≤ h. Except at the feed and end
z − z −jkR    
−4πjωρEρ (z, ρ) = − e I (z )+k2 I(z ) dz + points, the field is effectively zero as required by the boundary conditions.
−h R
 (25.2.9) To observe the importance of the correction term, that is, the principal-value integral
z z − h −jkR1 z + h −jkR2
+ 2I (0+) e−jkR0 − I (h) e + e in Eq. (25.2.6), the third-row graphs display the real and imaginary parts of Ez (z, a)
R0 R1 R2 versus z. Plotted separately are also the correction and standard terms, which appear
which may also be written as, always to have opposite signs canceling each other so that the net field is zero.
The graphs for Fig. 25.2.1 were generated by the following MATLAB code (for Figure
h  
z − z −jkR 25.2.2 simply set L=1):
− 4πjωρEρ (z, ρ)= I(z )(∂2z + k2 ) e dz (25.2.10)
−h R
L = 0.5; h = L/2; a = 0.005; k = 2*pi; eta = 377;
M = 100; [In,zn] = hdelta(L,a,M,’e’,’p’); % Hallen solution
25.2. Improved Near-Field Calculation 1229 1230 25. Coupled Antennas

l = 0.5λ, a = 0.005λ l = 0.5λ, a = 0.005λ, z = 0.2h l = 1.0λ, a = 0.005λ l = 1.0λ, a = 0.005λ, z = 0.2h
12 4 2.5 2
total
log(ρ) approx
10 standard term 2
3

|E (z,ρ)| (V/m)

|E (z,ρ)| (V/m)
8
|I(z)| (mA)

|I(z)| (mA)
1.5
total
6 2 1 log(ρ) approx
standard term
1

z
4
1
0.5
2
3−term fit 3−term fit
numerical numerical
0 0 0 0
0 0.05 0.1 0.15 0.2 0.25 0 0.2 0.4 0.6 0.8 1 0 0.1 0.2 0.3 0.4 0.5 0 0.2 0.4 0.6 0.8 1
z/λ ρ/λ z/λ ρ/λ
l = 0.5λ, a = 0.005λ, z = 0.2h |Ez(z,a)| l = 0.5λ, a = 0.005λ, z = 0.2h |Ez(z,a)|
8 2 40
real part real part
6 real part log(ρ) approx real part log(ρ) approx
imag part 80 imag part
30
4 imag part log(ρ) approx 1 imag part log(ρ) approx
60
2
20
0 40 0

−2 10
20
−4 −1
0
0
−6

−8 −2
0 1 2 3 4 5 0 0.05 0.1 0.15 0.2 0.25 0 1 2 3 4 5 0 0.1 0.2 0.3 0.4 0.5
ln(ρ/a) z/λ ln(ρ/a) z/λ
real part of Ez(z,a) imaginary part of Ez(z,a) real part of Ez(z,a) imaginary part of Ez(z,a)
20 20 20 20

0 0

0 0
−20 −20

−40 −40
−20 −20

−60 total −60 total total total

correction term correction term correction term correction term
standard term standard term standard term standard term
−80 −80 −40 −40
0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25 0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z/λ z/λ z/λ z/λ

Fig. 25.2.1 Calculated near field Ez (z, ρ) for l = 0.5λ. Fig. 25.2.2 Calculated near field Ez (z, ρ) for l = 1.0λ.

Inp = In(M+1:end); znp = zn(M+1:end); % keep upper-half only

G = @(x,r) exp(-j*k*sqrt(x.^2 + r.^2))./sqrt(x.^2 + r.^2); % kernel function
z = 0:h/100:h; Helm = @(z) -k^2*(A(1)*sin(k*h) + A(2)*cos(k*h) + A(3)*(cos(k*h/2)-3/4*cos(k*z/2)));
A = kingfit(L,Inp,znp,3); I = kingeval(L,A,z); % 3-term fit
z = 0.2*h; r = linspace(a,200*a, 1001); logr = log(r/a);
s = 1000; % scale in units of mA S = -j*eta/4/pi/k; % scale factor, note omega*epsilon = k/eta
plot(z,abs(I)*s,’-’, znp,abs(Inp)*s,’.’, ’markersize’,11); % upper-left graph
[wi,zi] = quadrs([-h,0,h],32); % quadrature weights and evaluation points
I1h = k*(A(1)*cos(k*h) - A(2)*sin(k*h) - A(3)/2 * sin(k*h/2)); % I’(h)
I10 = A(1)*k; % I’(0+) for i=1:length(r),
25.2. Improved Near-Field Calculation 1231 1232 25. Coupled Antennas

GHelm = G(z-zi,r(i)) .* Helm(zi); component can be obtained from Maxwell’s equation:

E1(i) = (wi’*GHelm) * S; % correction term
E2(i) = (- I1h * (G(z-h,r(i)) + G(z+h,r(i))) + 2*I10 * G(z,r(i))) * S; I (z) jωQ(z)
E(i) = E1(i) + E2(i); jωEρ = −∂z Hφ = − =
2πρ 2πρ
end
that is, for small ρ:
Eapp = E(1) - Helm(z) * logr * 2*S; % King-Wu approximation adjusted by Ez(z,a)
Q(z)
Eρ (z, ρ)= (25.2.14)
figure; plot(r,abs(E), r,abs(Eapp),’:’, r,abs(E2),’--’); % upper-right graph 2πρ
figure; plot(logr,real(E), logr,real(Eapp),’--’,...
The same result can also be derived from Eq. (25.2.10) by recognizing the small-ρ
logr,imag(E),’-.’, logr,imag(Eapp),’:’); % middle-left graph 
limit (z − z )e−jkR /R → sign(z − z )e−jk|z−z | , which satisfies the Helmholtz identity:
clear E E1 E2;

z = linspace(0,h,201); r = a; (∂2z + k2 )sign(z − z )e−jk|z−z | = 2∂z δ(z − z )

for i=1:length(z), Combining Eqs. (25.2.13) and (25.2.14) into the Faraday equation (25.2.12), we have,
GHelm = G(z(i)-zi,r) .* Helm(zi);
E1(i) = (wi’*GHelm) * S;
Q  (z) I(z) j I (z)+k2 I(z)
E2(i) = (- I1h * (G(z(i)-h,r) + G(z(i)+h,r)) + 2*I10 * G(z(i),r)) * S; ∂ρ Ez = ∂z Eρ + jωμHφ = + jωμ =
E(i) = E1(i) + E2(i); 2πρ 2πρ ω 2πρ
end
Integrating from ρ = a, we obtain the small-ρ King-Wu approximation:
 ρ
figure; plot(z,abs(E),’-’); % middle-right graph
figure; plot(z,real(E), z,real(E1),’--’, z,real(E2),’:’); % lower-left graph j 
figure; plot(z,imag(E), z,imag(E1),’--’, z,imag(E2),’:’); % lower-right graph
Ez (z, ρ)= Ez (z, a)+ I (z)+k2 I(z) ln (25.2.15)
2πω a
Next, we discuss the King-Wu small-ρ approximation [1730,1731]; see also McDonald Strictly speaking, we must set Ez (z, a)= 0 because of the boundary condition. How-
[1785]. First, we note that the Hφ and Eρ components in Eqs. (25.2.8) and (25.2.10) ever, in our numerical solution, we have kept the term Ez (z, a), which is small but not
were obtained by using Maxwell’s equations (25.1.2), that is, Ampère’s laws ∂ρ (ρHφ )= necessarily exactly zero, in order to compare the analytical calculation (25.2.15) with
jω ρEz and jωEρ = −∂z Hφ . We may also verify Faraday’s law, which has only a φ the numerical solution. The left middle-row graphs confirm the linear dependence on
component in this case: ln(ρ/a) with the right slope.
∂ρ Ez − ∂z Eρ = jωμHφ (25.2.12) For longer antennas, up to about l = 3λ, the four-term approximation discussed in
Sec. 24.6 can be used and leads to similar results. In this case, the following current
Indeed, this can be derived from Eqs. (25.2.3), (25.2.8), and (25.2.10) by using the identity:
expressions should be used:
  
∂ e−jkR ∂ z − z −jkR    
ρ + e = −jke−jkR I(z) = A1 sin(k|z|)− sin(kh) + A2 cos(kz)− cos(kh) +
∂ρ R ∂z R
         
kz kh 3kz 3kh
For a thin antenna, the small-ρ dependence of Hφ is obtained by taking the limit + A3 cos − cos + A4 cos − cos
 4 4 4 4
ρ → 0 in the right-hand side of Eq. (25.2.8). In this limit, we have e−jkR = e−jk|z−z | ,     
which is recognized as the Green’s function of the one-dimensional Helmholtz equation kh 15 kz
 I (z )+k2 I(z ) = −k2 A1 sin kh − k2 A2 cos kh − k2 A3 cos − cos
discussed in Sec. 24.3 that satisfies (∂2z +k2 )e−jk|z−z | = −2jkδ(z−z ). It follows then, 4 16 4
    
h h 3kh 7 3kz
 − k2 A4 cos − cos
−4πjkρHφ (z, ρ) = I(z )(∂2z + k2 )e−jkR dz → I(z )(∂2z + k2 )e−jk|z−z | dz 4 16 4
−h −h
h I (0+)= −I (0−)= kA1
= −2jk I(z )δ(z − z )dz = −2jkI(z)    
−h 1 kh 3 3kh
I (h)= −I (−h)= kA1 cos kh − kA2 sin kh − kA3 sin − kA4 sin
or, for small ρ, 4 4 4 4
I(z)
Hφ (z, ρ)= (25.2.13) We observe in the upper-right figures that the maximum values of |Ez (z, ρ)| occur
2πρ
roughly at distance:
Let Q(z) denote the charge density per unit z-length along the antenna, which is λ
ρ= (25.2.16)
related to I(z) via the charge conservation equation I (z)+jωQ(z)= 0. Then, the Eρ 20
25.3. Self and Mutual Impedance 1233 1234 25. Coupled Antennas

and this remains roughly true for antenna lengths 0.5 ≤ l/λ ≤ 1.3 and radii 0.001 ≤ The quantities Z11 , Z22 are the self impedances of the two antennas and are approx-
a/λ ≤ 0.007 and for a variety of distances along the antenna, such as, 0.2h ≤ z ≤ 0.7h. imately equal to the input impedances of the isolated antennas, that is, when the other
Thus, this distance may be taken as a rough measure of the distance beyond which antenna is absent. If antenna-2 is open-circuited, so that I2 = 0, then the second of
the standard terms begin to take over and the sinusoidal current approximation becomes Eqs. (25.3.2) gives (25.3.1).
justified. In order to derive convenient expressions that allow the calculation of the mutual
The mutual impedance formulas that we develop in succeeding sections are based and self impedances, we use the reciprocity result given in Eq. (24.5.6) for the short-
on the sinusoidal assumption, and therefore, they can be used more reliably for antenna circuit current and open-circuit voltage induced on a receiving antenna in the presence
separations d that are greater than that of Eq. (25.2.16). For example, to increase one’s of an incident field.
confidence, one could take the separations to be greater than, say, double the above If antenna-2 is open-circuited and the z-component of the electric field generated
value, that is, d ≥ λ/10. by antenna-1 and incident on antenna-2 is E21 (z), then according to Eq. (24.5.6), the
induced open-circuit voltage will be:
 h2
25.3 Self and Mutual Impedance 1
V21,oc = − E21 (z)I2 (z)dz (25.3.3)
I2 −h2
The mutual coupling between antennas cannot be ignored if the antennas are near each
other. The mutual impedance is a measure of such proximity effects [2,1799–1811]. where h2 = l2 /2, and I2 (z), I2 = I2 (0) are the current and input current on antenna-2
Consider two parallel center-driven linear dipoles, as shown in Fig. 25.3.1. Their when it is transmitting. It follows from definition (25.3.1) that:
distance along the x-direction is d and their centers are offset by b along the z-direction.
 h2
V21,oc 1
Z21 = =− E21 (z)I2 (z)dz (25.3.4)
I1 I1 I2 −h2

Assuming that the currents are sinusoidal,

 
sin k(h1 − |z|)  
I1 (z) = I1 = Im1 sin k(h1 − |z|)
sin kh1
 
sin k(h2 − |z|)  
I2 (z) = I2 = Im2 sin k(h2 − |z|)
sin kh2

then, according to Eq. (25.1.9) the electric field E21 (z) along antenna-2 will be:


jηIm1 e−jkR1 e−jkR2 e−jkR0
Fig. 25.3.1 Parallel linear dipoles. Ez (z)= − + − 2 cos kh1 (25.3.5)
4π R1 R2 R0

If antenna-1 is driven and antenna-2 is open-circuited, the near field generated by

where −h2 ≤ z ≤ h2 , and R1 , R2 , R0 are defined in Fig. 25.3.1:
the current on antenna-1 will cause an open-circuit voltage, say V21,oc on antenna-2. The

mutual impedance of antenna-2 due to antenna-1 is defined to be:
R0 = d2 + (z + b)2
V21,oc 
Z21 = (25.3.1) R1 = d2 + (z + b − h1 )2 (25.3.6)
I1

where I1 is the input current on antenna-1. Reciprocity implies that Z12 = Z21 . More R2 = d2 + (z + b + h1 )2
generally, if both antennas are driven, then, the relationship of the driving voltages to
the input currents is given by: Inserting Eq. (25.3.5) into (25.3.4) and rearranging some constants, we find the final
expression for the mutual impedance Z21 :
V1 = Z11 I1 + Z12 I2  h2
(25.3.2) jη
V2 = Z21 I1 + Z22 I2 Z21 = F(z)dz (25.3.7)
4π sin kh1 sin kh2 −h2
25.3. Self and Mutual Impedance 1235 1236 25. Coupled Antennas


e−jkR1 e−jkR2 e−jkR0   radius is zero. This allows us to estimate the self-impedance of an infinitely thin half-
F(z)= + − 2 cos kh1 sin k(h2 − |z|) (25.3.8) wavelength antenna by setting L = 0.5 and a = 0:
R1 R2 R0
Z = imped(0.5, 0)= 73.0790 + 42.5151j Ω
This is the mutual impedance referred to the input terminals of the antennas. If
one or both of the antennas have lengths that are multiples of λ, then one or both of
Similarly, for radii a = 0.001λ and 0.005λ, we find:
the denominator factors sin kh1 , sin kh2 will vanish resulting in an infinite value for the
mutual impedance. Z = imped(0.5, 0.001)= 73.0784 + 42.2107j Ω
This limitation is caused by the sinusoidal current assumption. We saw in Chap. 24 Z = imped(0.5, 0.005)= 73.0642 + 40.6319j Ω
that the actual input currents are not zero in a real antenna. On the other hand, in most
applications of Eq. (25.3.7) the lengths differ slightly from half-wavelength for which the A resonant antenna is obtained by adjusting the length L such that the reactance part of Z
sinusoidal approximation is good. becomes zero. The resonant length depends on the antenna radius. For zero radius, this
length is L = 0.48574823 and the corresponding impedance, Z = 67.1843 Ω. 

The definition (25.3.4) can also be referred to the maximum currents by normalizing
by the factor Im1 Im2 , instead of I1 I2 . In this case, the mutual impedance is Z21m =
Example 25.3.2: Consider two identical parallel half-wavelength dipoles in side-by-side arrange-
Z21 sin kh1 sin kh2 , that is, ment separated by distance d. The antenna radius is a = 0.001 and therefore, its self
 h2
jη impedance is as in the previous example. If antenna-1 is driven and antenna-2 is parasitic,
Z21m = F(z)dz (25.3.9)
4π −h2 that is, short-circuited, then Eq. (25.3.2) gives:
The self-impedance of a single antenna can be calculated also by the same formula
V1 = Z11 I1 + Z12 I2
(25.3.7). Evaluating the near-field on the surface of the single antenna, that is, at d = a, 0 = Z21 I1 + Z22 I2
where a is the antenna radius, and setting h2 = h1 and b = 0 in Eq. (25.3.6), we find:
Solving the second for the parasitic current I2 = −I1 Z21 /Z22 and substituting in the first,
 h1  h1
1 jη we obtain driving-point impedance of the first antenna:
Z11 = − E11 (z)I1 (z)dz = F(z)dz (25.3.10)
I12 −h1 4π sin2 kh1 −h1 
V1 Z12 Z21 Z2
Zin = = Z11 − = Z11 1 − 21
2
I1 Z22 Z11


e−jkR1 e−jkR2 e−jkR0  
F(z)= + − 2 cos kh1 sin k(h1 − |z|) (25.3.11) where we used Z12 = Z21 and Z22 = Z11 . The ratio Z21 2 2
/Z11 quantifies the effect of the
R1 R2 R0
coupling and the deviation of Zin from Z11 . For example, we find the values:

   d 0.125λ 0.25λ 0.50λ 0.75λ 1.00λ

R0 = a2 + z2 , R1 = a2 + (z − h1 )2 , R2 = a2 + (z + h1 )2 (25.3.12) |Z21 /Z11 |2 0.58 0.35 0.15 0.08 0.05

The MATLAB function imped implements Eq. (25.3.7), as well as (25.3.10). It returns
Thus, the ratio decreases rapidly with increasing distance d. Fig. 25.3.2 shows a plot of
both Z21 and Z21m and has usage:
Z21 versus distance d. 

[Z21,Z21m] = imped(L2,L1,d,b) % mutual impedance of dipole 2 due to dipole 1

[Z21,Z21m] = imped(L2,L1,d) % b = 0, side-by-side arrangement
For separations d that are much larger than the antenna lengths, the impedance Z21
[Z,Zm] = imped(L,a) % self impedance falls like 1/d. Indeed, it follows from Eq. (25.3.6) that for large d, all three distances
R0 , R1 , R2 become equal to d. Therefore, (25.3.8) tends to:
where all the lengths are in units of λ. The function uses 16-point Gauss-Legendre
integration, implemented with the help of the function quadr, to perform the integral e−jkd    
F(z)→ 2 − 2 cos kh1 sin k(h2 − |z|)
in Eq. (25.3.7). d
In evaluating the self impedance of an antenna with a small radius, the integrand
which, when inserted into (25.3.7), gives the asymptotic form:
F(z) varies rapidly around z = 0. To maintain accuracy in the integration, we split the
integration interval into three subintervals, as we mentioned in Sec. 24.10. jη(1 − cos kh1 )(1 − cos kh2 ) e−jkd
Z21 → , for large d (25.3.13)
π sin kh1 sin kh2 kd
Example 25.3.1: Because the function imped uses an even length (that is, 16) for the Gauss-
Legendre integration, the integrand F(z) is never evaluated at z = 0, even if the antenna The envelope of this asymptotic form was superimposed on the graph of Fig. 25.3.2.
The oscillatory behavior of Z21 with distance is essentially due to the factor e−jkd .
25.3. Self and Mutual Impedance 1237 1238 25. Coupled Antennas

Mutual Impedance, Z21 = R21 + j X21 The input impedance (25.3.10) deserves a closer look. Replacing the exponential
80
resistance R21 integrals in (25.3.16) in terms of their real and imaginary parts,
reactance X21
1/d envelope  π
E1 (ju)= −γ − ln u + Cin (u)+j Si (u)−
40 2

as defined in Eq. (G.5), then (25.3.10) can be expressed in the following form, where we
set Z11 = Zin = Rin + jXin , h1 = h, and l = 2h:
0

η A + jB
Zin = Rin + jXin = (25.3.17)
2π sin2 kh
−40
0 1 2 3 4
 
d/λ With the definitions l± = a2 + h2 ± h and L± = a2 + 4h2 ± 2h, we obtain:

Fig. 25.3.2 Mutual impedance between identical half-wave dipoles vs. separation. A = Cin (kl+ )+Cin (kl− )−2Cin (ka)
1  
+ cos kl 2Cin (kl+ )−Cin (kL+ )+2Cin (kl− )−Cin (kL− )−2Cin (ka) (25.3.18)
An alternative computation method of the mutual impedance is to reduce the inte- 2
grals (25.3.7) to the exponential integral E1 (z) defined in Appendix F, taking advantage 1  
+ sin kl 2Si (kl− )−Si (kL− )+Si (kL+ )−2Si (kl+ )
of MATLAB’s built-in function expint. 2
 
By folding the integration range [−h1 , h1 ] in half and writing sin k(h2 − |z|) as a
B = Si (kl+ )+Si (kl− )−2Si (ka)
sum of exponentials, Eq. (25.3.7) can be reduced to a sum of terms of the form:
 h1  1  
e−jkR −jksz + cos kl 2Si (kl+ )−Si (kL+ )+2Si (kl− )−Si (kL− )−2Si (ka)
G(z0 , s)= e dz , R = d2 + (z − z0 )2 , s = ±1 (25.3.14) 2 (25.3.19)
0 R
 
1 aL+
which can be evaluated in terms of E1 (z) as: + sin kl 2Cin (kl+ )−Cin (kL+ )+Cin (kL− )−2Cin (kl− )+2 ln
2 l2+
 
G(z0 , s)= se−jksz0 E1 (ju0 )−E1 (ju1 ) (25.3.15) These expressions simplify substantially if we assume that the radius a is small, as
is the case in practice. In particular, assuming that ka 1 and a h, the quantities
with  l± and L± can be approximated by:
u0 = k d2 + z20 − sz0
 a2 a2
l+  2h = l , l− = 
u1 = k d2 + (h1 − z0 )2 + s(h1 − z0 ) l+ l
(25.3.20)
Indeed, the integral in (25.3.7) can be written as a linear combination of 10 such terms: a2 a2
L+  4h = 2l , L− = 
L+ 2l
 h1 
10
F(z)dz = ci G(zi , si ) (25.3.16) Noting that Si (x) and Cin (x) vanish at x = 0, we may neglect all the terms whose
−h1 i=1 arguments are kl− , kL− , or ka, and replace kl+ = kl and kL+ = 2kl, obtaining:
with the following values of zi , ci , and si , where c1 = ejkh2 /(2j): 1   1  
A = Cin (kl)+ cos kl 2Cin (kl)−Cin (2kl) + sin kl Si (2kl)−2Si (kl) (25.3.21)
2 2
i zi si ci i zi si ci
  
1 h1 − b 1 c1 6 h1 − b −1 c∗
1 1  1 2a
B = Si (kl)+ cos kl 2Si (kl)−Si (2kl)]+ sin kl 2Cin (kl)−Cin (2kl)+2 ln
2 −h1 + b 1 c1 7 −h1 + b −1 c∗
1 2 2 l
3 −h1 − b 1 c1 8 −h1 − b −1 c∗
1
(25.3.22)
4 h1 + b 1 c1 9 h1 + b −1 c∗
1
We note that A is independent of the radius a and leads to the same expression for
5 b 1 −4c1 cos kh1 10 b −1 −4c∗
1 cos kh1
the radiation resistance that we found in Sec. 17.3 using Poynting methods.
An additional approximation can be made for the case of a small dipole. Assuming
The MATLAB function Gi implements the “Green’s function integral” of (25.3.14). that kh 1, in addition to ka 1 and a h, we may expand each of the above terms
The function imped2, which is an alternative to imped, uses (25.3.16) to calculate (25.3.7).
25.4. Coupled Two-Element Arrays 1239 1240 25. Coupled Antennas

into a Taylor series in the variable kh using the following Taylor series expansions of where ρ = x x̂ + y ŷ is the cylindrical radial vector. Inserting (25.4.1) and performing
the functions Si (x) and Cin (x): the x , y integrations, we obtain:
1 3 1 5 1 2 1 4 1  h1  h2
Si (x) x − x + x , Cin (x) x − x + x6 (25.3.23) μ e−jkR1 μ e−jkR2
18 600 4 96 4320 Az (z, ρ )= I1 (z )dz + I2 (z )dz (25.4.2)
4π −h1 R1 4π −h2 R2
Such expansions, lead to the following input impedance Z = R + jX to the lowest
non-trivial order in kl : where, as shown in Fig. 25.4.1, R1 , R2 are the distances from the z point on each antenna
 to the (x, y, z) observation point, that is,
η 1 4(1 + L)
Zin = Rin + jXin = (kl)2 +j (small dipole) (25.3.24)
2π 12 kl  
where L = ln(2a/l). The resistance R is identical to that obtained using the Poynting R1 = (z − z )2 +(x − x1 )2 +(y − y1 )2 = (z − z )2 +|ρ
ρ − d 1 |2
method and assuming a linear approximation to the sinusoidal antenna current, which   (25.4.3)
is justified when kh 1: R2 = (z − z )2 +(x − x2 )2 +(y − y2 )2 = (z − z )2 +|ρ
ρ − d 2 |2
   
sin k(h − |z|) k(h − |z|) |z| where d1 = (x1 , y1 ) and d2 = (x2 , y2 ) are the xy-locations of the antenna centers. The
I(z)= I0  I0 = I0 1 − (25.3.25)
sin kh kh h z-component of the electric field generated by the two antenna currents will be:

jωμ Ez (z, ρ )= (∂2z + k2 )Az (z, ρ )

25.4 Coupled Two-Element Arrays
Working with the rescaled vector potential V(z, ρ)= 2jcAz (z, ρ), we rewrite:
Next, we consider a more precise justification of Eq. (25.3.2) and generalize it to the
 h1  h2
case of an arbitrary array of parallel linear antennas. Fig. 25.4.1 shows two z-directed jη e−jkR1 jη e−jkR2
parallel dipoles with centers at locations (x1 , y1 ) and (x2 , y2 ). V(z, ρ )= I1 (z )dz + I2 (z )dz (25.4.4)
2π −h1 R1 2π −h2 R2
We assume that the dipoles are center-driven by the voltage generators V1 , V2 . Let
I1 (z), I2 (z) be the currents induced on the dipoles by the generators and by their mu-
(∂2z + k2 )V(z, ρ )= −2kEz (z, ρ ) (25.4.5)
tual interaction, and let h1 , h2 be the half-lengths of the antennas, and a1 , a2 , their
radii. Then, assuming the thin-wire model, the total current density will have only a Denoting by V1 (z) and V2 (z) the values of V(x, y, z) on the surfaces of antenna-1
z-component given by: and antenna-2, we obtain from Eq. (25.4.4):

Jz (x , y , z )= I1 (z )δ(x − x1 )δ(y − y1 )+I2 (z )δ(x − x2 )δ(y − y2 ) (25.4.1) V1 (z)= V11 (z)+V12 (z)
(25.4.6)
V2 (z)= V21 (z)+V22 (z)

The z-components of the electric fields induced on the surfaces of antenna-1 and
antenna-2 are obtained by applying Eq. (25.4.5) to each term of (25.4.6):

E1 (z)= E11 (z)+E12 (z)

(25.4.7)
E2 (z)= E21 (z)+E22 (z)

where we defined, for p, q = 1, 2:

 hq
jη
Vpq (z)= Gpq (z − z )Iq (z )dz
2π −hq (25.4.8)
2 2
(∂z + k )Vpq (z)= −2kEpq (z)
Fig. 25.4.1 Array of two linear antennas.
and the impedance kernels:
It follows that the magnetic vector potential will be:

 e−jkRpq
μ e−jkR Gpq (z − z )= , Rpq = (z − z )2 +d2pq (25.4.9)
Az (z, ρ )= Jz (x , y , z )dx dy dz , 
R = |r − r | Rpq
4π R
25.4. Coupled Two-Element Arrays 1241 1242 25. Coupled Antennas

If p = q, then dpq is the xy-distance between the antennas, and if p = q, it is the We will solve these numerically in Sec. 25.7. Next, we derive Eq. (25.3.2). Accord-
radius of the corresponding antenna, that is, ing to definitions (25.3.4) and (25.3.10), the mutual impedance between antenna-p and
 antenna-q can be restated as follows, for p, q = 1, 2:
d12 = d21 = |d1 − d2 | = (x1 − x2 )2 +(y1 − y2 )2
(25.4.10)  hp
1
d11 = a1 , d22 = a2 Zpq = − Epq (z)Ip (z)dz (25.4.14)
Ip Iq −hp
Thus, Vpq (z) and Epq (z) are the vector potential and the z-component of the electric
field induced on antenna-p by the current Iq (z) on antenna-q. and, more explicitly:
To clarify these definitions, Fig. 25.4.2 shows a projected view of Fig. 25.4.1 on the  h1  h1
xy plane. The point P with radial vector ρ is the projection of the observation point 1 1
Z11 = − E11 (z)I1 (z)dz , Z12 = − E12 (z)I1 (z)dz
(z, ρ ). When P coincides with a point, such as P2 , on the surface of antenna-2 defined I1 I 1 −h1 I1 I2 −h1
by the radial vector ρ 2 , then the distance (P2 O2 )= |ρ ρ2 − d2 | will be equal to the antenna  h2  h2
1 1
radius a2 , regardless of the location of P2 around the periphery of the antenna. Z21 = − E21 (z)I2 (z)dz , Z22 = − E22 (z)I2 (z)dz
I2 I 1 −h2 I2 I2 −h2
On the other hand, the distance (P2 O1 )= |ρ ρ2 − d1 | varies with P2 . However, because
the separation d12 is typically d12  a2 , such variation is minor and we may replace Using these definitions and Eq. (25.4.12), we find:
|ρ
ρ2 − d1 | by |d2 − d1 |. Thus, in evaluating V(z, ρ 2 ) on antenna-2, we may use Eq. (25.4.4)
 h1
with R1 , R2 defined by: 1  
Z11 I1 + Z12 I2 = − E11 (z)+E12 (z) I1 (z)dz
  I1 −h1
R1 = (z − z )2 +|d2 − d1 |2 = (z − z )2 +d212  h1
1   1
  (25.4.11) =− −V1 δ(z) I1 (z)dz = V1 I1 (0)= V1
I1 −h1 I1
R2 = (z − z )2 +|ρ
ρ2 − d2 |2 = (z − z )2 +a22
where, by definition, I1 (0)= I1 . Similarly, we can show the second of Eq. (25.3.2).
The mutual impedance defined in Eq. (25.4.14) actually satisfies the reciprocity sym-
metry condition, Zpq = Zqp . To write it in a form that shows this condition explicitly,
we replace Epq (z) by (25.4.8) and obtain the alternative symmetric form:

 hp  hq
jη Ip (z)Iq (z ) 2
Zpq = (∂z + k2 )Gpq (z − z )dz dz (25.4.15)
4πk −hp −hq Ip Iq

If we assume that the currents are sinusoidal, that is, for p = 1, 2,

 
sin k(hp − |z|)
Fig. 25.4.2 Array of two linear antennas. Ip (z)= Ip (25.4.16)
sin khp

Now, on the surface of the first antenna, the electric field Ez must cancel the field of then, in Eq. (25.4.15) the ratios Ip (z)/Ip and hence Zpq become independent of the input
the delta-gap generator in order for the total tangential field to vanish, that is, E1 (z)= currents at the antenna terminals and depend only on the geometry of the antennas.
−E1,in (z)= −V1 δ(z). Similarly, on the surface of the second antenna, we must have
E2 (z)= −E2,in (z)= −V2 δ(z). Then, Eq. (25.4.7) becomes:
25.5 Arrays of Parallel Dipoles
E11 (z)+E12 (z)= E1 (z)= −V1 δ(z)
(25.4.12)
E21 (z)+E22 (z)= E2 (z)= −V2 δ(z) The above results on two antennas generalize in a straightforward fashion to several
antennas. Fig. 25.5.1 depicts the case of K parallel dipoles in side-by-side arrangement
Combining these with the Eq. (25.4.8), we obtain the coupled version of the Hallén-
with centers at positions (xp , yp ), and driving voltages, lengths, half-lengths, and radii,
Pocklington equations:
Vp , lp , hp , ap , where p = 1, 2 . . . , K.
 
(∂2z + k2 ) V11 (z)+V12 (z) = 2kV1 δ(z) Assuming sinusoidal currents as in Eq. (25.4.16), we define the mutual impedances
  (25.4.13) Zpq by Eq. (25.4.14) or (25.4.15), where p, q take on the values p, q = 1, 2 . . . , K. The
(∂2z + k2 ) V21 (z)+V22 (z) = 2kV2 δ(z)
25.5. Arrays of Parallel Dipoles 1243 1244 25. Coupled Antennas

the knowledge of the currents Ip (z), one can obtain the radiation pattern of the array.
Indeed, the radiation fields are obtained from Eq. (17.1.6), that is,

e−jkr
E = θ̂
θEθ = θ̂
θ jkη Fz (θ, φ)sin θ
4πr
(25.5.5)
e−jkr
H = φ̂
φHφ = φ̂
φ jk Fz (θ, φ)sin θ
4πr

where the radiation vector F = ẑFz has only a z-component given by:


Fz (θ, φ)= Jz (r )ejk·r dr (25.5.6)
V
Fig. 25.5.1 Two-dimensional array of parallel dipoles.
But, in the thin-wire approximation, the total current density of the array is:

Hallén-Pocklington equations (25.4.13) generalize into: 

K
Jz (r )= Ip (z )δ(x − xp )δ(y − yp )
p=1

K 
K
(∂2z + k2 ) Vpq (z)= −2k Epq (z)= 2kVp δ(z) , p = 1, 2, . . . , K (25.5.1)
Inserting this into Eq. (25.5.6) and performing the x , y integrations, we obtain:
q=1 q=1


K  hp

where Vpq (z) is defined by Eqs. (25.4.8) and (25.4.9). The mutual distances are: Fz (θ, φ)= ejkx xp +jky yp Ip (z )ejkz z dz (25.5.7)
−hp
  p=1
(xp − xq )2 +(yp − yq )2 , if p = q
dpq = (25.5.2) Using Eq. (25.4.16) for Ip (z) and replacing kz = k cos θ, we obtain:
ap , if p=q

Multiplying Eq. (25.5.1) by Ip (z) and integrating along the length of the pth antenna, 
K
2Ip cos(khp cos θ))− cos khp
Fz (θ, φ)= ejkx xp +jky yp (25.5.8)
and using the mutual impedance definitions (25.4.14), we obtain the generalization of p=1
k sin khp sin2 θ
Eq. (25.3.2) to the case of K antennas:
The radiation intensity is given, in general, by Eq. (16.1.4):

K
Vp = Zpq Iq , p = 1, 2, . . . , K (25.5.3) ηk2  
sin θ Fz (θ, φ)2
U(θ, φ)=
q=1 32π2

where Iq is the input current at the center of the qth antenna. Eq. (25.5.3) may be written Replacing kx = k sin θ cos φ and ky = k sin θ sin φ, we obtain:
in a compact matrix form:
 2
V = ZI  
(25.5.4) η  K cos(khp cos θ))− cos khp jk sin θ(xp cos φ+yp sin φ) 
U(θ, φ)=  Ip e 
 
where Z is the impedance matrix. For example, in the case K = 4, we have: 8π2 p=1 sin khp sin θ 
⎡ ⎤ ⎡ ⎤⎡ ⎤
V1 Z11 Z12 Z13 Z14 I1
⎢ V ⎥ ⎢ Z ⎥⎢ I ⎥ Thus, the normalized gain of the array will be, up to a proportionality constant:
⎢ 2 ⎥ ⎢ 21 Z22 Z23 Z24 ⎥⎢ 2 ⎥
V=⎢ ⎥=⎢ ⎥⎢ ⎥ = ZI
⎣ V3 ⎦ ⎣ Z31 Z32 Z33 Z34 ⎦ ⎣ I3 ⎦  2
 
V4 Z41 Z42 Z43 Z44 I4  K cos(khp cos θ))− cos khp jk sin θ(xp cos φ+yp sin φ) 

g(θ, φ)=  Ip e  (25.5.9)

p=1 sin khp sin θ 
We note that Z is a symmetric matrix, Z = ZT , as a consequence of the reciprocity
relations Zpq = Zqp .
Given the driving voltages Vp , Eq. (25.5.4) may be solved for the input currents Ip , Equations (25.5.4) and (25.5.9) provide a complete solution to the problem of cou-
which completely define the assumed sinusoidal currents Ip (z) of Eq. (25.4.16). From pled antenna arrays, based on the sinusoidal approximation for the currents. In the
25.5. Arrays of Parallel Dipoles 1245 1246 25. Coupled Antennas

special case of identical antennas, Eq. (25.5.9) factors as usual into an array factor and Because both gains are defined over a 2π-angular range, they must be plotted with
an element factor: the MATLAB functions abp2 and abz2, or in dB, with dbp2 and dbz2.
 2  2
   
 K 
jk sin θ(xp cos φ+yp sin φ)   cos(kh cos θ))− cos kh 
Example 25.5.1: Three-element parasitic array. Undriven parasitic antennas located near trans-

g(θ, φ)=  Ip e   
p=1   sin kh sin θ  mitting ones can act as reflectors or directors, directing the radiation towards certain
preferred directions. Fig. 25.5.2 shows an array of three half-wavelength dipoles. The
The MATLAB function impedmat calculates the K×K mutual impedance matrix Z of geometry is the same as that of Example 22.3.3. The xy-coordinates of the elements are
such an array, given the antenna lengths and radii, lp , ap , and the coordinates (xp , yp ), d1 = (0, 0), d2 = (0.5λ, 0), and d3 = (0, 0.5λ).

for p = 1, 2, . . . , K. It has usage:

Z = impedmat(L,a,d); % mutual impedance matrix of array of parallel dipoles

where all the lengths must be given in units of λ. It calls imped to calculate the individual
matrix elements Zpq .
The input parameters L, a, d are the vectors of antenna lengths, antenna radii, and
(xp , yp ) pairs, or the xp positions, if the array is along the x-axis:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
L1 a1 x1 , y1 x1
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ L2 ⎥ ⎢ a2 ⎥ ⎢ x2 , y2 ⎥ ⎢ x2 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
L=⎢ .. ⎥ , a=⎢ .. ⎥ , d=⎢ .. ⎥ or ⎢ .. ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ . ⎦ ⎣ . ⎦ ⎣ . ⎦ ⎣ . ⎦
LK aK x K , yK xK Fig. 25.5.2 Three-element array.
The MATLAB function gain2s calculates the E-plane and H-plane array gains using
Let V = [V1 , V2 , V3 ]T be the driving voltages of the three elements. If only element-1 is
Eq. (25.5.9) and assumes that the input currents Ip have been obtained by solving
driven and the others parasitic, we may take V = [1, 0, 0]T .
Eq. (25.5.4). It has usage:
If the mutual couplings between the antennas are ignored, that is, the impedance matrix Z
[ge,gh,th] = gain2s(L,d,I,N,ph0); % gain of 2D array of dipoles with sinusoidal currents
of Eq. (25.5.4) is taken to be diagonal, then, the input currents, will be I = [I1 , 0, 0] and the
[ge,gh,th] = gain2s(L,d,I,N); % equivalent to φ0 = 0
parasitic elements will be completely passive as though they were absent. The radiation
where the input parameters L, a have the same meaning as in impedmat, and I is the pattern would be that of a single half-wave dipole. In particular, the azimuthal pattern
would be omnidirectional.
vector of input currents I = [I1 , I2 , . . . , IK ]. The output angle parameter th is either the
polar or the azimuthal angle and takes N equally-spaced values in the interval [0, 2π]. This is not the case if the mutual couplings are taken into account. The parasitic elements
The H-plane gain gH (φ) is defined to be the azimuthal gain on the xy-plane corre- act as reflectors, reflecting the radiation back towards the active element-1. By the sym-
sponding to θ = π/2, and the E-plane gain gE (θ) is defined to be the polar gain on any metry of the arrangement, the maximum directivity will be in the direction with azimuthal
angle φ = −135o . Fig. 25.5.3 shows the resulting H-plane and E-plane radiation patterns
fixed azimuthal plane φ = φ0 , that is,
demonstrating this behavior. The dashed gains were computed by solving the coupled sys-
tem of Hallén equations for the exact currents on each of the three antennas, as discussed
gH (φ)= g(π/2, φ), 0 ≤ φ ≤ 2π
(25.5.10) in Example 25.7.1.
gE (θ)= g(θ, φ0 ), 0 ≤ θ ≤ 2π Assuming equal radii, a = 0.001λ, the 3×3 impedance matrix Z is found to be:
⎡ ⎤
Note that by allowing θ to vary over [0, 2π], the E-plane gain can give both the 73.08 + 42.21j −12.52 − 29.91j −12.52 − 29.91j
⎢ ⎥
forward and backward gain. The polar angle range [0, π] covers the forward direction Z = ⎣ −12.52 − 29.91j 73.08 + 42.21j −24.62 + 0.78j ⎦
φ = φ0 , whereas, the range [π, 2π] covers the backward direction φ = φ0 + π, that −12.52 − 29.91j −24.62 + 0.78j 73.08 + 42.21j
is, we have the equivalence:
Then, the solution of Eq. (25.5.4) is:
g(θ, φ0 )= g(θ − π, φ0 + π), π ≤ θ ≤ 2π ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
I1 1 0.0133∠−7.46o
⎢ ⎥ −1 −1 ⎢ ⎥ ⎢ o ⎥
This follows from the trigonometric identities: I = ⎣ I2 ⎦ = Z V = Z ⎣ 0 ⎦ = ⎣ 0.0066∠18.23 ⎦
I3 0 0.0066∠18.23o
sin(θ − π)cos(φ0 + π)= sin θ cos φ0
sin(θ − π)sin(φ0 + π)= sin θ sin φ0 The typical MATLAB code used to generate these graphs was as follows:
25.5. Arrays of Parallel Dipoles 1247 1248 25. Coupled Antennas

Azimuthal gain Polar gain towards φ0 = 45o Azimuthal gain Polar gain towards φ0 = 45o
90o θ 0o θ 90o θ 0o θ
o o
120 60 30o 30o 120
o
60
o
30o 30o

o o
150 30 60o 60o 150
o
30
o
60o 60o

φ φ

o −9 −6 −3 o o −9 −6 −3 o o −9 −6 −3 o o −9 −6 −3 o
180 0 90 90 180 0 90 90
dB dB dB dB

o o o o
−150
o −30 120
o 120 −150
o −30 120
o 120

o o o o
o
−120 −60 150
o 150 o
−120 −60 150
o 150
o o o o
−90 180 −90 180

Fig. 25.5.3 H-plane and E-plane radiation patterns, V = [1, 0, 0]T . Fig. 25.5.4 H-plane and E-plane radiation patterns, V = [0, 1, 1]T .

L = [0.5, 0.5, 0.5]; % lengths Because of the identical geometry, the impedance matrix Z is the same as that of the
a = [0.001, 0.001, 0.001]; % radii previous example. But, the input currents are different:
d = [0,0; 0.5,0; 0,0.5]; % xy locations
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
I1 0 0.0133∠18.23o
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
Z = impedmat(L,a,d); % impedance matrix I = ⎣ I2 ⎦ = Z−1 V = Z−1 ⎣ 1 ⎦ = ⎣ 0.0173∠−19.04o ⎦
V = [1; 0; 0]; % driving voltages
I3 1 0.0173∠−19.04 o
I = Z\V; % input currents

ph0 = 45; % 45o azimuthal plane for polar gain

The only change in the previous MATLAB code was to use V = [0, 1, 1]T . 

[ge1,gh1,ph] = gain2s(L,d,I,360,ph0); % gain2s assumes sinusoidal currents

Example 25.5.3: One of the earliest experimental studies of parasitic reflectors was by Nagy
[1803]. One of his arrangements is shown in Fig. 25.5.5 in which the driven element is at
M = 40; % number of upper-half samples the origin and the other three elements are parasitic. The antenna lengths were l = 1.19
m, and their radii a = 0.395 cm. The operating wavelength was λ = 2.5 meters, (i.e.,
[I,z] = hcoupled2(L,a,d,V,M); % solves for currents on all antennas frequency of 120 MHz.)

[ge2,gh2,ph] = gain2d(L,d,I,360,ph0); % gain2d uses Hallén currents

figure; dbz2(ph,gh1,30,12); dbadd2(2,’--’,ph,gh2,30,12);

figure; dbp2(ph,ge1,30,12); dbadd2(1,’--’,ph,ge2,30,12);

Anticipating the symmetry about the 45o azimuthal plane, the E-plane gain was computed
with φ0 = 45o . As expected, the polar plot shows that the maximum gain is in the backward
φ0 direction, that is, toward φ0 + 180o = 225o = −135o . 

Example 25.5.2: Next, consider the case when element-one is parasitic, but elements two and
three are driven by equal voltages, V = [0, 1, 1]T . If the mutual coupling is ignored, then
the two active elements act as an array which is broadside to the line joining them, that
is, maximum directivity is in the 45o azimuthal direction, but with both the forward and
the backward (i.e., −135o ) directions being equal. This pattern is shown in the upper-right
Fig. 25.5.5 Four-element parasitic array.
graph of Fig. 22.3.4.
If the mutual couplings are taken into account, element-1 will act as a reflector, reflecting It follows that, l = 0.476λ and a = 0.00158λ. Elements two and four were placed symmet-
towards the φ0 = 45o direction and reducing the gain in the opposite direction. This is rically along the y-axis at distances ±0.535λ, and element three was on the negative side
demonstrated in Fig. 25.5.4. As in the previous example, the dashed gains correspond to of the x-axis at distance 0.248λ from the origin. Fig. 25.5.6 shows the calculated patterns.
the exact coupled Hallén solution. We observe that the three parasitic antennas act as reflectors, enhancing the radiation in
the φ = 0 direction.
25.5. Arrays of Parallel Dipoles 1249 1250 25. Coupled Antennas

Azimuthal gain Polar gain towards φ0 = 0o Example 25.5.4: Coupled Dolph-Chebyshev array. In this example, we study the impact of
90o θ 0o θ
120
o
60
o
30o
30o mutual coupling on the array design methods of Chap. 23. For a typical array spacing of
half-wavelength, the mutual impedance matrix is diagonally dominant and therefore, there
150
o
30
o o
60 60
o will be some but minor impact on the design.

φ Fig. 25.5.7 shows a 15-element array of z-directed half-wavelength dipoles with spacing
−9 −6 −3 −9 −6 −3
d = λ/2 arranged along the x-axis. The antenna radii are a = 0.001λ.
o o o o
180 0 90 90
dB dB

o o
−150
o −30 120
o 120

o o
−120o −60 150o 150
o o
−90 180

Fig. 25.5.6 H-plane and E-plane radiation patterns, V = [1, 0, 0, 0]T .

Fig. 25.5.7 Fifteen-element Dolph-Chebyshev array.

This array was later studied theoretically by Brown [1804], using the same methods as
those presented here. Brown treated monopole antennas (i.e, half dipoles above a ground
We take the feed voltages V = [V1 , V2 , . . . , V15 ]T to be Dolph-Chebyshev weights that
plane,) and therefore, the values of his mutual impedances are half of ours. The inputs to
would steer the azimuthal array gain towards φ0 = 120o and would achieve a 20-dB side-
the design equations were the parameters:
lobe level. These weights can be designed with the function dolph.
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ If the mutual coupling is ignored, the impedance matrix Z will be proportional to the
0.476 0.00158 0.000, 0.000
⎢ 0.476 ⎥ ⎢ 0.00158 ⎥ ⎢ 0.000, 0.535 ⎥ identity matrix because the antenna elements are identical. Then, the input currents I will
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
L=⎢ ⎥, a=⎢ ⎥, d=⎢ ⎥
⎣ 0.476 ⎦ ⎣ 0.00158 ⎦ ⎣ −0.248, 0.000 ⎦ be essentially equal to the driving voltages V and the array will behave according to the
0.476 0.00158 0.000, −0.535 desired design.
If the mutual coupling is taken into account, the currents must be calculated from the
The impedance matrix elements are: solution of ZI = V and some distortions on the desired angular pattern may occur because
Z is no longer diagonal.
Z11 = Z22 = Z33 = 63.42∠0.65o , Z12 = Z14 = 26.76∠−123.87o
Fig. 25.5.8 shows the azimuthal and polar gain patterns with and without mutual coupling.
Z13 = 43.56∠−34.69o , Z23 = Z34 = 24.78∠−141.96o
The primary effect is to distort the sidelobe levels so that they are no longer equal. But
Z24 = 14.74∠53.15o
they are still acceptable as a close approximation to the desired Dolph-Chebyshev pattern.
The typical MATLAB code used in this example was as follows:
With V = [1, 0, 0, 0]T , the solution of ZI = V is:

⎡ ⎤ ⎡ ⎤
I1 0.0135∠−26.26o
⎢I ⎥ ⎢ 0.0043∠74.61o ⎥ K = 15;
⎢ 2 ⎥ ⎢ ⎥
I=⎢ ⎥=⎢ ⎥ ph0 = 120; % steering angle
⎣ I3 ⎦ ⎣ 0.0126∠116.70o ⎦
I4 0.0043∠4.61o L = 0.5 * ones(1,K); % vector of antenna lengths
a = 0.001 * ones(1,K); % antenna radii
and we find for the ratios: d = (0:K-1)*0.5; % equally-spaced with λ/2 spacing

I2 I4 I3 V = dolph(0.5, ph0, K, 20).’; % Dolph design with 20-dB sidelobes

= = 0.3180∠100.87o , = 0.9343∠142.96o
I1 I1 I1
Z = impedmat(L,a,d); % 15×15 impedance matrix
I = Z\V; % input currents
These numerical results are in close agreement with Brown’s [1804]. The dashed Hallén
gains are not shown, as in the previous examples, because they are virtually indistinguish- [ge,gh,ph] = gain2s(L,d,I,400,ph0); % gains with coupling
able from the sinusoidal ones (for M = 40.) 

figure; dbz2(ph,gh); % azimuthal gain

figure; dbp2(ph,ge); % polar gain
25.6. Yagi-Uda Antennas 1251 1252 25. Coupled Antennas

Azimuthal gain with coupling Azimuthal gain without coupling

90o 90o
o o
120 60 120o 60o

150o 30o 150o 30o

φ φ
−30 −20 −10 −30 −20 −10
180o 0o 180o 0o
dB dB

−150o −30o −150o −30o

−120o −60o −120o −60o

Fig. 25.6.1 Five-element Yagi-Uda array.
−90o −90o

Polar gain with coupling Polar gain without coupling

θ 0o θ θ 0o θ
o o and act as “directors.” The reflector and directors direct the radiation preferentially
30 30 30o 30o
towards endfire, that is, along the x-axis.
60
o
60
o
60
o
60
o The Yagi-Uda array is widely used as a TV reception antenna and achieves fairly good
directivity with such a simple structure. Good directivity characteristics are realized
with certain choices for the antenna lengths and separations.
−30 −20 −10 −30 −20 −10
90o
dB
90o 90o
dB
90o The analysis of the Yagi-Uda array follows the steps of the previous section. We
assume that there are K dipoles, with the last K − 2 being the directors, and that the
currents are sinusoidal as in Eq. (25.4.16) because the antenna lengths are of the order
120o 120o
120o 120o of half-wavelength. Then, we compute the mutual impedance matrix Z and the input
currents I = Z−1 V. Because only the second element is driven, the vector of voltages is:
150o 150o 150o 150o
180o 180o
V = [0, 1, 0, 0, . . . , 0]T (25.6.1)
  
Fig. 25.5.8 H-plane and E-plane patterns with and without coupling. (K−2) zeros

Once we have the input currents I = [I1 , I2 , . . . , IK ]T , the gain of the array is com-
[ge,gh,ph] = gain2s(L,d,V,400,ph0); % gains without coupling puted by Eq. (25.5.9), which simplifies into the following form because the dipoles lie
along the x-axis:
figure; dbz2(ph,gh);
figure; dbp2(ph,ge);  2
 
 K cos(khp cos θ))− cos khp jkxp sin θ cos φ 

g(θ, φ)=  Ip e  (25.6.2)

The E-plane polar gains were computed on the plane of the desired steering angle, that is, p=1 sin khp sin θ 
φ0 = 120o . The figures show that maximum gain is at θ = 90o in the φ0 direction. In the
case without coupling, we set I = V inside gain2s because any proportionality constant
We assume that the lengths and separations are such that the maximum gain is
gets canceled out. 

towards endfire, that is, towards θ = 90o , φ = 0o . The forward and backward gains,
and the forward-backward or front-to-back ratio are defined as:
25.6 Yagi-Uda Antennas gf
gf = gmax = g(90o , 0o ), gb = g(90o , 180o ), Rfb = (25.6.3)
gb
A special type of parasitic array is the Yagi-Uda array shown in Fig. 25.6.1. The z-
directed dipoles are arranged along the x-axis. The second dipole is driven; all others It follows that the normalized gain will be gn (θ, φ)= g(θ, φ)/gf . Integrating it
are parasitic. over all solid angles, we obtain the beam solid angle and hence the directivity of the
The first dipole has length slightly longer than that of the driven dipole, and acts as a Yagi-Uda array:
“reflector”. The elements to the right of the driven dipole have lengths slightly shorter,  π  2π
4π
ΔΩ = gn (θ, φ)sin θ dθ dφ, D= (25.6.4)
0 0 ΔΩ
25.6. Yagi-Uda Antennas 1253 1254 25. Coupled Antennas

In dB, the directivity and forward-backward ratio are 10 log10 D and 10 log10 Rfb . Reflector case, H−plane gain Reflector case, E−plane gain
90o θ 0o θ
The MATLAB function yagi implements the above design steps. It computes the input o
120 60o
30o 30o
currents I as well as the directivity and forward-backward ratio. Its usage is:
150o 30o 60o 60o
[I,D,Rfb] = yagi(L,a,d); % Yagi-Uda array design
φ
The function always assumes that the second element is the driven element and sets
−12 −8 −4 −12 −8 −4
180o 0o 90o 90o
the value of V according to Eq. (25.6.1). The double integral in Eq. (25.6.4) is done with a dB dB

16-point Gauss-Legendre quadrature integration formula for each integration variable.

Example 25.6.1: Reflectors and directors. The simplest possible Yagi-Uda array has one driven −150o −30o 120o 120o

element and either one reflector and no directors, or a single director and no reflector.
Fig. 25.6.2 depicts the two cases. −120o −60o 150o 150o
−90o 180o

Director case, H−plane gain Director case, E−plane gain

90o θ 0o θ
o o
120 60 30o 30o

o o o o
150 30 60 60

o −12 −8 −4 −12 −8 −4
180 0o 90o 90o
dB dB

−150o −30o 120o 120o

Fig. 25.6.2 The simplest Yagi-Uda arrays.

−120o −60o 150o 150o
−90o 180o
If the reflector is slightly longer than the driven element, and if the director is slightly
shorter, then in both cases the radiation will be directed to the right, along the x-axis. Fig. 25.6.3 H-plane and E-plane gains of simple Yagi-Uda arrays.
Fig. 25.6.3 shows the resulting radiation patterns.
The length of the driven element was 0.50λ and that of the reflector and director, 0.54λ
and 0.46λ, respectively. The antenna radii were a = 0.003λ and their separation d = 0.1λ. Example 25.6.2: Three-element Yagi. Here, we consider a three-element Yagi-Uda array with
The mutual impedances were calculated with impedmat: one reflector, one driven element, and one director. The corresponding antenna lengths,
radii, and locations along the x-axis (with the driven element at the origin) were in units of



92.47 + 104.19j 75.68 + 11.63j 73.07 + 41.37j 59.77 + 4.35j λ:
Z= , Z=
75.68 + 11.63j 73.07 + 41.37j 59.77 + 4.35j 57.65 − 17.01j ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
0.50 0.003 x1 −0.125
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
L = ⎣ 0.48 ⎦ , a = ⎣ 0.003 ⎦ , d = ⎣ x2 ⎦ = ⎣ 0 ⎦
The typical MATLAB code that was used was:
0.46 0.003 x3 0.125
L = [0.54,0.50] % reflector case
a = 0.003*[1,1]; % radii
The azimuthal and polar gains are shown in Fig. 25.6.4. The dashed gains correspond to the
d = [0,0.1]; % x-coordinates of locations exact coupled Hallén equations, as discussed in Example 25.7.3. The computed directivity
and front/back ratio were D = 8.18 dB and Rfb = 18.69 dB. Thus, the array achieves a gain
Z = impedmat(L,a,d); % impedance matrix of D − 2.15 = 6.03 dB over a single half-wavelength dipole.
I = Z\[0,1]’; % input currents
The impedance matrix was:
[ge,gh,th] = gain2s(L,d,I,400); % gain computation ⎡ ⎤
73.07 + 41.37j 60.47 − 0.97j 36.25 − 25.53j
⎢ ⎥
figure; dbz2(th,gh,30,16); % azimuthal gain Z = ⎣ 60.47 − 0.97j 64.93 + 11.75j 53.72 − 2.71j ⎦
figure; dbp2(th,ge,30,16); % polar gain 36.25 − 25.53j 53.72 − 2.71j 57.65 − 17.01j

The driving voltages were in the two cases: V = [0, 1]T and V = [1, 0]T . 

The input currents and input impedance of the driving element were:
25.6. Yagi-Uda Antennas 1255 1256 25. Coupled Antennas

Azimuthal gain Polar gain Unoptimized H−plane gain Unoptimized E−plane gain
o
90 θ 0o θ 90o θ 0o θ
o o o o o o
120 60 30 30 120 60 30o 30o

150o 30o 60o 60o 150o 30o 60o 60o

φ φ
−30 −20 −10 −30 −20 −10 −30 −20 −10 −30 −20 −10
180o 0o 90o 90o 180o 0o 90o 90o
dB dB dB dB

−150o −30o 120o 120o −150o −30o 120o 120o

−120o −60o 150o 150o −120o −60o 150o 150o

−90o 180o −90o 180o

Optimized H−plane gain Optimized E−plane gain

Fig. 25.6.4 Azimuthal and polar gains of three-element Yagi-Uda array. 90o θ 0o θ
o o
120 60 30o 30o

o o o o
150 30 60 60
⎡ ⎤ ⎡ ⎤
I1 −0.0290 + 0.0176j φ
⎢ ⎥ ⎢ ⎥ V2 1
I = ⎣ I2 ⎦ = ⎣ 0.1062 − 0.0182j ⎦ , Z2 = = = 9.15 + 1.57j −30 −20 −10 −30 −20 −10
I2 I2 180o
0o 90o 90o
I3 −0.0801 − 0.0256j dB dB

The typical MATLAB code for this example was:

−150o −30o 120o 120o

L = [0.50, 0.48, 0.46]; % antenna lengths

a = 0.003*[1,1,1]; % radii −120o −60o 150o 150o
d = [-0.125, 0, 0.125]; % x-locations −90o 180o

[I,D,Rfb] = yagi(L,a,d); % solve ZI = V Fig. 25.6.5 Gains of six-element Yagi-Uda array.

[ge,gh,th] = gain2s(L,d,I,360); % compute gains at 1o increments

The directors were identical and equally spaced at spacing of 0.31λ. The computed direc-
M = 40; % number of upper-half samples
tivity and front/back ratio were 11 dB and 9.84 dB, respectively. The optimized case has
[I,z] = hcoupled(L,a,d,[0,1,0],M); % compute Hallén currents
slightly different lengths and x-locations:

[ge2,gh2,ph] = gain2d(L,d,I,360); % gain of Hallén currents L = [L1 , L2 , L3 , L4 , L5 , L6 ]= [0.476, 0.452, 0.436, 0.430, 0.434, 0.430]

figure; dbz2(ph,gh); dbadd2(2,’--’,ph,gh2);

d = [x1 , x2 , x3 , x4 , x5 , x6 ]= [−0.25, 0, 0.289, 0.695, 1.018, 1.440]
figure; dbp2(ph,ge); dbadd2(1,’--’,ph,ge2);
Typical MATLAB code was as follows:
The driving voltages were defined within yagi to be V = [0, 1, 0] . T


L = [0.476, 0.452, 0.436, 0.430, 0.434, 0.430];

a = 0.003369 * [1,1,1,1,1,1];
Example 25.6.3: Optimized six-element Yagi. Chen and Cheng [1818] applied King’s three-term
d = [-0.25, 0, 0.289, 0.695, 1.018, 1.440];
current approximation [4] and devised procedures for optimizing the choices of the an-
tenna lengths and separations of Yagi-Uda arrays. Fig. 25.6.5 shows the gains before and [I,D,Rfb] = yagi(L,a,d);
after optimization of a six-element Yagi-Uda array calculated with the functions yagi and
gain2s. The antenna radii were a = 0.003369λ. [ge,gh,th] = gain2s(L,d,I,360);

For the unoptimized case, the antenna lengths and x-locations were in units of λ: figure; dbz2(th,gh,30,40);
figure; dbp2(th,ge,30,40);
L = [L1 , L2 , L3 , L4 , L5 , L6 ]= [0.510, 0.490, 0.430, 0.430, 0.430, 0.430]

d = [x1 , x2 , x3 , x4 , x5 , x6 ]= [−0.25, 0, 0.310, 0.620, 0.930, 1.240] The optimized directivity was 12.54 dB and the forward/backward ratio 17.6 dB. 

25.7. Hallén Equations for Coupled Antennas 1257 1258 25. Coupled Antennas

25.7 Hallén Equations for Coupled Antennas The basis-function expansion for the qth current is

M
In Sects. 25.4 and 25.5, we developed the Hallén-Pocklington equations for coupled an- Iq (z )= Iq (zm )Bq (z − zm ), q = 1, 2, . . . , K (25.7.8)
tennas, that is, Eqs. (25.4.8)–(25.4.9) and (25.5.1). Here, we discuss their numerical solu- m=−M
tion. On the pth antenna, we have: We are going to use only the pulse and the triangular bases defined by Eqs. (24.8.4)
2 2 and (24.11.2), respectively, with sample spacing Δ = Δq . For the triangular basis, we
(∂z + k )Vp (z)= 2kVp δ(z), p = 1, 2, . . . , K (25.7.1)
must set Δq = hq /M instead of Δq = 2hq /(2M + 1).
where Vp is the driving delta-gap input and Vp (z) is the sum of the (scaled) vector Inserting (25.7.8) into (25.7.6) and sampling along the p-th antenna, that is, at the
potentials due to the currents on all antennas: points zn = nΔp , for −M ≤ n ≤ M, we obtain the discretized system:
 hq
jη  
K M

 K  Iq (zm ) Gpq (zn − z )Bq (z − zm )dz = Cp cos kzn + Vp sin k|zn |
jη  hq
K
2π q=0 −hq
Vp (z)= Vpq (z)= Gpq (z − z )Iq (z )dz (25.7.2) m=−M
q=0
2 π q=0 −hq
We define the N×N impedance matrix Zpq whose nmth matrix element is:
 hq
where we recall the definition of the impedance kernel: jη
Zpq (n, m) = Gpq (zn − z )Bq (z − zm )dz
2π −hq

e−jkR  hq (25.7.9)
Gpq (z − z )= , R = (z − z )2 +d2pq (25.7.3) jη
R = Gpq (zn − zm − z)Bq (z)dz
2π −hq
and dpq are the mutual distances or radii, as defined in Eq. (25.5.2). The use of the ap-
proximate kernel in (25.7.3) is well-justified for the off-diagonal terms (p = q) because with −M ≤ n, m ≤ M. In particular, for the pulse basis, these take the form:
 Δq /2
the distances dpq are typically much greater than the radii. However, for the diagonal jη
Zpq (n, m)= Gpq (zn − zm − z)dz (25.7.10)
term (p = q), one could use the exact kernel given as in (24.7.10) by, 2π −Δq /2
 π/2  and, for the triangular basis:
2 e−jkR
Gpp (z − z )= dθ , R = (z − z )2 +4a2p − 4a2p sin2 θ (25.7.4)  Δq 
π 0 R jη |z|
Zpq (n, m)= 1− Gpq (zn − zm − z)dz (25.7.11)
2π −Δq Δq
Following the discussion of Sec. 24.3, the solution of (25.7.1) is of the form:
Denoting Iq (m)= Iq (zm ), the discretized Hallén system becomes:
Vp (z)= Cp cos kz + Vp sin k|z|, −hp ≤ z ≤ hp (25.7.5)

K 
M
where we assumed that all the antennas are center-driven, and therefore, Vp (z) will Zpq (n, m)Iq (m)= Cp cos kzn + Vp sin k|zn | (25.7.12)
be even in z. Combining (25.7.5) with (25.7.2), we obtain the coupled system of Hallén q=0 m=−M

equations, for p = 1, 2, . . . , K:
where p = 1, 2, . . . , K. And, written in a more compact form:
 hq
jη 
K

K
Gpq (z − z )Iq (z )dz = Cp cos kz + Vp sin k|z| (25.7.6) Zpq Iq = Cp cp + Vp sp (25.7.13)
2π q=0 −hq
q=0

The K constants C1 , C2 , . . . , CK are determined by imposing the end conditions on where we defined the N-dimensional vectors:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
the K currents: Ip (hp )= 0, for p = 1, 2, . . . , K. Iq (M) cos kzM sin kzM
⎢ .. ⎥ ⎢ .. ⎥ ⎢ .. ⎥
To solve this system, we use a basis-function expansion of the form of Eq. (24.8.3) ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ . ⎥ ⎢ . ⎥ ⎢ . ⎥
and apply point matching. For simplicity, we take the same number of sampling points ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ Iq (1) ⎥ ⎢ cos kz1 ⎥ ⎢ sin kz1 ⎥
on each antenna, N = 2M + 1. Because the antenna lengths may be different, the sample ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
Iq = ⎢ Iq (0) ⎥ , cp = ⎢ cos kz0 ⎥ , sp = ⎢ sin kz0 ⎥ (25.7.14)
spacings will also be different. On the qth antenna we have, ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ Iq (1) ⎥ ⎢ cos kz1 ⎥ ⎢ sin kz1 ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ .. ⎥ ⎢ .. ⎥ ⎢ .. ⎥
2hq lq ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ . ⎦ ⎣ . ⎦ ⎣ . ⎦
zm = mΔq , Δq = = , −M ≤ m ≤ M (25.7.7)
2M + 1 N Iq (M) cos kzM sin kzM
25.7. Hallén Equations for Coupled Antennas 1259 1260 25. Coupled Antennas

and used the even symmetry in z. The vectors cp and sp depend on p through the sample Solving for Ip and multiplying by uT , we obtain the condition:
spacing in zn = nΔp , −M ≤ n ≤ M. 
The system (25.7.13) provides K coupled matrix equations by which to determine the uT Ip = Cp uT Z− 1 T −1
pp cp + Vp u Zpp sp − u T Z−1
pp Zpq Iq = 0
q=p
K sampled current vectors I1 , I2 , . . . , IK on each antenna. The N×N matrices Zpq are not
Toeplitz, unless the antennas are identical, in which case Δp = Δq and the Zpq (n, m) Defining the quantity up = Z−1
we solve this condition for Cp :
pp u,
depends only on the difference n − m. Of course, for p = q, Zpp is both symmetric and ⎛ ⎞
Toeplitz. 1 
Cp = ⎝ uT T ⎠
However, while not Toeplitz, the matrix Zpq is reversal-invariant because of the p Zpq Iq − Vp up sp
uT
p cp q=p
property Zpq (n, m)= Zpq (−n, −m), which follows from Eq. (25.7.9). Therefore, the
matrix system (25.7.13) can be wrapped in half by the procedure discussed in Sec. 24.9, Inserting Cp into Eq. (25.7.17) and rearranging terms, we obtain:
which replaced the matrix equation (24.9.6) by (24.9.7).  
 cp uT cp uT
Here, each N×N matrix Zpq is wrapped to size (M + 1)×(M + 1) by the same Zpp Ip + I−
p
Zpq Iq = Vp I −
p
sp (25.7.18)
process. The resulting system looks identical to (25.7.13), except the currents and right- q=p
cT
p up cT
p up

hand sides are essentially half those of (25.7.14):

⎡ ⎤ ⎡ ⎤ ⎡ ⎤ To simplify it, we define the (M + 1)×(M + 1) projection matrices:
Iq (0) cos kz0 sin kz0
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ cp uT
⎢ Iq (1) ⎥ ⎢ cos kz1 ⎥ ⎢ sin kz1 ⎥ Pp = I −
p
, p = 1, 2, . . . , K
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ (25.7.19)
Iq = ⎢ .. ⎥, cp = ⎢ .. ⎥, sp = ⎢ .. ⎥ (25.7.15) cT
p up
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ . ⎦ ⎣ . ⎦ ⎣ . ⎦
Iq (M) cos kzM sin kzM Then, Eq. (25.7.18) can be written in the form:

In particular, if all antennas are identical, then the wrapping process can be made Zpp Ip + Pp Zpq Iq = Vp Pp sp (25.7.20)
even more efficient using the Toeplitz-Hankel properties of the wrapped matrices, as q=p

discussed in Example 24.9.1. In any case, we will assume in the sequel that the system Thus, eliminating the constants Cp by enforcing the end conditions, amounts to
(25.7.13) has been wrapped in half. replacing the impedance matrices Zpq by the projected ones:
If the constants Cp were known, the solution of the system (25.7.13) could be ob- 
tained by writing it as a single block-matrix linear system of the form: Zpp , if q=p
Z̄pq = (25.7.21)
Pp Zpq , if q = p
ZI = Cc + Vs (25.7.16)
and the term sp by the projected one, s̄p = Pp sp . Then, Eq. (25.7.20) can be written in
where Z is the K×K block matrix whose pqth matrix element is the (M + 1)×(M + 1) the form:
matrix Zpq , and C, V are appropriate block-diagonal matrices. The vectors I, c, s are the 
K
concatenations of Ip , cp , sp . For example, in the case K = 3, the system (25.7.13) reads: Z̄pq Iq = Vp s̄p , p = 1, 2, . . . , K (25.7.22)
q=0
Z11 I1 + Z12 I2 + Z13 I3 = C1 c1 + V1 s1
Z21 I1 + Z22 I2 + Z23 I3 = C2 c2 + V2 s2 or, compactly in the block-matrix form:
Z31 I1 + Z32 I2 + Z33 I3 = C3 c3 + V3 s3
Z̄I = Vs̄ (25.7.23)
This can be written in the 3×3 block-matrix form:
⎡ ⎤⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤⎡ ⎤ with solution: ⎡ ⎤ ⎡ ⎤
Z11 Z12 Z13 I1 C1 I 0 0 c1 V1 I 0 0 s1 I1 V1 s̄1
⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ Z21 Z22 Z23 ⎦ ⎣ I2 ⎦ = ⎣ 0 C2 I ⎦ ⎣ c2 ⎦ + ⎣ 0 V2 I 0 ⎦ ⎣ s2 ⎦ ⎢ I2 ⎥ ⎢ V2 s̄2 ⎥
⎢ ⎥ ⎢ ⎥
Z31 Z32 Z33 I3 0 0 C3 I c3 0 0 V3 I s3 I = ⎢ . ⎥ = Z̄−1 Vs̄ = Z̄−1 ⎢ . ⎥ (25.7.24)
⎢ .. ⎥ ⎢ .. ⎥
⎣ ⎦ ⎣ ⎦
where I is the (M + 1)×(M + 1) identity matrix. IK VK s̄K
Next, we discuss the determination of the constants Cp . The condition Ip (M)= 0
The MATLAB function hcoupled implements the above solution procedure. First, it
can be written vectorially in the form uT Ip = 0, where u = [0, . . . , 0, 1]T , as was done in
constructs the impedance matrices Zpq by calculating the integrals in Eq. (25.7.9) using a
Sec. 24.9. Separating the pth term of the pth equation in (25.7.13), we have:
32-point Gauss-Legendre quadrature integration formula. Second, it wraps the matrices

Zpp Ip + Zpq Iq = Cp cp + Vp sp (25.7.17) Zpq in half and puts them together into the block-matrix Z. And third, it constructs the
q=p projected matrix Z̄ and the solution (25.7.24). Its usage is:
25.7. Hallén Equations for Coupled Antennas 1261 1262 25. Coupled Antennas

[I,z] = hcoupled(L,a,d,V,M,ker,basis); % solve Hallén equations for coupled dipoles Performing the z-integration, we finally get:

where L, a, d are the vectors of antenna lengths, radii, and xy-locations, and V is the
b

M 
K
sin(kz Δp /2)
vector of the driving voltages V = [V1 , V2 , . . . , VK ]. The parameters L, a, d have the jkz mΔp jkx xp +jky yp
Fz (θ, φ)= Ip (m)e e Δp (25.7.25)
same usage as in the functions yagi and gain2s. The string input ker takes the val- m=−M p=1
kz Δp /2
ues ’e’,’a’ for using the exact or the approximate kernel in the computation of the
diagonal elements of the impedance matrix. The basis parameter can take only the two where kx = k sin θ cos φ, ky = k sin θ sin φ, and kz = k cos θ, and b = 1 for the pulse
values ’p’,’t’ for pulse or triangular basis. basis and b = 2 for the triangular one. The corresponding normalized gain of the array
The output I is the (2M + 1)×K matrix whose pth column is the double-sided vector will be, up to a constant:  2
of current samples Ip (zm ), zm = mΔp , −M ≤ m ≤ M. Thus, the matrix elements of g(θ, φ)= sin θ Fz (θ, φ) (25.7.26)
I are I(m, p)= Ip (zm ). Similarly, the pth column of the output matrix z holds the The MATLAB function gain2d computes the E-plane polar gain and the H-plane
sampled z-locations on the pth antenna, that is, z(m, p)= mΔp . azimuthal gain from Eqs. (25.7.25) and (25.7.26). Its usage is:
The output matrix I is obtained by using the MATLAB function reshape to reshape
  [ge,gh,th] = gain2d(L,d,I,N,ph0,basis) % gain of 2D array of antennas with Hallén currents
the K(M + 1) -dimensional column vector solution (25.7.24) into a matrix of size (M +
[ge,gh,th] = gain2d(L,d,I,N,ph0) % equivalent to basis=’p’ (pulse basis)
1)×K, and then, symmetrizing it to size (2M + 1)×K.
[ge,gh,th] = gain2d(L,d,I,N,basis) % equivalent to ph0=0
A faster version of hcoupled is the function hcoupled2, which assumes that the
[ge,gh,th] = gain2d(L,d,I,N) % equivalent to ph0=0, basis=’p’
antennas are identical. It is faster because it makes use of the Toeplitz-Hankel structure
of the wrapped matrices Zpq to construct them more efficiently. Its usage is: where the current input I is exactly the same as the output matrix from hcoupled
or hcoupled2. The meaning of the outputs are exactly the same as in the function
[I,z] = hcoupled2(L,a,d,V,M,ker,basis)); % Hallén equations for coupled identical dipoles
gain2s discussed in Sec. 25.5. The string basis takes the values ’p’ or ’t’, for pulse
where I has the same meaning as in hcoupled, but z is now a single column vector, that or triangular basis.
is, zm = mΔ, −M ≤ m ≤ M. In both hcoupled2 and hcoupled, the final solution is The difference between gain2s and gain2d is that the former assumes the currents
   
obtained by solving the system (25.7.23), which is K(M+1) × K(M+1) -dimensional. are sinusoidal and I represents only the input currents, I = [I1 , I2 , . . . , IK ]. whereas in
In order to conveniently manipulate the block impedance matrices, we developed a the latter, the full (2M + 1)×K current matrix is needed, I = [I1 , I2 , . . . , IK ].
MATLAB function, blockmat, which is used extensively inside hcoupled2 and hcoupled.
Example 25.7.1: Hallén solution of parasitic array. Consider the three-element array of Example
It allows one to create block matrices and to extract or insert sub-blocks. Its usage is as 25.5.1 and shown in Fig. 25.5.2. The Hallén currents on each antenna can be computed
follows: by using hcoupled2 because the elements are identical. Fig. 25.7.1 shows the computed
    sampled currents with N = 2M + 1 = 81 or M = 40.
Z = blockmat(K,K,M+1,M+1); % create a K(M + 1) × K(M + 1) matrix of zeros
Zpq = blockmat(K,K,p,q,Z); % extract pqth submatrix of Z
Current on driven element Current on parasitic elements
Z = blockmat(K,K,p,q,Z,Zpq); % insert Zpq into pqth submatrix of Z 15 15
  Hallen Hallen
s = blockmat(K,1,M+1,1); % create a K(M + 1) -dimensional column of zeros sinusoidal sinusoidal
sp = blockmat(K,1,p,1,s); % extract the pth subvector of s
s = blockmat(K,1,p,1,s,sp); % insert sp into pth subvector of s
10 10
|I1 (z)| mA

|I2 (z)| mA
Once the sampled currents Ip (m) are known, the gain of the array can be computed
by finding the total current density, J(r)= ẑJz (r) :
5 5

K 
K 
M
Jz (r)= Ip (z)δ(x − xp )δ(y − yp )= Ip (m)Bp (z − zm )δ(x − xp )δ(y − yp )
p=1 p=1 m=−M

0 0
where we used Eq. (25.7.8). The corresponding radiation vector is: 0 0.05 0.1 0.15 0.2 0.25 0 0.05 0.1 0.15 0.2 0.25
z/λ z/λ
 
M 
K  hp
Fz (θ, φ) = Jz (r)ejk·r d3 r = Ip (m)ejkx xp +jky yp Bp (z − zm )ejkz z dz Fig. 25.7.1 Currents on driven and parasitic antennas.
m=−M p=1 −hp


M 
K  hp Because of the symmetry, the currents on the two parasitic antennas are the same. For all
= Ip (m)ejkx xp +jky yp ejkz zm Bp (z)ejkz z dz three antennas, the currents are essentially sinusoidal, justifying the use of this assump-
m=−M p=1 −hp tion. The gains computed with gain2d, and under the sinusoidal assumption with gain2s,
25.7. Hallén Equations for Coupled Antennas 1263 1264 25. Coupled Antennas

were shown in Fig. 25.5.3. The MATLAB code used to generate the currents and the gains We observe that the sinusoidal assumption is fairly accurate. The MATLAB code used to
was given in Example 25.5.1. 

generate the current graphs was as follows:

Example 25.7.2: Full-wavelength parasitic array. If one or more of the antennas has length L = [0.50, 0.48, 0.46]; h = L/2;
equal to a multiple of λ, the analysis methods based on the sinusoidal assumption break a = 0.003 * [1, 1, 1];
down because the impedance matrix computed with Eq. (25.4.15) becomes infinite. d = [-0.125, 0, 0.125];
V = [0, 1, 0]; % can be defined as column or row
On the other hand, the numerical solution of the Hallén system can still be carried through
giving a finite answer. Fig. 25.7.2 shows the gains and currents of the parasitic array of k = 2*pi;
Example 25.5.1, but all the antennas being full-wavelength elements, l = λ. The distance of M = 40;
the parasitic antennas to the driven element was also changed to d = 0.25λ from d = 0.5λ.
[I,z] = hcoupled(L,a,d,V,M);
Azimuthal gain Polar gain
90
o
0
o I1 = abs(I(M+1:end,1)); m1 = max(I1); z1 = z(M+1:end,1);
θ θ
120o 60o 30o 30o I2 = abs(I(M+1:end,2)); m2 = max(I2); z2 = z(M+1:end,2);
I3 = abs(I(M+1:end,3)); m3 = max(I3); z3 = z(M+1:end,3);

150o 30o 60o 60o

s1 = 0:h(1)/50:h(1); Is1 = m1*abs(sin(k*(h(1)-s1)));
φ s2 = 0:h(2)/50:h(2); Is2 = m2*abs(sin(k*(h(2)-s2)));
s3 = 0:h(3)/50:h(3); Is3 = m3*abs(sin(k*(h(3)-s3)));
−9 −6 −3 −9 −6 −3
180o 0o 90o 90o
dB dB
figure; plot(z1, I1, ’.’, s1, Is1, ’:’);
figure; plot(z2, I2, ’.’, s2, Is2, ’:’);
−30o
figure; plot(z3, I3, ’.’, s3, Is3, ’:’);
−150o 120o 120o

−120o −60o 150o 150

o Note that I1 , I2 , and I3 are obtained from the three columns of I, and z1 , z2 , and z3 from
−90o 180o the three columns of z. Only the currents on the upper-half of each antenna are plotted.
Current on driven element Current on parasitic element The sinusoidal currents are scaled to the maximum values of the corresponding Hallén
2 0.4
Hallen Hallen currents.
sinusoidal 0.35 sinusoidal
Current on reflector element Current on driven element Current on director element
120 120 120
1.5 0.3 Hallen Hallen Hallen
sinusoidal sinusoidal sinusoidal
100 100 100
|I1 (z)| mA

|I2 (z)| mA

0.25
80 80 80

|I1 (z)| mA

|I2 (z)| mA

|I3 (z)| mA
1 0.2
60 60 60

0.15
40 40 40

0.5 0.1 20 20 20

0.05 0
0 0.05 0.1 0.15 0.2 0.25
0
0 0.05 0.1 0.15 0.2 0.25
0
0 0.05 0.1 0.15 0.2 0.25
z/λ z/λ z/λ
0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
z/λ z/λ
Fig. 25.7.3 Currents on the Yagi antennas.

Fig. 25.7.2 Gains and currents of full-wavelength parasitic array.

These examples demonstrate the remark made earlier that the sinusoidal assumption is
justified only for antennas with lengths near half a wavelength.


The sinusoidal assumption for the driven element is fairly accurate except near z = 0,
where the current has an non-zero value. But on the parasitic element, the sinusoidal
assumption is completely wrong. 

25.8 Problems
Example 25.7.3: Three-element Yagi-Uda array. Here, we compute the currents on the three
25.1 Show that the asymptotic form of Eq. (25.3.7) for the mutual impedance between two parallel
antennas of the Yagi-Uda array of Example 25.6.2. Because the antennas are not identical,
dipoles separated by a distance d is given by
the function hcoupled must be used. The gains were computed with gain2s and gain2d
   
in Example 25.6.2. and shown in Fig. 25.6.4. The sampled currents on the three antennas jη kh1 kh2 e−jkd
are shown in Fig. 25.7.3. Z21 = tan tan , for large d
π 2 2 kd
25.8. Problems 1265

25.2 Using the higher-order terms in the series (25.3.23), show that the input impedance Zin =
R + jX of a small dipole is given as follows to order (kl)4 , where L = ln(2a/l):

η

1 1

η

4(1 + L) 1

2

1

11
 26
R= (kl)2 + (kl)4 , X= − L+ (kl)− L− (kh)3
2π 12 360 2π kl 3 3 180 30

25.3 Consider a small dipole with a linear current given by Eq. (25.3.25). Determine the radiation Appendices
vector, and the radiated electric and magnetic fields at a far distance r from the dipole. Cal-
culate the radiated power Prad by integrating the radial Poynting vector over a large sphere.
Then identify the radiation resistance R through the definition:

1
Prad = R|I0 |2
2

and show R is the same as that given by Eq. (25.3.24)

A. Physical Constants
We use SI units throughout this text. Simple ways to convert between SI and other
popular units, such as Gaussian, may be found in Refs. [123–126].
The Committee on Data for Science and Technology (CODATA) of NIST maintains
the values of many physical constants [112]. The most current values can be obtained
from the CODATA web site [1823]. Some commonly used constants are listed below:

quantity symbol value units

speed of light in vacuum c0 , c 299 792 458 m s−1
permittivity of vacuum 0 8.854 187 817 × 10−12 F m−1
permeability of vacuum μ0 4π × 10−7 H m −1
characteristic impedance η0 , Z0 376.730 313 461 Ω

electron charge e 1.602 176 462 × 10−19 C

electron mass me 9.109 381 887 × 10−31 kg

Boltzmann constant k 1.380 650 324 × 10−23 J K−1

Avogadro constant NA , L 6.022 141 994 × 1023 mol−1
Planck constant h 6.626 068 76 × 10−34 J/Hz

Gravitational constant G 6.672 59 × 10−11 m3 kg−1 s−2

Earth mass M⊕ 5.972 × 1024 kg
Earth equatorial radius ae 6378 km

In the table, the constants c, μ0 are taken to be exact, whereas 0 , η0 are derived
from the relationships: $
1 μ0
0 = , η0 = = μ0 c
μ0 c2 0
The energy unit of electron volt (eV) is defined to be the work done by an electron
in moving across a voltage of one volt, that is, 1 eV = 1.602 176 462 × 10−19 C · 1 V, or

1 eV = 1.602 176 462 × 10−19 J