C h a p t e r S e v e n

Power Dividers and

Directional Couplers

Power dividers and directional couplers are passive microwave components used for power
division or power combining, as illustrated in Figure 7.1. In power division, an input signal is
divided into two (or more) output signals of lesser power, while a power combiner accepts
two or more input signals and combines them at an output port. The coupler or divider may
have three ports, four ports, or more, and may be (ideally) lossless. Three-port networks take
the form of T-junctions and other power dividers, while four-port networks take the form of
directional couplers and hybrids. Power dividers usually provide in-phase output signals with
an equal power division ratio (3 dB), but unequal power division ratios are also possible. Di-
rectional couplers can be designed for arbitrary power division, while hybrid junctions usually
have equal power division. Hybrid junctions have either a 90◦ or a 180◦ phase shift between the
output ports.
A wide variety of waveguide couplers and power dividers were invented and characterized
at the MIT Radiation Laboratory in the 1940s. These included E- and H -plane waveguide
T-junctions, the Bethe hole coupler, multihole directional couplers, the Schwinger coupler,
the waveguide magic-T, and various types of couplers using coaxial probes. In the mid-1950s
through the 1960s, many of these couplers were reinvented to use stripline or microstrip tech-
nology. The increasing use of planar lines also led to the development of new types of couplers
and dividers, such as the Wilkinson divider, the branch line hybrid, and the coupled line direc-
tional coupler.
We will first discuss some of the general properties of three- and four-port networks, and
then treat the analysis and design of several of the most common types of power dividers,
couplers, and hybrids.

7.1 BASIC PROPERTIES OF DIVIDERS AND DOUPLERS

In this section we will use properties of the scattering matrix developed in Section 4.3 to de-
rive some of the basic characteristics of three- and four-port networks. We will also define
317
318 Chapter 7: Power Dividers and Directional Couplers

Divider P2 = ␣P1 P1 = P2 + P3 Divider P2

P1 or or
coupler P3 = (1 – ␣)P1 coupler P3

(a) (b)

FIGURE 7.1 Power division and combining. (a) Power division. (b) Power combining.

isolation, coupling, and directivity, which are important quantities for the characterization
of couplers and hybrids.

Three-Port Networks (T-Junctions)

The simplest type of power divider is a T-junction, which is a three-port network with two
inputs and one output. The scattering matrix of an arbitrary three-port network has nine
independent elements:


S11 S12 S13
[S] = S21 S22 S23 . (7.1)
S31 S32 S33
If the device is passive and contains no anisotropic materials, then it must be reciprocal
and its scattering matrix will be symmetric (Si j = S ji ). Usually, to avoid power loss, we
would like to have a junction that is lossless and matched at all ports. We can easily show,
however, that it is impossible to construct such a three-port lossless reciprocal network that
is matched at all ports.
If all ports are matched, then Sii = 0, and if the network is reciprocal, the scattering
matrix of (7.1) reduces to


0 S12 S13
[S] = S12 0 S23 . (7.2)
S13 S23 0
If the network is also lossless, then energy conservation requires that the scattering matrix
satisfy the unitary properties of (4.53), which leads to the following conditions [1, 2]:
|S12 |2 + |S13 |2 = 1, (7.3a)
|S12 |2 + |S23 |2 = 1, (7.3b)
|S13 | + |S23 | = 1,
2 2
(7.3c)
∗
S13 S23 = 0, (7.3d)
∗
S23 S12 = 0, (7.3e)
∗
S12 S13 = 0. (7.3f)
Equations (7.3d)–(7.3f ) show that at least two of the three parameters (S12 , S13 , S23 ) must
be zero. However, this condition will always be inconsistent with one of equations (7.3a)–
(7.3c), implying that a three-port network cannot be simultaneously lossless, reciprocal,
and matched at all ports. If any one of these three conditions is relaxed, then a physically
realizable device is possible.
If the three-port network is nonreciprocal, then Si j = S ji , and the conditions of input
matching at all ports and energy conservation can be satisfied. Such a device is known as a
circulator, and generally relies on an anisotropic material, such as ferrite, to achieve non-
reciprocal behavior. Ferrite circulators will be discussed in more detail in Chapter 9, but
 7.1 Basic Properties of Dividers and Douplers 319

we can demonstrate here that any matched lossless three-port network must be nonrecip-
rocal and, thus, a circulator. The scattering matrix of a matched three-port network has the
following form:


0 S12 S13
[S] = S21 0 S23 . (7.4)
S31 S32 0

If the network is lossless, [S] must be unitary, which implies the following conditions:
∗
S31 S32 = 0, (7.5a)
∗
S21 S23 = 0, (7.5b)
∗
S12 S13 = 0, (7.5c)
|S12 | + |S13 | = 1,
2 2
(7.5d)
|S21 |2 + |S23 |2 = 1, (7.5e)
|S31 | + |S32 | = 1.
2 2
(7.5f)

These equations can be satisfied in one of two ways. Either

S12 = S23 = S31 = 0, |S21 | = |S32 | = |S13 | = 1, (7.6a)

or

S21 = S32 = S13 = 0, |S12 | = |S23 | = |S31 | = 1. (7.6b)

These results shows that Si j = S ji for i = j, which implies that the device must be non-
reciprocal. The scattering matrices for the two solutions of (7.6) are shown in Figure 7.2,
together with the symbols for the two possible types of circulators. The only difference
between the two cases is in the direction of power flow between the ports: solution (7.6a)
corresponds to a circulator that allows power flow only from port 1 to 2, or port 2 to 3, or
port 3 to 1, while solution (7.6b) corresponds to a circulator with the opposite direction of
power flow.
Alternatively, a lossless and reciprocal three-port network can be physically realized
if only two of its ports are matched [1]. If ports 1and 2 are the matched ports, then the
scattering matrix can be written as


0 S12 S13
[S] = S12 0 S23 . (7.7)
S13 S23 S33

2 2
0 0 1 1 0 1 0 1
[S] = 1 0 0 [S] = 0 0 1
0 1 0 1 0 0

3 3
(a) (b)

FIGURE 7.2 Two types of circulators and their scattering matrices. (a) Clockwise circulation.
(b) Counterclockwise circulation. The phase references for the ports are arbitrary.
320 Chapter 7: Power Dividers and Directional Couplers

S21 = e j␪

0 e j␪ 0
[S] = e j␪ 0 0 1 2
S12 = e j␪
0 0 e j␾

S33 = e j␾

FIGURE 7.3 A reciprocal lossless three-port network matched at ports 1 and 2.

To be lossless, the following unitarity conditions must be satisfied:

∗
S13 S23 = 0, (7.8a)
∗ ∗
S12 S13 + S23 S33 = 0, (7.8b)
∗ ∗
S23 S12 + S33 S13 = 0, (7.8c)
|S12 | + |S13 | = 1,
2 2
(7.8d)
|S12 |2 + |S23 |2 = 1, (7.8e)
|S13 | + |S23 | + |S33 | = 1.
2 2 2
(7.8f)

Equations (7.8d) and (7.8e) show that |S13 | = |S23 |, so (7.8a) leads to the result that S13 =
S23 = 0. Then, |S12 | = |S33 | = 1. The scattering matrix and corresponding signal flow
graph for this network are shown in Figure 7.3, where it is seen that the network actu-
ally degenerates into two separate components—one a matched two-port line and the other
a totally mismatched one-port.
Finally, if the three-port network is allowed to be lossy, it can be reciprocal and
matched at all ports; this is the case of the resistive divider, which will be discussed in
Section 7.2. In addition, a lossy three-port network can be made to have isolation between
its output ports (e.g., S23 = S32 = 0).

Four-Port Networks (Directional Couplers)

The scattering matrix of a reciprocal four-port network matched at all ports has the follow-
ing form:
⎡ ⎤
0 S12 S13 S14
⎢S 0 S23 S24 ⎥
[S] = ⎣ 12 ⎦. (7.9)
S13 S23 0 S34
S14 S24 S34 0
If the network is lossless, 10 equations result from the unitarity, or energy conservation,
condition [1, 2]. Consider the multiplication of row 1 and row 2, and the multiplication of
row 4 and row 3:
∗ ∗
S13 S23 + S14 S24 = 0, (7.10a)
∗ ∗
S14 S13 + S24 S23 = 0. (7.10b)
 7.1 Basic Properties of Dividers and Douplers 321

∗ , and (7.10b) by S ∗ , and subtract to obtain

Multiply (7.10a) by S24 13
∗
S14 (|S13 |2 − |S24 |2 ) = 0. (7.11)
Similarly, the multiplication of row 1 and row 3, and the multiplication of row 4 and row
2, gives
∗ ∗
S12 S23 + S14 S34 = 0, (7.12a)
∗ ∗
S14 S12 + S34 S23 = 0. (7.12b)
Multiply (7.12a) by S12 , and (7.12b) by S34 , and subtract to obtain
S23 (|S12 |2 − |S34 |2 ) = 0. (7.13)
One way for (7.11) and (7.13) to be satisfied is if S14 = S23 = 0, which results in a direc-
tional coupler. Then the self-products of the rows of the unitary scattering matrix of (7.9)
yield the following equations:
|S12 |2 + |S13 |2 = 1, (7.14a)
|S12 | + |S24 | = 1,
2 2
(7.14b)
|S13 | + |S34 | = 1,
2 2
(7.14c)
|S24 | + |S34 | = 1,
2 2
(7.14d)
which imply that |S13 | = |S24 | [using (7.14a) and (7.14b)], and that |S12 | = |S34 | [using
(7.14b) and (7.14d)].
Further simplification can be made by choosing the phase references on three of the
four ports. Thus, we choose S12 = S34 = α, S13 = βe jθ , and S24 = βe jφ , where α and β
are real, and θ and φ are phase constants to be determined (one of which we are still free
to choose). The dot product of rows 2 and 3 gives
∗ ∗
S12 S13 + S24 S34 = 0, (7.15)
which yields a relation between the remaining phase constants as
θ + φ = π ± 2nπ. (7.16)
If we ignore integer multiples of 2π, there are two particular choices that commonly occur
in practice:
1. A Symmetric Coupler: θ = φ = π/2. The phases of the terms having amplitude β
are chosen equal. Then the scattering matrix has the following form:
⎡ ⎤
0 α jβ 0
⎢ α 0 0 jβ ⎥
[S] = ⎣ ⎦. (7.17)
jβ 0 0 α
0 jβ α 0
2. An Antisymmetric Coupler: θ = 0, φ = π . The phases of the terms having ampli-
tude β are chosen to be 180◦ apart. Then the scattering matrix has the following
form:
⎡ ⎤
0 α β 0
⎢α 0 0 −β ⎥
[S] = ⎣ ⎦. (7.18)
β 0 0 α
0 −β α 0
322 Chapter 7: Power Dividers and Directional Couplers

Input 1 2 Through

Isolated 4 3 Coupled

Input 1 2 Through

Isolated 4 3 Coupled

FIGURE 7.4 Two commonly used symbols for directional couplers, and power flow conventions.

Note that these two couplers differ only in the choice of reference planes. In addition,
the amplitudes α and β are not independent, as (7.14a) requires that

α 2 + β 2 = 1. (7.19)

Thus, apart from phase references, an ideal four-port directional coupler has only one de-
gree of freedom, leading to two possible configurations.
Another way for (7.11) and (7.13) to be satisfied is if |S13 | = |S24 | and |S12 | = |S34 |.
If we choose phase references, however, such that S13 = S24 = α and S12 = S34 = jβ
[which satisfies (7.16)], then (7.10a) yields α(S23 + S14 ∗ ) = 0, and (7.12a) yields β(S ∗ −
14
S23 ) = 0. These two equations have two possible solutions. First, S14 = S23 = 0, which is
the same as the above solution for the directional coupler. The other solution occurs for
α = β = 0, which implies that S12 = S13 = S24 = S34 = 0. This is the degenerate case of
two decoupled two-port networks (between ports 1 and 4, and ports 2 and 3), which is of
trivial interest and will not be considered further. We are thus left with the conclusion that
any reciprocal, lossless, matched four-port network is a directional coupler.
The basic operation of a directional coupler can be illustrated with the aid of Figure 7.4,
which shows two commonly used symbols for a directional coupler and the port definitions.
Power supplied to port 1 is coupled to port 3 (the coupled port) with the coupling factor
|S13 |2 = β 2 , while the remainder of the input power is delivered to port 2 (the through
port) with the coefficient |S12 |2 = α 2 = 1 − β 2 . In an ideal directional coupler, no power
is delivered to port 4 (the isolated port).
The following quantities are commonly used to characterize a directional coupler:
P1
Coupling = C = 10 log = −20 log β dB, (7.20a)
P3
P3 β
Directivity = D = 10 log = 20 log dB, (7.20b)
P4 |S14 |
P1
Isolation = I = 10 log = −20 log |S14 | dB, (7.20c)
P4
P1
Insertion loss = L = 10 log = −20 log |S12 | dB. (7.20d)
P2
The coupling factor indicates the fraction of the input power that is coupled to the out-
put port. The directivity is a measure of the coupler’s ability to isolate forward and back-
ward waves (or the coupled and uncoupled ports). The isolation is a measure of the power
 7.1 Basic Properties of Dividers and Douplers 323

delivered to the uncoupled port. These quantities are related as

I = D + C dB. (7.21)
The insertion loss accounts for the input power delivered to the through port, diminished by
power delivered to the coupled and isolated ports. The ideal coupler has infinite directivity
and isolation (S14 = 0). Then both α and β can be determined from the coupling factor, C.
Hybrid couplers are special cases√of directional couplers, where the coupling factor is
3 dB, which implies that α = β = 1/ 2. There are two types of hybrids. The quadrature
hybrid has a 90◦ phase shift between ports 2 and 3 (θ = φ = π/2) when fed at port 1, and
is an example of a symmetric coupler. Its scattering matrix has the following form:
⎡ ⎤
0 1 j 0
1 ⎢1 0 0 j ⎥
[S] = √ ⎣ ⎦. (7.22)
2 j 0 0 1
0 j 1 0
The magic-T hybrid and the rat-race hybrid have a 180◦ phase difference between ports
2 and 3 when fed at port 4, and are examples of an antisymmetric coupler. Its scattering
matrix has the following form:
⎡ ⎤
0 1 1 0
1 ⎢ 1 0 0 −1 ⎥
[S] = √ ⎣ ⎦. (7.23)
2 1 0 0 1
0 −1 1 0

POINT OF INTEREST: Measuring Coupler Directivity

The directivity of a directional coupler is a measure of the coupler’s ability to separate forward
and reverse wave components, and applications of directional couplers often require high (35 dB
or greater) directivity. Poor directivity will limit the accuracy of a reflectometer, and can cause
variations in the coupled power level from a coupler when there is even a small mismatch on
the through line.
The directivity of a coupler generally cannot be measured directly because it involves a
low-level signal that can be masked by coupled power from a reflected wave on the through
arm. For example, if a coupler has C = 20 dB and D = 35 dB, with a load having a return loss
RL = 30 dB, the signal level through the directivity path will be D + C = 55 dB below the
input power, but the reflected power through the coupled arm will only be RL + C = 50 dB
below the input power.
One way to measure coupler directivity uses a sliding matched load, as follows. First, the
coupler is connected to a source and a matched load, as shown in the accompanying left-hand
figure, and the coupled output power is measured. If we assume an input power Pi , this power
will be Pc = C 2 Pi , where C = 10(−CdB)/20 is the numerical voltage coupling factor of the
coupler. Next, the position of the coupler is reversed, and the through line is terminated with a
sliding load, as shown in the right-hand figure.

Pc V0 (Pmax, Pmin )
Sliding
C Load load
C
D C

Vi , Pi Vi , Pi
Γ

Changing the position of the sliding load introduces a variable phase shift in the signal re-
flected from the load and coupled to the output port. The voltage at the output port can be
324 Chapter 7: Power Dividers and Directional Couplers

written as

C
V0 = Vi + C||e− jθ ,
D

where Vi is the input voltage, D = 10(D dB)/20 ≥ 1 is the numerical value of the directivity, ||
is the reflection coefficient magnitude of the load, and θ is the path length difference between
the directivity and reflected signals. Moving the sliding load changes θ, so the two signals will
combine to trace out a circular locus, as shown in the following figure.

Im V0

V0 CΓVi
␪ Vmax
0
Vmin C ReV0
V
D i

The minimum and maximum output powers are given by

2
2
C C
Pmin = Pi − C|| , Pmax = Pi + C|| .
D D
Let M and m be defined in terms of these powers as follows:

2

Pc D Pmax 1 + ||D 2
M= = , m= = .
Pmax 1 + ||D Pmin 1 − ||D
These ratios can be accurately measured directly by using a variable attenuator between the
source and coupler. The coupler directivity (numerical) can then be found as

2m
D=M .
m+1
This method requires that || < 1/D or, in dB, RL > D.

Reference: M. Sucher and J. Fox, eds., Handbook of Microwave Measurements, 3rd edition, Volume II, Polytech-
nic Press, New York, 1963.

7.2 THE T-JUNCTION POWER DIVIDER

The T-junction power divider is a simple three-port network that can be used for power
division or power combining, and it can be implemented in virtually any type of transmis-
sion line medium. Figure 7.5 shows some commonly used T-junctions in waveguide and
microstrip line or stripline form. The junctions shown here are, in the absence of transmis-
sion line loss, lossless junctions. Thus, as discussed in the preceding section, such junctions
cannot be matched simultaneously at all ports. We will analyze the T-junction divider be-
low, followed by a discussion of the resistive power divider, which can be matched at all
ports but is not lossless.

Lossless Divider
The lossless T-junction dividers of Figure 7.5 can all be modeled as a junction of three
transmission lines, as shown in Figure 7.6 [3]. In general, there may be fringing fields and