Impedance Matching

Impedance Matching
1 Introduction
Impedance matching is the process to match the
load ZL to a transmission line by a matching
network, as depicted in Fig. 1. Recall that the
reflections are eliminated under the matched
Fig. 1: Impedance matching
condition. Impedance matching is important for
the following reasons:
• To achieve maximum power transfer and minimize power loss.
• To improve signal-to-noise ratio.
• To reduce amplitude and phase errors for power distribution networks, e.g., antenna arrays.
There are many choices regarding matching network design, but the following factors must be
considered in the selection of the network:
• Complexity
• Bandwidth
• Implementation
• Adjustibility

2 Matching with Lumped Elements (L Networks)

The L-section is considered the simplest type of matching
network. There are two possible configurations, as depicted
in Fig. 2. (a) is the network for Re[ZL] > Z0, while (b) is the
network for Re[ZL] < Z0. Note that in both configurations,
two components (jX, jB) are required in order to have
degree of freedom 2, since the load impedance is generally
complex. (a)
Consider Fig. 2(a). Let ZL = RL+jXL, then the impedance
seen looking into the matching network followed by the
load impedance must be equal to Z0, i.e.,
1
Z 0 = jX + .
jB + 1 /( RL + jX L )
Rearranging and separating into real and imaginary parts
yield (b)
B( XRL − X L Z 0 ) = RL − Z 0 ; X (1 − BX L ) = BZ 0 RL − X L Fig. 2: L-section matching networks.
Solving the above equations yields
X L ± RL / Z 0 RL2 + X L2 − Z 0 RL
B= .
RL2 + X L2
Note that the argument inside the second square root is always positive since RL > Z0. The series
reactance can be found as
1 X L Z0 Z
X= + − 0 .
B RL BRL
Note also that two solutions are generally possible. One must consider the above factors in deciding
which L network to use.
Likewise, for the network in Fig. 2(b), the matched condition is given by
1 1
= jB + .
Z0 RL + j ( X + X L )
Rearranging and separating into real and imaginary parts yield

1
Impedance Matching

BZ 0 ( X + X L ) = Z 0 − RL ; X + X L = BZ 0 RL .
Solving for X and B gives
± ( Z 0 − RL ) / RL
X = ± RL ( Z 0 − RL ) − X L ; B = .
Z0
Note that RL < Z0 in this case, so the argument of the square root is always positive.
Example 1 Design an L-section matching network to match a series RC load with an impedance ZL =
200 – j100 Ω, to a 100 Ω line, at a frequency of 500 MHz.

L section matching : reflection coefficient plot

1
solution 1
0.9 solution 2

0.8

0.7

0.6
|Γ|

0.5

0.4

0.3

0.2

0.1

0
0 1 2 3 4 5 6 7 8 9 10
Frequency [Hz] 8
x 10

3 Single-Stub Tuning
The impedance matching using L-sections discussed previously requires lumped elements that might
not be available, thus it is not practical in some cases. The single-stub tuning is the matching
technique that uses a single open-circuited or short-circuited length of transmission-line (a “stub”),
connected either in parallel or in series with the transmission feed line at a certain distance from the
load. Note that there are two design parameters, namely the length of the stub and the distance from
the load, which contribute degree of freedom 2, as in the matching with L-sections.
The choice of open-circuited stub or short-circuited stub depends on the type of transmission line
media. For microstrip lines, open stubs are preferred due to ease of fabrication, while for coaxial lines

2
Impedance Matching

or waveguides, short stubs are more desirable since such open-circuited stubs tend to radiate, resulting
in reactance changes.
3.1 Shunt Stubs
The single-stub shunt tuning circuit configuration is shown in
Fig. 3. Refer to the figure, to match the impedance, it is required
that
Y0 = Yin = Y1 + Ystub .
Since Ystub is purely susceptance (i.e., zero conductance), the real
part of Y1 must be equal to Z0. Furthermore, the susceptance of
Y1 must cancel out the susceptance of Ystub, resulting in Yin
becomes Y0. Using the Smith chart makes the design process
easier. The first step is to find the distance such that the
normalized admittance is on the 1+jb circle. Then find the Fig. 3: Single-stub shunt tuning
length such that the stub has susceptance –jb.
Example 2 For a load impedance ZL = 60 – j80 Ω, design two single-stub (short circuit) shunt tuning
networks to match this load to a 50 Ω line. Assuming that the load is matched at 2 GHz and the load
consists of a resistor and a capacitor in series.

0.9

0.8

0.7 solution #1
solution #2
0.6
|Γ|

0.5

0.4

0.3

0.2

0.1

0
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
Frequency [GHz]

3.2 Series Stubs

The single-stub series tuning circuit configuration is
shown in Fig. 4. Refer to the figure, to match the
impedance, it is required that
Z 0 = Z in = Z1 + Z stub .
Since Zstub is purely reactance (i.e., zero resistance), the
real part of Z1 must be equal to Z0. Furthermore, the
reactance of Z1 must cancel out the reactance of Zstub,
resulting in Zin becomes Z0. As in the shunt tuning
circuit design, using the Smith chart makes the design Fig. 4: Single-stub series tuning
process easier. The first step is to find the distance such

3
Impedance Matching

that the normalized impedance is on the 1+jx circle. Then find the length such that the stub has
reactance –jx.
Example 3 For a load impedance ZL = 100 + j80 Ω, design two single-stub (open circuit) series tuning
networks to match this load to a 50 Ω line. Assuming that the load is matched at 2 GHz and the load
consists of a resistor and an inductor in series.

0.9

0.8

0.7

0.6
|Γ|

0.5

0.4

0.3

0.2

solution #1
0.1
solution #2
0
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
Frequency [GHz]

4 Double-Stub Tuning
The single-stub tuner requires a variable length of line between the load and the stub, thus it is
difficult to make it “adjustable”. The double-stub tuning shown in Fig. 5 uses 2 adjustable shunt stubs
in fixed positions. However, the double-stub tuner cannot match all load impedances.
d d

Y0 jB2 Y0 jB1 Y0 Y’L Y0 jB2 Y0 jB1 YL

Open or Open or l1 Open or Open or l1

shorted
l2 shorted shorted
l2 shorted
stub stub stub stub

(a) (b)
Fig. 5: Double-stub tuning (a) Original circuit with the load an arbitrary distance from the first stub
(b) Equivalent circuit with the load transformed to the first stub.
The Smith chart solution can be illustrated in Fig. 6. First, locate yL and draw the rotated
1+jb circle with respect to the stub spacing d. Then move the load admittance onto the
rotated 1+jb circle (points y1, y’1) using the susceptance b1, b’1 of the stub. Next, move the
points y1, y’1 onto the 1+jb circle (points y2, y’2). Finally, add the susceptance b2, b’2 to
match the load impedance. Note that there are two possible solutions as in the case of
single-stub tuning.

4
Impedance Matching

Notice that if yL is inside the shaded

region in the figure, specified by g0+jb
circle, it is impossible to move this
admittance onto the rotated circle,
which means that it cannot be matched
by a double-stub tuner (i.e., there is no
solution). Therefore, this shaded region
forms a forbidden range of load
admittances that cannot be matched by
this double-stub tuner. Reducing the
space d can lead to the reduction in the
size of this forbidden range, however, d
must be kept sufficiently large for
fabricating two separate stubs. In
addition, spacings near 0 or λ/2 load to
matching networks that are very
frequency sensitive. In practice, stub
spacings are usually chosen as λ/8 or
3λ/8. Furthermore, if the length of line Fig. 6: Smith chart diagram for the operation of a double-
between the load and the first stub can stub tuner.
be adjusted, then yL can always be
moved out of the forbidden region.
Example 4 For a load impedance ZL = 60 - j80 Ω, design a shunt double-stub tuner to match
this load to a 50 Ω line. The stubs are to be open-circuited and are spaced λ/8 apart. Also,
the load is assumed to consist of a 60Ω-resistor and a 0.995pF-capacitor.

l1=0.4819, l2=0.3498, l1p=0.1465, l2p=0.2042

−

− −

− − 1

0.9
0.006+0.008i
0.8

0.7
− − −
ψ 0.6
ψ −
|Γ|

0.5

0.4

0.3
− − − −
0.2
Solution #1
0.1
Solution #2
− −
− − 0
1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3
−
− Frequency [Hz] 9
x 10

5 Quarter-Wave Transformer
Recall that, for a quarter-wavelength transmission line (l = λ/4), the input impedance becomes
Z in = Z 02 / Z L or Z 02 = Z in Z L .
Therefore, a quarter-wavelength transmission line can be used to convert a resistive load to match a
transmission line by choosing the proper characteristic impedance of the quarter-wavelength line. This
is called a quarter-wave transformer. The general configuration of this quarter-wave transformer is
shown in Fig. 6, where

5
Impedance Matching

Z12 = Z 0 RL .
To match an arbitrary ZL using the quarter-
wave transformer, one must somehow modify
the load such that it becomes purely resistive.
This may be done by adding certain lumped Fig. 6
elements, transmission line of certain length, tuning circuits or stubs.
Example 5 Repeat example 3 by using the quarter-wave transformer.

6
Impedance Matching

6 The Theory of Small Reflections

Quarter-wave transformers provide a simple mean of impedance matching, but cannot achieve broad
bandwidth. To obtain more bandwidth, multisection transformers can be used.
Single-section Transformer
Consider the single-section transformer shown
in Fig. 7, the partial reflection and transmission
coefficients are given by
Z 2 − Z1 Z − Z2
Γ1 = ; Γ2 = −Γ1 ; Γ3 = L ;
Z 2 + Z1 ZL + Z2
2Z 2 2 Z1
T21 = 1 + Γ1 = ; T12 = 1 + Γ2 = .
Z 2 + Z1 Z 2 + Z1
The total reflection can then be given in terms
of an infinite sum of partial reflections and
transmissions as follows:
Γ = Γ1 + T12T21Γ3e − j 2θ + T12T21Γ32Γ2e − j 4θ + L
∞
= Γ1 + T12T21Γ3e − j 2θ ∑ Γ2n Γ3n e − j 2 nθ
n=0 Fig. 7
Using the geometric series
∞
1
∑x
n =0
n
=
1− x
, for x < 1 ,

Γ can be rewritten as
T12T21Γ3e − j 2θ
Γ = Γ1 + .
1 − Γ2 Γ3e − j 2θ
Using Γ2=-Γ1, Τ21=1+Γ1, Τ12=1−Γ1 yields
Γ1 + Γ3e − j 2θ
Γ= .
1 + Γ1Γ3e − j 2θ
If the discontinuities between the impedances Z1, Z2 and Z2, ZL are small, then |Γ1Γ3|<<1, and
Γ ≅ Γ1 + Γ3e − j 2θ .
Multisection Transformer
Now consider the multisection transformer shown in Fig. 8. This transformer consists of N equal-
length (commensurate) sections of transmission lines. Partial reflection coefficients can be defined at
each junction as
Z1 − Z 0 Z − Zn Z − ZN
Γ0 = ; Γn = n +1 ; ΓN = L .
Z1 + Z 0 Z n +1 + Z n ZL + ZN

Fig. 8
We also assume that all Zn increase or decrease monotonically across the transformer, and ZL is real.
This implies that Γn will be real and of the same sign. Then the total reflection coefficient Γ can be
approximated as

7
Impedance Matching

Γ (θ ) = Γ0 + Γ1e − j 2θ + Γ2 e − j 4θ + L + ΓN e − j 2 Nθ .
Furthermore, assume that the transformer can be made symmetric, so that Γ0=ΓN, Γ1=ΓN-1, etc. (Note
that this does not imply that the Zn’s are symmetrical.) Then,
Γ (θ ) = e − jNθ {Γ0 (e − jNθ + e jNθ ) + Γ1 (e − j ( N − 2 )θ + e j ( N − 2 )θ ) + L}.
It follows that for N even,
 1 
Γ (θ ) = 2e − jNθ Γ0 cos Nθ + Γ1 cos( N − 2)θ + L + Γn cos( N − 2 n)θ + L + ΓN / 2  ,
 2 
and for N odd,
Γ (θ ) = 2e − jNθ {Γ0 cos Nθ + Γ1 cos( N − 2)θ + L + Γn cos( N − 2 n )θ + L + Γ( N −1) / 2 cos θ }.
From these results, one can notice that any desired reflection response (as a function of θ) can be
realized by choosing the proper Γn’s and using enough sections. Recall the fact that a smooth function
can be approximated by a Fourier series, if enough terms are used.
7 Binomial Multisection Matching Transformers
The passband response of a binomial transformer is optimum in the sense that, for a given number of
sections, the response is flat as possible near the design frequency. Thus, such as response is also
known as maximally flat. This type of response is designed, for an N-section transformer, by setting
the first N-1 derivatives of |Γ(θ)| to zero, at the center frequency f0. Such a response can be obtained if
Γ(θ ) = A(1 + e − j 2θ ) N .
Then the magnitude |Γ(θ)| is
N
Γ (θ ) = A e − jθ e jθ + e − jθ
N
= 2 N A cosθ .
Note that |Γ(θ)|=0 for θ=π/2 and that (dn|Γ(θ)|)/dθn = 0 at θ=π/2 for n = 1, 2, …, N-1. (θ=π/2
corresponds to the center frequency f0, for which l=λ/4 and θ = β l = π /2.)
Let f→0, then θ = β l = 0, and
Z L − Z0
Γ(θ = 0) = A(1 + 1) N = A2 N = ,
Z L + Z0
since for f = 0 all sections are of zero electrical length. Thus,
Z L − Z0
A = 2−N .
Z L + Z0
Now expanding Γ(θ) according to the binomial expansion yields
N
N!
Γ(θ ) = A(1 + e − j 2θ ) N = A∑ CnN e − j 2 nθ ,where CnN = . Since,
n =0 ( N − n)!n!
N
Γ (θ ) = A∑ C nN e − j 2 nθ = Γ0 + Γ1e − j 2θ + Γ2 e − j 4θ + L + ΓN e − j 2 Nθ , Γn = AC nN .
n=0

If Γn’s are assumed to be small, the following approximation can be applied:

Z n+1 − Z n 1 Z n+1 x −1
Γn = ≅ ln , since ln x ≅ 2 . Therefore,
Z n+1 + Z n 2 Z n x +1
Z n+1 Z − Z0 N Z
ln ≅ 2Γn = 2 AC nN = 2( 2 − N ) L Cn ≅ 2 − N CnN ln L .
Zn Z L + Z0 Z0
To calculate the bandwidth, let Γm denote the maximum value of reflection coefficient that can be
tolerated over the passband. Then,

8
Impedance Matching

Γm = 2 N A cos N θ m , where θm < π /2 is the lower edge of the passband. Thus,

 1  Γ 1 / N 
θ m = cos   m   , and the fractional bandwidth is given by
−1

2  A  
 
∆f 2 ( f 0 − f m ) 4θ m   1 / N 
4 −1 1 Γm
= = 2− = 2 − cos     .
f f0 π π 2  A  
 
Example 6 Design a three-section transformer to match a 50 Ω load to a 100 Ω line, and calculate the
bandwidth for Γm = 0.05.

8 Chebyshev Multisection Matching Transformers

In contrast with the binomial matching transformer, the Chebyshev transformer optimizes bandwidth
at the expense of passband ripple. The Chebyshev transformer is designed by equating Γ(θ) to a
Chebyshev polynomial, which has the optimum characteristics needed for this type of transformer.
Chebyshev Polynomial
The nth order Chebyshev polynomial is a polynomial of degree n, and is denoted by Tn(x). The first
four Chebyshev polynomials are
T1 ( x ) = x; T2 ( x ) = 2 x 2 − 1; T3 ( x ) = 4 x 3 − 3 x; T4 ( x ) = 8 x 4 − 8 x 2 + 1 .
6
Higher-order polynomials can be found using
the following recurrence formula: n=4
n=3
Tn ( x) = 2 xTn−1 ( x) − Tn−2 ( x) . 4

n=2
Some important properties of Chebyshev
2
polynomials are listed here:
n=1
1. For -1≤x≤1, |Tn(x)|≤1. In this range, the
Tn(x)

0
Chebyshev polynomials oscillate between ±1.
This is the equal ripple property, and this
-2
region will be mapped to the passband of the
matching transformer.
2. For |x|>1, |Tn(x)|>1. This region will be -4

mapped to the frequency range outside the

passband. -6
-1.5 -1 -0.5 0 0.5 1 1.5
3. For |x|>1, |Tn(x)| increases faster with x as n x

increases. Fig. 8: First four Chebyshev polynomials

Now, let x = cos θ for |x| < 1. Then it can be shown that the Chebyshev polynomials can be expressed
as
Tn (cos θ ) = cos nθ , or more generally as
cos(n cos −1 x) for | x |< 1
Tn ( x) = 
cosh(n cosh −1 x) for | x |> 1

9
Impedance Matching

Since equal ripple is desirable in the passband, it is necessary to map θm to x = 1 and π−θ m to x = -1,
where θm and π−θ m are the lower and upper edges of the passband. This can be accomplished by
replacing cos θ in the above equation with cos θ /cos θ m:
 cos θ    cos θ 
Tn   = Tn (secθ m cos θ ) = cos n cos −1   .
 cos θ m    cos θ m 
Then |sec θ m cos θ | ≤ 1 for θ m < θ < π−θ m, so |Tn(sec θ m cos θ )| ≤ 1 over this same range.
It follows that the first four terms of the Chebyshev polynomials can be written as
T1 (sec θ m cos θ ) = sec θ m cos θ ; T2 (sec θ m cos θ ) = sec 2 θ m (cos 2θ + 1) − 1 ;
T3 (sec θ m cos θ ) = sec3 θ m (cos 3θ + 3 cos θ ) − 3 sec θ m cos θ ;
T4 (sec θ m cos θ ) = sec 4 θ m (cos 4θ + 4 cos 2θ + 3) − 4 sec 2 θ m (cos 2θ + 1) + 1 .
The above results can be used to design matching transformers with up to four sections.
Design of Chebyshev Transformers
A Chebyshev equal-ripple passband can be synthesized by making Γ(θ) proportional to TN(secθm
cosθ), where N denotes the number of sections. Thus,
Γ (θ ) = 2e − jN θ {Γ0 cos N θ + Γ1 cos( N − 2)θ + L + Γn cos( N − 2 n )θ + L} = Ae − jN θ TN (sec θ m cos θ )
where the last term in the series is (1/2)ΓΝ/2 for N even and Γ(N-1)/2cosθ for N odd. The constant A can
be found from letting θ = 0:
Z L − Z0 Z − Z0 1
Γ(θ = 0) = = ATN (secθ m ) , or A = L .
Z L + Z0 Z L + Z 0 TN (secθ m )
Now if the maximum allowable reflection coefficient magnitude in the passband is Γm (i.e., the ripple),
then Γm = |A|, since the maximum value of Tn(secθmcosθ) in the passband is unity. Using the
approximation introduced in the previous section yields
1 Z L − Z0 1 Z
TN (secθ m ) = ≅ ln L . It follows that
Γm Z L + Z 0 2Γm Z 0
 1  1 Z L − Z 0    
secθ m = cosh  cosh −1   ≅ cosh 1 cosh −1  ln(Z L / Z 0 )  .
  
 N  Γm Z L + Z 0   N  2Γm 
Once θm is known, the fractional bandwidth can be calculated from
∆f 4θ
= 2− m .
f π
Each Γn can be determined by expanding TN(secθmcosθ) and equating similar terms of the form cos(N-
2n)θ. The following approximation can be applied to improve the accuracy:
Z − Z n 1 Z n+1
Γn = n+1 ≅ ln .
Z n+1 + Z n 2 Z n
Example 7 Design a three-section Chebyshev transformer to match a 100 Ω load to a 50 Ω line, with
Γm = 0.05.

10