You are on page 1of 14

As long as the load impedance Z L has a non-zero real

part, a matching network can always be found. Typically, many


choices are available, and we will examine some of the most
common and practical ones here. Several factors may influence
the choice of a matching network, including the following:

Complexity As with most engineering designs, the


simplest one that meet the design criteria is preferable.
A simpler design is usually less costly, more reliable,
and less noisy than a more complex design.

Bandwidth Any type of matching network can give a


perfect match (zero reflection) at a single frequency.
In many applications, it is desirable to match the load
over a range of frequencies. This is usually possible,
at the cost of increasing complexity of the design.

Implementation Depending on the type of


transmission line or waveguide being used, one type of
matching network may be preferable compared to
another. For example, adjustable short-circuit stubs
are usually easier to implement than are multi-section
quarter-wavelength transformers.

Adjustability In some applications, it is important to


be able to adjust the matching network to match a
variable load impedance. Some matching networks are
more easily adjusted than others.

Chapter 5: Impedance Matching & Tuning


In this chapter we turn to the application of the theory and
techniques assimilated in the previous chapters to the first of
several important and practical topics in microwave engineering:
that of impedance matching. Impedance matching is of vital
importance in design practical RF systems. The basic matching
configuration to be dealt with is illustrated below.

A load of impedance Z L is to be connected to a microwave


network via a transmission line of characteristic impedance Z 0 ; a
matching network is inserted between the transmission line and
the load. In the ideal case, the matching network is lossless, and
is carefully designed so that the impedance looking into its input
terminals is Z 0 . Then reflections are eliminated on the
transmission line (although there are usually reflections between
the matching network and the load). The design of a suitable
matching network is often called tuning, and the matching
network called a tuning network.
Impedance matching is important for a number of reasons,
among them:

Maximum power is delivered to the load when the load


is matched to the line (assuming a matched generator),
and power loss in the feed line is minimized.

Impedance matching of sensitive receiver components


(antennas, low-noise amplifiers, etc.) improves the
signal-to-noise ratio of the system.

Impedance matching of a power distribution network


(e.g., an antenna array feed network) will reduce
amplitude and phase errors.

5.1

Matching with Lumped Elements (L Networks)

The simplest matching network consists of a pair of


reactive elements (inductors and/or capacitors) in an L shaped
configuration. Such matching networks may be used to match and
arbitrary load impedance to a transmission line. The two possible
configurations are depicted below; they differ by the relative
placement of the series and shunt elements.
1

Separate into real and imaginary parts to obtain the pair of


equations:
B ( XRL ! X L Z 0 ) = RL ! Z 0

X (1 ! BX L ) = BZ 0 RL ! X L

If the normalized impedance zL = Z L Z 0 is inside of the unit


resistance circle zL = 1 + jx on the Smith chart, circuit (a) should
be used; otherwise, the circuit (b) should be used. The reason will
become apparent when we examine the Smith chart method.
In both cases, the two reactive elements of the matching
network are either inductors or capacitors, so that there are eight
such configurations in all. The use of L matching networks is
limited to realizable values of capacitance and inductance that
these discrete circuit elements must have for the design at hand.
These matching networks are typically limited to frequencies
below 1 GHz due to practical device limitations.
We first present analytical expressions for the values of the
matching network elements, and then discuss the use of the Smith
chart for finding appropriate element values.

Solving the first for B and substitution into the second results in a
quadratic equation for B :

X L RL Z 0 RL2 + X L2 ! Z 0 RL
B=
RL2 + X L2
Then the series reactance can be found as:

X=

Note that this indicates that there are two solutions for B and
hence for X . One may be preferable to the other in terms of
component values, bandwidth, or SWR, as we will see.
Now consider the second circuit, used in the case of a
normalized load outside of the unit resistance circle, or RL < Z 0 .
Following a similar method to that above, we obtain the solutions:

Analytic Solutions
Although the Smith chart method of design is usually
easier, the analytical solution may used in computer-aided
solutions, and provide a greater degree of accuracy. Assume that
Z L = RL + jX L .
First consider the case (a) above, which means that zL lies
inside of the unit-resistance circle of the Smith chart; that is,
RL > Z 0 . The impedance looking into the matching network must
be equal to Z 0 for a match:

Z 0 = jX +

1 XL Z0
Z
+
! 0
B
RL
BRL

X = RL ( Z 0 ! RL ) ! X L
B=

( Z 0 ! RL )

RL

Z0

Once again, we see that two solutions are possible.

1
jB + 1 ( RL + jX L )
2

Smith Chart Solutions


The Smith chart may be used to quickly obtain the
solutions for the L matching network. In particular, the
Impedance and Admittance Smith chart, or YZ Chart, is most
suitable for solving for the L matching network. The technique is
best presented by means of an example. See the example
provided, as well as that in the Text utilizing the usual Z Smith
chart. In the text, both solutions are computed, and compared in
terms of bandwidth. The reflection coefficient as a function of
frequency for both solutions is presented below, and it is apparent
that there is a significant difference in terms of bandwidth.

5.2

Single-Stub Tuning

This method relies on a matching circuit that consists of a


series section of transmission of particular length, followed by a
stub transmission line segment connected in either series or
parallel. The stub may be either open- or short-circuited at its
termination, as shown in the figure below.

The text also provides a brief discussion of lumped


reactive elements for RF circuits (RF capacitors and inductors)
and provides the figure below:

Shunt Stubs
The most common configuration is the parallel, or shunt,
stub connection of figure (a) above, due to its relative ease of
fabrication.
The principle of operation of single stub tuners is as
follows. The length of the series segment of transmission line is
chosen to move the location of the load impedance along a circle
of constant SWR (constant radius about the center of the chart)
until the circle intersects either the unit conductance circle (for the
shunt stub (a)) or the unit resistance circle (for the series stub (b)).
At this point, the impedance has the form 1 + jx for some value of
x . Then the necessary length of the stub is determined to cancel
the reactive value x , leaving an input impedance of 1 normalized,
or a match to the transmission line.
As with L networks, the technique is best presented by use
of examples, which are provided in the text and on Blackboard.
The text example presents two solutions, which are once again
compared in terms of bandwidth. The reflection coefficient as a
function of frequency is plotted below.

Analytical formulas for the series line length d and stub


length ! are found in the text; because of their cumbersome
nature they are not reproduced here.
Series Stubs
The stub line section may also be connected in series with
other line, as depicted in figure (b) above. The principles of
operation are the same as with the shunt stub designs, and are best
illustrated by example, as in the text. The results of two possible
designs are illustrated below.

5.3

Double-Stub Tuning

The single stub tuner is able to match any load impedance


with non-zero real part to a transmission line, but suffers the
drawback of requiring a variable length section of line in a series
4

connection between the load and the stub section. For a fixed
design this may not be a problem, but for an adjustable tuner,
creating such a variable length line might prove problematic. This
difficulty is overcome by use of the double stub tuner. Such a
tuner, depicted below, makes use of two open- or short-circuited
stubs separated by a fixed distance.

between the two stubs. What this means for the Smith chart
solution technique is that we will find it necessary to sketch the
rotated unit resistance circle on the chart. Now, the length of the
first stub will be used to place the impedance of the load plus that
stub on this rotated unit resistance circle. The figure below shows
the original and the rotated unit resistance circles for the text
example.

Smith Chart Solution

The Smith chart solution technique for design of a doublestub tuner is best illustrated by means of an example. Examples
are provided in the text and on Blackboard. Once again, two
solutions are possible, and the text example leads to the designs
illustrated below.

The principle of operation of the double-stub tuner is as


follows. As with the single stub tuner, we wish the impedance at
the second stub (counting from the load) to reside on the unit
resistance circle; i.e., z = 1 + jx for some reactance x . Then the
length of the second stub may be chosen to cancel this reactance
x , leaving a perfect match looking into the entire stub tuner. But
if the impedance at the second stub is on the unit resistance circle,
then the impedance looking into the first stub must lie on the unit
resistance circle rotated toward the load by the fixed distance d
5

load. However, these techniques themselves yield a frequency


dependence that often has the effect of reducing the bandwidth of
the design.
The single-section quarter-wave transformer of Chapter 2
is depicted below.

The length ! is chosen as a quarter of a wavelength at the design


frequency, and the impedance of the quarter-wave section is
chosen as:
Analytic Solution

Z1 = Z 0 Z L

The analytic solution for the double-stub tuner is given in


the text. It is somewhat unwieldy, but suitable for programming.

5.4

Note that at frequencies other than the design frequency, the


length of the line section is no longer a quarter of a wavelength, so
a perfect match is no longer obtained. We seek an expression for
the mismatch versus frequency below.
As usual, the input impedance looking into the matching
section is given by:

Quarter-Wave Transformer

In Chapter 2 we introduced the quarter-wave transformer


and analyzed it from an impedance viewpoint and a multiple
reflection viewpoint. Here we will approach the quarter-wave
transformer in a more general setting and show that it may be
extended to multisection designs. Such multisection designs may
be synthesized to yield optimum behavior over a desired
frequency band.
One drawback to the quarter-wave transformer is that it
only match real load impedances. A complex load impedance
may be converted to a real value by insertion of a series section of
line of appropriate length or by use of a series or shunt stub at the

Zin = Z1

Z L + jZ1 tan ! !
Z + jZ1t
= Z1 L
Z1 + jZ L tan ! !
Z1 + jZ L t

Here we have used the abbreviation t = tan ! ! = tan " , where


! = " ! = # 2 at the design frequency f0 . Then the reflection
coefficient is given by:

(
(

2
Zin " Z 0 Z1 ( Z L " Z 0 ) + jt Z1 " Z 0 Z L
!=
=
Zin + Z 0 Z1 ( Z L + Z 0 ) + jt Z12 + Z 0 Z L

)
)

Now, Z12 = Z L Z 0 , so this expression reduces to:

!=

Z L " Z0
Z L + Z 0 + j2t Z L Z 0

We are most interested in the magnitude of this reflection


coefficient, which is shown in the text to be:
! =

{1 + #$ 4Z Z
L

If we set a maximum value, ! m , of reflection coefficient


that is allowable, then we can define a bandwidth of the matching
transformer as:

( Z L " Z 0 )2 %& sec2 ' }

12

%#
(
!" = 2 ' $ " m *
&2
)

Use has been made of the identity 1 + tan 2 ! = sec 2 ! .


Now, for operation near the design frequency of f0 , we

This is due to the fact that the curve is symmetric about ! 2 , and
! = ! m at ! = ! m and also at ! = " # ! m . We now equate ! m to
the exact expression for ! and solve for ! m to obtain:

have ! " !0 4 and ! ! " 2 , so that sec 2 ! ! 1 . In this case the


formula for ! simplifies to:
! !

Z L " Z0
2 Z L Z0

cos#

for # near $ 2

cos! m =

This formula for the approximate mismatch of the quarter-wave


transformer near its design frequency is plotted below.

"m

2 Z L Z0
1 # " 2m Z L # Z 0

For TEM lines, this may be further simplified, since


! = " ! = # f 2 f0 . Then we obtain:

fm = 2! m f0 "
We are now in position to find the fractional bandwidth as:

% $
2 Z L Z0
!f
4
m
= 2 " cos "1 '
f0
#
'& 1 " $ 2m Z L " Z 0

We first consider the single section transformer shown


below. We will find an approximate expression for the overall
reflection coefficient ! .

(
*
*)

This quantity is usually expressed as a percentage. Note that it


increases as Z L gets closer to Z 0 .
While the above expression is correct for TEM lines, it
also yields a useful approximation to the behavior of non-TEM
lines and guides. The plot below shows the behavior of the
reflection coefficient versus normalized frequency for various
load/line mismatches.

The partial reflection and transmission coefficients are given by:

!1 =

Z 2 " Z1
Z 2 + Z1

! 2 = "!1
!3 =

5.5

The Theory of Small Reflections

2Z 2
Z 2 + Z1
2Z1
T12 = 1 + ! 2 =
Z 2 + Z1
T21 = 1 + !1 =

Z L " Z2
Z L + Z2

We will compute the total reflection coefficient ! seen by the


feed line by use of the multiple reflection method. The total
reflection is the infinite sum of partial reflection and transmission
terms:

In order to generalize our results for the quarter-wave


transformer to the case of multisection transformers, we need to
develop some approximate relations for the total reflection caused
by partial reflections from several small discontinuities. This
topic goes by the name of the theory of small reflections.

! = !1 + T12T21! 3e" j 2# + T12T21! 23 ! 2 e" j 4# +


$

= !1 + T12T21! 3e" j 2# % ! n2 ! n3 e" j 2#

Single-Section Transformer

n=0

We recall the sum of the geometric series:

Next we consider a multisection transformer as shown


below.

"x
n=0

1
for x < 1
1# x

Thus we can find the sum above in closed form:

T12T21! 3e" j 2#
! = !1 +
1 " ! 2 ! 3 e " j 2#

It consists of N equal-length (called commensurate) sections of


transmission lines. Our goal is to find an approximate expression
for the total reflection coefficient ! .
To that end, define the reflection coefficients at each
junction:

Simplify this expression using the definitions to obtain:

!=+

! 1 + ! 3 e " j 2#
1 + !1! 3e" j 2#

!0 =

Now if the discontinuities between the impedances Z1 , Z 2 and


Z 2 , Z L are small, then !1! 3 ! 1 and we may approximate this
result as:

Z1 " Z 0
Z1 + Z 0

!n =

Z n +1 " Z n
Z n +1 + Z n

!N =

ZL " ZN
ZL + ZN

We will assume that all Z n increase or decrease monotonically


across the transformer and that Z L is real. This implies that all
! n are real and of the same sign ( ! n > 0 if Z L > Z 0 and ! n < 0 if
Z L < Z 0 ). Using the result of the previous section, the overall
reflection coefficient is approximated by:

! ! ! 1 + ! 3 e " j 2#
In words, this equation states the intuitive fact that the total
reflection coefficient is dominated by the reflection from the
initial discontinuity between Z1 and Z 2 (that is, by !1 ), and the
first reflection from the discontinuity between Z 2 and Z L is the

! (" ) ! ! 0 + !1e# j 2" + ! 2 e# j 4" + " + ! N e# j 2 N"

Make a further assumption that the transformer is symmetrical, so


that ! 0 = ! N , !1 = ! N "1 , ! 2 = ! N " 2 , and so on (this does not
imply that the Z ns are symmetrical). This leads to the expression:

first correction factor (in the form of ! 3e" j 2# ). Of course, the


phase factor e! j 2" accounts for propagation of the incident wave
up and down the line. This approximation is surprisingly accurate
for small discontinuities.

! (" ) ! e# jN" ! 0 $% e jN" + e# jN" &' + !1 $% e j ( N # 2 )" + e# j ( N # 2 )" &' + "

Multisection Transformer
9

Then we have:

If N is odd, the last term is ! ( N "1) 2 e j# + e" j# , while for even N,


the last term is ! N 2 . This equation for ! (" ) is in the form of a

! (" ) = A e# j"

finite Fourier cosine series in ! , which can be written as:

5.7

The response of a binomial matching transformer is


optimal in the respect, for a given number N of sections, the
response is as flat as possible near the design frequency; hence the
common terminology maximally flat. The response is designed so
that the first N ! 1 derivates of ! (" ) go to zero at the center
frequency f0 . Such response is obtained if we let:

= 2 N A cos"

Chebyshev Multisectional Matching Transformers

In contrast to the smooth response of the binomial


transformer, the Chebyshev transformer response is characterized
by ripples in its passband. If such ripples are tolerable, the
Chebyshev design will outperform the binomial design in terms of
achievable bandwidth for a given number of sections. The
Chebyshev transformer is designed by equating ! (" ) to a
Chebyshev polynomial, which has the optimal characteristics
necessary for this type of behavior. We first look at the
Chebyshev polynomials, and then discuss their use in design of
matching transformers using small reflection coefficient theory.

Binomial Multisectional Matching Transformers

! N$
binomial coefficient # & , but of even more immediate use to us
" n%
are the necessary characteristic impedances of the various quarterwavelength long sections of line. Table 5.1 of the text is devoted
to listing the normalized impedances to use for multisection
binomial transformers with a total of N sections, for N varying
between 2 and 6, and normalized load impedances between 1 and
10. An example of the use of this table in designing a
multisection binomial transformer is to be found on Blackboard.

The importance of these results lies in the fact that we can


synthesize any desired reflection coefficient response as a
function of frequency (recall ! = " ! = # ! v ) by properly
choosing the ! ns and using enough sections N. This is clear since
Fourier series may be used to approximate arbitrary smooth
functions to any desired accuracy if enough terms are used. We
will investigate two important types of multisection transformers
in the sections to follow: the maximally-flat (binomial) response,
and the Chebyshev (equiripple) response.

! (" ) = A 1 + e# j 2"

e j" + e# j"

The calculations outlined in the text from this point are


rather tedious, but the results are what we are primarily interested
in here. The values of the ! ns are indeed proportional to the

$ ! 0 cos N" + !1 cos ( N # 2 )" + !


&
! + ! n cos ( N # 2n )" + ! N 2 2 for N even
# jN" &
! (" ) = 2e
%
&! 0 cos N" + !1 cos ( N # 2 )" + !
&! + ! cos ( N # 2n )" + !
n
( N #1) 2 cos" for N odd
'

5.6

Chebyshev Polynomials

10

The notation Tn ( x ) is used for the Chebyshev


polynomials. The first few are given below:

For x > 1 , Tn ( x ) increases faster with x as n


increases.
Now use the change of variables x = cos! for x ! 1 . It
may be shown that the Chebyshev polynomials may be expressed
as:

T1 ( x ) = x
T2 ( x ) = 2x 2 ! 1
T3 ( x ) = 4x 3 ! 3x T4 ( x ) = 8x 4 ! 8x 2 + 1

Tn ( cos! ) = cos n!

Higher-order polynomials may be found from these using the


recursion formula:

We may extend this formula to values x > 1 as follows:

Tn ( x ) = 2xTn !1 ( x ) ! Tn ! 2 ( x )

(
)
T ( x ) = cosh ( n cosh x )

The first four Chebyshev polynomials are plotted below:

Tn ( x ) = cos n cos !1 x

x "1

!1

x >1

Another desirable transformation is to map the interval


! " ! m onto the interval x ! 1 ; this is accomplished by replacing
the variable cos! by cos! cos! m = sec ! m cos! . Then we have:

Tn ( sec ! m cos! ) = cos n #$ cos "1 ( sec ! m cos! ) %&


One more useful transformation is obtained by expanding
powers of cos! using multiple angle identities. We make use of:
T1 ( sec ! m cos! ) = sec ! m cos!
T2 ( sec ! m cos! ) = sec 2 ! m (1 + cos 2! )
T3 ( sec ! m cos! ) = sec 3 ! m ( cos 3! + 3cos! ) " 3sec ! m cos!
T4 ( sec ! m cos! ) = sec 4 ! m ( cos 4! + 4 cos 2! + 3) " 4 sec 2 ! m (1 + cos 2! ) + 1

We note the following interesting behavior:

In the range !1 " x " 1 we have Tn ( x ) ! 1 and Tn ( x )


oscillates between 1 . This is the equiripple behavior
that we will see in the passband of the transformer.

For x > 1 we have Tn ( x ) > 1 . We will map this

We will use these results to design transformers of up to 4


sections, and also in later results.

region outside of the passband of the transformer.


11

!f
4#
=2" m
f0
$

Design of Chebyshev Transformers


We can now design a Chebyshev equiripple transformer
by making ! (" ) proportional to TN ( sec ! m cos! ) , where N is the
number of sections of the transformer. Thus we have:

The ! n are found by expanding Tn ( sec ! m cos! ) and equating like


terms in cos ( N ! 2n )" . We may also use the approximation:

! (" ) = 2e# jN" $% ! 0 cos N" + !1 cos ( N # 2 )" + &'


= Ae# jN" TN ( sec " m cos" )

!n !

Table 5.2 of the text contains tables of exact results for some
specific ! n for 2, 3, and 4 section transformers.

The last term of the series is ! N 2 2 for even N and ! ( N "1) 2 cos#
for odd N. As with the binomial transformer, find A by letting
! = 0 , so:

5.8

Z " Z0
! (0) = L
= ATN ( sec # m )
Z L + Z0

Tapered Lines

In the limit of an infinite number of infinitesimal


discontinuities, a discrete step impedance design becomes a
continuously tapered design, as suggested in the figure below.

Then we have:

A=

1 Z n +1
ln
2
Zn

Z L ! Z0
1
Z L + Z 0 TN ( sec " m )

If the maximum allowable reflection coefficient in the passband is


! m , then ! m = A . Then ! m may be found from:

*1
$ 1 Z L " Z0
sec ! m = cosh , cosh "1 &
% # m Z L " Z0
,+ N

*1
'$ ln Z L Z 0
"1
) / ! cosh , N cosh & 2#
( /.
%
,+
m

')/
( /.

Once ! m is known, the fractional bandwidth is given by:


By changing the type of taper, we can obtain different passband
characteristics. Several common tapers are discussed in the text.
12

5.9

The Bode-Fano Criterion

We have discussed several ways to match an arbitrary load


at a single frequency. Now consider a lossless network used to
match an arbitrary load over a non-zero bandwidth, as shown
below.

Several questions may arise:

Can we achieve a perfect match (zero reflection) over a


given bandwidth?

If not, how well can we do? What tradeoffs are there


between ! m , the maximum reflection in the passband,
and the bandwidth?

How complicated must the matching network be for a


given specifiction?
These questions may be answered by the Bode-Fano
criterion, which gives a theoretical limit on the minimum
reflection coefficient magnitude that can be obtained with a
matching network for certain common matching network types.
These upper limits may not be achievable in practice, but provide
a benchmark against which the matching network may be
compared.

Consider the parallel RC load of (a) above. The Bode-Fano


criterion states:
#

&

$ ln ! (" ) d" % RC
0

Assume that we want to design a matching network with


reflection coefficient as shown in the figure below:

13

We have in this case:


#

$ ln ! (" ) d" = $ ln !
0

%"

d" = %" ln

1
'
&
!m
RC

This leads to the following considerations:

For given load, larger bandwidth !" can be achieved


at the expense of higher reflection in the passband.

A perfect match ! m = 0 may only be achieved at the


expense of a band !" of frequencies.

As R and/or C increase, the quality of the match


decreases, since !" and/or 1 ! m must decrease. This
means that higher-Q circuits are harder to match than
lower-Q circuits.
Since ln1 ! m is proportional to return loss (in dB ), the
criterion may be viewed as stating that the area between return
loss curve and the line ! = 1 (or RL = 0 dB ) must be equal to a
constant.

14

You might also like