Introduction

I t is common knowledge that both analog and digi-

tal design engineers are continually developing and refining circuits for increasingly
higher operational frequencies. Analog circuits for wireless communication in the giga-
hertz (GHz) range and the ever-increasing clock speeds of computer circuits in high-
performance mainframes, workstations, and, of course, personal computers exemplify
this trend. Global positioning systems require carrier frequencies in the range of
1227.60 and 1575.42 MHz. The low-noise amplifier in a personal communication sys-
tem may operate at 1.9 GHz and fit on a circuit board smaller in size than a dime. Satel-
lite broadcasting in the C band involves 4 GHz uplink and 6 GHz downlink systems. In
general, due to the rapid expansion of wireless communication, more compact ampli-
fier, filter, oscillator, and mixer circuits are being designed and placed in service at fre-
quencies generally above 1 GHz. There is little doubt that this trend will continue
unabated, resulting not only in engineering systems with unique capabilities, but also
special design challenges not encountered in conventional low-frequency systems.
This chapter reviews the evolution from low- to high-frequency circuit operations. It
motivates and provides the physical rationales that have prompted the need for new engi-
neering approaches to design and optimize these circuits. The example of a cellular phone
circuit, components of which will be analyzed in more detail in later chapters, serves as a
vehicle to outline the goals and objectives of this textbook and its organization.
The chapter begins with a brief historical discussion explaining the transition from
direct current (DC) to high-frequency modes of operation. As the frequency increases
and the associated wavelengths of the electromagnetic waves becomes comparable to
the dimensions of the discrete circuit components such as resistors, capacitors, and
inductors, these components start to deviate in their electric responses from the ideal
frequency behavior. It is the purpose of this chapter to provide the reader with an appre-
2 Chapter 1 .
lntroductlon

ciation and understanding of high-frequency passive component characteristics. In par-

ticular, due to the availability of sophisticated measurement equipment, the design
engineer must know exactly why and how the high-frequency behavior of his or her cir-
cuit differs from the low-frequency realization. Without this knowledge it will be impos-
sible to develop and understand the special requirements of high-performance systems.

1.1 Importance of Radiofrequency Design

The beginning of electrical circuit design is most likely traced back to the late
eighteenth and early nineteenth centuries when the first reliable batteries became avail-
able. Named after their inventor A. Volta (1745-1827), the Voltaic cells permitted the
supply of reliable DC energy to power the first crude circuits. However, it soon became
apparent that low-frequency alternating current (AC) power sources can transport elec-
tricity more efficiently and with less electric losses when transmitted over some dis-
tance and that rerouting the electric energy could be facilitated through transformers
that operate in accordance with Faraday's induction law. Due to pioneering work by
such eminent engineers as Charles Steinmetz, Thomas Edison, Werner Siemens, and
Nikolas Tesla, the power generation and distribution industry quickly gained entry into
our everyday life. It was James Maxwell (183 1-1879) who, in a paper first read in 1864
to the Royal Society in London, postulated the coupling of the electric and magnetic
fields whose linkage through space gives rise to wave propagation. In 1887 Heinrich
Hertz experimentally proved the radiation and reception of electromagnetic energy
through air. This discovery heralded the rapidly expanding field of wireless communi-
cation, from radio and TV transmissions in the 1920s and 1930s to cellular phones and
Global Positioning Systems (GPS) in the 1980s and 1990s. Unfortunately, the design
and development of suitable high-frequency circuits for today's wireless communica-
tion applications is not so straightforward. As will be discussed in detail, conventional
Kirchhoff-type voltage and current law analysis tools, as presented to first- and second-
year undergraduate electrical engineering students, apply strictly only to DC and low-
frequency lumped parameter systems consisting of networks of resistors, capacitors,
and inductors. They fail when applied to circuits governed by electromagnetic wave
propagation.
The main purpose of this textbook is to provide the reader with theoretical and
practical aspects of analog circuit design when the frequency of operation extends into
the radio frequency (RF) and microwave (MW) domains. Here conventional circuit
analysis principles fail. The following questions arise:

At what upper frequency does conventional circuit analysis become inappropriate?

4 Chapter 1 lntroductlon

After the input signal has been digitally preprocessed, it is converted back to ana-
log form via a digital-to-analogconverter (DAC). This low-frequency signal is mixed
with a high-frequency carrier signal provided by a local oscillator. The combined signal
is subsequently amplified through a power amplifier (PA) and then routed to the
antenna, whose task is to radiate the encoded information as electromagnetic waves
into free space.
In the block diagram of Figure 1-1 let us focus on the transmitter PA. This could be
a 2 GHz PA for cellular phones that may be implemented as a dual-stage amplifier.
Details of the circuit diagram for the first stage PA are shown in Figure 1-2(a).

RF Blocking

BFG425W Interstage Matching

................................., Network
Input Matching
Network
Figure 1-2(a) Simplified circuit diagram of the first stage of a 2 GHz power
amplifier for a cellular phone.

We notice that the input signal is fed through a DC blocking capacitor into an
input matching network, needed to match the input impedance of the transistor (type
BFG425W of Philips Semiconductors), operated in common emitter configuration, to
the output impedance of the mixer that precedes the PA. The matching is needed to
ensure optimal power transfer as well as to eliminate performance degrading reflec-
tions. The interstage matching network must then match the output impedance of the
transistor to the input impedance of the second stage of the PA. Key components in the
matching networks are microstrip lines shown by the shaded rectangles in Figure
1-2(a). At high frequency these distributed elements exhibit unique electric properties
that differ significantly from low-frequency lumped circuit elements. We also notice
Importance of RadiofrequencyDeslgn

additional networks to bias the input and output ports of the transistor. The separation
of high-frequency signals from the DC bias conditions is achieved through two RF
blocking networks that feature so-called radio frequency coils (RFCs).
The actual dual-stage circuit board implementation is given in Figure 1-2(b),
which shows the microstrip lines as copper traces of specific lengths and widths.
Attached to the microstrip lines are chip capacitors, resistors, and inductors.

k
0.5 inch
4
- - - - - - - - - - - - - - r C
_ Interstage Matching
-7 Network
,
,

Figure 1-2(b) Printed circuit board layout of the power amplifier.

To understand, analyze, and ultimately build such a PA circuit requires knowledge

of a number of crucial RF topics discussed in this textbook:

Microstrip line impedance behavior is discussed in "Transmission Line Analysis"

(Chapter 2) and its quantitative evaluation is considered in Chapter 3, "The Smith
Chart."
The ability to reduce a complicated circuit into simpler constituents whose input-
output is described through two-port network description. This is discussed in
Chapter 4, "Single- and Multipart Networks."
Strategies of generically developing particular impedance versus frequency
responses as encountered in filter design. Chapter 5, "A Brief Overview of RF Fil-
ter Design," outlines the basic discrete and distributed filter theories, and Chapter
8, "Matching Networks," delves into a detailed circuit implementation as related
to Figure 1-2(b).
 Chapter 1 Introduction

High-frequency bipolar junction and field effect transistors as well as RF diodes

are investigated in "Active FW Components" (Chapter 6) in terms of their physical
basis followed by "Active Circuit Device Models" (Chapter 7), where large signal
and small signal circuit models are analyzed.
The overall amplification requirements, as related to gain, linearity, noise, and sta-
bility, are basis of "RF Transistor Amplifier Design" (Chapter 9).
In addition to amplifiers, Chapter 10, "Oscillators and Mixers," focuses on addi-
tional important FW circuit design concepts, as shown in Figure 1- 1.

A successful RF design engineer knows about and applies all these concepts in the
design, construction, and testing of a particular RF circuit project. As the preceding
example implies, our concern in this textbook is mostly geared toward analog FW cir-
cuit theory and applications. We purposely neglect mixed and digital RF signals since
their treatment would exceed the size and scope of this textbook.

1.2 Dimensions and Units

To understand the upper frequency limit, beyond which conventional circuit the-
ory can no longer be applied to analyze an electric system, we should recall the repre-
sentation of an electromagnetic wave. In free space, plane electromagnetic (EM) wave
propagation in the positive z-direction is typically written in sinusoidal form:

H y = H o ycos ( a t - pz) (1.lb)

where Ex and H, are the x-directed electric and the y-directed magnetic field vector
components, as shown qualitatively in Figure 1-3. Here Eox and HOy represent con-
stant amplitude factors in units of V/m and Mm.
These waves possess an angular frequency a , and a propagation constant P that
defines the spatial extent in terms of the wavelength h , such that P = 2 d h . Classical
field theory based on Maxwell's equations reveals that the ratio between electric and
magnetic field components is defined in terms of the so-called intrinsic impedance Zo

based on the material dependent permeability p = pop, and permittivity E = E ~ E , ,

with pO and being absolute permeability and permittivity of free space and p, and E,
denoting relative values. We also point out that the field components are orthogonal to
each other and both are orthogonal to the direction of propagation. This is known as
transverse electromagnetic mode (TEM) and, since we deal exclusively with RF, it is
Dlmenslonsand Unlts 7

Figure 1-3 Electromagnetic wave propagation in free space. The electric and
magnetic fields are recorded at a fixed instance in time as a function of space
(9,9 are unit vectors in x- and y-direction).
the only mode that is considered in this text. TEM wave propagation is in stark contrast
to the various transverse electric (TE) and transverse magnetic (TM) wave modes,
which are the underlying principles of MW and optical communication. In these cases
the field vectors are no longer perpendicular to the direction of propagation.
The phase velocity vp of the TEM wave can be found via

Relevant quantities, units and symbols used throughout the book are summarized in
Tables A-1 and A-2 in Appendix A. Although we are dealing here with rather abstract
concepts of electromagnetic wave quantities, we can immediately relate (1.1) to circuit
parameters by observing that the electric field, as the unit of V/m already implies, can
intuitively be understood as a normalized voltage wave. Similarly, the magnetic field,
given in units of A/m, is a normalized current wave.

C & M W
Example 1-1: Intrinsic wave impedance, phase velocity, and
wavelengths

Compute the intrinsic wave impedance, phase velocity, and wave-

lengths of an electromagnetic wave in free space for the frequencies
f = 30 MHz, 300 MHz, 30 GHz.
8 Chapter 1 lntroductlon

Solution: Relative permeability and permittivity of free space

are equal to unity. Therefore, from (1.2) we determine that intrinsic
impedance in this case is equal to

The phase velocity according to (1.3) is equal to

v =- 1 1
= - = 2.999~10~ m/s
" &
which happens to be the speed of light v p = c . The wavelength is
evaluated by the following expression:

where f is the operating frequency. Using equation (1.4), we find

that the wavelength for an electromagnetic wave propagating in free
space at a frequency of 30 MHz is equal to h = 10 m; at 300 MHz it
is already reduced to h = 1 m; and at 30 GHz the wavelength is a
minute h = 1 cm.

This example conveys an appreciation of how the wavelength

changes as a function of frequency. As the frequency increases, the
wavelength reduces to dimensions comparable to the size of circuit
boards or even individual discrete components. The implication of
this fact will be analyzed in Chapter 2.

1.3 Frequency Spectrum

Because of the vast scope of applications, engineers have to deal with a broad
range of frequencies of circuit operation. Over the years several attempts have been
made to classify the frequency spectrum. The first designations for industrial and gov-
ernment organizations were introduced in the United States by the Department of
Defense during and shortly after World War 11. However, the most common frequency
spectrum classification in use today was created by the Institute of Electrical and Elec-
tronic Engineers (IEEE) and is listed in Table 1-1.
Frequency Spectrum 9

Table 1-1 IEEE Frequency Spectrum

I Frequency Band I Frequency Wavelength
ELF (Extreme Low Frequency) 30-300 Hz 10,000-1000 km
VF (Voice Frequency) 300-3000 Hz 1000-100 km
IVLF (Very Low Frequency) 1 3-30 kHz 1 100-10 km
LF (Low Frequency) 30-300 kHz 10-1 km
MF (Medium Frequency) 300-3000 kHz 1-0.1 km
IHF (High Frequency) 1 3-30 MHz 1 100-10 m
VHF (Very High Frequency) 30-300 MHz 10-1 m
UHF (Ultrahigh Frequency) 300-3000 MHz 100-10 cm
I SHF (Superhigh Frequency) 1 3-30 GHz 1 10-1 cm
I EHF (Extreme High Frequency) 1 30-300 GHz I 1-0.1 cm
Decimillimeter 300-3000 GHz 1-0.1 mrn
P Band 0.23-1 GHz 130-30 cm
I L Band 1 1-2 GHz 1 30-15 cm
S Band 2-4 GHz 15-7.5 cm
C Band 4-8 GHz 7.5-3.75 cm
X Band 1 8-12.5 GHz 1 3.75-2.4 crn
KUBand 12.5-18 GHz 2.4-1.67 cm
K Band 18-26.5 GHz 1.67-1.13 cm
Ka Band 126.5-40 GHz 1 1.13-0.75 cm
Millimeter wave 1 40-300 GHz 17.5-1 nun
Submillimeter wave 300-3000 GHz 1-0.1 rnrn

Based on Table 1- 1 and calculations carried out in Example 1- 1 we note that the
VHFNHF band, as typically encountered in television sets, constitutes the point at
which the wavelength first reaches dimensions equivalent to the physical extent of the
electronic system. It is this region where we need to begin to take into account the wave
nature of current and voltage signals in the respective electronic circuits. The situation
becomes even more critical when for instance 30 GHz frequency in the EHF band is
considered. Without being able to assign exact limits, the RF frequency range is cus-
tomarily associated from VHF to the S band. The MW frequency range has been tradi-
tionally associated with radar systems operating in the C band and above.
10 Chapter 1 introduction

1.4 RF Behavior of Passive Components

From conventional AC circuit analysis we know that a resistance R is frequency
independent and that a capacitor C and an inductor L can simply be specified by their
reactances Xc and XL as follows:

XL = o L (1.5b)
The implications of (IS), for example, are such that a capacitor of C = 1 pF and an
inductor of L = 1 nH at low frequencies of 60 Hz represent, respectively, either an open
or short circuit condition because
1
- 2 . 6 5 ~ 1 0SZ~= m

-
Xc(60 Hz) = (1.6a)
2n.. 6 0 . 10-l2 =

X,(60 Hz) = 2n. - 60 - 10" 3 . 7 7 ~ 1 0 -Q~ = 0 (1.6b)

It is important to point out that resistances, inductances, and capacitances are not only
created by wires, coils, and plates as typically encountered in conventional low-fre-
quency electronics. Even a single straight wire or a copper segment of a printed circuit
board (PCB) layout possesses frequency dependent resistance and inductance. For
instance, a cylindrical copper conductor of radius a, length I, and conductivity ocond
has a DC resistance of

For a DC signal the entire conductor cross-sectional area is utilized for the current flow.
At AC the situation is complicated by the fact that the alternating charge carrier flow
establishes a magnetic field that induces an electric field (according to Faraday's law)
whose associated current density opposes the initial current flow. The effect is strongest
at the center r = 0 , therefore significantly increasing the resistance in the center of the
conductor. The result is a current flow that tends to reside at the outer perimeter with
increasing frequency. As derived in Appendix B, the z-directed current density J, can be
represented by

where = - j o p o c o n d , and J o ( p r ) , J , ( p a ) are Bessel functions of zeroth and first

order, and I is the total current flow in the conductor. Further calculations reveal that the
RF Behavior of Passive Components 11

normalized resistance and inductance under high-frequency conditions (f 2 500 MHz)

can be put in the form
R/RDc z a/(26) (1.9)
and
(oL)/RDCG a/(26)
In these expressions 6 is the so-called skin depth
6 = (nf~oco,d)-1'2 (1.11)
which describes the spatial drop-off in resistance and reactance as a function of fre-
quencyf, permeability p , and conductivity oCond. For the equations (1.9) and (1.10) to
be valid it is assumed that 6 <ai . In most cases, the relative permeability of the conduc-
tor is equal to unity (i.e., p, = 1 ). Because of the inverse square root frequency behav-
ior, the skin depth is large for low frequencies and rapidly decreases for increasing
frequencies. Figure 1-4 exemplifies the skin depth behavior as a function of frequency
for material conductivities of copper, aluminum, and gold.

Figure 1-4 Skin depth beeavior of copper a,, = 64.516~106,~ l maluminum

,
o, = 40.0~10 Slm , and gold a, = 48.544~10 Slm .

If we consider the conductivity of copper, we can plot the AC current density (1.8)
normalized with respect to the DC current density Jzo= Z/(na2) as schematically
shown for the axisymmetric wire depicted in Figure 1-5(a).
For a fixed wire radius of, let us say, a = 1 mrn we can now plot Jz/Jzoas a
function of radius r for various frequencies as given in Figure 1-5(b).
12 Chapter 1 lntmducllon

Hieh
--0-
current Low current
density density

r
Current Flow -Q a

Figure 1-5(a) Schematic cross-sectional AC current density representation

normalized to DC current density.

Figure 1-5(b) Frequency behavior of normalized AC current density for a

copper wire of radius a = 1 rnrn.
We notice the significant increase in current flow at the outer perimeter of the wire
even for moderate frequencies of less than 1 MHz. At frequencies around 1 GHz, the
current flow is almost completely confined to the surface of the wire with negligible
radial penetration. An often used high-frequency approximation for the z-directed cur-
rent density is
 As seen in (1.12). the skin depth 6 has a simple physical meaning. It denotes the
reduction in the current density to the e-' factor (approximately 37%) of its original
DC value. If we rewrite (1.9) slightly, we find

This equation shows that the resistance increases inverse proportionally with the
cross-sectional skin area, see Figure 1-6.

Figure 1-6 Increase in resistance over the cross sectional surface area. The
current flow is confined to a small area defined by the skin depth 6.
.I
To standardize the sizes of wires, the American Wire Gauge (AWG) system is
commonly used in the United States. For instance, the diameter of the wire can be
determined by its AWG value. A complete listing of all AWG values and their corre-
sponding diameters is given in Table A-4 in Appendix A. The general rule is that in the
AWG system, the diameter of the wire roughly doubles every six wire gauges starting
with 1 mil for a AWG 50 wire (see Table A-4).

m m w
Example 1-2: Conversion between wire diameter and AWG
size

Determine the radius of the AWG 26 wire if the diameter of the

AWG 50 wire is 1.0 mil (or 2.54x10-~m).
14 Chapter 1 lntroductlon

Solution: The increase in diameter is computed as follows:

AWG50 d = 1 mil
AWG44 d = 2 m i l s
AWG38 d = 4 m i l s
AWG32 d = 8 m i l s
AWG 26 d =16 mils
Thus we determined that the diameter of a AWG 26 wire is equal to
16 mils. Therefore, the radius is

8 mil = 8 x (2.54~10-'m) = 0.2032 mm

Even in today's increasingly metric world, AWG has retained

its importance, and knowledge of how to convert mil-based AWG
size wires into millimeters often proves indispensable.

1.4.1 High-Frequency Resistors

Perhaps the most common circuit element in low-frequency electronics is a resis-
tor whose purpose is to produce a voltage drop by converting some of the electric
energy into heat. We can differentiate among several types of resistors:

Carbon-composite resistors of high-density dielectric granules

Wire-wound resistors of nickel or other winding material
Metal-film resistors of temperature stable materials
Thin-film chip resistors of aluminum or beryllium based materials

Of these types mainly the thin-film chip resistors find application nowadays in RF and
MW circuits as surface mounted devices (SMDs). This is due to the fact that they can
be produced in extremely small sizes, as Figure 1-7 shows.
As the previous section has shown, even a straight wire possesses an associated
inductance. Consequently, the electric equivalent circuit representation of a high-
frequency resistor of nominal value R is more complicated and has to be modified such
that the finite lead dimensions as well as parasitic capacitances are taken into account.
This situation is depicted in Figure 1-8.
The two inductances L model the leads, while the capacitances are needed to
account for the actual wire arrangement, which always represents a certain charge sepa-
RF Behavior ol Passive Componenm 15

Ire 1-7 One- and quarter-watt thin-film chip resistors in comparison with a
conventional quarter-watt resistor.

Figure 1-8 Electric equivalent circuit representation of the resistor.

.
ration effect modeled by capacitance C, and interlead capacitance Cb. The lead resis-
tance is generally neglected when compxed with the nominal resistance R. For a wire-
wound resistor the model is more complex, as Figurc 1-9 shows.

Figure 1-9 Electric equivalent circuit representation for a high-frequencywire-

wound resistor.
16 Chapter 1 lntroductlon

Here, in addition to the lead inductances L2 and the contact capacitance, we need
to include the inductance L1of the wire-wound resistor, which acts as a coil, and the
stray capacitance C, established between the windings. The interlead capacitance C2
(or Cb in Figure 1-8) is usually much smaller than the internal or stray capacitance and
in many cases can be neglected.

M& * W
Example 1-3: RF impedance response of metal film resistors
i -
Find the high frequency impedance behavior of a 500 Q metal film
resistor (see Figure 1-8) with 2.5 cm copper wire connections of
AWG 26 and a stray capacitance C, of 5 pF.

Solution: In Example 1-2 we have determined that the radius of

an AWG 26 wire is a = 2 . 0 3 2 ~ 1 0m~. According to (1.10) the
inductance of the straight wire at high frequency is approximately
equal to L = R D c a / ( 2 0 6 ) . Substituting (1.11) for the skin depth,
we get the following expression of a lead inductance (we set the
conductivity of copper to be a,, = 64.5 16x10~R-I . m-l ):
r

where the length of the leads is doubled to account for two connec-
tions. The preceding formula for the computation of the lead induc-
tance is applicable only for frequencies where the skin depth is
smaller than the radius of the wire [i.e., 6 = (7cfpo)-lJ2cc a ] or in
terms of frequency f D 1 /(7cpoc,a2) = 95 kHz.
Knowing the inductance of the leads, we can now compute the
impedance of the entire circuit as

The result of the computation is presented in Figure 1-10, where the

absolute value of the impedance of the resistor is plotted versus
frequency.
As seen, at low frequencies the impedance of the resistor is
equal to R. However, as the frequency increases and exceeds
RF Behavior of Passive Components 17

1o - ~
1o6 10' lo8 1o9 lo10 1ol1 loL2
Frequency, Hz
Figure 1-10 Absolute impedance value of a 500-Sl thin-film resistor as a
function of frequency.

10 MHz, the effect of the stray capacitance becomes dominant,

which causes the impedance of the resistor to decrease. Beyond the
resonance at approximately 20 GHz, the total impedance increases
due to the lead inductance, which represents an open circuit or infi-
nite impedance at very high frequencies.

This example underscores the care that is required when deal-

ing with the ubiquitous, seemingly frequency-independent resistors.
While not all resistors exhibit exactly the same response as shown in
Figure 1-10, it is the single, ofen multiple, resonance point that
occurs when the frequency reaches into the GHz range.

1.4.2 High-Frequency Capacitors

In most RF circuits chip capacitors find widespread application for the tuning of
filters and matching networks as well as for biasing active components such as transis-
tors. It is therefore important to understand their high-frequency behavior. Elementary
circuit analysis defines capacitance for a parallel plate capacitor whose plate dimen-
sions are large compared to its separation as follows:
18 Chapter 1 Introduction

where A is the plate surface area and d denotes the plate separation. Ideally there is no
current flow between the plates. However, at high frequencies the dielectric materials
become lossy (i.e., there is a conduction current flow). The impedance of a capacitor
must thus be written as a parallel combination of conductance Ge and susceptance oC:

In this expression the current flow at DC is due to the conductance G, = odielA/d,

with odielbeing the conductivity of the dielectric. It is now customary to introduce the
series loss tangent tanA, = o ~ / and o insert
~ ~it into~ the
~ expression for Ge to yield

Some practical values for the loss tangent are summarized in Table A-3. The corre-
sponding electric equivalent circuit with parasitic lead inductance L, series resistance R,
describing losses in the lead conductors, and dielectric loss resistance Re = l / G e , is
shown in Figure 1- 11.

Figure 1-11 Electric equivalent circuit for a high-frequency capacitor.

W&MW
Example 1-4: RF impedance response of capacitor

Compute the high frequency impedance of a 47 pF capacitor whose

dielectric medium consists of an aluminum oxide (AL203)possess-
ing a series loss tangent of lo4 (assumed to be frequency indepen-
dent) and whose leads are 1.25 cm AWG 26 copper wires
(oCu= 6 4 . 5 1 6 ~ 1 0R-I~ . m-'1.
RF Behavior of Passlve Components 19

Solution: Similar to Example 1-3, the inductance associated

with the leads is given by

The series resistance of the leads is computed from (1.13) to be

Finally, in accordance with (1.16), the parallel leakage resistance is

equal to

The frequency response of the magnitude of the impedance based on

equation (1.15) for the capacitor is shown in Figure 1- 12.

Figure 1-12 Absolute value of the capacitor impedance as a function of

frequency.

In computing the parallel leakage resistance Re we have

assumed the loss tangent tanA, to be frequency independent. In
reality, however, this factor may significantly depend upon the oper-
ating frequency. Unfortunately, data sheets often do not, or only
very incompletely, report this behavior.
20 Chapter 1 Introduction

Since the loss tangent can also be defined as the ratio of an

equivalent series resistance (ESR) to the capacitor's reactance, many
data sheets list ESR instead of tan A, . The ESR value is thus given as
tan A,
ESR = -
oc
This indicates that ESR + 0 as tanA, +0 .
As already known from the RF resistor impedance response in
Example 1-3, the capacitor reveals a similar resonance behavior
due to the presence of dielectric losses andjnite lead wires.

The construction of a surface-mounted ceramic capacitor is shown in Figure 1- 13.

The capacitor is a rectangular block of a ceramic dielectric into which a number of
interleaved metal electrodes are sandwiched. The purpose of this type of packaging is
to provide a high capacitance per unit volume by maximizing the electrode surface
area. Capacitance values range from 0.47 pF to 100 nF with operating voltage ranging
from 16V to 63 V. The loss tangent is usually listed by the manufacturer as
t a n 4 5 lo5 at a 1 MHz test frequency. Again, this loss tangent can significantly
increase as the frequency reaches into the GHz range.

~exarnicmaterial
Figure 1-13 Actual construction of a surface-mounted ceramic multilayer
capacitor.
 - -

RF Behavlor of Passive Components

Besides capacitance and loss tangent, manufactures list a nominal voltage that
cannot be exceeded at a particular operating temperature (for instance, T 5 85°C). Fur-
thermore, the capacitance is temperature dependent, as further discussed in the problem
section of this chapter.

1.4.3 High-Frequency Inductors

Although not employed as often as resistors and capacitors, inductors generally
are used in transistor biasing networks, for instance as RF coils (RFCs) to short circuit
the device to DC voltage conditions. Since a coil is generally formed by winding a
straight wire on a cylindrical former, we know from our previous discussion that the
windings represent an inductance in addition to the frequency-dependent wire resis-
tance. Moreover, adjacently positioned wires constitute separated moving charges, thus
giving rise to a parasitic capacitance effect as shown in Figure 1-14.

Figure 1-14 Distributed capacitance and series resistance in the inductor coil.

The equivalent circuit model of the inductor is shown in Figure 1-15. The para-
sitic shunt capacitance C,and series resistance R, represent composite effects of distrib-
uted capacitance Cd and resistance Rd, respectively.

Figure 1-15 Equivalent circuit of the high-frequency inductor.

22 Chapter 1 Introduction

C & Mw
Example 1-5: RF impedance response of an RFC

Estimate the frequency response of an RFC formed by N = 3.5 turns

of AWG 36 copper wire on a 0.1 inch air core. Assume that the
length of the coil is 0.05 inch.

Figure 1-16 Inductor dimensions of an air-core coil.

Solution: The dimensions of the coil are shown in Figure 1-16.

From Table A-4 in Appendix A, we find that the radius of the
AWG 36 wire is a = 2.5 mils = 63.5 Fm. The radius of the coil core
is r = 50 mils = 1.27 mm. The length of the coil is 1 = 50 mils
= 1 . 2 7 ~ .The distance between two adjacent turns is
d = Z / N= 3 . 6 ~ 1 m
0~ .
To estimate the inductance of the coil we will use a well-
known formula for the inductance of an air core solenoid:

Strictly speaking, this formula is valid only for the case when r << I
and the number of turns N is large. In our case, the length of the coil
is comparable with its radius and the number of turns is relatively
small. Therefore, (1.17) will not give an exact value for the induc-
tance, but a rather good approximation. Substituting the given val-
ues into (l.l7), we obtain L = 61.4 nH .
To approximate the effect of the capacitance C,, we will use
the formula for an ideal parallel-plate capacitor (1.14). In our case
RF Behavior of Passive Components 23

the separation d between the plates is assumed to be equal to the

distance between the turns d = 1/N = 3.6x10-~m , and the area A
can be estimated as A = 2alWire,where lWi, = 2nrN is the length
of the wire. We conclude that

Since the radius of the wire is only 63.5 pm, we can neglect the skin
effect and compute the series resistance R, as a DC resistance of the
wire.

The frequency response of the RFC impedance just analyzed is

shown in Figure 1-17.

Figure 1-17 Frequency response of the impedance of an RFC.

RFCsJind widespread use for biasing RF circuits. Howevel; as

Figure 1-1 7 shows, the frequency dependency can form complicated
resonance conditions with additional elements in an RF system.
Indeed, certain matching circuits rely on the RFCs as tuning elements.
24 Chapter 1 lntroductlon

As can be seen from Figure 1-17, the behavior of the RFC deviates from the
expected behavior of an ideal inductance at high frequencies. First, the impedance of
the RFC increases more rapidly as the frequency approaches resonance. Second, as the
frequency continues to increase, the influence of the parasitic capacitance Cs becomes
dominant and the impedance of the coil decreases.
If the RFC had zero series resistance, then the overall impedance behavior at reso-
nance would reach infinity, but due to the nonzero value of R, the maximum value of
the impedance is of finite value. To characterize the impact of the coil resistance, the
quality factor Q is commonly used:

where X is the reactance and R, is the series resistance of the coil. The quality factor
characterizes the resistive loss in this passive circuit, and for tuning purposes it is desir-
able that this factor is as high as possible.

1.5 Chip Components and Circuit Board Considerations

The practical realization of passive components on printed FW circuit boards is
primarily accomplished in chip form and placed on specially fabricated board materi-
als. In the following section we examine the three most common passive chip elements
in terms of their sizes and electric characteristics.

1.5.1 Chip Resistors

The size of chip resistors can be as small as 40 by 20 mils (where 1 mil = 0.001
inch = 0.0254 mm) for 0.5 W power ratings and up to 1 by 1 inch for 1000 W ratings in
RF power amplifiers. The chip resistor sizes that are most commonly used in circuits
operating up to several hundred watts are summarized in Table 1-2.
A general rule of thumb in determining the size of the chip components from the
known size code is as follows: the first two digits in the code denote the length L in
terms of tens of mils, and the last two digits denote the width W of the component. The
thickness of the chip resistors is not standardized and depends on the particular compo-
nent type.
The resistance value range from 1/10 Q up to several M i l . Higher values are diffi-
cult to manufacture and result in high tolerances. Typical resistor tolerance values range
from f 5% to fO.01 % .Another difficulty that arises with high-value resistors is that they
are prone to produce parasitic fields, adversely affecting the linearity of the resistance ver-
sus frequency behavior. A conventional chip resistor realization is shown in Figure 1-18.
Chlp Components and Clrcult Board ConsidetaIlona 25

Table 1-2 Standard sizes of chip resistors

I Geometry I Size Code 1 Length L, mils 1 Width W, mils I

Marking
Protective coat
1 Resistive layer

/
End contact
Inner electrodes Ceramic substrate
Figure 1-18 Cross-sectional view of a typical chip resistor,

A metal film (usually nichrome) layer is deposited on a ceramic body (usually

aluminum oxide). This resistive layer is trimmed to the desired nominal value by reduc-
ing its length and inserting inner electrodes. Contacts are made on both ends of the
resistor that allow the component to be soldered to the board. The resistive film is
coated with a protective layer to prevent environmental interferences.

1.5.2 Chip Capacltors

The chip capacitors are implemented either as a conventional single-plate configu-
ration, as shown in Figure 1-19, or a multiple-layer design (see Figure 1-13).
Frequently, single-plate capacitors are combined in clusters of two or four ele-
ments sharing a single dielectric material and a common electrode, as shown in Figure
1-20.
 Chip capacitor Ribbon lead or wire

\ /
C i u i t lraces
Flgure 1-19 Cross section of a typical single-plate capacitor connected to the
board.

Dual capacitor Quadrupole capacitor

llfP
1 -7-
Flgure 1-20 Clusters of single-plated capacitors sharing a common dielectric
material.

The standard sizes of the capacitors range from a minimum of 15 mils square in a
single layer configuration to 400 by 425 mils at higher values. m i c a 1 values for com-
mercial capacitors range from 0.1 pF to several pF . The tolerances vary from f 2 % to
f5 0 % . For small capacitances tolerances are usually expressed in terms of pF instead
of percent; for example, we often encounter capacitorswith the nominal values given as
(0.5 f 0.25) pF .

1.5.3 SurfaceMounted Inductors

The most common implementation of surface-mounted inductors is still the wire-
wound coil. A typical example of such an inductor with air core is shown in Figure 1-21.
Modem manufacturing technology allows us to make these inductors extremely small.
Their dimensions are comparable to those of chip resistors and capacitors. m i c a 1 sizes
Chlp Components and Circuit Board Conslderatlans

Figure 1-21 Typical size of an RF wire-wound air-core inductor in comparison

with a cellular phone antenna (courtesy of Coilcraft, Inc.).
of the surface-mounted wire-wound inductors range from 60 by 30 mils to 180 by 120
mils. The inductance values cover the range from 1 nH to 1000 pH.
When thickness constraints of the circuit play a major role, flat inductors are often
employed that can be integrated with microstrip transmission lines. A generic configu-
ration of a flat coil is shown in Figure 1-22. Although such thin-wire coils have rela-
tively low inductances on the order of 1to 500 nH, it is the frequency in the GHz range
that helps push the reactance beyond 1 kR.The physical construction can be as small
as2mmby2mm.

Air bridge
Figure 1-22 Flat coil configuration.An air bridge is made by using either a wire
or a conductive ribbon.

Flat coils are used in both integrated and hybrid circuits. Hybrid circuits are very
similar to an ordinary circuit, but discrete semiconductor elements are placed on the
dielectric substrate in die form (without case) and are connected to the conductors on
28 Chapter 1 lntroductlon

the board using bond wires. After the entire circuit is assembled, it is then placed into a
single case to protect it from environmental interferences. Resistors and capacitors for
hybrid circuits can directly be implemented on the board by metal-film deposition. This
approach permits significant reduction in the size of the circuit.

1.6 Summary
In this chapter the evolution from low- to high-frequency systems is discussed and
placed in a historical context. A key concept when dealing with high-frequency applica-
tions is the fact that the electromagnetic wave nature begins to dominate over Kirch-
hoff's current and voltage laws. Issues such as propagation constant and phase velocity,
0 1
p = 2n/h and v p = - = -
K p
gain importance.
A consequence of the electromagnetic wave nature is the skin effect, which forces
the current to flow close to the surface of the conducting structures. The depth of pene-
tration from the surface can be determined via the skin depth equation:

a=- 1

With the skin depth we can approximately characterize the frequency dependent resis-
tance and reactance of components at RF frequency. As an example, the simple cylin-
drical lead wires exhibit resistances and reactances that become a function of frequency

These wires, in conjunction with the respective R, C, and L elements, form electric equiv-
alent circuits whose performance markedly deviate from the ideal element behavior. We
find that the constant resistance at low frequency is no longer constant, but displays a
second-order system response with a resonant dip. The dielectric material in a capacitor
becomes lossy at high frequencies (i.e., allows the flow of a small conduction current).
The degree of loss is quantified by the loss tangent, which is tabulated for a range of
engineering materials. Therefore, a capacitor exhibits an impedance behavior that fol-
lows an inverse frequency response only at low frequencies. Finally, inductors represent
an impedance response that follows a linear increase at low frequencies before deviating
from the ideal behavior by reaching a resonance peak and then turning capacitive.
A passive RF component vendor will always attempt to keep the physical dimen-
sions of resistors, capacitors, and inductors as small as possible. This is desired since
the wavelength of high-frequency voltage and current waves becomes ever smaller,
Further Reading 29

eventually reaching the characteristic sizes of the circuit components. As discussed in

subsequent chapters, when the wavelength is comparable in size with the discrete elec-
tronic components, basic circuit analysis no longer applies.

Further Reading

I. Bohl and P. Bhartia, Microwave Solid State Design, John Wiley, New York, 1988.
C. Bowick, RF Circuit Design, Newmes, Newton, MA, 1982.
D. K Chen, Fundamentals of Engineering Electromagnetics, Addison-Wesley, Reading,
MA, 1993.
R. A. Chipman, Transmission Lines, Schaum Outline Series, McGraw-Hill, New York,
1968.
L. N. Dworsky, Modern Transmission Line Theory andApplications, Robert E. Krieger,
Malabar, FL, 1988.
M. F. Iskander, Electromagnetic Fields and Waves, Prentice Hall, Upper Saddle River,
NJ, 1992.
T. S. Laverghetta, Practical Mircowaves, Prentice Hall, Upper Saddle River, NJ, 1996.
H. P. Neff, Basic Electromagnetic Fields, 2nd ed. Harper & Row, New York, 1987.
K. F. Sander, Microwave Components and Systems, Addison-Wesley, 1987.
K. F. Sander and G. A. L. Read, Transmission and Propagation of Electromagnetic
Waves, 2nd ed. Cambridge University Press, Cambridge, UK, 1986.

W. Sinnema, Electronic Transmission Line Technology, 2nd ed., Prentice Hall, Upper
Saddle River, NJ, 1988.
F. T. Ulaby, Fundamentals of Applied Electromagnetics, Prentice Hall, Upper Saddle
River, NJ, 1997.

Problems
1.1 Compute the phase velocity and wavelength in an FR4 printed circuit board
whose relative dielectric constant is 4.6 and where the operational frequency
is 1.92 GHz.
 Chapter 1 lntroductlon

The current flowing in a microstrip line (assumed to be infinite and lossless)

is specified to be i(t) = 0.6cos(9 x 109t- 5002) A. Find the (a) phase
velocity, (b) frequency, (c) wavelength, and (d) phasor expression of the cur-
rent.

A coaxial cable that is assumed lossless has a wavelength of the electric and
magnetic fields of h = 20 cm at 960 MHz. Find the relative dielectric con-
stant of the insulation.

The electric wave field of a positive z-traveling wave in a medium with rela-
tive dielectric constant of E, = 4 and with frequency of 5 GHz is given by
E, = E,,cos ( a t - k z ) V/m
(a) Find the magnetic field if E,, = lo6 ~ l m .
(b) Determine phase velocity and wavelength.
(c) Compute the spatial advance of the traveling wave between time intervals
t, = 3ps and tz = 7 p s .

Find the frequency response of the impedance magnitude of the following

series and parallel LC circuits:
L=lOnH

C = lOpF
C = lOpF
Compare your results to the situation when the ideal inductance is replaced
by the same inductance and a 5 Q resistance connected in series. Assume
that these circuits operate in the VHFIUHF frequency band of 30-3000
MHz.

For the circuit shown, derive the resonance frequency and plot the resonance
frequency behavior as a function of the resistance R.
Problems

1.7 Repeat Problem 1.6 for the following circuit.

1.8 For the following circuit we chose R cc (m)/2.

Find I V,/ViI as a function of frequency and identify the dominant circuit

portions for the low-, mid-, and high-frequency domains.

1.9 One of the objectives of Chapter 1 is to sensitize the reader to high-fre-

quency phenomena that are usually neglected in a low-frequency circuit ana-
lysis. One such phenomenon is the skin effect. To show its importance in RF
calculations, (a) compute the frequency behavior of an inductor formed by
10 turns of AWG 26 copper wire on a 5 m m air core. The length of the coil is
5 mrn. (b) repeat the computations by first neglecting the skin effect and then
including it.

1.10 The leads of a resistor in an RF circuit are treated as straight aluminum wires
(o,, = 4.0 x 107s/m) of AWG size 14 and of total length of 5 cm. (a)
Compute the DC resistance. (b) Find the AC resistance and inductance at
100 MHz, 1 GHz, and 10 GHz operating frequencies.

1.11 Compute the skin depths for copper (a,, = 64.516 x 1 0 ~ ~ / maluminum
),
6
(o,, = 40 x 10 S/m), and gold (o,, = 48.544 x 106s/m) at 1 GHz and
10 GHz, and find the resistance of a 10 cm wire with diameter of 1 rnm.

1.12 A typical PCB substrate consists of A1203with a relative dielectric constant

of 10 and a loss tangent of 0.0004 at 10 GHz. Find the conductivity of the
substrate.
 Chapter 1 Introduction

For the series RLC circuit with R = 1 Q , L = 1 nH, and C = 1 pF,

compute the resonance frequency and quality factor at +lo% of the reso-
nance frequency. Does the presence of the resistor affect the resonance fre-
quency?

A 4.7 pF capacitor with relative dielectric constant of 4.6 and series loss tan-
gent of 0.003 is used in a circuit operated at 10 GHz. For a combined copper
lead length of 6 cm and diameter of 0.5 mrn, determine (a) the lead resis-
tance and lead reactance, and (b) the conductance and the total im edance.
-7
Q-I~ . m .
The conductivity of copper is given as oc, = 6 4 . 5 1 6 ~ 1 0

A manufacturer data sheet records the series loss tangent of a capacitor to be

at 5 GHz. For a total plate dimension of 1 0 - ~ c r nand
~ plate separation
of 0.01 mm and a relative dielectric constant of 10, find the conductance.

A two-element impedance of the generic form

Z = R+jX
has to be converted into an equivalent admittance form Y = 1IZ such that
Y = G+jB
Find the conductance G and susceptance B in terms of resistance R and reac-
tance X.

A more elaborate model of a capacitor is sometimes represented by the fol-

lowing circuit:

Here the loss tangent is specified as consisting of two parts involving the
admittance Y, = l/R,+ j o C with a parallel-circuit loss tangent
tanA, = ~ R ~ { Y , } / I ~ { Y , } ~ and series impedance Z, = Rs + l / ( j o C )
with a series-circuit loss tangent tanAs = ~Re{Z,}/Irn{Z,}~ (it is noted
that Rs is different from Example 1-4). Show that for low-loss capacitances
we approximately obtain tanA=tanA,+tanA,, where
tanA = IRe{Z}/Im{Z}I and Z is the total impedance.

When recording the capacitance with a measurement equipment, the user

has often the choice to select a suitable circuit representation. For the series
representation, the instrument attempts to predict Rs and C, , while for the
Problems

parallel representation the prediction involves Rp and C p . Which mode

should be chosen if large capacitors of more than 100 p F are to be mea-
sured? Is this mode also suitable for small values of less than 10 pF?
Explain your answers.

1.19 The ability to store electric charge, expressed through the capacitance,
depends on the operating temperature. This behavior can be quantified
through the relation C = C o [ l+ a ( T - 20°C)] , where Co is the nominal
capacitance and a is a temperature coefficient that can be positive or nega-
tive. If the capacitance C at T = 20°C is recorded to be 4.6 pF, which
increases to 4.8 pF at T = 40°C, what is the temperature coefficient a ?
Determine the capacitance at 0°C and 80°C .

1.20 When measuring impedance at low frequency we connect the measurement

equipment to a device using a pair of wires and assume that the reading
reflects the impedance of the device under test (DUT). As we have seen in
this chapter, at high frequencies we have to take into account the influence of
the parasitic elements. The typical circuit representation of the measurement
arrangement is as follows.
z,, ZDUT
, Rs
A a
V
1
Measurement ;
Equipment Yi ==CP? DUT
a

A
" J
Cables and Fixture *
Measurement Device
Plane Plane
Here the fixture and cables are replaced by an equivalent circuit of the lead
impedance ( R s + j o L s ) and stray admittance ( G p+ joC,). Ideally, we
would like to perform the measurement at the device plane. However, due to
the influence of the fixture, the measurement plane is shifted away from the
DUT.
To measure accurately the impedance of the DUT, the test fixture with
connecting cables has to be taken into account. The methodology adopted by
most manufacturers is to compensate for these undesired, fixture-related
influences through an open- and short-circuit calibration. The first step is to
34 Chapter 1 .
Introduction

replace the DUT by a short circuit and record the resulting impedance. Due
to the influence of the fixture, the measured impedance will not be equal to
zero. Next, the short circuit is replaced by an open circuit and the impedance
is recorded again. These two measurements allow us to quantify the parasitic
influence of the fixture.
After calibration, we can connect the DUT and measure the input
impedance. The equivalent circuit in this case is as follows.
Zs =Rs+joL,

ni'

Knowing the values of the parasitic elements (Zs and Y p ) , we can now
compute the true impedance of the DUT.
Explain the procedure with all necessary equations, and specify under
what conditions such a calibration is possible. Next, develop the formula that
allows us to find the desired DUT impedance in the absence of the fixture.

1.21 The results of a frequency sweep impedance measurement of an unknown

passive device are shown in the following figure.

lo4t
c
6
- lo3.
0
U

u
E
10':

lo0 . .. . .......' . .

lo4 105 lo6 10' lo8 lo9 io10 10" 10"

Frequency, Hz
Problems 35

Based on the shape of the impedance response, develop a circuit that can be
used as an equivalent circuit to replicate this device under test. What device
can it be: resistor, inductor, or capacitor?

1.22 To measure the impedance of a passive component at RF frequencies is quite

a challenge. Conventional techniques such as bridge circuits and resonance
techniques fail beyond a few MHz. A technique pursued by several instru-
ment manufactures is the current voltage recording based on the following
simplified schematic.

Here the voltages are measured with vector voltmeters that allow the record-
ing of magnitude and phase. Explain how the impedance of the component
under test is determined and discuss the purpose of the transformer and
operational amplifier.

1.23 An RFC is constructed by winding four turns of AWG 38 copper wire on a

2 mm ceramic core diameter ( p r = 1 ) of 0.1 mm length. Based on Example
1-5, estimate the inductance, stray capacitance, resistance, and resonance
frequency.

1.24 Using data and the equivalent circuit diagram developed in the previous
problem, find values of the equivalent circuit parameters for the magnitude
of the impedance if the device is 100 SZ under DC conditions and 1257 Q
at 100 GHz. Assume the resonance frequency point to be at 1.125 GHz.

1.25 A quadrupole capacitor as shown in Figure 1-18 consists of four equal-size

electrodes of 25 mils square separated 5 mils from a common ground plane
through a dielectric medium of a relative dielectric constant of 11. Find the
individual and total capacitance that can be achieved.
 C H A P T E R 3

The Smith Chart

A transmission line changes its impedance

depending on material properties and geometric dimensions. Typical practical realiza-
tions include microstrip line, coaxial cable, and parallel-plate line. In addition, both the
length and operating frequency of the transmission line significantly influence the input
impedance. In the previous chapter we derived the fundamental equation describing the
input impedance of a terminated transmission line. We found that this equation incorpo-
rates the characteristic line impedance, load impedance, and, through the argument of
the tangent function, line length and operating frequency. As we saw in Section 2.9, the
input impedance can equivalently be evaluated by using the spatially dependent reflec-
tion coefficient. To facilitate the evaluation of the reflection coefficient, P. H. Smith
developed a graphical procedure based on conformal mapping principles. This
approach permits an easy and intuitive display of the reflection coefficient as well as the
line impedance in one single graph. Although this graphical procedure, nowadays
known as the Smith Chart, was developed in the 1930s prior to the computer age, it has
retained its popularity and today can be found in every data book describing passive
and active RFMW components and systems. Almost all computer-aided design pro-
grams utilize the Smith Chart for the analysis of circuit impedances, design of matching
networks, and computations of noise figures, gain, and stability circles. Even instru-
ments such as the ubiquitous network analyzer have the option to represent certain
measurements in a Smith Chart format.
This chapter reviews the steps necessary to convert the input impedance in its
standard complex plane into a suitable complex reflection coefficient representation via
a specific conformal transformation originally proposed by Smith. The graphical dis-
102 Chapter 3 The Smlth Chart

play of the reflection coefficient in this new complex plane can then be utilized directly
to find the input impedance of the transmission line. Moreover, the Smith Chart facili-
tates evaluation of more complicated circuit configurations, which will be employed in
subsequent chapters to build filters and matching networks for active devices.
The following sections present a step-by-step derivation of the Smith Chart fol-
lowed by several examples of how to use this graphical design tool in computing the
impedance of passive circuits.

3.1 From Reflection Coefficient to Load Impedance

In Section 2.9 the reflection coefficient is defined as the ratio of reflected voltage
wave to incident voltage wave at a certain fixed spatial location along the transmission
line. Of particular interest is the reflection coefficient at the load location d = 0 . From a
physical point of view this coefficient T o describes the mismatch in impedance between
the characteristic line impedance Zo and the load impedance Z L as expressed by (2.52).
In moving away from the load in the positive d-direction toward the beginning of the
transmission line, we have to multiply T o by the exponential factor exp (- j2pd) , as
seen in (2.64),to obtain T ( d ). It is this transformation from T o to T ( d ) that constitutes
one of the key ingredients in the Smith Chart as a graphical design tool.

3.1.1 Reflection Coefficient in Phasor Form

The representation of the reflection coefficient T o can be cast in the following
complex notation.

where 0 , = tan-' ( b i / T o , ) . We recall that pure short- and open-circuit conditions in

(3.1) correspond to T o values of -1 and + 1, located on the real axis in the complex T -
plane.

c m w -
Example 3-1: Reflection coefficient representations

A transmission line with a characteristic line impedance of

Z , = 50 R is terminated into the following load impedances:

(a) Z L = 0 (short circuit)

From Retlactlon Coefficient to Load Impedance 103

(b) ZL + 00 (open circuit)

(c) ZL = 50 i2
(d) Z, = (16.67 - jl6.67) 51
(e) Z, = (50 + j150) 51

Find the individual reflection coefficients To and display them in

the complex r -plane.

Solution: Based on (3.1) we compute the following numbers for

the reflection coefficients:
(a) To = -1 (short circuit)
(b) To = 1 (open circuit)
(c) r0= 0 (matched circuit)
(d) r0= 0.54L221°
(e) To = 0.83L34"
The values are displayed in polar form in Figure 3- 1.

Figure 3-1 Complex T -plane and various locations of To

The rejection coeficient is represented in phasor form as done

when dealing with the conventional voltages and currents in basic
circuit theory.
104 Chapter 3 me Smith Chart

3.1.2 Normalized Impedance Equation

Let us return to our general input impedance expression (2.69), into which we
substitute the reflection coefficient

This results in

In order to generalize the subsequent derivations, we normalize (3.3) with respect to the
characteristic line impedance as follows

The preceding equation represents a mapping from one complex plane, the zin-plane,
to a second complex plane, the r -plane. Multiplying numerator and denominator of
(3.4) by the complex conjugate of the denominator allows us to isolate real and imagi-
nary parts of zin in terms of the reflection coefficient. This means

can be separated into

and

Equations (3.6) and (3.7) are explicit transformation rules of finding zin if the reflection
coefficient is specified in terms of T, and T i . Therefore, the mapping from the com-
plex r-plane into the zin-plane is straightforward, as the following example under-
scores.
From Reflection Coefficient to Load impedance

C & M W
Example 3-2: Input impedance of a terminated transmission
line

A load impedance Z L = (30 + j60) R is connected to a 50 R

transmission line of 2 cm length and operated at 2 GHz. Use the
reflection coefficient concept and find the input impedance Zin
under the assumption that the phase velocity is 50% of the speed of
light.

Solution: We first determine the load reflection coefficient

Next we compute T ( d = 2cm) based on the fact that

This results in 2pd = 191.99" and yields for the refection coeffi-
cient
r = Toe-j2pd = r, + jTi = - 0.32 - j0.55 = J2/5e -j120.43"
Having thus determined the reflection coefficient, we can now
directly find the corresponding input impedance:

We note that the reflection coeficient phasor form at the load,

T o , is multiplied with a rotator that incorporates twice the electric
line length pd. This mathematical statement thus conveys the idea
that voltage/current waves have to travel to the load and return back
to the source to dejine the input impedance.

Example 3.2 could have been solved just as efficiently by using the impedance
equation (2.65) developed in Section 2.9.
106 Chapter 3 The Smith Chart

3.1.3 Parametric Reflection Coefficient Equation

The goal of our investigation is to pursue a different approach toward computing
the input impedance. This new approach involves the inversion of (3.6) and (3.7). In
other words, we ask ourselves how a point in the zin -domain, expressed through its nor-
malized real, r, and imaginary, x, components, is mapped into the complex r-plane,
where it then can be expressed in terms of the real, T, , and imaginary, Ti, components
of the reflection coefficient. Since T appears in the numerator and denominator, we
have to suspect that straight lines in the impedance plane zin may not be mapped into
straight lines in the T -plane. All we can say at this point is that the matching of the load
impedance to the transmission line impedance Zin = Z, or zin = 1 , results in a zero
reflection coefficient (i.e., r, = Ti = 0 ) located in the center of the r -plane.
The inversion of (3.6) is accomplished by going through the following basic alge-
braic operations:

At this point the trick consists in recognizing that T, can be written as a complete bino-
mial expression (see also Appendix C)

This finally can be cast in the form

In an identical way as done previously, we proceed to invert (3.7). The result for
the normalized reactance is

Both (3.10) and (3.11) are parametric equations of circles in the complex T -plane that
2 2 2
can be written in the generic form (r,- a ) + (Ti - b) = c . Here a, b denote shifts
along the real and imaginary T axes, and c is the radius of the circle.
From Reflection Coefficient to Load Impedance 107

Figure 3-2 depicts the parametric circle equations of (3.10) for various resis-
tances. For example, if the normalized resistance r is zero, the circle is centered at the
2
origin and possesses a radius of 1, since (3.10) reduces to T,+ T i = 1. For r = 1 we
2

find (T,- 1/212 + = (1/2)' , which represents a circle of radius 112 shifted in the
positive r, direction by 1/2 units. We conclude that as r increases, the radii of the cir-
cles are continually reduced and shifted further to the right toward the point 1 on the
real axis. In the limit for r + m we see that the shift converges to the point
r/(r + 1) + 1 and the circle radius approaches 1/( r + 1)24 0 .
It is important to realize that this mapping transforms fixed values of r only and
does not involve x. Thus, for a fixed r an infinite range of reactance values x, as indi-
cated by the straight lines in the z-plane, maps onto the same resistance circle. The
mapping involving r alone is therefore not a unique point-to-point correspondence.

z-plane r-plane
Constant resistance lines (r = const)
Figure 3-2 Parametric representation of the normalized resistance r in the
complex T -plane.

A different graphical display results for the circle equation (3.1 I), which involves
the normalized reactance. Here the centers of the circles reside all along a line perpen-
2 2
dicular to the T, = 1 point. For instance, for x = we note that (T, - 1) + Ti = 0 ,
which is a circle of zero radius, or a point located at l?, = 1 and T i= 0 . For x = 1
we see that the circle equation becomes ( r , - 1)2 + (Ti - 1) = 1 . As x + 0 the radii
2

and shifts along the positive imaginary axis approach infinity. Interestingly, the shifts
108 Chapter 3 The Smith Chart

can also be along the negative imaginary axis. Here for x = -1 we notice that the cir-
2 2
cle equation becomes ( r , - 1) + (Ti + 1) = 1 with the center located at T,= 1 and
Ti = -1 . We observe that negative x -values refer to capacitive impedances residing in
the lower half of the r -plane. Figure 3-3 shows the parametric form of the normalized
imaginary impedance. For better readability the circles are displayed inside the unit cir-
cle only. In contrast to Figure 3-2 we notice that fixed x-values are mapped into circles
in the r-plane for arbitrary resistance values 0 l r < w , as indicated by the straight
lines in the impedance plane.
The transformations (3.10) and (3.11) taken individually do not constitute unique
mappings from the normalized impedance into the reflection coefficient plane. In other
words, impedance points mapped into the r-plane by either (3.10) or (3.11) cannot
uniquely be inverted back into the original impedance points. However, since the trans-
formations complement each other, a unique mapping can be constructed by combining
both transformations, as discussed in the next section.

z-plane (r > 0) r-plane

Constant reactance lines (x = const)

Figure 3-3 Parametric representation of the normalized reactance x in the

complex -plane. r
3.1.4 Graphical Representation
Combining the parametric representations for normalized resistance and reactance
circles (i.e., Figures 3-2 and 3-3) for (rl5 1 results in the Smith Chart as illustrated in
From Reflectlon Coefflclent to Load Impedance 109

Figure 3-4. An important observation of the Smith Chart is that there is a one-to-one
mapping between the normalized impedance plane and the reflection coefficient plane.
We notice also that the normalized resistance circles r have a range 0 Ir < = and the
normalized reactance circles x can represent either negative (i.e., capacitive) or positive
(i.e., inductive) values in the range -= < x < += .
It should be pointed out that the reflection coefficient does not have to satisfy
Irl 5 1 . Negative resistances, encountered for instance as part of the oscillation condi-
tion for resonators, lead to the case Irl > 1 and consequently map to points residing
outside the unit circle. Graphical displays where the reflection coefficient is greater than
1 are known as compressed Smith Charts. These charts, however, play a rather limited
role in RF/MW engineering designs and are therefore not further pursued in this text.
The interested reader may consult specialized literature (see the Hewlett-Packard appli-
cation note listed at the end of this chapter).

z-plane r-plane
Figure 3-4 Smith Chart representation by combining rand x circles for Irl
In Figure 3-4 we must note that the angle of rotation 2pd introduced b the length
of the transmission line is measured from the phasor location of To = lroleW L in clock-
wise (mathematically negative) direction due to the negative exponent (-2 j p d ) in the
reflection coefficient expression (3.2). For the computation of the input impedance of a
terminated transmission line, the motion is thus always away from the load impedance
or toward the generator. This rotation is indicated by an arrow on the periphery of the
chart. We further observe that a complete revolution around the unit circle requires
110 Chaptef 3 The Smith Chart

where d = h / 2 or 180'. The quantity pd is sometimes referred to as the electrical

length of the line.

3.2 lmpedance Transformation

3.2.1 lmpedance Transformation for General Load

The determination of the impedance response of a high-frequency circuit is often
a critical issue for the RF design engineer. Without detailed knowledge of the imped-
ance behavior, RFJMW system performance cannot adequately be predicted. In this
section we will elaborate on how the impedance can be determined easily and effi-
ciently with the aid of the previously introduced Smith Chart.
A typical Smith Chart computation involving a load impedance ZL connected to a
transmission line of characteristic line impedance Zo and length d proceeds according
to the following six steps:
1. Normalize the load impedance ZL with respect to the line impedance Zo to deter-
mine zL .
2. Locate zL in the Smith Chart.
3. Identify the corresponding load reflection coefficient T o in the Smith Chart both
in terms of its magnitude and phase.
4. Rotate To by twice its electrical length pd to obtain T i n ( d .)
5. Record the normalized input impedance zin at this spatial location d.
6. Convert zin into the actual impedance Zi, .
Example 3-3 goes through these steps, which are the standard procedure to arrive at the
graphical impedance solution.

Y , & M w
Example 3-3: Transmission line input impedance determina-
tion with the Smith Chart

Solve Example 3-2 by following the six-step Smith Chart computa-

tions given in the preceding list.
Impedance Transtormatlon 111

Solution: We commence with the load impedance

ZL = (30 + j60) R and proceed according to the previously out-
lined steps:

1. The normalized load impedance is

2. This point can be identified in the Smith Chart as the intersec-

tion of the circle of constant resistance r = 0.6 with the circle of
constant reactance x = 1.2, as seen in Figure 3-5.

3. The straight line connecting the origin to point zL determines

the load reflection coefficient T o . The associated angle is recorded
with respect to the positive real axis.

4. Keeping in mind that the outside circle on the Smith Chart cor-
responds to the unity reflection coefficient (Irol= 1 ), we can find
its magnitude as the length of the vector connecting the origin to zL.
Rotating this vector by twice the electrical length of the line (i.e.,
2 x pd = 2 x 96" = 192" ) yields the input reflection coefficient
'in .

5. This point uniquely identifies the associated normalized input

impedance zin = 0.3 - j0.53.

6. The preceding normalized impedance can be converted back

into actual input impedance values by multiplying it by
Z0 = 50 Q , resulting in the final solution: 2, = (15 - j26.5)Q.

We recall that the exact value of the input impedance obtained

in Example 3-2 is (14.7 - j26.7) Q . The small discrepancy is
understandable because of the approximate processing of the graph-
ical data in the Smith Chart. The entire sequence of steps leading to
the determination of the input impedance of the line connected to
the load is shown in Figure 3-5.
112 Chapter 3 The Smith Chart

Figure 3-5 Usage of the Smith Chart to determine the input impedance for
Example 3-3.

These steps appear at jrst cumbersome and prone to error if

carried out by hand. Howevel; using mathematical spreadsheets and
relying on modern computer-based instrumentation, the calcula-
tions are routinely done in seconds and with a high degree of
accuracy.
Impedance Transformation 113

3.2.2 Standing Wave Ratio

From the basic definition of the SWR in Section 2.8.3 it follows that for an arbi-
trary distance d along the transmission line, the standing wave ratio is written

where T(d) = Toexp (- j2pd) . Equation (3.12) can be inverted to give

This form of the reflection coefficient permits the representation of the SWR as circles
in the Smith Chart with the matched condition r ( d ) = 0 (or SWR = 1) being the
origin.
It is interesting to note that equation (3.12) is very similar in appearance to the
expression for determining the impedance from a given reflection coefficient:

This similarity, together with the fact that for (T(d)(5 1 the SWR is greater or equal to
unity, suggests that the actual numerical value for the SWR can be found from the
Smith Chart by finding the intersection of the circle of radius (T(d)( with the right-
hand side of the real axis.

Example 3-4: Reflection coefficient, voltage standing wave

ratio, and return loss

Four different load impedances:

(a)ZL = 5 0 Q , (b)ZL = 48.5 SZ, (c)ZL = (75+ j25) SZ, and
(d) ZL = (10 - j5) SZ , are sequentially connected to a 50 SZ trans-
mission line. Find the reflection coefficients and the SWR circles,
and determine the return loss in dB.

Solution: The normalized load impedances and corresponding

reflection coefficients, return loss, and SWR values are computed as
follows:
 Chapter 3 The Smith Chart

(a) zL = 1 , r = ( z L - l ) / ( z L +1 ) = 0 , RLdB = W , SWR = 1

(b) zL = 0.97, r = ( z L - l ) / ( z L+ 1) = -0.015, RL,, = 36.3,

SWR = 1.03

(c) zL = 1.5 + j0.5, r = ( z L - I ) / ( z L + 1 ) = 0.23 + j0.15,

RLdB = 11.1, SWR = 1.77

(d) zL = 0.2 - j0.1, r = ( z L - l ) / ( z L+ 1) = - 0.66 - j0.14,

RL,, = 3.5, SWR = 5.05

To determine the approximate values of the SWR requires us to

exploit the similarity with the input impedance, as discussed previ-
ously. To this end, we first plot the normalized impedance values in
the Smith Chart (see Figure 3-6). Then we draw circles with centers
at the origin and radii whose lengths reach the respective impedance
points defined in the previous step. From these circles we see that
the load refection coefficient for zero load reactance ( x L = 0 ) is

The SWR can be defined in term of the real load reflection coeffi-
cient along the real T -axis:

This requires lrol= rr2 0 . In other words, for Tr 2 0 we have to

enforce rL 2 1 , meaning that only the intersects of the right-hand-
side circles with the real axis define the SWR.

As a graphical design tool, the Smith Chart allows immediate

observation of the degree of mismatch between line and load imped-
ances by plotting the radius of the SWR circle.
--

Impedance Transformatlon

Figure 3-6 SWR circles for various reflection coefficients.

3.2.3 Special Transformation Conditions

The amount of rotation by which the point of the normalized transmission line
impedance circles around the Smith Chart is controlled by the length of the line, or
alternatively the operating frequency. Consequently, both inductive (upper plane) and
capacitive (lower plane) impedances can be generated based on the line length and the
termination conditions at a given frequency. These lumped circuit parameter represen-
tations, realized through distributed circuit analysis techniques, are of significant practi-
cal importance.
The cases of open- and short-circuit line termination are of particular interest in
generating inductive and capacitive behavior and are examined in more detail next.
116 Chapter 3 The Smith Chart

Open Circuit Tkansformations

To obtain a pure inductive or capacitive impedance behavior, we need to operate
along the r = 0 circle. The starting point is the right-hand location ( T o = 1 ) with
rotation toward the generator in a clockwise sense.
A capacitive impedance -j X c is obtained through the condition

as direct comparison with (2.70) shows. The line length d l is found to be

1
d , = ~[c~l-~(---& + nn]
where n n ( n = 1 , 2 , . . . ) is required due to the periodicity of the cotangent function.
Alternatively, an inductive impedance jXLcan be realized via the condition

The line length d 2 is now found to be

Both conditions are schematically depicted in Figure 3-7. How to choose a particular
open-circuit line length to exhibit capacitive or inductive behavior is discussed in the
following example.

s m W 4
Example3-5: Representation of passive circuit elements
through transmission line section

For an open-ended 50 2 ! transmission line operated at 3 GHz and

with a phase velocity of 77% of the speed of light, find the line
lengths to create a 2 pF capacitor and a 5.3 nH inductor. Perform
your computations both by relying on (3.16) and (3.18) and by using
the Smith Chart.

Solution: For a given value of phase velocity, the propagation

constant is
Impedance Transtormatlon 117

Substituting this value into (3.16) and (3.18), we conclude that for
the representation of a 2 pF capacitor we need an open-circuit line or
stub with line length d = 13.27 + n38.5 rnrn . For the realization of
a 5.3 nH inductor, a d , = 32.81 + n38.5 mm stub is required.
The alternative method for computing the lengths of the
required stubs is through the use of the Smith Chart (see Figure 3-7).
At a 3-GHz frequency, the reactance of a 2 pF capacitor is
Xc = l / ( o C ) = 26.5!2. The corresponding normalized imped-

Flgure 3-7 Creating capacitive and inductive impedances via an open-circuit

transmission line.
118 Chapter 3 The Smlth Chart

ance in this case is zc = - j X c = -j0.53. From the Smith Chart

we can deduce that the required transmission line length has to be
approximately 0.172 of one wavelength. We note that for the given
phase velocity, the wavelength is h = v,/ f = 77 mm . This
results in a line length of d l = 13.24 mm which is very close to the
previously computed value of 13.27 mm. Similarly, for the induc-
tance we obtain zL = j 2 . The line length in this case is 0.426 of
one wavelength, which is equal to 32.8 mm.

Circuits are often designed with lumped elements before con-

verting them into transmission line segments, similar to the proce-
dure described in this example.

Short-circuit Transformations
Here the transformation rules follow similar procedures as outlined previously,
except that the starting point in the Smith Chart is now the To = - 1 point on the real
axis, as indicated in Figure 3-8.
A capacitive impedance -j X c follows from the condition

where use is made of (2.66).The line length d l is found to be

dl = l[n-tan-'(-) 1 + nn]
P w cz,
Alternatively, an inductive impedance jXL can be realized via the condition

The line length d2 is now found to be

At high frequencies, it is very difficult to maintain perfect open-circuit conditions

because of changing temperatures, humidity, and other parameters of the medium sur-
rounding the open transmission line. For this reason short-circuit conditions are more
preferable in practical applications. However, even a short-circuit termination becomes
Impedance Transformatlon 119

Figure 3-8 Creating capacitive and inductive impedances via a short-circuit

transmission line.

problematic at very high frequencies or when through-hole connections in printed cir-

cuit boards are involved, since they result in additional parasitic inductances. Moreover,
a design engineer may not have a choice if the circuit layout area is to be minimized by
requiring the selection of the shortest line segments. For instance, the realization of a
capacitor always yields the shortest length for an open-circuit line.

3.2.4 Computer Simulations

There are many computer aided design (CAD) programs available to facilitate the
RF/MW circuit design and simulation processes. These programs can perform a multi-
tude of tasks, varying from simple impedance calculations to complex circuit optimiza-
tions and circuit board layouts. One commercial software package that is used throughout
120 Chapter 3 The Smlth Chart

this textbook is called Monolithic and Microwave Integrated Circuit Analysis and Design
(MMICAD) (Optotek Ltd., Kanata, Ontario, Canada), which is a linear simulator pro-
gram with optimization tools. Another well-known program with advanced features is
EESof's Libra package (Hewlett-Packard Corporation, Westlake Village, CA, USA),
which is capable of performing linear as well as nonlinear analyses and optimizations.
It is not the purpose of this textbook to review and discuss the various CAD pro-
grams presently in industrial and academic use. However, to reproduce the subsequent
simulation results, Appendix I provides a brief introduction to the basic features of
MATLAB, which was chosen as a tool to carry out most simulations presented in this book.
The main reason for using MATLABis its wide-spread use as a mathematical
spreadsheet which permits easy programming and direct graphical display. This elimi-
nates the need to rely on complex and expensive programs accessible to only a few
readers. The benefit of a MATLAB routine will immediately become apparent when the
Smith Chart computations have to be performed repetitively for a range of operating
frequencies or line lengths as the following discussion underscores.
In this section we revisit Example 3-2, which computed the input reflection coeffi-
cient and input impedance of a generic transmission line connected to a load. We now
extend this example beyond a single operating frequency and a fixed line length. Our
goal is to examine the effect of a frequency sweep in the range from 0.1 GHz to 3 GHz
and a change in line length varying from 0.1 cm to 3 cm. The example MATLAB routine,
which performs the analysis of the transmission line length changing from 0.1 cm to
3 cm at a fixed operating frequency 2 GHz, is as follows:

smith-chart ; %plot smith chart

Set-ZO(50); %set characteristic impedance to 50 Ohm
s_load(30+j*60); %set load impedance to 30+j60 Ohm
vp=0.5*3e8; %compute phase velocity
f=2e9 ; %set frequency to 2 GHz
d=O.0:0.001:0.03; %set the line length to a range from 0 to
% 3 crn in 1 nun increments
betta=2*pi*f/vp; % compute propagation constant
Gannna=(ZL-ZO)/(ZL+ZO); % compute load reflection coefficient
rd=abs(Gamna); % magnitude of the reflection coefficient
alpha=angle(Gamma)-2*betta*d; % phase of the reflection
% coefficient
plot(rd*cos(alpha),rd*sin(alpha)); % plot the graph

In the first line of the MATLABcode (see file fig3-9.m on the accompanying CD)
we generate the Smith Chart with the necessary resistance and reactance circles. The
next lines define the characteristic line impedance Zo = 50 Q , load impedance
impedance Transformation 121

9
Z, = (30 + j60) L2 , operation frequency f = 2 x 10 Hz, and phase velocity
v, = 0.5 x 3 x 10 m/s . The command line d=o .o :o .0 0 1 :o . o 3 creates an array d rep-
8

resenting the transmission line length, which is varied from 0 mm to 3 cm in 1-mm

increments. After all parameters have been identified, the magnitude and phase of the
input reflection coefficients have to be computed. This is accomplished by determining
the propagation constant P = 2 nf /vp, load reflection coefficient
To = (ZL- Zo)/(ZL + Z,) and its magnitude ITo[, and the total angle of rotation
a = L(T,) - 2 p d . Finally, the display of the impedance as part of the Smith Chart is
done through the plot command, which requires both real and imaginary phasor argu-
ments ITo(cos(a) and Ir,l sin(a) . The final result is shown in Figure 3-9.

Figure 3-9 Input impedance of a loaded line of 2 cm length for a sweep in

operating frequency from 0.0 to 3 GHz. If the operating frequency is fixed at 2 GHz
and the line length is varied from 0.0 to 3 cm, the same impedance curve is obtained.

For the case where the length of the line is fixed to be 2 cm and the frequency is
swept from values ranging from 0.0 to 3 GHz, the only necessary modification to the
above input file is to set d=o. 02, followed by specifying the frequency range in incre-
ments of 100 MHz (i.e., f=o.o: le7:3e9). We should note that in both cases the electri-
cal length ( p d ) of the line changes from 0" to 144". Therefore, the impedance graphs
produced for both cases are identical.
At the end of the rotation, either by fixing the frequency and varying the length or
vice versa, the input impedance is found to be Zi, = ( 12.4 + j15.5) 52. It is reassuring
that for a fixed frequency f = 2 GHz and a line length range d = 0 . . . 2 cm, we ulti-
 - - - -

Chapter 3 The Smlth Chart

mately arrive at the same input impedance of Zin = (14.7 - j26.7) s2 as obtained in
Example 3-2.

3.3 Admittance Transformation

3.3.1 Parametric Admittance Equation

From the representation of the normalized input impedance (3.4),it is possible to
obtain a normalized admittance equation by simple inversion:

where Y o = l/Zo. To represent (3.23) graphically in the Smith Chart, we have several
options. A very intuitive way of displaying admittances in the conventional Smith Chart
or 2-Smith Chart is to recognize that (3.23)can be found from the standard represen-
tation (3.4)via

In other words, we take the normalized input impedance representation and multiply
-jz
the reflection coefficient by -1 = e , which is equivalent to a 180' rotation in the
complex r -plane.

Y , & M w
Example 3-6: Use of the Smith Chart for converting imped-
ance to admittance

Convert the normalized input impedance zin = 1 + j 1 = &e i(zA)

into normalized admittance and display it in the Smith Chart.

Solution: The admittance can be found by direct inversion, that is

In the Smith Chart we simply rotate the reflection coefficient corre-

sponding to zin by 180' to obtain the impedance. Its numerical
value is equal to yin as shown in Figure 3-10. To denormalize yi,
 Figure 3-10 Conversion from impedance to admittance by 180" rotation.

we multiply by the inverse of the impedance normalization factor.

Thus,

Rotations by 180 degrees to convert from the impedance to the

admittance representation require only a rejection about the origin
in the r-plane.

In addition to the preceding operation, there is a widely used additional possibility.

Instead of rotating the reflection coefficient by 180' in the 2-Smith Chart, we can
124 Chapter 3 The Smith Chart

rotate the Smith Chart itself. The chart obtained by this transformation is called the
admittance Smith Chart or the Y-Smith Chart. The correspondences are such that
normalized resistances become normalized conductances and normalized reactances
become normalized susceptances. That is,

and

This reinterpretation is depicted in Figure 3- 11 for a particular normalized impedance

point z = 0.6 + j1.2.

(a) Z-Smith Chart (b) %Smith Chart

Figure 3-1 1 Reinterpretation of the Z-Smith Chart as a Y-Smith Chart.

As seen in Figure 3- 11, the transformation preserves (a) the direction in which the
angle of the reflection coefficient is measured and (b) the direction of rotation (either
toward or away from the generator). Attention has to be paid to the proper identification
of the extreme points: A short-circuit condition zL = 0 in the Z-Smith Chart is
yL = 00 in the YSmith Chart, and conversely an open-circuit zL = in the Z-Smith
Chart is y , = 0 in the YSmith Chart. Furthermore, negative values of susceptance are
plotted now in the upper half of the chart, corresponding to inductive behavior, and pos-
itive values in the bottom half, corresponding to capacitive behavior. The real compo-
nent of the admittance increases from right to left.
Adrnlttance Transformation 125

To complete our discussion of the Y-Smith Chart, we should mention an addi-

tional, often employed definition of the admittance chart. Here the admittance is repre-
sented in exactly the same manner as the impedance chart without a 180" rotation. In
this case the reflection coefficient phase angle is measured from the opposite end of the
chart (see the book by Gonzalez listed in Further Reading at the end of this chapter).

3.3.2 Additional Graphical Displays

In many practical design applications it is necessary to switch frequently from
impedance to admittance representations and vice versa. To deal with those situations a
combined, or so-called ZY-Smith Chart, can be obtained by overlaying the Z- and Y-
Smith Charts, as shown in Figure 3-12.

Figure 3-12 The ZY-Smith Chart superimposes the Z-and Y-Smith Charts in one
graphical display.
126 Chapter 3 The Smlth Chart

This combined ZY-Smith Chart allows direct conversion between impedances and
admittances. In other words, a point in this combined chart has two interpretations
depending on whether the Z-Chart or Y-Chart display is chosen.

C & M W
Example 3-7: Use of the combined ZY-Smith Chart

Identify (a) the normalized impedance value z = 0.5 + j0.5 and (b)
r__X#-C

the normalized admittance value y = 1 + j2 in the combined ZY-

Smith Chart and find the corresponding values of normalized admit-
tance and impedance.

Solution: Let us first consider the normalized impedance value

z = 0.5 + j0.5. In the combined ZY-Smith Chart we locate the
impedance by using circles of constant resistance r = 0.5 and con-
stant reactance x = 0.5, as shown in Figure 3-12. The intersection of
these two circles determines the specified impedance value
z = 0.5 + j0.5. To find the corresponding admittance value we
simply move along the circles of constant conductance g and sus-
ceptance b. The intersection gives us g = 1 and jb = -jl (i.e., the
admittance for part (a) of this example is y = 1 - jl ). The solution
for the normalized admittance y = 1 + j2 is obtained in identical
fashion and is also illustrated in Figure 3-12.

The ZY-Smith Chart requires a fair amount of practice due to

its "busy" appearance and the fact that inductors and capacitors
are counted either in positive or negative units depending on
whether an impedance or admittance representation is needed.

3.4 Parallel and Series Connections

In the following sections several basic circuit element configurations are analyzed
and their impedance responses are displayed in the Smith Chart as a function of fre-
quency. The aim is to develop insight into how the impedanceladmittance behaves over
a range of frequencies for different combinations of lumped circuit parameters. A prac-
Parallel and Series Connections

tical understanding of these circuit responses is needed later in the design of matching
networks (see Chapter 8) and in the development of equivalent circuit models.

3.4.1 Parallel Connection of Rand L Elements

Recognizing that g = Zo/R and bL = + Z o / ( o L ) , we can locate the normal-
ized admittance value in the upper Y-Smith Chart plane for a particular, fixed normal-
ized conductance g at a certain angular frequency oL:

As the angular frequency is increased to the upper limit a,, we trace out a curve along
the constant conductance circle g. Figure 3-13 schematically shows the frequency-
dependent admittance behavior for various constant conductance values g = 0.3, 0.5,
0.7, and 1 and for frequencies ranging from 500 MHz to 4 GHz. For a fixed inductance
value of L = 10 nH and a characteristic line impedance Zo = 50 a , the susceptance
always starts at -1.59 (500 MHz) and ends at -0.20 (4 GHz).
In Figure 3-13 and the following three additional cases, the transmission line
characteristic impedance is represented as a lumped impedance of Zo = 50 rR . This is
permissible since our interest is focused on the impedance and admittance behavior of
different load configurations. For these cases the characteristic line impedance serves
only as a normalization factor.

Figure 3-13 Admittance response of parallel RL circuit for o, Io I o, at

constant conductances g = 0.3, 0.5, 0.7, and 1.
128 Chapter 3 The Srnlth Chart

3.4.2 Parallel Connection of R and C Elements

Here we operate in the lower Y-Chart plane because susceptance bc = Z o o C
remains positive. To locate the normalized admittance value for a particular, fixed nor-
malized conductance g and angular frequency oLwe have

Figure 3-14 depicts the frequency-dependent admittance behavior as a function of vari-

ous constant conductance values g = 0.3, 0.5, 0.7, and 1. The normalized susceptance
for C = 1 pF and characteristic line impedance Zo = 50 i2 always starts at 0.16
(500 MHz) and ends at 1.26 (4 GHz).

Figure 3-14 Admittance response of parallel RC circuit for a, 5 o I o, at

constant conductances g = 0.3, 0.5, 0.7, and 1.

3.4.3 Series Connection of Rand L Elements

When dealing with series connections, we can conveniently choose the 2-Smith
Chart for the impedance display. Identifying the normalized reactive component as
x, = o L / Z o , it is straightforward to locate the normalized impedance value for a par-
ticular, fixed normalized resistance r at a given angular frequency a, :

In Figure 3-15 the frequency-dependent impedance behavior is shown as a function of

various constant resistance values r = 0.3, 0.5, 0.7, and 1. For the same inductance of
10 nH and characteristic line impedance of 50 i2 as used in Figure 3-13, we now pick
Parallel and Serles Connections 129

reactance circles associated with 0.63 (500 MHz) and with 5.03 (4 GHz). Because the
reactance is positive and since we use the 2-Smith Chart, all impedances have to reside
in the upper half plane.

Figure 3-15 Impedance response of series RL circuit for o, I 6.1I o, and

constant resistances r = 0.3, 0.5, 0.7, and 1.

3.4.4 Series Connection of R and C Elements

We again choose the 2-Smith Chart for the impedance display. The normalized
reactive component is xc = +1/(oCZo), indicating that all curves will reside in the
lower half of the Smith Chart. The normalized impedance value for a particular, fixed
normalized resistance r at an angular frequency oLreads
1
zin(aL) = r-j-
~ L C ~ O
Figure 3-16 displays the frequency-dependent impedance behavior as a function of var-
ious constant resistance values r = 0.3,0.5,0.7, and 1. The capacitance of 1 pF in series
with the variable resistance connected to a characteristic line impedance of 50 SZ now
yields circles associated with the reactances of -6.03 (500 MHz) and -0.8 (4 GHz),
which intersect with the four resistance circles, uniquely determining upper and lower
impedance values.
 Chapter 3 The Smlth Chart

Figure 3-1 6 Impedance response of series RC circuit for oLI o I o, at

constant resistances r = 0.3, 0.5, 0.7, and 1.

3.4.5 Example of a T-Network

In the previous examples only pure series or shunt configurations have been ana-
lyzed. In reality, however, one often encounters combinations of both. To show how
easily the ZY Chart allows transitions between series and shunt connections, let us
investigate by way of an example the behavior of a T-type network connected to the
input of a bipolar transistor. The input port of the transistor is modeled as a parallel RC
network as depicted in Figure 3-17. As we will see in Chapter 6, RL approximates the
base-emitter resistance and CLis the base-emitter junction capacitance. The numerical
parameter values are listed in Figure 3-17.

+ - '
T-type network Transistor
input
Figure 3-17 T network connected to the base-emitter input impedance of a
bipolar transistor.
Parallel and Series Connections

To use the Smith Chart for the computation of the input impedance of this more
complicated network, we first analyze this circuit at 2 GHz and then show the entire
response of the circuit for a frequency range from 500 MHz to 4 GHz by employing the
commercial MMICAD software simulation package.
To obtain the load impedance, or the input impedance of the transistor, we use the
Y-Smith Chart to identify the conductance point corresponding to the load resistor
RL = 3 1.25 R . Assuming a 50 R characteristic line impedance, we determine the
normalized admittance for this case to be gA = 1.6, which corresponds to point A in
Figure 3-18.

Figure 3-18 Computation of the normalized input impedance of the T network

shown in Figure 3-17 for a center frequency f = 2 GHz .
132 Chapter 3 The Smith Chart

The next step is to connect the capacitance C L = 1.91 F in shunt with the resis-
9 -P
tor R L . At the angular frequency of oL= 2x2 x 10 s , the susceptance of this
capacitor becomes BCL = o L C L = 24 mS , which corresponds to a rotation of the
original point A into the new location B. The amount of rotation is determined by the
normalized susceptance of the capacitor bcL = BcLZo = 1.2 and is carried out along
the circle of constant conductance in the YSmith Chart (see Figure 3-18).
Re-evaluating point B in the Z-Smith Chart, we obtain the normalized impedance
of the parallel combination of resistor RL and capacitor C L to be z, = 0.4 - j0.3 . The
series connection of the inductance L , results in the new location C . This point is
obtained through a rotation from xB = -0.3 by an amount xL, = o L L 1 / Z o = 1.1 to
xc = 0.8 along the circle of constant resistance r = 0.4 in the Z-Smith Chart as dis-
cussed in Section 3.4.3.
Converting point C into a Y-Smith Chart value results in yc = 0.5 - jl.O. The
shunt connected capacitance requires the addition of a normalized susceptance
bc = a C Z o = 1.5, which results in the admittance value of yD = 0.5 + j0.5 or
point D in the Y-Smith Chart. Finally, converting point D into the impedance value
zD = 1 - j 1 in the 2-Smith Chart allows us to add the normalized reactance
xL2 = w L L 2 / Z o = 1 along the constant r = 1 circle. Therefore, we reach z , = 1
or point E in Figure 3-18. This value happens to match the 50 L? characteristic trans-
mission line impedance at the given frequency 2GHz. In other words,
Zin = Z , = 5 0 Q .
When the frequency changes we need to go through the same steps but will arrive
at a different input impedance point z i n .It would be extremely tedious to go through
the preceding computations for a range of frequencies. This is most efficiently done by
the computer.
Relying on the previously mentioned CAD program MMICAD we are able to
produce a graphical display of the input impedance in the 2-Smith Chart over the entire
frequency range in preselected increments of 10 MHz, as shown in Figure 3-19. This
figure can also be generated as part of the MATLAB software (see file fig3-l8.m on the
accompanying CD).
We notice that the impedance trace ranging from 0.5 to 4 GHz is in agreement
with our previous calculations at 2 GHz. Also, as the frequency approaches 4 GHz, the
capacitor of C = 2.39 pF behaves increasingly like a short circuit in series with a single
inductor L 2 . For this reason, the normalized resistance r approaches zero and the reac-
tance grows to large positive values.
Summary

Figure 3-19 CAD simulation of the normalized input impedance q, for the
network depicted in Figure 3-17 over the entire frequency range
500 MHz I f 2 4 GHz.

3.5 Summary
This chapter has derived the Smith Chart as the most widely used RF graphical
design tool to display the impedance behavior of a transmission line as a function of
either line length or frequency. Our approach originated from the representation of the
normalized input impedance of a terminated transmission line in the form

which can be inverted in terms of the reflection coefficient to yield two circle equations
(3.10) and (3.11), which take on the following expressions for the normalized
resistance r:

and for the normalized reactance x

Superimposing the circles described by both equations over the complex polar form of
the normalized impedance z-plane on the unit circle yields the Smith Chart. The key
feature to remember is that one full rotation is equal to halfa wavelength because of the
134 Chapter 3 The Smith Chart

exponent 2pd in the reflection coefficient expression (3.2). In addition to observing the
impedance behavior, we can also quantify in the Smith Chart the degree of mismatch
expressed by the standing wave ratio (SWR) equation (3.12), or

which can be directly obtained from the chart.

To facilitate computer-based evaluation of the Smith Chart, a wide range of com-
mercial programs can be utilized. Due to its ease of implementation on a PC and its
user-friendly interface, throughout this book we have used the package MMZCAD
developed by Optotek. However, for the relatively incomplicated circuits analyzed in
this Chapter, one can also create a custom-tailored Smith Chart and perform simple
computations by relying on mathematical spreadsheets such as Mathematics, MATLAB,
or MathCad. To demonstrate the procedure, a number of MATLAB modules have been
developed, and the use of these so-called m.$les as part of a basic Smith Chart compu-
tation is demonstrated in Section 3.2.4.
A transition to the admittance, or Y-Smith Chart, can be made via (3.23):

and it is found that the only difference to (3.4) is a sign reversal in front of the reflection
coefficient. Consequently, rotating the reflection coefficient in the Z-Smith Chart by
180" results in the Y-Smith Chart. In practice, this rotation can be avoided by turning
the chart itself. Superimposing the rotated chart over the original Z-Smith Chart pro-
vides a combined ZY-Smith Chart display. The benefit of such a display is the easy tran-
sition from parallel to series connection in circuit designs. This ease is demonstrated by
a T-network configuration connected to the input port of a bipolar transistor consisting
of a parallel RC network. To investigate the impedance bghavior as a function of fre-
quency sweep, however, is most easily accomplished through the use of CAD programs.

Further Reading

G. Gonzalez, Microwave Transistor Amplijiers: Analysis and Design, 2nd ed., Prentice
Hall, Upper Saddle River, NJ, 1997.
K. C. Gupta, R. Garg, and I. J. Bohl, Microstrip Lines and Slotlines, Artech House,
Dedham, MA, 1979.
J. Helszajn, Passive and Active Microwave Circuits, John Wiley, New York, 1978.
Problems 135

Hewlett-Packard Application Note 154, "S-Parameter Design," 1972.

H. Howe, Stripline Circuit Design, Artech House, Dedham, MA, 1974.
S. Y. Liao, Microwave Devices and Circuits, Prentice Hall, Englewood Cliffs, NJ, 1980.
MMZCAD for Windows, Reference Manual, Optotek, Ltd., 1997.
D. M. Pozar, Microwave Engineering, 2nd edition, John Wiley, New York, 1998.
P. A. Rizzi, Microwave Engineering, Passive Circuits, Prentice Hall, Englewood Cliffs,
NJ, 1988.
P. H. Smith, "Transmission-Line Calculator," Electronics, Vol. 12, pp. 29-31, 1939.
P. H. Smith, "An Improved Transmission-Line Calculator," Electronics, Vol. 17, p. 130,
1944.

Problems

3.1 Consider a load ZL = (80 + j40) R connected to a lossy transmission line

with characteristic line impedance of
0.1 + j200
20 =0.05 - j0.003
Determine the reflection coefficient and the standing wave ratio (SWR) at
the load.

A coaxial cable of characteristic line impedance Zo = 75 R is terminated

by a load impedance of 2, = (40 + j35) R . Find the input impedance of
the line for each of the following pairs of frequency f and cable length d
assuming that the propagation velocity is 77% of the speed of light:
(a) f = 1 GHz and d = 50 cm
(b) f = 5 GHz and d = 25 cm
(c) f = 9 GHz and d = 5 cm

The attenuation coefficient of a transmission line can be determined by

shortening the load side and recording the VSWR at the beginning of the
line. We recall that the reflection coefficient for a lossy line takes on the form
T ( d ) = r0exp(-kl) = roexp(-al)exp(-jpl) . If the line is 100 m in
length and the VSWR is 3, find the attenuation coefficient a in Nplm, and
dB1m.
136 Chapter 3 The Smlth Chart

3.4 A load impedance of Z L = (150 - j50) R is connected to a 5 cm long

transmission line with characteristic line impedance of Zo = 75 R . For a
wavelength of 6 cm, compute
(a) the input impedance
(b) the operating frequency, if the phase velocity is 77% the speed of light
(c) the SWR

3.5 Identify the following normalized impedances and admittances in the Smith
Chart:
(a) z = 0.1 + j0.7
(b) y = 0.3 + j0.5
(c) z = 0.2 + j O . l
(d) y = 0.1 + j0.2
Also find the corresponding reflection coefficients and SWRs.

3.6 An unknown load impedance is connected to a 0.3h long, 50 R lossless

transmission line. The SWR and phase of the reflection coefficient measured
at the input of the line are 2.0 and -20°, respectively. Using the Smith Chart,
determine the input and load impedances.

3.7 In Section 3.1.3 the circle equation (3.10) for the normalized resistance r is
derived from (3.6). Start with (3.7); that is,

and show that the circle equation

can be derived.

3.8 Starting with the equation for normalized admittance

prove that the circle equations for the Y-Smith Chart are given by the follow-
ing two formulas:
(a) For the constant conductance circle as

(b) For the constant susceptance circle as

Problems 137

3.9 A lossless transmission line (Zo = 50 R ) is 10 cm long (f = 800 MHz,

v, = 0 . 7 7 ~ )If. the input impedance is Zin = j60 R
(a) Find Z L (using the Smith Chart)
(b) What length of a short-circuit transmission line would be needed to
replace ZL?

3.10 A transmission line of characteristic impedance Zo = 50 R and length

d = 0.151 is terminated into a load impedance of ZL = (25 - j30) R .
Find To, Zin(d), and the SWR by using the 2-Smith Chart.

3.1 1 A short-circuited 50 R transmission line section is operated at 1 GHz and

possesses a phase velocity of 75% of the speed of light. Use both the analyt-
ical and the Smith Chart approach to determine the shortest lengths required
to obtain (a) a 5.6 pF capacitor, and (b) a 4.7 nH inductor.

3.12 Determine the shortest length of a 75 R open-circuit transmission line that

equivalently represents a capacitor of 4.7 pF at 3 GHz. Assume the phase
velocity is 66% of the speed of light.

3.13 A circuit is operated at 1.9 GHz and a lossless section of a 50 R transmis-

sion line is short circuited to construct a reactance of 25 R . (a) If the phase
velocity is 314 of the speed of light, what is the shortest possible length of
the line to realize this impedance? (b) If an equivalent capacitive load of 25
R is desired, determine the shortest possible length based on the same phase
velocity.

3.14 A microstrip line with 50 R characteristic line impedance is terminated into

a load impedance consisting of a 200 R resistor in shunt with a 5 pF capac-
itor. The line is 10 cm in length and the phase velocity is 50% the speed of
light. (a) Find the input impedance in the Smith Chart at 500 MHz, 1 GHz,
and 2 GHz, and (b) use the MATLAB routine (see Section 3.2.4) and plot the
frequency response from 100 MHz to 3 GHz in the Smith Chart.

3.15 For an FM broadcasting station operated at 100 MHz, the amplifier output
impedance of 250 R has to be matched to a 75 S2 dipole antenna.
(a) Determine the length and characteristic impedance of a quarter-wave
transformer with v p = 0 . 7 ~ .
(b) Find the spacing D for a two-wire lossless transmission line with AWG
 Chapter 3 The Smith Chart

26 wire size and a polysterene dielectric (E, = 2.55 ).

Consider the case of matching a 73 SZ load to a 50 SZ line by means of a

h / 4 transformer. Assume the matching is achieved for a center frequency of
f = 2 GHz. Plot the SWR for the frequency range 1/3 I f / f 5 3 .

A line of characteristic impedance of 75 52 is terminated by a load consist-

ing of a series connection of R = 30 i2 , L = 10 nH , and C = 2.5 pF .
Find the values of the SWR and minimum line lengths at which a match of
the input impedance to the characteristic line is achieved. Consider the fol-
lowing range of frequencies: (a) 100 MHz, (b) 500 MHz, and (c) 2 GHz.

A 50 SZ lossless coaxial cable (E, = 2.8) is connected to a 75 SZ antenna

operated at 2 GHz. If the cable length is 25 cm, find the input impedance by
using the analytical equation (2.71) and the Z-Smith Chart.

A balanced to unbalanced (balun) transformation is often needed to connect

a dipole antenna (balanced) to a coaxial cable (unbalanced). The following
figure depicts the basic concept.
Problems 139

As an alternative of using a transformer, one often uses the following

antenna connection.

(a) Explain why one leg of the dipole antenna is connected a distance h / 4
away from the end of the coax cable.
(b) For an FM broadcast band antenna in the frequency range from 88 to
108 MHz, find the average length where the connection has to be made.

3.20 Using the ZYSmith Chart, find the input impedance of the following net-
work at 2 GHz.

What is the input impedance of this network at 1 GHz?

3.21 A Zo = 50R transmission line is 0.5h in length and terminated into a load
of Z, = (50 - j30)R . At 0.35 h away from the load, a resistor of
R = 25R is connected in shunt configuration (see figure below). Find the
input impedance with the help of the ZY Smith Chart.
140 Chapter 3 The Smith Chart

3.22 A 50-R transmission line of 314 wavelength in length is connected to two

transmission line sections each of 75 R in impedance and length of 0.86 and
0.5 wavelength, respectively, as illustrated in the following figure.

The termination for line 1 is Z1 = (30+ j40) R and Z2 = (75-j80)R

for line 2.Employ the Smith Chart and find the input impedance.

3.23 Repeat the previous problem if all characteristic line impedances are
Z, = 50 R and all transmission line sections are h/4 in length.

3.24 A dipole antenna of impedance ZL = (75+ j20) R is connected to a 50 R

lossless transmission whose length is h/3 . The voltage source VG = 25V
is attached to the transmission line via an unknown resistance RG. It is
determined that an average power of 3 W is delivered to the load under load-
match
side matching (ZL = 50 R). Find the generator resistance R G , and
determine the power delivered to the antenna if the generator impedance is
matched to the line via a quarter-wave transformer.
Problems 141

3.25 Determine the values of the inductance L and the capacitance C such that
they result in a 50 i2 input impedance at 3 GHz operating frequency for the
following network.
v

3.26 An open-circuit transmission line (50 a)is operated at 500 MHz (vp = 0.7~).
Use the ZY Smith Chart and find the impedance 2, if the line is 65 cm in
length. Find the shortest distance for which the admittance is Y, = -j0.05S.

3.27 Find the minimum line length l I and the minimum length of the short-cir-
cuited stub l2 in terms of wavelength A , such that the input impedance of
the circuit is equal to 50 i2 .

3.28 Find the input impedance in terms of magnitude and phase of the following
network at an operating frequency of 950 MHz.

3.29 Repeat your computation and solve Problem 3.28 for a 1.5 GHz operating
frequency. Comment on the differences in your results.
142 Chapter 3 The Smlth Chart

3.30 A specific transmission line configuration is as follows:

The characteristic line impedance for all three elements is Zo = 50 R . The

load impedance has a value of Z , = (20 + j40) R , and the electrical
lengths of the corresponding line segments are = 164.3", = 57.7',
and Q3 = 25.5'.
(a) Find the input impedance.
(b) Find the input impedance if transmission line segment O2 is open
circuit.
(This problem and Problem 3.27 become very important in Chapter 8, when
we discuss the problem of matching a particular load impedance to a desired
input impedance.)
168 Chapter 4 Single- and Multipart Networks

The computations are based on the circuit in Figure 4-1 1 with L = 1 nH,
C = 10 pF ,transmission line length of I = 5 cm , and phase velocity equal to 65% of
the speed of light. The transistor is described by the following set of values:
rBE = 520 R , rm = 8Ok R , C,, = 10 pF, CBC = 1 pF, and g m = 0.192 s .

4.4 Scattering Parameters

In almost all databooks and technical literature regarding RF systems, the scatter-
ing or S-parameterrepresentation plays a central role. This importance is derived from
the fact that practical system characterizations can no longer be accomplished through
simple open- or short-circuit measurements, as it is customarily done in low-frequency
applications and as discussed at the beginning of this chapter. We should recall what
happens when we attempt to create a short circuit with a wire: The wire itself possesses
an inductance that can be of substantial magnitude at high frequency. Also, the open
circuit leads to capacitive loading at the terminal. In either case, the opedshort-circuit
conditions needed to determine Z-, Y-, h-, and ABCD-parameters can no longer be guar-
anteed. Moreover, when dealing with wavepropagation phenomena, it is not desirable
to introduce a reflection coefficient whose magnitude is unity. For instance, the terminal
discontinuity will cause undesirable voltage and/or current wave reflections, leading to
oscillations that can result in the destruction of the device. With the S-parameters, the
RF engineer has a tool to characterize the two-port network description of practically
all RF devices without requiring unachievable terminal conditions or causing harm to
the device under test (DUT).

4.4.1 Definition of Scattering Parameters

Simply put, S-parameters are power wave descriptors that permit us to define the
input-output relations of a network in terms of incident and reflected power waves.
With reference to Figure 4-14 we define an incident normalized power wave a, and a
reflected normalized power wave b, as follows:

1
- -z ~ z n )
b, = 2fio(vn (4.36b)

where the index n refers either to port number 1 or 2. The impedance Zo is the charac-
teristic impedance of the connecting lines on the input and output side of the network.
Under more general conditions the line impedance on the input side can differ from the
Scattering Parameters 169

line impedance on the output side. However, for our initial discussion, we will keep
things simple and assume that both impedances are the same.

Figure 4-14 Convention used to define Sparameters for a two-port network.

Inverting (4.36) leads to the following voltage and current expressions:

V, = &(a, + b,) (4.37a)

1
I = -(an-b,J (4.37b)
"
The physical meaning of (4.36)becomes clear when we recall the equations for power:

Isolating forward and backward traveling wave components in (4.37), we immediately

see

which is consistent with the definitions (4.37) since

V, = V; + V, = Z,I; - ZoI, (4.40)

Based on the directional convention shown in Figure 4-14 we are now in a position to
define the S-parameters:

where the terms are

170 Chapter 4 Single and Multlport Networka

Sll = - - reflected power wave at port 1

incident power wave at port 1

- transmitted power wave at port 2 (4.42b)

incident power wave at port 1

reflected power wave at port 2

incident power wave at port 2
a, = O

transmitted power wave at port 1

incident power wave at port 2

We observe that the conditions a2 = 0 and a l = 0 imply that no power waves are
returned to the network at either port 2 or port 1. However, these condition can only be
ensured when the connecting transmission lines are terminated into their characteristic
impedances.
Since the S-parameters are closely related to power relations, we can express the
normalized input and output waves in terms of time averaged power. With reference to
Section 2.10.2 we note that the average power at port 1 is given by

where the reflection coefficient at the input side is expressed in terms of Sll under
matched output according to the following argument:

This also allows us to redefine the VSWR at port 1 in terms of S1 as

1 + IS111
VSWR = -
1 - Is111
Furthermore, based on (4.39a) we can identify the incident power in (4.43) and express
it in terms of a, :
ScatWlng Parameters 171

which is the maximal available power from the generator. Using (4.46) and (4.44) in
(4.43) finally gives us the total power at port 1 (under matched output condition)
expressed as a combination of incident and reflected powers:

If the reflection coefficient, or SI1,is zero, all available power from the source is deliv-
ered to port 1 of the network. An identical analysis at port 2 yields

4.4.2 Meaning of SParameters

As already mentioned in the previous section, the S-parameters can only be deter-
mined under conditions of perfect matching on the input or output side. For instance, in
order to record Sll and S21 we have to ensure that on the output side the line imped-
ance Zo is matched for a, = 0 to be enforced, as shown in Figure 4-15.

Figure 4-15 Measurement of S,l and &,by matchin the line impedance 4 at
8
port 2 through a corresponding load impe ance Z,= 4.

This configuration allows us to compute Sll by finding the input reflection

coefficient:
n - 20
sll = rin= i'-
'in +O'

In addition, taking the logarithm of the magnitude of Sll gives us the return loss in dB
RL = -2010g(S1,( (4.50)
Moreover, with port 2 properly terminated, we find
 Chapter 4 Single- and Multlport Networks

Since a2 = 0,we can set to zero the positive traveling voltage and current waves at
port 2. Replacing V1 by the generator voltage VG1 minus the voltage drop over the
source impedance Zo , VG1 - ZoI gives

Here we observe that the voltage recorded at port 2 is directly related to the generator
voltage and thus specifies the forward voltage gain of the network. To find the for-
ward power gain, we square (4.52) to obtain

If we reverse the measurement procedure and attach a generator voltage VG2 to

port 2 and properly terminate port 1, as shown in Figure 4-16, we can determine the
remaining two S-parameters, S22 and S I 2 .

Figure 4-16 Measurement of S,, and S,, by matching the line impedance 4 at
port 1 through a corresponding input impedance Z, = 4.

To compute S22 we need to find the output reflection coefficient To,, in a similar
way as already discussed for S I 1 :
Zout - Zo
S22 = rout =
Zout + zo
and for SI2
Scattering Parameters 173

The term S12 can further be manipulated through the substitution of V2 by

VG2 - Z0Z2,leading to the form

2
known as the reverse voltage gain and whose square ISl21 is identified as reverse
power gain. While determining Sll and S22 can be directly computed as part of the
impedance definitions, SI2 an S2, require the replacement of the defining voltages by
the appropriate network parameters. In the following example, the S-parameters are
computed for a simple, three element network.

, v m w
Example 4-7: Determination of a T-network elements

Find the S-parameters and the resistive elements for the 3 dB attenu-
ator network shown in Figure 4-17(a) assuming that the network is
placed into a transmission line section with a characteristic line
impedance of Zo = 50 R .

Solution: An attenuator should be matched to the line imped-

ance and must therefore meet the requirement SI1 = S22 = 0 . As a
result, based on Figure 4-17(b) and consistent with (4.49), we set

Because of symmetry, it is immediately clear that R1 = R2. We

now investigate the voltage V2 = V i at port 2 in terms of
V1 = v:. According to the circuit configuration shown in Figure
4-17(c), the following expression is obtained
174 Chapter 4 Single- and Multipart Networks

Port 1 Port 2
(c)
Figure 4-17 S-parameter computation for a T-network. (a) circuit diagram;
(b) circuit for S,, and S, measurements; (c) circuit for S, and S,, measurements.

For a 3 dB attenuation, we require

Setting the ratio of V2/V1 to 0.707 in the preceding equation

allows us, in combination with the input impedance expression, to
determine R1 and R3 .After simplification it is seen that

$2-1
R1 = R2 = -Zo = 8.58 S2 and Rj = 2 f i . Zo = 141.4 R
JZ+1

The choice of the resistor network ensures that at the input and
output ports an impedance of 50 R is maintained. This implies that
this network can be inserted into a 50 R transmission line section
without causing undesired rejections, resulting in an insertion loss.
Scattering Parameters 175

The definitions for the S-parameters require appropriate termination. For instance,
if S,, is desired, the transmission line connected to port 2 has to be terminated into its
characteristic line impedance. This does not necessarily mean that the output impedance
Z,, of the network has to be matched to the line impedance Zo. Rather, the line imped-
ance must be matched to ensure that no wave is reflected from the load, as implied by
a2 = 0 . If this is not the case, we will see in Section 4.4.5 how Sll is modified.

4.4.3 Chain Scattering Matrix

To extend the concept of the S-parameter representation to cascaded networks, it
is more efficient to rewrite the power wave expressions arranged in terms of input and
output ports. This results in the chain scattering matrix notation. That is,

It is immediately seen that the cascading of two dual-port networks becomes a simple
multiplication. This is apparent in Figure 4-18, where network A (given by matrix [TIA)
is connected to network B (given by matrix [TIB).

Figure 4-18 Cascading of two networks A and 9.

If network A is described by the relation

and network B by

we notice, based on the parameter convention shown in Figure 4-18, that

 Chapter 4 Slngle and Multipart Networks

(4.59)

Thus, for the combined system, we conclude

which is the desired matrix multiplication. Therefore, the chain scattering matrix plays
a similar role as the ABCD-matrix discussed earlier.
The conversion from the S-matrix to the chain matrix notation follows identical
steps as outlined in Section 4.3.1. In particular, to compute T l for instance, we see that

Similarly,

Conversely, when the chain scattering parameters are given and we need to convert to
S-parameters, we find the following relations:
Scattering Parameters 177

Alternatively, a matrix manipulation as discussed in the next section could have been
carried out with the same result.

4.4.4 Conversion between Z-and SParameters

We have already seen how certain S-parameters can be defined in terms of input
and output impedances of a network [i.e., equations (4.49) and (4.54)]. In this section,
we go through a formal conversion between the Z- and S-parameter sets. Once this
interrelation is established, we are able to formulate conversion links between all six
network parameter sets (S, 2, I:ABCD, h, T).
To find the conversion between the previously defined S-parameters and the Z-
parameters, let us begin with the defining S-parameter relation in matrix notation [i.e.,
(4.4111
{b} = [Sl{a} (4.69)
Multiplying by fi0gives
fl0{b1 = WI = fl0[sl{a} = [sI{v+) (4.70)
Adding {v'} = fio{a} to both sides results in

where [El is the identity matrix. To compare this form with the impedance expression
{V} = [Z]{I} , we have to express {v'} in terms of {I}. This is accomplished by
first subtracting [s]{v+} from both sides of {v'} = &{a}; that is,

{v+}- ISI{V+I = &({a} - { b ~ )= zo{I} (4.72)

Now, by isolating {v+},it is seen that

Substituting (4.73) into (4.71) yields the desired result of

178 Chapter 4 Single and Multiport Networks

Identifying individual terms is now easily carried out. A complete summary of all net-
work coefficient sets is given in Appendix C.

4.4.5 Signal Flow Chart Modeling

The analysis of RF networks and their overall interconnection is greatly facilitated
through signal flow charts as commonly used in system and control theory. As origi-
nally introduced to seismology and remote sensing, wave propagation can be associated
with directed paths and associated nodes connecting these paths. Even complicated net-
works are easily reduced to input-output relations in which the reflection and transmis-
sion coefficients play integral parts. In this section we will briefly summarize key
principles needed for a signal flow network analysis.
The main concepts required to construct flow charts are as follows:
1. Nodes that are deployed to identify network parameters such as al, b l , a2,b,
when dealing with S-parameters
2. Branches that are needed when connecting the network parameters
3. Addition and subtraction of branch values in accordance with the directions of the
branches
We will now discuss these three items in detail. To this end let us consider a section
of a transmission line that is terminated in a load impedance Z , , as seen in Figure 4-19.

b+
(a) (b)
Figure 4-19 Terminated transmission line segment with incident and reflected S-
parameter description. (a) Conventional form, and (b) Signal flow form.
Scattering Parameters

Even though we could use voltage values as node identifier, it is the S-parameter
representation that finds widespread use. In Figure 4-19(b) the nodes a and b are con-
nected through the load reflection coefficient T L .This makes sense since the reflection
coefficient is the ratio b/a, so that it simply states that node b is found as a result of mul-
tiplying node a by T , . This is depicted in generic form in Figure 4-20.

(a) Source node a, which launches wave. (b) Sink node b, which receives wave.

(c) Branch connecting source and sink.

Figure 4-20 Generic source node (a), receiver node (b), and the associated (c)
branch connection.

In terms of notation, we can encode the situation shown in Figure 4-20 as

b = Fa (4.77)
A more complicated situation arises when we need to make the transmission line
circuit shown in Figure 4-19 more realistic by including a source term, as seen in Fig-
ure 4-21.

(a) (b) (c)

Figure 4-21 Terminated transmission line with source. (a) conventional form, (b)
signal flow form, and (c) simplified signal flow form.

Unlike Figure 4-19, the nodes a and b are preceded by two additional nodes that
we shall denote a' and b'. The ratio b'/a' defines the source reflection coefficient Tsas
already discussed in Section 2.1 1. Here we also see that b' is given by multiplying a'
with the source reflection coefficient. By relying on the concept of summation, we
define b' as the sum of bs and a'Ts. Thus, the source bs is
bs = b' - a'T, (4.78)
180 Chapter 4 Single- and Muitiport Networks

An explicit expression for bs is obtained by noting that

v, = V, + I,ZG (4.79)
based on an outflowing current convention (see Figure 4-21). This can be converted into
the form

Rearranging terms and division by so gives

When comparing (4.81) with (4.78), we immediately see that

An important conclusion can be drawn when expressing a' in (4.78) by rLb' SO that we
obtain

This is a known as a self- or feedback loop (see Figure 4-22), which allows us to repre-
sent the nodes bs and b' by a single branch whose value is given by (4.83).

Figure 4-22 A self-loop that collapses to a single branch.

All signal flow chart principles can therefore be reduced to six building blocks, as
summarized in Table 4-3.
By way of an example, let us analyze a more complicated RF circuit consisting of
a sourced and terminated dual-port network.
--

Scattering Parameters

Table 4-3 Signal flow chart building blocks

F
Description Graphical Representation

Nodal Assignment

L Branch

Series Connection

Parallel Connectior

Splitting of Branche

Self-loop

c & M w
Example 4-8: Flow chart analysis of a dual-port network

For the network shown in Figure 4-23 find the ratios of b l / a l and
a l / b s . Assume unity for the multiplication factor of the transmis-
sion line segments.
182 Chapter 4 Single and Multipart Networks

(a) Circuit representation

b, 1 1 a , S2, b2 1

OA:]-IT]G
1 b, S,, a2 1
(b) Signal flow chart
Figure 4-23 Sourced and terminated two-port network.

Solution: The process of setting up the individual ratios is

explained best by going through a step-by-step simplification for the
ratio a /bs employing the rules summarized in Table 4-3. Figure 4-
24 depicts the five steps.
Step 1: Splitting of the rightmost loop between b2 and a 2 , leading
to the self-loop S22rL
Step 2: Decomposition of the self-loop between branches al and b2,
resulting in the multiplication factor S2,/(1 - S22rL),which can be
combined with TL and S,,
Step 3: Series and parallel connections between a and bl , leading
to the input reflection coefficient

Step 4: Splitting the loop into a self-loop, resulting in the multiplica-

tion factor

Step 5: Decomposition of the self-loop at a , , leading to the

expression
!Scattering Parameters 183

Step 1

S21 .............
bs 1 1 a,!'1 - SZ2rLb2 ]

......a? .........1
bl!......S.!?
Step 2

Step 3

Step 4

Step 5
Figure 4-24 Step-by-step simplification to determine the ratio a , / b ,
 - - ---

Chapter 4 Slngle- and Muklport Network.

Rearranging and simplification leads to the final form:

The preceding derivation follows a pattern similar to jnding

the transferfunction of a control system or a signal processox Even
complicated circuits can be reduced eficiently and quickly to estab-
lish the nodal dependcies.

The preceding example points out what will happen if the matching condition for
recording the S-parameters is not satisfied. As we know, if we compute S1, we need to
ensure that a2 = 0 . However, if a2 f 0 , as is the case in the preceding example, we see
that Sl is modified by the additional factor S12S21 r L / ( 1 - S22rL).

4.4.6 Generalization of SParameters

In our discussion thus far it was assumed that the characteristic line impedance at
both ports has the same value Zo . However, this does not have to be the case. Indeed, if
we assume that port 1 is connected to line impedance Zol and port 2 to impedance
ZO2, we have to represent the voltage and current waves at the respective port
(n = 1 , 2 ) a s

V, = V; + Vi = K n ( a , + b,) (4.84)
and

where we immediately observe

Scattering Parameters 185

These equations allow the definition of the S-parameters as follows:

When compared to the previous S-parameter definitions, we notice that scaling by the
appropriate line impedances has to be taken into account. It should also be apparent that
although the focus of our derivations was a two-port network, the preceding formulas
can be extended to an N-port network where n = 1, . . . ,N .
A second consideration is related to the fact that practical measurements involve
the determination of the network S-parameters through transmission lines of finite
length. In this case we need to investigate a system where the measurement planes are
shifted away from the actual network, as depicted in Figure 4-25.

Figure 4-25 Two-port network with finite-length transmission line segments.

An incident voltage wave launched from the power supply will have to travel a
distance lI in order to reach port 1. Consistent with the notation introduced in
Section 2.9, we note that at port 1 the incident voltage is given as

and, at the generator side, as

The reflected voltage wave at port 1 can be cast in the form

and
186 Chapter 4 Single- and Multlport Networks

where, as usual, pl stands for the lossless propagation constant of line 1. In an identical
fashion, the voltage behavior at port 2 can be formulated by simply replacing Vin in
terms of Vout and Vl in terms of V2 as well as p, in terms of p2. The preceding equa-
tions can be combined in matrix form

which links the impinging waves at the network ports to the corresponding voltages
shifted by the electric lengths of the attached transmission line segments. For the
reflected voltage waves we get the matrix form

As the discussion in Section 4.4.1 taught us, the S-parameters are linked to the coeffi-
cients a, and b,, which in turn can be expressed through voltages (if we assume
z,, = z,, 1.

It is apparent that if transmission line segments are added, we have to replace the above
voltages by the previously derived expressions, leading to the form

This final reveals that the S-parameters for the shifted network are comprised of three
matrices. In terms of the coefficients, we see that

The physical meaning of this form is easy to understand. The first matrix coefficient
reveals that we have to take into account 2Plll or twice the travel time for the incident
voltage to reach port 1 and, upon reflection, return. Similarly, for port 2 we see that the
Scattering Parameters 187

phase shift is 2P212. Moreover, the cross terms, which are closely related to the for-
ward and reverse gains, require the additive phase shifts associated with transmission
line 1 (P1ll ) and transmission line 2 (P212), since the overall inputloutput configura-
tion now consists of both line segments.

Y , m w
Example 4-9: Input impedance computation of a transmission
line based on the use of the signal flow chart

A lossless transmission line system with characteristic line imped-

ance Z, and length 1 is terminated into a load impedance ZL and
attached to a source voltage VG and source impedance ZG, as
shown in Figure 4-26. (a) Draw the signal flow chart and (b) derive
the input impedance formula at port 1 from the signal flow chart rep-
resentation.

Figure 4-26 Transmission line attached to a voltage source and terminated by a

load impedance.

Solution: (a) Consistent with our previously established signal

flow chart notation, we can readily convert Figure 4-26 into the form
seen in Figure 4-27.

Figure 4-27 Signal flow chart diagram for transmission line system in
Figure 4-26.

(b) The input reflection coefficient at port 1 is given by

188 Chapter 4 Slngk and Multlport N.twork.

-j2pI
bl = TLe al
which is exactly in the form given in Section 3.1, with TL = To and
1 = d.Thus
-j2pI
= -i'n - ZO
rin(l) = r L e
i'n + ZO
Solving for Zin yields the final result

This example shows how the input impedance of a transmis-

sion line can be found quickly and elegantly by using signal $ow
chart concepts.

4.4.7 Practical Measurements of SParameters

Measurement of the S-parameters of a two-port network requires reflection and
transmission evaluations of traveling waves at both ports. One of the most popular
methods is to use a vector network analyzer. The vector network analyzer is an instru-
ment that can measure voltages in terms of magnitude and phase. Usually network ana-
lyzers have one output port, which provides the RF signal either from an internal source
or an external signal generator, and three measurement channels, which are denoted as
R, A, and B (see Figure 4-28).
The RF source is typically set to sweep over a specified frequency range. The
measurement channel R is employed for measuring the incident wave. Channel R also
serves as a reference port. Channels A and B usually measure the reflected and transmit-
ted waves. In general, the measurement channels A and B can be configured to record
any two parameters with a single measurement setup. An example of the test arrange-
ment that allows us to measure Sll and S2, is shown in Figure 4-28.
In this case the value of Sll can be obtained by evaluating the ratio A / R , and
S21 through computing B / R . To measure S12 and S22 we have to reverse the DUT. In
Figure 4-28 the dual-directional coupler allows the separation of the incident and
reflected waves at the input port of the DUT. The bias tees are employed to provide nec-
essary biasing conditions, such as a quiescent point for the DUT. Since the most com-
Scattering Parameters 189

50 Q

Bias tee Bias tee

Dual-Directional Dual-Directional
Coupler Coupler
I 7

Figure 4-28 Measurement system for S l l and S21 parameters using a network
analyzer.

mon use of network analyzers is the characterization of two-port devices, bias tees,
directional couplers, and necessary electronic switches as well as the RF sweep signal
generator are all integral parts of most modem analyzers.
As we can see, a practical test arrangement is more complicated when compared
with the simple ideal system described in Sections 4.4.4 and 4.4.6, where we assume
that the DUT is connected to perfectly matched transmission lines of equal (Section
4.4.4) or unequal (Section 4.4.6) characteristic impedance. In a realistic measurement
system we cannot guarantee either matching conditions or ideality of the components.
In fact, we have to consider all effects of the external components connected to the
input and output ports of the DUT. Furthermore, the primary reference plane for mea-
surements of complex voltages, which are then converted into S-parameters, is usually
somewhere inside of the networks analyzer. As a result, it is necessary to take into
account not only attenuation and phase shifts due to the external components, but also
portions of the internal structure of the network analyzer itself.
In general, the measurement test arrangement can be reduced to the cascade of
three networks depicted in Figure 4-29.
190 Chapter 4 Single- and Multipart Nedworks

- Error
i b o x A
.4
i
DUT
p
i
'
Error
b o x B i
g++ tH,
Measurement
Reference Plane
2 i
Desired Reference Plane
b2 ~easurement
Reference Plane

Figure 4-29 (a) Block dia ram of the setup for measurement of S-parameters of
a two-port network; (by signal flow chart of the measurement test setup.

In Figure 4-29 the signals R, A, B correspond to the reference port and channels A
and B of the network analyzer. RFi, is the output line from the signal source. The
branch denoted Ex represents possible leakage between the output of the signal source
and the channel B.
The network analyzer treats everything between the measurement reference
planes as a single device. Therefore, our task is reduced to finding a way to calibrate the
network analyzer in such a way that it becomes possible to eliminate the effect of all
undesired influences or parasitics. The main goal of a calibration procedure is to char-
acterize the error boxes prior to measuring the DUT. This information can then be used
by an internal computer to evaluate the error-free S-parameters of the actual DUT.
Assuming that the error boxA network is reciprocal, we can state E12 = E21.
Therefore, we have to find six parameters (Ell, E12,E22,EX,ER,and ET) to character-
ize the error boxes.
The simplest calibration method involves three or more known loads (open, short,
and matched). The problem with this approach is that such standards are usually imper-
fect and are likely to introduce additional errors into the measurement procedures.
These errors become especially significant at higher frequencies. To avoid the depen-
dency on the accuracy of calibration standards, several methods have been developed
(see Eul and Schiek and Engen and Hoer, listed in the Further Reading section at the
Scattering Parameters

end of this chapter). In this section we will only consider the so-called Through-
Reflect-Line (TRL) technique (see Engen and Hoer).
The TRL calibration scheme does not rely on known standard loads. Instead, it is
based on the use of three types of connections, which are shown in Figure 4-30.

-4
RE;;" E21 R

El2 1
1

(a) Through
F ET

12

(b) Reflect

(c) Line
Figure 4-30 Signal flow graphs of TRL method: (a) Through, (b) Reflect, (c) Line
configurations.

The Through connection is made by directly connecting ports 1 and 2 of the DUT.
Next, the Reject connection uses a load with high reflectivity. The reflection coefficient
does not have to be known because it will be determined during the calibration process.
The only requirement is that the load possesses the same reflection coefficient for both
input and output ports. The Line connection is made by connecting ports 1 and 2 via a
transmission line matched to the impedance of the error boxes. Usually, this impedance
192 Chapter 4 Single- and Multipart Networks

is close to 50 0.Before we continue with the actual analysis of each particular con-
nection type, let us first consider the system as a general two-port network.
From Figure 4-29(b) it is seen that the signal at node B is a linear combination of
the input RF signal and the signal at node F:

Applying the self-loop rule, we can write that signal at node F as

To compute the signal at port R, the same method as discussed in Example 4-8 can
be used. In this example we first replaced the loop with the signal F through a self-loop
and then performed the same transformation for the signal R. The result of these com-
putations is

Substituting (4.99) into (4.98) followed by the substitution of (4.98) into (4.97), we
obtain an expression for signal B:

Finally, the value for the signal at node A is obtained by using the summation rule:

If the measurement system does not introduce any errors, then E12 = E21 = ET = 1
and E l l = E22 = ER = EX = 0 . Substituting these values into (4.99), (4. loo), and
(4.101), we find that R = 1 , A = S l , and B = S12,which shows the validity of the
formulas.
Now we are ready to investigate the TRL connections in more detail. To avoid
confusion, let us denote the measured signals R, A, and B for Through by subscript 7',
for Reflect by R, and for Line by L.
Scattering Parameters 193

For the Through connection we know that Sll = S2, = 0 and S12 = S21 = 1 .
Setting E = E21 it follows that

For the Reject connection we have Sll = S22 = r and S12 = S21 = 0 . This results
in the equations

BR = Ex (4.103~)
Finally, for the Line connection we see that Sll = S22 = 0 and S12 = S21 = e -71 ,
where 1 is the transmission line length and y is a complex propagation constant
(y = a + j p ) that takes into account attenuation effects. The result is

Equations (4.102a)-(4.104b) allow us to solve for the unknown coefficients of the error
boxes E l l , E12, E22, EX, ER, ET, the reflection coefficient T , and the transmission
line parameter e-y'. Knowing the error coefficients we are then in a position to process
the measured data in order to obtain an error-free S-parameter set of the DUT.
194 Chapter 4 Singk and Multiport Networks

4.5 Summary
Networks play an integral part in analyzing basic low-frequency circuits as well as
RF/MW circuits. For instance, the admittance or Y-matrix for an N-port network can be
written in generic form as

where currents and voltages become the defining external port conditions. The evalua-
tion of the matrix coefficients is accomplished through appropriate terminal conditions:

The concepts of Z-, Y-, h-, and ABCD-matrix representations of networks can be
directly extended to high-frequency circuits. Unfortunately, we encounter practical dif-
ficulties in applying the required open- and short-circuit network conditions needed
when defining the respective parameter sets. It is for this reason that the scattering
parameters as normalized forward and backward propagating power waves are
introduced:

For a two-port network this results in the matrix form

Unlike open- or short-circuit network conditions, impedance line matching at the

respective port is now required to establish the S-matrix set. The S-parameters can be
directly related to the reflection coefficients at the input and output of the two-port
network (SI1,S22). Furthermore, forward and reverse power gains are readily identified
ls1212 1.
(1~211~9
Further Readlng 195

The S-parameters are also very useful descriptors when dealing with signal flow
diagrams. A signal flow diagram is a circuit representation involving nodes and paths
for the sourced and terminated transmission line as follows:

With signal flow diagrams even complicated systems can be examined in terms of spe-
cific input output relations in a similar manner as done in control system theory.
Chapter 4 finishes with a brief discussion of the practical recording of the S-
parameters for a two-port network (DUT) through the use of a vector network analyzer.
To compensate for various error sources associated with the measurement arrangement,
the so-called TRL method is presented. Here the Through, Reject, and Line calibrations
are shown to account for the various errors and therefore permit the recording of the
actual S-parameters needed to characterize the DUT.

Further Reading
C. Bowick, RF Circuit Design, Howard Sams & Co., Indianapolis, IN, 1982.
R. S. Elliot, An Introduction to Guided Waves and Microwave Circuits, Prentice Hall,
Upper Saddle River, NJ, 1997.
G. E Engen and C. A. Hoer, "Thru-Refect-Line: An Improved Technique for Calibrat-
ing the Dual Six-Port Automatic Network Analyzer," IEEE Trans. Microwave Theory
and Techniques, Vol. MTT-27, pp. 987-998, 1979.
H.J. Eul and B. Schiek, "Thru-Match-Reflect: One Result of a Rigorous Theory for De-
embedding and Network Analyzer Calibration," Proceedings of the 18th European
Microwave Conference, Stockholm, Sweden, 1988.
G. Gonzales, Microwave TransistorAmplijers, 2nd ed. Prentice Hall, Upper Saddle
River, NJ, 1997.
S-Parameter Design, Hewlett-Packard Application Note 154, 1972.
D. V. Morgan and M. J. Howes, eds., Microwave Solid State Devices and Applications,
P. Peregrinus Ltd., New York, 1980.
P. A. Rizzi, Microwave Engineering-Passive Circuits, Prentice Hall, Upper Saddle
River, NJ, 1988.
D. Roddy, Microwave Technology, Prentice Hall, Upper Saddle River, NJ.
 Chapter 4 Single- and Multlport Networks

Problems
4.1 From the defining equations (4.3) and (4.6) for the impedance and admit-
tance matrices, show that [Z] = [Y]-' .

For the following generic T-network, find the impedance and admittance
matrices.

Show that for a bipolar-junction transistor in a common-base configuration

under small-signal low-frequency conditions (whose equivalent circuit is
shown below) a hybrid parameter matrix can be established as follows:

where the individual transistor parameters are denoted in the figure.

Using the results from Problem 4.3, compute the equivalent circuit parame-
ters for a BJT in common-base configuration if the h-matrix is given as
r 7

Employ the conversion table for the different parameter representations of

the two-port network and find the h-matrix representation for a Darlington
pair shown in Figure 4-7 under the assumption that the transistors are speci-
fied by the same h-matrices derived in Example 4-2.
Problems

4.6 Using the definition of the ABCD network representation, find the Y-parame-
ter description.

4.7 From the results of Problem 4.3 and Example 4.2, establish the conversion
equations between the h-matrix parameters for the common-base and com-
mon-emitter transistor configurations.

4.8 Unlike the series connection discussed in Example 4-4, derive the ABCD-
parameters for a two-port network where the impedance Z is connected in
parallel.

4.9 Find the ABCD-parameters for a generic three-element pi-network, as

depicted in Figure 4-2.

4.10 Compute the ABCD-parameters for an RF transformer with turn ratio

N = N / N 2 , where N , is the number of turns a the primary winding and
N 2 is the number of turns of the secondary winding.

4.11 Prove that the h-matrix parameters for a high-frequency hybrid transistor
model shown in Figure 4-12 are given by (4.31).

4.12 In this chapter we have mentioned several h-matrix representations of the

bipolar-junction transistor for different frequency conditions. In all cases we
have neglected the influence of the parasitic components associated with the
casing of the transistor. The modification to the equivalent circuit of the tran-
sistor that takes into account these parasitics is shown below:

II

LB Intrinsic LC
B- ' Y Y \ : ~ Transistor C . Tvv\,C
Model
E'
II
II
II
II
-
CB, CEc
LE
 Chapter 4 Single- and Multlport Networks

Assuming that the intrinsic transistor model is given by a generic h-matrix,

derive the modified model that accounts for the casing.

Compute the return loss for a 25 !2 resistor connected to a 75 !2 lossless

transmission line.

Find the forward gain of the circuit discussed in Example 4-8.

Given that the input of an amplifier has a VSWR of 2 and the output is given
by VSWR = 3, find the magnitudes of the input and output reflection coeffi-
cients. What does your result mean in terms of S l l and S22?

Using the same approach as described in Section 4.4.4, show that the S-
parameters of the network can be computed from the known Y-parameters
using
[Sl = ([Yl + Y,[EI)-'(Y,[EI - [YI)
and the corresponding inverse relation
[Yl = Y,([EI - [SI)([Sl+ [El)-'
where Yo = 1/Zo is the characteristic line admittance.

The ideal transformer of Problem 4.10 can also be represented in S-pararne-

ter form. Show that the S-matrix is given by

where N = N1/N2.

For the following two circuits, prove that the S-parameters are given as

respectively, where r1= (1 + 2Z0/Zl)-' and r2= -( 1 + 2Yo/Y ,)-I .

Problems 199

4.19 For the following T-network inserted into a transmission line with character-
istic impedance of Zo = 5 0 R , the three resistances are
R1 = R2 = 8.56 R , and R j = 141.8 f2. Find the S-parameters of this
configuration and plot the insertion loss as a function of inductance L for
the frequency of f = 2 GHz and L changing from 0 to 100 nH.

4.20 In practice, the resistors in the T-network of the previous problem are not
frequency independent. At RF frequencies parasitic effects have to be taken
into account. Compute the S-parameters at 2 GHz when all resistors have a
0.5 nH parasitic series inductance. Assume L is fixed at 10 nH.

4.21 A BJT is operated in a 50 R circuit at 1.5 GHz. For the bias conditions of 4
mA collector current and collector-emitter voltage of 10 V, the manufacturer
provides the S-parameters in magnitude and angle as follows:

Find (a) the 2-parameter and (b) the h-parameter representation.