Microwave Engineering

- 4. The Physical Basis of -

Transmission Lines
Yo-Shen Lin
[email protected]

MWEng113– Y.-S. Lin

Outline
• Why transmission lines?
• Transmission line theory.
• Transmission line and waveguide.
• Lossy transmission line.
• Transmission Line and Reflections

2
MWEng113– Y.-S. Lin
 What is a Transmission Line?

3
MWEng113– Y.-S. Lin

Transmission Line: Definition

• Fundamentally, a transmission line is composed of
any two conductors that have length.
• The transmission line is a ideal “distributed” circuit
element that has a constant instantaneous impedance
(or characteristic impedance) and a time delay
associated with it.
– It match the actual, measured behavior of real signal traces
and interconnections much better and to much higher
bandwidth.
signal path

return path
length
4
MWEng113– Y.-S. Lin
 Why Transmission Line?

• Short Wire
I=1A
50V
50
Instantaneous Reaction

• Long Wire
I=?A

50V
50

Earth Moon
Delayed Reaction
382,171.2 (km)/300,000(km/s) = 1.27 sec

5
MWEng113– Y.-S. Lin

Why Transmission Line? (cont’d)

V

Low frequencies
 wavelengths >> wire length
 measured voltage and current not dependent on position along wire

 current (I) travels down wires easily for efficient power transmission

High frequencies
 wavelength » or << length of transmission medium

 measured envelope voltage dependent on position along line

 need transmission lines for efficient power transmission

 matching to characteristic impedance (Z ) is very important for low

0
reflection and maximum power transfer
6
MWEng113– Y.-S. Lin
 Forget the Word “GROUND”

GROUND

• Far more problems are created than solved by referring

to the second line as the ground.
• It is a good habit to use the term return path instead.
7
MWEng113– Y.-S. Lin

Charge Distribution Along a Transmission Line

• DC input

• AC input
I
Signal path

VAC E
Return path

V(z) = Vsignal - Vreturn

Note: If we know the “impedance the signal sees”, we can always
8 calculate the current associated with the signal. MWEng113– Y.-S. Lin
 Ex.: Microstrip Surface Current Density

Signal path

Top view
Cross-sectional view
Return path

9
MWEng113– Y.-S. Lin

Ex.: Microstrip Mode Field Pattern

Vector E-field Vector H-field

Cross-sectional view

10
MWEng113– Y.-S. Lin
 The Speed of Signal in a Transmission Line
• The speed of signal along a transmission line depends on how
quickly the changing electric and magnetic fields associated with
the signal can build up and propagate in the space around the
transmission line conductors.

• The propagation and interaction of these fields is described by

Maxwell’s Equations.
• The speed of signal is the speed of light in the material
1
v : depends on the materials that
11 0 r 0 r “surround the conductors”
MWEng113– Y.-S. Lin

Speed of Signal v.s. Speed of Electrons

• The speed of the electrons in a typical copper wire is actually
about 10-billion-times slower than the speed of the signal.
I = ΔQ / Δt = ( q  n  A  v  Δt ) / Δt = q  n  A  v
v = I / ( q  n  A ) = 1cm/sec!
where I = the current passing one point, in Amps
ΔQ = the charge flowing in a time interval, in Coulombs
Δt = the time interval
q = the charge of one electron = 1.6  10-19 Coulombs
n = the density of free electrons, in #/m3
A = the cross section area of the wire, in m2
v = the speed of the electrons in the wire, in m/sec

12
MWEng113– Y.-S. Lin
 Zeroth-Order Model of a Transmission Line

Charging a capacitor on each step!

C  CL  x
Q CV
I   CL  V  v
t x / v
where I = the current from the signal
Q = the charge in each foot step
C = the capacitance of each foot step
Δt = the time to step from capacitor to capacitor
CL = the capacitance per length
Δx = the distance between each foot step
v = the speed of walking down the line
13 V = the voltage of the signal MWEng113– Y.-S. Lin

Instantaneous Impedance and

Characteristic Impedance
• We call the impedance the signal sees with each step, the
instantaneous impedance of the transmission line.
V 1
Z 
I vCL
where
Z = the instantaneous impedance of the
transmission line
CL = the capacitance per unit length
v = the speed of light in the material
• Z depends only on the cross section of the line and the
material properties.
• For a uniform line, the instantaneous impedance is constant
down the line, and is called the characteristic impedance of
a transmission line.
14
MWEng113– Y.-S. Lin
 Characteristic Impedance of a Transmission Line

Ex.: Trace on PCB

1
Z
vCL

15
MWEng113– Y.-S. Lin

Uniform Transmission Lines

• Transmission lines are classified by their 2D geometry
• Uniform transmission line: when the cross section is the same
down its length; also called the controlled impedance line.

h
a
r
b w1
w
w2
Coaxial Twisted-pair Microstrip Coplanar Waveguide
(CPW)

CPW
16
MWEng113– Y.-S. Lin
 Discussions
• The major role of the transmission line is to transmit
a signal from one end to the other end with an
acceptable level of distortion.
– Reflections will be minimized and signal quality will be
optimized, if the transmission lines are uniform, or
have controlled impedance.
Ex.: Ringing noise due to
impedance mismatches and
multiple reflections

• The propagation characteristic of a transmission line

can be well characterized by its 2D cross-section.
17
MWEng113– Y.-S. Lin

Effect of Line Impedance on Signal

• The impedance, looking into the transmission line, is time dependent!

• When the rise time of a signal is shorter than the round-trip time of
flight of an interconnection, a driver will see the interconnect with a
resistive impedance equal to the characteristic impedance of the line.

18
MWEng113– Y.-S. Lin
 Effect of Line Impedance on Signal (Cont’d)
• In most system with clock frequencies higher than 200MHz, the
rise time is less than 0.5 nsec.
– This is why the transmission line behavior of all
interconnections must be considered.

19
MWEng113– Y.-S. Lin

More On Return Paths

• Signal trace passing through a multi-layered board:

• The two return planes create a transmission line!

• The return current leads to a voltage drop in the return path,
which is called ground bounce.
– Minimizing the impedance of the return path will minimize
the ground-bounce noise
20
MWEng113– Y.-S. Lin
 First-Order Model of a Transmission Line
• The zeroth-order model is a “physical” model but not an “electrical”
model.
• By modeling the current loops associated with the transmission line
using inductors, we obtain the first-order model of a lossless
transmission line:

21
MWEng113– Y.-S. Lin

First-Order Model of a Transmission Line (Cont’d)

signal path

• LLoop=(LS - LM)+(LR - LM)

return path
length = LS + LR - 2LM
LS LLoop

LM
C v.s. C

LR

• C is the capacitance between conductors for each section;

LLoop is the loop inductance of each section.
• This lumped circuit model is an “approximation” of an ideal
transmission line, as the size of capacitors and inductors
gets diminishingly small and the number of each increases,
22
the approximation gets better and better.
MWEng113– Y.-S. Lin
 Calculation of Characteristic Impedance
Z Z Z Z

Z  jL;
Zin Y Y Y
Y  j C .
An infinite ladder network

Zin  Z  Zin //(1/ Y ) …the instantaneous impedance

Z in / Y
Z in  Z   ( Z in  Z )(1 / Y  Z in )  Z in / Y
1 / Y  Z in
Z  Z 2  4( Z / Y ) Z 4
Z in   (1  1  )
2 2 ZY
Consider |ZY| << 1 and choose the positive value of Zin

23
 Zin  Z / Y  L / C  Z 0 …the characteristic impedance
MWEng113– Y.-S. Lin

Calculation of Time Delay

• Time delay of the ladder network: L  LL  z

TD  LC  z  LL C L
C  CL  z
z 1
vp   ;
TD LL C L

where ∆z = physical length of one section of the

ladder network
LL = per unit length inductance
CL = per unit length capacitance
• Time delay (TD) v.s. Z0:
Ltotal  TD  Z 0
Length
TD   Length  LL CL  Ltotal Ctotal ; TD
vp Ctotal 
Z0

24
Ex.: Z0=50 ohm, TD=1 ns, Length = 6 inches, CL= ? LL= ?
MWEng113– Y.-S. Lin
 Calculation of Propagation Constant
• Propagation constant (): describing the ratio of voltages
(currents) between any two points on a transmission line
Vn Z Vn+1 Z Z Z Vn Z Vn+1

Z  j LL dz;
Y Y Y Y Z0
Y  jCL dz.

 1  1
Z // Z 0 //
 0
Y   Vn 1  j CL dz Z0
Vn 1  Vn  
1 
] Z 0  j LL dz  Z 0 LLCL dz
Vn 1 2 2
 Z  [ Z 0 // ]  j LL dz  [ Z 0 //
 Y  j CL dz
Vn 1 Z0 1
dz  0     1  jLL dz / Z 0  1  j LL C L dz
Vn Z 0  jLL dz 1  jLL dz / Z 0
Vn 1  Vn dV
Vn 1  Vn (1  j LL C L dz )    j LL C L Vn    j LL C L V
dz dz
 j LL CL z
 V ( z )  V0 e  V0 e  z ;
25   j   j LLCL . …the propagation constant MWEng113– Y.-S. Lin

Discussions
• For ‘lossless’ line, the propagation constant  = j
contributes to the phase of signal
=>  : phase constant
V ( z )  V0 e   z  V0 e  j  z ;   j   j LLC L .
• Phase constant (  ) v.s. phase velocity (vp)
1
vp  ;
LLC L
 
   LLC L  ; or v p  .
vp 
• Compared with the zeroth order model
1 …same as the result predicted
Z 0  Z / Y  L / C  LL / CL 
v pCL by the zeroth-order model!
26
MWEng113– Y.-S. Lin
 Transmission Line Circuit Theory
i( z, t ) i ( z, t ) LL  z i ( z  z, t )

z  v ( z  z , t )
v( z, t ) v ( z, t ) CL  z

z
• LL & CL: per unit length loop inductance & capacitance
• KVL, KCL => v( z, t )   L i ( z, t ) ; i ( z, t )  C v ( z, t )
L L
z t z t
Time-harmonic dV ( z ) dI ( z )
  jLL I ( z );   jCLV ( z )
form: dz dz
2
Wave equation: d V ( z )   2V ( z )  0; d 2 I ( z)
2
  2 I ( z)  0
dz 2 dz

27    LLCL : phase constant

MWEng113– Y.-S. Lin

Traveling Wave Solutions

• Traveling wave solutions:
V ( z )  V  ( z )  V  ( z )  V0 e  jz  V0 e  jz

Travels in +z direction Travels in -z direction

    jz   jz V0  jz V0  jz

I ( z)  I ( z)  I ( z)  I e 0 I e
0  e  e
Z0 Z0
V0 V0 LL
where Z 0     : the characteristic impedance
I 0 I 0 CL
dV ( z )
   jLL I ( z )   jV0 e  jz  jV0 e  jz
dz
 LLCL   jz  LLCL   jz V0  jz V0  jz
 I ( z)  V0 e  V0 e  e  e
LL LL Z0 Z0
• Traveling wave solutions in time domain: …Correspond to voltage
v ( z, t ) | V | cos(t  z  V ) | V | cos(t  z  V )
0

0

0

0
 and current waves
propagated along the
28 i ( z, t ) | I | cos(t  z  I ) | I | cos(t  z  I )

0

0

0

0 transmission line!Y.-S. Lin
MWEng113–
 Cont’d
t = t1 t = t1+∆t v( z , t )  | V0 | cos(t   z  V0 )

For t   z  V0  const.  t  z , wave travel in +z

Similarly, t   z  V0  const.  t  z  , wave travel in -z
-

• Phase constant Phase velocity

2   1
    LLCL vp  
 vp  LLCL

*The results for Z0, , and vp from the infinite ladder network model
and the transmission line circuit theory are exactly the same!
29
MWEng113– Y.-S. Lin

Discussions
• Compared with field theory (for TEM mode only):
1 c 1
vp    ;       LL C L
 r LL CL
Et 
Z wave   : wave impedance
Ht 
V LL
Z0   : characteristic impedance
I CL

• Ex: For coaxial line:

μ, ε
 b 2
a LL  ln , C L  .
b 2 a ln b / a

30
MWEng113– Y.-S. Lin
 Discussions (Cont’d)
• LL v.s. CL
c 1 r r
vp   CL  7 ( pF / inch), LL  7 (nH / inch)
r LL CL LL CL
1 1 83
CL   r   r ( pF / inch)
v p Z 0 cZ 0 Z0
Z0
LL   0.083Z 0  r (nH / inch)
vp

Ex.: Z0=50 ohm, r=4, CL= ? LL= ?

31
MWEng113– Y.-S. Lin

Calculating Characteristic Impedance

• Engineering a particular target characteristic impedance
is really a matter of adjusting the line widths, dielectric
thickness, and dielectric constants.
• In general, there are three types of analysis we can use:
1. Rules of thumb
2. Approximations
3. 2D/3D field solvers

CPW

32
MWEng113– Y.-S. Lin
 Ex.: Design of Microstrip
• Characteristic of microstrip
– Effect of strip width W:
W , CL , LL  , Z 0  LL / CL 
– Effect of substrate height h:
h , CL , LL , Z 0  LL / CL 
– Effect of dielectric constant
 r , CL , LL no change, Z 0  LL / CL 
• What about stripline?
air r
r

33
MWEng113– Y.-S. Lin

Rule of Thumb for Z0

• : As a good rule of thumb, a 50-ohm microstrip in FR4 has a
line width twice the dielectric thickness. A 50-ohm stripline has
a total dielectric spacing between the planes equal to twice the
line width.
• : To the first order, the characteristic impedance of stripline
and microstrip will scale with the ratio of the dielectric thickness
to the line width. As long as this ratio is constant, the
characteristic impedance will be constant.

air r
r

34
MWEng113– Y.-S. Lin
 Analytic Approximations for Z0
• There are only three cross-section geometries that have exact equations.
All others are approximations
– These exact solutions are useful relationships to help calibrate 2D/3D field solvers,
since they are exact.

60Ω

where Z0 = the characteristic impedance, in ohms

a = inner radius of the coax, in inches; b = outer radius of the coax, in inches
r = radius of the round rod, in inches; s = center-to-center spacing of two rods, in inches
h = height of the center of the rod over the planes, in inches
35 r = dielectric constant of the materials
MWEng113– Y.-S. Lin

Cont’d
• An approximation for the characteristic impedance of
microstrip:
r

t: metal thickness

• An approximation for the characteristic impedance of

stripline:
r

36
MWEng113– Y.-S. Lin
 2D Field Solver for Calculating Z0
• When accuracy counts (error<10%), or when we are worrying
about second-order effects, a verified 2D/3D field solver
should be used. (The only way to know the accuracy of an
approximation is to compare it to the results from a verified
field solver.)
• The basic assumption made by all 2D field solvers is that the
geometry is uniform down its length.

37 Circles: 2D field solver; Line: analytic approximation.

MWEng113– Y.-S. Lin

Examples of 2nd-Order Effects

• 1. Extension of return plane for a microstrip
trace width: 10mil
trace thickness: 0.7mil
dielectric thickness: 5mil
dielectric constant: 4

: As a rough rule of thumb, the return

path should extend at least three times
the dielectric thickness on either side
of the signal trace for the characteristic
impedance to not exceed more than
1% of the value when the return path
is infinitely wide.

38
MWEng113– Y.-S. Lin
 Examples of 2nd-Order Effects (Cont’d)
2. Metal thickness 3. Solder mask thickness

Circles: 2D field solver; Line: 2 Ohms/mil.

*Solder mask
(Green colored)

39
MWEng113– Y.-S. Lin

N-Section Lumped Circuit Model

• Real interconnections match the behavior of an ideal
transmission line to high bandwidth. An ideal transmission
line is a very good model for real interconnections.
• Ex.: Measured (circles) and simulated (line) impedances
of a 1-inch-long, 50-ohm characteristic impedance,
open-ended transmission line:

1 inch

50 ohm

40
MWEng113– Y.-S. Lin
 How good is the Lumped-Circuit Approximation?
LLoop

• We can approximate the ideal lossless transmission C

line by a combination of LC lumped-circuit sections.
– How many sections required for a given accuracy?
Ex.: Simulated impedance of an open-ended ideal
transmission line (Δ) v.s. simulated impedance of a single
section LC lumped-circuit model (□).
Input impedance
100000
n=1

1000
|Z| (ohm)

Z0=50ohm, TD = 1ns o.c.

10

Z |Z(1,1)| |Z(2,2)|
Transmission line LC with n=1

.1
0 1 2 3 4
41 Frequency (GHz) MWEng113– Y.-S. Lin

Improving the Model Bandwidth

100000
• An n-section LC circuit is an n=2
approximation of an ideal
transmission line. 1000
|Z| (ohm)

– C = Ctotal / n;
L = Ltotal / n. 10

• The more sections, the higher |Z(1,1)|

TL
|Z(3,3)|
LC with n=2

the model bandwidth of the .1

0 1 2 3 4
approximation. 100000
Frequency (GHz)

– : Roughly one LC section n=8

for every 1/10th wavelength 1000
of the highest signal
|Z| (ohm)

frequency of interest are

required 10

|Z(1,1)| |Z(5,5)|
TL LC with n=8
.1
42 0 1 2
Frequency (GHz)
3 4
MWEng113– Y.-S. Lin
 • : We can generalize (or estimate) the highest frequency at
which we have good agreement – the bandwidth of the model –
as:
BWmodel ≈ n/4  f0 /2 ≈ n  f0 /10
or
n = 10  BWmodel / f0 = 10  BWmodel  TD
where
BWmodel = bandwidth of the n-section lumped-circuit model
n = the number of LC sections in the model
TD = the time delay of the transmission line
f0 = the resonant frequency for one wavelength = 1/TD

• : To reach a bandwidth of the model equal to the 1/TD, 10 LC

sections are required; or, there should be one LC section for
every 1/10th wavelength of the signal.
Ex.: TD = 1 ns, BWmodel required = 5 GHz, how many
sections of LC required?
Ex.: Up to what frequency does a transmission line look
43 like a simple LC circuit for TD = 1 ns? MWEng113– Y.-S. Lin

TEM v.s. Non-TEM Lines

• The transmission line theory is exact for transmission lines
that support TEM waves (transverse electromagnetic
wave)
• Transmission line in inhomogeneous medium does not
support TEM wave
– Use quasi-TEM approximation to facilitate design
TEM:
a

b
Coaxial

h Non-TEM:

r

w
44
Microstrip MWEng113– Y.-S. Lin
 Effective Dielectric Constant
εr1 E ( x, y )
• Effective dielectric constant, εre
εr2
εre = Cd / Ca
where Ca = capacitance with dielectric replaced
by air
Cd = capacitance with the actual dielectric
material distribution
– Can be regarded as an ‘average’ dielectric constant
for inhomogenous media
εre
εr1 ε0

εr2

45
Cd Ca C1 = Cd
MWEng113– Y.-S. Lin

Discussions
• Field distribution in dielectric and air of a microstrip
under different substrate height h:
h ↓, εre↑

h = 500 microns h = 250 microns h = 125 microns

εr = 4.5 εr = 4.5 εr = 4.5
46 *White curve: equipotential line
MWEng113– Y.-S. Lin
 Discussions (Cont’d)
• Field distribution in dielectric and air of a
microstrip under different trace width W:
E ( x, y )

W ↑, εre↑

47
MWEng113– Y.-S. Lin

Transmission Line Parameters for Non-TEM Lines

Cd
• With effective dielectric constant:  re 
Ca
1 c c
Speed of wave: v p   
0 r 0 re  re C d / Ca

Characteristic impedance: Z 0  1  1
v p Cd c Ca Cd
where
Cd = CL : capacitance per unit length with the dielectric presented
Ca : capacitance per unit length with the substrate replaced by air;
c : velocity of light.

 r1  re

r2 
48 Non-TEM TEM MWEng113– Y.-S. Lin
 Summary of Transmission Line Parameters
• For lossless line: l
– Characteristic impedance Z0
1 LL Z0 , 
Z0  
v p CL CL

– Propagation Constant   j 
 2 f 2
  
vp vp g
 c
where v p   …phase velocity
  re
 …guided wavelength
g  0
 re
– Electrical length    
49
MWEng113– Y.-S. Lin

Losses in Transmission Lines

• The first-order approximate model for transmission
line is an n-section LC model, which is often referred
to as the lossless model. It accounts the characteristic
impedance and time delay, but not the loss of the
signal.
• Loss mechanism:
– Radiation loss
• EMI
– Coupling to adjacent traces
– Impedance miss matches
• Impedance discontinuities can have a dramatic
impact in distorting the transmitted signals.
– Conductor loss Primary causes of attenuation
50 – Dielectric loss in transmission lines.
MWEng113– Y.-S. Lin
 Lossy Lines and Rise Time
• A signal with a very fast edge going into a long transmission
line with come out with a longer rise time.
Ex: Measured input and output signals through a 50-ohm
line of 36-inches long.
Input: 50 psec rise time
Output: 1 nsec rise time

51
MWEng113– Y.-S. Lin

Intersymbol Interference (ISI)

• The important consequence of rise-time degradation is ISI:
the precise waveform of the bit pattern will depend
on the previous bits that have passed by, which
leads to an increase of bit error rate (BER).

Ex.: 5GHz clock-driven pseudorandom bit stream. Left: bit pattern when the rise
time is much shorter than the bit period. Right: bit pattern when the rise time is
comparable to the bit pattern, causing “pattern-dependent” voltage levels or
“intersymbol interference”.
52
MWEng113– Y.-S. Lin
 Cause of Rise Time Degradation
• The reason the rise time is increased in propagating down a real
transmission line is specifically because the higher-frequency
components of the signal are preferentially attenuated more than
the lower frequencies.

53
MWEng113– Y.-S. Lin

Frequency Dependent Loss and ISI

• Ex.: Simulated transmitted signals at the output of a 30-
inch-long, 50-ohm trace on FR4 with
(a) no loss;
(b) frequency-independent loss;
(c) frequency-dependent loss.

54
MWEng113– Y.-S. Lin
 Conductor Resistance
• The series resistances a signal sees in propagating down
the signal and return paths is related to conductor’s bulk
resistivity and the cross section through which the current
propagates.

I I

• RDC    d / ( w  t ) for uniform current distribution in the

signal conductor at DC.
• The bulk resistivity (ρ = 1/σ) of metal is absolutely constant
with frequency until frequencies near 100GHz, but the
current will redistribute itself at higher frequencies due to
55 the skin effect. MWEng113– Y.-S. Lin

Skin Effect
• Skin Effect: Current will re-distributes itself over the conductor
cross-section at different frequencies
• Driving force for current distribution: Microstrip:
the lowest impedance path for the
current loop

Round wire

56
MWEng113– Y.-S. Lin
 Skin Depth and AC Resistance
• Skin depth (): the penetration depth of the electromagnetic
wave in good conductors over which the amplitude of the
electric field decays to 1/e of its value at the conductor
surface.

1

f
where σ is the bulk
conductivity of the
conductor

• Approximated current
distribution of a round
wire:
57
MWEng113– Y.-S. Lin

Approximation of AC Resistance
• Let R    d / (w   ) , the total series AC resistance RAC of
microstrip line is approximately
RAC  Rsignal  Rreturn  R  R / 3  1.33R  f at frequencies
above about 10 MHz…Frequency dependent!
•   2.5 1 , where  = the skin depth, in microns
f
f = frequency, in GHz

w 
3w

58 Approximated current distribution
MWEng113– Y.-S. Lin
 Dielectric Loss
• An ideal capacitor with air as dielectric has an infinite DC
resistance. Real dielectric materials have some resistivity
associated with them, which leads to leakage current
when we apply a DC voltage across it. It can be modeled
by a resistor.
• For most dielectrics, the bulk resistivity (ρ) is very high at
DC, typically 1012 ohm-cm. The bulk resistivity decreases
with frequency due to the increasing motion of permanent
electrical dipoles in the material, and the bulk conductivity
( = 1/ρ) increases with frequency.

59
MWEng113– Y.-S. Lin

Bulk Conductivity of Dielectrics

• For most dielectrics, the bulk conductivity ( = 1/ρ) is
constant from DC until some frequency is reached and
then it begins to increase with frequency.
• At high frequency, the conductivity increases with
frequency
  ( f )  2 f  0 r tan 
  0 r tan 
where
 = bulk AC conductivity
of the dielectric
f = frequency in Hz
0 = permittivity of free space
r = relative dielectric constant,
dimensionless
tan = the dissipation factor of the material,
60 dimensionless
MWEng113– Y.-S. Lin
 Dissipation Factor
• The dissipation factor (tan) is weakly frequency dependent
and will vary from lot to lot and board to board.
• To completely describe the electrical properties of a dielectric
material, both the dielectric constant and dissipation factor
are required.

61
MWEng113– Y.-S. Lin

Real Meaning of Dissipation Factor

• The current through a real capacitor, with an applied
sine-wave voltage, can be described by two components
– an out-of-phase component passing through C
– an in-phase component passing through R
• For time-harmonic case:
I  jCV  j r C0V
define  r   r  j r
where r = the complex dielectric
constant
 I  j rC0V   rC0V
 tan    r /  r
where  = the loss angle
62
MWEng113– Y.-S. Lin
 Dielectric Loss v.s. Frequency
V 1
• I leakage  Re( I )   rC0V  Rleakage1  
Re( I )  rC0
• For parallel-plate configuration:
A h
A
C   0 r   rC0
h r
1 h 1 
Rleakage 2     0
 A  C0
Rleakage1  Rleakage 2     0 r tan   2 f  0 r tan 

• Even though the dissipation factor itself is only weakly

frequency dependent, we see that the bulk-AC conductivity
of a dielectric will increase linearly with frequency. Likewise,
since the power dissipated by the leakage resistance is
proportional to the bulk-AC conductivity, the power
dissipation will also increase linearly with frequency.
63
MWEng113– Y.-S. Lin

Modeling Lossy Transmission Lines

• The two loss processes for attenuating signals in a transmission
line are the series resistance through the signal- and return-path
conductors and the shunt resistance through the lossy dielectric
material.
– Both of these resistors have resistances that are frequency
dependent
1
RL  RAC ( f )  f GL    rC0   tan  CL  f
Rleakage
RL∆z LL∆z

CL∆z GL∆z

RL : series resistance per length of the conductors

CL : capacitance per length
LL : series loop inductance per length
GL : shunt conductance per length from the dielectric
64
MWEng113– Y.-S. Lin
 Lossy Transmission Line Circuit Theory
i ( z, t ) i ( z, t ) RL  z LL  z i ( z  z , t )

v ( z, t )  v ( z, t ) G L  z
CL  z
v( z  z, t )

z z
v ( z, t ) i ( z, t )
• KVL, KCL =>   RLi ( z, t )  LL ;
z t
i ( z , t ) v( z , t )
 GL v( z , t )  CL
z t
• Time-harmonic form:
dV ( z ) dI ( z )
 ( RL  jLL ) I ( z );  (GL  jCL )V ( z )
dz dz
• Wave equation:
d 2V ( z ) d 2 I ( z)
2
  2V ( z )  0;
2
  2 I ( z)  0
dz dz
65   ( RL  jLL )(GL  jCL )    j : propagation constant
MWEng113– Y.-S. Lin

Cont’d
• Traveling wave solutions:
V ( z )  V  ( z )  V  ( z )  V0 e  z  V0 e  z
V V
I ( z )  I  ( z )  I  ( z )  I 0 e  z  I 0 e  z  0 e  z  0 e  z
Z0 Z0
where
V V RL  jLL : characteristic impedance
Z 0  0   0 
I0 I0 GL  jC L

• Time domain solutions:

v ( z, t ) | V0 | e z cos(t  z  V0 )  | V0 | ez cos(t  z  V0 )
i ( z, t ) | I 0 | e z cos(t  z  I 0 ) | I 0 | ez cos(t  z  I 0 )

• Ex.: Try to obtain the expressions of characteristic impedance

and propagation constant using the infinite ladder network
66 point-of-view. MWEng113– Y.-S. Lin
 Attenuation and dB
•Voltage wave travels in +z:
v  ( z, t ) | V0 | e  z cos(t   z  V0 )

 d A
• | V0 ( d ) || Vin | e n | Vin | e n
where V(d) = the voltage on the line at position d
Vin = the amplitude of the input voltage wave
An = the total attenuation, in nepers
n = the attenuation per length, in nepers/m
| V  (1m) || V  (0m) | e  n 1 | V  (0m) | 10-dB 1/20
| V  (0m) | | V  (0m) |
 n  ln  (Neper/meter);  dB  20 log  (dB);
| V (1m) | | V (1m) |
 | V  (0m) |   | V  (0m) | 
67  dB /  n  20*log    / ln     20 / ln 10   8.68588
 | V (1m ) |   | V (1m ) |  MWEng113– Y.-S. Lin

Discussions
• The characteristic impedance of lossy transmission line is
frequency dependent and complex!
RL  j LL
Z0 
GL  j CL
• The velocity of a sine-wave signal is also frequency dependent!

vp 
1
2 [ ( RL2   2 L2L )(GL2   2CL2 )   2 LLCL  RLGL ]

• A new term is introduced that describes the attenuation of the

sine-wave amplitude as it propagate down the line. This
attenuation is also frequency dependent!

n  1
2 [ ( RL2   2 L2L )(GL2   2CL2 )   2 LLCL  RLGL ]
68
MWEng113– Y.-S. Lin
 Low Loss Approximation
• Low loss approx.: RL << LL, GL << CL
Ex.: For a 50-ohm microstrip line in FR4 with 1-ounce copper
trace of width w (in m), above about 10 MHz:
AC resistance per length
RL  1.33 / (w   )  3.47 107  f / w (ohm/m)
Inductance per length for a 50-ohm line h

LL  Z 0 / v p  274 (nH/m) r
For  LL  RL : w
  274 109  3.47 107 f / w Microstrip
f  0.04 / w 2 (Hz)
• For line width wider than 0.1 mm, the low-loss regime is for
frequency well above 4 MHz! For line wider than 0.1 mm, the
low-loss regime begins at even lower frequencies. The “very
lossy” regime is actually in the low frequency, below the
69
frequency where skin depth plays a role.
MWEng113– Y.-S. Lin

Low Loss Approximation (Cont’d)

• The conductance of dielectric GL   tan  CL
• For GL  CL  tan   1
– For virtually all interconnect, the dissipation factor
is less than 0.02, and the interconnect is always in
the low-loss regime.
• We may conclude that the low-loss approximation is
a very good approximation for all important frequency
ranges of interest in microwave applications.

70
MWEng113– Y.-S. Lin
 Characteristic Impedance of a Lossy
Transmission Line
• For ideal lossy transmission line, the characteristic impedance is
given by
RL  j LL
Z0  : A complex number!
GL  j CL
• In the low-loss regime, the characteristic impedance reduces to
LL
Re( Z 0 )  Im( Z 0 )  0
CL
: same as the lossless case!

Ex.: The magnitude of the complex

characteristic impedance of a
3-mil wide, 50-ohm microstrip
in FR4.
71
MWEng113– Y.-S. Lin

Propagation Constant of a Lossy

Transmission Line
• For lossy line:

RL G RG
  ( RL  j LL )(GL  jCL )  j LLCL 1  j (  L )  2L L
 LL CL  LLCL

• By low loss approximation:

RL G  j  RL G 
  j LLCL 1  j (  L )  j LLCL 1    L 
 LL  CL  2   LL  CL  
1 CL L  1  RL 
 n   RL  GL L     GL Z 0  ;
2 LL CL  2  Z0 
   LLCL . : same as the lossless case!

72
MWEng113– Y.-S. Lin
 Signal Velocity in a Lossy
Transmission Line
• For ideal lossy line, the signal velocity is given by

vp 
1
2 [ ( RL2   2 L2L )(GL2   2CL2 )   2 LLCL  RLGL ]

• In the low-loss regime, the velocity can be

approximated by:
 1
vp  
 LL CL
: same as the lossless case!

Ex.: The dispersion due to

losses for a 3-mil wide,
50-ohm microstrip in FR4.
73
MWEng113– Y.-S. Lin

Attenuation in Lossy Lines

• The attenuation is a new term that describes a special
property of lossy transmission lines. The attenuation per
length, denoted by alpha, in nepers/length is given by:

n  1
2 [ ( RL2   2 L2L )(GL2   2CL2 )   2 LLCL  RLGL ]

• In the low-loss approximation:

1 RL
n  (  GL Z 0 )
2 Z0
  conductor   dielectric
• Although there is no intrinsic frequency dependence to the
attenuation, since RL and GL increase with frequency, n
also increases with frequency

•
74
MWEng113– Y.-S. Lin
 Attenuation in Lossy Lines (Cont’d)
• For conductor loss: RL  f   cond  f
• For dielectric loss,
 tan  CL Z 0  tan  CL  r  tan   r
 diel     f
2 2cCL 2c
– The attenuation due to the dielectric is only determined by the
dissipation factor of the material and is independent of geometry
for homogenous dielectric.
– The attenuation from dielectric will increase faster with
frequency than the attenuation from the conductor.
Ex.: Attenuation in a 8-mil wide,
50-ohm microstrip in FR4,
separating the attenuations
from the conductor, the
dielectric, and the total.
75
MWEng113– Y.-S. Lin

Measured Properties of a Lossy

Line in the Frequency Domain
• Ex: Measured (thick) and simulated (thin) insertion loss of a
4inch-long, nearly 50ohm stripline in FR4:
|S21| (dB)

= - |S21| (dB);

|S21| : Magnitude of transmission coeff.

76
MWEng113– Y.-S. Lin
 Transmission Line and Reflections
• Reflections at impedance changes, or discontinuities:
– If the instantaneous impedance along a transmission line
changes, some of the signal will reflect back in the opposite
direction and some of it will continue with a different
amplitude.
– The amount of signal that reflects depend on the magnitude
of the change in instantaneous impedance.
Vreflected Z 2  Z1
 
Vincident Z 2  Z1
where
Vreflected/Vincident = the reflected/incident voltage
Z1 = the instantaneous impedance (characteristic impedance)
of the first region where the signal is initially propagating on
Z2 = the instantaneous impedance (characteristic impedance)
of the second region where the signal just enters
77  = the reflection coefficient MWEng113– Y.-S. Lin

Why Are There Reflections?

• No voltage and current discontinuities are allowed
at the discontinuity!!!
Vincident  Vreflected  Vtransmitted ;
I incident  I reflected  I transmitted .
Vincident Vreflected V
 Z1 ,  Z1 , transmitted  Z 2 ;
I incident I reflected I transmitted

Vincident Vreflected Vtransmitted Vincident  Vreflected

  
Z1 Z1 Z2 Z2

Vreflected Z 2  Z1
   …reflection coeff.
Vincident Z 2  Z1
Vtransmitted 2Z 2
  T …transmission coeff.
Vincident Z 2  Z1
*With the reflected signal, the voltage and current are all continuous
78 across the interface! MWEng113– Y.-S. Lin
 Lightwave Analogy to Microwave

Incident
Transmitted

Reflected
Lightwave

Microwave

79
MWEng113– Y.-S. Lin

Reflection and Impedance Matching

• Why we don’t want reflections to occur?
– It will reduce the power transfer to the other end of
the transmission line!
– It will cause “ringing” noise to the signal!

Ex.: Ringing noise at the receiver end of

a 1-inch-long controlled-impedance
interconnect created because of
impedance mismatches and multiple
reflections at the ends of the line.

• To avoid reflections:
– Use uniform transmission lines (controlled impedance
lines)
– Use additional “impedance matching network” at the
80 discontinuities
MWEng113– Y.-S. Lin