Académique Documents
Professionnel Documents
Culture Documents
September, 2008
Course Overview
0.1
Course Description
Introduction to analog signal processing, with an emphasis on underlying concepts from circuit and system analysis: linear systems, review of
elementary circuit analysis, dierential equation models of linear circuits
and systems, Laplace transform, convolution, stability, phasors, frequency
response, Fourier series, Fourier transform, active lters and AM radio.
0.2
Prerequisite
0.3
Goals
To introduce fundamentals of analog signal processing, with major emphasis on circuit analysis, dierential equations, convolutions, Fourier methods,
and applications in ltering and AM radio.
0.4
Topics
0.5. ASSESSMENT
Characterization and solution of LSI systems via linear, constantcoecient dierential equations
Transient response of RLC circuits
Complex numbers and functions of a complex variable
One-sided Laplace transform
Impedance
Laplace transform solution of dierential equations
General form of solution to a dierential equation
Transfer function and block diagrams
Impulse as a generalized function
Convolution
Stability
Sinusoidal steady-state analysis and phasors
Steady-state analysis of circuits
Frequency response
Fourier series
Fourier transform
Design of active lters
AM radio
Sampling theorem and overview of digital signal processing
0.5
Assessment
0.6
Texts
0.7
Course Content
Chapter 1
Fundamentals of Circuit Analysis
1.1
1.1.1
Circuit Introductory
Circuit Variables
dW
dq
(1.1)
as ground). The voltage between any point and this reference point is call
the potential of that point. It is always assumed that + is at the point and
the - sign is at the reference point. Therefore, there is no need to indicate
+ and - signs for potential. Voltage between two points is the dierence
between the potential of the two points (see Figure 1.1).
Voltage between two charge reservoirs is analogous to height dierence
between two uid reservoir and the same way, the potential of each point
is analogous to its elevation compared to some reference (e.g., sea level).
Potential
V1
Elevation
+
V2
v = v12 = V1 - V2
v = v21 = V2 - V1
v = -v
h2
h1
Sea level
h = h1 h2
however, is dicult and not needed expect for some cases (propagation of
electromagnetic wave and light, antennas, etc.).
In most relevant engineering cases the problem can be greatly simplied
by noting that electric charges preferentially ow through a conductor (or
a semiconductor) as opposed to vacuum, air, or any insulator. In this case,
the system can be described as a circuit containing circuit elements and
connecting ideal wires. Circuit theory is the scientic discipline that describes behavior of circuits be and is built upon the following assumptions:
All of the electromagnetic phenomena occurs inside each circuit element. They communicate with the outside world only through the
voltage across and current that particular element.
Circuit elements are connected to each other with ideal wires that do
not impede ow of charge. They can be stretched (making them longer
or shorter, for example) without any eect on the circuit.
Net amount of charge cannot be accumulated in any circuit element
or any location in the circuit. If a net charge of q enters a circuit
element, the same amount of charge should leave the element. This
means: (a) a circuit element should have at least two terminals, (b)
Because current travels through the system at a good fraction of speed
of light, we can safely assume that the total current entering a circuit
element is exactly equal to the current leaving that element at any
instant time.
i
+
1
Circuit element
v
2
Ideal wire
i
Figure 1.2: Circuit element
with two terminals.
In the context of circuit denition above, the important physical quantities are current and voltages (and electric power). These are the circuit
variables that form the basis for communication between circuit elements.
The value of voltage between two points, v, is incomplete unless the +
and - signs are assigned to the two points. Similarly, the value of current
is incomplete unless a reference current direction is assigned. However, as
v and i are algebraic numbers, the assigned position of + and - for voltage
and direction for the current is arbitrary. The sign of v and i ips according
to this choice.
Internal of each circuit element impose a relationship between the current owing in the element and the voltage across that element. This relationship is called the element law or i-v characteristics. While the choice
of reference directions for current through and voltage across an element
is arbitrary, the element law or i-v characteristics of an element depends
on the reference directions. To see this point, note that there exists four
possible choices for voltage and current directions in a two-terminal circuit
element:
i
-
(a )
v
2
+
1
( b)
Figure 1.3: (a) Passive sign convention. (b) Active sign convention.
In two left cases, the current direction is marked such that it ows from
+ to - signs. This case is called the passive sign convention. In the two
right cases, current ows from - to + signs. This case is referred to as active
sign convention. Obviously, the i-v characteristic will be dierent in each
case as the sign of current or voltage is switched. To resolve this confusion,
we follow this convention:
10
i-v characteristics of two-terminal element are written assuming passive sign convention.
Use passive sign convention when marking current and voltages in a
circuit as much as possible.
So, in marking references for voltages and currents in the circuit, it is
best to arbitrarily choose either the voltage or current reference direction
and then use passive sign convention to mark the other variable. Note that
it may not be possible or practical to mark every element using passive
sign convention. In this case, a good rule is to mark every element except
voltage and current sources using passive sign convention.
Electric power produced or absorbed in an element
The power, denoted by P , is dened as
P =
dW
dt
(1.3)
where the unit is Watt (W = J/s). Since v = dW/dq and i = dq/dt (both
total derivatives), then:
P =
dW
dq
dW
=
=iv
dt
dq
dt
(1.4)
11
i
A
+
Figure 1.4: Absorbed
and supplied powers.
Example
Find the power absorbed or supplied by elements A and B if i = 25A and
v = 120V.
For element A,
P = vi = 25 120 = 3, 000W = 3KW
Since element A reference directions follow passive sign convention and
P > 0, element A is absorbing 3 KW of power.
For element B,
P = vi = 25 120 = 3, 000W = 3KW
Since element B reference directions follow active sign convention and P > 0,
element B is supplying 3 KW of power.
1.1.2
Kirchho Laws
12
i1
i2
i3
i4
13
vb
va
vc
KVLs and KCLs are constraints on circuit variables which arise because
of the circuit arrangement (attachment of connection wires). In addition,
internal of each circuit element impose a relationship between the current
owing in the element and the voltage across that element (element laws
or i-v characteristics). Combination of these two set of constraints results
in a unique set of values for the circuit variables (currents and voltages).
In a circuit with E elements, there are 2E circuit variables (i and v for
each element). We need 2E equations to nd these circuit variables. If the
circuit has N nodes, we can write (N 1) KCL equations (KCL on the
N th node is exactly the sum of KCL on the other (N 1) nodes). We can
also write E N + 1 independent KVLs and E i-v characteristic equations.
Totally, it is 2E equations.
If the i-v characteristic is a linear relationship between i and v, the
element is a linear element. If a circuit is made of linear element, the
resulting set of 2E equations in 2E variables for a linear algebraic set of
equations. In this course, we only use linear elements.
1.1.3
Resistor
i-v characteristic: v = Ri.
Resistance: R. Unit: Ohm ().
Conductance: G = 1/R. Unit: Siemens (S).
P = vi = Ri2
(1.6)
14
i = v/ R
+ v -
R
Figure 1.7: Resistor: Circuit symbol and i-v characteristic.
+ i
v
-
+
-
v = vs
vs
vs
15
+ i
v
is
is
i = is
v
+ i
v
v
+ i
v
-
16
Switch
+ i
v
-
i-v characteristic:
Switch open: Open circuit i = 0 for any v.
Switch close: Short circuit v = 0 for any i.
1.1.4
Example
25
+ v0 25
i0
i2
10
i1
+
70
180V
180V
i5
i0
10
+ v2 - i 1
+
v1
-
+ v3 70
i3
i4
+
v4
-
(b)
(a)
Figure 1.12: Example of circuit analysis using KVL and KCL. (a) Original circuit. (b) Circuit
with nodes and loops.
17
=
=
=
=
=
25i0
70i1
5i2
10i3
8i4
=
=
=
=
=
=
0
0
0
0
180
0
18
1.2
19
i
+
Rest of
the circuit
Subcircuit
i
+
Rest of
the circuit
v
Equivalent
subcircuit
20
i
+
v
-
R1
i1
+ v1 - +
v2
-
R2
i2
= R1 i + R2 i
= (R1 + R2 )i
= Req i
Req = R1 + R2
So, a subcircuit containing two resistors in series has an i-v characteristics of the form v = Req i and is equivalent to a resistor, Req = R1 + R2 .
The above can be easily extended: k resistors in series are equivalent to
k
21
Rj
(1.7)
j=1
v = R1 is = v2
i
+
v
-
i1
+ v1 - +
v2
-
is
The i-v characteristics of ICS states that its current is is, independent
of its voltage (v2 ). Above equations show that the i-v characteristics of
subcircuit is i = is and is independent of voltage v = v1 + v2 (as value of
v1 can be anything). The equivalent subcircuit is an independent current
source with strength is .
i
+
R
+
vs
v
-
22
vs1
+
vs2
v
-
Figure 1.17:
A voltage
source in series with a voltage source.
The i-v characteristics of IVS states that their voltages are vs1 and vs2 ,
respectively, independent of their currents. Above equation shows that the
i-v characteristics of subcircuit is v = vs1 +vs2 and is independent of current
i. The equivalent subcircuit is an independent voltage source with strength
vs1 + vs2 .
is 1
+ v1 -
+
vs2
v
-
i2
Figure 1.18:
A voltage
source in series with a current source.
23
v = v1 + vs2
The i-v characteristics of ICS states that its current is is1 , independent
of its (v1 ). Above equations show that the i-v characteristics of subcircuit
is i = is1 and is independent of voltage v = v1 + vs2 (as value of v1 can be
anything). The equivalent subcircuit is an independent current source with
strength is1 .
Note: A current source in series with any element reduced to a current
source. Series element all have the same current and the current source
requires the current through to be equal to its strength.
is 1
+
v
is 2
24
=
+
i=
Req
Req
R1 R2
So, a subcircuit containing two resistors in parallel has an i-v characteristics of the form v = Req i and is equivalent to a resistor, 1/Req =
1/R1 + 1/R2 .
The above can be easily extended: k resistors in series are equivalent to
one resistor with
k
1
1
=
(1.8)
Req
R
j
j=1
A Resistor in Parallel with a Current Source
i
+
v
-
+
v1
-
i1
R1
+
v2
-
is
This is the Norton form and cannot be reduced further. We will show
later that Norton and Thevenin forms are equivalent.
A Resistor in Parallel with a Voltage Source
KCL : i + i1 + i2 = 0
KVL : v = v1 = vs
25
i
+
v
-
+
v1
-
i1 i2
R1
-
vs
Figure 1.21:
A resistor
in parallel with a voltage
source.
i = is1 + is2
i
+
v
v1
-
is 1
+
v2
-
is 2
The i-v characteristics of ICSs state that their currents are is1 and is2 ,
respectively, independent of their voltage, v. Above equations show that
the i-v characteristics of subcircuit is i = is1 + is2 independent of value of
voltage v. The equivalent subcircuit is an independent voltage source with
strength is = is1 + is2 .
26
i
i1
+
v
-
vs
-
+
v2
-
is
Figure 1.23:
A current
source in parallel with a
voltage source.
27
i
i1
+
+
vs1
-
i2
+
vs2
-
RN
vT
iN
R1
vs
Req
v2 = iR2 =
R2
vs
Req
28
i
R1
+
v1
R2
+
v2
+
vs
-
Figure 1.26:
vider.
Voltage di-
Also,
R1
v1
=
v2
R2
This circuit is called a voltage divider as the two resistors divide the
voltage of the IVS between them proportional to their values. This circuit
can be extended by adding more resistors to the circuit and get more reference voltages. This circuit is used extensively in electronic circuits. The
basic reason is that power supplies are bulky and/or expensive. Typically,
one power supply with one voltage is provided. On the other hand, more
than one voltage may be needed for the circuit to operate properly.
Example: A battery operated radio has a 9V battery. Part of radio
circuits require a 6V supply. Design a voltage divider circuit to supply 6V
voltage to these circuits. The desired circuit is the voltage divider circuit
above with vs = 9V and v2 = 6V. Then,
v2 =
R2
R2
R2
6
vs 6 =
9
=
Req
R1 + R2
R1 + R2
9
This is one equation in two unknowns and one is free to choose one
parameter. For example, choosing R1 = 1K, we get R2 = 2K. Voltage
dividers are aected by the load current drawn from them (see gure). The
voltage divider formula can only be used for the circuit if il i.
29
i
R1
+
v1
R2
- iL
+
v2
+
vs
-
Load
Current Divider
The two resistors can be replaced by an equivalent resistor,
1/Req = 1/R1 + 1/R2
is
v
-
+
v
-
i1 i2
R2
R1
+
v
-
Thus
v = Req i
v = i1 R1 i1 =
v
Req
=
is
R1
R1
v = i2 R2 i2 =
v
Req
=
is
R2
R2
Also,
R2
i1
=
i2
R1
30
This circuit is called a current divider as the two resistors divide the
current of the ICS between them (inversely proportional to their values).
This circuit can be extended by adding more resistors to the circuit and
get more reference currents.
Wheatstone Bridge
R1
+
vs
-
A
R2
RA
+ vm + +
v2 vB
- -
B
RB
31
R2
vs
R1 + R2
vB =
RB
vs
RA + RB
KVL : v2 + vm + vB = 0 vm = v2 vB = vs
RB
R2
R1 + R2 RA + RB
vm = vs
R1 + R2 RA + RB
R1
Note that we do not need to know value of vs to nd RB . This way, RB
is measure within the accuracy of resistors, R1 , R2 , and RA .
1.3
In this section, we learn two methods, node-voltage and mesh-current methods, which reduce the number of equations to either number of KCLs or
number of KVLs.
1.3.1
Node-Voltage Method
32
v1
vb
v2
va
vc
v3
Reference node
( g r o u nd )
Figure 1.30: Node voltage: voltage between that node and the reference.
33
i1
v1
+
vs
-
is
i2
v2
i1
v1
+ v1 2 R
- v2 1 +
i2
v2
Current Source: The current owing through current source is independent of voltage across it and, therefore, is independent of node voltages.
From the gure, it is obvious that the current leaving node 1 is i1 = is and
current leaving node 2 is i2 = is .
Voltage Source: The voltage across a voltage source is constant and
is independent of current owing through it. The i-v characteristics of
the voltage source cannot be used to nd the current owing through it.
Voltage sources should be treated in a special manner that will be discussed
later.
Resistor: Resistors follow Ohms Law. We need to calculate the voltage
across the resistor rst and nd the current using the Ohms Law. For
example, if we want to nd current leaving node 1, i1 , we rst nd voltage
v12 :
v12
v1 v2
v12 = v1 v2 i1 =
=
R
R
Alternatively, if we want to nd the current leaving node 2, i2 , we calculate
v21 rst:
v21
v2 v1
v21 = v2 v1 i2 =
=
R
R
In summary, we have found that we can dene node voltages as circuit
variables and:
Voltage across each element can be written in terms of node voltages.
34
v1
is1
i2
R2
i2
i1
i3
R1
R3
v2
is2
Example: The circuit in Figure 1.32 has three nodes. We choose the
node at the bottom as the reference (ground).The voltages at the other
nodes are denoted by v1 and v2 . We write KCL at these two nodes:
v1 v2 v1 0
+
= is1
R2
R1
v2 0 v2 v1
KCL 2 : is2 + i3 + i2 = 0
+
= is2
R3
R2
KCL 1 : i2 + i1 is1 = 0
The above are two equations in two unknowns (node voltages v1 and
v2 ). Note that a circuit with 5 elements which, in principle, has 10 circuit
variables and 10 equations to be solved is reduced to 2 equations in 2
unknowns. The previous equations can be rearranged as:
1
1
1
+
+ v2
= is1
v1
R1 R2
R2
1
1
1
+ v2
= is2
v1
+
R2
R2 R3
35
G = ..
..
..
.. ,
.
.
.
.
Gn1 Gn2 . . . Gnn
v1
v
2
v = .. ,
.
vn
is1
i
s2
is = ..
.
isn
36
5A
v1
7A
12
v2
v3
12
17A
is , we obtain
4 1 0
1
G=
1 6 2 ,
12
0 2 7
v1
v = v2 ,
v3
2
is = 5
17
From
Gv = is
we obtain the elements of vector v as v1 = 12V, v2 = 24V, and v3 = 36V.
Using Ohms law, we get
i=
v2 v3
= 2A
6
v1
6K
v2
2K
-
3V
v3
+
20V
37
4K
6mA
38
= 20
v1
= 3
v2 v3
6K
v2
3V
v3=v2-3
+
2K
20V
4K
6mA
Noting v1 = 20, we can write the value of the node voltage on the circuit.
Also, using v2 v3 = 3, we can arrive at v3 = v2 3 as shown in the gure.
Now, we only have one unknown, v2 , which can be found by writing KCL
on the supernode:
KCL for supernode 2 & 3 : 6 103 +
v3 3
v2
v2 20
+
+
=0
4 103 2 103 6 103
The above is one equations in one unknown which can be solved rapidly.
39
Notes:
1. Voltage sources actually simplify the node voltage equations. By
using the i-v characteristic equations of IVS on the circuit diagram, we can
reduce the number of unknown node voltages (thus, number of equations
to be solved) to N 1 NIV S , where NIV S is the number of IVS.
2. The situation is much simpler for voltage sources that are connected
between a node and ground (such as the 20-V source in the above circuit).
As the location of the reference node is arbitrary, a good rule is: Choose
the reference node as the node with maximum number of voltage sources
attached to it.
3. While we do not need to write KCL at the nodes connected to a
voltage node in order to nd the node voltages, we need to use KCL if we
want to nd the current through a voltage source. For example, in the
circuit above, current i can be found by writing KCL at node 3 (after we
have found all node voltages):
i 6 103 +
v2 3
= 0
4 103
i = 4mA
Summary for Node Voltage Method
1. Identify nodes and supernodes.
a. Choose the reference node as the one with maximum number of
voltage sources attached to it.
b. Use i-v characteristic equations of IVS to nd node voltage values
and reduce the number of unknowns.
2. Write KCL at each node or supernode.
3. Solve node-voltage equations.
4. Calculate problem unknowns from node voltages. If you need to calculate the current in a voltage source you may have to write KCL at a
node connected to that voltage source.
40
1.3.2
Mesh-Current Method
R2
R1
i2
i3
i1
i2
R3
vs1
vs2
i3
i1
i2
41
Ri = vs
i2
i1
6
+
9V
-
16V
-
i2
+
-
6V
Figure 1.37:
method.
42
The problem unknowns, i and v can now be found from the mesh currents:
i = i1 i2 = 1A
v = 2i1 = 4A
Mesh-Currents Method for Circuits with Current Sources
1A
i1
2
+
10V
-
i2
2
2A
i3
6
43
The second equation can be found by noting that KVL can be written over any closed loop. We dene a supermesh as the combination of two meshes which have a current source on their boundary
as shown in the Figure 1.39. While KVL on mesh 2 or on mesh
3 both include the voltage across the 2-A current source that is
unknown, KVL on the supermesh does not include that, we have:
Supermesh 2 and 3 : 2(i2 i1 ) + 2(i3 i1 ) + 6i3 10 = 0
1A
i1= -1
2A
+
10V
i2=i3-2
i3
Totally, we have
= 1
i1
= 2
i i2
3
4i1 + 2i2 + 8i3 = 10
44
Principle of Superposition
If a linear circuit is driven by more than one independent source, the response of the circuit can be written as the sum of the responses of the
circuit to individual sources with all other sources having their strength set
to zero. Note that a source is set to zero does not mean removing it.
45
46
10
+
v
+
3A
15V
-
1A
(a)
10
+
v
+
15V
(b)
10
+
v
3A
(c)
Figure 1.40: Example of superposition. (a) Original
circuit. (b) Setting the current source to zero: open
circuit. (c) Setting the voltage source to zero: short
circuit.
47
i=0
i=0
+
+
+
Subcircuit
vT
v = voc
v = voc
(a)
Subcircuit
i = isc
i = isc
+
iN
v=0
-
RN
v=0
-
(b)
Figure 1.41: Two-terminal subcircuit and its Thevenin/Norton equivalent
circuits. (a) Thevenin equivalent circuit. (b) Norton equivalent circuit.
48
+
25V
-
20
3A
Thevenins Theorem: Find two of the following three: RT , voc , and isc .
a) Find RT by setting the sources to zero (see Figure 1.43)
5
+
25V
-
20
20
3A
3A
5//20 = 4
RT = 8
49
voc
i=0
v
+
25V
-
20
3A
voc
-
Using node-voltage method (note that the voltage drop across the 4
resistor is zero).
voc 25 voc 0
+
3 = 0
5
20
vT = voc = 32V
c) Find iN = isc (set v = 0)
25V
v2
4
+
+
25V
-
20
3A
isc
v=0
-
50
1.3.5
Real Sources
Rs
Ideal source
v = vs
R eal
source
vs
vs
is
v
-
(a)
Rs
v
-
(b)
Figure 1.46: Real source. (a) i-v characteristic of real source with a straight line. (b)
Circuit model of a real source.
1.4
51
v1
mv1
ni1
(a )
(b)
i1
qi1
pv1
v1
(c )
(d)
Note that the element located in the input with the controlling current
or voltage can be any element: a short circuit, an open circuit, or a resistor.
Controlled sources behave similar to ideal (or independent) sources. For
example, in the voltage-controlled voltage source in the above gure, the
output voltage is v1 no matter what current is drawn from the circuit. All
analysis method developed so far (KVL and KCL, node-voltage and meshcurrent methods, superposition, etc.) can be used for circuits containing
controlled sources, and by treating the controlled source similar to an ideal
source. In node-voltage and mesh-current methods, however, we need to
write an auxiliary equation which relates the controlling parameter to
node-voltage or mesh-current methods as is seen in the examples below.
52
Rc
Rs
io
ix
Rp
vs
-
+
-
rix
RL
vo
-
KVL : RS ix + RP ix vs = 0 ix =
Substituting for ix from the rst equation in the second and noting
vo = RL io , we get:
rRL
vo = RL io =
vs
(RS + RP )(RC + RL )
Example: Consider the circuit in Figure 1.49. Find v1 using node-voltage
method.
8i 1
-
16V
-
1 .2 5 A
+
+
v1
0.75v1
i1
Figure 1.49: Example of node-voltage method with controlled sources.
53
Circuit has 5 nodes and 2 voltage sources (one independent and one
controlled voltage source). Thus, the number of node voltage equations is
NN V = 5 1 2 = 2. We choose the reference node to be the one with
most voltage sources attached to it. Then, we get directly
vA = 16
vC = 8i1
8i 1
-
16 V
-
vA = 16V
vC = 8i1
vB
2
+
1. 2 5A
v1
0.75v1
i1
vD
Figure 1.50: Node voltages of the circuit in Figure 1.49.
Node vD :
vB 8i1 vB vD
+
1.25 = 0
2
4
3vB vD = 5 + 16i1
vD 0 vD 16 vD vB
+
+
0.75v1 = 0
4
2
4
vB + 4vD = 32 + 3v1
Two above equations are two equations in two unknowns (vB and vD ).
But they also contain the control parameters i1 and v1 . We need to write
two auxiliary equations relating these control parameters to our node
voltages:
i1 =
vD 16
,
2
v1 = vB 8i1
v1 = +vB 4vD + 64
54
4i
i
2K
32V
6K
Finding voc :
Since the circuit is simple, we proceed to solve it with KVL and KCL
(noting that i = 0) :
i1 = 0
KCL : i1 + i + 4i = 0
i2 = 0
KCL : i2 4i + i1 = 0
KVL : 32 + 2 103 i2 + 6 103 i1 + voc = 0
vT = voc = 32V
55
4i
i =0
2K
i2
32V
6K
i1
voc
Finding isc
Since the circuit is simple, we proceed to solve it with KVL and KCL:
KCL : i1 + i + 4i = 0
i1 = 5isc
KCL : i2 4i + i1 = 0
i2 = isc
3
3
KVL : 32 + 2 10 isc + 6 10 isc = 0
iN = isc = 4mA
4i
i =i s c
2K
+
32V
i2
6K
i1
v=0
56
circuits). For these cases, one has to either nd RT directly and/or directly
nd i-v characteristics of the subcircuits as is shown below for the circuit
of previous example.
Finding RT
4i
i
2K
i2
6K
i1
v
-
v
8 103
57
2K
+
i2
6K
32V
i1
1.5
58
Practical Ampliers
Amplier Saturation
Ampliers do not create power. Rather, they act as a valve adjusting
the power ow from the power supply into the load according to the input
signal. As such, the output voltage amplier cannot exceed the power
supply voltage (it is usually lower because of voltage drop across some active
elements). The fact that the output voltage of a practical amplier cannot
exceed certain threshold value is called saturation. A voltage amplier
behaves linearly, (i.e., Vo /Vi = Av : constant), as long as the output voltage
remains below the saturation voltage, Vsat
Vsat < Vo < Vsat
Note that the saturation voltage, in general, is not symmetric, i.e.,
Vsat,1 < Vo < Vsat,2 . For an amplier with a given gain, Av , the above
range of Vo translate into a certain range for Vi
Vsat < Vo < Vsat
59
since Vo = Av Vi , then
Vsat
Vsat
< Vi <
Av
Av
i.e., any amplier will enter its saturation region if Vi is raised above certain
limit.
Amplier Bandwidth
Typically, ampliers work in a certain range of frequencies. Their gain,
Av = Vo /Vi drops outside this range. The frequency response of amplier
is plotted as in Figure 1.58, where fL , fH are 3-dB cut-o frequencies. The
bandwidth of the amplier is dened as BW = fH fL .
In terms of frequency response, voltage ampliers are divided into two
categories. (1) AC ampliers which only amplify AC signals. (2) DC ampliers which amplify both DC signals and AC signals up to a certain
frequency.
60
|A |
Am
Mid-frequency range
A m / 1 .4
High-frequency
range
Low-frequency range
fL
f
fH
Rise Time
In an ideal amplier, if the input voltage is a unit step function, the output
voltage will also be a unit step function as shown. A practical amplier
cannot change its output instantaneously if the input changes suddenly. It
takes some time (a short but nite time) for the amplier output voltage to
reach its nominal level. The maximum rate of change in the output voltage
is called the rise time.
1.5.2
Feedback
Not only a good amplier should have sucient gain, its performance should
be insensitive to environmental and manufacturing conditions, should have
61
62
very large gain, (105 to 106 ), high input impedance (> 1 10M), and
low output resistance (< 100). They are constructed as a dierence
amplier, i.e., the output signal is proportional to the dierence between
the two input signals (see Figure 1.60)
Vo = A0 Vd = A0 (Vp Vn )
In Figure 1.60, Vs and Vs are power supply attachments. They set the
saturation voltages for the OpAmp circuit. Power supply ground should
also be connected to the OpAmp ground. + and - terminals of the OpAmp
are called, respectively, non-inverting and inverting terminals.
OpAmp Models
Because Ri is very large and Ro is very small, ideal model of the OpAmp
assumes Ri and Ro 0. Ideal model is usually a very good model
for OpAmp circuits. Very large input resistance also means that the input
current into an OpAmp is very small:
Ip In 0
An important feature of OpAmp is that because the gain is very high,
the OpAmp will be in the saturation region without negative feedback. For
example, take an OpAmp with a gain of 105 and Vsat = 15V. Then, for
OpAmp to be in linear region, Vi 15 105 = 150V (a very small
value). OpAmps are never used by themselves. They are always part of a
circuit which employ either negative feedback (e.g., linear ampliers, active
lters) or positive feedback (e.g., comparators). Examples below shows
several OpAmp circuits with negative feedback.
1.5.4
63
Inverting Amplier
The rst step in solving OpAmp circuits is to replace the OpAmp with
its circuit model (ideal model is usually used).
Vp = 0,
Vo = A0 Vd = A0 (Vp Vn ) = A0 Vn
Vo = 0
A0 R1
R1
Ao
Vo
R2
1
=
Vi
R1
1 + 1/A0 + R2 /(A0 R1 )
Since OpAmp gain is very large, 1/A0 1. Also if R2 and R1 are chosen
such that their ratio is not very large, R2 /R1 A0 or R2 /(A0 R1 ) 1,
then the voltage gain (voltage transfer function) of the OpAmp is
Vo
R2
=
Vi
R1
The circuit is called an inverting amplier because the voltage gain is
negative. The negative sign means that there is 180 phase shift between
input and output signals.
Note that the voltage gain of the inverting amplier is independent
of the OpAmp gain, A0 , and is only set by the values of the resistors R1
64
Vi 0
Vi
Ri =
= R1
R1
Ii
The input impedance of the inverting amplier circuit is R1 (although
input impedance of OpAmp is innite). The output impedance of the
circuit is zero because Vo is independent of RL (Vo does not change when
RL is changed).
Ii =
1.5.5
Non-inverting Amplier
Vp = Vi
Negative Feedback Vn Vp = Vi
Vn 0 Vn Vo
+
=0
R1
R2
Substituting for Vn = Vi , we get
R2
Vi + Vi Vo = 0
R1
Finally,
Av =
Vo
R2
=1+
Vi
R1
Vi
Ii
Vi
Vo
Ip
Vp
Vo
Vd
AVd
R2
65
Vn
R1
R2
R1
Input Resistance
Since Ii = Ip = 0. Therefore, Ri .
Output Resistance
Vo is independent of RL , so Ro = 0. Note that Ri and Ro = 0 should be
taken in the context that we are using an ideal OpAmp model. In reality,
the above circuit will have input and output resistances equal to that of
the OpAmp itself.
1.5.6
Voltage Follower
Vi
+
-
Vo
66
Inverting Summer
Rf
V2
V1
R2
R1
Vn
Vp
Vo
+
+
-
AVd
Vp = 0
Negative Feedback Vn Vp = 0
Vn V1 Vn V2 Vn Vo
+
+
=0
R1
R2
Rf
67
Finally,
Rf
Rf
V1
V2
R1
R2
So, this circuit adds (sums) two signals. An example of the use of this
circuit is to add a DC oset to a sinusoidal signal.
Vo =
1.5.8
Non-Inverting Summer
V2
V1
R2
Vp
R1
Vn
Vo
+
-
AVd
Rf
Rs
Negative Feedback: Vn Vp
Vp V1 Vp V2
+
= 0 Vp
R1
R2
Vn 0 Vn Vo
+
=0
R1
Rf
1
1
+
R1 R2
Vo =
Rf
1+
Rs
V2
V1
+
R1 R2
Vn
68
1.5.9
Dierence Amplier
Rf
V1
R1
Vn
Vp
V2
R2
Vo
+
+
-
AVd
R3
Negative Feedback: Vn Vp
Vp V2 Vp 0
+
=0
R2
R3
Vn Vp =
R3
V2
R2 + R3
Vn V1 Vn Vo
+
=0
R1
Rf
If one choose the resistors such that R3 /R2 = Rf /R1 , then
Vo =
1.5.10
Rf
(V2 V1 )
R1
Current Source
Negative Feedback: Vn Vp = Vs
Vs
Vn
= : constant
R
R
For a xed value of Vs , the current IL is independent of value of RL and
output voltage Vo . As such, this circuit is an independent current source.
The value of the current can be adjusted by changing Vs (for a xed RL .
Therefore, this circuit is also a voltage to current converter or a transconductance amplier.
iL =
Vp
Vs
Vn
69
+
IL
+
Vo
RL
-
Vs
R1
Vn
Vp
R2
R3
RL
IL
+
Vo
-
The problem with the above current source is that the load is not
grounded. This may not be desirable in some cases. This circuit here
is also a current source with a grounded load if Rf /R1 = R3 /R2 .
1.6
70
i=
dv
dq
=C
dt
dt
Note that the current i ows as long as the voltage (and charge, q = Cv)
changes in time. At the DC steady-state condition, dv/dt = 0 and capacitor
current, i = 0, and capacitor acts as an open circuit.
Energy stored in a capacitor, W , can be found from the denition of
electric power:
1
W (t) = Cv 2 (t)
2
71
t = t0
t < t0
i
Is
+
v
i
R
Is
+
v
-
t > t0
i
Is
+
v
-
+
have a discontinuity at t = t0 as i(t
0 ) = i(t0 ) and v(t0 ) = v(t0 ).
This behavior is in general true for all resistive circuit. There is a discontinuity in both voltage and current waveforms at the time that the switch
72
is thrown. This is not true for circuits with capacitors (or inductors). For
capacitors, the i-v characteristics equation i = Cdv/dt implies that if there
were a discontinuity in the voltage, an innite amount of current should
ow through the capacitor. Therefore, the voltage waveform across a capacitor should be continuous. This means that if a switch is thrown at time
+
t0 , we should have v(t
0 ) = v(t0 ). Note the capacitor current waveform can
have a discontinuity.
Inductor
A typical inductor is made of a a wire wrapped around a core of magnetic
material. As current ows through the wire, a magnetic eld is produced.
As such, an inductor can store energy in the form of magnetic eld. The
i-v characteristics equation for an ideal inductor is given by Faradays Law
(wire in an ideal inductor has no resistance):
di
Inductance, L. Unit: Henry (H)
dt
using passive sign convention. Energy stored in an inductor, W , can be
found as
1
W (t) = Li2 (t)
2
For inductors, the i v characteristics equation v = Ldi/dt implies that
if there were a discontinuity in the current waveform, an innite amount
of voltage should appear across the inductor. Therefore, the current waveform in an inductor should be continuous. This means that if a switch is
+
thrown at time t0 , we should have i(t
0 ) = i(t0 ). Note the inductor voltage
waveform can have a discontinuity.
v=L
Comments
Circuits containing one or more capacitors and/or inductors are called
dynamic circuits. Voltage and currents in dynamic circuits are, in
general, functions of time, i.e., their value change as time progresses.
All analysis method (KVL, KCL, circuit reduction, node-voltage and
mesh current methods) can be applied to dynamic circuits with no
modication.
Application of the analysis methods to a dynamic circuit results in
a set of linear dierential equations (as opposed to a set of algebraic
73
74
t = t0
Is
t < t0
Is
t > t0
Is
t = t0
1A
200
200
iL
+
vc
-
vc
1A
75
50
iL
+
vc
-
50
76
conditions). We draw the circuit for t0 < t (closed switch) and replace the
capacitor with an open circuit and the inductor with a short circuit. We
then proceed to solve the resistive DC steady state circuit.
The resulting circuit is current divider and iL and vC can be found readily
from current-divider formulas. Alternatively, using node-voltage method,
we have
vc 0 vc 0
+
1 = 0 vc = 40V
50
200
vc 0
= 0.8A
iL =
50
Note that since the circuit is in DC steady state, values of vC and iL are
constant in time. Using the no-discontinuity conditions for vC and iL , we
can nd the initial conditions for the time-dependent circuit as
vc (t = t+
0 ) = 40V
iL (t = t+
0 ) = 0.8A.
Chapter 2
Introduction to Signals and Systems
2.1
Classication of Signals
or
signal power (if signal energy is innite: signal amplitude does not
0, as |t| ) which is time average of energy (if it exists). We
dene power Pf of f (t) as
1 T /2 2
f (t)dt
Pf = lim
T T T /2
for a complex valued signal f (t), it is
1 T /2
Pf = lim
|f (t)|2 dt.
T T T /2
For periodic signals, the time averaging need to be performed only over
one period. Square root of signal power is rms (root mean square) value
of the signal.
77
78
2.2
2.2.1
Time Scaling
Combined Operations
2.3
79
The below signal models serve as a basis for representing other signals and
their use can simplify many aspects of the signal and systems.
2.3.1
The unit step function u(t) is very useful in representing causal signals
and signals with dierent mathematical description over dierent intervals.
This function is dened by
1 t0
u(t) =
0 t<0
2.3.2
t = 0
(t)(t T ) = (T ) (t T ) = (T )
which states that the area under the product of a function with a unit
impulse is equal to the value of that function at the instant where the
impulse is located (assuming the function to be continuous at the impulse
location).
80
2.3.3
Exponential Function
One of the most important functions in signals and systems is the exponential function est , where s = + j is thecomplex frequency. Therefore
est = et+jt = et (cos t + j sin t)
If s is the conjugate of s, then
et cos t = (est + es t )
2
The exponential function encompasses a large class of signals as
A constant k = ke0t (s = 0)
A monotonic exponential et
A sinusoid cos t ( = 0, s = j)
An exponentially varying sinusoid et cos t (s = j)
2.4
81
Every signal f (t) can be expressed as a sum of even and odd components because
1
1
f (t) = [f (t) + f (t)] + [f (t) f (t)] = fe (t) + fo (t)
2
2
2.5
Classication of Systems
A system processes input signals to modify them or extract additional information from them to produce output signals (response). A system may
be made up of physical components (hardware) or may be an algorithm
that computes an output signal from an input signal (software).
A system is characterized by its input, its output, and the rules of operation describing its behavior. Example, in electrical engineering, the rules
are the i-v characteristic of resistors, capacitors, inductors, transistors, etc.,
as well as the rules for interconnection such as Kirchhos laws. Using these
rules we derive mathematical equations relating the outputs to the inputs.
These mathematical equations represent a mathematical model of the system.
The study of systems consists of three major areas:
Mathematical modeling
Analysis problem: To determine the system outputs for the given inputs and a given mathematical model of system.
Design problem (or synthesis problem): To construct a system which
produce a desired outputs for the given inputs.
82
Linear systems are characterized by the linear property, which implies superposition. If several causes (such as various inputs and initial conditions)
are acting on a linear system, the total eect (response) is the sum of the
responses from each cause, assuming that all causes are absent. A system
is nonlinear if it is not linear.
Response of a Linear System
A linear system output for t 0 must be the sum of two components:
rst, the zero-input response component that results only from the
initial condition at t = 0 with the input f (t) = 0 for t 0
and then the zero-state response component that results only from
input f (t) for t 0 when the initial conditions (at t = 0) are assumed
to be zero.
The property of linear systems which permits the separation of an output
into components resulting from the initial conditions and from the input is
called the decomposition property.
(See example 1.9, p. 81)
In general, a system described by a dierential equation of the form
dn1 y
dm f
dm1 f
df
dn y
+
a
+
+
a
y
=
b
+
b
+
+
b
+ b0 f
n1
0
m
m1
1
dtn
dtn1
dtm
dtm1
dt
is a linear system. The coecients ai and bi can be constant or functions
of time.
2.5.2
83
The time-invariant systems are characterized by the fact that the system
parameters do not change with time. For example if the coecients ai and
bi of the linear system are constant, then the linear system is linear timeinvariant (LTI) system. If the system parameters are function of time, then
the system is time-varying system.
2.5.3
For memoryless (instantaneous) systems, the system response at any instant t depends only on the present value (at t) of the input. For systems
with memory (dynamic systems), the system response at any instant t depends not only on the present value (at t) of the input but also on the past
values of the inputs (value before t).
2.5.4
A system whose input and output signals are analog is an analog system; a
system whose input and output signals are digital is a digital system.
2.5.5
Systems whose inputs and outputs are continuous-time signals are continuoustime systems. On the other hand, systems whose inputs and outputs are
discrete-time signals are discrete-time systems.
2.5.6
84
2.5.7
A causal (or physical or non-anticipative) system is one for which the output
at any instant t0 depends only on the value of the input f (t) for t t0 .
In other words, the value of the output at present instant depends only
on the past and present values of the input f (t), not on its future values.
A system that violates the conditions of causality is called a noncausal or
anticipative system.
Any practical system operating in real-time (In real-time operations, the
response to an input is simultaneous with the input itself) must be causal.
Noncausal systems are unrealizable in real time (but realizable with some
time delay in response). Noncausal systems with independent variables
other than time (e.g. space) are realizable.
2.6
To construct a system model, we must study the relationships between different variables in the system. In electrical engineering, we must determine
a satisfactory model for i-v relationship of each element (such as Ohms
law) and the various constraints on voltages and currents when several elements are interconnected (such as KCL, KVL). For all these equations,
we eliminate unwanted variables to obtain equations relating the desired
output variable to the inputs.
(See example 1.10, p. 89)
It is convenient to use a compact notation D for the dierential operator
d
dt
dy
dt
Dy(t)
d2 y
D2 y(t)
2
dt
..
.
For example,
d2 y
df
dy
+
2y(t)
=
,
+
3
dt2
dt
dt
can be expressed as
(D2 + 3D + 2)y(t) = Df (t)
1
to represent integration
D
t
y( )d
1
y(t)
D
85
Chapter 3
Time-Domain Analysis of
Continuous-Time Systems
In this chapter, we consider time-domain method to analyze linear, time
invariant, continuous-time (LTIC) systems.
We consider linear dierential systems. This is the class of LTIC, for
which the input f (t) and the output y(t) are related by linear dierential
equations as
dn1 y
dm f
dm1 f
df
dn y
+b0 f (t) (3.1)
+a
+
+a
y
=
b
+b
+
+b
n1
0
m
m1
1
dtn
dtn1
dtm
dtm1
dt
where all the coecients ai and bi are constant. Using notation D to
represent d/dt, we obtain
(Dn +an1 Dn1 + +a1 D+a0 )y(t) = (bm Dm +bm1 Dm1 + +b1 D+b0 )f (t)
(3.2)
or
Q(D)y(t) = P (D)f (t)
(3.3)
where the polynomials Q(D) and P (D) are
Q(D) = Dn + n1 Dn1 + + a1 D + a0
P (D) = bm Dm + bm1 Dm1 + + b1 D + b0
(3.4)
87
3.1
Zero-Input Response
The zero-input response y0 (t) is the solution of equation (3.1) when the
input f (t) = 0 so that
n
(D + n1 D
n1
Q(D) = 0
+ + a1 D + a0 )y0 (t) = 0
(3.5)
(3.6)
(3.7)
The exponentials
ei t
(i = 1, 2, n)
in the zero-input response are the characteristic modes of the system. There
is a characteristic mode for each characteristic root of the system, and the
zero-input response is a linear combination of the characteristic modes of
the system
3.1.1
Repeated Roots
88
(3.8)
(3.9)
(3.10)
(3.11)
Complex Roots
For a real system, complex roots must occur in pairs of conjugates if the
coecients of the characteristic polynomial Q() are to be real. The zeroinput response corresponding to this pair of complex conjugate roots is
y0 (t) = c1 e(+j)t + c2 e(j)t
(3.12)
For a real system, the response y0 (t) must also be real. This is possible
only if c1 and c2 are conjugates. Let
c
c
c1 = ej and c2 = ej
2
2
Then
c j (+j)t c j (j)t
+ e e
e e
2
2
c t j(t+)
j(t+)
= e e
+e
2
y0 (t) =
= cet cos(t + )
(3.13)
89
3.2
Zero-State Response
h(t)
(t nt)
h(t nt)
(3.14)
f (n )(t n ) [f (n ) ]h(t n )
input
output
The last pair in (3.14) represents the system response to only one of
the impulse components of f (t). The total response y(t) is obtained by
summing all such components as depicted in Figure 3.1(d).
lim
n=
f (n )(t n )
The input f (t)
lim
f (n )h(t n )
n=
The output y(t)
(3.15)
f (t)
y(t)
(3.16)
90
91
Note that the assumption used to derive (3.17) is linear and time-invariant
system.
3.2.1
Convolution Integral
The convolution integral of two functions f1 (t) and f2 (t) denoted by f1 (t) f2 (t)
is dened by
f1 (t) f2 (t)
f1 ( )f2 (t )d
(3.18)
(3.19)
Distributive property:
f1 (t) [f2 (t) + f3 (t)] = f1 (t) f2 (t) + f1 (t) f3 (t)
(3.20)
Associate property:
f1 (t) [f2 (t) f2 (t)] = [f1 (t) f2 (t)] f3 (t)
(3.21)
Shift property:
If
f1 (t) f2 (t)
= c(t)
then
f1 (t) f2 (t T )
f1 (t T ) f2 (t)
= c(t T )
= c(t T )
and
f1 (t T1 ) f2 (t T2 ) = c(t T1 T2 )
f ( )(t )d = f (t)
(3.22)
(3.23)
Width property: If the durations (widths) of f1 (t) and f2 (t) are T1 and
T2 respectively, the the duration of f1 (t) f2 (t) is T1 + T2 .
92
(3.24)
f ( )h(t )d
t0
0
y(t) = f (t) h(t) =
(3.25)
0
t<0
Because of commutative property, we can also express (3.25) as
t
h( )f (t )d
t0
0
y(t) =
0
t<0
(3.26)
The results show that if f (t) and h(t) are both causal, the response y(t)
is also causal.
(See example 2.4, p. 123)
Convolutional Table
The below table lists several pairs of signals and their resulting convolution.
It helps to determine conveniently a system response y(t) to an input f (t),
without performing the integral.
Multiple Inputs
Multiple inputs to LTI systems can be treated by applying superposition
principle. Each input is considered separately, with all other inputs assumed to be zero. When all the inputs are applied simultaneously, the
system output will be the sum of all these individual system response.
93
94
3.2.2
95
96
3.3
Total Response
The total response of a linear system can be expressed as the sum of its
zero-input and zero-state components
n
Total Response =
cj ej t
j=1
zero-input component
f (t) h(t)
zero-state component
(3.28)
Note that for the repeated roots and complex roots the zero-input component should be appropriate modied.
Natural and Forced Responses
The zero-input component is composed of characteristic modes. However,
the zero-state component contains also the characteristic mode terms. This
is generally true for LTI systems. When we lump together all the characteristic mode terms in the total response, it gives us a component known as
the natural response yn (t). The remainder, consisting of noncharacteristic
mode terms, is known as the forced response y (t).
Total Response = Natural Response yn (t) + Natural Response y (t)
(3.29)
3.4
In classical method, to nd the total response of the system, we solve dierential equation to nd the natural component (also known as homogeneous
solution or complementary solution) and forced components (also known
as particular solution), instead of nding the zero-input and the zero-state
components.
3.4.1
97
or
Q(D)yn (t) + Q(D)y (t) = P (D)f (t)
but yn (t) is composed entirely characteristic modes, therefore
Q(D)yn (t) = 0
so that
Q(D)y (t) = P (D)f (t)
(3.30)
The natural response, being a linear combination of the system characteristic modes, has the same form as that of zero-input response; only
its arbitrary constants are dierent. These constants are determined from
auxiliary conditions (systems initial conditions).
3.4.2
98
t>0
(3.32)
n
Kj ej t t + H()e t
(3.33)
j=1
(3.34)
(3.35)
(3.36)
99
3.5
System Stability
It shows that
lim et =
Re < 0
Re > 0
(3.38)
3.6
101
(3.39)
In practical system, the order n of the polynomial Q(D) and the order m
of the polynomial P (D) are restricted by n m. In most general case,
m = n. Therefore, (3.39) can be expressed as
(Dn +an1 Dn1 + +a1 D+a0 )y(t) = (bn Dn +bn1 Dn1 + +b1 D+b0 )f (t)
(3.40)
In the textbook (pp. 261-263), it shows that the unit impulse response h(t)
is given by
(3.41)
h(t) = bn (t) + [P (D)yn (t)]u(t)
where bn is the coecient of the nth-order term in P (D) in equation (3.40),
yn (t) is a linear combination of the characteristic modes of the system
subject to the following initial conditions:
yn(n1) (0) = 1 and yn (0) = yn (0) = yn (0) = yn(n2) (0) = 0
(3.42)
(k)
:
:
:
:
yn (0) = 1
yn (0) = 0 and yn (0) = 1
yn (0) = yn (0) = 0 and yn (0) = 1
...
yn (0) = yn (0) = yn (0) = 0 and y n (0) = 1
(3.43)
If m < n, then bn = 0.
(See example 2.3, pp. 116-117, textbook)
Comment
It shows that, the expression of h(t) for all possible values of m and n
is given by
h(t) = P (D)[yn (t)u(t)]
(3.44)
where yn (t) is a linear combination of the characteristic modes of the
system subject to the condition in (3.42).
Other simple method for determination of h(t) is to use the Laplace
transform.
Chapter 4
Signal Representation by Fourier
Series
In this chapter, we consider the issue of representing a signal as a sum of
its components by using Fourier series.
4.1
Component of a Vector
q
cx
102
103
(4.1)
(4.2)
(4.3)
(4.4)
(4.5)
(4.6)
Component of a Signal
t1 t t2
otherwise
(4.7)
(4.8)
We now select some criterion for the best approximation. We know that
the signal energy is one of possible measure of signal size, which is dened
over the interval [t1 , t2 ] as
t2
t2
Ee =
e2 (t)dt =
[f (t) cx(t)]2 dt
(4.9)
t1
t1
104
(4.10)
t2
f (t)x(t)dt
(4.11)
t1
t
where Ex = t12 x2 (t)dt is energy of x(t). Therefore, the optimum value of
c that minimizes the energy of error signal in the approximation
f (t) cx(t)
is given by equation (4.11).
We dene the real signals f (t) and x(t) are orthogonal if c = 0, or
t2
c=
f (t)x(t)dt = 0
(4.12)
t1
4.1.3
Consider the issue of approximating a complex signal f (t) in terms of another complex signal x(t) over an interval [t1 , t2 ]:
f (t) cx(t)
where the coecient c is complex (in general). Similar to the case of real
signals, it shows that (see pp.175-176, textbook)
t2
1
c=
f (t)x (t)dt
(4.13)
Ex t1
t
where Ex = t12 |x(t)|2 dt is energy of complex signal x(t). Therefore, if two
complex signals f (t) and x(t) are orthogonal, then
t2
t2
f (t)x (t)dt = 0 or
f (t)x(t)dt = 0
(4.14)
t1
4.2
4.2.1
t1
105
(4.16)
Correlation Functions
When f (t) and g(t) are similar but they are time-shifted so that nonoverlapping in time. From (4.15), we obtain
1
f (t)x(t)dt = 0
(4.18)
cn =
Ef Ex
Thus, the correlation is zero because two similar signals are disjoint. For
this reason, instead of using the integral on the right-hand of (4.15), we
dene the crosscorrelation function of two real signals f (t) and g(t) as
f g (t) =
f ( )g( t)d
(4.19)
106
(4.21)
Autocorrelation Function
Correlation of a signal with itself is called autocorrelation. The autocorrelation function f (t) of a signal f (t) is dened as
f (t) =
f ( )f ( t)d
(4.22)
4.3
Real Signals
We dene orthogonality of real signal set x1 (t), x2 (t), , xN (t) over interval
[t1 , t2 ] as
0
m = n
xm (t)xn (t)dt =
(4.23)
En m = n
N
n=1
cn xn (t)
(4.24)
107
N
cn xn (t)
(4.25)
n=1
(4.27)
n=1
cn xn (t)
(4.28)
n=1
where the coecients cn are given by (4.26). The series on the right-hand
side of (4.28) is called generalized Fourier series of f (t) with respect to
the set {xn (t)}. The set {xn (t)} is called a set of basis functions or basis
signals.
It follow also that the energy of f (t) is now equal to the sum of the
energies of its orthogonal components c1 x1 (t), c2 x2 (t), . Thus
t2
2
2
2
f (t)dt = c1 E1 + c2 E2 + =
c2n En
(4.29)
t1
n=1
It is Parsevals theorem.
Complex Signals
The above results can be generalized to complex signal as follows: A set
of functions x1 (t), x2 (t), , xN (t) is mutually orthogonal over the interval
[t1 , t2 ] if
t2
0
m = n
xm (t)xn (t)dt =
(4.30)
En m = n
t1
108
(4.31)
t2
1
cn =
f (t)xn (t)dt
(4.32)
En t1
An interesting property of the coecients c1 , . . . , cn , . . . cN is the optimum value of any coecient is independent of the number of terms used in
the approximation of a signal f (t). It means we can continue to add terms
to the approximation without disturbing the previous terms.
where
4.4
(4.37)
109
or
f (t) = a0 +
t1 t t1 + T0
(4.38)
n=1
where 0 = 2
T0 .
Using (4.26) and (4.34)-(4.36), we can determine the Fourier coecients
a0 , an and bn as
1 t1 +T0
a0 =
f (t)dt
(4.39)
T0 t 1
2 t1 +T0
f (t) cos n0 tdt
n = 1, 2, 3, . . .
(4.40)
an =
T0 t 1
2 t1 +T0
bn =
f (t) sin n0 tdt
n = 1, 2, 3, . . .
(4.41)
T0 t 1
Compact Trigonometric Fourier Series
The trigonometric Fourier series in (4.38) contains sine and cosine terms of
the same frequency, then we can express in the compact form the trigonometric Fourier series as
f (t) = C0 +
Cn cos(n0 t + n )
t1 t t1 + T0
(4.42)
n=1
(4.43)
for all t
(4.44)
110
4.4.1
111
112
4.4.3
4.5
n = 0, 1, 2, . . .
f (t) =
Dn ejn0 t
(4.50)
f (t)ejn0 t dt
(4.51)
n=
where
1
Dn =
T0
t1 +T0
t1
1
jn
2 Cn e
1
jn
2 Cn e
1
Dn = (an jbn )
2
(See example 3.6, pp. 208, textbook)
(4.52)
(4.53)
4.5.1
113
For the exponential spectra, we plot the magnitude |Dn | and the angle Dn
of Dn as functions of .
Compare (4.39) and (4.51) for n = 0 shows that
D0 = a0 = C0
(4.54)
Dn = n
and
Thus,
Dn = |Dn |ejn
n = 0
Dn = n
(4.55)
(4.56)
(4.57)
Parsevals Theorem
The power of periodic signal f (t) in terms of its trigonometric Fourier series
is given by
1 2
2
C
(4.58)
Pf = C0 +
2 n=1 n
It is also shown that the power of periodic signal f (t) in terms of its exponential Fourier series is given by
Pf =
n=
|Dn |2
(4.59)
114
D02
+2
|Dn |2
(4.60)
n=1
(See example 3.9, pp. 215, textbook. Study how to compute harmonic
distortion).
Limitations of the Fourier Series Method of Analysis
We have studied in this Chapter a method of representing a period signal
as a weighted sum of everlasting exponentials whose frequency lie along
the j-axis in the s-plane. This Fourier series representation is valuable in
many applications. However, as a tool for analyzing linear systems, it has
some limitations as
The Fourier series can be used only for periodic signal. However, practical signals are aperiodic. This limitation can be overcome by representing aperiodic signals in terms of everlasting exponentials through
the Fourier integral (which may be considered as an extension of the
Fourier series). We will consider the Fourier integral in Chapter 5.
The Fourier series technique can be applied for stable systems. It
can not handle easily unstable or marginally stable systems. This
limitation can be overcome by using Laplace integral, which will be
discussed in Chapter 6.
Chapter 5
Continuous-Time Signal Analysis:
The Fourier Transform
In previous chapter, we considered the issue of representing periodic signals as a sum of sinusoids or exponentials. In this chapter we extend this
representation to aperiodic signals.
5.1
T0
(5.1)
Thus, the Fourier series representing fT0 (t) will also represent f (t) in the
limit T0 . The exponential Fourier series for fT0 (t) is given by
fT0 (t) =
n=
115
Dn ejn0 t
(5.2)
f(t)
0
(a)
fT0(t)
-T 0
T0
(b)
Figure 5.1: Construction of periodic signal fT0 (t) by periodic extension of aperiodic signal f (t).
where
1
Dn =
T0
T0 /2
T0 /2
fT0 (t)ejn0 t dt
117
(5.3)
and
2
(5.4)
T0
Observe that integrating fT0 (t) over (T0 /2, T0 /2) is the same as integrating f (t) over (, ). Therefore, (5.3) can be expressed as
1
f (t)ejn0 t dt
(5.5)
Dn =
T0
0 =
(5.6)
1
F (n0 )
T0
(5.7)
It means that the Fourier coecients Dn are (1/T0 ) times the samples of
F () uniformly spaced at interval of 0 . Therefore, (1/T0 )F () is the
envelope for the coecients Dn .
Substitution of (5.7) in (5.2) yields
F (n0 ) jn0 t
e
fT0 (t) =
T
0
n=
(5.8)
1
f (t) = lim fT0 (t) = lim
F (n)e(jn)t
T0
0 2
n=
(5.11)
The sum on the right-hand side of (5.11) can be seen as the area under
the function F ()e(j)t . Therefore
1
F ()ejt d
(5.12)
f (t) =
2
The integral on the right-hand side is called the Fourier integral. From
(5.12), it means that an aperiodic signal f (t) is represented by a Fourier
integral (rather then Fourier series). This integral is basically a Fourier
series (in the limit) with fundamental frequency 0. The amount of
the exponential ejnt is F (n)/2. Thus, the function F () given
by (5.6) acts as a spectral function.
We call F () the direct Fourier transform of f (t) and f (t) the inverse
Fourier transform of F (). We call f (t) and F () are a Fourier transform
pair. Symbolically, this statement is expressed as
F () = F[f (t)] and f (t) = F 1 [F ()]
or
f (t) F ()
where
F () =
f (t)ejt dt
(5.13)
and
1
f (t) =
2
F ()ejt d
(5.14)
F ()
(5.15)
where |F ()| is the amplitude and F () is the phase of F (). From (5.13),
we have
f (t)ejt dt
(5.16)
F () =
From (5.13) and (5.16), it follows that if f (t) is a real function of t, then
F () and F () are complex conjugates. Therefore
|F ()| = |F ()|
F () = F ()
(5.17)
119
Thus, for real f (t), the amplitude spectrum |F ()| is an even function, and
the phase spectrum is an odd function of . The transform F () is the
frequency-domain specication of f (t).
Note that Fourier spectrum of a periodic signal has nite amplitude
and exists at discrete frequency (fundamental frequency and its multiples),
while the spectrum of an aperiodic signal continuous.
Thus,
y(t) = H(s)est
Setting s = j in this equation yields
H(j) = H() =
h(t)ejt dt
(5.20)
(5.21)
(5.23)
Thus, transmission of a signal through a linear system is seen as transmission of various sinusoidal components of the signal through the system.
5.2
F ()
eat u(t)
1
,
a + j
a>0
eat u(t)
1
,
a j
a>0
ea|t|
2a
,
a2 + 2
a>0
teat u(t)
1
,
(a + j)2
tn eat u(t)
n!
,
(a + j)n+1
(t)
2()
ej0 t
2( 0 )
cos 0 t
[( 0 ) + ( + 0 )]
a>0
a>0
10 sin 0 t
j[( + 0 ) ( 0 )]
11 u(t)
() +
12 sgnt
2
j
13 cos 0 tu(t)
j
[( 0 ) + ( + 0 )] + 2
2
0 2
14 sin 0 tu(t)
0
[( 0 ) ( + 0 )] + 2
2j
0 2
0
,
(a + j)2 + 02
1
j
a>0
sinc
21
Wt
2
/2 2
2
2
sinc
2
4
(t nT )
n=
22 et
a>0
rect
2W
W
sinc(W t)
18
t
19
W
20
sinc2
2
a + j
,
(a + j)2 + 02
121
2W
0
( n0 ),
n=
2 2
2e /2
0 =
0
T
5.3
Operation
f (t)
F ()
Addition
f1 (t) + f2 (t)
F1 () + F2 ()
Scalar multiplication
kf (t)
kF ()
Symmetry
F (t)
2F ()
Scaling (a is real)
f (at)
1
F
|a|
a
Time shift
f (t t0 )
F ()ejt0
f (t)ejt0
F ( 0 )
Time convolution
f1 (t) f2 (t)
F1 ()F2 ()
Frequency convolution
f1 (t)f2 (t)
1
F1 () F2 ()
2
Time dierentiation
dn f
dtn
(j)n F ()
f (x)dx
Time integration
5.4
F ()
+ F (0)()
j
If f (t) and y(t) are the input and output of an LTIC system with transfer
function H(), then
Y () = H()F ()
(5.24)
123
(5.25)
(5.27)
Y () = F ()H()
(5.28)
H() = kejtd
(5.29)
But
Therefore
This is the required transfer function of the system for distortionless transmission. It follows that
|H()| = k
(5.30)
H() = td
This result shows that for distortionless transmission, the amplitude response |H()| must be a constant, and the phase response H() must be
a linear function of .
5.5
H() = ejtd
(5.31)
(5.32)
The response h(t) begins even before input applied at t = 0, then this ideal
low pass lter is noncausal and therefore physically unrealizable.
In the frequency domain, this condition equivalent to the well-known
Paley-Wiener criterion, which states that the necessary and sucient condition for the amplitude response |H()| to be realizable is
| ln |H()||
d <
(5.33)
1 + 2
If |H()| = 0 over any nite band, | ln |H()| = over that band, and the
condition (5.33) is violated. If, however, H() = 0 at a single frequency
(or a set of discrete frequencies), the condition in (5.33) may still nite.
According to this criterion, ideal lters (low pass, high pass, band pass
lters) are unrealizable.
125
The unit impulse response h(t) in (5.32) is not realizable. One approach
is to cut o the tail of h(t) for t < 0. The resulting causal impulse response
h(t)
is
= h(t)u(t)
h(t)
The truncation operation, however, creates some unsuspected problems as
discussed in next section.
In practice, we can realize a variety of lter characteristics to approach
ideal characteristics. Practical (realizable) lter characteristics are gradual,
without jump discontinuities in amplitude response. We will study them
in Chapter 7.
5.6
We often need to truncate data in diverse situation from numerical computation to lter design. Data truncation can occur in both time and
frequency domain.
Window Functions
Truncation operation may be regarded as multiplication a signal of a large
width by a window function of a smaller nite width.
Consider a signal f (t) and a window function w(t). If f (t) F ()
and w(t) W (), and if the windowed function fw (t) Fw (), then
fw (t) = f (t)w(t)
and
1
F () W ()
2
Truncation operation using window function causes two eects:
Fw () =
leakage or sidelobe. We can achieve a given leakage behavior by selecting a suitably smooth window (e.g. Hanning or Hamming windows).
For a given window width, the remedies for the two eects are incompatible. For example, among all the windows of a given width, the rectangular
window has the smallest spectral spread, but has high level sidelobes. which
decay slowly. A smooth window of the same width has smaller and faster
decaying sidelobes, but it has wider mainlobe. However, we can remedy
both the side eects of truncation by selecting a suitably smooth window
of sucient width.
There are several well-known window functions, such as Bartlett (triangular), Hanning, Hamming, Blackman and Kaiser, which truncate the data
gradually. These window oer dierent tradeos with respect to spectral
spread (A), the peak sidelobe magnitude (C), and the leakage rollo rate
(B) as indicated below.
Window
w(t)
Rectangular
t
rect
T
4
T
13.3
8
T
12
26.5
8
T
18
31.5
8
T
42.7
8
T
18
58.1
11.2
T
59.9
Bartlett
t
2T
Hanning
2t
0.5 1 + cos
T
Hamming
2t
0.54 + 0.46 cos
T
2t
+
0.42 + 0.5 cos
T
4t
+0.08 cos
T
Blackman
Kaiser
I0 1 4(t/T )2
I0 ()
1 10
B
[dB/oct]
C
[dB]
( = 8.168)
5.7
127
Signal Energy
It follows,
f (t) = f (t)
Therefore, from (5.36), the Fourier transform of the autocorrelation is its
energy spectral density |F ()|2
f (t) |F ()|2
(5.37)
Chapter 6
Continuous-Time System Analysis:
The Laplace Transform
The Fourier transform allows us to represent a signal f (t) as a continuous
sum of exponentials of the form ejt , whose frequencies are restricted to
the imaginary axis (s = j). This representation is valuable in the analysis
and processing of signals.
In system analysis, however, the Fourier transform has some disadvantages. First, the Fourier transform exists only for a restricted class of
signals, and therefore, can not be used for such inputs (such as growing
exponential signals). Second, the Fourier transform can not be used easily
to analyze unstable or marginally stable systems.
6.1
The basic reason for both diculties is that for some signals. such as
et u(t), ( > 0), the Fourier does not exist because the ordinary sinusoids
or exponentials of the form ejt are incapable of synthesizing exponentially
growing signals.
This problem can be resolved if we can use the basis signal of the form est
(instead of ejt ). This is known as the bilateral Laplace transform, where the
frequency variable s = j is generalized to s = + j. Such generalization
permits us to use exponentially growing sinusoids to synthesize a signal
f (t).
128
6.1.1
129
(6.1)
1
f (t) =
F (j)ejt d
2
Consider now the Fourier transform of f (t)et ( real)
t
t jt
f (t)e e
dt =
f (t)e(+j)t dt
F[f (t)e ] =
and
(6.2)
(6.3)
(6.4)
The pair of equations (6.7) and (6.8) is known as bilateral Laplace transform
pair or two-sided Laplace transform pair. The bilateral Laplace transform
is denoted symbolically as
F (s) = L[f (t)] and f (t) = L1 [F (s)]
or
f (t) F (s)
y(t) Y (s)
then
Y (s) = F (s)H(s)
(6.9)
Region of Convergence
The region of convergence for F (s) is the set of values of s for which the
integral in (6.8) converges. See example 6.1, pp. 366, textbook.
Unilateral Laplace Transform
The unilateral Laplace transform is a special case of the bilateral Laplace
transform, where all signals are restricted to being causal. The unilateral
Laplace transform is dened as
F (s)
f (t)est dt
(6.10)
0
(6.11)
We can nd the inverse transform from the below table. Most of the transform F (s) of practical interest are rational functions, that is, ratio of polynomials in s. Such functions can be expressed as a sum of simpler functions
of the form listed in the table by using partial fraction expansion.
131
f (t)
F ()
(t)
u(t)
1
s
tu(t)
1
s2
tn u(t)
et u(t)
1
s
tet u(t)
1
(s )2
tn et u(t)
n!
(s )n+1
8a
cos btu(t)
s
s2 + b 2
8b
sin btu(t)
b
s2 + b 2
9a
s+a
(s + a)2 + b2
9b
b
(s + a)2 + b2
10a
10b
re
at
cos(bt + )u(t)
n!
sn+1
0.5rej
0.5rej
+
s + a jb s + a + jb
10c
As + B
s2 + 2as + c
r=
=
b=
A2 c+B 2 2ABa
,
ca
2
AaB
arctan Aca2 ,
c a2
10d
eat
b=
B Aa
A cos bt +
sin bt u(t)
b
c a2
As + B
s2 + 2as + c
6.2
133
Operation
f (t)
F ()
Addition
f1 (t) + f2 (t)
F1 (s) + F2 (s)
Scalar multiplication
kf (t)
kF (s)
Scaling
f (at), a 0
1 s
F
a
a
Time shift
f (t t0 )u(t t0 )
F (s)est0 , t0 0
Frequency shift
f (t)es0 t
F (s s0 )
Time convolution
f1 (t) f2 (t)
F1 (s)F2 (s)
Frequency convolution
f1 (t)f2 (t)
1
F1 (s) F2 (s)
2
Time dierentiation
df
dt
sF (s) f (0 )
d2 f
dt2
s2 F (s) sf (0 ) f(0 )
d3 f
dt3
s3 F (s) s2 f (0 )
sf(0 ) f(0 )
Time integration
f ( )d
1
F (s)
s
f ( )d
1
1
F (s) +
s
s
Frequency dierentiation
tf (t)
dF (s)
ds
f (t)dt
Frequency integration
f (t)
t
Initial value
f (0+ )
Final value
f ()
F (z)dz
s
lim sF (s)