Chapter 1

Types of signal processing

Analog signal processing
Relies on analog circuit elements such
as capacitors, resistors, inductors,
transistors, and diodes

Digital signal processing

Relies on three digital computer
elements; adders, multipliers, and
memory

Relies on numerical computation for

operation and real time performance is
not guaranteed

Classification of signals
Continuous-time (CT) and
discrete-time (DT) signals

Even and odd signals

Periodic signals and nonperiodic

(aperiodic) signals

Energy signals and power signals

 Continuous-time (CT) and discrete-time (DT) signals
ECG and EEG are examples of continuous-time signals

Sample the values of a continuous-time signal at a standard time increment to

obtain a discrete-time signal

Ts is the sampling period

n is an integer

Sample a continuous time signal x(t) at time t=nT s yields

Such a sequence is known as a time series

Even and odd signals

Even signals Odd signals

A CT signal is an even signal if A signal is odd if

x(-t)=x(t) for all t x(-t) = -x(t) for all t

Even signals are symmetric about the Odd signals are antisymmetric about
vertical axis or time origin the time origin

Even-odd decomposition

x(t)=xe(t)+xo(t) eqn (1)

Given that xe(t) is even and xo(t) is odd

 xe(-t) = xe(t)
xo(-t) =- xo(t)

x(-t)= xe(-t) + xo(-t)

x(-t)= xe(t) - xo(t) eqn(2)
(1)-(2) (1)+(2)
x(t)-x(-t)= xo(t)+xo(t) x(t)-x(-t)= xe(t)+xe(t)
1 1
xo(t)= [ x ( t )−x (−t )] xe(t)= [ x ( t ) + x (−t ) ]
2 2
Sketching odd and even components of signal
Given the signal, sketch the odd and
even component

1
Even component: xe(t)= [ x ( t ) + x (−t ) ]
2

1
Odd component: xo(t)= [ x ( t )−x (−t )]
2

1
1. Sketch x (t)
2

1
2. Sketch x (−t)
2

3. Sketch even signals

1
Even component: xe(t)= [ x ( t ) + x (−t ) ]
2

4. Sketch odd signal

1
Odd component: xo(t)= [ x ( t )−x (−t )]
2

To verify if correct, sum of odd and

even should be x(t) given in qns
 Periodic and Nonperiodic Signals
Periodic Nonperiodic
A signal that repeats itself after a Waveform occurs at irregular intervals
standard time increment like a pulse

Periodic signal
x(t) is a continuous function of time that satisfies the condition
x(t)=x(t+T)
where T is a positive constant
Smallest value of T is the fundamental period

Fundamental frequency
1
f = in hertz (Hz)
T

Angular frequency
2π
ω=2 πf =
T
X[n] is a discrete function of time that satisfies the condition
X[n]=x[n+N]
where N is a positive integer
Smallest value of N is the fundamental period

Fundamental angular frequency of x[n]

2π
Ω=
N

Energy signals and power signals

Continuous Discrete

Energy Energy
∞ ∞

E= ∫ x (t)dt
2
E= ∑ x2 [n]
−∞ n=−∞

Time averaged power Time averaged power

t2 N −1
1 1
P = ∑ x [n ]
2
P = ∫ x (t) dt
2
T t1 N n=0
Energy signal has zero averaged power
Aperiodic
 Power signal has infinite energy
Periodic
Chapter 2
Operations performed on
Amplitude scaling
Continuous signal y(t) is a result of amplitude scaling of x(t), y(t) = xc(t)
Discrete signal, y[n]=cx[n]

Addition
Continuous signal y(t) is a result of addition, y(t) = x1(t)+x2(t)
Discrete signals, y[n]= x1[n]+x2[n]

Multiplication
Continuous signal y(t) obtained from multiplication, y(t)=x1(t)x2(t)
Discrete signal, y[n]=x1[n]x2[n]

Differentiation
d
y(t)= x(t)
dt

Integration
t
y ( t ) =∫ x (τ)
−∞

Time scaling
The signal y(t) obtained by scaling the independent variable time by a factor of a is
defined by
y(t)=x(at)

Compressed
If a>1, y(t) will be compressed

Expanded in time
If 0<a<1, y(t), is an expanded version of x(t)

For discrete,
 y[n]=x[kn], k > 0

If k>1, y[n] will be compressed

If 0<k<1, y[n] will be expanded

If K>1, some values of discrete-time signal are lost

Time shifting
y(t)=x(t-to) where to is the time shift

Shifting x(t) towards right

If to>0, y(t), is obtained by shifting x(t) towards the right, relative to the time axis

Shifting x(t) towards left

If to<0,y(t), is obtained by shifting x(t) towards the left, relative to time axis

Reflection

Signal y(t) represents a reflected version of X(t) about t=0

Even signal property x(-t)=x(t), hence reflected version is the same

Odd signal property x(-t)=-x(t), hence reflected version is negative of original
Similar observations for discrete signals

Precedence rule
1. Perform time shifting first
2. Followed by time scaling

Correct order

Wrong order

Unit step function

Denoted by u(t-c)

Unit ramp function

Denoted by r(t)=tu(t)
 For discrete signals

Impulse function

The impulse δ(t) = 0 except at origin

Applies to discrete signal as well

Properties of impulse function

Product of δ(t) and some signal x(t)
 Sifting property
∞ ∞

∫ x ( t ) δ ( t ) dt=x(0) ∫ δ ( t ) dt=x ( 0 ) .1=x (0)

−∞ −∞

Exponential Signals
X(t)=Beat
B and a are real parameters
B is amplitude of exponential signal measured at t=0

If a<0, exponential decays

If a>0, exponential grows

Continuous time sinusoidal signal

x(t)=Asin(ωt+ Ø)
A is amplitude
Ω is frequency in radians per second
Ø is phase angle in radians

Discrete-time sinusoidal signal

X[n]=Acos(Ωn+ Ø)

If signal is periodic, x[n]=x[n+N]

Substituting
X[n]=Acos(Ωn+ ΩN+ Ø)

Signal will only be periodic if its frequency is a rational number

For satisfying the conditions of discrete-time periodic signal, it is required that

2 πm
ΩN=2πm radians or Ω= radians/cycle where m and N are integers
N

Chapter 3
System properties
 Memory vs memoryless
 Invertibility
 Causality
 Stability
 Time-invariance
 Linearity

Memory vs memoryless
 Memory system has an output that depends on past or future values of
input signal
 Memoryless system only depends on present value of input signal

Examples
Involves time shifted variables,
hence memory system
Only depend on present value
Hence memoryless system

Invertibility
 If input can be obtained by reversing the output, system is invertible

Example
This is not invertible as the square
root of the output gives us ±input
instead of input

Causality
 System is causal if present value of output signal depends on present or
past values of input signal
 Non-causal system depend on one or more future values of input signal
 Causality is required for a system to be capable of operation in real time

Example
Casual because output only
depends on present and past
values only
Noncasual as x[n+1] is a future
value

Stability
1. For a system to be stable, output should be bounded when given a
bounded input at every instant of time

Example
 Input is bounded from -6 to 6 which
cause output to be bounded to e-6t to
e6t
Stable
Step function u(t) bounds input at 1 for
when t>0, but output will continuously
increase as time increase
Not stable

Time invariant
1. System is time invariant if a time delay in the input causes output to have
the same time delay

Example

Linearity
Additivity
1. X(t)=x1(t)+x2(t) has to produce the corresponding output y(t)=y1(t)+y2(t)
Scalability
2. Input x(t) is scaled by a constant factor will cause output to be scaled by the
same factor
Example
 Let x(t) = ax(t)

Let x(t) = (x1(t)+x2(t))

Let x(t) = ax(t)

Let x(t)=(x1(t)+x2(t))

Chapter 4
Convolution operation for discrete function

How to perform this operation?

1. Find all values of x[n] and h[n] (if not given, draw graph)
2. Draw table and fill in x[n] on top and h[n] at the side, underline value at n=0
3. Fill in values in the table by multiplying
4. Sum the diagonal and plot the value as the output
5. Tic tac toe goes brr brr
 Convolution operation for continuous signals
1. Choose 1 signal to perform reversal and time shift to obtain h(t- τ ) signal
2. Slowly slide h(t- τ ) to overlap the other signal by setting a boundary to the
value of t
3. Integrate using limits of the overlapping part and sub in the values of each
signal equation
Upper boundary
y (t)= ∫ x ( t ) h (t−τ ) d τ
Lower boundary
If graph is flat, If graph is ramp with If graph is ramp with
Eqn = 1 positive gradient negative gradient
Eqn = τ Eqn = t−τ

4. Repeat till all cases are done

Chapter 5
Eigenfunction of LTI system
If output of LTI system is a scaled version of input, the input function is called
eigenfunction and the scaling factor is called the eigenvalue

For continuous-time eigenfunction of LTI system

For discrete-time eigenfunction of LTI system

 Fourier series representation
Synthesis equation
∞
x ( t )=∑ x [k ]e jkωt
−∞
Analysis equation
t
1
2

x [ k ] = ∫ x (t)e− jkwt dt
T t 1

Calculate FS coefficient by inspection

1. Find ω from equation
2. Sub in ω into synthesis equation
3. Use Euler formula to expand sin or cos terms

4. Compare with the exponential term in synthesis equation to get x[k]

Direct calculations of FS coefficient

2π
1. Find ω using ω= ,can get T from equation
T
2. Sub in ω and x(t) into analysis equation
3. Solve the integral
4. Sub in values of K(odd or even) to find the values of x[k]
Fourier series representation
2π
1. Find ω using ω= ,can get T from equation/graph
T
2. Apply analysis equation, if x(t) has multiple values, split integration into 2
and change the limits
3. Solve integration
4. Break imaginary part into trigo form and sub in odd and even K values to
find all possible x[k] values

Combining the 2 concepts

 If the complex exponential term in Fourier series is the eigenfunction of an LTI
system, the output must be a linear combination of complex exponentials

Chapter 6
Fourier transform
FT Inverse FT
∞ ∞
1
X ( jw ) =∫ x (t )e ∫
− jwt jwt
dt X ( t )= x ( jw ) e dw
−∞ 2 π −∞

Chapter 7
Digital Signal Processing (DSP) system

Analog-to-digital conversion process

Nyquist-Shannon Sampling Theorem

F S – sampling frequency
F N – maximum frequency of signal
Nyquist frequency (folding frequency) Nyquist rate (Sampling frequency)
F F S=2 F N
FN= S
2

Sampling at this frequency causes No aliasing effect

aliasing effect
Aliasing effect

Aliasing effect refers to when two different signals become indistinguishable when
sampled

Discrete Fourier transform

N −1
X ( k )= ∑ x( nT ) e− jkΩnT
n=0
2π
Using Ω= (fundamental frequency)
NT
N −1 n
− jk 2 π
X ( k )= ∑ x( nT ) e N

n=0

 Write down x(T) values, should be given in question

 Sub in k = 0 to (N-1)
 n value is the same as k value
Amplitude spectrum Phase diagram
| X (K )|=√ (ℜ) +(ℑ)
2 2
ℑ
φ ( k )=tan−1 ( )
ℜ
Use fundamental frequency to find Ω , 2
Ω , and 3Ω Plot φ ( k ) vs Ω

Plot a line graph of | X ( K )| vs Ω

Inverse Discrete Fourier Transform

N −1
1
x ( nT )= ∑ X (k ) e
jk ΩnT
N K=0
2π
Using Ω= (fundamental frequency)
NT
 N −1 n
1 jk 2 π
x ( nT )=
N
∑ X (k ) e N

K=0

 Sub in n=0 to (N-1)

 K value = n value
 Should get back the same value as from DFT

Chapter 8
Fast Fourier Transform
DFT formula
N −1 n
− jk 2 π
X ( k )= ∑ x n e N

n=0
2π
−j
To simplify, use W =e N
N
N −1
X ( k )= ∑ x n W N
Kn

n=0
Using unit circle
Example of W 46

1. Split unit circle into parts, number of parts is the subscript of W

2. Starting on the positive X-axis, rotate between the points in clockwise
direction, the number of rotations is the superscript of W
3. If number of parts is odd, use calculator to find
4. Or just use calculator to find and don’t care this up to you
FFT method
1. Spilt X(n) until groups of 2 are obtained
2. Apply formula to each group
3. Sub in values of K into each group
4. Add respective values obtained from different groups with same K values
together to get Xo(k)

General formula to apply to each group

For level 1
K
X 0 ( k )=X 10 ( k ) +W N x11 (k )
 For level 2 (depends on how many splits)
K K
X 10 ( k ) =X 20 ( k ) +W N x 21 (k ) X 11 ( k )= X 22 ( k )+W N x 23( k )
2 2

Chapter 9
Z-transform formula
∞
X ( z )= ∑ x (n) z−n
n=−∞
Example: finite sequence

∞ 0
X ( z )= ∑ x ( n ) z = ∑ x ( n ) z −n
−n

n=−∞ n=−6
−(−5) −(−3) −(−5 ) −(−3) −(−1)
X ( z )=0+ z +3 z +5 z +3z +z +0
5 4 3 2 1
X ( z )=z +3 z +5 z +3 z + z

Example infinite sequence

∞ ∞
X ( z )= ∑ x ( n ) z =¿ ∑ (1) z −n ¿
−n

n=−∞ n=0
−1 −2 −3
X ( z )=1+ z + z + z +… - (1)

Apply geometric progression

Multiply by observable pattern, this case is z−1
−1 −1 −2 −3 −4
z X ( z ) =z + z + z + z +… - (2)

Region of convergence
Criteria that makes X(z) go to 0

|z−1|<1
Using (1) – (2)
X ( z )−z−1 X ( z )=1
X ( z ) ( 1−z )=1
−1

1
X ( z )=
( 1−z −1 )
z
X ( z )=
( z−1 )
 Properties of z-transform
Linearity Delay or shift
Can add sequence of x(n) together then x ( n)→ X ( z )
Z-transform or vice versa x ( n−m ) → Z−m X (z)
Differentiation Convolution
x ( n ) → X (z ) ∞

dX ( z) y ( n )= ∑ h ( k ) x (n−k )
nx ( n ) →−Z k=−∞
dz Y(z) = H(z)X(z)

Inverse z-transform
−1
x ( n )=Z [ X ( z)]

Example: partial fraction expansion method with linear poles

z−1
X ( z )=
1−0.25 z−1−0.375 z−2
z
X ( z )= 2
z −0.25 z−0.375
z
X ( z )=
(z−0.75)( z +0.5)

Poles are z = 0.75, z = -0.5

C1 z c z
X ( z )= + 2
z−0.75 z +0.5

( z− pk )
c k= X ( z )
z z= pk
|
c 1=
( z
(z−0.75)(z +0.5) z )
( z−0.75 )
|
z=0.75
=0.8

c 2= ( (z−0.75)(z
z
+0.5) ) ( z−0.75 )
z |z=−0.5
=−0.8

0.8 z 0.8 z
X ( z )= −
z−0.75 z+ 0.5

Refer to z-transform table and look or similar form of X(z) to get X(n)
 X ( n ) =0.8 [ ( 0.75 ) −(−0.5 ) ] u[n]
n n

Example: partial fraction expansion with poles that have power

2
z
X ( z )=
(z−0.5)¿ ¿
Cz D1 Z D2 Z
X ( z )= + +
(z−0.5) (z−1) ¿ ¿

c z=¿

power −1
1 d
Di= ¿
( power of root−i) d z power −i
1 d 2−1
D 1= ¿¿
( 2−1 ) ! dz 2−1
d
D 1= ¿
dz
D 1=¿
D1=−2
2−2
1 d
D 2= ¿¿
( 2−2 ) ! d z 2−2
D2=¿ ¿
D2=2

2 2 Z 2Z
X ( z )= − +
(z−0.5) ( z−1 ) ¿ ¿
Refer to z-transform table

x ( n )=[ 2 ( 0.5n ) −2 ( 1n ) +2 n ( 1 n ) ] u [ n]
Chapter 10
Finite impulse response (FIR) filter Infinite impulse response (IIR) filter
N −1 N
H ( z )= ∑ h(k )z ∑ bk z−k
−k

K =0 k =0
H ( z )= M
(1+ ∑ a k z−k )
k=1

Filter design steps

1. Specification of filter requirements
2. Calculation of suitable filter coefficient
3. Implementation of filter in software and/or hardware

Chapter 11
1. Specify the type of filter
Given in question
2. Obtain impulse response
Given in question
3. Select window
Write window type given in question
3.3 transition width
∆ f= , where ∆ f = (round up value to make it odd number)
N Sampling freq
Find N
Write down W(n) given in question
4. Calculate coefficient
Transition width
f c = passband edge frequency +
2

Normalise
fc
f c' =
sampling frequency

Sub in values into impulse response formula

'
ω=2 π f c

h ( n ) =h D ( n ) +w (n)

Make this table