Signals, Spectra, and Digital Signal Processing

1
Introduction

Module 001 Introduction

At the end of this module you are expected to:

1. Recognize the physical quantities that are referred to as signals and the
importance of signal processing.
2. Classify the characteristics of signals in both continuous time and
discrete-time form
3. Explain the basic block diagram of digital signal processing

Signals are all around us and they are sent and perceived in different ways. When we wave
to a person, we are basically sending a signal. When a person talks and we hear the
person’s voice, we actually hear an audio signal. It doesn’t end there; our brain which is the
processor of the things we receive will analyze and interpret these signals.

Digital Signal Processing

Digital Signal Processing or DSP is a discipline that spans electrical engineering,
computing, mathematics and the physical sciences. It is distinguished from other
areas in computer science by the unique type of data it uses: signals.
The study of DSP deals with the mathematics, algorithms and techniques used to
manipulate these signals after they have been converted into a digital form to acquire
the necessary information from the original signal.
Among the many applications of DSP are the enhancement of visual images,
recognition and generation of speech, compression of data for storage and
transmission. The analog signal (also known as the continuous time signal) which
comes in various forms such as smoke, dust particles, water level, humidity, pressure
etc. is converted into a computer-readable format or a digital signal (thru digitizing
not digitalizing) and processed according to the desired output thru the use of DSP
system. Digital signals are those for which both time and amplitude are discrete.

DSP in the past

Imagine how people live before the emergence of DSP, in communication, mails
being sent through “snail mails” before the electronic mails, storage of data using
paper trails before the floppy disks; switchboard operator “manually” connects long
distance calls before mobile phones are accessible by the public and even using the
internet by the help of data compression.
In military, remote sensing was created to detect intruders; image processing which
creates a clearer image of pictures captured in the outer space and many other
technologies which are developed even more in these days.

Course Module
 The complexity, speed, and capability of DSP chips have grown exponentially since
the early 1980s and show no sign of slowing down.

21st century DSP

There are many applications of DSP that are continuously being studied. Listed
below are some of them:
• Image processing
• Neural networks
• Pattern recognition
• Digital communications
• Biomedical applications
• Speech processing
• Underwater acoustics
• Radar signal processing
• Astronomy
• Geophysical data analysis

There are things unimaginable before that are now happening because of DSP. Non-
conventional methods of processing signals are reconstructing signals are emerging.
The discoveries in using distinct vibrations from a bag of chips or a plant captured
from a video can be used to reconstruct the sound produced when these things were
hit by the sound waves.
DSP technology can be beneficial wherever programmable and power efficient
computing is needed. In the IoT, DSP will be a key technology for Low Data Rate LTE
and LPWAN standards to ensure low power and flexibility to support multiple
standards in a single device.
Most systems used now are in trend of embedding DSP technology in SoC (system
on chip) and they have wide arrays of applications. While there are signals that need
to be analyzed and extract information from, digital signal processing can be a great
tool.

Classifications of Signals
Signals are primarily classified into two: continuous time signal and discrete time.
Continuous time signals are often referred to as analog signals and they take on
values in continuous interval (a, b), where a can be -∞ and b, ∞. Unprocessed physical
quantities such as the audio signal that we hear are in the form of continuous time. Figure 1
shows an example of a waveform of an analog signal generated using MATLAB. The
continuous time domain representation of a female saying,
“The quick brown fox jumps over the lazy dog.”
 Signals, Spectra, and Digital Signal Processing
3
Introduction

Figure 1. Waveform of an analog audio signal represented in time domain generated in MATLAB.

Discrete time-signals on the other hand, are signals which are the preprocessed
signals which are to be used in digital signal processing. The process involve in converting
a continuous time to discrete time signal is referred to as sampling. It is represented
mathematically by a sequence of numbers x, in which the nth number in the sequence is
denoted x[n] and is formally written as:
x = x[n], -∞ < n <∞
x[n] = xa nT, -∞ < n <∞

Figure 2. Waveform of continuous time signal , y = sin ωt generated in MATLAB

Figure 3. Waveform of discrete time signal , y = sin ωn generated in MATLAB

Notice the difference in the notation of a continuous time signal, to discrete time
signal, while continuous time signal uses t, discrete time uses n, which represent the
number of sample which will be furthered, discussed in the next module.
Both continuous time and discrete-time can be furthered classified with the following:
• Deterministic signal
Course Module
 Signal which can be expressed in mathematical form example is y = A sin ωt
where it can be described as a sinusoidal signal with amplitude, A and is a
function of time, t.

• Non-deterministic signal
Signal which cannot be expressed in simple mathematical form example is
random noise. Random signal are expressed using probability.

Figure 4. (top) Waveform of deterministic signal in the mathematical form y = sin ωt;
(bottom) Waveform of a random noise generated in MATLAB

• Periodic signal
Signal which exhibits periodicity or can complete a certain pattern in one
cycle. Using the same signal in Figure 4. the deterministic signal is also
periodic since it has a complete pattern in cycle. Mathematically, it can be
determined by the ω in y = A sin ωt where ω is the angular frequency in
radians per second . The angular frequency is equal to the frequency
multiplied by a factor of 2π. Periodicity is an important characteristic of
signal used in spectral analysis.
• Aperiodic signal
Signal which cannot complete a certain pattern in one cycle. Also, using the
same example in Figure 4, the random signal didn’t exhibit periodicity.

• Even signal
Signal which exhibits symmetry in the vertical axis. An even signal may be
expressed in continuous time as x (t) = x(-t) or x(n) = x(-n) for discrete time
form.

• Odd signal
Signal which exhibits rotational symmetry with respect to the origin. An odd
signal may be expressed in continuous time as x (t) = -x(-t) or x(n) = -x(-n)
for discrete time form.
 Signals, Spectra, and Digital Signal Processing
5
Introduction

Figure 5. (top) Waveform of a discrete-time even signal

(bottom) Waveform of a discrete-time odd signal generated in MATLAB
DSP System
A basic block diagram of DSP system is represented by an analog-to-digital
converter, a digital signal processor and an analog-to-digital converter is shown in Figure 6.

Analog-to -Digital Digital Signal Analog-to-Digital

Converter Processor Converter
Signal in Signal out
Figure 6. Basic block of a DSP system

Signal in or the analog input

All signal processing begins with an input transducer. The transducer converts an analog
signal to typically an electrical signal. In signal-processing applications, the transducer can
take many forms. In audio signal processing, a microphone acts as the transducer in the
system. In communications, an antenna converts electromagnetic waves also to an
electrical energy.

Analog-to-digital converter (ADC)

There are a lot going on inside an analog to digital converter, an in-depth approach will be
discussed in the next module but the main concept behind the ADC is that from the
electrical signal coming from the transducer, it is converted into a stream of 0s and 1s
which can be read by the digital signal processor.

Digital Signal Processor

Oftentimes, the digital signal processor is represented by computers. It is where the
analysis and decision takes place. Using computer algorithms, the signals are processed
through the instructions to achieve the desired output.

Course Module
 Digital- to analog converter (DAC)
A DAC is used to convert digital signals consisting of 0s and 1s to varying analog signals
(such as a voltage signal).
An audio signal for example may come in and come out in the same continuous time form-
a processed version designed according to the desired output. It may be free from noise,
amplified, with effects such as reverberation. There are some signals however, according to
the desired output may come in the form of heat and may come out as a trigger to an
automation system. Such applications are widely seen nowadays in field programmable
gate arrays (FPGA) and microcontrollers for specific purpose applications which replace
the size and latency of some computers.

Glossary:

Algorithm: is a step by step procedure that must be followed to execute a certain task. In
DSP, this involves signal acquisition and creating codes that would make decision on
what to do to the signal.
Analog signal: a signal which is perceived by our senses such as sound which can be
heard, heat that can be felt, light which can be seen and the like.
Audio signal: other terminology for sound signal
Compression: is the transformation of a collection of data typically into a smaller file
size.
Digital signal: a signal which is converted in the form understandable by computers; that
is 1’s and 0’s
FPGA: stands for field programmable gate array. It is an integrated chip which is
programmed according to function and can execute commands in high speed.
Frequency: cycles per second (Hertz) another term used to describe the periodicity of a
signal.
IoT: Internet-of-things. Technology whose feature is accessibility of information and
control thru clouds.
Low Data Rate LTE: LTE being ‘Long-term-evolution’, a technology which is a standard
for high-speed wireless communication for mobile phones and data terminals having low
data rate means an improved rate of which data is being transmitted and received while
considering the complexity.
LPWAN: Low-Power Wide-Area Network (LPWAN) or Low-Power Network (LPN) is a
type of wireless telecommunication wide area network designed to allow long range
communications at a low bit rate among things (connected objects), such as sensors
operated on a battery.
Radar: acronym for Radio Detection and Ranging is a system used to detect location,
direction, distance and speed of aircrafts and other objects by firing pulses that are
reflected off the object back to the source. The signal detected is noisy and needs DSP to
get accurate results.
 Signals, Spectra, and Digital Signal Processing
7
Introduction

Reverberation: similar to echo, it occurs when a sound continuously reflect on surfaces

causing it to be heard even the source already ceased
SoC: acronym for system on chip. It integrates all components of a computer or other
electronic systems to process analog signal, digital signal on a single IC or chip.
Spectral analysis: a procedure done to the signal, typically to observe its frequency
components. Random noise appears as high frequency signal when analyzed.
Sonar: acronym for Sound Navigation and Ranging. The function is practically the same
with radar, but is best used for underwater.
Transducer: a device that converts one form of energy into another

References and Supplementary Materials

Books and Journals
1. Proakis, J. G., & Manolakis, D. G. (2007). Digital signal processing. Pearson Prentice
Hall
2. Oppenheim, A.V., & Schafer, R.W. (2014)Discrete-Time Signal Processing. Pearson
Education Limited
Online Supplementary Reading Materials
1. 30 years of dsp: From a point solution for powering a toy to a pervasive technology.
http://www.newelectronics.co.uk/electronics-technology/30-years-of-dsp-from-a-
point-solution-for-powering-a-toy-to-a-pervasive-technology/45422/; Aug 14, 2017
2. There's still growing demand for DSP, say experts;
http://www.newelectronics.co.uk/electronics-technology/theres-still-growing-
demand-for-dsp-say-experts/117710; Aug 14, 2017
3. Digital Electronics Basics - Chapter 6: Analog to Digital & Digital to Analog Conversion.
http://www.ni.com/example/14498/en/. November 9, 2017

Online Instructional Videos

1. Good Vibrations: A Bag of Potato Chips and the Science of Sound.
http://www.wnyc.org/story/eavesdropping-using-vibrations-bag-potato-chips/;
November 9, 2017

Note:
• Save each file using the format: Week001-Module
• There should only be one file for module per week.
• Avoid plagiarism by paraphrasing information and texts coming from the Internet and
other sources.

Course Module
 Signals, Spectra, and Digital Signal Processing
1
Theory of Discrete Time Signals and Systems

Module 002 Theory of Discrete Time Signals

and Systems

At the end of this module you are expected to:

1. Classify discrete time signals and systems.
2. Convert continuous time signal to discrete time signal mathematically.
3. Apply the Nyquist Theorem to properly sample and reconstruct a signal.
4. Calculate the sampling resolution of a sampled signal.

In the basic block diagram that an analog or continuous time signal enters the analog-to-
digital converter before it could be processed. Why is conversion necessary? Can’t the
analog signal be processed directly to the digital processor? The digital signal processor
needs a digital signal, because it could only understand it in this form. Processing it using
algorithms to compute the necessary decision is based on the fact that the signal is
represented by values understandable by the computer.

Discrete time Signals

Recall the basic block diagram of DSP

Analog-to -Digital Digital Signal Digital - to -

Converter Processor Analog Converter
Signal in Signal out
Figure 1. Basic block of a DSP system
Signal in is a real world signal which is in continuous time form. It will be converted in to
digital form using ADC and the output of the ADC is of course, a digital signal. So what
happened inside the ADC?
There are specifications that ADC requires, for it to produce the necessary form
understandable by the digital signal processor.
1. Sampling frequency, fs
2. Sampling resolution, 𝜏

Discrete Time Signals

Below is a graph of a continuous time signal. In order to analyse a signal in certain
points for example, we take the values of the amplitude at those point, at the y axis with
respect to time which is the x axis as shown in Figure 2. The values of y at the time x is
equal to 0.2, 0.4, 0.8 is defined.

Course Module
 Figure 2. Waveform of continuous time signal , y = sin ωt generated in MATLAB

But what if we have a lot of points we want to analyse? It is not practical to do it at unequal
intervals. The best option is to divide it into several points. In this way, we can analyse both
the discrete value at that point.

Figure 3. Waveform of discrete time signal , y = sin ωn generated in MATLAB

Notice that the time of 1 second is divided into definite interval making it easier to locate
the value at that point, each discrete value is referred to as a sample. The process of
converting continuous time signal into discrete time signal is referred to as sampling. The
distance or the time difference between two consecutive samples is referred to as the
sampling interval, measured in seconds. The domain is still in time domain however one
second is uniform intervals. This domain is referred to as the n domain where n is the
number or the sequence of the sample.
A discrete time signal may also be represented by sequence of numbers. But the upper and
lower limit must be defined first. Using the lower limit to 0 and the upper limit as positive
infinity is referred to as causality. A causal signal will neglect the values at the negative
side. If both sides are considered, then the signal is called anti-causal.
Causality is also used to describe a system more of this in topics about Transform.
 Signals, Spectra, and Digital Signal Processing
3
Theory of Discrete Time Signals and Systems

Figure 4. A unit sample signal generated in MATLAB

In sequence form, it is expressed as x (n) = [ 1 0 0 0 …]. It is called a unit sample signal since
the signal has maximum amplitude of 1. The indication where the zero will start is an
arrow or an underline. Those values which are in placed before zero are considered
advanced while those which are after zero is considered delayed.

Example:
Represent the following signals graphically for
x1(n) = δ(n) + 0.5δ(n-1)

Solution:

Figure 5. A sequence of impulses generated in MATLAB

In sequence form, it is expressed as x (n) = [ 1 0.5 0 0 …]. Notice the underline in the
sequence. It indicates n = 0 and the value of the element at the position is 1 which is the
magnitude of the impulse. The same applies with the delayed impulse at n = 1, the
magnitude of the impulse is equal to 0.5.

Another commonly used type of discrete time signal is the unit step signal. Notice it is a
train of unit impulses which is theoretically infinite.
Course Module
 Figure 5. A train of impulses called unit step generated in MATLAB

Exercise:
Graphically represent the discrete time signal given by x2(n) = u(n) - δ(n)

Solution:

Figure 6. Difference of a unit step u(n) and a unit sample with no delay δ(n)
is equal to a unit step with one delay u(n-1) generated using MATLAB

The graph from the top represents a unit step signal while the second represent the unit
`sample. The third represents the difference between the unit step signal and a unit sample
with no delay results to a unit step with one delay.
Among other discrete time signals are the unit ramp signals, real valued exponential,
complex-valued exponential and sinusoidal signal.

Sampling frequency
The next question is: how many samples should be used per second? First let’s try to look
at the basic block diagram again. The continuous time signal will come in and be converted
to discrete time through ADC. Assuming that the signal can be understood by the digital
processor, then the decision will occur at the DSP as to what should be done to the signal
whether it’s noise removal or signal enhancement, and then converted back to a continuous
time signal by the use of the DAC. The process of retrieving the continuous form of the
 Signals, Spectra, and Digital Signal Processing
5
Theory of Discrete Time Signals and Systems

signal which could be free from noise or an enhanced version of the original signal is
referred to as reconstruction which happens in the DAC. Going back to the question, how
many samples should be used per second?
According to the Sampling theorem, also known as the Nyquist theorem, named after Harry
Nyquist, it states that, for a signal to be properly reconstructed, a signal must be sampled
twice the maximum frequency component of the signal. Trying to sample a signal under the
requirement of the sampling theorem will cause distortion called aliasing. Aliasing is a
phenomenon which occurs when the sampling frequency is below twice the maximum
frequency component of the signal. From the term itself, the signals appear in a different
form, the high frequency may appear as low, and the low may appear as high.
Trying to sample the signal equal to twice the maximum frequency component of the signal
will allow the signal to be properly reconstructed. Going extremely higher than twice will
also reconstruct the signal but is not practical.

Practical Application of Sampling Frequency

Audible signals are said to have a range of 20 Hz up to 2kHz. An audio signal coming from a
single may be composed of several frequencies within this range. But remember, it is the
highest frequency that will dictate how much sampling frequency is needed to avoid
aliasing. Notice that audio CD have sampling rate of 44.1kHz, that is slightly greater than
twice the maximum frequency component of the signal!
Exercise:
From the given analog signal:
xa (t) = 2 sin 3000πt + 5 cos 2000πt
a. Determine the frequency components of the signal.
b. Determine the minimum sampling frequency to avoid aliasing.
c. Suppose the signal is sampled at 6000 Hz, will there be aliasing? Show your solution.
d. Suppose the signal is sampled at 2000 Hz, will there be aliasing? Show your solution.

Solution:
a. To determine the frequency component of the signal. Remember, the angular frequency
is given as ω = 2πf.
So, xa (t) = 2 sin 3000πt + 5 cos 2000πt can be expressed as:

xa (t) = 2 sin 2π(1500)t + 5 cos 2π(1000) t

Therefore, the frequency components of the signal are:
f1 = 1500 Hz and f2= 1000 Hz

Course Module
b. To determine the fs or the Nyquist frequency, we must first determine the highest
frequency component of the signal. In this case, it is 1500 Hz. Remember, in order to
satisfy the Nyquist theorem, sampling frequency must be:
fs > 2 fmax
Therefore, the minimum sampling frequency would be 2fmax,
2 fmax = 2 (1500) = 3000 Hz or 3000 samples per second

c. The sampling frequency is above the minimum, so there will be no aliasing. Let’s try to
prove it mathematically.
(1500) (1000)
x (n) = 2 sin 2π 4000 n + 5 cos 2π 4000 n
The discrete time signal obtained is
(1500) (1000)
x (n) = 2 sin 2π 6000 n + 5 cos 2π 6000 n

3000 2000
x (n) = 2 sin π(6000)n + 5 cos π(6000)n or simply
𝜋 π
x (n) = 2 sin ( 2 )n + 5 cos ( 3 )n

If we try to reconstruct the signal, all we have to do is multiply the angular frequency
with the maximum n which is also 6000 because n will start from n = 1 up to 6000 in
this question, then replace it with t, the signal is the same as the original analog form,
without aliasing.
1 1
x (t) = 2 sin 2𝜋(4)(6000) + 5 cos 2𝜋(6)(6000)

x (t) = 2 sin 2𝜋(1500)t + 5 cos 2𝜋(1000)t

d. Here, the sampling frequency is 2000 Hz, below the minimum sampling frequency
required to avoid aliasing computed which is 3000 Hz. Let’s try to demonstrate aliasing.
The discrete time signal obtained is
(1500) (1000)
x (n) = 2 sin 2π 2000 n + 5 cos 2π 2000 n

(3000) (2000)
x (n) = 2 sin π 2000
n + 5 cos π 2000
n

Notice that in the latter form, we have an improper fraction; this has a significant role
later.
3
x (n) = 2 sin π( 2)n + 5 cos π n

Reconstructing it
1 1
x (n) = 2 sin 2𝜋(1 − 4)𝑛 + 5 cos 2𝜋(2)n

Since it is periodic at 2π, we can simplify the equation as

1 1
x (n) =- 2 sin 2𝜋(4)𝑛 + 5 cos 2𝜋(2)n

To complete the reconstruction, replacing the n with t and multiplying it with 2000
which is the sampling frequency would give us:
 Signals, Spectra, and Digital Signal Processing
7
Theory of Discrete Time Signals and Systems

1 1
x (t) =- 2 sin 2𝜋 (4) (2000)𝑡 + 5 cos 2𝜋(2)(2000)t
or

x (t) =- 2 sin 2𝜋(500)𝑡 + 5 cos 2𝜋(1000)t

Notice that since the sampling frequency is below twice the maximum frequency
component, there was aliasing. The first component was aliased while the second was
not, reason? The second frequency component was sampled twice of its frequency thus
it was able to still meet the sampling requirement while the first one didn’t. As
mentioned, aliasing will make a signal take another form. In this case, the high
frequency appeared as low. The 1500 Hz component which is the high frequency
component appeared as 500 Hz which is a low frequency.

Sampling Resolution
Sampling resolution corresponds to how many levels or gradations can be made to a
waveform.
Suppose, you have a signal that you want to represent in 1000 levels or gradations of
amplitude, how many bits can represent this? The answer is 10 bits. 210 will give us 1024
covering the required 1000 gradations. One sample can take 1024 levels.
To visualize sampling resolution, let’s have an example. Suppose a signal would be
represented using 2-bit ADC. There are 4 gradations to represent one sample.
The sampling resolution is computed as:
1
τmin = 2𝑛 −1 where

τmin = is the minimum sampling resolution

n = is the number of bits

1 1
τmin = or
22 −1 3

The four levels will be represented using the binary numbers which is now acceptable in
the digital signal processor. In this example, the levels are represented using normalized
value from 0 to 1 as the minimum and maximum levels, respectively. It has no units, only
the magnitude.

Binary Maximum
representation resolution

Course Module
 00 0

01 1
3

10 2
3

11 3
or 1
3

Figure 7. A quantized version of the discrete time signal generated using MATLAB
Practical Application of Sampling Resolution
In practice, audio signals are sampled at 8 bits, 16 bits and 24 bits.
Both the sampling frequency and the sampling resolution allow the signal to take the
waveform necessary for reconstruction. A good sampling frequency will allow the signal to
be properly reconstructed. The number of bits is dictated by the number of gradations of
amplitude to represent the levels of a signal. High resolution will allow the signal to take
the desired waveform, in other words, providing adequate sampling frequency and high
resolution will produce the best results.

Glossary:

Frequency: expressed in Hertz (Hz). Cycles per second

Gradation: are the levels usually expressed in uniform interval,
Period: is the inverse of frequency. Seconds per cycle
Waveform: the shape or form of the signal

References and Supplementary Materials

Books and Journals
 Signals, Spectra, and Digital Signal Processing
9
Theory of Discrete Time Signals and Systems

1. Proakis, J. G., & Manolakis, D. G. (2007). Digital signal processing. Pearson Prentice
Hall
2. Oppenheim, A.V., & Schafer, R.W. (2014)Discrete-Time Signal Processing. Pearson
Education Limited
Online Supplementary Reading Materials
1. Sample Rates . http://wiki.audacityteam.org/wiki/Sample_Rates. November 14, 2017
2. Digital Audio Basics: Sample Rate and Bit Depth.
https://www.presonus.com/learn/technical-articles/sample-rate-and-bit-depth.
November 14, 2017
3. Digital Electronics Basics - Chapter 6: Analog to Digital & Digital to Analog Conversion.
http://www.ni.com/example/14498/en/. November 9, 2017

Online Instructional Videos

1. Good Vibrations: A Bag of Potato Chips and the Science of Sound.
http://www.wnyc.org/story/eavesdropping-using-vibrations-bag-potato-chips/;
November 9, 2017

Note:
• Save each file using the format: Week002-Module
• There should only be one file for module per week.
• Avoid plagiarism by paraphrasing information and texts coming from the Internet and
other sources.

Course Module
 Signals, Spectra, and Digital Signal Processing
1
Introduction to Discrete Transforms`

Module 003 Introduction to Discrete

Transforms

At the end of this module you are expected to:

1. Interpret a signal’s composition by its frequency content.
2. Examine the effect of Fourier Series expansion of a rectangular pulse.
3. Determine the spectra of the signal by inspecting the characteristics of
the signal

Imagine yourself going in a different country, you want to buy something but unfortunately
you weren’t able to exchange the money with the currency used in that country. Of course,
you can still purchase whatever you need, however it is much more efficient if you have it
exchanged. In mathematics, scientific notation has been of great use when it comes to
expressing too large or too small values numerical values. These two examples given may
represent the analogy of transforms – although they may still possess the same nature
however they can be expressed into a more useful form.
Signals, as discussed in the previous modules may be represented in time-domain. For
example, an audio signal - may it be speech or music can be represented in the time domain
and be expressed in seconds. In other words; a sound may be described according to how
long it lasted, the amplitude or volume of the sound at a particular time.

Transformation to Frequency Domain

Another way of representing signal is by its frequency content. Going back to our example
which is an audio signal, from the lessons we had in Physics class, the sound we hear is a
combination of different frequencies – a high pitch sound is represented by high frequency
audio signal while a low frequency sound is represented by low frequency signal. This
representation of signal using its frequency content is referred to as frequency domain
representation.
A simple representation of a linear system will be like this: an input will pass through a
system that will result to an output.

INPUT OUTPUT
SYSTEM

Course Module
In time domain representation, the input which is typically denoted by x(t) will pass
through a system or the impulse response function h(t). The output y(t) is the result of
convoluting the input, x(t) and the impulse response, h(t). Mathematically expressed as:

y(t) = x(t) * h(t)

x(t) y(t)
h(t)

Convolution in the time domain is multiplication in the frequency domain. Mathematically,

it is expressed as:
Y(ω) = X(ω) H(𝜔)
where ω is a frequency domain. Another domain that would allow representation of time
domain signal to complex frequency domain is the z domain.
So in designing a system, it is more efficient to use transform since convolution in the time
domain is multiplication in the frequency domain and vice versa.

Fourier Analysis
When you strum the strings of a guitar to play a chord, you need to press down on the
strings with the tips of your fingers. You will likely hear a single sound collectively called as
a guitar chord. The guitar chord is the sum of the sound produced by the vibrations of each
strings. Each string has different characteristics such as tension, thickness, etc. therefore it
gives different pitch. Although we perceive a single sound, it is actually the sum of the
sounds produced by each strings with different pitch.
According to the French mathematician and physicist, Jean-Baptiste Joseph Fourier (1768-
1830), any continuous periodic signal is could be represented by sum of sinusoids.
Following our example, Fourier analysis would explain how a single guitar chord can be
represented by harmonic oscillations.
From the time-domain representation of a signal, it can be transformed or represented to
frequency domain by using Fourier Transform. The term Fourier transforms can be broken
into four categories:

1. Fourier Series for periodic, continuous-time signals

2. Fourier Series for aperiodic, continuous-time signals
3. Fourier Transform for periodic, discrete-time signals
4. Fourier Transform for aperiodic, discrete-time signals

Fourier Series: Periodic- Continuous time Signal

Sinusoids and complex exponentials play a major role in the analysis of signals and
systems.
A simple harmonic oscillation is mathematically described by the following continuous-
time sinusoidal signal:
 Signals, Spectra, and Digital Signal Processing
3
Introduction to Discrete Transforms`

xa(t) = A cos(Ωt + ϴ), -∞ <t <∞ Eq 1

where the x(t) with a subscript a denotes an analog signal, A is the amplitude of the
sinusoid. Ω is the angular frequency in radians per second (rad/s), and ϴ is the phase in
radians. Oftentimes, Ω is expanded in the form
Ω = 2 πF
where f is the frequency in cycles per second (or Hertz), hence an analog signal can also be
represented by
xa(t) = A cos(2 πF + ϴ), -∞ <t <∞ Eq 2

𝜋
Figure 1. Graph of xa(t) A cos Ωt where A = 1, f = 2 and ϕ = generated using MATLAB
6

Remember that for every fixed value of F, the signal is periodic. The signal will therefore
contain the fundamental period and harmonics
xa (Tp + t) = xa (t)
where Tp = 1/F is the fundamental period of the sinusoidal signal.

Remember using Euler’s identity in expressing sinusoids as a complex exponential function

is given as
e + jϕ = cos ϕ + j sinϕ
Note that a sinusoidal signal can be obtained by adding two equal-amplitude complex-
conjugate exponential signals, sometimes called phasors.
𝐴 𝐴
xa(t) = A cos(Ωt + ϴ) = 2 e j(Ωt + ϴ) + 2 e -j(Ωt + ϴ) Eq 3
Continuous time harmonically related exponentials are expressed as:

sk(t)= 𝑒 𝑗𝑘𝛺𝑜 𝑡 = 𝑒 𝑗𝑘2𝜋𝐹𝑜 𝑡 , k = 0, + 1, +2, … Eq 4

1
We note that for each value of k, sk(t) is periodic with fundamental period(𝑘 𝐹𝑜) = Tp/k or
fundamental frequency kFo. Since a signal that is periodic with period Tp/k is also periodic
Course Module
with period k (Tp/k) = Tp for any positive integer k, we see that all of the sk(t) have a
common period of Tp.
From the basic signals, we can construct a linear combination of harmonically related
complex exponentials of the form

𝑘= −∞ 𝑐𝑘 𝑠𝑘 (t) = ∑𝑘= −∞ 𝑐𝑘 𝑒
xa(t) = ∑∞ ∞ 𝑗𝑘𝛺𝑜 𝑡
𝑘 Eq 5

where ck = 0, + 1, +2, … are arbitrary complex constants.

1
The signal xa(t) is periodic with fundamental period T, = 𝐹𝑜, and its representation in terms
of Eq 4 is called the Fourier series expansion for xa(t).

The expression for the Fourier coefficients in terms of the given periodic signal is:
𝑐𝑘 = ∫𝑇 𝑥(𝑡)𝑒 −𝑗𝑘2𝜋𝐹𝑜𝑡 𝑑𝑡 Eq 6
𝑝

which can also be written as:

𝑡 + 𝑇𝑝
𝑐𝑘 = ∫𝑡 𝑜 𝑥(𝑡)𝑒 −𝑗𝑘2𝜋𝐹𝑜𝑡 𝑑𝑡 Eq 7
𝑜

where 𝑡𝑜 is an arbitrary but mathematically convenient starting value.

The so called Dirichlet conditions guarantee that the Fourier series expansion will
converge. The following conditions must be satisfied:
1. The signal has a finite number of discontinuities in any period.
2. The signal contains a finite number of maxima and minima during any period.
3. The signal is absolutely integrable in any period∫𝑇 | 𝑥(𝑡)|𝑑𝑡 < ∞
𝑝

Frequency Analysis of Continuous-Time Periodic Signals

In general, the Fourier coefficients 𝑐𝑘 are complex valued. If the periodic signal is real, ck
and c-k are complex conjugates. That is,

𝑐𝑘 = |𝑐𝑘 | e j ϴk
then
𝑐−𝑘 = |𝑐𝑘 | e -j ϴk

The analysis and synthesis equation for Fourier Series of continuous periodic signals are
hereby given as

Analysis Equation (direct transform)

∞
𝑋(𝐹) = ∫−∞ 𝑥(𝑡)𝑒 −𝑗2𝜋𝐹𝑡 𝑑𝑡 Eq 8
Synthesis Equation (inverse transform)
 Signals, Spectra, and Digital Signal Processing
5
Introduction to Discrete Transforms`

∞
𝑥(𝑡) = ∫−∞ 𝑋(𝐹)𝑒 𝑗2𝜋𝐹𝑡 𝑑𝐹 Eq 9

Fourier Series of a Rectangular Pulse Train

Notice that as the number of harmonics increases, the waveform of the rectangular pulse
becomes more defined but not perfectly rectangular. The presence of the ringing due to the
addition of high frequencies is referred to as the Gibbs’ phenomenon.

Figure 2. Graph of the Fourier series of a rectangular pulse train generated using MATLAB

Fourier Series for Discrete-Time Periodic

Suppose that we are given a periodic sequence x(n) with period N, that is x(n) = x(n +N)
for all n, The Fourier series representation for x(n) consists of N harmonically related
exponential functions
𝑒 𝑗2𝜋𝑘𝑛/𝑁 , , k = 0, + 1, +2, … N-1
and is expressed as

𝑁−1
x(n) = ∑𝑘=0 𝑐𝑘 𝑒 𝑗2𝜋𝑘𝑛/𝑁

where the {𝑐𝑘 } are the coefficients in the series representation.

To derive the expression for the Fourier coefficients, we use the following formula:
𝑁−1
𝑁 k = 0, ±N, ±2N
∑ 𝑐𝑘 𝑒 𝑗2𝜋𝑘𝑛/𝑁 = {
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
𝑘=0
Course Module
 The Fourier coefficients in terms of signal x(n) is expressed by the formula:
𝑁−1
1
𝑐𝑘 = ∑ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁
𝑁
𝑘=0

Sample problem:

Determine the spectra of the signals

1. x(n) = cos √2 πn
𝜋
2. x(n) = cos 3 𝑛
Solution

1. The fundamental frequency, fo is not a rational number which means that the signal is
aperiodic (for ωo = √2 π therefore fo = 1/ √2, it is not periodic.).

Fourier series expansion ( FSE) applies to periodic signals thus this signal cannot be
expanded using FSE. Nevertheless, the signal does possess a spectrum. Its spectral
content consists of the single frequency component at ω = ωo = √2 π

2. In this case, fo = 1/ 6 and hence x(n) is periodic with fundamental period N = 6.

Therefore, to determine the Fourier Series coefficients, we have

5
1
𝑐𝑘 = ∑ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/6
6
𝑘=0

However, x(n) can be expressed as

2𝜋𝑛 1 1
x(n) = cos = 2 𝑒 𝑗2𝜋𝑛/6 + = 2 𝑒 −𝑗2𝜋𝑛/6
6

The basic difference between the Fourier Series representation of continuous-time signal
and discrete-time periodic signals is that continuous-time periodic signal can consist of an
infinite number of frequency components, where the frequency spacing between two
successive harmonically related frequencies is 1 / Tp and where Tp is the fundamental
period whereas in discrete-time signal, of fundamental period N can consist of frequency
components separated by 2π/N radians or f = 1/N cycles.

References and Supplementary Materials

Books and Journals
1. Oppenheim, A.V., & Schafer, R.W. (2014). Discrete-Time Signal Processing. Pearson
Education Limited
 Signals, Spectra, and Digital Signal Processing
7
Introduction to Discrete Transforms`

2. Proakis, J. G., & Manolakis, D. G. (2007). Digital signal processing. Pearson Prentice
Hall
Online Supplementary Reading Materials
1. Fourier Analysis of a Rectangular Pulse.
https://www.mathworks.com/matlabcentral/fileexchange/25571-fourier-analysis-
of-a-rectangular-pulse November 24, 2017
2. Square Wave from Sine Waves.
https://www.mathworks.com/help/matlab/examples/square-wave-from-sine-
waves.html. November 24, 2017
3. Practical Introduction to Frequency-Domain Analysis.
https://www.mathworks.com/help/signal/examples/practical-introduction-to-
frequency-domain-analysis.html November 24, 2017

Online Instructional Videos

1. Fourier Series of a Square Wave. https://www.coursera.org/learn/audio-

engineering/lecture/xI3Ca/fourier-series-of-a-square-wave. November 24, 2017

Note:
• Save each file using the format: Week003-Module
• There should only be one file for module per week.
• Avoid plagiarism by paraphrasing information and texts coming from the Internet and
other sources.

Course Module
 Signals, Spectra, and Digital Signal Processing
1
Introduction to Discrete Transforms`

Module 004 Introduction to Discrete

Transforms

At the end of this module you are expected to:

1. Examine the relationship between periodic and discrete frequency
response and discrete-time periodic sequence
2. Illustrate discrete Fourier transform and discrete time Fourier transform
of signals
3. Examine the Fast Fourier Algorithm

Recall the analysis and synthesis equation for Fourier Series of continuous periodic signals
are hereby given as

Analysis Equation (direct transform)

∞
𝐸𝑞 1: 𝑋(𝜔) = ∫ 𝑥(𝑡)𝑒 −𝑗2𝜋𝐹𝑡 𝑑𝑡
−∞

Synthesis Equation (inverse transform)

∞
𝐸𝑞 2: 𝑥(𝑡) = ∫ 𝑋(𝜔)𝑒 𝑗2𝜋𝐹𝑡 𝑑𝜔
−∞

They could be also expressed as

∞
𝑋(𝜔) = ∫ 𝑥(𝑡)𝑒 −𝑗𝜔𝑡 𝑑𝑡
−∞
∞
𝑥(𝑡) = ∫ 𝑋(𝜔)𝑒 𝑗𝜔𝑡 𝑑𝜔
−∞

since 𝜔 = 2πF. Notice that the domain transformed from time to frequency domain.
This is for continuous-time signals. In order for us to convert a continuous-time signal to
discrete-time time, sampling is performed. It is simply expressing the period into equal
intervals. The value at each interval is called a sample. Therefore t will be replaced by n
where n represents the samples.
Graphically, we could represent the transform like this:

Course Module
 Figure 1. Illustration of the relationship of X(ω) to x(n). Image generated using MATLAB

The sampling of the continuous-time periodic signal to its discrete-time involves

approximations. Mathematically, we could approximate the integral of the continuous time
by getting the summation of its discrete equivalent. Hence we can now directly express the
transform the Discrete Fourier Transform from the above equation as:
𝐿−1
X(ω) = ∑𝑛=0 x(n)𝑒 −𝑗𝜔𝑛 0 ≤ω ≤ 2π
where L is the length of x(n).
Sampling X(ω) at equally spaced frequencies
2𝜋𝑘
ωk = , k = 0, 1, …, N-1
𝑁
𝑁−1 2𝜋𝑛𝑘
X(k) = X(ω) = X(ω) = ∑ x(n)𝑒 −𝑗 𝑁
𝑛=0

is the N point DFT and the inverse DFT is

𝑁−1
1 2𝜋𝑛𝑘
x(n) = ∑ x(n)𝑒 −𝑗 𝑁
𝑁
𝑛=0

The Discrete Time Fourier Transform (DTFT) of is just DFT with zero padding (which will
be illustrated later). It is the extended DFT from finite to infinite. Generally, the DTFT of the
sequence is described by:

𝑋̂(ω) = ∑ x(n)𝑒 −𝑗𝜔𝑛

−∞

We observe that there are dualities between the following analysis and synthesis
equations:
1. The analysis and synthesis equations of the continuous-time Fourier transform.
2. The analysis and synthesis equations of the discrete-time Fourier series.
 Signals, Spectra, and Digital Signal Processing
3
Introduction to Discrete Transforms`

3. The analysis equation of the continuous-time Fourier series and the synthesis equation
of the discrete-time Fourier transform.
4. The analysis equation of the discrete-time Fourier transform and the synthesis equation
of the continuous-time Fourier series.
Note that all dual relations differ only in the sign of the exponent of the corresponding
complex exponential. It is interesting to note that this change in sign can be thought of
either as a folding of the signal or a folding of the spectrum.
Example:
Determine the Discrete Fourier Transform of the sequence
1, 0≤𝑛 ≤𝐿−1
𝑥(𝑛) = {
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Solution:

In the past module, discrete-time signal is represented graphically and through a sequence.
The discrete-time signal given in the problem above is expressed using functional
representation. This means that the value of x(n) when the intervals are satisfied will be a
unit step.

Therefore, the Fourier Transform of the given problem can be expressed as

𝐿−1 𝐿−1
X(ω) = ∑𝑛=0 x(n)𝑒 −𝑗𝜔𝑛 = ∑𝑛=0 𝑒 −𝑗𝜔𝑛
Evaluating the integral,
𝐿−1

X(ω) = ∑ 𝑒 −𝑗𝜔𝑛
𝑛=0

𝟏 − 𝒆−𝒋𝝎𝑳
=
𝟏 − 𝒆−𝒋𝝎
which can be further expressed using identities.
Both the Fourier series and the Fourier transform for discrete-time signals are periodic
with period ω = 2π. As a result of this periodicity, the frequency range of discrete-time
signals is finite and extends from ω = -π to ω = π radians, where ω = π corresponds to the
highest possible rate of oscillation.
The figure below will illustrate this using Fast Fourier Transform or FFT in MATLAB. FFT
computes the discrete Fourier transform (DFT) of X using a fast Fourier transform (FFT)
algorithm.

Course Module
 Figure 2. Illustration of x = sin(2*pi*15*t) + sin(2*pi*40*t) DFT when N = 512. It shows a more defined
spectrum of the signals. Frequencies are easily identified.

Figure 3. Illustration of DFT of x = sin(2*pi*15*t) + sin(2*pi*40*t) when N = 64.

The frequency spectrum is spread compared to N = 512

Ideally, number of nfft point should be greater than the length of the signal. This will
‘reconstruct’ the signal better and going under L will cause an overlapping of samples
frequency and may lose information.
If N is selected such that N = L. then the DFT becomes
𝐿, 0
𝑋(𝑘) = {
0, 𝑘 = 1,2, … 𝐿 − 1

It is important to note that adding more nfft points doesn’t improve the resolution in the
sense that more values are added. It is just improved by the so called zero padding.
In summary, an FFT algorithm involves:
1. Storing the signal x(n) into a column vector
2. Multiply the resulting array by the phase factors
3. Compute the L-point DFT of each column
A summary of the properties of DFT could be seen in the table below.
 Signals, Spectra, and Digital Signal Processing
5
Introduction to Discrete Transforms`

Property Time Domain Frequency Domain

Notation x(n), y(n) X(k), Y(k)

Periodicity x(n) = x(n + N) X(k) = X(k+ N)

Linearity a1x1(n) + a2x2(n) a1X1(k) + a2X2(k)

Time Reversal x(N-n) X(N-k)

Multiplication of two x1(n)x2(n) 𝟏

X1(k) ∗ X2(k)
𝑵
sequences

Parseval’s Theorem 𝑁−1 𝟏

X(k) Y’(k)
𝑵
∑ x(n)y′(n)
𝑛=0

Table 1. Properties of DFT

Some other properties of DFT include the circular time and frequency shift as well as
the circular convolution.

References and Supplementary Materials

Books and Journals
1. Oppenheim, A.V., & Schafer, R.W. (2014). Discrete-Time Signal Processing. Pearson
Education Limited
2. Proakis, J. G., & Manolakis, D. G. (2007). Digital signal processing. Pearson Prentice
Hall
Online Supplementary Reading Materials
1. Discrete Fourier Transform. https://www.mathworks.com/help/signal/ug/discrete-
fourier-transform.html. November 30, 2017
2. Fast Fourier transform. https://www.mathworks.com/help/matlab/ref/fft.html.
November 30, 2017
Online Instructional Videos
1. Lecture 4: The Discrete-Time Fourier Transform. https://ocw.mit.edu/resources/res-
6-008-digital-signal-processing-spring-2011/video-lectures/lecture-4-the-discrete-
time-fourier-transform/. November 30, 2017
Note:
• Save each file using the format: Week005-Module
• There should only be one file for module per week.

Course Module
• Avoid plagiarism by paraphrasing information and texts coming from the Internet and
other sources.
 Signals, Spectra, and Digital Signal Processing
1
Z-Transform`

Module 005 Z-Transform

At the end of this module you are expected to:

1. Examine the discrete-time representation of the signal in the z-plane
2. Analyze the region of convergence of a signal.
3. Convert the time-domain representation of a signal to complex-frequency
domain.

The transformation of a signal from continuous-time to discrete-time form through

sampling doesn’t just involve the conversion of the nature of the signal. This may also allow
us to analyze the stability of the system through the use of the z-transform.

Z- Transform allows a complex plane representation of the signal or the system. Recall that
a complex plane (or z-plane) is a graphical representation of the real and imaginary
components .The graphical representation of the signal and the system may already
describe the characteristics of the signal or the system.
If the Laplace Transform is used to analyze the continuous-time form of the signal, its
discrete-time counterpart is the Z-Transform.

Direct Z-transform

Examine the direct Z-transform of a discrete-time signal defined as the power series:

Eq 1 X(z) = ∑∞
𝑛= −∞ 𝑥(𝑛)𝑧
−𝑛

where z is a complex variable.

Recall from a continuous-time form of a signal, we get the value of the signal from the
precious module by getting the integral of the function. As mentioned in the previous
modules, through sampling a signal takes a discrete-value at a specified interval defined by
the sampling rate.

Going back to the direct z-transform, this is the same as integrating all the components of
the signal from negative to positive infinity as explained in the previous modules.

Course Module
In order to get the value of a continuous-time signal, integration is performed. However, if
the signal is represented in discrete-time form, the discrete values are integrated or more
appropriately termed as summed to take the form of the signal.

Figure 1. A signal represented in continous-time and discrete-time form.

Mathematically, the Z-transform and inverse Z is represented by the following notations:

Eq 2 X(z) = Z { x(n) }

𝑧
Eq 3 x(n) X(z)

Graphically, it can be represented by a complex plane. Recall that a complex value is plotted
in a z-plane in this way:

Im(z)
z = r 𝑒 𝑗𝛳

Re(z)

Notice that the Z-transform is equivalent to the Discrete-Time Fourier Transform if r = 1.

This relationship is extremely helpful with the analysis of the response of the signal and the
system.

Example:
Recall how signals can be represented using sequence and graphically:
x(n) = δ(n) + 0.5δ(n-1)
 Signals, Spectra, and Digital Signal Processing
3
Z-Transform`

The Z-Transform of the signal given above is just straightforward:

Solution:

X(z) = ∑ 𝑥(𝑛)𝑧 −𝑛
𝑛= −∞

The x(n) which the magnitude of the impulse has values of 1 and 0.5 at n = 0 and n=1,
respectively.
Therefore, the Z-transform of the signal is:

X(z) = 𝑧 −0 + 0.5 𝑧 −1
or simply
X(z) = 1 + 0.5 𝑧 −1

which is just much simpler to express if the signal is represented in sequence form:

x(n) = [ 1 0.5]
which is similar to

X(z) = 1 + 0.5 𝑧 −1
with the exponent representing the shift of the impulses.

Course Module
However, z-transform comes along with the region of convergence (ROC) since its dealing
with representation in the z-plane.
Region of convergence or simply referred to as the ROC is the set of values of z where the
value of X(z) will be finite or in other words, convergent.

Region of Convergence (ROC)

Using the same example above, the z-transform of x(n) = δ(n) + 0.5δ(n-1) is
X(z) = 1 + 0.5 𝑧 −1 , to complete the Z-transform, the ROC must be determined.

The value of z that will make X(z) finite is ROC is entire z-plane z=0. The statement is true
because if z will be substituted to the expression X(z), the signal will become infinite.
Therefore, the complete z transform of
x(n) = δ(n) + 0.5δ(n-1)
is

X(z) = 1 + 0.5 𝑧 −1 , 𝑅𝑂𝐶: 𝑒𝑛𝑡𝑖𝑟𝑒 𝑧 − 𝑝𝑙𝑎𝑛𝑒 𝑒𝑥𝑐𝑒𝑝𝑡 0

Generally, the ROC of a finite duration causal signal is entire z-plane except 0.
Example 1:

Suppose the signal is represented by x(n) = δ(n+1) + 0.5δ(n+2)

Find its Z-transform of the signal.

Solution:

Again the Z-transform of this is just straightforward, taking to account the shift of the
impulses as the exponent of the z.
X(z) = 𝑧1 + 0.5 𝑧 2
The ROC of the signal in this example would be, entire z-plane except = ∞ . It is only when
𝑧 = ∞ that X(z).
Therefore, the complete z transform of
X(z) = 𝑧1 + 0.5 𝑧 2 , 𝑅𝑂𝐶: 𝑒𝑛𝑡𝑖𝑟𝑒 𝑧 − 𝑝𝑙𝑎𝑛𝑒 𝑒𝑥𝑐𝑒𝑝𝑡 ∞

Generally, the ROC of a finite duration anti-causal signal is entire z-plane except ∞.
Example 2:
Find the Z-transform of the signal given below:

x(n) = 3 δ(n+2)+ 2δ(n+1) + δ(n) + 0.5δ(n-1)

 Signals, Spectra, and Digital Signal Processing
5
Z-Transform`

Solution:
X(z) = 3 𝑧 2 + 2𝑧1 + 1 + 0.5 𝑧 −1
The region of convergence for this example is when z =∞ and z = 0.
Therefore, the complete z transform of
X(z) = 3 𝑧 2 + 2𝑧1 + 1 + 0.5 𝑧 −1 , 𝑅𝑂𝐶: 𝑒𝑛𝑡𝑖𝑟𝑒 𝑧 − 𝑝𝑙𝑎𝑛𝑒 𝑒𝑥𝑐𝑒𝑝𝑡 0 𝑎𝑛𝑑 ∞

Generally, the ROC of a finite duration two-sided signal is entire z-plane except 0 and ∞.

Example 3

Find the Z-transform of the signal given below:

x(n) = 0.5n u(n)

Solution:

It is easier to understand the Z-transform by expressing the signal first in sequence

form:

x(n) = 0.50 δ(n) + 0.5 δ(n-1) + 0.52 δ(n-2)+ …

Using the definition of the Z-transform

∞

X(z) = ∑ 𝑥(𝑛)𝑧 −𝑛
𝑛= −∞

But notice that there is no indication that the Z-transform is bilateral which means that the
n extends from -∞ to ∞ (non-causal). The u(n) where there is no advance (from -∞ to -1) ,
is considered a unilateral Z-transform which has limits from 0 to ∞ (causal).
∞

X(z) = ∑ 0.5𝑛 𝑧 −𝑛
𝑛= 0

or

𝑛= 0 (0.5𝑧 ) = = 0.5 z + 0.5 z + 0.5 z + …

X(z) = ∑∞ −1 𝑛 0 0 -1 2 -2

This is a geometric series expressed as:

Course Module
 1
𝑋(𝑧) =
1 − 0.5𝑧 −1

The value of the z where the geometric series is convergent would be at |z| > 0.5 which
has the same meaning of the ROC. Therefore, the Z-transform of x(n) = 0.5n u(n) is

1
𝑋(𝑧) = , ROC at |z| > 0.5
1−0.5𝑧 −1

Inverse Z-transform

It is best to understand the properties and common z-transform pairs for ease of
computations. Below is a table of Z- transform pairs and the properties of Z-transform.

COMMON Z TRANSFORM PAIRS

X(Z) ROC
x(n)
1, 𝑛=0 1 All z
𝛿(𝑛) = {
0, 𝑛≠0
z-k z≠ 1
𝛿(𝑛 − 𝑘)
u(n) 1 |z| > 1
1 − 𝑧 −1

1 |z| > |a|

an u(n)
1 − 𝑎𝑧 −1

𝑎𝑧 |z| > |a
nan u(n)
(𝑧 − 𝑎)2

an sin (ωn)
𝑎𝑧 sin (ω) |z| > |a|
𝑧2 − 2𝑎𝑧 cos 𝜔 + 𝑎2

an cos (ωn)
1 − 𝑎𝑧 cos (ω) |z| > |a|
𝑧2 − 2𝑎𝑧 cos 𝜔 + 𝑎2

Table 1. Table of unilateral Z-transform pairs

Example:

Find the inverse z-transform of the signal

X(z) = 0.5 + 0.5z -2

Knowing what the common z- transform pairs are, the inverse z of the signal given
above is:
 Signals, Spectra, and Digital Signal Processing
7
Z-Transform`

x(n) = 0.5δ(n) + 0.5δ(n-2)

Properties of Z- transform

The properties of the z-transform is summarized with the given table below

Property n Domain z Domain

Notation x(n), y(n) X(z), Y(z)

Linearity a1x1(n) + a2x2(n) a1X1(z) + a2X2(z)

Time Shifting x(n-k) z –k X(z)

Time Reversal x(-n) X(z-1)

Convolution of two x1(n) * x2(n) X1(z) X2(z)

sequences

Multiplication of two x1(n) x2(n) X1(z) * X2(z)

sequences

Table 2. Summary of properties of Z-transform

Linearity generally involves two properties - homogeneity and additivity. A third one
which is a requirement in DSP is shift invariance.

Homogeneity is a property which refers to the scaling or multiplying by a constant (i.e.

a1x1 (n)). A change in the amplitude in the input will result an amplitude change in the
output.

Additivity is the property which refers to whatever is added in the input will have a
corresponding change in the output.

Shift invariance refers to the shift in the input will have a corresponding shift in the
output.

Difference Equation

Following the properties of the z-transform, the input and output represented by x(n)
and y(n) in time domain, are X(z) and Y(z) in the z-domain, respectively.

There would be times where the input will be feedback into the output as shown in the
Figure 2.

Course Module
 Figure 2. The present output y(n) is the result of the summation of the input, x(n) and a past output y(n-delay).

The relationship of the past output and present and past input samples and the present
output sample is described by the difference equation.

Generally, the equation of the output referred to here is defined by the equation below:

y(n ) = b0 x(n) + b1x(n-1) + b2 x(n-2) + ….bnx(n-N) - a1y(n-1) - a2 y(n-2) - …- amy(n-M)

where an are the coefficients of the past output samples and bn are the coefficients of
the present and past output samples.

This will be best described in the latter modules which will discuss the rational z-
transform of transfer functions.

References and Supplementary Materials

Books and Journals
1. Oppenheim, A.V., & Schafer, R.W. (2014). Discrete-Time Signal Processing. Pearson
Education Limited
2. Proakis, J. G., & Manolakis, D. G. (2007). Digital signal processing. Pearson Prentice
Hall
Online Supplementary Reading Materials
1. z-transform.
http://web.stanford.edu/~kairouzp/teaching/ece310/secure/Chapter5.pdf.
December 20, 2017
Online Instructional Videos
1. Lecture 5: The z-Transform. https://ocw.mit.edu/resources/res-6-008-digital-signal-
processing-spring-2011/video-lectures/lecture-5-the-z-transform/ December 20,
2017
Note:
• Save each file using the format: Week006-Module
• There should only be one file for module per week.
• Avoid plagiarism by paraphrasing information and texts coming from the Internet and
other sources.