An Introduction to 𝑧 -space, the

𝑧 -transform and Digital Filter

Design
1. Introduction
• The z-transform is the discrete-space equivalent of the Laplace transform
• This transform is most widely applied in the analysis and design of infinite
impulse response (IIR) filter
• Also termed recursive filters because of the feedback nature of their operation
• Although it shares many similarities with the Laplace transform, there are
certain distinctions between these two transform of which we should be
familiar in order tat we can use them effectively
• These differences arise due to a seemingly obvious but fundamental
reason: because discrete signals must be sampled at some finite rate, they
must by definition be band limited
2. 𝑧 -Transform and its relationship to the
Laplace Transform
• The 𝑧-transform is similar in form to the DFT, and for a discrete signal x[n], the unilateral
version is given by ∞

𝑋 𝑧 =𝑍 𝑥 𝑛 = ෍ 𝑥 𝑛 𝑧 −𝑛
𝑛=0
where 𝑍 denotes the 𝑧 –transform and 𝑧 is given by
𝑧 = 𝑒 𝜎𝑇 𝑒 𝑗𝜔𝑇 = 𝑒 𝜎+𝑗𝜔 𝑇 = 𝑒 𝑠𝑇
• Hence,
𝑧 −1 = 𝑒 −𝑠𝑇
where 𝑇 represents the sample period
• Furthermore, the polar coordinates of a point 𝑋 𝑟, Φ in 𝑧 – space are given by
𝑟 = 𝑧 = 𝑒 𝑠𝑇
Φ = arg 𝑧 = 𝜔𝑇
ln 𝑟
𝜎=
𝑇
2. 𝑧 -Transform and its relationship to the
Laplace Transform
• Furthermore, using
𝑧 −1 = 𝑒 −𝑠𝑇
It can be shown that
2 1 − 𝑧 −1
𝑠=
𝑇 1 + 𝑧 −1

• We can always use the property of the z transform:

𝑍 𝑥 𝑛 =𝑋 𝑧
𝑍 𝑥 𝑛 − 1 = 𝑋 𝑧 𝑧 −1
Example 1
• Take the 𝑧 –transform of the following discrete signal:
𝑥 𝑛 = 4, −1, 7, 3, 2

• Using the transfer function in the s-domain as follows, change to the

z-transform and find the difference equation of the discrete signal
2𝑠+3
a. 𝐻 𝑠 = 𝑠+1 𝑠+6
1
b. 𝐻 𝑠 = 1+4𝑠
2. 𝑧 -Transform and its relationship to the
Laplace Transform
• Using previous equations it is possible to map a point in Laplace space
into z-space
• Now this latter domain is usually represented by the unit circle
• So-called because it has a radius of 1
• As with Laplace space, the horizontal axis is real and the vertical axis is
imaginary
• Any point on the circumference on the unit circle denotes a frequency,
the angle of which is given by 𝜔𝑇
• For example, assume we are sampling a signal at 10 Hz, i.e. 𝑇 = 0.1 s,
and we wish to represent frequencies at 1.25 Hz, 2.5 Hz, 3.75 Hz, 5 Hz
and 6.25 Hz
• These are indicated by points A to E in the figures and table in the next slide
2. 𝑧 -Transform and its relationship to the
Laplace Transform
Point on unit circle Frequency, Hz Frequency, 𝝎 Angle, 𝝎𝑻,where 𝑻 = 𝟎. 𝟏
A 1.25 7.854 𝜋/4
B 2.5 15.708 𝜋/2
C 3.75 23.562 3𝜋/4
D 5 31.416 𝜋

.
E 6.25 (3.75) 23.562 −3𝜋/4

. .
𝑗
B
C A

. 𝜔𝑇

E . −𝑗
2. 𝑧 -Transform and its relationship to the
Laplace Transform
• We know that the highest frequency that can be represented by a
𝑓
signal sampled at 𝑓 Hz is Hz
2

• The Nyquist point is denoted by 𝜋

• As a result, although point E represents a frequency of 6.25 Hz and

therefore and angle of 5𝜋/4, in fact if the system in our example
attempted to digitise it, it would appear as a frequency of 3.75 Hz (ie
as point C) with an angle of 3𝜋/4
3. The 𝑧 –Transform as a power series and its
role as a delay operator
• Imagine we have a causal signal 𝑥 𝑛 , given by
𝑥 𝑛 = 5, −2, 3, 7, −1 for 𝑛 = 0, … , 4
• Then the 𝑧-transform is given by
4

𝑋 𝑧 = ෍ 𝑥 𝑛 𝑧 −𝑛 = 𝑥 0 𝑧 0 + 𝑥 1 𝑧 −1 + 𝑥 2 𝑧 −2 + 𝑥 3 𝑧 −3 + 𝑥 4 𝑧 −4
𝑛=0
• This gives
𝑋 𝑧 = 5 − 2𝑧 −1 + 3𝑧 −2 + 7𝑧 −3 − 𝑧 −4
• An observation is that 𝑧 is essential a time shift operator
• Multiplication by 𝑧 advances a signal point by one sample interval, and multiplication
by 𝑧 −1 delays it by the same amount
• There is a very simple way to demonstrate the time-shifting property of the 𝑧-
transform; simply apply it to an impulse function and a delayed impulse function
Example 2
• Find the 𝑧-transforms of the following:
a. The impulse function 𝛿 𝑛
b. The weighted, delayed impulse function 𝐴𝛿 𝑛 − 3
4. Digital Filters, Diagrams and the 𝑧-transfer
function
• There are two main kinds of linear digital filter that operate in the time
domain; the Finite Impulse Response (FIR) and the Infinite Impulse
Response (IIR)

• There are only ever three operations that a processor uses to execute
them: time shift, multiplication and addition

• Because these three operations are of such pivotal importance not just to
digital filters but to whole of DSP in general, all real-time DSP devices are
designed to execute them as fast as possible
4. Digital Filters, Diagrams and the 𝑧-transfer
function
• A digital filter may be represented as an equation, sometimes called a
difference formula

• Or it may also be depicted as a diagram, the visual nature of which

often aids in the understanding of the flow of signals and coefficients,
and hence how the filter operates

• Filter block diagrams, as they are called portray the three key
operations of digital filters as shown in the next slide
4. Digital Filters, Diagrams and the 𝑧-transfer
function
• A square enclosing the symbol 𝑧 −1 (or sometimes 𝑇) denotes that a delay operation is applied to
the incoming signal

𝑧 −1
• A triangle enclosing a symbol or number indicates that the incoming signal is multiplied by this
term or value

ℎ𝑛

• Addition or summation is represented by two or more input signals to a circle enclosing a ∑

∑
4. Digital Filters, Diagrams and the 𝑧-transfer
function
• Example of digital filter: (a) FIR (b) Typical IIR filter block diagrams

(a) (b)
Exercise 3
• Draw an FIR filter as follows:
3

𝑦 𝑛 = ෍ℎ 𝑘 𝑥 𝑛−𝑘
𝑘=0
4. Digital Filters, Diagrams and the 𝑧-transfer
function
• The non-recursive FIR discrete-time equation is given by
𝑀−1

𝑦 𝑛 = ෍ ℎ 𝑘 𝑥 𝑛−𝑘
𝑘=0
• The equation above shows that the impulse response is time limited, i.e. it
comprises 𝑀 coefficients, or taps
• In contrasts, the IIR version involves a feedback structure and is therefore given
by
∞ 𝑀 𝑁

𝑦 𝑛 = ෍ℎ 𝑘 𝑥 𝑛−𝑘 = ෍𝑎 𝑘 𝑥 𝑛−𝑘 −෍𝑏 𝑘 𝑦 𝑛−𝑘

𝑘=0 𝑘=0 𝑘=1
• The IIR equation above reveals that although there is a finite number (𝑀+1) of
transversal coefficients and a finite number 𝑁 of recursive coefficients, because
of feedback the impulse response is infinite, or eternal
4. Digital Filters, Diagrams and the 𝑧-transfer
function
• As with the Fourier transform and the Laplace transform, the
principle of duality operates between the time domain and the 𝑧-
domain
• Hence:
𝑌 𝑧 =𝐻 𝑧 𝑋 𝑧 =𝑍 ℎ 𝑛 ∗𝑥 𝑛
• Rearranging equation above, we find in general, the transfer function
is given by
∞
𝑌 𝑧
𝐻 𝑧 = = ෍ ℎ 𝑛 𝑧 −𝑛
𝑋 𝑧
𝑛=0
4. Digital Filters, Diagrams and the 𝑧-transfer
function
• If we are dealing with an FIR filter, the transfer function is simply a single
polynomial of 𝑧 −1
𝑀−1

𝐻 𝑧 = ෍ ℎ 𝑛 𝑧 −𝑛 = ℎ 0 + ℎ 1 𝑧 −1 + ⋯ + ℎ 𝑀 − 1 𝑧 − 𝑀−1

𝑛=0
• In contrast, the transfer function of an IIR filter involves the ratio of two
polynomials:
𝑎 0 + 𝑎 1 𝑧 −1 + ⋯ + 𝑎 𝑚 𝑧 −𝑚 ∑𝑀𝑚=0 𝑎 𝑚 𝑧 −𝑚
𝐻 𝑧 = =
−1
1 +𝑏 1 𝑧 + ⋯+ 𝑏 𝑛 𝑧 −𝑛 1 + ∑𝑁𝑛=1 𝑏 𝑛 𝑧 −𝑛

• In the IIR transfer equation above, the numerator polynomial represents

the non-recursive, or FIR part of the filter, and the denominator
polynomial represents the recursive part
4. Digital Filters, Diagrams and the 𝑧-transfer
function
• This can be verified with respect to the IIR filter equation
• If there is no feedback, then the denominator becomes 1 and we are left with
a single polynomial
• The second feature of interest is the change in sign (+) in the
denominator, respecting the 𝑦 𝑛 coefficients, 𝑏 𝑛
• Refers back to the IIR equation:
𝑀 𝑁

𝑦 𝑛 = ෍𝑎 𝑘 𝑥 𝑛−𝑘 −෍𝑏 𝑘 𝑦 𝑛−𝑘

𝑘=0 𝑘=1
• In the IIR equation above the 𝑦 𝑛 term are show as negative
• This may readily be understood using the following reasoning
4. Digital Filters, Diagrams and the 𝑧-transfer
function
• Say we have a recursive system, given by
𝑦𝑛
=𝑎 0 𝑥 𝑛 +𝑎 1 𝑥 𝑛−1 +𝑎 2 𝑥 𝑛−2 −𝑏 1 𝑦 𝑛−1
−𝑏 2 𝑦 𝑛−2 −𝑏 3 𝑦 𝑛−3
• It can be shown that
𝑌 𝑧 ∑2𝑚=0 𝑎 𝑚 𝑧 −𝑚
𝐻 𝑧 = =
𝑋 𝑧 1 + ∑3𝑛=1 𝑏 𝑛 𝑧 −𝑛
• The chance in sign of the denominator polynomial is something that
must not be overlooked when moving from the transfer function to
expressing the filter in terms of its difference equation
Exercise 4
• Proof that the transfer function of
𝑦𝑛
=𝑎 0 𝑥 𝑛 +𝑎 1 𝑥 𝑛−1 +𝑎 2 𝑥 𝑛−2 −𝑏 1 𝑦 𝑛−1
−𝑏 2 𝑦 𝑛−2 −𝑏 3 𝑦 𝑛−3
is given by
𝑌 𝑧 ∑2𝑚=0 𝑎 𝑚 𝑧 −𝑚
𝐻 𝑧 = =
𝑋 𝑧 1 + ∑3𝑛=1 𝑏 𝑛 𝑧 −𝑛
Exercise 4
• Say we have a recursive system, given by
𝑦𝑛
=𝑎 0 𝑥 𝑛 +𝑎 1 𝑥 𝑛−1 +𝑎 2 𝑥 𝑛−2 −𝑏 1 𝑦 𝑛−1
−𝑏 2 𝑦 𝑛−2 −𝑏 3 𝑦 𝑛−3
• Re-arranging to group 𝑦 𝑛 and 𝑥 𝑛 yields:
𝑦 𝑛 +𝑏 1 𝑦 𝑛−1 +𝑏 2 𝑦 𝑛−2 +𝑏 3 𝑦 𝑛−3
=𝑎 0 𝑥 𝑛 +𝑎 1 𝑥 𝑛−1 +𝑎 2 𝑥 𝑛−2
• Taking the 𝑧-transform of both sides (and using the law of linearity), we
obtain
𝑌 𝑧 + 𝑏 1 𝑌 𝑧 𝑧 −1 + 𝑏 2 𝑌 𝑧 𝑧 −2 + 𝑏 3 𝑌 𝑧 𝑧 −3
= 𝑎 0 𝑋 𝑧 + 𝑎 1 𝑋 𝑧 𝑧 −1 + 𝑎 2 𝑋 𝑧 𝑧 −2
Exercise 4
𝑌 𝑧 1 + 𝑏 1 𝑧 −1 + 𝑏 2 𝑧 −2 + 𝑏 3 𝑧 −3
= 𝑋 𝑧 𝑎 0 + 𝑎 1 𝑧 −1 + 𝑎 2 𝑧 −2

• So finally, we have
𝑌 𝑧 𝑎 0 + 𝑎 1 𝑧 −1 + 𝑎 2 𝑧 −2
𝐻 𝑧 = =
𝑋 𝑧 1 + 𝑏 1 𝑧 −1 + 𝑏 2 𝑧 −2 + 𝑏 3 𝑧 −3
∑2𝑚=0 𝑎 𝑚 𝑧 −𝑚
=
1 + ∑3𝑛=1 𝑏 𝑛 𝑧 −𝑛