Digital Control System

Mechatronics Engineering
Future University

Dr. Sayed Ahmed Zaki

Electrical Engineering Department
Faculty of Engineering
Cairo University
 Digital Control System
Overview – Discrete System Analysis
Lecture Outline

• Stability Definition

• Relation between s-plane and z-plane

• Discretization of a continuous time system

– Discretization of a delayed system

• Hybrid block diagram manipulation

Stability
• Definition of external stability:
A bounded input would result in a bounded output (BIBO)
• Stability condition:
Assume a bounded input 𝑢 𝑖 , i.e. 𝑢 𝑖 ≤ 𝑀 < ∞, where 𝑀 is a
constant. Let ℎ 𝑘 be the impulse response coefficients.
• For a causal system, the output is given by
∞

𝑦 𝑖 = ℎ 𝑘−𝑖 𝑢 𝑘
𝑘=0
∞

𝑦 𝑖 = ℎ 𝑘−𝑖 𝑢 𝑘 ≤ ∞
𝑘=0
∞ ∞ ∞

𝑦 𝑖 = ℎ 𝑘−𝑖 𝑢 𝑘 ≤ ℎ 𝑘−𝑖 𝑢 𝑘 ≤𝑀 ℎ 𝑘−𝑖

𝑘=0 𝑘=0 𝑘=0
∞
• The output is bounded iff ℎ 𝑘−𝑖
𝑘=0 < ∞, i.e. the sum of the impulse
response coefficients should be finite.
Stability Check
• Since the sum of the impulse response coefficients should be finite, the
response must converge to zero.
• Since the response always depends on the poles magnitudes to power k,
where k is the time index, the magnitudes of the poles must be less than
one.
• Routh Table cannot be used since it checks whether there is a root in the
left hand side of the complex plane while we are interested in checking
whether there is a root outside the unit circle.

• A linear discrete-time system is stable if and only if all its poles are
inside the unit circle in z-domain.
 Relation between s-plane and z-plane
(constant settling time)

𝑗𝜔 𝑍 = 𝑒 𝑠𝑇 , 𝑠 = 𝜎 + 𝑗𝜔
jIm(z)

𝜎1 Circle of radius 𝑒 𝜎1𝑇

Re(z)
𝜎

The relation 𝑧 = 𝑒 𝜎1 𝑇 𝑒𝑗𝜔𝑇 is an equation of a circle of radius 𝑒 𝜎1 𝑇

in z-domain. If 𝜎1 = 0, 𝑖. 𝑒. 𝑗𝜔-axis, we get the unit circle in z-
domain.
As the poles of the system get closer to the unit circle in
z-domain, the response gets slower
 Relation between s-plane and z-plane
(constant frequency lines)
z= 𝑒 𝑠𝑇 , 𝑠 = 𝜎 + 𝑗𝜔
𝑗𝜔 𝑗Im(z)

𝑗𝜔1
𝜎 𝜔1 𝑇 Re(z)

The relation 𝑧 = 𝑒 𝜎𝑇 𝑒 𝑗𝜔1 𝑇 is an equation of a line of an angle 𝜔1𝑇.

As 𝜔1 increases, the angle of the line in z-domain increases and the
Response becomes more oscillatory.
 Relation between s-plane and z-plane
j Im(z)
j Im(s)
𝑍 = 𝑒 𝑠𝑇 𝑧 =1
𝜔𝑠
2
Re(z)
Re(s)
−𝜔𝑠
2

On the j𝜔-axis 𝑠 = 𝑗𝜔 → 𝑧 = 𝑒 𝑗𝜔𝑇 , this is an equation of

a circle of radius one.

2𝜋 𝜔 𝜔
Note that 𝜔𝑠 = 𝑇 , this means that 2𝑠 is mapped to 𝜋 and − 2𝑠 is
mapped to −𝜋. The strip to the left in s-domain is mapped inside
the unit circle in z-domain.
Relation between s-plane and z-plane

The stable region in s-domain is mapped inside the unit

circle in z-domain. 𝜔
Frequencies in the range 0, 𝑠 are mapped on the upper
2
half of the unit circle.
𝜔𝑠
Frequencies in the range 0 + 𝑘𝜔𝑠 , + 𝑘𝜔𝑠 , 𝑘 = 0, 1, ⋯
2
are mapped on the upper half of the unit circle. This leads
to aliasing.
The mapping of negative frequencies in s-domain is similar
to those in the positive range.
𝜔𝑠
The frequency is called Nyquist frequency.
2
Constant damping ratio and damping
frequency
1
0.5/T
0.6/T 0.4/T

0.8 0.7/T 0.1 0.3/T

𝜻: 𝐜𝐨𝐧𝐬𝐭.
𝝎𝒏 : 𝐜𝐨𝐧𝐬𝐭. 0.2

0.6 0.3
0.8/T 0.4 0.2/T
0.5
0.4 0.6
0.7
0.9/T 0.8 0.1/T
0.2
0.9

1/T
0
1/T

-0.2
0.9/T 0.1/T

-0.4

0.8/T 0.2/T
-0.6

-0.8 0.7/T 0.3/T

0.6/T 0.4/T
0.5/T
-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Dead Beat Response and Time Delay
𝑧−0.2
Consider the unit step response of the system 𝐺 𝑧 =
𝑧3
Step Response
1

0.9

0.8

0.7
If all poles are at the origin, the response
0.6
reaches steady state after a number of
Amplitude

0.5 samples equals the number of poles. This

0.4 called dead beat response.
0.3 The output remains unaffected by the input
0.2 for a number of samples equals the
0.1
difference between the number of poles
and zeros. This is called the delay.
0
0 2 4 6 8 10 12 14 16 18 20
Time (seconds)
Typical Impulse Response
 Discrete models of sampled data
systems
𝑢 𝑘𝑇 𝑦 𝑡 𝑦 𝑘𝑇
D/A 𝐺 𝑠 A/D

The D/A acts as a zero order hold (ZOH). The A/D acts as a
sampler.
𝑢 𝑘𝑇 𝑦 𝑘𝑇
ZOH 𝐺 𝑠
T
𝒀𝒛
In order to find the transfer function 𝑼 = ℤ 𝑮𝒁𝑶𝑯 𝒔 𝑮 𝒔 ,
𝒛
we have to determine 𝐺𝑍𝑂𝐻 𝑠 .
Transfer Function of the Zero Order Hold

If a unit impulse is applied to a ZOH, the output will be as in the figure.

𝑔𝑍𝑂𝐻 𝑡
1

-T T 2T 3T

This can be expressed as 𝑔𝑍𝑂𝐻 𝑡 = 𝑢 𝑡 − 𝑢 𝑡 − 𝑇 ,

where 𝑢 is the unit step.

𝟏 𝟏
The Laplace transform of 𝑔𝑍𝑂𝐻 𝑡 is 𝑮𝒁𝑶𝑯 𝒔 = − 𝒆−𝑻𝑺
𝒔 𝒔
 Graphical Calculation of the time response for a unit
step input
𝑒 𝑒∗ 𝑢 𝑦
1
𝑟 𝐺𝑍𝑂𝐻 𝑠
𝑇 𝑇𝑠
−

𝑟 𝑒 𝑒∗
1 1 1

T 2T 3T T 2T 3T T 2T 3T
𝑢 𝑦
1 1

T 2T 3T T 2T 3T
 Transfer function of a sampled data
system
𝑢 𝑘𝑇 𝑦 𝑘𝑇
D/A 𝐺 𝑠 A/D

𝑢 𝑘𝑇 𝑦 𝑘𝑇
ZOH 𝐺 𝑠
T
𝑌 𝑧
The transfer function 𝑈 = ℤ 𝐺𝑍𝑂𝐻 𝑠 𝐺 𝑠 is given by
𝑧
𝟏 − 𝒆−𝑻𝒔 𝑮 𝒔
𝑮 𝒛 =ℤ 𝑮 𝒔 = 𝟏 − 𝒛−𝟏 ℤ
𝒔 𝒔

The default in this course is to assume that a system is

preceded by a ZOH. So, G(z) is given as shown above.
 An Example of Calculating G(z)
𝑎
Assume that a sampled system has 𝐺 𝑠 = . Find 𝐺 𝑧 .
𝑠+𝑎
−1 𝐺 𝑠
Recall that 𝐺 𝑧 = 1 − 𝑧 ℤ .
𝑠
𝐺 𝑠 𝑎 1 1
First, we calculate ℤ =ℤ =ℤ − .
𝑠 𝑠 𝑠+𝑎 𝑠 𝑠+𝑎
1 𝑧
Note that ℤ = since this is a unit step function.
𝑠 𝑧−1
1 𝑧
and ℤ = since this is an exponential function
𝑠+𝑎 𝑧−𝑒 −𝑎𝑇
𝑒 −𝑎𝑡 , 𝑡 ≥ 0.
This leads to
−1 1 1 𝑧−1 𝑧 𝑧 1−𝑒 −𝑎𝑇
G 𝑧 = 1−𝑧 ℤ − = − = .
𝑠 𝑠+𝑎 𝑧 𝑧−1 𝑧−𝑒 −𝑎𝑇 𝑧−𝑒 −𝑎𝑇
Time delay systems

𝑎
Consider the system 𝐺 𝑠 = 𝑒 −𝜆𝑠
𝑠+𝑎
Let 𝜆 = 𝑙𝑇 − 𝑚𝑇 where 𝑙 is an integer and 0 < 𝑚 < 1.
𝜆 is called the time delay of the system.

Physical systems that have a delay include pipes, belt

conveyers, and any system that transports material.
pipe
If we monitor the liquid level in the
pump tank, the delay corresponds to the
level time required by the liquid to go from
the pump to the tank.
Time delay systems (contd.)

Example: Suppose 𝜆 = 1.5 𝑠𝑒𝑐𝑠. and 𝑇 = .2 𝑠𝑒𝑐.

We can write
0.1
𝜆 = 8 × 0.2 − × 0.2 = 8𝑇 − 0.5𝑇, 𝑖. 𝑒 𝑙 = 8 𝑎𝑛𝑑 𝑚 = 0.5
0.2
−𝜆𝑠 𝑎
Recall: 𝐺 𝑠 = 𝑒 ,𝜆 = 𝑙𝑇 − 𝑚𝑇
𝑠+𝑎

𝑧−1 −𝑙𝑇𝑠 𝑚𝑇𝑠 𝑎

Hence, 𝐺 𝑧 = ℤ 𝑒 𝑒
𝑧 𝑠(𝑠+𝑎)

Note 𝑒 −𝑙𝑇𝑠 = 𝑧 −𝑙 So, 𝐺 𝑧 =

𝑧−1 −𝑙
𝑧 ℤ 𝑒 𝑚𝑇𝑠
𝑎
𝑧 𝑠 𝑠+𝑎
Time delay systems (contd.)
𝑧 − 1 −𝑙 𝑚𝑇𝑠
𝑎
𝐺 𝑧 = 𝑧 ℤ 𝑒
𝑧 𝑠 𝑠+𝑎
𝑧 − 1 −𝑙 𝑚𝑇𝑠
1 1
𝐺 𝑧 = 𝑧 ℤ 𝑒 −
𝑧 𝑠 𝑠+𝑎
𝑧 − 1 −𝑙
𝐺 𝑧 = 𝑧 ℤ 𝑢𝑠 𝑡 + 𝑚𝑇 − 𝑒 −𝑎(𝑡+𝑚𝑇)
𝑧
𝑧
Note that ℤ 𝑢𝑠 𝑡 + 𝑚𝑇 = ℤ 𝑢𝑠 𝑡 =
𝑧−1
𝑧
and that ℤ 𝑒 −𝑎 𝑡+𝑚𝑇
= 𝑒 −𝑎𝑚𝑇 ℤ 𝑒 −𝑎𝑡 = 𝑒 −𝑎𝑚𝑇
𝑧−𝑒 −𝑎𝑇
𝑧 − 1 −𝑙 𝑧 −𝑎𝑚𝑇
𝑧
𝐺 𝑧 = 𝑧 −𝑒
𝑧 𝑧−1 𝑧 − 𝑒 −𝑎𝑇
 Time delay systems (contd.)

𝑧+𝛼
We conclude that 𝐺 𝑧 = 1 − 𝑒 −𝑎𝑚𝑇
𝑧 𝑙 𝑧−𝑒 −𝑎𝑇
𝑒 −𝑎𝑚𝑇 −𝑒 −𝑎𝑇
where 𝛼 =
1−𝑒 −𝑎𝑚𝑇
The original system has a pole at 𝑠 = −𝑎. It is mapped to
𝑧 = 𝑒 −𝑎𝑇 , i.e. it follows the relation 𝑧 = 𝑒 𝑠𝑇 .

The number of poles exceeds the number of zeros by 𝑙.

So, the discrete system has a delay equals to 𝑙 samples.

The order of the discrete-time model is finite and equal to 𝑙 + 1.

The order of the continuous-time model (with a delay) is infinite.

 Block Diagrams

We will consider two cases:

1- All blocks are in z-domain

𝐶 𝑧 𝐺 𝑧

2- Some blocks are in s-domain and the others are in z-domain

𝐶 𝑧 𝐺 𝑠
 Block diagrams with all blocks are in z-domain

Series blocks 𝐶 𝑧 𝐺 𝑧
𝑈 𝑧 𝑌=𝐶 𝑧 𝐺 𝑧 𝑈

Parallel blocks 𝐶 𝑧
+
𝑈 𝑧
𝑌 = 𝐶 𝑧 +𝐺 𝑧 𝑈
+
𝐺 𝑧
Feedback systems 𝐺
𝑅 𝑧 𝐺 𝑧 𝑌 𝑧 = 𝑈
- 1 + 𝐺𝐻

𝐻 𝑧
Basic rules to discretize continuous-time
systems
T T T
𝑈 𝑠 𝑌∗ 𝑠
𝐻1 𝑠 𝐻2 𝑠
𝑈∗ 𝑠 𝑋 𝑠 𝑋∗ 𝑠 𝑌 𝑠

∗ ∗ ∗
We can write 𝑌 𝑠 = 𝐻2 𝑠 𝑋 𝑠
∗ ∗ ∗ ∗
⟹ 𝑌 𝑠 = 𝐻2 𝑠 𝐻1 𝑠 𝑈 𝑠
∗
Note that 𝐻1 𝑠 𝑈 𝑠 ∗
= 𝐻1∗ 𝑠 𝑈∗ 𝑠 = 𝐻1 𝑧 𝑈 𝑧
∗
∗
Hence, 𝑌 𝑠 = 𝐻2 𝑠 𝐻1∗ ∗
𝑠 𝑈 𝑠 = 𝐻2∗ 𝑠 𝐻1∗ 𝑠 𝑈∗ 𝑠

i.e. 𝑌 𝑧 = 𝐻1 𝑧 𝐻2 𝑧 𝑈 𝑧
Basic rules to discretize continuous-time
systems
T T
𝑈 𝑠 𝑌∗ 𝑠
𝐻1 𝑠 𝐻2 𝑠
𝑈∗ 𝑠 𝑋 𝑠 𝑌 𝑠

∗ ∗ ∗
Note that 𝑌 𝑠 = 𝐻1 𝑠 𝐻2 𝑠 𝑈 𝑠
∗ ∗
This leads to 𝑌 𝑠 = 𝐻1 𝑠 𝐻2 𝑠 𝑈∗ 𝑠

i.e. 𝑌 𝑧 = ℤ 𝐻1 𝑠 𝐻2 𝑠 𝑈 𝑧

An important note: ℤ 𝑯𝟏 𝒔 𝑯𝟐 𝒔 ≠ 𝑯𝟏 𝒛 𝑯𝟐 𝒛
 Example 1 G(s)

-V +V
A/D Computer D/A
R -
C(z)=1
Mechanical load
H(s)
Power Amplifier motor
Filter
Y: position
The equivalent block diagram

R Y
G(s)
-
𝑌 𝑧 =?
H(s)
Example 1 (contd.)
R G(s)
Y
- E

H(s)

From the block diagram 𝐸 = 𝑅 − 𝐻𝑌 (1)

Also 𝑌 = 𝐺𝐸 ∗ (2)
From (1)-(2): 𝐸 ∗ = 𝑅 ∗ − 𝐻𝑌 ∗
= 𝑅 ∗ − 𝐻𝐺𝐸 ∗ ∗
(3)
Eq. (3) leads to: 𝐸 ∗ = 𝑅 ∗ − 𝐻𝐺 ∗ 𝐸 ∗ (4)
∗ 𝑅∗
Eq. (4) leads to: 𝐸 = (5)
1+ 𝐺𝐻 ∗
∗ ∗ ∗ 𝐺∗ ∗ 𝐺 𝑧
From (2)-(5): 𝑌 = 𝐺 𝐸 = 𝑅 = 𝑅 𝑧
1+ 𝐺𝐻 ∗ 1+ℤ 𝐺𝐻
Example 2
E
R 𝒀∗
G(s)
-
𝑌 𝑧 =?
H(s)

From the block diagram 𝑌 ∗ = 𝐺𝐸 ∗ ∗

= 𝐺∗𝐸∗ (6)
From the block diagram 𝐸 = 𝑅 − 𝐻𝑌 ∗ (7)

From Eq. (7): 𝐸 ∗ = 𝑅 ∗ − 𝐻 ∗ 𝑌 ∗ (8)

Substituting for 𝐸 ∗ from (8) into (6): 𝑌 ∗ = 𝐺 ∗ 𝑅∗ − 𝐻 ∗ 𝑌 ∗
𝐺 ∗ 𝐺 𝑧
This leads to: 𝑌 ∗ = 𝑅 ∗ 𝑌 𝑧 = 𝑅 𝑧
∗
1+𝐺 𝐻 ∗ 1+𝐺 𝑧 𝐻 𝑧