ELE8311: Transfer Function

Dr. Hassan A. Bashir

ELE8311: Digital Control Engineering

3 Transfer function

Transfer function 𝐺𝐺(𝑧𝑧) for discrete-time systems can also be derived from their difference
equations. In addition, poles and zeros of 𝐺𝐺(𝑧𝑧) are defined in the same way as in continuous-
time domain.

Let the transfer function be

𝑁𝑁(𝑧𝑧) 𝑘𝑘(𝑧𝑧 + 𝑧𝑧1 )(𝑧𝑧 + 𝑧𝑧2 ) … (𝑧𝑧 + 𝑧𝑧𝑛𝑛 )

𝐺𝐺(𝑧𝑧) = =
𝐷𝐷(𝑧𝑧) (𝑧𝑧 + 𝑝𝑝1 )(𝑧𝑧 + 𝑝𝑝2 ) … (𝑧𝑧 + 𝑝𝑝𝑛𝑛 )

If 𝑁𝑁(𝑧𝑧) and 𝐷𝐷(𝑧𝑧) share no common factors, then the roots of 𝑁𝑁(𝑧𝑧) are the zeros and the
roots of 𝐷𝐷(𝑧𝑧) are the poles of 𝐺𝐺(𝑧𝑧).

3.1 Transfer function of connected systems

Typically, linear systems involve connection of subsystems in series, parallel, or as feedback

loops.

A. Parallel systems:

𝑔𝑔(𝑘𝑘) = 𝑔𝑔1 (𝑘𝑘) + 𝑔𝑔2 (𝑘𝑘)

Applying the linearity property of Z-transform yields:

𝐺𝐺(𝑧𝑧) = 𝐺𝐺1 (𝑧𝑧) + 𝐺𝐺2 (𝑧𝑧).

B. Series systems:

𝑔𝑔(𝑘𝑘) = 𝑔𝑔1 (𝑘𝑘) ∗ 𝑔𝑔2 (𝑘𝑘)

Applying the convolution property of Z-transform yields:

𝐺𝐺(𝑧𝑧) = 𝐺𝐺1 (𝑧𝑧)𝐺𝐺2 (𝑧𝑧).

C. Feedback loop

Fig. 3.1: Sampled Feedback Control

1
 ELE8311: Transfer Function

Let the z-transform be:

𝑌𝑌(𝑧𝑧) = 𝐺𝐺(𝑧𝑧)𝑈𝑈(𝑧𝑧) = 𝐺𝐺(𝑧𝑧)𝐺𝐺𝑐𝑐 (𝑧𝑧)𝐸𝐸(𝑧𝑧)

= 𝐺𝐺(𝑧𝑧)𝐺𝐺𝑐𝑐 (𝑧𝑧)(𝑅𝑅(𝑧𝑧) − 𝑌𝑌(𝑧𝑧))

∴ 𝑌𝑌(𝑧𝑧) + 𝐺𝐺(𝑧𝑧)𝐺𝐺𝑐𝑐 (𝑧𝑧)𝑌𝑌(𝑧𝑧) = 𝐺𝐺(𝑧𝑧)𝐺𝐺𝑐𝑐 (𝑧𝑧)𝑅𝑅(𝑧𝑧)

𝐺𝐺(𝑧𝑧)𝐺𝐺𝑐𝑐 (𝑧𝑧)
𝑌𝑌(𝑧𝑧) = 𝑅𝑅(𝑧𝑧)
1 + 𝐺𝐺(𝑧𝑧)𝐺𝐺𝑐𝑐 (𝑧𝑧)

Hence, the closed loop transfer function 𝑇𝑇(𝑧𝑧) is 𝑌𝑌(𝑧𝑧). Also, the relation between the response
and system error is:

𝐸𝐸(𝑧𝑧) = 𝑆𝑆(𝑧𝑧)𝑅𝑅(𝑧𝑧)

where

1
𝑆𝑆(𝑧𝑧) = .
1 + 𝐺𝐺(𝑧𝑧)𝐺𝐺𝑐𝑐 (𝑧𝑧)

The effect of a sampler

Similarly, for the systems in Fig. 3.2 (a), (b) and (c), we have:

𝐺𝐺1 (𝑧𝑧)𝐺𝐺2 (𝑧𝑧)

𝐶𝐶 (𝑧𝑧) = 𝑅𝑅(𝑧𝑧)
1+𝐺𝐺1 (𝑧𝑧)𝐺𝐺2 𝐻𝐻(𝑧𝑧)
(a)

𝐺𝐺1 (𝑧𝑧)𝐺𝐺2 (𝑧𝑧)

𝐶𝐶 (𝑧𝑧) = 𝑅𝑅(𝑧𝑧)
1+𝐺𝐺2 𝐻𝐻1 (𝑧𝑧)+𝐺𝐺1 (𝑧𝑧)𝐺𝐺2 𝐻𝐻2 (𝑧𝑧)
(b)

𝐺𝐺1 (𝑧𝑧)𝐺𝐺2 (𝑧𝑧)𝐺𝐺3 (𝑧𝑧)

𝐶𝐶 (𝑧𝑧) = 𝑅𝑅
1+𝐺𝐺2 (𝑧𝑧)𝐺𝐺3 𝐻𝐻1 (𝑧𝑧)+𝐺𝐺2 (𝑧𝑧)𝐺𝐺1 𝐺𝐺3 𝐻𝐻2 (𝑧𝑧)
(c)
Fig. 3.2: Feedback Control variants
2
 ELE8311: Transfer Function

Example 3.1

Determine the equivalent sampled response sequence and z-transfer function for the cascade
of the two analog systems shown in Fig. 3.3 when

i. The systems are directly connected.

ii. The systems are separated by a sampler.

1 2
𝐻𝐻1 (𝑠𝑠) = , 𝐻𝐻2 (𝑠𝑠) =
(𝑠𝑠 + 2) (𝑠𝑠 + 4)

Fig. 3.3: Analog Cascaded System

Solution

i. When no sampler between the systems in Fig. 3.3, the overall transfer function is

2
𝐻𝐻(𝑠𝑠) = 𝐻𝐻1 (𝑠𝑠)𝐻𝐻2 (𝑠𝑠) =
(𝑠𝑠 + 2)(𝑠𝑠 + 4)

1 1
𝐻𝐻(𝑠𝑠) = −
𝑠𝑠 + 2 𝑠𝑠 + 4

The time sequence is therefore

ℎ(𝑡𝑡) = 𝑒𝑒 −2𝑡𝑡 − 𝑒𝑒 −4𝑡𝑡

and the sampled transfer function is

ℎ(𝑘𝑘𝑘𝑘) = 𝑒𝑒 −2𝑘𝑘𝑘𝑘 − 𝑒𝑒 −4𝑘𝑘𝑘𝑘 , 𝑘𝑘 = 0,1,2, …

Hence, the z-transfer function is

𝑧𝑧 𝑧𝑧 (𝑒𝑒 −2𝑇𝑇 − 𝑒𝑒 −4𝑇𝑇 )𝑧𝑧

𝐻𝐻(𝑧𝑧) = − =
𝑧𝑧 − 𝑒𝑒 −2𝑇𝑇 𝑧𝑧 − 𝑒𝑒 −4𝑇𝑇 (𝑧𝑧 − 𝑒𝑒 −2𝑇𝑇 )(𝑧𝑧 − 𝑒𝑒 −4𝑇𝑇 )

ii. When the analog transfer functions in Fig. 3.3 are separated by a sampler, then
each has a z-transfer function given by:

𝑧𝑧 2𝑧𝑧
𝐻𝐻1 (𝑧𝑧) = , 𝐻𝐻2 (𝑧𝑧) =
𝑧𝑧 − 𝑒𝑒 −2𝑇𝑇 𝑧𝑧 − 𝑒𝑒 −4𝑇𝑇

3
 ELE8311: Transfer Function

The overall transfer function is therefore

2𝑧𝑧 2
𝐻𝐻(𝑧𝑧) =
(𝑧𝑧 − 𝑒𝑒 −2𝑇𝑇 )(𝑧𝑧 − 𝑒𝑒 −4𝑇𝑇 )

Using partial fraction to expand yields

2 𝑒𝑒 −2𝑇𝑇 𝑧𝑧 𝑒𝑒 −4𝑇𝑇 𝑧𝑧
𝐻𝐻(𝑧𝑧) = −2𝑇𝑇 � − �
𝑒𝑒 − 𝑒𝑒 −4𝑇𝑇 𝑧𝑧 − 𝑒𝑒 −2𝑇𝑇 𝑧𝑧 − 𝑒𝑒 −4𝑇𝑇

Taking the inverse transforms yields the following sequence

2
ℎ(𝑘𝑘𝑘𝑘) = [𝑒𝑒 −2𝑇𝑇 𝑒𝑒 −2𝑘𝑘𝑘𝑘 − 𝑒𝑒 −4𝑇𝑇 𝑒𝑒 −4𝑘𝑘𝑘𝑘 ]
𝑒𝑒 −2𝑇𝑇 − 𝑒𝑒 −4𝑇𝑇
2
= �𝑒𝑒 −2(𝑘𝑘+1)𝑇𝑇 − 𝑒𝑒 −4(𝑘𝑘+1)𝑇𝑇 �, 𝑘𝑘 = 0,1,2, …
𝑒𝑒 −2𝑇𝑇 − 𝑒𝑒 −4𝑇𝑇

3.2 Transfer function of the Zero-order Hold (ZOH)

The transfer function of the zoh is obtained by evaluating the difference between the Laplace
transform of a unit step and the time delay theorem, i.e.

1 𝑒𝑒 −𝑠𝑠𝑠𝑠
𝐿𝐿{1(𝑡𝑡)} = , 𝐿𝐿{1(𝑡𝑡 − 𝑇𝑇)} =
𝑠𝑠 𝑠𝑠

Thus, the transfer function of the zoh is

1 − 𝑒𝑒 −𝑠𝑠𝑠𝑠
𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑠𝑠) =
𝑠𝑠

Hence, for an analog plant 𝐺𝐺(𝑠𝑠) sampled with a zoh 𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑠𝑠), the overall transfer function
𝐺𝐺𝑍𝑍𝑍𝑍 (𝑠𝑠) is

𝐺𝐺(𝑠𝑠)
𝐺𝐺𝑍𝑍𝑍𝑍 (𝑠𝑠) = 𝐺𝐺(𝑠𝑠)𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑠𝑠) = (1 − 𝑒𝑒 −𝑠𝑠𝑠𝑠 )
𝑠𝑠

In discrete domain, we have

𝐺𝐺(𝑠𝑠)
𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑠𝑠) = (1 − 𝑧𝑧 −1 )𝑍𝑍 � � (3.1)
𝑠𝑠

4
 ELE8311: Transfer Function

Example 3.2

Assume the cruise control of an automobile shown in Fig. 3.4 is represented by the following
model:

Fig. 3.4: Cruise control model of an automobile

𝑀𝑀 𝑣𝑣̇ (𝑡𝑡) + 𝑏𝑏𝑏𝑏(𝑡𝑡) = 𝑢𝑢(𝑡𝑡)

where, 𝑢𝑢 is the input force, 𝑣𝑣 is the velocity of the car, and 𝑏𝑏 is the viscous friction coefficient.
Find the discrete transfer function given a ZOH sampler.

Solution

From the mathematical model of the system, we have by taking the L transforms:

𝑉𝑉(𝑠𝑠) 1
𝐺𝐺(𝑠𝑠) = =
𝑈𝑈(𝑠𝑠) 𝑀𝑀𝑀𝑀 + 𝑏𝑏

Compared to a standard first order system transfer function, we have

𝐾𝐾 𝐾𝐾/𝜏𝜏
𝐺𝐺(𝑠𝑠) = =
𝜏𝜏𝜏𝜏 + 1 𝑠𝑠 + 1/𝜏𝜏
𝐺𝐺(𝑠𝑠)
where 𝐾𝐾 = 1/𝑏𝑏 and 𝜏𝜏 = 𝑀𝑀/𝑏𝑏. Then, partial fraction expansion of yields
𝑠𝑠

𝐺𝐺(𝑠𝑠) 𝐾𝐾 𝐴𝐴 𝐵𝐵
= � �� + �
𝑠𝑠 𝜏𝜏 𝑠𝑠 𝑠𝑠 + 1
𝜏𝜏

Therefore, solving for the constants, we have

𝐺𝐺(𝑠𝑠) 𝐾𝐾 𝜏𝜏 𝜏𝜏
= � �� − �
𝑠𝑠 𝜏𝜏 𝑠𝑠 𝑠𝑠 + 1
𝜏𝜏

Hence, the desired z-transfer function is obtained from equation (3.1) as:

5
 ELE8311: Transfer Function

𝐾𝐾 𝜏𝜏 𝜏𝜏
𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧) = (1 − 𝑧𝑧 −1 )𝑍𝑍 �� � � − ��
𝜏𝜏 𝑠𝑠 𝑠𝑠 + 1
𝜏𝜏
𝑇𝑇
𝑧𝑧 − 1 2𝑧𝑧 − �1 + 𝑒𝑒 −𝜏𝜏 �
= 𝐾𝐾 �1 + 𝑇𝑇 � = 𝐾𝐾 � 𝑇𝑇 �.
− −
𝑧𝑧 − 𝑒𝑒 𝜏𝜏 𝑧𝑧 − 𝑒𝑒 𝜏𝜏

Example 3.3

Find the 𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧) for the series RL circuit with the inductor as the output.

Solution

Using voltage divider rule on a series RL circuit yields

𝑉𝑉𝑜𝑜 𝐿𝐿𝐿𝐿 𝜏𝜏𝜏𝜏 𝐿𝐿

= = , 𝜏𝜏 =
𝑉𝑉𝑖𝑖𝑖𝑖 𝑅𝑅 + 𝐿𝐿𝐿𝐿 1 + 𝜏𝜏𝜏𝜏 𝑅𝑅

Hence, using 3.1 we have:

1
𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧) = (1 − 𝑧𝑧 −1 )𝑍𝑍 � �
1
𝑠𝑠 + 𝜏𝜏

𝑧𝑧 − 1 𝑧𝑧 𝑧𝑧 − 1
= × 𝑇𝑇 = 𝑇𝑇
𝑧𝑧
𝑧𝑧 − 𝑒𝑒 −𝜏𝜏 𝑧𝑧 − 𝑒𝑒 −𝜏𝜏

Exercise 3.1

Determine the 𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧) for the vehicle position control system shown Fig. 3.3, where, 𝑢𝑢 is the
input force, 𝑦𝑦 is the velocity of the car, and 𝑏𝑏 is the viscous friction coefficient.

Hint: the model is: 𝑀𝑀 𝑦𝑦̈ (𝑡𝑡) + 𝑏𝑏 𝑦𝑦̇ (𝑡𝑡) = 𝑢𝑢(𝑡𝑡)

1 𝜏𝜏(𝑧𝑧−1)
Ans: 𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧) = 𝐾𝐾 �𝑧𝑧−1 − 𝜏𝜏 + 𝑇𝑇 �.
−
𝑧𝑧−𝑒𝑒 𝜏𝜏

Exercise 3.2

Determine the z-transfer function of an armature-controlled DC motor represented by the

following system of dynamic equations.

6
 ELE8311: Transfer Function

𝑑𝑑𝑑𝑑(𝑡𝑡)
𝐽𝐽 𝜃𝜃̈(𝑡𝑡) + 𝑏𝑏 𝜃𝜃̇ (𝑡𝑡) = 𝐾𝐾𝑡𝑡 𝑖𝑖(𝑡𝑡), 𝐿𝐿 + 𝑅𝑅𝑅𝑅(𝑡𝑡) = 𝑢𝑢(𝑡𝑡) − 𝐾𝐾𝑒𝑒 𝜃𝜃̇(𝑡𝑡), 𝑦𝑦(𝑡𝑡) = 𝜃𝜃(𝑡𝑡);
𝑑𝑑𝑑𝑑

where the system output 𝑦𝑦 is the position of the shaft 𝜃𝜃, I is the armature current, 𝑢𝑢 is the
source voltage (system’s input), 𝐽𝐽 is the inertia of the motor, 𝑏𝑏 is the viscous friction
coefficient, 𝐾𝐾𝑡𝑡 is the armature constant, 𝐾𝐾𝑒𝑒 is the motor torque constant, 𝑅𝑅 is the electric
resistance and 𝐿𝐿 is inductance.
𝑌𝑌(𝑠𝑠) 𝐾𝐾 𝐾𝐾
Hint: the transfer function is 𝐺𝐺(𝑠𝑠) = 𝑡𝑡
= 𝑠𝑠[(𝐽𝐽𝐽𝐽 +𝑏𝑏)(𝐿𝐿𝐿𝐿+𝑅𝑅)+𝐾𝐾 = 𝑠𝑠(𝑠𝑠+𝑝𝑝 .
𝑈𝑈(𝑠𝑠) 𝐾𝐾 ]
𝑡𝑡 𝑒𝑒 1 )(𝑠𝑠+𝑝𝑝 2 )

Ans:

1 𝑝𝑝1 + 𝑝𝑝2 1
𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧) = 𝐾𝐾 � − 2 2 + 2
𝑝𝑝1 𝑝𝑝2 (𝑧𝑧 − 1) 𝑝𝑝1 𝑝𝑝2 𝑝𝑝1 (𝑝𝑝2 − 𝑝𝑝1 )(𝑧𝑧 − 1)(𝑧𝑧 − 𝑒𝑒 −𝑝𝑝 1 𝑇𝑇 )
1
− 2 �.
𝑝𝑝2 (𝑝𝑝2 − 𝑝𝑝1 )(𝑧𝑧 − 1)(𝑧𝑧 − 𝑒𝑒 −𝑝𝑝 2 𝑇𝑇 )

3.3 Time Response of a Discrete-time System

For discrete-time systems, the time response is the solution of the difference equation
governing the system. For LTI systems, the response consists of the forced or input
component and that due to initial conditions. While the input response for a given discrete
system is derived via a convolution sum of its input and its response to a unit impulse signal,
the response can also be obtained by algebraic product of their z-transforms.

Procedure:

i. Find the z-transform for the system and the input

ii. Multiply the two transforms
iii. Invert the z-transform to obtain a temporal output sequence.

Example 3.4

For the given discrete-time signal, find the impulse response ℎ(𝑘𝑘).

𝑦𝑦(𝑘𝑘 + 1) − 0.5𝑦𝑦(𝑘𝑘) = 𝑢𝑢(𝑘𝑘), 𝑦𝑦(0) = 0

Solution:

Taking the z-transform yields

𝑧𝑧𝑧𝑧(𝑧𝑧) − 𝑧𝑧𝑧𝑧(0) − 0.5𝑌𝑌(𝑧𝑧) = 1𝑈𝑈(𝑧𝑧)

7
 ELE8311: Transfer Function

At 𝑦𝑦(0) = 0, we have

𝑌𝑌(𝑧𝑧) 1
𝐻𝐻(𝑧𝑧) = =
𝑈𝑈(𝑧𝑧) 𝑧𝑧 − 0.5

Taking the inverse transform yields:

(0.5)𝑘𝑘−1 , 𝑘𝑘 = 1,2,3, …
ℎ(𝑘𝑘) = �
0, 𝑘𝑘 < 1

Exercise 3.3

Derive the system transfer function and its response due a unit-step input.

𝑦𝑦(𝑘𝑘 + 1) − 𝑦𝑦(𝑘𝑘) = 𝑢𝑢(𝑘𝑘 + 1).

𝑧𝑧 𝑘𝑘 + 1, 𝑘𝑘 = 0,1,2,3 …
Ans: 𝐹𝐹(𝑧𝑧) = (𝑧𝑧−1)2 and 𝑦𝑦(𝑘𝑘) = �
0, 𝑘𝑘 < 0

Example 3.5

Find the z-transform from the following transfer function

5(𝑠𝑠 + 4)
𝐺𝐺(𝑠𝑠) =
𝑠𝑠(𝑠𝑠 + 1)(𝑠𝑠 + 2)

Solution

5(𝑠𝑠 + 4) 10 15 5
𝑍𝑍 � � = 𝑍𝑍 � + + �
𝑠𝑠(𝑠𝑠 + 1)(𝑠𝑠 + 2) 𝑠𝑠 𝑠𝑠 + 1 𝑠𝑠 + 2

10𝑧𝑧 15𝑧𝑧 5𝑧𝑧

= − + .
𝑧𝑧 − 1 𝑧𝑧 − 𝑒𝑒 −𝑇𝑇 𝑧𝑧 − 𝑒𝑒 −2𝑇𝑇

Example 3.6

Determine the output response for the sampled system shown in Fig. 3.5. Assume 𝑘𝑘 = 10,
𝑇𝑇 = 0.5s for a unit step input 𝑟𝑟(𝑡𝑡).

Fig. 3.5
8
 ELE8311: Transfer Function

Solution

From the system’s block diagram, we have

𝐶𝐶(𝑘𝑘) 𝐾𝐾(5𝑠𝑠 + 1)
=
𝑒𝑒 ∗ (𝑘𝑘) 𝑠𝑠(2𝑠𝑠 + 1)

Therefore,

2𝑠𝑠 2 𝐶𝐶(𝑠𝑠) + 𝑠𝑠𝑠𝑠(𝑠𝑠) = 5𝐾𝐾𝐾𝐾𝑒𝑒 ∗ (𝑠𝑠) + 𝐾𝐾𝑒𝑒 ∗ (𝑠𝑠)

2𝑑𝑑 2 𝑐𝑐(𝑡𝑡) 𝑑𝑑𝑑𝑑(𝑡𝑡) 5𝐾𝐾𝐾𝐾𝑒𝑒 ∗ (𝑡𝑡)

+ = + 𝐾𝐾𝑒𝑒 ∗ (𝑡𝑡)
𝑑𝑑𝑡𝑡 2 𝑑𝑑𝑑𝑑 𝑑𝑑𝑑𝑑

Now, in discrete domain, we have:

2𝑑𝑑 2 𝑐𝑐(𝑡𝑡) 𝑐𝑐(𝑘𝑘) − 2𝑐𝑐(𝑘𝑘 − 1) + 𝑐𝑐(𝑘𝑘 − 2)

≈ 2 ;
𝑑𝑑𝑡𝑡 2 𝑇𝑇 2
𝑑𝑑𝑑𝑑(𝑡𝑡) 𝑐𝑐(𝑘𝑘) − 𝑐𝑐(𝑘𝑘 − 1)
≈ ;
𝑑𝑑𝑑𝑑 𝑇𝑇

5𝐾𝐾𝐾𝐾𝑒𝑒 ∗ (𝑡𝑡) 5𝐾𝐾�𝑒𝑒 ∗ (𝑘𝑘) − 𝑒𝑒 ∗ (𝑘𝑘 − 1)�

≈ ;
𝑑𝑑𝑑𝑑 𝑇𝑇
𝐾𝐾𝑒𝑒 ∗ (𝑡𝑡) ≈ 𝐾𝐾𝑒𝑒 ∗ (𝑘𝑘).

Thus, the difference equation is therefore,

2 + 𝑇𝑇 4 + 𝑇𝑇 2 5𝐾𝐾 + 𝐾𝐾𝐾𝐾 ∗ 5𝐾𝐾 ∗

2
𝑐𝑐(𝑘𝑘) − 2
𝑐𝑐(𝑘𝑘 − 1) + 2 𝑐𝑐(𝑘𝑘 − 2) = 𝑒𝑒 (𝑘𝑘) + 𝑒𝑒 (𝑘𝑘 − 1)
𝑇𝑇 𝑇𝑇 𝑇𝑇 𝑇𝑇 𝑇𝑇
5𝐾𝐾𝐾𝐾 + 𝐾𝐾𝑇𝑇 2 ∗ 5𝐾𝐾𝐾𝐾 ∗ 4 + 𝑇𝑇 2
𝑐𝑐(𝑘𝑘) = 𝑒𝑒 (𝑘𝑘) + 𝑒𝑒 (𝑘𝑘 − 1) + 𝑐𝑐(𝑘𝑘 − 1) − 𝑐𝑐(𝑘𝑘 − 2)
2 + 𝑇𝑇 2 + 𝑇𝑇 2 + 𝑇𝑇 2 + 𝑇𝑇

For the given parameters, we have:

𝑐𝑐(𝑘𝑘) = 11𝑒𝑒 ∗ (𝑘𝑘) + 10𝑒𝑒 ∗ (𝑘𝑘 − 1) + 1.8𝑐𝑐(𝑘𝑘 − 1) − 0.8𝑐𝑐(𝑘𝑘 − 2)

And from the block diagram

𝑒𝑒 ∗ (𝑘𝑘) = 𝑟𝑟(𝑘𝑘) − 𝑐𝑐(𝑘𝑘) = 1 − 𝑐𝑐(𝑘𝑘)

Therefore,

9
 ELE8311: Transfer Function

41 4
𝑐𝑐(𝑘𝑘) = 1.75 − 𝑐𝑐(𝑘𝑘 − 1) − 𝑐𝑐(𝑘𝑘 − 2).
60 60

Then, assuming 𝑐𝑐(−1) = 0, 𝑐𝑐(0) = 0, the subsequent response 𝑐𝑐(𝑘𝑘) of the system can be
obtained recursively as follows:

41 4
𝑐𝑐(1) = 1.75 − 𝑐𝑐(0) − 𝑐𝑐(−1) = 1.75
60 60
41 4
𝑐𝑐(2) = 1.75 − 𝑐𝑐(1) − 𝑐𝑐(0) ≈ 0.554
60 60
41 4
∴ 𝑐𝑐(𝑚𝑚) = 1.75 − 𝑐𝑐(𝑚𝑚 − 1) − 𝑐𝑐(𝑚𝑚 − 2).
60 60

3.4 The Closed-loop Transfer Function

For a closed-loop block diagram shown in Fig. 3.6, let the 𝐶𝐶(𝑧𝑧) block represent the
comparator and digital controller, and the 𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧) block represents the ADC-analog
subsystem-DAC transfer function. Then for any input 𝑅𝑅(𝑧𝑧), if the output response is 𝑌𝑌(𝑧𝑧),
similar to the analog systems, the closed-loop transfer function for the system 𝐺𝐺𝑑𝑑 (𝑧𝑧) is given
by:

Fig. 3.6: Closed Loop Transfer Function of A Digital System

𝐶𝐶(𝑧𝑧)𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧)
𝐺𝐺𝑑𝑑 (𝑧𝑧) =
1 + 𝐶𝐶(𝑧𝑧)𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧)

and the closed-loop characteristic equation is:

1 + 𝐶𝐶(𝑧𝑧)𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧) = 0.

The roots of the characteristics equation are the closed-loop system poles which can be
selected for desired time response specifications.

Note, however, that for some digital systems, the transfer function may not be completely
expressed as a function without the input. The following example demonstrates such cases.
10
 ELE8311: Transfer Function

Example 3.7

Find the Laplace transform of the analog and sampled output for the system shown in Fig.
3.7.

Fig. 3.7
Solution:

The Laplace transform of the analog variable 𝑥𝑥(𝑡𝑡) is

𝑋𝑋(𝑠𝑠) = 𝐻𝐻(𝑠𝑠)𝐺𝐺(𝑠𝑠)𝐷𝐷(𝑠𝑠)𝐸𝐸(𝑠𝑠)

But, 𝐸𝐸(𝑠𝑠) = 𝑅𝑅(𝑠𝑠) − 𝑋𝑋 ∗ (𝑠𝑠), substituting this in 𝑋𝑋(𝑠𝑠) yields

𝑋𝑋(𝑠𝑠) = 𝐻𝐻(𝑠𝑠)𝐺𝐺(𝑠𝑠)𝐷𝐷(𝑠𝑠)[𝑅𝑅(𝑠𝑠) − 𝑋𝑋 ∗ (𝑠𝑠)]

Thus, the impulse-sampled variable 𝑥𝑥 ∗ (𝑡𝑡) has the following Laplace transform

𝑋𝑋 ∗ (𝑠𝑠) = (𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻)∗ (𝑠𝑠) − (𝐻𝐻𝐻𝐻𝐻𝐻)∗ (𝑠𝑠)𝑋𝑋 ∗ (𝑠𝑠)

Notice that now, several components in the above expression cannot be separated, thus

(𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻)∗ (𝑠𝑠)
𝑋𝑋 ∗ (𝑠𝑠) =
1 + (𝐻𝐻𝐻𝐻𝐻𝐻)∗ (𝑠𝑠)

Hence, the error is given by

(𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻)∗ (𝑠𝑠)
𝐸𝐸(𝑠𝑠) = 𝑅𝑅(𝑠𝑠) −
1 + (𝐻𝐻𝐻𝐻𝐻𝐻)∗ (𝑠𝑠)

Now, from the block diagram, the analog output 𝑌𝑌(𝑠𝑠) = 𝐺𝐺(𝑠𝑠)𝐷𝐷(𝑠𝑠)𝐸𝐸(𝑠𝑠), which is

(𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻)∗ (𝑠𝑠)
𝑌𝑌(𝑠𝑠) = 𝐺𝐺(𝑠𝑠)𝐷𝐷(𝑠𝑠) �𝑅𝑅(𝑠𝑠) − �
1 + (𝐻𝐻𝐻𝐻𝐻𝐻)∗ (𝑠𝑠)

The sampled output 𝑌𝑌 ∗ (𝑠𝑠) is therefore

11
 ELE8311: Transfer Function

(𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻)∗ (𝑠𝑠)
𝑌𝑌(𝑠𝑠) = (𝐺𝐺𝐺𝐺𝐺𝐺)∗ (𝑠𝑠) − (𝐺𝐺𝐺𝐺)∗ (𝑠𝑠)
1 + (𝐻𝐻𝐻𝐻𝐻𝐻)∗ (𝑠𝑠)

The sampled output can be rewritten in z-domain via 𝑧𝑧 = 𝑒𝑒 𝑠𝑠𝑠𝑠 transformation

(𝐻𝐻𝐻𝐻𝐻𝐻𝐻𝐻)(𝑧𝑧)
𝑌𝑌(𝑧𝑧) = (𝐺𝐺𝐺𝐺𝐺𝐺)(𝑧𝑧) − (𝐺𝐺𝐺𝐺)(𝑧𝑧) .
1 + (𝐻𝐻𝐻𝐻𝐻𝐻)(𝑧𝑧)

The above equation demonstrates that there is no expression for the transfer that excludes
the input.

3.5 Disturbances in a Digital System

Those variables present in the system response which are not originally part of the system
model are considered as disturbance variables. Disturbance signals can be deterministic, such
as load torque in a position control system, or stochastic, such as sensor or actuator noise.
However, disturbances are generally analog and are inputs to the analog subsystem in a
digital control loop.

Consider the system shown in Fig. 3.8, where the disturbance input 𝐷𝐷(𝑠𝑠) has an associated
transfer function 𝐺𝐺𝑑𝑑 (𝑠𝑠). Note that since the system is linear, the reference input can be
assumed to be zero.

The Laplace transform of the impulse-sampled out is

(𝐺𝐺𝐺𝐺𝑑𝑑 𝐷𝐷)∗ (𝑠𝑠) (𝐺𝐺𝐺𝐺𝑑𝑑 𝐷𝐷)∗ (𝑠𝑠)

𝑌𝑌 ∗ (𝑠𝑠) = =
1 + (𝐺𝐺𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 )∗ (𝑠𝑠)𝐶𝐶 ∗ (𝑠𝑠) 1 + (𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 )∗ (𝑠𝑠)𝐶𝐶 ∗ (𝑠𝑠)

Fig. 3.8: Digital System with an Analog Disturbance

In z-domain, we have

(𝐺𝐺𝐺𝐺𝑑𝑑 𝐷𝐷)(𝑧𝑧)
𝑌𝑌(𝑧𝑧) = (3.2)
1 + (𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 )(𝑧𝑧)𝐶𝐶(𝑧𝑧)
12
 ELE8311: Transfer Function

Example 3.8:

Suppose the transfer function of the blocks in Fig. 3.8 are:

𝐾𝐾𝑝𝑝 1
𝐺𝐺(𝑠𝑠) = , 𝐺𝐺𝑑𝑑 (𝑠𝑠) = , 𝐶𝐶(𝑧𝑧) = 𝐾𝐾𝑐𝑐 .
𝑠𝑠 + 1 𝑠𝑠

Determine the steady-state response of the system to an impulse disturbance of strength 𝐴𝐴.

Solution

Let’s evaluate the forward gain along the disturbance path as:

𝐾𝐾𝑝𝑝 𝐴𝐴 1 1
𝐺𝐺(𝑠𝑠)𝐺𝐺𝑑𝑑 (𝑠𝑠)𝐷𝐷(𝑠𝑠) = = 𝐾𝐾𝑝𝑝 𝐴𝐴 � − �
𝑠𝑠(𝑠𝑠 + 1) 𝑠𝑠 𝑠𝑠 + 1

Now, the z-transform corresponding to impulse response sequence is

𝑧𝑧 𝑧𝑧
(𝐺𝐺𝐺𝐺𝑑𝑑 𝐷𝐷)(𝑧𝑧) = 𝐾𝐾𝑝𝑝 𝐴𝐴 � − �
𝑧𝑧 − 1 𝑧𝑧 − 𝑒𝑒 𝑇𝑇

To take care of the ZOH sampler, recall that

𝐺𝐺(𝑠𝑠)
𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧) = (1 − 𝑧𝑧 −1 )𝑍𝑍 � �,
𝑠𝑠

Therefore, we have

1 − 𝑒𝑒 −𝑇𝑇
𝐺𝐺𝑍𝑍𝑍𝑍𝑍𝑍 (𝑧𝑧) = 𝐾𝐾𝑝𝑝
𝑧𝑧 − 𝑒𝑒 −𝑇𝑇

From the disturbance response in equation (3.2) we have

𝑧𝑧 𝑧𝑧
𝐾𝐾𝑝𝑝 𝐴𝐴 �
𝑧𝑧 − 1 𝑧𝑧 − 𝑒𝑒 𝑇𝑇 �
−
𝑌𝑌(𝑧𝑧) =
1 − 𝑒𝑒 −𝑇𝑇
1 + 𝐾𝐾𝑐𝑐 �𝐾𝐾𝑝𝑝 �
𝑧𝑧 − 𝑒𝑒 −𝑇𝑇

Using the final value theorem, the steady-state response is

𝐾𝐾𝑝𝑝 𝐴𝐴
𝑦𝑦(∞) = lim(𝑧𝑧 − 1)𝑌𝑌(𝑧𝑧) = .
z=1 1 + 𝐾𝐾𝑐𝑐 𝐾𝐾𝑝𝑝

Notice that similar to analog systems, increasing the controller gain reduces the error due to
the disturbance.

13