EENG6760 Digital Signal Processing and Control

EENG6760 CONTROL LECTURE 14

Design By Emulation

Dr Xinggang Yan

Senior Lecturer
School of Engineering
University of Kent

Room 105, Jennison Building

Email: x.yan@kent.ac.uk
  Design By Emulation:
Tustin’s Rule
Pole-Zero Mapping

 W-plane Design

UNIVERSITY OF KENT Slide 2

 Design by Emulation

• If we introduce sampling into continuous operations, it needs to transform continuous transfer

functions to discrete transfer functions.

• For a given continuous transfer function 𝐺(𝑠), find the best discontinuous transfer function
𝐺(𝑧) such that the signal produced by passing an input signal through 𝐺(𝑧) is the same as that
produced by passing the same signal through 𝐺(𝑠). (Discrete equivalent)

Convert
𝑮 𝒔 𝐆(𝐳)
 DISCRETE APPROXIMATIONS

Could use substitutions

z esT , s  1 ln z
T
but awkward, complicated.

Instead we use simpler, “approximate” transformations

which usually give adequate descriptions.

Two main methods:

(1) numerical integration
(2) pole-zero mapping
 Numerical Integration
𝒀 𝒔 𝒂
Given a continuous transfer function: =𝑮 𝒔 =
𝑼 𝒔 𝒔+𝒂
This represents the ODE
y (t ) ay (t )au(t )
with zero initial conditions.

For a digital representation, we need to solve the ODE at the sample instants:
dy
 au(t )  ay (t )
dt

kT kT
 dy 
 

dt  

au(t )  ay (t ) dt
 dt 
(k 1)T (k 1)T
k  1, 2, 3 ...
𝒌𝑻

𝒚 𝒌𝑻 = 𝒚 𝒌 − 𝟏 𝑻 + න 𝒂𝒖 𝒕 − 𝒂𝒚 𝒕 𝒅𝒕
𝒌−𝟏 𝑻
Numerical Integration (Continued)
Trapezium
Rule
𝒌𝑻

𝒚 𝒌𝑻 = 𝒚 𝒌 − 𝟏 𝑻 + 𝒂 න 𝒖 𝒕 − 𝒚 𝒕 𝒅𝒕
𝒌−𝟏 𝑻
approx
Str. f(t)
line kT




f (t )dt 
T
2
f (kT )f  (k 1)T 
(k-1)T kT t (k 1)T

y (kT )  y  (k 1)T  
aT
2

u(kT ) y (kT )u (k 1)T  y (k 1)T  
𝒂𝑻
Z-transforming 𝒀 𝒛 = 𝒀 𝒛 𝒛−𝟏 + 𝑼 𝒛 − 𝒀 𝒛 + 𝑼 𝒛 𝒛−𝟏 − 𝒀 𝒛 𝒛−𝟏
𝟐


Y (z)   z 1 

aT
2 
1 z 1  


Y ( z ) 
aT
2
1 z 1U (z)
 
Tustin’s Rule

Y (z)   z 1 


aT
2
1 z 1 Y (z)  aT2 1 z1U(z)
Multiplying across by z
 aT  aT 
Y (z )

a
zY (z)  1 (z 1)Y (z)   z 1U(z)
2 U (z ) 2 (z 1)  a
  2
T (z 1)
𝒂
𝑮 𝒛 =
𝟐𝒛−𝟏
+𝒂
𝑻𝒛 + 𝟏
𝒂 𝒂
Result: For a given 𝑮 𝒔 = , its discrete equivalent is: 𝑮 𝒛 = 𝟐 𝒛−𝟏
𝒔+𝒂 +𝒂
𝑻 𝒛+𝟏
𝟐 𝒛−𝟏
𝒔=
𝑻 𝒛+𝟏
𝑮(𝒔) 𝐆(𝐳)
TUSTIN’S RULE
 Example- TUSTIN’s Rule
𝒔−𝟐
For a given continuous transfer function 𝑮 𝒔 = , find its discrete equivalent 𝐺 𝑧 with
𝒔+𝟏
sampling period 𝑇 = 0.2 using Tustin’s Rule.

𝑧−1
10 −2
𝑮 𝒛 =𝑮 𝒔 ∣ = 𝑧 + 1
𝟐 𝒛−𝟏
𝒔=𝟎.𝟐 𝒛+𝟏 𝑧−1
10 +1
𝑧+1

8 𝑧 − 1.5 𝑧 − 1.5
= = 0.7273
11 𝑧 − 9 𝑧 − 0.8182
11
𝟐 𝒛−𝟏
𝒔=
𝑻 𝒛+𝟏
𝑮(𝒔) 𝐆(𝐳)
TUSTIN’S RULE
 Pole-Zero Mapping

Every pole and zero of G(s) in the s-plane has its equivalent position in
the z-plane through the mapping

𝒛 = 𝒆𝒔𝑻
It’s reasonable to form G(z) from G(s) by mapping the poles and zeros in
the s-plane to poles and zeros in the z-plane.

G(s)  G(z)
s 0 z 1
 Pole-Zero Mapping Rules 𝒛 = 𝒆𝒔𝑻

1. Map zeros according to 𝒛 = 𝒆𝒔𝑻

If 𝑮(𝒔) has a zero at 𝒔 = 𝒂, then, 𝑮(𝒛) should have a zero at 𝒛 = 𝒆𝒂𝒕

2. Map poles according to 𝒛 = 𝒆𝒔𝑻

If 𝑮(𝒔) has a pole at 𝒔 = 𝒂, then, 𝑮(𝒛) should have a pole at 𝒛 = 𝒆𝒂𝒕

3. Match the gain:

𝑮 𝒔 𝒔=𝟎 = 𝐆 𝐳 𝐳=𝟏
 𝒛 = 𝒆𝒔𝑻
Pole-Zero Mapping Example
𝒔+𝟎.𝟓
Example. Consider 𝑮 𝒔 = Find its equivalent digital transfer function G(z) with
𝒔+𝟐
sampling period T=0.1.

Step 1: The zero of 𝐺(𝑠) is 𝑠0 = −0.5. So 𝑧0 = e−0.05 = 0.95 should be the zero of G(z).

Step 2: The pole of 𝐺(𝑠) is: 𝑠𝑝 = −2. So 𝑧𝑝 = e−2×0.1 = 0.82 should be the pole of G(z).

𝑧 − 0.95
G 𝑧 =𝐾
𝑧 − 0.82

Step 3: Match the gain 𝑮 𝒔 𝒔=𝟎 = 𝐆 𝐳 𝐳=𝟏

0 + 0.5 1 − 0.95
=𝐾 , 𝐾 = 0.9
0+2 1 − 0.82
𝒛 − 𝟎. 𝟗𝟓
𝐆 𝒛 = 𝟎. 𝟗
𝒛 − 𝟎. 𝟖𝟐
 The w-plane Design
Step 1
ZOH F s 
Form a plant model F(z) from

Step 2
Transform F(z) to w-plane. “Reverse emulation”. The commonest method is to use Tustin’s rule, the bilinear
transformation.
An approximation: 𝑇
2 1−𝑧 −1 𝑤+1
𝑧= 2
𝑤= 𝑇
Then, 𝑇 1 + 𝑧 −1 1− 2𝑊
𝑻
𝒘+𝟏
𝒛= 𝟐
𝑻
𝟏− 𝟐𝑾
F(z) F(w)

Step 3
Use our usual s-plane design techniques in the w-plane. eg (Bode Diagrams) 𝑤 = 𝑗𝑣
Step 4
Inverse transform the resulting controller using w  2
 z  1
T  z  1
Design Example (W Plane)
i/p D z ZOH G s  o/p
+
1
G s  
_ -
T=0.05secs
_
s2
1  e  sT 
G z  Z ZOH  G  s   Z 
 s3 
 1 
  
1  z 1 Z  3 
s 
from sheet

T 2 z  z  1
 
1 z 1
2  z  13


T2  z  1  0.00125
 z  1
2  z  1  z  1
2 2

T
w 1
Transform into w-plane using z  2
T
1 w
2
1  0.025w
G w  
w2

Use frequency domain analysis in the w-plane by letting w  jv  cf s  jw 

 𝟏−𝟎.𝟎𝟐𝟓𝒘
Bode Plot for D(w)=1, 𝑮 𝒘 =
𝒘𝟐

The system is not stable if the compensator D(w) is chosen as D(w)=1.

Then we can use the design method for continuous systems to design D(w)
 600 w  2  1  0.025w
Bode Plots D w   , G w  
w  60  w2

Inverse transform from

w 
2  z  1
T  z  1

This gives a controller

D z  252
 z  0.9 
 z  0.2 

You will not be asked any question for w-plane design in the exam