Discrete Approximation

of Continuous Systems
CSE 421 Digital Control
Lecture 3

1
Introduction
• One approach for Digital Controller Design is an
indirect one. The procedure is as follows:
1. Design a continuous-time controller H(s)
2. Find a discrete system H(z) approximating the
behavior of H(s) using numerical integration
3. Realize H(z) on a computer.

2
Example
• Given the first-order system
U (s) a
H (s)   ,
E (s) s  a
• Cross-multiplying and using inverse Laplace
transform we get

u (t )  au (t )  ae(t )

• We integrate to obtain u(t),

t
u (t )    au ( )  ae( )d
0

3
• In discrete time, we can express this integral in two
intervals: [0, kT−T] and [kT−T, kT]
kT T kT
u (kT )    au ( )  ae( )d    au( )  ae( )d
0 kT T
kT
uk  uk 1    au( )  ae( )d
kT T
       
area over kT T t  kT

• This area may be approximated in three ways:

• Forward rectangular
• Backward rectangular
• Trapezoidal
4
 a
Forward & backward rules sa

kT
u k  u k 1    au ( )  ae( )d
kT T

 u k 1  [au k 1  aek 1 ]T
 (1  aT )uk 1  aTe k 1
U ( z) aT a
 
E ( z ) z  (1  aT )  z  1 
 a
 T 

u k  uk 1  [au k  aek ]T
U ( z) aTz a
 
E ( z ) (1  aT ) z  1  z  1 
 a
 Tz  5
Trapezoidal rule
kT
u k  u k 1    au ( )  ae( )d
kT T

T
 uk 1  [ au k 1  aek 1  auk  aek ]
2
 aT   aT 
 z   
U ( z)  2   2  a
 
E ( z )  aT   aT   2 z  1 
1   z  1    a
 2   2   T z 1

6
Mapping between s and z plane
• Comparing the original H ( s)  a ,
sa
with the three H(z) we obtained, the following
approximations for s are obtained.

• This may be viewed as mapping from the s-

plane to the z-plane.
7
Comparison of Integration
Rules
• It is instructive to see where in the z-plane these
approximations transform the left-half s-plane
(the region of stable s-plane poles).
• To this end, solve for the inverse mapping, then
substitute for s = jω.

8
Comparison of Integration
Rules
• The mapping of the left-half s-plane are as shown below.
• With forward rectangular rule, there is a possibility that a
stable continuous system results in an unstable discrete
approximation.
• With trapezoidal rule, stable continuous system H(s) will
result in a stable discrete system H(z) and vice versa.
• Among these three methods, only trapezoidal rule is used
in MATLAB. It is called “Tustin” approximation.

9
Example 1
• With MATLAB, find the discrete approximation of the
following continuous system. Use trapezoidal rule
approximation and a sample period T = 0.05.
10
H ( s) 
s  10
numc = [0 10]; % numerator & denominator
denc = [1 10];
sysc = tf(numc,denc); % Create continuous LTI system
T = 0.05; % Sampling period
sysd = c2d(sysc,T,’tustin’) % Discretize using
trapezoidal rule; Rectangular rules are not supported
0.2 z  0.2
• Transfer function: H ( z) 
z  0.6 10
Example 1, continued
• Sometimes you need to extract the numerator and
denominator from the LTI description of the
system.
• Using the result obtained above, here’s how to use
tfdata:

[numd,dend] = tfdata(sysd,’v’)
numd = 0.2 0.2
dend = 1.0 -0.6

• NOTE: Don’t forget the ’v’ or you’ll be in for grief!

11
Example 2
• The continuous-time PID controller has the transfer
function
U ( s)  1 
 K c 1   Td s 
  E (s)  Ti s 
 
• Where the gain Kc, the integral time Ti, and the
derivative time Td are the controller parameters.
• Derive the equivalent discrete-time controller
transfer function using the backward Euler
approximation.
12
Answer z 1
• First, we use the substitution s  Tz
 Tz z 1   (T / Ti ) z z 1 
GPID ( z )  K c 1   Td   K c 1   (Td / T ) 
 Ti ( z  1) Tz   ( z  1) z 
 z ( z  1)  (T / Ti ) z 2  (Td / T )( z  1) 2 
 K c  
 z ( z  1) 
• The discrete PID controller is given by the following
transfer function:

U ( z ) K 0  K1 z 1  K 2 z 2
 1
,
E( z) 1 z
 T Td   T   Td 
K 0  K c 1   , K1  K c   1  2 d  2
, K  K c 
 Ti T   T  T  13
Implementation of Difference
Equations in Real Time
• It is time to consider how to construct computer
programs that will realize the difference equations we
obtain.
• Consider the second-order discrete transfer function:
U ( z ) b0  b1 z 1  b2 z 2
H ( z)  
E ( z ) 1  a1 z 1  a2 z  2
• with corresponding difference equation

uk  a1uk 1  a2uk  2  b0 ek  b1ek 1  b2 ek  2

• We assume that there is an interrupt occuring every T
seconds. When an interrupt occurs, variable FLAG is
set to one.
14
Implementation of Difference
Equations in Real Time
10 while (FLAG == 0) wait;

20 e = adc_in(); % Read error ek

30 u = a1*u1 + a2*u2 + b0*e + b1*e1 + b2*e2; % compute uk

40 dac_out(u); % Write output u(k) to D/A converter.

50 e2 = e1; % Propagate variables backwards for next sample

60 e1 = e;

70 u2 = u1;

80 u1 = u;

90 Goto 10; % Execute continuously. 15