IV Design of Discrete Time

Control System by
Conventional Methods
Topics to be covered
1. Introduction

2. Mapping between the s plane and z plane

3. Stability analysis

4. Transient and steady state response

5. Design based on root locus method

6. Design based on frequency response method

7. Analytical design method

IV.1. Introduction
Note: Three different design methods are introduced: 1) root locus based method, 2) frequency
response method in w plane, 3) analytical method.

Many conventional continuous-time system design method can be applied to discrete-time

systems.

1
 IV.2. Mapping between the s plane
and z plane
The complex variables z and s are related by the equation
z  e Ts
s    jw
z  e Ts  e T   jw   e T e jTw  e T e j Tw2k 
2
Note: frequencies differ in integral multiples of the sampling frequency , are mapped into the
T
same location in the z plane. jw axis in the s plane corresponds to z  1 . The interior of the unit
circle corresponds to the left half of the s plane. The exterior of the unit corresponds to the right
half of the plane.

Note:
1) z  wT the angle of z varies from   to  as w varies from   to  . as point moves from
1 1 2
 j ws to j ws ( ws  ) z  wT varies from   to  This is the primary strip.
2 2 T
2) Each other strip with range of ws will trace the z plane in one circle.
3) Mapping between s plane and z plane is not unique

2
 Mapping between the primary strip and the unit circle

a) constant attenuation line  is constant maps in to the circle of radius z  e T

3
Setting time relate to the region on the left of  1 , which corresponds to the interior of the circle
with radius z  e T 1
b) Constant frequency loci:
Constant frequency locus w  w1 , in the s plane maps into the radial line of constant
angle Tw1 in the z plane.

c) Constant damp ratio line maps into a spiral in the z plane.

4
 s  wn  jwn 1   2  wn  jwd , where wd  wn 1   2

2 wd w
  j 2 d
Twn  jTwn 1 2 1 2 ws ws
ze e
Ts
e

2 wd

1 2 ws wd
Thus z  e and z  2
ws

Thus the magnitude of z decreases and the angle of z increases linearly as wd increases,
the locus in the z plane becomes a logarithmic spiral.

The region bounded by constant frequency lines and constant attenuation lines and the mapping.

Constant  lines

5
 Diagram of the orthogonality of the constant  and wn mapping to the z plane.

Example 4.1.: specify the region in the z plane that corresponds to a desirable region in s plane.

6
 IV.3. Stability analysis
Closed loop pulse transfer function system:
C ( z) Gz 
 4.1
R z  1  GH  z 
The stability of 4.2 may be determined from the location of the closed loop poles in the z plane,
or the roots of the characteristic equation.
P z   1  GH z   0 4.2
Note:
1) For the stable system, all closed loop poles must lie in the unit circle in the z plane.
2) If a simple pole lies at z  1 , then the system becomes critically stable. Also if a single
pair of conjugate complex poles lies on the unit circle in the z plane, the system is
critically stable. Any multiple closed loop pole on the unit circle makes the system
unstable
3) Closed loop zeros do not affect the absolute stability and therefore may be located
anywhere in the z plane.
Method for testing absolute stability:
1) Schur-cohn stability test
2) Jury stability test
3) Test based on the bilinear transformation coupled with the Routh stability criterion.
4) Liapunov stability analysis
Note: Both Schur-cohn stability test and Jury stability test may be applied to polynomial
equations with real or complex coefficients. When the coefficients are real, Jury test are much
simpler than schur-cohn test.
The Jury Stability Test:
P z   a0 z n  a1 z n 1   an 1 z  an where a0  0
The table is given as follows:

7
 an a n 1k
bk  , k  0,1,2,  , n  1
a0 a k 1

bn 1 bn 2k
ck  , k  0,1,2, , n  2
b0 bk 1


p3 p2k
qk  k  0,1,2
p0 pk 1

Stability criterion by the Jury Test: A system with the characteristic equation P z   0
rewritten as P z   a0 z n  a1 z n 1   an 1 z  an where a0  0
is stable if the following conditions are all satisfied:

1) a n  a 0
2) P z  z 1  0
 0 for n even
3) P z  z  1 
  0 for n odd
4) bn 1  b0
cn 2  c0

q 2  q0

Example 4.2: determine the stability of following Ch equation

P z   z 2  z  0.6321  0

8
Stability analysis by use of the bilinear transformation and routh stability criterion

Bilinear transformation defined by :

w 1 z 1
z w
w 1 z 1

bilinear transformation maps the inside of the unit circle into the left half of the w plane.

assume w    jw

if z  1 
  jw  1
1
  jw  12
1
  12  w 2
 1  4  0    0
  jw  1   jw  12   12  w 2

Note: once we transform P z   0 into Q w   0 , it is possible to apply the Routh stability

criterion in the same manner.

Example 4.3: determine the stability of following Ch equation using bilinear transformation

P z   z 2  z  0.6321  0

9
 IV.4. Transient and steady state
response

Transient response specifications:

1) Delay time t d the delay time is the time required for the response to reach half the final
value the very first time.
2) Rise time t r The rise time is the time required for the response to rise from 10% to 90%,
or 5% to 95%, or 0% to 100% of its final value.
3) Peak time t p , the peak time is the time required for the response to reach the first peak of
the overshoot.
4) Maximum overshoot M p The maximum overshoot is the maximum peak value of the
response curve measured from unity.
C t p   C  
maximum percent overshoot =   100%
C  
5) Settling time t s the settling time is the time required for the response curve to reach and
stay within a range about the final value of a size specified as an absolute percentage of
the final value, usually 2%.

Note: The transient response of a discrete system to the Kronecker delta input, step input, ramp
input, and so on, can be obtained easily by use of MATLAB.

10
Steady-State Error Analysis:

recall

The loop transfer function is written in general form:

M
k  ( s  zi )
Gc ( s )G ( s )  i 1
Q
s  ( s  pk )
N
k 1
N
The term s in the dominator represents a pole of multiplicity N at the origin. The number
of integrations is often indicated by labeling a system with a type number that is simply is
equal to N.

Note: as the type number increases, the steady state error reduces, however, stability problem
aggravates. Concepts of static error constants can be extended to the discrete-time control
system.

The discrete system loop transfer function is written in general form:

C( z) B z  k  ( z  zi )
  i 1
R z  z  1 Az  z  1N 
N Q
( z  pk )
k 1

B z 
where contains neither a pole or a zero at z=1. Then the system will be classified as
Az 
type N system.

 E *  s 
 Rs +  E s   C (s )
1  e Ts
G P (s)
s
  T 
R(z)
- C(z)

bt 

H(s)

Consider above system: et   r t   bt 

from final value theorem, we have

lim ekT   lim1  z 1 E z  4.3
k  z 1

11
for above system configuration :

define
 G p s 
G z   1  z 1 Z   and
 s 

 G s H s 
GH z   1  z 1 Z  p 
 s 
then we have:

C z  G z 

R z  1  GH z 
and

E  z   R z   B z   R z   GH  z E  z 
4.4
 E z   R z 
1
1  GH  z 

substitute into 4.3, we have

lim ekT   ess  lim1  z 1 E z   lim1  z 1  R z 

1
k  z 1 z 1 1  GH z 

1
Case 1: Unit step input: r (t )  u(t )  R( z ) 
1  z 1

ess  lim1  z 1 E z   lim1  z 1 

1 1 1
 lim
z 1 z 1 1  GH z  1  z  z1 1  GH z 
1

We define the static position error constant K p  lim GH z 

z 1

thus

1
ess  4.5
1 Kp

Note: The steady state error in response to a unit step input becomes zero if K p   , which
requires that GH z  has at least one pole at z=1.
Tz 1
Case 2: ramp step input: r (t )  tu(t )  R( z ) 
1  z  1 2

12
 Tz 1
ess  lim1  z 1 E z   lim1  z 1 
1 T
 lim
z 1 z 1 1  GH z  1  z 
1 2 z 1 1  z GH  z 
1

We define the static velocity error constant K v  lim

1  z GH z 
1

z 1 T
thus

1
ess  4.6
Kv

Note: the steady state error in response to a ram step input becomes zero if K v   , which
requires that GH z  has at least double poles at z=1.

T 2 1  z 1  z 1
1
1
Case 3: Unit acceleration input: r (t )  t 2 u(t )  R( z ) 
21  z 1 
3
2
T 2 1  z 1  z 1
1

ess  lim1  z E z   lim1  z 

1 1 1 T2
 lim
z 1 z 1 1  GH z  21  z 1 3 z 1
1  z 1 2 GH z 
We define the static acceleration error constant K a  lim
1  z  GH z 
1 2

z 1 T2
thus
1
ess  4.7
Ka
Note: The steady state error in response to a unit acceleration input becomes zero if K a   ,
which requires that GH z  has at least three poles at z=1.

Summary:

Table of steady state errors

Type number Step input Ramp input Acceleration input
0 A infinity infinity
ess 
1 K p
1 0 A infinity
ess 
KV
2 0 0 A
ess 
Ka
3 or more 0 0 0

13
Static error constants for typical closed-loop configurations of discrete time control system

14
Response to disturbance:

 N  z 

 Rz 
+ C z 
 G D z  G z 
-

(a)

+ C z 
G z 
 N  z 
-

GD z 

(b)
assume the reference input is zero, or R z   0
C z  G z 
 , which can be obtained from the equivalent model in b.
N z  1  G z G D z 

G z 
E  z   R z   C  z   C  z    N z 
1  G  z G D  z 

z 1
  z 1

e ss  lim 1  z 1 E  z   lim 1  z 1 1  GGz Gz  z  N z 
D

Note: To reduce the disturbance impact, we can increase the gain G z G D z 

15
 IV.5. Design based on root locus
method
Note: The root locus method developed for continuous-time systems can be extended to discrete-
time systems without modifications, except that the stability boundary is changed from the jw
axis in the s plane to the unit circle in the z plane.

Angle and magnitude conditions

characteristic equations may have either of following forms:

1  G z H z   0 and 1  GH z   0 which can be written as :

1  F z   0

Angle condition: F z   1802k  1 k  0,1,2 

Magnitude condition: F z   1

General procedure for constructing root loci.

1) Obtain the characteristic equation

1  F z   0 then rearrange the equation:

K z  z1 z  z 2  z  z m 
1 0
z  p1 z  p2  z  pn 
2) Find the starting points and terminating points of the root loci. As K increases from 0 to
infinite, a root locus starts from an open loop pole and terminates at a finite open loop
zero. Locate the poles pi and zeros zi on the z-plane with selected symbols. By
conversion, we use ‘x’ to denote poles and ‘o’ to denote zeros.
Remarks:
Loci begins at the poles and ends at the zeros. The number of separate loci is equal to the
number of poles.
Root loci must be symmetrical with respect to the horizontal real axis because the
complex roots must appear as pairs of complex conjugate roots.

3) Determine the root loci on the real axis. Root loci on the real axis are determined by
open-loop poles and zeros lying on it. The root locus on the real axis always lies in a
section of the real axis to the left of an odd number of poles and zeros.
4) Determines the asymptotes of the root loci.

16
The loci proceed to the zeros at infinity along asymptotes centered at  A and with angles  A . The
number of zeros M is less than the number of poles n by N=n-M. the N sections of loci must end
at zeros at infinity. These sections of loci proceed to the zeros at infinity along asymptotes as K
approaches infinity. These linear asymptotes are centered at a point on the real axis given by

n M

 poles of P(z )   zeros of P(z )   ( pi )  i (zi )

A  i1 1
n M n M
The angle of the asymptotes with respect to the real axis is

2k  1
A  1800 , k=0,1,2…(n-M-1),
nM

5) Determine the breakaway point on the real axis (if any). In general, the tangent to the loci
at the breakaway point are equally spaced over 3600 .
We may evaluate
dK
0
dz

6) Determine the angle of departure of the locus from a pole and the angle of arrival of the
locus at a zero, using the phase angle criterion.
The angle of locus departure from a pole is the difference between the net angle due to all
other poles and zeros and the criterion of  1800 (2k  1) , and similarly for the locus angle
of arrival at a zero.

7) Find the points where the root loci cross the unit circle

8) Any point on the root loci is a possible closed-loop pole.

17
Cancellation of the poles of G z  with zeros of H z 
if F z   G z H z  and the denominator of G z  and numerator of H z  have common
factors then the corresponding open loop poles and zeros will cancel each other, reducing
the degree of the characteristic equation.

zc za
Example 4.4 H z  in the feedback loop : H z   and G z  
zb z  c z  d 
zc za za
G ( z ) H z    Pole z   c  has been cancelled.
z  b z  c z  d  z  d z  b 

zc za
Example 4.5 H z  in the feed forward loop : H z   and G z  
zb z  c z  d 
zc za za
G ( z ) H z    Pole z   c  has been cancelled.
z  b z  c z  d  z  d z  b 

1
Example 4.6 System has loop gain G  z   Tz , draw the root locus.
1  z 1  
1
1  KG z   1 
KTz
1  z 1   0
Note: sampling Time will affect the K value. Sometimes, it will affect the pole and zero locations
as well.

Effects of sampling Period T on transient response characteristics

Note: Increase the sampling period T will make the system less stable and eventually will make
it unstable.

18
The damp ratio  of the closed loop pole can be analytically determined from the location of the
closed loop pole in the z plane.
s  wn  jwn 1   2  wn  jwd , where wd  wn 1   2

2 wd w
  j 2 d
Twn  jTwn 1 2 1 2 ws ws
ze e
Ts
e

2 wd

1 2 ws wd
Thus z  e and z  2
ws

19
Example 4.7 (B-4-8) Consider the digital control system shown in figure, plot the root loci.
Determine the critical value of gain K for stability. The sampling period is 0.1 sec what value of
gain K will yield a damping ratio  of the closed loop poles equal to 0.5? with gain K set to
yield   0.5 , determine the damped natural frequency w d and the number of samples per cycle
of damped sinusoidal oscillation.

+   K  z  1 C z 
z  1z  0. 6065 
 R z 
-

20
 IV.6. Design based on frequency
response method
Note: To obtain the frequency response of G z  we need to only substitute e jwT for z. The
function G e jwT  is commonly called the sinusoidal pulse transfer function. It is
periodic, with the period equal to T.
1
Example 4.8 (B-4-8) Consider system defined by G  z   Tz with input A sin kwT . Obtain
1  z 1 
the steady state output.

G e jwT  
T T T
 
e jwT
 1 coswT   j sin wT   1 coswT   1  j sin wT 

G e jwT  
T
coswT   1 2
 sin wT 
2

 sin wT  
G e jwT     tan 1  
 1  coswT  
Thus the steady state output:

  sin wT   
x ss kT   A
T
sin  kwT  tan 1   
coswT   1 2
 sin wT 
2
   1  cos  wT  

Bilinear transformation and the w plane

Note: z transform maps the primary and complementary strips of the left half of the
s plane into the unit circle in the z plane, conventional frequency response methods,
which deal with the entire left half plane, do not apply to the z plane.

w transformation will solve the problem:

21
 T
1 w
z 2 , where T is the sampling period.
T
1 w
2

2 z 1
the inverse transform: w 
T z 1

Note: the primary strip of the left half of the s plane is first mapped into the inside
of the unit circle in the z plane and then mapped into the entire left half of the w
plane.
ws
s varies from 0  j along jw axis in the s plane
2
z varies from 1  1 , along the unit circle
w varies from 0   along the imaginary axis in the w plane.

ws w
s plane primary strips:  w s
2 2
w plane frequency 0  v  

Although w plane resembles the s plane geometrically, the frequency axis in the w
plane is distorted. The fictions frequency v and the actual frequency w are related
as follows:

2 z 1 2 e jwT  1 2 wT 2 wT
jv    j tan  v  tan
T z  1 z e jwT T e  1 T
jwT
2 T 2

22
 T
Following figure shows the relationship between v and w
2

b0 z m  b1 z m1    bm
G z   , where m  n
z n  a1 z n 1    a n

T
w 1
take the transformation z  2
T
1 w
2
then,
 0 w m  1 w m1     n
G w  
 n  1 w n 1     n

Nyquist stability and bode diagram can be applied.

Note: high frequency asymptote of the logarithmic magnitude for G  jw  and G  jv 

may be different.

Advantages of bode diagram approach:

1) low frequency asymptote of the magnitude curve is indicative of the static
error constants K p K v or K a
2) Specifications of the transient response can be translated into those of the
frequency response in terms of phase margin, gain margin, bandwidth, and
so forth.
3) Design in simpler manner.

23
Review phase lead, phase lag, phase lag-lead:
1) Phase lead is commonly used for improving stability margins. The phase
lead compensation increases the system bandwidth. Thus the system has
faster speed to respond. It may be subjected to high-frequency noise due to
its increased high frequency gains.
2) Phase lag compensation reduces the system gain at higher frequencies
without reducing the system gain at lower frequencies. The system
bandwidth is reduced and the system has slower response. Steady state
accuracy can be improved. High frequency noise can be attenuated.
3) Sometime, phase lag compensator is cascaded with a phase lead
compensator. Low frequency gain can be increased, the bandwidth can be
maintained. PID controller as an example. (PD as lead and PI as lag)
Design Procedure in the w plane.
T
1 w
z 2
T
1 w
1) First obtain G z  , then transform G z  to G w  through 2 , (T
should be properly chosen) rule of thumb is to sample at the frequency 10
times that of the bandwidth of the closed loop system)
2) Substitute w  jv and plot the bode diagram for G  jv 
3) Read the bode diagram static error constants, the phase margin, the gain
margin.
4) Assume the low frequency gain of the controller GD w  is unity, determine
the system gain by satisfying the requirement for a given static error
constant. Using conventional design technique to design GD w 
5) Transform the GD w to GD z  through the bilinear transformation given by
2 z 1
w
T z 1
6) Realization the pulse transfer function GD z  by computational algorithm.

24
Example 4.8 (B-4-15) Using the bode diagram approach in the w plane, design a digital
controller for the system shown in figure. The design specifications are that the phase margin be
50 degree, the gain margin be at least 10 dB, and the static velocity error constant K v be
20 sec-1 . The sampling period is assumed to be 0.1 sec. after the controller is designed,
calculate the number of samples per cycle of damped sinusoidal oscillation.

 Rs + 1  e  Ts  C (s )
  K
 G D  s
s s s  0 .5
-

25
 IV.7. Analytical design method
Design of digital controller for minimum settling time with zero steady state error

Define the z transform of the plant that is preceded by the zero-order hold as G z  , or
1  e Ts 
G z   Z  G P s 
 s 
The open loop transfer function becomes G D z G z  (feed forward)
Define the closed loop pulse transfer function

C z  GD z G z 
  F z  4.8
R z  1  GD z G z 

Since it is required that the system exhibit a finite settling time with zero steady state
error, the system must exhibit a finite impulse response. Hence, F z  must be of the
following form:

F z   a0  a1 z 1    a N z  N , where N  n , n is the order of the system.

Note: F z  must not contain any terms with positive power in z.

solve 4.8 we will have

F z 
G D z   4.9
G z 1  F z 

Note: Both F z  and GD z  must be physically realizable

1) The order of the numerator of GD z  must be equal to or lower than the order of the
denominator. (Otherwise, the controller requires future input data to generate current
input. )
2) If the plant G p s  involves a transportation lag e  Ls , then the designed closed loop
system must involve at least the same magnitude of the transportation lag. (Otherwise,
the close loop system will respond before the input is given)
3) If G z  is expanded into a series in z 1 , the lowest -power term of the series
expansion of F z  in z 1 must be at least as large as that of G z  .
4) In addition to the physical realizability conditions, attention should be paid to the
stability. We must avoid cancelling an unstable pole of the plant by a zero of the
digital controller. Similarly, the digital controller pulse transfer function should not
involve unstable poles to cancel plant zeros that lie outside the unit circle.

26
 G1 z 
Assume G z  involves an unstable pole G z   , where a  1
za
G1 z 
G D z 
C z  G D z G z  z  a  F z 
 
R z  1  GD z G z  1  G z  G1 z 
, thus 1  F z  must have z  a as zero.
za
D

za
 1  F z  
z  a  G1 z GD z 

Note: since GD z  should not cancel unstable poles of G z  , all unstable poles must be included
in 1  F z  as zeros.
Zeros of G z  lie inside the unit circle may be cancelled with poles of GD z  . However, all zeros
of G z  that lie on or outside the unit circle must be included in F z  as zeros.

Design process:

E z   R z   C z   R z 1  F z  4.10
for unit step input R z  
1
1  z 1
Tz 1
for unit ramp input R z  
1  z 1 2
T 2 z 1 1  z 1 
for unit acceleration input: R z  
21  z 1 
3

P z 
In general, R z   , substitute into 4.10
1  z 
1 q 1

E z   R z   C z   R z 1  F z 
P z 
 1  F z 
1  z 1 q1
to ensure that the system reaches steady state in a finite number of sampling periods and
maintains zero steady state error, E z  must be a polynomial in z 1 with a finite number of
terms., we chose the function 1  F z  to be the form
1  F z   1  z 1 q 1 N z  , where N z  is a polynomial in z 1 with a finite number of terms.
then,
E z   P z N z  , which is a polynomial in z 1 with a finite number of terms.
once we have F z  , then substitute into 4.9 to get GD z 

F z  F z 
G D z   
G z 1  F z  G z 1  z 1 q 1 N z 

27
For a stable plant G P s  , the condition that the output not exhibit intersampling ripples after the
settling time is reached my be written as follows:

C (t  nT )  cons tan t , for step input


C (t  nT )  cons tan t , for ramp input


C (t  nT )  cons tan t , for acceleration input

Example 4.9 (B-4-18) Consider the control system shown in figure. Design a digital controller
GD z  such that the system output will exhibit a deadbeat response to a unit step input (that is,
the settling time will be the minimum possible and the steady state error will be zero; also the
system output will not exhibit intersampling ripples after the settling time is reached) the
sampling period T is assumed to be 1 sec.

 Rs + 1  e  Ts 1  C (s )
 G D  s s2
s
-

28