Chapter 1

Digital system representation and

analysis
 Contents
• Sampled data systems
– Need for sampling
– Types of sampled data systems
• Digital system representation and analysis
• Stability anaslsys in digital systems
 Feedback control
• Closed loop Control system
– Input to a controlled system/plant is determined as a
function of the system/plant output
– Some mechanism to sense the plant output
– Controller to computer the desired input to the system so
as to achieve a desired goal
• Disturbance rejection
• Steady state error
• Transient performance
• Sensitivity to parameter variation
– Actuator drives the plant taking input from controller
 Sampled data systems
• Digital control system always involves sampled
data systems
• Sampling is the process of converting a
continuous analog signal to a discrete signal

x(t) t=KT x*(t)=x(KT)

 Sampled data systems
• Sampled signals
– Easier to transmit
– Easily regenerated
– Easy for coding
– Can be easily multiplexed
• Sampled control
– Multiple use of expensive equipment
– Usually in controlled system data are available/modified
in certain instants only
• Radar control, chemical analysis, SCR
– Naturally discrete
• Stepping motor, encoders
 Sampled data systems
• Sampler is introduced to improve dynamic
behavior of control loop
– Better reliability
– Better sensitivity behavior
– No drift
– Noise reduction
– Less weight
– Less hardware cost
– Less software/ maintenance cost
 Sampled data systems
• Types of sampling
– PAM
– PWM
– PFM
• PAM preserves linearity of system and is used
in control
• There are three basic types of sampled data
systems
 Types of sampled data systems
• Sampled data system used with continuous
controller and continuous plant
 Types of sampled data systems
• Sampler used with digital controller but
continuous plant
 Types of sampled data systems
• Sampler used with digital controller and
continuous plant
 Sampled data systems- Analysis
• Ideal sampler- have infinite switching
frequency
• The output of the ideal sampler x*(t) is a
series of impulse signals whose amplitude is
equal to the original signal x(t)
• X*(t) is then given as
x * (t )  x(t ) (t )  x(t  T ) (t  T )  ...
• In the Laplace transform,

X(s) X*(s)
 Sampled data systems
• The Laplace transform of the sampled signal
x*(t) is defined as

X * ( s)  
k 0
x(kT )e  kTs (1.1)

• Example: determine the value of X*(s) for

x(t)=u(t) and x(t)  et
• It may be difficult to find closed form of X*(s)
from the equation given in (1.1)
 1 
X * (s)  
at poles of X ( )


residues of X( )
1 - e -T(s - ) 

 Sampled data systems
• Example: Find X*(s) when X(s) is given by
1
X ( s) 
( s  1)(s  2)
• Another closed form expression for X*(s) is
k 
1 x(0)
X * ( s) 
T 
k  
X ( s  jk s ) 
2

• Properties of X*(s)
– 1. X*(s) is period in s with period jws
– 2. If X(s) has a pole at s=s1, then X*(s) will have
poles at s=s1+jnws
 Sampled data systems
• The frequency spectrum of x*(t) is obtained by
using Fourier transform
• X*(jw) is obtained by replacing s=jw in the
Laplace transform since x(t)=0 for t<0.
• Hence X*(jw) is given by
j ( j )
X * ( j )  X ( j ) e
• Effect of sampling on frequency spectrum is to
repeat or replicate the frequency spectrum
centered at ks with k=1,2,3…
 Sampled data systems
• The sampling theorem- sampling should be
performed at frequency at least twice the
maximum frequency of the original signal x(t)

– Where fs is sampling frequency and fmax is maximum

frequency of signal
• When sampling frequency is less than Nyquest
frequency, aliasing occurs.
• For reconstruction of sampled data signals, filter
is required- low pass filter
 Sampled data systems- reconstruction

• Data hold
– in all sampled data systems, data hold systems are
used
– It approximates an ideal low pass filter
– It converts the sampled signal into a continuous
time signal
– There are three types of data hold circuits
• Zero order, first order and second order data hold
 Data reconstruction
• If the input of a sampler is given by e(t) and e n(t)
is used to denote the reconstructed signal, then

– Where e’(nT) is the derivative of e(t) at nT and

e’’(nT) is second derivative
• Order of the data hold is determined by the
number of terms used to approximate en(t)
• First term only –zero order hold
• Two terms only- first order hold
• Three terms only – second order hold
 Data reconstruction
• Taking numerical approximation of e’(t) and
e’’(t)

– Zero order hold is given as

en(t)=e(nT)
1 e Ts
E0 ( s )  
s s

– The transfer function is then

E0 ( s ) 1  e Ts
  Gh 0 ( s )
E (s) s
 Data reconstruction contd…

• In the frequency domain, the ZOH is given as

sin(T / 2)  jT / 2
Gh0 ( jw)  T e
T / 2
 Data reconstruction contd…
• First order hold
– Contains the first two terms of Taylor series
expansion of en(t)
en(t)=e(nT)+e’(nT)(t-nT) nT<t<(n+1)T
– Where the derivative e’(nT) is given by
e(nT )  e[( n  1)T ]
e ' (nT ) 
T

– Hence the transfer function of FOH is given by

2 2
1  Ts 1  e Ts  1  jwT 1  e  jwT 
Gh1 ( s )    Gh1 ( s )   
T  s  T  jw 
 Data converters

• Data conversion from analog to digital and

vice versa is a critical part in digital control
– Analog to digital converter
• Interface sensors to digital controller
– Digital to analog converter
• Interfaces digital controller to actuator/plant
• DAC is a practical hold circuit
– Reference voltage( full scale voltage)
– Resolution
 Data converters
• ADC converts analog signal to digital
– No of bits of the ADC
– Conversion time
– Various types of converters are used
• Select converter based on the cost/conversion time
 Exercise
• 1) If the input of an ideal sampler is
e(t)=2cos(1t) where 1 <s/2 , draw the
Fourier transform of the output E*(jw).
Discuss also the effect of varying 1
• 2) Draw the frequency response
characteristics of a FOH.
• 3) List at least 3 commercially available ADC
and DAC and compare their performances
 Description of discrete time systems
• Linear time invariant causal system involves
– Summation unit
– Delay unit
– Amplification unit
• The interconnection of the three elements
form discrete time system
– Summation unit
 Description of discrete time systems
– Amplification

– delay units

• Block diagram description of a first order and

second order discrete time systems
– First order system
y (k )  bo u (k )  b1u (k  1)  a1 y (k  1)
 Description of discrete time systems
– Block diagram of first order system

y (–k )For
 bsecond
o u ( k )  b1u ( k  1)  b2 u ( k  2)  a1 y ( k  1)  a1 y ( k  2)
order system
 Description of discrete time systems
– Block diagram of second order system

• Generally four methods are used

– Difference equation
– Transfer function method
– Impulse response method
– State space method
 Description of discrete time systems

• 1) Difference equation method

– General form is
y (k )  a1 y (k  1)    a n y (k  n)  bo u (k )  b1u (k  1)    bn u (k  n)

• The initial conditions y(-1),y(-2) etc should be given

– Difference equation may be obtained from
differential equations describing systems or
continuous controllers
• Example: digital PI controller
• 2) Transfer method
– A transform method used for digital systems is Z-
transform
– It is a transform for number sequences
 Description of discrete time systems
• The Z transform is defined as
E ( z )   e(k )   e(0)  e(1) z 1  e(2) z 2  ...
k 
E ( z )   e(k )   k 0
e( k ) z  k

• Example: 1) find E(z) if e(k)=1 for all k

– 2) Find the z transform of the signal obtained by
sampling the signal e(t )  e  at
• The inverse z-transform is a transform given
by
1
e(k )   1  E ( z )  
E ( z ) z k 1dz
2j

 Description of discrete time systems
• Properties of z-transform
– Addition and subtraction

– Multiplication by a constant

– Real translation: if E(z) is z transform of e(k)

 e(k  n)u (k  n)  z  n E ( z )

 k  n 1 
 e(k  n)u (k  n)  z  E ( z ) 
n



k 0
k 1
e( k ) z 

 Description of discrete time systems
• Properties of z-transform contd…
– complex translation

 
  ak e(k )  E ( z  a )
e(k )  k ak
• Example: find the z transform of
– Initial value
• Given E(z) as Z transform of e(k) , then
e(0)  lim E ( z )
z 

– final value
• Given E(z) as Z transform of e(k) , then
lim
n 
e(n)  lim ( z  1) E ( z )
z 1
 Description of discrete time systems
• Properties of Z transform contd
– Example: Apply the final and initial value
properties to the Z transform of sample of unit
step function and check its values
e(k )  1 for all k  0
z 1
E ( z)  
z  1 1  z 1

– Applying initial value property

e(0)  lim E ( z )  1
z 

– Applying
lim final
e ( n )  value
lim ( z  1) E ( zproperty
)  ( z  1)
z
1
n  z 1
z 1
 Description of discrete time systems
• 2. Transform methods
• Similar to continuous time systems:
– the transfer function of a discrete time system is
given by H(z) which is a ratio of Z-transform of
output Y(z) to the Z transform of the input U(z)
Y ( z ) [ y (k )]
H ( z)   u(k)  y(k)  0 for all k  0
U ( z ) [u (k )]

• Can be written as ratio of two polynomials

 Description of discrete time systems
• 3) Impulse response method
– Consider a unit impulse sequence
1 k  0
 (k )  
 0 k0

– When the input to a discrete system is the unit

impulse sequence,
– the output of the system y(k)=h(k), H(z)=[h(k)]
assuming initial condition of system is zero
(k) y(k)=h(k)
Discrete time
system
 Description of discrete time systems

• 4. State space method

– State equation is set of first order difference
equations
– the state space or state equation of discrete time
systems is
x(k  1)  Ax(k )  Bu (k )
y (k )  Cx(k )  Du (k )
– Taking the Z transform of input and output
H ( z )  C  zI n  A 1 B  D

– The impulse response h(k) is obtained from

inverse transform of H(z)
 Analysis of discrete time systems
• Analysis of discrete time systems is finding the
response of discrete time systems
– Can be performed in four different ways
• Difference equation method
• Z transform method
• Impulse response method
• State space methods
• Each method has its own advantage and
disadvantage
• Difference equation method is useful for
computer programming
 Analysis of discrete time systems

• Analysis based on difference equation

– Difference equations can be solved using
• Manual method – to compute output y(k) for each k
value manually.
• Computer programs – to write a computer program
that can compute y(k) for each k
• Z transform method – convert the difference equation
to an algebric equation in z and solve for y(k) in closed
form
• Example: find y(k) for the sequence
y (k )  u (k )  u (k  1)  y (k  1) for k  0
1 for k even
u(k)  
 0, for k odd
 Analysis contd…
• Manual method
– y(0)=u(0)-u(-1)-y(-1)=1-0-0=1
– y(1)=u(1)-u(0)-y(1)=0-1-1=-2 and so on
• Computer program – write a program using
any programming language
• Z transform method
– Consider the following difference equation
y (k )  a1 y (k  1)    a n y (k  n)  bo u (k )  b1u (k  1)    bn u (k  n)

– Using real translation property,

Y ( z )  a1 z 1Y ( z )    an z  nY ( z )  boU ( z )  b1 z 1U ( z )    bn z  nU ( z )
b0  b1 z 1    bn z  n
Y ( z)  U ( z)
1  a1 z 1    an z  n
 Analysis contd…
• For the example given
Y ( z )  E ( z )  z 1 E ( z )  z 1Y ( z )
1  z 1
Y ( z)  E( z)
1  z 1

• Then Y(z) can be solved using transform

method
 Analysis contd…
• Analysis based on transform method
– This is computing the output y(k) from the inverse
transform of Y(z)
y (k )  1[Y ( z )]
– There are four methods to solve the inverse z
transform
• Power series method – divide num by den
• Partial fraction method- expand E(z)/z
• Use the Inversion formula given by
1
e(k )   1  E ( z )  
E ( z ) z k 1dz
2j


• Discrete convolution
 Analysis contd…
• Example: 1) find the value of e(k) when E(z) is
given by
z
E( z) 
z 2  3z  2
– Using power series method
• use long division to obtain e(k) , 0,1,3,7,15…
• Example 2: find e(k) when E(z) is given by
z
E( z) 
( z  1)( zexpansion
– Using partial fraction  2)

e ( k )  1  2 k
 Analysis contd…
• Exercise : find y(k) when Y(z) is given by

– a) 3.894 z 3.894 z
Y ( z)  
z 2  0.6065 ( z  j 0.7788)( z  j 0.7788)

1
Y ( z) 
– b) ( z  2)( z  1)
 Analysis contd…
• Using the inversion formula
– When computing y(k) using the inversion formula,
the following equivalent formula is used
y (k )   residues of Y(z)z 
k -1

at poles of [Y(z)z k-1 ]

– Example: find y(k) when Y(z) is given by

z
Y ( z) 
z 2  3z  2
– Residue at z=1 is -1 and at z=2 is 2k
 Analysis contd…
• Exercise: find y(k) when Y(z) is given by
z
Y ( z) 
( z  1) 2

• Discrete convolution technique

– Write the output Y(z) as product of two simpler
algebraic expressions,
Y ( z )  Y1 ( z )Y2 ( z )

– then find the inverse of Y1(z) and Y2(z)

– Find Y(z) as discrete convolution using
nk
y (k )   y ( n ) y ( k  n)
n 0
1 2
 Analysis contd…
• Example: compute y(6) using discrete
convolution when Y(z) is given by
z
Y ( z) 
z 2  3z  2

– Assume that
z 1
Y1 ( z )  Y2 ( z ) 
z 1 z2
– y1(k)= 1 for all k and y2(k)=2 power of (k-1)
– Then y(6)=127
 Analysis contd…
• Analysis using impulse response method
– When the impulse response of a discrete system
h(k) is given, the output y(k) can be obtained using
• A) Y(z)=H(z)U(z) and then using the inverse method
• B) y(k)=h(k)*u(k)
n
y (k )   h(n)u(k  n)
n  

• Analysis based on state equation- given state

space equation of a discrete systemk 1
k
x ( k )  A x ( 0)  A
n 0
k  n 1
Bu (n)

k 1
k
y (k )  CA x(0)  C A
n0
k  n 1
Bu (n)  Du (k )
 Analysis contd…
• Example: find the transition matrix, states and
output of a discrete time system when the
state equation is given by
 0 1 0  1
A  B  C 
 2 3 1 1

– The transition matrix is (k)=-1[z(zI-A)]-1

– The states x(k) and output y(k) can be computed
using formulas given above
 State equation method
• Derive the state equation and signal flow
graph for the system given by
Y ( z ) b2 z 2  b1 z  b0
H ( z)   2
U ( z) z  a1 z  a0

• Rewrite the equation in power of z-1,

 Stability of digital systems
• Open loop discrete time systems
– Consider a sampled data systems shown below
 Introduction
• Open loop discrete time systems
– Using the transform method, C(z)=E(z)G(z)
– Example: In the system shown above, if the plant
Gp(s)=1/s(s+1), find C(z) and c(kT)
• Solution:
1  e Ts  1  e T z
G ( z )    T
E ( z) 
 s ( s  1)  z  e z 1
1  e T z
C ( z) 
z  e T z 1

C (kT )  1  e  kT
 Introduction
• Open loop discrete time systems
– For the system above, the steady state value is 1.
This is obtained by using the final value theorem.
– The DC gain is the G(z) as z1
– When there are cascaded systems, the output of
the cascaded system may vary depending on the
location of the samplers
– Example: consider the following three sampled
data systems, find C(z) for the three cases
 Introduction
• Open loop discrete time systems

C ( z )  E ( z )G1 ( z )G2 ( z )

C ( z )  E ( z )G1G2 ( z )

C ( z )  G1 E ( z )G2 ( z )
 Introduction
• Open loop discrete time systems
– When the system has digital filters, the open loop
output is obtained by using the Z-transform of the
difference equation
– Example: for the system shown below, find the
output

C ( z )  E ( z ) D( z )G ( z )
 Introduction
• Open loop discrete time systems
– Exercise: Determine the expression for the output
of the following systems
 Introduction
• Closed loop discrete time systems
– Closed loop discrete systems contain samplers
either in the forward or feedback path or in both
paths
– The output can be derived by using the methods
used for open loop.
– Consider the closed loop system below

G ( z ) R( z )
C ( z) 
1  GH ( z )
 Introduction
• Closed loop discrete time systems
– When the closed loop system contains digital
filters, the system output is obtained by using the
cascade of the digital filter and analog parts
– For example: for the system below, the output is
given as

G ( z ) D( z ) R( z )
C ( z) 
1  D( z )GH ( z )
 Introduction
• Closed loop system
– When the system contains multiple samplers, the
closed loop system output is obtained by
• Drawing the signal flow graph of the original system
• Writing equations to each sampler input
• Combining the equations

G1 ( z )G2 ( z ) R ( z )
C ( z) 
1  G1 ( z )G2 ( z )  G2 H ( z )
 Introduction
• System time response for discrete time
systems
– Is obtained using the inverse transform method
ones the output z transform is obtained
– Example: for the temperature control system
shown below, determine the output when the
input is unit step

G( z)
C ( z)  R( z )
1  G( z)
 Introduction
• System time response
– Using the Z transform of unit step input and
0.3625
G( z) 
z  0.8187
– The closed loop output and its inverse become
0 . 3625 z
C (z) 
( z  1 )( z  0 . 4562 )
 Introduction
• System time response
– For a general closed loop system with single
sampler and single loop, the output is given as
m

G( z)
K  (z  z ) i

C ( z)  R( z )  R( z )
1  G( z ) n

 (z  p ) i

– Then using partial fraction expansion

k1 z z
C ( z)     kn  CR ( z)
z  p1 z  pn

– Which yields C (kt )   i 1

k i ( pi ) k  1[C R ( z )]
 Introduction
• System time response
sT
– Mapping of the s-plane in the z-plane, z  e
• the left half of the s-plane maps to a unit circle in the z-
plane
• The right half maps to a point out of the unit circle in z
plane
 Introduction
• System time response
– Generally, for a system given by transfer function
 n2
G(s)  2
s  2 n s   n2

– S-plane poles are given by

s   n  j n 1   2

– Z plane poles are

z  e sT  e nT    nT 1   2
s1, 2
 Introduction
• System time response
– If the z-plane complex pole is given as z  r  
• The damping ration, natural frequency and time
constant can be obtained as
 ln r

ln 2 r   2
1
n  ln 2 r   2
T
1 T
 
n ln r
 Stability Analysis

• Definition
– Consider a nonlinear discrete system given by
x(kT  T )  f [ x(kT ), kT , u ( kT )) x ( k oT )  xo

– Assume that u(kT)=0 for k>ko and x(kT) is the

~
solution when the initial condition is x(koT) and x (iskT )
~
the solution when the initial condition is x (k 0T )
• The solution x(kT) is stable if for some >0,
there exists (,ko)>0 such that
x k 0T   ~
x  k 0T     x kT   ~
x  kT  for all k  k 0
 Stability Analysis
• Asymptotic stability
– The system above is said to be asymptotically stable iff it
is stable and
x kT   ~
x  kT   0 as k   when x0  kT   ~
x0  kT 
is sufficiently small
• Stability from characteristic equation
– Consider a system whose output after partial fraction
expansion is given by
k1 z z
C ( z)     kn  CR ( z)
z  p1 z  pn

– The system will be stable if all poles are inside the unit
circle
 Stability
• Characteristic equation
– For a system whose output is given by
G( z)
C ( z)  R( z )
1  GH ( z )

– The characteristic equation

1  GH ( z )  0 is

– For stability
• Roots of the characteristic equation must lie with in the
unit circle
• If any root is on the unit circle, system is marginally stable
• If any root is out side the unit circle, system is unstable
 Stability
• Characteristic equation
– When the system is described in state space, as
x(k  1)  Ax(k )  Br (k )
y (k )  Cx(k )  Dr (k )

– The system transfer function is given as

Y ( z)
 C[ zI  A]1 B  D
R( z )

– Then the characteristic

zI  A equation
0 is given by
 Bilinear Transformation
• Is a transformation which transforms the z-
plane to a new plane, w-plane
• The transformation is given by
1  (T / 2) 2 z 1
z w
1  (T / 2) T z 1
• By this transformation, the unit circle in z-
plane transforms to the imaginary axis of w-
plane
 Bilinear transformation
• The stable region of the w-plane is the left half
plane.
• This enables the RH, Bode plot and other
frequency domain analysis to be performed
on discrete systems
• On the unit circle in z-plane, z  e jwT

2 T
w j tan
T 2
• Hence the frequency in s-plane  and the
frequency in w plane are related by
2 T
ww  tan
T 2
 RH criteria
• Is a systematic way of determining the stability
of a continuous system without actually
solving the characteristic equation
• In digital systems, simple RH can not be
applied but following method is used
– Convert the z domain transfer function to w plane
by using Bilinear transformation
– Determine the characteristics equation
– apply RH as continuous system
 Jury’s stability test
• Is a similar method to the RH which can be
directly applied to characteristic equation in z
domain
• Consider a characteristic equation given by
C ( z )  an z n  an 1 z n 1    a1 z  a0 an  0
z0 z1 z2 z nk z n 1 zn
a0 a1 a2 an  k an 1 an
an an 1 an  2 ak a1 a0
b0 b1 b2 bn  k bn 1
bn 1 bn  2 bn  3 bk 1 b0
c0 c1 c2 cn  k
cn  2 c n 3 cn  4 cn  2
 Jury’s stability test
• The entries in the third row and above are
calculated by
a0 an  k b0 bn 1 k c0 cn  2  k
bk  ck  dk 
an ak bn 1 bk cn  2 ck

• The necessary and sufficient condition for

stability is then
C (1)  0 b0  bn 1
(1) n C (1)  0 c0  c n  2
(1) (2)
a0  a n d 0  d n 3

m0  m 2
 Jury’s stability test
• Example: For the system shown below,
determine the stability of the system when
T=0.1s

• The characteristic equation of system is given

by 1  KG( z)  0 where G(z) is again given by
z 1  1 
G( z)   2 
z  s  s  1 


 1  z (T  1  e T ) z  (1  e T  Te T )
 2 

 s  s  1  ( z  1) 2
( z  e T
)
 Jury’s stability test

• Substituting T=0.1 and solving for G(z)

0.00484 z  0.00468
G( z) 
( z  1)( z  0.905)
• Hence the characteristic equation is
1  KG( z )  z 2  1.905z  0.905  0.00484Kz  0.00468K

• Jury’s table is
z0 z1 z2
0.00468K+0.905 0.00484z-1.905z 1

• Applying group 1 constraints, for stability K>0

and K<20.3
 Jury’s stability test
• Steps
– From the given system description obtain the
characteristic equation
– Apply group 1 constraints, it any one fail, then
system is unstable, stop
– If group 1 constraints are valid, start filling Jury’s
table and check the condition at each row. If
condition fails, system is unstable stop
– If all constraints are fulfilled, then system is stable
• Example: for a system having ff characteristic
equation, check stability using Jury’s test
C ( z )  z 3  1.8 z 2  1.05 z  0.2
 Frequency response techniques
• Root locus method- consider a closed loop
system shown below.

• The characteristic equation of the system is

1+KGH(z)=0
• The root locus of this system will be drawn
using the standard steps of continuous system
 Frequency response techniques
• Root locus method
– Example: Draw the root locus of the following
system if D(z)=K

– The1characteristic
 KG ( z )  0 equation of the system is
z 1  1 
G( z)   2 
z  s  s  1 
– Where
 Frequency response techniques
• Example: using Z-transform table and T=1, the
final characteristic equation becomes
K (0.3679z  0.2642)
1  KG ( z ) 
( z  1)( z  0.3679)
• The root locus becomes
Root Locus
1.5

0.5
Imaginary Axis

-0.5

-1

-1.5
-5 -4 -3 -2 -1 0 1 2
Real Axis
 Frequency response techniques
• Bode plot- is another frequency response
technique
– Graphical method which draws the magnitude and
phase of the closed loop response as a function of
frequency
– For discrete time systems two methods can be
followed
• Form the characteristic equation of the system in z
• Change to w plane and plot bode plot or
• Draw the bode plot in z domain and convert it to s
domain using the conversion formula