'rablc of Contents

'•f > :
CllJ\PTER I '/
~
/.
Introduction to Discrete Time Control System

Art. Topic Page

I. I Introduction
1.2 Digital Control System
1.3 Quantizing and Quantization Error
1.4 Data Acquisition and Data Distribution System .
1.4.1 Analog Multiplexer
1.4.2 Demultiplexer
1.4.3 Sample and Hold Circuits

CHAPTER II

The z Transform

2.1 Introduction 14
2.2 The z Transfonn 14
2.2. l Region of Convergence for z Transfonn 15
2.3 z Transfom1 of Elementary Function 16
2.4 Impo1tant Properties and Theorems of the z Transform 23
2.5 The Inverse z Transfonn 32
2.6 z Transform Methods for Solving Difference Equations 43
2. 7 Real Convolution Theorems 46
2.8 Complex Convolution Theorems 4.7
2.9 Parseval 's Theorems r 48
2.10 Some Examples 49
2.11 Problems 55
---
CHAPTER III

·z PJanc Analysis of Discrete-Time Contro·l System

3.1 Impulse Sampling
58
3.2 Data-Hold Circuit
3.2. l Zero-Order Hold 60
3.2.2 Transfer Function of First Order Hold
61
63
3.3 Obt11i11i11g the z Trn11sform by the onvolution Integral 67
3.3. 1 onvolution Integral in the Lc fl J falf of the p-Planc 69 ..
3.3.2 lntcgrnl in the Right Hnlf Plane 71
3.3.3 Obtaining z Transfonn of Functions Involving the Term (l-c-,.5)/s 73
3.4 Rcconstrncting Original Signals from Sampled Signals 75
3.5 The Pulse Transfer Function 77
3.5.1 Staned Laplace Transform of the Signal Involving Both Ordinary
and StmTed Laplace Transfonn
3.5.2 General Procedures for Obtaining Pulse Transfer Function
-
80

81
3.5.3 Pulse Transfer Function of Cascaded Elements 84.
3.5.4 Pulse Transfer Function of Closed-Loop Systems 86
3.5 .5 Pulse Transfer Function of a Digital Controller 87
3.5.6 Closed-Loop Pulse Transfer Function of a Digital Control System 88
3.5.7 Transfer Function of a Digital PID Controller 89
~ Realization of a Digital Controllers and Digital Filters 92
3.6.1 Direct Programming • 94
3.6.2 Standard Programming 95
3.6.3 Series Programming . 97
3.6.4 Parallel Programming 98
3.6.5 Ladder Programming 99
3.7 Problems 103

CHAPTER IV

Design of D_iscrcte-Time Control Systems by Conventional Methods

4.1 Mapping of the Left Half of the s-Plane into the z-Plane 106
4 .2 Stability Analysis of Closed-Loop System in the z-Plane., 107
4.3 Methods for Testing Absolute Stability 109
4.4 Stability Analysis by Use of the Bilinear Transformation and Routh 114
Stability Criterion
4.5 Transient and Steady-State Response Analysis 115
4.6 Sfoaa)(-State Error .Analysis 116
4 .7 Design 13ased,on the Root-Locus Methods 121
4.8 Design Based on the Frequency Response Methods . 126
4.8.1 Response of a Linear Time-Invariant Discrete-Time System 127
to a Sinusoidal Input '
4.8 .2 Bilinear Transformation and thew-Plane 130
4.8.3 Bode Diagrams ::.:.:.. . ;.• 133
4.9 Analytical Design Method 137
4.10 Problems 152
 CHAPTER V

State-Space Analysis
5.1 Introduction 154
5.2 Concept of the State-Space Method 154
5.3 State-Space Representation of Discrete-Time System 156
5.4 Solving Discrete-Time State-Space Equation 171
5.4. I Solution of the Linear Time-Invariant Discrete-Time
State-Space Equation 171
5.4.2 State Transition Matrix I 71
5.4.3 z Transform Approach to the Solution of Discrete-Time 172
State Equation
5.4.4 Solution of the Linear Time-Varying Discrete-Time
State Equation 175
5.5 Pulse Transfer Function Matrix 177
5.6 Discretization of Continuous-Time State-Space Equations 178
5.6.l Review of Solution of Continuous-Time State-Space Equation 179
5.6.2 Discretization of Continuous-Time State-Space Equations 180
5. 7 Problems 185

References 187
 CHAPTER- I

Introduction to Discrete-Thne Control Systems

1.1 Introduction:
The rapid advance-S in use of digital controllers and computers in control system are the
motivation for digital con~ or discrete time systems. Digital controls are used for
ma,imum producti, ity, efficiency, precision control etc. The current trend towards digital
rather than. ~al~g control of dynamic systems is mainly due to the availability of low-cost
...
-
Types of Signals:
1. Continuous-time (analog) signals:
It is defined O\ er a continuous range of time. The Fig: 1.1 shows the continuous-time
signal

x(t)

0 t
Fig 1.1
( Continuous-time analog signal )
In this type of signal the amplitude may assume a continuous range of values.

2. Co11ri1111ous-rime}qJJcpztized s[gnals:
The continuous time signals ~emg represen_ted by a distinct set of values-quan ti zed
values s quantized in amplitude only not in ti~e). The Fig . •1:2 shows the continuous -
time quantized signal.
 :x(t)

0
Fig 1.2

(Continuous-rime quantized signal)

The range of magnitude is cli,ided into a finite number of disjoint intervals hi which arc not
necessarily equal.

3. Sampled-Data Signals:

- ---- -
The signal which is defined at discrete intervals of time is called the discrete time
signal. (ThatJs, one in which independen,'Vafiabte t is quantiz:d). If the ~ p_l~~1cle can
take a continuous range of values, then the signal is called sampled-data signal. A
sampled-data signal can be generated by sampling an analog signal at discrete intervals
of time. The Fig. 1.3 shows this type of signal.

x(t)

0
2 3 4 5

Fig 1.3
( Sampled-data signal )

•
 4. Di ital Signals:
/,,,,. .
It is a discrete-time signal with(quant1zed • de Cl early, _it is a signal quantized
amp 11tu
- -- -- -d.- . ✓, t II requires quantization of
both in amplitude and i!! tim~ The us~ oft11e 1g1ta con roet _
signals both in amplitude and in time as shown in Fig. 1.4.

x(t)
4

..._

0 t
2 3 4 5 6

Fig 1.4
( Digital signal) ✓
Most of the signals in the world are analog or continuous type so if the digital controllers are
to be used then signal conversion from analog to digital becomes necessary.
Tenninology such as discrete-time control systems,samplcd data control systems and digital
control systems refer to the same type or very similar type of control systems.
IP

The
,_
control systems we examine in this course are mostly linear and time-invariant. Linear
~

system is one in which a principle of superposition applies. If y1 be the response of the

system to input X1 and Y2 the response to input x2, then the system is linear if and only if, for
every scalar a and p, the response to input ax1+Px2 be ay 1+py2. A linear system may be
described by linear differential equ.;tions. A linear timc-invaria t s stem is one in which the
coefficients in the differential equation or difference equation do not vary with time i.e. one
in which the properties of the system do not change with lime.

Discrete-Time Control Systems and Coutinuous-Time Con·trol Systems:

• • - - · - - - - -- - - - -----C--
(i) Discrclc-~!:!Ie control s J1ems_are those jn which. one or more variables can change
only at discrete instant of time. Th~se instants, \~hich are denoted by kT or Tk (k=O, I,
2, 3 .......). -

3
 .
Discrete-time .
control systems differ. •
from continuous-t'rme control systems.. in those
ii)
•. •
s1gnnls . . control systems arc m
for d1scrcte-hme • samp led data form or i.._-,:..
di 1 ital
_ __ form_.

If a digital computer is involved in a control system as a digital controller any

r _ _ __ _ ,

sampled data must be converted into digital data.

(iii) C_9ntinu~time systems, whose signals ares qntinug us in_time, may be_described by
EifTeren~ ~quati~ns ~ut in the _discrete-time systems, which involve sampled

-signals
-
or digital signals may_be_described by ~iff~nce_eguatio~ afier ap~~opriate
- -- ._......
discretization of continuous time signal~

~f.~ Whenever a digital controll:_r~ lved, the signal used needs to be in

digital form. Since most real world signals are i_n analog form, thus sig~_~l_z~e~s to be
converted into digital fonn. The process of con_yerjing_analog si~als to dig~al one •sJrnown
- =-- -
?S_5!mp_H ~~cess. As for examQ.l.e, a(sampling process is needed wherever a large-sea~
controller or computer is time-s are y scvera p ants in order to save cost. Then a control
signal is sent out to each plant only periodically and thus signals become a sampled data

-
signal.

T~e term "discreti~ rather than "sampling", are fi:cquently used m the analysis of
~1._!ltiple input-multiple-output system, although both the terms mean the same thing.

\ It is important t<? note that occasionally the sampling operation or discretization is entirely °'
fictitious and has been introauced only to simplify the analysis of control systems that
actuaily contain 9nly eon~inuous-time_signals. In fact, we oJl~n use a s~itable discrete-time
model for a continuous-time syst~m. An example is a digital computer simulation of a
contin.uous-time system. SucE.3 di_gital computer simulated systc_m can be analyzed to yield
parameters that will-optimize a given performance inde~.
• ,- - . . ) '

4
,
 Digita l Control ~ms
The configuration of a basic digital control system is shown in Fig. 1.5.
error
signal
input output
~

~
Digital
.
~ Actuator ...
~
Plant ~
~

Controller
...
Sensor or
Transduser

Fig 1.5
( Block diagram of a digital control system)

The input to and output of digital controller is an analog signal.

. . .. . - ...

Digital controller: The configuration of a digital controller is shown in Fig. 1.6.

Error
Signal
Analog Digital Digital Analog
.
~ S/H& .. Digital .. DIA ~ Hold I
.
ND
~

Comp~ter Circuit I

\~ Fig 1.6 t
--- - - - [cCQ0]1,. .- - -·
_ - -J'
( Block diagram of a digital controller)

The error signal is converted.to_the 9igital form by _sample and hold circuit and analog to
- · ----- 1,.- - - -

digital £Onverter. The cqnversion i~ done at the sampling time. The digit~ computer
-- - - - - . . ~ .
processes.the sequence of number by means of an algorithm and produces a new sequence of
numbers.
------

--
The DIA converter and the hold circuit convert the sequence of numbers in numerical code
·- --- - - -~ -- - - - - - ----- -- -
into a ~cewise continuous time signa!:_ I_h~ output of the bold -circuit, a continu
--- -- -
-time
- - -- - - -
signal is fed to the pl_~nt, either directly or through the actuator, to control the

5
 Sa111pli11g or Discreti:ation: The operation that transfonns continuous-time signals into
discrete-time data is called sampling or discxetjzation. The reverse operation, the operation
# I l l transfonns disc1:cto-timc ~~ co;1ti;1:0-~me signal_is called 6)i>1i.J
The S/H and AI D conve11er convert the analog signal into a sequence of numerically coded
binary words. Such conversion is called coding or encoding. The DIA conversion process is
cal led decoding.

Some Defi11itio11s:
a) Sanmle a11-d-Nold:
!t d~scribes a circuit that ta~:_s an ana_!__og ~ c l : holds it at a const
specific period of time. Usually, the signal is electric~l-b e mechanical and optical
signals are also possible after converted into electrical signals:

b) Analog-to-Digital Converter (AID)·

It is also called an encoder that converts analog signals into digital signals, usua lly a
--- - -
j ~u1:1eri~ally coded signal. ~ is needed as an interface between analog signal and
digital computer. S/H is usually an integral part of ND. ND conversion is
[lpprox1matioil.\since analog sigl!a~ _£an_take_infinite_nuJ11~ ( val!1cs while the
_ _ ~~gi~ig~9:uu1otJhe apprQfilmation i~_c~ll~g ..the quanti~n. /

c) f!.!.ig~tal-to-Analog Converter (D!Ai]

\
It is also called a decoper which converts digital signals into an analog signals-needed
~ an interfa~between digital computer and analog component ( either an actuator or
a plant).

d) Plant or Process:
A plant is any physical system to be controlled. Examples: chemical reactor,
spacecraft, machines, robots etc.

The operation to be controlled is called a process. Examples: chemical, economical

and biological process.

.J
 '5)/ Tra11sd11cer:
A~ransduccr is a device that c~nverts an input signal into an output signal o·f ~nother '
fonn. Such as a device that converts a rcssure signal into a vollagLJJ11put. The
output signal, in general, depends on the past history of the input.

Classification of Tra11s<111cers:

i) Analog Transducer: It is a transducer in which the input and output signals arc
continuo~nction of time. , /
ii) Sampled-data Tra11sc/11cer: It is one in which the input and output signal~ occur _only
@ctisc'.ete instants oftim~but the magnitudes ofth~ signais a r e ~
iii) D~gital Transducer: A digital transducer is one in which the input and output signals
oc~ur only· ~t} iscrete instants of time and the signal magnitudes ar~ quantiz~d.
V
Type,s of Sampling Operations: lb{,? ~i/-.,

_l. ~ic...Sampliug.,· In the case of periodic sampling, the sampling instants are equally
spaced, or tk == kT where k = b, 1,2. .... It is the most convenient type of the sampling
--
operation.
;ff}· Multip_k-_prder..SamplingJ he ~ of the tk 's is repeated periodically, that is, t k+r- tk

·.,r . -~ nstant for all k. ' P~

~~ e , , , c~amplinfi'Jn a control system having multiple \cops, the largest tin
/ , _constant involved in one loop may be quite different from that of other loo~Hen
it might be betterto sample slowly in a loop involving large time con tam's. While in
a loop involving only s~all time constants, the sampling rate must be fast. In this
case, two concurrent sampling operation occur at tk = pT 1 and qT 2, where T 1, T 2 are
constants and p, q are integers.
4. Random Sampling: In this case, the sampling operation is random, or tk is a random
- - - -- V
variable.

_✓. _j
 Qunntizing nnd Qunntization Error
The main functions involved in ND conversion are sampling, quantizing ar)d coding. When
. -- ,
the value of any sam~ bJ:twecn two adjacent ~tted; output states, it must be. read
on the permitted state nearest the actual value of output signal. The process of representmg a

-
quantization.
---
continuo~s or analog ~ n11l by finite ·number of discrete state is called ampli~ud~ (
~ -- ~

The output state of each _quantize~~ te is then des~ribe~_bt ~ num~rical .co~~. such a process t
is calJed cnco~ coding. Encoding is a process of assigning a digital word or cct<le' o •
each discrete state. ~

l Qua11tizi11g: The standard number system used for processing digital signal is th~binary
,,..
number system. In this system, the code group consists of n pulses each indicating either
'ON' (1) or 'OFF' (0). The n "ON-OFF'' pulses can represent 2" amplitude levels or output
states.
_The quantization I~~cl Q is t~e range be~ween two adjacent decision pointymd is given by
lR v
Q=·FsR ,._ LsP.. -
2"
-Where, FSR is full scale range.
Left most bit is called most significant bit (MSB) and has the highest value while the right
-- -
mos/ bit called least significant bit (LSB) has the least weight. LSB is the quantization level

w
' Due to the fact that the_bi__ts in~ digital word is finite, ND conversion
Q11a11tizatio11 Error:
results in a firyite resolution. The analog signals must be rounded off to a quantization level -
quantization error. The error varies between 0 al!d ± ½Q, N? matter how many bits one uses,
there is always some kind of quantization error when going ND conversion.

The Fig. 1. 7 shows a block diagram of a quantizer together with its input-output

-
characteristics. For an analog input x(t), the output y(t) takes on only a finite number of
l~hich ar..,:!nte~~Itiples of the qu~ization level Q.

8
 y

2Q
x(t
►I
y(t)
Quantizer ] ►
Q

0 Q 2Q X

Fig. 1.7
( Block diagram of a quantizer and its input-.o utput characteristics )
Since the quantizing process is an approximating· process in that the analog quan tity is
- . : . . . . - - - - - --
approximated by a finite digital number, the quantization error is a round-off error. Clearly,
~ y e_q q a ~ i s the sm~JlerJ~e o_und-off_err_or. /
The Fig. 1.8 shows an analog input x(t) and the discrete output y(t), which is in the form of
staircase function. The quantization error e(t) is the difference between the in~ut signal. and
the quantized output, or
e(t) = x(t) - y(t)
The magnitude of the quantized error is

- O~jc(t)j~_!_Q
2
For a small quantization level Q, the quantization error is like a random noise.

y(t)

x(t}

x(t)

Fig 1.8
( Analog input x(t) and discrete output y{t) )
9
 v••f Data Acquisition nnd Data Distribution Systems:
The Fig. 1.9 shows the block diagram of a data acquisition system .

Physical
./ "fl
- - - - - , '/ ,------,
Variable Transducer Sample
Amplifier Low Pass Analog
Filter Multiplexer &Hold

To Digital
AID
Controller
Convener
Fig 1.9
( Block diagram of data acquisition system )
In this system, the input to the system is a physical vruiable such as_ ppsition, velocity,
~ cceleration, temperature _o!:__~ressure. Su_ch a physical variable is first converted into an
electrical signal (a voltage or current signal) by a appropriate transducer. The amplifier that
follows the transducer performs one or more of the following functions.
(i) ~t amplifies th:._voltag£_9_utput of the transducer.: _/
(ii) It converts the current signal into a voltage sign✓.
(iii) It buffers the ~ignal/ •

The low pass fitter that follows the amplifier attenuates the high frequency signal'
.
components, such as noise signals (The electronic noises are random in nature and may be
--- - -
reduced by low-pass filter). The output of the low-pass fitter is an analog signal. T~is signal
is fed to th~ analog m~ltiplexer. T~e output of the multiplexer is fed to the sample and ~old
circuit, whose output is fed to analog to digital converter. The output of the converter is in
r digital f~~ and it is fed to the digital computer (controll; r). ~ - - - --
- --- - -

The Fig. 1.10 shows the block diagram of data distribution system.
✓ --
F~ital
✓ K Dl'>

--- -
controller
Register Demultiplexer D~ Hold ...r
~
~ ~
~
r r r
Co(lY.C.[t To
actuator
-.. .... , .....__ .:~:
Fig I.IO
( Block diagram of data distribution system )

.J
 . . . process 1s
The reverse of the data-acqms1t1011 . the data- d'1stn'but'1on process• It consists
.. of
.
register, .
a demult1plexer, . .
d1g1tal-to-analog converter and ho Id circut
• ·ts. It converts the signal

in digital form (binary numbers) into analog fonn. ':[he output of DI A converter is fed to the
hold circ~it. !J1; output of the h~ld circuit is fed to the analog actuator, which in tum,
directly controls the plant under consideration...: ~
-·---

1.4.1 Analog A111/tiple.1:a:..- 1

The analog multiplexer is a device that performs the function of time-sharing in AID
converter among many analog channels. If many ~gnals are ~o be_processed_by_a single
?igita! control!~! then these input signal~must be fed to the controllc~through a multipexer.

2.

~
3---./WWvt._<)--+---O-
... (I.)
:;j c::
0.. ~
c:: (ll
1-----+--t--7 To Sampler
...... ..c
u ------·

Fig 1.11
( Schematic diagram of an analog multiplexer )

The schematic diagram of an analog nrnltiplexer is shown in Fig. I. I 1. It is a multiple switch

-- - - - ---- . -
(an electronic switch) that sequentially switches among many analog input channels in some
prescrib~ fashion. W~en the ~witch is on in a given input channel, the input si gnal is
1
connected to the' output of the multiplexer for a specified period of time. During the
connect~n time the sample and hold circuit samples the signal voltage (analog signal) and
holds its value, while the AID converter converts the analog value into digital datV

l.4.2 De11mltipJe.,'t:er:
,__/,,

The opposite of multiplexer js demultiplexer which separates the composite output digital
data from the digital controller into t h ~ e l s .
11
~ 1.4.3 '
Sample-a11d-/Jo/d Circuit:
A ~nplcr-ina d igital system c o ~ a l o g signal into a train of amplitudc-n~odulatc<l
-Y Th~ hold circu~t l~ol~s the value of the sampled p~ signal over a spcc1 1c pcno of
t ~ The sample-and-hold is necessary in the ND converter to produce a number that
accurately represents the input signal at the sampling instant.
,_,

The Fig. 1.12 shows the simplified diagram of the sample-and-hold circuit. 1he S/H circuit is
an analog ,circuit in which ~ (input voltage is acquirei)tnd then stored /on a high quality
capacito__~:}.vith low leakage rnd low dielectric abso~ h aracteristics.
• ' l vs1/ - l J 6- ...v

o~-t£:,_____.__eyf6r--_,__---,

Arnp. 1
I
!
Amp. 2
1
An l In C- ' l /Qu..W-

~1-0-g-p-ut_ _(b_:..-----_~
_\ _._ _ _ _-+-!_ _ _-+-C-H.....--- -~

1
- An-al-o-g
O-ut:ut

Sampler

Sampler -and-hold
Command
.----
Fig 1.12
( Sample and hold circuit)

As shown_in the Fig 1.12 an clectroni:_ s~~itch is connecte? lo the ~old capacitor. S)perational
amplifier 1 is[iui input buffer amplifier with a high input impedanceJQperational amplifier i-·
is the output amplifier which buffers the voltage on the hold capacitor. \Vhen the swi.tch is
- - - - -
• closed the capacitor charges up to the average level of the input signal that means the charge
acitor in the circuit tracks the input voltage. '!bis mode of operation is known as
==--.........~d-e_. -~O\~, when the switch is open the capacjtor starts to discharge very slowly

____
through the stray leakage hence the capacitor voltage holds constant for a specified time
.........:.._

12

IJ
 ..
r criod Hild this operation mode is k11ow11 as ~ HWC..
'fl
JC
r··tg. 1.13 sho ws the tracking

11111 11· : 111tl hold 111111ld --- 1 l uhl 111<,11~

!j111 01 1lc• 10
holil o ff s~l droor ~

. .'
.
C
~

-~ .T
--~: _ _ J _ ____ _
------ - -
;;
Q
1
:i I
0 I /\prr111rc
'O
►t 1 ◄- 1i,11r
..
C -
I I 0 11 1pu l
:.i
u I I si~n:11
C I I
I I
I I

TrJck ing ·-►11 flolJ

m(Hh? ◄ -- , notJc

t
I !old con>r11311J
is Hivcn her c

Fig I. IJ
( Tracking mod..: and hold mode)
---
In foct, the switching from the tracking mode !o the hold mode is not instantaneous. _When
' '

!he :,old command is given while the circuit is in the tracking ~1'!_odc, then the circuit will stay
~ -
• J - . ' - .- • ...- - - - •

in ~11! trncking mode fo~ a S.il..QLL\~e before reacting to tl1c hold conrn1~11d. }'he time intcrv;il
- . . . - . -
du ·111;~ which the swi!ching takes place is c:1llcc! !lie .aperture time.
j --- •
j

I
0

- - ----. - - -- • -- • - - - .-- -
Thc 011tput voltage during the hold mode may decrease slightly. The hold mode droo1~ may
-
_be rri~uced by using a hi_gh-input-impcdarfce optp11t bu er amplifier Such an Oltlput buffer
am?lificr must have very_!ow_bi~s cu~·c11t. / ·
.J
Thi: sample-and-hold operation.is controllssl by.-a. criodic clock,---

13

I
 CIIAPTER-11

The z-Trausforn1
2.1 Introduction
-------
A mathematical tool commonly used f~r the analysis and synthesis of di screte-time control
systems is the z transfonn. The role of the z transfonn in discrete-time systems is simil ar to
that of the Laplace transfonns in-continuous-time systems.

In a linear discrete-time control system, a linear difference equation characterizes th~

,-
dynamics of the system. To determine the systems response to a given inpti( such a
difference equation must be solved with the z transform method, the solutions to linear
difference equation equations become algebraic in nature.

Discrete-Time Siguals: The discrete-time signals arise if the system involves a _sampling
--- - -- - - - --·-- .

-- - - - - - - - - ---
operation of continuous-time signals. The sampled signal is x (OJ, x (T), x (2 T), .......... where
T is the sampling period. Such a sequence of values arising from the sampling operation is
--==-~ ==--~ -- -- -

--- -----
usually written as x (kT). If the system involves an iterative process caJTied out by a di gital
-
COI.!!IDJter, the signal involved is a number sequence x (OJ,~). x (2J, .......... The sequence of
number is usually written as x (kJ. Although x(k) is a number sequence, it can be considered
as a sampled signal of x (tJ when the sampling period T is 1 sec.

2.2 The z Transform

The z transform..of ~-time function. x_(!), ~ here t is nonnegative, or of a scquef}~e of values
x (kT), where k takes zero or- positive integers
-
and T_i~ the sampling period·, is~;fi~ect' by-the
following equation: ""
~-
.
~.. . • X(z
. ) = Z [x (t)] = Z [x (kT)] = L
x(kT)z-k ....... ...... (2.1)
k =0
For a sequence of numbers x(k), the z transform is defined by

a:,

X(z)=2[x(k)]= I~0 .. ........... (2.2)

k =O

14
 The 2
transform defined by Eqs. (2.1) and (2.2) is referred lo as the one-sided z lransfonn .
..
The symbol 2 denotes "The z transform or•. In the one-sided z transform, we assume
x (t) == 0 for t < 0 or x (k) = 0 fork < 0 .

The z transfom1 ofx (t), where - oo < t < oo, or of x (k), where k takes integer values (0, ± 1, ±
2, ........), is defined by

X(z) = 2 [x (t)] = 2 [x (kT)] = L"' x(kT)z•k ...... ..........(2.3)

k=-oo
00
X(z) = 2[x (k)] = 2[x (k)]= Ix(k)z·k ... .... ... .... (2.4)
k=-oo
The z transfom1 defined by Eqs. (2.3) and (2.4) is refe1Ted to as the two-sided z transfonn. In
the two-sided z transform, the time function x (t) is assumed to be nonzero for t < 0 and the
sequence x (k) is considered to have nonzero values for k < 0. Both the one-sided and two -
sided z transforms are series in powers o f v

2.2.1. Regio11 of C(!11verge11tfor z-Tra11sfo1·m:

As we know that

"'
X(z) =2 [x(k)]= :Ix(kT)z-k
-- - -

- -
is a power series which may not conv ge for all values of z. That is, the infinite sum may
not be finite. For any given' sequence, the set of values of z for which the z transfonn
-
converges is called the ~egion of convergence (ROC):::::_

Example 1. Find the region of convergence for the function x (k) == 1, k=0,1,2, ......
..,. .

Solution: As we know that,

1
X(z)=2[z(k)]= - -
-1
~~
1-z

1
converges if l z- I~ I i.e. I z I> 1 (ROC)
15

H- I ( ,I

( l (, \
 Fig. 2.1 (Region of convergence)

Example 2. x (k) == ak u (k), where u (k) is a unit step function:-

~
1
Solution. Now, Z x(k) == - -1
l - az·
1
ROC is jaz- j L 1

.. • i.e. jzj) a

Example 3. x (k) = ( l/2)k u (k) + ( -1/3 )"u (k), where u (k) is the unit step function.
/ ./
I I
Now,Zx(k)= +---
1- 12 z-1 1 + 13 z-•
ROC'sare
1
l~z-•1Ll&-iz· Ll

l 1
1.e. lzl )-,
2
lzl )-
3

Hence, the region of convergence for this function is lzl > !3

2.3 z Transforms of Elementary Function

~ Unit-step f1111ctio11: Let us find the z transfonn of the unit-step function

16
 X 1 ~ { 1 (I) , t ~ 0
0, t O
The z transfonn is give by,
., .,
X(z) =2 [1(t)) = Itz•k = L:Z.1c

k=O k =O
=1+ z·' + z•2 + z·3 + ............. .
=-

=-
z-1

Sii) U11it-Ramp F1111ctio11: Let us find the z transform of the unit ramp function

(t)=: {j ;
X t~0
O· t<0
'
The z transform of unit ramp function is give by,
., .
X(z)= 2 [t] = Ix(kT)z-k
00 .,

= Ikrz·lc =T I kz•k
k=O k=O
=T -I

=Ti
n - z-')
-T z. I)
- , (z - 1) 2 )
The figure below shows the sampled unit-ramp signal.
xro x~n
....,x(4T) -

x(3T)
~
-
x(T)

--.
I

0 T 2T 3T 4T ST )

Fig 2.2 ( Sample unit-ramp signal) 17

Polyuominl F1111ctio11 a": Let us obtain the z trnnsfonn of x (k) as defined by

ak k "7 0, I, 2, ..... .
x(k)== '
{
0, k<0
Where a is a constant.
The z transform is then given by,
"' 00
X(z)=2[a1:]= L x(k)z-k =Iakz•k
0

-_ 1 + az -I + a 2 z - 2 + a 3 z - 3 + .......... .
1
=---

S}_µ)
u Expo11ential F1111ctio11: Let us find the z-transfom1 of

x (t)= { e•at ; t~ 0
0 ; t <0
Since x (kT) = e•aH, k = 0,1,2,3, .....
we have
.,
X(z) = ~[e-11 ] = Ix(kI)z-1c
k• o

-
00
.=L e - •kT z -le
kso
---
1
= - - - -1
1- e - ar z-

=~z-e
Jjv) Sinusoidal F1111ctio11: Consider the sinusoidal function

sinw t ,t~0
x(t) =
{ 0 , t(0

13
A we know that
d1~1 = cos wt + j sin wt

c ~ml = cos wt - J• sm
•
w t.

Thus we have

Since the z ·transfonn of the exponential function is

We can write

X(z) = 2[sinwT] =z[L (eiOJI -e-iax )]

=~( 1 __- I )
2 J 1-eiwTz- 1 1-e-io,,tz- 1

(ei"'T - e- i.,r)z-1
=-- -'---- -- - - - -
2j l-(e JwT +e- 1.,1) 2-1 +z-2

z- 1 sin wT
=--------2
1- 2z - 1 cos w T + z -

' zsihwT
2
z - 2z cos w T + 1
;,_/ ,,q. ,d~:(:'<-~, '>-..,_.-,,
.1,··. ·;:=, ,/ .I/ . ,, .•
y xample ~ tain the1}[~~nn"6tlfe· cosine function
/ " {co'sco·t ,t ~ o'
x(t) =
0 ,t < 0
Solution. The z transform is given by ' \

- 19
 -- 2-(e -J"'T +ef"'T)z -1
21-(el"'T +e-J"'T)z-' +z-2
l-z-1 coswT
= - - 1- -- - -
1- 2z- cosw T + z- 2
2
z - z cos w T / ,
- z' -2zcosmT+I~

Example 5. Obtain the z transform of

,./
1
X(s)=--
s(s + 1)

Solution. Whenever a given function is in s-domain then there are mainly two approaches to
determine the z transform. The first approach is to convert x (s) into x (t) and then find the z
transform of x (t). And the second approach is to expand x (s) into partial fractions and use a
z transfonn table to find the z transform of the expanded tcnns. There are still other
approaches, will be explained later.

The inverse Laplace tra_nsform ofx (s) is

X (t) = 1 - e·t, t ~ 0.
. , ''
Hence, 11 •
~

1 1
X(z) = 2[1 -e- 1] = --1 -
l-z- 1-e-Tz-1
(l-e-r)z-1
= - _ . c __ _; _ __ _

( I - z -i )( 1 - e -r z _, )
(1-e-r)z
=
(z-l)(z-e-T)

20
 -- T 2 [ z., +4z•2+9z. J+ 16z •4 + ..... .]
2 1
z • + ..... .]
- T z ., [ 1+4 z• +9z +16 •2 1 ,;·
('I
(
1
..
,
= TT~z:.b + ~-\) / ~ ~ c,(' l.

(1- z-1)3~ ~u: '5< (

{' '.,.-
...
2
5>
_ T z(z+ 1)
- (z-1) 1

Table ofz tra11sforms:

X (s) X (t) X (kT) or x X (z)

v/
(k) ,

I. - - Kronecker
delta~ l
. ,. I, k = 0
'
O,k-:1;0
2. - - Oo (n- k)
1, n = k z-k

0, n * k
3. 1 1
- ../ --
.•
s l(t) l(k) 1 -z
'
-
-1 l
-aT
4. s+a e/ e-akT 1-e z -1
1
- Tz-'
5. ~ i: KT (
(1- z- 1 )2
./
21

\
I. ·, -·
• "'
 - -
('
/.,.
T 2z -1(1 +z- 1)
(})
' '7
J ..
6. ·t2 (kT)2 (I'- 2-1)J

6 T 3 z- 1{1 +42- 1 + z-2 )

7.
-
s4 T3 (l-z- 1 /
(k1:)3
a (1 - e-•T)z- 1
8. s(s + a) I -e-al 1 - e•ak\ (1:....z-l)(l-e-aTz-1 ) •e -,/'

~_,.1 l'"". J

b-a (e -aT -e -bT )z - I

9. (s+a)(s+b) e--31 - e-bl e-Jkl-e-bkl (1-e-ffz-1)(1-e-brz- 1)
' )\ ✓{. ~,, .,,.

1 T -aT -I
L-
L-1,.., ~
10. (s + a) 2 te•al KTe·~kT (l _ ~ -aT 2_:1 )2
1
t...--- .,/ s- ~ 1-(1 +aT)~
11. (s +_av ( I - at)e·0 \ ( l - akT) e•akT (1- £-aT z~I)2
~ ,(.J-·) ..,.•·-·--;,
• ~ - rift
2 T 2 e - aT{] +-aT~-1) z -I
12. (s + a) 3 tl e•at (kT)2 e•akT (1- e-aT 2- 1 ) 3
..___,
a2 [(aT-1 + e-ar) + (1-e-•T - aTe-ar )z- 1 ]z- 1
13. s 2 (s+a) at-1 +e•at AkT-1+e•akT (1 - z- 1 )2(1 - e-.r z-1 )
nd l'\.f (.yl'](A ',

OJ z-'sinOJT
14. s2 +OJ2 sin ro t sin wkT v 1 - 2z- 1cosOJ T + z- 2

s l - z- 1cosro T
15. s2 +m2 ' cos rot cos rokT ✓ 1 - 2z-1cosm T + z-2

OJ c-•T z- 1sinOJ T
\!::.-I
(s+a)2 +(t;2 1- 2c-•T 1 21 2
16. e•at sin rot e•aki sin rokT ✓ ___, z- cosm T + e- • z- ~

s+a 1- c-•T z- 1cosm T

2
17. (s +a/+ m
~

-
e•al COS ©t e-akt cos rokT
/
1- e-aT Z-1cosm T + e- 2• 1 z- 2
" e--=.,

1
✓
18. ak 1 - az -1

z-1
k-1
a .v
19. 1- az -1
k = 1,2,3,....
22
 -1
z
.,
20. ka
k-1 (l - az - I )2

z- 1(1 + az- 1)
21. k2 l ·l (l-az- 1) 3

z - 1(1+4az- 1 +a 2 z-2 )
22. k3 i" 1 (l-az- 1) 4

z-1 (1 + l laz- 1 + l la 2z- 2 + a 3z- 3 )

23. k4 l-1 (J-az- 1)5

1
24. akcos k 1t l+az- 1

k(k-1) z-2
25. 2! (1-z-')3

k(k - 1) ... (k - m + 2) - m+ I
2

26/ (m-1)! (l-z- 1)"1

k(k-1) k-2 z-2

.. .. a
27. 2! (1- az-• )J

k(k-1) ... (k - m + 2) k-m ► t 2

- mtl
a
28. (m -1)! (1 - az - 1)"'
x(t) = 0, for t<0.
x(kT) = x(k) = 0, for k<0.
Unless otherwise noted, k = 0,1,2,3, ... ... ..

2.4 Important Properties and Theorems of the z Transform

iJ)' Afultiplication by a co11sta11t: If X (z) is the z transfonn of x (t), then

2'[ax (t)] = a 2' [x (t)] = aX (z)
\\There a is a constant.
Proof: By definition of z transform, we have
,,, . ,,,
2'[ax(t)] = Iax(kT)z-k =aI:x(kT)z-1:

;;; aX(z)

23
(ii) Linearity of tile z trn11sform: The z transfonn possesses an important property:
linearity. If f (k) and g (k) are z transfonnable and a and p arc scalars, then x (k) formed by a
linear c~mbination x (k) = a f (k) + pg (k) has the z transfonn
X (z) = a F (z) + p G (z).
Where F (z) and G (z) are the z transfonns off (k) and g (k), respectively.

Proof: The z transform of function x (k) is given by

X(z) = 2[x(k)] = 2[a ((~) + /J. g(k)]

..,
= 1)a f(k) + p g(k)]z-k
= a 2[f{k)] + p 2[g(k)]
= a F(z) + fJ G(z)

(iii) Multiplicatio11 by ak: If X (z) is the z transfonn of x (k), then the z transform of
- - --
ak x (k) can be given by X (a·1z):
1
2[a'<~X(a· z).

Proof: As we know that

00

2[akx(k)]=I a1cx(k)z-1c
k=O
00

= Ix(k)(a-'zrlc
k =O _ __

=X(a-'z)
--

(iv) Shifting Theorem: This theorem is also referred as the real translation theorem. If
0
x (t) = I fort< 0 and x (t) has the z transfonn X (z), then
2[x (t - nT)] = z·" x(z) .... ....... (2.5)

n-1

and 2[x(t + nT)] =z [X(~)-

0
L x(kT)z-k] ........... (2.6)
k sp

where n is zero or positive integer.

.
24
 .

~( Proof: As we know that

.,
2[x(t - nT)] = I:x(kT- nT)z-k
k ~
00

= z-nIx(kT- nT)z-(k-n)
k• O • · -

Let m = k- n
:. k=m+n
when k = 0, m = -n
k = oo, m = oo

So, 2[x(t - nT)] = z-"

..
L x(mT)z-m .__/
m•-n

Since, x (mT) = 0 form < 0, so we can change the lower limit of m = -n to m = 0

Hence,
.,
2[x(t - nT)] = z-" Z::x(m1Jz-m
= z"X(z)

Now, to prove Eq. (2.6).

As we know that,

2[x(t + nT)] = L"' x(kT + nT)z-k

= z" [f
k• O
x(kT + nT)z-<k•n>]

= z"I f
t m•n
x(mT)z-m ].

where k + n = m (suppose)
so that when k = 0, m = n and
when k = oo, m = oo.

25
 2[x(t + nT)] = z"r :t
~o
x(mT)z-"' - f
m• O
x(mT)z-m ]

= z"[X(z)--Ix(kT)z-k] (proved)
k =Q

Example 7. ~d the z transfonns of unit-step functions that arc delayed by 1 sampling

_/period and 4 sampling periods, respectively, as shown in Fig.2.3 (a) and (b). I

x(t)

0 T 2T 3T 4T ST 6T
(a)
( Unit-step function delayed by l sampling period )

x(t)
l(t-4T)

t
0 T 2T 3T 4T ST 6T

(b)

( Unit-step function delayed by 4 sampling period )

Fig 2.3

Solution. Using the shifting theorem we have

Also,

26
 •

. £x3mple . 0 1:un e ramfum f

1 ,.
f (k) = a • . .• ==- L - .3,........
'
= 0. k$ 0
Solution. The z tri..nsf m1 of a" is
I

=-- \\llerek=l, -.3 ...... -

I -az -l

Example 9. Consider the function y (k . which is a sum of functions x (h),

~ -here h = 0.1.-......k. such that

\
y(k) = L x(h) k = 0, 1 2..........

where y (k) = 0 fork< 0. Obtain the z transform of y (k).

Solution. GiYen that,

\:
y(k) = Ix(h)

= x(0) + x(l) + ..... + x(k -1) + x(k)

and also, y(k-1) = x(0) + x( 1) + ..... + x(k - I)

Hence, y (k) - y (k- 1) = x (k), k = 0,1 2,.....

or
Y(z) - z-1Y(z) =X(z)
: 1
:. Y(z) = X(z)
(1 -z-1)
27

J
) {complex Tra11sfotin11 Theorem: lfx (I) has the z lransfonn X (z), then the z transfo"'.' ~
e· x (t) can be given b·y X (z c0 \ This is known as the complex transl ati on theorem.
------
Proof: As we know that

2 [e-11 x(t)]= fk-0

x(kT)e-al.T z-k

CIO

= Ix(kT)(zc•Trk

Thus, ,:e j conclude that to find the z transform of e•at x (t), we can just replace z in X (z)
byzeJ •

•
Example 10. Obtain the z transfonns of e•at sin cot and e•.it coscot, respectively, by using the
complex translation theorem.

Solution. As we know that

. ] z -J sin {JJ T
Z [sm mt = - --1 -- - - -2
1- 2z- cos w T + z -e

Let us now substitute z eaT for z to obtain the z transform of e•at sin cot, as follows:

-at • ] e-•Tz-1 sin m T

Z [e smm t = - -- - 1- - - - -2 - -2
1- 2e -ar z- cos m T + e • aT z-

Similarly, ,

Let us now substitute z eaT for z to obtain the z transform of e•at cos (J)t, as follows:

28
 Example 11. ~ n the z transfonn s oft c·~

Solution. As we know that

Thus,

(vi) Initial Value Theorem: I~b~t)Jp~ t!1e _z_ transfonn XJg and if lim X (~) ex ists, then
the initial value x (0) of x (t) or x (k) is given by ~ '2::-) c;o

x (0) = Jim X (z)

z➔ 00

Proof: As we know that

"'
X(z) = Ix(k)z-k
1.- 0

= x(0) + x(l )z-1 + x(2)z- 2 + .....

Letting z➔co, we get

lim X(z) = x (0)

Z➔ «>

Hence, the behavior of the signal in the neighbourhood of t =0 or k = 0 can thus be

determined by the behavior of X(z) at z = co. This theorem is .useful for checking z tran sform
. -
calculations for possible errors. Since x (0) is always kn<_?-~~. a check ·for the initial value by
Jim X (z) can easily spot• errors in X (z), if any exist
Z ➔CO

29

---
 I

Example 12. Detcnnine the initial value x (0) if the z transfonn of x (t) is given by .

Solution. By using the initial value theorem, we find

Note: The given _X (z) was the z transform of x (t) = 1- e-1. So, when t = 0, x(0)=0 which
agrees with the result obtained by initial value theorem.

(v) Filial Value Theorem: Suppose that x (k), where x (k) = 0 for k < 0, has the z
transfonn X (z) and that all the poles of X (z) lie inside the unit circle, with the possible
exception of a simple pole at z = 1 (condition for stability). Then, the final value ofx (k), that
is, the value of x (k) ask approaches infinity, can be given by
lim x (k) = x (oo) = Jim [(1 - z- 1) X (z)]
Z➔ I

Proof: As we know that

..,
2 [x(k)] = X(z) = L x(k)z-k .... . .. . ..(2.7)
k:O

and
"'
2 [x(k - 1)] = z-1X(z) = ~ x(k -1 )z-k . . ..... .. . (2.8)
~ -
-
k~-0 : _ _ - - - -

-
Hence, subtracting Eg·. (2.8) from (2.7), we get
.., ""
L x(k)z-k - L x(k-l)z-k = X(z)- z- x(z) 1
.. . ... .. .. (2.9)
---=--z--
k=O

Now, taking the limit as z ➔ 1 to the both sides, we get

30
 lim [ Ix(k)z-1c-
"' "'
Ix(k-l)z-1c ] = lim [(l-z- 1 )X(z) ] ....... (2. 10)
z ➔ l k• O k• O z - >l

The lefi-hand side of Eq. (2.10) becomes,

[x (0)- x (- 1)] + [x (1)-x (0)] + [x (2)- x (l)] + ...... = x (cx:) = Jim x (k)
\

• Hence,
lim x (k) = lim ((1 - z- 1) X (z)] ..... (2.11)

This Eq (2.11) proves the final value theorem. The final value theorem 1s useful m
detennining the behavior of x (k) as k➔oo from its z transform X (z).

Example 13. Obtain the final value x (ex:) of

1 1
X(z) = 1- z-1 1- e-aTz-• a>0

by using the final value theorem.

Solution. By using the final value theorem.

lim r ]
x(oo) = L(l-z -•)X(z)
-~ z ➔ 1 1~ 7
- I
lim [ _1 ( 1 1 \] '
=z ➔
1 (1 - z ) 1 -I
-z - 1
-e- aT z-I JI
=1
• • .... • ,,, ~ ·# -

• .. • .. : . ; ; - ~ :· : . 'J

Note: The given X (z) was the z transform ofx (t) ~ t·-··~" 11
• So, x (ex:) = 1. Which agrees with "
the result obtained by final value theorem.

I -
31
 2.5 The Inverse z Transform
The notation for the inverse z transfom1 is 2· 1. The inverse z transfonn of X (z) yields the
coITesponding time sequence x (k).

When X (z), the z transfonn of x (kT) or x (k) is given, the operation that detem1ines the
corresponding x (kT) or x (k) is known as inverse z transformation and is denoted by
x (k) = 2" 1 [X (z)]

\jt ~hould be noted that the inverse z-transfonn yields the sampled function x (k), not the
continuous-time function x (t).

Here, four methods, by which we can detennine the inverse z transform, are explained.

"'
(i) Direct Divisio11 Jt1etlrod: The defining equation of the z transfonn i.e. X(z) = L x(kT)z -k
- ~=0

suggests that the inverse z transform can be determined simply by expan£ing the function
into an infinite series in powers of z· 1. This method is useful when it is desired to find only
the first several tenns of x (kT) or x (k) .

.
As we know that
..,
X(z) = I x(kT)z-k

= x(O) + x(T)z-1 + x(2T)z -2 + ...... + x(kT)z -k + ...... .

or,

= x(O) + x(l )z-1 + x(2)z- 2 + .. ,..... + x(k)z-• + .......

.
Then x (kT) or x (k) is the coefficients of z•k terms. Hence, the values of x (kT) or x (k) for
k = 0, 1,2,..... can be detennined by inspection.

32
Example 14. Find x (k) fork = 0, 1,2, .... when X (z) is given by

Solution. The function X(z) can be written as a ratio of polynomials in z·' as follows,

X(z) = 2 z
z -0.5z-0.5
z-•
= -- - 1- --
1- o.sz- - 0.5z- 2

And divided as follows:

o.sz-2 + 0.5z-3
o.sz-2 -0.25z-3 - 0.25z-4
0.75z- 3 + 0.25z-4
0.75z-3 - .......... .

Thus, X (z)_= 2· 1 + 0.5 z·2 + 0.75 z·3 + .......

By comparing this infinite power series expansion of X (z) with

110

X (z) =Ix(k)z -1t , we · c -

k=O
X (0) = 0

x(l) = l.
X (2) = 0.5

X (3) =0.75

\. \

~h ~ <:,D(\ l)('l'\(y d e-,i-,e~ • Tn ~fnna.t .q · r- f, 1 , '•r-c

Q c\o~rc\ { Y ?1 " H, r'Y'I -~ctr 1 <•/ ) l PY r 1- , n

33
 Example 15. f-ind x(k), when X(z) is given by

X(z) = _!_
z +1

Solution. X(z) can be rewritten in the power series in 2· 1 as follows:

z-1
X(z) = -- = z - 1 -z-2 + z -l - z- 4 +~ ..... .
l +z-1
By comparing the infinite series expansion of X(z) with '

X (z) = L"' x(k)z-k , we obtain

X (0) =0
, x(l)=+l
X (2) =-1

X (3) =+ 1
x(4) = -1

This is an alternating signal of 1 and -1, which starts from k = l. The Fig.2.4 below shows a
plot of this signal.

x(k)

2 3 4 5
-1 0 1 k

, .
Fig 2.4
.,.. .· · .....
• ( Alternating signal of 1 and - I starting froR'I k=~,·): •
Example 16. Obtain the inverse z transfonn of X (z) = I + 2z~ •~ 3 z·2 + 4_ z_:
Solution. Since X (z) has a finite number of terms, it corresponds to a signal of finite length.
By inspection, we get

34
 X (Q) = 1
x(l) = 2
X (2) = 3
x (3) = 4 and all other x (k) values are zero.
(iii) Partial-Fractio11- Expa11sio11 NIet/,od: i° n efljifl~~l1f\_J °Wh'CCLr\<Jri S eh --J_J,,.~ --z. - tyOr'\-!

~ 'YN~i,-<,G\ )( Le..~ l"Tfl (j ~ ~ ~Nl ,

YY\ m-
Xl-z...) ~ bo --z- 4- b '<- -lr- • • -+b ff\. ( m~ f' l
(a} For simple poles: __,,/
-z.. .i.a ... n- ~ . ~o
Let us consider X (z) as given by 'bo --z- - < -1-11} - r - m~·.,' _,
..i.o""
- - - - --Lh-i.
--- .....

b m b m-1
X(z)= oz + ,z + ........ + b m-lz+bm
msn
Zn +a,z
n- 1
+ ........ +an_,z+an '

Let ~1s now-factorize the denominator of X(z) and find the poles of X(z):

X(z) = b oz m + b ,z m-l + ........ + b m-,z+ b m

(z._- P1 )(z-p2 ) ........ (z- Pn)

.
We then expand X(z) into partial fractions so that each term is easily recognizable in a table
z
of z transform.

where

(b) For multiple poles: If X(z) involves a multiple pole, for example, a double pole at z = p 1
z

and no oth~r ·;~i~:.~ ~~~· X(z) · ~vijiiye t~ torm

z ·•
X(z) c1 c
--= +- 2-
z (z-p.)2 ;-:::-Pi

Now, the coefficients Ct and c2 are determined from

35
 c = [(z-p )2 X(z)]
I I Z ■ p 1i
z _,
I •
)' , /,
(
_,, /I•II

Example 17. Obtak-the inverse z transform of

X(z) = 2z3 + z
(z-2)2(z-1)
by use of the partial-fraction expansion method.

Solution. Let us expand X(z) into partial fractions as follows:

Where k·1 =[ 2z2+1

2
(z-2)2]
(z-2) (z-1) Jz=~

_ 2z 2 +1 /
- z-1 1z=2
= 8 +I
=9

k, ={![~:~:;]L
2
= (z-1)(4z)- (22 + 1) I
(z-l)2 z=2
_..,/'
8-9
=
= -1

36
 k - 2z 2 + 1
3 - (z-2)2 z=l
I
=3

X(z) .9 1 3
-- = 2- - +--
z (z - 2) z-2 z-1
9z z • 3z
or, X ( z ) = - - - - - + -
(z-2)2 z-2 z- 1
- -c.-, \ '3. -=- 9\,,. ~ _,
• • <. ,- ::?-<..-. r... ,-
-i... 7--•
From the table of z transtonn, we can wntc ' - -z...-\
'

where k = 0, 1,2,3, ..... .

,
Example 18. Obtain the inverse z transform of X(z) = {l - e-•T )z , where a is a
/ (z-l)(z-e-•T )
constant and T is the sampling period, by use of partial fraction expansion method.

Solution. Lel u~ expand X(z) into partial fractions as follows.

z

Where,
1- e-•T
k 1 = - --aT
-
I
z=I
z-e
=l
and,
} -aT

k2 = : : I Iz=e-aT
=-1
'
X(z) I.
--= - -----
z z-1 z-e-aT

By use of z transfonn Table, we can write

37
 x (kT) =l - e •JkT wl1ere k == 0, 1,2, .......
~

(iii) Jm·ersiou Integral Afethod: As we know that

~

.,
X(z) = Lx(kT)z-k .......... (2.12)
l=O

= x(O) + x(T)z·1 + x(2T)z- 2 + ..... + x(kT)z-k ..........(2.13)

By multiplying both sides by z'-· 1 we get ✓

X (z) zk•I = x (0) zk-J + x (T) zk-2 + x (2T) i ·' + ..... + x (kT) z·1 + . . . . . . .. (2.14)
. ✓ ...,/

Let us integrate both sides along with the circle in counterclockwise direction,

{xcz)zk•I dz= {x(O)z k•l dz+ ... +{x(kT)z"1 dz+ .... .. ........ (2.15)

where c is a circle with its center at the origin of the z-plane such that all poles of X(z) zk•l
are inside it.

, Now, applying Cauchy's theorem, all the terms on the right-hand side of Eq. (2.15) become
zero except. /
{x(kT)z-:dz -/

{x(z)z'-'dz=~ . . . ..... . (2.16)

I
· 11 we o b ta1p
From w 111c
--
./ ·

.... . .. .(2 . 17) ·I

The coefficient x (kT) associated with the tenn 2· 1 is the residue.

The Eq. (2.17) is the inversion integral for the z transfonn, which is equivalent to stating that

\
38
 . x (kT) = L [ residues ofX(z) 2 k-l at the poles of X(z)].

Let us now evaluate residues:

(a) For a simple pole: If the denominator of X(z)zk•I contains a simple pole z = Zi then the
corresponding residue k is

k = lim [(z -zi) X (z) z k - l ~

(b) For a multiple pole: If X (z)l·1 contains a mult1ple pole Zj of order q , then the residue k
is given by ~

1\../"lim dq- 1
k = - -· __ r(z - Z )q X(z)z k-l]
(q-l)!Z ➔ ZjdZ~-•~ j
' .
•
2
2
Example 19. Obtain the inverse z transform of X(z) ~ T by using inversion
(z - 1) 2 (z - e-•)
~ integral method.

Solution. x (k) = :E residues of X(z)zk•I at the poles of X(z).

Here, X(z) l ·1 has a simple pole at z = e•aT and a double pole at z = 1.

- ,... . ...
:. X (kT) = k1 + k2

Now, the residue k 1 can be found as,

39
 at po Ic z =c -•T ]

k+l
k2 =residue of z at double pole z = 1
· (z-l) 2 (z-e-•T)
1
1 Iim d [ 2 k+ ]
~(2 - l)!z ➔ ldz (z -e-•T)
= lim [<z-e-•T)(k+l)z• -zk+']
z➔l (z-e-•T )2
k e-•T
=

Hence, ·

1
Example ~Qbtain the inverse z transfonn of X(z) = Oz by using inversion
v< (z - l)(z - 2)
• integral method.
Solution. Since X(z) zk-t has two poles at z = 1 and z = 2. So, there are two residues k 1 and
k2.
Now, to evaluate the residue k 1,
'

k, = Jim [IOz(z - l)zk-•]

z ➔ l (z - l)(z - 2)

lim [ 102 1c ]
- z ➔ I (z - 2)
=- 10

40
Similarly,

k = Iim [1 Oz(z - 2)zk-1]

z ➔ 2 (z-l)(z-2)
2

1
= z ~2[(~Q~:)]
= 1Q2k

Hence, x (k) = - 1.0 + 10(2k)

= 10 ( -1 + 2\ k = 0,1,2, ...... .

E x ~ (kT) by using the inversion integral method when X(z) is given by

Solution. x (kT) = I:[ Residues of X(z) zk-l at the poles of X(z)].

Here the poles are at z = 1, and z = e-aT

Now, let us evaluate k 1,

k = lim [z k(1 - e-•T)]

1 z➔1 z-e-•T
=l
and,

= -e - al:T
:. X (kT) = 1 - e-akT, k = 0,1,2, ......

41
 Table of /111porta11t Properties mu/ Theorems of the z-Tnmsform:

x(t) or x(k) Z[x(t) or Z[x(k)]

1. ax(t) AX(z) .
2. ax1(t)+bx 2(t) aX 1(z)+bX2(z)
3. x(t+T)or x(k+ l) zX(z)-zx(O)
4. x(t+2T) z"X(z)-zlx(O)-zx(T)
5. x(k+2) zLX(z)-z"x(O)-zx( 1)
6. x(t+kT) zKX(z)-zKx(0)-£1' · 1x(T)-... -zx(kT-T)
7. x(t-kT) z·Kx(z)

8. x(n + k) zKX(z)- z"x(O) - zK· 1x( 1)- . .. .. . - zx(k-1)

9. x(n-k) z""X(z)

1LJ.O: tx(t) d
-Tz-X(z)
dz
\.1-V kx(k)
-z~X(z)
dz .
J -2~ e·3~x(t) X(ze
31
)
·. i

13. e·aKx(k) I X(ze

3
)

14.-- al(x(k)
..__/

x(1) l

l.J5:' kaKx(k)
r -: z~x( ~)
dz a

16. x(O) Jim X(z) if the limit exists

Z➔OO

17. x(oo) Jim (( t-z·')X(z)] if (t-z· 1)X(z) is analytic on

z➔ l and outside the unit circle

18. Vx(k)=x(k)-x(k-1) ( 1- z·1 )X(z)

19. ti.x(k) =x(k + 1)- x(k) • (z - l)X(z) - zx(O)

20. n 1
I x(k) _
1
X(z)
k=O 1-z

21.
-
a x(t,a) a
-x(z,a)
aa aa
42
~22. k"'x(k)
.
(-z d~r X(z)
n
23. X(z)Y(z)
I x(kT)y(nT - kT)
k■O

24. .,
X(l)
Ix(k)
k ■O

2.6 z-Transform Method for Solving Difference Equations

Re,•iew: we have already derived

2:,.x (k- nl == z·" X(z) and

4. x (k + nJJ~ z" [ X(z)- ~ x(W' ]

Then x (k + 1), x (k + 2), x (k + 3), ...... and x (k-1), x (k-2), x (k- 3), ...... can be expressed
in tenns of X(z) and the initial conditions as follows:

4 4
(i) 2[x (k + 4)] = z X(z) - z x(O) -z3 x(l) - z2 x(2) - z x(3)

(ii) 2[x (k + 3)] = z3 X(z) - z3 x(O)- z2 x(l) - z x(2)

(iii) 2[x (k + 2)] = 1/ X(z) - z2 x(O)- z x(l)

(iv) 2 [x (k + I)] = z X(z) - z x(O)

(v) Z[x (k )) = X(z)

.__/

(vi) 2 [x (k - I)]= z· 1 X(z)

~

(vii) Z[x (k - 2)] = z·2 X(z)

'--...__,/

43
 (viii) 2 [x (k - 3)] = z·3 X(z)

(ix) 2 [x (k - 4)] = z-4 X(z)

Exnr~1plc 22. Solve the following difference equation by use of z transfonn method:
x(k f 2) + 3x (k + 1) + 2x (k) = O. Given that x(O) = 0, x(I) = I

Solution. As we know that

2[x (k + 2)] == z2 X(z) - z2 x(O) - z x(l)

2[x (k + l)] = z X(z)- z x(O)

2[x (k )] = X(z)

Now, taking z transfonn to the both sides of given difference equation, we get

<
2 ) 2 ,,
z X(z -z x(O)-zx(l)+3zX(z)-3zx(0)+2X(z) \ O _
,:':

or,
(z2 + 3z + 2) X(z) - z = 0
(putting x(O) = 0 and x(l) = 1)

z z z
X(z) = -z-2 _+_3_z_+_2 = -z+-1 - _
z _+_2 ,_ __

Taking inverse z transfonn, we get

x (k) = (-1/ ~ (-2/, k = 0,1,2, .....

. Example 23. Solve the following differential equation

x~k + zj.x (k +I)+ 0.25 x (k) = u(k + 2) wher~ x (0),; I, x (1) = 2, u(k) is input function
and is given by u(k) = 1, k = 0,1,2, ......

Solution. The given difference equation is

44
 x (k + 2) - x (k + 1) + 0.25 x (k) = 44-'(k +2)
By taking z transfonn to the both sides, we get
2 2
z 2 X(z)-z x(0)-zx(l)- zX(z)fzx(0)+0.25X(z)=z 2 U(z)-z u(0)- z u(I)
2
or, z X(z) j - ii- zX(z) -i+ 0.25 X(z) = z 2 U(z)-/ /
I I ~~
or, X(z) [z 2 - z + 0.25) = '.l§'A- z2 _z_ "I
7· z- ,

=
z-1
2
X(z) z +2z-2
or, . - =
z (z - 1)(z 2 - z + 0.25)
X(z) z 2 +2z-2
or, --=
z
(z-l{z-H

Let us now expand X(z)/z into partial fraction,

To evaluate k1, la and la, we get

3
kl =4,k2 =-3,k3 =2
4z 3z 3 z
So, X(z) = z - 1 - z - .!_ + 2 (z - !)2
• 2 2

On taking inverse z tr~~~'- ·•

•

x(k)=4-3(
l)k 3 (1 B a,
+ k ) -y;?',,,
2 2
2
Wherek = 0, 1, 2, ....

45
2.7 Real Convolution Theorem
Let us consider the functions x 1(t) and xi(t), where
x1(t) = O, fort<O

for l < 0

Ass umc that x,(t) and x,(t) arc z transfonnablc and their z lransfonns arc X,(z) and X,(z,(
respectively. Then

X,(z)X, (z) = 2[t,x,(hT)~,(kT- ~!)] .......... (2.18)

=2[x,1(k) * X2(k)] ,
This equation is called the re~] convolution theorem.

Proof: As we know that

ao k
=I Ix,(hT)x 2 (kT-hT)z-k QI '
k=O h=O
"" k
=L Ix,(hT)x 2 (kT-hT)z-k ..........(2.19)
k• O h• O

Where the condition has been used is x2(kT - hT) = 0 for h > k. Now, let us define ':1 = k - h
such that when k = 0, m = -h and when k = ro, m = oo. Then

z[t, x, (hT)x, (k~ - hT)] = t. ;'.:~T)z-•J: ;(m:)z-m .......... (2.20)

Since x2 (mT) = 0 form< 0. Hence we can write,

., .
= 2:X, (hT)z-b Ix(mT)z-m
hs O m=O J.. .
= X, (z)X 2 (z) ······ ····<2.21)
46
2.8. Complex Convolution Theorem
.,
The complex convolution theorem is useful in obtaining the z transform of the product of two
sequences x1 (k) and x2(k). Suppose both x,(k) an'd x 2(k) are zero fork < 0. Assume that

X1(z) = 2[x1(k)], lzl > R 1

X2(z) = 2[x2(k)], Jzl > R2

where R1 and R2 ~e the radii of absolute convergence for x,(k) and x 2(k), respectively. Then
the z transform of the product of x 1(k) and x 2(k) can be given by

2[x, (k)x 2 (k)] = ~

2
ff X 1
2
(o)X 1 (b'- 1z)d8 ......... (2.22)

Where R 2 <Io I< I; I

1

Proof: As we know that the z transform of x 1(k) x2(k) is,

..,
2[x 1 (k)x 2 (k)] = L x (k)x (k)z-k
1 2
. ........ (2.23)
k• O

The series on the right hand side of Eq (2.23) converges for Jzl > R, where R is the radius of
absolute convergence for x,(k) x2(k).

From invci:_sion integral method,

X2(k) = 2~ {x2(z)zk-ldz• .: . •

= _l_ Jx (0)0 1- 1do ....... .. (2.24)

2ffj 1: 2

Substituting x 2(k) from Eqs. (2.24) into(2.23), we get

47
 ......... (2.25)

Noting that Eq (2.23) converges uniformly for the region lzl > R, we may interchange the
order of summation and integration, then Eq. (2.25) can be written as,

........(2.26)

Since,
,0

Ix, (k)(o-1zrk = X1(8-1z)

k• O

We have,

.. . ...... (2.27)

Where c is a contour (a circle with centre at origin) which lies in the region given by
1
/81 > R2 and 18· zl > R1 or

......... (2.28)

2.9. Parseval's Theorem

The inequality derived in the previous theorem is

1
If this inequality satisfies for I z I= 1 or R 2 < 181 L -
R,
Then by substituting /zl = 1 in Eq (2.27) of the previous theorem, we get

48
Let us set x 1(k) = x2(k) = x(k) then

2
2[x (k)]= _l.
. 2!()
f>-X(8)X(c5 - )dc5
1 1

1
= - -. rfz- 1 X(z)X(z- 1)dz ......... (2.29)
27rj 1: . :
or,

. .. .......(2.30)

This Eq. (2.30) is known as Parseval's theorem and it is useful for obtaining the summation
ofx2(k). ...,,,-

2.10 Some Examples

Example 1. Obtain the z transform of k 2 .~-

.
Solution. By definition of z transform,

2 [k 2 ]=f k 2
z -k =
z-
1
+4z-2 + 9z- 3 + .....
k; O

=z-1 (l + z - 1)(1+3z-1 +6z- 2 +lOz- 3 + ......)

z- 1 (l+z-1 )
= (1-z-')3

49
Example 2. Obtain the z transform of the curve x(t) shown in fig. 2.5. Assume that the
sampling period Tis 1 sec.

x(t)

1.0

5 6 7 8•

Fig 2.5
( Curve ofx(t))
Solution. From the above figure we obtain
x(0) = 0 .
x(l) = 0.25
x(2) = 0.5
x(3) = 0.75
x(4) = t,' k = 4, 5, 6, ....... .
\

Then the z ~ransforrn of x(k) can be given by

00

X(z) = Ix(k)z-•·

=0.25z -I + O.5z- 2 + 0.75z - 3 +z -4 +z- S +z -6 + .....

1
= 0.25(2-1 + 2z- 2 + 3z-3 ) + z-4 - -_-1
1-z

We can verify this by the following way. As we know that the curve x(t) can be written as

50
 x(t) = ¾t -¾ (t - 4)1(t - 4) ~ '\>
where 1(t -4) is the unit-step function occurring at t = 4. Since the sampling period T = I sec,
the z transform of x(t) can be given by,

X(z) = 2'[x(t)] = z[ z[
¾ t]- ¾<1 -4)1(1 -4)]

=
4(1-z-1)2 4(1-z- 1) 2
. = 1 z-• (l -.z- 4 )
4 (l-z-1 ) 2

(same result as obtained earli~r).

2
Example 3. Obtain the inverse z transform of X(z) = z(z + 2) by use of all the methods
(z-1)
described in this chapter.

Solution.
Met/rod 1 (Direct dfrisio11 met/rod): Let us write X(z) as a ratio of two polynomials in 2· 1.

X z _ 1+2z-1 _ 1+2z-•
( ) - (}
-z -1 )2 - 1- 2z-I +z -2

Dividing the numerator by the denominator, we get

Hence,
x(0) = 1
x(l) = 4 • . ._ . I,
·• .
x(2) = 7
x(3)= IO

51
 M~I 2 (Parli[1l-fractio11-expa11sio11 met/Jot/): Let us expand X(z) into the following
r,(rtial fraction,

y, ( 1.,./ -

'"2

Now,

Hence we obtain,
x(O) = 1
x(k) = 3k + 1, k = 1,2,3,.......

from which we can combine as

. x(k) = 3k + 1, k = 0,1,2, ....... .

Metltod 3( /11versio11 integral method):

~ r e , we can write

X(z)z1:-1 = (z + 2)~1:
(z-1) ·

• Since X(z) l· 1 has a double pole at z = 1. So, we can find x(k) by,

. (z + 2)zk ]
x(k) = residue of at double pole z = 1
[ (z-1) 2

52
o.

x(k)= l lim ..5!_[ (z - !)2(z + 2)zk]

(2-1 )! z ➔ I dz (z _ I )2
lim d r
=z➔ ldzl(z+2)zk]
=3k + 1, k =0, I, 2, .......... .

E~le 4. Consider the difference equation ~

x (k + 2) - 1.379 x (k + 1) + 0.3679 x (k) = 0.3679 u (k + 1) + 0.2642 u (k)
where x(k) is the output and x(k) = 0 fork ~ O and u(k) is the input and is given by
u(k) = 0, k<0
u(O) = j L
u(l) = 0.2142
u(2) = - 0.2142
u(k) ·= 0, k = 3,4,5, ..... .
Determine the output x(k).

Solution. By taking the z transfonn .,of the given difference equation, we obtain
...,
0

- z x(l◊)"-
2
[z X(z) - z2 x(Of
, 7
I .3679[z X(z) - z x(O))
I
+ 0.3679 X(z)
= 0.3679
•
[z U(z) - z u(O)] + 0.2642 U(z)
.
...... .... (2.31)

By substituting k = -1 into Jhe given diffe~ence equation, we find .,

x( 1) - 1.3679 x(O) '~ 0.3679 x(;
- /
1)= 0.3679. (0) + 0.2642 u(:1)
/'

Since x(O) = x(-1) = 0, u(-1) = 0 and u(O) == 1, we obtain

x(l) = 0.3679 u(O) = 0.3679

By substituting the initial data

x(O) = 0, x(l) = 0.3679, u(O) = 1 into Eq (2.31), we get

z2 X(z) - 0.3679z - l .3679zX(z) + 0.3679:X(z) = 0.3679U(z) - 0.3679z + 0.2642 U(z)

53
Solving for X(z), we get

X(z) = 0.3679z + 0.2642

2
z - l .3679z + 0.3679 U(z)

The z transform of the inp_ut u(k) is

U(z) = 2[ u(k) ] == 1 + 0.2142 2· 1 - 0.2142 z·2

Hence, \

36792
2642
X(2) == 2
0. +0. (1+0.21422·1 -0.21422-2 )
2 - l .3679z + 0.3679
_ 0.36792- 1 +0.34302- 2 -0.0222lz-3 - 0.05659z- 4
1-1 .36792-1 + 0.36792- 2
=0.36792-1 + 0.84632- 2 + z-3 + 2 - 4 + 2 -s + ......

Thus, the inverse z transform of X(z) gives

x(0) = 0
x(l) == 0.3679
x(2) = 0.8463
x(k)== l;

~
Example 5. Consider the difference equation x(k +2) = x(k + 1) + x(ky<
where x (0) = 0 and x (1) = 1. Note that x (2) = I, x (3) = 2, x (4) = 3, ..... The series 0, 1, I,
2, 3, 5, 8, 13, ..... is known as the Fibo_nacci series. Obtain the general solution x(k) in a
closed form.

Solution. By takingJhe z transfonn of this difference equation, we get

~ ..
z2 X(z) - z 2
x(0) - z x(l) = z X(z) - z x(0) + X(z)

Solving for X(z) gives

X(z) = z x(0) ~ 2\( I) - zx(0)

2

z- -z - 1
Now, by substituting the initial data x(0) = 0 and x(l) = 1 into this last equation, we have

54
 X(z) = 2 z
z -z-1

1 z, ,.. z
- ✓5 _1_+_✓
-=-
5 ~ 1- ✓5
z- - z- -
2 2

l 1 - I
= ✓§ 1- 1+ ✓5 z-1 ~ 1- ✓5 - I
2 z- - 2- z

The inverse z transfonn of X(z) is

x(k) ~ l[(1+/5)\(i-tr]
k = 0, 1, 2, ........... .

2.11 Problems

1. Obtain the z transfonn for the following functions.

2
\_~ (t) = t (b) x(t) = t4
(c) x(t) = teat (d) x(t) = t3 e•at
V
2
.(e).,... x(t) = t e•at (f) x(t) = 1 .. e•at
(g) x(t) = e •at - e · bt (h) x(t) = (1 - at) e•at
(i) ~(t) = at - 1 + e•at
..(j) x(k). = ak•t, . k = 1,2,....
(k) x(k)'= k ak•I (I) x(k) = ~2 l.1
✓

(mf x(k) = ak cos k1t (n) x(k)-= k3 ak•t

•
Ans.
(a) T 2 2·1 (1 + 2· 1) (b) T4 2· 1 (1 + 11 z· 1 + 11 z·2 + z· 3 )
(l -z- 1)3 (1 - z•l)5

55
 ~T .\ ,2aT
( ) ..L--J'-----

(c) T2 c•aT (1 + e•aT z·') z·l (f) ( l - e•aT ) z· 1

(1 - e•aT z·1)3 (1-z"1)( 1- e•aTz· 1)

(i) [(aT - I + e·aT) + (1 - e•aT -aT e•aT) z·l] z•l G) ..L

(1 - z·1)2 (1 - e·aT z 1) !-a z·l

-I

(k) (l-:z-1)2

1
(m)} -I
-az

2. Find the inverse z transform of the following:

z 1
(a) (b)
z+a z+a

1 3z 2 + 2z + 1
(c) (d)
z 2 -3z+2
z-a
1
3z 2 + 2z + 1 2z
J) 2
z +3z+2 ~ (2z - 1) 2

(g) z-0.4
i 2
z +z+2

56
 z -4
(i)
(z -l)(z -2) 2

Ans.

(a) (-a/
(b) (-af 1u(k-l)
(c) ak•1u(k-1) 1
(d) 38(k)-6u(k-1)+17(2f (k-1)
1
(e} 38(k) + 2(-lf u(k-l)-9(-2f 1u(k-1) (f) k(l/2/

(g)- ~
✓7
(-✓2/sin0k + ;(-✓
0
v7
2)k-i sin0k-1u(k - 1)
0 =tan ·1 ✓7 •

1
(h) - ka k
a

(i) - 3u(k - 1) + 3(2)•-, u(k-1) - (k-1)(2)k-l u(k-1) ~

c . - 0btai n the solution of the following difference equation in tenns of x(O) and x( I):
~ k+2) + (a+b) x(k+ l) +ab x(k) =Owhere a and bare constants and k=O, l , 2, . ...
Also assume case (i) a=b; case (ii) a;cb.

Ans. (i) x(k) = x(O) (-a/ + [ax(O) + x(l) }<(-af'

(ii)x(k)= bx(O) + x(l)(- a) k + ax(O)+x(l\-b)'
b-a a- b

4. Solve the difference equation x(k+2)+3x(k+l)+2x(k)=u(k); x(O)=l, x(k)=O fork< 0.

Ans.

x(k)=!_I(-l)k +2(-2/
6 2 3
5. Solve the difference equation

x(k + 2) - 3x(k +I)+ 2x(k) =4 k ; x(O) = 0, x(l) = 1

57
 CHAPTER - II I

Z-Plane Analysis of Discrete-Time Control Syste1n

Yimpulse Sampling

Let us consider a fictitious sampler commonly called an impulse sampler. The output of this
sampler is considered ,to be a train of impulses that begins with t = o, with the sampling
period equal to T and the strength of each impulse equal to the sampled value of the
continuous-time signal at the corresponding sampling instant. A pictorial diagr~ jmpulse
sampler is shown.in figure.
'd\.Lt) x•(t)

. 1

I •
~---x-*(-t)--•

X(s)
X*(s)

Fig 3.1 (Impulse Sampler) ~ \'\-4 t ( -ie.JJ,J • l,

(Mathematically, an impulse is defined as having an infinite magnitude with zero width, it is
graphically represented by an arrow with an amplitude representing the strength of impulse)
- - ~ - - - - - " - - - - --

The impulse-sampled output is a sequence~ f ir!!p~ses with the stren th of each impulse
equal to the magnitude of x(t) at the corresponding instant of time. That is at time t = kT, The
impulse is x(kT) 8 (t-kT). The notation to represent the impulse-sampled output is x•(t) . The
sampled signal x•(t), a train of impulses, can thus be represented by the infinite summation.

x * (t) =
-
I x(kT)b'(t - kT) -
k=O
l\ I -Y.J
or, x * (t) = x(O)o(t) + x(T)b'(t - T) + ..... + x(kT)b'(t - kT) + ....... ......... (3 .1)

#- ruik. -kAJ- ~ li-- k i) :; () ''-v-tvU1M

. I. t ::. k: 1

58
 Let us define a train O~J?rnpulscs as Ot(l).
or,
CX)

"r(t) = Io(t- kT)

k=O
Hence, the sampler output is equal to the product of the continuous-time input x(t) and the
train o~pulscs 8r(V

The sampler may be consi~ercd a modulator with the input x(t) as the modulating signal and

~ I imGs; 6 t ~sthe c~::(~: shown in f i : r t r t it

,____ V ,---1-----.Carrie .f i
- - - + - -- ..i Modulator
\
x(t) _ - -- x*(t) Output
Modulating Signal

Fig 3.2 (Impulse sampler as a modulator)

Taking the Laplace Transfom1 of Eq.(3.1), we get

..,-- -
• X * (S) =..L'[x *(t) ]=x(0)..L'[o(t) ]+ x(T)..L'[c5(t-T) ]+ x(2T)..L'[o(t-2T) ]+ ........... .

= x(0) + x(T)e -Ts + x(2T)e-2Ts + ........

-
= I x(kT)c-kTs -=- . .. ........ .... . (3.2)
k=O
Let us define e Ts = z
1
s= - ln z
T o,n
- f ><._ ( '.<.T)
This Eq.(3 .2) becomes K- = D ~ K'

ex> -k
X*(s)
1
= ~x(kDz ··········· ··········<3.3)
s =- lnz k-O
- -- .!.--" T

59
 The Eq.(3 .3) is the Z-transfo1m of the sequence x(O), x(t), x(2T) ...... generated from ~(t) at t =
kT, k = 0,1,2 .......... .

Hence we may write,

X*(s) =X(z) ................. (3.4)

1
s = - lnz
. T

Hence we can conclude,

X*(s)=X*(..!_1nz)==X(z)= Ix(kT)z-k
T k==O

R.21 Data:Hold Circuit

/2.~-hold. is a_ process of generating a coutinuous-timc signal h(t) from a discrete-time
sequenc~ x(kTI. A hold circuit converts the sam led signal into-2..SQ!ltinlliLtJs-tin~ si 0 nal,
which approximately reproduces the signal applied to the sampler.

The signal h(t) during the time interval kT~ t ~ _Q0:l)T may be approximated by a

The signal h(kT) must be equal to x(kT)

-, -__
or,
h(kT) = x(kT) 4 'l o._o () wt-
. •b' c;-- -;:IJ . IC--T ~ VJ ,::
• )., ( t'
Hence,Eq.(3.5)canbewrittenas, ,l {plfrri 3·.~;~!-.: .... .. ,.· ·; •• •• • ,se.-fd. h( ~r ) -:::)l lr-:-t J
. • ... __. _ · • . ~a , t_(Y-.~l -;:- ~IJ:
- • (·k.T) ,
• h(kT +,)=an-r n +an_ r n-1 + ....... +:, r+x ...... •........ (3 .6)
1 1

it is called a ~ T'
--
If the data-hold circuit is an nth-order polynomial, it is called an nth-order hold. Thus if n= 1,

The simplest data-hold is obtained when n = 0 in Eq.(3.6) that is when,

60
 h(kT + T) = x(kT)

3.2.1
r.
\

Proof

x(kT h(t)
x(t)
I
✓
l

kT t
(a)

_:_:___,/
x(t) ~. ;--· _Zero-order
x_(k_T)---4--1. _ _ ___, ►
h(t)
Hold
Sampler
(b)

Fig 3.3 (Sampler and zero-order hold) J

The figure above shows a ~ and a zero-order hold. The input signal x(t) is sampled at

------ • • -- - -- - ' \ t " I

discrete instants and sampled signal is passed through the zero-order hold. The zero-order
hold circuit smoothes the sampled signal to produce the signal h(t), which is constan.t from
theUast sampled valu"s)until the next sample is available, that is

h(kT + t) =x(kT), forO 5t < T ....................... (3 .8)

.
shown below in Fig. 3.4(a). As we know that the integral of an ,,impuise function is a
constant, we may assume that the zero-order hold is an integrator and the input to the zero-
order hold circuit is a train of impulses. Then a mathematical model for a real sampler and

61
zero-order may be constru t d
(~ / l- T) :
l.
Q~(

.
1
I l{; ( <
~ ·c·,,
vr
I
e e as shown in Fig. 3.4(b), where Gho(s) is the transfer funetion
of the zero-order hold and x"(t) . . •
' is the impulse sampled signal of x(t).

h(t)

x(t)
__/ x(kT) Zero-order \
.....
h1(t)
T hold ',
v 'Y..' l - V

r
I

(a) ~ . ) t

x(t)
h2(t) I
I
I ...
>
( l- . '2 •"'
_ _ / r-x*(t) ZOH - -l Y--)

or X*(s) Gho(s)
H2(s) ' I

(b) .
Fig 3.4, (A real sa pler, zero order
hold and'o)mathematical model that
consists of a!_l_im ulse sam ler and
transfer function Gh0 (s))

Consider the s:ampler _m=td zero order hole-&hewn-in Fi~{3 .4 (a) Assume that the signal x(t) is
...
zero for t<0. Then the output h1(t) is re1ated to x(t) as follows:
. .J
•"
' , ..).,_i;; I '
~l , "i I
/

1/
=
CX) ~ s~----- --,
hl (t) = x(O)[l (t) -1 (t - T)] + x(T)[! (t - I) -l(t - 2T)] + x(2T)[l (t - 2T) - I (t - 3T)] + ......

Lx(kT)[l(t-kT)-l(t-(k+l),-ij (_ lv) Q ................ (3.9)

k=O/ d
,,.
Since,
- - - - k T s- \
,,t[1 (t - kT)] = e s j IP
The Laplace transfonn of Eq.(3.9) becomes,
....
-- '
[ ] 00 ---- ; =-kTs -e-(k+l)Ts
,ilhl (t) = H (s) = k~O x(kT)---s---
1 I
62
 1-e-Ts o:i
= L x(kT)e - kTs .............. (3 .10) "
s k=O
✓

1ow
'
consider the mathen1 r
a ica1 model of Fig. 3.4(b). The output of this m~del must be same
as that of r~zer~or er ho~ .

oL'h2 (t) ==H2 (s) = Hl (s)

-- 1-~-Ts CX)
Thus, H 2 (s) = _ _ tx(kT)e-kTs - ....................... (311)
s k=0 · ' ~ ·
From Fig 3.4(b), .
..y'
H~s) = Gh 0 (s)~*J;?
. , ;;' kT 7
Smce X*(s)= Ix(kT)c- s j
k=0
Hence from Eq. (3 .11 ),

1 -e -Ts
H 2 (s) =--X*(s) ................... (3.12)
s

Hence the transfer function of the zero-order hold may be given by,
-7'
-Ts 1 •- I

G (s)= -1 -_e_
ho s .,, ~

• 3.2.2 Transfer Fu11ctio11 of First-Order Hold: 66 \:::, c.. ·<0.

5
Theorem: The transfer function of the first-order hold may be given by
2
G (s)=(l-e- TsJ Ts+l ... ............. (3 .13)
hl s T

Proof- The. output of the nth order hold circ~it is given by

j(kT +,)=an -r 11 + a _ ,n- l + ....... + a -r + x(kT) ....... :-. ., .......... (3.14)

11 1 1
Let us now substitute n = 1 for the first order date-hold, we get

h(kT + -r) =a 1r + x(kT) ...................... (3.15)

Wheref ;5; t < ~and k = 0, 1, 2, ......... .

63
By applying the condition that
h( (k -1 )T) == x( (k -1)T)
. l:J
The constant a 1 can be detcmlincdr• .,._,
as 1011 ows: .,.
h((k-l)T)= - a T+~(kT) - ((k "'- "
1 - x - 1)T) ,,( )
.J..- /

or, a1=x(kT) - x((k- l)T) <J;/

- T 4f "-'

Hence, Eq.(3.15) becomes ,- ,,( I 1.

0 -" •
h(kT + ,) =x(~T) + x\k1)- x((k-:- l)T.) i, •. ' 't ,

'<. T • ,,, ~ ··········· ..... (3.l6)

Whcr~ 0 $ t < T. The extrapolation process of the first order hold is based on Eq.(3 .1G). The
continuous-time output signal h(t) obtained by use of the first-order hold is a piecewise-linear
signal, as shown in figure below.

x(kT h(t)

,,,,.....

T '2.T '.f f
-, T
~
~
x(kT) First-order h(t)
hold ►
I
'
Fig 3.5 (Input and output of a first• order hold)

To derive the transfer function of the first-order hold circuit, it is convenient to assume a
•·
simple function for x(t). For example, a unit step-function, a unit-impul~mction, and a
unit-ramp function would be the g~hoices for x{t)._
~

64
 TTE an' ft 11 1 ft C ::>Gi & &Al=-

11
--· - - - - - - ; . r I-() ' I Id
LetU s choose a unit-step function as x ). Then, for t~cJ!i!!~plci:Jmd first-order ,o
shown 1~ 6(a), the output h(t) of the first-order hold consists of straight lines that arc
extrapolations of the two preceding sampled values. The output h(t) is shown in the diagram.

The output curve h(t) may be written as follows:

................ (3.17)

}-e·Ts 1-e·Ts
= - - + - 2-
s Ts
=(1-e· Ts) Ts+l
Ts 2

x(t) x(kT) 'l.--h(t)

•
I
/

L
,, /

0 "T 2."T 3T LtT r- -T T 2T t

t ' t

~ x(kT)
►I
First-order
hold
h(t)
►
I

(a)
?.. . ')

1,G"\ :.. t + \ ~ l 1>.. l\ -

(.1_,\ - L (· -
~ t-1 ~ )'~C•) - -r 65

• ( I. 1) ' ~~ - (t:.. :q "\ H-- )

\__ - - ~ - - J
... --1 l~ ., J
 x(t)
h(t)

-T T 2.T

h(t)
►

(b)

Fig 3.6( Real sampler cascaded with first-

order hold and mathematical dodel
consisting of impulse sampler and Gh I (s))

The Fig 3.6(b) shows a mathematical model of the real s a ~

__sascaded with the fi rst-order
hold shown in Fig 3.6(a). The model consists o the impulse sarnpJec:.arr'd Gh1(s), The transfer
function of the first-order hold. The output signal of this model is the same as the output of
the real system. Hence, the output H(s) is also given by Eq.(3.18).
~

The Laplace transfonn of the input x*(t) to the first-order hold Gh1(s) is

X*(s)= ~l(kT)c-kTs =
~
1
I - e-Ts
~}10-.,7Vt-,zr- (!,.'l) t-i'\J.-1 x_0:.1): l ./ ~ ;{ ~ -ir
k•O

Hence, the transfer function Gh1(s) of the first-order hold is given by

Gh (s)= H(s) =(l-€1s)2 Ts+l

1
X*(s) Ts 2

• ~c-:~
n)' T~+l

Note that a real sarnpl~r combined· with a first-order hold is equivalent to an impulse sampler
combined with a transfer function

(l-e-Ts) 2 (Ts+l)
(Ts 2 )

66
3.3 Obtaining the Z-transf01.
. • rn by the Convolution Integral
Let us consider the impulse sa 1
mp er as shown in figure below:

x*(t)
_x_(t_)_ _--'-/
- - - - - -?

Fig 3.7(Impulse sampler)

The output of the impulse sampler is given by,

x * (t) = L"' x(t)8(t - kT) = x(t)Io(t- kT) .............. (3.19)

k=O k=O
- -x-
As we know that

..t[o(t - .kT) ~ =e••n

We have,

..t[fo(t-kT)] = l+e-Ts + e-2Ts+ .......... = 1T

k=O (}-e- 5)

Here, X*(s) is the Laplace transform of the product of two time functio!} (t) and
"'
- - - --
Z:o(t-kT) . Note that it can't be equal to the product of the t~ o corresponding Laplace
l•Q____-
...
transforms.

The Laplace transform of the product of two Laplace-transformable functions f(t) and g(t)
can l?e given by,
·"'
J
.l[f(t)g(t)] = f(t)g(t)e-s'dt
0

J c+joo
= -? • fF(p)G(s-p)dp ............ (3.20)
-trJ c- j:o

67

j
Note that inversion integral is,
c· -
1 c+jco
f(t) = - . J F(s)e st ds, t>0
27rJ c-jco
..---

Where c is the abscissa of convergence for F(s). Thus,

l .., c+ i..,
Jf
.1 f(t)g(t) = ~ F(p)eP'dp g(t)e-s1dt
7rJ O c-joo

Because of the uniform convergence of the integrals considered we can invert the order of
integration.
1 c+ jco oo
.1 f(t)g(t) =-. J. F(p)dp Jg(t)e-(s-p)tdt.
21lJ C-JCO Q
.,
Noting that fg(t)e-<s-p> dt =G(s-p)
0

We obtain
. 1 c~j,o

,,f f(t)g(t) =~ fF(p)G(s-p)dp

1lJ c- joo

I o(t-kT)
,0

Let us substitute x(t) and for f(t) and g(t) respectively. Then the Laplace

transfonn x·cs) is,

X *(s) =.t[x(t) Io(t-kT)]

k=O

1 c+j., 1

= -2. f X(p) 1- - T(s-p) dp ................. (3 .21)

ilJ c-j,o e
'Where the i!}tegration is along the line from c-joo to c+joo and this line is parallel to the

imaging axis in the p-plane and separates the poles of X(p) from those of )rcs--p) . The
1
Eq.(3.21) is known as convolution integral. Such an integral can be evaluated in terms of
residues by forming a closed consisting of the line from c-joo to c+joo and a semicircle of
infinite radius in the left or right half-plane.

68
 JJ,l Co111•olutio11 futcnmt •
0 11111
ie l eft Ila ff of the p-J>lrm e:
,,..
)(

)(

)(

)(

rI
X
)(
X

0 nc
X
X
◄ ·c ...
X
Poles of
X(p) X

)(

X
A!. we have, Fig 3.8 (Closed contour in lhi.: lcfl half of 1hc p pbnc) )(

I r +i., ~

X * (s) = -2 .
m
J X{p) 1- e-1_
. ll>- 11)
ciJ)
1
r oles o ( ----::c---;
1 - t Th . pl
· :, <'-J "

=1- cf x(p) 1 •d) - -

I X(p) d)
'>m'
- · :1
I- c · T(S-J>) f
I 211].· 1- c - T(s-11) I
'L

He1 l! we asslimc that X(s) ~ q(s) and p(s) is of higher degree in s than q(s), that is
p(s)
lim X(s) = O
s- •o,

Usi ng the con lour shown above, along the sc111icircle, as s - >oo , the integral->0, that is
1
X * (s) = - -cf X(p) d1
2Rj I - C- T(, - 11) I

1
=~
""'[Residues of I X~~ ) at the poles of X(p)
- c -1t--,,1

f ,clli g z = cTs

--
X(z) = I[rcsiducs of t -(~{~ at the po !cs of X{p)]

69
 By changing the complex variabl . .
' e notation from p to s we obtain,
X(z) = I[rcsiducs of X(s)z ]
~
z-e at the poles of X(s)

Case (i) For a simple pole: If a pole at s=s; is a simple pole then the corresponding residue kJ
is

ki = lim[(s-s .) X(s)z]
S➔s, J z-eTs

Case (ii} For a multiple pole: If a pole at s = s; is a multiple pole of order ni, then the residue
ki is

1
1
k . = - - - lim d n.t -1 [ s-s ni X(s)_z
(n . - l)!S➔s. n.-1 ( i) ,r,_ Ts
l
1 1 ds 1 ~e
~

Example 1. Given

;::-- X(s)= , I
s-· (s+l)
Obtain X(z) by use of the convolution integral in the left half plane.

Solution. The function X(s) has a double pole at s= 0 and a simple pole_at s= -1
Hence,
.ft
X(z) = I[resi uf of ~ ~t the poles of X(s)]

•- -1- lim [d - 1 L--z ] . [ 1 z Ts ]

-s-:--? - - c - - - - ~ + ltm (s+l)
- (2-l)!s➔0 ds ; ·(s'+-l)t'i-eTs/ s➔-1 • s2(s+l) z-e
.-. . -z[z-e15 +(s+ l)(-T)e 15 ] z
- ltm _.!,___--:-..:.._---;;:,-----;;--~
2 15 2
+-1- /
2 T
- s-o (s+l) (z-e ) (-1) z - e·

-z(z - 1- T) z z 2 (T-1 +e-1 ) + 2(1-e· 1 -Te-1 )

= 1 + -T =
(z-1)· z-e (z - 1) 2 (2 - e· 1 )

=
·-
70
r,x:m1plc 2. Given X(s) = s+3
(s1 l)(s -12)

Qb.riin X(z) by the method of convolution integral in the left half plane.

Solution: The function X(s) has poles at s= -1 ands= -2

Nov:,

X( z) = ~[ • Iucs o f -X(s)z
L., res1( -.-. al the poles of X(s) ]
z-e ••

.
s-►-t,§"'1
= 11111
(~.+"l)(s + 3)
.
z .
+ 11111
(s + 3)(,s,+"2). / z
)(s+ 2) ,. -cTs] s+2Ls+ I)(s-+ 1) z - cTs
•
]
z
= --_-=-
T
z- e
3.3.2 /11teg~a/ ill tlte Right I/tiff Pla11e: i_.

l(

0 Re
X
X

roles ol
Xlpl

~ . .
Ii ;,• ,: ....~•-i-:~:_/ -•
.
..... , ..

• rig 3.9 (Closed con1mir in ihc ri~ht half of the p plane)

1 .
·oles
TllCµ• of _
l - e -11,-,,) may be obtained by solving

71
--
e - T(s-p) =l
- T(s - p) = ±21r jk
1 .
s - p = ±- 2iZJk
T
.21C
P =s ± J -T k' (k =0, l ' 2,3)

Thus there is an infinite number of poles.

In order to _evaluate the given convolution integral, let us choose the contour which cor_1sists
of the line from c-joo to c+joo and semicircle •R of an infinite radius in the right half of the p-
1
plane as shown in Fig 3.9. The closed contour encloses all poles of but it docs not
}-e -T(s- p)

enclose any poles of X(p).

Now, x•(s) can be written as,

I C• j«> X(
X*(s)=- J p) d
2m . }- - T(s-p) p
•:, C- J,o e

_ _ l ,f X(p) d
- 21(J. 'jl - e -T(s- p) p - -2
1
.
J1 X(p) -T(s-p)
d
p
1(J -e
'•

Let us consider the integral along the semicircle •R· Consider the degree of denominator of
X(s) is at least z greater than the degree of numerator, it can be shown that the value of
integral is zero, or

_I J X(p) -O
21(J• 1 -T(s-p) -
-e
'•

Therefore,

1(J -e
f
X*(s) =-21. 1 x~~~-p> dp

The integral along the closed contour can be obtained by evaluating residues. Hence,

72
 X * (s) =

P• s+ j( ~ )
T
1 .,
- ~X( .21r
- TLJ s+ J-k)
k-o:, T

X(z) = [.!_
T
f k-ro
2
X(s + j 1r k)] -
T 1
s• - lnz
T

J ,_/ Jc--
Obtailliug Z Tra11sforms of F1111ctio11s J11vofl,i,zg the Term (l-:c-T )/s :
5

~ . .
S'uppose the transfer function G{s) follows the ZOH. Then the product of the transfer
function of the ZOH and G(s) become
• 1-e-Ts
X(s) = --G(s)
s
Now, the function X(s) can be written as

................ ······ (3.22)

Where I
I

G (s) = G(s)
I S

Consider the function,

.................... .. (3.23)

The inverse Laplace transform of Eq. (3.23) can be given by convolution integral,
... - "" . . . ·, . .
, _::-·,.,- -x 1 (t) = Jg;;(t - r)g 1 (-r)dr
0

\Vhere go(t) =.,[,-le-Ts= 5(t..-!f)

g1(t)=.,C-1G1(s)

73
Thus,
I

X 1 (l)= Jo(t-T-r)g 1 (r)dr

0

By letting,

2[g 1(t)] = G1(z)

2 [x1 (t)] = 2[g1 (t -r)] = z-1G (z)

1
Referring to Eqs. (3.22) and (3.23), we get
X(z) = 2[G 1 (s)-X 1 (s)]

= 2[gl (t)]-2[x 1 (t)]

= G 1 (z)-z-1G 1(z)
= (1 -z-1)G 1(z)
=(l-z-1)2[G 1(s)]

= (1- z-1 )2[ G~s)]

l
···· ·· ······••••(3 .24)
I
I

Hence, we can conclude that if X(s) involve a factor (1-e•Ts) then, in finding the z transf~rm
of X(s), the term (1-e·Ts) = (t-z·1) may be factored out so that X(z) becomes the product of
(l-z" 1) and the z transform of the remaining term.

Example 3.. Obtain the z transform of

1'- e-Ts l
X(s) = · -. -
,s ~

X(z) = (1 - z-1 )2[ (s)'

·- _ ( s

74
 _(1-c - ')z-•
) - C - T i'. - I

3.4 Reconstructing Original Signals from Sampled Signals

Stimpliug theorem: Let us assume a continuous signal x(l) has lhc frequency spectrum as

sh.: wn in figure below. This signal docs nol contain any rrcqucncy components ·above 0>1

rucVscc.
I X(iw l I

- r
I
1
"'
- w, 0 ,.,, w

Fig 3.10 (A frequency spectrum)

The theorem states lhat, " If w = '21r , where Tis lhe samI)\ing period, is greater than 20>1, or
• T

wl •re 2w 1 corresponds to lhe rrcquency spcctr!llll of the continuous signal x(t). lhen the

si~.nal x(t) can be rcconslructcd completely from the sampled signal x" (t)".

I\; w,: know thal the Laplace transform of the sampled signal x*(l) is _given by,

.................. (3 .25)

75
.-

Uy substi tuting s = jtl) we w·111 ,

• ' gel l11e frequency spectrum as below

X * Urv) = Tl
l
f XUcv + j cv k)'
"' \ I

The :o llowing Fig 3.1 1 shows the plot of IX * Ucv)I versus (J) .
I x·1,.,,1 I

Pig J.11 (l'lut nl'thc f'n.:qucm:y spc1:trnm I >..:°U(I) ) I versus 1,1

for u1, > 2 (1) 1 )

2,r
I
Ear;h ~lot of X*Ucv) versus I {J) consists of IX * Ucv)\ repealed every cv~ =Trad/sec. In the

frequency spectrum of Ix*U<v)I , the component lXUw)l/r is called the primary component
ancl the other component IXLJ(cv±rv,k)~h is called complemenla1y component.

1r c,,,>2rni. no two components of IX * (iw)I will overl ap, and the sampled frequency

spl:ct rum wi II be repealed every <•l~ rad/sec.

1x·1,,.,11

,.,

Fig 3 .12 (!'lot of the frequency spcctnim Ix·(jw) I versus 01

for Ill, '- !t1I I )
76
 If ros < 2ro1, the original shape of IX . * .
Gcv)j no longer appears in the plot of jX Uw) versus
I
w because of the superpositio 11 . .
of the spectrum. Hence, we see that the continuous time
signal x(t) can be reconstructed fi . .
rom the impulse sampled signal x*(t) only if Ws>2w,.

3.5 The Pulse Transfer Function

The transfer function for ti1 ·
e continuous-time system relates the Laplace transfom1 of the
continuous-time output to th t Of h • ·
a t e contmuous-time input, while the pulse transfer function
relates the z transfonn of the t . ·
------ - - - .____ ou put at the sarnpl!ng instants Lo.JhaLoi.tbe..s~pje.d_inp.ut.
Let us first discuss the convoluti·on su mma
·- 110n.
. -
-------- -.
- • - ..

Co11volutio11 S11mmatio11 •• Let us cons1·der a discrete-lime

· · fi1gure beIow.
· system as shown m

x(t) y(t)
G(s) l _ ➔ ~
L_l'I 7 y*(t)

Fig 3.13 ( Continuous - time system with an impulse sampler at the input)

Here, the sequence of impulses x*(t) is the input to the continuous-tim;f ant whose transfer
function is G(s). The output of the plant is a continuous signal y(t). lf"the output there is an
• anoiher sampler, which is synchronized in phase witbj nllli sampler and_QP.erate_g_~t the_same
~ g_period; thenJhe...outpuLis a train of impulses.

If y(t) = 0 for t<O, then z transfonn of y(t) is

..
2 [y(t)] =Y(z) = L y(kT)z-k ................ (3.26)

The output y(t) is related to the input x(t) by the convo~ integral,
~

77
 Or,
I

J
y(t)= g(t-,)x(r)d,
0

= Jx(t - r)g(r)dr ............. (3.27)

0

\\There g(t) is th e impulse - response function of the system. For discrete-time system; we
have a convolution summation, which is similar to the convolution integral. Since

.. ..
X * (t) = I x(t)o(t - kT) =I x(kT)o(t - kT) ............... (3.28)
k•O k•O

is a train of impulses, the response y(t) of the system to the input x*(t) is the sum of the
individual impulse responses. Hence,

g(t)x(O), 0s ts T
g(t)x(O) + g(t - T)x(T), T s t s 2T
y(t) = g(t)x(O) + g(t - T)x(T) + g(t - 2T)x(2T), 2T S t S 3T

g(t)x(O) + g(t - T)x(T) + .......... + g(t - kT)x(kT), kT s t s (k + 1)T

Noting that for a physical system a response can not precede the input, we have g(t) = 0 for
t<O or g(t-kT)=O for t<kT. Hence the preceding equations may be combined into one
equation:

y(t) =g(t)x(O) + g(t- T)x(T) + g(t -2T)x(2T) + ............ + g(t-kT)x(kT)

k
=L g(t-hT)x(hT) ................ (3.29)
h• O

The values of the output y(t) at the sampling instants t = kT (k = 0, 1,2,3, ........) are given by
It
y(kT) = Ig(kT-hT)x(hT) ............... (3 .30)
h=O

k
I x(kT - hT)g(hT) '.............. (3.31)
h=O

78
The summation in Eq.(3.30) or (3.31) is called the convolution summation. ote that the
simplified notation
y(kT) =x(kT) * g(kT) ............... (3.32)
is also used for the convolution summation.

~ , , _ . Since we assumed that x(t)=O for t<O, we have x(kT-hT)=O for~-

Also, since g(kT-hT)=O for h>k, we may assume that the values·of h in Eqs. (3.30) and (3•3 I)
can be taken 'from O to co rather than O to k without changing the value ~f the summation.
Therefore, Eqs. (3.30) and (3.31) can be rewritten as,

O'J

y(kT) = Ig(kT-hT)x(hT) ....................... (3.33)

h•O

..,
=I x(kT - hT)g(hT) ..................... (3.34)
h• O

Where k=O, 1,2, ........ ..

Hence, the z transform of y(kT) becomes

Y(z) = I y(kT)z-k
.
l■O

.., ..
= I I g(kT-hT)x(hT)z-1t
lt • O h • O
., ..,
=LL g(mT)x(hT)z-<m•hl
., .,
= Lg(mT)z-m L x(hT)z-h
nt•O h ■O

................. (3.35)
YCJJ = G(z)X(z)
Where m = k-h and

G(z) =
.. .
Lg(mT)z-m =z transform of g(t)
m• O

79
Eq. (3.35) relates the pulsed Oul put X( ) From
Y(z) of the system to the pulse input z•
Eq.(3.35), we get

G(z) = Y(z)
X(z) ...................... (3.36}
/

Hence, the Eq. (3.36) gives the ratio of output Y(z) to the input X(z), called the pulse transfer
function of the discrete-time system. The figure below shows a block diagram for the pulse
transfer function G(z), together with the input X(z) and the output X(z).

J
X(z) Y(z)
---~7 G(z)

Fig 3.14 (Dlock diagram for a pulse-transfer-function system)

3.5.1 Starred Laplace Transform of the Signal Jnvofri11g Botlt Ortli11mJ a11d Starred
Laplace Transforms:
.- ---- - -
In analyzing discrete-time control systems, we oflen find that some signals in the system arc
starred-(meanmg that signals.,.2r.£J.mg_ul~e sampled) and others are not. To obtain pulse
transfer functions and to analyze discrete-time control systems, therefore, we must be able to

obtain the transfonns of output signals of systems that contain sampling operations in various
places in the loops. Let us consider an impulse-sampled system as shown in figure below.

x(t) • x*(t)
y(l)
~--------~ G(s) >
X(s) Br ~ Y(s)

Fig 3.15 ( Impulse-sampled system)

Then the output Y(s) is given by

1} 80
 Y(s) =G(s)X * (s) ................... (3.37)

Here, Y(s) is U1e product of X*(s), which is periodic with period 21r ' and G(s), which is not
{J)I

periodic. The fact that the impulse-sampled signals are periodic can be seen from the fact that
X*(s) =X*(s±jcv k), 5 k = O,l,2 ,....... ................... (3.38)

Let us now take the starred Laplace transfonn of Eq (3.37) we may factor out x·(s) so that,
Y * (s) =[G(s)X * (s)] *

=[G(s)]* X * (s)
=G * {s)X * (s) ............ (3.39)
This fact is very important in deriving the pulse transfer function and also in simplifying the
block diagram of the discrete-time control system.
~ 1 General Procedures for Obtui11i11g Pulse Transfer F1t11ctio11s:
Let us consider a system that has an impulse sampler at the input as shown in fi gure (a). Let
~.s obtain t lie puJs~ transfer function. The pulse t;ansfer fun~tion G(z) of the system shown in
- - .
figure (a) is

Y(z) = G(z) ' /

X(z)
Next, consider the system as shown in fi gure (b). The transfer fu nction G(s) is given by

Y(s) =G(s)
X(s)
x(t) x*(t) y(t)

~h
?.
-;;1 G(s).

c~l
8r Y(z)
(a)
x(t) y(t)

X(s) 1 G(s)
Y(s)
>

(b)

Fig 3.16 ( Continuous-time system with an impulse sampler

at the input and the continuous-time system)
81
 The pulse transfer functi f th 1 · ·
on ° e system (b) is not equal to z @(s).1since there is n~ ~ t
sampler. The presence or ab • . ••
-~ . ' scnce of the input sampler is crucial in determining the pul_se
transfer function of a system. •

Now, for system (a),

The Laplace ti'ansfom1 of the output·y(t) is ,
Y(s) =G(s)X*(s).__..,.-
Hence, by taking the starr d L I
c apace transfom1 of Y(s), we have
Y*(s) = G*(s)X*(s)
~ '
or,

'Jz) = <3.z)~(~}
For the system (b), the Laplace transform of the output y(t) is
Y(s) = G(s)X(s)
By taking starred Laplace transform, yields
Y ~ = [G(s)X(~) f r= [GX(s) ]*
or,
- . -

In discussing the pulse transfer function, we assu~e tliat'ther_.e-is a ~ampler at the input of the
element in consideration.·Thep~~~7~-;; ;-~7; -s-;;~j~~~j~;~u-tput of the clement
(or the system) docs not affect the pulse transfer function, because, if the sampler is not
.
,..hysically resent at the output side of the system, it is always possible to assut~
.fictitious sampler is present at the output. . his means that, allhou h Jlll signal is
- -
_f.ontmuous, we_c.an..conside.uhe..l'.alues.of the muµulonl~at t = kT (k = 0, 1,2, .. .... ) and thus
get sequence x_~T). •

Note that only for the case where the input to the system G(s) is an i;11pulse sampled signal is
the pulse transfer function given by
G(z) = 2 G(s)

82
 - - ------ - -

Exnmplc 4. Obtain the puls t r · · by

• e ransLer G(z) of the system where G(s) 1s given
l
G(s)=-
s+a
Note that ther~ is a sampler at the input of G(s). V

Solution. Since there is a sampler at the input of G(s), the pulse transfer function is
G(z) = 2 [G(s)]

Here, ZG(s) = 2[_!_)

s+a
I
=---

Hence, G(z)
I -e -aT 2 -I

VJ('ll)
Example 5. Obtain the pulse transfer function of the system where G(s) is given by
~ 1 -Ts
G(s)= -e - -
}
, .., r; -: :--
(

s s(s+l) -
Note that there is a sampler at the input of G(s).
\,.,,-

Solution. Since there is a sampler at the input of G(s), the pulse transfer function
I
/ G(z)= 2G(s).
I
Here,

. ..
•i: ~J~ • __

From the z transform table, we can write

83
J.5.3 Pulse Transfer Functio,1 0,re
J ascatlcd Efcmeuts:
{L~___.- "
Consider the systems sh • . J rs re
own tn Fig 3. 17 (a) and (b). Let us assum~__ilii!tjhc~amp c
synchronized and have the s·1 . • · ff-" 3 17 (a)
°
; ; . t . . . - - - - - - - - - - - " = -= lllc..samphng-period. The pulse transfer function ig •
will be G(z) H(z) while the pulse transfer function of Fig 3.17 (b) will be GII(z), wi1ich is
different from G(z) H(z).

x(t) x*(t) y*(t)

u(t) u*(t)
71 G(s) H(s) >
8r 8r
or

(a)
x(t) x*(t)
l y*(t)

~
71 u«-J
6T 1 G(s)
1 H(s)
OT
>

(b)

Fig 3.17 ( Sampled system with a sampler between cascaded

elements G(s) and H(s) and sampled system with no sampler
between cascaded elements G(s) and H(s)
For Fig (a),

U(s) = G(s)X * (s)

Y(s) = H(s)U * (s)

Hence, by taking the starred Laplace transform of each of these two equations, we get

U*(s) = G *(s)X* (s)

Y*(s) = H*(s)U*(s)

Consequently,
Y*(s) = H *(s)U*(s) = H*(s)G*(s)X*(s)

or,
Y*(s) = G*(s)H*(s)X*(s)
~ l_..A ........

In terms of z transform notation,

Y(z) = G(z)H{z)X{z) ,-

84
 ]lcncc, the pulse trnnsfcr runcti·o . .
n 1s given by,
Y(z)
X(z) = G{z)I l(z) ........ (3.40)

Let us now consider the syst • fi

cm in 1gurc {b). from the diagram we find
Y(s) =G(s)H(s)X*(s) =GH(s)X *(s)

Where,
GH(s) = G(s)H{s)

Taking the starred Laplace transform of Y(s), we have

Y*(s) = [GH(s)r X*(s)

In tenns of z transfonn notation I

Y(z) =GH(z)X(z)
Hence, the pulse transfer function is given by,

Y(z) =GH(z) ....... (3.41)

X(z)
Here, the pulse transfer functions for system (a) and (b) are different.

rWmple 6. Obtain the pulse transfer function for system (a) and (b) shown above where
,/ ~ 1 l
/ G(s) = - andH(s) = -
s+a s+b

Solution.
_For s~m (a), the pulse transfer function is

{Jfj)=~z)

Hence Y(z)
' X(z)
= G(z)H(z) = z[--]z[-
s+ a
1 1
-]
s+ b
1 I
= 1-e-•T z- 1 1- e-bTz-1

85
For system (b ), the pulse transfer function is Y(z) = GIJ(z)
X(z)
Y(z) r, [
Hence, X(z) = 2 LG(s}H(s}] = 2 _!__ _!__]
s+as+b

= 2[2--(
b-a ~-~
1 1 )]

~ b~a[ 1-e~•y {i ~ '

Clearly, we see that the pulse transfer function of two systems are different; that is,
G(z)H(z)" GH(z). S'~ y, (?) ::: ,__ c "10 1 1-1 1 iJ }
-Pulse Transfer Fu11ctio11 of Closed-loop Systems:
~et us cons,ider the closed-loop control system as shown in figure below.
f- . _g
t

R(s) E(s)
"~
.. G(s)
/> qs)

H(s)

Fig 3.18 (Closed-loop control system)

From the diagram,

E(s) =R(s) - H(s)C(s)

C(s) = G(s)E * (s)
Hence,
E(s) = R(s) - H(s)G(s)E * (s)
. ..
86
Then, by taking the statTed Laplace t _ r .
rn11s1onn, we obtain
E*(s) = R *(s) - GH *(s)E*(s)
or,

E*(s)==
R *(s)•
• 1+qH*(s)
Since
C*(s) == G*(s)E*(s)
__.:-:--
We obtain ~

C*(s) = G*(s)R *(s)

...- l+GH*(s)

In tem1s of the z transfonn notation I

C(z) = G(z)R(z)
l+GH(z)
Hence, the pulse transfer function for the present closed-loop system is
C(z) G(z)
- = - --
R(z) l+GH(z)

~ · \
Pulse Trausfer F1111ctio11 of a Digital Co1ttroller:
The pulse transfer function of a digital controller may be obtained from the required input-
output characteristics of the digital controller.

In general, the output m(k) may be given by the following type of difference equation.:
-·
m(k)+ a 1m(k-l) +a 2 m(k - 2)+ ........ +a 0 m(k-n)

= b 0 e(k) + b 1e(k-l) + ......... + b 0 e(k- n) .... ..... (3.42)

Where e(k) = input to digital controller

Taking the z transfonn of Eq. (3.42), we get

M(z)+a 1z-1M(z) + a 2 z-2 M(z)+ ........ + a z-"M(z)
0

== b 0 E(z) + b 1z- 1E(z) + .........+ b0 z-n E(z)

87

f
I
 ¼e :bi tt -d c:::::~

or, ·

(l + a,'l.-1 +a~'/. -J + .......... + a i'.-")M(·,) = (l) ·t I> ., -• .1- + b z-")E(z)

II ' • ll 1' • • • • •• • • II

#J <?ao_;etl-loop Pulse Tn111sfcr F1111ctio11 of11 Digit11/ Co11tr11I System;

1/The f-tg. 3.19 (a) shows the block diagram of a digital control system. Herc, the sampler, AI D
converter. digital controller, zero-order hold, nnd 0/ A converter produce a continuous-time
cor.trol signal u(t) to be fed to the plant. The rig. 3. l 9(b) shows the transfer functions of
blocks involved in the system.

Samr,ltr Zrro or<Jcr frl cir)

019il • I 11
/\ID C')nlrollcr hotel ~
0 / /\

_Rf,I
___..,.., __,_
C'l>l
l I J~ 2..-{ '' G, hl
Cl<I

Gisi

Fig 3.19 (Block diagram of a digital control syst~rn and

equivalent block diagram showing transfer functions of
blocks).
G,,'\s) rcptcscnts the trc1nsfcr function of digital controller.
Referring 10 lhc Fig 3. 19 (b), le! us define

-Ts
1
- c G" (s) = G(s)
s

88
 or,

C * (s) =G * (s)E * ( )G • ~l
S D (s) ( (\) -, ? (CJ)}. ~ ;I.{()
and
E*(s) =R_*(s)-C*(s) \ l() r _ rl (J - \ (( I { )' (J (, ? (1
~

C*(s) =G *(s)G o. (s)[R *(s)-C*(s)] (: ,'\ ln ( \ + t. lS f l I o-:. r 'J

or,
~ )l·Ls1: J!:
I) "'cr ).
,.J.. ~ l<,J lL- ~ ' ((

or,
C*(s) = G*(s)G 0 *(s)
R*(s) l+G*(s)G 0 *(s)
C(z) _ G O (z)G(z) V
R(z) -l+G 0 (z)G(~ \
This equation gives the closed 1 I • · ·
- oop pu se transfer function of a d1g1tal controller.

~Transfer Fuuctiou of a Digital PID Controller:

Just as m contmuous systems, there arc three basic types of control: Pror,ortional, Integral
and Derivative, hence the name, PIO.
Proportional Colltrol: A discrete implementation of proportional control is identical to
continuous, that is, where the continuous is
m(t) = Ke(t) \n \ ') = \ c\) ~ \,,; j\ e C. \:

The discrete is
m(kt) = Ke(kt)
Taking the z transfonn, we get
.M(z) = Iq:(z) ----- (3.44)
.. ,.
,. . , ,. , -
/11,rtr-i( Co1111-ol:· In this, the control action is proportional to the integral of the actuating
error. For continuous system, the output of the integral controller is given by,

f
m(t) =K[__!_ e(t)dt]
T; o
The corresponding discrete equivalent is obtained by the t~s:znidaJ smnmatio.n as,

89
or,

Ti h- 1
I
m(kT) = K[.! c((h-l)T)+e(hT)]
2 .
Define
e((h-1 )T) + e(hT)
2 = f(hT), f(O) = o

Then

~ e((h-l)T)+e(hT) k
LJ
h• I
2 = L f(hT)
h• I

Taking the z transfonn of this last equation, we get

~[±c((h-l)T)+e(hT)]~Jtf(hl)l = Z[f_ 1, -'\ I" )1

h•I 2 1f:I j \.\ _,

l [
= l-z-• F(z)-f(O)
] ·- •,· \, ; -c-h • /A..-'~ +
l
= -1 - -• F(z)
-z

l+z- 1
Notice that F(z) =l[f(hT)] = - - E(z)
2

1 2 1
Hence, M(z) = K[_I.__ + - 1 ]E(z) ----- (3.45)
2Ti 1- z -

Derivative Co11trof: In this, the control action is proportional to the'-derivative of the

actuating error signal. For the continuous system output of the derivative controller is given
by

We can obtain the discrete equivalent by approximating the derivative tenn by a two-point
difference form.

m(kT) ~ K[Td e(kt)-e~k- I )f:)]

90
,,.

Taking the z transfonn we get,

Td
M(z) = K T (l -z-')E(z) ----- (3.46)

Hence the output of the PIO co11t10 · 11 . 1 10 proportional,

er 1s obtained by riddi ng the output "uc
intcoral and derivative controllers • 1-'1'e nee, we get

l
"'

1
)_ [ T l + z- T ----- (3.4 7)
M(z - K l + 2Ti 1- z-I + ; (l - z-1) E(z)

The Eq (3.47) may be rewritten as follows:

~T\ (r------1 • T
I J
M(z) = K \1- - +- - - + ~ (1- z·') E(z)
~\ 2Ti \ Ti 1- z - I T ).
J K .
= [~P+ l-~- 1 +Ko(l-z- )]E(z)
1

Where,

K P -K KT - Kl . .
- - - - K - - = Proporhonl gam
2Ti 2
KT .
K 1 = - = mtegral gain
Ti
KTd . . .
K 0 = - - = denvahvegam
T
Hence, the pulse transfer function for the PIO controller becomes

G 0 (z) = M(z) = K +....!S._+ K 0 (l-2· 1) ----- (3.48)

E(z) r l-z-1
ExaJ.e 7. Considc.r the control system with a digital PID cl!i,troller shown in figure below
rJir:nsfer function of the plant is assumed to be
1
G ( s ) - -- and the sampling period T is assumed to be l Sec.
p - s(s+l)

r(t) e(kT)
m(kT) C(t)
Digital PIO Zero-order • Plai1t
Controller Hold
T=l ~ •
lf_ ~Gh(s) Gp(s)

Fig 3.20 (Block diagram of a control system)

91
Also assume that kr = 1 ki ==0 2 d k
' • an o =0.2
Obtain the closed-loop pulse tran fi r. .
s er ,unction of the system.

Solution. The transfer functio

n °r the zero-order hold circuit is given by,
l -e -s
Gh(s)=-
s
(Since th.:_ time period T == 1 Sec.)

1 1
] = G(z) = 0.36792- + 0.2642z-
2
yl"-J ?
. 0~
•ow, ~ - • -'
s(s + l) ._....- (l-0.3679z-1 )(1..:.z-1)
_____ (3.49)

Wi th kp = 1' ki = 0.2, and ko = 0°.2, the pulse transfer function of the digita~ controller
becomes

G (z)= 1.4-l.4z-1+0.2z-2
D I -1 -=- 1/- -'I-- "I ----- (3.50)
-z \-"Z--1

Hence, the closed loop pulse transfer function becomes

C(z) G O (z)G(z)
R(z) =l+G 0 (z)G@iQ,r

0.51512- 1 - 0.14522-2 -0.2963z-3 +0.05282--.1

=-----------------
1- 1.8528z-1 + 1.59062-2 -0.6642z-3 + 0.0528z-~
----- (3 .51)

3.6 Realization of Digital Controllers and Digital Filters

Realization of digital controllers ar.d digital filters may involve either soflwarc or hardware
or both. The realization of pulse transfer function means determining the physical layout for
the appropriate combination of arithmetic and storage operations.
Computer programs for the digital computer will help to obtain the software realirntio~Y
whereas in a h~~warc realization we build a special-purpose processor using such circuit) ,
as digital adders multipliers, and delay elements.

A digital filter processes a digital signal by passing desirable frequency components of the
digital input signal and rejecting undesirable ones.

92
ln this section, we deal the bl k d. . . d clements
. . oc iagrnm realization of digital filters using cay I .
adders and multtphcrs. This block diagr"m 1· 1· b d bas1·s for the design of
• u rca 1za 10n can c use as a
software or hardware.

ln the block diagram realization a pulse transfer function of z·l represents a delay of one time
unit.

x(k) x(k-1)
X(z) >
Fig 3.21 (Pulse transfer function showing a delay of one time unit)

The general form of a pulse transfer function between the output and the input is given by
bo +biz-I+ b2z-2+ ........+ bmz-m
G(z) = -Y(z) =__;;_- . . . : . . _ -- = - ----..:.:::____ n>m ----- (3.52)
X(z) l +a 1Z -I +a z- 2 + ........+anz -n ' -
2

Where ai's and bi's are real coefficients (some of them may be zero). The pulse transfer
function is in this form for many digital controllers.

As for example, the pulse transfer function for a PID controller is given by

G 0 (z)=Kp + K,_, +K 0 (1 - z- 1)
1-z
_ [Kp(l-z- 1 )+K 1 +K 0 (l-z-1 ) 2 ]
- (l-z-1) .
KP -Kpz-1 +K1 +Ko(l-22-1 +z-2)
= (l - z-1)

b 0 + b,z- 1 + b 2 z-2
= -I
l +a 1z +a 2 z
-2

Where,
bo = Kp + K1 + Ko
bi = - (Kp+2Ko)

93
.UJ✓Direct Progm,11111i11g:
In the Ji reel programming method the cocflicicnls a·; and bi appear directly as multipli ers. Let
us consider the general form of the pulse transfer function,

Y(z)+a l z- 1 Y(z)+·t< z -2 Y(z)+

2 .1 • • •• ' • · .' •
+a z -" Y(z)
ll

or,

X(1l
--1.0--i-□1 ··--►[J
r.;7J
--□

;
-- ------{]--. ,

94
l~l"l '. in tl i1 t Pll)grn m1ni11g ic, li, ati on, we rc,lli1c the in1111c1.1tnr :111d clcnornina tor of the
,h-: 1r:111slLT f11nt: li n 11 usi 11 l' s . . . . .. •• •,l!I of Ill
P' • • ' · 1.: p , 11 ,II ' !-- ·Is or d ·lay ck111c111:-. '1'111.: 11111111.!1:iln t 11 ,c.;' • •
cl l:ly clements nncl dc110111inntor' uses
• ·rr,ercnl set o r 11 cc
a t11 I Iay elc1nc11t. ·1lw'i the tntal
1111 hcr of delay clement s use,t • 1· . . .
1)1 ' Ill< II CCt progra1111111ng IS m+n.

3. (,..'' • Stm1tlar,f />1'J!g !:.!!_!J1tt1 ; 11g:

·1·11,, number or delay ele111e111 . I • . . . I r , m to n
• --·- - - ~- - _ s isc<. - 111 c11rect pro gramm ing can be , e~1uccc ,rorn 1

(where
11
~ !J ~he he~~~~~·d programmin g method. Th is method uses a mini n:urn -
~ e_n~ cr of dch~y~ lemcnts. Let 11s cons ider the p~ lsc trans fer funct io n as
Y(z) l\, +h?.-, +h ,,.-1 + ..... ...... + h ,. - 11,
-- = •- - - . . c - ---=--- I ll
----- (3.53 )
X (z) (I + a , z -, + n ' z - 2 + ......... + a z _,,
~ - II
I ,rP",r ,b.,

Rearranging the pulse t7ansfcr function as fo llows:

Y(z) YV,) l l(z)
- -== - - - -
X(z) H(z) X(z)

Where,

Y(z) b -l -2 -m
0 + 1z + b 2 z + ......... + bmz
-- = b ----- (3 .55)
H(z)
and
H(z)
X(;,.)
-------- -- - ----- (3 .56)

Now, let us draw block diagrams for the systems given by Eqs (3 .55) and (3.56).
Rewriting Eq (3.55),
Y(z) =b 11 H(z) + b, z- 1 H(z) + b ~z-1 H(z)+ ......... + b 111 z- H(z)
111
----- (3.57)

-Gi--- - - - - - -

lf!1I
-•□l~C~-►GJ--- I I
r1,1

fig 3.23 (Olock diagram rc:ilizalion orEq. 3.57) 95

~l)\\. ('(:\\' rit illl' Eq (3.56),

11(1.) · X(1,) ll 'l. 'tt,) . '

I ,,. - ,I /
~'
11( / ,) .,,,,, ,, , - j1 i', " 11(1,) -----{J .5~)
11

[J
/11,i

- - .. t '

.,
-i·
..
t' ]· ·----□ - ----{]
•I ◄

fig 3 •24 (n1ock diagram rcalizatio1U>fEq. 3.58)

No", the comhi11.1tion or these

· 1wo 11
) oc..: •
1 d1.1gra111s . give
w ill • the following
• block cJiagram
• (c) .

---Fl- - - -- ------
► [] ---------
- --[]--- - --

Fig 3.2 4 (13lock diagram rcaliz:ition or Eq. 3.53) b c :ird programming )

Hc1~ec, in this method, we use only n delay clemen ts. The "CecCG · ·cnts a I , a2, ........ a,, appear

as rccdh:1ck rkmcn~ d_ thc codlicic11l s h11, ,hl.!_:.:.:,: •~11 appear as recd forward clements.

Thus, the stamlard programming is preferred as ii uses less number of delay elements.

96
•'
,,.

J,r,.3 Series Progra111111ing:

1,~ .Ill' -.rrirs prngr;un,t1ing 111 L:1'1od, we w i ll i111p k ,11 c 11f Ilic p11lsc lransli.:r rum; fion Ci(1.) i lS a

series ,;onn~c t'1011 0 f fiu•st -order and/or second-order pulse transfer functions.
• If G(z) can be
wrillcn as a product of pulse transfer functions Gi(z), Gi(z), .......... Gp(z), or
G(,.) = G 1 (z)G ~ (,.)....... .... G ,. (z) ----- (3.59)

Then the digital filler for G(z) may be given as a series c.onneclion of the component digital
filt ers G,(z), G~(z), .......... G,,(z), as shown in figure below:

Y(k)
--~ G,(z) 1---- 7
Y(z)

fig 3.26 (Digital filler G(z) decomposed inlo a series connection ofG,(z), G1(z),...... G,,(z))
Lei us consider the pulse transfer function
G(z) = G 1 (;,;)G 1 (z)

_
I+ b.z· 1 I +c.z-' + f z"1
I I I
----- (3.60)
- J+a .z· 1 l + c.z-1 +d.z· 2
I I I

Y (z) l+b .z -1
The :)Jock diagram representation of - 1 - = - - ' -1 is shown below.
X1 (z) I +a;z-

( ~Q.. ,_r)
X(tl
--►
Y( ✓ ,
r Y'\ 1,., . . . .""' oJ

Fig 3.27

•-1 f - 2
Y,(z) l+c.z + .z
1 1
and that for - •• i$ shown he low
X ? (.z ) - l+c .z -I + d .z -2
- I I "

97
 1 ·-----
~-. Ir --GJ-

--@
Fig 3.28

Thus hy drawing lbcsc lwo cliagrnrns in series, we gel 1h.e block diagram rcprcscntiltion for
scri::s programming.

3.ti.4 Parallel Progm111111i11g:

In ;!tc parallel programming methou the pul se lransl'cr li.111cti on is expanded into partial
fra:tions. If G(z) is expanded on a sum of A, G,(z), G2(z),........ Gq(z) or,
----- (J .C, I)

Where /\0 is a co11sla11t lhi..:11 tl1i..: block diagra 111 li.ll' thi..: digital lillL:r G(z) can bi..: oblaincJ as a

pa..allcl connections of q+ I digital fillers as shown in Fig 3.29.

X(d

Fig 3.29 (D igital lillcr Gl'l.) decomposed as a parallel co n nect i on of A , G 1( z ), G1

( z ) , . ..... G ~ ( z ) )

I.cl ll:> conside r the pul se transfer fun ction for lhc digital filt er is

+ ri z
-I
bi ci
G(z)= - ,.;.._,_+ , _,
• l + a z -• I +c,z.- + d, :t: -
I

98
 [,cl 115 now draw Ilic blocl· diagrn 111 for

_\}t} .::: h, _
X,(r.) l +a,z-1

Fig 3.30
Again drawing the block cl iag1

Y,(z) e + f z- 1
- -- = I I

X, (:t.) I I l: '/. I Id '/. ~

I I

□----.

1/(1)
__
J' lkl,..
Y [1)

'-----[J
---CJ ◄-

Fig 3.J I

Now, by drawing these two diagrams in parallel, wc get the block diagram representat ion for
parallel progr:11n111i11g.

3.6.5 lmlder ProgN1111111i11g:

In lli is method, we expand the pulse transfer runction G(z) into the fol lowing continucd-
fract :on from.

99
 ----- (3.62)

Let us define

I= 1,2,:..... , n-1 ----- (3 .63)

i = 1,2,......, n- 1 ----- (3.64)

G ce>(z) = I
" I ' ----- (3 .65)
- Bz+
n -
._____A"

Then G(z) may be written as

----- (3.66)

Let us now take a simple case.when n = 2

G(z) = A 0 + - - -- ~ - - ----- (3. 6 7)

Biz+ I
Al+ 1
B z+ -
2 A
2

... .
v No.w, by the use.of functions G, CA)(z), ~~zj°'~md'(i2~8 >(~). the transfer function G(z) may be
••
written as follows:
I
G(z)=A 0 + 1
Bz+ - --
1 A1 +G/Bl(z)

100
 =A 0 + l
B1z+G/"l(z)
= AO + G Iffi) ( z)
----- (3.68)

8
Notice that Gi< >(z) may be written as

G.<o>(z)= Y;(z) = l
----- (3.69)
' X i (z) Biz+G?l(z)
or,
Xi(z) = Yi (z)Biz+ Yi (z)Gi<Al(z)

1
or, Y;(z)= ~i z- X;(z)- ~ - z-1G/"l(z)Yi (z) ----- (3. 70)
I

• r (ll) .
The block dtagram 1or Gi (z) given by Eq. (3.70) is shown below.

1 •I
z
B .I

Fig 3.32 (Block diagram for G?1(z) given by Eq. (3.70))

Now, G?>(z) can be written as,

----- (3.71)

or,

or,

1 I
Y;(z) =-X;(z)--Y;(z)G; . 11s1( z) . .... . (3.72)
A; Ai
The block diagram for G?>(z) given by Eq. (3.72) is shown below,

IO l
,,,

Y ,(z)

1-- - ---.-- - -?
/\

G, 1 I(II)(i'.)

. ( )
f-ig 3.33 (nlock diagram fnr G;4'°'1(z) given by Eq. (3.72))

..
'

l·lc.1ce for lhe case 11 == 2, by combining componc11l cligilal filters as shown in fi gure (;i)

bclcw, it is possible lo draw the block diagram or the digital filler G(z) as shown in fig 3.34
(h) .

0
•-rn..,___..
©
0
-►1-.n_..
.- ~

fig 3.34 ( Component block d_iJgr:ims for l:iddcr program1~ing of G(z) gi_vcn by Eq
3.62 when 11=2 :ind co111hinal1on of componcnl block diagrams show111g ladder
progr:m1111ing of G(z)))

102
 ~
Prol>lcms
.
J.7.
C'nnsi d1.:r a ,.crn-onkr hol 1 . . Ic
• prccclbl hy a sa111pkr. The following figure shows 11
l
input X(l) lo lhc sa1111 I . . · I
1 111d
er ' the 011lp11l y( l) or the zero-order hold . Obtain tic
expression ror y(l) Then 1- 1 \' . . ·d ·
• lllt (s) and obtain the tra11s!cr f'unction ol the zero-or Ct
hold.
VIII

'I .
v
\ - c · T,
Gh (s) = - - - '
...\- V 1...T ' J -
, 'l'- , ' ...
" s
::)
I -h
Consider the runction X(s)- - c .
,._./ s '--.--L._J.__1_--1.._
0 T
_L.._ .1_..___.__ _ ,,_

2r JT 4T 5T GT 7T r
v
show that s= 0 is not a pole or X(s). Also show that Y(s) =-~ - Ts has a simple pole
s· -
al s ~ 0. ' '~
➔ _,..
I 7

,. / l ~ I s
-1-- -z-1
·" )
,
:. ~-
..... : : - - - ' '
,.) _/
J. ,. ~ Ot1ta1iy, rt z trnnsfonn or X(s) = , by using residue method.
,i<'} __/ . (s + l t(s+2)
2-2c -Tz- 1 - Tc -Tz -1 2
Ans.
1 2
(l - c-tz- ) l -e -nz- 1

j~
J ;,:,1in the discrete-time ou tput C(z) or ~he closecl-l~op control system shown

--,··re below. Also, obtain the co1'.li11uous-1i111c output C(s).

111

R(s) C(s)
) °' 1~ -- - - .
G1(s) G2(s) 1......----;>,

I l(s)

/\ns.

103
 H(s)

Ans. C(z) = Gjz)G 2 (z)

G, *(s)R *(s)
R(z) 1+ G i (z)G iH(z) -C(s)=G 2 (s)--'---- -
l+G 1 *(s)G 2 H*(s)
For the sampled data t .
con ro 1 system shown m figure below, find the output c(k) for
r(t) = unit step. •

c(t)
ZOH
s+1

Ans c(k) = -j0.51(0.5 + j0.62)k + j0.51(0.5- j0.62/

7. Consider the digital filter defined by
✓
1 2 3
G(z) = Y(z) = 0.2 + l.2z- + 2.3z- + 4z-
X(z) 1 + 5z- 1 + Gz- 2 + 7z- 3 + 8z- 4

Draw the direct realization diagram and standard realization diagram.

~
8. Consider the digital filter defined by ~
Y(z) 3(z-l)(z 2 +1.2z+l)
G(z) = - = __;_....;_:__ _ __
X(z) (z+O.l)(z 2 -0.3z+0.8)
Draw a series realization diagram and a parallel realization diagram/
I
104
 Assume t\rnt a .u1g1ta 11 1ter 1.s given
1· • \ r. . by lhc fo llowing cl1. ffercnce equation:
9,
y(k) +. a,y(k - 1) + a 2 y(k - 2) = b I x(k) + b 1 x(k-1)
Draw block drngrams for the !i\ler by using
( \) dirett progranm11ng, (2) standard programming, ancl
(3) \adder programming.
Consider the digital filter defined by
\0.
2+ 2.2z- 1 +0.2z-2
G(z)== l + 0.4z-1 -0.l2z- 2
. d the ladder scheme.
Realize this [i\ter in the series scheme, the parallel scheme, an
 CHAPTER-IV

Design· of Discretc-T·ime control Systems by Convent1ona

· J
Methods

1 Mapping of the L ft J-I If f

- _ e a o the s-Plane into the z-Plane ../
1ear dynamic system is st abl e I•f aII t11e poles of the transfer function
. he . m
. tl~e lcfi-ha If O,r
the s-2lane.
,,
In •the
_
z-plane ti_1e 1eft -half S-P..lane c~nes onds to the umt· circle
· centcred aL tI1e
ori.00 in
-
or' the left-half
.._
s-plane maps mot
• t l1e ms1dc
• · of the umt-' 1~
· the z-plane.

. TS
Smce ' .......
z=e and s=cr+jffi
We can \\'rite,

and L z=Tffi.
Since cr is negative in the left-half of the s-plane, the left-half of the s-planc corresponds to

The jffi axis in the s-plane where cr=O, corresponds to the unit circle in the z-plane (z= l ). That
- is the imaginary axis in the s-plane ~~sp_onds to the u~it_~jrc_le in_the z-plane, and the
interior
-- o.f.the_unit..circle_corre_sp_onds...to the left-half
,. . -
of- ...the-....s-plane
- . ..r--

Note that since L z= ffiT the angle of z varies from -oo to oo as ffi varies from -oo to CJJ .

Consider a representative point on the jffi axis in the s-plane . As the point moves from _

_ jlm to jlw on the jffi axis, where IDs is the sampling frequency, we have I~ = l and_}:: 2
2 s ·2 s '
varies from -rr to rr in the counterclockwise direction in the z-plane. As the point moves from

j ½ws to j ~ ws on the jro axis, the corresponding point in the z-plane traces out the unit
.,,.
.'
106
 •
·ir It.: once in the counter clockwisc <r.
¥
. Thus, as the point in thcl:I~
11ccllon. vcs [rain -
11 11
, ,, ,. 11 ,,,.n11lhcj111:1:-;is.wclran.: lh·c • .• . •'
11111 1 t:i 1t·.k 111 the z - p l :1m.: :111 •111'1• 111•\c 11 u 111 \i1,•I• ll r1 · - cs.-
0
th 11151
~Cl \
' ml eacl1 st11··P O f width
· •
Ws in lhc left-hair or s-p\anc maps into e • '<l c f
the unit circle in the z-p\anc. This implies that the le ft-hal f of the s-plane may be divided into
1
,•in ir.linitc no. of periodic st,·,· P as s11own below. The primary
• strip 1
• extcn< s ,rom - J2<
• vs to• ~
5
.-1 ( !; . The complemcnlary slriJ.) extei,d
, ~-0111 J•·-I CtJ to J-CtJ
11 •3 •3 W to J-CtJ
J- • •••••• and from
J2 5 2 s 2 s' 2 s 2 s

I -1
- i J o1 Io - I 2· ro , ......... ..
.- s • s
,,,, t
., ~
. , •• , • • .• • • • !'li , )
.. . . ....:J. ··•., ; •1•~•: •,,• I
·I •
1
,··,uu11l1:11w1 ,r, 11 v
,..,. .,. ) J+l.tP,:

1111
L p l.inc

--- - --- - - - ·>-

' nc

(: 1 1mpl,:11wlll Jf 'f
-
!»h tj•

( \ 111 1pli:11u:111.:,v :,.

:.11111

::. :.:.:. u~~:~ :•;•/.~;·.<·:::~~-:~ i !,.,~,

'.!

Fig 4 .1 (Periodic strips in the s-plane and lhc corresponding region in the z-plane)

4.:! Stabilit y Analysis of Closed-Loop System in the z-P lanc

Stability analysis of a closed loop system :-

Le•. us consider a closed-loop discrete time system as shown below:-

C(s)

R( s)

1:ig 4 .2 ( Closed loop 1.:011\ro l syste m) 107

r'
-
.,

rh, puls, trnns ~·r fun ·tion r this clos 11

•Cl oop systcn1 is,
(,)
R(z) l + Gl l(z) .......... (4 .1)
The stabilit) of the system defi db • J
me YEq (4. 1) as well as other types of discrete-time contro
systems may be dctcn11ined fr . t
om t1le locat1011s of the closed loop poles in the z-plane, or tie
roots of the characteristic equation.
P(z) = l + GH(z) = o
35 follows:
1. For the system to be t· bl ----=------ ·
s a e, the floscd- loop _E.9fos or the roots of the charac ten·st,·c
equation mu st lie within the ~ circle in the z-plane. Any closed-loop pole outside
the unit cit_:fle_n1akes the system unstable.
2. If a simple pole lies at z== I, then the system becomes critically stable. Also, th e
system becomes_s:ritically stable if a single pair of conjugate complex poles lie on the
unit circle in the z-plane. Any multiple closed-loop pole on the unit circle makes the
system unstable.
... .
.)
Closed-loop zeroes do not affect the absolute stability and therefore may be located
anywhere in the z-plane.

Example I.Consider the closed-loop control system shown below. Detennine the stability of
/r the system when k=l .
C(s)
R(s)
k I - e- s
s s(s + 1)

Fig 4.3 (Closed loop control system)

(_ r'!l Y'(-f: )
Solution. \ { ~ t J ('1 )
1- e-s 1 '(I (1 J : j
G(s) == s s(s + 1) ·v.
108
!'IOW, G~ =2[1 -;-• s(s~t)]

= 0.3G79z + 0.2642
(z - 0.3679) (z -1)
The closed loop pulse transfer funct' f . .
ion o t11e given system 1s,
C(z) = G(z)
R(z) l + G(z)
Hence, the characteristic equation is
'
l+G(z)=O.
which becomes,
(z-0.3679) (z-1) + 0.3679z + o. 2642 = 0
Oor, z2-z + 0.6321 =O.
The roots of the char equation are f~und to be
z1 =0.5 + j 0.6181 and z2;:::0.5- j 0.6181

Since, \z11= lz2 \ < 1.

The system is stable.

\ j.3 Methods for Testing Absolute Stability

/'J ~ -lt is an algebraic criterion for determining whether or not the roots of
the characteristic polynomial lie within a unit circle there by determining system stability.
Consider that the characteristic polynomial

The Jury's test consists of wo parts.

( 1) A simple test for necessity.
(2) A second test for sufficiency.
•
The necessary condition for stability ar~
P(l)>O; P(- l)>Oforneven
< 0 for n odd.
The sufficient condition for stability can be established through the following md hod.
Prepare a table of coefficients of the characteristic polynomial as below:

109
 Iz·
::11

3.:;.)
I~
::11

:1_
I ·····"·······
••••••••••••·
z·'·-
I
:ln•?
r-rl

:ln-1
-, Zn -

iln
··- :ln-3 n,
b1 b-
••••••••••••• :l.2
b _1 bn.1
b.:.. ••••••••••••• b n-~ r
bn.J
b r.4
•••••••••••·• b1 bu
Ct
C
- C3 .............. Cn-2
Cn.J
Cn....;
C:;.5
••••••••••••• Co

_n-5 Po Pl
P:! p3
_n....i p3 Pl Pt Po
.,
_n-., qJ qi
---= q2

where

a a
b = 0 n-k
k an a
k
b
C = Q
k b
n-1
The sufficient condition for stability are.

laol < lanl

lbol > lbn -11

/ !col> lcn-21
lqol > lq2I

Example2. Ext ine the stability of the following characteristic equation.

P(z) = 2z~+7z
• 3 '
+ 10z-+4z+ l =0
Solution. For this characteristic equation
P(l) = 2+7+10+4+1= 24 > 0 satisfied
,r
P(-1) = 2-7+ 10-4+1 = 2 > 0 satisfied

C, j 0
_: b
110
 zt
z2 z3 z·'
10 7
2 2 ;1 10
2
4
3 -10 ,
K -10 -I
.- C)

4 -1 Q -10 -3
5 20 . --.
20

The sufficient condition arc,

-laol < la41 Le. l < 2

1-31 > I- 11 \. "7 ~

0

satisfied
0

1~>1201 '(., t-1 ( nQ!..sal.isfted -

The system is therefore unstable.
..==
E~e 3. Examine the stability of the following characteristic equation:
4 3
{_ _ P(z) =z - 1.2z +0.07z2+0.3z- 0.08 =0
Solution . The necessary conditions arc.
P(l) = 1-1.2+0.07 +0.3-.0.3-0.08 = 0.09 > 0
P(-1) =l + 1.2+0.07-0.3-0.08
= 1.89 > 0.
satisfied (for n even)
For the sufficient condition let us make a table as shown below.

Row zO z' z2 z~ z'~

l -0.08 I 0.3 0.07 -1.2 l ,/
2 Il -1.2
\__
0.07
-:,,-
0.3 ( -0.08 .
3 -0.994 l .176 -0.0756 -0.204
-0.204 -0.0756 1. 176 -0.994
4
5 0.946 -l.184 0.315 ----
ao ... 0.08.. 84= 1
satisfied

~~ 111
 bo- 0.994 b3::..o._ 04

lbol> lb3l , satisfied

co=0.946, c2=0.3 l 5

satisfied
Hence. the system is satisfied 1• e 11
• • a the ro 0 ts 11e
· ms1de
• • the unit circle in the z-plane.

Example -t Examine the stabilit f .

y O the followmg characteristic equation .
P(z)=z3-l. lz2-0. 1z+0.2=0
Solution. P( 1)== 1-1.1 -0.1 +0.2=0
........... (4.2)
P(-1)=,=- l 8 < o
• ,n=3a: odd, satisfied. n
The Eq (4.2) indicates that at le t . .... /,. / \'- •
as one root ~ ~ o r e the system can be critically
stable. The remainin°~ tests dete rmme
• wI1eti1er the system 1s. critically
. stable or unstable.

Row zO z1 z2 z3

I 0.2 -0. 1 -1.1

2 1. -1.1 -0.1 0.2
3 -0.96 ~ - 0~
i. •Ot

Here,

I0.21< 1 and
I- 0.91> I- 0.121conditions are ~
Hence, the system is critically stable.

G(z) k(0.37692 + 0.2642)

(z - 0.3679)(2- l)
Considerthediscrcte-tim~ unity feedback control system (with sampling period T= l sec).
Whose open loop pulse tran~fe~ function is given ab . Detennine the range of gain k for
.
the stability by use of the jury stability test and also_ find the frequency of sustained
oscillation. J
( ) •. <
Solution .The closed loop transfer function is,
112
7

C(z) = 2 k(0.36792 + 0.2642)

R(z) z ~
(~.3679k - l .367~.3679 + 0.2642k
Thus, the charactenstic equation is ,tJ
'
P(z) = z~ + (0.3679k-l .3679) z. 0 3679 i \
• + 0.2642k = O
Applying the condition, ' - - -- •
~O? Cr-

or, J(0.3679 + 0.2642 k)J < 1

or, k < 2.3925

c.:.--- .... ....... (4.3)
The second condition is,
P(l) > 0, P(-1) > 0, for n=2 .
So, P( l) = l + 0.3679 k - 1.3679 + 0.3679 + 0.2642k > 0
or, 0.6321 k > 0
or, k >0 ...........(4.4)
Again~ P(~I)= 1-0.3679 k + 1.3679 +.3679 + 0.2642 k
= 2.7358 - 0.1037 k > O
which fields, 'e-,
26.382 > k. · ..........(4.5)
Hence, from the inequalities (4.3), (4.4) and (4.5) for the system to be stal}ie,
'
~3925 > k> .0
Hence, the range of! for stability is between Oand 2 3925.
If gain k=2.3925 is set then the system becomes critically stab)~ that n:ean the sustained
oscillations exist at the output.
with k=2.3925, the characteristic e uation becomes,
z2-0.4877z + 1 =0
L-Z -:::.
z = 0.2439 ±. j0.96~8
-- - (JJ - 21r '
Now, w -= 2 _Lz = -Lz [T=lscc]
/, d 21r 2Jt -

= ta;~.'...l o.9~ ~y-321,4 rad/see

o.2~v J

..._____--"'-=--_ ____ - --
I

Note: In ·s-plane, a constant damping-ratio can be given by,

11 3
 s = -OCtJ + j Ct)
n n
D u -- -uw .r.:
+• _
' n JWd ,;
Where,

Ct) =CtJ ✓1-c52

d n
0,,, in z-plane, _,.;
TS
z =e = exp (-8co 0 T+jcodT)

Hence,

(JJ
< z = 2;r _Q_
Ct)
s

4.4 Stability Analysis by use of the Bilinear Transformation and Routh Stability

-
V _;;=
Criterion

In the stability analysis of discrete time control system, another frequently used method is
Bilinear transformation-coupled with R-H criterion. This, method requires transfom1ation
from z-plane to co-plane then apply R-H criterion.
The bilinear transformation is defined by
+lCtJ
z---
-w-1 ✓

z +l
which gives, CtJ = -- J
z- 1
Example 6. Use bilinear transformation and R-H stability criterion to determine the stability
condition for the following characteristic equation.
P(z) =z3- 1.3 z2 -0.08z + 0.24 =O
CtJ + l
Solution . Let us substitute, z = -Ct)-- we get,
1

w + 1)3 -l....(CtJ +l)2 -O.Ojw : l) + o.24 = 0

( CtJ - 1 \CtJ- 1 \w l
or, -0.14 co 3 + 1.06 co 2 + 5.10 co+ 1.98 =O
11--1
or, c~:/ - 7.57,lrn2 - 36. Dw -14._14 :::Q
-r1i,· 1(,11 1111 ;111.iy : '
.\
(ll

2
(1)

II
(I)

Thcie is one sign change from lto-7 57 I Tl .

. • • 1c system 1s unstable.
There is one sign change in first I
co umn, that 1111~:111s
-
one root lie in the right half of thew-

lil,u;c or one root lie outside the unit circle 111

• II IC i'.-p.1anc.

4.5 Transient and Steady-State Response Analysis

Th•; unil-SICp rcspon:;c curve showi ng the transient response spcc ilication IJ., t,, Ip, Mandt, is
shewn here.
r (tl

Ill
ll!t

I
I
I
I ' I
I I I • I
I I I
I I I I
It I j / I I I I
() ~
/:I - -l . ____ I __ _____ _ I
I - .. -·- - - - . . - ·- ··-· ----~
◄ ►
' ·
1 I

,,.
r,

Fi!; -1 .-1 (Uni I sll'p rl'sponsc curve showing lr;insicnl rcspnnsc sp,·cilic;itions IJ. I,. 11,. 1\1P.
:1111! t )

I. Dda.r rime :- It is the time required to reach hal f Lhc linal value at the very first time.

1t is denoted hy t,1.
2. Rise time :- For over damped system it is from 0'1/c, to I 00%. For under damped

system it is from I 0% lo 90% rise ti:ne is commonly useu. ll is denoted by 11.

3. J'eak time! :- Ti me requi red for the response lo each th e fi rst peak or the overshoot. It

is denoted by l r.
11 5
 Maxim 11111 overs/1001 :-
1
t ., ,
is lbe 111 axi1 nurn
JH.:c1k value of the rc,p0n'/,; c1.,r,c
111casmcd from t111i1y Ir 111 ,,1
• <.: 111w1 ·I - th-• 1·n :,,,
. " v,, uc of the rc~pomc differ•; from unity
nwx1111um overshoot is corn
111011 1
Yused as below.
01 M • c(t ) - c(uJ)
io ax1mum overshoot = p ,,
c{(/)) x lOO ¼
5. Settling lime :- The scttlii 1 • • ,
ig time 1s the time required for the response curve to rcacn
and stay within a range b b J
a out tI1c final value of a size specified as an a so ute
percentage of the final value, usually 2%.

4.6 Steady-State Error Analysis ( ,eJr)

The Steady - State pcrforniance or a stable control system is generally judged by the s cady-
state error due to step, ramp and acceleration in uts. Any physical control system inherently
suff:rs ste:_dy- st ate ~~r in_~espo~1se to ~~rtain types of inputs. That is, a system .!]1ay have no
- -
steady-state crro~ with_sts:p inputs, but the same system may exhibit non zero steady-sta e
•-- • • . - -=----·- --
error in res_p.91)se to ram in ut. Whether or not a given system wi ll exhibit steady-s ate error
in response to a given type or inpu t depc~ds on the type of open-loop transfer function of the
system.

Consider the continuous - time control system whose open-loop transfer function G(s) H(s) is
given by,
·- - -
k(Tas + 1)(Tbs+ 1)..... (Tns + 1)
G ~ = --;N
:-;--- - - " - - -- -
- S (T1s+l)(Ti5+l).:...(Tps+l)
The tenn SN is the denominator represents a pole of multiplicity Nat the origin.

A system is said to be type 0, typel, type 2, ifN=O, 1':,=l, N=2, .... respc_£tivcly.
Now, the discretc-tim; control~tem can be classified according to the number of open-
loop pole~ an open loop pole at z=l corresponds to an integration in the loop).
Suppose, the open-loop transfer function is given by the equation.
. _ _ B(z)
Open-loop pulse transfer function = N A{ )

-
(z - 1) 2

116
 B(z) .
\/}lerc 7Jiicontains neither a pol •ri d a
\ / / (z). __ e nor a zero at z=t . Then the system can be class, ,c as

Pe o system a type 1 system or t -I =2

ty ' a Ypc 2 system accordin g to whether 1 ==O, - '
respectively.

Let us consider a discrete time control syst

cm as s11own below :-

r(t) e(t)
c(t)
1- e '
f,. c.2)
s

~ - - - - - ---l ll(s)

Fig 4.5 (Discrete- time control system)

Consider the system is stable so that we can apply final value theorem.
From the diagram
e(t) = r(t) - b(t)
Final value theorem,

lim c(kT) = lim [(1-z- l)E(z)) ············<4.6)

k ➔ oo z ➔ l

from the figure define

1- c-Ts ]
G(z) =2 s Gp(s)
[

G (s)
=(1-z - 1) 2 - p-
s
G (s)H(s)
p
and GH(z) = (l -z-1) 2
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)

-------- 117
or, E(z) = 1+ GH(z) R(z)
.......... (4.7)
substituting the value of E(z) to E (
q 4.6), We net
Jim [
ess =
z➔ 1
( 1- z- l )
b

l
~ R(z)
l ../
Let us consider three types of in ut . . . .
. . p • unit-st ep, unit ramp and unit-accclcration-inp11t as tn the
case of contmuous-ttmc control system.
Static P!!_sitiol_! Error~o11staut :-
--
~or a unit step input r(t) = l(t),

R(z) = l
(I-z- 1)

So, ess = lim

z -> 1
[c I - z· I) 1
1+ GH(z) (1- z·')
l l
lim
=
z -> 1 l + GH(z)
Let us define state position error constant kp as follows :-
lim
kP = GH(z)
z ➔ l • 1
e =--
Then we can write, ss 1+ k
p

The steady-state actuating error in response to a unit step input becomes zero if kp =oo, which
requires that GH(z) have at least one pole at z=l
Static Velocity Error Co11sta11t :- For a unit ramp input,
r(t) = t 1(t),

Tz· 1
R(z) = -1 2
(1-z )

So,

11 8
1.ct us ddin the stntic vclocit ..
. y c, 'or constant kv ns
kv = hm[~ ]
z - >1 T

l .
Then, e ss =-
k
V

If k,, = oo, then the steady-state .

actuatmg cnor •
quires GH(z) to posse d m response to a unit-ramp input is zero. This
re ss a oub]c pole at z==l.

Static Acceleratio11 Error Coust . .

Gilt.- For a umt acceleration input
l '
r(t) = - t 2 l(t)
2

= lim [
Z➔ l

Now, define static acceleration error constant ka as

The steady-state actuating error in response to a unit acceleration input becomes zero if
ka =cc. This requires GH(z) to possess a hiple pole at z=1.
-.,,-

The following table shows the system types and the corresponding steady-state errors in
response to step, ramp and acceleration inpu(s for the discrete-time control system.

119
 y tel11 Steady State ITrror i11!{c!ip01lSC lo
Step input r (1) ~I R .
amp input r (l) -·t
Acc-c-lcr" l1·0 11
0

I 2
r (t) =-- l
Type 0 1/(1 +kp) 2
Cl.)
Type 1 0 00

rype2 llkv
0 00

0
Ilka

The table below shows the static error constants for typical closed-loop configurati~ns of
discrete-time control system.

Closed loop configuration

Values of kp, kv, and ka
r-

k = lim GH(z)
.... P z➔I
~

lim (1-z-')GH(z)
'---,I
J ll{s) L
JI""~
. -
kV = z ➔ 1 T

lim (l - z-1)2GH(z)
k, = z ➔ 1 y2

lim
k = G(z)H(z)
P z➔l

~~T k -
Jim (l _.:.·2 - 1)G(z)H(z)

t I H(s) hr---
v-z ➔ l

k =
a z➔1
. T

lim (1- z- 1/ G(z)H(z)

T
2

k - lim G1(z)G2H(z)
➔
0

2
P z 1
~ ~ ~ G (s)
Iim (l - z-1)G1(z)G2H(z)
k - .
v - z➔ l T
t
- - -
j •1
H{s) !14.,._ __
J,
lim (l - z- 1)2 G1(z)G 2H(~
ka = z ➔ 1 Tl

20
,'

-~e
d ll(s) _ k = Jim (1-z- 1)G 1(z)G 1 (:)H(z)
V
1
II

z - >I T

k = lirn (l-z- 1)G 1(7.)G 2 (z)H (z)

3
z➔ I T1

4.7 Design Based on the Root-Locus l\lcthods

Ge11 eral Rules for Co11strncting Root Loci:-
\. ,. Obtain the clwractcristic equation
I +F(z) =O aml rearrange as follows:

k(z + z 1)(z + z2 ) .......(z+ Zm)

I + - --=-----~---.:.'...'....'.. - O
(z+p1)(z+p2) .......(z+pn) -

Where k is the gain and k > 0.

I
I

from the fac~orcd form or the open-loop pule transkr fonction, locate the open loop po!es
and zeroes in z-plane.

2. Find the originating points and terminating points of the root loci. The points on the
root loci corresponding to k=O are open loop poles and those corresponding to k = Cl)
arc open loop zeroes.
(z+p1) (z+p2) ............. (z+pn) + k (z+zi) (z+z2) ......... (i'.+zm)=O
n m
or, I (z + p.) + k I (z + z. ) = o
J=l ___.J
• • }
1=
I
----
When k=O, this equation has roots at - p, U= l ,2, .... n) which arc open loop poles. The root
locus branches therefore start. at open loop poles.
I n I'Q
when k = oo, - L (z+p.)+ I: (z+z.)
1
=0
k j=l J i=:_1
This equation has roots at -Zi (i=l, 2... m) which are open loop zeroes. The root locus
:......---
branches therefore terminate at open loop zeroes.

121
,·

..,~i:. 111
1
II, th1.: op1.:n loop tr1111sfl.!r ~ ,
111 • • • llll ct 10 11 1ia s ( . . . . . .. , _
cs of root locus lcrminnt , . . 11
11 1

;,,croc:i :i t l llcrcforc.: (11 11 1

bn111CIl
) 111 f ,1111 y . )

c ill tnfimty alon' ll

g le nsyniptotcs.

;\ point on the real axis lies

.
011 l11c root Iocus 1f
• the number of open-loop poles PIus •
zeroes on lI1e real axis to lhc ri I r .
g 1l O this point is odd.

The (n-m) branches of root 1 . .

octis which tend to infinity, do so along strai ght li ne
asymptotes
. • g·1ven l1y,
whose angles ·ire
(2q + 1) 180
¢,, = (n-m) ,q=0,1,2, .....,(n-m-1)

where, n = no, or Ii nit~ poles of F(z)

nr no, of finite zeroes of F(z)

5. The asymptotes cross the real axis at a point known as centroid, dctcnninc<l hy the
/relation.

_I I
real parts of poles- reals parts of zeroes
-a,, - - - - -- -- ----=~- ----=--- -- -
. no.of poles - no.of zeroes

6. The breakaway points (points at which multiple roots of the characteristic equation
dk
occur) o ~ the root locus arc the solution of dz = 0

If the characteristic equation

l +F(z) =O is written as

l +k B(z) =O
A(z)

-A(z)
then, k = B(z)
.
dk A'(z)B(z}- A(z)[3'(z) =
So, -=- 0
dz B\z)

.122
 r' ihi' \':tl111• 111' 1' ('nrn·::pnndj 111 , 1 .
t • II ,I 10111 I 'I 1· dk
II II () • • • I I
· · l Since k · .. . d1 i:: po::11 1vl'. pni111 1. 1 11 is Ilic: in;:1,
,,,:iv polll • 1s ,lssumcd lo b , •
• t: ll011-11cg·1t j \' • • J'
1
ll1 point z= zo is not a bn.:·ik w· . ' c, Ilic value or k Illus obtained is negative
I
11 ' •1Ypo1111.

7.

<11 "'" 0. 1- (0 1·10 2 I O_rl 0~)

(~ 11 -= 180 ·In., -(0 rl·Orl O_rl 05)

Si111ilarly, Ilic a11gk or arrival al a11 OJH.:11 loop 1.crn is giv(.; 11 by,

,,,, c-:: ± 180 (2q I I) -11,

\
8. • Find the points when; !he rnol loci crnss the imaginary ;ixis. The ~ints where the roo\
loci intersect the i111agi11ary axis can be found by selling z::: jv in the characteristic
t:quation, equaling both the real part and the imaginary part lo z.cro, and solving ror v
anti k. The valucs or v .irnl \.; thus found give the loc:1tio11 at which the root loci cross
llw i111:il'i11:1rv :t:-. i:; :111d till' ,·:ihll' nr 1111· l'Offc::p111uli111 1•.:ii11 k n·:;pccliV l! ly.
1
•
-' .
 The open loop ga111
, • k 111
• polc-,cro
• forn, nt any point ' -0 on the• root locus ,s
· g,iven h)'.

product of phasor length frolll z 10 open loop poles

~ 0
k product of phasor length fron, z to open loop zeroes ·
0

Drnw ruot locus d111~r11111, in the z-pl1111c for the system

, • ,
111 f'0 llowi11g figllr<.: fur
the followmg I ee sampling periods. T= I sec,
• hr • • •T= 2 sec, ' \ T= 4 sec.
:

--
r(I)

C(I)

Fig 4 _7 (Closed loop control system)

,.
I
Solutioq •

1-c -Ts
(i(z) = 2 [. s

+ (1- e-T -Te -T )z-21

k[( T-1 + e-T )z-1 _....:.._-=-,---=i-~ .
_.,,,,,J = -- -- ( I - z--1 )( I - c-Tz-1)

I. • T=l : For T=l,

Sampling penod

0.3679~~z + 0.7_1_81)
G(i'.) ~' - (z-.l)(z-U.3679)

-07181
zeroes at z-- ·

an<l poles atz-- 1' Z = 0.3679,

124
or.

\\'liidi r.i,·L'S, z~0.<1•17 1>, -2•.K I I ~

,) .__ . ~
~

Sc. draw a circle or radius al centre ;it · 0.7 I~ I. _i,j/

(z - I )(z - .3679)
') = (U<>7'>(~ .;.- (i.11 K
.i l
I

litll(l', Ilic critical 1.!:litl is rolll1d lo he k .a 2.:;n5 ( li.1r st:1h i lity).

----
ry.; ~ \?~{wr't'J
- -""t
\1 11. 1 . ,, .. 1.,

', , ' 11 • 1,rcl

I \I
1K • &•
,
I

\ - o 11~1
K •O
'
1- -
o ::G1'.J , h
K• O
;,/
-- ►
,.._.
\ I

\'
\
o r~ ~

2. Samplt11g.
. • •
p~riod I ··-, \\··1ll 1 T ,c/- G(z)•
• • __ ., ht:Cl)lllCS

.. • . . . l z:::: 0 4 ?l:> 3 and ,.=-=-1.5247.

I he lm;ak way po1nl ,11 c .i • •
 1
K • 1.•l~'..I "' l
, 111, r ,,,(!,. \
~

II ... : ,, . \

I'

Fi!; •I.CJ {IZnot lrn.:us di ·1u r •11 1·

':: ' 11 or th1: • ,.
S) s l1:111 \\' ll"ll
'- 'I'•• .!."' SC\: l
With Sa111pli11g period T 4 sec.

G(z) = 3.0183 (:r. + 0.3010)

·•:_.
(z- l )(z-0.0183)

The break way points arc .it z:: () ]4JS

· • •• • nncl z:a_ 0.9455

l•n I

J • II
,.A. .. fJ

-~-'
'"";;- ~,.
,_/ "'oJ•JJ~

• 'onmJ

Fig 1. IO (!{0111 Iurns diagram !'or lhc s ysll'lll w h 1:11 T ,1s1:t:

1
)

j / ~ ·1 1 •. ••••
Th: crili_c:d gain k li.ll' stability is\.; -0.%53 . Fron icsc l·llJ't:t: c ;1st:s comm crcc 11011cc lha l the
Slllalkr lite s:11l1pling period is. the larger 1l 1

~-8 Drs ii.:11 Bas(·d 011 1h l' Fn·q 1H·11 ry l~l'.SJH111 st· I\ 1d h o<I
Thc .frcquc11cy response concept plays the same rok: in digital control systems as it do_cs i 11 •
' ' '

co11ti .111m1s-limc control systc.1ns. The ram iliarity with 11odc diilgrams (logarilhrni~ pl~ts) i~ .·
lll'tTssary j 11 the c:--tc 11 sin11 of tile ro11vcntional rn.:q11cncy n.:spn11sc tcclrniqui:s lo the an:dysis :

and design or discn.:tc-ti111c control systems.

 11 11 •1, response methods have very Ii' .
fr•'i r ' • cqucnt ly hccn 11. cd in lhc co111pcnsa1111 1b 11•11 'J1,c
., . ~
• •• , •• • • .
1tll' lilts 1s lht.: s1mpl1ctty ol the Ille .
1 d
I ,•,~(111 • • . t lo s. \Vil ,le pcrforniin,,t, frcc1111.: 11cy n.:sponsc tc:-.1.
1I••·n.:1c-t1me system, ti ts i111p 01 lant . ~ . . .. ti c
11l,tl lite systl:111 ha\'c a low-p:1ss Itltcr bt: or 1.:
1111
:1 ,~c • • 1 1
11 so that s1dcba11ds arc fil tered O 1 . . . · ··mt
., ,ph:r • • ll • 111c11 the response or the linear li me 111va n,
~· 10 a sinusoidal input 11rcserves ti r · I cl
s)'S1cn1 ' . le ircqucncy and modifi es only the nmpli tuc e an
ph:1sc or the input signal. Tints, the ,1mplitude and the phase arc the only two qua11tities that
111 ust be dealt with.

~.S.I /fr.,'}HJ/l~i<! ofa LinC'ur Time-ln\'i1ria11t Discrete-1'i111e System to" Si1111.rnit!al Input:-

const'der the stable linear time-invariant discrete-time system shown in Figure below. The. .
• 1to
1npU be system G(z) before sampling is .
u{l) = sin c,1t

u(t)~ _ u ( _ kT
_)_ _...PI Stable linear time -
. invariant discrete-time
T system

Fig 4.11 (Stable 'linear time-invariant discrete - time system)

The sampled signal u(~T) is

u(kT)=sin kwT .
The z trnnsfonn of the sampled input is
z sin roT

The response of the system is given by

z sin wT
X(z) = G(z)U(z) = G(z) (z - ci.,r) (z - c· ,.,r )

=
az . 1 + az- Jt1if + [tenns due to poles of G(z)] .. ......... (4.8)
Z- c'llJ z- C

•
Multiplying both sides . (4 ·S) by (z-d '•>T) /z, we ohtnin
of Eq
• T
- J"' ) Z - C j"'T
sincvT a(z-c +- ~ [tcnns due to poles or G(z)]
G( z ) - -~
-j1uT
=a + z -e i,,,t z
z-c

127
 . second tcnn on the right-hand .
111c s1 0c of this last c .
,,r Since the system considered 1 . quation approaches zero as z approaches
d • . \ere is stable, the . . .
aches zero as z approaches ci(:)T th1rd lem1 on the nght-hand side also .
3ppro • Hence b I .
• .. ' y ettmg 2 approach c'0'r, we have
- G(·'/. ) -
·1 -
'
Sll\(1)
--
1I -
G( t;\ l••I )
z - c-J.. I l•c•~• - - ~
-J
The coefficient a the complex conjugate of . .
a, 15 then obtam as follows: \
G(c-J111T )
• ,·
·1 ::: - - - - -
-J

Eq (4.S) can now be written as

Mejo z Mc-io z
X(z) =
2j z - eiruT - 2j 2
_ eiruT + [tcnm due to poles of G(~))

. iVI ( cin: c- ,11 - J

or. /'(z) = 2j ; ·_ c'~'' - ; ~ c:-; 1• ·I [terms due to poles of G(z)]
.,
The inverse z transform of this last equation is

·(kT)- M ( ,,..,T j ll · i\•iT .,n -I

x - --;,: c c - c e ) + Z [terms due lo poles of G(z)] .... ... (4.9) ..
-J .
/ . .· ..
The last term on the_right-hand side of Eq (4.9) represents the transient response. Since the·
system G(z) has been assumed to be stable, all transient response lem1s will disappear at .
sk:idy stale 1111d we will gel llil' following slcady-slak rl'SIH1111;c x"(kT):

xss(kT)= ~ [c j<\ .,T • O> -e·Jt\,,,T+OJ]=Msin(kwT+O) .... .... (4.10)" '

-J
where M, the gain of the discrete-time system when subjected to a sinusoidal input, is given
by

and 0, the phase angle, is given by

0=0(ro)= LG(J<•>T)

In tenns of G(c? 1T) Eq (4.10) can be written as follows:

x" ( kT) - (i( l'.jr,,T ) I

si 11 ( k r,,
·1· 1
/G, (c jr,,T )
I
 ic,,T) • • •
c shown tIwt G( c mdccd gives ti . • . ,.
\\'C hn'' • . • ic m:ign1t11clc nnd phase or· 1hc frcq11 c11cy r •t,;prn 1 •
Thus to ob tam the freq ucncy re • ~•r r. . . .
of Gtz). ' • . .. sponsc of G(z), we ncc<l only to su~s~itute d· · ,or z •
). The [unction G( c1''11 ) is c0111111 1 • • · •·
jrl G(Z ' on Y called the sinusoid:tl pulse tr111isl'er lu11c.:lloll .
i thing that '.
NO ' 11••> t(2nfl') r =d(l''!' Jln = d•,iT
i:

We find that the sinusoidal pulse transfer function G (d~',) is periodic, with\llC ;Jeriod equal
' . .,

10 'f,

e~a. n,plc 8 .Consider the system

' defined by
x(kT) == u(kT) +ax (k- l )T), 0< a <1 .. . '

w11cn.: u(kT) is Ilic inpul nnd x(kT) lht: outpu1.· Obtain the s1c,1dy-sta!c.: oulpul x(kT) when ~he : ·. ·.
input u(kT) is the sampled sinusoid; or. u(kT) = A sin kcoT. • • • ••. ::; :·.
:: .

Solution . The z transfonn of the system equation is

1
1(z) =U(z) +ai X(z)
By defii~ing G(z) =X(z)/U(z), _we have
X(Z) I
• G(z) = U(z) = I - az- 1

Let:us substitute doiT for i in G(Z). Then sinusoidal pulse transfer function .G(e j~T) can~~ ...
obtained as
. ,. l 1
G(c'IV'):::; - 1. 1 = . . .
1- ae "' 1- a cos cvT + J a sincvT

The amplitude of G( d0ff) is

I
IG(CJr~r )I= M = --;:== = = =
. J1 + a 2 - 2acoscvT
and the phase angle of G(d 0>T) is
.r
LG(e'"' ) = 0 =-tai1 - I -asin-cvT
-
• 1- acoscvT
Then the steady-state output Xss (kT) can be written as follows:

xss(kT) = AM sin (kwT+8)

=-;== A • (k T
= = = = Sm ' CV -tan
_1 a sin tVT )
2
~l+a -2acoscvT · 1-acoscvT
11 9
.. ... .,-..,.
.i :: •._ : J ; " ,..:

az u;u,,ear Tra11.ifor111atio11 ti IC! ,. I3 c,orc • • : :

Pla11 c- we cnn nclva111a gcously ,,1pp• IY.our ,••
(Ill{/ 11' r
J.o• 11
•
r: .
,,ell· developed ,requency-response
. . · method s to the analysis and design of <liscrctc-1_ • •1me ,'

,ontrol systems, certain modifications

. in th e z plane approach arc necessary• . .•Since
• in
• .~he z•••:.

. as z = eiCc>T' if
pJon' the frequency appears
. • we treat frequency
• response in the z plane;. the ••
si11 1' rcilY of the lognn1hn11c plots will 1,c complc1cly lost. Thus, 1hc di reel npp_lication• of ••
1 1
freqtle cy-rcsponse · methods is 'not worthy of consideration. In fact·'·1 since the • z. .-:•
11
rans rrnation maps the primary and comp 1ementary stnps
! • •of the lcft half of the s plane_
; • mto
• ...• •
10
1
thC 11· · circle in the z plane, co11vcnti ona• I 1rcqucncy-rcsponsc
r. n1cthods, which deal with
• • the• :•_
111 1
,,,iira Jell half plane, do not apply to the 2 plane. •

'fhe 1·rr: culty • however' can be ovcrc ome by transfonnation

• the pulse transfer
• function
• •111 11c
J • :•
d 111
, plane into that in thew plane. The transformation, commonly called the w transformation, . ..
! bilinear transformation, is defined by •••• _••
1+ (T / 2)w .......... (4.11)
z == 1- (T / 2)w
I ..,,✓ . , •

Where Tis sampling period involved in the discrete-time control system under considCration.•• •
By Conv~rti ng a gi vcn pulse trans fer function in the z pIanc into a rationa I funeti on of ,y, _ih{ •••
.. ' :,.•
frcqncncy-rcsponsc methods can be extended to discrete-time control systems, lly solving
Eq (4.11) for w, we obtain the inverse relationship. :•:
... .. ..... (4.12)
z z-1
w=--
T z+l
-
Through the z transformation and the w transformation, the primary strip of the leri half of .:
the s plane is first mapped into the inside of the unit cire le in the z pIan c and then map peel·•••• •
inio 1i1e entire lefi half of the w plane. The two mapping processes are depicted. in figure
. ·•. •

below. (Note that in the s plane we consider only the primary strip). Notice that·the origin of :'. :
~ .'
the z plane is mapped into the point w -2rf in thew plane. Notice also that a varies froni-0. ·_:.. ·:.

lo /<o,/2 along the jw axis in the s plane, z varies from I to -1 along the unit eircJ_c in the z . • ;

plane and w varies from Oto oo along the imaginary axis in thew plane.

130
 11 "' J 111,n ,

) • ••• •••• .. ••••• .. • •) • II

, , . \..~

l • I

Ill
f I

- the hJl ha l r or the \\' phnc

1\ lI.1111 1!\t ' coi--responds lo Ilic illl it~ inary a~i s or the s plane. there
1
.,1Il.. diffcn.:nccs
· hdwcc11 the lwo I)la1H: s.. ·1·1le, c•111d
· , • d1l• •• • lhal the bcilav1or
kn.:llct: Is • 111
• the s

1;111: (l\'Cf lhL: rrcq11c111.:y ran ge 2 <ti_ .--__,, <tJ :.; <ti, 111:ips 111 the range -rfJ < v < rfJ, whl:rc v
1) 2
..
1~ iht: lil·1i1i1111s frl·q11c111.:y i11 w 11h• 11 1.: • Tlii·:,. 1m:a11 s 11 1a 1, al I I1out:I I II 1c I·n.:q m:ncy rcspon :-.c

ch;Hadcrislics or the analog filter will be reproduccJ in the discrete or digital · filter. thi.:
rrcqll'.:ilCY scale Oil wh idi the rcspo11sc occurs will bc co111pn.:sscd rrom all i:1fillite interval i11

lhL' an:iln): li ltcr to a linitc i11tc1Yal i11 the di1:iIal lilt cr.

Once 1hc pulse tra11srcr function (j(z) is 1ra11sror111cd i111n G(w) by means or the "
1ra11sk,n11atio11. ii may be treated as a conventional trans !cr runction in w. Conventional

lil·1;ucIK)' respon se tccl111 iqucs ca\\ then be uscd in thew plane, und so the well -establ ished

frcq1ic11cy-rcspo11se th:sig11 Lcclrniqucs ca n be app_licd lo tile dcsign of discrete - lime control

As noted earlier. v rep resents the lictilious frequency. 8y replacing w by jv, conventional
frc<pcncy-rcsponsc Lcchniqucs may be used to draw the Bode cli i1gram for the transfer

r1111~1io11 in w.

J\llho.1~h thew pl ane resembles 1he s pla1_1c gcomclrically, tile frequency axis in thew plan e
~ ,.
i-; 1· •t . , I 7'1 . 1- -t · · )tt s frcl ucncv v and the actu:il rrequcn cy c,> .ire related :1s fo llows:
· l IS 1)1 1U . l\.: ll 111 1 • I •

\\·I "

1 ·oil'
j \;ill I.I I
; . ... . .,.,.-

. '
..

? rJI'
_ ::-1nn-
,- T
I
2 ' .......... (4.13)

4 13) civcs the relationship between the nctunl frequency<,> and the nc1i1ious frcq11 ~11cy
!)l ( • .
tc that as the actual frequency w moves from -~cv lo O the· fictitious frequency v
,,. r-Jo . 2 s _ .

froll1 -co to 0, and as Cl> moves from Oto ~ w v moves from Oto Cf.l .
010\'C~ 2 J,

• g to Eq (4.13) th e _actual frequency Cll can be translated into the fictitious frequency v.
p.cfcrr1n .

le 9 . Consider the transfer.function system shown is figure below. The sampling·

£xa111P
. 0 Tis assumed to be 0.1 sec. Obtain G(w).
~~ • .

__ / I -e
s
-T,

1---~ ,:io 1--►

·. .

'-------.J

I
I
! Fig 4.13 (Transfer function system)

Solution . The z transform of G(s) is.

l - c -rs IO ]
G(z)=Z - - - -
[ s s + 10

_1 z[ lO ]
= ( 1- z ) s(s + 10)

0.6321
- z-0.3679
l3y use of the bilinear transformation we get,

1+ (T / 2)w 1+ 0.05w
2
= 1- (T / 2)w = 1- 0.05w

G(z/ can be transformed into G(w) as follows:

0.(d 2 l 0.6321 { I - 0.05w) .·.,
G(w) = 1+ O.O5w = O.6321 + 0.O684Ow
_ . - - -- 0.3679
l-O.O5w
I - 0.05w
- 1> 1 ,11 •
•• ·- w ·t-~.2-tl
D2
Notice that the locntio 11 or 11
,e pole or the J)I • . )hnc is
') l.l
\' •• ' -
I ' The "',11· 11 \ '· ii
o II
.
ll t,; I
ant ts s-=--10 and th:it of the 1wlc 111 I1,cw I '
i 11 I 11.: s pla11i: •. II 'I I . 1Jntl1
. pole locations and ti . is l illHI that i11 tltt.: w pl:1111.: is 'J.2'11. ( tits .
t1ll- le g.i1n values arc • ·1 However.
\\') has a zero at w ===?fT _ ' Simi ar in tlic plant and the w plane).
1
G( - - -0, although ti1 I
c P ant docs not have any zero.
li1n . 1'
w - > tJ C,(w) - 1111 10
. s - > () !H· I 0
This fact is very uscful in checkin, .
g the numerical calcu lat ions 111 transforming G(s) into
G(w).

To summarize, the w transfon • .. . . • ·

nation, a bilinear transformation, maps the 1ns1dc ol the urnt
th 2
circle of e plane into the lcfl half of the w plane. The overall result due to th e
transformations rro1n the s )I • 1- . . .
I ,lllc 1110 the z plane and from the z plane 11110 the w plane 1s th.ii
the w pl:111c and the s iihii. . . . • • • · ·
' c ,lie s11111lar O\'cr Ilic region of' i11 1crcst i11 tll1.: s pla ne . 1111 !> 1s
because some of the distortio ns c,1uscd
. . from the s plane 11110
by the lransformat1011 · the z pl ane

'
arc partl y compensated ro r h , ti 1e tra11s,orma11
j
r: •
on )
·
from the z plane mlo thew plane.
Nlllt.: th'at i r

G(7.)=
boz"' + b,z m- 1 + ... ... + b"'
1 rn $ 11
/ ' I :I 1z"
• •
I .. .. . I ,I

\Vhi:rc tile a,'s and h,'s arc co11sta111s. is tra11sli..1nm·d i11lo tilt.: w plane hy lhL· lr,111s f"onnalio11.
I + tT / 2)w
z= · -- - - -
l-(T/2)w
then G(w) takes the fo1111

G(w) = f-'o ll'

fl II fl 11 • 1
+ f-'oll' + .... ... + n
/J
a Q \\," + a I II'n-1 + ...... . Ta,
,

Where the a i's and the Bi's arc constants (some of them nrny by zero). Thus, G(w) is a ratio of
pl1lyno111ials in ,v. where the degrees of the numerator and denominator may or may 1101 he
e4ual.

,t8.3 Hotle Diagm111s:- Design by means or Bode diagrams lws been wide ly used in
di.:aling \\'ilh si11glc-i11put-si11glc-output continuous-time control systcllls. ln part icu lar. ir the
transfer function is in a factored form, the simplicity and ease with which the asymp tot ic
Bode diagram can be <lraw11 and reshaped arc well known.
,\'!, '!,lnkd t'arli cr, 'th,
CC'll\V('lll i I r!
lllln
frcq 11 , •'✓ 11 1c 1,.. 111 •·frr
fu11cli1)1'S in the w Phti, •• cnty -rc1: pn11 Hc 111c1ho d ; appl y lo ; ' •• ,
1
, c. {ccall lhnt th 13 . I . II c,;
. . . . c ode d1agra111 consists of two separalo P ot S, 1
l01~.1nth1111 c magmtudc jGU )j ,.. . . •
v vcrsu~:log v. I I G(') ... log v
'k ·t ·I • , 1· I . - ,Int I 1c pha se :u1glc L JV vc1 s u s •
S 1.: l 1111 1;. o t 1c 1 ognrrth .. . .
llll<.: 111a11111 l ucle · • I
k 011
th1.' principle O r add • , , . . ~, 0 11 the fac1ori 11g o I' G(j v), so that It wor s
is )nscd
lllb the IIH!iv1dual r, , . . . . . . . .
~ - •. . . . ,Wlorccl tcn11s 111slcad ol'rnulliply111 g ,11d, v1d11al tcn11 s.
1
J•mrn 1,11 ,1symµto1 1c ()Iott · . •
ing lcclu • . •
. k Hques can be applied, nnd therefore the magnitude curve
can be qu1c ly drawn by · . . •
. llSlng st raight-line asymptotes. Using the Bode diagram , a di gi tal
colllpcnsatory or d1gil'II c" 110 . •
. . ' ~,n 11 er may ' b_e <Icsrgncc
• I with
• convent1011al
• • tee I1111ques.
design •
It 1s important to note lli·t( ti :-,
. ,
. 'Y I'Jc n d'tffcrcncc in the high.frequen cy rnag111lu
•. lCrc. Ill'\ • d cs 1·_or
GUrn) curve and GUv). The 1i· ij 1'.. .. . . .
ig frequency asymptote of lhc lognnlhm1c magrntude curve for
G(jv) mny he a constant _ de -1 ,11 . .
Cl lt.: inc (lha1 1s, a horiw11tal line).
On the other hand, if lim G( ) _ . . . •
s- ,,,, s - 0, then the magn1t11de ol GU(J)) always npproachcs zero
(·"' llfl) as rn approaches infin it F . .• • . <l
. Y: or cxampl~, rcfcrnng to the previous example we obta111e
G(w) for G(s) as follows: •

1
G(w) = 9.241( - O.OSw)
w + 9.241

The high-frequency magnitude of GGv) is'°

Jim IGUv)
. I== lim ( 1- ·0.05jv)
9.24 l . = 0.4621
V ➔ CO V ➔ er., JV+ 9.241

While the high-frequency magni.!udc of the pla~t is

11111 I I II I ,),, "". ' l

◄" • . .. j 111 I I II ,

The di!Tcre11cc in the □ode di1gra~1s at the higi1-freqL;ency end can be explained as follow~.
!; ' . ' •.• .
• , f~~ • ·1
First, recall thnt we arc inl cr~slc~.only ~i,_i the frequency range O $ OJ $ - rv,, which
,. -:; 2
. ~' •• "d ' .
• .. ,. ': l
corresponds O~ v ~ co . Then, noting hat v= ~ i~_.th~.w plane corr~sponds to m = in the
; ·., ·.:· .\ • ".· .
2 (J}s

. . ;,
constant. (It is important to note that these two .values arc generally not equal to each other.)
•.1,'

134
 from the pole-zero point or view it
•
b . I . I.
cnn c s:11cl that when GUv) 1s a non •
zero cons1:111t at·
" :e: co it is implied that G(w) co t , ..
ns ants the same number of poles an<l zeros.
In general, one or more zeros of G(, ) 1· . . I Tl e presence of a
v 1c 111 the nght half of thew p ane. 1
;,'Cl'll i11 1he right h:it1· nr th, 1 I . . I c; 1n111~ rc.:r
1 c " P a n c 111ca11s 111111 G(w) i H 11 1m1111111 11111111n P 11 1 11

fu11c1ion. Therefore, we must bc careful 111

. drawing the phase angle curve in
• the Bode
diagram.

Procedures to drmv Bode diagrams:-

Let us consider a digital control system as sho~vn in figure below.

C(z) •
R(z)
Go(z) Hold Plant

Fig 4.14 (Digilal conlrul sysrcm)

. -
I. First, obtain G(z}, the z transform of the plant preceded by a hold. Then transfonn •
G(z) into a transfer function G(w) through the bilinear transfonnation.
1 + (T / 2)w
z=
1-(T/2)w
. ...·
2. Substitute CD= jv into G(ro) and plot the Bode diagram for GUv).
Example 10 . Draw a Bode diagram for the control system shown below. Assume that the
sampling period is 0.2 sec, or T=0.2 and the value of k=2.

r(t) c·( t)
Gu*(s) I -c -Ts k
or C(z.) .
R(z) s s(s + I L .
Go(z)

Fig 4.15 (Digital control system)

135
 = 0.0l 87J'. ~ ~ o.93 -G) !
L(Z-1 )(Z-0.8187) j
_ k(0.01873z·+ 0.01752)
- >
I -C-l.8187z-'-0.8187
I

I - -
~ext, we tr~msfonn the pulse transfer function G(z) into :1 transfer function G(w) by 11,1ean~ of ·:
the bilincnr transfonn:ition.

z -- .l .+. (T . - . - 1 1- 0.Jw'
.. / .2)\\'
. - 1 (T / ~)\v .· I O. Jw

Thus .

/ o.OI 87f l .!. O. l ,v !+ 0.01752 ·1

G(w) =
I
L •
1- 0.1w ) .
-

1+ 0.lw l · -l.818l l + O.lw ·+ 0.8187

( 1-0.lw .- U-0.lw )
k( - 0.000333w 2 - 0.09633w"" 0.9966
= w 1 ~ 0.9969w

k( I+ - \\' X 1- -\V \ 1
:__ ~ 300 10 )
= .w{w +'l)

Ily s~tt ing h =2, we plot the Bode diagram of G(w).

0

2(-Q,()(){)3J3\V! - 0.09633\V T 0.9966)

Ci(w) ::: \'/ 0.9969w

'.ZJ( l +w- X1--

. w "!
\, 300 \ 10 ;
\\'(\\' + 1)

136
F~t!•11·c 4.1 (1 shows lhc bode diagram ror the ~ys1c111 . For the niagnituclc cmve we have used
s1rai \~ht - li11,: asy 1Hpl11ks. The 111ai:11iludc a11d the ph :1sc a11 1;k ur (i(jv) :ire show1 1 by dasll cd
d
1.'.llr'.r::s. TIH.: phase l\largin ca11 he read fro 111 lk limk diagr:1111 (dash ccl curv ' S) as 3o" all lilt:

gain 1•1argin as 14.5 dB .

Fig 4. \(i(\lt!.k cliagr;im for lhc: cxampk \0)

..t.9 t Analylit.'lll Design l\lcthod

ll t!1is ;cction we spccilically present an analytical tksign 111dhod for digital controllers that
will force lhe error sequence, "hen subjccteu lo a speci 11c type of time-domain input, to
bcc):ne.zero aner a fin ite number of sampling periods and, in fact, lo become zero and stay
1.cr11 ;•flcr lhc mi11inw111 possilik 1111111\icr or sa111pling peri od s.
If 111,~ response or a closed-loop contro l system to a step inpul exhibits the minimum possible'
scllli11g 1i111c {lhal is, lhc oulpul rc,1ches lhc final value in the 111inirnu111 time anti stay ther e),

no ~lcady -stale error, and no rippl es helwL:c11 lhL: samp li n!', instants, then thi s lypc or •
resp1)11sc is commonly called a deadbeat res!1011sc.

t:r.'
 J)esi~II of Di~ita/ Co11trol/crs for Minimum SC'ttli11g Time 11•it!, 7.C'ro Stl'{tt/y-Stute frmr:-
Considcr the digital cot 1t 1 . I • h • the
' ro system shown in figure below. The error signal e(L), w llC 15
difference between the· ( · · )T
mput r L) and the output c(l), is sampling every time interva •

r(I)
~c(kT) r----~
T - Oigilal
1-..--- ►
Con1rnllcr -►
hold

Fig(a)

C(z)
R(z)
U(z)
G(z)

Fig(b)
Fig 4.18 (Digital control system and the diagram showing equivalent digital control system

The input to Lhc digitnl co11troller is the error signal c(kT). The oulpul ol' the digital controller
is the control signal u(kT). The _control signal u(kT) is fed to the zero-order bold, and the
output of the hold, u(t), which is a piecewise continuous-time signal, is feel to the plant. It is
desired to design a digital controller G0 (z) such that the closed-loop control system \yill
exhibit the minimum possible selling time with zero steady-stale error in response to a step, a
ramp, or an acceleration input. It is required that the output not exhibit inter sampling ripples
after the steady state is reached. The system must satis_fy riny other specifications, if rec'1uircd,
such as a specification for the static velocity error constant.

Let us define the 7. transform of the plant that is preceded by the zero-ore.I er hold as G(,,), or

Then the open-loop pulse transfer function becomes Gil (z) G(z), as shown in figure (b),

138
Next. di.:li11e lhc desired cl ,I1
I ·
USCl • uop pulse transfer fu11ctio11 as l·(z).
C(z) _ G 0 (z)G(z.)
R(z) - l +Go (z)G(z) = F(z) ..........(4.14)

Since it is rcquin.:tl lllat the s~slcm exhibit a finite scllling lime willl zero steady - 5 lale error,
th e 5Yst em muS t exhibit a finite impulse response. Hence, the desired closed-loop pulse
transfer function must be of the followiilg 1orn1.
r

II /. N I II / N. I
1-'(1) - II I , I . .. ...... I "N
zN
or,

F(z) = ao + alz-1 + ....._.... .,. .·1 N z-N

. .... .. .... (4.15)

\~here N '2:. n and n is the order of the system. [Note that F (z) must not contain any tenns
with positive powers in z, since such tenns in the series expansion of F (z) imply that the
nulpllt pn.:ccdcs lhc i11p11t, which is 1101 possible for a physii.:ally realizable syslcm.l In our
I

design approach, we solve the closed-loop pulse transfer funclion for the digital controller
Gn(z). That is, we find the pulse transfer function Gn(z) thnt will satisfy Eq (I, 14). Solving
Eq (i( 14) for G11 (z), we obtain,

G (z) = F(z) ........ .(4,16)

D G(z)[ 1- F(z)]
The designed system must be physically realizable. The conditions for physical realizabi~it~ ., .· •
place rerl:1i11 co11slrni11ts .<111 the l'loscc!-loop p11lsc transfer f'um:li1111 F(,.) :ind the diµ,ilal
co11trullcr plus transfer l'llnction U1> (z). The condition for physical realizability may be statc<l
as follows:
l. The or<lcr of the numerator of Gn(z) must be equal to or lower than the order of the
<lcnominator. (Otherwise, the controller requires future input data to produce the current ·.- .
output).
2. If the plant Gr (s) involves a transportation lag e•t.5. then the designed closctl-loop system ~
must involve al least the same magnitude or the transportation lag. (Otherwise, the
closed-loop system would have to respond before an input was given, which is
impossible for a physically realizable·system).

139
'.\. Ir (1(1.) is expanded into n series in z· ' tl1e lo l I • o. f 1·l1c seri es expansion of
• •, wes -power crn1 •
F(z) in i' must be at least as large as that of G(Z). For example, if an expansion of G(z)
• • -l b 4 I 5)
into a scnes m z cgins with the z·' term, then the first tenn of F(z) gi ven by Eq ( •
11111st h!.! zero, or ao must
•
Ct]ti·tl
'
o·, ti la t •Is, t11e expansion
, has to he o rl I1c ,ol
r. •ll 1•
1 2
F( z) -a
- 12· + a2z• • +............ + aNz N
where N ~ n and n •is the order of the system. This means that the plant c_annot respond
instantaneously when~ control signal of finite magnitude is appl ied: the respond comes at a
delay ol'al least one sampling period if the series cxp;nsion of G(z) begins with a term in z·'.
\

In addition lo the physical rcalizabilty conditions, ~vc must pay allcntion to the stabili ty .
aspects of the syS tem. Specifically, we must avoid canceling an unstable pole of the plant
by a zero of the digital controller. If s~ch a cancellation is allempted, any err~r in the _pole-
~ero cancellation will diverge as time elapses and the system will become unstable . .
Si111il11rly. tile di~ilal ct111tnillcr pulse tn111sl'er f\ 1m·1io11 should 1101 i11volv • 1111slahh: poles 10 . •
1:1111n:l pl:111\ zeros that lie out.s ide the u11it circle,
I
I

Next. let us investigation _what will happen to the closed-Joor pulse trnnsfer function f(z) if ·
. .
G(z) involves an unstable (or critically stable) pole, that is, a pole z= a outside (~r .on) the·.· ..
. . . .. .
unit circle. [Note that the following argument applies equally, if G(z) involve~ t\~O or more •. •
1111st:1hle nr critically stable - poks,l I.cl \If; dcfi11c

'i, ( / )
( 1( /,)
z -· tJ
where G1(z) docs not include a term that cancels with z •a.Then the closed - loop pulse . ·••
transfer function becomes

.......... (4.17)

Since we require that no zero of Gu(z) cancel the unstable pole of G(z) al z = u, we must
have.

140·
•i·1,·it
' is. 1-F(z) must have z• = ,., .,
u. uS a 2C • • . • OfG(zJ,
• ,o. Also notice that from equation (4) if ;,cros
lo 11 ot cancel poles of G (z) ti • I ,
c D • , 1c zeros of G{z) becomc zeros ofr.( ) [r.(z)
1· z . 1• •
may invo vc
ad litionnl zeros.]

Let us s11111111nrize what we have st·1t 1 . ..

• • e( conccrn111g stab1 lily.

I Since the digital cont II G ·· • , bl )'

• • ro er o(z) should not cancel unstublc (or cnt,cally Sla e
poles of G(z), all un st able (or critically stable) poles of G(z) must be included 'in.·.
1- F(z) as zeros.

. .'.- ....
2. Zeros of G(z) that lie inside the unit circle may be canceled with poles of Go · (z).
However, zeros of G(z) that lie on or outside the unit circle must not be canceled wil'1
poles of Go(z). Hence, all zeros of G(z) that lie on or outside the unit circ.lc must be
included in F(z) as zeros.

Now wc shall proceed with the design. Since e(kT) = r(kT) - c(kT), rcforri11g to Eci'( 4. 14) ,vc
have,
I

E(z) = R(z} - C(z) =R(z) [I- F{z)] ......... (4.18)

Note that for a unit-step input r(t) =l(t)
1
R(z) = _
1- z l
For,a unit-ramp input t(t) =ti (I),
.,. -1
'/.
R(z) = - - - -
(l-z- 1)2

And for a unit-acceleration input r(t) = ½t2 I(t),

I /, Il
·1•i / , - I ( I
R(z} = -· ·-· ------
2( 1- z- 1)3 •
Thus, in general, z transfor~llS of such time-domain polynomial inputs n1ay be written as
P(z)
R(z) = -I q + I ..........(4.19)
(1 - z )
1
where P(z) is a polynomial in 2· . Notice that for a unit-step input P(z) =1 and q=O; for a unit-

ramp input, P(z) =Tz·' and q=l; and for a unit-acceleration input, P(7.) =½ T2;,-: 1( 1+z· 1) :ir!d
q=2.

14 l
 £3)' substituting Eq (4.19) into Eq (4 1S) .
• we obta111
E(z) = P(z)( 1- F(z)]
,_
(1-z-l)q+I ..........(4.20)
To ensure that the system rcacl , . st. . . • · I· . J : .'
ics c,idy state in a linite number or s;1111pl111g pcnol 5 •111 :
maintains zero steady-state err E( ) . • b f
or, • z must be a polynomial in 2· 1 with a fi111tc num er 0
tcnns. The, by referring to Eq (4 ?Q) ••
·- , we choose the function J-f(z) to be of the fon11. ..
l-f(z)=( l-z· 1r+ 1N(z)
........... (4.21)
where N(z) is a polynomial in z-1 with a finite number of le,rn1s. Then, .
.•.·, .
E(z) =P(z) N(z) .. .
......... (4.22) .
1
which is a polynomial in z· with a finite number of terms. This means that the error .signal
becomes zero in a finite number of sampling periods. .' , "
From the preceding analysis, the pu!se transfer fu~ction of the digital controller can b~
dcl;pnn incd as follO\vs. By ~rst letting F(z) satisfy the physical rcnlizabi lity and stability · •• •_·
1 conditions and then substituting Eq (4.2 I) into Eq (4. i6) we obtain.

Go(z) = F(z) .......... (4.23)

G(z)(I- z-l)q+lN(z)
Eq (4.23) gives the pulse transfer function of the digital controller that will produce_z.e ro
steady-state error after a finite number of sampling periods.
For a stable Gp(s), the condition that the output not exhibit intcrsampling ripple~ after the·
scllling ti111c is reached may be written as follows:
c(t~ nT) = constant, for step inputs

c(t ~ nT) = constant, for ramp inputs

c(t ~ nT) = constant, for acceleration inputs

The applicable condition must be satisfied when the system is designed. In designing the ·· _.
. ..
system, the condition on c(t), c (t). or c (t) must be interpreted in tc_rms of 11(t). Note that the .:
plant is continuous time and the input to the plant is u(t), a continuous - time funct_ion; :_.
therefore, to have no ripples in the output c(t), the control signal u(t) at steady state must be
• ...
either constant or monotonically increasing (or monotonically decreasing) for ste~'. ramp, 01:

acceleration inputs.

' j
 l:x111nplc l l • Consider the digital cont 1· I . I "tic r1 la11t trnnsfcr
•· 0 system 111 figure below, w 1crc 1

function Gp(s) is given by I

Gp (s) = -s(s_+_I)

1(1) c(t)
c{kT)
/..,, r - -- - - .
---- T -► lligi1al
C11111rnlb

Fig 4.19 (A Digi1al con1rol system)

Design a digital controller Gn(z) such that the closed-loop system will exhibit ad~ad_b~at _
response to a unit-step input. (I~ deadbeat response the systc,ii should not exhibit i_nt_cr
~an1pling ripples in the output after the settling time is reached.) The snmpling period T is
w:'.,11 11wd li 1 lw I i:n· . Thl'11, w:i11p, 1111' dip,ilnl r111lli11lln < i 1, (1) rn • dl'!: ip_1wd, l11 vc:::1i11,:1l<i° lltr

n:sponsc or this system to a llllil-ra111p input. •

Solution . The first step in the design is to detem1ine the z transform of the plant that is
prcc·cded by the zero-order h<:>ld.

G(z) = 2[1- c-Ts • I

s s(s+I)
l

-(\
- -z
•
- 1) [ z-
---- - --
1

( 1--;,_- l /
1 - + --- 1 --
( ~ ) l · OJ<>7 1J:t.- I
l
0.3679(1 + 0.7181 z-l )z-l
.......... (4.24)
== ( 1- z- 1)( 1- 0.3679z- I)

Now redraw the block diagram of the system as shown in figure below.

143
ll(t.) E(z)
U{z)
G(z)

Fig 4 .20 (Diagram showing cqui\'alcnt digi1al control system)

Deline the closed-loop pulse transfer function as f-(z), or

C(z) G 0 (z)G(z)
R(z) = I+ Go (z)G(Z) = F(z)

Notice that if G(z) is expanded into a series in 2 · 1 then the first term will be 0.36792· 1. I lcncc,
F(z) must begin with a tern1 in 2· 1.

Hence, ,we can assume F{z) to be of the following fom1.

I

' F(z) =a1z-1+a2z-2 ........... (4.25)

Since the input is a step function, we can write l-f(z)=( l-z" 1) N(z) ........ (4.26)

Since G(z) has a critically stable pole at z=l, the stability requirement states that 1-F(z) ~ust
have a zero at z= I. However, the function 1-F(z) already has a tern, l-7: 1 and therefore
I
satisfies the rcqui rcmcnl.

Since the system should not exhibit inter sampling ripples and the input is a step function, we ·
require c(t ~2T) to be constant. Noting that u(t), the output of the zero-order hold, is a
. .
continuous-time function, a constant c(t ~2T) requires that u(t) also be constant fort ~2T. In
terms of the z transfom1, U(z) must be to the following type of series in z· 1.
U(z) =b0 +b17: 1+b(z·2 +z· 3+z·.i+ ....... )
Where bis constant. Because the plant transfer function Gp (s) involves an integr~tor, b must _ •
be zero. (Otherwise, the output cannot stay constant). Conscqucnlly, we have
1
U(z) = bo+b1z·
From Fig (b), U(z) can be given as follows:
 U(z) = C(z) = C(z) R(z) - . . R(z)
G(z) R(z) G(z) - F(z)G(z)

I( I . -I)
= F(z) -
1
-z (I-OJ67.9z-1) J-0.3679z- 1
1- z- 0 3679 -1 1 = F(z) - - - - -- 1- 1
• • (l+O.?l8Jz )2- 0.3679(1+0.718lz- )z-
1
for U(z) to be a series in 2· w·t~ I . .
1 1
on Ytwo tenns, F(z) must be of the following fonn .
1
F(z) :-e(l 10.71811." ) z·'F, ........... (4.27)
where FI is a constant. Then U(z) can be written as follows:
U(z) = 2.7 I 8I( I-0.3679z· 1)Fi ..........(4.28)
Eq (4,28 ) gives U(z) in terms of F1. Once constant F1is determined, U(z) can be given as a
1
series in z· with only two terms.

Now we shall determined N(z), F(z), and F,. By substituting Eq (4.25) into Eq (4.26).·wc
obtain.
•I ., I
I-a 1z -u~z - = ( 1-z' )N(z)
The le fl-hand side of its last equation must be divisible by 1-z·'. Ir we divide the lcft-han<l
side by I 1-z·', the quotient is l+(l-a 1)2· 1 and the remainder is (l-a 1-a2)z' 2 . Hence, N(z) ·is
dctcnni;,cd as.
N(z) =I+( l-a1)2· 1 .......... (4.29)
and ·the remainder must be zero. This requires that
..... ..... ('U fl)

Also, from Eqs (4.25) and (4.27) we have

F(z) =a1z· 1+a1z"2 = (1 +0.7181 z1)z· 1F,
Hence,

Division of the lcfi-hand side ·o f this last equation by 1+0.7 I 8 I z· 1 yields the quotient a, a~u.l
the remainder (a 2-0.718Ia,)z·' . By equating the quotient with F, and the remainder with ·
zero, we obtain.

and
a2•0.718 l a1 ;;;Q ..........(4.31)
Solving Eqs (4.30) and (4.31) for a, and a1 gives

145
 a, =0.5820,
Thus. F(z) is dctcnnincJ :is
F(z) =0.582oz· 1+ 0.41802·2
.........(4.3 2)
:ind
F, =0.5820
l:q (4.2<J) ~ivcs i

"''; 1-1 1 11,,11rn.,· 1

........(4 .33)

The digital controller ptilse Irans,cr

r fiunction
· Go{z) 1s
· then dctcrn11neJ
· as follows .

GD (z) = F(z) -· _
u{z)( I - z- f JN(z)

= (l+0.718lz-')z- 1 (0.5820)
0.3679(1+0.7J81z- 1)z-' _, _,
(1-z ')(l-0.3679z ' ) {1-z )(l+0.4180z )

}.5820- 0.582Qz-l
=
I + 0.41802- 1
Willi IIIL' digi1al co111111llcr 1!111s dcsi,:11cd, th~ closed-loop pulse transfer function become
follows.
C(z) ,
R(z) = F(z) = 0.58202- 1
+ 0.4 I 80z-·

0.5820(z + 0.7181)
=
The system output in response to a unit-step input r(t) = 1 can be obtained as follows.
C(z) =F(z) R(z)

? .. - 1
=(0.5 8_o_ +0.4J8o_.. 1 > I _,
1-z
= 0.5820l + l
., ·2 .
T z-J + z• -I+......... .

c(O) =O
c(l)=0.5820
c(k) =I. k=2,3,4, .. ....

Notice that substitution of 0.5620 for F1 in Eq (5.28) yields

i
 l l(,.) ~ - . 71, I ( l-0J(17'h'') (0.5820)
- I .5 , -0-0.58 20 z' '

TlnI:-. th...: control sigll'\I u(k) l . . .. ·, l ing

~ ' >cconics zero fork ~ 2, as required. Tllcrc is 110 111 tcr s,ulll
ripple in the output after the sctt'i t· .
ng 1111c ts reached.

i ext. let us im cs ti gate the r f . .

csponsc o !Im system to a unit-ramp mput.
( '(z} c- F(,.) R(z)

-1
= (0.5820 z· 1+0.4 l 80 2 -2) __z---,-"'T'"
(1-z-1)2
2
==0.5820 2· +1.5820 2· 3+2 .5820 z-4+3.5820 z· 5+.......... .
For th e unit-ramp response, the control signal U(z) is obtained as follows . referring lo Eqs
(4.24) an<l (4.32),

. ) _ ~ (z) F(z) f(z) z-I

U(
/ z - G(z) == G(z)R(z)= G(z)(l - z-1)2.

2
-1
= ( l.5820 - 0.5820z-l) _1
1- z
= I .5820 z·' + z•2+ •z•3 +z-4+ .......
The signal u(k) becomes constant (b==I) fork?: 2.· Hence, the system output will _not exhibit

inter sampling ripples:

Note that the st.11ic velocity error constunt k" for the present system is

Kv =
2
lim [l-z-
➔
1
1
. T G0 (z)G(z)
l
Jim [ _1 F(z) ]
= z➔ 1 (t-
2
) (1- z-l )N(z)

2
= lim 0.58202- l + 0.4180z- = _
0 7052
z -► I 1 ·I· 0.41801.- 1

Thus, the steady-state error in the unit-ramp rcspon~c is

l
Css = - = 1.4180
•• kv

• 14 7
 Exninpk 12 • Consider II tk:si , ,. 1t 1lr11
. . • L\ 11 I rnbkrn tli c s111111.: as thal or previous cx:1 111plc cxcc1 '
thl' st:111r ,. lnrny crrnr con . . k . . · l
. . . Sl,UH 1• 15 spcciftcd. (lkca11sc or thi s :1dcli1 io11 con st raint. 1 JC
sl.'111111~ t1111c will he lon!-\er tlnn , .. . . . . n is
. '" ' - sec.) I he block diagram of' the d1 g11:il ontrol sy 5ic, . •
sho\\'ll 111 figure below. The 11 . . . .
I ant transfer func11011 G1i(s) under cons1dcra11011 1s.
I
Gp(s) ::: -- - -
s(s + 1)
The design specifications n (1) 1 . . . . ·
< re t rnt the closed-loop system is to exh1b1t n firntc sett 1mg time
with zero steady-state error • ti • • I ·b· · t
in le u111t-step response, (2) that the output 1s not to ex 11 11 111 er
sampli11° ripple af1cr the s ttl' • • • • • · k•
~ c mg tune 1s reached, (3) that the static vcloc1ty error constant \·
1
4
is to be sec· , a nd (4) that the settling time is to be the minimum possible that will ~a;_isfy all
th ese spcci fications. The sampling period T is assumed to be I sec. Design ·a, digi t~!

controller Gn(7.) that satisfies the given specifica tio ns. After the conlrollcr is desi gned.
i!wcsligatc lhc response of the system to a unit-ramp input.

Solution , The z transform of the plant that is preceded by the zero-order hold was obtained
I

l
in the pl'cvious example as.

G(z) = 2 l - e-Ts • 1
[ s s(s +1)

0.3679( l + 0.718 lz-l )z-l

= (l-z- 1)(1-0.3679z- 1)

Define the closccl-loop pulse transfer function as F(z).

C(z) = GD (z)G(z) = F(z)

R(z) I+ G () (z)G(z)

Si 11cc the first term in the expansion of G(z) is 0.36 79z: 1• f,(z) 1_1111s t hcgi n wi th a term in ,." 1.
r.( z)
1· =a 1 z•I +a~z .~ +........ -,·a~z
, .• ~

where N ~ n and n is Lhe order of the system (that is, n =2 in the present case) . Because of the
added constraint, ,vc may assume N > 2. We shall try N =3. Thus, we assume.
1
1-(
• z) =a1z· l •I a~z • •I a.1 •z.J .. ... .. ... (4 ·-34)
(If a satisfactory result is not _obtained, we must assume N > 3.) Since the inpti't is a step
function we require that.
1
J-F(z) =( l-z· ) N(z) ... ... ... (4.35)

14S
N11l' 1hal the pn:sc111 or;\ crit ' ,. II . ;( Z)
ic,1 Y. tahl c 1101,. . . .. ,, r.,r (1111ct1 011
. ' . . I 1:(• ) I C ,ll z n ) Ill the pl:1111 pul. C (1,111.,,c
r~lJ\ltl t:!i • z lo ia, c n zc1·c) t I z·'
• a z =- I 11 I r • tc.:rlll • •
. ~ . . • • owcvcr. lhc r1111cli o11 1-f,(z) alrc:1dy w., "
nn \ 1hc1c 01c snt1sfics the stnb 'I• . ,
. I ity rcqu1rcmcnt. .
The n.:q11in.:111c111 that the ~lat •. , ·I . • .
ic \ c oc11y err . .1
1 .- 11 , 1 •1 s follow s.
lim [ _ z - I ] 01 co11s1a111 be t sec c:111 ic w, 1 c1 ' ·
1
k,, = z ➔ I -:f"- Go(z)G(z)

= lim [( I - z.-1 F(z) ]

Z-► I )- - - - -
( I - z- I )N(z)
F(l)
= N(l) :::: 4

Where we used q=O Not· I ·

• ice t iat Eq (4.35) we have F(l) == I. Hence, k1, can be written as
follows .
I
kV - - 4
- N(I) = ......... (4.36) : ·,

Since tlic sySlem output should 1101 exhibit inter sampling ripples ancr· the selling time is
reached, we require U(z) to be of the following fom1. , · 1'
1
U(z) = bo+b1z- + b1z·2+b(z· 3+z-1+z" 5+........ .)
I3ccausc the plant transfer function Gp(s) involves an integrator, b must be zero.
Consequently, we have.
U(z) = bo+b,z- 1+ b2/? .
Also, from figure (b), U(z) can be given by

U(z) = C(z) = C(z) R(z) = F(z) R(z) -:· .

G(z) R(z) G(z) G(z)

I - 0.3679z- l
= f-(z) - - - - - - r - - - = -
0.3679(1 + 0.718lz- 1)z- l

For U(z) lo be a series in z- 1 with three tcm1s, F(z) must be of the following form .
F(z) =(1+0.718lz· 1)2· 1F1(z) ......... (4.37) •
1
where r, 1 (z) is a first-<kg rcc polynomial in z· Then U(z) can be wriltcn as follows .
•

1
U(z) =2.718l(l-0.3679z" ) F1(z) .. ........ (4.38)
From Eqs (4.34) and (4.35), we have
I-F(z)-;::l-a 1z-1 -a2z .2-a3 z.J = (l -z·l)N(
. • • z)

1-19

_j
 ........ :.(4 .39)

......... (4.40)
Nole· that from 12q (4 36) we r . I
. • c4lurc N(I) = ¼ . Therefore, by substituting-( 1=.I i~ 110 • ~q
(4.39) \\ e obtain. •
2a1+a2=2.75
.......... (4.41)
Also, Eq (4.3 7) can be rcwrittc~ as
I

r:(z) = a1z•1_ a2z•2-a.1z·J === (l+0.7 181 z1)z"1F1(z)

-+ ·I .,
:11 • a2z +a3z - -==( l+0.718 Jz·')r (z)
1
Division of the lcfi-hand side of thi~ last equation by I+0.7 181 2 - 1 yields the quo~icnl
1

[ti,+(a2-0.718la1)i and tl{c remainder [a .q.718l(a~-0.718!a )]z" By equating th~ quotient.

]

3
1
2
.

with F1 fz) and the rema inder with zero, we obtain.

F1(z) =a1+(a2-0.7 I 81 a1)z" 1
and
a3-0.718 l(a2-0.718 Ia 1)=0 ......... (4.42)
Solving Eq (4.40), (4.41) and (4.42) for a1, a2 and a3 gives
a1 =1.26184, a2 =0.22633, a3 =-0.48816
Thus, F(z) is determined as
f(z) = 1.261 S4z· 1 + 0.22633z· 2 - 0.488 I Gz·3
and
Fi(z) = 1.26184 - 0.67979z' 1
l
Eq (4.39) gives
N(z) =l - 0.261 S4z· I - OAS 8 I 7 z -2

The digital controller pulse transfer function Go(z) is then detem1 ined as follows:
F(z)

150
 0 1
()~ 'J I ~ •718 ~- ~ )1." 1(1.2ri1iM - 0.<,7'J801.- I) --
•• (179(l+0.71SJz- 1)%- l • ' • -2
(l-z-l)(l-0.3G79z-l)z-1 (1-z-1)(!-0.26 184z-1-o.48.817z )

== 3.429Si!_- O.S]S?z-l )(I- 0.3679z-l)

. . (1-0.84187. l)(l+0.5799z-l)
\\ itll the <l1g1tal controller thu d . . • · ut
. . s esigncd, the system output in response to a u111t-Slcp 111 ~
r(t) =I 1s obtained as follows.
('(z) =f-(,.) R(z)

''

Hence,
c(O) =O
c(I) = 1.2618
c(2) = 1.4882
I
c(k) =I, k =3,4,5, ........... .
The unil-:<;lcp response s~quence has a maximum overshoot of approximately 50%. The
settling time is 3 sec.
Notice tlial from Eq (4.38) we have
U(z) =2.7181(1-0.3679 2· 1)(1.26184-0.67979 z·1)
= 3.4298-3. t 096 2· 1+.6798 2· 2
Thus, til e control signal u(k) becomes zero for k 2:: 3. Conscquc11tly, there arc •no
intersampling ripples in the response. Notice that the assumption of N=3, lh~t is, tht:
• !'

assumption of F(z) as given by equation 1, is satisfactory.

Next, let us investigate the response of this system lo a unit-rump input.
C(z) =F(z) R(z) , . ,

z-1
2
1
=(1.26184z· +0.22633z· -0.48816z· 3) _
(] - z 1)2
= 1.26 IS z·2+ 2. 7500 z•J·l-3.7500 z~1 +.. .......... .
In the unit-ramp response, the control signal U(z) is obtained as follows.

. /.
t5 I
 l (I ) F(;,.) .- 1
7
( i(I.) I _ ,. I I _ ,. I

(J . 42n - J . I OCJG✓.·l _ 0 -2 i'.- 1

.679Sz )- -
-f
1-z
= 3.•L9 z•t 0.3_0_ z·· z· . --l + ·5+
z z ......... .
·1h . sib11:d u(k) bc.:c.:0111~.s c.:011s1a111 (I> • 1) fork c: J. I lcm:c.:, the syslt:111 output will not t:xltil> it
int amp ling ripples .

.t 10 Problems :
1. Consider the system desc.:ribcd by
y(k) -0.6 y(k-1) -0.S I y (k-2) + 0.67 y(k-3)-0.12 y(k-4) =x{k)
\\ here x (k) is the input and y(k) is the output of the system. Dc!crminc the stability
of the system. .._,/
Ans. Critical stable.
2. ~onsider the characteristic cqu:llion of the second-order system.
'
f(z) =a2Z-+a1z ao=O ; a2 > O
Detem1ine the stability of the system.
Ans. Unstable.
J. Consider the sampled. data system as shown in figure below. Determine its
characteristic equation 111 the z-domain and ascertain its stability via the bilinear
transfonnation.

c(I)
r( I) c(I) c•( t) 5
s(s + l)(s + 2)
&r{T=lsec)

Ans. Stable

• . 152 .
 __ _____·-
...._

4. Dr:1w root I 1·. . • r •c below r • I ht:

ocus < ,,,grams lfl the z plane for the sysIc111 ,i;/Jown 111 I H1" 101

ltillowing three sampling periods. T== I sec, T- 2 sec, and T== 4 sec.

__/ k c(t)
,----,►
s(s + I)

5. Draw the Bode diagram in the w-plane for the transfer function of the pl?nt
1 .
G(s) === s(s+
' 2) which is preceded by a zero-order hold.

153
 CHAPTER- V

I' State . . Space Analysis

5. vnlroduction
Conventional methods su 11 • J fi
. . . c as the root-locus and frequency-response methods arc usefu or
dcalmg with smgle-input- s • . · I
mgle -output systems. These methods are simple and require on Y
a reasonable number 0 f . . . ..
. computations, but they arc applicable only to linear t1mc-rnvan,mt
systems hav111g a sin •I • . - · J•
• g c mput an~ smglc output. They arc based on input-output rcUitions 11 P
of the systems that • I1 . • · J
' is, t c transfer function or the pulse transfer function. Convcnt,ona
mcthods do 1101 ,11111 1 1O . ·• , I
' Y non 1incar systems except in simple cases . Also. the convcntion,t
methods do 1101 a I • • • ·1
PP Y lo t11c design of optimal and adaptive control systems, wh1c 1 arc
,·
.mostly time varying and/ or nonlinear.

A mndi:rn
, = ''O"' • I ·'«y_...!~t!.!L!!!!.!Y
• I
-1ill'..C ·
many 111p111s and man 0111 rnrs. For 111u I11• 1c lllJHt
• 1s am I

:!_1lllliple outputs, \VC need state space methods

~ · Concept of the State-Space l\lethod --- .

-
The stalc-_~Racc method is based on the description of system equations in tcn11~ of_!!)irst-
- ----- ------ ___ . ·--- - --- -
' - -- --
qr~r difference equation. Which may be combined into a first-order vector-matrix difference
- - - - - - - - -- - - ---- - - - - - - '
equation. The use of the vector-matrix -l)Otatjo11 greatly simplifi_cs_Jh~_!l!~thcn~a~~ _
- - - ~----- - - - ---- . ' . . . .
representation of the cqua_tions of the ~ystcms. :-·,

The use of the state-space concept enables us to design control systems with respect to giv-e'b
performance imJcxcs. In the state-space the design c;in be carried out for a class of input ~.
instead of a specific input function such as the impulse function, step function, or sinusoidal
li.1tll:tio11 . Thi.: slate-space methods also enables us lo include i11itial co11clitio11s in the desi gn.
which is riot possible in the conventional methods.

Definitions
/ .,
St~he stat~- of ad 'namic syste~ is th smallest set of variables (called state variables)
such that the knowlc<l~_c of th cse vanabl~~lJ=lo, comp.!£tely determines the behavior of the
~ ---- - - zz..: : : : : :

154
system for nny time t ~ 10 It is I0 l . only limited to
. • • )C noted that the concept or n stnlc l:i 110 1
phym~nl systems, but ulso 11111)1•. 11 . . . social systems
ICI\) c to h10\og1cnl systems, eco1101111c sy st em.
nnd oth1.:rs. •

Stat(' J nria/,/cs • The st t •b . k• , up the

. -- • a c vana lcs of a dynamic system arc the variables ma mg
smallest set o~ variablss tl~at c.letcnnine the stat~ of the dynamic system. If al lc~st n variables
xi, xi, ..... , Xn arc needed to describe completely 1hc behavior of the dynamic sy st cm (so th at
~1~~J!ieJ npl~l i~ given_lur .I ~ lo and the initial state at t=to is speci lied·, t.!!c li1tt1re st atc o[ pi~
- -
sySlem_is_~011}Plct_cly determined), then such n va~iables arc a sci of swtc variabl.9s.

Sime V£'ctor: Jt _n state variables arc needed to descri_hc complctflY _~l~c l?clrnvior of a gi~cn
SY st em, t.hcn these Jl stat~ variables can be considered the n components of a vcclo_r X. S_uch
a.vector i_s lh~s-a vect~r_th~t dctcnni~cs uniquely the system stale x(t) for any ti_~e t ~ to,
once the state at t=to is given and the input u(t) fort~ to is specified.

I.
S~a.te Space: lhe n-dimensional space whose coordinate axes consist of the X1 axi.s, X2 a~is,
...... , ~n axis is called a state space. Any state can be represented by a po'int in the state space.

State- Space Equations : Ir~ state-space ~nalysis we need to _concern with _three' types of
variables that are inv?lved in the model_ing of dynamic syst~ms : il}Put variables, o_utpu_!
I

variable and the state variables.

For time-varying (linear or n~!1-~ne~) discrete -time systems, the state equation may ~e
written as.

~
--
(k +I)= qx(k), u(k). k]

and the output equation as

- ·--
.... ... ...(5.1)

'
✓y(k) = g( x(k1_u(kl;l .......... (5,.2)

For Ji near time-varying discrete-times ~1_12s, the state equation and output cquati~n may be
- ~-- - -·- . - .
simplified to .
~l~.}
x(k + 1) = G (kJ + H (k) u(k) ............ (5.3)
r - - - - -- - __y
l
y(k) == C(k)x(k) + D(k)u(k)
............ (5.4)

155
,·

\\'l 1l 'I\' .

\(\, ) II \ ·1·rt111
( ·. l:1k , ·, ·,·1,11 I
\'( \-) II\ \ ' l'L' l111
('. 111tp11( \'1T l111 J~
11 (k) : r - vector
(lllput ,·cclor)~
(i(k) :: 11 x 11111atrix
(state matrix)
H( kl ••• 11 x r 111:tlrix
(input 111:ilrix ) J
l '(k) =.: 111 x ') _111ati-ix
(ll ~1tp111111;1l rix ;
IJ(k) =: Ill X r lllalrix
(direct lr:i11smi ssio11 mat rix)
h11 Ilic I i1 11c - i 11,·;lfi ;uit s . . . • • •. 111
• Y ~IL!n til l · ~ l;ik l·, p1 ;11 i,111 ;111d 11 111p111 L·q11 a l1P11 L·; 111 li t: s,nqildil d
:qk+ .I J -=-G:xlk) + l lt1t l,; ) ........... .( 5.5 )

) ( k l == C\ ( k ) + Du { k ) .. ... .. ... ... ( 5. 6)

Thi: Fi g. 5.1 shnws 111 "•' \)loci". t11•a~r;1111 rep resentation or the t ,·1scrc tc-11111
• e con t 1·ol -<·ystc111
~

,, ___
Ill
..-

Fig 5.1 ( ~lock diufir~1111 of o lmc.:Jr lirnc inv,irinnl di scrclr.- lirnc co11lrol system
rnprcscnlcc! in the sl<1 c sp:ic:~ )

5.3
_____-:-- __
SI atc-!:ipacc HcprC!iClllatio11 or Discrctl'-Ti111e Sysll'lll
__,...-,---- - -.--,,-~:----
Lei us .:nnsidcr tile d1scrc\ c- l1111c sys ti.: 111 described by

y (k) 1 a,y(k-1) + ;1~y (k-~) !· .... ;1" y (k-11 )

h.. I I( k) l li I \I ( k - I ) I .. ..... I b,. II l k - 11 ) .......... ( ) .7 )

\Vh ,; rl' ,i(k) i s 1hc 111p111 and ylk) is tin~ oulpul of Ilic s ys k1n :11 tile ktl1 samplin g insl:llll. Noll.:

tli :it : tlllll' 1,1· tin; cm:lfo..:i c11t s :1, (i I .~. ... .. 111 and b,i U 0. 1.2...... , 11) m:1y he / .LTO . Th e l:q (5. 7:
111
;p • h· "' ' ittcll i n tliL· 1i 1r 111 11 1' 1111l :, t: 11:11 1s l t ·1 l 11 11t· t 11 111 :i s

1:i(j
 Y(z) h + h z- l
0 I • +.. .. .. ... +b i'.- 11
= n
(5.8}
l ll1.) I + :il1. ··l .1 ·t . • _
····••• •• •••
. .... .. ,111 :: 11

Y(z) h0z11 + h zn - I
• I • + ... .. .+h
or. -U-(z-.) - n 1 n .. .. ....... .(5 .9)
z + .
alzn- + +"
.. . .. "ll

As there arc many wavs lo re·11•. . . • , svstcn 1

•• ' ii.c state-space representations for the d1scrclc 11111 c • •
d~~crihcd hy Eqs (5 .7). (5 .8) (.c c . . . . .. • ·
• or .}. )). \\ c will present the rollow111g n.:prcsc111, 1t1011 s,
li) Controllahlc canonical form
(ii) Obscrv~ canonical foml
(iii} Diagonal canonical fonn
(i\') Jordan canonical form
• -<!
.• . 1J1YC'. u-(~r/v..,..,.,'>,i) . ·.
(i) Co11trollablc ca110111ca/ form :A contr~ system is controllab le if every slate van c1blcs
c~n be controlled in a finite time period by some _unconstn'.incd con~ol signal.
'i'hc Eqs (5.8) or (5.9) may be put in conlrollnblc canonical form ns shown below :

0 xI (k) 0
x1(k+I) 0 0
x (k) ()
:--: (k+l) () () ........... 0 2
2
= + u(k)
xn_ 1(k+ l ) 0 0 0 ........... 0
X (k)
n-1
x11 (k + l) -a• • 11 -a 11 ,- -a -al x 11 (k) I
1 n-2
'
.. ..... (5.10)

XI (k)
x-, (k)
y(k) = [ l_>n - a_,_lb~/ b
0
_
1 - a 11 _I bo:· .. :b 1 - a 1bo ] -; + b0 u(k) ............. (5. 11)

x 11 (k)

Eqs (5.10) and (5 . 11) arc the state equation and output equation respectively.

157
 .. ......... (5 . l 2)

Sol11(i1111. Tit'-' 1•ivl'II~\·st1.:111 c ·111 l1. 1·1·

• • ' t: 111111 1 1cd to

Y(z) b0' +b 1z- 1+b 7z-2+ ......+ b .Z-n

U(z) = I - , 11 .......... (5.13)
l+a 1z ,+a 22· ~+.. ., +"nl) z-n
, I •

nr.

.. .... (5 .15)
. I

Let us define,

......... (5. 1(,}

Then, lhc Eq (5 .15) becomes,

l
Y(z) =b0 U(z)+ Y1(z) .......... .(5 . 17)

Eq (5 .16) can be rewritten as,

Y1(z)
- -.- --_- _.,.---
1- -. -.- .- 2
(hi - a b(/~ + (h 2 - .1 2 b0 )z +... .. +(h 11
. ·-:·=-n
-,1 11 h )1.
1 0

........... (5.1 S)

So.

.. ... .. ... (5 . 1<)) .

1SfJ
Let u. no,, define the st t .
a c variables as follows :

X1{i'.) ,.,, /"Q (i'.)

X ~(i'.) =:: 7--~H IQ (7.)

Xn-1(2) = z· 2Q (z)
Xn(z) == z· 1Q (z)
'lh.·11 \\'~' l'all writ~

z \i(Z) == Xi(Z)
Z X2(7.) = X3(Z)

z Xn.1(z) = X1i(z)

In tcnns of di ffcrencc equation~. the preceding n-1 equations become

x 1(k:t- 1) = x~(k)
, .. .
X2(k I· I) =~ X_1( k) ........ (5.22)

X11-1(k+ I) =xn(k)

Now, by substituting Eq (5 .21) into th_t: Eq (5.~}. we_can write

'/. Xn(i'.) = - alXu(z) -• aJX,, .1(1) - ..... - a,)~i(;,.} +lJ(z)

or, ... .. ... (5 .23)

Also. Eq (5.20) can be wriLLcn as, - - -·

Y 1(z) = (h 1-a1bo) X,,(z)+· (b2 -a~b11} X,,., (z)!·.... + (b11-anbo) X1 (,.)

or,
y 1(k) = (b 11 -a 11 h0 ) ~,(k) ·'-- (bn .1-a11.rho) x~(k) +... .... ·l(h1-a1hri) x 11 (k) ... .... .(5 .24)

from (5 .17) and (5.24), we get

y(k) = bou(k) +(b 11 -anbo) X1(k)+Cbn-1-an,1bo)X2(k)+ .. ...... +(b1-a 1b0 ) x11 (k) ....... (5.25)
Combining Eqs (5}2) ~111d (5.23), we get the state equation as bclc)\\:
'

159
 ; l 1' I

:-. Ilk I l ) 0
()
X ( k I I) (I
0 xI (k) ()
()
() x (k) ()
:::: 2
u(k) .. 5.2(, )
:-- n - l t k l) 0 () ()
0 X (k)
:-. l k t I) I
11 - I
II - a -a
11 ll I - a - ·1 x (k)
. - 11 - 2 ·1 ll
and from Eq (5.25) we oct 11
' " ic output equation as below:

.. ... .. (5.27)

Exar·/lconsidcr lhc d. ·
~·C iscrctc-t,mc system defined by
Y{z) _ 0.36? (z+0.717) ;
U(z) - (z-1) (z-0.368) )

Solution . The gi\'cn system cnn be modified as

Y( i.) _ OJ(>8;,:1+ 0.26 ~- 2

l .(z) , I l.3(18z-l+U.3(1Sz· 2

,Y(z) U(z) ,
or, - 7- = - - - - , - - - - - = - = Q(z)
OJGSz ~+0 .2Ci4z-- l-1.368z- 1 +0.368z- 2
or. Q (z) = .~z) + I .36sz· 1giz) - OJ<i8:(~ Q (z) ......... (5.28)
• '.,~ ,. - v1"/
and Y •z) = 0.J6sz· Q (z) + 0.264z.~Q (z) ....... .(5.29)
- - ·~ . --~

f-.1 (Z) -:.' z.. ~' ( ?)

z;' .Q.(z:) ::- z. X,Cz..)
. ,

So.
2·
~
1
Q(z) = X2(z)

zX 1(z)=X2(z) --
~(?.) -::: z_ X2,C~)

or, x 1(k ·t l) ~ ~dk) ··· ·· ·····(5.30)

-------·
Eq (5.2S) can be rc\Hittcn as.
r
z \~(z) == U(z) + 1.368 X~(z) -0.36S X 1(z)
- )

160
 "\
x,lk t l) r 11(k) + l (C\ ........... (SJ I J
•• h) X~(k) • OJ(,8 Xt (k)

Nlrn·. lh,m Eq (.- .29), \\·c get

-
Ylk) (U(,8 x,(k) ·10.2<,4 X1(k) .......... (5.32)

I knee, the slate-space mod 1 ftl .

e O le system IS,
X (k+l)l [
[•I ()
:-;2(k + I) == -O .J(18 t.)(18
I l [X (k)l [()]
x~(I·() + I u(k)
.......... (5.33)

and y(k) = [0.264 0.368) [xi(k)l ...... .... (5.34)

x2 (k)

(ii) {)h.\·(•ri•u/J/e C{IIJ011icul form

~
: Consider the pulse trnnskr fi111ctio11 systt.:111 dr.;lincd I'Y

\' 1,.) = ho -~h1z~ ; ' •••••:l''.•i'.'~ 11

_ ( "'"'/-6{ P;.~"'~'.Y,J. (s 35 J
.U(z) l+alz-1+ ... ... +a~-n

TI1c state-space representation of the discrete - time system given by equation may be put in
the followi ng form:

x (k + I) 0 0 0 0 - an ,,
1 x (k) b -a · b
x2 (k+ l) 0 0 0 - a 1 n n 0
n- 1 b
x2(k) n- l - a 11 - 1.b·, 0
= + u(k)

x (k+l)
X
n-1 (k) b2 - a2-bO
n-1 0 0 1 0 -a2
x 11 (k) bl-albO
x 11 (k+l) 0 0 0 - al

....... (5.36) •

' , x (k)
1
.,. x (k)
'' ,.
2
y(k) = [o 0 0 ·I] + b0u(k) ....... (5.37)
X
11 -- I
(k)
x11 (k)

161
,·.

lht: slal '•Spn 'l' r 'p1cst 111:11io11 g1v ·11 hy Eqs (5J(,) :ind (5J7 )
·;int111i ·a I forlll .

1, 1-onf: l{l'\\'l'iting the 1n1\s- 11··,111s r1cr ,1111

r .
c11 011 as follows :
Y(1) l- a 1,.-ly(;,) .1 a ,.-2Y( · ) . . . . - llv(· )
2 '/, -f ....... 1,111'/. I '/.

= huUV)+hlz- lU(z)+b 2z-

· 2U(·Z-f
) ·........ .··l . - 11 'I)
iJ 11 Z \

or,

Let us define the state \'ariahles as follows:

I

1
X 11 (z) = z- [b 1U(1.)-a Y(1.) + X _ (1.)]
1 11 1
1
X 11 _ 1(z) = z- [b 2U(z)-a Y(z)+\ _ (z)]
2 1 2
..... . .(5 .39)

\':?. ( 7.) - z- I [ b 11. (I.) - a n -1 y ( z) + XI (z)]

1
X 1(z) = ,- [bnU(z)-a~]

I lcncc rrom Eq (5.38), \\'C gel

....... (5.40)
Y(z) = boU(z)-'- Xn {j
' 5
from (5.39),
z \ 1(1.) = b 11 U(1.) - a,.Y(z)

or, z Xi(z) = bn U(z) - an [boU(z)+ Xn(z))

= bn U(z)- an bo U(z) - anXn (7.)
== (b 11 -anbn) U(z) - aa Xn(Z)

:. :,; i(k+ I)=_ ;1 11 :-;,,(k) ~(hn-anbo) u(k)

1G2
~i1n1\ail y .
l . J I
!\•\\.-1 \ ) r
,, r , J',.I Y1,,; ~ rr T
d Jt1 I ·,.J) 't .

. \ . 1 Ol . . . . . 11..111 ·kr fu11t:1iu 11 i11

l•.'.\. l\ lllll ~ •• )\.Ill\ lhe Slilh:-space lCJHcs1.:11lalion ol',lllc fo llow1 11g pulse ' s /
1
ohscr,Jablc canonical fom\ .. • ,

. I
l
Y(z.) _ 0.36Sz- 1' + 0.264.,.-i
U(z) - l - l J(18,.-l + OJ(i81.-2'

Solution . The given pulse transfer functio n can be rcwrittc1; as,

:Y(z)- l.~68·,.- 1Y(zl +0.3(,8z- 2 Y(z)

2
= 0.368z- 1U(z) + 0.264z- U(z)

2
or, \'(z) = l.3(>8z- 1Y(z)- 0.368z- Y(z)
2
+ OJ68z- \U(z) + 0.2641,- U(z)

or, Y(z) = 7,- 1lL?_CiSY_(z.) + o~z)1

Y(z) = z,- 1(1 .368 Y(z) + 0.3_68 U(z)

or,
~

Let us define ,
X (z) ~ 7.- 1[1.36S Y(z) + 0.~6S U(z) + Xl (,.)]
2
i XI <.z) == z-ll 0.264 U(z) - 0.368 Y(i'.)1

: . Y(z) == X 2 (?.)

163
llcncc. the stale-space re " . . . ,jvcn ;is (o ll oW!i:
prcsentat1011 of the gi\'cn pulse tr:111sfer f1111ctior1 is g

[ ~l(k_+l)l==l() - OJ(18l[X1{k)] [0.2(14]

x2 (~ + I) I 1.368 x2 (k) + 0.3.68 u(k)

:ind . ytk) :-: [o 1J[x1(k)l

X,(k}

(iii) Diagonal ca 110111•cal fo~·m: Consider the pulse transfer function

· f tl ,e sy stem is ~0 i,·cn °
h~· ( f /bt,,t-,' '<.,( V>-tJ c,{(A. -~7-f d--r'-1 ( ""1 /

Y(z) b z" + b Izn-1

-- - 0 , +... .. +bn
....... .. .(5.4 l)
U(z) zn + a zn-1
I +....... +an

If tllc poles of the pulse transfer function arc all distinct, then tile state-space representation
may be put in diagonal canonical fonn as follows :
x1(k+I) Pr 0 () xI (k)
x.,(k1l) (} p., () :-; 2 (k)
:::
+ u(k) ... ... .... (5.42)

x 11 (k+I) () ()
Pn X11(k)
and

x (k)
1
x.,(k)
y( k) = [C' I C1 .. ... C11 ] - .. .... .. ..... (5 . D)

X11 (k)

Proof: The given pulse transfer function can be modified as follow s:

\'(;,)
I z n- 1+.... .. .·I 1111
(ii);,
11 I 1
1
- ). = - -n- . 11-l ,- • - ,-:__ _
l j{ i'.. z+a 1,. •...... · 1
,11

164
 111 • all the poles f I Y(z) Y(z) • • be cx11:indcd
0 tic pulse transfer function __:_ are distinct. -U(· ) c,in • •
· U(z) i
in\n th l'l)\\ O\\ 111g
• form:

Y(7.) c ........ (5.45)

- ==b + 1
U(z) 0 z - Pt

where,

......... (5.46)
or,

Let us define the state variables ::is follows:

1
X (z) = - - U(z)
1 z - P1
l - ........... (5-.47)
X, (z) = --U(z)
- z- P2

1
X 11 (z) = --U(z)
Z-P n

or,
z X (z)
1
= p 1X 1(z) + U(z)
........... (5.4S)
z X (z) = p X/z) + U(z)
2 2
' .
, '

z X11 (z) = P;1X1~(z)-:- U(z)

and
.. ....... (5.49)

165

/ '
 - --------·-----

:--,tk I \) • P1~1(k) I lttk)

:-:Jk+l)= P2xJk)+ u(k) ........ .(S.50)

and
........ (5.51)
y(k)-= Ci Xi (k)+ C2 X2 (k)+ ...... + Cn Xn (k) + bo u(k)
Hence, th c st ate-space representation can be given as Eqs (5.42) and (S.4J).

• r function
l~
t..xamp Ie 4• Ol1ta111
•
a state-space representation of the fo ll owrng
• ' 1rnlsc-trn11s1cr
•
system in the diagonal canonical form

Y(z) l+Gz1 +82- 2

U(z) = l + 4z-l + 3z·2

Solution. Th; given pulse transfer function can be rewrlucn as

I

?
'
Y(z) z- +6z+8
--= ?
U(z) z- + 41. + 3
') ~
_ 7. + )
=l+ 2 '
i + 4z+ 3
2z+5
- 1+ •
- (z + l)(z+3)

Y(z) cl c2
or, - - = \ + - + - -..
U(z) z+l z+.,

Where,
_I 2z+5 (z+ l)l
cl -l(z+l)(z+3) Jz=-1

166
 - <H-5
·:\ ·I· I 2
Y(z) - 3/2 1/ ,
or, -1+ - + -
U(z) Z+l ~

or, Y(z) == U(7.) + '}}_-3__ (· ) I/ 2

• ,+ ~ U(z) )
7
7. + I
Let us de Ii nc the state v·,ri.·il l --
.- -- -- - ' ' 1 cs as follows:

~ x,(z) = ~V(z) - ·

X?(z) == _!_U(z) 2 J -i
- z+ 3 ,- ...._ __..-,
or, z X.1(z) == U(z) _ x ,(z) .,
1
z X2(z) == U(: ) • 3 X (z) A/, .
2 ,.
and 3 1 .
Y(z) == U(z) + 2Xl (z) + 2X2 (z)
I l
I

Hence,
x 1(k+l) == u(k)-x (k)
1
x2 (k + 1) == u(k) - 3~ (k)
2
-
3 ·, l ·
and y(k) = u(k) + x (k) + ~,-th1'
1- 1 ' 1- -

Hence, the state -sp~cc representation is given by

x 1(k + l)] == [-1 0 I xl(k)]+[l] u(k)

[ x (k + l) 0 - 3 x (k) I
2 2

and y(k) = [ ¾ m:~::i] + u(k)

' 167
 ' . --

x111 (k + 1) 0 '
::: 0 0 0 0
-- .- ------
--- --- --- ----
xm+l (k + 1) 0 0 ·o 0
x11 (k + \)
0 0 0 0 0 Pn
0
0
• I
I
1;
+ u(k) ..... ..... (5 .53)

.... ...... l5,5-1}

Pi·oof : Since the system pulse transfer function can be written in the fom1

168
 ·I· Ill l· l C
z- )
- 1 +_ 111 •1- '>
.._ 1- e ll
1\1+ 1 z- J) . ....... -1•: - -- ... ... ....(5 .55 )
111+2 z- p11

Then, we obtain

Y(z) = b U(z) +
0 (z
cl
rn U(z) +
c2 cm (
c
m+l U(z)
- Pl) . (z - P/11- I U( +.......+
z) z - p 1U z) + z - Pm+ 1

Cm+2
+ -=--=-U(z)+ Cn
z - p -, •.......+--=- U(z) • ..... .(5.56)
m+_ z-pn
; I ;-'

Let us define the first · ·

• J m state vanables Xi(z), Xi(z), ......... .Xm(z) by the equat10ns
I

Xt (z) = t U(z)
(z -p l)m

1
X2 (z) = m-1 U(z)
(z- P1)

l .
Xm(Z);;:: ( ) U(z)
z-p
1.
and the remaining (n-m) state variables Xm+i(z), Xm+2(z), ....... , Xn(Z) by the equation

l
X +'> (z) = - - - U(z) · .......... (5.57) •
Ill - z-p . +?
111 -

1
Xn (z) = - U(z)
z-p
n

169
 xi (z) 1
X2(i'.)-z-pl

X2 {z) l
X3(z) = z- Pt

........ (5.58)

Hence • by taking
. •
mvcrsc .z transform, we obtarn
•

xl (k +I)= P1X1 (k) + Xz(k)

x2(k + 1) = P1X2 (k) + X3(k) ........ (5.59)

xm-1 (k +I)= Pt.xm- 1(k) + Xm(k)

Xm(k + 1) = p 1xm(k) + u(k)

xm+l (k +I)= Pm+l\n + I (k) + u(k)

x 0 (k+l) = p11 x 11 (k)+u(k)

and the output equation is. .
Y(z) = cl XI (z) + CzXz(Z)+ .......+cmXm(Z) + cm+lxm+l (z)
........ (5.60) ',
+ cm+ X m+ (z);t-......+c 0 X n (z) + bo U(z)
2 2

Uy taking the invcrsi..: z transform, we obtain

y(k) == c x (k) + c x (k)+ .......+cmxm(k) + cm+lxm+l (k)
1 1 2 2
+c 2x 2<k)+ ......+CnXn(k)+bou(k)
m+ m+

Hence, from Eqs (5.59) and (5 .60), we obtain the state-space representation given by Eqs

(5.53) and (5.54).

170
 Solvl1111h 1)1 SCl'P(('
· 'l'I
inc S111t"-
5 ./ J \' I 1 " pare l•.,11111111011
• • t~/' t/11• J /11 ,· . . .
• o Ill 011 ' ', Ff111(lt/OJI .' /II
.I I . • c '" I l1111•-fn1'11r/a111 J)/,\'l'l'l'll'-'J'IJ/tt' Slllll'·,\JJIU '- • ,
gene, n • t 1c <hscrctc-timc c • . . durc which is
. . quations arc solved by means of a recursion procc .
quite simple anc.l co, .
1Vcnicnt f, ct· •
. or igital computation.
Lct us cons1c.lcr the st .
, .
a1c cquat10
' 11 anc.I output equation as follows:
x(k+l) = Gx(k) +llu(k) .......... (5.61) .
. y(k) =Cx(k) +Du(k) ......... (5.62)

The Eq (5.61) can bes 0 l

ved by the method of recursion as follows:
x(l )=Gx(O)+Hu(O)
x(2)=:Gx( 1)+Hu( l) .. :

2
= G x(O) +GHu(O)+Hu(l)
x( 3) = Gx(Z)+Hu(2)=G3x(O)+Gi Hu(O)+G.Hu(l)+Hu(2)

By repeating this procedure, we get

I
I k k-1 k .
x(k) = G x(O) + .L G '-J-lHuU) k=l,2,3 ...... (5.63)
J=O

Hence, it is clear that Eq (~.63) consists of two parts, one rcprcscnti1~g the contribution of i_hc
initial state x(O) and the other the contribution of the input uU), where j=O, 1,2, ..... ,k-l. Tl:e •
output y(k) is given by

k k-1 k •
y(k) = CG ·x(0)+Cj~O G ~J-lHuU)+Du(k) ........ (5.64)"

& .S III te .Tra11sitl011 Ma;rl< : It is possible to wri tc the so lu tio n o( the ho mo gen co us stat c
Y euat1on . '
. .
.••
x(k+ 1) = Gx(k) .......... (5.65)
as
x(k) =G1.P(k), .......... (5.66)
Where \l 1(k) is a unique n x n matrix satisfying the condition
1.1' (k+ 1) = G1.l'(~). . I.Jl(O) = I
........ (5 .67) .. ..
Ckarl). 'l'(k) cnn be ,,,r·,,, ,.
l.:ll as
'l'(k) == k .. .. ...... (5.68)

Frnm Eq (5.66) it is 1 . . -~ ll'tlion of the

. .. ' • c C,lt lliat the solution of Eq (5.65) is sirnply a tran s on '
1111ttal state. Thcrcfo . . . . M, •ir
re, l1~c unique matrix \I'(k) is cnllcd the Stale Trrt 11511' 0" a11 • •

As we know that the sol .

ution of the state equation is,

k k-1 k .
•X(k) = G x(O)+I G·-J- 1 . ......... (5.69)
j==O HuU)

Hence, in tcm1s of .tt1c sale

t trans1t1on
. . matrix \I'(k), Eq (5.69) can be written
• as

k-l
x(k) = tl1(k)x(O) + .L 'l'(k - j- I)I IuU) ..... .... (5 .70)
J==O

k-1
= \V(k)x(O) + .I \JIU)Hu(k - j - 1) ...... .. .(5.71)
J=O

Again tile output equation

k - k-1 k .
y(k)=CG x(O)+C).: G -J-lHuU)+Du(k) .......... (5 . 72)
J=O

In tenns of state transition matrix, Eq (5.72) can be written as

k-1
y(k) = Cl'(k)x(O) + C .I 'l'(k - j- I )l luU) + Du(k) .......... (5.73)
J=O .
k-1
= C4'(k)x(0) + C .I 'FU)Hu(k - j-1) + Du(k)
J=O

5.4.3 z - Transform Approach to tile Solutiou of Discrete -Time Slate Equation :-

Let us consider the discrete-time system describe by the following equation,
x(k+~ )+Hu(k) ........(5.75)
Taking the z transfonn of both sides of Eq (5.75), we get

172
 z X(z)-zx(O)" GX(z) II U(7.) t
where ,.(z), .._ ::::
[ k)]
:x (

Th~11 nnd
U(z) == 2 [u(k)]

(_zl-G) X(z) " ZX(O) +HU(z)

Prcmult1py1ng bot! .
1 oftl.
l Sites I . .
"(~)
•" z ==(zl•G) ·1 . l!s nst equation by (zl-G)' 1, we ohtnm
• • ZX(O)+ Vl•Gr 1HU(z)
Takmg the inverse 2 l .. ... ... (5.7(,)
• ransfonn of both sides of Eq (5.76), we get
x(k):::2· 1[ - 1]
(zl. G) Z x(O) + z· I[ (zl. Gf I HU(z)] ...... .. (S. 7?)
Now, as we know the I . . .
so ution of discrete time equation x(k+ I) = Gx(k) + 1-Iu(k) 15
1
x(k)::: Gkx(O) + \: Gk-j-I .
j;'o HuU) ....... .. (5.78)
Comparing Eq (5.77) with (S )
. •78 , we get
Gk ,,z·l[(zl-Gfl,]
......... (5.79)
anti

k-1 k- ·-1 [i
.L
J=O
G J HuG) = 2-l (zI -G)-l HU(z)
......... (5.80)
Where k==l,2,3, ........

Exam~Obtain the state tra~sition matrix of the following discrete-time system:

__ x(k+ I )=Gx(k)+Hu(k)
y(k)=Cx(k)

where G = [_ 0° 16 _\].H =[:Jc= [ I o]

Then obtain the state x(k) and the output y(k) when the input u(k) = _l for k
=O, 1,2, .. .... Assume that the initial state is given by,

173
 Solution . As we kn
ow that the i;tnt 11. . .
• c a11s1t1on matrix 11' (k) is

'\ \}'(k) - Gk
, - == 2•1[(zl-G)"1z]
' 1 hcrcforc, we nrst ol t •
' l 1'1ll (zl-Q)"!

(zl-G)-l==[z .· -1]-l _
0.16 z+ I

-[ •z + 1
(z + 0.2)(2 + 0.8) (z + 0.2)(z + 0.8)
4 I 5 5
j - ·3 3 + - 3
7.+0.2 + z+ 0,8 z+0.2 z+0.8
==
_ P.8
3
+
Q.&
3· - ·3l + f j
z+0.2 z+0.8 z+ 0.2 z+0.8

The state transition matrt'x t 1, {k) 1s

• now o btame
• d as r. IIows.
10 • •
I
/ 4'(k) ==Gk==z· 1[(zI-G). 1zL

4 z 1 z
= z-1 J z + 0.2 . J z + o.s 3sz +z0.2 - 35 z +z0.8 l
[ - 0.8 z +-0.8 -z- - I - - z
+-
4 z
3 z + 0.2'- • 3 z + 0.8 3 z + 0.2 3 z + 0.8
~
~ • '
-
4( 5 l 5 •k ]
' - 3 -0.2) - 3(-0.~) k k k
3(- 0.2) - 3(- 0.8)
..... .... (5.81)
- 0.8 k • 0.8 I k 4 k
[ -3(-0.2) ' +3(-0.8)" -3 (- 0.2) +3(-0.8)

Eq (5.8 l) _gives the state transition matri~

z
Next, compute x(k). The transform of x(k) is given by
1 1
• ~ ) ] = ~ = (zI-G)" zx(0)+,._(zI-G)" HU(z)
• . =(zI-GY '[zx(O) +HU(z)J
.: - = ;...I ... .

l z
Since U(z)=
_1-z
-1 = -· -
y . ,. 22

We obtain zx(O) + HU(z) = [ :2 ]+[ 2

~ I] =
z-1
-z2+22
z --: 1
z- 1

174
 " .. . . . . ··~ -
...
Hence :'\(z)==(zl-Gr lr· .
zx(O) I IU(;,)J

- (~ + 2)z
== (z + O.~) -r~-.·1-0,-8)-(i'.- .-~G
- (-;,.- ~ l.34z)z
(z + 0.2) (z + 0~8-~(z-~ I)

r -17
22
\ 6- z 25
9 z •• 2
7.+0.2 +- - + lL
z+ 0.8 z-1
3.4 -17.6 7
I),
6 i'. z
-..L. 9 IS z
z+0.2 ' z + 0.8 + -z-1
Thus, the state\ cctor ""(k) 1·s .::,lven
o· t)y

- -17 (-0 ? { + -22 -0 8 { +~5

-
x(k)=Z"" 1[X(z)J = _6 _:) ~) 1'
J.4
[ -(-0.2) k 17.6 k 7
--(-0.8) +-
.
p.,;;;= 9
r-,!wlly, the output y(k) is obtained as follows :
18
~

---

5.4.4 Sol11tio11 of Linear Time-Va,yi11g Discrete-Time St11te Eq11atio11s : Consider the .

. .
following linear timc-\'arying discrete-time state equation and output equation:
x(k+ 1) =G(k) x(k) +H(k) u(k) ......... (5.82)
y(k) = C(k) x(k) + D(k) u(k) ...... :_.(5.83)

The solution of Eq (5.82) may be found easily by recursion, as follows:

x(h+ J)=G(h) x(h) + H(h)u(h)
x(h+2) = G(h+ l) x (h+l) +II(h+l) u(h+l)
= G(h+ I) G(h)x(h)+G(h+ I) H(h) u(h) +H(h+ l )u(h+ l)

175
Lei us define th c state
. trn11sitio11 ..·. . .
• Jb
• crn · dd111Ct y
Eq (5.82) as ll'(k . m,1t11x (funda111c.:ntal matnx) for tl1c syst
,h). It is a unic1uc .
nrntnx satisfying the comlitio11s
tJJ (k+l,h) ==G(k) tp (k I)
1
' ' 4'(h,h) =I •
where k=h,h+ I ' h+?-, .... It can be s • . • t rile
1 • ccn t11at the state transition matnx , 1 1s• given JY
• lfl(k I)
equation.

I.F(k,h)==G(k-1) G(k-2) G(Il ), .. :.:... (5.84) •

.... k >h
Using
.
1.11 (k , I1,
) thc solution ofEq (5 S2) b
.• ccomcs

x(k) :::lP(k h) k-1 ..._:. ... (5.85)

' x(h)+ j~h 4'(kj+I) HU) uU), k >h

Notice that the first tcm1 on the right-hand side of Eq (5.85) is the contribution or_the initial
state X(h) to the current state X(k) and that the second tcm1 is the contribution of the input
u(h), u(h+ l ), ........ u(k-1 ).
. .'

Eq (5.8,5) can be verified easily. Referring to Equation (5.84), we have

I
....... (5.86)
l.jl(k+ 1,h) =G(k)G(k-1 ) .... G(h)==G(k) '11 (k,h)

If we substitute Eq (5.86) into

k
x(k+ 1) =lf'(k+ 1,h) x(h) + .I lI'(k+ 1j+ 1) H(j)uU)
J=h

we optain
k-1
x(k+ I) =G(k) \p (k,h)x(h)+ .L 4' (k+ 1j+ I) HU) uU) +\f'(k+ l ,k+ 1) H(k) u(k)
J=h

k-1 ]
=G(k) \}'(k, h)x(h) + .I 4-'(k,j + l)HU)u(j) + H(k)u(k)
[ J=h

==G(k) x(k) +I-I(k) u(k)

Thus, we have shown that Eq (5.85) is the solution of Eq (5.82)

Once we get the solution x(k), the output equation, Eq (5.83), becomes as follows: :
k-1
y(k) ==C(k) l!I (k.h) x(h)+ j~h C(k) 'F(kj+l) HU)uU)+D(k) u(k). ' k >h

176
IfG(k)isnonsin,
.
1
ll nr for all k ,·:iluc , . i' I) , jsts tli cll tl 1c
im crsc oC\11 (k h) ti • co11s1clt.:rcd, so that tl!c: i11vcr~c of I (I, , 1 ex • • •
• • noted by \1, Ii . . .
\I r 1(k t,) - \l l ,k), IS given as t'oll ows·
' I - / (h,k) •

:::: [G(k-1) G(k-2) .... G(li)r1

== c·'·(h)G·' (h + I) .... G.'(k-1) ........ (5 .87)
, umtll:iry on \I' (I · 1 .
\, i). I\ su111111a r\• I . ) • • ti c fol lo w 111 ~~:
I. \l'(k,k)==I J 011 t l e state tra11 !:ii liu11 matnx \I' (k.ll gives J -

2. \IJ(k,h)==G(k-1 )G(k-2) ... G(h), k >h

3°. lf the inverse of\P (kl) .
1 , exists, then
\p·l
(k,h) =\!'(h,k)
4. If G(k) 1s • nonsinoula r II k •
~ < r_1or a , values considered, then
\I' (k ' i) - \l'(k J') \!' U,1),
. for uny iJ,k
• ! f G(k) is singular for any value ofk, then

\I'(k,i) = \f1(kJ) \I'(j,i); fork >j >i

S.S Pulse Transfer Function Matrix

A •Single-input-single-output discrete-time system !1iay be modeled by a p~dse transrer

- -
function . Extension of the pulse-tr;i.nsfer-function c.oncept to a multiple-input-multiple~oi.1tput
discrete-time syslem gives _us the . pulse-transfer-function matrix. In this section we ' .shall •
•

investigate the relationship •between state-space representation and representation b~i the
' :
pulse-transfer-function matrix.

Pulse Tramfer F1111ctio11 Jlfatrix : The state-space representation of an n th-order linear

time-invariant discrete-time system with r inputs and m outputs 'can be given by .•

• ()'(
_x(l<+l) =:=.9~k) t it uJ 1-_ ........ (5.88)
/
'<' y(k) ==Cx(k) +Du(k) • ........ (5.89)
,07 . . .
Where x(k) is an n-vector, u(k) 1s an r-vector, y(k) 1s an m-vcctor, G is an 11 x n .matrix, H is
I

an n x r matrix, c is an m x n matrix, and D is an m x r matrix. Taking the z tran'~ro 1711 or

Eqs (S.SS) and (5.S9) , we obtain
zX(z) __ z:x(O) =GX (z) +HU(z)
Y(z) =CX (z) +DU(z)

177
 - -------..:_..:....
·. > ·

Notin g !hat the deli 111·t·

10
11 or 11i 1Hrl . • ofzcr jni I j :i/
cond i ti on J • • lr ani&r f1111cti o11 alls for Ihe ;1 s~aI111p 1I 011 •
• • lClt; we also •
,1. Sl!l)Jt; th ·,t ti • '. I 1- 1·11
X(· ) _ ' ic 1111t1 nl stale x(O) is zero. th en we o > •1
• 7. - (zl- i)" ' HU(z)
and

\ (z):::::[C(zl-G)" ' H+D] U(z) - .

Where f-(z)U (z)

P( z) :::::C(zl-G)"'H+D
... .. .... .(5.90)

F(z) is called the pulse tr r . . r;

ansicr funct1011 matrix. It is an m x r matnx. The pulse trans er
function matrix f( z) clla t .
' rac cri zcs the input-output dynamics of the given discrcl e-tim e
system.

Since the inverse of ITl'Hrix (.., I G) • .

• ' ,. - can 1JC wnttcn as
(zl-G)"' = adj(zl-G) . -.. .
[zI-GI
I
the pulse transfer function matrix F(z) can be given by the equation
F z = C adj (zl - G)H
() lzI-GI :·D
.·.

C!ca rly. the poles of f-(z) me the :t.cros of lz! - Gl = O. This means th al the characlerist;c
cc;ualion of the discrete-time sy~tcm is gi ven by
• /z!- GI = 0 •
or.
Where lh t.: coeffic ients a, depend on the e:lcmcnts of G.

~~
5.6) Dis cr~ti:.aHon of Continuous -Time State-Space Equations
Jn di gi tal con trol of continuous-time plants, we need lo co11vc:11 r.:ontinuous-timc state-space
equa tions into di screte-time state-space equations, Such conversion can be done ·by
in lroduci ng fi ctitious sampler~ and fi ctitious holdi11g devices into continuous-tin'le systc; is .
The cirnr inlroJuccd by discrctiza!ion may be made negligible by using a suflicicntly sn~all
i ~amp ling Jcriod compared with the significant time constant of the system.

178
5• 6•1 [,\>e,,,·,,l ,rs
' "''' oJ ol11tio 11 0,r C'011 111
• I , II
5 1' 1
first re v1·cw the
. J t 11011s-Time Stute Equatirms: We
m:itnx cx1mncnti 11 1, c"' Th. . .· •
• c ll\,lll IX cxpoI1c111i11I is ddi11cd by

cAl ::: l +Al+ -A-1-

l ? ?
+ .....
l
+-A kt k + = ~ "~
21. k! ...... k~O

Because of lhc convergence of the infinite scncs I

Cl)
A k l k / k!, tl1c .,'-'c1·1·cs can be
K=O
diffc.rcntiated term by term lo give

i C
At
:::
2 J\3t2
A+A t + - +
J\ktk-1
+---+
dt ·2! ....... (k-1)! ....

A2t2 Ak-ltk-1 ] At
=A[ l+At+-+....+ - - - - + ..... =Ae
2! (k-1)!

A2t2 Ak-ltk-1 ] At
= I+ At+--+ ......+ - - - - +..... A= e A
[ 2! (k-1)!

I
I • •
The matnx exponential has the property that
eA(t+s} =eAteAs

This can be proved as follows:

e
AtcAs = ( -; ~
k';O
Aktk)(
k!
I
k=O
Aksk)
k!
= }:
k=O
Ak[ i tisk-i ]
i=O i!(k - i)!

=
oo
I
k (t +
A .,;.__- =c
s/
A(t+s)
k=O k!

In particular, if s==-t, then

. cA1 e·''t=c·''1eAt =eM1-1>=1

. r, 1 is c·'''. Since the ·inverse o f cAt• aJways exists,

• eAt·1s nonsrngular.
•
Thus, the inverse o e .
It is important to point out that
e(A+~)1 == e'\l eo1,
if AB=BA • 1·

if AB ;t BA

179

I
I
We shall next obtain ti 1 .
1 . e so 1ution of the co11tin11ous-timc state equation
X == Ax+ nu
.......... (5 .91)
Where x is the state vcct 01. all n x 11 constant
. (n-vcctor), u the input vector (r- v ctor), A
matnx, and B an n
' x r constant matrix.

By writing Eq (5.9!) as

x(t) -Ax(t) == Bu(t)

and premultipl)'ino b0 ti • • . 11 ·
~ 1 SH1cs of this last equation by c· \ we obtain

c- At [ X(t) - Ax(t)] ~ :J-Atx(t)] ~ c-At Bu(t)

Integrating the preceding equation between Oand l gives

t
c-Atx(l) ==x(O) +f c·ArBu(r)dr
I
0
; t
or, x(t)=cAtx(O)+J cA(t-r)u(r)dr ... ..... (5.92)
0

Eq (5.92) is the solution of Eq (5.91). Note that the solution of the state equation starting with
the initial state x(to) is

A(t-t ) . t A(t )
X(t)=e O x(t )+ fe -r U(r)dr ......... (5.93)
0 t
0
1
----·
• .6.2) Discreti:ation of Co11ti11uo11s-Time State-Space Eq11alio11s : In what follows we
shall present a procedure for discrctizing continuos-timc state-space equations. We assume
that the input vcdnr u(l) rharn;cs only al

sampling operation here is fictitious. We shall derive the discrete-time stnle equation and
.--
equal!µ~~ sampling instants. Nott; tl!_..1t . tile · • •

output equation that yield the exact values at t=kT, where k=O, 1,2, .....
Consi lcr the continuous-time state equation and output equation

x = Ax+Bu ......... (5.9-+)

y = Cx+Du
....... .. (5.9 ")

180

r-.
In the following ana!J sis t0 . . n kT and (k+ I)T
• • ·' clanfy the prcsc11tatio11, we 11sc the notat io •
111stcad ofk and k+l T . fi
. . ., • he d1 scrctc•timc representation of Eq (5.94) will tal·c tile onn
x(k + l ) I) - G(·t·) (5 96)
- x(kT) +I l(T) u(kT) " ...... •

r ote that the matrices G nd , . m Jling pcriocJ

. a H depend on the sampling period T. Once the sa I
T 1s l!xcd G a11ct H
' ' arc constant matrices.

We assume tll''t tile •

• sampled and fed to a •zci·o-or<ler
111pt11 u(t) 1s 110 11
( so 11, "... t all til e
- ------ u

components o f u(t) t' vc samp ling

arc co~~stant over the interval between -any two_ £ _ ~--. '
·ostants, or --

u(t) =u(kT), -- for kT :St< kT + T ........ .. (5 .97)

Since ---- -- -

......... (5.98) .•

and
I
I

. . AkTkT - A
.. x(k T) = c AkT ,~(0) e
•
J c r Bu( r)d-r ..........(5 .99)
' 0

Mtiltiplying Eq (5.99) by e·AT;nd subtracting ii from Eq (5.98) gives us

x((k+l)T)=e
AT ~
t:Jf·~. A(k+l)T 1~•lff A
JT c- rBu(r)dr
__.../,.

Since from Eq (5.97) u(t) =u(kT) for kT :S t < kT +T, we may substitute u('t) =u(kT)=
constant in this last equation.
- T ----
• - - - - - -AT · AT - At 1
x((k + 1)T) = e x(kT) + e_: J e Bu(kT)dt
• 0 J

T ·-
=·cAT x(kT) + J ct A,1.Bu(kT)d;{.
0 J
....... .. (5.1 00)

Where A. :::;; L:-=t..-H" we define

' - - \T
G (T) = e'
........ (5.101)

181
 H (T) ~ 0 c'" d,\) B, • ,, ,.,,., (5. I 02)

then Eq (5 .100) b
ccomes
X((k+ l )T) =G(T) x(kT + - , .... .. .. (5.103 ) •
Which is Eq (5 96 ) I l(T)u(k f) . T ·111d H(T).
• ). Thus, Eqs (5 .101) and (5. I 02) give the desired matrices G( ) ' . . . :
. .
Note that G(T) and H . . . (S 95 ) the output
•. ' ( f) depend on the sampling period T. Refcmng to Eq • ' .
equation becomes, ,

y(kT) =Cx(kT) + Du(kT)

• • ---- ---
•.....,.... (5.104) •

Where matrices c. 10 • .. nUng •1;criod T:

U11( arc constant matrices and do not depend on t11c sat;1!;, .
lf rnatri
,
x A i5 nonsingular,
•
• • • t·fi d
then H(T) given by equation (9) can be s11np I ic . to .
• .
. -· ·-
. H(T)=(T eA.?..d).)B=J\- 1(eAT _I)B=(eAT -l)A-1B
0 · -----y- . - - --
---
I
I

Example 6. Consider the continuous-time system given by

G(s) = Y(s) =·_ I_ ,

1
U(s) s+ a
' '
Obtain the continuoLIS-tim~ stale-space representation of the system·, Then discrcti~c the state
equation and output equation and obtain the discrete-time state-space representation of the
system. Also, obtairy the pulse transfer function for the system.

Solution : The continuous-time state-.space.representation of the system is simply, · :,

-u l-\--(-:-.C\ ·' !\· ..

1l.11

1_82
 - ·

Hence th ct·
' c 1scrcti
7.C(
I version
. of ti
le sys1c111 equntions is

X(k + I ) ::: c-aT ·(k I - e- aT

X ) + - - - u{k)
U). j( , vJ a

The pulse transfer fu1 . .

ict1011 for tlrn; system is
F(z) ==C(zl-G) - 1H

{2. I

This result ' of course, agrees with

.
the z transfonn of G(s) where 1l. 1s
•
prcce ded_ by a- sampl
- r-
and zero-order hold. •
------

Example 7 .Obtain the discrete-time slate and output equations and the pulse transfer
function (when the sampling period T=l) of the following continuous-time system: •
Y(s) 1 ..
G(s) = - - = - - ::' .
U(s) s(s+~
which may be rcprcs_ented in state space by the equations
'L •'

[;J;[~ ~V~H~J ". 12

5 Lf ~, +- <7.s~ «) :- u ( SJ
• r

'1 ~\'/ -vt--9

UI ~ 8 /
f LLV" '1 ~ I ~- ►\
-- I
_·.

. ~ ~ll : _r1 1 ~:xci..

y- [1 oJ[ :;] ">Ci(~ y(I/ - u G1-1.·~-L
-- ------.
/

Solution : The desired-time slate equation will have the fonn

x((k+ l)T)=G(T) x(kT) +H(T) u(kT) '/
1crc matrices G(T) and H(T) arc obtained as follows:

• 183
Thus,

2
x.·_ 1((k +. l)T) ] - [ ' -l •(l - c-2T)][x (kT)] [1[T+ t ; - 2 T~ )
[ x ((k . l )l' - L l + _ u(kT)
•2 ,- .) O c-2T x1 (kT) 2 l - 2T
. - ·-
2 (! -. e )

The outrut equation becomes

When the sampling period is 1 sec, or T=l, the discrete-time state equat ion and the output
equation become, rcsrcct ivcly,

x 1(k+l)] = [l 0.-4323-:rx 1(k)] [0.28381 . ,,

[ x2 (k +I) 0 0.13531 x (k) + 0.4323j'u(k)
2

The pulsc-(n111s rcr-ru11ction representation of this system can be obtained as fo llows:

F(z) =C(z!.::_9L;H.;-D

rz- 1 r0.2S3S]
1

~
;::(t l
0
o
- 0.4323 1-
z-0.1353_ lo.43~3 +o

184
 -- _ 0.-1)23
::: [ I o] : - I (:-=if(: ~0. 1353 0.28JSJ
0 I [ (J.4323
:-0.1 353
~
= : ~ ~ ( ·l)
(Z-l)(z-0.1353) !)

== ~.~~DS;( l +0.14S5z- 2
{I - z- I) ( I - O. I 353z- I)

5.7 Problems.
I. .
Uh1ai11 a statc-spa1.:t: r1.:pr1.: st;11tat il)ll of' till.! fo llowing pulst: 1rn11sti;r 1'1 11 1<.: tio11 sy.; tt:111 in
the controllable canonical form.

Y(7.) z-l+Jz-2
,- - - - - . ---,a
: U(z) - I +sz-1 + 6%-2-
2- Obtain a state-space representation of the following pulse transfer function system i 1 1
_

the observable canonic fom1 .

? . .,
Y(z) z- - + 4z--'
U(z) = 1+6z- l + ll z- 2 +G:-J

3. Obtain a state-space representation of the following pulse transfer functi on system in

the diagonal c:monic form.

Y(z) 1+Gz-l +82- 2

U(z) = 1+4z- 1 +3z- 2
8 Obtain :i sl:1tc-space represe ntation or the system described by the equation .
y(k+2) + y(k+ l )+O. l 6y(k) =u(k+ I) +2u(k)

5. Obtain a state~space representation of the follow ing system in the diagonal canonical
fo nn.
-I _?
Y(z) z + 2z -
U(z) = I +0.7z- l +0.1 2z- 2

H35

..,,_
 ,
- - - - - - - . . . . ; . _,

Find the 1n1ls, t 1•• 1 . 1• .

c , 11s ~r f1111 c11011 or th1.: sy stem< Ic fin1 ed IJy.
x(k+ l) :::: 1x(k) +I Iu(k)
Y(k)--: ( X( k) ·I D11( k)

- al I 0
where. G = - a,
0 l , H = hhl
2
I
- a3 0 0
h3
C" =(l 0 o]. r., =ho
7.
Consider the discrcte-ti111c state equations.

Obtain Ilic state transition matrix\(' (k).

186