Discrete-Time State Space Analysis

8.3 Discrete-Time Models

Discrete-time systems are either inherently discrete (e.g. models of bank accounts,

national economy growth models, population growth models, digital words) or they

are obtained as a result of sampling (discretization) of continuous-time systems. In

such kinds of systems, inputs, state space variables, and outputs have the discrete

form and the system models can be represented in the form of transition tables.

The mathematical model of a discrete-time system can be written in terms of a

recursive formula by using linear matrix difference equations as

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–41
Here represents the constant sampling interval, which may be omitted for brevity,

that is, we use the following notation

Similarly to continuous-time linear systems, discrete state space equations can

be derived from difference equations (Section 8.3.1). In Section 8.3.2 we show

how to discretize continuous-time linear systems in order to obtain discrete-time

linear systems.

8.3.1 Difference Equations and State Space Form

An th-order difference equation is defined by

  

   

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–42
The corresponding state space equation can be derived by using the same technique

as in the continuous-time case. For phase variable canonical form in discrete-time,

we have

... .. .. .. ... .. ... ..

. . . . .
... ...


  ...

Note that the transformation equations, analogous to the continuous-time case, are

given in the discrete-time domain by

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–43
   

  

  

  

...

 

(which gives the phase variable canonical form)

   

(which by eliminating gives the state space output equation).

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–44
8.3.2 Discretization of Continuous-Time Systems

Real physical dynamic systems are continuous in nature. In this section, we show

how to obtain discrete-time state space models from continuous-time system models.

Integral Approximation Method

The integral approximation method for discretization of a continuous-time linear

system is based on the assumption that the system input is constant during the

given sampling period. Namely, the method approximates the input signal by its

staircase form, that is

The impact of this approximation to the solution of the state space equations,

for , is given by

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–45
 
 "!

 
# # 
 

We can conclude that

$ #
  
$ #  %&! '
  

Note that $ and $ are obtained for the time interval from to . It can easily

be shown that due to system time invariance the same expressions for $ and $

are obtained for any time interval. The procedure can be repeated for time intervals

with initial conditions taken as .

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–46
 For the time instant and for ( , we have

)+*-,%.0/21

*31
4 4

From the above equation we see that the matrices 4 and 4 are given by

1
4 5
)6*-,.0/21 1 1
4 57
*31 ( (

The last equality is obtained by using change of variables as .

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–47
 In a similar manner, the formula for the system output at implies

Comparing this equation with the general output equation of linear discrete-time

systems, we conclude that

8 8

In the case of discrete-time linear systems obtained by sampling continuous-time

linear systems, the matrix 8 , can be determined from the infinite series
<
= =
8 9#: ; ;
=?>@

The matrix 8 can also be obtained either using the Laplace transform method or the

method based on the Cayley–Hamilton theorem and setting in 9A .

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–48
 To find B , we perform integration (see Appendix A—matrix integrals)

B C#D EC#D EF EF B EF

which is valid under the assumption that A is invertible.

Example 8.8: The discrete-time state space model of a continuous-time system

for the sampling period is equal to , is obtained as follows.

B E D
# EGHD E#D EGID
EGHD E#D EGHD E#D
F EGHD E#D
B B EF G E#D EGHD
B B

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–49
 Euler’s Method

Less accurate but simpler than the integral approximation method is Euler’s

method. It is based on the approximation of the first derivative at

Applying this approximative formula to the state space system equation, we have

or

where
J J

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–50
8.3.3 Solution of the Discrete-Time State Equation

We find the solution of the difference state equation for the given initial state

and the input signal . From the state equation

K K

for it follows

K K

K K KL K K K

K K MK KL K K K K
...

NPOQ R
K K KN KNPO  O Q K
R?ST

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–51
 Using the notion of the discrete-time state transition matrix defined by

U UV

we get
VXWY
U U U
Z?[]\

Note that the discrete-time state transition matrix relates the state of an input-free

system at initial time ( ) to the state of the system at any other time ,

that is

U UV

It is easy to verify that the discrete-time state transition matrix has the following

properties

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–52
 (a) ^ _^ ^
b3cedbgf bgfHdbih b3cjdb3h
(b) ^ ` _ ^ ` a ^ a _ ^ ^ ^
b b
(c) k^ ^ ^ k k^

(d) ^ ^ ^ ^

The last property follows from

^ ^ ^ ^
It is important to point out that the discrete-time state transition matrix may be
b
singular, which follows from the fact that ^ is nonsingular if and only if the matrix
^ is nonsingular. In the case of inherent discrete-time systems, the matrix ^
may be singular in general. However, if ^ is obtained through the discretization
procedure of a continuous-time linear system, then
d a d a d l#m
lm
^

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–53
 The output of the system at sampling instant is obtained by substituting

into the output equation, producing

oPpq
n n n n n
r?s]t

Remark 8.1: If the initial value of the state vector is not but t , then

the solution has to be modified into

oPpq
t n t n n t
r?st

Note that for , the following modification must be used

oPpq
n n n
r?sut
oXpq
n n n n n
r?sut

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–54
 Remark 8.2: The discrete-time state transition matrix defined by wv can
be evaluated efficiently for large values of by using a method based on the

Cayley–Hamilton theorem and described in Section 8.5. It can be also evaluated

by using the -transform method, to be derived in the next subsection.

8.3.4 Solution Using the -transform

Applying the -transform to the state space equation of a discrete-time linear system

w w

we get

w w

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–55
 The frequency domain state space vector can be expressed as

x yz x yz x

The inverse -transform of the last equation gives , that is

yz x yz yz x yz x

We conclude that

x yz x yz x{

and
x x yz

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–56
 The inverse transform of the second term on the right-hand side is obtained

directly by the application of the discrete-time convolution, which produces

 |}
|} ~ |} ~ ~ ~
€?‚

We have the required solution of the discrete-time state space equation as

 |}
~ ~ ~
€?u‚

From
~ ~

the system output response is obtained

 |}
~ ~ ~ ~ ~ ~
€?u‚

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–57
 The frequency domain form of the output vector is obtained if the

-transform is applied to the output equation, and is eliminated, leading to

ƒ ƒ „ ƒ ƒ „ ƒ ƒ

From the above expression, for the zero initial condition, i.e. , the

discrete matrix transfer function is defined by

ƒ ƒ ƒ „ ƒ ƒ

Example 8.9: Consider the following discrete-time system

ƒ ƒ ƒ ƒ
† †‡

The discrete-time state transition matrix in the frequency domain is obtained as

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–58
 ‰Š ‰Š
ˆ ‰Š
‹Š ‹Œ ‹Š ‹Œ
 Ž‘’ 
‹Œ Ž‘”“ Ž•– “ Ž—”“ Ž—–“
‹Œ ‹Š ‹Š ‰ ’0“  ”
Ž•” “ Ž•– “ Ž—”“ Ž—–“

The time domain state transition matrix is given by

™ ‹ ‹
‰Ž•˜  “
 Ž—–“ ‰Ž• “ Ž—–“
ˆ ‰Š ˆ ‰Š  ”  ™” 
Ž•” “ Ž—–“ Ž•”“ ˜ Ž—–“
š š š š
Š Š Š ™Š š
˜ š š™ ˜ š
Š ™Š Š ™Š
˜ ˜
Let us find the response of this system due to
š

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–59
 Since the system state transition matrix is already determined, we can use the

derived formula, which produces

œXž
Ÿ
› ›
Ÿ?u¡
ž œ ž œ œPž ž Xœ  Ÿ ž ž Xœ  Ÿ ž
¢ £ ¢ £ Ÿ
ž œ ž œ ž œX Ÿ ž ž œX Ÿ ž
¢ £ Ÿ?¡ ¢ £

This can be accepted as the final result. Note that using known formulas for

series summation (Appendix B), the above formula can be further simplified, and

eventually a closed form solution might be obtained for . However, if we find

in the frequency domain , then in most cases, the inverse –transform will

produce a nice closed formula for . In this example, we have

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–60
 §
¤ ¤ ¤ ¨
¥ ¦ ¨
§
¤ ¤ ¤ ¨
¥ §¦ ¨
¤ ¤ ¤ ¨ ¤ ¤
¥ ¦ ¨ ¥ ¦
© ¥Hª ®¦ ª © i¤ ¥eª j° ª ¦jª
ª «¤
ª«•­¬ ª«—¯¬ ª«•­¬ ª«•¯ ¬
ª ª ¨jª ¦®ª ¦jª
ª«•­¬ ª«—¯¬ ª«•­¬ ª«—¯¬ ª«¤

Applying the inverse –transform we obtain the state response

¤ ± ¤ ± ¤ ± ¤ ± ±
¥ ±¦ ¥ ¦
¤ ± ¤ ¤ ± ¤ ± ±
¥ ¦ ¥ ¦
¤ ± ¤ ± ±
¥ ¦ ¤
¤ ± ¤ ± ± ¥
¥ ¦

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–61
 For the system output response we have
³ ³
³
²

8.3.5 Discrete-Time Impulse and Step Responses

The impulse and step responses of multi-input multi-output discrete-time systems

are defined for zero initial conditions, and calculated using the formulas
³P´ ²
¸ ¸ ¸
µ?¶·

Since the input forcing function is a vector of dimensions , we can define the

impulse and step responses for every input of the system. Introduce the discrete-time

system input function whose all components are zero except for the th component,

that is

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–62
 ¹ ¹ º

Note that
¹ ¹ ¹
»

where ¹ is the th column of the matrix » . The system state and output responses

due to the th component of the input signal are given by

¹ ¼P½¾ ¹ ¹ ¹ ¹ ¹ ¹
» »
¿?ÀÁ

where ¹ is the th column of the matrix » . For ¹ , these formulas

¹
produce the system output impulse response, , due to the delta impulse function

on the th system input and all other inputs equal to zero, that is

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–63
 Â ÃXÄÅ Â
Ê ÊÃPÄÅ Â
ÆÈÇ]É
   Â
Ê Ê ÊÃXÄÅ Â Â

Similarly, with  , we can define the system output step response

due to the unit step function on the th system input and all other inputs set to zero
 ÃXÄÅ ÃXÄÅ Æ
ËÍÌÏÎÏÐ Ê Â Â ÊÃPÄ  Ä Å Â
Æ?ÇuÉ Æ?ÇuÉ
 ÃPÄÅ Æ
ËÑÌÏÎÒÐ Ê ÊÃXÄ  Ä Å Â Â
Æ?ÇÉ

It follows that
 à    Â
ËÑÌÏÎÒÐ ËÒÌÓÎÒÐ ËÒÌÏÎÑÐ
Æ?ÇÉ

The slides contain the copyrighted material from Linear Dynamic Systems and Signals, Prentice Hall 2003. Prepared by Professor Zoran Gajic 8–64