Linear System Theory

 Dr. Vali Uddin

 Hamdard University

 [email protected]

Lecture 2 1
Convolution Integral

 The output of the LTI system is given as

u(t) Linear y(t)

System

t
y( t )   g( t, )u( )d
t t0 t
y( t )   g( t  )u( )d   g( )u( t  )d
t0 t0

Lecture 2 2
Transfer-Function Matrix
 The above convolution leads to the use of
Laplace transform
– This then converts the differential equations
into algebraic equations for easy solutions
as below 
ŷ(s)   y( t )est dt
0
    st
ŷ(s)     g( t  )u( )d e dt
t 0  0 

 
ŷ(s)     g ( t  )u ()d e s ( t ) e sdt
t 0  0 
Lecture 2 3
   
  
ŷ(s)     g( t  )e s ( t )
dt u( )esd
0  t  0 

  s 
ŷ(s)     g( )e d u( )esd

0     
  

ŷ(s)     g( )e d u( )esd
s
0  0 

      s 
 s 
ŷ(s)    g( )e d   u( )e d 
 0  0 

ŷ(s)  ĝ(s)  û(s)

Lecture 2 4
 u(t) y(t)=g(t)*u(t)
System
^u(s) ^y(s)=g(s)u(s)
^ ^

• ĝ (s) ~ Transfer function, the Laplace transform of

the unit impulse response
– For a MIMO system, we have
ŷ(s)  Ĝ(s)  û(s)

 ĝ11(s) ĝ12 (s) ĝ1p (s) 

ĝ (s) ĝ (s) ĝ 2 p (s) ~ Transfer-function
Ĝ (s)   21 22  matrix, or transfer
 
ĝ (s) ĝ (s)  matrix
 q1 q2 ĝ qp (s) 

Lecture 2 5
Introduction to System Theory and
Linear Algebra

State Variable Description

 Start with a general lumped system:
x ( t )  h( x ( t ), u( t ), t )
y( t )  f ( x ( t ), u( t ), t )

• If the system is linear, the above reduce to:

x ( t )  A( t )x ( t )  B( t )u( t )
y( t )  C( t )x ( t )  D( t )u( t )

Lecture 2 6
• If the system is linear and time-invariant, then:
x ( t )  Ax ( t )  Bu ( t )
y( t )  Cx ( t )  Du( t )

• To find an LTI system's response to a particular

input u(t), we can use Laplace transform:
sx̂ (s)  x 0  Ax̂ (s)  Bû(s)
ŷ(s)  Cx̂ (s)  Dû(s)

Lecture 2 7
• Solve the above linear algebraic equations:
x̂ (s)  sI  A 1 Bû(s)  sI  A 1 x 0
 
ŷ(s)  CsI  A 1 B  D û(s)  CsI  A 1 x 0
Transfer function matrix Ĝ(s)
• More will be said for linear time varying
systems later

Lecture 2 8
Linearization
 There are many results on linear systems
while nonlinear systems are generally
difficult to analyze
– What to do with a nonlinear system
described by

x ( t )  h( x ( t ), u( t ), t )
y( t )  f ( x ( t ), u( t ), t )

Lecture 2 9
• Linearization. How? Under what conditions?
– Using Taylor series expansion based on a nominal
trajectory, ignoring second order terms and higher
– Effects are not bad if first order Taylor series
expansion is a reasonable approximation over the
duration under consideration
 Suppose that with xo(t) and uo(t), we
have
x o ( t )  h( x o ( t ), u o ( t ), t )  x o ( t )

– Suppose that the input is perturbed to u o ( t )  u( t )

Lecture 2 10
 – The solution is xo(t) +x(t), withx(t) satisfying

x o ( t )  x ( t )  h( x o ( t )  x ( t ), u o ( t )  u( t ), t )
h h
 h( x o ( t ), u o ( t ), t )  x u  ...
x o u o

 h1 h1 h1   h1 h1 h1 

 ..   .. 
 x1 x 2 x n   u1 u 2 u p 
h h 2 h 2   h 2 h 2 h 2 
h  2 h  ..
u p 
..
  x1 x 2 x n ,  u1 u 2
x   u  
: : .. :  : : .. : 
 h h n h n 
 n ..   h n h n
..
h n  ~ Jacobians
 x1 x 2 x n   u u 2 u p 
 1
Lecture 2 11
 Then the perturbed system can be
described by
h h
x ( t )  x u ~ A linear system
x o u o

– The above is valid if the first order Taylor series

expansion works out well within the time duration
under consideration
• What to do with the output y(t) = f(x(t), u(t), t)?
– The output equation can be similarly linearized, but
many times there is no need for linearization unless
with output feedback
Lecture 2 12
Modeling of Selected Systems

 We shall briefly go over the following

systems
– Electrical Circuits
– Integrator/Differentiator Realization
– Operational Amplifiers

Lecture 2 13
Electrical Circuits i L
State variables? + + +
y
u(t) C - v R
– i and v - -

• How to describe the evolution of the state variables?

di
L uv di 1 1
dt  v u
dt L L
dv v dv 1 v
C i  i
dt R dt C RC

State Equation: Two first-order differential

equations in terms of state variables and input

Lecture 2 14
 i
Output equation: y=v  0 1    0u
v

In matrix form:  di   0 1 
 1
     i 
dt 
 dv   1
L
     L u
1 v  
     0
 dt   C RC 

Lecture 2 15
 Steps to obtain state and output
equations:
Step 1: Pick {iL, vC} as state variables
di
L L  vL
Step 2: dt Express vL and iC in terms of state
dv C variables and input using KVL and KCL
C  iC
dt
di L 1
 v L (state var iables , input , nothing else )
Step 3: dt L
dv C 1
 i L (state var iables , input , nothing else )
dt C
Step 4: Put the above in matrix form
Step 5: Do the same thing for y in terms of state variables and
input, and put in matrix form
Lecture 2 16
Example R1
i1
L1 L2
i
 State variables? u(t)
+ + 2
C - v R2
+
y
-
– i1, i2, and v, -

• State and output equations?

di1 R 1 1
di   1 i1  v  u
L1 1  u  R1i1  v dt L1 L1 L1
dt
di 2 di 2 R 1
L2  v  R 2i 2   2 i2  v
dt dt L2 L2
dv dv 1 v
C  i1  i 2  i1  i 2
dt dt C C
 di1   R1 0 
1
1
 dt   L1 L1   i     i1 
 di 2   1     1
1 L
  0 
R2
 i2  0 u y  R 2i 2  0 R 2 0i 2 
L2       
 dt   L2
 v 
 dv   1 1   v   0 
 dt   C  0   
C 
Lecture 2 17
Integrator/Differentiator Realization
 Elements: Multipliers, differentiators, and
integrators
Multiplier Differentiator Integrator
y(t) =
f(t) y(t) = af(t) f(t) y(t) = df/dt f(t) t
a S 1/S  f ( )d  y( t 0 )
 t 0
• Are they LTI elements? Yes
• Which one has memory? What are their dimensions?
– Integrator has memory. Dimensions: 0, 0, and 1, respectively
• They can be interconnected in various ways (assuming no
loading effect) to form LTI systems
– Number of state variables = number of integrators
– Linear differential equations with constant coefficients
Lecture 2 18
 y(t)  What are the state
+ variables?
1/s 1/s
u(t) - x2 x1
 Select output of
2 integrators as SVs
• What are the state and
x 1  x 2 output equations?
x 2  u  2x1 y = x2

 x 1   0 1  x1  0  x1 
x    2 0 x   1 u y  0 1    0u
 2    2   x 2 
A B C D
• Linear differential equations with constant coefficients

Lecture 2 19
  Steps to obtain state and output equations:
Step 1: Select outputs of integrators as state variables
Step 2: Express inputs of integrators in terms of state
variables and input based on the interconnection of
the block diagram
Step 3: Put in matrix form
Step 4: Do the same thing for y in terms of state
variables and input, and put in matrix form

u u – No. of state variables?

s s
u(t) c – 0
b
+ + + y(t) y  u  bu  cu
+

Lecture 2 20
 .
u-ay-by
+
-. .. 1/s . 1/s
u y y y
ay
a
+ +
b
by

Input/Output Description
y   by  ay  u

s2 ŷ(s)   bŷ(s)  as ŷ(s)  û(s)

s 2

 as  b ŷ(s)  û(s) ŷ(s)  2
1
s  as  b
û(s)

ĝ(s)
Lecture 2 21
 State Variable Description
-bx1-ax2+u x2 x1
+
- 1/s 1/s
u y
a
+ +
b
x 1  x 2 x 2  bx1  ax 2  u

 0 1   0
x    x  u y  x1  1 0 x  0 u
 b  a 1
1
 s  1   0 1
ĝ(s)  CsI  A  B  D  1 0
1   0  2
 b s  a  1 s  as  b

Lecture 2 22
 Operational Amplifiers (Op Amps)
v0 = A (va - vb), with -VCC  v0  VCC
Vcc, 15V
Non-inverting
terminal, va Vo
Output, vo
ia + Vcc
io va - vb
o
ib -
Inverting Slope = A
terminal, vb -Vcc
-Vcc, -15V

Usually, A > 104

– Ideal Op Amp:
 A   ~ Implying that (va - vb)  0, or va  vb

 ia  0 and ib  0

– Problem: How to analyze a circuit with ideas Op Amps

Lecture 2 23
 i2 i1 + i 2 = 0
R1 R2
- u1  u 2 y  u 2
+ u2 +  0
u1(t) i1 + + y R1 R2
- u2(t) -
- R2  R2 
y u1  1  u 2
R1  R1 
Delineate the relationship between input and output
Input/Output description
 Key ideas: Pure gain, no SVs

– Make effective use of ia = ib = 0 and va = vb

– Do not apply the node equation to output terminals
of op amps and ground nodes, since the output
current and power supply current are generally
unknown
Lecture 2 24
 C
Node 1, v(t)2
Node 2, y(t)
R2
Input/Output Description
+
+ ~ Use node equations
+ R1 y
u(t) C1
+ -
- - -

uv yv d y  v  vy

 C1y
Node 1:   C 2 0 Node 2:
R1 R2 dt R2

v  y  R 2C1y v  y  R 2C1y d v  y 
 R 2C1y
dt
u  ( y  R 2C1y ) y  ( y  R 2C1y )
  R 2C1C2y  0
R1 R2
R1  R 2 1 1
y  y  y u Input/Output description
R1R 2C2 R1R 2C1C2 R1R 2C1C2
Lecture 2 25
 C2 -v State and output equations
2 R2
+
+
What are the state
+ R v1 + v2
1 v1
+
y variables?
u(t) C1
+ -
- - -
v1 State variables: v1 and v2

dv1 v1  v 2   v1 v 2 dv 2 u  v1  v 2  v 2

C1   C2  
dt R2 R2 dt R1 R2

 dv1   0 1 
 dt   R 2C1   v1   0 
 dv 2     1      1 u
 1  1 1   v 2   
    R C
 dt   R1C2 C2  R1 R1  
1 2

 v1 
y  v1  1 0    0u
v 2 
Lecture 2 26
THE END

Lecture 2 27