1 Zeros of LTI Systems

1.1 SISO Systems

N (s)
H(s) = = C(sI − A)−1 B + D
D(s)

Definition: s0 ∈ C is a zero if N (s0 ) = 0.

Assumption: Matrix · ¸
sI − A −B
Q(s) =
C D
is nonsingular for some s ∈ C.

Theorem: Assume s0 is not a pole. Then s0 is a zero of

H(s) = C(sI − A)−1 B + D if and only if Q(s0 ) is singular.
From Linear Algebra:
? Matrix V = XY Z is nonsingular
· ¸ iff X, Y and Z are nonsingular.
A 0
? The triangular matrix is nonsingular iff A and C are nonsingular.
B C
Proof: Verify that
· ¸
sI − A −B
Q(s) =
C D
sI − A 0 (sI − A)−1
· ¸· ¸· ¸
0 sI − A −B
=
C I 0 C(sI − A)−1 B + D 0 I
sI − A 0 (sI − A)−1
· ¸· ¸ · ¸
0 sI − A −B
= .
C I 0 H(s) 0 I
If s0 is not a pole then (s0 I − A) is nonsingular. Therefore
· ¸ · ¸
s 0 I − A 0 s 0 − A −B
(s0 I − A)−1 , , and
C I 0 I
are nonsingular. Therefore Q(s0 ) is singular iff H(s0 ) = 0, i.e., iff s0 is a zero.

WARNING: What if s0 is a pole?

MAE 280A 1 Maurı́cio de Oliveira

1.2 MIMO Systems
Extension to MIMO Systems: s0 is such that H(s) looses rank.
If H is square, this means H(s0 ) is singular.

Normal rank: normalrank Q(s) := maxs∈C {rank Q(s)}.

Definition: s0 ∈ C is a zero if rank Q(s0 ) < normalrank Q(s).

Assumption: Matrix Q(s) has full normal rank.

Theorem: Assume that Q(s) has full-column normal rank. Then s0 is a zero of
H(s) = C(sI − A)−1 B + D if and only if there exists x 6= 0, u such that
µ ¶ · ¸µ ¶
x s0 I − A −B x
Q(s0 ) = = 0. (1)
u C D u

Proof:
Sufficiency: If ∃ s0 ∈ C and x 6= 0, u such that (1) holds, then Q(s0 ) does not
have full-collumn rank, and s0 is a zero accordingµto ¶the definition.
x
Necessity: If ∃ s0 ∈ C is a zero then there exists 6= 0 such that (1) holds.
u
All we have to prove is that x 6= 0. Assume (1) holds with x = 0. Then
µ ¶ · ¸
0 −B
Q(s0 ) = u = 0.
u D
µ ¶
x
But since Q(s) has full-collumn normal rank this implies u = 0 and 6= 0,
u
which is a contradiction.

Theorem: Assume that Q(s) has full-row normal rank. Then s0 is a zero of
H(s) = C(sI − A)−1 B + D if and only if there exists y 6= 0, v such that
· ¸
¡ ∗ ∗¢ ¡ ∗ ∗ ¢ s0 I − A −B
y v Q(s0 ) = y v = 0.
C D

MAE 280A 2 Maurı́cio de Oliveira

1.2.1 Example 1
Consider
  
−1 0 0 1 0 · ¸
1 0 1/2
A =  0 0 0 , B =  1 0 , C= , D = 0,
0 1 1
0 0 0 0 1
and compute
¸ s + 1 0 0 −1 1 0
   
· · 1 1
¸
1 0 1/2  2s
H(s) = 0 s 0 1 0 = s+11 1 .
0 1 1 s s
0 0 s 0 1

Hij (s) has no zeros for all i, j! However

normalrank Q(s) = rank Q(2) = 5,

and
rank Q(1) = 4 < normalrank Q(s).
Indeed
2 0 0 −1 0 1
  
µ ¶  0 1 0 −1 0   2 
 
x  
Q(1) = 0 0 1 0 −1  −2 = 0.
u   
 1 0 1/2 0 0   2 
0 1 1 0 0 −2
Conclusion: s0 = 1 is a zero of H(s)!

MAE 280A 3 Maurı́cio de Oliveira

1.2.2 Example 2
Consider
· ¸ · ¸
−1 0 1 0 £ ¤ £ ¤
A= , B= , C = −1 1 , D= 1 1 .
0 0 0 1
and compute
¤ s + 1 0 −1 1 0
· ¸ · ¸
£ £ ¤ £ s s+1
¤
H(s) = −1 1 + 1 1 = s+1 s
0 s 0 1

s = 0 is a zero of H11 (s) and s = −1 is a zero of H12 (s)! However

normalrank Q(s) = rank Q(2) = 3.

but there exists no s0 ∈ C such that rank Q(s0 ) < 3.

Conclusion: H(s) has no zeros!

MAE 280A 4 Maurı́cio de Oliveira

2 Summary for LTI Systems
... and brief introduction to Matlab.

2.1 Example
Transfer function:
s+1
H(s) =
s2
+ 5s + 6
Observer state space cannonical realization:
· ¸ · ¸
−5 1 1
ẋ = x+ u,
−6 0 1
£ ¤
y= 1 0 x
Poles: ¯ ¯
¯s + 5 −1¯
¯ ¯ = 0, s2 + 5s + 6 = (s + 2)(s + 3) = 0
¯ 6 s ¯
Distinct eigenvalues λ1 = −2, λ2 = −3, ⇒ A is diagonalizable.
Eigenvectors:
· ¸ µ ¶ · ¸
3 −1 1  1 1
(−2I − A)v1 = v = 0, v1 =  T = ,
6 −2 1 3  3 2

· ¸ µ ¶ ⇒ · ¸
2 −1 1  −2 1
(−3I − A)v2 = v = 0, v2 =  T −1 = .
6 −3 2

2  3 −1
Diagonal state space realization:
· ¸ · ¸
−2 0 −1
ẋ = T −1 AT x + T −1 Bu = x+ u,
0 −3 2
£ ¤
y = CT x = 1 1 x.
Time response:
Z t
At
y(t) = Ce x0 + CeA(t−τ ) Bu(τ )dτ,
0
Z t
¤ e−2t 0 ¤ e−2(t−τ )
· ¸ · ¸· ¸
£ £ 0 −1
= 1 1 −3t x0 + 1 1 −3(t−τ ) u(τ )dτ,
0 e 0 0 e 2
Z t³ ´
−3(t−τ ) −2(t−τ )
£ −2t −3t ¤
= e e x0 + 2e −e u(τ )dτ.
0

MAE 280A 5 Maurı́cio de Oliveira

 Impulse Response
1.2

0.8

0.6
Amplitude

0.4

0.2

−0.2
0 0.5 1 1.5 2 2.5
Time (sec)

MAE 280A 6 Maurı́cio de Oliveira

 Step Response
0.25

0.2

0.15
Amplitude

0.1

0.05

0
0 0.5 1 1.5 2 2.5 3
Time (sec)

MAE 280A 7 Maurı́cio de Oliveira

 Linear Simulation Results
1

0.8

0.6

0.4
Amplitude

0.2

−0.2

−0.4

−0.6

−0.8
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec)

MAE 280A 8 Maurı́cio de Oliveira

 Linear Simulation Results
2

1.5

1
Amplitude

0.5

−0.5

−1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Time (sec)

MAE 280A 9 Maurı́cio de Oliveira