You are on page 1of 15

1.

Unstable Pole-Zero Cancelled System


E(z)

z+2

U (z)

z + 0.5

e(n)

Y (z)

y(n)

z + 0.5

z+2

Solution of cancelled system:


xs(k + 1) = 0.5xs(k) + e(k)
ys(k) = xs(k)
Iteratively solving,
ys(k) = (0.5)k xs(0) +

k1
X

(0.5)me(k m 1)

m=0

Stable.

Digital Control

Kannan M. Moudgalya, Autumn 2007

2.

System without Pole-Zero Cancellation


E(z)

z+2

U (z)

z + 0.5

Y (z)

z+2

Can verify for first block:


First block: x1(k + 1)
u(k)
Second block: x2(k + 1)
y(k)
Substituting, x2(k + 1)


x1(k + 1)
x2(k + 1)

0.5x1(k) + 1.5e(k)
x1(k) + e(k)
2x2(k) + u(k)
x2(k)
2x2(k) + x1(k) + e(k)


  
0.5 0
x1(k)
1.5
=
+
e(k)
1 2 x2(k)
1



 x1(k)
y(k) = 0 1
x2(k)

Digital Control

=
=
=
=
=

Kannan M. Moudgalya, Autumn 2007

3.

System without Pole-Zero Cancellation


E(z)

z+2

U (z)

Y (z)



 z + 0.5  
  
z+2
x1(k + 1)
0.5 0
x1(k)
1.5
=
+
e(k)
x2(k + 1)
1 2 x2(k)
1



 x1(k)
y(k) = 0 1
x2(k)
k1
X
= CAk x(0) +
CAmbe(k m 1)
m=0

Can show that this is equal to




k1
X

3
x
(0)
1
+
(0.5)me(k m 1)
(0.5)k (2)k
x1(0) + 1.5x2(0)
2
m=0

x(0) 6= 0, results in y(k) being unbounded! Whence 2?

Digital Control

Kannan M. Moudgalya, Autumn 2007

4.

Diagonalization

A: square matrix j : jth eigenvalue xj : jth eigenvector


Ax1 = 1x1
..
Axn = nxn
Stacking these side by side,

|
|
1
0
|
|
...

A x1 xn = x1 xn
|
|
|
|
0
n
AS = S
Assume the eigenvectors to be independent S 1 exists
A = SS 1
Digital Control

Kannan M. Moudgalya, Autumn 2007

5.

Diagonalization
A = SS 1
A2 = SS 1SS 1 = S2S 1
A3 = S2S 1SS 1 = S3S 1
..
Ak = Sk S 1

Easy to evaluate RHS:


k
k

1
0
1
0
...
...
=

k =
0
n
0
kn
This approach is used to arrive at the solution.

Digital Control

Kannan M. Moudgalya, Autumn 2007

6.

Condition for Cancellation of Poles and Zeros


E(z)

z+2

U (z)

z + 0.5

Y (z)

z+2

Solution of system after cancellation:


ys(k) = (0.5)mxs(0) +

k1
X

(0.5)me(k m 1)

m=0

Solution of system if there is no cancellation:




k1
X

3
x
(0)
1
+
(0.5)me(k m 1)
(0.5)k (2)k
x1(0) + 1.5x2(0)
2
m=0

Two solutions are identical only if


x1(0) = 1.5x2(0) or
x1(0) = x2(0) = 0.

Digital Control

Kannan M. Moudgalya, Autumn 2007

7.

Aryabhattas Identity

We need to solve the polynomial equation


X(z)D(z) + Y (z)N (z) = C(z)
for X and Y , with D, N and C specified.
Has a solution iff the GCD of D(z) and N (z) divides C(z)
There are infinitely many solutions to Aryabhattas identity
Unique solution under special conditions. Suppose D and N
are coprime with degrees dD > 0 and dN > 0. If
0 dC < dD + dN
there is a unique least degree solution given by
dX(z) < dN (z)
dY (z) < dD(z)
Digital Control

Kannan M. Moudgalya, Autumn 2007

8.

Algorithm for Aryabhattas Identity


X(z)D(z) + Y (z)N (z) = C(z)

is the same as



 D(z)
X(z) Y (z)
= C(z)
N (z)
Solved by comparing the coefficients of powers of z 1. Same as
V (z)F (z) = C(z)
Can be written as

[V0 + V1z 1 + + Vv z v ][F0 + F1z 1 + + FdF z dF ]


= C0 + C1z 1 + + CdC z dC
v is an unknown. All Vi are also unknowns.
Digital Control

Kannan M. Moudgalya, Autumn 2007

Algorithm for Aryabhattas Identity


[V0 + V1z 1 + + Vv z v ][F0 + F1z 1 + + FdF z dF ]
= C0 + C1z 1 + + CdC z dC
is the same as

F0 F1 FdF 0 0
0 F0 F1 FdF 0
[V0 V1 Vv ] ..

0 0 F0 F1 FdF
= [C0 C1 CdC ]
with maximal v. Ensures all combination of products are used.
Multiplying, first two equations are obtained as
V0F0 = C0, V0F1 + V1F0 = C1
9.

Matrix equation is written as


VF =C
Digital Control

Kannan M. Moudgalya, Autumn 2007

10.

Algorithm for Aryabhattas Identity

Recall from the previous slide:

[V0 + V1z 1 + + Vv z v ][F0 + F1z 1 + + FdF z dF ]


= C0 + C1z 1 + + CdC z dC
Can be written as,
VF =C
IF F is right invertible, the solution is,
V = C F 1
For this, rows of F have to be linearly independent.
Choose v as the largest integer, satisfying this requirement.
From V , V (z) and then
 X(z), Y (z) can be determined,
because, V = X Y
Digital Control

10

Kannan M. Moudgalya, Autumn 2007

11.

Summary: Algorithm for Aryabhattas Identity


X(z)D(z) + Y (z)N (z) = C(z)



 D(z)
X(z) Y (z)
= C(z)
N (z)
V (z)F (z) = C(z)

Can be written as

F0 F1 FdF 0

 0 F0 F1 FdF
V0 V1 Vv
..
0 0 F0 F1

= C0 C1

0
0

FdF

CdC

Write this as V F = C , solve for V as V = C F 1


Digital Control

11

Kannan M. Moudgalya, Autumn 2007

12.

Example: Algorithm for Aryabhattas Identity

Solve for X(z) and Y (z) satisfying


X(z)D(z) + Y (z)N (z) = C(z)
with
D(z) = 1 5z 1 + 4z 2
N (z) = z 1 + z 2
C(z) = 1 z 1 + 0.5z 2
Construct F :


   
 
1
2
1 5z + 4z
1
5 1
4 2
F (z) =
=
+
z +
z
z 1 + z 2
0
1
1
which is of the form F0 + F1z 1 + F2z 2.
Digital Control

12

Kannan M. Moudgalya, Autumn 2007

13.

Example: Algorithm for Aryabhattas Identity

Recall


   
 
1
2
1 5z + 4z
1
5 1
4 2
F (z) =
=
+
z
+
z
z 1 + z 2
0
1
1
which is of the form F0 + F1z 1 + F2z 2. Explore v = 1:




1 5 4
F = F0 F1 F2 =
0 1 1
Although independent, v is not the largest. Explore v = 2:

1 5 4 0


0 1 1 0
F0 F1 F2 0

F =
=

0 F0 F1 F2
0 1 5 4
0 0 1 1
These rows are independent. Hence v is still not the maximum.
Digital Control

13

Kannan M. Moudgalya, Autumn 2007

14.

Example: Algorithm for Aryabhattas Identity

Explore the possibility of v = 3:

1
0

F0 F1 F2 0 0
0
F = 0 F0 F1 F2 0 =
0

0 0 F0 F1 F2
0
0

5
1
1
0
0
0

4
1
5
1
1
0

0
0
4
1
5
1

0
0

4
1

No longer independent. Remove one row and solve:

1
1 5 4 0 0


 0 1 1 0 0

V = 1 1 0.5 0 0
0 1 5 4 0
0 0 1 1 0
0 0 1 5 4
Digital Control

14

Kannan M. Moudgalya, Autumn 2007

15.

Example: Algorithm for Aryabhattas Identity

V is a 1 5 vector.
To account for the row removed from F , we add a zero to
V . We obtain


V = 1 3.25 | 0.75 3 | 0 0
where we have separated coefficients of powers of z 1 with
vertical lines.




We have V0 = 1 3.25 and V1 = 0.75 3 .
We obtain
X(z) = 1 + 0.75z 1
Y (z) = 3.25 3z 1
These form the solution to Aryabhattas identity
Digital Control

15

Kannan M. Moudgalya, Autumn 2007

You might also like