Sei sulla pagina 1di 21

Z-Transform

Z-Transform
Linear time-invariante systems:

input: x(n); output: y (n); impulse response: h(n)

X
y (n) = h(k)x(n − k)
k

For example, if x(n) = z n :

X X X
y (n) = h(k)x(n − k) = h(k)z n−k = z n h(k)z −k
k k k

So:

y (n) = H(z)z n
Then, z n are ”eigenvectors” with H(z) as ”eigenvalue”.

Z-Transform
Because of the linearity it is convenient to look at linear
combinations:
X
x(n) = ak zkn
k
X
y (n) = ak H(zk )zkn
k
X
H(z) = h(k)z −k
k

is called the z-transform of h.


It is closely related to the DFT:
X 2π
X (k) = x(n)e −j N
kn

Z-Transform
DTFT:

X
X (ω) = x(n)e −iωn
n=−∞

so, it is the Z-transfor for z = e iω .


Recall: e iω = cos(ω) + isin(ω).
So, if we restrict the z-transform to complex numbers z so that,
|z| = 1 then they are the same thing.

Z-Transform
Z-transform. Example

Let u(n) be the unit step. It is 1 if n ≥ 0 and 0 otherwise.


Consider now x(n) = an u(n).

∞ ∞
X X 1 z
X (z) = an u(n)z −n = (az −1 )n = =
n=−∞
1 − az −1 z −a
n=0

This can be done if |z| > |a|.

Z-Transform
Z-transform. Properties

- It is linear.
- Convolutions go to products (like with the Fourier transform).

x(n) = x1 (n) ∗ x2 (n) =⇒ X (z) = X1 (z)X2 (z)


Because of this, if we recall that h is the impulse response:

y (n) = h(n) ∗ x(n) =⇒ Y (z) = H(z)X (z)


and then

Y (z)
H(z) =
X (z)

Z-Transform
Some system properties.

Causality:
The output depends only on present and past inputs.
It is necessary and sufficient that h(n) = 0, n < 0.
Stability:
If the input has bounded infinity norm then the output does also.
It is necessary and sufficient that the 1−norm of h is bounded.

In frequency domain the region of convergence of H(z) must


contain the unit circle.

Z-Transform
Causal linear time-invariant systems.

An LTI is called causal if:

a0 y (n) + a1 y (n − 1) + ... + aM y (n − M) =

b0 x(n) + b1 x(n − 1) + ... + bN x(n − N)


By taking the Z-transform on both sides, and using the fact that
Z (x(n − k)) = z −k Z (x(n)) we get:

a0 Y (z) + a1 z −1 Y (z) + ... + aM z −M Y (z) =

b0 X (z) + b1 z −1 X (z) + ... + bN z −N X (z)

Z-Transform
Causal linear time-invariant systems.

And then:

Y (z) b0 + b1 z −1 + ... + bN z −N
=
X (z) a0 + a1 z −1 + ... + aM z −M
By convention a0 = 1 and, also, we can take b0 out:

Y (z) 1 1 + c1 z −1 + ... + cN z −N
=
X (z) b0 1 + a1 z −1 + ... + aM z −M
But, as we know, this H(z).
Then we can study the zeros and poles of that expression.
For example, a filter that contains no other y (n − 1), y (n − 2)...
will yield an H(z) with no poles.
An a filter that contains no other x(n − 1), x(n − 2)... will yield an
H(z) with no zeros.
Z-Transform
Causal linear time-invariant systems.

So, the idea is that, by designing rational function with the zeros
and poles that we want, we can, by using Y (z) = H(z)X (z) either
kill or amplify certain frequencies.

Z-Transform
Frequency Response

If we remember the definition of H(z) and the fact that we can


replace z = e iω .
We define:
Frequency response of a system: H(e iω ).

Magitud response of a system as : |H(e iω )|.

Phase response of a system as : ∠H(e iω ).


iω )
We know that we can write: H(e iω ) = |H(e iω )|e i∠H(e .

Z-Transform
Example: ideal delay system

Suppose that y (n) = x(n − nd ).


Take as input x(n) = e iωn . Therefore y (n) = e iωn e −iωnd .
By the definition of H(e iω ) we see that in this case:

H(e iω ) = e −iωnd
Page 41.

Z-Transform
Example: moving average

M2
1 X
y (n) = x(n − k)
M1 + M2 + 1
k=−M1

In this case:
1
h(n) = if − M1 ≤ n ≤ M2
M1 + M2 + 1
Therefore, the frequency response is given by:
M2
1 X

H(e ) = e −iωk
M1 + M2 + 1
k=−M1

The absolute value of the frequency response decays at high


frequencies.
So, moving averages attenuate fast variations. Page 45, graph.

Z-Transform
Example: lowpass filter

Suppose that we want to create an h that does not modify low


frequencies and kills high frequencies.
We can define:

1 if 0 < |ω| < ωc
Hlp =
0 if ωc < |ω| < π
How do we find h so that H has that form?
Z ωc
1 1
hlp (n) = e iωn dω = (e iωc n − e −iωc n )
2π −ωc 2πin
1
= sin(ωc n)
πn

Z-Transform
IIR & FIR

Page 251

Z-Transform
Relationship between phase and magnitud (page 270)

|H(e iω |2 = H(e iω )(H(e iω ))∗


But z −1 = z ∗ if z = e iω . Then:
1
|H(e iω |2 = H(z)H( )
z∗
which is:
−1 )(1 − c ∗ z)
b0 Π M
k=1 (1 − ck z k
C (z) = −1 )(1 − d ∗ z)
a0 ΠN
k=1 (1 − d k z k
So, if we know that H(z) is causal we can, from the poles of C (z)
identify the poles of H.
However, the same can’t be said for the zeros (since there is no
restriction for the zeros of a stable system).

Z-Transform
Relationship between phase and magnitud (page 270)

The poles and zeros of C (z) are conjugate reciprocals. Recall:

1 c d
= 2 2
−i 2
c + di c +d c + d2
So:

1 c d
( )∗ = 2 2
+i 2
c + di c +d c + d2
So, conjugate reciprocals lie in the same direction but with
reciprocal absolute values.
See example 5.11 (page 271).

Z-Transform
All-Pass Systems

Example 5.12: how many stable, causal systems with three poles
and three zeros?
Only 4.
If we do not restrict the number of poles and zeros the number of
possible H(z)s is unlimited.
The frequency response magnitude of

z −1 − a∗
Hap (z)
1 − az −1

is independent of ω, |Hap (e iω )| = 1.
A system that does this is called all-pass system.

Z-Transform
Minimum Phase Systems

So, in order for the system to be stable and causal the poles of the
transfer function have to be inside the unit circle.
However, there are no restriction on the zeros.
In some cases it is useful to impose the condition that the inverse
1
system (the one with transfer function H(z) ) is also stable and
casual.
These type of systems are called minimum-phase systems.

Z-Transform
Minimum Phase Systems

If we now have the function C (z) and we know that the associated
H(z) is a minimum-phase system then H(z) is uniquely
determined.
So, from the square of the magnitude of the frequency response we
can’t uniquely determine H(z) since we could take H and multiply
it by all-pass factors and those can’t be seen from the C (z).
Stable, causal systems can be decomposed as:

H(z) = Hmin (z)Hap (z)

Z-Transform
Minimum Phase Systems

Page 280

Z-Transform

Potrebbero piacerti anche