Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
n=
x(n)z
n
donde z es una variable compleja (z C).
Nomenclaturas: Tambien se denota como X(z) Z{x(n)} y
la relaci on entre x(n) y X(z) se indica con:
x(n)
z
X(z)
Convergencia: Como la TZ es una serie innita de potencias,
solo existe cuando se da la convergencia de la serie de potencias
Region de convergencia (Region of Convergence, ROC).
Denida por el conjunto de valores de z que hacen X(z) nita.
X(z) =
n=
|x(n)z
n
| <
Consecuencia: Siempre que hablemos de una TZ hay que dar su
ROC.
46
donde z es una
variable compleja.
3.1. Denicion y propiedades. Causalidad y estabilidad
Denicion: La Transformada Z de una se nal discreta x(n) se
dene como la serie de potencias:
X(z) =
n=
x(n)z
n
donde z es una variable compleja (z C).
Nomenclaturas: Tambien se denota como X(z) Z{x(n)} y
la relaci on entre x(n) y X(z) se indica con:
x(n)
z
X(z)
Convergencia: Como la TZ es una serie innita de potencias,
solo existe cuando se da la convergencia de la serie de potencias
Region de convergencia (Region of Convergence, ROC).
Denida por el conjunto de valores de z que hacen X(z) nita.
X(z) =
n=
|x(n)z
n
| <
Consecuencia: Siempre que hablemos de una TZ hay que dar su
ROC.
46
Representaciones ms usuales
Feng-Li Lian 2004
Z - 3
Signals & Systems NTU-EE
The z-Transform
! The z-Transform of a General Signal x[n]:
Feng-Li Lian 2004
Z - 4
Signals & Systems NTU-EE
The z-Transform
! z-Transform & Fourier Transform:
Feng-Li Lian 2004
Z - 3
Signals & Systems NTU-EE
The z-Transform
! The z-Transform of a General Signal x[n]:
Feng-Li Lian 2004
Z - 4
Signals & Systems NTU-EE
The z-Transform
! z-Transform & Fourier Transform:
Juega el mismo papel en procesado digital
de seales que la Transformada de Laplace
en el anlisis de sistemas continuos.
Obtencin de expresiones entrada-salida.
Simplicacin de estructuras.
Implementacin de estructuras.
Resolucion de ecuaciones en diferencias
Puente entre el diseo analgico y digital
(transformacin bilineal e impulso-invariante)
Usos ms comunes.
2
Procesado Digital de Seales.4 Ingeniera Electrnica.
Universitat de Valncia. Profesor Emilio Soria.
Regin de Convergencia.
En la expresin de la transformada Z
aparece un sumatorio cuyos lmites son !
Ese sumatorio puede converger
(o no!!!) para determinados
valores de la variable compleja z.
Denimos la Regin de Convergencia, tambin conocida como R.O.C, de una
transformada Z a la regin del plano complejo donde la transformada Z converge
Ejemplo
!
R
XX
(m) =
1
2 " #
S
X
w ( ) " e
jmw
$#
#
%
dw
!
S
Y
w ( ) = H( j " w)
2
" S
X
w ( )
!
x(n) = A " e
#" n
!
x(n) = A " sin #" n +$ ( )
!
x(n) = A " e
j" w" n+# ( )
!
y
n
{ } = x
k
" h
n#k
k=#$
$
%
&y(n) = x(k) " h(n # k)
k=#$
$
%
!
x(n) = a
n
" u(n)
!
X(z) = a
n
" u(n) " z
#n
n=#$
$
%
= a
n
" z
#n
n= 0
$
%
= a " z
#1
( )
n
n= 0
$
%
!
r
n
k= 0
"
#
=
1
1$ r
% r <1
!
X(z) =
1
1" a # z
"1
!
R
XX
(m) =
1
2" #
S
X
w
( )
" e
jmw
$#
#
%
dw
!
S
Y
w
( )
= H( j " w)
2
" S
X
w
( )
!
x(n) = A " e
#" n
!
x(n) = A " sin #" n +$
( )
!
x(n) = A " e
j" w" n+# ( )
!
y
n
{ } = x
k
" h
n#k
k=#$
$
%
&y(n) = x(k) " h(n # k)
k=#$
$
%
!
x(n) = a
n
" u(n)
!
X(z) = a
n
" u(n) " z
#n
n=#$
$
%
= a
n
" z
#n
n= 0
$
%
= a " z
#1
( )
n
n= 0
$
%
!
r
n
k= 0
"
#
=
1
1$ r
% r <1
!
X(z) =
1
1" a # z
"1
Hay que recordar
que (Condicion de
R.O.C)
!
R
XX
(m) =
1
2" #
S
X
w
( )
" e
jmw
$#
#
%
dw
!
S
Y
w
( )
= H( j " w)
2
" S
X
w
( )
!
x(n) = A " e
#" n
!
x(n) = A " sin #" n +$
( )
!
x(n) = A " e
j" w" n+# ( )
!
y
n
{ } = x
k
" h
n#k
k=#$
$
%
&y(n) = x(k) " h(n # k)
k=#$
$
%
!
x(n) = a
n
" u(n)
!
X(z) = a
n
" u(n) " z
#n
n=#$
$
%
= a
n
" z
#n
n= 0
$
%
= a " z
#1
( )
n
n= 0
$
%
!
r
n
k= 0
"
#
=
1
1$ r
% r <1
!
X(z) =
1
1" a # z
"1
!
R
XX
(m) =
1
2 " #
S
X
w
( )
" e
jmw
$#
#
%
dw
!
S
Y
w
( )
= H( j " w)
2
" S
X
w
( )
!
x(n) = A " e
#" n
!
x(n) = A " sin #" n +$
( )
!
x(n) = A " e
j" w" n+# ( )
!
y
n
{ } = x
k
" h
n#k
k=#$
$
%
&y(n) = x(k) " h(n # k)
k=#$
$
%
!
x(n) = a
n
" u(n)
!
X(z) = a
n
" u(n) " z
#n
n=#$
$
%
= a
n
" z
#n
n= 0
$
%
= a " z
#1
( )
n
n= 0
$
%
!
r
n
k= 0
"
#
=
1
1$ r
% r <1
!
X(z) =
1
1" a # z
"1
Si y slo si
Feng-Li Lian 2004
Z - 5
Signals & Systems NTU-EE
The z-Transform
! Example 10.1:
Feng-Li Lian 2004
Z - 6
Signals & Systems NTU-EE
The z-Transform
! Example 10.2:
1
Unit circle
e
m
z-plane
a
Figure 3.3 Polezero plot and region
of convergence for Example 3.1.
From Discrete-Time Signal Processing, 2e
by Oppenheim, Schafer, and Buck
1999-2000 Prentice Hall, Inc.
R.O.C
3
Procesado Digital de Seales.4 Ingeniera Electrnica.
Universitat de Valncia. Profesor Emilio Soria.
Regin de Convergencia.
Hay que destacar que diferentes seales discretas
pueden tener la misma transformada Z pero
diferentes R.O.C.
1
Unit circle
e
m z-plane
a
Figure 3.4 Polezero plot and region of convergence for Example 3.2.
From Discrete-Time Signal Processing, 2e
by Oppenheim, Schafer, and Buck
1999-2000 Prentice Hall, Inc.
!
R
XX
(m) =
1
2" #
S
X
w
( )
" e
jmw
$#
#
%
dw
!
S
Y
w
( )
= H( j " w)
2
" S
X
w
( )
!
x(n) = A " e
#" n
!
x(n) = A " sin #" n +$
( )
!
x(n) = A " e
j" w" n+# ( )
!
y
n
{ } = x
k
" h
n#k
k=#$
$
%
&y(n) = x(k) " h(n # k)
k=#$
$
%
!
x(n) = a
n
" u(n)
!
X(z) = a
n
" u(n) " z
#n
n=#$
$
%
= a
n
" z
#n
n= 0
$
%
= a " z
#1
( )
n
n= 0
$
%
!
r
n
k= 0
"
#
=
1
1$ r
% r <1
!
X(z) =
1
1" a # z
"1
Feng-Li Lian 2004
Z - 5
Signals & Systems NTU-EE
The z-Transform
! Example 10.1:
Feng-Li Lian 2004
Z - 6
Signals & Systems NTU-EE
The z-Transform
! Example 10.2:
Feng-Li Lian 2004
Z - 5
Signals & Systems NTU-EE
The z-Transform
! Example 10.1:
Feng-Li Lian 2004
Z - 6
Signals & Systems NTU-EE
The z-Transform
! Example 10.2:
1
Unit circle
e
m
z-plane
a
Figure 3.3 Polezero plot and region
of convergence for Example 3.1.
From Discrete-Time Signal Processing, 2e
by Oppenheim, Schafer, and Buck
1999-2000 Prentice Hall, Inc.
R.O.C
|z|<a
R.O.C
|z|>a
Propiedades.
La R.O.C consiste en un anillo centrado
en el origen.
La R.O.C no contiene ningn polo de la
transformada Z.
Si x(n) es nita la R.O.C es todo el plano
complejo excepto, posiblemente z=0 y
z"!
Si x(n) es causal, x(n)=0 # n<0, la
R.O.C es el exterior de una
circunferencia cuyo radio se
corresponde con el polo de la
transformada Z mayor en valor
absoluto. Si es estrictamente
anticausal, x(n)=0 # n>0, es el interior
(un crculo) de la circunferencia con
radio igual al polo de menor valor
absoluto de la transformada Z. En otras
situaciones se tiene un anillo.
4
Procesado Digital de Seales.4 Ingeniera Electrnica.
Universitat de Valncia. Profesor Emilio Soria.
Propiedades de la Transformada Z.
6/7
e
l
e
c
3
6
0
0
Summary of Properties
Property Time domain z-domain ROC
Linearity ax
1
[n] +bx
2
[n] aX
1
(z) +bX
2
(z) R
x
1
R
x
2
Time-shift x[nn
0
] z
n
0
X(z) R
x
Scaling in z z
n
0
x[n] X(z/z
0
) |z
0
|R
x
Dierentiation
in z
nx[n] z
dX(z)
dz
R
x
Time-reversal x[n] X(1/z) 1/R
x
Conjugation x
[n] X
(z
) R
x
Symmetry
(real)
Im{x[n]} = 0 X(z) = X
(z
)
Symmetry
(imag)
Re{x[n]} = 0 X(z) = X
(z
)
Convolution x
1
[n] x
2
[n] X
1
(z)X
2
(z) R
x
1
R
x
2
Initial value x[n] = 0, n < 0 x[0] = lim
z
X(z)
Aqu el superndice !indica que z=0 z=! pueden aadirse/eliminarse. 1/R
x
hace referencia a
los puntos en los que 1/z pertenece a R
x
5
Procesado Digital de Seales.4 Ingeniera Electrnica.
Universitat de Valncia. Profesor Emilio Soria.
Transformadas Z comunes.
TABLE 3.1 SOME COMMON z-TRANSFORM PAIRS
Sequence Transform ROC
1. [n] 1 All z
2. u[n]
1
1 z
1
|z| > 1
3. u[n 1]
1
1 z
1
|z| < 1
4. [n m] z
m
All z except 0 (if m> 0)
or (if m< 0)
5. a
n
u[n]
1
1 az
1
|z| > |a|
6. a
n
u[n 1]
1
1 az
1
|z| < |a|
7. na
n
u[n]
az
1
(1 az
1
)
2
|z| > |a|
8. na
n
u[n 1]
az
1
(1 az
1
)
2
|z| < |a|
9. [cos
0
n]u[n]
1 [cos
0
]z
1
1 [2 cos
0
]z
1
+ z
2
|z| > 1
10. [sin
0
n]u[n]
[sin
0
]z
1
1 [2 cos
0
]z
1
+ z
2
|z| > 1
11. [r
n
cos
0
n]u[n]
1 [r cos
0
]z
1
1 [2r cos
0
]z
1
+r
2
z
2
|z| > r
12. [r
n
sin
0
n]u[n]
[r sin
0
]z
1
1 [2r cos
0
]z
1
+r
2
z
2
|z| > r
13.
a
n
, 0 n N1,
0, otherwise
1 a
N
z
N
1 az
1
|z| > 0
From Discrete-Time Signal Processing, 2e
by Oppenheim, Schafer, and Buck
1999-2000 Prentice Hall, Inc.
6
Procesado Digital de Seales.4 Ingeniera Electrnica.
Universitat de Valncia. Profesor Emilio Soria.
Polos de la Transformada Z-Tipo de seal.
Existe una relacin directa
entre los polos de la
transformada Z y el tipo de
seal que da lugar a dicha
transformada.
APLICACIN DIRECTA EN
EL DISEO DE SISTEMAS
DE CONTROL DIGITAL
2000 by CRC Press LLC
FIGURE 6.4.3
2000 by CRC Press LLC
Although their Z-transformis identical their ROC is different. Therefore, to nd the inverse Z-transform
the region of convergence must also be given.
FIGURE 6.4.3 (continued)
Z
Z
I I
n
n
I I
n
n
n
n
n
n
u nT u nT z
z
z
z
z
u nT T u nT T z
z
z
z
z
z
z
( ) { } = ( ) =
!
=
!
>
! ! ! ( ) { } = ! ! ! ( )
= ! !
"
#
$
$
%
&
'
'
= ! = !
!
=
!
<
=
(
!
!
=!(
!
!
!
=!(
=
(
)
)
)
)
0
1
1
0
0
1
1 1
1
1
1 1
1
1 1
1
2000 by CRC Press LLC
Example
The Z-transform of the functions of u(nT) and u(nT T) are
FIGURE 6.4.3 (continued)
7
Procesado Digital de Seales.4 Ingeniera Electrnica.
Universitat de Valncia. Profesor Emilio Soria.
Transformada Z unilateral.
Si la secuencia es causal; esto es x(n)
=0 # n<0
Transformada Z unilateral:
Esta transformada exige que las se nales esten denidas en <
n < permitiendo su uso en sistemas que no est an en reposo
(condiciones iniciales = 0). Viene denida por:
X
+
(z) =
n=0
x(n)z
n
El lmite inferior es la diferencia, siempre nulo, independiente de
que la se nal sea causal (x(n) = 0, para n < 0) o no.
Algunas propiedades importantes:
1. No contienen informacion sobre x(n) para n < 0.
2. Es unica solo para se nales causales.
3. Z
+
{x(n)} = Z{x(n)u(n)}.
4. Como x(n)u(n) es causal, la ROC de Z{x(n)u(n)} y, por tanto,
la ROC de Z
+
{x(n)} es siempre exterior a un crculo. CON
TRANSFORMADAS UNILATERALES NO ES NECESARIA
LA DEFINICI
ON DE LA ROC.
55
Se tiene entonces lo que se conoce
como Transformada Z unilateral.
Hay que destacar que cuando se tenga
este tipo de transformadas no hay que
especicar la R.O.C ya que ser
siempre el exterior de un crculo;
evidentemente ya que slo se
considera la parte causal de x(n)
Otras peculiaridades:
1. Retardo temporal:
Si
x(n)
z
+
X
+
(z)
entonces
x(n k)
z
+
z
k
[X
+
(z) +
k
n=1
x(n)z
n
]
2. Adelanto temporal:
Si
x(n)
z
+
X
+
(z)
entonces
x(n + k)
z
+
z
k
[X
+
(z)
k
n=1
x(n)z
n
]
3. Teorema del Valor Final:
Si
x(n)
z
+
X
+
(z)
entonces
lm
n
x(n) = lm
z1
(z 1)X
+
(z)
56
Si se designa el par secuencia discreta-
Transformada Z unilateral por
Se tienen entonces las siguientes
propiedades.
Otras peculiaridades:
1. Retardo temporal:
Si
x(n)
z
+
X
+
(z)
entonces
x(n k)
z
+
z
k
[X
+
(z) +
k
n=1
x(n)z
n
]
2. Adelanto temporal:
Si
x(n)
z
+
X
+
(z)
entonces
x(n + k)
z
+
z
k
[X
+
(z)
k
n=1
x(n)z
n
]
3. Teorema del Valor Final:
Si
x(n)
z
+
X
+
(z)
entonces
lm
n
x(n) = lm
z1
(z 1)X
+
(z)
56
Retardo temporal.
Adelanto temporal.
Otras peculiaridades:
1. Retardo temporal:
Si
x(n)
z
+
X
+
(z)
entonces
x(n k)
z
+
z
k
[X
+
(z) +
k
n=1
x(n)z
n
]
2. Adelanto temporal:
Si
x(n)
z
+
X
+
(z)
entonces
x(n + k)
z
+
z
k
[X
+
(z)
k
n=1
x(n)z
n
]
3. Teorema del Valor Final:
Si
x(n)
z
+
X
+
(z)
entonces
lm
n
x(n) = lm
z1
(z 1)X
+
(z)
56
8
Procesado Digital de Seales.4 Ingeniera Electrnica.
Universitat de Valncia. Profesor Emilio Soria.
Transformada Z inversa. Mtodos (I)
Expansin en fracciones
2000 by CRC Press LLC
(6.7.1)
where the inverse Z-transform of the polynomial can easily be found by inspection.
A proper function (M < N) is of the form
or
(6.7.2)
Because N > M, the function
(6.7.3)
is always a proper function.
Partial Fraction Expansion
Distinct Poles
If the poles p
1
, p
2
, , p
N
of a proper function F(z) are all different, then we expand it in the form
(6.7.4)
where all A
i
are unknown constants to be determined.
The inverse Z-transform of the kth term of (6.7.4) is given by
(6.7.5)
If the signal is causal, the ROC is z > p
max
, where p
max
= max{ p
1
, p
2
, , p
N
}. In this case, all terms
in (6.7.4) result in causal signal components.
Example
(a) If F(z) = z(z + 3)/(z
2
3z + 2) with z > 2 then
F z
N z
D z
b b z b z
a z a z
c c z c z
N z
D z
M
M
N
N
M N
M N
( )
=
( )
( )
=
+ + +
+ + +
= + + + +
( )
( )
! !
! !
!
!
! !
( )
0 1
1
1
1
0 1
1
1
1
!
!
!
F z
N z
D z
b b z b z
a z a z
a M N
M
M
N
N
N
( )
=
( )
( )
=
+ + +
+ + +
" <
! !
! !
0 1
1
1
1
1
0
!
!
,
F z
N z
D z
b z b z b z
z a z a
N N
M
N M
N N
N
( )
=
( )
( )
=
+ + +
+ + +
! !
!
0 1
1
1
1
!
!
F z
z
b z b z b z
z a z a
N N
M
N M
N N
N
( )
=
+ + +
+ + +
! ! ! !
!
0
1
1
2 1
1
1
!
!
F z
z
A
z p
A
z p
A
z p
N
N
( )
=
!
+
!
+ +
!
1
1
2
2
!
Z
!
!
!
#
$
%
&
%
'
(
%
)
%
=
( ) ( )
>
( )
( )
! !
( )
<
( )
#
$
%
&
%
1
1
1
1 p z
p u nT z p
u nT T z p k
k
n
k
n
k
if ROC: causal signal
p if ROC: anticausal signal
k
2000 by CRC Press LLC
(6.7.1)
where the inverse Z-transform of the polynomial can easily be found by inspection.
A proper function (M < N) is of the form
or
(6.7.2)
Because N > M, the function
(6.7.3)
is always a proper function.
Partial Fraction Expansion
Distinct Poles
If the poles p
1
, p
2
, , p
N
of a proper function F(z) are all different, then we expand it in the form
(6.7.4)
where all A
i
are unknown constants to be determined.
The inverse Z-transform of the kth term of (6.7.4) is given by
(6.7.5)
If the signal is causal, the ROC is z > p
max
, where p
max
= max{ p
1
, p
2
, , p
N
}. In this case, all terms
in (6.7.4) result in causal signal components.
Example
(a) If F(z) = z(z + 3)/(z
2
3z + 2) with z > 2 then
F z
N z
D z
b b z b z
a z a z
c c z c z
N z
D z
M
M
N
N
M N
M N
( )
=
( )
( )
=
+ + +
+ + +
= + + + +
( )
( )
! !
! !
!
!
! !
( )
0 1
1
1
1
0 1
1
1
1
!
!
!
F z
N z
D z
b b z b z
a z a z
a M N
M
M
N
N
N
( )
=
( )
( )
=
+ + +
+ + +
" <
! !
! !
0 1
1
1
1
1
0
!
!
,
F z
N z
D z
b z b z b z
z a z a
N N
M
N M
N N
N
( )
=
( )
( )
=
+ + +
+ + +
! !
!
0 1
1
1
1
!
!
F z
z
b z b z b z
z a z a
N N
M
N M
N N
N
( )
=
+ + +
+ + +
! ! ! !
!
0
1
1
2 1
1
1
!
!
F z
z
A
z p
A
z p
A
z p
N
N
( )
=
!
+
!
+ +
!
1
1
2
2
!
Z
!
!
!
#
$
%
&
%
'
(
%
)
%
=
( ) ( )
>
( )
( )
! !
( )
<
( )
#
$
%
&
%
1
1
1
1 p z
p u nT z p
u nT T z p k
k
n
k
n
k
if ROC: causal signal
p if ROC: anticausal signal
k
Ejemplo
2000 by CRC Press LLC
(6.7.1)
where the inverse Z-transform of the polynomial can easily be found by inspection.
A proper function (M < N) is of the form
or
(6.7.2)
Because N > M, the function
(6.7.3)
is always a proper function.
Partial Fraction Expansion
Distinct Poles
If the poles p
1
, p
2
, , p
N
of a proper function F(z) are all different, then we expand it in the form
(6.7.4)
where all A
i
are unknown constants to be determined.
The inverse Z-transform of the kth term of (6.7.4) is given by
(6.7.5)
If the signal is causal, the ROC is z > p
max
, where p
max
= max{ p
1
, p
2
, , p
N
}. In this case, all terms
in (6.7.4) result in causal signal components.
Example
(a) If F(z) = z(z + 3)/(z
2
3z + 2) with z > 2 then
F z
N z
D z
b b z b z
a z a z
c c z c z
N z
D z
M
M
N
N
M N
M N
( )
=
( )
( )
=
+ + +
+ + +
= + + + +
( )
( )
! !
! !
!
!
! !
( )
0 1
1
1
1
0 1
1
1
1
!
!
!
F z
N z
D z
b b z b z
a z a z
a M N
M
M
N
N
N
( )
=
( )
( )
=
+ + +
+ + +
" <
! !
! !
0 1
1
1
1
1
0
!
!
,
F z
N z
D z
b z b z b z
z a z a
N N
M
N M
N N
N
( )
=
( )
( )
=
+ + +
+ + +
! !
!
0 1
1
1
1
!
!
F z
z
b z b z b z
z a z a
N N
M
N M
N N
N
( )
=
+ + +
+ + +
! ! ! !
!
0
1
1
2 1
1
1
!
!
F z
z
A
z p
A
z p
A
z p
N
N
( )
=
!
+
!
+ +
!
1
1
2
2
!
Z
!
!
!
#
$
%
&
%
'
(
%
)
%
=
( ) ( )
>
( )
( )
! !
( )
<
( )
#
$
%
&
%
1
1
1
1 p z
p u nT z p
u nT T z p k
k
n
k
n
k
if ROC: causal signal
p if ROC: anticausal signal
k
con R.O.C |z|>2
2000 by CRC Press LLC
Therefore,
(b) If F(z) = z(z + 3)/(z
2
3z + 2) with 1 < z > 2, then following exactly the same procedure
However, the pole at z = 2 belongs to the negative-time sequence and the pole at z = 1 belongs
to the positive-time sequence. Hence,
Example
To detrmine the inverse Z-transform of F(z) = 1/(1 1.5z
1
+ 0.5z
2
) if (a) ROC: z > 1, (b) ROC: z
< 0.5, and (c) ROC: 0.5 < z < 1, we proceed as follows:
or
(a) f (nT) = 2(1)
n
(1/2)
n
, n ! 0 because both poles are outside the region of convergence z > 1
(inside the unit circle).
(b) f (nT) = 2(1)
n
u(nT T) + (1/2)
n
u(nT T), n " 1 because both poles are outside the region
of convergence (outside the circle z = 0.5).
(c) Pole at 1/2 provides the causal part and the pole at 1 provides the anticausal. Hence,
F z
z
z
z z
A
z
A
z
( )
=
+
!
( )
!
( )
=
!
+
!
3
2 1
2 1
1 2
A
z z
z z
A
z z
z z
z z
1
2
2
1
3 2
2 1
5
3 1
2 1
4 =
+
( )
!
( )
!
( )
!
( )
= =
+
( )
!
( )
!
( )
!
( )
= !
= =
,
F z
z
z
z
z
f nT n
n n
( )
=
!
!
!
( )
=
( )
!
( )
" 5
2
4
1
5 2 4 1 0 or
F z
z
z
z
z
( )
=
!
!
!
5
2
4
1
f nT
n
n
n
n
( )
=
!
( )
"
!
( )
# !
$
%
&
'
&
4 1 0
5 2 1
F z
z
z z
z
z z
A
Bz
z
Cz
z
( )
=
! +
=
!
( )
!
(
)
*
+
,
-
= +
!
+
!
2
2
2
1 5 0 5
1
1
2
1 1
2
. .
F z
z
z
z
z
( )
=
!
!
!
2
1 1
2
f nT u nT T u nT n
n
n
( )
= !
( )
! !
( )
(
)
*
+
,
-
( )
!.< < . 2 1
1
2
Computational Elements
In the design of a discrete-time lter, we will see that this can be
done conveniently as an LTI system with a rational transfer function.
We already know that we can represent a dierence equation
which is initially at rest as an LTI system with a rational transfer
function.
In the design of lters, we rst derive the transfer function and
work back to the dierence equation.
To implement a dierence equation, we need three elements:
+
PWQVR UST //
x
1
[n]
+x
2
[n]
x
2
[n]
//
x
1
[n]
//
a
x[n]
//
ax[n]
z
1 //
x[n1]
//
x[n]
Addition Multiplication Delay
The z-transform oers insight into possible arrangements of these
elements.
2/7
e
l
e
c
3
6
0
0
Computational Elements
In the design of a discrete-time lter, we will see that this can be
done conveniently as an LTI system with a rational transfer function.
We already know that we can represent a dierence equation
which is initially at rest as an LTI system with a rational transfer
function.
In the design of lters, we rst derive the transfer function and
work back to the dierence equation.
To implement a dierence equation, we need three elements:
+
PWQVR UST //
x
1
[n]
+x
2
[n]
x
2
[n]
//
x
1
[n]
//
a
x[n]
//
ax[n]
z
1 //
x[n1]
//
x[n]
Addition Multiplication Delay
The z-transform oers insight into possible arrangements of these
elements.
2/7
e
l
e
c
3
6
0
0
Computational Elements
In the design of a discrete-time lter, we will see that this can be
done conveniently as an LTI system with a rational transfer function.
We already know that we can represent a dierence equation
which is initially at rest as an LTI system with a rational transfer
function.
In the design of lters, we rst derive the transfer function and
work back to the dierence equation.
To implement a dierence equation, we need three elements:
+
PWQVR UST //
x
1
[n]
+x
2
[n]
x
2
[n]
//
x
1
[n]
//
a
x[n]
//
ax[n]
z
1 //
x[n1]
//
x[n]
Addition Multiplication Delay
The z-transform oers insight into possible arrangements of these
elements.
Multiplicador (cte)
Retardo.
Sumador
La primera en ser analizada es la que se conoce como forma
directa I; aplicando la ecuacin en diferencias se tiene
3/7
e
l
e
c
3
6
0
0
Computational Structures
Direct Form I
From the recursive form of the dierence equation, it becomes rela-
tively easy to see how the system could be constructed from adders,
multipliers and delay elements.
The Direct Form I construction is:
+
HOI NJMKL
+
HOI NJMKL
+
HOI NJMKL
z
1
z
1
z
1
x[n]
//
b
0
//
//
b
1
//
//
b
M1
//
//
b
M
OO
OO
OO
+
HOI NJMKL
+
HOI NJMKL
+
HOI NJMKL
z
1
z
1
z
1
//
y[n]
//
1/a
0
oo
a
1
oo
oo
a
N1
oo
oo
a
N
OO
OO
OO
//
However, this not the only possible realisation. 15
Procesado Digital de Seales.4 Ingeniera Electrnica.
Universitat de Valncia. Profesor Emilio Soria.
Anlisis de sistemas L.T.I usando la Transformada Z. Estructuras.
La siguiente es la que se conoce
como forma directa II o cannica
5/7
e
l
e
c
3
6
0
0
Direct Form II
To see another possible realisation, write
Y(z) =
b
0
+ +b
M
z
M
W(z), W(z) =
X(z)
a
0
+ +a
N
z
N
.
By calculating w[n] as an intermediate step, we can actually save
on the number of delay elements required.
This leads to the Direct Form II realisation (here, N = M):
+
HOI NJMKL
+
HOI NJMKL
+
HOI NJMKL
z
1
z
1
z
1
//
x[n]
//
1/a
0
oo
a
1
oo
oo
a
N1
oo
oo
a
N
OO
OO
OO
+
HOINJMKL
+
HOINJMKL
+
HOINJMKL
w[n]
//
b
0
//
//
b
1
//
//
b
N1
//
//
b
N
OO
OO
OO
//
y[n]
3/7
e
l
e
c
3
6
0
0
Direct Form II
To see another possible realisation, write
Y(z) =
b
0
+ +b
M
z
M
W(z), W(z) =
X(z)
a
0
+ +a
N
z
N
.
By calculating w[n] as an intermediate step, we can actually save
on the number of delay elements required.
This leads to the Direct Form II realisation (here, N = M):
+
HOI NJMKL
+
HOI NJMKL
+
HOI NJMKL
z
1
z
1
z
1
//
x[n]
//
1/a
0
oo
a
1
oo
oo
a
N1
oo
oo
a
N
OO
OO
OO
+
HOINJMKL
+
HOINJMKL
+
HOINJMKL
w[n]
//
b
0
//
//
b
1
//
//
b
N1
//
//
b
N
OO
OO
OO
//
y[n]
!
H(z) =
Y(z)
X(z)
=
b
s
" z
#s
s=o
P
$
1# a
k
" z
#k
k=1
N
$
%
&
'
(
)
*
!
Y(z) =
A
k
(1" p
k
# z
"1
)
s
k
$
+O(X(z))
!
g(n) = p
n
" u(n)
g(n) = n
s#1
" p
n
" u(n)
!
w(n) =
1
a
0
" x(n) #a
1
" w(n #1) +... #a
N
" w(n # N)
( )
!
y(n) = b
0
" w(n) +...+b
M
" w(n # M)
5/7
e
l
e
c
3
6
0
0
Direct Form II
To see another possible realisation, write
Y(z) =
b
0
+ +b
M
z
M
W(z), W(z) =
X(z)
a
0
+ +a
N
z
N
.
By calculating w[n] as an intermediate step, we can actually save
on the number of delay elements required.
This leads to the Direct Form II realisation (here, N = M):
+
HOI NJMKL
+
HOI NJMKL
+
HOI NJMKL
z
1
z
1
z
1
//
x[n]
//
1/a
0
oo
a
1
oo
oo
a
N1
oo
oo
a
N
OO
OO
OO
+
HOINJMKL
+
HOINJMKL
+
HOINJMKL
w[n]
//
b
0
//
//
b
1
//
//
b
N1
//
//
b
N
OO
OO
OO
//
y[n]
!
H(z) =
Y(z)
X(z)
=
b
s
" z
#s
s=o
P
$
1# a
k
" z
#k
k=1
N
$
%
&
'
(
)
*
!
Y(z) =
A
k
(1" p
k
# z
"1
)
s
k
$
+O(X(z))
!
g(n) = p
n
" u(n)
g(n) = n
s#1
" p
n
" u(n)
!
w(n) =
1
a
0
" x(n) #a
1
" w(n #1) +... #a
N
" w(n # N)
( )
!
y(n) = b
0
" w(n) +...+b
M
" w(n # M)
Esta estructura se basa en utilizar
los retardos de la variable
intermedia, w(n) a la hora de
calcular la salida del sistema y(n).
Este hecho se traduce en un ahorro
en el nmero de retardos
necesarios para la implementacin
del sistema discreto de tal forma
que esta estructura es la que menos
retardos necesita.
16
Procesado Digital de Seales.4 Ingeniera Electrnica.
Universitat de Valncia. Profesor Emilio Soria.
Anlisis de sistemas L.T.I usando la Transformada Z. Estructuras.
En estas dos estructuras, en el caso de tener
polos complejos se asocian por pares
conjugados para dar estructuras de orden 2.
De forma encubierta estamos aplicando
propiedades de la convolucin (asociativa y
distributiva para ser exactos).
7/7
e
l
e
c
3
6
0
0
Parallel Form
Under certain conditions, we have found that we can write a rational
transfer function in a partial fraction expansion:
H(z) = H
1
(z) + +H
N
(z), H
k
(z) =
A
k
1 d
k
z
1
.
This leads directly to the parallel form of implementation:
x[n]
+
HOI NJMKL
OO
//
y[n]
+
HOINJMKL
z
1
//
oo
d
1
OO
//
A
1
//
H
1
(z)
+
HOI NJMKL
z
1
//
oo
d
N
OO
//
A
N
H
N
(z)
Again, complex conjugate poles should be combined.
7/7
e
l
e
c
3
6
0
0
Parallel Form
Under certain conditions, we have found that we can write a rational
transfer function in a partial fraction expansion:
H(z) = H
1
(z) + +H
N
(z), H
k
(z) =
A
k
1 d
k
z
1
.
This leads directly to the parallel form of implementation:
x[n]
+
HOI NJMKL
OO
//
y[n]
+
HOINJMKL
z
1
//
oo
d
1
OO
//
A
1
//
H
1
(z)
+
HOI NJMKL
z
1
//
oo
d
N
OO
//
A
N
H
N
(z)
Again, complex conjugate poles should be combined.
!
H(z) =
Y(z)
X(z)
=
b
s
" z
#s
s=o
P
$
1# a
k
" z
#k
k=1
N
$
%
&
'
(
)
*
!
Y(z) =
A
k
(1" p
k
# z
"1
)
s
k
$
+O(X(z))
!
g(n) = p
n
" u(n)
g(n) = n
s#1
" p
n
" u(n)
!
w(n) =
1
a
0
" x(n) #a
1
" w(n #1) +... #a
N
" w(n # N)
( )
!
y
k
(n) = x
k
(n) "c
k
# x
k
(n "1) +d
k
# y
k
(n "1)
!
y
k
(n) = A
k
" x
k
(n) +d
k
" y
k
(n #1)
7/7
e
l
e
c
3
6
0
0
Parallel Form
Under certain conditions, we have found that we can write a rational
transfer function in a partial fraction expansion:
H(z) = H
1
(z) + +H
N
(z), H
k
(z) =
A
k
1 d
k
z
1
.
This leads directly to the parallel form of implementation:
x[n]
+
HOI NJMKL
OO
//
y[n]
+
HOINJMKL
z
1
//
oo
d1
OO
//
A1
//
H1(z)
+
HOI NJMKL
z
1
//
oo
dN
OO
//
AN
HN(z)
Again, complex conjugate poles should be combined.
Estructura en paralelo; ahora H(z) se puede factorizar de la siguiente forma
6/7
e
l
e
c
3
6
0
0
Cascade Form
Yet another realisation is possible when H(z) is factorised as (again
assuming M = N):
H(z) =
b
0
a
0
H
1
(z) H
N
(z), H
k
(z) =
1 c
k
z
1
1 d
k
z
1
.
This leads to a cascade form:
+
HOI NJMKL
z
1
//
b
0
/a
0
//
x[n]
oo
d
1
OO
+
HOINJMKL //
//
c
1
OO
H
1
(z)
+
HOI NJMKL
z
1
oo
d
N
OO
+
HOINJMKL //
//
c
N
OO
H
N
(z)
//
y[n]
We combine complex zeros and poles with their complex con-
jugate pairs to avoid implementing complex arithmetic in hard-
ware.
That is, for conjugate pairs of zeros, c
k+1
= c
k
, and poles, d
k+1
=
d
k
, we combine H
k
(z) and H
k+1
(z) to obtain
H
k,k+1
(z) =
1 2Re{c
k
}z
1
+ |c
k
|
2
z
2
1 2Re{d
k
}z
1
+ |d
k
|
2
z
2
.
6/7
e
l
e
c
3
6
0
0
Cascade Form
Yet another realisation is possible when H(z) is factorised as (again
assuming M = N):
H(z) =
b
0
a
0
H
1
(z) H
N
(z), H
k
(z) =
1 c
k
z
1
1 d
k
z
1
.
This leads to a cascade form:
+
HOI NJMKL
z
1
//
b
0
/a
0
//
x[n]
oo
d
1
OO
+
HOINJMKL //
//
c
1
OO
H
1
(z)
+
HOI NJMKL
z
1
oo
d
N
OO
+
HOINJMKL //
//
c
N
OO
H
N
(z)
//
y[n]
We combine complex zeros and poles with their complex con-
jugate pairs to avoid implementing complex arithmetic in hard-
ware.
That is, for conjugate pairs of zeros, c
k+1
= c
k
, and poles, d
k+1
=
d
k
, we combine H
k
(z) and H
k+1
(z) to obtain
H
k,k+1
(z) =
1 2Re{c
k
}z
1
+ |c
k
|
2
z
2
1 2Re{d
k
}z
1
+ |d
k
|
2
z
2
.
6/7
e
l
e
c
3
6
0
0
Cascade Form
Yet another realisation is possible when H(z) is factorised as (again
assuming M = N):
H(z) =
b
0
a
0
H
1
(z) H
N
(z), H
k
(z) =
1 c
k
z
1
1 d
k
z
1
.
This leads to a cascade form:
+
HOI NJMKL
z
1
//
b
0
/a
0
//
x[n]
oo
d
1
OO
+
HOINJMKL //
//
c
1
OO
H
1(z)
+
HOI NJMKL
z
1
oo
d
N
OO
+
HOINJMKL //
//
c
N
OO
H
N(z)
//
y[n]
We combine complex zeros and poles with their complex con-
jugate pairs to avoid implementing complex arithmetic in hard-
ware.
That is, for conjugate pairs of zeros, c
k+1
= c
k
, and poles, d
k+1
=
d
k
, we combine H
k
(z) and H
k+1
(z) to obtain
H
k,k+1
(z) =
1 2Re{c
k
}z
1
+ |c
k
|
2
z
2
1 2Re{d
k
}z
1
+ |d
k
|
2
z
2
.
Otra posible estructura es en
cascada; si la funcin de
transferencia H(z) se puede
factorizar de la siguiente forma
!
H(z) =
Y(z)
X(z)
=
b
s
" z
#s
s=o
P
$
1# a
k
" z
#k
k=1
N
$
%
&
'
(
)
*
!
Y(z) =
A
k
(1" p
k
# z
"1
)
s
k
$
+O(X(z))
!
g(n) = p
n
" u(n)
g(n) = n
s#1
" p
n
" u(n)
!
w(n) =
1
a
0
" x(n) #a
1
" w(n #1) +... #a
N
" w(n # N)
( )
!
w
k
(n) = x
k
(n) +d
k
" w
k
(n #1)
y
k
(n) = w
k
(n) #c
k
" w
k
(n #1)
!
y
k
(n) = A
k
" x
k
(n) +d
k
" y
k
(n #1)
Aplicando una estructura
canonica a cada bloque se
tiene
17
!
h(n) =
1
2
"
#
$
%
&
'
n
( u(n) ) H(z) =
1
1* 0.5( z
*1
x(n) =
1
3
"
#
$
%
&
'
n
( u(n) ) X(z) =
1
1*
1
3
( )
( z
*1
!
Y (z) =
1
1" 0.5# z
"1
#
1
1"
1
3
( )
# z
"1
=
A
1" 0.5# z
"1
+
B
1"
1
3
( )
# z
"1
!
Y (z) =
3
1" 0.5# z
"1
"
2
1"
1
3
( )
# z
"1
!
y(n) = 3"
1
2
#
$
%
&
'
(
n
) 2 "
1
3
#
$
%
&
'
(
n
*
+
,
,
-
.
/
/
" u(n)
!
H(p) =
A
p+b
!
h(t) = A " e
#b" t
t > 0
!
h(n "T ) = A " e
#b" n"T
" u(n)
!
H(z) =
A
1"e
"b#T
# z
"1
La transformada inversa
de H(p) viene dada por
!
h(n) =
1
2
"
#
$
%
&
'
n
( u(n) ) H(z) =
1
1* 0.5( z
*1
x(n) =
1
3
"
#
$
%
&
'
n
( u(n) ) X(z) =
1
1*
1
3
( )
( z
*1
!
Y (z) =
1
1" 0.5# z
"1
#
1
1"
1
3
( )
# z
"1
=
A
1" 0.5# z
"1
+
B
1"
1
3
( )
# z
"1
!
Y (z) =
3
1" 0.5# z
"1
"
2
1"
1
3
( )
# z
"1
!
y(n) = 3"
1
2
#
$
%
&
'
(
n
) 2"
1
3
#
$
%
&
'
(
n
*
+
,
,
-
.
/
/
" u(n)
!
H(p) =
A
p +b
!
h(t) = A " e
#b" t
t > 0
!
h(n "T ) = A " e
#b" n"T
" u(n)
!
H(z) =
A
1"e
"b#T
# z
"1
Si se
muestrea
!
h(n) =
1
2
"
#
$
%
&
'
n
( u(n) ) H(z) =
1
1* 0.5( z
*1
x(n) =
1
3
"
#
$
%
&
'
n
( u(n) ) X(z) =
1
1*
1
3
( )
( z
*1
!
Y (z) =
1
1" 0.5# z
"1
#
1
1"
1
3
( )
# z
"1
=
A
1" 0.5# z
"1
+
B
1"
1
3
( )
# z
"1
!
Y (z) =
3
1" 0.5# z
"1
"
2
1"
1
3
( )
# z
"1
!
y(n) = 3"
1
2
#
$
%
&
'
(
n
) 2"
1
3
#
$
%
&
'
(
n
*
+
,
,
-
.
/
/
" u(n)
!
H(p) =
A
p+b
!
h(t) = A " e
#b" t
t > 0
!
h(n "T ) = A " e
#b" n"T
" u(n)
!
H(z) =
A
1"e
"b#T
# z
"1
Procesado Digital de Seales.4 Ingeniera Electrnica.
Universitat de Valncia. Profesor Emilio Soria.
Anlisis de sistemas L.T.I usando la Transformada Z.
Se puede relacionar el mundo digital con el mundo analgico relacionando las Transformadas ms
utilizadas en esos dominios como son la Transformada Z y la Transformada de Laplace; existiendo dos
formas de hacer este cambio. Estas transformaciones son las que se conocen como Transformacin
bilineal y Transformacin impulso-invariante.
Impulso-invariante.
El bloque bsico de anlisis de sistemas lineales continuos usando transformadas
de Laplace es el denido por (es anlogo a lo que ocurre en procesado digital!!!!).
!
h(n) =
1
2
"
#
$
%
&
'
n
( u(n) ) H(z) =
1
1* 0.5( z
*1
x(n) =
1
3
"
#
$
%
&
'
n
( u(n) ) X(z) =
1
1*
1
3
( )
( z
*1
!
Y (z) =
1
1" 0.5# z
"1
#
1
1"
1
3
( )
# z
"1
=
A
1" 0.5# z
"1
+
B
1"
1
3
( )
# z
"1
!
Y (z) =
3
1" 0.5# z
"1
"
2
1"
1
3
( )
# z
"1
!
y(n) = 3"
1
2
#
$
%
&
'
(
n
) 2 "
1
3
#
$
%
&
'
(
n
*
+
,
,
-
.
/
/
" u(n)
!
H(p) =
A
p+b
!
h(t) = A " e
#b" t
t > 0
!
h(n "T ) = A " e
#b" n"T
" u(n)
!
H(z) =
A
1"e
"b#T
# z
"1
Transformacin bilineal.
Ahora el paso es comparar el operador
integrador en el mundo analgico y en
el digital; en el analgico H(p)=1/p es un
integrador; qu sistema digital hace una
operacin semejante?. Para determinarlo
de forma sencilla hay que plantearse el
signicado de la integral: rea encerrada
por una curva y el eje de abcisas
!
h(n) =
1
2
"
#
$
%
&
'
n
( u(n) ) H(z) =
1
1* 0.5( z
*1
x(n) =
1
3
"
#
$
%
&
'
n
( u(n) ) X(z) =
1
1*
1
3
( )
( z
*1
!
Y (z) =
1
1" 0.5# z
"1
#
1
1"
1
3
( )
# z
"1
=
A
1" 0.5# z
"1
+
B
1"
1
3
( )
# z
"1
!
Y (z) =
3
1" 0.5# z
"1
"
2
1"
1
3
( )
# z
"1
!
y(n) = 3"
1
2
#
$
%
&
'
(
n
) 2"
1
3
#
$
%
&
'
(
n
*
+
,
,
-
.
/
/
" u(n)
!
H(p) =
A
p +b
!
h(t) = A " e
#b" t
t > 0
!
h(n "T ) = A " e
#b" n"T
" u(n)
!
H(z) =
A
1"e
"b#T
# z
"1
!
A(n) = A(n "1) +
x(n) + x(n "1)
( )
2
#T
!
A(z) " 1# z
#1
( )
=
T " X(z)
2
" 1+ z
#1
( )
!
h(n) =
1
2
"
#
$
%
&
'
n
( u(n) ) H(z) =
1
1* 0.5( z
*1
x(n) =
1
3
"
#
$
%
&
'
n
( u(n) ) X(z) =
1
1*
1
3
( )
( z
*1
!
Y (z) =
1
1" 0.5# z
"1
#
1
1"
1
3
( )
# z
"1
=
A
1" 0.5# z
"1
+
B
1"
1
3
( )
# z
"1
!
Y (z) =
3
1" 0.5# z
"1
"
2
1"
1
3
( )
# z
"1
!
y(n) = 3"
1
2
#
$
%
&
'
(
n
) 2 "
1
3
#
$
%
&
'
(
n
*
+
,
,
-
.
/
/
" u(n)
!
H(p) =
A
p+b
!
h(t) = A " e
#b" t
t > 0
!
h(n "T ) = A " e
#b" n"T
" u(n)
!
H(z) =
A
1"e
"b#T
# z
"1
!
A(n) = A(n "1) +
x(n) + x(n "1)
( )
2
#T
!
A(z) " 1# z
#1
( )
=
T " X(z)
2
" 1+ z
#1
( )
!
H(z) =
T
2
1+ z
"1
1" z
"1
#
$
%
&
'
(
!
p =
2
T
1" z
"1
1+ z
"1
#
$
%
&
'
(
!
H(z) =
T
2
1+ z
"1
1" z
"1
#
$
%
&
'
(
!
p =
2
T
1" z
"1
1+ z
"1
#
$
%
&
'
(
18