Inversion of the z-Transform

• Focus on rational z-transform of z −1.

• Apply partial fraction expansion.

• Like bilateral Laplace transforms, ROC must be used to

determine a unique inverse z-transform.

Let
B(z) b0 + b1z −1 + · · · + bM z −M
X(z) = =
A(z) a0 + a1z −1 + · · · + aN z −N
and assume that M < N .
ECE352 1

Inversion of the z-Transform (cont.)

If M ≥ N :
M −N
X
−k B̃(z)
X(z) = fk z +
A(z)
k=0

where B̃(z) has order one less than the denominator polynomial.

• Partial fraction expansion is obtained by factoring the denominator

polynomial into a product of first-order terms.

b0 + b1z −1 + · · · + bM z −M
X(z) = QN
a0 k=1(1 − dk z −1)
N
X Ak
= , if all poles dk are distinct
1 − dk z −1
k=1
ECE352 2

Inversion of the z-Transform (cont.)

z Ak
• Ak (dk )nu[n] ←→ 1−dk z −1
, with ROC |z| > dk .

z Ak
• −Ak (dk )nu(−n − 1) ←→ 1−dk z −1
, with ROC |z| < dk .

If a pole di is repeated r times, then there are r terms in the

partial-fraction expansion associated with that pole:

Ai1 Ai2 Ai r
−1
, −1
, ··· ,
1 − di z (1 − diz ) 2 (1 − diz −1)r
ECE352 3

Inversion of the z-Transform (cont.)

z
• A (n+1)···(n+m−1)
(m−1)! (d i )n
u[n] ←→ A
(1−diz −1)m
, with ROC |z| > di.

z
• −A (n+1)···(n+m−1)
(m−1)! (d i )n
u[−n − 1] ←→ A
(1−diz −1)m
, with ROC
|z| < di.

• ROC of X(z) is the intersection of the ROCs associated with

the individual terms in the partial-fraction expansion.
ECE352 4

Inversion of the z-Transform: Examples

Example 7.9, p574: find the inverse z-transform of

1 − z −1 + z −2
X(z) =
(1 − 12 z −1)(1 − 2z −1)(1 − z −1)

with ROC 1 < |z| < 2.

Using partial fraction expansion:

A1 A2 A3
X(z) = + +
1 − 12 z −1 1 − 2z −1 1 − z −1
1 2 −2
= + +
1 − 12 z −1 1 − 2z −1 1 − z −1

where A1,A2, and A3 are solved the same way as in Laplace transform:

A1 = X(z)(1 − 1/2z −1)|z=1/2 = 1.

ECE352 5

Inversion of the z-Transform: Examples (cont.)

ECE352 6

Inversion of the z-Transform: Examples (cont.)

Applying the given ROC

1 n z 1


• The first term (pole at z = 1/2) is a RSS. Thus, 2 u[n] ←→ 1− 12 z −1
.

• The second term (pole at z = 2) is a LSS. Thus,

z 2
−2(2)nu[−n − 1] ←→ 1−2z −1 .

z −2
• The third term (pole at z = 1) is a RSS. Thus, −2(1)nu[n] ←→ 1−z −1
.

Combining these terms gives

 n
1
x[n] = u[n] − 2(2)nu[−n − 1] − 2(1)nu[n].
2
ECE352 7

Inversion of the z-Transform: Examples (cont.)

Example 7.10, p575: Find the inverse z-transform of

z 3 − 10z 2 − 4z + 4
X(z) = 2
, with ROC |z| < 1
2z − 2z − 4

• X(z) given in terms of z, instead of z −1.

• X(z) is not a proper function of z −1.

Factoring z 3 from the numerator and 2z 2 from the denominator

gives
−1 −2 −3
 
1 1 − 10z − 4z + 4z 1
X(z) = z = zY (z)
2 1 − z −1 − 2z −2 2
ECE352 8

Inversion of the z-Transform: Examples (cont.)

• Factor 12 z is easily incorporated using the time-shift property.

• The term in parentheses, Y (z), must be converted into two

terms, a polynomial function of z −1 and a proper function of
z −1, as

−1
−1 −5z −2
Y (z) = (−2z + 3) +
(1 + z −1)(1 − 2z −1)
−1 1 3
= (−2z + 3) + −1
− −1
, with ROC |z| < 1
1+z 1 − 2z
ECE352 9

Inversion of the z-Transform: Examples (cont.)

Thus, we have

1
X(z) = zY (z)
2
−1 1 3
Y (z) = (−2z + 3) + −1
−
1+z 1 − 2z −1
(apply tables on p784−785)
y[n] = − 2δ[n − 1] + 3δ[n] − (−1)nu[−n − 1] + 3(2)nu[−n − 1]
1
x[n] = y[n + 1]
2
3 1
= − δ[n] + δ[n + 1] − (−1)n+1u[−n − 2] + 3(2)nu[−n − 2].
2 2
ECE352 10

The transfer function

• For LTI discrete-time systems with input x[n] and output y[n]:
? y[n] = x[n] ∗ h[n]
? Y (z) = X(z)H(z), where system transfer function H(z) is
viewed as
Y (z)
H(z) = .
X(z)

• In order to uniquely determine the impulse response from the

transfer function, must know ROC.

• If ROC is not known, other system characteristics such as

stability or casuality must be known.
ECE352 11

The transfer function - Examples

Example 7.13, p580: Find the transfer function and impulse of a causal LTI
system if the input is
x[n] = (−1/3)nu[n]
and the output is
y[n] = 3(−1)nu[n] + (1/3)nu[n].

1
X(z) = , ROC |z| > 1/3
1 + (1/3)z −1
3 1
Y (z) = +
1 + z −1 1 − (1/3)z −1
4
= , ROC |z| > 1
(1 + z −1)(1 − (1/3)z −1)
ECE352 12

The transfer function - Examples (cont.)

Thus, the transfer function is obtained as

4(1 + (1/3)z −1)

H(z) = −1 −1
, with ROC |z| > 1
(1 + z )(1 − (1/3)z )

Partial-fraction expansion:

A B
H(z) = +
1+z −1 1 − 13 z −1
2 2
= +
1+z −1 1 − 13 z −1

Taking inverse z-transform we obtain the system impulse

response
h[n] = 2(−1)nu[n] + 2(1/3)nu[n].
ECE352 13

The transfer function - Examples (cont.)

Problem 7.8, p580: An LTI system has impulse response
h[n] = (1/2)nu[n]. Determine the input to the system if the output
if given by y[n] = (1/2)nu[n] + (−1/2)nu[n].

• The z-transform of system output

1 1
Y (z) = −1
+ −1
, ROC |z| > 1/2
1 − (1/2)z 1 + (1/2)z

• System transfer function

1
H(z) = −1
, ROC |z| > 1/2
1 − (1/2)z
ECE352 14

The transfer function - Examples (cont.)

The z-transform of the system input is

X(z) = Y (z)/H(z)
1 − (1/2)z −1
= 1+
1 + (1/2)z −1
2 z n
= −1
←→ x[n] = 2(−1/2) u[n].
1 + (1/2)z
ECE352 15

The transfer function and difference equation

• For a system described by the difference equation:

N
X M
X
ak y[n − k] = bk x[n − k]
k=0 k=0
PM −k
Y (z) b
k=0 k z
H(z) = = PN
X(z) k=0 ak z
−k

• The transfer function of an LTI system described by a

difference equation is a ratio of polynomials in z −1.

• This form of the transfer function is termed a rational transfer

function.
ECE352 16

The transfer function and difference equation - Examples

Example 7.14, p581: Determine the transfer function and the impulse response
for the causal LTI system described by the difference equation

y[n] − (1/4)y[n − 1] − (3/8)y[n − 2] = −x[n] + 2x[n − 1]

Taking z-transform of both sides gives

Y (z) − (1/4)z −1Y (z) − (3/8)z −2Y (z) = −X(z) + 2z −1X(z)

Y (z) −1 + 2z −1
H(z) = =
X(z) 1 − (1/4)z −1 − (3/8)z −2
−2 1
= −1
+
1 + (1/2)z 1 − (3/4)z −1
h[n] = − 2(−1/2)nu[n] + (3/4)nu[n]
(causal system)
ECE352 17

The transfer function and difference equation - Examples

(cont.)
Example 7.15, p581: Find the difference-equation description of an LTI system
with transfer function
5z + 2
H(z) = 2 .
z + 3z + 2
Dividing both numerator and denominator by z 2, we obtain

Y (z) 5z −1 + 2z −2
H(z) = = .
X(z) 1 + 3z −1 + 2z −2

Cross multiply and then inverse z-transform:

Y (z) + 3Y (z)z −1 + 2Y (z)z −2 = 5X(z)z −1 + 2X(z)z −2

y[n] + 3y[n − 1] + 2y[n − 2] = 5x[n − 1] + 2x[n − 2]
ECE352 18

System causality and stability

• Similar to continuous-time LTI systems, but there are

differences.
z
• System transfer function H(z) ←→ h[n], system impulse
response.

• In order to uniquely determine h[n], must know ROC or other

knowledge of the system characteristics.

• Causal system → h[n] = 0 for n < 0 → H(z) is right-sided

transform.

• Stable system → h[n] absolutely summable → DTFT of x[n]

exists → ROC includes unit circle (z = ejΩ).
ECE352 19

System causality and stability (cont.)

• Assume a pole at z = dk .
? If |dk | < 1 (pole inside unit circle), the pole contributes an
exponentially decaying term to the impulse response.
? If |dk | > 1 (pole outside unit circle), the pole contributes an
exponentially increasing term to the impulse response.

• Conclusion: If a system is causal and stable, then all poles of

H(z) are inside the unit circle.
ECE352 20

System causality and stability (cont.)

ECE352 21

System causality and stability (cont.)

ECE352 22

System causality and stability - Examples

• Read Example 7.16, p584

• Read Example 7.17, p585

Problem 7.10, p585: A stable and causal LTI system is described

by the difference equation

1 1 5
y[n] + y[n − 1] − y[n − 2] = −2x[n] + x[n − 1].
4 8 4
Find the system impulse response.
ECE352 23

System causality and stability - Examples (cont.)

Taking z-transform of both sides of the difference equation, we
obtain
1 −1 1 −2 5
Y (z) + Y (z)z − Y (z)Z = −2X(z) + X(z)z −1
4 8 4
Y (z) −2 + 54 z −1
H(z) = =
X(z) 1 + 14 z −1 − 18 z −2
−2 + 54 z −1
=
(1 − 14 z −1)(1 + 12 z −1)
A B
= 1 −1 +
1 − 4z 1 + 12 z −1
ECE352 24

System causality and stability - Examples (cont.)

Poles at z = 1/4 and z = −1/2, both are inside the unit circle. Because
system is causal, both poles corresponding to right-sided terms.

−2 + 45 z −1
A = =1
1 + 12 z −1
z −1 =4

−2 + 54 z −1
B = = −3
1 − 14 z −1
z −1 =−2
1 −3
H(z) = +
1 − 14 z −1 1 + 12 z −1
 n  n
1 1
h[n] = u[n] − 3 − u[n]
4 2
ECE352 25

Freq. response from poles and zeros

• If ROC of an LTI system transfer function includes the unit circle, frequency
response can be obtained as H(ejΩ) = H(z)|z=ejΩ .

• For a rational transfer function assuming a pth-order pole at z = 0 and an

lth-order zero at z = 0 expressed as
QM −p
b̃z −p k=1 (1 − ck z −1)
H(z) = QN −l
−l −1 )
z k=1 (1 − dk z

where b̃ = bp/al, the frequency response is obtained by substituting

z = ejΩ:
−jpΩ
QM −p −jΩ
jΩ b̃e k=1 (1 − ck e )
H(e ) = QN −l
−jlΩ −jΩ )
e k=1 (1 − dk e
−j(N −M )Ω
QM −p jΩ
b̃e k=1 (e − ck )
= QN −l jΩ
k=1 (e − dk )
ECE352 26

Freq. response from poles and zeros (cont.)

For a particular frequency Ω0, the overall magnitude is evaluated
in terms of the magnitude associated with each pole and zero as
QM −pjΩ0
jΩ0 |b̃| k=1 (e − ck )
|H(e )| = QN −l
jΩ0 − d )
k=1 (e k

and the overall phase is evaluated in terms of the phase

associated with each pole and zero as

arg{H(ejΩ0 )} = arg{b̃} + (N − M )Ω0 +

M
X −p N
X −l
arg{ejΩ0 − ck } − arg{ejΩ0 − dk }
k=1 k=1
ECE352 27

Applications to Filters and Equalizers

• Distortionless transmission:
? A scaling of magnitude
? A constant time delay
Let x(t) be the input to an LTI system. If the system is
distortionless, output must be y(t) = Cx(t − t0), where C is a
constant and t0 is the transmission delay.
? The impulse response of the system is: h(t) = Cδ(t − t0).
? Fourier transform of y(t): Y (jω) = CX(jω)e−jωt0 .
ECE352 28

Distortionless transmission

• The system transfer function: H(jω) = Ce−jωt0

? Magnitude response: |H(jω)| = C
? Phase response: arg{H(jω)} = − ωt0
ECE352 29

Ideal low-pass filters

Consider
e−jωt0 |ω| ≤ ωc

H(jω) =
0 |ω| > ωc
where

• the constant is set to C = 1.

• a finite delay t0 is chosen.

• ωc is the cutoff frequency.

ECE352 30

Ideal low-pass filters (cont.)

To evaluate the filter impulse response h(t), we take the inverse
Fourier transform:
Z ωc
FT 1
H(jω) ←→ h(t) = ejω(t−t0)dω
2π −ωc
−ωc
jω(t−t0)
1 e
=
2π j(t − t0)
−ωc
sin(ωc(t − t0)) ωc ω
c

= = sinc (t − t0)
π(t − t0)) π π

sin(πωt)
where the definition of sinc(ωt) = πωt is applied.
ECE352 31

Ideal low-pass filters (cont.)

Frequency response of ideal low-pass filters. (a) Magnitude

response. (b) Phase response.
ECE352 32

Ideal low-pass filters (cont.)

Time-shifted form of the impulse response of an ideal,

noncausal, low-pass filter for ωc = 1 and t0 = 8.
ECE352 33

Design of filters
Impulse responses of ideal filters are noncausal and infinite
length. These filters are nonimplementable. Practical filters allow

• Passband ripple: 1 −  ≤ |H(jω)| ≤ 1 for 0 ≤ |ω| ≤ ωp, where

ωp is the passband cutoff frequency and  is a tolerance
parameter.

• Stop band ripple: |H(jω)| ≤ δ for |ω| ≥ ωs, where ωs is the

stopband cutoff frequency and δ is the tolerance parameter.

• Transition band: ωs − ωp, a finite width.

ECE352 34

Ideal low-pass filters (cont.)

Tolerance diagram of a practical low-pass filter. The passband,

transition band, and stopband are shown for positive frequencies.
ECE352 35

Approximating functions – Butterworth prototype

Butterworth function of order K:

2 1
|H(jω)| =  2K , K = 1, 2, 3, · · · ,
ω
1+ ωc

• ωc: cutoff frequency.

• K: filter order.

• For prescribed values of tolerance parameters  and δ, the

passband and stopband frequencies are:
 1/(2K)

? ωp = ωc 1−
1−δ 1/(2K)


? ωs = ωc δ
ECE352 36

Approximating functions – Butterworth prototype (cont.)

Magnitude response of Butterworth filter for varying orders.

ECE352 37

Approximating functions – Butterworth prototype (cont.)

• Butterworth filters are maximally flat at ω = 0 (i.e., the first

2K − 1 derivatives of |H(jω)|2 at ω = 0 are equal to zero).

• For any given set of specifications (ωp, ωs, , δ), K and ωc can
be calculated, and hence |H(jω)|2 can be determined.

Given Butterworth function |H(jω)|2, the transfer function H(s)

maybe obtained by using the following procedures:

• Find the 2K pole locations of H(s)H(−s)|s=jω = |H(jω)|2:

s = ωcejπ(2k+1)/(2K), for k = 0, 1, · · · , 2K − 1.

• For a stable system, the K poles on the left half plane belong
to H(s).
ECE352 38

Approximating functions – Butterworth prototype (cont.)

• Read example 8.3, p627.

• Problem 8.3, p628: Find the transfer function of a Butterworth

filter with cutoff frequency ωc = 1 and filter order K = 2.
? The 2K = 4 poles of H(s)H(−s) are determined √ √
to be
2 2
(s = ωcej(2k+1)/(2K)
√ √
, k = 0, 1, 2, 3): s1,2 = 2 ± j 2 ,
s3,4 = − 22 ± j 22 . √ √
? Poles of H(s) are: s3,4 = − 2 ± j 22 . Thus,
2

1
H(s) = √ √ √ √
2 2 2 2
(s + 2 +j 2 )(s + 2 −j 2 )
1
= √
s2 + 2s + 1
ECE352 39

Approximating functions – Chebyshev prototype

Magnitude response of Chebyshev filter for order (a) K = 3 and (b) K = 4 and
passband ripple = 0.5 dB. The frequencies ωb and ωa in case (a) and the
frequencies ωa1 and ωb, and ωa2 in case (b) are defined in accordance with the
optimality criteria for equiripple amplitude response.
ECE352 40

Approximating functions – Chebyshev prototype (cont.)

• Equiripple in the passband.

• Monotonic in the stopband.

• Approximation functions with an equiripple magnitude

response are known collectively as Chebyshev functions.

• A filter designed on this basis is called a Chebyshev filter.

ECE352 41

Frequency transformations

• So far, have considered only low-pass filters.

• High-pass, band-pass, and band-stop filters can be designed

by an appropriate transformation of the independent variable.
? Low-pass to high-pass transformation: s → ωsc , where ωc is
the desired cutoff frequency of the high-pass filter.
s2+ω02
? Low-pass to band-pass transformation: s → Bs , where B
is the bandwidth of the band-pass filter and ω0 is the
midband frequency of the band-pass filter, both measured in
radians per second.