EE141 Digital Signal Processing (Fall 2003)

Solution Manual

Prepared by Xiaojun Tang and Zhenzhen Ye

Text: A.V. Oppenheim, R.W. Schafer, and J. R. Buck, Discrete-Time Signal

Processing, 2nd edition, Prentice-Hall, 1999.

HW#1: P2.1 (a), (c), (e), (g); P2.4; P2.24

HW#2: P2.5; P2.18; P2.29 (a), (c), (e)
HW#3: P2.40; P2.41; P3.27 (a), (c)
HW#4: P3.6 (d), (e); P3.20; P4.1; P4.3
HW#5: P4.5; P4.7; P5.2; P5.3
HW#6: P5.10; P5.12; P5.15
HW#7: P6.7; P6.8; P6.11; P6.25; P7.15

HW#1: P2.1 (a), (c), (e), (g); P2.4; P2.24

P2.1
(a) T(x[n]) = g[n]x[n];
• Stable if g[n] is bounded;
• Causal – output is not decided by future input;
• Linear – T(ax[n] + by[n]) = ag[n]x[n] + bg[n]y[n] = aT(x[n]) + bT(y[n]);
• Time variant – T(x[n-m]) = g[n]x[n-m];
• Memoryless – output only depends on x[n] with same n;

n + n0
(c) T ( x[n]) = ∑ x[k ] ;
k = n − n0

• Stable;
• Causal only if n0 = 0, else non-causal;
• Linear – T(ax[n] + by[n]) = aT(x[n]) + bT(y[n]);
n − m + n0 n + n0
• Time Invariant – T ( x[n − m]) = ∑ x[k ] =
k = n − m − n0
∑ x[k '−m] ;
k '= n − n0

• Memoryless only if n0 = 0;

(e) T ( x[n]) = e x[ n ] ;
• Stable;
• Causal;
• Nonlinear – T(ax[n] + by[n]) ≠ aT(x[n]) + bT(y[n]);
• Time Invariant – T ( x[n − m]) = e x[ n − m ] ;
• Memoryless – output only depends on x[n] with same n;

(g) T(x[n]) = x[-n];

 • Stable;
• Non-Causal – output depends on future input;
• Linear – T(ax[n] + by[n]) = aT(x[n]) + bT(y[n]);
• Time Variant;
• Not Memoryless;

P2.4
3 1
As y[n] − y[n − 1] + y[n − 2] = 2 x[n − 1] , the Fourier transform is
4 8
3 1
Y (e jw ) − Y (e jw )e − jw + Y (e jw )e − 2 jw = 2 X (e jw )e − jw
4 8
When x[n] = δ [n] ⇒ X (e jw ) = 1 , thus
2e − jw  1 1 
Y (e jw ) = − jw − 2 jw
= 8 − jw
− 
− jw 
1 − (3 / 4)e + (1 / 8)e  1 − (1 / 2)e 1 − (1 / 4)e 
 1 n  1 n 
y[n] = 8   −   u[n]
 2   4  
 

P2.24
As h[n] = [1 1 1 1 –2 –2] for n from 0 to 5 and x[n] = u[n-4], the system response is:
y[n] = x[n] * h[n] = [0 0 0 0 1 2 3 4 2 0 …..]; The sketch is shown as follows:

HW#2: P2.5; P2.18; P2.29 (a), (c), (e)

P2.5
(a) The roots for polynomial 1 − 5 z −1 + 6 z −2 = 0 are 2 and 3, so the homogeneous
response for the system is:
y[n] = A1 2 n + A2 3 n
(b) As y[n] − 5 y[n − 1] + 6 y[n − 2] = 2 x[n − 1] and x[n] = δ [n] , the impulse response of
the system is:
2e − jw  1 1 
H (e jw ) = − jw − 2 jw
= 2 − jw
− − jw 
1 − 5e + 6e  1 − 3e 1 − 2e 
⇒ h[n] = 2(3 n − 2 n )u[n]

(c) As y[n] − 5 y[n − 1] + 6 y[n − 2] = 2 x[n − 1] and x[n] = u[n] , the step response of the
system is:
2 z −1 1 4 3
Y ( z) = H ( z) X ( z) =
(1 − 5 z + 6 z )(1 − z ) 1 − z 1 − 2 z 1 − 3z −1 , z > 3
−1 −2 −1
= −1
− −1
+

⇒ y[n] = (3 n +1 − 2 n + 2 + 1)u[n]

P2.18
(a) h[n] = (1 / 2) n u[n]
Causal, the output of the system does not depend on future input;

(b) h[n] = (1 / 2) n u[n − 1]

Causal, the output of the system does not depend on future input;

n
(c) h[n] = (1 / 2)
Non-Causal, the output of the system depends on future input;

(d) h[n] = u[n + 2] − u[n − 2]

Non-Causal, the output of the system depends on future input;

(e) h[n] = (1 / 3) n u[n] + 3 n u[−n − 1]

Non-Causal, the output of the system depends on future input;

P2.29
As x[n] = [1 1 1 1 1 1/2] for n from –1 to 4,
(a) x[n-2] = [1 1 1 1 1 1/2] for n from 1 to 6; The sketch is:
(c) x[2n] = [1 1 1/2] for n from 0 to 2; The sketch is:

(e) x[n-1]δ[n-3] = x[2]; The sketch is:

HW#3: P2.40, P2.41, P3.27 (a), (c)

P2.40
x[n] = cos(πn)u[n] = (−1) n u[n]
j
h[n] = ( ) n u[n]
2
∞ ∞ n
j j
y[n] = h[n] * x[n] = ∑ h[k ]x[n − k ] = ∑ ( 2 )
k = −∞ k = −∞
k
u[k ](−1) n − k u[n − k ] = (−1) n ∑ (− ) k
k =0 2
n +1
1 − (− j / 2)
= (−1) n
1 − (− j / 2)
1 − (− j / 2) n +1 1
Since lim =
n → ∞ 1 − ( − j / 2) 1+ j / 2
The steady state response to the excitation x[n] = (−1) n u[n] is
1 cos(πn)
(−1) n =
1+ j / 2 1+ j / 2

P2.41
∞
Given a periodic impulse train x[n] = ∑ δ [n + kN ] , we can write its Fourier transform as
k = −∞

2π ∞
X ( e jω ) =
N
∑ δ (ω + 2πk / N )
k = −∞
(1)
(Refer to Signal and Systems, 2nd edition by A.V. Oppenheim and A.S. Willsky, Page 371 for its proof)

In problem, 2.41,N=16, so its Fourier transform is

2π ∞
X ( e jω ) = ∑ δ (ω + 2πk / 16)
16 k = −∞
(2)

Let Y (e jω ) denotes the output of the system, then

Y (e jω ) = X (e jω ) H (e jω ) (3)
If | ω |< 3π / 8 , H (e jω ) = e − jω 3
2π ∞ 2π
= e − jω 3 ∑
16 k = −∞
δ (ω + 2πk / 16) =
16
[δ (ω ) + e j 3π / 8δ (ω + π / 8) + e − j 3π / 8δ (ω − π / 8)] (4)

If | ω |≥ 3π / 8 , H (e jω ) = 0 , thus Y (e jω ) = 0 , (5)
2π
So Y (e jω ) = [δ (ω ) + e j 3π / 8δ (ω + π / 8) + e − j 3π / 8δ (ω − π / 8)] (6)
16
Take the inverse Fourier transform, we can get
1 1
y[n] = (1 + e j 3π / 8 e − nπ / 8 + e − j 3π / 8 e nπ / 8 ) = (1 + e −( n −3)π / 8 + e ( n −3)π / 8 )
16 16
(7)
1 π 1 1 π
= (1 + 2 cos (n − 3)) = + cos( (n − 3))
16 8 16 8 8

Note: Take a look at (3), H (e jω ) is band limited, X (e jω ) is infinite pulse train. If we

multiply them together, we can only consider those pulses falling into the band
( − 3π / 8 , 3π / 8 ). The rest pulses are cancelled due to the multiplication with 0. There are
 2π 2π 2π
three pulses falling into the band δ (ω ) , δ (ω + π / 8) , δ (ω − π / 8) , so we get
16 16 16
(6)

P3.27
(a)
1 A B C D
X ( z) = = + + −1
+
1 1 1
(1 + z −1 ) 2 (1 − 2 z −1 )(1 − 3z −1 ) (1 + z −1 ) 2 1 + z −1 1 − 2 z 1 − 3z −1
2 2 2
X (z)’s poles are z=-1/2, 2, 3, if it is stable, the ROC is | z |∈ (1 / 2,2)
1 1 1
A = X ( z )(1 + z −1 ) 2 z = −1 / 2 = −1 −1 z = −1 / 2 =
2 (1 − 2 z )(1 − 3 z ) 35
1 1568
C = X ( z )(1 − 2 z −1 ) z = 2 = z =2 = −
1 1225
(1 + z −1 ) 2 (1 − 3z −1 )
2
1 2700
D = X ( z )(1 − 3z −1 ) z =3 = z =3 =
1 1225
(1 + z −1 ) 2 (1 − 2 z −1 )
2
Also, Let z −1 = 0 at both sides,
X ( z ) z −1 =0 = 1 = A + B + C + D
58
Thus, B = 1 − A − C − D =
1225
1 1 / 35 58 / 1225 1568 / 1225 2700 / 1225
X ( z) = = + − −1
+
1 −1 2
z ) (1 − 2 z −1 )(1 − 3z −1 )
(1 +
1 −1 2
(1 + z ) 1+ z
1 −1 1 − 2 z 1 − 3 z −1
2 2 2
Since the ROC is | z |∈ (1 / 2,2) ,
1 1 58 1 1568 n 2700 n
x[n] = (n + 1)(− ) n u[n + 1] + (− ) n u[n] + 2 u[− n − 1] − 3 u[−n − 1]
35 2 1225 2 1225 1225

Note: To get the inverse Z-Transform of second-order term or multiple order term, we
dX ( z )
can use the differentiation property nx[n] ↔ − z (Refer to page122 of textbook for
dz
its proof)
1
E.g. right side sequence x[n] = a n u[n] ↔ (ROC: | z |>| a | )
1 − az −1
1
d( −1
) az −1 z −1
na n [n] ↔ − z[ 1 − az ]= , So na n −1
[ n ] ↔
dz (1 − az −1 ) 2 (1 − az −1 ) 2
(c)
z 3 − 2z 2
x[n] ↔ X ( z ) = = z 2 + 2z +
z−2 1 − 2 z −1
X(z) has its only pole at z=2. If x[n] is a left-sided sequence, the ROC is | z |< 2
x[n] = δ [n + 2] + 2δ [n + 1] − 2(2) n u[−n − 1]
or x[n] = δ [n + 2] + 2δ [n + 1] − 2 n +1 u[−n − 1]

P3.1
(g)
9
1 1
( ) n (u[n] − u[n − 10] = ∑ ( ) nδ [n] is a finite length sequence, so its ROC is | z |≠ 0 . The
2 n =0 2
solution in the textbook is right.

HW#4: P3.6 (d), (e); P3.20; P4.1; P4.3

P3.6
1 − (1 / 2) z −1
(d) X ( z ) = z > 1/ 2
1 − (1 / 4) z − 2
• Partial Fraction Expansion:
1 − (1 / 2) z −1 1 − (1 / 2) z −1 1
X ( z) = −2
= =
1 − (1 / 4) z (1 − (1 / 2) z )(1 + (1 / 2) z ) 1 + (1 / 2) z −1
−1 −1

n
 1
⇒ x[n] =  −  u[n]
 2

• Power Series Expansion:

1 − (1 / 2) z −1
X ( z) = −2
= 1 − (1 / 2) z −1 + (1 / 4) z − 2 − (1 / 8) z −3 + (1 / 16) z − 4 + L
1 − (1 / 4) z
n
 1
⇒ x[n] =  −  u[n]
 2

• Fourier Transform exists as the ROC including unit circle.

1 − az −1
(e) X ( z ) = z > 1/ a
z −1 − a
• Partial Fraction Expansion:
1 − az −1 1 az −1 1 a2 − 1/ a a
X ( z ) = −1 = −1 − −1 = −1 − a − −1 = −1
−a+
z −a z −a z −a z −a z − a 1 − (1 / a ) z 1 − (1 / a ) z −1
n +1 n −1 n +1 n −1
1 1 1 1
⇒ x[n] = −  u[n] − aδ [n] +   u[n] = −  u[n] +   u[n − 1]
a a a a

• Power Series Expansion:

 1 − az −1 1 az −1
X ( z ) = −1 = −
z − a z −1 − a z −1 − a
n n +1
1  1  1 1 1   1  1  1
−1
=  − 1 + z −1 + 2 z − 2 + 3 z −3 + L ⇒ x1[n] =  −   u[n] = −  u[n]
z − a  a  a a a   a  a  a
n −1
az −1 1 1 1 1
− −1 = z −1 + z − 2 + 2 z −3 + 3 z − 4 + L ⇒ x 2[n] =   u[n − 1]
z −a a a a a
n +1 n −1
1 1
⇒ x[n] = x1[n] + x 2[n] = −  u[n] +   u[n − 1]
a a

• Fourier Transform exists when the ROC including unit circle, which means
a < 1.

P3.20
(a) As the ROC of X(z) is |z| > 3/4, and the ROC of Y(z) is |z| > 2/3, the ROC of H(z)
should be |z| > 2/3 ;
(b) As the ROC of X(z) is |z| < 1/3, and the ROC of Y(z) is 1/6 < |z| < 1/3, the ROC of
H(z) should be |z| > 1/6 ;

P4.1
As xc (t ) = sin[2π (100t )] and T = 1/400 sec,
 nπ 
x[n] = xc (nT ) = sin[2π (100nT )] = sin  
 2 

P4.3
 nπ 
As xc (t ) = cos[4000πt ] and x[n] = cos   ,
 3 
1
(a) Let x[n] = xc (nT ) ⇒ T =
12,000
5
(b) T is not unique, for example, T =
12,000
HW#5: P4.5; P4.7; P5.2; P5.3

P4.5
(a) From Nyquist Sampling theorem, to avoid aliasing in the C/D converter, the sampling
1
frequency Ω s = ≥ 2Ω m = 2 * 5000 Hz = 10 4 Hz , so Ts ≤ 10 −4 s
Ts
 f cutoff π /8 4
(b) Ω cutoff = Ωs = 10 = 625Hz
2π 2π
f cutoff π /8
(c) Ω cutoff = Ωs = 2 × 10 4 = 1250 Hz = 1.25kHz
2π 2π
Ω f
Note: The relation between digital frequency f and analog frequency Ω is = ,
Ω s 2π
where Ω s is the sampling frequency, f is in radians.

P4.7

(a)
xc (t ) = s c (t ) + αs c (t − τ d )
X c ( j Ω ) = S c ( j Ω )(1 + α e − j Ω τ d )
Consider sampling, x[n] = x c (nT ) , in frequency domain (refer to Eq4.19 in textbook, P147),
2πk
X (e jΩT ) = ∑ X c ( j (Ω −
1 ∞
))
T k = −∞ T
2πk 2πk
X (e jω ) = X (e jΩT ) = ∑ X c ( j (Ω −
1 ∞ 1 ∞
)) = (1 + αe − jΩτ d ) ∑ S c ( j (Ω − ))
T k =−∞ T T k = −∞ T
1 ∞
ω 2πk
= (1 + αe − jωτ d / T ) ∑ S c ( j ( − ))
T k = −∞ T T
(b)
H (e jω ) = 1 + αe − jωτ d / T
(c)
π π π π
1 1 1 1
∫ ∫ ∫ ∫
− jωτ d / T
h[n] = H ( e jω
) e jnω
dω = (1 + αe ) e jnω
dω = e jnω
dω + αe − jω ( n −τ d / T ) dω
2π −π 2π −π 2π −π 2π −π
sin(nπ ) sin[(n − τ d / T )π ]
= +α
nπ (n − τ d / T )π
sin(nπ ) sin[(n − 1)π ]
if τ d = T , h[n] = +α = δ [n] + αδ [n − 1]
nπ (n − 1)π
sin(nπ ) sin[(n − 1 / 2)π ] sin[(n − 1 / 2)π ]
if τ d = T / 2 , h[n] = +α = δ [ n] + α
nπ (n − 1)π (n − 1)π

P5.2
10
y[n − 1] − y[n] + y[n + 1] = x[n]
3
10
z −1Y ( z ) − Y ( z ) + zY ( z ) = X ( z )
3
 Y ( z) 1 z −1 − 3/8 3/8
H ( z) = = = −1 −1
= −1
+
X ( z) 10 (1 − (1 / 3) z )(1 − 3 z ) 1 − (1 / 3) z 1 − 3z −1
z −1 − + z
3
(a) H (z) has two zeroes: 0, ∞ ; two poles: 1/3, 3

(b) The system is stable, so the ROC includes the unit circle. The ROC is 1/3<|z|<3
3 1 3
h[n] = − ( ) n u[n] − 3 n u[− n − 1]
8 3 8

P5.3
1
y[n − 1] + y[n − 2] = x[n]
3
1
z −1Y ( z ) + z − 2Y ( z )) = X ( z )
3
Y ( z) 1 1 z
H ( z) = = −1 −2
= −1 =
X ( z ) z + (1 / 3) z z (1 + (1 / 3) z ) 1 + (1 / 3) z −1
−1

The poles are: 1 / 3 , there are two ROC, 0 <| z |< 1 / 3 , | z |> 1 / 3
1
(1) 0 <| z |< 1 / 3 : h[n] = −(− ) n +1 u[− n − 2] , choose (d)
3
1
(2) | z |> 1 / 3 : h[n] = (− ) n +1 u[n + 1] , choose (a)
3
HW#6: P5.10; P5.12; P5.15

P5.10
As one of the zeros of H(z) is at z = ∞ , the corresponding pole of Hi(z) will be also at
infinity. The existence of a pole at z = ∞ implies that the system is not causal.
P5.12
(a)

As the poles of H(z) are 0.9, -0.9, ROC includes the unit circle, the system is stable.

(b)
(1 + 0.2 z −1 )(1 + 3z −1 )(1 − 3 z −1 )
H ( z) =
(1 + j 0.9 z −1 )(1 − j 0.9 z −1 )
− 9(1 + 0.2 z −1 )(1 + (1 / 3) z −1 )(1 − (1 / 3) z −1 ) ( z −1 + 1 / 3)( z −1 − 1 / 3)
=
(1 + j 0.9 z −1 )(1 − j 0.9 z −1 ) (1 + (1 / 3) z −1 )(1 − (1 / 3) z −1 )
14444444244444443 14444244443
H1 ( z ) H ap ( z )

= H 1 ( z ) H ap ( z )

P5.15
Generalized Linear Phase – GLP;
Linear Phase – LP;

(a) As h[n] = 2δ [n] + δ [n − 1] + 2δ [n − 2] ⇒ H ( jw) = (1 + 4 cos w)e − jw ,

A( jw) = 1 + 4 cos w and α = 1, β = 0 , it is a GLP, but not LP as A( jw) is not always
nonnegative for all w;

(b) It is not a GLP or LP as it is not a symmetric filter;

(c) As h[n] = δ [n] + 3δ [n − 1] + δ [n − 2] ⇒ H ( jw) = (3 + 2 cos w)e − jw ,

A( jw) = 3 + 2 cos w and α = 1, β = 0 , it is a GLP and also LP as A( jw) is always
nonnegative for all w;

(d) As h[n] = δ [n] + δ [n − 1] ⇒ H ( jw) = 2 cos( w / 2)e − j ( w / 2 ) ,

A( jw) = 2 cos( w / 2) and α = 1 / 2, β = 0 , it is a GLP, but not LP as A( jw) is not always
nonnegative for all w;

(e) As h[n] = δ [n] − δ [n − 2] ⇒ H ( jw) = 2 sin we − j ( w−π / 2) ,

A( jw) = 2 sin w and α = 1, β = π / 2 , it is a GLP, but not LP as A( jw) is not always
nonnegative for all w;

HW#7: P6.7; P6.8; P6.11; P6.25; P7.15

P6.7
1 1
The difference equation is: y[n] − y[n − 2] = x[n − 2] − x[n]
4 4
1 1
Z-Transform: Y ( z ) − Y ( z ) z − 2 = X ( z ) z − 2 − X ( z )
4 4
1
− + z −2
Y ( z)
Transfer function: H ( z ) = = 4
X ( z) 1
1 − z −2
4
x[n] -1/4 y[n]

z −1

z −1
1/4

P6.8
y[n] − 2 y[n − 2] = 3x[n − 1] + x[n − 2]

P6.11

z −1 (1 − 2 z −1 )(1 − 4 z −1 ) z −1 − 6 z −2 + 8 z −3
H ( z) = =
1 −1 1
1− z 1 − z −1
2 2
(a)
y[n]

z −1
1/2

z −1
-6

z −1
8
(b) x[n] y[n]

z −1
1/2

z −1
-6

z −1
8

P6.25
1 −1
1− z
2 1
(a) H ( z ) = [ H 1 ( z ) + H 2 ( z )]H 3 ( z ) = [ + (1 + 2 z −1 + z − 2 )]( )
3 −1 7 − 2 1 − z −1
1− z + z
8 8
The Z-Transfer function:
9 9 11 7
2 + z −1 + z − 2 + z −3 + z − 4
H ( z) = 8 8 8 8
11 5 7
1 − z −1 + z − 2 − z −3
8 4 8
(b)
The difference equation:
11 5 7 9 9 11 7
y[n] − y[n − 1] + y[n − 2] − y[n − 3] = 2 x[n] + x[n − 1] + x[n − 2] + x[n − 3] + x[n − 4]
8 4 8 8 8 8 8
(c)
x[n] 2 y[n]

z −1
11/8 9/8

z −1
-5/4 9/8

z −1
7/8 11/8

z −1
7/8
P7.15
Specifications:

(a) Pass band ripple: δ p = 0.05 , A p = 20 log δ p = −26.02dB

Stop band ripple: δ s = 0.1 , As = 20 log δ s = −20dB
Pass band edge: ω p = 0.25π
Stop band edge: ω s = 0.35π
Cutoff: ω c = 0.3π
The peak approximate error 20 log10 δ < −26.02dB
Among the windows in Table7.1 (Page 471), Hanning, Hamming, Blackman can be used
(b)
8π
Hanning: 0.1π = , M = 80 , L = M + 1 = 81
M
8π
Hamming: 0.1π = , M = 80 , L = M + 1 = 81
M
12π
Blackman: 0.1π = , M = 120 , L = M + 1 = 121
M
Note that the estimation is not accurate. We can use MATLAB to find the minimum of
filter order to meet the requirements.