CONTENTS

Chapter 1

Chapter 2

11

Chapter 3

51

Chapter 4

89

Chapter 5

191

Chapter 6

211

Chapter 7

233

Chapter 8

293

Chapter 9

, 319

Chapter 10

333

Chapter 11

355

Chapter 12

375

iii

Chapter 1

1.1
(a)
(b)
(c)
(d)
(e)

One dimensional, multichannel, discrete time, and digital.

Multi dimensional, single channel, continuous-time, analog.
One dimensional, single channel, continuous-time, analog.
One dimensional, single channel, continuous-time, analog.
One dimensional, multichannel, discrete-time, digital.

1.2
(a)
(b)
(c)
(d)

/
/
/
/

=
=
=
=

2 ^ i = 2J5 = periodic with Np = 200.

J f e ( i ) = 1 => Periodic with Np = 7.
| = | => periodic with Np = 2.
- => non-periodic.

(e) / = # () = 8 =* p eriodic th N? = 10 1.3

(a)
(b)
(c)
(d)
(e)

Periodic with period Tp T?-.

/ = => non-periodic.
/ = y|^ => non-periodic.
cos(|) is non-periodic; cos(Ig-) is periodic; Their product is non-periodic.
cos(Ij-) is periodic with period JVp=4
5 n
' ( i r ) *s P er idic w ^ h period Np=16
cos(Z2. + ) i s periodic with period 2Vp=8
Therefore, x(n) is periodic with period NP=16. (16 is the least common multiple of 4,8,16).

1.4
(a) w = ^ - implies that f = jf- Let
a=

GCDof (A, AT), i.e.,

Jb = * ' a , A r = Ar' Q .

Then,
Jb'
/ = , which implies that

*' = *:
a
1

(b)
N
k
GCD(k,N)

N0

7
0 1234567
71111117
17777771

(c)
N
k

Nv

16
0 1 2 3 4 5 6 7 8 9 10 11 12 . . . 16
1 6 1 2 1 4 1 2 1 8 1 2 1 4 . . . 16
1 6 8 16 4 16 8 16 2 16 8 164 . . . 1

GCD(k,N)

1.5
(a) Refer to fig 1.1
(b)

Figure 1.1:

x(n)

xa(nT)
za(n/F,)
35z'n(7rn/3) =>

-(-)

(c)Refer to fig 1.2

Q""Q

O
,'20

t(ms)

$...->

Figure 1.2:
(d) Yes.
100u\
z ( l ) = 3 = 3 s i n ( - - ) => F, = 200 samples/sec.
F.

1.6
(a)
x(n)

=
=

ylcos(2TFon/F s + 0)
Acos(27r(T/Tp)n + 6)

But 77TP = / => x(n) is periodic if f is rational.

(b) If x(n) is periodic, then f=k/N where N is the period. Then,

Td

= (jT)

= k(%r)T = kTp.

Thus, it takes k periods {kTp) of the analog signal to make 1 period (Td) of the discrete signal.
(c) Td = kTp => NT = ibTp => / = ife/AT = T/T p = f is rational => x(n) is periodic.

1.7
(a) Fmax = 10kHz => F, > 2Fmax = 20kHz.
(b) For F, = 8*#z, F f o l d = F 3 /2 = 4fc#2 => 5ifctfz will alias to 3kHz.
(c) F=9kHz will alias to 1kHz.

1.8
(a) F m ax = 100kHz, F, > 2 F m a x = 200Hz.
(b) /fold = h =

125

^"
3

1.9
(a) F m a x = 360Hz, FN = 2 i W = 720Hz.
(*>) F fold = ft = Hz.
(c)
x(n)

x fl (nT)
^(n/F,)
sm(4807m/600) + 3'n(7207rn/600)
sin(47rn/5) 3sin(4Trn/5)
-2sin(4irn/5).

x(n)

Therefore, w = 4TT/5.
(d) ya(t) = x(F,t) = -2sm(4807rt).

1.10
(a)
Number of bits/sample
F,

=
_

/op 2 1024=10.
[10,000 bits/sec]
[10 bits/sample]
1000 samples/sec.
bOOHz.

=
=

io\d

(b)
^max

FN

1800*
^
2ir
900Hz]
= 2Fmax = 1800#z.

(<0
/i

h =
But / 2
Hence, x(n)
^a

\)

A -

^ ~

=
=

600* 1
2* {F/
0.3;
1800* 1
{
2*
F/
0.9;
0.9 > 0 . 5 = / 2 = 0.1.
3co$[(2*)(0.3)n] + 2cos[(2*)(0.1)n]

m " ~ g m i n _ 5 - ( - 5 ) _ _io_
m-l

~~

1023

1023-

1.11
x(n)

za(nT)
/l00xn\

V 200 /
4

f250nn\

V 200 /

=
T

3cos

(=)-. (2=)

' = ISO =>(') = (/!*)

ya(<)

3cos(500?rt) - 2s:n(750irt)

1.12
(a) For F, = 300tfz,
x(n)

3cos (-T-J + lOsin(Tn) cos (--]

3CO5

(T)-3COS(T)

(b) x r (/) = 3cos(100007rt/6) - cos(100007rt/3)

1.13
(a)
Range
m

=
_
=
=
=

(b) m = 1 + ^

xmax - ^min = 12.7.

, range
1 + T
A
1 2 7 + 1 = 128 =>/oy 2 (128)
7 bits.

= 636 => /op2(636) => 10 bit A/D.

1.14
*

( 2 0 f!ES!) x ( 8 Ji!

sec
,bits
160
sec

^foid = f2
Resolution

=
=

sample

= l0Hz

lvolt
28-l
0.004.

1.15
(a) Refer to fig 1.3. With a sampling frequency of 5kHz, the maximum frequency that can be
represented is 2.5kHz. Therefore, a frequency of 4.5kHz is aliased to 500Hz and the frequency of
3kHz is aliased to 2kHz.
(b) Refer to fig 1.4. y(n) is a sinusoidal signal. By taking the even numbered samples, the
sampling frequency is reduced to half i.e., 25kHz which is still greater than the nyquist rate. The
frequency of the downsampled signal is 2kHz.
5

Ft - SKHz, F0-500HZ

SO

F - 5KHz, F0-2000HZ

100

50

F - 5 K H z , F0-3000H2

100

F t > SKHz. F0-4500HZ

100

1.16
(a) for levels = 64, using truncation refer to fig 1.5.
for levels = 128, using truncation refer to fig 1.6.
for levels = 256, using truncation refer to fig 1.7.
(b) for levels = 64, using rounding refer to fig 1.8.
for levels = 128, using rounding refer to fig 1.9.
for levels = 256, using rounding refer to fig 1.10.
(c) The sqnr with rounding is greater than with truncation. But the sqnr improves as the
number of quantization levels are increased.
(d)
levels
theoretical sqnr
sqnr with truncation
sqnr with rounding

64
128
43.9000 49.9200
31.3341 37.359
32.754
39.2008

256
55.9400
43.7739
44.0353

The theoretical sqnr is given in the table above. It can be seen that theoretical sqnr is much
higher than those obtained by simulations. The decrease in the sqnr is because of the truncation
and rounding.

FO - 2KH2, Fs-50kHz

tov* - 64, using trunatfon, SQNR - 31.3341 dB

Figure 1.5:

levels 128. using truncation. SQNR 37 3590B

100
>n

150

200

100
>n

150

200

Figure 1.6:

levels x 256. using truncation. SONR-43 7739dB

100
>n

150

200

Figure 1.7:

lewte - 64, uting roundng, SQNR-32.754dB

Figure 1.8:

tmmi% 128. tning roundng, SQNR39.2006d3

Figure 1.9:

levels = 256, using rounding, SQNR=44.0353dB

100
>n

150

200

Figure 1.10:

10

Chapter 2

2.1
(a)

x(n) = j . . . O , i , ? , l , l , l , l , 0 , . . . j
Refer to fig 2.1.
(b) After folding s(n) we have

Z0
1/3

-3

.2

-1

Figure 2.1:

x ( - n ) = | . .0,1,1,1,1, | , i , 0 , . . . | .
After delaying the folded signal by 4 samples, we have

x(-n + 4) = I.. .0,0,1,1,1,1,|, i,0,... j .

On the other hand, if we delay x(n) by 4 samples we have

x(n - 4) = j . . . 0,0, i , | , 1,1,1,1,0,.. X .

Now, if we fold x(n 4) we have
ar(_n-4)=|...0,l,lllJl,|>i,0J0,...J

11

(c)

*(-n + 4)=|...0,l,l,l,l,|,i,0,...J
(d) To obtain x(n -f k), first we fold x(n). This yields x(-n). Then, we shift x(n) by k
samples to the right if k > 0, or k samples to the left if k < 0.
(e) Yes.
1
2
x(n) = -6{n - 2) + -(n + 1) + u(n) - u(n - 4)

2.2
*(n) = {...0,1,1,1,1,1,1,0,...}
(a)
x(n-2) = {...0,0,1,1,1,1,1,1,0,...}
(b)
x(4-n)=j...0,1,1,1,1,1,1,0,

(see 2.1(d))
(0
x(n + 2) = i . . . 0 , 1 , 1 , 1 , 1 , 5 , 5 , 0 , . . . !
(d)

x(n)ti(2-n)={...0,1,1,1,1,0,0,...}
()

x ( n - l ) 6 ( n - 3 ) = i...0,0,1,0,...1

(0
x(n2)

{...0,x(4),x(l),x(0),x(l),x(4),0,...}

{...0,1,1,1,1,1,0,...}

(g)

, /\ _ *(n) + *(~n)

*e( n )

x(-n)

'

{...0,5,5,1,1,1,1,0,...}
f

1 1 1 , , , * 1 *A
12

(h)

,.() = *()-*(-)
v '

-{

1 1 1 1 1 1
0.-i.-i.-2.0.0..2'i'i

2.3
(a)
C 0, n < 0
ti(n) - u(n - 1) = 6(n) = I 1, n = 0
I 0, n > 0

0>)

* ( - * > = ( 1,

n>0

2.4
Let

*o(n)= - [ x ( n ) - x ( - n ) j .
Since
x(-n) = x e (n)
and
x 0 (-n) = -x 0 (n),
it follows that
x(n) = x e (n) + x 0 (n).
The decomposition is unique. For
x(n)= 12,3,4,5,61,
we have
*.(n)=|4,4,4,4,4|
and
x0(n) = j - 2 , - l , 0 , l , 2 | .
13

2.5
First, we prove that
oo

x e (n)x(n) = 0

n=oo
oo

oo

x e (n)x 0 (n)

n oo

xe(-m)x0(-m)

m=-oo
oo

*e(m)x 0 (m)

m=oo
oo

= - E x'(n)x(n)
n=oo
oo

*e(n)x e (n)

n=oo

Then,

E *'() = E [*.() + *.()]*

n = oo

n = oo

*J()+ E

ns-oo

*5()+ E

ns-oo

fi=oo

2.6
(a) No, the system is time variant. Proof: If
*( n ) y(n)
*(-*)-Vi(n)

x n

=
=
=

( *)
x[(n-*)2]
x(n 2 + k2 - 2nk)

y ( n - *)

(b) (1)

x(n)= jo.l.l, 1,1,0,... I

(2)

y(n)=x(n 2 ) = {...,0,1,1,1,0,...}
(3)

y ( n - 2 ) = {...,0,0,1,1,1,0,...}
14

2x c (n)x 0 (n)

(4)
x ( n - 2 ) = | . ..,0,0,1,1,1,1,0,..A
(5)

y2(n) = T[x(n - 2)] = / . . . , 0,1,0,0,0,1,0,..A

(6)
J/2(n) ^ y(n 2) => system is time variant.
(0 (1)

*(n)= jl, 1,1,1 j

(2)

y(n) = {l T ,0,0,0,0,-1}
(3)

y ( n - 2 ) = {o,0,1,0,0,0,0,-lj
(4)

x ( n - 2 ) = jo,0,1,1,1,1,1 J
(5)

y2(n) = | o , o , i , o , o , o , o , - i |
(6)

yi(n) = y{* ~ 2).

The svstem is time invariant, but this example alone does not constitute a proof.
(d) (1*)
y(n) = nx(n),
: ( n ) = | . . . ,0,1,1,1,1,0,. .A
(2)

y( n ) = | . . . ,0,1,2,3,..A
(3)

y(n - 2 ) = | . . . , 0 , 0 , 0 , 1 , 2 , 3 , . . A
(4)

:(n-2)=|...,0,0,0,1,1,1,1,..A
15

(5)
y 2 (n) = T [ x ( n - 2 ) ] = { . . . , 0 , 0 , 2 , 3 , 4 , 5 , . . . }
(6)
SfcC") 7^ y(n 2) => the system is time variant.

2.7
(a) Static, nonlinear, time invariant, causal, stable.
(b) Dynamic, linear, time invariant, noncausal and unstable. The latter is easily proved.
For the bounded input x(k) = u(Jb), the output becomes

*)=I>={'+2, "t-l
since y(n) oo as n oo, the system is unstable.
(c) Static, linear, timevariant, causal, stable.
(d) Dynamic, linear, time invariant, noncausal, stable.
(e) Static, nonlinear, time invariant, causal, stable.
(f) Static, nonlinear, time invariant, causal, stable.
(g) Static, nonlinear, time invariant, causal, stable.
(h) Static, linear, time invariant, causal, stable.
(i) Dynamic, linear, time variant, noncausal, unstable. Note that the bounded input
x(n) = u(n) produces an unbounded output.
(j) Dynamic, linear, time variant, noncausal, stable.
(k) Static, nonlinear, time invariant, causal, stable.
(1) Dynamic, linear, time invariant, noncausal, stable.
(m) Static, nonlinear, time invariant, causal, stable.
(n) Static, linear, time invariant, causal, stable.

2.8
(a) True. If
vi(n) = 7 i [ * i ( n ) ] and
v 2 (n) = T i M n ) 3 ,
then
cnxi(n) + a 2 x 2 ( n )
yields
c*ivi(n) + a2t>2(n)
by the linearity property of T\. Similarly, if
yi(n) = T 2 [vi(n)] and
!/2(n) = T 2 [v 2 (n)],
then
ftfiO-O + &v 2 (n) y(n) = 0iyi(n) + / ^ ( n )
by the linearity property of T 2 . Since
vi(n) = Ti[xi(n)] and
16

v 2 (n) = T 2 [x 2 (n)],
it follows that
Aix 1 (n) + A 2 x 2 (n)
yields the output
A1T[x1(n)) +

A2T{x2(n)l

where T = TiT 2 . Hence T is linear.

(b) True. For 7i, if
x(n) v(n) and
x(n ib) v(n k),
For T 2 , if
f ( n ) y(")

andr(n fc) y(n k).

Hence, For T1T2, if
x(n) y(n) and
x(n - k) -* y{n - k)
Therefore, T = T1T2 is time invariant.
(c) True. 7i is causal => v(n) depends only on x(k) for ib < n. 7^ is causal => y(n) depends only on v(ib) for k
n. Therefore, y(n) depends only on x(ib) for ib < n. Hence, T is causal.
(d) True. Combine (a) and (b).
1
(e) True. This follows from hi(n) * /2(n) = /2(n) * /i(n)
'*
(f) False. For example, consider
7i : y(n) = nx(n) and
T2 : y(n) = nx(n + 1).
Then,
72[Tx[6(n)]]

T2(0) = 0.

7i[T2[*(n)]]

Ti[(n+1)]

- $ ( n + l)

5* 0.

(g) False. For example, consider

T\ ' y( n ) = *( n ) + & and
T2 : y(n) = x(n) 6, where 6 ^ 0 .
Then,
T[x(n)] = 7 2 pi[*(n)]] = 7 2 [x(n) + 6] = x(n).
Hence T is linear.
(h) True.
Ti is stable => v(n) is bounded if x(n) is bounded.
T2 is stable = y(n) is bounded if u(n) is bounded .
17

Hence, y(n) is bounded if x(n) is bounded =$ T = T\T2 is stable.

(i) Inverse of (c). T\ and for T2 are noncausal => T is noncausal. Example:
Ti : y(n)
T2 : y(n)
=>T:y(n)

=
=
=

x(n + 1) and
z(n - 2)
x(n - 1),

which is causal. Hence, the inverse of (c) is false.

Inverse of (h): T\ and/or T2 is unstable, implies T is unstable. Example:
7i : y(n) = ex^n\

stable and T2 : y(n) = /n[x(n)], which is unstable.

But T : y(n) = x(n), which is stable. Hence, the inverse of (h) is false.

2.9
(a)
y(n)

22

h(k)x(n-k),x{n)

0,n<0

koo
n+N

y(n + N)

n+N

h(k)x(n + N-k)=

22

k = 00

*(Jb)r(n - it)

i=-oo

n+N

= J2 h(k)x{n-k)+ 22 M*)*(n-*)
t=-oo

*=n+l
n+N

= y()+ *(*)*(-*)
*=n+l

For a BIBO system, limn-.oolM n )l = 0- Therefore,

n+N

lim_oo 22

M * ) x ( n k) = 0 and

*=n+l

limn^ooyCn + AT) =

y(N).

(b) Let x(n) = x0(n) + au(n), where a is a constant and

x 0 (n) is a bounded signal with lim x0(n) = 0.
n-^oo

Then,

y{n) = of] M*)(n-*) + E *(*)*("*)

Jb=0

i=0

= a >(*) + .(*)
*=0

clearly,

x n
n

l( )

< =* E n 2fo(n) < ( f r o m ( c ) below) Hence,

limn^oolj/oCn)! = 0.
18

and, thus, lim n _ 0 0 y(n) = a^_Qh(k)

= constant.

(0
y(") = J>(*)(-*)
k

oo

E^w = >(*)*(-*)
- o o L fc

= X>(*W')2>(-*)*(-*)
k

But
x ( n - * ) * ( n - / ) < x 2 ( n ) = 2?,.
n

Therefore,

For a BIBO stable system,

|/>(*)|<M.
k

Hence,
Ey < M2EX, so that
Ey < 0 ii Ex < 0.

2.10
The system is nonlinear. This is evident from observation of the pairs
*3(n) - $te(n)

and

*2(n) *- !/2(n).

If the system were linear, yiin) would be of the form

y 2 (n) = {3,6,3}
because the system is time-in variant. However, this is not the case.

2.11
since
xi(n) + x2{n) = 6(n)
and the system is linear, the impulse response of the system is

Vi(n)+ !&(*)= JO, 3,-1,2, l j .

If the system were time invariant, the response to X3(n) would be

{3,2,1,3,1}.
But this is not the case.
19

2.12
(a) Any weighted linear combination of the signals x,(n), i = 1,2,..., iV.
(b) Any x,(n Jb), where k is any integer and i = 1,2,..., N.

2.13
A system is BIBO stable if and only if a bounded input produces a bounded output.

y(n) = J>(*)*(n-*)
k

\y(n)\ < 2>(*)ll*(-*)l

fc

< Jlf,J>(*)l
k

where \z(n k)\ < Mx. Therefore, |y(n)| < oo for all n, if and only if

5>(*)l<oo.
k

2.14
(a) A system is causal <= the output becomes nonzero after the input becomes non-zero. Hence,
x(n) = 0 for n < n0 => y(n) = 0 for n < n0.
i

(b)
n

y(n) = ^2 M * ) z ( n ~ *) w*iere * ( n ) = 0 for n < 0.

oo

If h(k) = 0 for ib < 0, then

n

y(n) = Y j /i(ib)x(n ib), and hence, y(n) = 0 for n < 0.

o
On the other hand, if y(n) = 0 for n < 0, then
n

^2 h(k)x(n - k) =>fc(ib)= 0, ib < 0.

oo

2.15
(a)
For a = 1, ] T an

AT-M + 1

N
flM +

flM+l

flJ*

flAf+1

fora^l, ] T a n
+

(l-a)>
20

M+l

N _ aN _

flJV+l

= 0,|a| < 1, and N - oo,

00

a | < 1.
n=0

y(n) = >(*)*(n-*)
k

I>w = E !>(*)*(*-*) = *(*)

n

fc

fc

x(n

n=-oo

y(n) = fc(n)* x(n) = {1,3,7,7,7,6,4}

y(n) = 35,
n

J>(*) = 5.

*(*) = ?

y(n) = {1,4,2,-4,1}

y(n) = 4,

>(*) = 2,

*(*) = 2
k

y() = {o,|,-i |,-2,o,-|,- 2 j

(n) = -5,

*() = 2.5,

>(n) = -2
n

y(n) = {1,2,3,4,5}

y(n) = 15,

h(n) = 1,

x(n) = 15
n

y(n) = {0,0,1,-1,2,2,1,3}

>() = 8.
n

5>0) = 4,
n

>(n) = 2

y(n) = {0,0,1,-1,2,2,1,3}
E v ( n ) = 8,

5 > ( n ) = 2,

J>(n) = 4

y(n) = { 0 , 1 , 4 , - 4 , - 5 , - 1 , 3 }
21

 > ( n ) = -2,
n

fc(n) = - l ,

^x(n) = 2

y(n) = u{n) + u(n - 1) + 2u(n - 2)

y(n) = oo,
n

E h(n) = oo,

E x(n) = 4

y(n) = {1,-1,-5,2,3,-5,1,4}
y ( n ) = 0,
n

*h(n) = 0,
n

^x(n) = 4
n

y(n) = {1,4,4,4,10,4,4,4,1}

y ( n ) = 36,

^li(n) = 6,

Ex(n) =6

y(") = [2(i) n -(i)>(n)

E^) = !> E*(n) = r, E^) = 2

x(n)

t1'1'1}

A(n)

J6,5,4,3,2,lj
n

y(n)

=
i=0

y(o) =

x(o)h(o) = e, ,

y(i)
y(2)

x(0)/i(l) + x(l)/i(0) = l l
x(0)/j(2) + x(l)/(l) + x(2)h(0) = 15

y(3)
y(4)

=
=

y(5)

x(0)/*(4) + x(l)h(Z) + x(2)/(2) + x(3)/i(l) + x(4)/i(0) = 14

x(0)/i(5) + x(l)h(4) + x(2)A(3) + x{Z)h(2) + x(4)h(l) + x(5)/i(0) =

y(6)
y(7)

=
=

x(l)/(5) + x(2)/i(4) + x(Z)h{2) = 6

x(2)/i(5) + x(Z)h(4) = 3

y(8)
y(n)

=
=

x(3)/(5) = 1
0,n>9

y(n)

|6,11,15,18,14,10,6,3, l |

x(0)/i(3) + x(l)h(2) + x(2)/(l) + x(Z)h(0) = 18

22

(b) By following the same procedure as in (a), we obtain

y(n) = |6,11,15,18,14,10,6,3, l |
(c) By following the same procedure as in (a), we obtain
y(n) = j l , 2 , 2 , 2 , l |
(d) By following the same procedure as in (a), we obtain

y(n) = jl,2,2,2,lj
2.18
(a)
f

1 2

4 5

h(n)

jl,l,l,l,ll

y(n)

x(n)*h(n)
(I
10

n\

20

11

n\

(b)
*( n )

/i(n) =
y(n) =

3[()-w(-7)],
u(n + 2) - u(n - 3)
x(n).*/i(n)

-ri[ti(n) - u(n - 7)] * [ti(n + 2) - ti(n - 3)]

-n[u(n) * u(n + 2) - u(n) * u(n - 3) - u(n - 7) * u(n + 2) + u(n - 7) * u(n - 3)]

y(n) =

\6{n + 1) + <5(n) + 2S(n - 1) + ~ ^ ( n - 2) 4- b6(n - 3) + ^ 6 ( n - 4) + 66(n - 5)

3
o
o

+56{n - 6) + 56(n - 6) + ^6{n - 7) + $(n - 8)

2.19
y(n) =

>(*)*(n-*),

*(n)

j Or"3, Of- 2 , Or" 1 , 1, Or, . . ., Or 5 1

h(n)

j l , 1,1,1, l |
23

y() =

X>(n~*),-3<n<9

0, otherwise.

lK-2'

y(-i
y(o
y(i
y(2
y(3
y(4
y(5
y(6
y(7
y(8
y(9

=
=

Xr(-3) + x ( - 2 ) =
- 33 _ - 22 , _ _- 1l

Therefore,

lK-3"
a"3 + a-2,

a~ + a~ + a ,
a"3 + a"2 + a ~ 1 + 1
- 3 + a-3

~ 2 + a~ 1 + I + 0 + Q3

a"1 + l + a + a 2 + a^33
a 4 + a 3 + a 2: +, a_ +, _1
a + a2 + a 3 + a4 + a5,
a2 + a3 + a4 + a5
<3 + a 4 + a 5 ,
r4 + a5s
=

Q5

2.20
(a) 131 x 122 = 15982
( b ) { l T , 3 , l } * { l T , 2 , 2 } = {1,5,9,8,2}
(c) (1 + 3z + z 2 )(l + 22 + 2z 2 ) = 1 + 52 + 92 2 + 8z 3 + 22 4
(d) 1.31 x 12.2 = 15.982.
(e) These are different ways to perform convolution.

2.21
(a)
y(n) = ^2aku(k)bn-ku(n

- k) = bn ^ ( a t " 1 ) *

k=0

k-0

6"(n+l)u(n),

a =

(b)
*(n)

{l,2,l,l}

A(n)

{l,-l,0,0,l,ll

y(")

{1,1,-1,0,0,3,3,2,1}
24

(0
x(n) =

|l,l,l, l,l,0,~l|,

h(n) = {l,2,3,2,l}
y{n) =

|l,3,6,8,9,8,5,1,-2,-2,-lj

(d)
x(n)

h'{n) =
h(n)
y(n)
y'(n)

/l.l.l.l.l},
{o,0,l,l,l,l,l,l}

= h'(n) + h'(n - 9),

- y/(n) + y / ( n - 9 ) , where
=

{M,l,2,3,4,5,5,4,3,2,l}

2.22
(a)
y.(")
yi(n)

Sft(n)
ya(n)
y4(n)
ys(n)

x(n)*hi(n)
x(n) + x ( n - 1)
{1,5,6,5,8,8,6,7,9,12,12,15,9}, similarly
{1,6,11,11,13,16,14,13,15,21,25,28,24,9}
{0.5,2.5,3,2.5,4,4,3,3.5,4.5,6,6,7.5,4.5}
{0.25,1.5,2.75,2.75,3.25,4,3.5,3.25,3.75,5.25,6.25,7,6,2.25}
{0.25,0.5,-1.25,0.75,0.25, -1,0.5,0.25,0,0.25, -0.75,1,-3, -2.25}

(b)
y3(w) =

^yi(n), because

M")

-^i(n)

y4(n) =

-.^(n), because

/i4(n) =

-/2(n)

(c) y2(") and y4(n) are smoother than yi(n), but y4(n) will appear even smoother because of the
smaller scale factor.
(d) System 4 results in a smoother output. The negative value of As(0) is responsible for the
non-smooth characteristics of ys(n)
(e)
, , fl 3
1
, . 1 1
13
9\
( n ) =
1,
, , ,1,
,
*
\2'2'
'2'
' 2 2 ~2 2'~2J
y2(n) is smoother than y6(n).
25

2.23
if
yi(n) =
yi(n) =
x(n) =

nyi(n - 1) + Xi(n) and

n y 2 ( n - l ) + x 2 (n) then
axi(n)+6x 2 (n)

produces the output

y(n) = ny(n - 1) + x(n), where
y(w) = ayi(n) + 6y2(n).
Hence, the system is linear. If the input is x(n 1), we have
y ( n - l ) = ( n - l ) y ( n - 2 ) + x ( n - l ) . But
y(n - 1) = ny(n - 2) + x(n - 1).
Hence, the system is time variant. If x(n) = u(n), then |x(n)| < 1. But for this bounded input,
the output is
y(0) = 1,
y(l) = 1 + 1 = 2,
y(2) = 2x2 + 1 = 5 , . . .
which is unbounded. Hence, the system is unstable.

2.24
6(n) =
6(n - k) =
x(n)

=
=

7(n) af{n 1) and,

7 (n - *) - 07(71 - ib - 1). Then,
)

z(*)*(n-*)

*(*)[7(n-*)-a7(n-fc-l)]

Jk=-oo

x(n)

f;

*(*)7(n-*)-af; x(*)7(n-*-l)

i = 00

i = 00

00

x(n)

00

x(*)7(n-*)-a

*(-l)T(n-)

k=-oo

Jfc=-oo
00

E
Thus, Ci

[*(*)-a*(*-l)]7(n-*)

= x(Jfc) - ax(ib - 1)

(b)
y(n)

T[x(n))

= T[f;

c 4 7 (n-*)]

t=-oo

f;

ckT[f(n-k))

k=-oc
oo

= 51 c*0(n"*)
26

(c)
h(n)

T[6(n)]

T[y(n) - ay(n - 1)]

9(n) - ag{n - 1)

2.25
With x(n) = 0, we have

y{*~i)

+ \y{n-i)

y(-i)

-|y(-2)

y(0)

(-|)2y(-2)

y(i)

(-|)3K-2)

y(it)

(-) i+2 y(2) < zero-input response.

o

2.26
Consider the homogeneous equation:
y(n) - gV(n - 1) + -y(n - 2) = 0.
The characteristic equation is
* ~ ^ A + 7 = 0.A = - , - .
6
6
2 3
Hence,

(n) = c1(i) + c2(i)

The particular solution to
x(n) = 2 n u(n) is
lfr(n) = * ( 2 > ( n ) .
Substitute this solution into the difference equation. Then, we obtain
A(2")ti(n) - k^){Vl)u(n

- 1) + *(i)(2"- 2 )ti(n - 2) = 2"u(n)

For n = 2,
4

5k

*-T

, 8

6=4^*=5-

Therefore, the total solution is

y(n) = yp(n) +

yjk (n)

= f ( 2 > ( n ) + C!(i)"u(n) +

To determine ci and C2, assume that y(2) = y(1) = 0. Then,

y(0) = 1 and
27

Ci()nu(n).

y(i) = gy(o) + 2 = ~
Thus,
1 => C l + C 2 = - "

- + ci + c2
16 1
T+2C

l +

1
3C2

and, therefore,
Cx = - l , C

-.

The total solution is

y(*) =

|(2)" - (|r + | ( | r u(n)

2.27
y(n) - 3y(n - 1) - 4y(n - 2) = x(n) + 2x(n - 1)
The characteristic equation is
A2 - 3A - 4 = 0.
Hence, A = 4, 1 and
yh(n) = c 1 ( n ) 4 n + c 2 ( - l ) n .
Since 4 is a characteristic root and the excitation is
x(n) = 4 n (n),
we assume a particular solution of the form
Mnu(n).

yp(n) =
Then

Jfcn4nu(n) - 3k(n - l)4 n _ 1 ti(n - 1) - 4ib(n - 2 ) 4 n _ 2 u ( " ~ 2)

= 4 n u(n) + 2(4) n " 1 t*(n - 1)
For n = 2,
*(32 - 12) = 4 2 + 8 = 24 k = -.
The total solution is
y(n)

yp(n) + yh(n)
-n4n + c14f,+C2(-l)'
o

u(n)

To solve for c\ and c2, we assume that y(1) = y(-2) = 0. Then,

y(0) = 1 and
y(l) = 3y(0) + 4 + 2 = 9
Hence,
d -+- c2 = 1 and
28

24 A
-r- + 4ci - c2 = 9
2 1

4ci - c 2 = -
Therefore,
Cl

26
.
= -andc2 =

1
~

The total solution is

u(n)

y() =

2.28
From 2.27, the characteristic values are A = 4, 1. Hence
yh(n) = c 1 4 n + c 2 ( - l ) n
When x(n) = 6(7V), we find that
y(0) = 1 and
y(l) - 3y(0) = 2 or
y(i) = 5.
Hence,
c\ + c 2 = 1 and 4ci c 2 = 5
This yields, ci = | and c2 = \. Therefore,

I4" " it"1)" ()

fc(n) =

2.29
(a) I i = N\ + Mi and L2 = N7 + M 2
(b) Partial overlap from left:
low # ! + Mi

high JVx + M 2 - 1

Full overlap: low N\ + M 2

high 7V2 + Mi

Partial overlap from right:

low iV2 + Mi + 1

high N2 + M 2

(c)

x(n)

|l,l,l,l,l,l,l|

h(n)

{2,2,2,2}

JVi

"2,

^2

4,

Mi

-1,

M2

2,

Partial overlap from left: n = 3

Full overlap: n = 0
Partial overlap from right:n = 4
29

n = 1

Li

n= 3
n= 6

I2

2.30
(a)
y(n) - 0.6y(n - 1) + 0.08y(n - 2) = x(n).
The characteristic equation is
y - 0.6A + 0.08 = 0.
A = 0.2,0.4 Hence,

ln

2n

With x(n) = <5(n), the initial conditions are

y(o) =

i,

y(l)-0.6y(0)
Hence, ci + c 2

=
=

0 =} y(l) = 0.6.
1 and

-ci + -

0.6 =>C! = - l , c 2 = 3.

Therefore /i(n)

-(l)"+2(|r

()

The step response is

n

5(n)

^/(n-*),n> 0
i=0

{"i[r-fo[<r'-]}>
(b)
y(n) - 0.7y(n - 1) + 0.1y(n - 2) = 2x(n) - x(n - 2).
The characteristic equation is
A ' - 0 . 7 A + 0.1 = 0.
A = ~, | Hence,

ln
yh(n) = c1-

ln
+ c2- .

With x(n) = 6(n), we have

V(0)
y(l)-0.7y(0)
Hence,ci + c 2
1

=
=
=

!
Cl +

2,
0 = y(l) = 1.4.
2 and
7

, .
=

M =

14

These equations yield

10
4
Cl =

T' C 2 = "3h(n)

=
30

3V

3V

(n)

The step response is

s(n) =

h(n-k),
fc=o

i=0

=
=

t=o

Tib-^-Ub"^
*=o
i=0

10 1 "

(2 +1 1)u(n)

y<5 " "

1 1

"

{ (5n+1 1)u(n)

zi

"

2.31
H{n)

f 1 ! 1 l \
\}'2'i'8'l6/

y(n) =

{l,2,2.5,3,3,3,2,l,o}

x(0)/i(0) =
ix(0) + x(l)

y(0) => x(0) = 1

y(l)=>x(l)=^

By continuing this process, we obtain

()={l.|,|.J.|.- j
2.32
(a) h{n) = /n(n) * [h2(n) - h3{n) * h4(n))
(b)
( n - l)u(n-2)
2u(n) - 6(n)

/i3(n) * h4{n)
h2(n) - h3{n) * /i4(n)

I6(n)

I6(-l)+l6(n-.2)

Hence /i(n) =

]-6(n) + i$(n - 1) + ^6(n - 2)1 * [2u(n

i6(n) + ^ ( n - l ) + 26(n-2) + 5 u ( n -

(0
x(n)
y(n)

{1,0,0,3,0,-4}
f1 5

25 13

31

c n rt n

2.33
First, we determine
s(n)

u(n) * h(n)
00

i=0
oo

= .~*
=

r-,n>0
a 1

For x(n) = u(n + 5) u(n 10), we have the response

s(n + 5) - s(n - 10) =

an+6 - 1
-u(n + 5)
a1

an_9 - 1
-u(n - 10)
a 1

From figure P2.33,

y(n)

Hence, y(n)

x(n) * h(n) - x{n) * h(n - 2)

an+6 ~ 1 /
^
an~9 - 1u n 1 0
7u(" + 5)
r~ ( )
a n
1
a

1
a +4 - 1
a"-11 - 1
Tu(n + 3) +
u ( n - 12)y
a1
a- 1

2.34
fc(n) =

[u(n)-u(n-M))/M
00

*(n) = #(""*)
*=-oo

A/

2.35

W")l =
n = 00

|oP

n=0,neven

n=0

"
Stable if Id < 1
32

1-lap

2.36
h(n) anu(n). The response to u(n) is
00

yi(n) = (*)*(*-*)
fc=o

i=0

Then, y(n)

y^n) - yi(n - 10)

-i-[(l-a"+1)u(n)-(l-a^9)
1a

2.37
We may use the result in problem 2.36 with a = \. Thus,
y(") = 2

-(3)"+l

u(n) - 2

i-(5)- 9

2.38
(a)

y(n) =

*(*)*(*-*)

k= -oo

= E(5) l 2-'
1=0 ^

= 2"[i-(^r +1 ](|)
= f[2n+1-(i)n+1](n
(b)

y() = *(*)*(-*)
*=-oo

= >
fc=0

= f)(5)* = 2.<0
fc=0

33

u(n

y(n) =

JTh{k)
00

n-1

00

= (?>*-<3>*
2

2(i)-,n>0.

2.39
(a)
h e (n)

=
=
=

ni(n)*n2(n)*n3(n)
[6(n) - 6(n - 1)] * ti(n) * n(n)
(u(n)-u(n-l)]*n(n)

$(n)*/(n)

*(n)

(b) No.

2.40
(a) x(n)6(n no) = x(no). Thus, only the value of x(n) at n = no is of interest.
x(n) * 6(n no) = x(n no). Thus, we obtain the shifted version of the sequence x(n).
(b)
y(n)

=
=

Linearity :xi(n) yi(n)

*2(n)-lfe(n)
Then x(n)
y(n)

=
=
=
=
=

f;

n(*)x(n-*)

h(n) * x(n)
n(n)*xi(n)
M n ) * x 2 (n)
a x i ( n ) + /?x 2 (n) -* y(n) = n(n) * x(n)
n ( n ) * [ a x 1 ( n ) + /?x 2 (n)]
ah(n) * xx(n) + ^n(n) * x 2 (n)
<*yi(n) + /?y2(n)

Time Invariance:
x(n) y(n) = n(n) * x(n)
x(n-n0)-*yi(n) =
n(n)*x(n-n0)
=
=

/(*)x(n-n0-*)
y(n - n 0 )

(c) h(n) = 6(n - n 0 ).

34

2.41
(a) s(n) = -ais(n - 1) - a?s(n - 2) - ... - aNs(n - N) + b0v(n). Refer to fig 2.2.
(b) v(n) = fb3 [s(n) + ais{n - 1) + a2s(n - 2) + ... + aNs(n - N)]. Refer to fig 2.3
v(n)

s(n)

Q>

*1

-a 2

Figure 2.2:

2.42
y() =

- j v ( n - l ) + x(n) + 2 x ( n - 2 )

!K-2)

- y ( - 3 ) + x(-2) + 2 x ( - 4 ) = l

y(-l)

-iy(-2)+x(-l) + 2x(-3)=|

y(0) =

- j y ( - i ) + 2*(-2) + x(0) = H

y(i) = -iy(o) + *(i) + 2x(-i) = j . e t c

2.43
(a) Refer to fig 2.4
(b) Refer to fig 2.5

35

s(n)
1/bb

< >
z-1

< >

Y
,-i

*
Figure 2.3:

2.44
(a)
x(n)

y() =
y(0) =

i 1,0,0,...!
2^n_1)
x(0)=l,

+ z n

( )+

(n-1)

y(i) = ^y(o) + x(i) + x(o) = ?

y(2)
n

y{n)

1
3
-y(l) + x(2) + x(l) = T . Thus, we obtain
I
4

(. 3 3 n

"" 1 ' 2 ' 4 ' 8 , 1 6 , 3 2 , " J

(b) y(n) = \y{n - 1) + x(n) + x(n - 1)

(c) As in part(a), we obtain
,
f
,(B, =

5 13 29 61
1

i '2 77 18 -j

(d)
y(n) =

\
,

u(n) * A(n)

= (*)*(n-*)
k

36

=Q

1/2

-1/2

Q>

y(l

*.

3/2

Figure 2.4:

= X>("-*)
y(0) =

M0)=1

y(l)

M0) + M 1 ) = |

y(2) =

13
h(0) + Ml) + M 2 ) = j , etc

(e) from part(a), h(n) = 0 for n < 0 => the system is causal.
00

3
1 1
]T |/i(n)| = l + r ( l + r + - + --.) = 4 = J ' system is stable
n=0

2.45
(a)
V(n)
=>&(n)

= ay(n - 1) + 6z(n)
n
= 6o u(n)

f>(n) =
n=0

=6

1-a.
37

x(n)

y(n)

"0

-1

,-1

,-1

Figure 2.5:

(n)

J>(n-*)
k-0

*(oo)

1 - a"*1
1-a

1-a
1-a.

=>6

( n )

= 1

a in both cases.

V(n)
y(n)-0.8y(n-l)

=
=

0.8y(n - 1) + 2x(n) + 3x(n -- 1 )

2x(n) + 3x(n - 1)

The characteristic equation is

A-0.8

A =
Vh(n)

38

0.8.
c(0.8) n

Let us first consider the response of the sytem

y{n) - 0.8y(n - 1) = x(n)
to x(n) = S(n). Since y(0) = 1, it folows that c = 1. Then, the impulse response of the original
system is
h(n)

=
=

2(0.8) n u(n) + 3 ( 0 . 8 ) n - 1 t i ( n - l )
26(n) + 4.6(0.8) n " 1 u(n - 1)

(b) The inverse system is characterized by the difference equation

x(n) = -1.5x(n - 1) + -y(n) - 0Ay(n - 1)
Refer to fig 2.6

0.5

-0.4

-1.5
Figure 2.6:

2.47
y(n) - 0.9y(n - 1) + x(n) + 2x(n - 1) + 3x(n - 2)
(a)For x(n) = <5(n), we have

y(o) =
y(i)
y(2)
y(3)
y(4)
y(5)

=
=
=
=
=

i,
2.9,
5.61,
5.049,
4.544,
4.090,
39

(b)
(0)

y(0)=l,

*0)
5(2)

=
=

y(0) + y(l) = 3.91

y(0) + y(l) + y(2) = 9.51

*(3)

y(0) + y ( l ) + y ( 2 ) + y(3)= 14.56

*(4)

y ( n ) = 19.10
o
5

*(5)

v ( n ) = 23.19

(c)

A(n)

(0.9) n ti(n) 4- 2(0.9) n ~ 1 ti(n - 1) + 3(0.9) n " 2 ii(n - 2)

6(n) + 2.96(n - 1) + 5.61(0.9) n - 2 u(n - 2)

2.48
(a)
y(n)

= -*(n)+-x(n-3) + y(n-l)

for x(n) = 6(n), we have

., * _ / 1 1 1 2 2 2 2
\
l ;
"
"
\3'3,3,3,3,3,3,"7
(b)
y(")
with x(n)
y(-l)

=
=
=

2y(n~1)+8y(n"2
6(n), and
y ( - 2 ) = 0, we obtain

) +

(n)

\ ' ' 2 ' 4 ' 1 6 " 8'128'256'1024' j

X ( n

2 )

(c)

y(n) - 1.4y(n-l)-0.48y(n-2) + x(n)

with x(n) = 6(n), and
y ( - l ) = y ( - 2 ) = 0, we obtain
h(n) = {1,1.4,1.48,1.4,1.2496,1.0774,0.9086,...}
(d) All three systems are IIR.
(e)
y(n)
The characteristic equation is
A 2 - 1 . 4 A + 0.48

1 . 4 y ( n - l ) - 0 . 4 8 y ( n - 2 ) + x(n)

0 Hence

A =
yh(n)
=

0.8,0.6. and
n
n
C l (0.8) + c 2 (0.6) For x(n) = <5(n). We have,
40

c\ + c2
0.8ci + 0.6c2
=> ci
c2
h(n)

1 and

= 1.4
= 4,
= 3. Therefore
= [4(0.8) n -3(0.6) n ]u(n)

2.49
(a)
/i(n)
h2(n)
/i 3 (n)

= c06{n) + c^in - 1) + c 2 6(n - 2)

= 6 2 ^(n) + 6 i 6 ( n - l ) + 6 0 6 ( n - 2 )
= a 0 6 ( n ) + ( a i + a 0 a 2 ) 6 ( n - 1 ) + aia 2 6(n - 2)

(b) The only question is whether

h3(n)

=
=

Let ao

Co,

= ci,

ai + 02^0

a 2 ai

+ a 2 c 0 c\
a2
=> c 0 a 2 - c i2 + c2

h2(n) = hi(n)

c 2 . Hence

= 0
= 0

For Co ^ 0, the quadratic has a real solution if and only if

c\ 4CQC2 > 0

2.50
(a)
y()

For y(n) -y(n 1)

2

- y ( n - l ) + x(n) + x ( n - l )
= 6(n), the solution is
s.

M) = (jNnJ + ijrMn-l)
(b) M " ) * [6(n) + 6(n - 1)] = (i) n ti(n) + ( ^ " " ^ ( n - 1).

2.51
(a)
convolution: yi(n) . = < 1,3,7,7,7,6,4 >
correlation: 7i(n)

=
41

< 1, 3,7, 7,7, 6,4 >

(b)

9
convolution: y2(n)

j - , 0, - , - 2 , - , - 6 , - - , - 2 I

| > 0, - , - 2 , - , - 6 , - - , - 2 j

correlation: 71 (")

Note that j/2(n) = 72(n), because /2(-n) = h2(n) (c)

convolution: ^3(n)

< 4,11,20,30,20,11,4 >

correlation: 7J(n) =

< 1,4,10, 20,25,24,16 >

(c)

convolution: y 4 (n)

< 1,4,10,20,25,24,16 \

correlation: 7 4 (n)

I 4,11,20,30,20,11,4 >

Note that h3(-n)

hence, 7 3 (n)
and/i4(-n)
=>74(n)

=
=
=

h4(n +
y 4 (n +
^3(n +
ys(n +

3),
3)
3),
3)

2.52
Obviously, the length of h(n) is 2, i.e.
h(n)

{/i 0) /,}

h0

3h0 + /i

=> /o

l,/i = 1

2.53
N

(2.5.6)

y{n)

- a * y ( n - *) +
*=1
N

(2.5.9)

w(n)

6*x(n - *)

4=0

- ^ a t u ; ( n - * ) + x(n)
=i
M

(2.5.10)

y(n)

^6fcu;(n-*)

From (2.5.9) we obtain

JV

x(n) = u;(n) + y^afct^(n - it)

(A)

fc=i

By substituting (2.5.10) for y{n) and (A) into (2.5.6), we obtain L.H.S = R.H.S.
42

2.54

-*.

y(n)-4y(n-l) + 4y(n-2)

x(n) - x(n - 1)

The characteristic equation is

A2-4A + 4

A =
Vh(n)

2,2. Hence,
ci2n+c2n2n

The particular solution is

Ifr(n) =

fc(-l)"u(n).

Substituting this solution into the difference equation, we obtain

* ( - l ) n t i ( n ) - 4ib(-l) n " 1 u(n - 1) + 4k(-l)n-7u(n

- 2) = ( - l ) n u ( n ) - ( - ^ " " ^ ( n

For n = 2, Jfc(l + 4 + 4) = 2 = ib = | . The total solution is

Cl2

y(") =

+ c 2 n 2 n + - ( - l ) n u(n)

From the initial condtions, we obtain y(0) = l , y ( l ) = 2. Then,

c1 +

CI

2cj + 2c2 - C2

= 5'
=

3'

2.55
From problem 2.54,
h(n)= [ Cl 2 n + c 2 n2 n ]ti(n)
With y(0) = l,y(l) = 3, we have
Cl

2c! + 2c 2
^

C J

Thus h{n)

=
=
=
=

1
3
1
2
2" + - n 2 n u(n)
2

2.56
x(n)

=
=
=

z(n)*6(n)
x(n) * [ u ( n ) - u ( n - 1)]
[x(n)-x(n-l)]*ti(n)
oo

(z(*)-*(*-l)]u(n-*)
43

2.57
Let h(n) be the impulse response of the system
k

s(k) =

h(m)
m= oo

=>h(k)

S{k)-S(k-1)

oo
* = -oo
oo

[,(*)-,(*-l)]*(n-*)

* = -oo

2.58
w n \ _ / *> n0-N<n<n0
^
\ 0, otherwise
_ f 1,
- | o,

y\n)

+N

-N <n< N
otherwise

oo

7xx(/)=

53

*(n)x(n-l)

n = oo
i

The range of non-zero values of 7XX(/) is determined by

n0-N<n<n0

+ N

n0-N<n-l<n0

+ N

which implies
-2N
<1<2N
For a given shift /, the number of terms in the summation for which both x(n) and x{n - I) are
non-zero is 27V + 1 |/|, and the value of each term is 1. Hence,
7(/)=|

-|/|,
0 )

-2N<l<2N
otherwise

For 7ry(0 we have

m = |/
7*y(0

2 N + 1

-|/-n0|,

Q>

n0-2N
< l< n0 + 2N
otherwise

2.59
(a)
7(/)

7rx("3)

7r*(-2)

53

x(n)x(n-/)

n = -oo
x(0)x(3)=l

x(0)x(2) + x(l)x(3) = 3
44

7(-l)

x(0)x(l) + x(l)x(2) + x(2)x(3) = 5

7xx(0)

x2(n) = 7

n=0
AlS0Trx(-/)

7xx(0

Therefore 7xx(0

{ 1,3,5,7,5,3, l |

(b)

Tw(0

V()lKn-0

n = oo

We obtain
7yy(0

{1,3,5,7,5,3,1}

we observe that y(n) = x(n + 3), which is equivalent to reversing the sequence x(n). This has
not changed the autocorrelation sequence.

2.60
7xx(0

x(n)x(n-/)

n = oo

-- {

7xx(0)

2JV + 1 - | / | ,
0,
2N+1

-2N<l<2N
otherwise

Therefore, the normalized autocorrelation is

Pxx(0

^~[(2N

0, otherwise

l-\l\),-2N<l<2N

2.61
(a)
00

7(/)

Y,

I B

( )'( B -')

n= -oo
oo

5Z [5(n) + T i s ( n ~ *i) + 72s(n-*2)]*

n = oo

[s(n - /) +

7l5(n 7 2 2 )7(0

/ - *i) + 72*(n - / - * 2 )]

(1 + 7? +

+ 71 [7*.C + *l) + 7(/ - *i)l

+ 7 2 [ 7 ( ' + *2) + 7 . . ( ' - * 2 ) ]

+7172 [7C + *i - *2) + 7..C + *2 - *i)]
(b) 7xx(0 has peaks at / = 0,Jki,ifc2 and (ki + Jfc2). Suppose that ki < Jt2. Then, we can
determine 71 and k\. The problem is to determine 7 2 and fc2 from the other peaks.
(c) If ->2 = 0, the peaks occur at / = 0 and / = kx. Then, it is easy to obtain 71 and k\.
45

2.62
(a)
(b) variance = 0.01. Refer to fig 2.7.
(b) Delay D = 20. Refer to fig 2.7.

0.5

-0.5
50

100

150

200

200

Figure 2.7: variance = 0.01

(c) variance = 0.1. Delay D = 20. Refer to fig 2.8.
(d) Variance = 1. delay D = 20. Refer to fig 2.9.

-20

20

Figure 2.8: variance = 0.1

(e) x(n) = { 1,1,l,-hl,H-l,-hl,-hl, 1,-H1, l T H-l,-hl,1,1,-hl}. Refer to fig 2.10.
(f) Refer to fig 2.11.
46

1
0.5

-0.5
-1

SO

100
> n

150

200

200

Figure 2.9: variance = 1

2.63
(a) Refer to fig 2.12.
(b) Refer to fig 2.13.
(c) Refer to fig 2.14.
(d) The step responses in fig 2.13 and fig 2.14 are similar except for the steady state value after
n=20.

2.64
Refer to fig 2.15'.

47

200

200

20
15
10

f
E

-5

-10

/VV V
-20

0
> n

20

Figure 2.10

1.5

0.5

~ 0.5
c
!

-0.5

-0 5
-1

50

100

150

-1.5

200

> n

11 IB

nft

50

100

>n

Figure 2.11:

48

150

200

impulse response h(n) of the system

- 0 5.

Figure 2.12:

zero-cute eiep reapers* (n)

1.5
1.4

"

1.3

^1-2
A

1 1.1

09
0.8

A
1 / \ r\ w _
1 / \J

'II

10

19

20

25

Figure 2.13:

49

30

39

49

50

Figure 2.14:

Figure 2.15:

50

Chapter 3

3.1
(a)

x(z) = Y,x(n)z~n
n

3z 5 + 6 + z _ 1 - 4 z - 2 ROC: 0 < \z\ < oo

(b)

x(z) = Y.x^2~n
n

= Bj)"*""
n=5

n=5
m=0

=
E(^-'r + 5
= (-=-)
2 '

1-iz-i

1
z""5
1
(V ) ;
i 1 rROC:|z|>3 2 l - kz' ' 2

3.2

x(z) = Y,xWz~n
n

= f)(l + n)*-"
n=0
OO

OO

= E*- + En*~"
n=0

n=0

OO

But Vz- f l
n^:
=0

1 - z_1
51

ROC: \z\ > 1

and ) nz-

'"'

ROC: |z| > 1

n=0

Therefore, X(z)

(1 - z " )
1

,-\
(1-r-1)2

2 +

(l-*"1)2
(b)
X(z)

^ K - K O Z "
oo

oo

^anz-

+ ^a-nz"

n=0

But a n * n

ad 5I ~ *~

n= 0

1
ROC: Izl > \a\
-az-1

oo

a n

B0C
=W>W
(i-J,..p

1
1-az"1

Hence, X{z)

1
1-^z"1

2-(a+-Mz"1
R

(1--)(!-1,-.)

C : ''I >

~ l

(c)

*M = f>5)"*-"
n=0

= rnr^ > w > 2

(d)

*(*)

Y J na n smu;onz"
n= 0

= I>"

gju'o" _

e-;uion

2j

n= 0

at->wz-x
(1-ac-J^oz-1)2

jtuor-l

2j

ae'-"z
(l-aeJvoz-1)2

[ a z - 1 - ( a z - 1 ) 3 ] si'nwo
(1 - 2QCOSWQZ-1 +

(e)

X(z)

y j na n cou;on^

n=0

= 2>"
n=0

2
52

a2z~2)2

,\z\>a

-jw -l

JVOf-l

ae
ae 0r
1 2 +
(l-ac^02~ )
(l-ae~>u'02-1)2[az-1 + (az'1)3] sinwp - 2a 2 2~ 2
* > a
(l-2acosu;o2-1 + a22-2)2
'
(f)
X(2)

A^rncos(woTi

+ <t>)z-n

n=0
oo

= ^Er"
n=0

e>*

,-j*

l_rc;u;o2-l

l_rc-Jwor-l

cos^ rcos(wo <^)2_1 "

4
1 2rcoswoz~i + r 2 2 - 2

2 > r

(g)

n=l

E n r-r

-1

2- n

, - 1 _i_

= -

1,-2

i5
,-1

Therefore, X(z)

2"

L(1_I2-1)2

+i2-

(l_i2-l)3j

.-1

W>5

(i-iz->)3'
(h)
1

1
n = 10

n=0

(i)102-10

1
l_l*-i
i , - l1\ l O
i-a*)

1-12-

i _ i z - i

1*1 > 2

The pole-zero patterns are as follows:

(a) Double pole at z = 1 and a zero at z = 0.
(b) Poles at 2 = a and z = . Zeros at 2 = 0 and 2 = j ( a + ).
(c) Pole at 2 = - | and zero at 2 = 0.
(d) Double poles at 2 = ae^w and 2 = ae~jw and zeros at z = 0, 2 = a.
(e) Double poles at 2 = aeJU, and 2 = ae~JU, and zeros are obtained by solving the quadratic
acoswoz2 2a 2 2 + a3coswo = 0.
(f) Poles at 2 = re3W and 2 = ae~iw and zeros at 2 = 0, and 2 = rcos(u;o 4>)/cos<j>.
(g) Triple pole at 2 = 5 and zeros at 2 = 0 and 2 = 3- Hence there is a pole-zero cancellation so
53

that in reality there is only a double pole at z i and a zero at z = 0.

(h) X(z) has a pole of order 9 at z = 0. For nine zeros which we find from the roots of

i-(^-')10

= o

or, equivalent^, ( - ) 1 0 - z 10

Hence, zn

1 i2z
-e > ,n = 1,2,.. .,*.

Note the pole-zero cancellation at z = | .

3.3
(a)

n=0

n = oo

= Trbr +n=0
Ds)"'- 1
3

3'
5
6_

(i-iz-i)d-i*)
The ROC is i < \z\ < 2.
(b)

n n
n n
f^(b
z-Y,2
*3

**(*) =
n=0

n=0

1
l-2z~1'

l_Iz-i
_52-l

(l-ir-i)(l-2z-i)
The ROC is \z\ > 2.

(0
Xz{z)

x 1 (n + 4)z" n

n= oo

zAXi{z)
(l-lz-i)(l-lz)

The ROC is J < |z| < 2.

(d)
X 4 (z)

xjC-n)*-"

54

]T *i(m)m = oo

(1

_IZ)(1 _!,-!)

The ROC is I < \z\ < 3.

3.4
(a)
X(z)

^n(-l)"z-
n= 0

= -*-i>"*n=0

1
1+z"1

dz

-l

> 1

(1 + z - 1 ) 2 '
(b)

x(z) = 5>2*-
n=0
=

d2
dz*

T
r=0

1
1-z-1

dz2

-1
1

(1-z- )

2r - l
(1-z-1)3

z - 1 ( l + ^" 1 )
( 1 - Z - 1 ) 3 ,W>i
(c)
-1

-na"

X(z) =

-l

= -* E <).n = oo

1-az"1

dz
-l

az
( 1 - a z -rnj.W<
i)

(d)
oo

X(z)

(-l)nco5%r-B
n=0

55

From formula (9) in table 3.3 with a = - 1 ,

*<,) = -

A+J^i

l + 2z- 1 cos + z - 2

1 + iz"1
2
l + 2 - i + z " 2 -, ROC: Izl > 1
(e)

n=0

1
1 + 2"1'

UI>1

(0
x(n)

ji,0,-1,0,1,-lj

X(z)

l-r-2 + r-4-2-5,

z^O

3.5
Right-sided sequence :x r (n)

0,n<no
-l

Xr(z)

*,()*-"+X>r(n)z*
r=0

The term J2n=n xr{n)z~n converges for all z except z = oo.

The term In=o xr{n)z~n converges for all \z\ > ro where some ro. Hence Xr(z) converges for
r 0 < \z\ < oo when no < 0 and \z\ > ro for no > 0
Left-sided sequence :xj(n)

Xi{z)

0, n > no
JT

x,(n)z-" + J T x , ( n ) r - "

n= -oo

n=l

The first term converges for some \z\ < r/. The second term converges for all z, except z = 0.
Hence, Xi(z) converges for 0 < \z\ < r/ when no > 0, and for \z\ < r/ when no < 0.
Finite-Duration Two-sided sequence :x(n)

0,n > n0andn < n t , where no > ri\

ro

X(z)

x(n)z-n

nsnj

x(n)r~" + j r \ ( n ) * - "
n=0

The first term converges everywhere except z = oo.

The second term converges everywhere except z = 0. Therefore, X(z) converges for 0 < \z\ < oo.

3.6
n

y(n) =
56

*(*)

=> y(n) - y{n - I)

x{n)

Y(z)

X(z)
X(z)
1-z-1

Hence,y(z)-y(z)2-

/ ^ / (1)".

n n
nz
*(*) = E(^)
3' "^ +'
< ^ (h~
2

\n_ n

n=0

n = -oo

1
1 - iz"1

1
+

1-

i*

5
6

(i_iz-i)(i_iz)
=o2

a
= rrrpr.5<W<
2'
3

~^

/ ^
Hence,y(n)

/ -2(3-)"+ ()"' " >

=

n < 0

y(n) = *(*)
CXI

y(z)

x(fc)u(n-fc)

x(n)* w(n)

*(r)l/(z)

u(n) * u(n)

*(*)

\J

u(fc)u(n k)

k = -oo
n

=
Hence,r(n)

andX(z)

(*) = (n+l)u(n)

u(n) * ti(n)
(1

_*_1)2,H>1
57

3.8
y{n) = x(n)eiwn. From the scaling theorem, we have Y(z) = X(e-jwz).
zeros are phase rotated by an angle WQ.

Thus, the poles and

3.9
*(n) = 5Mn) + (-l) n ()]

From the final value theorem

x(oo)

lim(2-l)X+(r)

hm(* + - i
-L)
x
*-i
2+1 '
2

3.10
(a)

1+224
1-22"1 +2-2
1 + 42-1 + 7 2 ' 2 + 102_3+...

j 1,4,7,10,..., 3n + 1 , . . . I

X(z)

Therefore,x(n)

(b)
X{z)
Therefore,x(n)

22 + 5z 2 + 82 3 + . . .

< . . . , - ( 3 n + 1 ) , . . . , 11, 8, 5, 2,o[

3.11
X(z)

(l-22-1)(l-^-1)2
A
B
Cz~l
( 1 - 2 2 - 1 ) + -r.
( l _ r - l rr
) + (1-2-M2
4,=-3,C=-l

Hence,x(n)

[4(2)n - 3 - n] u(n)

3.12
(a)
v ;

^ 0,
58

even
n o dd

51 *i(n)*""

*i(*)

n = oo
oo
=

n = oo
oo

=r
fc = - o o

X(2 2 )
=

(b)

x2(n)

= x(2n)

X2(z)

oo

2(n) 2 - "

x(2n)z-

n = oo

= f: x(*)*-
k= -oo

2 - ^ , K even
k= -oo

= I f; (*)x-* + i f; w(-z*)-
fc = - o o

* = -oo

= i[X(v^ + X(-v^)]

3.13

X(2)

A
Hence,x(n)

1-32 -1
1 + 32-1+22-2

A
(l + 2-i)
2,B = - 1

B
(i+2z-i)

[2(-1)"-(-2)"]U(B)

(b)

X(z)

1 - 2 - 1 + ^2-2

A ( l - i 2 - 1 ) + B(J2- 1 )
1 - 2 - 1 + ^2-2

A =

1,5=1
*\.-i

Hence, X(z)

1 - ; & ( *)*

l-S^teot})*-'+ ( L^2,-2
)'!
59

;k(""T)*
+ l - 2 ^ ( c o * } ) z - i + (L^l 2*, - 2
Hence, x(n)

r c o s

r s m

()

(c)
-7

,-6

x(n)

u(n - 6) + u(n - 7)

(d)
,-2

1 + *- 2
A'(z)

+2

1 + 2- 2

2-

x(n)

1 + 2- 2
cos nu(n) + 2cos (n - 2)u(n - 2)

(n)

26(n) cos-nu(n)

1 + 6Z-1 + z'2
7
4 ( l - 2 r - 1 + 2 2 - 2 ) ( l - i2'z " 1 )

(e)
X{z)

A(l-z-i)
l-2r-1+22"2
3

23

~5'B-To,c

+ l-2*~1+2z-2

\z~l

17
20

3 . 1 ._
Tt
2 3 / 1 n. n . v
V( - = );n c o s - n + (=)
sm-n
5 v/2
4
10 V V2
4

Hence, x(n)

1-

17,lxn
+ (-)
20 V

(f)
X(z)

x(n)

2 - 1.5. , - i
1 - l.bz'1 + 0 . 5 z - 2
1
1
l-z-1
1-iz-i
(n)

(g)
*(*)

l + 2z-1+z-2
l + 42~1+4z-2

V(l + 2z-M(l + 2 r - 1 ) ;

x(n)

=
=

,-2
2z - l
1
1 + 2*"
(1 + 2Z- 1 ) 2
<5(n)-2(-2)n-1u(n-l) + ( n - l ) ( - 2 ) n - 1 u ( n - i ;
6(n) + ( n - 3 ) ( - 2 ) n - 1 u ( n - l )

60

(h)
1

X(:

( , + ! ) ( , + !)

4 (z - \)(z - -=V^)(z -

' ,,)

1
(l + f z - 1 + I1, ,--22w) ,- i
8'
4 ( i - l 2**- i )A( *i _ 2 - i + ' I2z' - 2 )

^(1-K 1 )'

A
Hence, x(n)

Cz - 1

1-Z-l + iz- + l - Z - i + iz"

1
7
3
5 =
C =
" 2 '
8 '
i
5

( - ) ^ c o S - ( - 1) +

X(z)

1-Iz-i

j ( - 1) + ; ( 5 )

J ( 5 ) T

1 + iz"1
,-i

1+iz"
x(n)

4 1 + iz"1

(-i)B(n)+i(-5)B-l(n-l)

(j)
X(z)

=
"

1 az- l
.-l

all-iz-ij
-l

1-^z"
x(n)

az
l-hz-i

_I(I)
() + (I)-
tt(n-l)
a a u
a
(-i)"+lu(n) + ( i ) " - l u ( n - l )
a
a

3.14
X(z)

Iflz|>2,x(n)

If-<|z|<2,x(n)
Ifl*l<5,*(n)

5z - l

(l-2z-i)(3-*-M
1
1
1 +
1-2Z"
1-iz"1

2n - (\r

u(n)

-(^)n"(-n-l)-2nu(-n-l)
(i)"ti(-n-l)-2"u(-n-l)
61

r-l

u(n-l)

y-)^u(n-l)

xi(n)
_

l + (5)B](n)

x2(n)

y(r)

y(n)

xi(n

X1(z)X2(z)

=>^l(2

x2(n

(n)
1
1-2"1'

W + Cj)""^)
=

1+

X1(2)X2(2)
3

1 - 2"!

1-2"1
y(n)

1 - iz-1

3-<5)"

u(n)

xi(n)
=>X!(2)
x 2 (n)
=> X 2 ( 2 )
V(2)

1
1 - i2"1'
cosTrnu(n)
1 + 2-1
l+22~1 + 2-2
Xi{z)X3(z)
1 + 2"1
( l - i 2 " 1 ) ( l + 2 2 - 1 + 2-2)
A(l + z->)
l + 2 2 - 1 + 2-2

y4
V(n)

1-Jz- 1

3'fl"3

3 ^ n + -(-r
62

(n)

X!(n)

nu(n)
,-i

(I-'"1)2'
2nu(n-l)

x 2 (n)

2Z- 1
1-2*"1
*i(*)X a (*)
2z~ 2
( 1 - 2-1)2(1 _ 2 z ~ )

-2z~l

-2

1-2-1 ~ (1-2-1)2

y()

1-2Z-1

[ _ 2 ( n + l ) + 2 n + 1 ]ti(n)

z[X+(z)-z(0)}

z+[x(n+l)]

2X+(Z)-2X(0)

Therefore,

x ( n + l ) z - n + 2x(0)

zX+(z)

n=0
oo

(z-\)X

oo

+ (z)
n=0

n=0

= x(o)+Ez(n+1)-Ex(n)

\imz^X+{z){z-\)

n=0

=
=
=

n=0

/im m _ o o [x(0) + x(l) + x(2) + . . . + x(m)

-x(0)-x(l)x(2)-...-x(m)]
/im m ^oox(m + 1 )
x(oo)

x'(n)z-n

= oo

[x(n)(z*)-"]*

nss oo

= *"(0
-[X(z) + X-(z*)]

i [ 2 { x ( n ) } + 2{x'(n)}]
rx(n) + x-(n)l
=

z[Re{x(n)}]
63

-[X(z)-X'(z')]

x(n)-x'(n)
2j
z[Im{x(n)}]

= z
=

Xk(z)

S>.-

n = oo,n/kinteger

= J2 x^z mk
m = oo

e j,B0 "x(n)*- n

5Z ^(")(c" J U ; o ^)' n

i = -oo

n = -oo

XCze"-'"'0)

X(z)

log(l-2z),\z\<-

Y(z)

dX(z)
dz

-1

l-iz-i'

=>y(")

(^)n.n<0

Then,x(n)

y(n)

X(z)

Y(z)

lzl<i

| Z | <

n I

'oj(i-5*-').W>5
Z

dX(z)
dz

,
l-iz-''|

Hence,y(n)

-\(\)"-^n-l)

x(n)

fa

--(i)nu(n-l)

-**" .|*l>i
Z | >
2

n 2.

64

3.19
(a)
xi(n)

Xi{z)

rnsinwQnu(n),

0< r < 1

rsmiuoz"
- 11 +
j _ r- 22 zr - 2
1 2rcos\VQZ~

Zero at z = 0 and poles at z = re^Wo = r(coswo jsinwo).

(b)
X2(z)

(1 _ reJw*z-l)(l
z

re-Jvoz-1)

- 11 +_ L ^r22zr - 2
1 2TCOSWQZ~

(c) X\(z) and ^2(2) differ by a constant, which can be determined by giving the value of X\{z)
at 2 = 1.

3.20
Assume that the polynomial has real coefficients and a complex root and prove that the complex
conjugate of the root will also be a root. Hence, let p(z) be a polynomial and z\ is a complex
root. Then,
anz? +an-1zf-1
+ - . . + a 1 2 1 + a0 = 0
(1)
The complex conjugate of (1) is
an(zl)n

+ a . 1 ( z j ) n " 1 + + i(*I) + ao = 0

Therefore, 2* is also a root.

3.21
Convolution property:

z{xi(n)*x2(n)}

*!(fc)x 2 (n-t)
.*=-

n = 00
00

= r '<*> E >(-*)*-"
Jt = - o o

n=:-oo

00

X l (fc)z-*X 2 (z)

fc = -oo
X!(2)X2(2)

Correlation property:
r(/)

xi(n)x2(-n)
z{x!(n)*x2(-n)}
X1(2)z{x2(-n)}
X 1 (2)X 2 (2" 1 )
65

3.22
X(z)

22

1 + 2 +

z*
+

2!

X(z)

,-3

,-2

+ 1 + 2- l
x(n)

3!+"
2!

3!

6(n)+-

3.23
(a)

Hence, x(n)

1
1 + 1.52"1 - 0 . 5 2 - 2
0.136
0.864
1-0.28Z" 1 + 1 + 1.782"1
[0.136(0.28)n + 0.864(-1.78) n ] u{n)

(b)
X(z)

1
l-O.bz-i+OAz-2

1-0.25*"1
l

+ 0.3412:

l-Q.5z- +0.6z~

Then, x(n)

'

0.73262 - l
1-0.52~1+O.62-2

(0.7746) [cosl.24n + 0.3412si'nl.24n] u{n)

partial check: x(0) = l , x ( l ) = 0.5016,x(2) = -0.3476,x(oo) = 0. From difference equation,

x(n) - 0.5x(n - 1) + 0.6x(n - 2) = 6(n) we obtain, x(0) = 1, x(l) = 0.5, x(2) = -0.35, x(oc) = 0.

3.24
(a)
X{z)

For \z\ < 0 . 5 , x ( n )

For \z\ > l,x(n)
For 0.5 < \z\ < l,x(n)

1
1

1 - 1.52" +0.52-2
2
1
1-2"1
1-0.52"1
[(0.5)"-2]u(-n-l)
[ 2 - ( 0 . 5 ) " ] u(n)
- ( 0 . 5 ) n u ( n ) - 2 u ( - n - 1)

(b)
X(z)

For \z\ > 0.5,x(n)

1
(1-0.52-1)2
0.52" 1
(1-0.52-1)2

22

2(n + l ) ( 0 . 5 ) n + 1 u ( n + i ;
66

For |*| < 0.5, x(n)

(n+l)(0.5)nu(n)
- 2 ( n + l ) ( 0 . 5 ) n + 1 u ( - n - 2)
-(n+l)(0.5)nu(-n-l)

=
=
=

3.25
X(z)

l_^2-i+2-2
27
8

1-i*-1

1-32-1

ROC: - < \z\ < 3,x(n)

o

3.26
X(z)

x(n)z""

n = oo
oo

*i(n)*5(n)*-n

n = oo

^]fXi{v)vn-ldvx'3(n)z'n

E
^ *^c

n= -oo

ttfxMdv
27r

f; x;(n)(i)-
r

Un'oc

3.27
Conjugation property:

n=-oo

= oo

X'(Z')
Parseval's relation:
oo

5^

OO

^IC^SC**)

^-.fXMvn-'dvx'2{n)

n= -oo

67

3.28
!

/ N

I2"62

where the radius of the contour c is r c > \a\. For n < 0, let w = j . Then,

\w~n-1

If

2ffj yc* u; - i
where the radius of d is ^-. Since ^- < \a\, there are no poles within c7 and, hence x(n) = 0 for
n < 0.

3.29
x(n) = x(N 1 n), since x(n) is even. Then
N-l

X(z)

5^x(n)z""
n=0

x(0) + x(l)z~l

+ ...+ x(N-

W"1)/2

,()

2 ) z - " + 2 + x(N - 1 ) 2 " N + 1

2 (N-l-2n)/2 + z -(N-l-2)/2J

A>ven

n=0

If we substitute z~l for z and multiply both sides by z ~ ( ^ - 1 ) we obtain

z

-("-DX(z-i)

x ( z )

Hence, X(z) and X ( 2 _ 1 ) have identical roots. This means that if z\ is root (or a zero) of X{z)
then j - is also a root. Since x(n) is real, then z\ must also be a root and so must j r
section (a)
Xx(z)
X2(z)
7(2)
Hence, xi(n) *x 2 (n)

=
=
=
=
=

2 2 + ^ + l + 2 - 1 + z" 2
l + r - i + z-2
Xx{z)X7{z)
z 2 + 22 + 3 + 3 2 - 1 + 3 z - 2 - - 2 r - 3 + 2" 4
y(n)

|l,2,3,3,3,2,l|

Xf{z)
X}(z)

=
=

l + z ^ + z" 2
l + z-^z"2

y+(z)

1+ 2 * - 1 + 3 * - 2 + 2z-3 + z-4

Hence, y(n)

{1,2,3,2,1}

By one-sided transform:

(b) Since both xi(n) and *2( n ) a ^ causal, the one-sided and two-sided transform yield identical
results. Thus,
Y(z)

Xx{z)X2{z)
1

(l-Jz-i)(l-Jz-i)
68

l - i , - i

Therefore, y(n)

i - i

- i

3(i) n -2(iHu(n)

(c)
By convolution,
y{n)

xi(n)*x2(n)

J4,11,20,30,20,11,4!

=
=
=
=
=

2 + 3z-1+42-2
2 + 2"1
X+(z)X+(z)
4 + 8 z _ 1 + l l z - 2 + 42" 3
< 4, 8,11,4 \

By one-sided z-transform,
X?(z)
X}(z)
Y+(z)
Therefore, y(n)

(d) Both xi(n) and X2(n) are causal. Hence, both types of transform yield the same result, i.e,
\ + z-1 + z-2 + z-3 + z-4
l + z-i+z"2

Xi(z)
X7(z)

=
=

Then, Y(z)

Therefore, y{n)

X1(z)X2(z)

1 + 22" 1 + 3 r - 2 + 3 z " 3 + 3 2 - 4 + 2 2 _ 5 + 2" 6

j 1, 2, 3,3, 3,2,1 i

3.30
X+(z)

=
n=0
00

= 2>n=0

3.31
From the definition of the Fibonacci sequence, y(n) y(n 1) + y(n - 2),y(0) = 1. This is
equivalent to a system described by the difference equation y(n) = y(n 1) + y(n 2) + *(n),
where x(n) = 6(n) and y(n) = 0,n < 0. The z-transform of this difference equation is Y(z) =
z~lY{z) + r _ 2 y ( z ) = * ( * ) Hence, for X(z) = 1, we have
1

YW

,4
l_a^tiz-i

B
!-l^|
69

-i

V5+1 _

where A

Vh-\
2Vb

2y/b

1 - y/5
2\/5

y/S+l,y/$+\sn

Hence, y(n)

-2vr(-2)uin)-^7E
1

l + v/5, n .n
;

v/5

I-N/5,1

1-N/5

->/5

+ 1

)"(")

ti(n)

'

3.32
(a)
^ + U ) + \ [z-'Y+iz)

+ i/(-l)] - i \z-2Y+{z) + ^ ( - 1 ) + y(-2)] = 0

(a)
Hence Y+(z)

U
- 1 - ^*
*

Therefore, y(n)

0.154
0.404
1
1-0.3U'
1 + 0.81Z"1
[0.154(0.31)" - 0.404(0.81)"] u(n)

(b)
Y+(z) - 1.5 [z _ 1 y + (2) + l] + 0.5 [z~2Y+(z) + z" 1 + 0] = 0
Y+(z)

1.5-0.52" 1
1-1.5Z" 1 +0.52- 2
2
0.5
1-z-1
1-0.52" 1
[2 - 0.5(0.5)"] u(n)

Therefore, y(n) =

[ 2 - (0.5)"*1] u(n)

=
(c)

y + (2)-o.5[z- 1 y + (z) + i] =

v+(*) =

1 - iz-1
1.5- I z " 1
(l_I2-i)(1_0.5r-i)

1-0.52"
Hence, y(n)

l - ^ -

1(0.5)"-2()"]u(n)

(d)

y + ( z ) - i [ z - 2 y + ( z ) + i]

1-z-i

y + (*) =

i -i
l
^
i
*
(l-z-')(l-lzl-*"

Hence, y(n)

i
1-^z-

)
_7_
24
1

1 + iz"1

i _!(!)
+ i-(-I)"]
u(n)
V
;
V
;
V

70

24

'

3.33
(a)
Y(z) [1 - 0.2Z"1]

X{z) [l - 0.3*- 1 - 0.022" 2 ]

Y{z)
X{z)

_
"

(l-0.1z-1)(l-0.2z-1)
1-0.22"1

1-O.U- l

(b)
Y(z)

y(*)
X(z)

X{z)[l-O.U-1}
,1-0.12"
.,,-i

Therefore, (a) and (b) are equivalent systems.

3.34
X(z)
=> xi(n)
or X2(n)

l
1-02"1
anu(n)
a n u(-n1)

=
=
=

Both xi(n) and X2(n) have the same autocorrelation sequence. Another sequence is obtained
from X ( z - 1 ) = r i r
X{z'1)

*
l-az

1-

1
l-^"

(-)nu{n)
a
We observe that X3(n) has the same autocorrelation as x\(n) and X2(n)
Hence x 3 (n)

6(n) -

3.35

n=-l

n=0

= TT3FT + rr-prr. C: | < 1*1 < 3

V(2)

ff(*)X()

= (i-,-)(r-t-')(i-fz-)-Roc:i<M<
_

13
6

3
2

2
3

"

1-2"1

1-32"1

1 - fz-1

Therefore,
-3nu(-n-l) +

13

2 2

3V

u{n)

3.36
(a)
h(n

i^in)

H(z

1 - iz-1

x(n
X(z
H(z)X(z)

Y(z

7^x

l-iz-i

4*

l - i z - i + Iz-2

3\/3

l-iz-' + i -2

Therefore,
-(-)n + - ( - ) n c o s
H - ^ ( - ) n s m u(n)
7V3;
7V2;
3
7 v2;
3

y(")

0>)
h(n)

(jVx*)

H(z)

l-Iz-i

x(n)

(\)nv(n)

1
1 i*-1
H(z)X(z)

X(z)
Y{z)

(\)-nu(-n-l)

1
1-2Z"1

3*

(l-iz-i)(l-iz-i)(l-2z-i)

"2

-4
3

1 - -i z " 1 + "1 - J\ z - ir + 1 - 2 Z - 1
Therefore,

?(ir-i)r

!Kn)

u(n)+-2"ti(-n-l)

(c)

y(") =

-0.1y(n - 1) + 0.2y(n - 2) + x(n) + x(n - 1)

H(Z)

l + O.lz-i-O^z-2

72

x(n)

X(z)

Y(z)

(5)-()
1
-^"1
H(z)X(z)
]

1 + z'1

+ o.iz^-o^z-2)

( l-iz-^o

28

fi

"L _~i 2 - i

-1

1-0.4Z"

1 + 0.5Z- 1

Therefore,
y(n)

3;

3 5;

3v2y J

y(n)

Ix(n)-ix(n-l)

=>Y{z)

I(l-*-i)X(z)

"

10
1 + 2" 2

X(z)
Hence, Y(z)
y(n)

=
=
=
=

y(n)

Y(z)

X(z)
Y(z)

"

^
Trn
i r ( n - 1)
5co~u(n) 5cos
u(n 1)
irn . irn
ocos
bstn- u ( n - l) + 56(n)
I
2
2J
re, x 10 . .7rn 7T.
56(n)+-^5zn( -f-)ti(n-l)
10 . .irn
7r.
n
+
) t t
^" <T 4

( n )

-y(n - 2) + 10x(n)

10
XM
l + z-2A(-}
10
=
1 + Z- 2
100
=
(1 + 2 - 2 ) 2
50
50
1
""" l + j z " 1 ' 1 - J Z " 1

-25jz"1
(1+J2-1)2
J

25>r" 1
. ' . .
(1-jZ-l)2
.

Therefore,
y(n)

=
=
=

fc(n)

{50 [ T + ( - ; ) " ] - 2 5 n L T + ( - i ) n ] ) ( n )
(50-25n)(in + (-j)n)ti(n)
( 5 0 - 2 5 n ) 2 c o *ny u ( n )

(T) n (n)
o
73

H(z)

u(n) u(n 7)
l-z~n

x(n)
X{z)

tf(z)*(z)
l-z-n
(l-fz-i)(l-z-i)
l-z-i

l-z->

l-|z-i

1-fz-V

Therefore,

6-2(5r

y(")

*(n)

x(n)

=
=

ti(n)-V

'

5 - 2 ( - ) n-7 u ( n - 7 )

(^(n)

(-l)n,
cosm,

- oo < n < oo
oo < n < oo

periodic sequence and its z-transform does not exist.

y(n)

\H(wo)\cos[irn + &(wo)],w0 = it

H{z)

H(ir)

1 - Je-i
1

1+1

= 3'
Hence, y(n)

*(n)
#(z)

e =

-cosjrn,
o

oo < n < oo

(5)"(n)
1 - iz-1
1

2Z

x(n)

(n+l)(-ru(n)

X(z)

1
1-iz-l

iz"1
(l-iz"l)2

Y(z)
1
(l-$z-l)(l-lz-l)2
4

1 _ 1 2 -1

+ (l-iz"l)2
74

-3
1-Jz-l

-7

Therefore,

4 ( j ) " - n ( i r - 3 ( i ) ' u(n)

3.37
l-2z-1 + 2z"2-r-3

Tii \

*<*> =

(1

l
, Pi = 2'

l
55
-2.8
1-1,-t
P>2 2 =
=

(b) //(*)

i +

h(n)

(n) + 5 ( | r - i 4 ( | y

!_I2-i

i<W<i

_iz-i)(l _!,-!)'

lj'v/3

(a) ^1,2

. .

5<W<1

(1--.)(!-fa-.)(i-!,-.)

-i

u(n)

3.38
y()

x(n)
X(z)
Y(z)

0.7y(n - 1) - 0.12y(n - 2) + x(n - 1) -f x(n - 2)

z-' + z-2
^X(z)
l-OJr-i+O.^z-2
nu(n)
zr - i
( l - z - -M) 2
:- 2 + 2 - 3
(l-z-i)'(l-^-)(l-^-2)

System is stable
4.76z - 1
(1_z-i)2 +

Y(z)
y(")

-12.36
_,-i)

( 1

+
( 1

-26.5
_A2-i)+

( 1

38.9
_22-1

4.76n - 12.36 - 26.5(^-) n + 38.9(-) n u(n)

lu
o

3.39
(a)
? y ( n - l ) - i y ( n - 2 ) + x(n)
1

Y(z)
Impulse Response: X(z)

i-S*-' + i - 2
=

Y(z)
=/(*)

X{z)

1
1-iz-i

=
75

1-iz"1

Since the poles of H(z) are inside the unit circle, the system is stable (poles at z = i , \).
Step Response: X(z)

Y(z)

V(n)

1
1-z"1
8
3

1-z"1

1-Jz-1

+1-

HM^"

iz" 1

ti(n)

(b)
y(")

y ( n - l ) - - y ( n - 2 ) + x(n) + x ( n - l )

//(z) has zeros at z = 0,1, and poles at z = ~^-. Hence, the system is stable.
Impulse Response: X(z)

Y(z)

l-^-^Qgfz-1
1 - 2(v/2)- 1 co5jz" 1 + ( ^ ) 2 z - 2

y(n) = h(n)

( - ^ ) cos n + sm n u(n)
4
4

Step Response: X(z)

7T

1 - z" 1 +

\z-2

IT

1-z"1
1 + z- 1
(l-z~i)(l-z-i + iz-2

+
y(n) =

(^r

If

+
T

stn n cos nu(n) + 2u(n)

4
4

(c)

ff(z)
=> /(n)

=
=

z-'jl

+ z-1)

(1-Z")3

n 2 u(n)

Triple pole on the unit circle ^ the system is unstable.

Step Response: X(z)

Y(z)

y(n)

1
1-2-1

"

z-^l-t-z"1)
(1-z'1)*
lz-1(l+4z"1+z-2)
l z - ^ l + z" 1 )
+
T
1 4
2 (1-z-1)3
3
(1-z* )

( - n 3 + - n 2 + -n)u(n)

in(n + l)(2n+l)ii(n)
76

r-i
1
+ -
6 (1-z"1)2

(d)
y(n)

0 . 6 y ( n - 1) - 0.08y(n - 2) + x ( n )

l - o . e z ^ + o.osz- 2
Impulse Response: X(z)

H(z)

1
l

(i-i'-MU-i*- 1 )

=> zeros at z = 0, poles at pi = \,pi = | system is stable.

tf(z)

-1
l-iz-i
2(

Step Response: X(z)

Y(z)

=
=

y(n)

fr-<sr]<:

l-z-1
( 1

Y(z)

+l - | z - i

_i2-l)(l_lz-l)(1_2-l)
25
12

i
4

_4
3

l - z - ^+l - i z - ^ l+- f z - i
4V5;

12

v y

3V

(e)
y(n)

Y(z)

"

0.7y(n-l)-0.1y(n-2) + 2x(n)-x(n-2)
2-z-2
l - O J z ^ + O . l z "^ 2* ( * )
2-z"2

(l-iz-i)(l-lz-)X(x)

zeros at : = 0,2, and poles at z = ^ , 5 . Hence, the system is stable.

Impulse Response: X(z)

H(z)

2+

h(n)

2^(n)-|(i)"-1u(n-l)+~(ir-1u(n

Step Response: X(z)

1
1-z-1

Y(z)

-5
3

11

Jl

- i2 zi " 1
t;

4fi

1-2"1
=

2-z -2
( l - , - i ) ( l - l *2 - i ) ( l - l z - i )
3

V(n)

.-1

1
11 - i-z"
5 z

1-iz-l

5
10 1
+
2 T(2}

-23
6

l-Jz-

23 1 1
"6'(5)

U(n)

3.40

x(z) =

' *]

^i^

(l-iz-iKi-P^-MCi-p^-1)'
77

P=4+I

(a)
X l (n)
Xx{z)

=
=

x ( - n + 2)
z-2X{z-1)
z " 2 ( l + z)
(l-iz)(l-pz)(l-p-z)'

ROC:|z| < 2

Zero at z = - 1 , Poles at z = 2, , ^ and z = 0.

(b)
e

x 2 (n)

x(n)

X2(z)

= *(e=*S)
1 + e^z"1
(l-ieJrz-1)(l-pcJfz-1)(l-p-eifz-1

All poles and zeros are rotated by | in a counterclockwise direction. The ROC for X%{z) is
the same as the ROC of X(z).

3.41
*(n)

(^""H-^jr-M"-!)
,-1
1
41-iz-1

X(z)

l_I2-i
1-iz"1
1l - i z " 1
Z

y(")

(^)nti(n)

11 - i3zZ - 1
(a)

#(z)

y(*)*(*)
1 - iz"1
1

( l - i z - i ) 2( l - I z - i )
3
2
1-iz-i
1-Iz-i

M) =

3(i) n -2(i)"]u(n)

(b)
H(z)

1-K1

=
1

V(n)

12 Z

^ 12Z

~ y ( n - l ) - l y ( n - 2 ) + x(n)-ix(n-l)

(c) Refer tofig3.1.

(d) The poles of the system are inside the unit circle. Hence, the system is stable.
78

x(n)

y(n)

\J
r

\f

z"

-1/2

l i n

If

z1

-1/12
Figure 3.1:

3.42
H(z)
If a\ 4d2
Pi,2

|Pl,2|2

=
<
_
=

1
1 + <l\Z~l + Q2Z~2

0, there are two complex poles

-oi

JV4Q2-QI

\jAai - a\

<T> +(

< 1

a2 < 1
If a^ a4a2 > 0, there are two real poles
a\ y/a] - 4Q2
Pi,2

^ - 4 a 2
2
-ai - yja\ - 4a 2

<

land

>

-1

ai +

ai 02
a\ + 02

<
>

1 and
1

Refer to fig 3.2.

79

Figure 3.2:

3.43
#(z) =

r ' + l*-2

i - | z - + 4*-

W
7

r//_\

_-l

A(n)

' 7
~2 V

y(z)

H(z)X{z)

X(z)

Y(z)

y(n)

2.

u(n-l)

2V5;

(b)

1-z-1

-l

1-z
2
+
+

-3
!.j2-i

l_i2-i

i I(Ir_3(2)
8 sV
V

()

(c) Determine the response caused by the initial conditions and add it to the response in (b).
y(n)--y(n-l)+y(n-2)
80

Y + (z) - I [Y + (z)z-> + 1] + 2 [Y + (z)z- 2 + z" 1 + 2] =

25
Y+(z)

0
_ 25
ii

2-T-I
25'
i t
*

o-i^xi-i--5' 1 )
25

1 1

y+()

12 2

_L(i.) _ *() u()

25v5;

25V

Therefore, the total step response is

25 + i l r l x n . ? I / 2 . n
u(n)
8
200 V
25V

y(")

3.44

[ay(z) + x(z)]z- 2 = y(2)

-2

r?*(*)
2

1-OZ-

Assume that a > 0. Then

l

//(z)

a + (l-jaz-l){l
1
2al-N/a^"1

a
h{n)

-^

Step Response: X(z)

( n ) +

1-z

y(0

2^

Jaz~l)
1
1
+' 2al + y/az-1
+

l ( V

"VH)nlu(n)

-l

(l-z-1)(l-^-1)(l + ^ 1

(^T)

1-z"1
V(n)

2(a->A)
+

2(c+,/o)

1-v/^z"1

a - 1 + 2(a->/a)

1
+

1+ v^z-1

(v^)n +

2(a + Va)

i-VZ)'

(n)

3.45
y(n)

Y{z)

- a i y ( n - l ) + 6 0 x(n) + 6 ! x ( n - 1 )

TT^X(Z)

(a)
//(z)

bn0 +16iz
. => A(n) = 6o(-ai) n u(n) + M - a i ) " ~ M n - 1)
1 -f a\z~l

&o +

(ti-fcoai)^"1

h{n) = M ( n ) + (6i - M i ) ( - a i ) n _ 1 u ( n - 1)

(b)
Step Response: X(z)

1-z"1
81

Y(z)

60 + &i* - 1
( l - z - i K l + ajz-1)
d\bo b\
6 0 + 6]
1

1 + 0 1 1 - zV()

1 4-a^"1

1 + a!

60 + &i
1+ai

fli^o
^1
(-ai)n
1 + ai

ti(n)

(c) Let us compute the zero-input response and add it to the response in (b). Hence,
Y+(z)^al[z-1Y+(z)

+ A]

0
a\A

=>yZ\(n)

-ai>l(-ai)nw(n)

The total response to a unit step is

60 + &i
y(n)

aibo b\ aiA(l

+ a\)

-ai)

1 + a,

1 + a,

(d)
x(n)

co$wonu(n)
1 Z~1COSWQ
1 2Z~1COSWQ

7(z)

(6 0 + 6 I Z

B(l

l + ai2_1

Z~1COSWQ)

)(1 -

(1 + a 1 r " 1 ) ( l - 2z- 1 cosu;o + z - 2 )

A

Then, t/(n)

-+- 2 ~ 2

-1

[A(ai)

z~lcoswo)

1 2Z~1COSWQ

C(z

+ z~2

^OSUJO)

1 2Z~1COSWQ

+ Bcoswo + Csiniuo] w(n)

where A, f? and C are determined from the equations

A + B

60

+ (sintfo)C'

61 6 0 cosu;o

J4 (a\ COSWQ)B + (sinu>o)C

b\coswo

(2cosu>o).4 -f (a\ COSWQ)B

3.46
- y ( n - l ) + 4*(n) + 3 x ( n - l )
4 + 32"1 ,

Y(z)

1-

i^v;

e>w n u( n)
1
1 - - e*w o * - l
4 + 3z"-1

x(n)
X(z)

(1

A
5
2

82

ei^zB

-1 ' 1 -- ti^oz'- 1

1where A

z~ 2

4ejWo + 3

Then y(n)
The steady state response is
/'m n _ 00 y(n) = y{n)

[A{frac\2)n

Be3*on

eju>0 _

+ Be***"] u(n)

3.47
(a)
H(z)

=
=

(z - r e ' e ) ( z - r e - > e )
z{z + 0.8)
l - 2 r c o s e z _ 1 + r2*-2
C
(l + 0.8z)
1.8
= 2.77
1 - 2rcosO + r 2

(b) The poles are inside the unit circle, so the system is stable.
(c) y(n) = -0.8y(n - 1) + Cx{n) - l.b\fiCx(n - 1) + 2.25Cx(n - 2). Refer to fig 3.3.

x(n)

7(n)

-0.8

-1.5/3

2.25
Figure 3.3:

83

3.48
If h(n) is real, even and has a finite duration 2N + 1, then (with M - 2N + 1)
H(z)
since /(n)
tf(z)

h(0) + h^z-1

+ h{2)z~2 + . . . + h(M -

/i(M - n 1), then

z-(^-i)/2(/,(0)[2(^-i)/2 + r-(^-i)/2'

\)z~^-i)/2

+/(1) [z< M - 3 >' 2 + z - ( w " 3 ) / 2 ] + . . . + h(M - 1/2))

with M = 27V + 1, the expression becomes
H(z)

z _JV (M0) [z N + z~N]

+M1)

+M2)

+ ... + h{N))
N-l

z~N | /*(#) +
n=0

V.

/(n)2^" n + ] T / i ( n ) z - ( N " n )
n=0

7V

z- {/(AT) + P(z) + P ( r - ) }

Now, suppose z\ is a root of H(z), i.e.,

H(zO = zf/v{M7V) + P(^i) + P K 1 ) } = o

Then, /.(/VJ + PCrO + P^f 1 )

= 0.

This implies that H(j-) = 0 since we again have

W + P ^ ^ + P ^ ) = 0.

3.49
z- 1

ROC: - < Izl < 2

*<*)=7
(z7+ni)(z + 3 ) ( z - 2 ) '

(b) The system can be causal if the ROC is \z\ > 3, but it cannot be stable.

(0
, ,

(1) The system can be causal; (2) The system can be anti-causal; (3) There are two other
noncausal responses.The corresponding ROC for each of these possibilities are :
ROCi : \z\ > 3;
ROC 2 : |*| < 3;
ROC 3 : \ < \z\ < 2;
ROC 4 : 2 < \z\ < 3;

3.50
x(n) is causal.
(a)
X(z)

Y.x(n)z~n
n=0

limz^tX)X{z)

=
84

x(0)

(b)(i)X(z) = ^ I = > / : m i _ 0 0 X ( z ) = oo => x(n) is not causal.

(ii) A'(z) =
sequence.
(in) X(z)
sequence.

(1 A / - 3 ) 3
t_

JU-I

= ^^3

=> / i m , _ 0 O X ( i ) = 1 Hence X(z) can be associated wih a causal

=> /im^eoXCz) = 0. Hence X(z)

can be associated wih a causal

3.51
The answer is no. For the given system hi(n) anu(n) => Hi(z) = 1 _ a 1 < . 1 , \a\ < 1. This system
is causal and stable. However when /2(") = anu(n + 3) => / ^ ( z ) = i_ a /-i
but is not causal.

tne

system is stable

3.52
Initial value theorem for anticausal signals: If x(n) is anticausal, then x(0) = lirrizoX(z)
Proof: X{z) = E L - o o x(n)z-n = x(0) + x ( - l ) z + x ( - 2 ) z 2 + . . . Then / i m , _ 0 * ( z ) = x(0)

3.53
5(n) = ( i r " 2 u ( n + 2)

h(n)

ff(r)

s(n)-s(n-l)

(I)"-2tl(n + 2 ) - ( i r - 3 u ( n + l )

3 4 6(n + 2 ) - 5 4 6 ( n + l ) - 18(^) n u(n)

,
-18
8 U 2 - 5 4 z + 1 - *-z~l
1

3Z

3*

//(z) has zeros at z = 0,1 and a pole at z = 5.

(b) /i(n) = 8l6(n + 2) - 546(n + 1) - 18() n u(n)
(c) The system is not causal, but it is stable since the pole is inside the unit circle.

3.54
(a)
1

2TJ

forn>0,x(n)

2, nn-_l i

/e 2 - 4

(i)n
85

for

n<0,x(-l)

= J-I

2*J Jc
z

dz

z{*-\)
z

**J Je 22{Z - i )
2

By continuing this process, we find that x(n) = 0 for n < 0.

(b)

*<*> = irrj=r.W<5
,(B)

/ ^LdZl

where c is contour of radius less t h a I

For n > 0, there are no poles enclosed in c and, hence, x(n) = 0. For n < 0, we have
2*j Jc z(z1
r

Z s

i)

lo = -2

Alternatively, we may change variables by letting w z

1

/ *

. Then

""" , 1

)<*u/,

2*7 / e , u;+i(l - 1;)

dw

I 2w~n~1 J
t>
dw
2*j Jc.
w-2)
-(2)'n,
n<0
1

=
(c)

*> = fF.w>Ei
:(n)

=
=

Forn>0,x(n)

z"" 1
dz, c has a radius greater than
2*J Jc
1
|a|

l lzi^LZ^dz

2*jJe

z - \

- I ( I ) - ( i - a)
a a
a
a

a
86

Forn = 0,x(n)

-1
27r?/e a

(*-<0 <fz

-ir-a

J-a

(a2+l-a2)
a
a
For n < 0, we let w = z *. Then
*(n)

2xj
I-aw~l
2irj Je,
0, for n < 0

tu2 )rfw.

(d)
X(z)

l-^-1
1
1_ i z - i _ U - 2 ' l z l > 2
6'
J_

10

1+iz-1

1 - J2-l

x(n) = J L / j f i l d Z

-L/jifld.

where the radius of the contour c is greater than |z| = 4. Then, for n > 0
x(n)

Forn<0,x(n)

10V

"(n)

10v 3 '

3.55
X(z)

71
7TT
TTT>
(1 - az)(l a z * )
1
-1
1-az"1
1-Iz-i

x(n)

-: 6

* '

1 / zn
For n > 0, r 4
dz
2*; Je z - a
2*J /C * - i
1 / zn
For n < 0, : i>
dz
2TTJ 7e z - a

-/A

2TTJ jTc z - a

an and

0 and

a < |z|< - , 0 < a < 1

a

dz -r
2*j / e z -

rdz
1

dz = a"

3.56
r20

X{Z)

J
M <2
,
"
( , - I ) ( z - 2 ) S ( z + )2(z + 3 ) ' 2< < M
87

x(n)

'-1,20

2*>JcU^TJTTZ

88

Chapter 4

4.1
(a) Since xa(t) is periodic, it can be represented by the fourier series

*.(0 = f ; cte>2**/T
where ck

Asin(irt/r)ej7*kt,Tdt

-T /

J o

dt
x

r e j*(i-2fc)t/T

e->(i+2t)t/T

j2r U f ( l - 2 * )
4

-jf(l+2A)J

1
1
1 - 2 * + 1 + 2*
1A

7T

TT(1-4*2)
oo

x f l (<)c" i 2 * ( / '"' ) t c/<

oo
oo

/-oo

kt _= -oo
oo

J OO

dt

= E

cMF--)

* = -oo

Hence, the spectrum of xa(t) consists of spectral lines of frequencies *,* = 0 , l , 2 ,

amplitude |c*| and phases Lc^-

(b) px = $f0zi(t)dt = I / ; ^ ' ( V ) A =

(c) The power spectral density spectrum is |ct|2,Jb = 0 , l , 2 , Refer to fig 4.1.
(d) ParsevaFs relation
P,

= lf
=

M2 =
t = oo

*l(t)dt

TJ 0

c*

id! v
JT-

ir2
89

2^

( 4 *

l)2

with

tf
ICjl-

*-l

lc2l

ico r

-2

Figure 4.1:
4,4'

,
1 +

Hence,

2
2
3*+15?

1.2337(Infinite series sum to )

|c*| 2

AA2

(1.2337)

4.2
(a)
/le-atu(0,
Xa(F)

a>0

^-atg-,2^1^

c -(a+>2F)t

- a - J2TTF

A
a + j'27rF
l*a(F)|

v/a 2 + (2TTF) 2

90

lXa{F)

-tan-l(

)
a

Refer to fig 4.2

(b)

A = 2, a = 4

Figure 4.2:

Xa(F)

I Aeatc-**Fidt+
Jo

\Xa(F)\

LXa(F)

I Ae-aXt->2* Ft dt
Jo

+' a + j2*F
a-j2nF
2aA
a2+2*F2
2aA
a 2 -I- (2vF)2
0

Refer to fig 4.3

4.3
(a) Refer to fig 4.4.
X[ }

Xa(F)

\ 0,

otherwise
othen

[ (l+t-)e-*7*Ftdt+
J -T

fT(l--)e-'2* dt
J 0

Alternatively, we may find the fourier transform of

*) = *') = { r -;;<-,0
91

A-2.A.6

Figure 4.3:

^A
Figure 4.4:

92

AA

2/x

1/T

= j

Y(F)

y(t)e-^Ftdt
dt

2sn 2 7rFr
jvFr
and X{F)

( -TFT)

\X(F)\

sinirFr\

LXa{F)

Ck

sinirFr\

J L / " " x(t)e-^kt>T>dt

hJ

-T,/2

= i - f / (1 + -) C -> 2 '^ T ^+ / T (l--)e-> 2 ' i t IT,

T

Jp [ i -T

/ 0

r IsinirkTf IT,
j.p
Tp [ xkr/Tp

) and (b), we have c* =

f-Xaif-)

x(n)

j . . . , 1,0,1,2,3,2,1,0,1, ...J

Ck

= 5E*()"i2"*"/ 6
n=0
r

-3*

-Jirt

3+ 26"^"+ c"^~+e

Hence, CQ =

-J4tr>

irk n
2irk
3 + 4cos-- + 2coso
<5
9
4
4
n
*
n
-,c1 = -,c2= 0 , c 3 = i , c 4 = 0 , c 5 = -

P, = i t ^ l 2
n= 0

-jtOwll

+ 2e

i ( 3 2 + 22 + l 2 + 0 2 + l 2 + 22)
6
93

19
16

Pj = E k W f
n=0

2
2
(T7r
16 + (fr + o + (rr + o + (fr)
Thus, Pt =

19
16
Pj
19
16

x(n) = 2 4- 2cos7rn/4 + COSTTTI/2 + -cos3irn/A, = TV = 8

Cfc

i^z(n)e-'""/4
n=0

{H,2+>v5,1,2_2v5.i.a-f^.u+|^}

x(n)

Hence, CQ

2 , C i = C-j = 1 , C 2 = Ce - , C

1
3

= C5 = - , C

^ = 2>(or
i=0

1 1 1

= 4+1 + 1 + - + - + +
4

16 16

53
8

x(n)

. 7r(n-2)
4s*n
-

4$in
Ck

2ir{n-2)

= lE'W^'6
n=0

= lrrt.
=
E

M 2)

" e""'""''

n=0

73 l
94

= 0

. 2irk

V3

. irk

,-j2ir*/3

+ sin
(-J2) sin
6
3

Hence, c0

0, ci = - j 2 e ~ j 2 , r / 3 , c 2 = c 3 = c4 = 0, c5 = cj

and | C l | = |c 5 |

2 ; |c 0 | = |c 2 | = k 3 | = | c 4 | = 0

7r -h

ZCI

Uh
ZCQ

=
=

JT

2*

5TT

-5*
ZC2 = lCz = ^C4 = 0

(b)
/ ^
x(n) =
Ck

2irn
. 2irn
cos-- + sin- => JV = 15
3
5

C u + C2i

where c u is the DTFS coefficients of cosQ1 and C2* is the DTFS coefficients of sin-.
2irn

ia>

But

- j j "s -

cos = - ( e ^ ^ +

Hence,
* = 5,10
otherwise

i fc

Clk

Similarly,
.

2irn

Sm

"5"

1 ,

}i'

27(e

-i2 .

"c

Hence,
C2k

0,

otherwise

Therefore,
2j

Ck = Cifc + C2*

2',

* - >

* = 5
* = 10

k = 12
0,
otherwise
l8iniin Hence, N 15. Following the same method

_
(c) x(n) = cos^-sin 2
-_ 2I 5| - n lfiin
5
15
as in (b) above, we find that

2*"" 15
-1

ck=<

jj,
I 0,

4 = 8,13
otherwise

(d)
AT

5
-j3l

Ck =

7 *(")<
n=0
1

- c

-ja>

-j4>

+ 2c

-iw>

2e

2j -sm( ) - 2 s m (
5
95

-!>

-e

Therefore, CO
CQ =
=

o,
2

. ,4*
'
2j - s m (/ * \) + 2sm(
)
o
o
5
0

C\

C2

C3

=
=

C4

47r

/ \ /2TI\
n(y)-2n(y)

2j
5
-c7
-ci

(e)

A7

Ck

= :*(>

6
3irifc

n=0

-l*k

- l + 2e '
6 .

-e

-iTIrk

-e

->4">

1
l
(_)_2C05( )
6 +4co5
1
Therefore, CQ
^
CO =
2
2
Cl
3
=
0
C2
-5
=
C3
C4
C5

0
2
3

(0
TV =

c* = ^5I*(n)<

-llwnk

n=0

1 f. , -!>
r 1+ c s
O L

,**

5cos(T)c

2
5

Therefore, Co

2
JT
-C05(-)

ci
C2

- ^

=1.

^2-

3*

-&

scos(T)c

c3 = r c o s (-r) c *

(g) .V = 1
(h)

c4

2
,4T
-cos{)e

ck= x(0) = lorco = 1

N

96

+ 2e

= 5E*<>"'""
n= 0

Ck

=> Co

2V
0,ci = 1

4.7
(a)

;n

( ) = YlCk e*^
k=0

Note that if c*
e

, then

*=0

n=0

8,

0, p ^ - n

I-

Since cjt

p= - n
Jiwfc

-)3irfc

+ e

We have i ( n )

c I

^_c=J|ziLl

+ 27

4<5(n + 1) + 46{n - 1) - 4j6{n + 3) + 4j6(n - 3), - 3 < n < 5

(b)
73
CQ

0,Ci =

V3
,C2 =

V^

,C3 = 0 , C 4 = -

v^
,C5 = -

.
, C 6 = C7 = 0

:(n) = 5I CjtcJ ~^"

= 73

Trn
. irn ] j(3-a)
sin + s i n I e
2
4

(c)

(") = J I Cke

t = -3

1 is-

2+e ^ + e

?rn
2 + 2cos- +
4

1 =is_

ffn
1 3?
cos+-cos
2
J
4

4.8
(a)
If*

N-l

n=0

97

0,N,2N,...

j3s*

-j

If*
N-l
V*

e ;'2irin

IN

0,N,2N,
-t* Ink

1 _
n=0
=

>2(0)

eJ2*k/N

(b) Refer to fig 4.5.

(0
k=l

s (2)

k*2

sj(l)

s2(l)

2(2)

s (1)
3 -,
s (3)
3
s (5)
3

i (5)
2

k=4

k=5

s (5)
4

s
s

4
4

(1)

s (0)
4
s (3)
4

5 (0)

(4)

Figure 4.5:
N-l

N-l

2 n/N-j2rin/N
5>(").
=
E^'
**
n=0
n=0
JV-1
V " eJ'2T(i-,)n/N
n=0
W, * = I

0,**i
Therefore, the {sjt(n)} are orthogonal.

4.9
(a)
x(n)

u(n) u(n 6)
98

x(w) = Yl x^y~iwn
n = oo

1 - e~Jw
x(n)

2nu(-n)
0

X(w)

E 2"<

n = oo

=
m=0

2
2-e''

*(n)

(^) n ti(n + 4)

n= -4

= (X^re'^HV 4 *
m=0
44cj4u,

1 - ^c--"

x(n]
X(u;)

ansinwonu(n),

|o| < 1

= S>"

2j

n= 0

2j

-;'(u/-u/o)

"' r=0

n=0

_1_
2j

-j(u/+u/ 0 )

1 - 2ocosu; 0 e-> u ' + a2e~32w

x(n)

\a\n$inw0n,

oo

\a\ < 1

oo

Note that J^ lx(n)l =

^2

ns-oo

n = -oo

= , so that Isintynnl

99

\a\n\sinw0n\

1.

Mn = Y, 1^)1--

Therefore, the fourier transform does not exist.

(n
(B)
=
v

'

X(w)

( 2 - ( i ) - , ll<
\ 0,

othe
otherwise

x(n)e"^"

n= - 4
4

= E

a-( 5 )"

,-jwn

n= - 4
2e;4u-

-e~3w

_ i [-4e>4u' + 4e~i4w - $e>3w + c _ j 3 u ' - 2e j2tt ' + 2c" j 2 , i ; - ejw +

=

2eJ'4u'
. .
r -ul + j r[AsinAw + 3stn3u> + 2sin2w -f sinw]

1 e~J

(g)
X(w)

x(n)e-^n

n = oo

-2e>2w - eJU' + e ^ + 2e - - ,2u '

2j [2sni2u> + smu>]

(h)
{

/ A(2M+l-\n\)t
*~ \ 0,

\n\<M
|n|>M

= A J2 (2A/ + 1 - |n|)e-^n
n= -M

(2M + 1)4 + > l ( 2 M + 1 - AJCe-^* + e]wk)

k=\
M

(2M + 1)4 + 2A YJ&M

+ 1 - k)coswk

fc=i

4.10
(a)
r(n)

J^X(w)<Jndw

/,'Wo

1
2TT

y_

100

/*

2 * y 0

(fu;

v - wo
For

*<

ejwndw

jn
fe-jw0n

dw

Jvin

-in

in
jn

= sinnwo , n ^ 0

Hence, x(n)
(b)
X(u>)

cos7(w)

I(e>2" + 2 + e->2u')
4

x(n)

/
2TT

XCuOe^du;

y_w

[27r^(n + 2) + 47r6(n) + 27r6(n - 2)]

OTT

\ [( + 2) + 26(n) + 6(n - 2)]

(c)

x(n)

A'(tx;)eJU;,l<itx;

-!- /
1

/^0+^f

2^7^.^
l ^ / s t n ^ W ^ ^ ^
7T \
n6u;/2
/

(d)
i

x(n)

/*/

Re I I
2*

2ejwndw +

yjQ
/*/8

/>3ir/d

eiwndw+

,7ir/8

e>wndw+l

A/8

J6w/6

Jit 16

f3w/8

r7/8

*tr

e>wndw)
J

/
2cosumfu; + /
coswndw + /
cosumdu; + /
2coswndw
J0
A/8
Jew/A
Jlw/i
1 [
lirrx
. 6TTTI . 37rn
. irn
\sirx
+ sin
sin
sin-
nir
8
8
8
8

4.11
x e (n)

x(n) + x ( - n )

= {5.0.1,2,1,0,1}
101

= {^0,-2,0,2,0,1}
3

Then, XR(w)

xt{n)e~iwn

51
n= - 3
3

jXj(w)

z0(n)e->n

j ;
n= - 3

Now, y(u;)
y(n)

Xi{w) + XR(w)e>7w.
F-l{Xi{w)}
+

=
=

Therefore,
F-l{XR(w)ti3w}

(g)
X(w)

x{n)e-i"n

^
nzz oo

_2e j 2 u ' - e*" + cJU/ + 2e - j 2 w ;

2j [2sin2w + sinit;]

(h)
A(2M+l-|n|),

*<> = { o , (

|n| < M
\n\> M

Af

*(,) =

53 *(n)e-^"
n= -Af
A#

A 5 3 (2M + l - | n | ) e - J u m
n= - M

(2Af + l)i4 + A 5 ^ ( 2 M + 1 - k)(e->wk + e ^ *

M

(2M + 1)A + 2A 5 3 ( 2 A / +

~ *)coswk

*=i

4.12
(a)
x(n)

J^X(w)^ndw

=
1

P"0

ir wo
For n ?fc 0, /

e^du;

e>wn\Z
102

/"

jn
Jwo

Hence, x(n)

sannu>o
_~""'~" ,n ^ 0
n?r

(b)
X(u>)

cos2(tu)
(

-i)2

i(cJ2"+2+e--'2")
x(n)

2*7-

AXtiOe^dti;

-!-[27r6(n + 2) + 47r6(n) + 27r6(n - 2)]

i [ 6 ( n + 2) + 2<5(n) + 6 ( n - 2 ) ]
4

(c)

X(n) =

^J\x(w)^ndw
2 * y w o _^r

6u;

szn(n6u;/2)
n6w/2

'W*1

(d)
x(n)

Rel I
2*

[Jo

2eJwndw + /
Jw/8

/w/8

2cosu/ncJu; + /

costimdtu + /

Jw/8

7irn

. 6irn

l
e>wndw +
e>wndw)
J6*/S
Air/8
J
r 7 " - /*

r3/8

Jo

1 f

e>wndw+

Je*/8

. 3irn

. irn~\

4.13
*(n)

x(n) -t-x(-n)
2

{5.0.1.2,1,0.1}
x 0 (n)

x(n)-x(-n)

= {i,0,-2,0,2.0,i}
103

coswndw + /
JT*I&

2cosu/nd

Then, XR(w)

xe(nK}wn

n= - 3
3
n= - 3

Now, y(;)
y(")

X 7 (u;) + XR{w)ei2w.

Therefore,

= rM^/N + r^HW^)
=

-jx0{n)

+ xe{n + 2)

= {i^-^'+kH2'0'!}
4.14
(a)

*(")

1_
= 2n
^

e>xundw+

/
J6*/10

eiwndw + 2

I
J-9W/10

ejwndw + 2
JBrflO

J-w

2TT

-(_eJ W i o + e-> 9 *"/ 10 + e>*n - e-J'

jn

[sinirn sinSTrn/10 sm97rn/10]

nn
[sin4irn/b + sin97rn/10]
H7T

(b)
x(n) =

-!- / X(w)e^wndw+^2* y_

2TT

2* y_ *

f
y0

X(w)ejwndw

2x JQ *

1
irn .-in/2
s i n c "* '
irn
2

(c)

x(n) =

- W
2
* Jwc-f

2e>wndw+

irn
_2_ sin{wc -\

2e3wndw
2ir J_Wc_*

2;
)n 5171(11;,; - )n

7rn

104

e>wndw

4.15
/ f 1, 0 < n < M
i\ ) ~ | o, otherwise

x n

X,(w) = J ] c - ^
n=0

l_c-;w(A#+i)
1 - e-'%

,,
2

f 1, -Af < n < - 1

' ~ \ 0, otherwise

-l

X2(w) = e"^
n= - M
Af

""
n=l

X{w)

1 - ejwM
1 - e'w
= X1(w) + X2(w)
I 4. eiw _

eJ

1 e-JwiM

+ l) e)w(M + l) _^_ elM _|_

w

e-jwM

2 - e~i - ei
2coswM 2cosw(M -f 1)
2 2cosw
2sin{wM + y)cos7r
2sin 2 f
sin(M + ^)w
5m(|)

4.16
(a) A'(0) = n x ( n ) = - 1
(b) lX{w) - TT for all w
(c) i ( 0 ) = f- / J , X(w)rfu; Hence, / % X(u;)rfu; = 27rX(0) = - 6 *
(d)
oo

*() = ^

*(n)e-'""' = ( - l ) " * ( n ) = - 3 - 4 - 2 = - 9

n = oo

(e) J_\ \X(w)\2dw = 2*Zn k()l2 = (2*)(19) = 38*

4.17
(a)
X(w)

x(n)e-'"""

105

X(0) = >(n)
n

dX{w)

. ^

= -;'5^nx(n)
n

Therefote,c

(b) See fig 4.6

L ^ - l

X(0) = 1 Therefore, c = S = 0.

dxr.w)
dw

Figure 4.6:

4.18
X!(n)

=
F

a n u(n)
1
1 - ae~]<

Now, suppose that

X4(n)

= !(-!)! flM(w)
F

[l-ae-^)*
holds. Then
106

*'-^r + x * ( u ; )

ae~iw
1
(l~ac-J')fc+1 ' ( l - a e " ' " ' ) *

* - > " " = (5^x(n)e-><-"'> B r = * ' ( ~ " )

V ( _ n ) e -i*

^y(n)e^"

x e ^ n = X'(u;)

*(n)e"'->(n-l)e-'"
n

7(u;)

=
=

"

X{w) + X(w)e->w
(\-e-JW)X(w)

y(n) =

x(*)

ts=-oo

Hence, X(u>)

=
=
=

y(") - y( - i)
*(n)
(1 - e ^ ^ u ; )
X(u/)
1 - e->"

VW =

y( w ) = x(2n)e"^
n

= *(f)
y() = *(?)"'""
2
n

X(2u;)
107

2lx^e'iwn

*,(u/) =

e>2w + e>w + 1 + e~jw + e ^ 2 "

1 + 2cosw 4- 2cos2u;

=
=

Y,x*(n)e~JWn

X2(u>) =
=

e>4

^2*

e->2 +

e-j4w

14- 2cos2u; 4- 2cosAw

X3(w)

^x3(n)e-^n

1 + 2cos3u> + 2cos6u/

and X3(w) = Xx(3w).

Refer tofig4.7
j an integer
otherwise

X*(u;)

x*(n)e-^n
n, an integer

A"(fcu;)

Xl(n)

x 2 (n)
-Y2(w)

i ( e " n / 4 + c-JTn/4)x(n)

- L ^ W * + c-^ n / 2 )x(n)
2;

* ( " - ) + *("+fl
108

-Ti/3

n/3

Figure 4.7:
(c)

x 3 (n)

i(e>"/ a + e - ^ B ' 2 ) x ( n )

X3M = I [*(,-I)+ *(, +I)

(<*)

X4(n)

5(^*"+-^")x(n)

X(u> - *)

4.22
N-l

^ = ^ n=0
E ^ ) c _ i 2 , r t n/ N
7V-1

oo

E(-~/N)
N E
n=0
L/=-oo

oo

tf-i-tN

"

e-j2wkn/N

= E E *(-)e-'J"l(m+W/-V
/=-oo m=-lN

109

N-l-IN

m=-lH

Therefore, c\

_
. ,
Let xN(n)

J^X(-J^)

sinwen
,

-N<n<N

TTl

,
/ \
where x(n)

=
=

w(n)

x(n)w(n)
sinwen
.
oo < n < oo
?rn
~~ ~
1,
- N <n< N

0,

sinwen
Then
Trn

X(w)

=
=
=

XN(w)

Xi(w)

otherwise

1,
0,

\w\ < wc
otherwise
X(w)*W(w)
fPi

/ x(e)w(w - e)de

'w< sin(2N +
\)(w-e)/2
de
sin(w - e ) / 2

x(2n+

l)e~]wn

= x(*)e-^*/V"/2

1_

X 2 (u;)

a e ;W2

x ( n + 2)e , r n / 2 e-J u m
n

- ^ x ( * ) e " - ' ^ t t ' + ^ / 2 V 2w

t

-X(u, + ^ ) e ^

X3(w)

^x(-2n)c-^n

110

-jkw/2)

-z(*)e->*
w

(d)
X,(u/)

J^i(e' 0 - 3 * B + c-J 0 - s * ,1 )*(n)e-' ,, ' B

= ij:*()ki(- -0.3ir)n
=

- [X(u> -

0.3JT)

-j(u/+0.3*)n

+ X{w + 0.3*)]

(e) Xb{w) = X{w) [X{w)e->w] = X7(w)e-ix

(0
X6(tu)

X(u;)X(-u;)
1
(1
-ae-3w)(l-aeJw)
1
(1 2acosw + a2)

4.25
(a) yi(u-) = T,nyi(n)e-'wn
= n n even *( n ) e ~ J U , n
obtained by combining the results of (b) and (c).
(b)

The

Courier transform Yi(w) can easily be

y2(n)

x(2n)

VbM

J]y2(n)e-'""
n
n

5^i(m)e-- ; ' ,, ' ,T,/2

m

Refer to fig 4.8.

;o

w-ff "

even
otherwise

Y3(w)

X>(n)e-jwn
n

*(n/2)e->n even
,j2wm
m

X(2w)
111

-71

3n/2

-TC/2

Figure 4.8:

We now return to part(a). Note that yi(n) may be expressed as

yi(n)-|

i/2),
0>

n even
n odd

Hence, Yx(w) = Y2{2w). Refer to fig 4.9.

4.26
(a) Because the range of n is (00,00), the fourier transforms of x(n) and y(n) do not exist.
However, the relationship implied by the forms of x(n) and y(n) is y(n) = x 3 (n). In this case,
the system Hi is non-linear.
(b) In this case,
1

X(w)

1-

Y{w)

\e-iw'

1 - \e~iw
Y(w)
X(w)
1 - \e~Jw

Hence, H(w)

1 " l-'' w
=>

System is LTI

Note however that the system may also be nonlinear, e.g., y(n) = x 3 (n). (c) and (d). Clearly,
there is an LTI system that produces y(n) when excited by x(n), e.g. H(w) = 3, for all u;, or
(e) If this system is LTI, the period of the output signal would be the same as the period of the
input signal, i.e., N\ = AV Since this is not the case, the system is nonlinear.
112

^(*)

-7i -7n/8

""/S

*#

*#

7r

^8

^2

3^4

^(w)

- n -3n/4 -7i/2 ^ 4

Figure 4.9:

4.27
(a)
M

Y,w*(n)e~3wn

wR(w) =

n=0
M

=E

,-P"

n=0

1 _ e-HM + l}w

1 -e-i

(b) Let u>T(n) = hR(n) * hR(n - 1),

h

/ N . / 1.
~ \ 0,

( n )

0 < n < f - 1
otherwise

Hence,
Wr(u;)

/fj(u;)e-^
sin -11)
sin*
113

,-jwM/2

(c)

Let c(n)
Then, C(w)
Wc(w)

1.,
2irns
3(1+-)
L

>

!,

"\

1,,

2ir,

7T < It; < 7T

= T- r c ( e )^(^ - e ) j e
5 WHM

+ 5 W(u/ - ~ ) + \\VR{w + ^ )

Refer to fig 4.10

-27I/M

-2n/M+l

2n/M+l

Figure 4.10:

4.28
(a)
A(n)
ff(u,)

(5)B(n)

= jt,(hn*'iwn
n=0

= E(|e-^r
n=0

114

2n/M

1 - ie->'
\H(w)\

=
[ ( 1 - \cosw)1 + (Jstniy) 2 ]*
1
T cosu;

Z//(u>)

-tan - 1

e{w)

For the input x(n); =

istnw
2
1 ^cosu;

cosn
10
1

]3wn

;3iri

- ( C > + C ><>

r/

X(u>)

y(u-)

7T

3TT.

*(u;"To)+

xn

\H(%)\cos

M<

3TT.

3TT

*(,--) + *(,+ _ )
'3icn ^ , 3 * \
io- + e(To>

,1,1,0,1,1,1,0,1,1,1,0,
T
x(n) + z ( - n )
2
x(n)-x(-n)
2
2>e(n)e-"""

First, determine x e (n)

and x 0 (n)
Then, X(u;)
Xj(w)

\H(w)\

e(w)

To)

ff(u>)X(w)

*<>
y(n)

3TT

(u;+

and y(u;)

{XR(w) + X](w)},
tan
-iXj(w)

XR(w)
H(w)X{w)

y(n)

Y(w)

x(n) + x ( n - 1)
-(l+e-^Wu;)
115

H(w)

-(l+e-'w)

Refer to fig 4.11.

(b)

-2

-1

0
> m

Figure 4.11:
x(n) - x(n 1)
lK)

Y(w)

i(l-e"^)X(ti

H(w)

5(1-e"^)
( s mv,^ e-,- ^ / V ^ 2

=
Refer to fig 4.12.
(c)

x(n + l ) - x ( n - 1)

v()
=

^-e-nXiw)

= {sinw)e>*12
Refer to fig 4.13.
(d)
irfn)

x(n + 1) + x ( n - i;
116

Figure 4.12:

Figure 4.13:

117

Y(w)
H(w)
=

cosw

Refer to fig 4.14 (e)

1.5

>m

Figure 4.14:

x(n)+x(n-2)

y(n)

Y{w)

1(1+ *->*")*( u>

H(w)

i ( l + e-j2")

(cosw)e~iw

V(n)

x(n)- - x ( n - 2 )
2

Y(w)

iu- e->

H(w)

io-

(sinu;)c-- ; ' u ' +;T/2

Refer to fig 4.15.

(f)

2tt;

)X(

e-j2u/)

Refer to fig 4.16 (g)

y(n)

V(u;)

x(n) + x ( n - l) + x ( n - 2 )
i ( l + e--'w + e--'' 2w )X(u;)
118

35

Figure 4.15:

0.8
?06

?0.4
i

N.

N.

0.2

o.s

1.5

>w

Figure 4.16:

119

2.5

3.5

H(w)

\H(w)\
/tj(

i(i

+ e - i + e ->a*)

\{l +

\(l + 2cosw)\

2cosw)e-jw

\ f ~w>
^W' ~ \ -K - w,

l + 2cosw>{
1 + 2cosw < I

Refer to fig 4.17.

(h)

Figure 4.17:

y(n)
Y(w)
H{w)

=
=
=
=

x(n) - x(n - 8)
(\-e-j6w)X(w)
(l-c->8u')
2(sin4w)ej{'/2-4uj)

Refer to fig 4.18.

Y(w)
H(w)

2z(n-l)-x(n-2)
(2c~' u '-e-"' 2u ')A:(u;)
2cou; cos2w j(2sinw sin2w)

\H(w)\
e{w)

[(2cosw cos2w)2 + (2sinw sin2w)7]

-tan'1

, /2smu; sm2u>\

\2cosw cos2w )
120

Figure 4.18:

Figure 4.19:

121

Refer to fig 4.19.

(J)

x(n) + x(n - 1) + x(n - 2) + x(n - 3)

Y(w)

- ( 1 + e->w + e'j2w
4

e-j3w)X(w)

H(w)

\ [e-jw(e>w + e " ^ ) + e ^ ' V " + e">')]

i(e->"'+e-> 2 , , ')co*;
E \ -)c- i 3 / 2
(cosiy)(cos
Refer to fig 4.20.
(k)

Figure 4.20:
..
V(n)

x(n) + 3 x ( n - 1) + 3x(n - 2) + x(n - 3)

-

Y{w)

l{l + 3e-Jw +

3e-i2w+e-i3w)X(w)

H(w) = i ( l + C "^) 3
=

(cosu;/2) 3 e--' 3u '/ 2

Refer to fig 4.21.

(1)
y(n)

x(n - 4)

Y(w)

e-> 4u, X(u;)

//(u;)

e~>Aw

\H{w)\

(u;)

Aw
122

Figure 4.21:
Refer to fig 4.22.
m
y(n)
Y(w)
H{w)
H(w)\

e(w)

=
=
=

x(n + 4)
ei^Xiw)
4w
e>

=
=

1
Aw

Refer to fig 4.23.

(n)
x ( n ) - 2 x ( n - l) + x ( n - 2 )
y(")
Y(w)

i ( l - 2 e - > u ' + c--'2tt')A'(u;)

4

(sn2u;/2)e-J^-*)

H(w)

Refer to fig 4.24.

4.30
(a)
x(n) + x ( n - 10)
e-jl0w)X{w)

Y(w)

(l +

H(w)

(2cosbw)e-}bw
123

1.5

0.5
n

0.5

1.5

>*

Figure 4.22:

Figure 4.23:

124

2.5

3.5

Figure 4.24:
Refer to fig 4.25.
(b)
=

"(f)
j,(n)

0
57T

(2cos )e~ J 3*
=

(6c

oST)S,n(- +
5TT

(6cos

)s,n(

- - _
47TT

3-lo")

(c)

H(0)

2
2

y(n)

27rn

20+ l O c o s + -

4.31
/>(n)

H(w)

=
=

tate Response: //()

n

Therefore, ySs{ )
Transient Response:

6(n) + 2 ( n - 2 ) + 6(n-- 4 )
1 + 2e~j2w + e - j 4 w

(l+e-J2tt')2
4(co5tx,)2e-J'2"

0,(n>4)

125

'-1

1.5

05

3.5

2.5

> w

Figure 4.25:

ytr{n)

=
=

lOe^u(n) + 20czisf21 u(n - 2) + 1 0 e r i ^ l i u ( n - 4)

106(n) + jl06(n - 1) + 10<5(n - 2) + j'10<5(n - 3)

4.32
(a)
x(n) + x(n 4)
(l +
e-j4u')X(w)
{2cos2w)e-j2w

V(n)
Y(w)
H(w)

Refer to fig 4.26.

(b)
y(n)

It ,

But cos (n 4)
2
=

It

7T

7T ,

cosn + cosn + cos(n 4) + cos-(n 4)

2
4
2
4
7T
. 7T
.
= cosncos2ir + sin nsin2ir
2
2
7T

cosn
2
and cos [n 4) = cosncosir sinnsinir
4
4
4
= cosn
4
Therefore, y(n) = 2cosn
\c) Note that H{\)

= 2 and / / ( J ) = 0. Therefore, the filter does not pass the signal

126

cos(jn).

0.5

25

1.5

3.5

Figure 4.26:

4.33
- [x(n) - x(n - 2)]

y(n)

Y(w)

H(w)

l(l-e')

(sinw)ej^-^

H(0)

o,*(f) = i

Hence, y{n)

3cos(^n +60)

ytr(n)

~(l-e~^)X(w)

4.34
x(n) = Acos^n
(a) y(n) - x(2n) = Acos\n => w = J(b) y(n) = x 2 (n) = 4 2 cos 2 J n = \A"* + A 2 cosfn. Hence, u> = 0 and u; =
(c)

y(n)

x(n)co57rn

Acosncosirn
4
A
5*
A
3*
cosn + cosn
2
4
2
4

=
Hence, it; =

3*

57T

and it; =
4
4
127

4.35
(a)
y(n)

-[x(n) + x ( n - l ) ]

Y(w)

j(l + e ^ ) X H

COs(-)e->*

Refer to fig 4.27.

(b)
1

0.8

o.6 "
?0.4

0.2

()

0.5

1.5

2.5

3.

>w

^
A

'-1.5

'
9

1.5

0.5

2.5

Figure 4.27:
y(n)

Y(w)

\H(w)\

0(u>) =

"2 [*(")-*("-!)]
-l(l-e~nX(w)
n|
e*<*-*>

Refer to fig 4.28.

(c)
y(n)
Y(w)

= -[x(n) + 3 x ( n - l ) + 3x(n-2) + x ( n - 3 ) ]
o

i ( l + e-^) 3 X(u;)

cos*(^)e->*

128

3.5

Figure 4.28:
Refer to fig 4.29.

4.36
y(n)
Y{w)
H(w)

x(n) + x(n - M)
{l +
e-JwM)X(w)

H(w)

0, at

or w
\H(w)\

= (*+-)*,

(2*+l)7r/M,

k = 0,1,

Jt = 0 , l ,

\2cos-\

4.37
y(n) = 0.9y(n - 1) + bx(n)
(a)
Y(w)

Y(w)
X(w)

H(w)

\H(0)\

0.9e-iwY(w)+bX{w)

1 - O.Qe-J"'
1,=6 = 0.1
wM

S(iv) =

7T

>

wM

129

cos^f- > 0
cos^f- < 0

i
I

0.5

1.5

2.5

, ,

3.5

> w

Figure 4.29:
(b) |tf(u, 0 )| 2 = i => 1 6 1 _ 1 6 . 8 c o ^ 0 = | = wo = 0.105
(c) The filter is lowpass.
(d) For |#(u;o)| 2 = 5 => wQ - 3.036. This filter is a highpass filter.

4.38
(a)
JV-l

Px

=
n=0
V-l

k=0
-1

cl + 2j2\ck\
k= l

P* - 2|c t 0 | 2
P* - 2|c k0 | 2

Spurious power
THD

2kt0|2
P*
(b) for / 0 = i , refer to fig 4.30
for / 0 = X , refer to fig 4.31
(c) for / 0 = J , refer to fig 4.33
for / 0 = i , refer to fig 4.34
for / 0 = 25^, refer to fig 4.35
The total harmonic distortion(THD) reduces as the number of terms in the Taylor approxi
mation is increased.
130

trm-4

Figure 4.30:

Figure 4.31:

131

tmrmtm 2

Z o

-20

20
K
A

: o
-20

88fM- 2 S

300

100

200

100

200

300

o
M

! -1
100

200

300

Figure 4.32:

ong costhd-2 22e-16

trTTtt-2th4-0 06186

100
0
100
200
terms5mO-0 6054
50
0

-50
-100
trms-7 ftflU) 069241
100

-100.
^trms-8 tfto 0026571

I -100
-200

Figure 4.33:

132

terms-3 thd-0 4379

ortg cos 1hd>0

terms*3 ihd-0 4439

terms-2 thd-0 07905

Figure 4.34:

ong cos thd6 6616-16

100
100
0

100

I -100

I
-100

-200
0

1
I terms.* %&.0 5271

trms-5^U) 6077
-200

100

0
100

-200
I
100

0
100
200

Figure 4.35:

133

-100
trms^S-0 6238
-200

4.39
(a) Refer to fig 4.36
(b) /o = *

100

200

300

> n

> n
Figure 4.36:

bits
THD

4
9.4616e-04

6
5.3431c - 05

8
3.5650c - 06

16
4.2848c-11

bits
THD

4
9.1993e-04

6
5.5965e - 05

8
3.0308c - 06

16
4.5383e-ll

(0 /o = TJoo

(d) As the number of bits are increased, THD is reduced considerably.

4.40
(a) Refer to fig 4.37
(b) Refer to fig 4.38
The response of the system to Xj(n) can be seen from fig 4.39

4.41
(a)

H(w) =

A(n)e"'"-"

n = oo

n=0

134

f-0.25

f-0.2

-0.5

100

OS

-1
0

006

0.1

0.15

02

0.25

toq(Hl)

Figure 4.38:

135

0.3

0.39

0.4

0.49

0.9

f-0.25

-0 2

Figure 4.39:
2C

1 " i*'

l-e>

\H(w)\
IH(w)

5 3cosw
4
5 Scosw
0

x(n)
X(w)

3*71

=
=

Y(w)

*(n)

y(n)

T"

'

3irn.

37rn

Hu~ ) + ( " +

* < u; <

J/(u;)X(ti;)
4?r
3irn.
..
3irn
* ( " - ) + <5(u;+)
5-3cos^-

, the output is simply y(n) =

where A

cos-

Acos

3*71

3*\

= "(">)!,= = # ( T )

= {...,-1,1,-1,1,-1,1,-1,1,-1... A
cosnn, oo < n < oo
4 _ 1
4
5 - SCOSTT ~ 8 ~ 2
1
zosirn

136

Y(w)

1[6(W-T)

6(W+W))

4.42
(a)
xa(t)

c--'2^0'

ATa(F)

/
Jo

=
=

za{t)e-'7'Ftdt

e-j2wFote-j2Ttdt

Jo

e-J2r(F+Fo)tdt

l^
Jo

c -j2>r(F+Fo)t

-J2*(F

Xa(F)

J2*(F + F 0 )

z(n)

e ~~n

F0)^

(b)

X{f)

x(n)<)e-i**J*

n=-oo
oo
L

e-

-r*-e-j2*fn

n= 0

= V J ' 2 ' r ( F + ) n
n=0

1
l-c-.W+ft)
(c) Refer to fig 4,40
(d) Refer to fig 4.41
(e) Aliasing occurs at Fs 10Hz.

4.43
(a)
y(n)

x(n) - 2coswox(n 1) + x(n - 2)

/i(n)

<5(n)-2cosu> 0 6(n-l) + 6 ( n - 2 )

(b)
ff(u;)

1 - 2coswQe->w + e ^ 2 "

(i_e--'u'0e-^)(l-^u'Vu')
, . j w + t^o . w - u>o
4c J B s i n - s i n -
2
2
2e~JW(cosw COSWQ)

\H(w)\

2\cosw - COSWQ\

=^ |//(tt;)j

0 a t t i ) = wo

137

KVU

1000

800

600

I
400

200

200

400

600

800

1000

1200

Figure 4.40:

F*-10

Fa-20

12
10

i:
:

10

15

Ft-40

30
IT
^20
A
1

10
ft

JL

20

40

ISO

60

Figure 4.41:

138

Refer to fig 4.42.

:o
w0 pi/3

Figure 4.42:
when w0 = ir/2,H{w)
at u> = ir/3, # ( * / 3 )

1 - e>2w
l - e > a , r / 3 = le> , r / 3

y(n)

|//(7r/3)|3cos(^n + 3 0 o - 6 0 )

3cos(^n-30)

4.44
y(n)

=
=

z(n)-x(n-4)
2e-j2wejr/7sin2w

Refer to fig 4.43.

(b)
x(n)

cosn + cos n,

y(n)

2cos^n,

/ f ( | ) = 2,

(c) The filter blocks the frequency at it; = y.

4.45
y(n)

//(n) = 0

-[x(n)-x(n-2)}
139

*#() = 0

r
2

1.5

0.5

2.5

3.5

>w

Figure 4.43:

H(w)

_(l_e-j2)

e-^ei'^sin w

x(n)

5 + 3sin(-n

+ 60) + 4sin(irn + 45)

H(0)

0,

y{n)

3 s m ( ^ n + 60)

ff(|)=l,

//(*) = 0

4.46
(a)
y(n)

x(2n) => This is a

y(n)*-,wn

oo

]T *(2n)e->"*

n = oo

=
=

i,

M<2

o,

!<M<*

(b)
y(n)

x 2 (n) => This is a non-linear, time-invariant system

Y(w) =

X(w)*X(w)
J7T

140

Refer to fig 4.44.

(c)

-nil

71/2

Figure 4.44:

y(n)
Y(w)

(cosTrn)x(n) ^
1

This is a time-vary in

[7r6(u; TT) -f 7r6(u; + 7r)] * X(w)

2TT

i [X(w - ) + X(w + TT)]

0,

1
2'

, ,

3*

37T

. ,

<M<*

4.47
/(n) = (^)ncos-nu{n)
(a)
l-icosfz"1
l-2(i)cosJz-1+(i)22-2

//(z)

l-^z"
_

1 -

4 *

141

^16-

(b) Yes. Refer to fig 4.45

(c) Poles at z \e3^, zeros at z = ^ .

x(n)

V
.(

''

z' 1

y(n)

"

\J~2I4

>^2/Z

'

-1
z
-1/16

Figure 4.45:
. - a af l . - >

V?1

# ("0 =
1

'" 1
~

r - . Refer to fig 4.46.

16c

(d:
x(n)

X(z)

Y(z)

(-)nti(n)

1-IzX(z)//(z)

+
2
1 - ^42Z- 1 ^+ 16
-J-2Z
1

y(n

2 ( 4 ^ ll + C S I n + ^ + ^ s i n ^ n ]

4.48
y(n) = i ( n ) - z ( n - 10)
142

u(n

13

12

In

\
09

n fi

05

15

35

25

> w

Figure 4.46:

\H(w)\

=
_
=

l-e->10w
2e- j5tt 'e J ? sinbw
2\sinbw\,

Q(w)

bw, for sinbw > 0

H(w)

bw + 7T, for sinou; < 0

2

Refer to fig 4.47.

(b)

(1)

Hence, y(n)

2cosn + 3\/3sin(-n - )
1U

(2)

tf(0)

0,

Hence, y(n)

#(y) =0

4.49

M) =

J\{w)^dw
_1_
27T

I J^dwJ-*r
143

e-}wnd

-i

10

.. 1

\
\

A
1

' -1
9

1.5

o.s

2.5

> w

Figure 4.47:
. 3TT

. 7r

sin n sin n
8
8

irn

2 . it
*
s t n ncos n
nn
8
4
(b) Let
Tin

Then,
u

0,

<|U;|<TT

and
/i(n) = hi(n)cos n
4

4.50
V(n)

-y(n-l) + x(n)+-*(n-l)

V(')

i ^ y W + ^CO + i*-1^*)
Y(*)

tf(z)

X{z)
1 x

^Z"1
2z

(a)

tf(z)
/i(n)

1 - iz"1

- 1

2(^)nu(n)-6(n)
144

3.5

(b)
H(w)

/i(n)* jwn
n=0

1 - ie"->'

- 1

1 + KJ"
1=

\e~^

#(*)!*-

(c)

*<f)

1-

\t->?

i + ii
le-;2tan-4

Hence, y(n)

c o s (

2n+4"

4.51
Refer to fig 4.48.
|X(w)|lor()

W*)lor(b)

|X(w)|ter(c)

4.52
=

Ww)|tor(d)

Figure 4.48:

//(;)

(l-cJ'Tz-1)(l-c--'Tr-i;
145

H(w)

=
=
=
=

l-v/22"1**-2
l-x/2e-;u;+e-2jt
2e-iw(cosw-)
2
x(n)-V2x(n-l) + x(n-2)

for x(n

y(o

sin u{n)
4
x(0)= 0

v(i

x(l)-V2x(0) + x(-l) =

V2

y(2

x(2) - \/2x(l) + x(0) = 1 - v / 2 - y + 0 = 0

y(3

x(3) - v^x(2) + x(l) = Y

y(4

4.53
(a) H(z] = ^ 0 ^ - 1 - Refer to fig 4.49.
(b)

Figure 4.49:

H(w)

ib

l + 0.9e-""
146

" ^

+ ^

= 0

Ww)l

, .
e(w)

21"" II

v/1.81+ 1.8cosw
_i sfnu;
, 0.9sinw
ion ' 1 cosu;+
-1 '/an"
- 1 1 -+- 0.9cosu>

(d) y(n) = -0.9y(n - 1) + ^ [x(n) - x(n - 1)]

(e)
ff(|)
y(n)

0.014e>e<*>

0.028cos(Jn+134.2)
6

4.54
(a) H(z) = 6 o i j ^ I i - Refer to fig 4.50.
(b) For a = 0.5,6= 0.6,

H(z) = 6O|IQ5*-* Since the pole is inside the unit circle and the

Direct form I:
x(n)

y(n)

G-O

bbQ

Direct form II

Or
V

x(n>,

V
,

-1
z
-a

Figure 4.50:
filter is causal, it is also stable. Refer to fig 4.51.
(c)
H(z)

=>\H(w)\'

147

l + 0.5z - l
1-0.52"1
4- cosu;
A2
6 j4i
T cosu;

b0

y(n)

z-plane

Figure 4.51:
The maximum occurs at w 0. Hence,

H{w)\w=0

o i
4

96g = 1
=*6o

(d) Refer to fig 4.52.

(e) Refer to fig 4.53.
obviously, this is a highpass filter. By selecting b = 1, the frequency response of the
highpass filter is improved.

4.55
\H(w)\'

[1 + r 2 - 2rcos(u; - 0)] [1 + r 2 - 2rcos(u; + 0)]

d
1
dw \H(w)\2

[2rn(u; - 0)(1 + r 2 - 2rcos(u> + ))

+2rsin{w + 0)(1 + r 2 - 2rcos(u; - 0))]

(1 -fr 2 )(sin(u- - 0 ) + sin(w + e))

0
2r [sin(u; - Q)cos(w + 0 ) + sin(w + 0)cos(u' - )]
148

|H(w)|
1

2.5

08

f 6
^0.4

"

0.5

1.5

3.5

w
phaM

Figure 4.52:

z-plane

Figure 4.53:

149

(1 +

r2)2sinwcosQ

Therefore, cosw
Wr

2rsin2w
4rsinwcosw
14-r2
cosG
2r
1 + r2
-l
cosQ
COS
2r

4.56
ix(n) + - x ( n - l ) + - x ( n - 2 )

y(n)
H(w)
=

'

e-i,nS?!L
'"cos
<>w
cos

\H(w)\

e(w)

2
LH{w) = -w

Refer to fig 4.54

4i

-2

-1

0
> w

Figure 4.54:

4.57
(a)

x(n) = (-)() +
4

u(-n-l)

150

X[z)

_i_

r i

ROC:i<|,|<l

4'

Hence, H(z)

Y{z)
X(z)

= YTI^-

R0C:

1*1 < l

(b)
Y{z)

4Z

(1
1-

')

-f
-i*- i

3
5

1.

' 1 + r- 1

V(")

-^)nti(n)-^(-l)nti(-n-l)

y(n)

6 0 x(n) + 6 i a r ( n - l ) + 6 2 * ( n - 2 )

ff(u>)

6o + 6 i e - ^ + 6 2 e-- ,2u '

4.58

(a)
=

bo + bie-i2?

tf(0)
For linear phase, 60
select 60
These conditions yield

=
=
=

6o + 6 i + 6 2 = l
62.
62 (otherwise 61 = 0).

60 = 61 = 62

\e-jw{\

H&

Hence, H(w)

+b2e-jir

+ 2cosw)

:b)
H(w) = - ( l + 2cosu;)
.
_ f -u>,
(u;) - I _ ^ + ^

for 1 + 2costf > 0

for 1 + 2costi; < 0

Refer to fig 4.55.

4.59
(a)

y(n)

2A/T7 2 - * ( n ~* }
151

=0

Figure 4.55:

H(w)
K ]

V * e~jwk
2M + 1 . *-"
M

1
2M+1

1+

2y^coswk
*=i

(b)
A/-1

M:
M-l

jLcofJIfu+JL 1 + 2 ^

*(>) =

cosu;*

t=i

The filter in (b) provides somewhat better smoothing because of its sharper attenuation at
the high frequencies.

4.60
(a)
,-}2*Ft

X(f)

Joo

f 3 \ f* *l(* - T)z2{T)dT e->2TFtdt

J oo [.J oo

f* *7{T)\r
J oo

Xi(F)

zl{t-T)e'*W-')dt

Woo

l"

x2(r)e-^FTdr

J oo

Xl[F)X2(F)
152

e-i**F7dT

(b)
\d\ = T + t,-T

*()

< t <0

/.;
IdA = r - * , - r X > 0

{i.

otherwise

Refer to fig 4.56.

(c)

x(t)

T
Figure 4.56:

Xl(F) = X7{F) =
=

I* e-j2*Ftdt
sirnrFr 2

(- *F

4.61
H(z)

l + z + z2 + . . . + 2l
1-z9
1-2"1

153

H(w)

1 - e-J9w
1 - e~iw
e-j9w/2 5 z n 9 l t ; / 2
e ~ > / 2 siniv/2
sin9w/2
j4w
esinw/2
sin9w/2
' stnw/2 '
4w, when sin9w/2 > 0
4u; + 7T, when sin9w/2 < 0

=
1

"
6(t/;)

=
=

9_.L

#(u>)

0, at w= - , J b = 1,2,...,8
y

The corresponding analog frequencies are *F.

* = 1,2,3,4,or kHz, kHz, |kHz, fkHz.

4.62
Refer to fig 4.57.

Figure 4.57:
(l_ei3'/4z-i)(1_e-;3r/42-i)

H(z)

H(w)

*(*)!.'

O-^- 1 ) 2
154

H{0)

\H(w)\

I2

( T ^
U G ? = 1
4

2 + v/2
?=- = 0.073

4(2 + 72)

4.63

1 rcos(w - 9) -f jrsin(u> - 0)
(a)
|#,(u;)|
201og 10 |/f,(;)|

{ [ l - r c o s ( u ; - 0 ) ] 2 + [rsm(u;-0)] 2 }5

[l + r 2 - 2 r c o s ( u > - 0 ) ] *

101og10[l - 2rcos(w -9) + r2)

Hence proved.
(b)
n i \
Q2[w)

=
=

4 - l i m a g . part
tan
:
real part
,._-i
rsin{w-9)
tan
1 rcos{w 9)

Hence proved.
(c)
y

aw
1
1 + , r'$tni(w-0)

[1 - rcos(w 9)]rcos(w 9) rsin(w 9)(rsin(w 9))

[i _ rcos(w - 9)]2

T2 rcos(w 9)
1 + r 2 - 2rcos{w - 9)
Hence proved.
(d) Refer to fig 4.58.

4.64

*'(w) = 1 - J'e-i '

r<1

(a)
\HP(w)\

1
2

{[1 - rcos{w - 9)] + [rsin{w - 0)] 2 }*

1
\HAw)\
155

phase theta=0

magnitude theta-0

group delay theta=0

-10
" magnitude iheta=1.57r
10|

-20
_l magnitude lneta=3.1 A
~
10

"

phase theta=1.571

group delaylheta=1.57T
20

1|

"

-10
"group delaylheta=3.145

phase theta=3.142

156

-10
-5

\Hp(w)\dB

201og10(

-201og 1 0 |//,(u)|
-\Mt(w)\dB
Hence proved.
(b)
Hp(w)

ep(w)

1 rcos(w 9) jrsin(w 9)
[1 - rcos{w - 9)]2 + [rsin{w - 9))2
, rsin(w 9)
-tan x
,
'-1 rcos{w 9)

-ez{w)

=
Hence proved.

(0
rUw)

<tep(w)
dw
d{-e,{w))
dw
det(w)
dw
-

Hence proved.

4.65
Hz(w)

{l-reJe->w)(l-re-j'e-ju')

=
=

A{w)B{w)

(a)

\H,W\

\n,M\dB

=
=
=

|-A(;)6(u;)|
\A(w)\\B(w)\
201og 10 |tf,(u,)|

101og10[l - 2rcos(w -9) + r 2 ] + 101og10[l - 2rcos{w + 9) + r 2 ]

(b)
LHt{w)

r'Aw)

LA{w)+LB{w)
rsin(w + 9)
, rsiniw
6)
y
tan-1
^ r + *an - i
1 rcos(w + 9)
1 rcos(w 9)

(c)

det{w)

dw
r>A(w) +

rf(w)

r rcos(w 9)
1 + r 2 - 2rcos(u; - 5 )
157

r 2 rcos{w + 9)
1 + r 2 - 2rcos(if + 9)

(d)

HJw)
Therefore, \Hp(w)\

1
Ht{w)
1

\n,{w)\

\HP(w)\dB
on the same lines of prob4.62

-|^(u)|dB

ep(w)

8z(u>) and

*!M

-r/W

(e) Refer to fig 4.59.

magnitude theta-0

magnitude theta=1.571

phase theta-0

group delay theta=0

phase theta=1.571

group delay theta=1.571

1.5

Figure 4.59:

4.66
(a)
\Hi(w)\2

=
158

(I-a)<
(1 acosw)2 + a2sin2tv

(i-g)2

|//l(ti')|

= - => COSW],

1 + a 2 2acosw
4a - 1 - a 2
2a

(b)
i u A M2

m*)\

("2")
(1-a)
2

i rr /

\i2

(l + cosw)2 + sin2w

,1-0,2

(l_ae08w)2
2

a28in*w

2 ( 1 + cosu;)
1 + a 2 2acosw

2a

1+a2

'

By comparing the results of (a) and (b), we find that cosw? > cosw\ and, hence u>2 < w\
Therefore, the second filter has a smaller 3dB bandwidth.

4.67
h(n)

cos(won + 0 )
= COSWQTICOSQ sinwQnsinQ

use the coupled-form oscillator shown in figure 5.38 and multiply the two outputs by cos
and sin&, respectively, and add the products, i.e.,
yc(n)cos& + yt(n)sinQ

COS(WQTI

+ 6)

4.68
y(n)

e ^ y ( n - l ) + x(n)

1)

=
=

(coswo + jsinwo) [yn(n - 1) + jyi(n - 1)] + x{n)

yR(n - l)cosw0 - yi(n - l)sinw0 + x(n)

yR(n - 1) + jyi(n-

+j [j/fl(n

l)sinw0 + yi(n -

1)COSWQ)

(b)Refer to fig 4.60.

(c)
Y{z)

y(n)

e^z^Y^ + l
1
ejnwou(n)

Hence,

[COSWQTI + jsinwQn] u(n)

J/K(TI)

coswonu(n)

yi(n)

sinwonu(n)

(d)
n

y/(n)

0 i

1 #

6 7

J
159

-I

-f

COS W/

Figure 4.60:

4.69
(a) poles: pi, 2 = rejUJ<
zeros: z\ 2 = eJU,
(b) For w = WQ,H{WQ) = 0 For
H(w) = 1. Refer to fig 4.61.
(c)
\H(w)\2

C2

G2

where u^o =

d\H(w)f

# ^0, the poles and zeros factors in H(w) cancel, so that

|1 - e J ^ e - ^ l 2 ! ! II - re> u 'e--' u; | 2 |l -

-^e-^l2

re-iwe-Jw\2
2 ( 1 - COS(w WQ))
2(1 - rcos(w+ U'Q))
2
1 -f r 2rcos(w u>o)J [ l + r 2 2rcos{w + u>o).

Then
o

0 => W = 7T

dw

i*(*)i2 =

4 G 2

< n ^)

= 1

I(l + r+r 2 )

(d) Refer to fig 4.62.

(e)
160

Figure 4.61:

Irr?
y(n)

-2r cos w,

-2cos

-1

Figure 4.62:
161

|ff(Of

h _ re> u 'e-> u '| 2 |l - re->e-.> u '| 2

In the vicinity of w = WQ, we have

\H(w)\2

% G2

1 _ reJwe-Ju'\2
2(1 co5(u; WQ))
G2
. 1 + r 2 2rcos(w tuo).
1 + r 2 - 4G 2
2r - 4G 2
_1/l + r 2 - 4 G 2 ,
C 0 5
"
< 2r-4G2 >

=
cos(w wo)
Wl,2
BzdB W\ W2

2C S

2r-4G 2

2cos-'(l-(

v^

2 2(1 )2

V ^

2vT^7

4.70
For the sampling frequency Ft = 500samples/sec, the rejected frequency should be'uii
2 T T ( | ^ ) = |TT. Hence,

2 " ( ^ ) = ^7r. The filter should have unity gain at w? =

H(-*)

= 0

and H{\-K)

H(w)

G(l-eJ^c">u')(l-e--'^e--'u;)

G e ' ^ ^ c o s u ; - 2cos ]

H(U)

Hence, G

2G\[cos(*)-coS(^*)}\=l
COSXTT cosiv
25

4.71
From (5.4.22) we have,
H(w)
\H(wQ)\

Hence, 6Q

60

1 - e->2w
(1 _ rei( - ))(l - rc->^o _ u ; ))

602

|l-e-'2u'|2
= 1
(1 - r) [(l - rcos2w0)2 + (rsm2u;o) 2 ]

w>

^ ( 1 - r) 2 (l - 2rcos2u>0 +T 2 )
2|sinu;o|
162

4.72
From Q

and

COSQ

(n + l)u' 0
(n - l)u>0

+ cos0

2cos

cos-, we obtain
2
2
2cosnu>ocosu>o
cosu;on, it follows that
2co$woy(n) or equivalently,
2coswoy(n 1) y(n 2)

cos(n + l)u>o + cos(n - \)WQ

with y(n)
y(" + i) + y ( n - i )
V(n)

4.73
. <* + /?
o
2s:n-cos
, we obtain
2 "" 2
nwo and /? = (n 2)u>o, we obtain
2sin(n l)u>ocosu>o
.Asinwon, then
2coswoy(n - 1) - y{n - 2)
Asinwo, y(2) Asin2wo
0

sina + sinfl
when a
sinnifo + sin(n 2)wo
If y(n)
y(n)
Initial conditions: y(1)

4.74
For h(n

1 Z~1COSWQ

H(z
Hence, y(n
For h(n

1 2COSWQZ~1

2-2

=
=

2coswoy(n 1) y(n 2) 4- Az(n) Acosxvox(n 1)

y45nnu;ou(n)

1 2cosu>o2-1 + 2 - 2
2cosi^oy( 1) + y{n 2) + Ax(n) >isinti;ox(n 1)

2 ^mtuo

H(z
Hence, y(n

Acoswonu(n)

4.75
Refer to fig 4.63. yi(n) = Acosnwou(n),y^(n)

Asinnwou{n)

4.76
(a) Replace 2 by 2 . We need 8 zeros at the frequencies w = 0,j,^,^-,ir

H(z)

Hence, y(n)

1-02"8

Y(z)
X(z)
ay(n - 8)-1-x(n) - z(n - 8)
163

Hence,

z" 1

r
Lv

-ACOSWQ

\ 2rcos %L
''

Asin^

z" 1
-1

2w

Figure 4.63:
(b) Zeros at 1, e ^ T . e ^ ' i . e ^ ^ , - 1
Poles at a i . a i e ^ . a i e * ' * , < ! * * > , - 1 . Refer to fig 4.64.
(c)
2|cos4u;|
\ / l 2acos8tu + a 2
1
cos4u;>0
f - f a n "-i_aim8uL_
"n8w
cos4u; < 0

w=

Refer to fig 4.65.

4.77
We use FJL = lcycle/day. We also choose nulls of multiples of -fa - 0.071, which results in a
narrow passband of ib0.067. Thus, M + 1 = 14 or, equivalently M 13

4.78
(a)
H(w)

\H(w)\2

=
164

1 ~ ie- ;^,
1 -ae->w
(1 - cosw)7 + (sinw)2
(1 acosw)2 + (asinw)2

Unit circle

Figure 4.64:

magnitude of notch filter

Figure 4.65:

165

+ jy ~ lcosw

1 -+ a 2 2acosw
z
for all w
a2
1

Hence, |#(ti;)|
For the two-pole, two-zero system,
\w>

Hence, \H(w)\

(1 _ ieJWoe-iw)(\
- Ic-^c--'")
w
w
(1 - re-3 *e-i ){l
- reie-iw)

1 - }co8w0e->w + jre-J 2 "'

1 - 2rcosu>0e-->u' + r 2 e-> 2 w
1
-=

(b) H{z) = 1 . 2 r e o f l i ; o 2 -i t r r a I -a
Hence, we need two delays and four multiplies per output point.

4.79
(a)

60
_ 6*
200 * ~ 50
(l-e'ttz-1)(l-e-J'ttz-1)6o

60(l-2cos-^r"1 +2~2)

H(w)

26oe"',1'(costi; co )

\H(0)\

260(1 - cos^)

WQ

H(z)

=1

1
2(l-COs|f)

bo

(b)
=

60

(1 - re> M 2 - i ) ( l -

26Q(1-CO5||)

1^(0)1

1 - 2rcos + r 2
1 - 2 r c o s f f + r2
2(l-cos|)

re-Jz-1)

= 1

60

h(n)

{/i(0),//(l),M2),M3)} where fc(0) = -fc(3), Ml) = - M 2 )

Hence, Hr{w)

2(/(0)st'n-y + / i ( l ) s t n | )

2fc(0)atny + 2 M D n | ) = 5

4.80

Hr{l)

166

=
1.85h(0)+ 0.765/i(l)
-0.765/i(0) + 1.85/(1)

Mi) =

2h(0)sin^-

+ 2h(l)sin^)

= 1

1_
2
1
0.56,/i(0) = 0.04

4.81
(a)
H(z)

= b0

H{w) = fc0
\H(w)\
5T,

l#(^)l

= 60
=

(1 - z'l){\ + z~l)(l - 2cos*z-1 + z~ 2 )

(1 - \.6cos^-z~l + 0.64z" 2 )(l - \.6cos%z-1 + 0.64z" 2 )
(2je-^5inti;)(2e-^)(co5u; - cos^-)
(1 - 1.6cos^e-)w
+ 0.64e"-' 2u ')(l - 1.6cos^e-> u ' + 0.64c~J2u')
4|sinu>||cosu; c o s ^ . |
|1 - 1 . 6 c o s ^ e - > + 0.64e-> 2u, ||l - 1.6cos*-e-Jw + 0.64e-> 2w |

l=6o = 0.089

(b) # ( z ) as given above.

(c) Refer to fig 4.66.
The filter designed is not a good approximation of the desired response.

_*

'

0.06

0.1

0.15

02

0.25

0.3

0.1

Figure 4.66:

4.82
Y(w) = e-jwX{w)

dX{w)
dw

(a)
Forx(n)

6(n),X{w)
167

= \.

0.4

0.45

0.5

Hence,

dX W

} ^
aw

h(n)

0, and Y(w) = e"-"

i- /

y(u;)e^ n du;

27T 7 _ T

2W-,r

it;

1
^(n-l)!*
2?rj(n - 1)
sirnr(n 1)
*(n-l)

(b) y(n) = x(n - 1) jnx(n). the system is unstable and time-variant.

4.83
H(w) =

*(n)e-;um

n = -oo

1,

\w\ < We

0,

wc < \w\x

n = oo

n = oo

m= oo

=
Hence,

CM-I

/f(2u;)

' H<^H>x-^

4.84
y(n) = x(n) x(n) * /(n) = [S(n) /(n)] * x(n) The overall system function is 1 - H(z) and the
frequency response is 1 H(w). Refer to fig 4.67.

4.85
(a) Since X(w) and Y(w) are periodic, it is observed that Y(w)
y(n)=e>'"z(n) = (-l)"z(n)
(b)x(n) = ( - i r y ( n ) .

4.86
y(n) = 0 . 9 y ( n - l) + 0.1x(n)
168

= X{w x).

Therefore,

H(w)

l-H(w)

=
0

l-H(w)

H(w)

=
w

Figure 4.67:
(a)
H(z)

Hbp(w) = H(w - - )

0.1
1-0.92-1
0.1
l-0.9e-^w-?)
0.1
1 - j0.9e-J M '

(b) /i(n) = 0.1(0.9e>*)u(n)

(c) Since the impulse response is complex, a real input signal produces a complex-valued output
signal. For the output to be real, the bandpass filter should have a complex conjugate pole.

4.87
(a)
Let g(n)
Then, G(w)

D =

nh(n)
dH(w)
; dw

\g(n)f
2TT

y_,

\G(w)fdw
169

-!- /
1

dH(w)
dw
Therefore,

But

G(w)G'(w)dw
r

(dH(w)\m
dw

\.dH(w),

1^ * > i * 2 ?
dH(w)
dw

'' H{\

+ \H(w)\'

de(w)
dw

i2

dw

(b) D consists of two terms, both of which are positive. For \H(w)\ ^ 0, D is minimized by
selecting Q(w) 0, in which case the second term becomes zero.

4.88
y(n) = ay(n - 1) + bx(n), 0 < a < 1
H(z) =

6
1-az"1

(a)
tf(u,)

|//(0)|

1 - ae-J

1*1

6 =

(l-a)

(b)
1^)1
=> 262

=
=

cosw

62
1
r + a - 2acosu, " 2
1 + a 2 - 2acosw
2

[l + a 2 - 2 ( 1 - a ) 2 ]
2a

= <4.-!-.)
ttf3

COS

4a - 1 - a1
2a

(c)

w3

Let / ( a )

Then/'(a)

co5"l(l-^-^-)
2a
:
(
q
i
)
1 2a
a2-l
2a 2
1-a2
>0
2a 2

Therefore / ( a ) is maximum at a = 1 and decreases monotonically as a 0. Consequently,

u>3 increases as a 0.
170

(d)
6 =

(l-a)
2
- i 4a - 1 - a
U>3 =
COS ;
2a
The 3-dB bandwidth increases as a > 0

4.89
x(n) + ax(n M),a > 0
H(w)

V 1 + 2acoswM + a 2
_j asinwM
tan
1 -I- acoswM

\H(w)\
e(w)

ae'iwM

1+

Refer to fig 4.68.

NMO, dpha-0.1
2
1.5
?

|
0.5
n
0

0 06

0.1

0.15

02

025
>l

0.3

0.35

Figure 4.68:

4.90
(a)

Y(z)

\{X{z)^z- 'lX{z)]

H(z)

Y(z)
X(z)

^(1 + ^ 1 )
*+ l
2z
171

0.4

0.45

0.5

Zero at z = 1 and a pole at z 0. The system is stable.

(b)

Y(z) = U-X(z) + z"lX{2)]

H(z)

2
Y(*)

X(z)

= (-1 + , - )
2- 1

1z
Zero at 2 = 1 and a pole at z = 0. The system is stable.
(c)
Y(z)

^(1 + ^ " 1 ) 3
1 (1 + 2) 3
8

23

Three zeros at z = 1 and three poles at z = 0. The system is stable.

4.91

Y(z)

X(z) + bz-2X(z)

For 6 = l,H{w)

l + ej2u'+e--'4u'

(1+2005^)6-^

|l + 2cosu;|

\H(w)\

Refer to fig 4.69.

^ ~ \ TT - u;, 1 + 2co*u; < 0
b=-l,H(w)
=
l-e->w+e-'iw
= (2costi; l)e-jw
\H{w)\ =
\2cosw-l\
* '

\ ir u;,

- l + 2cosu;<

Refer to fig 4.70.

4.92
y(n) = z ( n ) - 0 . 9 5 x ( n - 6 )
172

z-4X(z)

Figure 4.69:

o
-1
-2

05

1.5

2.5

Figure 4.70:

173

3.5

(a)
Y{z)
H{z)

=
=

z6
2

=
=

X{z)(l-0.9bz-6)
(l-0.95z~6)
ze - 0.95
z6
0.95
(0.95)*c^ 2 '*/ 6 ,Jb = 0 , 1 , . . . , 5

6th order pole at z = 0. Refer tofig4.71.

(b)Refer tofig4.72.

=(P.95J,(

Figure 4.71:
(c) Hin(z) = j r ^ . r = (0.95)*. Refer tofig4.73.
(d)Refer tofig4.74.

4.93
(a)
r

H(z)

l _

-i

- l _

- 2
r

-l

(l_ii^lz-i)(l-L=jZI2-i)
174

Figure 4.72:

r=(0.95^/6

Figure 4.73:

175

Figure 4.74:
l

7s

If ROC is ^ _ J . < M < ^ t l ,

From H(z),

If \z\

>

h(n)

1-

"7S

is ROC, then

1 UV5
v'T
2

1 ,l->/5vn
(*)
v^
2 j

then

Mn)

= -7!("T-)u(n)-75(^_)"("n" '

If \z\

<

h(n)

the difference equation is

y(n)

\/5 1
1 - --- is ROC, then
_ 1

1^5

1_ 1 ^ V 5

u(-n-l)

y(n-l) + y(n-2) + x(n-l)

(b)

tf(z)

The difference equation is

y(n)

tf(z)

If ROC is \z\

>

h(n)

If ROC is |z|

<

1_

1
a
e-* z-*

e - 4 a y ( n - l ) + *(n)
(1 - e - ^ " 1 j(l - e}ie-az-l){l
l-c-az~1
1, then

' l-jc"^-1

+ e-az"1)(l +>e-flz-1)

+' l + e - z - 1

l+

i [ l + (;) n + ( - l ) n + ( - j ) n ] e i i ( n )
4
1, then

AW = - j [i + o r + (-i)B + (-i)n] ""(--1:

4

176

je-'z-1

4.94
Y{Z)
X{Z)

=
=
=

Therefore, H{z)

l - 2 " 1 + 3 2 - 2 - 2 - 3 + 62- 4
(1 + 2" 1 + 22~2)(1 - 22" 1 + 32"2)
l + 2"1+22"2
Y(z)
X(z)
l - 2 2 _ 1 + 32- 2

h(n) = {}.-2,3J
4.95
y( n )

2^n_1^

x(n)

(\)nu(n)

+ r n

( ^

X(z)
1

i-j*->
X(Z)

1
' v^
*x,(*)

( 1

>i2-l

_l2-l)

) ( 1

X(2)X(2- )
1
(1-1,-1)(1-1Z)

-42-1
(l-i2-l)(l-42-l)

Hence, r r r (n)
fl^z)

16
1
15 1 _ I z - i

^(1)"^)

16
1
15 1 - 4 2 - 1
^(4)nu(-n

- 1)

H(z)H(z-1)
1
( 1

_ i

- i )

( 1

_ i

-22(l-j2-i)(l-2z-)
4
1
4
1
1
3 1 - iz"
3 1-22"1
Hence, rhh(n)

!(I) U () + l ( 2 ) n ( - n - 1)

*(*)

X(z)y(2- 1 )
1
( 1

_I

-i

) (

l_i

2 )

(l_i

177

16
1
16
1
128
17 1 - 2 * - 1 + TT^
15 1 - 4 ^
* "- 17 + 105 1 -

\z~l
4

Hence, rxy(u)

Ryy(z)

15(2)u(-n - 1) - ~ ( 4 ) " t i ( - n - 1 ) + i H | ( i ) % ( n )
Y(z)Y(z~1)
1
(l-l2-i)(l-lz->)(l-\z)(l-$z)
64
1
128
1
64
1
2 1 1 - 2 Z " 1 + rrz-.
105 1 - 4:r
*"1 + 2 1 1 - i z - i

Hence, r y y (n)

128
105 1 - i z - i

51(2)(--i)-i^(4)"(-n - 1)+^(i)"u(n) - ^(i)(n)

4.96
(a) M") = { 1 0 1 , 9 , - 7 , - 8 , 0 , 5 , 3 }
The roots(zeros) are 0.8084 jO.3370, -0.3750 >0.6074,-1.0, -0.7667
All the roots of H(z) are inside the unit circle. Hence, the system is minimum phase.
(b) h{n) = {5,4, - 3 , - 4 , 0 , 2 , l}H(z) = 5 + Az~l - 3z~ 2 - 4z'3 + 2z~ 5 + z ' 6
The roots(zeros) are 0.7753 j0.2963, -0.4219 j'0.5503, -0.7534 jO. 1900
All the roots of H(z) are inside the unit circle. Hence, the system is minimum phase.

4.97
The impulse response satisfies the difference equation
N

Y,akh(n-k)

(n),a0=l

aoh(0) = 1

k=0
N

n = 0,=>]Ta*/i(-*)
k=0

a0

n = l,=>a 0 /i(l) + a 0 /i(0)

0!

n = #,=> a0h(N) + aih(N - l)+ ... +aNh{0)

=>

1
A(0)
0
-a 0 /(l)

/i(0)

-/(!)
/i 2 (0)

yields a*

It is apparent that the coefficients {a n } can be determined if we know the order N and the values
/(0),/(l),.. .,h(N). If we do not know the filter order N, we cannot determine the {a}.

4.98
h(n) = b06(n) + bi6(n - D) + brfin - 2D) (a) If the input to the system is x(n), the output is
y(n) 6ox(n) + bix(n - D) + bzx(n 2D). Hence, the output consists of x(n), which is the input
signal, and the delayed signals x(n D) and x(n 2D). The latter may be thought of as echoes
of xln).

178

(b)
H(w)

b1e-iwD+b2e->2u,D

bo +

bo + b\C0swD + b2Cos2wD j(bisinwD

|i/(tx;)|

yj bo2 + bf + b22 + 261(60 + 62)cosu;D + 26062cos2u;D

t7^Un

tan

+ &2s"*2u>D)

60 + b\coswD + &2COs2u>D
(c) If |60 + 6 2 | < < |6i|, then the dominant term is b\e~iwD
2

\H(w)\ = yjb0 + &! +

&22

and

+ 2&1(60 + 62)cosu;D

and |//(u')| has maxima and minima at tx; = j}irt k 0 , 1 , 2 , . . .

(d) The phase Q{u>) is approximately linear with a slope of D. Refer to fig 4.75.

4i

\\
0

\
0 05

01

\
0 15

0.2

0 25

\
0 35

0.3

0.4

0 45

05

Figure 4.75:

4.99
v

'

A(z)

l + az

(a)
1 + (6 - a)*" 1 + (a 2 - a6)z" 2 + (a 2 6 - a 3 ) 2 " 3 + (o 4 - a 3 6)r" 5 +
Hence, /i(0)
A(l)
h(2)
A(3)
A(4)

1,
b-a,
a2 ab,
a26-a3,
a 4 - a3b
179

(b)
y{n) + ay(n- 1)
For x(n)
h(n) + ah{n- 1)
Multiply both sides by h(n) and sum. Then
rhh(0) + arhh(l)
r/h(l) + arhh(0)
rhh(2) + arhh{l)
rh/>(3) + ar h h (2)
By solving these equations recursively, we obtain
rhh{0)
rfcfc(l)
rhh(2)
r**(3)

x(n) + 6x(n - 1)
6(n) +

b6{n-l)

h(0)+bh(\)
bh(Q)
0
0
62-2a6+l
1-a2
(a6-l)(a-6)
1-a2
(a6-l)(a-6)
"a
IT?
(
a
6
l)(a-6)
2

4.100
x(n) is a real-valued, minimum-phase sequence. The sequence y(n) must satisfy the conditions,
y(0) = x(0), |y(n)| = |x(n)|, and must be minimum phase. The solution that satisfies the
condition is y(n) = ( l) n x(n). The proof that y(n) is minimum phase proceeds as follows:

Y(z)

J^y(n)z-n

=
n

(-l)"*(n)z-
n
n

X{-z)

This preserves the minimum phase property since a factor (1 a* -1 )(l + az~ 1 )

4.101
Consider the system with real and even impulse response h(n) = {ij, 1. j } and frequency response
H{w) = 1 + %cosw. Then H(z) = z~l(\z2 + z + \). The system has zeros at z - - 2 y/Z.
We observe that the system is stable, and its frequency response is real and even. However, the
inverse system is unstable. Therefore, the stability of the inverse system is not guaranteed.

4.102
(a)
g{n)
/(n)

=
=

h(n)*x{n)=>G{w)
= H(w)X(w)
h(n)*g{-n)=>
F(w) =
H(w)G(-w)
180

Y(w)

F(-w)

Then, Y(w)

H(-w)G{w)

=
=

H(-w)H(w)X{w)
H"(w)H{w)X(w)

\H(w)\2X(w)

But Ha(w) = |/f(t^)| is a zero-phase system.

(b)
G(w)
F(w)

Y(w)

=
=
=
=

H(w)X(w)
H(w)X(-w)
G(w) + F(-w)
H(w)X(w) +
H{-w)X(w)
X(w){H(w) + H'(-w))
2X(w)Re(H{u>))

But Hb(w) 2Re {H(w)} is a zero-phase system.

4.103
(a) Correct. The zeros of the resulting system are the combination of the zeros of the two systems.
Hence, the resulting system is minimum phase if the inividual system are minimum phase.
(b) Incorrect. For example, consider the two minimum-phase systems.
Hi(z)

1 - z _ 1
1

and H?(z)

3*

-2(1-r^-1)
1 - iz~l
1

3Z
6*

Their sum is H\{z) + #2(2)

lTf"*'
52

1 -

bich *s n o t minimum phase.

4.104
(a)
\H{w)\'.2

7 COSli;

10

H(z)H(z-l)\,

=
Hence, //(z)ff(z- 1 )

_=

!-*<* + *-

10
9

11-

')
2Z

(b)
l"(to)| 2
H(z)H(z-1)

=
=

2(1 - s )
1 + a 2 -- 2acost^
2
2( 1 - a )
1 + a 2 - a(z+
181

z~l)

_u
H{z)H{z-')
=
Hence, *(,)
V
'
or H(z)

2(l + a ) ( l - a )
(1 - az~l)(l

^ Z p
1-az"1

1 az

- az)

4.105
#(z)

- O.Se-Mh-1)^

(1 - O ^ ' V ^ l

(1 + 0.64z" 2 )(l - 4 = z _ 1 + 2.25z" 2 )

v2

- 1.5e'* /4 < r- 1 )(l - 1.5 C -''' 4 *

(a) There are four different FIR systems with real coefficients:
Hi(z)

(l+0.64r-2)(l-4=^'1+2252-2)
v2

H2{z)

(H-0.64r~2)(l--?=^'1+2.25z-2)
v2

(l+0.64z-2)(l-4=^"1+2.25z-2)
v2
2
# 4 (*) = ( l + 0 . 6 4 z ~ ) ( l - 4 = * ' 1 + 2 . 2 5 z _ 2 )
v2
# ( z ) is the minimum-phase system.
H3(z)

(b)
#!(*)

l-4=2'
\/2
3

+ 2.89z-2-i^z-3+1.44z-4
v^
1 92
1

{
H2{z)

0 . 6 4 z 2 - ^ z + 2.44--^=z-1+2.25z-2

1 92

0.64,-^,2.44,-^,2.25}
H3{z)

1 Q2

2.25z2-4=^ + 2 . 4 4 - ^ z " 1 + 0 . 6 4 z - 2
V2
y/2
3
1 92
1

2 25

HA{z)

- 'V5' 2 . 44 '-^"' 0 - 64 }

1 92
3
1 . 4 4 z 4 - - W + 2.89z2--?=z + l
v2
v2
1 92
3
1

(0

Ei(n)

{1,5.5,13.85,15.70,17.77}

E2{n)

{0.64,2.48,8.44,12.94,18.0}

3(n)

{2.25,6.75,12.70,14.55,14.96}

E4{n)

{1.44,3.28,11.64,16.14,17.14}

Clearly, /ia(n) is minimum phase and /2(n) is maximum phase.

182

4.106
H z

w1

() =

i + ElU***-*
H(\~lz)

(a) The new system function is H'{z)

H'(z)

jJ-

1 + Ef.i ***-*
If pk is a pole of H(z), then \pt is a pole of H'(z).
Hence, A < i ' a | is selected then |p*A| < 1 for all k and, hence the system is stable.
(b) y(n) = - E ? = 1 akXky(n - k) = x(n)

4.107
(a) The impulse response is given in prlOfig 4.76.
(b) Reverberator 1: refer to fig 4.77.

12

06

"S
i>06
A

I
04

0.2
0

500

1000

1500

2000

2500

>n

Figure 4.76:
Reverberator 2: refer to fig 4.77.
(c) Unit 2 is a better reverberator.
(d) For prime number of D\, D?, D3, the reverberations of the signal in the different sections do
not overlap which results in the impulse response of the unit being more dense.
(e) Refer to fig 4.78.
(f) Refer to fig 4.79 for the delays being prime numbers.

4.108
(a) Refer to fig 4.80.
(b) Refer to fig 4.81.

183

fnpuls* rwporwatorurwt1

05

IlliiL Hi
100

J M ,,iil.

200

iH^Umy

300
400
>n
imputed rMportMtorunit2

Figure 4.77:

phtM ratponM tor unll

4
> w
ph*M rponMtorun2

Figure 4.78:

184

500

600

1.2

08

500

1000

1500
>n

Figure 4.79:

Figure 4.80:

185

2000

2500

Figure 4.81:

4.109
10kHz
20kHz
10k
= 0.5
20k
7.778fc
= 0.3889
20A:

B
F.
*i

8.889Jb
23

20k
6.667Jb

24

H(z)

201;

= 0.4445
= 0.3334

(z - 0.5)(* - 0.3889)(z - 0.4445)(z - 0.3334)

(b) Refer to fig 4.82.

(c) It satisfies the objectives but this filter is not recommended in a practical application because
in a speech application linear phase for the filter is desired and this filter does not provide linear
phase for all frequencies.

4.110
Refer to fig 4.83.
Practical;
r = 0.99
r = 0.9
r = 0.6

wr = J
wr = I
wr = 0

Bandwidth = p g = 0.0245
Bandwidth = | | = 0.49
Bandwidth = 1.1536

Theoretical:
186

2-

a-

J
0.5

1.5

2.5

3.5

Figure 4.82:

r = 0.99,
r=0.9,

wr = |
wr = |

Bandwidth = 2(1 - r) = 0.02

Bandwidth = 2(1 - r) = 0.2

For r very close to 1, the theoretical and practical values match.

4.111
(1 - 0.9^ 4 * z - 1 ) ( l - 0.9e-' ^ " ^ ( l - l.Sei-69z~l){l
- 1.5e" j0 6 "z
B(z)
A(z)
(z - 0.9eJAw)(z - 0.9e^ A *)(z - 1.5eJ **){z - 1.5e-' 6,r )

H(z)
H(z)

(z-0.9ej04*)(2-0.9c->04,r)
(2-1.5c>06,r)(z-1.5e--'06,r)

Let Si (2)
B 2 (r)
A(z)

#min( 2 )

#ap(z)

24

Bi(z)B2(z)
A(z)
(z - 0.9e>Av)(z - 0.9c"-'0 ^ K z " 1 - 1.5e>-6w)(z-1 Bi(x)
B2(z-1)
(z1.5eJ6')(r-1.5e->06T)
1
( z - - \.beJ**){z-1 - 1.5e->06)
187

l.be'i0-6')

40
30

! *
' 10
0-10-20

-3

-2

0
>w(md)

Figure 4.83:
H&p(z) has a flat magnitude response. To get a flat magnitude response for the system, connect
a system which is the inverse of #niin( z )> ' e '
Hc(z)

1
H

nnn(z)

(z - 0.9e> 4 *)(z - 0.9e--> ^ ( z - 1 - 1.5e'0 6 *)(z~ 1 - 1.5e--> 6 T )

(b) Refer to fig 4.84 and fig 4.85.

188

p o a w a r o plots lor Hc(z)

9015

270
pote-zaro p t o * for compensated eyatem
901-5

mag ol compensated system

mag tor Hc(z)

0
> w(rad)

-2

/\

l/l i
0
->w(rad)

phase of compansatad system

phase tor Hc<z)

0
> w(rad)

Figure 4.85:

189

Chapter 5

5.1
Since x(n) is real, the real part of the DFT is even, imaginary part odd. Thus, the remaining
points are {0.125 + jO.0518,0, 0.125+ jO.3018}

5.2
(a)
x 2 (/)

x 2 (/)

Therefore, x i ( n ) ( j ^ ) x 2 ( n )

x2(/),

0</<AT-l

x2{l + N),

- (JV - 1) < I < - 1

n(y/),

0</<7

5 t n ( y ( / + 8)),

n(^|/|),

Y^ x 2 (n - m)

-7</<-l
|/|<7

m=0

'n( Y |n|) + sn( y | n - l|) + .. . + i n ( ~ | n - 3|)

{1.25,2.55,2.55,1.25,0.25,-1.06,-1.06,0.25}

co${--n),

3*
-cos(-n),
8

[2u(n)-l]cM(yn),

(b)
x 2 (n)

Therefore, X!(n)(V)x 2 (n)

3ir
o

f -J

0< /< 7
-7</<-l

|n|<7

x 2 (n - m)

ni*0

{0.96,0.62,-0.55,-1.06,-0.26,-0.86,0.92,-0.15}

(c)

for(a)Xi(fc)

]Txi(n)e--'in
n=0

191

{ 4 , 1 - ;2.4142,0,1 - >0.4142,0,1 + J0.4142,0,1 + ;2.4142}

similarly,
X2{k)

DFTofx1(n)(7)x2(n)

{1.4966,2.8478,-2.4142,-0.8478,-0.6682,-0.8478,
-2.4142,2.8478}
Xi(k)X3{k)

{5.9864,2.8478 - J6.8751,0, -0.8478+ jO.3512,0,

-0.8478 - jO.3512,0,2.8478 + J6.8751}

{1.3333,1.1612 - jO.2493,0.9412 - jO.2353,0.8310 - jO.1248,

*2(*)

0.8,0.8310 + jO.1248,0.9412 + j'0.2353,1.1612 + jO.2493}

{ 1 . 0 , 1 . 0 + j 2 . 1 7 9 6 , 1 . 0 - j 2 . 6 l 3 1 , 1 . 0 - jO.6488,1.0,
1.0 + jO.6488,1.0 + j'2.6131,1.0 - j'2.1796}

Consequently,
DFTofx!(n)(T)x2(n)

For sequences of part (b)

Xi{k)

X1{k)X2(k)

{1.3333,1.7046 + j'2.2815,0.3263 - J2.6947,0.75 - j'0.664,0.8,

0.75 + jO.664,0.3263 + j'2.6947,1.7046 - J2.2815}

5.3
x(Jb) may be viewed as the product of X(k) with
E V M - I

*>-

'

\ 0,

0<k<ke,
Ke<k<N-kc

N-kc<k<N-l

F{k) represents an ideal lowpass filter removing frequency components from (Kc + 1 ) ^ - to r.
Hence x(n) is a lowpass version of x(n).

5.4
(a)

*i(*)

y [*(*-!)+ *(* + !)]

alsoX2(Jb) =

?L[6(k-l)-6{k+l)}

SoX3(k)

and x 3 (n)

Xl(k)X7{k)

^ j [*(*-!)-(* + !)]

yn(-n)

(b)
Rtv(k)

Xi(k)X;{k)
4j

=>

rxy{n)

-
192

sin(-n)

(c)

R**(k)

Xx(k)Xl{k)

= [ * ( * - ! ) + (*+!)]
=>

r r i (n)

-2-cos( n)

(d)
AT2

[ 6 ( * - l ) + 6(Jb+l)]

- / >
yy( n ) =

co

<2* \
*(^n)

5.5
(a)
AT-1

n=0

- 2N
4
N

(b)

2>(n)*5(n) = - l g ( ^ ^ + e - J ^)( c ->^-^^)

n= 0
JV-1

n=0

N-1

(c)IIo*i(n)*5(n)= 1 + 1 = 2

5.6
0.42- 0.25 (e>T&Tn + e-jT&n)

u;(n)

N-l

HV-1

N-l

TN-1

0.42 ] T e - j ^ n k - 0.25
n=0

e^'^"*-'^* +

n=0

n=0
JV-l

n
n
+ 0.04 V* e J ' ^ r r ^ * + V
Ln = 0

+ 0.04 ^ ^ T " + c ->T^r n )

n=0

0.42Ar<5(Jt)

193

c ->7fe-" c -

J nk

n
e -ii*?T e -.Jfc*

-0.25
+0.04
0.42JV$(Jb)
-0.25

1 - cos(j^)

- cosWjfc

+ )) + e o . ( ^ )

l-co5(27r(^ + ^))

+0.04
l-co(2*(]?iT + J7))

5.7
Xe(fc)

X]ix(n)(e'-2=^+e-'2=^)e-a*1

n=0
AT-1

n=0

n=0

- * ( * - *o) m od* + 2X(k

similarly, X,(Jr)

*)mod*

5.8
y(n)

=
=

zi(n)4 2 (n)
3
*i ( m )mod4 z 2( n ~

)mod4

m=0

{17,19,22,19}

5.9
Xi(*)
X2{k)
=>X3(k)

=
=
=
=

{ 7 , - 2 - j , l , - 2 + j}
{ l l , 2 - j , l , 2 + ;}
Xx(k)X2(k)
{17,19,22,19}

5.10
x(n)
x(n)x'(n)
194

N-l

E = *(n)x
n=0

n=0

^2JV
4
TV
2

xi(n)

x(n-5)modg

X(t)c-ji^

x 2 (n)

x(n-2)mod8

X2(jfc)

X(*)c"-' 1 V iL

s(Jb)

W>X(k)

s(n)

T(-l)lX(i)^
6

t n

AT = 6

*=o

k{n 3)
= =ii,*(*w;
o
6

s(n)

=
=

(n-3)mod6
{3,4,0,0,1,2}

ipn{X{k)Ym{k)]
i [IDFT {X(Ar)} + IDFT {**(*)}]
-[x(n) + x-(-n)mod]
f ... x(l) + x(5) x(2) + x(4)
*(4) + x(2) x(5) + x ( l ) \
|*(0),
2
'
2
'X(3)'
2
'
2
/

3,3,3,i
195

(c)
v(n) =

I D F T [ - ( * ) _ *'MX
2;

By similar means to (b)

v(n) =

|o,-ij,j,0,-i, | j |

2N-1

3/V-l

5.13
(a)
Xx(k)

x(n)W#
n= 0
3N-1

N-l

= "W^ft + *(U)W3N + *(")*&

r=0

n=N

n=0

n=0

E*( n )[ 1 +

3+

n=2N

n= 0

3*]w;;*

n=0

2k>

(l + Wj + W s " ) ^ ^ )

(b)
Xi(i)
X3(*)

=
=

2 + W2*
2 + W6* + 2tV|* + W63* + 2W*k + W|*

(2 + W$) + W?*(2 + W2*) + W64*(2 + w}

5.14
(a)
y(n)

=
=

xl(n)\5jx2{n)
{4,0,1,2,3}

2
3
4
0
1

1
2
3
4
0

(b) Let x3(n) = {x 0 ,xi,.. ,x 4 }. Then,

0
1
2
3
4

4
0
1
2
3

3
4
0
1
2

" *0 "

*1
*2
*3
X4

"1 "
0
0
=
0
0

Solving yields sequence x 3 (n) = {-0.18T,0.22,0.02,0.02,0.02}.

196

5.15
Define H\{z) = H~1(z) and cpnesponding time signal /ii(n). The use of 64-pt DFTs of y(n)
and hi(n) yields x(n) = y{n)\^)hi(n)
whereas x(n) requires linear convolution. However we
can simply recognize that
X{z)

Y{z)H,{z)

sox(n)

=
=

Y(z)-0.bY(z)z~1
y ( n ) - 0 . 5 y ( n - 1),

with y ( - l )

0 < n < 63

5.16
N-l

Y,h(n)e~j*kn

H(k) =

n=0

4
4

G(k)

H(k)

= >+(;'-i,0 + (r- , , *)' + f 4 1 6 - 4 j 4 16 + 4J

\
repeatiot,mes

= l3'-n-'y-n-'

= ils:}+{(^)^-5:}+{i^2:}
+

{(^)^E}]

*o-l

where Y^

2_. *
i=0

But J^

1, yielding

9(0 = J

4
3

(U^)+|+(1^)]

256
255
197

0(*o)

s(2*o)

,,.

4
3+J

64
255
1

16
255
1 T4

(^H->(**)]
V 17 J + 5
./16-4A

^ 17 ) \
4

./l6 + 4j\l

and ^(n)

Therefore, g(n) * /i(n)

255
0 for other n in [0,4fc0)

256
1
= < - - , 0n, 0n , . . . , 0 , . . . , 0 , . . . , 0 , . - ^ 7 7 , 0
255
T
T
T
255
*o

2*0

3*o

4*o

y(.) represents a close approximation to an inverse system, but not an exact one.

5.17
X(k)

z(n)e->^
n=0

{6, -0.7071 - jl.7071,1 - j , 0.7071 + jO.2929,0, 0.7071 - jO.2929, 1 + j ,

-0.7071 + jl.7071}
{6,1.8478,1.4142,0.7654, 0,0.7654,1.4142,1.8478}
IX(k)

j o , -1.9635, ,0.3927,0, -0.3927, j , 1.9635 j

5.18
x(n) = J2 Hn-iN)
i'=-oo

y( ) = 53M"Mn-m)
= J>(m)

^26(n-m-iN)

j2h(*-iN)

Therefore, y(.) is a periodic sequence with period N. So

N-l

Y(k) =

'EyinWtT
n=0

H(w)\ws#k
198

5.19
Call the two real even sequences xe\(.) and x e 2 (.), and the odd ones x0\(.) and x o 2 ( ) (a)
Let x c (n)

Then, Xe(k)

whereXel(*)

[x e l (n) + x 0 i(n)] + ; [x e2 (n) + x o2 (n)]

DFT{xel(n)}+DFT{xol(n)}+jDFT{xe2(n)}+;DFT{xo2(n)}

[Xel{k) + X0l(Jb)] + j [X e2 (*) + * ( * ) ]

R e [ ^ * ) ] + Re[X,(-*)]
Re[X c (*)] - Re[X,(-fc)]

^oi(*J
v

,.>

Ae2()

Xo2(Jb)

Im[Xt(fc)] + Im[X c (-*)]

-

lm[Xc(k)} -

lm[Xc(-k))

(b)
s,(0)
-s,(7V-n)

=
=

x<(l)-*i(tf-l) = 0
- x , ( A r - n + l) + x , ( A r - n - 1 )

=
=

Xi(n+1)-Xi(n-1)
Si(n)

(c)

x(n)

[i!(n) + 83(n)] + j (x 2 (n) -I- sA{n)]

The DFT of the four sequences can be computed using the results of part (a)
AT-l

Fori = 3,4,Si(*)

J28iWWNn

=
n=0
JV-1

= JWn+l)-^-!)]^"
n= 0

W-kXi(k)-W^Xt(k)

=
=
Therefore, ^ ( i b )
X(k)

2J8in{jfk)Xi(k)

<

3(*)

2jsin(ftk)

'<*)

(d) JV3(0) and X 4 (0), because 5in(fr *) = 0.

5.20
S-l

X(k) = 5>(n)W#
n=0

199

= 'x( n )<"+x(n + )< ( " + * )

n=0

n=0

-1

[*(n)-x(n)W*]W#
n=0

If * is even, W2fc = 1, and X(k) = 0

(b) If k is odd, W2* = - 1 , Therefore,
X(Jb) =

2x(n)W#
n=0

2^x(n)^
n=0

For* =
X(2/+l)

2/ + 1,

l= 0,...,--l

2Y,z{n)W%W%
!L _ p t DFT of sequence 2x(n)W

5.21
(a) F, = FN = 25 = 6000 samples/sec
(b)
F.
1
6000

s= *
=

M
6000
50
120 samples

(c) LT = 6MoXl20 = 0.02 seconds.

5.22
x(n)

X(k)

Lifr

i e -J-* f

>(n)e-'**
r=0
JV-1 -

JV-1 ,

2"

' ^

n=0

0<n<N,

"=0

56(Jfc-l) + 56(Jb-9),
200

0<Jb<9

AT = 10

5.23
(a) A'(*) = E l ' o 1 S ( n ) e - ^ f c " = 1 ,
(b)

0 < it < JV - 1

N-l

X(k) = *( " n0)e-''**B

n=0

e"^*"0.,

0<*<W-1

(c)

X(Jt)

^anc-J^*

n=0

n=0

l-ae-**k
1-a"

(d)
X(Jb)

^ c - ^

k n

n=0

1-c-J^*
(e)
N-l

X{k) = V'#"* e - J '#* n

r=0

n=0

N6(k-k0)

(0
x{n)

Ie>^ n f c + I e - > ^ n * 0

From (e) we obtain X(k)

[6(k - k0) + 6{k - N + k0)]

(g)
i(n)

eJ^n*0-

Hence X(k)

^ [6(k - k0) - 6{k - N + kQ)}

201

e'j^nko

N-l

X(k)

5 ^ x(n)e"-'^' nfc ( assume N odd

n=0

e " ^ 2 * + e " > # 4 k + . . . + e-'"*<"- ! >*

1-"'**
1

N-l

<) = ] ? *(*>* *"*

fc = 0

N-l

E *(*)* ^"* = Nx(n)

fc=0

X(0) + X ( l ) + X(2) + X(3)

X(0) + X ( l ) e ' * + X ( 2 ) e ' * + * ( 3 ) > *

X(0) + X ( l ) e J T + X ( 2 ) e j 2 T + X ( 3 ) e j 3 , r

=
=

8
12

X(0) + X ( l ) e J ' ^ + X ( 2 ) c J 3 * + X ( 3 ) e ' s r

1 1 1 1
1
i -1
-j
1-1
1-1
1 -;' - 1
j

"4
8
12
4

' X(0) '

X(l)
X(2)
. *(3) .

X(0)

X(0) '
X(l)
s
f
X(2)
X(3) .

=
n=0

7
X(l)

=
n=0

-2->
JV(2)

x>(n)e-'-'
1

X(3)

-2 + j
202

' 7
-2-;
1
-2 + j

5.25
(a)
X(w)

=
=
=

,-jrt>n

X (n]

e>2w + 2e>w + 3 + 2e" i u ' + e ^ 2 "

3 + 2cos{2w) + 4cos(4u;)

(b)
V(*)

J>(n)e-''*"*
n=0

it
IK
3 + 4cos(-Jb) + 2cos( * )

=
(c)V(t) = X(ti;)L=a5*ss^

This is apparent from the fact that v(n) is one period (0 < n < 7) of a periodic sequence
obtained by repeating x(n).

5.26
Let x{n)

Hn + IN)

J=-00

Hence, x(n) is periodic with period N, i.e.

x(n)

=
=

ThenX(ib)

1,
0,

n=
0,N,2N,...
otherwise

N-l

J I z ( n ) e " ; ^ n f c = 1,
n=0
N-l

and-x(n)

^AT

J T 6(n + IN)

-^ ] T

X ^ * " *

N-l

Hence,

J=-oo

"

'^nfc

fc=0

5.27
(a)
Af-l
n=0

n=0

Now X(u-)

5Zx(n)e- J U ' f l ,

203

0 < it < N - 1

Af-1

=:
n=0

Af-1

y(t)
Therefore, Y{k)

(b)

Y(k) =

XWI..U*

y(t) = XHL = ^ t
=

*(*),

* = 2,4,...,AT-2

(c)
Xi(k)

X(* + l)

=>X!(r)

x(n)e-^n

*(n)W

xx(n) + Xi(n + y ) ,

0,

^(ib)

Y(i),

Let y(n)
T h e n X ( * + l)

elsewhere

Jb = 0 , 2 , . . . , J V - 2

AT
where V(it) is the -pt DFT of y{n)

5.28
(a) Refer to fit 5.1.
(b)

*<)-'" =

We-*
n = oo
1

-L
I

1
L
u m

a+ a V

+ a"e-Jum

l\\anco${wn)

a+

n=l
I

s=

x(0) + 2 ^ x ( n ) c o s ( u m )
n=l

204

0 < n < N - 1',

Figure 5.1:
(c) Refer to fit 5.1.
(d) Refer to fit 5.1.
(e) Refer to fit 5.2.
(f) N = 15. Refer to fit 5.3.

5.29
Refer to fig 5.4. The time domain aliasing is clearly evident when N=20.

5.30
Refer to fig 5.5.
(e)
Xam(n)

x(n)cos(2nfcn)
S-i

Xam(w)

^x(n)coS(27r/cn)e--'2^'
n=0
N-l

~E

\e-j2irU-fc)n

e-W+l'*

n=0

Xam(w)

-[X(w-We)

X(w+We)]

5.31
(b) Refer to fig 5.6. The DFT of x(n) with N = 128 has a better resolution compared to one
with N = 64.
205

Figure 5.2:

5.32
(a)

v(in) = ~-PUV)*xun)
= (r 0 n(~^)e- i2?l ) * [2*(n - o)]
where smcz

VOn)

Tosmc I

'

(b) u>o-P = 2irk for an integer Jb, or u>o- I*.y-ir

(0
N-l

y(u) = e"o i , - ; u i n
n=0

Larger N => narrower main lobe of |V(tx;)|. To in Y(jQ) has the same effect.
(d)

Y(k)

|y(t)i =

Y{w)\ws#k

|n*(t-/)|
Ism

JV6(* - /)

206

Figure 5.3:
(e) The frequency samples fyk fall on the zeros of Y{w). By increasing the sampling by a factor
of two, for example, we will obtain a frequency sample between the nulls.
Y(w)L=2ftk=jrk>

*=0,1,...,2N-1

207

x(n)

1.5

1
i

i 0.5
n

X(w)wi#iN=100 6

:(n) with N="n5b

Figure 5.4:

208

xc(n)

to
X

100

. ,200
xam(n)

-1

300

Xa^wjwithtfifes

300

Xa$|w)withrJifto

150

E 0
X

-2

M M % 0 with fSPfto

30

60

nnn
100

209

200

300

DFTofx(n)withN=64

800

20

DFTofx(n)withN=128

210

40

60

80

Chapter 6

6.1
Since (ej&k)N
e>2*k = 1, e J # * satisfies the equation XN 1. Hence e'$r fc is an N root
of unity. Consider 3n=T0 eJ ^ i n c J w / n . If k ^ /, the terms in the sum represent the N equally
spaced roots in the unit circle which clearly add to zero. However, if it = /, the sum becomes
E n J o l s = ^ - see fig 6.1

z-plane

unit circle
Roots for N=12
Figure 6.1:

6.2
(b) Let VV% = W+6 where VV is the truncated value of W9S. Now W*j - ( W j + 6 ) ' W# + /6.
211

Generally, single precision means a 32-bit length or 6 = 5x10" 10 ; while 4 significant digits means
6 = 5xl0~ 5 . Thus the error in the final results would be 105 times larger.
(c) Since the error grows as 16, after N iterations we have an error of N6. If W$ is reset to -j
after every ql % iterations, the error at the last step of the iteration is 16 =
error reduced by approximately a factor Aq.

& 6. Thus, the

6.3
N-\

X(k)

x n W

( ) Nn

0<k<N-l

n=0
* - l

AT-l

*(n)W# +

n=0

UtX'(k')
Then, X'(k')

r:s0

X(2k + 1),

Using the fact that W}f'

0 < k' < y - 1

[ x ( n ) < k ' + 1 ) " + x(n + ^ ) < + * > < 2 * ' + 1 >

n=0
n

z{n)Wtr

L
n

W' ,

*'(*') = E
n=0

=E

W# = 1
rn xxrk'n
z(n)W%
W*jf + x(n +

j)W^nWJ^W^

w#w n

x(n)-x(n + y )

n=0

6.4
Create three subsequences of 8-pts each

Y(k) =

yC)M*n+ E

n=0,3,6,...
7

n=l,4,7,...

n=2,5,...

y(3t)tt^ + E

y(3' + l ) ^ n

t=0

i=0

vH^ n + E ^ n ) ^ n
i=0

Yi(*) + w y 2 ( * ) + w # y 3 ( * )

where Y i ^ ^ represent the 8-pt DFTs of the subsequences.

6.5
X(z)
X(k)

=
=

l + z - ^ . - . + z"
X(2)|
^
212

y(3 + 2)W$ W #

x'(n)

{2,2,1,1,1}

]C*(" + 7m),

'( )

n = 0,l,...,4

Temporal aliasing occurs in first two points of x'(n) because X(z) is not sampled at sufficiently
small spacing on the unit circle.

6.6
(a) Zk = 0 . 8 ^ ' W + T ) see fig 6.2

(b)

z-plane

circle of radius 0.8

Figure 6.2:
X(k)

X(z)\1=th

= *(n)[o.8e'W+*]]
n=0

s(n)

x(n)0.8e-^n

6.7
Let M

,
213

L = 2. Then

F(0,g)

^i(0,m)Wj'
n=0

r=0

which are the same as Fi(k) and F2{k) in (6.1.26)

G(0,q) =
G(l,q) =
X(0, 9 ) = x(*) =
=
Xl,g) = x(*) =
=

F(0,g) = Fx(k)
W^F(1,) = F2(*)Wfc
G(0,g) + G(l,9)W2o
fi(*) + F2(*)W$
G(0,) + G(l,g)lV2l
Fl(k)-F2(k)WkN

6.8
w6 =

-L(i-j)

Refer to Fig.6.9. The first stage of butterflies produces (2, 2, 2, 2, 0, 0, 0, 0). The twiddle factor
multiplications do not change this sequence. The nex stage produces (4, 4, 0, 0, 0, 0, 0, 0) which
again remains unchanged by the twiddle factors. The last stage produces (8, 0, 0, 0, 0, 0, 0,D).
The bit reversal to permute the sequence into proper order unscrambles only zeros so the\result
remains (8, 0, 0, 0, 0, 0, 0, 0).

6.9
See Fig. 6.13.

6.10
Using (6.1.45), (6.1.46), and (6.1.47) the fig 6.3 is derived:

6.11
Using DIT following fig 6.6:
1 s t stage outputs

(III

II

\2,2,2,'",2J

2 n d stage outputs

: | l , 1(1 + W*),0,1(1 - W82), 1,1(1 + W82),0,1(1 - Wtf)J

3 r d stage outputs

: ^2, j ( l + Htf + VV| + W|),0,1(1 - W* + W* - W}),0,

1(1 - Wi + Wi - W*),0,1(1 - Wi - Wi + Wtfj
214

x<0)

MIS)

Figure 6.3:
Using DIF following fig 6.11:
,st
1 s t stage outputs
nd

stage outputs

3 r d stage outputs

f 1 1 1 1 1 1 l T , r l 1 TI , 2
lr3\
|_,-,-,-,-,-,-^,-W|,-W||

: | l , 1,0, 0, i ( l + W*),0. i(Wtf + W|), i ( l - W|), i(W*

: {2, 0,0, 0,1(1 + W + Wi + VV|), 1(1 - Wj + W* - W?
1

1(1 - w* + IV - WJ), i ( l - Wi - Wi + W*}

6.12
Let

*S

^[x(0)

1 1 1 1
1 -j - 1
j
1-1
1-1
1
j -1 - j
x(4) x(8)

x(12)]'

x2i[x(l)

x(5) x(9) x(13) ]'

X3 [ i(2)

x(6) x(10) x(14) ]

215

3U = [ (3)

x(7)

F(0)
F(4)
F(8)

x(ll)

x(15) ]'

= Ax_x

L ^(12) J
F(l)
F(5)
F(9)

= y4x 0 =

F(13) J
F(2)
F(6)
= Ax* =
F(10)
F(14) J
F(3)
F(7)
F(ll)
F(15)

= AxA

-4
0
0
0

As every F(J') = 0 except F(0) = - F ( 2 ) = 4,

x(0) "

*(7)
x(8)

. *02) .

AT

S\X A

F(0) *
F(l)
F(2)
. ^(3)

"0"
8
0
8

which means that A"(4) = X{\2) = 8. X(k) = 0 for other K.

6.13
(a) "gain" = WgW(-l)Wi
= -W* = j
(b) Given a certain output sample, there is one path from every input leading to it. This is true
for every output.
(c) X(Z) = x(0) + Wix(l) - Wix(2) + WiWix(3) - Wx(4) - W 8 W|x(5) + W8W2*(6) +
WWiW2x(7)

6.14
Flowgraph for DIF SRFFT algorithm for N=16 is given in fig 6.4. There are 20 real, non trivial
multiplications.

6.15
For the DIT FFT, we have

X(k) = J2 ^(2n)Vy|* + 5Z *( 2n + l)^N n + 1 ) t

n=0

nsO

216

OX(4)
OX(l2)

Figure 6.4:
The first term can be obtained from an y-point DFT without any additional multiplications.
Hence, we use a radix-2 FFT. For the second term, we use a radix-4 FFT. Thus, for N=8, the
DFT is decomposed into a 4-point, radix-2 DFT and a 4-point radix-4 DFT. The latter is
4-i

4-i

4-i
{ n+l)k

x(2n + l)W J

= J2 * (

4n

l W

+ ) N % + 5Z

n=0

n=0

X(k)

) Nkwk

4-1

for* = 0 , 1 , . . . , ^ - 1 are performed

* - i

= V z{2n)Wtf + V x(4n + l)WjW* + V x(4n + 3)W$*IV*

n=0

N
* ( * + T>

3 W

n=0

The computation of X(k), X(k + ) , X{k+%),X{k+*Z)

from the following:
* - i

z(4n +

n=0

n=0

4-1

4-1

= Z * ( 2 n ) ^ V l ) " + I Z * ( 4 n + l ) ( - i ) ^ + y ] * ( 4 n +3)^*0-)^*
n=0

ns=0

n=0

4-1

4-1

N = ^x(2n)^+^x(4n+l)(-lX^+^^n + 3)(-l)^^
*(* + )
n=0

n=0

n=0

4-1

4-1
X k+J

r)

= *(2n)W*(-l)" + *(4n+l)(j)WRtV* + x(4n + 3)(-W#W^*

n=0

n=0

n= 0

The basic butterfly is given infig6.5

217

ofDIF-SWTThwuffly

K0)O-

OWO)

**)0-

-Oxen

K2)^
*6)0-

-0X(2)

0^>N3

>

04)

i(l)01(5)0i(3) O -

Drr/surr

Th piffc looto like Itat nmpot of

-OX(3) MN-P"*XDFFFT.Th*twiddlpfKn

-0X(5)

*r

^Xfft)

\y

-0X(7)

iP) O -

Figure 6.5:

6.16
x

=
=
e =
XR
x/ =
Total

xR + jxi
(a + jb)(c + jd)
(a b)d
1 add , 1 mult
e + (c rf)a 2 adds 1 mult
e + (c + d)b
2 adds 1 mult
5 adds 3 mult

6.17
N-l

X(z) = X>(n)z-n
n=0
N-l

Hence, X(z 4 )

x(n)r-ne"^fcn

n=0

where zk = r e " ^ f c , i b = 0 , 1 , . . . , N - 1 are the N sample points. It is clear that X(zu),k

0 , 1 , . . ., N - 1 is equivalent to the DFT (N-pt) of the sequence x{n)r-n,n
[0, N - 1].
218

6.18

rto-i

LN-l

4=0

1
L7V
1

k=LN-k0+l

to-1

IJV-1

k=.0

k=LN-k0+l

"fi X(k)W[? +

X)

E^(*)WTJ}B+

L*=o

X{k + N-LN)WiT
X(t)W-i'-w+tw'"

fc=JV-*0-i

*o-l

Therefore Lx'(Ln)

N-l

N *=0
= *(n)

t= N-t0+l

L = 1 is a trivial case with no zeros inserted and

fl

. 1 1

.11

6.19
N-l

X(k) = J ] *(nX n
n=0

Let F(0,

* = 0 , 1 , . . . , N - 1 be the DFT of the sequence on k X(k).

N-l

F(t) = *(*)**#

TV-l

*(nXn

r=0
W-l

r*(n+t)

= 2>-> E <
i=0

n=0

N-l

5Z z ( n W n + Omod N
n=0
JV-1

x(n)6(W-l-n-0

1 = 0,1,. ..,7V - 1

n=0

{x(Ar-l),x(AT-2),...,x(l),x(0)}

6.20
2N-1

y(fc)

y(n)^n

n=0

219

* = 0,1,...,2JV-1

2N-1

n=o,n even
N-l

= y(2m)W*m
m=0
Af-1

= X>(m)W#
mssO

=
=

*(*),
X{k-N)}

t[0,AT-ll
ke[N,2N-l]

6.21
(a)
w(n)

ln

2irn

0< n < N - 1

1 _ 1 i*sv
|_I
(e
+ e -i#a)
2

N-l

W() = (n)z-"
n=0
JV-1

= i-i(^ + e-i)
n=0

1l-z-N
2 1-r1
4

ll-Cz-V^r)^
4 i_r-V'T^T

1_z-ic-;7fe-

0>)
Xu,(n)

X(ib)

tu(n)x(n)
W(k)NX(k)

6.22
The standard DFT table stores N complex values Wj^,

k = 0 , 1 , . . . , TV - 1. However, since

W^* = - W , we need only store W

it = 0 , 1 , . . . , - 1. Also, wff* = - j W which is
merely an interchange of real and imaginary parts of Wj^ and a sign reversal. Hence all essential
quantities are easily obtained from Wj
A= 0,1,...,^ 1

6.23
The radix-2 FFT algorithm for computing a 2N-pt DFT requires 20-hg22N = N + Nlog2N
complex multiplications. The algorithm in (6.2.12) requires 2 [ y / o ^ 2 ^ + y ] = y+/o^2jV complex
multiplications.
220

6.24
since H(z)

I + EI**-*
H{^k)

_EiU**^i

i/(*),

t = 0,...,JV

Compute AT + 1-pt DFTs of sequences {60, &i,. &MiO,0,.. .,0} and {l,a\,..
N > M), say B{k) and A(k)
k = 0,...tN

.,as) (assumes

H(k) ~ 51*1

6.25
8

y(*) = E ^ n w *
n=0

= E v(*W* + E v(nW*+ E v^W*

n=0,3,6

n = l,4,7

n = 2,5,8

3i

E y(3m)W 9 - + E v(3m + l ) w f
m=0

m+1)

r (3m +

* + E ^3m +2>^

m=0

2)k

m=0

= E y(3^)^m + E y(3m +iw**? + E ^ 3 m + 2 ) w 3 m k ^

Total number of complex multiplies is 28 and the operations can be performed in-place. see
fig 6.6

6.26

*(*) = E*(w*
n=0

n=0

n=3

n W n

= E *( ) *

n=6
2

x n

+ E < + )^9 ^3 + E x (" + 6)W9ni W32*

n=0
2

n=0

nk

*(3/) = E ^ ^ ^ E ^ ^ ^ ^ E ^ * 6 ) ^ '
nsO
2

X(3/+1) = E X H ^ 3 n , ^ 9 n + E X ( n
=

J2wZ[x(n)+W}x(n

+3

)^3'^9 n % 1 + E X ( n

+ 2) + WZx(n + 6)}W2>

221

+ 6 Vy ,

) 3 %n^32

<<0)O

-OX(0)

id) O

OX(3)

'(2)0

OX(6)

c(3)(>

Ox(i)

<(4)0

w,

OX(4)

<(5)0

OX(7)
OX(2)

t(6)C>
2

-OX(5)

i(7)0

t(8)0-

OX(8)

Figure 6.6:

x(3/ + 2)

^ V y 9 2 n [ x ( n ) + ^ 3 2 r ( n + 3) + Vy31x(n + 6)] W3n'

n=0

The number of required complex multiplications is 28. The operations can be performed in-place.
see fig 6.7

6.27
(a)Refer to fig 6.8
(b)Refer to fig 6.9
(c) D1F is preferable for computing all points. It is also better when only X{0), X(\),X(2),
X(3)
are to be calculated. The rule is to compare the number of nontrivial complex multiplies and
choose the algorithm with the fewer.
(d) If M N and L N, the percentage of savings is
Xl00% (1
Xl00%
10
9N

* %OTN '

= " ^)

6.28
(a)Refer to fig 6.10. If data shuffling is not allowed, then -X(0),..., X(3) should be computed
by one DSP. Similarly for X(4)t...,
X(7) and X(8), , *(11) and X ( 1 2 ) , . . . , X(15). From the
flow diagram the output of every DSP requires all 16 inputs which must therefore be stored in
each DSP.
(b)Refer to fig 6.11
(c) The computations necessary for a general FFT are shown in the figure for part (a), Ng =
222

Ox(0)
OX(3)
OX(6)

Ox(i)
OX(4)
OX(7)
OX(2)
OX(5)
OX(8)

Figure 6.7:
^Iog2N. Parallel computation of the DFTs requires

1N ,

hg

N 1

2M *M+2-<T*
1=1

Complex operations, as is seen in the figure for (b). Thus

5

_=

*$

NB
ZlogiN
Mlog2N
log7N - log2M + 2(M - 1)

6.29
Refer to fig 6.12

*<n) =

Tft,XWWkn

Jb=0

223

x(0)-

X5
.\>

jfi* - *

X(0)

X(8)

>

<X^

X(4)

X(2)
*-X(10)
X(6)

*Xws
.1 '

'

X*.
.1 '

X(14)
X(l)

'" X(9)

*Xe

X(5)

*-X(13)

x*
P

IP

X(3)

X(ll)
X(7)
X(15)

Figure 6.8:

= I E x(k)w*kn + \ E *(<w kn
k even

odd

m=0

m=0

= I E ^(2m)W4-mn + i E *<2m + l)W7mnW7

= E [ * < 2 m ) + * < 2 m + w " ! ^" m n
m=0

r 3

*()

A-(2m)lV-mn + Wg-n E ^(2m + l ) W - m n

T=0

m=0

*(n + 4) =

- E

0<n <3

A"(2m)W-mn - W7 n

Lm=0

X{2m + l)W 4 - mn

0< n < 3

m=0

This result can be obtained from the forward DIT FFT algorithm by conjugating each occurrence
of W%N
Wff1 and multiplying each output by g (or \ can be multiplied into the outputs of
each stage).

6.30
*(") = j[E*(*)W7*"
k=0

224

Figure 6.9:

fc=4
3

Y, X(tWskn + ( - l ) " E X{k + 4)W7*n

A= 0
3

x(2/)

fcsO
3

x(*)vv4-'*+5>(* + 4)ttv'k

x(2/+l)

/ = 0,1,2,3

i =0

,* = 0
3

Y,X(k)WA-lkWzk -Y,X(k + 4)W;lkWi"

k=0

/ = 0,1,2,3

u=o
Similar to the DIT case (prob. 6.29) result can be obtained by conjugating each W*N and scaling
by i Refer to fig 6.13

6.31
x(n)

x*(N-n)

IDFT(**(n))
n=0

= 77 E^-")^'"
n=0

225

X(15)

Figure 6.10:

= *<"W (W ->
m'=0

Since the IDFT of a Hermitian symmetric sequence is real, we may conjugate all terms in the
sum yielding
N-l

IDFT(x*(n))

=
m'=0

n=0

j}X(k)

In general, the IDFT of an N-length sequence can be obtained by reversing the flow of a forward
FFT and introducing a scale factor jj. Since the IDFT is apparently capable of producing the
(scaled) DFT for a Hermitian symmetric sequence, the reversed flow FFT will produce the desired
FFT.
226

X(2)

X(10)

X(6)
X(14)

Figure 6.11:

6.32
S-l

X(k)

Y,x(m)WkNm

N-l

J2 x(m)W^mW^kN

since W^kN = 1

m=0
N-l

x{m)Ws
m=0

This can be viewed as the convolution of the N-length sequence x(n) with the impusle response
of a linear filter.
hk{n)

Wrjvnu(n), evaluated at time N

00

Hk(z) = > * n z - n
n=0

l-Wfrz-i
X(z)
yk(n) = W ^ y t ( n - l ) + x(n),
yt(,V) = *(*)
227

y*(-l) = 0

X(0)
X(4)
X(2)

Figure 6.12:

6.33
(a) 11 frequency points must be calculated. Radix-2 FFT requires i y^/oy 2 1024 5000 complex
multiplies or 20.000 real multiplies. FFT of radix-4 requires 0.75x5000 = 3,750 complex multi
plies or 15,000 real multiplies. Choose Goertzel.
(b) In this case, direct evaluation requires 106 complex multiplies, chirp-z 22xl0 3 comples mul
tiplies, and FFT 1000+ ^ x l 3 = 33xl0 3 complex multiplies. Choose chirp-z.

6.34
In the DIF case, the number of butterflies affecting a given output is y in the first stage, ^ in
the second, .... The total number is
1 + 2 + . . . + 2 " 1 = 2 " ( 1 - ( - ) ) = tf - 1

Every butterfly requires 4 real multiplies, and the eror variance is y^. Under the assumption that
the errors are uncorrelated, the variance of the total output quantization error is
6*_ _ NV
^ = 4(^-1)^- = ^-12 ~ 3
228

X(0)-U&

x(0)

tr-xw

Figure 6.13:

6.35

Re[X n+1 (*)]

\xn+i{k)+X:+l(k)

= \xn(k) + l-wzxn(i) +l-x-n(k)- ^w-^x'M

=
since \Xn(k)\ < - ,
since | X , ( * ) | < - ,

\Re[Xn(k)]\

Re[Xn(ib)] + Re[W^Xn(/)]
1

< 5

|Re[WJ7X(0]| <

so |Re[IVtfX(')]l <
Therefore |Re[XB+1(fc)]| <

1
5

i
|R*[X(*)]| + \Re[WJ^Xn(l))\ < 1

The other inequalities are verified similarly, (b)

Xn+1(Jb)

Rc[X(*)] + iIm[X(*)]
[cos^m)

=
Therefore, |* + i(*)|

- jsin(^m)][Re[Xn(l))

+ ;Im[Xn(/)]]

Re[X(*)] + co(.)Re[*(/)] + sin(.)Im[X(/)]

+ j {Im[X(*)] + cos(.)lm[Xn(l)} + sin(.)Re[Xn(/)]}
|*(*)| + \Xn{l)\ + 4
229

where A

also | X n + 1 ( / ) | 2
Therefore, if A > 0,
max[|X n + 1 (*)|,|X n + 1 (/)|]

2cos{.){Re[Xn{k))Re[Xn(l)}
+
lm[Xn(k))lm[Xn(l))}
+2sin(.) {Re[Xn(k))lm[Xn(l)]
- Im[X n (*)]Re[X n (/)]}
\Xn(k)f + \Xn(l)\2 - A(*)

\Xn+l(k)\
{\Xn(k)\2 + \Xn(l)\7 + A}*
max[|X n (*)|,|X n (/)|]

=
>

By similar means using (*), it can be shown that the same inequality holds if A < 0. Also,
from the pair of equations fro computing the butterfly outputs, we have
2Xn(k)

2Xn(l)

X + i(*) + X+i(/)

WZmXn+i(k)-WjmXn+l(l)

By a similar method to that employed above, it can be shown that

2max[|X n (J:)|, \Xn(l)\) > max[\Xn^(k)l

|X + 1 (/)|]

6.36
Refer to fig 6.14.
(d) (1) The frequency interval between successive samples for the plots in parts (a), (b), (c) and
(a)N-64doie

20

40

(b)N-64dc8

60

20

(c)N-128dcie

40

60

80

(d)N-4dc-7.664*-M

(d) are , ^ , ^and

respectively.
(2) The dc values computed theoretically and from the plots are given below:
theoretical
practical

part a
16
16

part b
8
8

part c
16
16

part d
0
8.203e - 14

Both theoretical and practical dc values match except in the last case because of the finite word
length effects the dc value is not a perfect zero.
230

(3) Frequency interval = j$-.

(4) Resolution is better with N = 128.

6.37
(a) Refer to fig 6.15.
(b) Refer to fig 6.15.
r-0.9. YW

r.0.9, O0.92. W(K)

150

Figure 6.15:
(c) Refer to fig 6.15.
(d) Refer to fig 6.15.
(e) Refer to fig 6.16.

r = 0.5, Y(k)

W(k)

X 10

3
o
3
*<

(0

1
n

50

150
Figure 6.16:

231

100

150

Chapter 7

7.1
(a) H(z) = 1 + 22" 1 + 32" 2 + 4 2 " 3 + 32" 4 + 22" 5 + z~6. Refer to fig 7.1
(b) H(z) = 1 + 22" 1 + 32~ 2 + 3 2 " 3 + 22" 4 + 2 " 5 . Refer to fig 7.2

x(n)

Figure 7.1:

7.2
Refer to fig 7.3
1 + 2.882- 1 + 3.40482" 2 + 1.742 -3 + 0.42 - 4

A4(z) = H(z)

0.4 + 1.742 -1 + 3.40482" 2 + 2.882 - 3 + z~A

B*{z)
Hence, K\
Az{z)

0.4
A4(z) - kABA(z)
\-k\
l + 2.62~ 1 +2.4322- 2 + 0.72- 3
233

x(n)

Figure 7.2:

B3(z)

0.7 + 2.4322- 1 + 2.62 - 2 + z~3

Hence, K3

0.7
A3(z) -

*n*)
B2(z)
Then, K2

AM
Therefore, K\

k3B3(z)

l-*i
1 + 1 . 7 6 2 " 1 + 1.22"2
1.2+ 1.762 -1 + 2 - 2
1.2
A2(z) - k2Bi(z)
l-kl
1 + 0.82 - 1

=
=
=
=
=
=

0.8

Since A'2 > 1, the system is not minimum phase.

7.3
V(z)

X{z)+l-z-'V{z)

v(n)

z(n)+-v{n-l)

Y(z)

2(3X(2) + V(2)] + 2 2 - ^ ( 2 )

H(z)

h(n)

Y(z)
X(z)
8-z-1
1 -0.52"1
8(0.5)nu(n)-(0.5)n"1ti(n-l)
234

z->

x(n)

,->

7-1

2.88

3.4048

r
0.4

1.74

r^

y(n)

(a)

f 4 <n) - y(n)

(b)

Figure 7.3: (a) Direct form, (b) Lattice form

7.4
H(z)
h(n)

5+

32'

l + J*- +

l + 2z'

\-\z-'

U{n) + ^ - i r - ^ C n - 1) + () n u(n) + 2 ( i ) ' - 1 u ( n - i;

7.5
H(z)

6+fz'-|z-2
(

I^i)(l_Iz-i)

1 +

6+|z'-|z-2
Refer to fig 7.4

7.6
For the first system, H(z)
H(z)

"
=

l_6l2-i '
\-b2z^
l~(6i+62)z-1
(l_6lr-i)(l-62z-i)
CO + C i Z - 1
2 - I ) ( l _ a 2

For the second system, H(z)

( 1

clearly, co

=
=

-(&i + 62)

Ci

rfi
a2

=
=

235

d l

1
6i
62

-l)

Figure 7.4:

7.7
(a)
y(n)

a i y ( n - l ) + a 2 y ( n - 2 ) + 6 0 x(n) + 6 l x ( n - 1) + 62x(n - 2)

H(z)

60 + 612*1 + 6 2 *~ 2
1 + a i z - 1 -I- a-iz,~2

(b)

l + 2z"1 + *-2

#(z)

Zeros at z

I + LSZ-^+O^Z-2
1,-1

Poles at z

-0.75 j'0.58

Since the poles are inside the unit circle, the system is stable.

Zeros at z

_
~
=

l + 2z-i + z-2
l + z-^-2z-2
1,-1

Poles at z

2,-1

H{Z)

The system is unstable.

(c)
x(n)

cos(-n)
236

1
1 + r-1 -0.99z"2
1
1 + e~Jw - 0.99e-i2w

H(z)

H(w)

*<!>

100c" ; ^

Hence, y(n)

J ir
100cos(-n-

IT

-)

7.8
y(n)

- y ( n - 2 ) + x(n)

H(z)

l-iz-2

(a)
*<)

H(z)

~2

(5)- + (-j)-

+l+

l_i2-i

(n)

|2-i

(b)

x(n)

X(z)

X(z)

Y(z)

(ir + c-ir
1

iz"1

l-iz-2
*(*)#(*)
1
1 + iz"1

y()

+ 1+

1-iz-1

(n).

1-i*"1

(1-ir-1)2

(3)- + (-i)--(5r + n(-i)'

(l+i*-1)

u(n)

(c)Refer to fig 7.5

(d)
ff(u/)

1 _ Le-j2w
y/l7 - SC0s2l

L-tan

-l

Refer to fig 7.6.

7.9
(a)

"<*>

"

1 + i;-1
l - ^ - 1 + iz-2
237

sin2w
4 cos2w

Direct form 2
x(n)

-o-

cascade form

x(n)

*(H 1

( i 1

r rh V

1/2

D
1/4
Parallel form

rJ V

^J*y(n)

1/2
x(n)

*
i/V

^
D

iT
Figure 7.5:

+ *-'
( 1

11

_i2-i)(i_i2-i)

- i2 2i - l

+ 1-12-1

Refer to fig 7.7

(b)
H(z)

0.7(1-0.362-2)
I + O.I2-1 - 0 . 7 2 2 - 2
0.7(1 - Q , 6 2 - i ) ( l + 0.62-1)
( l + 0 . 9 2 - ! ) ( l - O.82-I)
0.1647
0.1853
0.35 1+0.92-1
1-0.82-1

Refer to fig 7.8

(c)

H(z)

3 ( 1 + 1.22-! + 0 . 2 2 - 2 )
1 + 0.12-1 - 0 . 2 2 - 2
3(l + 0.22-1)(l + 2 - i )
(l + 0.52~i)(l-0.42-1)
7
1

-3 + 1 - 0 . 4 2 - 1

Refer to fig 7.9

(d)
238

1 + 0.52-1

-1/2

1
D

'

Magrwtud* ot H(w)

Figure 7.6:

H(z)

2(l-2-1)(l + v/2V1+2-2)
(1 + 0.52- 1 )(l - 0.92- 1 + O.82-2)
2 + (2>/2 - 2)2-! + (2 - 2y/2)2~ 2 - 22~ 3 )
1 - 0.42" 1 + 0.362-2 + 0.4052" 3
A
B + Cz'1
1+0.52-1
1-0.92"1+O.82-1

Refer to fig 7.10

(e)
H(z)

1 + 2- 1
11 -

Z "
12

- 42Z ~ 2

1 + 2- 1
( l - 0 . 8 l 2 - 1 ) ( l + 0.3l2- 1 )
-0.62
1.62
+
1
1+0.312"1
I-O.8I2
Refer to fig 7.11
(f) H(z) 1l7-i*I^5^1a =* Complex valued poles and zeros.Refer to fig 7.12 All the above
systems are stable.

7.10
Refer to fig 7.13
H(z)
(1)
(2)
(3)

V(z)

W(z)

Y(z)

1
1 2rcoswoz~1 + T7Z~2
X(z) rsinwoz~1Y(z)
V(z)-rcosw0z-xW(z)
rcoswoz~lY(z) rsinwoz'1
239

W(z)

Direct form Q:

y ( n ) i<>

Cantfe:

Ptellkl:

GHQ
V,

x(n)-

r^

-y(n)

rO

z
1/3

1/2

-,<>

1/2

(n)

103

1M

7/3

L^>
Figure 7.7:
By combining (1) and (2) we obtain
,-i

(4)

WM = !

*^r*U) - . """""

1 rco5it>oi
Use (4) to eliminate W(z) in (3). Thus,
Y{z)[l - 2rcosw0z-1

+ (r2cos2u>0 + r 2 i i n 2 w 0 ) : ' 2 ]
Y(z)
X(z)

1 rcoswoz *

7.11
M*)
M(z)

Bi(z)

=
=

Bo(z) = 1
Ao(z) + k1B0(z)z-1

.+ i-

ss

I-"'

B,(z)

=
=
=

Ai(z) + ib2Bl(z)
l + 0.3z _1 + 0.6z-2

Mi*)

A2(z) + ^ 3 B 2 (r)

M(z)

0.6 + 0.3z" l + z" 2

240

X{z)
1
1 -2rcosw0z-1

-1

+- Lrr22z, - 2

Direct form II:

-tfn>*<n) ~ ~ \ 0

~JT Cly~* n)

Pmlkl:

CMcadc:

(MfrC? -

-0.1853

2-1
0.9

aL

0A

x(n)

JLZLMMMJ

0.9

^nVi(Y)--y(n)
0.1647

L-O
Figure 7.8:

1 - 0.12z _1 + 0.39z - 2 - 0.7z" 3

- 0 . 7 + 0.39z _1 - 0.12z" 2 + z - 3
,4 4 (z)

A3(z) +

k4B3{z)
1

Therefore, //(z)

1 - z'
+ 0.52z - 2 - 0.74z~ 3 + i z - 4
150
3
53 1
1
2
3
C(l - z+ 0.52z" - 0.74z" + - z " 4 )
150
3

where C is a constant

7.12
Refer to fig 7.14

Hk{z)
Vk{n)
w7k{n)

bok + bikZ'1 +bzkz-2

1 + a i k z - 1 + a2*z~ 2
bokXk(n) + wik(n - 1)
b\kx{n) - alkyk(n) + w2k{n - 1)
bikx{n) - a2kyk(n)
241

Otfcct fonn U:

Dtftrt form I:

x(n)

^ - y ( n ) x(n)

FtemUtl:

Cwcadc:

-(^j^H^jrG^
0.5

02

0.4

Figure 7.9:

7.13
YJMl = G * XIN
DO 20 J=1,K
YJ=B(J,0) * XIN + Wl(J)
W1(J) = B(J,1)*XIN - A(J,1)*YJ + W2(J)
W2(J) = B(J,2)*XIN - A(J,2)*YJ
YJMl = YJMl + YJ
20
CONTINUE
YOUT = YJMl
RETURN

7.14
YJMl = XIN
DO20J=l,K
W = -A(J,1) * WOLDl - A(J,2) * WOLD2 + YMJl
YJ = W + B(J,l)*WOLDl + B(J,2)*WOLD2
WOLD2 = WOLDl
WOLDl = W
YJMl = YJ
20
CONTINUE
YOUT = YJ
RETURN
242

DifBCt fonn D:

Direct form 1:
yin)

x(n)

Pmttrf:
Cttctdc:

'MZH^M^j-O-^u,
OS

-0.81

Figure 7.10:

7.15
z) = A2(z)
B2(z)

l + 22- 1 + - z " 2

- + 2z-1 + 2 - 2

*2

Ai(z)

=
=
=

*i

1
3
A2{z) - k2B2{z)
\-k\
.+
3
2

| . -

7.16
Ai(z)

Bi(z) J
A3(z)
Bi(z)

-h

2 l

~ I *i \ [ ~
1

AAz)
z-'B^z)
243

i ,-i
1 + ^2

U^

l+Ir-i_I2-2
- I + I . - 1 + .-2

DwAfansI:

OtffBCt IORD u:

y(n)x(n)

Cmmk

ftnlkk
y(n)

0.81

r<?
0.81

x(n)

-031

0.3

kb

_UM

Gy* y(n)
-0.62

Figure 7.11:

B3(z) J

1 1 f
l

A2(z)

[ 1 1 J [ *- S 2 (*)

ffi(*) = i*3(*) = l + *

-3

zeros at z = l,e J '*

(b)
// 2 (z)

A2(z)-z-1B2(z

= l + ik'-ik2-*-3
2

The zeros are z

-1

3
3
- 5 jy/U
1,
6

(c) If the magnitude of the last coefficient \k#\ = 1, i.e., ks = 1 , all the zeros lie on the unit
circle.
(d) Refer to fig 7.15. We observe that the filters are linear phase filters with phase jumps at the
zeros of H(z).

7.17
(a) Refer to fig 7.16

244

Direct form II, cascade, parallel:

Direct form I:

x(n)
1

I
A
-1 J 4-

'
z-1

'' \ !
^r
z-1

"

\ )
ir

y(n>x(n)

2"

+\

ii

''

-Q
CH

2-1

-0.5

-1/2

Figure 7.12:
x(n)
Mn)
gx(n)
/ 2 (n)

=
=
-

=
=
g7(n) =
=
h(n) = / 3 ( n ) =
=

6(n)
6(n) + 0 . 6 5 6 ( n - l )
0.656(n) + 6 ( n - 1)
/i(n)-0.34^(71-1)
6(n) + 0 . 4 2 9 6 ( n - l ) - 0 . 3 4 ( n - 2 )
-0.34/ 1 (n) + < 7 l ( n - l )
-0.34<5(n) + 0.4296(n- l) + 6 ( n - 2 )
/ 2 ( n ) + 0.8<7 2 (n-l)
6{n) + 0.1576(n - 1) + 0.00326(n - 2) + 0.86(n - 3)

(b) H(z) = 1 + 0.1572" 1 + 0.0032z~2 -I- 0.82" 3 . Refer to fig 7.17

7.18
(a)
Cz{z)
n\i)

M*)

Aziz)
Bs(z)

l + 0 . 9 z - 1 - 0 . 8 2 " 2 + 0.5z~ 3

0.5-0.82_1+0.92-2 + 2-3

k3

0.5

245

y(n)

w(n:U

rsin^)

>

/^N

v(n)

Figure 7.13:

A3(z) -

^(*)

**(*)
*2

=
=
=

*(*)

Bi{z)
*1

C 3 (z)
0s(*)

=
=
=
=

*3

=
=

c2(*)

=
=

1 - ^
1 + 1.732"1 - 1.672"2
- 1 . 6 7 + 1 . 7 3 2 - 1 + 2~ 2
-1.67
A7(z) - k2B2(z)
l-k%
1 + 1.622 -1
1.62+ 2 " 1
1.62
l + 2 2 _ 1 + 3 2 - 2 + 222 + 32" 1 + 2 2 ~ 2 + 2 - 3
2
C3(z) -

*2

Ci(z)

=
=

k3D3(z)

- 1 + 12- 2

r
4

-1

-2

1
3
C2(z) - k2D7(z)
\-k22

1+ 3

fll(*)

,1+
1

D 2 (2)

k3B3(z)

=
246

4+Z

-i

Ok

x,(n)

X (n)

jwi k (n)

,-1

b^

s \

-*2k

x(n) =

Hj(z)
xjCn)

-^_V>y(n)%.,n,

Hz)

Figure 7.14:
3

Cs(z)

t>0 + V\Di(z)
1

+ V2D2(z)

1 + 22" +3z~

VzD3{z)

+ 2z~

From the equations, we obtain

VQ

V\

V2

=
=

107
48
13
4
-1
2

The equivalent lattice-ladder structure is: Refer to fig 7.18

(b) ^3(2) = 1 + 0.9z" l - O.82"2 + 0.52" 3 ,
|Jbi| > 1 and |* 2 | > 1 => the system is unstable.

7.19
Refer to fig 7.19

Y(z)

C(z)

H(z)

[rsinBX(z) + rcoseY(z) - rsinBC(z)] z~l

[-rcosBXiz)
+ rsinBYiz) + rcoseC{z)} z~x
Y(z)
X(z)
247

Figure 7.15:

Hence, h(n)
and y(n)

=
=

rsinQz'1
1 - 2rcosGz~1 + r2z~2
rn sin(Qn)u(n)
rsinQx(n 1) + 2rcosBy(n 1) - r2y(n 2)

The system has a zero at z 0 and poles at z

re^'e.

7.20
H(z)

1
1 - 2rcosu;o2~1 + r2z~2
rcoswo - j 2tinxv*

rcosu;0 -rj

rcoswo - J 2 nu , 0 B
X(2)
2 (rcoswo + jrsantfo)
t>i(n) + jv 2 (n)

S(z)
s(n)
P

rcostuo

=>*i

r5n u;o

<*2

9i + J92
rcostuo
rcoswo

=>?i
92

t>i(n+l)
v2(n)

=
=

2,inw0

1 +z - (rcoswo + jrsinwo) + z - (rcoswo - jrsinwo)

2sinwo
aivi(n) - a 2 v 2 (n) + qix(n)
rcotyovi(n) r s i n w o t ^ n ) + rcoswox(n)
Q2Vi(n) + aiV2(n) +q2x(n)
rsmwoVi(n) + rcoswoV2(n) +
x(n)
2sinwo
248

Figure 7.16:
or, equivalently,
t(n + l) =

rcoswo
rsintvo

rsinwo
rcoswo

y(n)
or, equivalently,
y(n)

t/(n) +

TCOSWQ

reoswo

x(n)

2jinu/0

s(n) + s*(n) + x(n)

2v1(n) + x(n)

[2 0]i;(n) + x(n)

where
wi(n)
v 2 (n)

v(n) =

7.21
(a)
Adz
Bi(z
A?(z
B7{z
A3(z

=
=
=

0.6
1 + 0.62" 1
0.6 + z ' 1

=
=

Ai{z) +
k2Bx{z)z-1
l + 0.78z_1+0.3z-2

=
=

0.3 + 0.78z _1 + 2 " 2

A2(z) + k3B2(z)z-1

= l + o.gsz^ + o.egz^ + o.sz-3

B3(z
H(z) = A4(z

0.5 + 0 . 6 9 z - 1 + 0 . 9 3 z - 2 + z - 3

A3(z) +

1 + 1.38z

k*B3(z)z-1
-1

+ 1.311z"2 + 1.3372"3 + 0.9z - 4

249

x(n)

y(n)

Q.157

0.0032

,-l
0.8

Figure 7.17:
(b) Refer to fig 7.20

7.22
(a)
From (7.3.38) we have
y(n)
But, y(n)
Hence, *2
and, fci(l + 2)

=
=
=
=

*i

- * i ( l + * 2 ) v ( n - l ) - * 2 V ( n - 2 ) + x(n]
2rcosw0y(n 1) - r7y(n - 2) + x(n)
>"2
2rcosi^0
2rcosu;o
1 + r2

Refer to fig 7.21

(b) When r = 1, the system becomes an oscillator.

7.23
H(z)

For the all-pole system

1-O.Sz-1
+0.1SZ-2
1
l + 0.l2- -0.72z-2
B(z)
A(z)

. , we have
A(z)
*l(l + *2)
*2

250

0.1
-0.72

x(n)

Figure 7.18:
=>*!
For the all-zero system,

0.357

*2

Cfa)

=
=
=
=

Ai(z)
B2(z)
*2

Ai(z)

Bx(z)
*i

=
=
=

M*)

Ciiz)

0.72
l-0.82-1+0.152-2
1-0.82-1 +0.152-2
0.15-0.82"1 + 2 - 2
0.15
A2(z) - k2B2{z)
\-k\
1-0.6962"1
-0.696+2~x
-0.696
Bo(z) = 1

t>mm(2)

m=0

=
=

Vo + VxBi(z)

=
=

0.15
1

=> to

=
=

Vl

1-0.82

The solution is:

v2
v\ - 0 . 1 8 v 2
v0 -0.696ui +0.15v 2

Thus the lattice-ladder structure is: Refer to fig 7.22

251

-0.8
1.5
-0.68
0.15

_1

V2B7{z)

+0.152-2

rcos 6

rsinB J

y(n

r1

^rsine
x(n)
^sin 6
J -rcos 6 ^ -

c(n)

z-l

rcos 6

Figure 7.19:

7.24
M

y(n) = 6 k *(n-*)
k=0

Leti;j(n+1)

vM(n+l)

v.-+i(n),
i(Af)

Then, y{n)

b0x{n) + ^

i= 1,2,...,M-1
M

fc.t^+i_,(n)

i=i

Then
"0
0

vi(n+l)
v 2 (n + 1)

1
0

0
1

0
0

0"
0

..

1
0

v2{n)

'- o

=
v M - i ( n + l)
vM(n+l)
.

..
..

0
0

...

t>M-i(n)

v2(n)
y{n) = [ bM

bM-\

...

hi ]

+ 6 0 x(n)

General representation:
t>(n + l )

FVJin) + qx{n)

y(n)

V(n) + b0x(n)

The general state-space realization for this system is shown in figure 7.36 w
252

Direct fornL

z-1

l-l

(n)1

.I

7*1

1-311

138

0.9

1337

y(n)

T attitr farxn:

Figure 7.20:

7.25
y(n)

y ( n - l ) + 0 . 1 1 y ( n - 2 ) + z(n)

<H

- 1

a2

-0.11

6o

Type 1 representation:

0
0.11

1
1

<? =

0.11
1

9 =

9 =

0.11
1

d= 1

Type 2 representation:
=

0
1

0.11
1

d- 1

Refer to fig 7.23

7.26
y(n)

y ( n - l ) + y ( n - 2 ) + *(n)

ai = a2

60

Type 1 representation:

0 1
1 1

'

0
1

>

1
1

' 0
1

d=l

Type 2 representation:
F =

0
1

1'
1

"1 "
l
253

d= 1

Figure 7.21:
Diagonal form (see example 7.5.2)
F =

1 + 75
2

l->/5

3+75

9=

<7 =

""75

Cf= 1

3-\/5
2

7.27
vi(n+l)
v2(n+l)

0
1

y(n)

0.11
1

= [ 0 1 ]

^i(z)

zV7(z)

y(z)

=>ff(z)

0.11
0

f2(")

v 2 (n)

+ i(n)

0.11V2(z) + 0.11X(z)
Vi(z) + Vb(*) + X(z)
V2(z) + X(z)
Y(z)
X(z)
1
l-z^-O.llz"2
n
12

fc(n)

1 - l.U-1
n
11

1 + O.lz- 1

^(U) + i(-0.ir]u(n)

7.28
v(n+l) =

Q2

ai

254

Jl() +

9i
92

x(n)

x(n)

x(n)

+ )

^y(n)

Figure 7.22:

|A/-ZI =

A C*i
a?

2
A Qj

= (\-aly-(Q2y

(A - a j ) 2 = (o 2 ) 2 => A =

QI

2 _

=o

ja2

7.29
Refer to fig 7.24 To obtain the transpose structure, we reverse the directions of all branch transmittances and interchange the input and output. Thus we obtain the structure shown below:
Refer to fig 7.25

7.30

(> H{Z) = i: 0 t:lto6?:^

Refer to

fi 7 26

(b)
ai

-0.8

a2

0.64

60

255

Type I Realization:

Type 2 Realization:

x(n)

Figure 7.23:

6,

v/2

2
0.25

Type 1 representation:

0
-0.64

1
0.8

9 =

-0.39
0.8-^

d= 1

Type 2 representation:
0 0.64
1 0.8

F =

-0.39

i = [ 0 . 8 - f J' 9-

<f = 1

Refer to fig 7.27

(c)
(2)

Therefore, h(n)

l-0.707z-1-|-0.25z-2
1 - O.82-1 + o . 6 4 z - 2
0.609(1
-cos()z-1)
25
64 ' l - 2 ( 0 . 8 c o s | ) z - 1 + (0.8) 2 2- 2
-0.208sm(f)z-1
+l - 2 ( 0 . 8 c o s ) z - 1 + ( 0 . 8 ) 2 z - 2
0.39l6(n) + [0.609(0.8) n cos(-n) - 0.208(0.8) n sm(-n)]u(n)
256

Figure 7.24: State-space realization of coupled-form structure

From the state-space representation, we have
h ( n ) = [ -0.39

0.8-

0
0.64

1
0.8

n-l

u{n-l)

+ 6{n)

(d) Coupled-form, state space realization:

^
Hence, Ak
pt

0.0464 + jO.255
0.0464 - ;0.255
+
""
r - (0.4 + J0.4V3)
z - ( 0 . 4 - jOAy/3)
= 0.0464 + jO.255
qkx = 0.0464
qti = 0.255
= 0.4 + ;0.4>/5
afci = 0.4
afc2 = 0.4v^
+

Therefore,
F =
(e)

0.4
-0.4>/3

-0.4>/3
0.4

<*)*<*> = l:$-->Refer

-It

0464
255

9-

d= 1

to fi 7 28

*-

(b)
F =

0
-0.64

1
0.8N/2

1
9 =

[ 1 1

:c)

257

' -0.39 "

9 -

0.3 V5

d= 1

Figure 7.25:

h(n) = [ -0.39

0.3\/2 ]

0
-0.64

1
0.8\/2

9 =

0.212
0.133

n-l

u(n-l)

+ 6(n)

(d)
F =

OAy/2
0.4^2

-0.4\/2
0.4v/2

9 =

7.31
(a) H{z) = T ^ - i - o . 8 ^ - + o . 6 4 . - '

(b) //(z) = T ^ T + UlXXSZ.-

Refer l fi 7 29

'

(c)
#(*)

ai

1 + z- 1
1 - 1.63Z" + 1.21r~2 - 0.32z" 3
-1.63

a2

121

a3

-0.32

60
b,

=
=

1
1

258

d=l

Direct form II:

Transposed form :

-Gy-

y(n)

x(nj

x(n)

-y(n)

$^S^0
D

where a=0.8, b V I /2, c = 0.25 and d = -0.64

Figure 7.26:
Type 1 representation:
F =

0
0
0.32

1
0
-1.21

r o]

0
1
1.63

'

0
1

" 0.32
-1.21
2.63

d= 1

Type 2 representation:
F =

0 0
1 0
0 1

0.32
-1.21
1.63

" 0.32
-1.21
2.63

9-

' 0"
0
1

d=\

7.32
H(z)
where ifi(z)

=
=

0.5ffi(*)ff2(*)
1 + 0.4z _ 1 - l.2z~2 + 22~ 3

H7(z)

l-0.6z-l+0.2z-2

+ Q.5z->

H(z) can be realized as a cascade or two lattice filters (an all-zero lattice and an all-pole lattice)
or as a lattice-ladder filter. Let us choose a cascade configuration:
For H\(z) we have
A3(z)

1 + 0 . 4 z _ 1 - 1 . 2 2 ~ 2 + 2z" 3

B3(z)

2 - 1.2r'I+0.4z~2 + r"3

259

Type I Realization:
039

V,(B)

-0.64

ss *>
!

7-1

Yfn)

i-T +\y(n)

O^VJ^V^

0.8

Type 2 Rrattztkar

Figure 7.27:

*s

M*)

_
=
=

B2(z)
*2

Ai{z)

=
=
=
=

2
A3{z) - Jb33(2)
,

14

-1

-2

3
15
2
3
A2(2)-*2B2(2)

>-"'

14
~15

F*(z)

l - 0 . 6 2 - 1 + 0.22 _ 2 + 0.5

Gs(z)

=
=

*i

For Hi(z) we have

*3

Fiiz)

0.5 + 0.22 - 1 - 0 . 6 2 _ 2 + 2

0.5
F3(z) - k3G3(z)
l-k*

,
1

14

15
260

-1
2

+ -2

-2
'

Direct form II

Transposed form :

-y(n)

x(n)

where a=0.8\/"2. b-\/2/2,c = 0.25andd = -0.64

M2>
Figure 7.28:

G2(z)

Fi(z)

14
25

-77

The all-zero lattice has reflection coefficients

k2
*1

14
15

The all-pole lattice has reflection coefficients

*3

k2
14
25
261

_2

2
3
F2(z) - k2G2(z)
\-k\

k2

* :ii

14 . i

Parallel:
J2L31

Cascade:
x(n)_

< * >

''

ri

K3

1/2

jjsL

0.5

x(n)

-0.64

\ffi

G>
Figure 7.29:

Refer to fig 7.30

7.33
|A/-| =
>2 _

,-1
21

-0.64

(a).

2.96

ali.

A
0.81

A + 0 81

-1
A-l

= A(A-1) +0.81 = 0

_ o => A1)2 = 0.5 j l . 5 0

(b)

*W = *w-)-* = x ? r ^ r 5 3 i

(*) = z 2 _ 0 . 8 1
z+
262

z-\
1

-0.81
z

<y

1/2

y(n)

1Figure 7.30:

l-z-1
I - 2 - 1 -t- 0.81r~ 2

0n(n)

*12(*)

2
(0.9) n cos(0.98n)- -sin(0.98n) u(n)
o
-0.8*
2 2 - z + 0.81

4>i2(n)

i(0.9) n 5t'n(0.98n)u(n)

*(*)

02i(n)

z* - z + 0.81
-1.08(0.9) n sm(0.98n)u(n)

$22(*)

z2
z - z + 0.81

<^22(")

2
(0.9) n cos(0.98n) + -5t*n(0.98n) "(*)
0

-1
0.81

=
= 1
= 0
-1

1 - Z "

1 - z " 1 + 0.81z~2
(0.9) n cos(0.98n)u(n) + [-0.5cos(0.98(n - 1)) - sin(0.98(n - 1))] (0.9)n"1ti(r - 1)

the zero-input repsonse yi%{n) we have

*[(*)-t/(0)]

= EYXz)
263

V{z) =
Y*i(:)=9V(:) =
V(n) =

( / - ) " ^11(0)
li*L-D-lzW)
rl{y-(:)}

For the zero-state repsonse yi3(n) we have

zV_(z) = FV_(z) + qX(z)
Y*M
= lV{*) + X{z)
YMM = g(zL-F)-1qX(z)

+ X(z),

.-i

V(n) =
y{n) =
=

z-l{YMt(z))
y(n) + y(n)
(0.9)n [2cos(0.98n) + l.l'n(0.98n)] ti(n)

(e) Refer tofig7.31

Figure 7.31:

7.34
Using the same method as in 7.33 we obtain the following answers:
(a)
C(z) =

*2 + 3z + 2

264

roots: z

2,-1

(b)
*(n) =

0n(n)
<p2l(")

<^12(")
022(")

[-l(-2)"+2(-l)B]u(n)
[-l(-2)"+2(-l)"]ti(n)
-2<^i 2 (n)
[2(-2)"-(-l)"]u(n)

<t>\i(n)
<t>i2{n)
<t>2\{n)
<!>22{n)
(c)

H(z)

Mn)

1
z + 3* + 2
[-(-^"-^(-lrMn-l)
2

(d)
y(n)

y(n) + y()

h ( - 2 ) " + ( - l ) > ( n ) + [ ~ ( - 2 r - i ( - l ) n + i]u(n)

(e) Refer to fig 7.32

*-y(n)

Figure 7.32:

7.35
Using the same method for the solution as in 7.33, we obtain the following answers:
(a)
C(z)

z2 + 0 . 6 * - 0 . 0 7
265

roots: z

0.1,-0.7

(b)
<t>n(n) = 4>22(n) =
* ( n ) = * 2 i(n)

[i(0.1) B + i(-0.7)"]u(n)
[j(0.1) n - (-0.7)"]ti(n)

(c)
H(z)

h(n)

1
+1
z-0.1
6(n) + ( 0 . 1 ) n _ 1 u ( n - l )

(d) jrfn) = (0.1)"u(n) + [ ^ - ^(0.1)"]u(n) Refer to fig 7.33

-0.3

Figure 7.33:

7.36
(0
ff(z)
a2

1 + z- 1
1-0.9Z" 1 -*-0.08z" 2
-0.9
0.08
266

6o =

*i

Type 1 representation

1 1
0
-0.08 0.9

[ 01
=

[ -0.08 ]

1.9

d= 1

Type 2 representation:

0 -0.08 1
1
0.9

[ -0.08 1
1.9

[ 1 1

d= 1

(b) Refer to fig 7.34

Parallel:

i)

AlT

Jr
x(n)

"
!

-1 +

-11/7

x,( N

-J

18/7

0.8

Figure 7.34:

H(z)

1 + r- 1

(l-o.u-^i-o.sz-1;
7

l-O.lz-1
F =

0.1
0

0
0.8

1*

' =[oi]'

9=

Cascade:
H(z)

1-0.8Z" 1

1 + .-1
(l-O.Sz-^l-O.lz-1)
267

d=l

=
1
2

Hy(z)H2(z)

=
=

[0.8],
[0.1],

H(z)

h(n)

& = [l],
2 = [1],

,=[1.8],
2 = [0.1],

<*i = l
d2 = l

(c) Method 1:
11

7
1-O.lz"1

18
7

1-0.8*"1

11

16

- y ( 0 . 1 ) " + y ( 0 . 8 ) " u(n)

Method 2: From (a) we have,

v(n + l)

Fv(n) + ox(n)

y( )

v'v(n) + <fx(n)

V(*)
Hence, H(z)

=
=

'(*-JD" 1 !***)+ **(*>

^ ( z / - F ) " 1 ? + <f

0.08
1.9

tf(z) =

z
0.08

-1
z-0.9

X(z) = l

+ 1

l + z- l
l-O.gz-i+O.OSr-2
which is the same system function as in (a).
H(z) =

7.37
(a) H(z) =
(b)60=l,

1+*'" 1
fr

r,

3'

<i

" 0
1
8

3
- 4 ,

1 '
3
4 -

.=*
9 =

d= 1

<7 =

M
*(*) = *(*/-) - 1 =

Z - \

*~*

(l-^-'Ki-i*-1)

0(n) =

-1

2_ 3 + i
*
4*^8
2

(l-ii-'Ki-i^-1) .

- ( | ) " u ( n ) + 2(J)"u(n)
- ( i ) " + 1 u ( n ) + J(J)"u(n)

4(J)"ti(n) - 4(J)"u(n)
2(i)"u(n) - (J)"tt(n)

(d)

*<*)=[- Ml

-tim

+ 1

#(*) = i - i ; - u ^ - 3
(e) det(z7 - ) = z2 - \z + = (z - %){z - \) = 0. z = , J are both inside the unit circle.
Hence, the system is stable.
268

7.38
H(z) = gt(zl-;F)-lq

+d

*<*>=[o ! ] [ : , ;_v][?"' + 1
H(z)

l-z-i-O.llr-2
li
J_
12

l - l . l * ,-- 11
Therefore, h(n)

,
"r

12_

1 + O.lz" 1

7.39
(a) Refer to fig 7.35
(b)

i<n>

x(n)

>

O^
-5/16

Figure 7.35:
H(n+1) =

1(1) =

0
16

_-L
16

_ Li

1
l

v(0) +
269

v(n) +

*()

x(0) =

y(n)

Mo) =

[ i i 2]v(n) + 2x(n)
[ i i 2] (0) + 2x(0) = 2

Mi) = [ H 2]t;(l) + 2x(l) = 2

proceeding in the same way, we obtain
h(2)

fc(5)

_5
8
0
25
128
25
128

(c:

H{z) =

fc'^Z-)-*+<*
2-f 4 z - 1 + 2 z ~ 2

Therefore, y(n)

- y ( n - 1) - ~-y(n - 2) + 2x(n) + 4x(n - 1) + 2x(n - 2) ,

16

'(

>

(d) For x(n) = S(n), we obtain y(0) = 2,y(l) = 2,y(2) = | , etc. These results check with (b)
(e). Refer to fig 7.36

x(nj\

Figure 7.36:

270

7.40
(a) Refer to fig 7.37

Figure 7.37: All-zero lattice

^ ( n + l)
V2(n + 1)

=
=

x(n)
kix(n) + vi(n)

V3(n + 1)

kix(n) +kik2vi(n)

=
=

*/y_ 1 x(n) + * i * N - i V i ( n ) + *2*JV-it>2(n)+ +vjv-i(n)

x ( n ) + * i v i ( n ) + * 2 M n ) + + *NfJv(n)

vx(n + l)
y(n)

F =

+ V2{n)

0
1

0
0

. 0 0 1
. 0 0

*i*2

*1*3

*2*3

kiktf-i

:
=

1 0 J

*2*N-I

= [ kx k7

. . . kN ]

(b) If we proceed as in (a) we obtain

271

d=l

F =

-ks

k\

-fc 2

1 - k\

-k2

1 to

"*1

*2

9 =

k$

1 -t

9* =

*N

"*3

*N-1

-* J

*N-1

d=l

7.41
(a)
For positive numbers, range is
01.00...0x21001
n
or 7.8125xl0~ 3
negitive numbers
10
x2l 01

ii^i

01.11...1x20111
u
2.5596875xl0 2

10.00...0x2 0 1 1 1

n
or - 7 . 8 1 6 3 x l 0 - 3

n
-2.56xl02

(b)
For positive numbers, range is
01.00...0x210000001

01.11...1x20imm

3.4028234X103*

10.00...0x2 0 1 1 1 1 1 1 1

-3.4028236xl0 3 8

23

23

or 5.8774717xl0~ 39
negitive numbers
10. I l . . . l x 2 1 0 0 0 0 0 0 1
23

23

or -5.8774724xl0~ 3 9

7.42
(a) Refer to fig 7.38

HR(Z)
,
poles z p l i 2

(l + o i * - 1 + o -220\ - l

-ax y/a\ - 4a2

**

>

for stability
(i)a\-4a2
if ai > 0,

- a i - yja\ - 4a 2
2

"

=* Ja\ - 4a 2

<

272

2 - ai

l?)

,-i

-1

Figure 7.38:

if ax < 0,

ai < 2 and ai 02 <

- o i - \Ja\ - 4o 2
<
2

Jo f - 4 o

1
1

< 2-fai
=> ai > 2 and ai + 02 > - 1
2

( .. )( _a^ )2 + ( v ^ E 2 )

<

<

Refer to fig 7.39. The region of stability in the ai 02 plane is shaded in the figure. There are
nine integer pairs (01,02) which satisfy the stability conditions. These are (with corresponding
system functions):
(0,-1)
(0,0)

HRM
HRM

=
=

(0,1)
(1,0)

HRM
HRM
HRM
HRM

=
=
=
=

HRM
HRM
HRM

=
=

(1,1)
(2,1)
(-1,0)
(-1,1)
(-2,1)

(I-;-2)-1
1

(1+*- 2 )- 1
(1 + z- 1 )- 1
(l +

-l+2-2)-l

(l + 2*-1 + z-2)"1
(I-*"1)"1

(l-z-l + * - V
( l ~ 2 z - 1 + 2~2)-1

(b)
HRx{z)

HRMHR7{Z)

HR6(Z)

HR4(Z)HR4(Z)

HR9(Z)

HR7(Z)HR7(Z)

273

The**bkvnaf(4.a )

Figure 7.39:
(c) Only the following cases can make h(n) FIR:
(i)
hR(n)

6{n)
N

Then H{z) = ]T "

=o
N

y(n) = 5Z x ( n ~ 0
i=0

(ii)

hn(n)*hF{n)
Then H{z)
y(n)

=
=
=

S(n)
1
x(n)

(d) see above.

7.43
Refer to fig 7.40
Note that 4 multiplications and 3 additions are required to implement H\(z).
of Homer's method is in evaluating H\{z) for a specific 20. Thus, if
Hi(z)

=
=

bo + bohz-1 + 60&i&2*~2 + M i M 3 2 ~ 3
60 + 2 - 1 (6 0 6i + * _ I (& 0 M2 + z~1b0bib7b3))
274

The advantage

rvv4

Figure 7.40: Structure of H\{z)

the 3 multiplications and 3 additions are required for the evaluation of 7.1 in the field of z.
If the various powers of z are prestored, then Horner's scheme has no advantage over the direct
evaluation of 7.1. Refer to fig 7.41
This requires 4 multiplications and 3 additions. The linear-phase system is written as
H{z) = z2az + za2 + oi + z~la0 + z~2a\ + r " 3 a 2 + *~ 4 a 3
By applying Horner's scheme, we can rewrite this as
H{z) = z3(a3 + z'\a2

+ z~l{ax + z~\aQ

+ z~\ax

+ z~l{a2 +

z-la3))))))

Assuming that z'1 and z are given, a direct evaluation of H(z) at z = ZQ requires 8 multiplications
and 6 additions. Using Horner's scheme based on 7.43, requires the same number of operations
as direct evaluation of H{z). Hence, Horner's scheme does not offer any savings in computation.

7.44
(a) When Xi and x2 are positive, the result is obvious. If Xi and x2 are negative, let
xi

0 r\\ n2

0 n\ n2 .
x2
X3

where c

nb + 0 0 0 . . . 0 1

=
=

0 m\ n2
1 mi rn2 . mb + 0 0 0 . . . 0 1

Z\ + *2

-0n10...0 + 0m10...0 + c

00 n 2 . n*+ 00 m 2 . . . m b + 0 0 0 . . . 0 1 0
275

Figure 7.41: Structure of H{z) = b0z~3 + i 0 M ~ 2 + t o ^ i M " 1 + M 1 M 3

If the sign changes, there are two possibilities

(i)

m = mi
=>. m = mi
. 1
=>lil
=> |x3|
(u)(ni = l , m i = 0 , c = 0)
=>(|0ni0 ... 0)io|> 2
and |cio|
=*> | i 3 |

=
=
>
>
or
or

>
>

0
1
1
. . 1
,X2|>
j '
2
1, overflow
(nj = 0,mi = l,c = 0)
( | 0 m i 0 . . . 0) 1 0 | > ^
1
2
1, overflow

(b)
*1

*2

*3

*1 + *2

0100
0110
- 0 1 1 0 = 1 0 10
10 10, overflow

*1 + *2 + *3

0 10 0, correct result
276

7.45
(a)
H(z)

l#(0| 2

-a+ 2
l-az-1

-l

1 - ae-iw '
(a + cosw)2 + (sinw)2
(1 acosw)2 -f (asinw)2
a2 2acosw + 1
= 1 Vw
1 2acosw + a 2

(b) Refer to fig 7.42

(c) If |d| = | d|, where d means the quantized value of a, then the filter remains all-pass.

y(n)

.=a
-1

Figure 7.42:
(d) Refer to fig 7.43
(e) Yes, it is still all-pass.

7.46
(a)y(n)=[2(i)"-(i)"]ti(n)
(b) Quantization table
'"32
277

x= 1

y(n)

Figure 7.43:
31

^
> X

2329 ^"
32^Z
32 "

>
>
>

29
32
27
32

15
16
14
X =
16

x =

1
32

"1 + 3 l

14
16

* - - >

Therefore I(n)

(l.^.'^.O

y( n )

- ^ y ( n - ! ) + *(")

_, *
v{n)

12 7

= lI'ie'Ii'Ti'Te' ' '-]

f 3 7_ 15 _31_ _63_ "1
\}'4,16'64'256'1024'"'J

y{n)

-/ x

f 3 7 12 16 rt ^ 1
= \}'4'^'6l'25i''0'- j

y{n)

0 0

occur when number becomes small.

278

7.47
y(n)

0 . 9 9 9 y ( n - l ) + e(n)

e(n) is white noise, uniformly distributed in the interval [p, p-]

E{y2(n)}
(l-0.9992){y2(n)j

0.9992{y2(n-l)} + {e2(n)}

E {e 2 (n)}

/ .

Therefore, E{y2(n)\
' iy v "

_* A

r where A = 2 " 8
12

(4)2
2
1 2 v 2 8 ' 1 - 0 ..999
6.361x10 - 4

7.48
(a) poles zpi 0.695,
(b) Truncation

2 p2 = 0.180 Refer to fig 7.44

y(n)

11

D
0.69.5

ILLS.

Figure 7.44:

0.695
279

- = 0.625
8

0.180
poles zp\ ~ 0.625,

= 0.125

2 p2 = 0.125

(c) Rounding

poles zpi = 0.75,

0.695

- = 0.75
8

0.180

i = 0.125

zp2 = 0.125

(d)
|0.75-0.695|
Rounding is better
\Ha{w)\

<

|0.695-0.625|

[(1.483 +1.39costi;)(1.0324 + 0.36cosu;)]-*

\Hb{w)\

[(1.391+ 1.25cosu>)(1.0156 + 0.25cosu;)]-*

\Hc(w)\

[(1.563+1.5cosu>)(l.0156+ 0.25cosii;)]-*

7.49
(a)
Hi(z)

1-22"1

M) = {l,-i}
- i1\)- i
(1--Z-

H2(z)

M")

(i)nu(n)

H3(z)

(i + j z - 1 ) - 1

M")

(-i)nu(n)

Refer to fig 7.45 Cascade the three systems in six possible permutations to obtain six realizations.
(b) Error sequence ej(n) is uniformly distributed over interval (^2"*, 52"*). So o\% = 2- for
any i (call it c\)
(c) consider cascade H\ - Hi - Hz Refer to fig 7.46

= {i.o.i,o.(l)'.o,...}
n=0

<r 2
e

Li-(A)2
3.0745<r2

280

n=0

1
i-(i)2

Figure 7.45:
using similar methods:
H\ Hi H3

2.0745crl

Hi H\ H3

3.3882<7^

Hi H3 Hi

3.25880-*

H3 H1 Hi

3.26270-2

H3 Hi H\

3.32160-J

7.50
y(n) = Q[0.l6(n)] + Q [ a y ( n - l ) ]
(a)

C?[0.U(n)]

y(o) =

<?[o.i]=i

y(")

y(i)
y(2) = y(3) = y(4)

= <^ =
=

no limit cycle
(b)
y(n)

Q[0.1<5(n)] + Q[0.75y(n-1)]
281

l(n)

2<P)

Figure 7.46:

y(0)

Q[0.1] = g

y(i)

vl

32J

y(2)

vl

32J

y(3) = y(4)

1
8

limit cycle occurs

7.51
(a) al = r ( 0 ) = 3 => Ax = ^ (b)
A

2-6
A^
12
12x2 12

so SNR

10/oyio

1O/O0 1 O (12X2 1 2 )

46.91dfl
282

3 (n)

(c) left-justified.

a] = <**<)+ <,;,*'()
n=0
2

Now* .

12 V 2 8

el

and V / i 2 ( n )
V

(4)

n=0
2

so o-J

- = - ,
1 - 0.752
7'
7 V12 1 2* ;

12l26' ;

17
344,064
and SNR

=
=

I0logie0-r

43.064B

7.52
Define pe rcos9, pt rsinO for convenience, (a)
-P>y(n-

l) + ci(n) + x(n) + p e t>(n- l) + e 2 (n)

p , i ; ( n - l) + e 3 (n) + p c y ( n - l ) + e 4 (n)

=
=

v(n)
y(n)

(b)
-p3z-lY(z)+El(z)

+ X(z) + pcz-lV(z)

+ E2(z)

V(z)

p , * " ^ * ) + *(*) + ^ 2 - ^ ( 2 ) + 4 (z)

y(z)

y(2)

l-2pePz-Cr^[X(2)

ffi(z)X(*)

ffi(z)[i(*)+

+^ 2 (2)[ 3 (*) + ^4(*)]

when # i ( z ) and // 2 (z) are as defined in the problem statement
hi(n)

h2{n)

=
=

pi^^-isindnuin-l)
sinv
r sinn6u{n - 1)
rnsinndu(n)

- rnsin(n+

=
=
=

l)0u(n) + p c ^ - T r n - 1 s i n 0 n u ( n - 1)

sin?

5tn0

rn
^( ) + -r-r[s**n(n + 1)0 - cos0s:nr0]u(n - 1)
sind
6(n) + rncosn8u(n - 1)
n

rncosn$u(n)

(c)

A2
2

*-*

283

+ l ( 2 ) + 2(z)]

*(*)]

l
12 v

-6\2
;

2-26

2* 2 ]T/ l 2 (n) + 2<7 2 > 2 : (n)

r=0
oo
2n

2^[r

n=0

sin2n0 + r2,Wn0]

n=0

= 2*1 l _ r 2 n
2-26

1 - rIn

7.53
(a)
Mn)

(^) n "(n)

h,(n)

(i)nM")

l2(-)"-(-)W

h(n)
\.2

= 2<r 2 1 / 2 (n) + 2<r 2 2 > 2 2 (n)

n=0

64

35

n=0

16 ,
15

(b)
2

<r:

^ , , 2 ,

^-i^w-^E^
64

,T* 4 - V
35*' 1 + a 4 7 ' 2

7.54
Refer to fig 7.47

'l

<

26

" Vi
'(M-

- D<

(M- -D 2
12

7.55
//(;)

B(z)
A(z)
284

26

ifn)

^1-2

- iMJ Q^M-l (n)

Figure 7.47:
, (l-0.8e>T)(l -0.8c-**)
ri
(l-0.5z-i)(l+$r-i)
(l-0.8e>T)(l-0.8e->*)
Hl{z)H2{z)
(a)
--1

At w = 0 , z

*-Jw

_1

(l-0.8eJT)(i-0.8e--'T)

(1-0.5)(1 + J)
Gi

.1381

(l+0.25)(l-f)

'2(l-0.8eH)(i-0.8c-Ji)
G2
(b) Refer to fig 7.48.
(c) Refer to fig 7.49.
Refer to fig 7.50.
Refer to fig 7.51.

285

.7920

Direct for m I:

Direct form II and cascade structure:

_jn)

x(n)

-3/8

-126.
-1

-5/32

AJfu

Figure 7.48:

7.56

*i

-3
8_
4r
27
32

*2

9
5_
T T
32

Refer to fig 7.52a.

(b)
A{z)

*2

*1

1
(l-0.5*-)(l+J*-)
1

6
_1
6
1.
~5

286

Cascadetorm, mputee response

15
1
05

'

V"

nf
10

100

Cascadetorm,step response
1.5
1 4
13
] 1.2

1 1

10

20

30

40

50
> n

60

70

80

Figure 7.51:

(n)

fln)

r-\

fjn)

/^"N

Figure 7.52:

289

f,( B ).y(n)

90

100

Direct torm I. impulse response

Direct torm I, step response

90

Figure 7.49:
Refer to fig 7.52b.
(c) Refer to fig 7.53.
(e) Refer to fig 7.54.
(f) Finite word length effects are visible in h(n) for part f.

7.57
(a) Refer to fig 7.55.

Hx{z)

H3(z)

15
16
9
10

1 -

1*-

-1

S3
80

14

(b) Refer to fig 7.55.

(c) N = 100;
x = [1 zeros(l,N-l)];
c=15/16;
al = 9/10;
a2 = 83/80;
pi = 0.5;
p2 = -1/3;
y l ( l ) = al*x(l);
y2(l) = a2*x(l);
y(l) = c*x(l) + y l ( l ) + y2(l);
for i=2:N,
yl(i) = al*x(i) + pl*yl(i-l);
y2(i) = a2*x(i) + P 2*y2(i-1);
287

-1

100

Direct form II. impulse response

15
1
05
0
-0.5

1v10

20

30

40

50

60

70

80

90

100

Direct form II. step response

100

Figure 7.50:
y(i) = c*x(i) + yl(i) + y2(i);
end
(d) Refer to fig 7.56.

288

Figure 7.53:

IRtorpan b

IRtorpan a

0.5

0 -

-05_

-0.5

Figure 7.54:

290

Parallel form structure

{>
*(n)

jw_

Hj(z)

H^Z)

Parallel form structure astng 2a4-order coupled-form state-apace tectiooa

1

1/2

A^

x(n)

-d2-

Figure 7.55:

Parallel torm impulse response

i
i
i
i

25

~ 2
?15
1

05

ok- 10

i
2

t
0

5 0 6
>n

100

Pa/aHel structure coupled torm impulse response

1

08
c*0 6
A

! 04

02

nJ

10

5 0 6
> n

Figure 7.56:

291

100

Chapter 8

8.1
(a) To obtain the desired length of 25, a delay of 25jpi
Hd(w)

le~jUw,

0,

hd(n)

Then, h(n)

12 is incorporated into Hd(w). Hence,

0<\w\<^
6

otherwise
J

Hd(w)e->"ndw

i- (

5n|(n-12)
? r ( n - 12)
hd(n)w(n)

2* J-i

where w(n) is a rectangular window of length N = 25.

(b)H(w) = n 4 = o M n ) e " J U ' n => Plot \H{w)\ and IH(w). Refer to fig 8.1.
(c) Hamming window:
90
0

-501-

' -100

150
0

2-

0.05

0.1

0.15

0.2
0.25
0.3
>Frq(Hx)

0.35

0.4

0.45

0.5

Vwww
0.05

0.1

0.15

0.2

0.25

0.3

Figure 8.1:

w(n)

(0.54-0.46co5 nn

293

0.35

04

0.45

0.5

h{n)

hd(n)w(n)

for 0 < n < 24

Refer to fig 8.2.

(d) Bartlett window:

2-

-As|
0

K K NNNN"

0.09

0.1

\|

s| N

N
0.15

0.2
0.25
0.3
>Frq(Hi)

0.35

0.4

0.45

0.5

Figure 8.2:

,()

!_ 2(2^12)

Q<n<24

24

Refer to fig 8.3.

0.06

0.1

0.15

0.35

0.4

0 45

a o
-2

0.2
0.25
0.3
>Frq<Hx)

Figure 8.3:

294

0.5

8.2

Hd(w) = le->12-,

M<.

l<\w\<*

0
~

tf

If

o,g<IH<3

M")

u^e"-"""^

l7lj

*(n-12)

*(n-12)

(b) Rectangular window:

u^(n)

=
=

1,
0<n<24
0, otherwise

Refer to fig 8.4.

(c) Hamming window:

0.06

0.1

0.15

'

02
0.25
0.3
>fm*Hz)

'

0.35

0.4

Figure 8.4:

ti/(n)

h(n)

(0.54-0.46cos^J

12

hd(n)w(n)
24

H(w) = J>() jwn

n=0

Refer to fig 8.5.

(d) Bartlett window:
w(n)

1 - ("

12)
12

295

0<n<24

0.45

0.5

_J

0.06

0.1

0.15

02

0.25

0.3

0.35

0.4

0.45

0.5

Figure 8.5:
Refer to fig 8.6.
Note that the magnitude responses in (c) and (d) are poor because the transition region is
wide. To obtain sharper cut-off, we must increase the length N of the filter.

8.3
(a) Hanning window: w{n) - \{\ - cos^)t
0 < n < 24. Refer to fig 8.7.
(b) Blackman window: w(n) = 0.42 - 0.5cos*J + 0.08co*^. Refer to fig 8.8.

8.4
(a) Hanning window: Refer to fig 8.9.
(b) Blackman window: Refer to fig 8.10.
The results are still relatively poor for these window functions.

8.5
M

Hr(w)

4,

ffr(0)=l,

2MnWu,(^fi-n)]
l

n=0
1

Hr{\)=\

3
2]T/>(n)co*[ti,(--rO]
n=0
1

At u = 0, tfr(0) = 1

2^2 h(n)cos[0]

2[/>(0) + Ml)]

n=0

296

(1)

0.06

0.1

0.15

02

0.05

01

0.15

0.29
0.3
>Frq(Hi)
i

0.2
0.25
0.3
> Frq(Hx)

0.35

0.4

0.45

0.35

0.4

0.45

0.5

0.5

Figure 8.6:

7r

,ir.

Atu,= -,ffr(j) = -

2Mn)cos[|(?-n)]

n= 0

-h(0) + h(l)
Solving (1) and (2), we get
h(0)

0.354

Mi)

0.073 and
0.427

h{2)

M3)
Hence, h(n)

(2)

Ml)
= MO)
=

{0.073,0.427,0.427,0.073}

8.6
M = 15.ffr< ) - (

ff-M

o,

* = 4,5,6,7

M ^ ^ ) + 2^Mn)cosu;(^-i-n)
n= 0

h(n)
/x(n)

=
=

li(M-l-n)
/i(14-n)

i/r(u;) = h(7) + 2*52 h(n)cosw(7

~ n)

oo

n=0

Solving the above eqn yields,

h(n)

{0.3189,0.0341, -0.1079, -0.0365,0.0667,0.0412, -0.0498,0.4667

0.4667, -0.0498,0.0412,0.0667, -0.0365, -0.1079,0.0341,0.3189}

297
l

KKK
-2

l\ K

K K\J N

\ NN N N

0.09

0.1

0.15

02

0.29
0.3
>Frq(Hz)

0.35

0.4

0.45

0.5

Figure 8.7:

O,L
f 1.
* = 0,1,2,
M = 1 5 . / / r ( ^ ) = { 0.4, * = 4
15
{ 0,
* = 5,6,7

, . M 1.

^-s . . .

Hr(w)

H) + 2

h(n)
h(n)

=
=

/(M-l-n)
/(14-n)

Hr(w)

M?) + 2

.M 1

/i()c05u;(

n)

n= 0

/i(n)cosu;(7 - n)

n=0

the above eqn yields,

h(n)

{0.3133, -0.0181, -0.0914,0.0122,0.0400, -0.0019, -0.0141,0.52,

0.52,-0.0141,-0.0019,0.0400,0.0122,-0.0914,-0.0181,0.3133}

VM =

Hence, H(F)

=
298

dxa(t)
dt

j2*Fe>7ltFt
j2irF

!-ioo r

\ hhlMMM^I !'

-2
-A.

N
0

0.05

0.1

y N Vl N N N

N \

^
0.15

0.2

0.25
0.3
> Frq(Hz)

036

0.4

0.45

0.5

Figure 8.8:
(b)
\H(F)\

2irF

LH{F)

I,

F>0
F<0

2'

Refer to fig 8.11.

(c)

#(u>)

jtx;,

|//(ti,)|

|u;| < TT

w < 0

Refer to fig 8.12.

we note that the digital differentiator has a frequency response that resembles the response
of the analog differentiator.
(d)
y(n)

z(n)-x(n-l)

H{z)

l-*"1

H(w)

l-e-'w

e"Jt(2;n-)

2|n||

|ff(u/)|

m) = I
Refer to fig 8.13.
Note that for small it;, sinj

m y and H(w) % jwe~J$,

299

which is a suitable approximation

Figure 8.9:
to the differentiator in (c).
(e) The value H{w0) is obtained from (d) above. Then y(n) = A\H(xvo)\cos(won + 6 + | - - ^ )

8.9
i

Hd(w)

hd(n)

M n )

we~il0w,

-W-iiowf

J- f

cos7r(n 10),
(n-10)
0,
n = 10
cosirin 10)
(n-10) '

0,

0<w<ir
-TT<W<0

i/d(u;)e--'w'nrfu;
n^lO

0<n<20,n#10

n = 10

With a Hamming window, we obtain the following frequency response: Refer to fig 8.14.

8.10
H(s) haa two zeros at z\ = - 0 . 1 and zi = oo and two poles pi,2 = - 0 . 1 j3. The matched
z-transform maps these into:
lx

c - 0 1 T = e - 0 0 1 = 0.99

z2

e'T

p,

c ( - 0 1 + ' 3 ) T = 0.99e j 0 3

p2

0.99e--' 0 3

=0

300

'

f
g -5
' -10|-15
I

'

'

'

'

0.06

0.1

0.15

02

0.25 0.3
> Fmq(H2)

0.35

0.4

'

0.45 0.5

0.06

0.1

0.15

0.2

0.25
0.3
> Frq(Hx)

'

0.35

0.4

0.45

0.5

Figure 8.10:

Hence, H(z) =

1-rz"1
, u/o = 0.3 r = 0.99
1 - 2rcoswoz~1 + r 2 z - 2

From the impulse invariance method we obtain

H{s)

H(z)

1
s + 0.1+>3

1
s + 0.1-;3
1

l_e-0.lTe;3Tz-l
TCOSWQZ-1
l

1 2rco$tvoz~ 4-

1_e-0.1Te-j3T2-l

T2Z~2

The poles are the same, but the zero is different.

8.11
(fiu - 0|)

Ha(s)

s - ( n -f2/) + nufi/
21-r1
s = T1 + ; - 1

//(z)

(n-ni)

f (i - ^ K i + z-i)

1 2

(f ) (i - r- ) + (ntt - n,)(f )(i - z-)(i + z-1) + fiun,(i + z- 1 ) 2

2(a-^)(l-z-2)
[4 + 2(Q -0) + a(3) - 2(4 - a0)z~1 + [4 - 2(a - /?) + a ] z " 2
where a

In order to compare the result with example 8.3.2, let

u;u

UUT 5
301

B-p/6

2
. 0
A
I
I

-1

"
1

-0 06

-0 06

-0 04

-i

-002
0
0 02
- > Frq(Hz)

004

0 06

0 06

01

Figure 8.11:

wi

il\T
5

Then, (z)

* = "iTo.1^"?

In our case, we have o = ftuT =

/? = ft,T =

(exampie 8 32 j

2tan-- = 2.753
2tan~

= 1.453

By substituting into the equation above, we obtain

H(z)

2.599(1-z"2)
10.599 + 5.401Z"2
0.245(1 - z ~ 2 )
1 + 0.509Z" 2

8.12
Let T = 2
(a) ff(*) = = y(n) = y{n - 1) + *(n) + x(n - 1)
(b)

(c)

|J/(u,)|

t f,

- * < t/> <

(d) The digital integrator closely matches the magnitude characteristics of the analog integrator.
The two phase characteristics are identical.
(e) The integrator has a pole at w = 0. To avoid overflow problems, we would have [x(n)] = 0,
i.e., a signal with no dc component.
302

-1

Figure 8.12:

8.13
(a)
H(z)

H(z)U = i

U + *-- 11\)3

(l + z - 1 ) ( l + 2 r - 1 + r - 2 )
A( l _ i - i l _ I , - i
I2-2)
2
) (
+

( l - J 2 - i ) ( l - J r - i + Jz-)

, &! = 2, 62 = 1, ai = 1, ci
64
(b) Refer to fig 8.15

8.14
(a) There are only zeros, thus H(z) is FIR.
(b)
Zeros: z\

*2

4
3'
3
~4'

*3,4

25,6

=
=

*7

Hence, r2
303

r**

-5.-1

"2'd2-i

Figure 8.13:

*4

Z*.

z.
*7

and H(z)

z " 6 /Z ^J /-- 11 )

. - 6

Therefore, #(tu) is linear phase.

(c) Refer to fig 8.16

8.15
From the design specifications we obtain

6
fp

f.
Assume 1 = 1 . Then, Q p

= 0.509
99.995
4 _ 1
= 24 ~ 6
6 _ 1
= 24 ~ 4
=
=

and ft,

=
=

V =

2
2tannfp = 1.155

"mf

2tanxfa = 2
6
- = 196.5
304

Figure 8.14:

k = 5 1 = 1-732
Butterworth filter: Nm^n

>

-^j- = 9.613 = N = 10

Chebyshev filter: Nmin

>

^ r T T = 5 - 2 1 2 =>

Elliptic filter: stno

sin/)

^min
nun

= 0.577 => a = 35.3

- = 0.577 => 0 = 0.3

>7
fc(smo) k(cos0)
> 17
VL)
m = 3 7 8 = N = 4
k(cosa) k(sinp)

8.16
From the design specifications we have
e =
6 =

0.349
99.995

/, = T = 015
/. = 1 = 0.25
Qp

2tan^-

Qt

w.
2<an-y = 2

5T

it

r:

= 1.019

286.5
= 1.963

305

x(n)'

|
z-1
3

"^v. -i

J.

'
x-1

IJ.

"^L 1/2

1 -1/8

\1

Q^H-Q
B

0.5

1M

Figure 8.15:
>
Butterworth filter: iV
nun
Chebyshev filter: A ^ u
Elliptic filter: A ^ ^

>

( = 8.393
N = 9
logk
cosh~ n
? ,7 = 4.90 => N = b
cosn~lk

t(i) *<\A7i> =>

t ( v /rr-T)'

*<j>

8.17
Passband ripple = ldB => i = 0.509
Stopband attenuation = 60dB => 6 = 1000
w?

0.3TT

u;,

0.35TT

QB

w = 1.019
2*an-f

fi,

2 t a n ^ - = 1.226

- = 1965.226

rj

k = 2 i = 1.203
AT.m m

ft
c o s / r 1 ^ _ 8.277
= 13.2 => AT = 14
cos/i" 1 * ~ 0.627
306

yv = 4

x(n)

<D^)
25/12

(4/3)2 (3/4) + 4 cos 2 60= 481/144

Figure 8.16:

Special software package, such as MATLAB or PC-DSP may be used to obtain the filter coeffi
cients. Hand computation of these coefficients for N = 14 is very tedious.

8.18
Passband ripple = 0.5dB => e = 0.349
Stopband attenuation = 50dB
wp

0.24TT

ID,

0.35ir

ftp

2 r o n ^ = 0.792

Q,

2 t a n - i = 1.226

rj =
k
N.m i n

6
- = 906.1
? i = 1.547
Qp
cosh~lr) _ 7.502
= 7.48 => N = 8
cosh~lk " 1.003

Use a computer software package to determine the filter coefficients.

307

8.19
(a) MATLAB is used to design the FIR filter using the Remez algorithm. We find that a filter
of length N = 37 meets the specifications. We note that in MATLAB, the frequency scale is
normalized to ^ of the sampling frequency. Refer to fig 8.17.
(b)6l = 0.02, 62 = 0.01, A / = & - JJL = 0 .05

12i

01

0.2

0.3

0.4

0.5
>t

0.6

07

08

09

Figure 8.17:

With equation (8.2.94) we obtain

M

-20/og 1 o( N /?r?7) - 13
+ 1 % 34
14.6A/

With equation (8.2.95) we obtain

1.7371
11.166

Ax>(M2)
/(M2)
and M

Doo(6i67) - / ( M 2 ) ( A / ) 2
+ 136
A/

Note (8.2.95) is a better approximation of M.

(c) Refer to fig 8.18.
Note that this filter does not satisfy the specifications.
(d)The elliptic filter satisfies the specifications. Refer to fig 8.19.
(e)

order
storage
No. of mult.

FIR
37
19
19
308

IIR
5
16
16

Figure 8.18: M=37 FIR filter designed by window method with Hamming window

8.20
(a)
h(n)

io.l,2,3,4,5,4,3,2,1,0,..A
10

J2hWz~n

H{z) =

n=0

H(w)

2 _ 1 + 22" 2 + 32" 3 + 4 2 ' 4 + 52" 5 + 42~ 6 + 32" 7 + 22" 8 + 2"

e~}9w[2cos4w + 4cos3w + 6cos2w + 8cosw + 5]

(h)\H(w] = \2co$4w + 4cos3w + 6cos2w + Scosw + 5|. Refer to fig 8.20.

(c)
fc0 = h ( l ) = l
r hh (0)

= / i 2 ( n ) = 85
n= l
8

rhh(l)

J^Mn)M + l) = 80
n= l
7

n= l

rhh(0)
rhh(l)

rhh(l) 1 f ai 1 = f ~rhh{l)
rhh(0) \[a2 \
[ -r**(2)
i

=>ai =

8.21
(a)
dc gain: Ha{0)
309

272
-l65'fl2

124
l65

Figure 8.19:

3dB frequency:
or

\Ha{jQ)\ 2

a2 + n?

For all Q, only

tf(joo)

1i
2
1_
2
a
0

ha(r)
=>e-at

-fc(0) = e
t
=

-l

=> r

ha(nT)
e-nTu(n)
1
=
1

H(w)

1_

H(0)

2
3dB frequency: \H(w
c)\*
("<)|
Ta

(1 - Q coswc)

:i

H(u;)U = 0
1
1 - e-aT
=

Since |tf (w)| 2

310

\\H(0)\>
2(l-e~QT)2

+ (c- 5*nTi;c)
Hence, wc

e-aTe-ju,

1,

2s:'n~

LQT^

{sinn-)
1
1 - 2e- T cosu; + e - 2 a T

Figure 8.20:

it oscillates between

1
and
(l-e-*T)2
(l + e " a T ) 2
but never reaches zero
h(r)

_
=

>

_ ,-1
e, - a r T =
e
1
-~
aT

r is the smallest integer that is larger than

(c)

H(z)

since | # a ( ; f i c ) | 2

Let a
Then ff(z)

l-^-1

_L ,v

aT(l + z-1)
2 ( l - r - 1 ) + QT(l + z - 1 )
aTjl + z-1)
2+ oT+(aT-2)2-1
1
0

DC Gain: //(z)U=i
At z - - l ( i y = ir),H(z)

u;c

a
2

- , we have Qc a
2
=
2tan~1aT2
2-oT
2 + aT
1-a
(l + o)z - 1
1 + 1-02"1
2
l-o
[6(n) + (l + a ) a n - 1 t i ( n - l ) ]
311

1-a

MO)
MO)
=> (1 + a)a n - l

'"Tfc-1
/na
/n(
2-oT
2+aT

M
8.22
(a)

M) =
T_
2ir

^'""cfti; + /
. Trn
2T

T_
nir

e>" dw

. 27rnl
bT J

(b)
Let ht{n)
Then, h(n)

hd(n)w(n),
- 100 < n < 100(M = 101)
h,{n - 100) will be the impulse of the filter for 0 < n < 200

(0
' o,

0<w<

j 1 OOw
i

Hd{w) = <

o,
e~

jlOOw

>
L L

T <w<T

2*

-31*
200
irk

=
hen, H(k)

=
=
=

H(k)

=
=

100T
0 < Jb < 40
0,
50 < k < 150
0,
160 < k < 200
0,
e-^ ,
e-

iwkT

40 < k < 50
150 < * < 160

H(w) will match Hd(w) at 201 points in frequency. The filter will contain large ripples in
between the sampled frequencies. Transition values should be specified to reduce the ripples in
both the passbands and the stopband.
312

8.23
(a)
D7T

wi

Qi

T2
Wi
tan-^ ( for T = 2)

wu

12

w.
= tan^2

Qu
Analog: lowpass to bandpass

s2 + fljO,
s *

Bilinear: Analog to digital

2 - 1 _ 1-2-1
2 + 1 " 1 + 2-1
combine the two steps:
2
(

+ aa

( i - 2 - 1 ) 2 + n t < Q,(i-h2- 1 ) 2
(l-z-1)^-^)
Therefore, H(z)

(i-i->)a+n.ni(i-n-')a

i-,t-)a+n.ni(i++ ^2 ((i-*-'i(A.-ft.>

)a

+1

(b)
=

ran

<

1.43

<

1.8

<

1.82

1.7

= 1.7

filter (4) satisfies the constraint

8.24
1/1

H(z)

-12i

(I-*"")

|d--)
6

1 + 2"

1 - 2" + 2 -

y~\

1 + 2" + 2 - 2

1 ,-3+
+ iz-7+*;Z
2
- 33

I ( l - 2 - ) ( l - 2 ^ 1 ) ( 2 + 2-1 + ^
313

1 / 1 , 1
2

2- 4
~ ^ +. r2 - + 2 " + 2.-5
+ , - 1 + 2r-2

2 + 2 -1

1,~1
1

-2 + , r 3 +

.-4'

This filter is FIR with zeros at z

1, e J * , ej$,

e]bi,

-0.5528 jO.6823 and 0.3028 j'0.7462

(b) It is a highpass filter.

(0
H(w)

i ( l - e-j6w)(l
6

- e-]W){2 + e"-"" +

?C-J2U, +

"(f)

1.
2
2
2

8.25
(a)
fJL

fH

^
= 0.36
2500
= 0.44
2500

Refer tofig8.21.
(b) The ideal lowpass filter has a passband of 0.04 < / < 0.04. Hence,
W(f)

X(f)

03A.40.44

0.4

0.04

V(f)

0.04

0.04

Figure 8.21:

Hd(w)

-iis

_o.08ir < w < 0.08TT

otherwise

"I .

Hence,
0 08*

hd(n)

= ^- f
2K J_Q 08*

314

l w wn
e-> * e> dw

0.04

0.4

1 - j 3 w+
2

e-j4

i-0 O B I

2sin0087r

( n ~ 15)
(n-15)
hd{n)wH{n)
2?r(n- 15)
0.54-0.46co5
30

h(n)

WH{TI)

h(n) is the impulse response of the lowpass filter H(w)

8.26
(n-2)Mn) = 0

rdh(0)

n= 0
oo

rdfc(l)

6(n - 2)/i(n - 1) = /i(l) = - 0 . 5

n= 0
oo

] > ^ ( n - 2)/i(n - 2) =/i(0) = 1

r = 0

Therefore,
1.25
-0.5
0

-0.5
1.25
-0.5
60

*i

b2

0 '
-0.5
1.25

"6o
0.5

6i

62

-0.0471
-0.1176
0.7529

Emtn = JV(n)-& t r -h (*)

n=0

=
=

t =0

1-[0.7529+-(0.1176)]
0.1883

8.27

* W - 2^^

10,

K n

<

tf-1

x{n) is a periodic sequence with period N. Hence, y(n) is also periodic with period N. Let
H7(z)
and/i2(n)
Then. h2{n)*y{n)

v
l + at-*

=
=

{ l , a i , a 2 , .. ,a P }
x(n),
n = 0, 1 , . . . , N - 1
315

If p + 1 < TV, the TV equations above are sufficient to determine t^, a 2 , ,a P and their order. If
p + 1 > TV, it is not possible to determine the {a*} and the order p.

(0
Let/ii(n)

and/i 2 (n)
Then, h\(n) * x(n)

=
=

|&0t&i,...,6tf>

{l.ax,. . . , o p }
h2(n) * y(n)

If hi(n) is known, then w(n) = hi(n)x(n) is also aperiodic signal with period TV. If TV > p + 1 ,
then h2(n) or a\ya2,.. .,ap can be determined by the above equation. If either TV < p + 1 or
hi(n) is not known, then, it is not possible to determine the filter parameters, (d) (1) The set of
linear equations are:
Af-l

J]fc(*)r(*-0

= ,(/),

/ = 0,1,..., Af-l

fc=0
oo

where r r x (/)

Y^ z(n)z(n /) and
n=0
oo

ryr(l)

5>(n)*(n-/)

t=0

(2) Refer to fig 8.22.

(3) Refer to fig 8.23.
Total squared error

Figure 8.22:
(4) v{n) = y(n) + 0.01tu(n). Refer to fig 8.24.

316

m ol original h(n)

1000

M-11E-8 093

2000

1000

3000

M.12E-8 084

2000

3000

M.13E-8101

3000

3000

Figure 8.23:

fit ol original h(n

M*8E-8 42

M.9E-8 305

MeJNh

Py^E^SB 30(X> llfi^E^i 3000

tiWfrSSSb

05

1000

2000 3000

1000

2000 3000

Figure 8.24:

317

M9UP&1

100

200

300

Chapter 9

9.1
Since = $oio _ 3 i s

an

i n t e g e r ) then F,=2B

= 40Hz

9.2
Fc
B

=
=

100
12
F 4 ^

r = T

=
B'

F,

=
=

12 '
[8.831 = 8
F + 2lil-2r
106
8
53
4
2'
53
yHz

9.3
x,(<)

[x(0 + >*(0]e" j 2 * / e l

Since [z(<) 4- jx(<)] is complex, x/(*) is complex. The condition for xi(t) to be real is

x;(-f).
9.4
x(n)
x 2 (n)

X(u/)
X(uO*X(u;)
319

Xi(f)

The output yi(<) is basically the square of the input signal ya{t).

X(w)

-3n

-2n

-n

For the second system,

X(w) * X(w)

In

3n

spectrum of sampled Xg (t),

(i.c).s(n)-j? C n T )

Figure 9.1:
x\{t) < X(u>) * X(u;), the bandwidth is basically IB. The spectrum of the sampled signal is
given in fig refprob9.4a.
(b)

x(n)

y()

yi(0

co$40irt
40
cos 50
4m
cos
~5~
x.22 (n)
247rn

cos

2+

2COS5

1
2
1
:: +
2

1 2*n
2
5
1
-cos207rt
2

87rn

= *:<o
s(n)

cos740irt

- + -cosSOirt
2
2

1
1 80
- + -cos
2 2
50

320

Hence, y2{t)

=
=

1
2+
1
2+
- +

or Ft

30,

x(n)

cos

cos

1
8
C O S
2

1 2-irn
2CS
-cos20xt

Aim

3
2irn

yi(0

x 2 (n)3
,27rn
cos
3
1
1
4TH
+
C 0 5
2
2

1
1
2irn
2 + 2 COS
1
!
nn^
- + -cos207rt

sa(t)

*2(0

cos2407rt

- + -co807rt
2 2
1 1
80*n
2 2
30
1
1
2

2+ 2C05

r + -cos20irt

y(n)

=
=
=
=

'

Hence, j/2(0

9.5
sa{t)

*a(0 + oxfl(-r),

sa(n)

x a (n) + a x a ( n - - )

^ ^ i
Xa{w)
r is an integer, then we may select
If

H{Z)

1+ oe-^
C

1+

l-az-

w h e r e

9.6

f; *2(*) = ^ J^xwdw
fc =

321

|o|<l

-00

f=L

n = oo

2^/_*'**<A)|2A

X2(f)eft
-oo
in/ty

\Xa(f)\7df

-oo
Pi

Therefore,

x 2 (n)

/ ' \Xa(f)\2df

9.7
(a)
d(n) x(n) ax(n 1)
E[d(n)] = E[x(n)] - aE[x(n - 1)] = 0
E[d*(n)} = <rl = { [ x ( n ) - a x ( n - l ) ] 2 }

where p r ( l )

<r2 + a 2 < 7 2 - 2 a [ * ( n ) x ( n - 1 ) ]

=
=

<r2+a2<7*-2a7x(l)
<rj(l + a 2 - 2 a p , ( l ) )

= Ifiil
"

7.(0)

(b)
^[cr2(l + a2-2apx(l))]

2a-2pr(l) = 0

a
For this value of a we have

Px(l)

<r2[l+p2(l)-2p2(l)]

^[i-Px(D]

(c) u\ < a\ is always true if | p r ( l ) | > 0. Note also that \px{\)\ < 1.
(d)
d(n)

E[d2(n)]

x(n) - aix(n - 1) - a 2 x(n - 2)

{[x(n)-a1x(n-l)-a2x(n-2)]2}
322

a\

<rJ(l + a; + aJ + 2 a 1 ( o 2 - l ) p , ( l ) - 2 a 2 p x ( 2 ) )

d 2
a,
da i

=> ai

p,(l)[l-p,(2)]
1-P2(1)

da 2

*2

= 0

^ a2

Then, <rj mtn

P,(2)-p(l)
1-Pi(l)
1 - 3p(l) - p(2) 4- 2pl(l)px(2)

+ 2/>j(l) + jg(l)/g(2) - 2pj(l)p r (2)

[1-P?(1)]2

9.8
x(t)

Acos2irFt

-A(27rF)5m27rFr

-27rAFsin27rF*

-^Imax

2,AF<-

Hence, A

>

2irAFT
2xAF
F.

^Pdt

Refer to fig 9.2.

9.9
Let Pd denote the power spectral density of the quantization noise. Then (a)

Pddf
-

2B

= \
<T 2

SQNR

10/ooio-f-

58 10/flf

o\F*

"5fe
<T2F

Thus. SQNR will increase by 3dB if F, is doubled.

b) The most efficient way to double the sampling frequency is to use a sigma-delta modulator.
323

Figure 9.2:

9.10
(a)
Se(F)
\Hn{F)\

F,
=

2\sin

F,

\Hn(F)\2Se(F)dF

I.

4sin 2 ( ) T M F
cos=)dF
2TTB,

1<T\.1TB
- \^
L

. 2irB.
sin77

Ft

Fs

(b)
2irB
F,
2irB
sin- F
a

For

1,
2TTB _ 1 2nB

F,
324

6 l F,

Therefore,^

_ ! [ _ _ _ _ _ ( _ ) ]

3*" ( 77 )

{[X(z)-Dq(z)]T^-Dq(z)}T^-pT

=
Dq(z)

Dq(z)-E(z)
1

z- X(z)

(l-z-1)2E(z]

Therefore, H,(z) = z~x

and Hn(z) =

|ff B (F)|

4in2(^)

2 (l-co(H^))

'" =

JBBlHnin7lldF
2

)0B[^)2]2T.dF

JO

(l-z-1)2

5 7 r < T e ( 77 )

x(n)

xa(t)

2*
cos-n
N
x(n)|n=^
2rt
NT

C0527T(^)/

Therefore, Fo = -77
analog sinusoids can be generated. There are Ar possible different starting phases.
325

9.13
oo

h(t)

/
=

H(F)e}2,tFldF

oo

f [c(F - Fe) + cm(-F -

Fe)]e>2*FtdF

Joo

c{t)e>2*F<x +

c*{t)e-i2rF<t

2Re[c(t)e' 2irFet ]

(b)
H(F)

C(F-Fc)

X(F)

\[U(F-Fc)

Y(F)

C-(-F-Fe)
+

U-(-F-Fe)}

X(F)H{F)

i [ C ( F - F e ) [ / ( f - F c ) + t T ( - F - F C ) C ' ( - F - F c )]
+ i [ C ( F - Fe)U'(-F

But C ( F - FC)U'(-F

- Fe)

^-1[C(F)f/(F)]

U(F - FC)C'{-F
/

- F e ) + t / ( F - FC)C'(-F

- F c )]

- Fe) = 0

c{r)u(t - r)dr = v(t)

J oo

Hence, y(0

v{t)e>2*F<% + i v " ( 0 e ' i 2 " / ' e l

Rfi[r(<)c i2 * Fet ]

9.14
(a) Refer to fig 9.3.
(b) Refer to fig 9.4.
(c) Refer to fig 9.5. The first order hold interpolator performs better than the zero order
interpolator because the frequency response of the first order hold is more closer to the ideal
interpolator than that of the zero order hold case.
(d) Refer to fig 9.6.
(e) Refer to fig 9.7. Higher order interpolators with more memory or cubic spline interpolators
would be a better choice.
9.15
(a) zp(t) J2nL-oo x<>(* "" n - ^ )
fourier series representation are

ls a

Per>dic signal with period Ts. The fourier coefficients in a

T.

ck

-r
U
-,

~i7wki

xp{i)e^rr-dt
J-l?
/-

oo

= f IT E
* J~~2"

^(t-nTt)e-^-dt

n= -oo

* n = oo J

326

Zero Order hold: N = 32 thd - 0.1154

First Order hold, N = 32 thd=0.1152

2ero O R $ hold:flP=64 trSR* O.23IP

fcrst OrSSr hoW.tf * 64 trSU.232

fero Order flSd: N = 1^8id = 0.4^5?

Pirst Order fi&d, N = l 2 8 W o . 4 6 & 5

300

300

>n

>n
Figure 9.3:
1

r-nT.
i.+
t -Z*
T

_,,.' t , r , )

l a ^ ^ ^ ^ '

~ $Xa{T?
= YtXa(k6F)
(b) Let
'
IfT, >2r,

otherwise

xp(t)w(t)
XP(F)*W(F)

x.(0
*.(F)
A'a(F)

\ 0,

r,
327

sinxFT,
*FTS

Zero Order hold: N = 32 thd = 0.1154

First Order hold, N = 32 thd*0.1153

2ero Orc?er hold: ftfl 64 thP= 0.23?

forst OrlSr hold, tf * 64 tlxSU.23350

Zero Order S9d: N - 1ZB tnd = 0.468s

Pirst Order Sold, N 1 2 8 ^ = 0 . 4 6 $

100

200

100

300

200

300

>n

>n
Figure 9.4:

= T. *>

sinir(F j-)T,

k = -oo

(c) If T < 2r, there will be aliasing in every period of zp{t).

consequently, xa(t) cannot be recovered from xp{t).
(d) From (b) Xa{F) = _ ,

Xa{k6F)'^^^
6F

328

Hence, xa(t) ^ xp(t)w(t)

and

First Order Hold

Zero Order Hold

10

20

30

40

10

50

20

30

40

50

Rrst Order Hold, filter spectrum

Zero Order Hold, filter spectrum

2
1.5

0.5
/
-1

-0.5

0.5

\
1

-1

Figure 9.5:

329

-0.5

0.5

Zero Order Hold, interpolated output

50

.
-

| 20

IG)xl<-

40

10
n

___w/

V__
100

200

300

400

500

600

First Order Hold, Interpolated output

600
Figure 9.6:

330

Zero Order Hold, xi(n)

10

20

30

Zero Order Hold, y(n)

40

10

First Order Hold, xi(n)

20

30

40

First Order Hold, y(n)

Figure 9.7:

331

Chapter 10

10.1
(a) Let the corresponding baseband spectrum be called Xb(Q). Then
X

= \[xb(n

*W

2000TT)

+ Xb(Q +

2000TT)]

With frequencies normalized to F r ,

W

Fr

. The sequence x(n) has DTFT

oo

X(w') =

Xa(u>'-2*q)

oo

H iX'(w' ~ 087r "" 27r) + X^w> + 087r _

9=-oo

27r<

?)]

modulation by COS(0.8TT) causes shifts up and down by 0.8* (and scaling by 5) of each com
ponent in the spectrum. Refer to fig 10.1. Ideal LPF preserves only the baseband spectrum
(of each period). Refer to fig 10.2 The downsampling produces the figure in fig 10.3, where
w" = jr $jr- 10u/. Note that there is no aliasing in the spectrum | y ( u / ' ) | because the
decimated sample rate, in terms of w', is ff- > 0.04*.
(b) The assumed spectral amplitude normalization in fig 10.1 implies that the analog FT (mag
nitude spectrum) of xa{t) is (refer to fig 10.4). The given sample rate is identical to Fy above,
Fy = 250Hz. The DTFT of samples taken at this rate is Y(fi) = y - , XB(ft - &y) w h e r e
Qy = 2irFy. On a scaled frequency axis w" = QTV = , Y(w") - j-Y.qx(w"
~ 9 27 0Consequently y(n) = y(n).

10.2
(b) After decimation Y(w') = \X{\)

*-r-

(c)

DTFT{x(2n)}

x ( 2 n ) e -jw2n
-J
n

= x(2n)e"' V
n

=
333

Y(w')

lX(w')l

Assumes peak of X (.) normalized to unity

X (w') shifted to 0.&

b

period 2n
0.8n

-0.8*

-n

i-

^
P-0.1671

Figure 10.1:

10.3
(a)Refer to fig 10.5
(b)
Let w'

w'

T,' W=T,
YW) = <5)-'"""+ E

n+1

jM^H*^)]

n odd

n even

+ \ > ( * ) + x( 9 + l ) ] e - ^ " ( 2 ' + i :

x(p)e-^"^

X{2w") + ie-^"[JV(2u;") + e> 2 ""*(2u/')]

X{2w")[l + cosw"]

'

\ 0,

otherwise

0.2*
^

'

\ 0,

otherwise

_ / 1, 0 < K ' | < O.lTT

0, otherwise

-{
Y{w

->=u

COSW",

0 < |U>"| < O.lTT

otherwise
334

,-jm'n

IW(w')l

0.5

Y
!

-0.4n

O.ln

H(w,)

i
-n

period 2 n

0.5

/ \
0.4n

Figure 10.2:

(c) Refer to fig 10.6

1 + cosw",

0.35TT < \w"\ < 0.45*

or 0.55TT < \w"\ < 0.65*

0,

otherwise

10.4
(a) Let u/ = ,
w" = $ 7 - Refer to fig 10.7
Let x"(n) be the downsampled sequence.
x"(n)
X"{w")

x(nD)
1
w"
X(^)

As long as Z)ty^ < *, -V(u;') [hence x(n)] can be recovered from X"(w")[x"(n)
using interpolation by a factor D:

= x(Dn)]

X(w') = D r ( 2 } u ' )
The given sampling frequency is w't = ^-. The condition Dw'm < TT 2 u ^ < ^- = u;^
(b) Let xa(t) be the ral analog signal from which samples x(n) were taken at rate Fx. There exists
a signal, say x'a(t'), such that x'a(t') Xa(j-).
x(n) may be considered to be the samples of x'(t')
taken at rate fx = 1. Likewise x"(n) = x(n>) are samples of x'(t') taken at rate f'J = ^- = ^ .
From sampling theory, we know that x'(f') can be reconstructed from its samples x"(n) as long
as it is bandlimited to fm < 377, or wm < jf> which is the case here. The reconstruction formula

x'(0 == 5>"(*)M*'- *0)

k

where

335

0.08n

w*

lY(w")l = 0.1IV(0.1w")l

Figure 10.3:

Refer to fig 10.8

Actually the bandwidth of the reconstruction filter may be made as small as w'm, or as large
as ^ - w'm, so hr may be
_ sin(w'et')
(w'ef)

MO =

where w'm < w'c < = - w'm. In particular x(n) = x'(t' = n) so

x(n) = ^2z{kD)hr(n

- kD)

(c) Clearly if we define

(P)

p is an integer multiple of D
other p

*e = U '
then, we may write 10.4 as
x(n) =

^2v(p)hr{n-p)

so x(n) is reconstructed as (see fig 10.9)

336

IXa(Q)

i^
h 400n

-c

Figure 10.4:

10.5
Letu,

- ,

= _

X.M = ji,(n)e-'" B
n

]Tx(2m)e-'" 2 m
m

= |E*(f>
i

To recover x(n) from x,(n): see fig 10.10

(b)
Recall w' = 2w
XiW)
= 2*4(n)e-^,n
n

337

^
*,(n)e-'v*
n even

Q.

y(m)
Fy=FX

Figure 10.5:
since x,(n) = 0 when n odd

= *.<f >
see fig 10.11
No information is lost since the decimated sample rate still exceeds twice the bandlimit of
the original signal.

10.6
A filter of length 30 meets the specification. The cutoff frequency is wc = f- and the coefficients
are given below:

Mi) =
h(2)

/i(30) = 0.006399
/i(29) = -0.01476

M3) = M28) = -0.001089

= h(27) = -0.002871
M5) = h{26) = 0.01049
/i(6) = M25) = 0.02148
h(7) = M24) = 0.01948
h(S) = M23) = -0.0003107
h(9) = h{22) = -0.03005
i(10) = M 2 1 ) = -0.04988
h{4)

338

X(.)

0.771

0.971 71

l.lTl

1.371

0.3571

0.4571

0.5571

0.6571

271
71

w
w

Figure 10.6:

Mil)
M12)
M13)
M14)
h(\b)
Pk{n)

= /i(20) = -0.03737
= /(19) = 0.01848
= /(18) = 0.1075
= h(17) = 0.1995
= /(16) = 0.2579
= M n + ifc), it = 0,1

corresponding polyphase filter structure (see fig 10.12)

10.7
A filter of length 30 meets the specification. The cutoff frequency is wc = | and the coefficients
are given below:
Mi
h(2
h(Z
h(4
h(b
/(6
A(7
/i(8
h(9
h(10
Mil
M12

/i(30)
M29)
M28)
fc(27)
h(26)
/i(25)
M24)

0.006026
-0.01282
-0.002858
0.01366
-0.004669
-0.01970
0.01598

M23)
h(22)
M21)
M20)
M19)

0.02138
-0.03498
-0.01562
0.06401
-0.007345
339

Figure 10.7:
M13) =
/i(14) =
/(15) =
pk(n) =

/i(18) = -0.1187
h(\7) = 0.09805
/(16) = 0.4923
h{2n + k),
k - 0,1; n = 0, 1,.

,14

corresponding polyphase filter structure (see fig 10.13)

10.8
The FIR filter that meets the specifications of this problem is exactly the same as that in probl0.6.
ents are
Its bandwidth is \ . Its coefficients
g(n,m)

h{nl + (m>)/)

h{nI + mD - [ ] / )

/(2n + 5 m - 2 [ ^ ] )

g(0,m)
.7(1, m)

=
=

{/i(0),h(l)}
{h{2),h(2)}

Sf(14,m)

(M28),/(29)}
340

IX(w ')lperiod2n/D

(W)

271/D-w'

2n/D

Figure 10.8:
A polyphase filter would employ two subfilters, each of length 15
p 0 (n)
Pl(n)

=
=

{h{0)th(2),...,h(2S)}
{/i(l),M3),...,M29)}

10.9
(a)
x(n)

{x0,xux2,...}

z2(n)

{x 0 ,Z2,^4,-}

y2(n)

{xo,0,X2,0,X4,0,...}

zi(n)

{x0,0,x1,0,x2,0,...}

yi{n)

{x 0 ) xi,X2,...}

so yi(n)

yi(")

D 1 = 2. Decimation first

Interpolation first

(b) suppose D = dk and 7 = :* and d, i are relatively prime.

x(n)

{x0,xi,x2,...}

*2(n)

{*0>*d*>*2<> }

y 2 (n)

< x 0 ,0, . . . , 0 , X d t , 0 , . . . , 0 , x 2 d t , . .

Decimation first

i*-l

Interpolation first
341

i*-l

x'*(n)=x(kn)

ID

x(n)

v(n)

hj(n)

Figure 10.9:

*i(n)

<

xo,0,...,0,xi,0,...,0,Z2,0,...,0,
i*-l

ik-l

ik-l

= < IO,0,...,0,Xd)0,...,0,.

yi(n)

d-1

d-1

Thus j/2(n) = yi(n) iff d = dA: or it = 1 which means that D and / are relatively prime.

10.10
(a) Refer to fig 10.14

yi(")

=
=

h(n) * wi(n)
h(n)*x(nD)

H(zD)
H{zD)
342

X(w)
pei

4/2

-71

x.(n)_

-271/3

vfnl

41

1 ^,

-7C/3

71/3

x(n)

h(n)

2n/3

where
H f (w)
1

-71/2

71/2

Figure 10.10:

/o,0,...,0,/f 1 ,0 1 _ t ^ 1 0,/.(2),
D-l

D-\

nD-1
SO W-2(n)

h(k)x(n-k)

/j(*D)x(n - kD)

k=0

M*)x(n-*D)
y?(n)

W2{nD)

J2h{k)x{nD-kD)
h(k)x[(n-k)D]
fc=0

Soyi(n)

y2(n)

(b)
u'i(n) =
yi(") =

">i(p)>

" = P^(P a n integer )

343

period 2 7i

271/3

Figure 10.11:

w2{n)

=
=
=

0,
other n
x{p),
n = pl
0,
other n

Let h{n) be the IR corresponding to H{zI)

yj(n)

h(*)u>2(n-*)
oo

= h(*/)u;2(n-*/)
= /(*Wn-H)
for n

pi

V2(n)

= >(*W(p-*)/)
*=o

= M*)*(P-*)
=

t=o
u;i(p)( see above )

for n

pi

y* (n)

/(*)<> = 0
ksO

344

F
x
Figure 10.12:
so we conclude yi(n)

y2{n)

10.11
H{z)

^/i(2n)z-2n + ^/(2n+l)2-2n-1
n

^M2n)(2 )-"+z-1^/l(2n+l)(z2)-'1

= X>(2n)z-n

Therefore Ho(z)

^M2n+l)^n

Hi(z)
(b)
ff(r)

Mn)z"nD+X>(n+l)*~nZ?~l +
n

D-l

*=0

345

x(n)

-O

P^n)

F
x

= F xX / D

Figure 10.13:
Therefore Hk(z)

= J^ h(nD

+ jb ) z "

(c)

H(z)

1 - 02"1
oo

E a r~
n

n=0
oo

Ho(z) =

T.

2nz-n

n=0

1
a
1 - 2 *- 1
oo

Hx(z)

T.

2n+lz-n

n=0

l-a^"1

10.12
Dx

25,

F0

10 kHz ,

346

D2 = 4

Fi = 77-Q = 400 Hz

x(n)

x(n)

w^n)

H(z)

H(P)

wjn)

\D

yj(n)

y2(n)

Figure 10.14:

Passband 0 < F < 50

Transition band 50 < F < 345
Stopband 345 < F < 5000
Fi

7^- = 100 Hz
D2

Passband 0 < F < 50

Transition band 50 < F < 55
Stopband 55 < F < 200
For filter 1, 6X

A/

^Y - 0 0 5 ,

Mi
For filter 2, 6i

A/
M2

62 = 10 - 3

345-50
_ 2
nn
= 2.95xl0~ 2
10,000
-\0log6i62 - 13
+ 1 = 71
14.6A/
0.05,
62 = 10" 3
55-50
= 7.5xl0" 3
400
-IO/O0M2 ~ I 3
rrhh
+ 1 * 275
14.6A/

The coefficients of the two filters can be obtained using a number of DSP software packages.
347

10.13
To avoid aliasing Fte < %. Thus D - I = 50.
Single stage
6X

A J/

Mi

Two stages
Di

stage l.Fi

0.1,
62 = 1 0 - 3
65-60
,
B ,
r = 5x10 4
10,000
-IO/O0M2 - 13 , n n
+
14 6 A /
' " 37
25,
D2 = 2
10,000
'
= 400

7i = 2,

72 = 25

25

Passband 0 < F < 60

Transition band 60 < F < 335
Stopband 335 < F < 5000
6X

A/

stage 2:F 2

10" 3
62 =
4
2.75xl0" 2
Mi = 84
400
= 200
0.1,

Passband 0 < F < 60

Transition band 60 < F < 65
Stopband 65 < F < 100
61

A/

10" 3
*2 = - r 4
0.1875
M 2 = 13
0.1,

Use DSP software to obtain filter coefficients.

10.14
6 + (n) is nonzero for 0 < n < 2N - 2 with N even. Let c(n) = 6 + [ n - (TV- 1)]. So c(n) is nonzero
for -(/V - 1) < n < N - 1. From (10.9.43)
B+(u;) + ( - l ) N - 1 B + ( u ; - 7 r )

oe"^^"1)

orB+(r) + (-l);v-1B+(-2)
Therefore, C(z)z"( - > + ( - l ) ; v - 1 C ( - z ) ( - 2 ) - ^ - 1 )
or C{z) + C(-z)
c(n) + c(n)
when n ^ 0c(n)
when n is odd c(n)

=
=
=
=
=
=

a*-*"- 1 *
a z ^ ^
a
a6(n)
c(-n)
c(-n)

when n is even but n ^ 0, c(n)

(half-band filter)

;v 1

when n = 0,c(n)
348

a
=

10.15
one stage:
61 =
A J/

M-

0.01,
62 = 1 0 - 3
100-90
f .
= 10 3
10,000

-1O/O0M2-13

= diij+1 *2536
two stages: F 0

2xl0 5 Hz

/1

1,

Fi

- ~ = 2xl0 4 Hz

/2 = 2

/1

Passband 0 < F < 90

Transition band 90 < F < 19,900
Therefore A /

19,900-90
- - =5
= 0.09905
2xl0

and <5n

-,

f>n = h

- I O W - 1 3
14.6A/

F2 = ^r = 1:cl 4
'2

Passband 0 < F < 90

Transition band 90 < F < 9,900

M2 =

Hz

Therefore A J/

9,900-90
'
= 0.4905
nA
2xl0 4

and 621

14.6A/

622 = h

+ 1 S S ?

10.16
Suppose the output of the analysis section is xao(m) and xai(m).
become yo(m) and yi(m). Thus

After interpolation by 2, they

(y),

m even k = 0,1
m odd

z(m)

yo(m)*2/i(m) +

when m is even, say m

z(m) = z{2j)

=
=

2,;,
2y 0 (m) * h(rn) - 2y](m) * h(m)

The final output is

349

/(m)]

2Y,yo(k)h(rn-k)-2TTtyl(k)h(m-k)
k

yi(m)*[-2(-l)

2]Tyo(2/)/z(2; - 2/) - 2 i ( 2 / ) M 2 j - 2/)

= 2 xa0(l)h{2(j - /)] - 2 *fli(0M2(j " /)]

=

2[xao(j)-xai(j))*h(2j)

2[x a 0 (j)-X a iO')]*po(j)

2[*. 0 (j) + *.i(j)]*Pi(J)

In the same manner, it can be shown that

*(2j+l)

10.17
Refer to fig 10.15, where M n ) is a lowpass filter with cutoff freq. j - . After transposition (refer

r*

Interpolator 1

n )

,
;

Interpolator L

Figure 10.15: I = III? .. ./^.L-stage interpolator

to fig 10.16). As D = 7, let A = / i + i - i , then D= DiD2...DL.
Obviously, this is equivalent to the transposed form above.

Refer to fig 10.17

10.18
Suppose that output is y(n). Then Ty = jTx. Fy = f- = y - = F r . Assume that the lowpass
filter is h(n) of length M = kl (see fig 10.18)

350

tD)

h(n)

Figure 10.16:

10.19
(a)
for any n

i-\

II + j

{0 < j < I - I)

l-\

!>("-*)

^PkiU

PiiU)

=
=

Pi(0
h(j + ll)

*(n)
/-i

Therefore, h(n)

^p*(n-Jb)
*=0

(b) z-transform both sides

/-l

#(') = E z "W)
*=o
(c)

y E E ^ ^ :
3l(w->)

/=0

351

w-fc

j-k)

1
1
1 ^

Ijjn)

h{n)

DL

1
1

tD,

Figure 10.17: L-stage decimator

7-1

= j J2 H M* + m/)ej2Wm:-m
= J^ h{k + m/)z-m
m

= J>(m)z-"
m

Pk(z)

10.20
(a) Refer to fig 10.19.
(b)
Bandwidth
cut off freq
sampling freq of x(n)
sampling freq for the desired band of frequencies
Therefore, D
(c) Refer to fig 10.20.
(d) Refer to fig 10.21.
352

3
7T

2
2TT

2jr _
2

coefficient storage^
F

Wn>

g(n.O)

n=0,l

K-1

S^1*

n=0,l

K-1

gOU-1)

n=0,l

K-1

output
buffer
length I

y(n)

'input
buffer

length K

hr-o

2
K-1

n=0

F=
V

Figure 10.18:

spectrum of x(n)

spectrum of y(n)

Figure 10.19:

353

x(n)

|X(w)|

40
30
A
i
i

20
10
0

500

1000
>n

1500

500

1000

Figure 10.20:

p*cmm of (n)

200

400

600

Figure 10.21:

354

800

1000

1200

1500

Chapter 11

11.1
(a)
25

_
l

H(z)

and <r*

25

so x(n)

x(n 1) - x ( n 2) + w(n)

l - z - i + iz-

l
, ~ ll +
-L i\z, - 2
(b) The whitening filter is H~~lt-r\
{z)
= 11 -_ z~

11.2
For a stable filter, denominator (1 ^zl) must be chose. However, either numerator factor
may be used.

i/(z) =

_ f
v

or ^ - j y
'

[min

pk]

(b) Must invert the min. pk. filter to obtain a stable whitening filter.

H'\z)

11.3
(a)

H(z)

_!_ v
whitening filter, H~1(z)
zeros: z
pole: z

=
=
355

1 + 0.9Z" 1

l - l.ez-i + o.esz-22

1 - l.Gz-i + O.Mz-'
1 + 0.9*" 1
0.7 and 0.9
-0.9

st(w)

=
=

alH(w)H(-w)
11 + 0 . 9 e ~ ^ l 2
| 1 _ l . 6 c - ^ + 0.63e" 2 -' u '| 2

<?:,

A(z)

1+

24Z

5'

32

1_
3

kz
B3(z)

3 + 8* + 24*
I
2
1
3
r2 ++ 8'
r*"1 + *~2
A2(z) - k7B2(z)

k3
B2(z)

+Z

i-*3
= I + ix- l
*1

'M*)

l + 2z-1 + i z - 2

B 2 (r)

Ai(z)

ki

2,-'

z-'

1
3
^2(^)-fc2B2(z)
1 Ar|

.+ .-

3
2

*(*)

Bi(z)

A2(z)

1+
\

K'
^

Ai(r) + fc2Bi(2)r - 1

= 1 + r -r
,

356

1-1

-2

B*(z)

H{z) = A3(z)

1
1
+ 2"
+
-3
32
A2(z) + k3B2(z)z - l
l + z-3

The zeros are at z

-he**

Refer to fig 11.1

(b)

Figure 11.1:
If Af3 =
H(z) = A3(z) =
=
T,

1, we have
A3(z) - B2{z)z-X

The zeros are at 2 =

1 - - , > - 2- 3

1+-2

Vil

1, - j -

(c) If \kp\ = 1, the zeros of H(2) = Ap{z) are on the unit circle. Refer to fig 11.2.

11.7
1 + 0.62 - 1
Bi(z)

0.6+z_1
357

unit circle

Figure 11.2:
A2(z)

B2(z)
A3(z)

=
=

Bz(z)
Hz{z)

=
=

h(n)

AM

kiB^z-1

l + 0.78z_1+0.3r~2
0.3 + 0.78z _1 + 2 - 2
A2(z) + 0.52B7(z)z-1
1 4- 0.93z - 1 + 0.69z - 2 + 0.5z - 3
0.5 + 0.69Z"1 + 0.93z - 2 + z'3
A3(z) + 0.9B3(z)z-1
1 + 1.38*"1 + 1 . 3 H z - 2 + 1.3372"3 + 0.9z - 4
j l , 1.38,1.311,1.337,0.9,0,...j

11.8
Let y(m) = x(2n p m). Then, the backward prediction of x(n p) becomes the forward
prediction of y{n). Hence, its linear prediction error filter is just the noise whitening filter of the
corresponding anticausal AR(p) process.

11.9
x(n + m)

y^]a p (k)x(n - k)
*=i

358

e(n)

x(n + m) - x{n + m)

x(n + m) + ^ P ap(A:)z(n - fc)

0,

- 7 r x ( / + m),

[ e ( n ) x > + m)]

t=i

E[e(n)x'(n-l))

=>a p (*) 7rx (ib-/)

/=l,2,...,p
/=l,2,...,p

*=1

The minimum error is

|2
{|e(n)|^}

= 7*r(0) + J2 ap(kh**(m + *)
*=1

Refer to fig 11.3.

x(n+m)

Si
forward
z -m-l

x(n+m)

linear
predictor

Figure 11.3:

11.10
p-i

x(n-p-m)

=
e{n)

-^bp(k)x{n-

k)

x(n - p- m) - x(n - p- m)
359

p-i

Y,bp(k)x(n-k)

x(n-p-m)

0,

-7xx(/-p-m),

<fc=0

E[e(n)x'(n-l)}

/ = 0,2,...,p-l

P-I

=>5>(*h('-*)

/ = 0,2,...,p-l

fc=0

The minimum error is

{|e(n)| 2 }

E[e(n)x*(n p m)]
P-I

7xx(0) + ^2 bp(kh"(P

+ - *)

k=0

Refer to fig 11.4.

Backward
linear
predictor

jtfnX

x(n-p-m)

-p-m
x(n-p-m)

Kr

Figure 11.4:

11.11
The Levinson-Durbin algorithm for the forward filter coefficients is
7xx(m) + 7*' ,2m-I
a m (*)

-Em
am-\{k)^kmai'rn_l{m~k),
k = 1 , 2 , . . . , m - 1; m = 1,2,.. ,P
360

butfc,^*)
or am(k)

a'm{m-k),

Therefore, b'm{0) = km

C(m-lb)

Equivalently, 6 m (0) = k

M*)

* = 0,2,...,m

m(m-k)

7xr("^) + 7 *

.6" 1

' * "

bm_1(rn-l-k)^kmbm.l(k)
7 ^ ( ^ ) + 7^_1^_i

1
& m -i(* - l J + ^ ^ - i C m - t )

This is the Levinson-Durbin algorithm for the backward filter.

11.12
Let
Ln =
Then,

fc

Lm-l
7*'
-Un-1

'

7
lm-l

7xx(0)

m-l
0

.m-i

'

km0

{in =

<L-1
m

) JJ

L cm(m)

Hence,
L,,-!^-!+m_1dm_1+6m(m)7jn*_1

Cn-x

li!-iin-i+Ti;-iin-i+Mni)7..(0)

cm(m)

Butl^.^.!

=>Ln-ln-l

Hence, _ !

-M^OIm-iT^.!

Also, I ^ L i T ^ . !

o^-j

Therefore, 6 m (m) I ^_ i a! n *_ 1 +6 m (m)7xx(0)

cm(m) - T ^ . ^ - i

c,
tn-l
6*
-6m(m)7^_1

solving for 6 m (m), we obtain

bm (m)

^ W + lil-i^-i
Cm(m)-j^_lbm.l
^m-l

we also obtain the recursion

bm(k)

11.13
Equations for the forward linear predictor:
I_mm m

361

&m-l(*)-|-&m(m)am-l(m- *)'
ib = 1 , 2 , . . . , m - 1

where the elements of c m are yxx(l + m),

/ = 1,2,..., p. The solution of am is

a (m)

"

7(o) +2 2. lfl !;-,

cm(m)-7^_1am_1

am(fc)

where a m is the solution to I ^ a ^ ,

am-i(*) + om(m)Q^_1(m-*),
/b = 1,2,.. . , m 1;
m=l,2,...,p
7

The coefficients for the m-step backward predictor are 6,,, = a^.

11.14
1
k*

v =

VJVl

1
k*

_ ' 1

~lk'm

-kr
-1

kr
1

*m 1 f 1 0

1 [ 1

t m l . f

l"|itm|2

k'm 1 J "" L

= (l~|tm|2)

*,

1 J I - 1 J I *m 1

1
0

0
-1

-(1-Im|2)j

= (l-|*m|2)I

11.15
(a)'
[/ m (n)x(n - ,)]

E[J2am(k)x(n-k)x(n-i))
k=0

0, by the orthogonality property

J^a'm(k)E[x(n

(b)
E[gm(n)x(n-i)}

- m + k)x(n - )]

= m(*hr(*-m + i)
*=0

0,

i = 0,1,..., m - 1

(c)
E[fm(n)x(n))

E{fm(n)[fm(n)-JTam(k)x(n-k)]}
*=i

362

=
=

{l/m(n)| 2 }
Em

E{gm(n)[gm(n)

m-l

E[gm(n)x(n - m))

6 m (*)x(n - k)}}

4=0
2
^{|Pm(n)|2i
}

(d)
[A(")/j(n)]

^ ( n M n J + ^a^tMn-fc)]}

=
=

E{ft(n)x(n)}
max(*\j)

where i > j has been assumed

(e)

[//;("-<)]

E{ft(n)[x(n-t)

a3(k)x(n-t-k))}

when 0 < < < - j , x{n t 1), x(n t 2 ) , . . . , x(n t j) are just a subset of x(n 1), x(n 2 ) , . . . , x(n i) Hence, from the orthogonality principle,
E{fi(n)f3(n-t)}

=0

Also, when - 1 > t > ii j holds, via the same method we have
E[Mn)f}(n-t)}

=0

(f)
E\9i(n)9j(n-t)]

% ( n ) [ i ( n - t - j) + 6,(ib)x(n - t - *)]}
*=o

when 0 < t < i j , {x(n t),x(n t 1 ) , . . . ,x(n t j)} is a subset of {x(n),.. .,x(n i' + 1)}
Hence, from the orthogonality principle,
E[9i(n)9j(n-t))

=0

Also, when 0 > f > i j + l we obtain the same result (g)

for = j,E{ft(n

+ i ) / j ( " + J)}

{/.'(* + 0 }

for i j , suppose that i > j . Then

{/<(+ 0/>(" + i)} =

=

W i ( n + 0[*(n + ;) + Sflj(*)(n + i-*)]}

0
363

(h)
suppose i > j

t=o
=
=

[<;, ( n + >(")]
Ei

for i > j
{/<(")*}

E{/i(n)[x(n-;) + ^6;(*)x(n-lb)]}

E{/,(n)[fc J (0)x(n)]}

*,-[/, (n)*(n)]

kjEt

^te(n)[(n) + ^ o , - ( l : ) x ( n - f c ) ] }

k=0

for i < j ,
{/*(")*}

(J)
-i

{/i(n)*(n-l)}

=
=

{/i(B)Wn-l-j) + J ] M t M n - l - t ) ] }
*=o
(/i(n)i(n-l-i)]

E{ft(n)[gi+l(n)

6 i + 1 (*)x(n - k)])

k=0

-E[ft(n)bi+1(0)x(n))

(k)
i+1

{p,(n-l)x(n)}

{p,(n-l)[/,+1(n)-^al+1(i:)x(n-t)]}
Jb = 1

-E[5i(n-l)al+1(i+l)x(n-l-i)]

ki+iEi

Ei/^n + l M C n - i V ! > ( * ) * ( " - * - * ) ] }

{/i(n)[x(n-l-j) + M * ) x ( n - l - * ) ] }

{/<(+!)*(-,)}
(1)
suppose ' > j
Wi(n)9i(n-1)}

364

Now, let i < j . then

E{fx(n)9j(n-l)}

E{gj(n-l)[x(n)

^at(k)x(n-k)]}
*r = 1

E{9j(n-l)x(n)}

kj+iEj

from (d)

11.16
(a) E[fm(n)x'(n - i)} = 0,
1< i < m
(b) E[gm(n)x'{n - i)] = 0,
0 < < m - 1
(c) E[fm(n)x'(n)}
= , [ ? m (n)x'(n - m)] = ,
(d)[/li(n)/;(n)] = r m a x ( i I i )
(e)
W(nW(n-t)] = 0 , f o r { i f ^ ^ 7 i ' .

(0

| > j

^(n)y;(n-oi=o,i{j|;|;::j +li ;:>;:

(g)

(h) E[g,(n + i)<7j(n + j)} =

Emax(i,j)

(i)

^ ( n ) , ; ( n ) ] = { ^ |.*j
( k ) ^ ( n - l ) x - ( n ) ] = -*;+1i
(1)

^ ( n ) , ; ( n - l ) ] = { ^ ; + i E ; ) ]<i
11.17
iGn
iO

0
7xx(0)

fii

*1 =

7(1)
7xx(l)

7xx(2)
7xx(2)

7xx(3) '
7xx(3) m

7.r(l)

7xx(2)

7xx(3)

7xx(0)

>xx(l)

7xr(2)

7xx(l)
7xx(0)

Hi = L * ;
365

*1

Zii =

0
0

0
7 ( 0 ) + Aj7*r(l)

0
0

G2 =

0
0

7xx(2) + * i 7 r x ( l )
7xx(l) + ifcr7rr(2)

7rx(2) + fci7rx(l) 7*x(3) + *i7xx(2)

7xr(0) + fcl7xx(l) 7xx(l) + ^t7xx(2)

Iheretore, * 2 _ - 7 J 0 ) + * i 7 . . < i ) "

Let,

^.(i)-V.<55

n2 =
V,G, =

0
0

0
0

7xx(3) + *i7xx(2)
7rx(2) + ^ 7 x x ( 3 )

'

*2

*5

0
ib2[7xx(2) + ibi7xx(l)] + 7xx(0) + A:17xx(l)

,4
B

where A = 7xx(3) + *i7xx(2) + *i*27xx(2) + *27xx(l), and

B = *27xx(3) + *i*27xx(2) + *i7xx(2) + 7xx(l)
and therefore, G, =

0
0

0 0 7xx(3) + fci(l + ib2)7xx(2) + ib27xx(l)

0 0 *27xx(3) + M l + *2)7xx(2) + 7xx(l)

and
*3 = -

[7Jx(l)-7x"x(0)]7xx(3)-fC
[7xx(0)7xx(2) - 7j r (l)]7xx(3) - 7xx(l)7 x r (2) + D

where C = 27xx(0)7xx(l)7x(2) - 7xx(l)7^(2) - T J . O ) and

D = 7xx(0)7xx(l)7xx(2) + T J , ( 1 ) " 7(l)7j,(0)
This is the same result obtained from the Levinson Algorithm.

11.18
The results of section 11.1 apply directly to this problem. We may express r x x ( / ) as

r(/) = ri|ff(/)i2
where / / ( / ) is a filter with transfer function

H(z)=exp[Y,v(rn)z-m]
m=\

The prediction error filter whitens the input process, so that the output process is white with
spectral density c^ = exp[v(0)]. Therefore, the minimum MSE is

EL

- / : .
=

<r?.,dw

<rl /

366

dw

-f

But v(0)
Therefore, E^

lnTxx(f)df

2*exp[

j^lnrxx(f)df)

=>G(z)

Since d(n)

(1 - a z " 1 ) ( l - az)
1
1-az"1
x(n -I- m), we have

E{d(n)x(n

11.19

ldx(k)

- k)}

E{x(n+m)x(n-k)}
Therefore, Tdx(z)

7(m + t)
~M T::{Z)
zm<rl(\-az)
(1 - a z - ^ C l - az)

Tdx(2)

G(z'>)

1 -az-1
az

- ^ ( i - a z - 1 ) 1 az_

rt"

^opt(')

/*opt( )
the output is y(n)

h0pt(n)*x(n)
amx{m)

MMSEoc

7rx(0)-^/iOpt(t)7ax(t)
*=o

7xx(0)-a r n 7dr(0)
7xx(0) - a m 7xx(m)
1-a2
l-a2m
1-a2

1-a2

11.20
(a)
<2o =

0
0

am6(n)

iC?,
il

1l

-L

64

367

=>*! =

"

A -*

Hi =

G2 =

0
0

Hiai =

- 1 --2liJ 4
',8

12S

0 0 -i

re

&

I
=

2 <2 2 =

=>* 2 = r

1 -*

L 3

S w

o o I if
G, =

0 0 0
0 0 0

(b)Refer to fig 11.5

Figure 11.5:

368

=>*3

47
128

11.21
r(/)

trr(k)e-^k

ks-eo
m

Am{f)

= Ea m (p) C -> 2 *^
p=0

a=0

J Jr:(f)AM)K(f)df = E E E ^ W ^ K ^ / t-W+>+ridf
3

*
9?
P
oo
m
n

*^~3

fc = - o o p = 0 9 = 0
m
n

I]^7ir(?-P)am(p)a;(?)
p=0 9 = 0

= EE^/
=
=
=

?)X"(/ + P)]fl-(P)an(9)

^ { E a m ( p ) x - ( / + p)Ea;(?)x(/ + p)i
lp=0
j=0
J
E{/m(/ + m)/:(/+n)}
Em6mn

where the last step follows from prob. 11.16 property (g)

11.22
Ai(z)
Bx{z)
A2(Z)

=
=
=

1 + 0.62 - 1
0.6+ 2 _ 1
1
J 4 1 (2) + t 2 B 1 ( 2 ) 2 "

B2(z)

=
=

l + 0.782 _1 + 0.32- 2
0.3 + 0 . 7 8 2 " 1 + 2 - 2

5 3 (2)

=
=
=

A2(z) + k3B2(z)z~1
l + 0.932- 1 + 0.692" 2 + 0.52- 3
0.5 + 0.692~ 1 +0.932- 2 + 2 - 3

A4(z)

= ^(zJ + ^BsC*)*- 1

Az{z)

H(z)

1 + 1.382"1 + 1.3112"2 + 1.3372"3 + 0.92" 4

1
AA{z)

11.23
A2{z)

1 + 0 . U-"11 - 0 n. 77202- - 2
369

*2

B2(z)

M{z)

*i

Bi(z)

M*)
Ci(z)

Hence, 0Q
Pi
02

=
=
=
=
=
=
=
=
=
=

-0.72
-0.72 + 0.1z _ 1 + z - 2
A2(z)-k2B2(z)
1 + 0.357Z -1
0.357
0.357+ Z" 1
Bo(z) = 1
0oBo{z) + 01B1(z) + (32B3(z)
A, + /?! (0.357 + z'1) + /? 2 (-0.72 + O.lz" 1 + z'2)
l - 0 . 8 z _ 1 + 0.15z - 2
1.399
-0.815
0.15

Refer to fig 11.6

input

*- output

Figure 11.6:

11.24
Refer to fig 11.7 hx{n) mininizes E[e2(n)] (wiener filter) length M - 2 (a)
T{w)

crl\H{w)\2
370

*e(n)

Figure 11.7:

T(z)

0.49
|l-0.8e-J"|2
0.49
{l-O.Sz-^il-O.Sz)

We can either formally invert this z-transform, or use the following idea: The inverse ztransform of 11.1 will have the form
7 ( m ) = 7 ,.(0)(0.8)l m l
From the AR model for s(n) it is easy to show
7(0)

solve for 7,,(0)

0.87(l) + 7,(0)
0.8 7 (1) + * 2
0.87(0)+7,(l)
0.8 7 (0)
49

Now 7,,(m)

=
=
=

E[x(n)x(n m)]
{[s(n) + u ; ( n ) ] [ s ( n - m ) + u ; ( n - m ) ] }
7.(m) + (T^(m)

and7(l)

=
=
=
=

(b)
d(n)

s(n)

7dx(0 = 7,x(0

=
=

E[s(n)x(n - /)]
E{s(n)[s(n-l)
+

7(0
371

w{n-l)]}

So the normal equations are

i

494

, 49
494

36 5

49

4$ 4
L 36 5

49

" r 36

J'(0) = 0.462,

h\l) = 0.248

(c) { = MMSE 2 = * - 0.462 x | | - 0.248 x |36| x

} = 0.462
~ 5

11.25
Tdz(z)

T(z)
0.49
(1-0.82"1)(1 - 0 . 8 -0

r(*)

r(*) + i

Tdx(^)
G(z-i)

Htiz) =
ht(n) =
+ = MMSEoo

- 0 . 4 5 .0
1.78(1-0.452"
(l-0.82"1)(l-0.8z)
(1-0.452-1)
(l-0.8z-1)
0.49
;i-0.8r"1)(l-0.45z)
0.766
0.345z '
I - O . 8 2 - 1 + 1-0.452
0.766
1-0.8*-1
1 1-0.8Z"1
0.766
1.781-0.45Z- 1 I - O . 8 2 - 1
0.43
1-0.452"1
0.43(0.45) n u(n)

f\r..{z)-He(z)T.t(z-l)]i

1 }
0.28
2xj * e ( z - 0 . 4 5 ) ( l - 0 . 8 z )
0.438

11.26
Using quantities in prob. 11-25,

*(*) =

Txx(z)
0.275
( 1 - 0 . 4 5 2 " 1 ) ( 1 - 0.452)

tfe = MMSEne

^fViM-H+cizVtoiz-1)]
372

0.275
1 dz
0.45)(l-0.452"
)
2*j Jc {z 0.345

11.27
7 ( m ) = (0.6)'ml

2
0.6
0.36

0.6
2
0.6

r /i+(o) i
fc+(i) =
h+

0.36
0.6
2

L wJ

/ + (l) = 0.15,

/i+(0) = 0.455,

1
0.6
0.36

/i + (2) = 0.055

& = MMSE 3 = 1 - 0.455 - 0.15 x 0.6 - 0.055 x 0.36 = 0.435

Increasing the length of the filter decreases the MMSE.

11.28
7..(0)
7rx(l)
7(2)

7(1)
7xx(0)
7xx(l)

1
-1
0.6

-1
1.6
-1

7*x(2) '
7xx(l)
7xx(0)

" 1 '
0
=
0

r 7xx(o) i

0.6
0
1

7*r(0) = 2.5641,

1
-1
0.6

7xx(l)

L 7(2) J

" 1"
0
0

7xx(l) = 1.6026, 7rr (2) = 0.064

For m > 3, -yxx{m) = 7xx(m - 1) - 0.67 r r (m - 2)

For m < 0,7 x r (m) = 7xx(-m)

11.29
T..(z) =
Let a p (0)

(i + E U *p(k)z-k)(\ + EJ=i M*)*k)

1

(E^oM^-'KEkoM*)**.
373

r(*) = r(z) + ol
()()

r(n)is ARMA(p,p). Suppose

Comparing parameters of the two numerators

k=Q

*lf^ap(k)ap(k

k=0

+ q) =

t=0

*lf^bp{k)bp(k

+ q)

= l,2,...,p

Jb=0

There are p + 1 equations in p + 1 unknown parameters <?\, bp(l),...,

374

6 p (p). Note that bp(0) = 1.

Chapter 12

12.1
(a)

\i^-Eir\JT\it)e-"'F,dt\2
=

H m r ^ E \- P x ( * ) - " " ' * f * - ( r ) e > " " d r

[2To J_T
-/-To

l . m r . . ^ ^ C f

[x(()*-(r)] e -""-"-'>d<r

* * 0 J To ~ ~ To

l i m r . - , . - ^ f

l i n * . - . - ^ /'+T0 /T

fX

/%(< -

2*0 Jt-T0

T)e-W-^dtdr

TlI (a)e->

"-->d<da

J-To

yxx(a)e~^F^da

J 00

7rx(F)

(b)
1 *'*

7n(m)

n=0

7(m)e-' 2 "" =

^ E * ( n + m)*-(n)e-' a '/'

Af-1

n+N

= E Jj E x(/)x-(n)->2'A'-)
n=0

l=n-N

= ^EX>(*)*>v-i2"'v2''"
n=0

/=0

= ^iE x W e " ; 2 , r / n i 2
n=0

375

12.2
N-|m|-l N-|m|-l
2

E[hXT(m)\ } = - ^

]T

. E

n=0

[**(")*("+ )*("V(n' + m)]

n'=0

^ 2 E E { [ * * ( + rn)]E[x(n')x-(n' + m)]
n

n'

+ [x'(n)x(n / )][x*(n' + m)x(n + m)]

+ E [ x x ' ( n ' + m)];[x(n')x(n + m)]}
n

n'

+7xr( ' + m - n b ( n + m - n')]

Let p n n. Then
[| 7rx (m)| 2 ]

2
7 r (m)

I [ 7 r x ( m ) ] | 2 + i j Y, Jl^riPhlAP
n p

^ E t 7 " ( p ) 7 ( P ~ m bxx(p + m)]

" m)7xx(P + m)]

Therefore,
var[ 7ri (m)]

p = oo

12.3
(a)

{
"

"

jv-|m|-l

x-(n)x(n + m ) .

n=0

N-|m|-l

"I ^

E *<'+"')
JJ

n'=0

^ 2 E E E{x'(n)x(n + m)x(n')x>' + m')}

n

n'

772 E E { [ * * ( " + m))E[x(n')x'(n' + m')]

n

n'

+E[x'{n)x{n')]E[zm{n' + m')x(n + m)]

+f?[x'(n)x"(n' + m')][x(n')x(n + m)]}

= ^EEw m W m ')+*( n - n 'w m - m ')

n

r'

+<5(n' + m' - n)6(n + m - n')]

Hence, E\pxx{h)P::{h)]

7V-1

N-l

m= -(N-l)m' = -(iV-l)

376

E[lxx(m)lxx{m')}e'^m^e-^m'

m'

+ m' - n)6(n + m - ')] e -J 2 , r m ''e--' 2 , r m ''*

+6(n'
~4

n'

= **<! +

sinv{fi+f2)N
.7V*mx(/i + / 2 )J

sm7r(/! - / 2 ) J v *

+ lATin*(/i

i2'

- /2)

(b)
cov[p x z (/ 1 )p(/ 2 )]

[p(/i)px.(/a)]-^[p(/i)]f:ip(/2)]

[Prx(/l)Prx(/2)]-^

-,{

'simr(f1+f2)N'

.JVsin7r(/ 1 +/ 2 )

sinir(fi - f2)N
Nsinn{fi - f2)

(c)

var[p x r (/)]

cov[prr(/1)prr(/2)]|/l=/3=/

o\

/ sinz7rr;v X\
/sin2irfN

1 + \Nsin2irf

12.4
Assume that x(n) is the output of a linear system excited by white noise input w(n), where
a\ - 1. Then pxx(f) = Txx(f)pww{f).
From prob. 12.3, (a), (b) and (c), we have
E[pxx(f1)pxx(f2)}

T(fi)T**(f2)E\Pw*{fi)PwM7)]
simr(f1+f2)N

= r(/ 1 )r(/ 2 )<l +

co\\pxx{fi)pxx{/2)]

'wr(/i - /2)7V
Nsimrifx - f2)

r rr (/i)r xr (/ 2 )cov[p u;1l ,(/ l )p UMl; (/ 2 )]

= r(/i)r(/2)j
var[p(/)]

'nir(/i + f2)N
.Nsinirifi + / 2 ) .

cov[p(/1)p(/2)]|/l-/3-/

Tf

(sin2nfNfN_\
1 + \Nsin2irf

12.5
Let yk{n)

z{n)*hk{n)
N-l

*(m)e'

m=0

tf-1

j3>m

e'""*"5' ^ P x(m)e = J ^'

m=0
N-l

yt(n)|n=N

m=0

]T x ^ e- J .^' .
*(*)
377

>m

sinir{fi - f2)N
Nsinvifx - f2]

Note that this is just the Goertzel algorithm for computing the DFT. Then,
N-l

\yk(n)f = \X(k)\>

j2irfcm

k)
|^z(m)e^
d

|2

m=0

12.6
From (12.2.18) we have

W(f) =
.

Af-lAf-1

7777

w(*)V(n')e->W-^

n=0 n'=0

W{f)df

ATc/EL^^^ 7 1 ')/ e-i2*/(B"B')4f

= 7 ^ E "()">>( ~n')
n

n'
Af-1

1? M)IJ

= 1

n=0

by the definition of U in (12.2.12)

12.7
(a) (1) Divide x(n) into subsequences of length 4p and overlapped by 50% to produce 4it subse
quences. Each subsequence is padded with 4p zeros.
(2) Compute the M-point DFT of each frame or subsequence.
(3) Compute the magnitude square of each DFT.
(4) Average the 4k M-point DFT's.
(5) Perform the IDFT to obtain an estimate of the autocorrelation sequence.
(b)
M-l

X3(k)

^x3(m)e-zir
m=0

2^

i(m)c ^ ^ +

m=0

,=

A/-1

=
=

Af-1

*1(m)e-iMP+e-'"'r*
m=0

*a(*)

2 ^ * 2 ( m - )e *TT"
,v _ i 2 ^ a l

x^m')*"^

m'sO
iwk

Xi{k)^e' X2{k)

(c) Instead of zero-padding, we can combine two subsequences to produce a single M-point
subsequence and thus reduce the number of sequences form 4* to 2k. Then, we use the relation
in (b) for the DFT.
378

12.8
(a) Obviously, A / = 0.01. From (12.2.52), M = | | = 90.
(b) From (12.2.53), the quality factor is QB = l.l/VA/. This expression does not depend on M;
hence, there is no advantage to increasing the value of M beyond 90.

12.9
(a) From table 12.1, we have
QB

=*A/

Q
w
Qw

==

=>A/

QBT

l.UNAf
QB

1.117V
111
1.39/VA/
Qrv

1.39/V
139
2.34/VA/
2.34/V

234

(b)
For the Bartlett estimate,

Qs = M
=* Af

_N_
/v
= 100
QB

For the Welch estimate with 50% overlap,

Qw

16/V

- s-
For the Blackman-Tukey estimate,
1.5/V
M
..5/N

=*M = 4 ^ = 150

12.10
(a) Suppose PB (f) is the periodogram based on the Bartlett method. Then,

^(f)

= ^lEX'(m)e"j2'/n|2'
r>=0

n=0

^(/)

= (l-)Pil)(/)
= wP{Tl\f)^(\-w)P(B1)(f)
379

* = 0,l,...,Jb- 1

P\f)
Therefore, E{P^\f)}

= {\-w)[wPg\f) + P%\f)]
=
{l-w)J^mwm-kP{Bk\f)
=

w)y,Mwm-kE\PBk\f)\

(1 tsl

(1-w)

l-WM

f*

Af

"hi
^

var{Pif)(/)} =

<fa

a)M'
r rx (a) \sinir(f
r"." v ' *"'"

(l-wM)

sinir(f a)M
sinir(f - a)

rxx(o)

(fa

[ nir(/-a)

E{[Pg\f)?}-[E{PHf)})'

var{Pif)(/)} = { [ ( l - ^ ) ^ M ^ f ( / ) p }
*=1

-{E[(l-w)Y,Mvm-kPBk)(f)]Y
k-l

(1~^):

^ M u ^ - ' ^ i P J ^ / ) } 2 - {E[PBk\f))}
*=i

Mw^M-k\ar[PBk\f)}

(1 - wf ] T
*=i
,2Af

/atn2y/M\:
VMsm2ir/,/
/sin2xfMX

1 + \Msin2xf)
(b)

/ '

r(o)lV(/-o)rfo

J-| ^

M-l

where W ( / )

u;(n)c"> 2 ^

n-0

var[P^)(/)]

(l-u;)22^2(M-fc)var[Pii)(/)]
*=i

2M

(1-u/)

\ l + uj TlAf)

12.11
Let fii'i be defined as follows:
.(0/
rSrV(O)
rS(-l)

.(')/
rW(l)
rg>(0)
.()

rW(0) J
380

Then,
M-\

E'\f)R^E(f)

M-l

^rS(t-*>-'''<*-**>'
M

k = 0 k' = 0

*=0 m = k - ( M - l )

-(M-l)

Therefore,/**>(/)

Pg{f)

1 *
-][><(/)*(/)
4=1

12.12
To prove the recursive relation in (12.3.19) we make use of the following relations:
N-l

Em =

[|/ m (n)| 2 +|*(n-l)| 2 e]

where/m(n)

/m-i(n) +

p m (n)

^/m-i(n) +

andJE.!

(1)

(n-l)
ffm-i(n-l)

(2)

N-i

53

[l/m-i(n)|2-H<7,n-i(n-l)|2]

n=m1

| / m - i ( m - l ) | 2 + |(/m.1(m-2)|2
AT-l

+ [|/m-i(n)| 2 + |* n - 1 (n-l)| 2 ]
n=m
Ar 1

"

1 -

A1SO, 5Z[/m-l(n) + ^ _ ! ( n - 1)] =

--*mm-i

n=m

We substitute for fm(n) and <jm(n 1) from (2) into (1), and we expand the expressions. Then,
use the relations for Em-i and km to reduce the result.

12.13
x(n)

-x{n 1) + w(n) w(n 1)

E[x{n)]

^E[x(n-l)]

since E[w{n)]

0, it follows that E[x(n)] = 0

E[w(n))-E[w(n-l)]

To determine the autocorrelation, we have

MO) = jM-i) + *(0)-6(-i) = - i

Mi) = 5M0)+ (i)-*(0) = - i
381

p = q = l,

= ~2'

o = l,

&i = - l

Hence, 7xx(0)

- 7 ( l ) + ,(1 + I )

7xxd)

57(0) + <ri(-l)

and 7 ( 0 )

-<r

7rx(l)

7xx(m)

-a17(m-l)

ITxx(m)

7rr(-m)

-\cl

12.14
x(n)
E[x(n)]

=
-

u>(n)- 2u>(n- I) + w(n - 2)

0 since E[w(n)} = 0

7xx(m)

*^PM*+m.

0<m<g

fc = 0

where q - 2,

6Q = 1,

*i = - 2 ,

62 = 1

Hence, 7 xx(0)

*i

*=

6<r

"

7xx(i) = ^i 516*6*+i = - 4 c r i
2

7xx(2) = <r5I 6 k 6 j k + 2 = ^
k=0

7xx(m)
7*x(-m)

=
=

0,
\m\ > 3,
7xx(m)

12.15
(a)

r(*) = 2>(m)*-
m

2z"2(z4-2r3 + 3 r 2 - 2 z + l ]

Gil-z-1

1JV3 1JV3
2
'
2
The minimum-phase system is

The four zeros are

H(z)
Hence, / / ( r )

(b) The solution is unique.

382

+Z'2),
1

where G = \/2
2

>/2(l - z" + z~ )

12.16
(a)
oo

Txx{z)

7xx(m);

m = oo
,2

( 6 - 3 5 z - 1 + 6 2 z - 2 - 3 5 z " 3 + 6z- 4 )
62

=
=

^ ( 1 - 3 * - l ) ( l ~ 2*" 1 )(1 - \^)(l

The four zeros are z

3,2, - , -

The minimum phase system is H(z)

^=(1 - z ~ 1 ) ( l - 2 _ 1 )
v62
2
3

fi

- i*-1)

(b) The maximum phase system is H{z) = ^ r ( l 5 z - 1 + 6z~ 2 )

(c) There are two possible mixed-phase systems: H\{z) = -775(3 7 z - 1 + 2z~ 2 )
^(2-7z->+3z-2)

12.17
(a)

Thh{f)

1-0.8Z"1
H{z)H{z-')\^t.t
1 + c~>2^

1 + e>2T>

l-0.8e-->2,r/ 1 -0.8e>2*'
COS1 IT f

7rr()

(-)M

r(/)

1.64-

1.6COS2TT/

(ly-ie-i"/m = oo

0.75

ryv(/)

1.25-C0527T/

r(/)rfcfc(/)
3cos2?r/
(1.64 -

1.6COS2TT/)(1.25

cos2xf)

(b)

ryy(/) =

54

1.64 l.6cos2ir/

1.25 - cos2xf

9
=

7vv(m)

150

H
1.64 - 1.6cos2:r/

150(0.8)^1-50(^)^1
383

50

1
1.25 -

CO$2TT/

#2(2)

c) (rl = 7 ( 0 ) = 150 - 50 = 100

12.18
proof is by contradiction.
(a) Assume the \km\ > 1. Since Em (1 - |fc m | 2 )m-i> this implies that either Em < 0 or
Em-\ < 0. Hence, a\. < 0, and

o ' r ^ a = a

=*r,

is not positive definite.

(b) From the Schur-Cohn test, Ap(z) is stable if |Jbm| < 1. Hence, the roots of Ap(z) are inside
the unit circle.

12.19
(a)
x(n)

aix(n - 1) a2ir(n - 2) - a3x(n - 3)

Butx(n)

i l * ( n - l ) + x(n - 2) - ^ x ( n - 3) + u;(n)
i

{[x(n) x(n)] 2 } is minimized by selecting the coefficients as a\ = ^,02

= ~ ^ ' a 3 = 74

(b)
7rx(m)

-^a*7xx(-*),

m > 0

fc=i
p

- 5 ^ o * 7 ( m - *) + <r*,

m= 0

Since we know the {at} we can solve for 7xx(m), m = 0,1,2,3. Then we can obtain 7xx(m)
for m > 3, by the above recursion. Thus,
7xr(0)

4.93

7xx(l)
yxx(2)

=
=

4.32
4.2

7(3)

=
=
=

3.85
3.65
3.46

7xx(4)
7(5)
(c)
- / x

r, , *

14

-1

_,

384

-2

14 _o

-3

M(z) -

M*)

k3B3(z)

=
=
=

k2
B2(z)

Mz)

1 -0.569*-1 -0.351*-2
-0.351
-0.351 -0.569Z-1
A2(z) - k2B2{z)
\-k\

+ z'7

1
= 1 - 0.877*"
-0.877

*i

12.20
(a)
7xx(0)
7("1)
7rr(~2)

7xx(l)
7xx(0)
7rx("l)

Hence,

7xx(m)
7xx(m)

The values of the parameters dr

MA(2)
MA{A)
MA{$)

7xx(2)
7rr(l)
7xx(0)

1
0
-0.81

0.817xx(m-2),

{2.91,0,2.36,0,1.91,0,1.55,0,...}

> bkbk +m are as follows:

=
=
=

{2.91,0,2,36}
{2.91,0,2,36,0,1.91}
{2.91,0,2,36,0,1.91,0,1.55,0}

r*ii
0
0

m> 3

(b) The A/.4(2), MA(4)andMA(S)

models have spectra that contain negative values. On the
other hand, the spectrum of the AR process is shown below. Clearly, the MA models do not
provide good approximations to the AR process. Refer to fig 12.1.

12.21
7xx(m)= {l.656<r,0 l 0.8hr,0,...}.
For AR(2) process:
' 1.656*2
0
0.81<r

0
1.656(7=
0

0.81*2, '

1.656*5, _

. 2 .

" 9*1
=

0
0

The solution is
9
a\
a2
For the AR(4) process, we obtain g
a

=
=
=

1.12
0
-0.489
1.07 arid
{1,0

385

-0.643 ,0,0 314}

a>
X3

..

p.

. . - ^r

/
/

24

/-

I
1

a
S? 3

fc
?2
i
i

y _ _ _ _ - - ^

1
n

0.05

0.1

0.15

0.2
0.25
0.3
> frequency(Hz)

_i

0.35

_i

0.4

0.45

Figure 12.1:
1.024 and
{1,0,-0.75,0,0.536,0,-0.345,0,0.169}

For the AR(8) process, we obtain g

a
Refer to fig 12.2.

12.22
(a)
_
"

K }

4al(2-z-*)(2-z)
(3-z-*)(3-r)

The minimum-phase system function H(z) is

2 2 - z_1

4 1 - f iz , - i
9 1-iz-i
(b) The mixed-phase stable system has a system function
H{Z)

~ 3lT?T
2 1-2Z-1
9 1-iz-i

12.23
(a)
i4 2 (z)

l - 2 r c o 5 e z _- 11 +, -r2lz. - 2
386

0.5

MA(2)

AR(2)

0.2
0.4
> frequency(Hz)
AR(4)

0.6

0.2
0.4
> frequency(Hz)
AR(8)

02
0.4
> lrequeocy(H2)

0.2
0.4
> frequeocy(Hz)

Figure 12.2:
-

-2

B2{z)

Ai{z)

r2 -2rcosBz~1
+
A7(z) - k2B2(z)

1-

\-k\
2rcosS

-2

-i

1 + r*
2rcosS
=
1 + r2

Hence, k\
(b) As r ' 1, *2 1 and ki cosQ

12.24
fll(l)

Hence, i4 3 (z)

-1.25, a 2 ( 2 ) = 1.25, a 3 (3) = - 1

1 - 1.252"1 + 1.25z"2 - z " 3

First, we determine the reflection coefficients. Clearly, Jb3 = - 1 , whcih implies that the roots
of / M 2 ) a r e o n l n e u m t circle. We may factor out one root. Thus,
1

2
A3(z) = ( l - z" ^
MK
O -i; -* "^1 +' r- *)

where a

(l-Z-^Cl-OZ-^Cl-Q'z-1)

l+j>/63
8
387

Hence, the roots of Az(z) are z l , a , and a*.

(b) The autocorrelation function satisfies the equations
7*.(m) + ; > 3 ( * h ( m - * ) = { o 2 "

7(0)
7xx(l)
7xx(2)
7xx(3)

7(1)
7xx(0)
7xx(l)
7xx(2)

7xx(2)
7xx(l)
7xx(0)
7xx(l)

7xx(3) "
7xx(2)
7xx(l)
7xx(0)

7<~m<

1
-1.25
1.25
-1

0
0
0

(c) Note that since ^3 = 1, the recursion EJm = & . i ( l - l*m|2) implies that Ef3 = 0. This
implies that the 4x4 correlation matrix TXx is singular. Since 3 = 0, then er = 0

12.25
1
-0.5
0.625
-0.6875

7xx(0)

7xx(l)
7xx(2)
7xx(3)
Use the Levinson-Durbin algorithm

=
=
=

i(l)

*iC0

-k1

Ei

(l-a?(l))7xx(0) = ?

a 2 (2)

Oj(l)

Therefore, AQ(Z)

7xx(l) _ 1
7xx(0)
2

- * , - !

=>*2 =

7xx(2) + a 1 (l) T *x(l) _

Ei
a 1 (l) + a 2 ( 2 ) a 1 ( l ) = i
,
1 +

1 -1
1 -2
4*
-2Z

E2 =
s(3)

s(2)

s(l)

Therefore,.^*)

(l-a2(2))E1

7xx(3) + o 2 (l) 7 xx(2) + a 2 (2)7xx(l)

2
a2(2)+a3(3)a2(l) = - 5
a2(l) + a3(3)a 2 (2) = 0

1
=> *3 =

1
2

388

E3

(i_a2(3))2=27

12.26
(a) (1)
1 - e--"
1 + 0.816-^
\H(w)f<T2w
1 - e-iw
IVH.
l
l + 0.81e-J u '

H(w)
Tzx{w)
Txx(w)
(2)

(l-c-j2u')

H(w)
Txx(w)

(3)

H(w)

Txx(w)

1 -0.81c--""
1.6561- 1.62cosu;

(b) Refer to fig 12.3.

(c) For (2),
f ^Lo6*6k+m,
7xx() = < 0,
I 7(-).

since 6o =

0<m<2
m > 2
m< 0

1, fci = 0 and 62 = 1, we have

7xx(0)

2crl

7xx(2)

-ffl

7xx(-2)
7(m)

=
=

-*l
0,
m^0,2

For (3), the AR process has coefficients ao = l , a i = 0 and 02 = 0.81.

1
0
0.81

0
1.81
0

7xx(0)
7xx(m)
7xx(m)

=
=
=

0.81
0
1

' 7xx(0) "

7(1)

L Txr(2) J

2.9*1
0,
m odd
2.9(0.9)|m|<7*,
389

m even

0.2
0.4
> trequency(Hz)

0.2
0.4
> frequency(Hz)
(3)

0.2
0.4
> frequency(Hz)

Figure 12.3:

12.27
(a)

rrx(*) = JT 7 (m)*- m

1-1*

1-J*-I
11
(1_I2)(1_I2-1)
since Txx(z)
H(2)

e3H{x)H{x-1),
0.968
1-Jz-i

is the minimum-phase solution. The difference equation is

x ( n ) = - x ( n - l) + 0.968u;(n)
390

where w(n) is a white noise sequence with zero mean and unit variance.
(b) If we choose

H(z)

*-i*
,-i

z-.-l

then, x(n)

Az- l
1-4Z"1
4x(n - 1) - 4x0.968u;(n - 1)

12.28
7xx(0)

7xx(l)
7xr(2)

=
=

0
-a2

7xx(3)

*> = - S S =
Ai(z)
=> *i = 0

(l-aj(l)h(0) = l

=
=
=

7xx(2)-fai(l)7xx(l)
,
= a
a 1 (l) + a 2 (2)a 1 (l) = 0
1 + a 2 z~ 2

,9*
a 2 (2)
a 2 (l)
Therefore, A2(z)
=> k2 = a 2
2
,<n
a3(3)
a 3 (2)
a 3 (l)
Therefore,^*) = A2{z)

=
=
=
=

(l-a2(2))1 = l - a 4
7xx(3) + Q 2 (l)7rx(2) + a 2 (2) 7 xx(l)
ft
7;
= 0
a 2 (2) + a 3 (3)a 2 (l) = a 2
a 2 (l) + a 3 (3)o 2 (2) = 0
1 + a2z~2

1 - a4

=> Jb3 = 0

E3 = E2

12.29
(a) For the Bartlett estimate,
M

{b)M

391

0.9
A/
0.9
= 90
0.01
0.9

= 535=45

(c)for(a), QB

for(b), QB

M
2400
= 26.67
90

IL
M
2400 = 53.33
45

12.30
Ap(z) = Ap.x{z)

kpBp-X{z)z - l

where Bp-\(z) is the reverse polynomial of

Ap-i(z).
For |ibp| < 1, we have all the roots inside the unit circle as previously shown.
For |jbp| = 1, Ap{z) is symmetric, which implies that all the roots are on the unit circle.
For |Jbp| > 1, Ap(z) A,(z) + eBp-i(z)z~l,
where A,(z) is the symmetric polynomial with all
the roots on the unit circle and Bp-\(z) has all the roots outside the unit circle. Therefore, Ap(z)
will have all its roots outside the unit circle.

12.31
M2,r/-o-9|2

r m- ^

zxU) - v* \eJ7wJ _ j 0 .9| 2 |e> W

j0.9| 2

(a)
r,

/ x

2~0.9

Z~l

-0.9

z + 0.81 z - 2 + 0.81
Therefore, ff (r)7

-r2 2 + 0.81

z-l(l-0z-1)
1 + 0.81*- 2
(b) The inverse system is
1 + 0.812" 2
z-^l-O-Sz-1)

1
H(z)
This is a stable system.

12.32
N-l
- l3in

X(k) =

*(n)e^

n=0

(a)

E[X(k))

[x(n)]<r^=0
n

392

N-l

>
x

>

n=0

Nerl

(b)
E{X(k)X-(k-m)}

E[*(n)xm(n')}e=^ef3wn'V~n)

Y, L
n

n'

n'

7V<7*,

0,

m = pTV
otherwise

p = 0,l,2,..

12.33
"Txjvim) =

;[t;*(n)v(n + m)]

] T Yl fyk>E[wm(n - k)w(n + m - it')]

i'=0t=0

Then, r v , ( / )

<fme-;'w'

<ri
m= 9

12.34
7xx(m)

[x*(n)x(n + m)]

>l2'{cos(u;in 4- <t>)cos[wi(n + m) + <f>]}

A2
-E{coswim + cos[wi(2n + m) + 2<j>)}

COSW\Tl

12.35
(a)
x(n)
y(n)

=
=

=> x(n)

y(n) - v(n)

0.81y(n - 2) - 0.81v(n - 2) + w{n)

y(n)-v(n)

0 . 8 1 x ( n - 2 ) + iu(n)
x(n) + v(n)

393

Therefore, y(n)
so that y(n) is an ARMA(2,2) process

0.81y(n - 2) + v(n) - 0.81t;(n - 2) + w(n)

(b)
p

x(n)

akx(n - fc) + w(n)

*=i

y(n)

x(n) + v(n)

s x(n)

y(n) - v(n)

y(n)-i/(n)

- ^ak[y{n

- k) - v{n - *)] + w(n)

k=i
p

V(n) + X ^ a * y ( n " *)

t, n

( ) + y ^ a*v(n - fc) + n;(n)

Hence, y(n) is an ARMA(p.p) process

p

Note that X(z)[l + a * * ~ * ]

Txx(z)
andryy(z)

W()

1
AP(z)
^//(^(z"1)

<ri/f(z)/f(*- ) + rj
-2
,

*"

Xp(zMP(z-i)

-2

i4 p (zM P (2-)

12.36
(a)
K
7rr(m)

{[^AiC05(u;kn + ^ t ) +

u; n

k-i

fc'=i

5 Z 5 ^ AkAk>E{cos{wkn
*

( ) ] [ 5 Z Ak'COs{wk>{n + m) + <j>k>) + w{n + m)]}

+ 0fc)cos(u;fc.(n + m) + <^t)} + E[w{n)w(n + m)]

k'

*= 1

(b)

TZT(w) =

Y,

-y(m)e-^
+ <r2,

i =l

m = -oo

394

A A2
=

2Z -T-faH - uuk - 2nm) + 2*6(w + Wk - 2irm)} + <r

i =l
K

-Y^AliHw-^k
1

- 2irm) + 2ir6(w + wk - 2:rrn)] + a2w

*=i

12.37

=>Lyya-Xa
or y y a

= 0
= Xa

Thus, a is an eigenvector corresponding to the eigenvalue A. Substitute y y a = Aa into . Then,

S = A. To minimize , we select th smallest eigenvalue, namely, <r.

12.38
(a)
7xx(0)
7(D
7(2)
By the Levinson-Durbin algorithm,
i(l)

P+<r2w
Pcos2nfi
PcosAvf

7,.(1)
T,(0)

=
=

PC0527T/X

/>+*
*i

=
=

i(l)

(1-*J)7(0)
P2sin22*h + 2/V + <r

/> + *

(b) A:2 = ^2(2)

a2(2)

aj(l)

7(2) + ai(l)7(l)
1
Pclcos4nfi - P7sin22irf1
P2sin22irf1+2P(Tl
+ <r4J
a l (l) + a2(2)a,(l)
Pcos2irfi
P2sin22irfi Palcostrfi.
2
2
P
sin
2Tf
+2P<rl
+ cr4/l
1
P+vl

ibi = a i ( l ) as given above.

(c)

If <r^,
0, we have
a2(l)

-(co2ir/ l )(l+l)

=
a 2 (2)

-2cos2*fi
1
395

k2
jtt

=
=

1
-cos27r/i

12.39
(a)
N-l

7ry(no)

$ > ( n - n o ) [ y ( n - n o ) + u,(n)]
n=0
N-l

:[7xv("o)] =

[y 2 (-"o)]
n=0

^;[A2coS2ti;o(n-no)]

0 < n < M - l

n= l

Mil
var[7xy("o)]

2
[7xy( n o)]( )

E{y(n no)(y(n no)+u;(n)ly(n no)[y(n no)+w{n )])

=
n

"

n'

MA7

"

'"

(b)
SNR

{E[7ry(no)]} ;
J / Ni
var[7 iy
var[7 iy (n 0 )J

(*)*_
MA'

M>i2
2*2,
(c) As M increases, the SNR increases.

12.40
Refer to fig 12.4.

12.41
Refer to fig 12.5.

396

'"

'

periodogram Pxx(f)

autocor of w(n)

80
60

40

20

-20

-10

10

20

avg periodogram Pxx(f)

Figure 12.4:

397

200

400

600

BartlettwithM=:50

theoretical psd with M 100

-20
-40

Blackman-Tukey psd with lag=25

120

398