\ I

- f

( .

\ .

; .. ~'.' i

; ~

l . ~

\., ;-

'I

,/ .

INSTRUCTORS MANUAL

for

Discrete-Time Signal Processing, Second Edition

by Alan V. Oppenheim and Ronald W. Schafer with John. R. Buck

Solutions Prepared by

Li Lee, Jordan Rosenthal, Maya Said, and Greg Slabaugh

Matlab Links and Exam Problems Prepared by John R. Buck

Pearson Education! s) Pt 482, F.: E Patparzaru

I

I '

I

I

I 1

I

i

Lte 1

j

,r
":.___
,,;f:~,
','to -~
,"'
,tl
... ,
(;
(i
fi
f'
;::
'{;_
l'
~
<f~
~
"'--
;
t
f ( ( { [

:i

3

2.1. (a) T(x[nJ} = g[n]x[n]

• Stable: Let Ix[nJl ~ M then IT[x[nJl ~ Ig[nJiM. So, it is stable if Ig[n]1 is bounded.

• Causal: ydn] = g[n]xt[n] and Y2[n] = g[n]x2[n], so if xdn] = x2[n] for all n < no, then Yl [n] = Y2[n] for all n < no, and the system is causal.

• Linear:

T(axl[n] + bX2[nJ) = g[n](axdn] + bX2[n]

= ag[n]xdn] + bg[n]x2(n] = aT(xdn]) + bT(x2[nJ}

So this is linear.

• Not time-invariant:

T(x(n - no]) = g(n]x[n - no]

¥ yIn - no] = g(n - no]x(n - noJ

which is not TI.

• Memoryless: y(nJ = T(x(nJ) depends only on the nth value of x, so it is memoryless.

(b) T(x[n]) = L~=no x[kJ

• Not Stable: Ix[nJI ~ M -+ IT(x[n])1 ~ L~=no Ix[kJl ~ In - noiM. As n -t 00, T -t 00, so not stable.

• Not Causal: T(x[nJ) depends on the future values of x(nJ when n < no, so this is not causal.

• Linear:

n

T(axdnJ + bX2[nJ) = L axdkJ + bX2[kJ

n n

= a L xdnJ + b L x2[nJ

k=no k=no

The system is linear .

• Not TI:

n

T(x[n - no]) = L x[k - noJ

k=no

n-no

= L x[kJ

k=O

n-no

¥ y(n - noJ = L x[k]

k=no

The system is not TI.

• Not Memoryless: Values of y[nJ depend on past values for n > no, so this is not memoryless.

(c) T(x[nJ) L~~~~no x[kJ

• Stable: IT(x[n])1 ~ L;~~~no Ix[kJl ~ L;~~~no x[kJM ~ 12no + 11M for Ix[nJi ~ M, so it is stable.

• Not Causal: T(x[nJ) depends on future values of x[nJ, so it is not causal.

, .....

4

• Linear:

n+no

T(aXl[nJ + bX2[nJ) = L axdkJ + bX2[kJ

k=n-no

n+no

= a L

n+no

xdkJ + b L xdkJ = aT(xdnJ) + bT(x2[nJ)

k=n-no

k=n-no

This is linear.

• TI:

n+no
T(x[n - no] = L x[k - no]
k=n-no
n
= L x[k]
k=n-no
= y[n - no] This is TI.

• Not memory less: The values of yIn] depend on 2no other values of z , not memoryless.

(d) T(x[nJ) = x[n - no]

• Stable: IT(x[nJ)I = Ix[n - no]1 ~ M if Ix[n] ~ M, so stable.

• Causality: If no ;::: 0, this is causal, otherwise it is not causal.

• Linear:

T(axdn] + bx2[n]) = axt(n - no] + bX2[n - no] = aT(xdn]) + bT(x2[nJ)

This is linear.

• TI: T(x[n - nd] = x[n - no - nd] = yin - nd]. This is TI.

• Not memory less: Unless no = 0, this is not memoryless.

(e) T(x[nJ) = ez[n]

• Stable: Ix[nJi ~ M, IT(x[n])1 = lez[nll ~ e1z[nll ~ eM, this is stable.

• Causal: It doesn't use future values of x[n], so it causal.

• Not linear:

T(axl[nJ + bX2[n]) = eaz1[n]Hz2[nl = eaztln]ebZ2[n]

:j:. aT(xdnJ) + bT(x2[nJ)

This is not linear.

• TI: T(x[n - no]) = ez(n-no] = yin - no], so this is TI.

• Memoryless: yin] depends on the nth value of x only, so it is memoryless.

(f) T(x(n]) = ax[nJ + b

• Stable: IT(x[n])1 = lax[nJ + bl ~ alMI + Ibl, which is stable for finite a and b.

• Causal: This doesn't use future values of x[n], so it is causal.

• Not linear:

T(cxdn] + dX2[nJ) = acxdn] + adx2[n] + b =1= cT(xdnJ) + dT(x2[nJ)

This is not linear.

e f.\ ;

fl

f:

II .' •

• f

f I r I r

5

• TI: T(x[n - noD = ax[n - no] + b = y[n - no]. It is TI.

• Mernoryless: y[n] depends on the nth value of xfn] only, so it is memoryless.

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

• Stable: IT(xfn])l ::; Ixf-nJi ::; M, so it is stable.

• Not causal: For n < 0, it depends on the future value of x[nJ, so it is not causal.

• Linear:

T(axdnJ + bX2fnJ) = axd-nJ + bX2xf-n] = aT(xtfnJ) + bT(x2[nJ)

This is linear.

• Not TI:

T(x[n - no)) = x[-n - noJ

:f; yfn - noJ = x[-n + no]

This is not TI.

• Not memoryless: For n :f; 0, it depends on a value of x other than the nth value, so it is not memoryless.

(h) T(x[nJ) = x[n] + urn + IJ

• Stable: IT(x[nDI ::; M + 3 for n ;::: -1 and IT(x[n])l ::; M for n < -1, so it is stable.

• Causal: Since it doesn't use future values of x[nJ, it is causal .

• Not linear:

T(axl[n) + bX2[n)) = GXl[n] + bX2[n) + 3u[n + 11 :/; aT(xtfnJ) + bT(x2fnJ)

This is not linear.

• Not TI:

T(x[n - no) = x[n - no1 + 3ufn + 1] = y[n - no1

= xfn-no]+3ufn-no+l]

This is not TI.

• Memoryless: yfn1 depends on the nth value of x only, so this is memoryless.

2.2. For an LTI system, the output is obtained from the convolution of the input with the impulse response of the system:

00

y[n] = L hfk]xfn - k1

k=-oo

(a) Since h[kJ :f; 0, for (No S n ::; Nd,

Nl

y[n1 = L h[k]x[n - k]

k=No

The input, xfnJ :/; 0, for (N2 ::; n ::; N3), so

x[n - kJ ::j= 0, for Nz ::; (n - k) ::; N3

6

Note that the minimum value of (n - k) is N2. Thus, the lower bound on n, which occur/l.Vr k = No is

(

Using a similar argument,

(

Therefore, the output is nonzero for

(

( (b) If x[nJ =j:. 0, for some no::; n ::; (no + N - 1), and h[nJ =j:. 0, for some nl ::; n ::; (nl + M - 1), the

results of part (a) imply that the output is nonzero for: (

(no + nd ::; n ::; (no + nl + M + N - 2)

So the output sequence is M + N - 1 samples long. This is an important quality of the convo~utt~'1 for finite length sequences as we shall see in Chapter 8.

2.3. We desire the step response to a system whose impulse response is

h[n] = a-nuf-nJ, for ° < a < 1.

The convolution sum:

(

00

y[nJ = I: hfkJx[n - kJ

k=-oo

The step response results when the input is the unit step:

x[nJ = urn] = { 1, 0,

for n ~ ° for n < 0

Substitution into the convolution sum yields

00

yfnJ = L a-kuf-kJufn - k]

k=-oo

For n::; 0:

00
y[n] = I: a-k
k=-oo
00
= I: ak
k=-n
a-n
=
I-a For n > 0:

0
yfn] = I: a-k
k=-oo
00
= I:ak
k=O
1
I-a e c' ;t:)

« I.,

€! f) (;)

• f!

(")

:( ;

-c,

'7 ~._i

7

2.4. The difference equation:

3 1

y[n]- 4y[n - 1] + gy[n - 2] = 2x[n - 1]

To solve, we take the Fourier transform of both sides.

The system function is given by:

The impulse response (for x[n] = o[nJ) is the inverse Fourier transform of H(eiw).

H(eiW) = -8 8

1 + le-iw + 1 - le-iw

4 2

Thus,

2.5. (a) The homogeneous difference equation:

y[n]- 5y[n - 1] + 6y[n - 2] = 0

Taking the Z-transform,

1 - 5z-1 + 6z-2 = 0 (1- 2z-1)(1 - 3z-l) = O.

The homogeneous solution is of the form

(b) We take the a-transform of both sides:

Thus, the system function is

H(z) =

Y(z) X(z)

=

=

1 - 5z-I + 6z-2

-2 2

1 - 2z-I + -1---3-z--""'-I'

where the region of convergence is outside the outermost pole, because the system is causal, Hence the ROC is [z] > 3. Taking the inverse z-transform, the impulse response is

8

(c) Let x[nJ = u[nJ (unit step), then

1

X (z) = -1-_-Z--~l

( (

and

Y(z) = X(z)· H(z)

=

(1 - z-l )(1 - 2Z-1)(1 - 3z-1)·

Partial fraction expansion yields

1 4 3

Y(z) = 1 _ z-l - 1 - 2z-1 + 1 - 3z-1 .

The inverse transform yields:

y[nJ = u[nJ - 4(2)nu[nJ + 3(3)nu[nJ.

2.6. (a) The difference equation:

1

y[nJ - 2y[n - IJ = x[nJ + 2x[n - IJ + x[n - 2J

(

Taking the Fourier transform of both sides,

Y(eiw)[1 _ ~e-jwJ = X(eiw)[l + 2e-iw + e-i2wJ. 2

Hence, the frequency response is

Y(eiw) X(eiw)

1 + 2e-iw + e-i2w 1 - le-iw

2

\

=

(b) A system with frequency response:

1 + le-iw + ~e-i2w

2 4

Y(eiw) X(eiw)

=

cross multiplying,

. 1· 3.2 . 1· ·3

Y(eJW)[1 + 2e-JW + 4e-J Wj = X(eJW)[1 - 2e-}W + e-J w],

and the inverse transform gives

1 3 1

y[nJ + 2y[n - IJ + 4y[n - 2J = x[nJ - 2x[n - IJ + x[n - 3J.

2.7. x[nJ is periodic with period N if x[nJ = x[n + NJ for some integer N. (a) x[nJ is periodic with period 12:

7r •

=::::} 2rrk = 6" N, for integers k, N

Making k = 1 and N = 12 shows that x[nj has period 12 .

. ~

0
.
e
.
,,, ;
»»
f'
.
c
f'
fl
e)
c
.1
.'
e,
c'
()
f
.'
e)
•
f
(
(
•
I
I
I
•
I t ,
~'
;
t 9

(b) x[n] is periodic with period 8:

37T

==::} 27Tk = 4N,for integers k,N

==::} N = ~k, for integers k, N

The smallest k for which both k and N are integers are is 3, resulting in the period N being 8, (c) x[n] = [sin(7Tn/5)]/(7Tn) is not periodic because the denominator term is linear in n.

(d) We will show that x[n] is not periodic. Suppose that x[n] is periodic for some period N:

ei(7in) = ~(7i)(n+N) = ei(7in+2d)

==::} 2rrk = ~N,for integers k, N ==::} N = 2V2k, for some integers k, N

There is no integer k for which N is an integer. Hence x[n] is not periodic.

2.8. We take the Fourier transform of both h[n] and x[n], and then use the fact that convolution in the time domain is the same as multiplication in the frequency domain.

H(dW) 5
= 1 + !e-jw
2
Y(dW) = H(dW)X(dW)
5 1
= 1 + !e-iw 1- le-jw
2 3
3 2
= 1 + !e-iw + 1- le-jw
2 3
y[n] 1 1
= 2(3 )nu[n] + 3( - 2 )nu[n] 2.9. (a) First the frequency response:

Y(ejW) _ ~e-iwY(dW) + .!.e-2jwY(dW) = .!.~-2jw X(ejW) 663

Y(ejW)

=

X(ejw)

le-2jw 3

=

Now we take the inverse Fourier transform to find the impulse response:

Pearson [d:lr;atio!1( ~) Pt

t.:'!, l"' ,J [< P3':]),; ,,'

Delr ,": 'OOc!.'~ ; I

~ 2Z146067~70 i'<L<;.2.:H'

r,

, .. (

10

For the step response s[n]:

(

00

s[n] = L h[k]u[n - k]

k=-oo

n

= L h[kJ

k=-oo

1 - (1/3)n+I 1 - (1/2)n+l

= -2 1 _ 1/3 urn) + 2 1 _ 1/2 urn]

= (1 + (~)n - 2(~)n)u[n]

(b) The homogeneous solution Yh[n] solves the difference equation when x[nJ = O. It is in the form Yh[n] = L: A(c)n, where the c's solve the quadratic equation

c2 - ~c+ ~ = 0

6 6

So for c = 1/2 and c = 1/3, the general form for the homogeneous solution is:

Yh[n] = Al(~)n + A2(~)n

(c) The total solution is the sum of the homogeneous and particular solutions, with the particular' solution being the impulse response found in part (a):

yin] = Yh[n] + Yp[n)

1 1 1 1

= AI(2)n + A2(3)n + -2{ 3)nu[n] + 2(2)nu[n]

Now we use the constraint y[O] = y(l] = 1 to solve for Al and A2:

y[OJ = Al + A2 - 2 + 2 = 1

y[IJ = AI/2 + A2/3 - 2/3 + 1 = 1 Al + A2 = 1

Ad2 + A2/3 = 2/3

With Al = 2 and A2 = -1 solving the simultaneous equations, we find that the impulse response i

is

2.10. (a)

y[n] h[n] * x[n)

00

= L aku[-k - l]u[n - k)

k=-oo

=

( ,t a', n::; -1
-1
L ak, n> -1
k=-oo
{ an
1 - l/a' n S-1
l/a n> -1
1 - l/a' =

(~

r:
r>
r:
(,
.:
(
(-
r:
(,
~
, ~
r
, ~
,~
r:
I~ -~

11

(b) First, let us define v[n] = 2nu[-n - 1]. Then, from part (a), we know that

{ 2n+l n < -1

w[nJ == urn] * v[nJ == 1, ' n;-1

Now,

y[nJ == urn - 4J * v[nJ
== w[n - 4J
== { 2n-3, nS3
1, n>3 (c) Given the same definitions for v[n] and w[n] from part(b), we use the fact that h[n] == 2n-1u[-(n_ 1) - 1J = v[n - 1] to reduce our work:

y[n] = x[n] * h[n]

= x[n] * v[n - 1] = w[n -1]

= {2n, n S 0 1, n > 0

(d) Again, we use v[n] and w[n] to help us.

y[n] = x[nJ * h[n]

= (u[n] - urn - 10]) * v[n] = w[n] - w[n - 10]

= (2n+lu[_(n + 1)] + urn]) - (2n-9u[_(n - 9)] + urn - 10])

{ 2(n+l) - 2(n-9), n S -2

= 1_2(n-9), -1SnSS

0, n 2: 9

2.11. First we re-write x[n] as a sum of complex exponentials:

Since complex exponentials are eigenfunctions of LTI systems, H(ej1r/4)ejrrn/4 _ H(e-j1r/4)e-j1rn/4

y[n] == 2j

Evaluating the frequency response at w = ±tr/4:

1 -j1r/2

- e ,= 2(1 _ ') = 2V2e-j1r/4

1 + 1/2e-)1r J

'1r/2

1 - eJ , = 2(1 + j) = 2V2ei1r/4 1 + 1/2e)1t

We get:

2V2e-j1f/4d1rn/4 _ 2v'2d1r/4e-j1rn/4 2j

= 2v'2sin(trn/4 -tr/4).

yfn] =

, tJ IT 11

__...;__ -

2.12. The difference equation:

Yin) = ny[n - 1) + x[n)

Since the system is causal and satisfies initial-rest conditions, we may recursively find the response to

any input. (1

y[n) = 0, for n < 0 y[O) = 1

y[I) = I

y[2J = 2

y(3) = 6

y[4] = 24

(a) Suppose x[n) = b[n):

y[n] = h[n) = n!u[n]

y[n] = 0, for n < 0 y[O] = a + b y[l] = a + b y[2] = 2(a + b) y[3] = 6(a + b)

(b) To determine if the system is linear, consider the input:

x[n] = ab[n) + bOrn)

performing the recursion,

y[4] = 24(a + b)

Because the output of the superposition of two input signals is equivalent to the superposition 0/1 the individual outputs, the system is LINEAR. ( 'j'.

(c) To determine if the system is time-invariant, consider the input:

x[n] = o[n - 1]

the recursion yields

y[n] = 0, for n < 0

y[O] = ° y[1] = 1

y[2] = 2 y[3] = 6 y[4] = 24

Using h[n) from part (a),

h[n - 1) = (n - 1)!u[n - 1J f:. y[n]lx[n]=6[n-l] Conclude: NOT TIME INVARIANT.

(I

~

2.13. Eigenfunctions of LTI systems are of the form an, so functions (a), (b), and (e) are eigenfunctions.

Notice that part (d), cos(won) = .5(ejwon + e-jwon) is a sum of two an functions, and is therefore not an eigenfunction itself.

2.14. (a) The information given shows that the system satisfies the eigenfunction property of exponential sequences for LTI systems for one particular eigenfunction input. However, we do not know the system response for any other eigenfunction. Hence, we can say that the system may be LTI, but we cannot uniquely determine it. ====> (iv).

(b) If the system were LTI, the output should be in the form of A(I/2)n, since (1/2)n would have been an eigenfunction of the system. Since this is not true, the system cannot be LTI. ====> (i).

(c) Given the information, the system may be LTI, but does not have to be. For example, for any input other than the given one, the system may output 0, making this system non-LTI. ====> (iii). If it were LTI, its system function can be found by using the DTFT:

Y(ejW) H(dW) = X(ejw)

1

h[n]

2.15. (a) No. Consider the following input/outputs:

Xl [n] = o[n] ====> x2[n] = o[n - 1] ::::}

\

Even though x2[n] = xdn - 1], Y2[n] 1: YI[n - 1] = (Vn-Iu[n - 1]

(b) No. Consider the input/output pair x2[n1 and Y2[n] above. x2[n1 = 0 for n < 1, but Y2[0] 1: o.

(c) Yes. Since h[n1 is stable and multiplication with u[n1 will not cause any sequences to become unbounded, the entire system is stable.

~.16. (a) The homogeneous solution Yh[n] solves the difference equation when x[n] = O. It is in the form Yh[n1 = I: A(c)n, where the c's solve the quadratic equation

1 1

C - -c+ - = 0 4 8

So forc = 1/2 and c = -1/4, the general form for the homogeneous solution is:

Yh[n1 = AI(~)n + A2(-~)n (b) Taking the z-transform of both sides, we find that

Y(z)(1 - ~Z-l - ~Z-2) = 3X(z)

and therefore

H(z) = Y(z)
X(z)
3
= 1 - 1/4z-1 - 1/8z-2
3
= (1 + 1/4z-1 )(1 - 1/2z-1)
1 2
= +
1 + 1/4z-1 1 - 1/2z-1 Pears-in I ":C2tion( s) Pr 1M. 482, r , I~ Paeparg;lJ:!

Deliw: ; 0002 I Del fa

~'" 2214606"1· 7UC.E~;.: ; 46(J71

14

,_ . .:....

the causal impulse response corresponds to assuming that the region of convergence extends outside the outermost pole, making

he(nJ = «-1/4t + 2(1/2t)u[nJ

The anti-causal impulse response corresponds to assuming that the region of convergence is inside the innermost pole, making

hac[nJ = -«-1/4r + 2(1/2)n)u[-n - 1J

(c) he[nJ is absolutely summable, while hae(nJ grows without bounds. (d)

Y(z) = X(z)H(z) 1

1

=

1- iZ-1 . (1 + iz-l)(1 - iz-l)

1/3 + 2 + 2/3

1 + 1/4z-1 1 - 1/2z~1 1 - 1/2z-1

1 1 1 2 1

= 3(4 )nu[nJ + 4{n + 1)( 2 )n+lu[n + 1] + 3( 2 )nu[nJ

=

y[nJ

2.17. (a) We have

r[nJ = { 1,

0,

for 0 ::S n ::S M otherwise

Taking the Fourier transform

M R(dW) = L e-jwn

n=O

1 - e-j..,(M+l)

=

1- e-jw

(b) We have

w[nJ = { HI + cose;r)'

0,

for O::S n S M otherwise

" .'

We note that,

1 27rn

w[nJ = r[nJ . 2[1 + cos( M )].

Thus,

15

(c)

IR( e-i co )1

-It

It (j)

IW(ej co )1

-It

It (j)

2.18. h[nJ is causal if h[nJ = 0 for n < O. Hence, (a) and (b) are causal, while (c), (d), and (e) are not. 2.19. h[nJ is stable if it is absolutely sum mabie.

(a) Not stable because h[nJ goes to 00 as n goes to 00. (b) Stable, because h[n] is non-zero only for 0 S; n S; 9. (c) Stable.

-1 00

L Ih[nJI = L 3n = L(I/3)n = 1/2 < 00

n

n=-oo

n=l

(d) Not stable. Notice that

5

L I sin(1Tn/3) I == 2v'3

n=O

and summing Ih[nJI over all positive n therefore grows to 00.

(e) Stable. Notice that Ih[nJI is upperbounded by (3/4)lnl, which is absolutely summable. (f) Stable.

{ 2 -5 < n <-1

h[nJ = 1:: 0 S; ~ S; 4

0, , otherwise

So L Ih[nJl = 15.

2.20. (a) Taking the difference equation y[nJ = (I/a)y[n - IJ + x[n - IJ and assuming h[O] = 0 for n < 0:

h[OJ = 0
h[I] = 1
h[2] = l/a
h[3] = (I/a)2
h[n] = (I/a)n-1u[n - IJ
(b) h[n] is absolutely summable if II/al < lor if [c] > 1 Pearson !:"dt·(,"{>;''''I~i Ft

tl ~ ~l-IJ1 .. '- lJ i· . ,

4b2, f' ; F. f"'r;' DeU':-; :

~ . 22J46067- '

I'

_ .. ..:... .

2.~J. For an arbitrary linear system, we have

yin] = T{x[n]},

Let x[n] = 0 for all n.

yin] = T {x[n]}

For some arbitrary Xl [n], we have

Using the linearity of the system:

T{x[nJ + xdn]} = T{x[nj} + T{xl[n]} = yin] + Ydn]

Since x[n] is zero for all n,

T{x[n] + xl[n]} = T{xl[n]} = Ydn] Hence, y[n] must also be zero for all n.

2.22. We use the graphical approach to compute the convolution:

yin] = x[n] * h[n)

00

= L x[k)h[n - k)

k=-oo

(a) yin] = x[n] * h[n]

y[n] = o[n - 1] * h[n] = h[n - 1)

. r II •.

• •

o 2 3 n

(b) y[n] = x[n] * h[n]

5

n

(c) Yin] = x[n] * h[n]

5 5

o I 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

(d) y[n] = x[n) * h[n]

17

3 3

2

n

2 3 4 5

2.23. The ideal delay system:

yin] = T{x[n]} = x[n - no]

Using the definition of linearity:

T{axl[n] + bX2[n]} = axdn - no] + bX2[n - no] = aydn] + by2{n]

So, the ideal delay system is LINEAR. The moving average system:

(

I M2

y[n] = Tx[n] = M M L x[n - k]

1 + 2 + I k=-MI

by linearity:

I M2

T{axl{n] + bX2[n]) . = M M L (axdn] + bX2[n])

1 + 2 + I k=-M1

1 M2 I M2

= Ml +M2 + 1 L axdn] + Ml +M2 + 1 L bX2[n]

~-~ ~-~

= aydn] + by2[n]

Conclude, the moving average is LINEAR.

2.24. The response of the system to a delayed step:

yin] = x[nJ * h{n]

00

= L x[kJh[n - k]

k=-oo

00

= L u[k - 4Jh[n - k]

k=-oo

00

yin] = Lh[n - k]

k=4

Evaluating the above summation:

For n < 4: yin] = 0

For n = 4: yin] = h[O] = 1

For n = 5: yin] = hill + h[O] = 2

For-n = 6: yin] = h[2] + hill + h[O] = 3

For n = 7: y[n] = h[3] + h[2] + hill + h[O] = 4

For n = 8: yin] = h[4] + h[3] + h[2] + h[l] + h[O] = 2

For n 2: 9: yin] = h[S] + h[4] + h[3] + h[2J + hill + h[O] = 0

Pearson td;!~ation( s) Pt ! i'~1 482, f': E fL'rpeJTf,(. f I Delb:-ll0092 inOlj

... 22146067-70 l'.;u: .~.l± .J71

· ....,,;:.

18

2.25. The output is obtained from the convolution sum:

y[n] = x[n] * h[n]
00
= L x[k]h[n - k]
k=-oo
00
= L x[k]u[n - k]
k=-oo The convolution may be broken into five regions over the range of n:

yfn] = O,for n < 0

n

yIn] = Lak

k=O

1 - a(n+l)

I _ a ,for 0 ~ n ~ N,

=

Nt

yIn] = Lak

k=O

1 - a(Nt+l)

= , for Nl < n < N2

I-a

Nt n

y[n] = L ak + L a(k-N2)

k=O k=N2

1 - a(Nt+l) 1 _ a(n+l)

= +----

I-a I-a

2 - a(Nt+l) _ a(n+l)

1- a ' for N2 ~ n ~ (N1 + N2)

=

y[n]

Nt Nt+N2

= L ak + L a(k-N2)

k=O k=N2

Nl

= Lak+ LN1am

k=O m=O

2.26. Recall that an eigenfunction of a system is an input signal which appears at the output of the system scaled by a complex constant.

19

(a) x[n] = 5nu[n]:

00

y[n] = L h[k]x[n - k]

k=-oo

00

k:-oo

n

= Sn L h[k]S-k

k=-oo

Becuase the summation depends on n, x[n] is NOT AN EIGENFUNCTION. (b) x[n] = ei2wn:

00
yin] = L h[k]ei2w(n-k)
k:-oo
00
= ei2wn L h[k]e-i2wk
k=-oo
= ej2wn . H (ei2w) YES, EIGENFUNCTION. (c) eiwn + ei2wn;

00 00

yin] = L h[k]eiw(n-k) + L h[k]ei2w(n-k)

k=-oo

*:-00

00 00

= eiwn L h[k]e-iwk + ei2wn L h[k]e-i2wk

k=-oo

k=-oo

= eiwn. H(eiw) + ei2wn . H(ei2w)

Since the input cannot be extracted from the above expression, the sum of complex exponentials is NOT AN EIGENFUNCTION. (Although, separately the inputs are eigenfunctions. In general, complex exponential signals are always eigenfunctions of LTI systems.)

(d) x[n] = sn:

00
yin] = L h[k]5(n-k)
k=-oo
00
= sn L hlk]s-k
k=-oo
YES, EIGENFUNCTION.
; . (e) x[n] = snej2wn;
00
yin] = L h[k]5(n-k)ei2w(n-k)
k=-oo
00
= snei2wn L h[k]S-*e-i2wk
*=-0::
YES, EIGENFUNCTION. Pearson Education( sl Pt. U •. 482. F.1 E Patparc-,lj,')

DeIL,'-' ~ ,;

4 ~ L f!C11a

22146067,·70 F" ,.,,)"! ,,1""T'l'1

f-" ~'t ._t"", _l J - , "'-

20

2.27.

• System A:

1

x[n) = (2)n

This input is an eigenfunction of an LTI system. That is, if the system is linear, the output will be a replica of the input, scaled by a complex constant.

Since yin) = (i)n, System A is NOT LTI.

• System B:

x[n) = ein/8u[n)

The Fourier transform of x[n) is

0:0

X(dW) = L ejn/8u[n]e-iwn

n:::-oo

0:0

= L e-j(w-k)n

n=O

1

\ ;

=

1 - e-i(w-k)'

The output is yin) = 2x[n], thus

( jW) _ 2

Ye -( 1)' 1-e-]W-s

Therefore, the frequency response of the system is

Hence, the system is a linear amplifier. We conclude that System B is LTI, and unique.

• System C: Since x[n] = ejn/8 is an eigenfunction of an LTI system, we would expect the output to be given by

yin] = 'Ydn/8,

where 'Y is some complex constant, if System C were indeed LTI. The given output, yin) = 2ejn/8, indicates that this is so.

Hence, System C is LTI. However, it is not unique, since the only constraint is that

2.28. x[nJ is periodic with period N if x[n] = x[n + N] for some integer N. (a) x[nJ is periodic with period 5:

==> 27rk= 27r N, for integers k, N

5 .

Making k = 1 and N = 5 shows that x(n] has period 5.

(b) x(nJ is periodic with period 38. Since the sin function has period of 27r:

x[n + 38] = sin(7r(n + 38)/19) = sin(7rnj19 + 27r) = x(nJ

:'

=

.~

21

(c) This is not periodic because the linear term n is not periodic.

(d) This is again not periodic. ejw is periodic over period 21T, so we have to find k, N such that

x[n + N] = ej(n+N) = ej(n+2rrk)

Since we can make k and N integers at the same time, x[n] is not periodic.

2.29.

(a)

'I I I I I (

(b)

112

... T

"·123456··

11111..

I 234 5

'I I r"2

o

(c)

(d)

... ·0 12···~~

.. III II ....

-I 0 I 2

.. · · J · · · · ·

(e)

2.30. (a) Since cos(7rn) only takes on values of +1 or -1, this transformation outputs the current value of x[n] multiplied by either ±l. T(x[nJ) = (-l)nx[n].

• Hence, it is stable, because it doesn't change the magnitude of x[n] and hence satisfies boundedin/bounded-out stability.

• It is causal, because each output depends only on the current value of x[n].

• It is linear. Let yr[n] = T(xl[nJ) = cos(1Tn)xr[n], and Y2[n] = T(x2[nJ) = cos(1Tn)x2[n]. Now

T(axr[n] + bX2[nJ) = cos(1Tn)(axr[n] + bX2[nJ) = aydn] + by2[n]

• It is not time-invariant. Hy[n] = T(x[n]) = (-l)nx[n], then T(x[n -1]) = (-l)nx[n -1]:j:. y[n -1].

(b) This transformation simply "samples" x[n] at location which can be expressed as k2•

• The system is stable, since if x{n] is bounded, x[n2] is also bounded.

• It is not causal. For example, Tx[4] = x[16].

• It is linear. Let yr[n] = T(xr[n]) = xr[n2], and Y2[n] = T(x2[nJ) = x2[n2J. Now

T(axr[nJ + bx2[nJ) = axIfn2] + bX2[n2J) = aYl[n] + by2[n]

• It is not time-invariant. If y[n] = T(x[n)) = x[n2], then T(x[n - 1)) = x[n2 - 1] :j:. y[n - 1].

4Ki. ; Oct' .

22_

. = (c) First notice that

00

Lo[n - k] = urn]

k=O

So T(x[nJ) = x[n]u[nJ. This transformation is therefore stable, causal, linear, but n., ,

invariant.

To see that it is not time invariant, notice that T(o[n]) = o[nJ, but T(o[n + 1]) = O,

(d) T(x[nJ) = L~n-l x[k]

• This is not stable. For example, T(u[nJ) = 00 for all niL

• It is not causal, since it sums forward in time.

• It is linear, since

00

00

00

L axdk] + bxz[k] = a L xdk] + b L xz[k]

(

k=n-l

k=n-l

k=n-l

• It is time-invariant. Let

00

yin] = T(x[nJ) = L x[k],

k=n-l

then

00

T(x[n - no]) = L x[k] = yin - no]

k=n-no-l

( 2.31. (a) The homogeneous solution Yh[n] solves the difference equation when x[n] = O. It is in the

Yh[n] = L A{C)fi, where the c's solve the quadratic equation (

1 2

2+-c- - =0

15 5

So for c = 1/3 and c = -2/5, the general form for the homogeneous solution is:

Yh[n] = Al{!)n + A2(-~)fi

3 5

(b) We use the z-transform, and use different ROGs to generate the causal and anti-causal fmp

( ,

responses:

H{z) _ 1 = 5/11 + 6/11

- (1 - ~z-l)(1 + ~z-l) 1 - ~z-l 1 + ~z-l

5 1 6 2

hcfn] = U(3)nu[n] + il{-s)fiu[n]

hac[n] = -151 (~)fiu[-n - 1]- 161 (_~)nu[-n - 1]

(c) Since hcfn] is causal, and the two exponential bases in hc[n] are both less than 1, it is ali. It

summable. hac[n] grows without bounds as n approaches -00. .

(d)

Y(z) = X(z)H{z) 1

1

=

1 - ~Z-l . (1 - ~z-l)(1 + ~z-l)

-25/44 + 55/12 + 27/20

1 - 1/3z-1 1 + 2/5z-1 1 - 3/5z-1

-25(1)fi [] 55( 2)n [] 27(3)n [ ]

= -- un+--- un+-- un

44 3 12 5 20 5

=

y[n]

23

2.32. We first re-write the system function H(eiw):

Let yrfnJ = x(nJ * g(nJ, then

x[nJ

1Tn ej1fn/2 + e-j1fn/2

= cos(T) = 2

G(ej1f/2)ei1rn/2 + G(e-i1f'/2)e-j1fn/2

=

2

Evaluating the frequency response at w = ±1T /2:

(1 + -j1f + 4 -j21f)

= e-jf e e = 8e-i1f/2

1 + le-)1f

2

Therefore,

and

2.33. Since H(e-jW) = H*(eiw), we can apply the results of Example 2.13 from the text,

To find H(ei'*), we use the fact that H(elW) is periodic over 21T, so

Therefore,

31T 1T 21T 31T ll1T

yIn] = cos ( Tn + 4" + 3") = cos( Tn + 12 )

2.34. (a) Notice that

x[n] = xo[n - 2] + 2xo[n - 4] + xo[n - 6]

Since the system is LTI,

yIn] = yo[n - 2] + 2yo[n - 4] + yo[n - 6J,

and we get sequence shown here:

2

Pearson r r!"cation( s) P: I ig

Dell- • "necn

,,·22] .t'j

~. 24 ~.

(b) Since

yo[n] = -lxo[n + 1] + xo[n -1] = xo[n] * (-5[n + 1] + 5[n - 1]), h[nJ = -5[n + IJ + 5[n - 1J

2.35. (a) Notice that Xl [nJ = X2[nJ + X3[n + 4], so if TO is linear,

T{xdn]} = T{X2[nJ} + T{X3[n + 4]} = Y2[nJ + Y3[n + 4J

From Fig P2.4, the above equality is not true. Hence, the system is NOT LINEAR. (b) To find the impulse response of the system, we note that

o[nJ = x3[n + 4J

Therefore,

T{o[n]} = Y3[n-t4]

= 35[n + 6] + U[n + 5J

( c) Since the system is known to be time-invariant and not linear, we cannot use choices such a .

o[nJ = xdnJ - x2[n]

and

1

5[nJ = "2x2[n + IJ

to determine the impulse response. With the given information, we can only use shifted inputs. ';

2.36. (a) Suppose we form the impulse:

Since the system is linear,

A shifted impulse results when:

The response to the shifted impulse

Since,

L{J[n]} 1:- L{J[n -I]} The system is NOT TIME INVARIANT.

(b) An impulse may be formed:

since the system is linear,

L{o[nJ}

1 1

= "2Y1[n]- "2Y2[n] + Y3[nJ

= h[nJ

25

from the figure,

yr[nJ = -b[n + IJ + 3b[nJ + 3b[n - IJ + o[n - 3J Y2[nJ = -o[n + IJ + o[nJ - 30fn - IJ - o[n - 3J Y3[nJ = 2bfn + 2J + o[n + IJ - 3bfnJ + 2b[n - 2J

Combining:

h[n] = 2b[n + 2J + o[n + IJ - 2bfnJ + 3b[n - IJ +2b[n - 2J + o[n - 3J

3

2

2

-2

o I 2 n

2.37. For an LTI system, we use the convolution equation to obtain the output:

00

y[nJ = L x[n - kJh[kJ

k=-oo

Let n =m+N:

00

y[m + NJ = L x[m + N - kJh[kJ

k=-oo

00

= L x[(m - k) + NJhfkJ

k=-oo

Since x[nJ is periodic, x[n] = x[n + rN] for any integer r. Hence,

00

y[m + N] = L x[m - k1h[kJ

k=-oo

= y[mJ

So, the output must also be periodic with period N.

2.38. (a) The homogeneous solution to the second order difference equation,

3 1

y[n] - 4y[n - IJ + sy[n - 21 = 2x[n - 11,

is obtained by setting the input (forcing term) to zero.

y[n1- ~y[n - 1] + !y[n - 21 = 0

4 8

Solving,

3 -1 1_2

1- -z + -z = 0 48'

Psarsnn r rlL'~ation( s i {)" i I

~.s2, l' , F

Dell, -: ,OOC):':

. j ,

26

=

.~

( 1 -1 1 -1)

1 - ZZ )(1 - 4z = 0,

and the homogeneous solution takes the form

Yh[n] = Al(~)" + Az(~)n,

for the constants Al and Az.

(b) Substituting the intial conditions,

and

We have

2AI + 4A2 = 1 Al + Az = 0

. Solving,

Al = -1/2

and

A2 = 1/2.

( c) Homogeneous equation:

( .

1

y[n]- y[n - IJ + 4y[n - 2J = 0

Solving,

1

1 - Z-l + _z-z = 0

4 '

( 1 -1) 1 -1)

1 - ZZ (1 - ZZ = 0,

and the homogeneous solution takes the form

Yh[n] = Ad~)n.

Invoking the intial conditions, we have

Yh[-IJ = 2Al = 1 Yh[O] = Al = 0

Evident from the above contradiction, the initial conditions cannot be met. (d) The homogeneous difference equation:

1

y[n] - y[n - 1] + 4y[n - 2] = 0

Suppose the homogeneous solution is of the form

Yh[n] = Al(~)n + nBl(~)n, substituting into the difference equation:

1

Yh[n] - Yh[n - IJ + 4Yh[n - 2] = 0

Al(!)n + nBl(!)n - Al(!)n-l - (n -I)B1(!)n-l

2 2 2 2

IA (l)n_z I( )B (l)n-z 0

412 +4n-21Z =.

(____;__ .

27

(e) Using the solution from part (d):

and the initial conditions

and

we solve for Al and BI:

Al = 0 BI = -1/2.

2.39. (a) For xdnJ = J[nJ,

ydOJ = 1

yd1J = ay[OJ = a

For x2[n] = J[n - 1J,

Y2[0] = 1

yz[l] = ay[O] + x2[1] = a + 1 f:. ydO]

Even though x2[nJ = xdn - 1J, Y2[n] f:. Y2{n - 1J. Hence the system is NOT TIME INVARIANT. (b) A linear system has the property that

T{aXl[nJ + bxz(nJ} = aT{xtfn]} + bT{x2[n]}}

Hence, if the input is doubled, the output must also double at each value of n. Because y[O] = 1, always, the system is NOT LINEAR.

(c) Let X3 = axtfn] + .Bxz(nJ.

For n ~ 0:

Y3[n] = x3[n] + aY3[n - 1]

= aXI[n] + .Bx2[n] + a(x3[n - 1] + Y3[n - 2])

n-I n-I

= a L akxtfn - kJ +.B L akx2[n- k]

k=O k=O

= a(h[nJ * xI[nJ) + .B(h[nJ * x2[nJ) = aYl[n] + .BY2[n].

For n < 0:

Y3[nJ = a-I(Y3[n + 1J - x3[nJ)

n n

= -0 2: cikx1[n - kJ -.B L akxz[n - kJ

.1:=-1 .(:=-1

= ayl [n] + .Byz(nJ.

For n = 0:

Y3[n] = ytfn] = yz[nJ = O.

Conclude,

Y3[n] = aydn] + .BY2[n], for all n.

Therefore, the system is LINEAR. The system is still NOT TIME INVARIANT.

Pearson [[!ccation's) P1 , td

482, f' .• E Patp;!'I.> r. I Delb!~ 110092 ! lid fa

- . 22146067.70 Fitx2214i.()7

28

2.40. For the input

x[n] = cos(1rn)u[n] = (-ltu[nJ,

the output is

00
y[n] = 2: (j/2)ku[k](-1)(n-k)u[n - kJ t
k=-oo
¥
n
(-It 2:(j/2)k(_1)-k
k=O
n
= (-It 2:(_j/2)k
k=O
n C - (_j/2)(n+l))
= (-1) 1+j/2 For large n, (-j /2)(n+1) -t O. Thus, the steady-state response becomes

y[n] (_I)n
=
I + j/2
= cos(1rn)
1 + j/2'
2.41. The input sequence,
00
x[nj = L o[n + 16kj,
k=-oo has the Fourier representation

00 00

X(eiw) = L L o[n + 16kje-iwn

n=-ook=-oo

= ~ ~ o( 21rk).

16 L.t W + 16

k=-oo

Therefore, the frequency representation of the input is also a periodic impulse train. There are 16 frequency impulses in the range =t: :::; W ~ 1r.

We sketch the magnitudes of X(eiw) and H(ejW):

-61t -41t -21t 0 21t 41t 61t ro

16 16 16 16 16 16

From the sketch, we observe that the LTI system is a lowpass filter which removes all but three of the frequency impulses. To these, it multiplies a phase factor e-j3w•

The Fourier transform of the output is

1 1·~ 2rr

-o(w) + -e-) 16 o(w - -)

16 16 16

1·~ 2rr

+-eJI6o(w+ -)

16 16

Thus the output sequence is

1 1 2rrn 3rr

yIn] = 16 + 8 cos( 16 + "8)'

2.42. (a) From the figure,

yIn] = (x[n] + x[n] * hl[n]) * h2[n]

= (x[n] * (o[n] + hdn]) * h2[n].

Let h[n] be the impulse response of the overall system,

yIn] = x[n] * h[n].

Comparing with the above expression,

h{n] = (0 In] + hdn]) * h2[n] = hdn] + hdn] * h2[n]

= anu[n] + tJ(n-l)u{n - 1].

(b) Taking the Fourier transform of h[n] from part (a),

00

H(dW) = L h[n]e-j",n

n=-oo

00 00

= L anu[n]e-:j"'n + tJ L a(n-I)u[n - l]e-j",n

n=-oo

n=-oo

00 00

= L ane-j",n + tJ La(l-l)e-j"'l,

n=O l=O

where we have used f. = (n - 1) in the second sum.

1 tJe-j'"

----:.-+ .

1 - ae=r« 1 - ae-]W

1 + tJe-j'"

_";__-.:-, for lal < 1. 1- oe=>

H(d"') =

=

Note that the Fourier transform of anu[n] is well known, and the second term of h[n] (see part (a» is just a scaled and shifted version of anu[n]. So, we could have used the properties of the Fourier transform to reduce the algebra.

(c) We have

H(d"')

=

1 + tJe-jw l-ae-jw'

DdL, < 100S2

... 22146067·70 .bA.2~t .. " '

30

cross multiplying,

Y(eiW)[l - oe-iwJ = X(eiw)[l + ,6e-jwJ taking the inverse Fourier transform, we have

y[nJ - oy[n - 1J = x[nJ + ,6x[n - 1J.

(d) From part (a):

h[nJ = 0, for n < O.

This implies that the system is CAUSAL.

, IF If the system is stable, its Fourier transform exists. Therefore, the condition for stability is 'ttfe same as the condition imposed on the frequency response of part (b). That is, STABLE, if 101 4j_.

2.43. For (-1 < a < 0), we have

(a) real part of X(eiw):

=

1 - 2a cos(w) + a2

1-acos(w)

( .

(b) imaginary part:

x, (eiW) = 2~' [X(eiw} - X·(eiw}] -asin(w)

=

1 - 2acos(w} + a2

("t

(c) magnitude:

.l

= C _ 2aCO~(w) +a2) 2

(d) phase:

" .

LX(eiw) = arctan ( -asin(w) ) 1 - acos(w}

2.44. (a)

00

n=-oo

00

= L x[n]

n=-oo

= 6

(b)

00

X(e3W}lw=,.. = L x[n]e-i,..n

n=-oo

00

= L x[nJ(-l)n

n=-oo

= 2

-

31

(c) Because x[nj is symmetric about n = 2 this signal has linear phase.

A(w) is a zero phase (real) function of w. Hence,

LX(eiw) = -2w, -1[ ::; W ::; 1[

(d)

for n = 0:

/~ X(eiW)dw = 27rX[Oj = 411" (e) Let y[nj be the unknown sequence. T~en

Y(ei"') = X(e-iw)

= Lx[njeiwn

n

n

= Ly[nje-i",n

n

;'

Hence y[n] = x[-n].

... .

r -4 0

-I

(f) We have determined that:

XR(ei"') = Re{X(ei"')} = A(w) cos(2w)

= ~A(w) (ei2'" + e-i2",) 2

Taking the inverse transform, we have

1 1 1 1

2a[n + 2J + 2a[n - 2J = 2x[n + 4J + 2x[nJ

112 T I

.... • T _ •

. I fT.

1

. I

4

o

.--!L 32

2.45. Let x[n] = b[n], then

The output of the ideallowpass filter:

The multiplier:

causes a shift in the frequency domain:

The overall output:

yin] = e-i1rnw[n] + win] Y(dW) = H(ei(w-1f» + H(dW)

Noting that:

Y(eJW) = I, thus yin] = b[n].

2.46. (a) We first perform a partial-fraction expansion of X(eiw):

X(dW) =

(I - ae-iw)(1 - aeiw)

1 aejw

---,-. + .

1 - ae-JW 1 - aeJW

= anu[n] + a-nu[-n - 1]

= a1nl

=

x[n]

(b)

lr

271" J-1r X(dW)cos(w)dw

= .2_ r X(dW) eJw + e-jw dw

271" J-7r 2

= ~ (.2_ r X(dW)eiwdw +.2_ r X(dW)e-jWdw)

2 271" J-1r 271" J-7r

1

= 2(x[n - 1] + x[n + 1])

= ~(aln-ll + a1n+ll)

2

2.47. (a)

y[nJ = x[nJ + 2x[n - IJ + x[n - 2J = x[nJ * h[nJ

= x[nJ * (o[n] + 2b[n - 1] + o[n - 2]) h[n] = o[n] + 2o[n - 1] + o[n - 2]

(b) Yes. h[nJ is finite-length and absolutely summable.

33

(c)

H(ejW) = 1 + 2e-jw + e-2jw ·1·1·

= 2e-;W( _e;w + 1 + _e-;W)

2 2

= 2e-iw(cos(w) + 1)

(d)

IH(eiW)1 = 2(cos(w) + 1) LH(eiw) = -w

Magnitude

P ase

1t 0)

4

-1t

-1t

1t W

(e)

h1[nJ = .2...1 Hl(eiW)eiwdw
211" <2,,> .
= .2...1 H(ei(W+")dwndw
211" <2><>
= .2...1 H(d(W)ej(w-,,)ndw
211" <2,,>
= e-jlrn.2...1 H(ei(W)dwndw
211" <2,,>
= -lnh[nJ
= o[nJ - 2o[n - IJ + o[n - 2J . 2.48. (a) Notice that

s[nJ = 1 + cosixn) = 1 + (_l)n S(eiW) = 211" I: o(w - k1l")

k

(b) Since y[n] = x[nJs[nJ,

Pearson r"'~r:;ltion(s) Pt· tflt 482, fi.i 1: Patparg.i.. 1.

[; elb . -, ; C'092 l."·

~ 22J46067,7U f.; :'

34

=

=

= 2_ j1r S(ei8)X(ei(w-8»dw

27r -1r

= X(eiw) + X(ej(w-1r»

Y (ejW) contains copies of X (ejW) replicated at intervals of tt . (c) Since w[n] = y[n] + (1/2)(y[n + 1J + y[n - 1]),

Y(elW) + ~ (eiwY(ejw) + e-jwY(ejW») 2

= Y(ejW)(l + cos(w»

(d) The following figure shows X(ejW), Y(ejW), and W(ei"') for a < 2 and a> 2. Notice that

X(ejW) = {I, Iwl S n]«,

·0, 7r/a S Iwl 2: tt

So, for a > 2, y(&W) contains two non-overlapping replications of X(eiw), whereas for a <. "aliasing" occurs. When there is aliasing, W(eiW) is not at all close to X (el"'). Hence, a mu~ti: greater than 2 for w[nJ to be "close" to x[nJ.

X. Y. and W for a<2

2r---~------------~---'

1.5 r~1 !:S

0.5

-2

o til

2

2.5

1.Sf-

-2

o Cll

2

6r---~----~------~--~

-2

o til

2.49. (a) We start by interpreting each clue. (i) The system is causal implies

2

X. Y. and W for a> 2

6r---~----~----------~

-2

o w

2

6r---~----------------~

10

-2

o w

2

-2

o w

2

h[nJ = 0 for n S O.

(ii) The Fourier transform is conjugate symmetric implies h[nJ is real. (iii) The DTFT of the sequence h[n + IJ is real implies h[n + 1J is even.

From the above observations, we deduce that hfnJ has length 3, therefore it has finite duration

~

-~-

35

( "i

(b) From part (a) we know that h[nJ is length 3 with even symmetry around h[lJ. Let h[OJ = h[2] = a and h[l] = b, from (iv) and using Parseval's theorem, we have

2a2 + b2 = 2.

(

(

From (v), we also have
2a - b = O.
Solving the above equations, we get
h[O] 1
= V3
h[l) 2
= V3
h[2J 1
= V3
,or
h[O] 1
= -V3
h[l] 2
= -V3
h[2J 1
= - V3' (

(

(

(

(
j 2.50. (a) Carrying out the convolution sum, we get the following sequence q(n):
,
I
4 4
3 3 3 3
(
( q[nJ
1 1 1 1
(
;- n
I
0 1 2 3 4 5 6 7 8 9 10
( (b) Again carrying out the convolution sum, we get the following sequence r[n): (
4 4 4 4 4 4 r[nJ
I 3 3
1 1
n
( 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
-4
( -8
-12
-16
-20 Pearson r,'

482, r !~ ~"I);1

;-- ~ .(\: _! _-

36

(c) Let a[n] = v[-nLand bin] = w[-nJ, then:

+(XJ
a[n] * bin] = I: a[kJb[n - k]
k=-(XJ
+(XJ
= I: v[-k]w[k - n]
k=-oo
+00
= I: v[r]w[-n - rJ where r = -k
r=-oo
= q[-n]. We thus conclude that q[-n] = v[-nJ * w[-nJ. 2.51. For (-1 < a < 0), we have

X( "') _ 1

e - .

1 - ae=s-'

(a) real part of X(ejW):

•

=

1- acos(w)

1 - 2acos(w) + a2

(b) imaginary part:

=

-asin{w)

1 - 2acos(w) + a2

(c) magnitude:

1

= C _ 2aco~(w) + a2 ) 2

(d) phase:

. ( -asin(w) )

LX(eJW) = arctan

1 - acos{w)

2.52. x[nJ can be rewritten as:

x[nJ

ej "2ft. e " j 'ff2"

= --+--.

2 2

We now use the fact that complex exponentials are eigenfunctions of LTl systems, we get:

37

yin]

2.53. First x[nJ goes through a lowpass filter with cutoff frequency O.57r. Since the cosine has a frequency of O.67r, it will be filtered out. The delayed impulse will be filtered to a delayed sine and the constant will remain unchanged. We thus get:

[ J = 3sin{O.57r(n - 5» 2

wn () +.

7rn-5

yin] is then given by:

[] 3sin{O.57r{n - 5» sin{O.57r(n - 6»

yn = -3 .

7r{n -5) 7r(n - 6)

2.54.

x[n] 157rn 7r
= cos(-- --)
4 3
7rn 7r
= cos(-- --)
4 3
7rn 7r
= cos{- +-)
4 3
... ~ e-iiciT
eJ"feJ•
= +
2 2 Using the fact that complex exponentials are eigenfunctions of LTI systems, we get:

yin]

=

." ei*ei".n e-i*e-i"."

= e-J4{ 2 + 2 )

." 7rn 57r

= e-J4 cos("4 + 24)'

2.55. Since system 1 is memoryless, it is time invariant. The input, x[n] is periodic in w, therefore win] will

. also be periodic in w. As a consequence, yin] is periodic in w and so is A.

2.56. (a)

yin] = h[n] * (e-iwonx[nj) +00

= L e-iwOkx[kJh[n - kJ.

k=-oo

38

.-=-

',~Let x[n] = axdn] + bX2[n], then:

y[n] = h[n] * (e-jwon(axdn] + bX2[nJ))
+=
= L e-JwOk(axdk] + bX2[k])h[n - k]
k=-=
+= +00
= aL e-jwokxdk]h[n - k] + b L e-jwokx2[kjh[n - k]
k=-oo k=-oo
= aYdn] + by2[n] where yr[n] and yz[n] are the responses to xdn] and x2[n] respectively. We thus conclude that. system S is linear.

(b) Let xz[nj = x[n - no], then:

+co

= L e-jwo{n-k)x2[n - k]h[k]

k=-oo

+00

= L e-jwo(n-k)x(n - no - k]h[k]

k=-oo

::f. y[n - no].

We thus conclude that system S is not time invariant.

(c) Since the magnitude of e-jwon is always bounded by 1 and h[n] is stable, a bounded input x[n] will always produce a bounded input to the stable LTI system and therefore the output y[n] wil. be bounded. We thus conclude that system S is stable.

(d) We can rewrite y[n] as:

y[n] = h[n] * (e-jwonx[n])

+00

= L e-jwo(n-k)x[n - k]h[k]

k=-oo

+00

= L e-jwoneiwokx[n - k]h[k]

k=-oo

+00

= e-jwon L eiwokx[n - k]h[k].

k=-oo

System C should therefore be a multiplication by e-jwon.

2.57. (a) H1(ejW) corresponds to a frequency shifted version of H(ejW), specifically:

We thus have:

.,~

39

This is a high pass filter.

-

1

~--~--------~------~--~~~w

-1T -0.81T

o

0.81T 1T

(b) H2(eiw) corresponds to a frequency modulated version of H(eiw), specifically:

Hz(eiw) = H(dW) * (o(w - 0.51T) + o(w + 0.51T»

where Iwl ~ 1T.

We thus have:

{ 0 , Iwl < 0.31T Hz(dW) = 01, 0.31T s Iwl ~ O.h , O.h < Iwl ~ 1T.

This is a bandpass filter.

H2(eiw)
1
w

-1T -0.71T -0.31T 0 0.31T O. Tt: 1T

(c) H3(eiw) corresponds to a periodic convolution of Hlp (eiw) with another lowpass filter, specifically:

where H(ejW) is given by:

H(dW) = {01 , Iwl < 0.l1r

, 0.l1r ~ Iwl ~ 1T

Carrying out the convolution, we get:

{ 0.1 , Iwl < O.11r

H3(dW) = -~ + 0.15 , o.i» s Iwl s 0.31T o , O.31T < Iwl ~ 1T.

Pearson r ~:f.t!'Jn(~} Pt /t{! 48]:.} F I'

.'

l./,,_i~ -

I)

r--r--\ 0.1

2.58. Note that X(ejW) is real, and Y(ejW) is given by:

, O<w<1r

, -1r < W < O.

w[n] = x[n] + jy[nJ, therefore:

Using the above, we get:

Y( jW) _ { X(ejW)

J e - -X(e1W)

, O<w<1r

, -1r < W < O.

We thus conclude:

, O<w<1r +« < W < O.

-1r

1r

-+------------~--~------~~ W

o

2.59. (a) Using the change of variable: r = -k, we can rewrite R:,[n] as:

00

R:,[nJ = L x'[-rJx[n - rJ = x*[-nJ * xfnJ.

r=-oo

We therefore have:

gfnJ = x*[-nJ. (b) The Fourier transform of x* [-nJ is X" (ejW), therefore:

R:,(eiW) = X"(dW)X(dW) = IX(eiwW·

2.60. (a) Note that x2fnJ = - L!~~ xfn - kJ. Since the system is LTI, we have:

k=4

Y2fn] = - Ly[n - kJ.

k=O

41

(b) By carrying out the convolution, we get:

{ -1

h[nJ = -2

o ,

, n = O,n = 2 n=1

o.w.

2.61. The system is not stable, any bounded input that excites the zero input response will result in an unbounded output.

The solution to the difference equation is given by:

y[nJ = YZir[nJ + Yzsr[nJ

where Yzir[n] is the zero input response and Yzsr[nJ is the zero state response, the response to zero initial conditions:

YZir[n]

-_ a(_l)n h al d

2 were a is a constant determined by the initi con ition.

1

= (2" )nu[n] * x[n].

Yzsr[n]

An example of a bounded input that results in an unbounded output is:

x[n] = o[n + 1].

The output is unbounded and given by:

1 1 1

y[n] = (- )n+1u[n + 1]- _( _)n

2 2 2 .

2.62. The definition of causality implies that the output of a causal LTI system may only be derived from past and present inputs.

The convolution sum:

00

y[n] = L h[k]x[n - k]

k=-oo

-1 00

= L h[k]x[n - k] + L h[k]x[n - k]

k=-oo

k=O

Note that the first summation represents a weighted sum of future values of the input. Thus, if the system is causal,

-1

L h[k]x[n - k] = O.

k=-oo

This can only be guaranteed if h[k] = 0 for n < O.

Using reverse logic, we can show that if h[n] = 0 for n < 0,

00

y[n] = L h[k]x[n - k].

k=O

Since the convolution sum specifies that the input is formed from past and present input values, the system is, by definition, causal.

Pearson fdur,:aticn(s) P11 t~.

~, .

---,t ~1

42

2.63. The system could be LTI. A possible impulse response is:

h[n)

2.64. Let the input be x(n) = <s[n -1), if the system is causal then the output, y[nJ, should be zero for n < '.J

Let's evaluate y[O): < -t

, .

y[O) = 2_ /_+tr Y(eiw) dw
211" -7r
= 2_ t: e-jw e-jw/2 dw
211" _ ..
2
=
311"
i O. This proves that the system is not causal.

2.65. xdn] is even-symmetric around n = 1.5, furthermore since LX1[n] < 0 and we want A1(0) ~ 0, we

need to include a 11" factor in the phase. An appropriate choice for 81 (w) is therefore: \1'

7r

"2

w

o

-7r

x2[n] is odd-symmetric around n = 3, therefore:

tt

82(w) = -3w + "2 Iwl < n,

- :~

43

1'-' •

82(w)

1T

-1T

2.66. (a)

E(dW) = HI (dW)X(eiw)
F(eiw) = E(e-iw)
= Hl(e-iw)X(e-iw)
G(eiw) = HI (dW)F(dW)
= HI (eiw)Hl (e-iw)X (e-iw)
Y(eiw) = G(e-iw)
= HI (e-iw)Hl (dw)X(eiw). I (b) Since:

We get:

(c) Taking the inverse transform of H(eiw), we get:

h[n] = hI [-n] * hI [n].

2.67. (a) Using the properties of the Fourier transform and the fact that (-l)n = ej7rn, we get:

V(dW) = X(d(W+K»)
W(dW) = Hl(dW)V(dW)
= HICdW)X(d(w+7r)
Y(dW) = W(d(w-><»)
= HI (d(w-7r»)XCdW) H(eiw) is thus given by:

(b)

402, r _F P t:,

(

44

H(eiW)=={ 0 , Iwl<1f-We

1 , tt - We < Iwi ~ 7r.

H(eiw)
1
w

-1f -1f + We

o

1f - We

1f

2.68. If xtfn] = x2[n], wdn] and w2[n] will not be necessarily equal.

( ,

wdn] =, xIf-n - 2] w2[n] = xz[-n + 2] f:. xz[-n - 2]

A simple counterexample is xdn] = x2[n] = o[n]. Then:

wl[n] = o[n + 2] w2[n] = o[n - 2].

2.69. (a) The overall system is not guaranteed, to be an LTI system. A simple counterexample is:

Yl[n] = x[n] Y2[n] = x[n]

y[n] = Yl[n]Y2[n] = x2[n]

which is not a linear system, therefore the system is not LTI.

(b)

Yi(eiW) = HI (eiw)X(eiw) Y2(eiW) = H2(dW)X(ejW) Y(eiW) = Yi (ejW) d'2(dW).

Using the above relationships, ~e get:

Y(eiW) = {unspecifiedo ' 0 < Iwl < 0.611" , 0.61f $ Iwl $ 1f.

2.70. The first difference:

y[n] = V'(x[nJ) = x[n]- x[n - 1].

=

.~

=. =

~

45

(a) To determine if the system is linear:

\i'(axdn] + bX2[nJ) = axdnJ + bX2[nJ - axdn - 1] - bX2[n - 1J = a(xdnJ - xdn - 1]) + b(X2[nJ - X2[n - 1]) = \i'(axdn]) + \i'(aX2[n]).

Therefore, the system is LINEAR.

To determine if the first difference is time invariant:

\i'(xdn - 1]) = x[n - 1J - x[n - 2J = y[n - 1J.

The system is TIME INVARIANT.

(b) The impulse response is obtained by setting the input to x[n] = <S[n],

, yin] = h[n] = o[n] - orn - 1J

(c) Taking the Fourier transform of the result of part (b), we find that the system function is H(ejW) = 1- «>.

Thus the magnitude of the frequency response is

IH(e1W)1 = J(1- e-jw)(1 - e-jw) = J2 - 2cos(w).

--

We calculate the phase of the frequency response:

H(e1W) = (1 - cos(w» + j sinew)

Thus,

H(_1W) ( sinew) )

L e:- = arctan

. 1 - cos(w)

':;82. f !' l"T:i"

r',,: .. . ," .

~~ ,':)'z..

46

rr

"2

------~------+_----~------~ w

o 7r

(d) In general,

\7(x(nJ) = x(nJ * (J(nJ - J[n - 1]) = x[nJ - x[n - 1J.

So, for x[n] = f[nj * g[nJ,

\7(x[nJ) = f[n] * g(nJ * (J(nj - J[n - 1]) = f(nj * \7(g(nJ)

= \7(J(nJ) * g(nJ.

Where we have used the commutivity of the convolution operator to obtain the last two equalities. (e) We desire the inverse system, hi[n], such that

hi[n] * \7(x[nJ) = x[nj

The inverse system must satisfy:

hdnj * h[nj = J[nj,

in the frequency domain,

Recall from part (c),

So,

. 1

H-(e'I,,) - --~

• - 1 - er>:'

and

Hence, the unit step is the inverse system for the first difference.

2.71. For impulse response h[n], the frequency response of an LTI system is given by

00

H(eil,,) = L h[nje-jwn

n=-oo

(a) Suppose the impulse response is h·[nj,

n=-co

_,.JI

(b) We have

00

n=-oo

If h(n] is real,

n=::-oo

Hence, the frequency response is conjugate symmetric.

2.72. The analysis equation for the Fourier transform:

00

X(eiW) = L x[n]e-iwn

n=-oo

(a) The Fourier transform of x·(n],

n=-oo

00

L x·(nJe-jwn =

(b) The Fourier transform of x·(-n],

00

L x·(-n]e-jwn =

n=-oo

00

L x·[l]eiwl

1=-00

(too X[l]e-iW1) •

= x·(dW).

=

2.73. From property 1:

00

X·(e-jW) = L x·[n]e-jwn

n=-oo

for x(nJ real, x(nJ = x*(nJ, so

00

X*(e-iw) = L x(nJe-jwn

n=-oo

= X(dW).

Thus, the Fourier transform of a real input is conjugate symmetric.

x(d"') = XR(ejW) + jx/(dW) X·(e-iw) = XR(e-iw) - jX/(eiW)

Pearson Er'ncrttion(s) Pt, J t!l1.

48]. f (I ;;','

Ccil:J:""

<9}~ 2:~14c·067-}OhF.:~',·

= 48 ~

From property 7, X(eiw) = X'(e-jW) for x[nJ reaLThus,

XR(eiw) + jX/(elW) = XR(e-iw) - jX/(e-iw).

We may infer

From property 7:

So,

2.74. Theorem 1:

Theorem 2:

Theorem 3:

Theorem 4:

Theorem 5:

property 8: XR(ejW) = XR(e-;W) property 9: X/(ej",) = -X/(e-;W)

X(eiw) = IX(ej"')lejLX{ej~) X·(e-iw) = IX(e-jW)le-jLX{e-)~)

property 10: IX(ejW)1 = IX(e-jW)1 property 11: LX(eiW) = -LX(e-jW).

00

L (axr[nJ + bX2[n])e-jwn =

00

L axdnJe-jwn +

00

L bX2[n]e-jwn

n=-oo

n::-oo

n=-O(;

00 00
L x[n - nd]e-jwn = L x[e]e-iw(l-n,d
n=-O() (=-00
00
= elwnd L x[e]e-jwl
(=-00
= elwndX(elW) 00

L x[n]elwone-jwn =

00

L x[n]e-j(w-wo)n

n:-O()

n=-oo

00

L x[_n]e-iwn =

n=-oo

(=-00

00

2: nx[n]e-iwn

n=-oo

49

2.75. The output of an LTl system is obtained by the convolution sum,

00

yIn) = L x[kJh[n - kJ.

k=-oo

Taking the Fourier transform,

Y(ejW) = nJ;oo C~oo x[kJh[n - k)) «>: = kfoo x[kJ C%;oo'h[n - kJe-jWn)

= k~OO x[k)e-jwk C~oo h{n - k)e-jW(n-k))

Hence,

Y(elW) = X(elW)H(elW).

2.76. The Modulation theorem:

the time-domain representation,

yin)

= ~ r dO r dwX(d8)W(el(w-8)elwn

(27r)}_1f J-1f

= 2. r dO X (el8)w[n)ei8n 211' J-1C

= x [n]w[nJ

2.77. (a) The Fourier transform of y·[-n] is Y·(ejW), and X(eiw)Y(ejW) forms a transform pair with x[n] * yin]. So

G(elW) = X(elW)Y·(elW)

and

gin] = x[n) * y·[-nJ

form a transform pair.

(b)

00

L (x[n] * y·[-nJ) e-jwn

n=-oo

=

00 00

L L x[kJy·[k - n]e-jwn

n=-ook=-oo

for n = 0:

50

(e) Using the result from part (b):

x[nJ

sin(11"nj4) 211"n sin(mj6)

=

y'[nJ

511"n

We recognize each sequence to be a pulse in the frequency domain:

,l

1 2"

-+------~~--~--~------~~ w

-11"

" 0 "
-"4 "4
Y(eiw)
1
5
" 0 "
-6" 6" 11"

, t

-11"

Substituting into Eq. (P2.77-1):

I'}:

00

= 2_ r X(dw)Y*(eiw)dw 211" J-"

= 2_ [( ~ ) ( ~ ) ( 211" )]

211" 2 5 6

1

60

L x[nJy*[nJ

n=-oo

=

2.78. X(eiw) is given by:

(a)

ys[nJ

= {x[nJ, n even

0, n odd

= ~ (1 + d"n) x[nJ

2.79. (a)

51

which transforms to

(b)

Yd[n] = x[2n]

Yd(ejW) = ~ [X(ej'f) + X(ej('i-+7r))] = Ys(ej'f)

(c)

[n] = {X[n/2], n even

Ye 0 n odd

,

00

~%(-N, -w) = L x[n - N]x*[n + N]e1wn•

n=-oo

00

~;(N,w) = (L x[n + N]x·[n - N]e-jwn)*

n=-oo

Pearson r ,.J!'tati'm( $) Pt Ud. 482, }., E Paipc 1 F-. t :

Delh 1< 10092 ; nc .. ... ~ , r)f') J46067 7n t., ,. ","

~~._ 1 v, .l~, ''''

00

= L (x[n FNJx'[n - NJe-jwn)*

n::::-oo

00

= L x'[n + NJx[n - NJeiwn

n=--oo

= 4?z(-N, -w).

(b)

00

<P:t(N,w) = I: Aan+Nu[n + NJAan-Nu[n - NJe-jwn

n=-oo

00

= A2 L a2ne-jwn

n=N

( <'.

00

= A2 L (a2e-jw)n

n=N

A2 (a2e-jw)N 1 - a2e-jw a2N e-jwN

= A2 ..

1 - a2e-)W

(c)

00

X(d(v+(w/2») = L x[n]e-j(v+w/2)n.

n=-oo

00

X*(d(v-(w/2») = L x*[n]ei(V-w/2)n.

n=-oo

Let S = .1.. f1r X(ei(v+(w/2»)X·(ei(v-(W/2»)ei2vN dv then'

2". -,,- , .

S =

1 11f 00 00

- L x[n]e-i(v+w/2)n L x·[k]ei(v-w/2)keJ'2vN dv

21r _,.. .

n=-oo k=-oo

1 00 00 1"-

21r L L x[n]x·[k]e-iw(ntk) ei(k-n+2N) dv

n=-oo k=-oo -,,-

=

= ~ ~ ~ x[n]x.[k]e-iw(nr) 2sin(1r(k - n + 2N»

21r ~ ~ k - n + 2N

n=-ook=-oo

00

= L x[n]x*[n - 2N]e-iW(2"i2N)

n=-oo

00

= L x[n + N]x·[n - N]e-iwn

n=-oo

= q,z(N,w).

We thus conclude that:

q,:t(N,w) = 2~ j_: X(ei(v+(w/2»)X·(eJ'(V-(w/2»))eJ'2vN dv.

53

2.80.

w[nJ = x[nJ + y[nJ

The mean of w[nJ:

mw = E{w[nJ}

= E{x[n] + y[nJ}

= E{x[nJ} + E{y[n]}

= mx +my

The variance of w[nJ:

a~ = E{(w[n] - mw)2} = E{w2[nJ} - m~

= E{(x[nJ + y[nW} - m~

= E{x2[n]} + 2E{x[n]y[n]} + E{y2[n]} - m; - 2mxmll - m;

If x[n] and y[n] are uncorrelated:

a~ = E{x2[nJ} - m; + E{y2[n]} - m;

= a2 + a2

x "

2.81. Let ern] be a white noise sequence and E{s[n]e[mJ} = 0 for all n and m.

E{y[n]y[n + m]} = E{ s[n]e[n]s[n + m]e[n + mJ} = E{ s[n]s[n + m]e[n]e[n + mJ}

Since s[n] is uncorrelated with ern]:

E{y[n]y[n + m]} = E{ s[n]s[n + m]}E{ e[n]e[n + m]} = a;a;o[m]

2.82. (a)

<p:x[m] = E(x[nJx[n + m))

= E«s[n] + e[n]) (s[n + m] + ern + m)))

= E(s[n]s[n + m]) + E(e[n]e[n + m)) + E(s[n]e[n + m)) + E(e[n]s[n + m))

= <pss[m] + <Pee[m] + 2E(e[n))E(s[nJ) since s[n] and ern] are independent and stationary. = <pu[m] + <pu[m] where we assumed ern] has zero mean.

Taking the Fourier transform of the above equation, we get:

(b)

<P%e[m] = E(x[nJe[n + mJ)

= E«s[nJ + e[n])e[n + mJ)

= E(s[nJ)E(e[nJ) + <Pee[m] since s[n] and ern] are independent and stationary. = <pu[m] where we assumed ern] has zero mean.

Taking the Fourier transform of the above equation, we get:

Pearson F ducatlonl sl Pt , tt1 4S2, r: E Pa·pil~:. r: o.n.. -1100Ci2 i nd [;1 ... 22146067.]0 fax.2~:146U-·

54 ...

(c)

<pu[m] = E(x[n]s[n + mJ)

= E«s[n] + e[nJ)s[n + mJ)

= <pu[mJ + E(e[n])E(s[nJ) since s[n] and ern] are independent and stationary.

= <pss[m]

where we assumed ern] has zero mean.

Taking the Fourier transform of the above equation, we get: q,,,s(eiw) = q,ss(ejW).

,T

2.83. (Throughout this problem, we will assume lal < 1.)

(a)

<Phh[m] = h[m] * h[-m].

Taking the Fourier transform, we get:

q,hh(dW) = H(dW)H(e-jW)

1 1

=

(1- ae-jW) (1- aeiw)

1 1 1

= --z( . + . ).

1 - a 1 - ae-lW 1 - aer«

Taking the Inverse Fourier transform, we get:

alnl <Phh(m] = -1 -z·

-a

(b) Using part (a), we get:

IH(dWW = H(dW)H*(dW)

= H(eiW)H(e-iw) since h(n] is real = q,hh(dW)

1 1

=

(1 - ae-jW) (1 - aeiw)

1 1 1

= 1 - aZ (1 - ae-iw + 1 - aeiw)'

(c) Using Parse val's theorem:

+00

L Ih(nW

n=-oo

=

n=-(X)

n=O

1

=

55

/ ~"

2.84. The first-backward-difference system is given by:

yIn] = x[n] - x[n - 1].

(a)

q)yy[m] = E(y[n]y[n + mJ)

= E«x[n] - x[n - 1]) (x[n + mJ - x[n + m - 1])

= E(x[nJx[n + m)) - E(x[n]x[n + m - 1]) - E(x[n - l]x[n + mJ)

+E(x[n - l]x[n + m - 1])

= q)xx[mJ- q)x:r[m - 1J - q):rz[m + 1) + q):rx[m) = 2q)xx[mJ - q):r:r[m - 1J - q):rx[m + 1J

= 2u;6[mJ - u;6[m - 1J - u;6{m + 1J.

(

To get the power spectrum, we take the Fourier transform of the autocorrelation function:

= 2u2 _ u2e-jw - u2eiw

x x x

== 2u; - 2u;cos(w)

= 2u;(1- cos{w».

(b) The average power of the output of the system is given by q)vv[O):

(c) The noise power increased by going through the first-backward-difference system. This tells us that the first backward difference amplifies the noise of a signal.

Pearson Educatlonl s) Pt . Ltd,

I

56

2.85. (a)

00

E{x(nJy(n)} = E{x[nJ L h(nJx(n - kJ}

k=-oo

00

= L h(kJE{x(nJx[n - kJ}

k=-oo

00

= L h(kJ¢:rx(kJ

k=-oo

Because x(nJ is a real, stationary white noise process:

Therefore,

00

E{x(nJy[nJ} = (/; L h(kJo(kJ

k=-oo

(b) The variance of the output:

(/; = E{(y(nJ - my)2} = E{y2[n]} - m;.

When a zero-mean random process is input to a determistic LTI system, the output is also l mean: 1:

y[nJ = x[nJ * h[nJ

00

= L x[kJh[n - kJ.

k=-oo

Taking the expected value of both sides:

00

my = L E{x[n]}h[n - kJ

k=-oo

my = 0, if m", = O.

So,

(/2 = E{y2[n])
y
= E {m~oo h[m]x[n - m] kt;oo h[kJx[n - kJ}
00 00
= L L h[mJh[k]E{x[n - m]x[n - k]}
m=-ook=-oo
00 00
= (/2 L L h[mJh[kJ8[m - k]
'"
m=-ook=-oo
00
(/2 = (/2 L h2[mJ.
y '"
m=-oo ~r _

2.86. Using the solution to problem 2.85: (a)

00

<T; = <T; 2: hi[k]

k=-oo

this statement is TRUE, because x[n] is a white noise sequence. (b) Since y[n] is not a white noise sequence, this statement is FALSE. (c) Let

These systems are cascaded:

n

= Lakbn-k, n 2: 0 k=O

= bn (1 - (a/b)n+1) urn] 1 - (a/b)

w[n] = x[n] * h[n].

Since x[nJ is zero-mean, mw = 0 also.

.,

00

= <T; I: h2[k].

k=O

2.87. (a) x[n] is a stationary white noise process.

00

y[n] = L h[k]x[n - k], n 2: 0

k=-oo

E{y[n]} = E {ktoo h[kJx[n - k] } , n2:0
n
= L h[k]E{ x[n - k]}
k=-oo
{ m", 2::;=-00 h[k], n2:0
= 0, n<O (b)

~Y!l [nl' n2] = E{y[nl]y[n2]} .

= E Ltoo h[k]x[nl - k] m~oo h[m]x[n2 - m] } , n 2: 0

nl n2

= L L h[kJh[m]E{x[nl - k]x[n2 - m]}

k=-oom=-oo

nl n2

= L L h[k]h[mJ~:r:r[nl - k, n2 - m].

k=-oom=-oo

Pearson Education! sl Pt ~ tut

482, F.J E Patp;,;"-; ~ De.U'",~, ~

58

(c)

n

lim mr ~ h[k]

n-tcx) L__;

00

mx 2: h[k]

k=-oo k=-oo

my.

nl n2 00 co

nl'~~oo 2: L h[k]h[m]¢r:r[nl - k, nz - m] = L 2: h[k]h[m]¢:r:r[k, m].

k=-oo m=-oo k=-oo m=-oo

(d)

00
E{y[n]} = m., L anu[n]
k=-oo
m",
=
I-a !

\ ,

2.88. (a) No, the system is not linear. In the expression of yin], we have nonlinear terms such as xZ[n] alid divisions by x[n], x[n - 1] and x[n + 1].

(b) Yes, the system is shift invariant. If we shift the input by no, mr[n] shifts by no as well as O';[n]

and O';[n], therefore yin] shifts by no and the system is thus shift invariant. \

(c) If x[n] is bounded, m.,[n] is bounded so is O';[n] and 0'; [n]. As a result, y[n] is bounded and

therefore the system is stable. I}

(d) No, the system is not causal. Values of the output at time n depend on values of the input at t~.e n + 1 (through o-;[n] and m.,[nJ). Since present values of the ouput depend of future values of trk input, the system cannot be causal.

(e) When O'![n] is very large, O';[n] is zero, therefore:

y[n] = m.,[n]

1 n+1

= 3 L x[k]

k=n-l

which is the average of the previous, present and next value of the input.

When O'![n] is very small (approximately zero), then:

y[n] = x[n].

yin] makes sense for these extreme cases, because in very small noise power, the ouput is equal' the input since the noise is negligible. On the other hand, in very large noise power, the input is too noisy and so the output is an average of the input.

2.89. (a)

E{x[n]x[nJ} = ¢.,.,[OJ.

(b)

~%X(eiW) = X(eiW)X*(eiW)

= W(eiw)H(eiW)W·(dW)H·(ei"') = ~ww(ei"')IH(eiwW

Z 1

= O'w 1 - cos(w) + 1/4'

59

(c)

<pzx[nJ = <pt.ow[nJ * h[nJ * h[-n]

= (T~ ( (~) n urn] * (~) -n u[-n])

2.90. (a)

<pyz[n] = E{y[kJz[k - nJ}

= E L~oo h[rJxfk - rJ m~oo h[mJv[k - n - mJ} Note that <pxv[nJ = E{x[PJv[p - en, therefore:

00 00

¢vz[nJ = L L h[rJh[m]E{x[pJv[p - (n + e - s)J)

r=-oom:::-oo

= h[-nJ * h[nJ * ¢z,,[n].

~vz(ejW) = IH(eiwW~zv(dW).

(b) No, consider x[nJ white and

vrn] = -x[n] ¢ztl[n] = -(T;<5[n] ~zv(d"') = -(T;.

Noting that IH(ei"'W is positive,

Hence, the cross power spectrum can be negative.

2.91. (a) Since t[nJ = ern] - ern - 1],

~ ff (eiW) is given by:

~ff(d"') = Hl(eiW)Hl(e-jW)~ee(dW) = (1 - e-jW)(I - ejwk;

= (T;(2 - ejw - e-jW)

= (T;(2 - 2cos(w».

Pearson Education( s J p. I t~. 482. Fr F P3~::J<

.. ,

(b) 4>ff[m] is the inverse Fourier transform of <p!f(ejwr~sing part (a), we get:

4>f1[m] = a; (2a[m] - a[m + 1J - a[m - 1]).

(c)

<Pgg(ejW) = H2(ejw )H2 (e-jW)<p f f (ejW)

= {a;(2 - 2cos(w» , Iwi < We

o , We < Iwi ~ zr.

... _(-'-0-. ..... -,_. __ 0-,. .. :

(d)

q2 = i:- j" <pgg(ejW) dw
9 7r _"
1 jWC
= 27r -We q;(~-2cos(w»dw '\
= 2a; ( . ( »
- We -SIn We .
7r =

-~

Solutions - Chapter 3

The z- Transform

Pearson r~uc(ltiof1( ~ \ n~ 482, fl. FPc" !),'

61

If""

f J,q

62

I

f I

I.

I

~

3.1. (a)

(b)

(c)

(d)

(e)

(f)

(g)

3.2.

therefore

63

[(l)n ] 00 (l)n 00 ( l)n 1

Z 2 u(nJ =]; 2 z-n =]; 2z = 1 _ ~z-l

1 Izl> 2

2z 1

= - -- = --,---

1 - 2z 1 - lZ-l

2

1 Izl < 2

1 Izl < 2

Z(o(n]] = zO = 1 all z

Z[o[n - I]J = Z-l

Izi > 0

Z[o[n + 1]] = z+I

0:::; [z] < 00

[(I)n ] 9 (l)n 1- (2z)-lO

Z 2 (u[n] - urn - 10]) = ~ 2z = 1- (2Z)-1

Izi > 0

x[n] = {nN" 0:::; n:::; N -1 = n urn] - (n - N)u[n - N] s s»

n x[n]

d d 1

¢:} -z dz X (z) ::} n urn] ¢:} -z dz 1 _ z-l [z] > 1

z-l

¢:} (1 _ z-1 )2, Izl > 1

_ z-N-1

¢:} X(z)· z no::} (n - N)u[n - N] ¢:} (1 _ z-1)2 Izi > 1

n urn]

x[n - no]

-1 -N-1 -1(1 -N)

z -z z-z

X(z)= (1-Z-1)2 = (1-z-1)2

Pearson [.lncation( s) Pt U~. 482, 1 t PC'"

"'1.70 flax:2 '

64

3.3. (a)

(b)

(c)

=- =

.:::::::._

0< lal < 1

-I 00

Xa(z) = 2..: a-nz-n + 2..:anz-n

n=-oo n=O

00 00

= 2..:anzn + 2..:anz-n

n=1 n=O

o z 1 z(I - (2)

= -- + -_- = --'----'--

l-az I-ar! (l-az)(z-a)'

1 [o] < Izi < ~

Xc/..z)

pole zero v : ....- cancel

lJa

{I, 0 S n S N - I N-l 1 _ z-N zN _ I

Xb = 0, N S n ::;. Xb(z) = L z-n = 1- z-l = zN-l(z _ 1)

0, n < 0 n=O

z =J 0,1

pole zero ....-cancel

.. -

_v

65

3.4. The pole-zero plot of X(z) appears below.

X(z)

3

(a) For the Fourier transform of x[nJ to exist, the z-transform of x[nJ must have an ROC which includes the unit circle, therefore, It I < [z] < 121·

Since this ROC lies outside h this pole contributes a right-sided sequence. Since the ROC lies inside 2 and 3, these poles contribute left-sided sequences. The overall x[nJ is therefore two-sided.

(b) Two-sided sequences have ROC's which look like washers. There are two possibilities. The ROC's corresponding to these are: It I < Izl < 121 and 121 < [z] < 131·

(c) The ROC must be a connected region. For stability, the ROC must contain the unit circle. For causality the ROC must be outside the outermost pole. These conditions cannot be met by any of the possible ROC's of this pole-zero plot.

3.5.

X(z) = (1 + 2z)(1 + 3z-1)(1 - z-l) = 2z + 5 - 4z-1 - 3z-2

00

= L x[nJz-n

n=-oo

Therefore,

3.6. (a)

1

X (z) - --:---:- 1 + lz-l 2

1 Izl> 2

Partial fractions: one pole -+ inspection, x[nJ = (- t )nu[nJ Long division:

1 1 -1 + 1 -2

- 2z iZ

+ ...

1 + tz-1 1

1 + 1 -1 2z

1 -1 - 2z

1 -1 - 2"z

_ !z-2 4

+ 1 -3 SZ

=::} x[nJ = ( _~) n u[nJ

Pears (,

Del

66

(b)

1

X(z) = 1 1

1 + zZ-

1 [z] < 2

Partial Fractions: one pole -t inspection, x[nJ = - (- ~) nu[ -n - 1 J

Long division:

2z - 4z2 + 8z3 + ... ~z-l+1 1

1 + 2z - 2z

- 2z - 4z2

+ 4z2

+ 4z2 + 8z3

(c)

Partial Fractions:

Long division:

1 + 3 -1 + 1 -2

.4z SZ

(d)

Partial Fractions:

===> x[nJ = - (-~) n u[-n -I]

1 1-1 - -z

X(z) - 2

- 1 + !!Z-1 + lz-2

4 8

1 [z].> 2

X(z) =

x[n] =

+ (-136+I)z-2

+ ...

+ 1 -2 SZ

(-~ _ ~)z-1 (_~ _ !)Z-l

1 -2 - SZ

+ ~(_~ _ !)z-2

1 (3 1)-3

+ S -4 - 2" z

1- lZ-l

X(z) = 2

1- lz-2 ..

1 Izl> 2

X(z)

1 - lz-l 1

= 2 __ --:-_

1 - lz-2 - 1 + lz-l

4 2

1 Izl> 2

67

Long division: see part (i) above.

(e)

Partial Fractions:

X(z) = -a _ a-I(l- a2) [z] > la-II

1 - a-Iz-1

x[nj = -ao[nj- (1 - a2)a-(n+l)u[nj

Long division: _1

a

-a +z-l I 1

+ ...

1 - az-1

(a-1 - a)z-l

3.7. (a)

x[n] = u[-n-l] + (~) n urn]

-1 1

=> X(z) = 1 _ z-l + 1 _ lz-l

2

Now to find H(z} we simply use H(z} = Y(z}jX(z); i.e.,

Y(z} _lz-l

H(z) - __ - ~~~2 ,..--_-::-:- X(z) - (1-!Z-1)(I+z-1)

1

2" < [z] < 1

H(z) causal => ROC Izi > 1.

(b) Since one of the poles of X(z), which limited the ROC of X(z) to be less than 1, is cancelled by the zero of H(z), the ROC of Y(z) is the region in the z-plane that satisfies the remaining two constraints [z] > ! and Izi > 1. Hence Y(z) converges on Izi > 1.

(c)

1 1

Y(z) = -3 + 3

1 - lZ-l 1 + Z-l 2

Izl> 1

Therefore,

I

r < 3.8. The causal system has system function

I,

t

~

1 (1)" 1

yIn] = -- - urn] + -(-I)"u[nJ

323

1- z-l H(z) - --::--- 1 + !Z-l 4

and the input is x[n] = n r urn] + u[-n - 1]. Therefore the z-transform of the input is

1 X(z) = 1 _ lz-l 3

1

3 < Izl < 1

Pearson f-rl""atir'1( S) Pt l t~ 482. T ,IT her'

Delb:·ll009:: : i. r ,

_ ' 22146067~70 I ;(], __ ; } ·cr

68 __,

'~(a) h[n] causal =}

(3)n ( 3)n-l

h[nJ = -4 u[nJ - -4 urn - 1J

(b)

3

4 < Izl

Therefore the output is

8 (1) n 8 ( 3) n

y[n] = -- - urn] + - -- u[nJ

13 3 13 4

'It

(c) For h[n] to be causal the ROC of H(z) must be ~ < [z] which includes the unit circle. Therefolt

. h[nJ absolutely summable. 1

. ~

3.9.

1 + z-l 2 1

H(z) - -~--~-

- (1 - ~z-l)(1 + iz-l) (1 - ~z-l) (1 + tz-l)

(a) h[nJ causal =} ROC outside Izl = ~ =} Izi > ~.

(b) ROC includes [z] = 1 :::} stable.

(c)

y[n]

il ('

Y(z)

1 4

= -3' + 3'

1 + !Z-l 1 - 2z-1 4

(I

X(z)

1 + z-l

=

(1 + iz-l)(1 - 2z-1)

= Y(z) = (1 - ~z-l) Izl < 2

H(z) (1 - 2z-1)

= -(2)nu[-n - 1] + !(2)n-lu[_n] 2

1

4' < Izi < 2

(1

I:

x[n]

(d)

h[n] = 2 (~) n urn] _ (_~) n urn]

3.10. (a)

x[n] = (~) n urn -10] + (~) n urn -10] = (~) n urn] + (~) n urn]

_ [( (4) n + (~) n) (u[n] _ urn _ 11])]

The last term is finite length and converges everywhere except at z = O. Therefore, ROC outside largest pole t < 14

3.11.

.. ~.

69

(b)

{I, -10 < n < 10 x[n] = 0, other-;'ise-

Finite length but has positive and negative powers at z in its X (z). Therefore the ROC is 0 < Izi < 00.

(c)
x[n] C) -n
= 2nu[-n] = 2 u[-n]
x[-n] +-t X(I/z)
(~) n urn] ROC is 1
:::} Izl> 2
c)-n Izl < 2
2 u[-n] :::} ROC is
(d)
x[n] = [ (~) nH _ (ejJr/3)n] urn - 1] x[n] is right-sided, so its ROC extends outward from the outermost pole &7</3. But since it is non-zero at n = -1, the ROC does not include 00. So the ROC is 1 < [z] < 00.

(e)

x[nJ = urn + 10J - urn + 5J = {I, -IO~n~-6 0, otherwise

xfn] is finite-length and has only positive powers of z in its X(z). So the ROC is [z] < 00.

(f)

(l)n-l

xfn] = 2 urn] + (2 + 3j)n-2u[_n - IJ

x[nJ is two-sided, with two poles. Its ROC is the ring between the two poles: ~ < Izl < I 213j I, or ~ < Izi < ilJ·

00

x[nJ causal :::} X(z) = L x[nJz-n n=O

;, . which means this summation will include no positive powers of z. This means that the closed form of

X(z) must converge at z = 00, i.e., z = 00 must be in the ROC of X(z), or limz-+oo X(z) :f; 00.

(a)

could be causal

Pearson Erl!lr~ation(s) Pt I ... tao

482, f'.; E Patpar,;::::r: L

Deih:-: J OOG;? f nrh

.. 221 ry ~f'

3.12. (a)

(b)

. {z - 1)2

lim 1 = 00

z r-e oo (z - 2)

(c)

(d)

(z _ 1)6

lim 4 - 00

z-+=(z - ~)S -

could not be causal

could be causal

could not be causal

1 1-1 - -z

X1(z) = 1 + ~Z-1

The pole is at -2, and the zero is at 1/2.

(b)

1 - lz-1

X (z) - 3

2 - (1 + ~z-l)(1 _ ~z-1)

The poles are at -1/2 and 2/3, and the zero is at 1/3. Since x2[n] is causal, the ROC is extends from the outermost pole: Izl > 2/3.

Xl (z)

112

(c)

1 + z-l - 2z-2 X3{Z) = 1 _ llz-l + z-2 6

The poles are at 3/2 and 2/3, and the zeros are at 1 and -2. Since x3[n] is absolutely summable, the ROC must include the unit circle: 2/3 < Izl < 3/2.

X3 (z)

3.13.

71

G(z) = sin(z-l)(1 + 3z-2 + 2z-4)

-3 -5 -7

= (Z-l _ :_ + :_ - :_)(1 + 3z-2 + 2z-4)

3' 5! 7!

= Lg[nJz-n

n

g[IlJ is simply the coefficient in front of Z-ll in this power series expansion of G(z):

1 3 2 g[IlJ = -11! + 9! - 11!·

3.14.

H(z) =

1

1- !z-2 4

1

=

(1- ~z-l)(l + ~z-1)

0.5 0.5

= + --'---=--

1 - lZ-l 1 + lz-l

2 2

Taking the inverse z- Transform:

1 1 1 1

h[n] = -( - )nu[n] + -( -- )nu[n]

2 2 2 2

So,

1 A1 =-; 2

3.15. Using long division, we get

1 1 -10

- l'024z

1- !z-1

2

n=9

= 2)~)nZ-n

n=O

H(z) =

Taking the inverse z-transform,

h[nJ= { ar,

0,

n = 0, 1,2, ... , 9 otherwise

I

t ' Since h[n] is 0 for n < 0, the system is causal.

f 3.16. (a) To determine H(z), we first find X(z) and Y(z): i

1

X(z} =

1

1 - lz-1 3

1- 2z-1

=

1

3" < Izl < 2

Pearson r ~ocatiords) Pt 1 h~ 482, r' F Palp~lq~ r:

Delb l-l 1 O()Q2 ; n (j J <J

... 2Z146067c70 l'ax:22i4607\

_- '"

72

Y(z) =

,- 5

1 - !'Z-l 3

5

1 - ~Z-l 3

=

2 [z] > :3

Now

H(z) =

Y(z) X(z)

1 - 2z-1 1 - tz-l

2 Izl> :3

=

The pole-zero plot of H(z) is plotted below.

H(z)

(b) Taking the inverse z-transform of H(z), we get

h[n]

(c) Since

H(z) = Y{z) = 1 - 2z-1 , X(z) 1 - *z-l

we can write

2

Y(z)(l - 3z-1) = X{z)(l- 2z-1),

whose inverse z-transform leads to

2

y[nJ - -y[n - 1J = x[n] - 2x[n - 1) 3

(d) The system is stable because the ROC includes the unit circle. It is also causal since the impulse response h[nJ = 0 for n < O.

3.17. We solve this problem by finding the system function H(z) of the system, and then looking at the different impulse responses which can result from our choice of the ROC.

Taking the z-transform of the difference equation, we get

5

Y(z)(l- 2Z-1 + z-Z) = X(z)(l- z-l),

and thus

H(z)

Y(z) 1 - Z-l

= X(z) = 1 - ~z-l + Z-2

I

L

73

=

=

(1 - 2z-1)(1 - ~z-l)

2/3 1/3

---+-~-:- 1-2z-11-1z-1 2

If the ROC is (a) Izl < ~:

h[nJ = -~2nu[-n - IJ - ~(~ )nu[-n - 1J

3 3 2

==:} ~[OJ = O.

(b) ~ < [z] < 2:

(c) Izl > 2:

2 1 1

h[n] = 32nu[n] + 3(2Tu[n]

==:} h[O] = 1.

(d) Izi > 2 or Izi < ~:

3.18. (a)

H(z) =

1+2%-I+z-2 (1 + ~z-l)(1 - z-l)

1 8

= -2+ 3" + 3" 1+!z-11-z-1 2

Taking the inverse z..transform:

1 1 8

h[n] = -26[n] + 3( -2)fiu[n] + 3u[n].

(b) We use the eigenfunction property of the input:

y[n] = H(dlr/2)x[n],

where

1 8

= -2+ 3" + 3"

1 + !e-j1r/2 1 - e-j7r/2

1 8

= -2+-L..+-L 1-!i 1 + j

-2j

Putting it together,

74

~

---~-

- .. c.:...

3.19. The ROC(Y(z» includes the intersection of ROG(H(z» and ROC(X(z». (a)

1

Y(z) - -~------=-- (1 + ~z-l)(l- ~z-l)

The intersection of ROGs of H(z) and X(z) is Izl > ~. So the ROC of Y(z) is Izl > ~. (b) The ROC of Y(z) is exactly the intersection of ROGs of H(z) and X(z): ~ < [z] < 2. (c)

1

Y(z) - -~----:-- (1 - ~Z-1)(1 + ~z-l)

The ROC is [z] > ~.

3.20'. In both cases, the ROC of H(z) has to be chosen such that ROC(Y(z» includes the intersection of ROC(H(z» and ROC(X(z».

(a)

1 - ~Z-l

H(z) = 4

1 + '?'z-l 3

The ROC is Izi > ~.

(b)

1

H(z) = 1 I

1- -z.6

3.21. (a) The ROC is Izl > t· y[n] = 0

n<O

y[n]

n n 1 _ a-(n+l) 1 _ an+1

= 2:x[k]h[n - k] = 2:an-k = an 1 _ a-I = 1 _ a 0 :s: n < N - 1

k=O k=O

N-l N-I l-a-N I-N

= 2: x[k]h[n - k] = 2: an-k = an 1 _ a-I = an+l : ~ r : n > N

k=O k=O

y[n]

(b)

00 1
H(z) = 2:anz-n = Izl> lal
1- az-l
n=O
N-I 1 -N
X(z) 2: -z Izl > 0
= z-n -
- 1- Z-l
n=O Therefore,

1- z-N 1 z-N

Y (z) - = ..,.....---:-:-.,---"..,-

- (1 - az-1)(1 - z-l) (1 - az-I)(1 - z-l) (1 - az-l)(1 - z-l)

Izl> lal

Now,

I 1 ) ( )

1 l-a-1 r::a 1 1 a

(1 - az-1)(1 - z-l) = 1 - az-1 + 1 - z-l = (1 - a 1 - Z-l - 1 - az-1

( -~

'';

3.22. (a)

3.2S. (a)

75

So

yin] = (1 ~ a) [urn] - an+1u[n] - urn - N] - an-N+1u[n - N]]

1 - dn+1 1 _ an-N+1

= 1 - a urn] - 1 _ a urn - N]

o n<O

yin] l_an+1 O:::;nSN-l
= ~
an+1 (l_a-N) n > N
a-I
00
yin] = L h[k]x[n - k]
k=-oo (b)

=

=

yin]

H(z) =

(1 - iz-1 )(1- fz-1)

5 + lZ-1

= -4+ 2

1 - ~z-l + !z-2 48

2 7

= -4 - + --.,...--

1 - !'z-l 1 - !'Z-l

2 4

h[n] = -4o[n]- 2 (~) n urn] + 7 (~) n urn]

Pearson f"'ucatiorf s) Pt t h) 482, fl F Pa ~j:'i

76

(b)

3 1 1

y[n]- -y[n - 1] + -y[n - 2J = x[n] - -x[n - 2] 482

3.24. The plots of the sequences are shown below.

3.25.

(a) Let

Then

(b)

Obtain a proper fraction:

00

a[nJ = L o[n - 4k],

k=-oo

0:>

A(z) = L z-4n

k=-o:>

b[n]

= ~ [ei"n + cos (~n) + sin (~ + 21m)] urn] ~ [(_1)n + cos (~n) + 1] urn]

{ i' n = 4k, k ~ 0

2' n = 4k + 2, k ~ ° 0, otherwise

=

=

B(z)

= f ~z-4n + f ~z-(4n+2)

n=O n=O

3/2 + 1/2z-2

=

Izl > 1

1- Z-4

O.tl
'20.6 ...
'iii'
0.'
02
0
-'0 -lI ..,; .... -2 0 2
n
'5
'2
:c
... r r
05
0
-10 ... -lI .... -2 0 2 6
n
Z2 Z2
X(z) - =
- (Z - a)(z - b) z2 - (a + b)z + ab '0

10

77

1

z2-(a+b)z+ab Z2

Z2 - (a + b)z + ab (a + b)z - ab

X(z)

( b) b (a+b)a-ab (a+b)b-ab

1 a + z - a 1 a-b b-a

= + = + + _:....::.,-----

(z - a)(z - b) z - a z - b

a2 b2 2 -1 2 -1 )

= 1 + a-b _ a-b = 1 + _1_ ( a z _ b z

Z - a z - b a - b 1 - az-1 1 - bz-1

a2 b2

= a[nJ + -_an-1u[n - 1J - --bn-1u[n - 1]

a-b a-b

= &[nJ + (_1_) (an+! - bn+1)u[n - 1J a-b

x[n)

3.26. (a) x[n] is right-sided and

1 - lZ-l

X(z) = 3

1 + lz-l 3

Long division: 1

+ 2 -2 gZ

+ ...

+ 2 -2 gZ

Therefore, x[n] = 2(-l)nu[n) - &[n)

(b)

X(z) = 3 = 3z-1 = 4 _ 4

Z - 1 - lz-l (1 - !'z-l)(l + lz-l) 1 - lz-l 1 - !'Z-l

4 8 2 4 2 4

Poles at ~, and -i. x[nJ stable, => Izl >! => causal. Therefore,

I

r

I '

i

I

x[nJ = 4 (~) n urn] _ 4 (_~) n u[nJ

(c)

X(z) = In(l - 4z)

1 Izl < 4"

= - t(~)i = - t ~(4z)-l

i=l' (=-00

Therefore,

1

x[n] = -(4)-nu[-n - 1) n

Pearson f f'~'cation( sl Pl. t td 482, r.: E Paq.'<lTp"r:'

DelL :-l100Y2 ; nc1 1.'1

§If . 22146067·70 F~"L .. -.

78- ..

. ~ (d)

1

X(z) = 1

1 - 3z-3

[z] > (3)- t =} causal

By long division:

1 + 1 -3 3z

+ 1 -6 gZ

+ ...

1

1

+ 1 -6 gZ

n = 0,3,6, ... otherwise

3.27. (a)

X(Z) =

1

1

2" < [z] < 2

Therefore,

_ 1 (_l)n+1 58 (_1)" 1568" 2700 nt

x[nJ- 35(n+1)"'2 u[n+1]+ (35)2 "'2 u[nJ+ (35)2(2) u[-n-1J- (35)2(3) u[-n-1]

(b)t

1 Therefore, x[nJ = n! urn].

(c)

Z3 - 2z 2

X(z) = = z2 + 2z + ----:-

Z - 2 1 - 2z-1

Izl < 2

Therefore,

x[n] = o[n + 2] + 2o[n + 1] - 2(2)nu[-n - 1]

3.28. (a)

d

nx[n] {::} -z dxX(z)

x[n - no] {::} z-noX(z)

X(z) = (1 :!~~1)2 = 12z-2 [-z ~ C _ iz-1)]

x[n] is left-sided. Therefore, X(z) corresponds to:

(1) n-2

x[nJ = -12(n - 2) 4 u[-n + 1J

79

ROC includes Izl = 1

(-ll o[n + 2k + IJ (2k + I)!

1

Izl> 1

1- Z-7

0:: I 1

, 2: ,z- =:::;.. x[n] = -I I' + o[n]

-=on.:: n.

1 1 (1)l

= '5"-: - - z-l ._ t 2

.!:=-::c

1 Izi < 2

!'("'f;00f .. ) Pt l. t~

. " -_ U ,JVJI' ;"\ ;.' _QIl

iJa r pa cg. l 1. I DelL:-llO092 iridia

... 22146067·70 Fax 221,ttS07?

_. 80

3.31. (a)

.**"

x[nJ = anu[nJ + bnu[nJ + enu[-n - 11 lal < Ibl < lei
X(z) 1 1 1 Ib/ < Izi < lei
= + -
1 - az-1 1 - bz-I 1 - ez-1
X(z) 1- 2ez-I + (be + ae - ab)z-2 Ibl < /z/ < lei
=
(1 - az-I)(1 - bz-I )(1 - er1 ) Poles: a, b, c,

Zeros: ZI, ZZ, ex) where ZI and Z2 are roots of numerator quadratic.

(b)
x[nJ = n2anu[n]
xI[nJ 1 Izl > a
= anu[nJ ¢} Xl (z) = -1
1- az
xz[nJ = nxdnJ d az-1
= nanu[nJ ¢} X2(Z) = -z;rX1(z) = ( -1)2
Z 1- az
x[nJ = nX2[nJ 2 n [ J d d ( az-1 )
= n a un¢} -z dz X2(z) = -z dz (1 _ az-1)2
X(z) = -az-1 (1 + az-1) Izl > a
(1 - az-1)3 (c)

Izl > a

Izi > a

x[nJ = en· (cos ~n) urn] - en· (cos l~n) urn -11 = en· (cos l~n) (u[nJ- urn -1]) = 6[nJ Therefore, X(z) = 1 for all ]e].

.3.32. From the pole-zero diagram

y[nJ = x[-n + 3J = x[-(n - 3)J

3 lzl> 4

=

z(2 - 2z + z2)(t + z)

1/

1; i·

l

.~ ...

81

Poles at 0, -~, 1 ± j, zeros at 00

x[nJ causal ee- x[-n + 3J is left-sided ee- ROC is 0 < Izi < 4/3.

3.33. From pole-zero diagram

(a)

zeros poles

±l . -J 1 2

i'OO

(b)

poles at ±!j zeros at ±1

X(Z)=z2+11 z-:2

(l)n 4z2 + 1

yin] = - x[n] :::;. Y(z) = X(2z) = 1

2 2z -:2

V(z)

w[n] = cos (~n) x[n] = ~(ei".n/2 + e-i1rn/2)x[n] W(z) = !X(e-ill"/2z) + !X(eill"/2z) = !X(-jz) + !X0z)

2 2 2 2

1 (- z2 + 1) 1 (- z2 + 1) z2 - 1

W(z) = - . 1 + - . 1 = (2 1

2 -JZ - 2' 2 JZ - 2' 2 z + i)

Pearson Ert:q::ation( s) Pt . Lt". 482. f _j E Patpar!;:r:!

Delb -.'.' 0092 i '~fa

~ 22146067-70 f',lxL ,.,.

3.34.

3.35.

82

(a)

(b)

_=_

Y'(z)

3 - 7z-1 + 5z-2 1 3

H(z) == == 5 + --::----:-

1 - 2.z-1 + Z-2 1 - 2z-1 1 - lZ-1

2 2

h[n] stable => h[n] == M[n]- 2nu[-n -1] _ 3 (~) n urn]

n

y[n] = h[n] * x[n] = 2: h[k]

k=-oo

=

n2:0

n

- 2: 2k = _2n+1 k=-oo

n < 0

-1 n () k 1 e )n+l ( ) n

- L 2k + 5 - 'L) ~ == 4 - 3 - 2 1 == -2 + 3 ~

k-oo k=O 2 I - 2 2

== -2u[n] + 3 (~) n urn] _ 2n+1u[-n -11

Y(z) = 1 H(z) = -2 1 + 2 1 + 3 1

1 - z-l 1 - z-l 1- 2z-1 I - lZ-1

2

y[n] = -2u[nJ - 2(2)nu[-n _ 1] + 3 (~) n urn]

I

2" < Izl < 2

H(z) = 1 - z3 -1 (1 - z-3) Izl> 1
---z
1- z4 - 1- z-4
urn] 1 z Izl> 1 r -:: ~
¢:? =--
1- z-1 z-1 11:-
.. <;
U(z)H(z) Z-l _ Z-4 r:.~
= f
(1 - z-4)(1 - z-1) ~1:
z-l Z-4
= Izl> 1
1- z-l 1- z-4
00
urn] * h[nJ = urn - 1] - I>~[n - 4 - 4k]
k=O ~41 ~.

I I

f

i'

I

i 3.38.

.-.

83'

3.36.

1 x[nJ = u[nJ ¢:} X(z) = -l---z---:-j

Izl> 1

(l)n-l (l)n+l 4

y[nJ ="2 urn + 1J = 4"2 urn + 1J ¢:} Y(z) = 1 _ ;z-l

1 Izl> "2

(a)

H(z) = Y(z) = 4z(1 _ Z-l)

X(Z) 1 _ ~z-l

1 IzJ> - 2

(b)

h[nJ

4z 4 1

1 1 [z].> -

1 - '2z-1 1 - '2Z-1 2

(l)n+l . (l)n

= 4 2' urn + 1J - 4 2' u[nJ

= 40[n+l]_2(~)nU[n]

H(z) =

(c) The ROC of H(z} includes Izi = 1 => stable.

(d) From part (b) we see that h[n] starts at n = -1 => not causal

3.37.

1 1

X(z) = 3' + 4"

1 - lz-l 1 - 2z-1 2

has poles at z = ~ and z = 2.

Since the unit circle is in the region of convergence X(z) and x[nJ have both a causal and an anticausal part. The causal part is "outside" the pole at t. The anticausal part is "inside" the pole at 2, therefore, x[O] is the sum of the two parts

! !z 1 1

x[O] = lim 3 + lim _.f_ = - + 0 = -

z .... oo 1 - !Z-l z .... o Z - 2 3 3

2

Z-l + z-2 2

Y(z) - -~_-~-

- (1- ~z-l)(l + kz-l) 1- z-l

Izl> 1

Pears-r r ducatinn! s) Pt ~ ltd,

84

. .;.,

Therefore using 'a contour e that lies outside of Izl = 1 we get

y[lJ =

1 i 2{z + l)zndz

21r j C (z - !)( z + t)( z - 1)

2(!+1)(!) 2(-~+1)(-~) 2{1 + 1)(1)

(~+t)(t-l) + (-t-t)(-t-l) + (1-t)(1+t) 18 2

= -- - - + 6 = 2 5 5

=

3.39. (a)

Z10

X (z) - ---,---.,,----=----:--___=_

- (z - t)(z - ~PO(z + ~)2(z + ~)(z +~)

Stable ee- ROC includes [z] = 1. Therefore, the ROC is ! < [z] < ~.

(b) x[-8] = E[residues of X(z)z-9 inside e), where e is contour in ROC (say the unit circle).

X[8]=E[residuesof( 1)( 3)10( z 3)2( 5)( 7 insideunitcircle]

z-z z-z z+z z+z z+z)

First order pole at z = ! is only one inside the unit circle. Therefore

3.40. (a) After writing the following equalities:

V(z) = X(z) - W(z) W(z) = V(z)H(z) + E(z)

we solve for W(z):

H(z) 1

W(z) = 1 + H(z) X(z) + 1 + H(z) E(z)

. ;

(b)

.. -1

H(z) = 1-rl = z-l

1 + H(z) 1 + 1~:~1

1 = 1 _ z-1

1 Z-1 + 1-z-1

(c) H(z} is not stable due to its pole at z = 1, but Hl(Z} and H2(Z} are.

3.41. (a) Yes, h[n] is BIDO stable if its ROC includes the unit circle. Hence, the system is stable if rmin < 1

and Tma:z: > 1. 1:_1:-:

(b) Let's consider the system step by step.}: (i) First, v[n) = a-nx[n]. By taking the z-transform of both sides, V(z) = X(az).

(li) Second, v[n] is filtered to get wIn]. So W(z) = H(z}V(z} = H(z)X(az).

(iii) Finally, yIn] = anwln]. In the z-transform domain, Y(z} = W(z/a) = H(z/a)X(z). In conclusion, the system is LTI, with system function G(z} = H(z/er.) and gIn] = anh[n].

85

( ;

(c) The ROC of G(z) is armin < Izl < armax. We want rmin < 1ja and rmax > 1ja for the system to be stable.

3.42. (a) h(n] is the response of the system when x[nJ = 5[n). Hence,

10

h(n] + L akh[n - k) = 5[n) + /35[n - 1),

k=l

Further, since the system is causal, h[n] = 0 for n < O. Therefore,

10

h[O) + L akh[-k) = h[O) = 5(0) = l.

k=l

(b) At n = 1,

h[l] + a1h[O] = 5[1] + /35(0)

/3 - h[lJ

=> al = h[O] = /3 - h[l)

(c) How can we extend h]n] for n > 10 and still have it compatible with the difference equation for S?

Note that the difference equation can describe systems up to order 10. If we choose

1T

h[n] = (0.9)n cos( 4n)u[n),

we only need a second order difference equation:

The z-transform of h[n] can be found from the z-transform table:

1 0.9

-72

H{z) = (1- 0.gejlrj4z-l){1- O.ge-j1rj4z-l)

3.43. (a)

1 1

X{z) = 1 _ tz-l 1- 2z-1 '

6 6

Y(z) = 1 1 -1 1 _ lz-1 '

- Zz 4

1

2 < Izl < 2

3 Izl> "4

_! .. -1

H(z)

Y{z) (1-fr1)(l-!r1)

= X{z) = -i" 1

(l-fZ-l )(1-2..-1)

1- 2z-1 3

= 1 _ lz-1 ' Izl > 4"

4

l

P r , ~"1,.,...,1,' \ ru It!:.

ears on ,D1Jf:8;"c;;"'I' \ '

4821 r EVa;,

De:L:~l:OO~~ (':;~

_ '22146067e7C f'c\:;:

86

Pole-zero pIoI of H{z)

1.5

3/4 2

I

- I .5,-' _~---:---:--:-_--,--::_:-'

-1 _0.5 0 0.5 1.5 2

Real pan

(b)

(3) n (3)n-1

h[nJ = 4" u[nJ - 2 4" urn - 1J

(c)

3

yin] - -yin - 1J = x[nJ - 2x[n - 1J 4

(d) The system is stable because the ROC includes the unit circle. It is also causal since h[n] = 0 for n < O.

3.44. (a)

1 4

X(z) = -3 + 3

1 - ~Z-l. 1 - 2z-1

The ROC is ~ < Izl < 2.

(b) The following figure shows the pole-zero plot of Y(z). Since X(z) has poles at 0.5 and 2, the poles at 1 and -0.5 are due to H(z). Since H(z) is causal, its ROC is Izl > 1. The ROC of Y(z) must contain the intersection of the ROC of X(z) and the ROC of H(z). Hence the ROC of Y(z) is 1 < Izl < 2.

Pole-zero plot 01 Y(z)

1.sr----,--------,

2

-1

-1.s~_I~_O::":.s:---:-O --:::0.5-:---:----:1:"";.s--:2::-' Real paJ1

(c)

H(z)

Y(z)

=

X(z)

=

1 -1

(1-r1 )(Hl~-1 )(1-2rl) 1

(1-12% i)(1_2rl)

87

(1 + Z-1 )(1 - tz-1) (1 - z-1)(1 - tz-1)

2 2

= 1 + 3 + -3

1-z-11+1z-1 2

=

Taking the inverse z-transform, we find

2 2 1

h[nJ = o[n] + -u[nJ - -( -- )nu[n]

3 3 2

(d) Since H(z) has a pole on the unit circle, the system is not stable.

3.45. (a)

ny[nJ = x[nJ
dY(z) = X(z}
-z--
dz
Y(z) = - / Z-1 X(z)dz (b) To apply the results of part (a), we let x[nJ = urn - 1], and w[nJ = y[n].

/ -1
W(z) = - z-l Z 1 dz
1- z :
/ 1
I ~. = - dz
I' z(z- 1)
/ -1 1
= - -+--dz
z z-l
= In(z) -In(z - 1) 3.46. (a) Since y[n] is stable, its ROC contains the unit-circle. Hence, Y(z) converges for ~ < Izi < 2. (b) Since the ROC is a ring on the z-plane, y[n] is a two-sided sequence.

(c) x[n] is stable, so its ROC contains the unit-circle. Also, it has a zero at 00 so the ROC includes 00. ROC: Izi > ~.

(d) Since the ROC of x[n] includes 00, X(z) contains no positive powers of z, and so x[nJ = 0 for n < O. Therefore x[nJ is causal.

(e)

x[OJ = X(z)I==oo

A(l- tz-1)

= (1 + ~z-l )(1 _ ~z-l) 1::=00

= 0

(f) H(z} has zeros at -.75 and 0, and poles at 2 and 00. Its ROC is Izi < 2.

Pearson [rh'c?tir;r(r:) 1].82" r ~!~

fit .

Ltd.

8a:-

1.5

Pole-zero plot of H(zL

2

ROC: "k2

-1

-1.5L- J_-:------:--,-'

-1 -C.5 0 0.5 1.5

Real part

(g) Since the ROC of h(n] includes 0, H(z) contains no negative powers of z , which implies that h(n] = 0 for n > O. Therefore h(n] is anti-causal.

3.47. (a)

00 ,

X(z) = Lx(n]z-n

n=O

00

X (00) = lim" x(n]z-n = x(o]

z-+oo ~ n=O

Therefore, X(oo) = x(O] ¥- 0 and finite by assumption. Thus, X(z) has neither a pole nor a zero at z = 00.

(b) Suppose X(z) has finite numbers of poles and zeros in the finite z-plane. Then the most general form for X(z) is

M

00 II(z-q)

X(z) = L x[n)z-n = KzL..:.:k~::;,.1=-- __

n=O II (z - dk)

K=1

where K is a constant and M and N are finite positive integers and L is a finite positive or negative integer representing the net number of poles (L < 0) or zeros (L > 0) at z = O. Clearly, since X (00) = x[O) ¥- 0 and < 00 we must have L + M = N; i.e., the total number of zeros in the finite z-plane must equal the total number of poles in the finite z-plane.

le.

.,'

3.48.

X( ) = P(z) z Q(z)

where P(z) and Q(z) are polynomials in z. Sequence is absolutely summable ::} ROC contains Izl = 1 and roots of Q(z) inside Izi = 1.

These conditions do not necessarily imply that x(n] is causal. A shift of a causal sequence would only add more zeros at z = 0 to P(z). For example, consider

X(z)

=

1 Izl > 2

z 1

= -- = z . --;----:-

Z - 1 1 - lz-1

2 2

(1) n+1

=> x[nJ = 2 urn + IJ ::}

right-sided but non-causal.

3.49.

3.50. (a)

3.51. (a)

I

i

\ ,

"

3.52.

Therefore,

Therefore,

89

x[nJ = o[nJ + ao(n - N) lal < 1 X(z) = 1 + a«:"

a2 z-2N a3 z-3N

X(z) = 10gX(z) = log(l + az-N) = ««:" - 2 + 3 - ...

00 (_l)k+l

x[n] = L k ak5[n - kN]

k=l

x[n] = x[-n] => X(z) = X (~)

(b)

i.e., ljzo is also a zero of X(z).

x[nJ real => x[nJ = x'[nJ => X(z) = X*(z')

X(Zo) = 0 = X(zo) i.e., Zo is also a zero and by part (a) so is ljzo.

00 ( 00 )*

Z[x*[nJ] = n~oo X*[nJz-n = n~oo x[nJ(z*)-n = X*(z')

00 00

Z[x[-nJJ = L x[-nJz-n = L x[nJ(z-l)-n = X(Z-l)

n=-oo

n=-oo

Therefore

(b)

(c)

(d)

Z[Re{x[n]}] = Z [x[n] ~ x.[nJ] = ~ [X(z) + X*(z*)]

Z[Im{x[n]}] = Z [x[nJ ;/*[nJ] = ;j [X(z) - X*(z*)J

00

xl(n) = (-l)nx(n) => X1(z) = L (_l)nx(n)z-n = X(-z)

The poles and zeros are rotated 180 degrees about the origin.

n=-oo

i Iii...

48:::', f.i t. Pc; r

Delh > ~ 10092 ,riC' ,:i ... 22J46067~70 r,txL: j.:!.( (

90 .~ .

3.53. (a)

N-l

~" . - L x[nJ sin(nw)

B:r;(w) = tan-1 (Im{X(eJW}») ==} tanB:r;(w) = -:-:-,n_=O.:__ _

Re{X(eJw)} N-l

L x[nJ cos(nw)

N-l N-!

tanB:r;(w) I: x[nJcos(wn) = - L x[nJsin(nw)

n=O

n=!

N-!

tan8:r;(w)x[OJ + I: x[n)(tanBx(w) cos(nw) + sin(nw» = 0

n=l

N-l

tan B:r;(Wk) + xtO] ~ x[nJ(tanB:r;(wk)COS(nwk) +sin(nwk» = 0

for N - 1 values of Wk in the range 0 < Wk < tt .

(b) x[nJ = J[nJ + 2J[n - IJ + 3J[n - 2J ==} X(z) = 1+ 2z-1 + 3z-2

8:r;(w) = tan-1 ( -2sin(w) - 3sin(2w) ) 1 + 2cos(w) + 3cos(2w)

Consider the values 8:r; (I) = 5; and B:r; (23") = 56'" which give the equations

tan8:r; (~) + x~J [X[I] (tan8:r; (~) cos ~ + sin~)

+ x[2] (tan e. (~) cos 11" + sin 11" ) ] = 0

tanB:r; (2;) + :0 [X[I] (tanB:r; (2;) cos 2; + sin 2311")

+ x[2] (tan8:r; (2311") cos ~ + sin 4;)] = 0

1

1 + x[O] (x[I] . 1+ x[2]. -1) = 0

- ~ + xtO] (X[I] + 2~ +~) + x[2] (2~ - ~) = 0

x[O] + x[I] - x[2] = O} {X[I] =

-x[O] + 2x[I] - x[2] = 0 ==} x[2] =

2x[0]

3x[0]

Therefore

x[n] = x[O](J[n] + 2J[n - 1] + 3J[n - 2])

where x[O] is undetermined. ~.54.x[nJ = 0 for n < 0 implies:

00 00

lim X(z) = lim L x[nJz-n = x[O] + lim L xfn]z-n = x[O]

z-+oo %-+00 z-+oo

n=O n=!

r;

i

l '

I

_ "'F ' __

91

For the case x[nJ = 0 for n > 0,

o 00

lim X(z) = lim ~ x[n]z-n = x[O] + lim ~ x[-nJzn = x[OJ

z-+o z-tO L.--t z-+o L__,

n=-(X) n=l

3.55. (a)

00

00

cxx[nJ = L x[kJx[n + k] = L x[-k]x[n - kJ = x[-nJ * x[nJ

k=-oo

k=-oo

Cxx(z) = X(z-I)X(z) = X(z)X(Z-I)

X(z) has ROC: TR < Izl < TL and therefore X(z-l) has ROC: Til < Izl < Till. Therefore Cxx(z) has ROC: max[Til,TR] < [z] < min[Till,Td

(b) x[n] = anu[n] is stable if lal < 1. In this case

1 X(z) = 1 _ az-l

1

[c] < Izl and X(z-l) = -I- az

Therefore

1

1

1 - az-1 1 - az = (1 - az-1)(1 - a-I Z-1 )

1 1-,,2 =

1 - az-1

1 1_,,2

This implies that

1 _

cxx[n] = 1 _ a2 (anu[n] + a nu[-n - 1])

Thus, in summary, the poles are at a and a-I; the zeros are at 0 and 00; and the ROC of Cxx{z) is [c] < Izl < la-II·

1/0

(c) Clearly, xl[n] = x[-n] will have the same autocorrelation function. For example,

1 X1(z) = -l-az

-1 1 1 )

Izl < la I=> CX1X1(z) = -1--1 -1 = Cxx(z -az -az

(d) Also, any delayed version of x[n] will have the same autocorrelation function; e.g., x2[n] = x[n - m] implies

92

.~

3.56. In order to be a z-transform, Xc(z) must be analytic in some annular region of the z-plane. To-determine if X (z) = z" is analytic we examine the existence of X' (z) by the Cauchy Riemann conditions. If

X(z) = X(x + jy) = u(x, y) + jv(x, y)

then for the derivative to exist at z, we must have

au B» au av

- = - and

ax By By ax

In our case,

X(x+jy) =x-jy

and thus,

au = 1 t= av = -1

ax ay

unless x and yare zero. Thus, X'(z) exists only at z = O. X(z) is not analytic anywhere. Therefore,

x[nJ = ~ i=r= does not exist. 27rJ

3.57. If X(z) has a pole at z = Zo then A(z) can be expressed as a Taylor's series about z = Zo.

00 An(zo)

A(z) = A(Zo) + L -,-(z - zo)n

n. n=l

where A(zo) = O. Thus

Res [X(z) at z = zoJ

B(z) I

= X(z)(z - zo)lz=%o = A(z) %=%0

=' B(z)(z - Zo) I

00 An(zo) n

"--(z - zo) %=%0 L.., n!

n=l

= B(z) I B(zo)

00 An( ) = A'(zo)

A'(zo) + " ~(z - zo)n-l %=%0

L.., n!

n=2

._E!L_.

.~.

93

Solutions - Chapter 4

Sampling of Continuous-Time Signals

I

f1

i.f<

" I'

If

l

Pearson Erl:." ;,~:r 1·1 f' I ~,~ 482, f ~. I:

nc,L,

.. -_- ' .~

94

4.1.

4.2. The discrete-time sequence

95

x[nJ = xc(nT)
= sin (27r(100)n 4~0)
= sin (~n) 7rn

x[nJ = cos( "4 )

results by sampling the continuous-time signal

Since w = nT and T = 1/1000 seconds, the signal frequency could be:

or possibly:

4.3. (a) Since x[n] = xc(nT),

(b) No. For example, since

T can be 7/12000.

4.4. (a) Letting T = 1/100 gives

7r

no = - . 1000 = 25011" 4

11"

no = (211" + -) . 1000 = 225011". 4

1I"n

= 40001l"nT

3

T =

1 12000

11" 711"

cos( "3n) == cos ( "3n),

x[n] = xc(nT)

= sin ( 201l"n l~) + cos (401l"n 1~0 )

= sin (~n) + cos C;n)

(b) No, another choice is T = 11/100:

x[n] = Xc (nT)

= sin ( 201Tn 1~) + cos ( 40rrn 11;0 )

= sin (l1;n) + cos (22;n)

= sin (1I"5n) + cos C;n)

4.5. A plot of H(ejW) appears below.

96

..

-n18 nl8

(a)

Xe(t) = 0, ,11112: 211"·5000

The Nyquist rate is 2 times the highest frequency. ::;. T = lO,~OO sec. This avoids all aliasing in the CjD converter.

t

,

(b)

1 10kHz
=
T
w = T11
7r _1_11
=
8 10,000 e
11c = 211" . 625radjsec
Ie = 625Hz (c)

1 20kHz
=
T
w = T11
7r _1_11
=
8 20,000 e
11c = 27r' 1250radjsec
Ie = 1250Hz 4.6. (a) The Fourier transform of the filter impulse response

Hc(11) = i: hc(t)e-jOt dt

= 1000 a-ate-jOt dt 1

=

a+j11

So, we take the magnitude

1~{jn)1 l/a

97

(b) Sampling the filter impulse response,}d~J:r),:!he:tiFtmtfRpe filter is described by

00

Hd(ejW) = 2:::: Te-anT e-jwn n=O

T

=

Taking the magnitude of this response

. T

IHd(eJW)1 = 1 •

(1 - 2e-aT cos(w) + e-2aTp

Note that the frequency response of the discrete-time filter is periodic, with period 27r.

IHd(e jU)1

-41t

-21t

o

(c) The minimum occurs at W = 7r. The corresponding value of the frequency response magnitude is

T

1. (1 + 2e-~T + e-2aT) 2

T

=

1 + e=?'

T

~ <

I

4.7. The continuous-time signal contains an attenuated replica of the original signal with a delay of 'd, xc(t) = sc(t) + o:sc(t - 'd)

(a) Taking the Fourier transform of the analog signal:

XcUO) = Sc(jO) . (1 + o:e-jTdO)

Note that XcUO) is zero for 101 > 'KIT. Sampling the continuous-time signal yields the discretetime sequence, x[n]. The Fourier transform of the sequence is

X(dW) = f f: Sc(j; + j2;r)

r=-oo 00

+ -TO: ~ S (jw .27rT) -jTd{lf+~}

L..t c T +) T e .

r=-oo

Pearson Fducation(s) Pt·; Lt •• 482, f. E. Pa::parrr:i

Dc!h , .... 100C?2 ~~~KJ 221 ~,(;CC7-,7() ~

98

.. ~

(b) The desired response:

H(j0.) = { 1 + ae-jTdfl, for 10.1 ::; f

0, otherwise

Using w = 0.T, we obtain a discrete-time system which simulates the above response:

(c) We need to take the inverse Fourier transform of the discrete-time impulse response of part (b).

(i) Consider the case when Td = T:

=

2_ r (~jwn + aeiw(n-l» dw 21r J-1T

sin(rrn) asin[rr{n - l)J

_"'--..:_ + --;--'--.-.:.!.

rrn rr(n-1)

J[n] + aJ[n - 1]

h[n] =

(ii) For Td = T /2:

h[n]

= _!_ r (eiwn + aeiw(n-t» dw 2rr J-1r

sin(rrn) . asin[rr(n - !)]

= + 1

rrn rr(n-z)

[] asin[rr(n - !)] - In+ rr{n-!)

4.8. A plot of Xc(j0.) appears below.

(a) For xc(t) to be recoverable from x[n], the transform of the discrete signal must have no aliasing.

When sampling, the radian frequency is related to the analog frequency by

w=flT.

No aliasing will occur if the sampling interval satisfies the Nyquist Criterion. Thus, for the bandlimited signal, xc(t), we should select T as:

T < 1

- 2 X 104

99

(b) Assuming that the system is linear and time-invariant, the convolution sum describes the inputoutput relationship.

00

yin] = L x[kJh[n - k]

k=-oo

We are given

n
yin] = T L x[k]
k=-oo
oc
= T L x[kJu[n - kJ
k=-oo Hence, we may infer that the impulse response of the system

h[nJ = T . urn].

(c) We use the expression for yin] as given and examine the limit

n

lim yIn] = lim T· ~ x[k]

n-+oo n-+oo ~

k=-oo

00

= T· L x[k]

k=-oo

Recall the analysis equation for the Fourier transform:

00

X(dW) = L x[nje-iwn

n=-oo

Hence,

lim yin] = T· X(eiW)lw=o

n--+oo

(d) We use the result from part (c). Noting that

X(dW) = f f Xc(j; + j~r).

r=-oo

Thus, we have

. I ~ Xc(j2T7rr)

T· X(e'W) w=o = L..

r=-oo

From the given information, we seek a value of T such that:

For the final equality to be true, there must be no contribution from the terms for which r ~ O. That is, we require no aliasing at n = O. Since we are only interested in preserving the spectral component at n = 0, we may sample at a rate which is lower than the Nyquist rate. The maximum value of T to satisfy these conditions is

T < 1

- 1 x 1()4

Pearson r r;,'f;ati"rhd Pt I { 48:2, i' ,E I,! :

Delb ,- ~ ; 0092

4.9. (a) Since X(eiw) = X(ei(w-1<», X(ei"!,) is periodic with period tt . (b) Using the inverse DTFT,

x[nJ = _!_ f X(ejW)ejwndw
27r (21r)
= _!_ f X(ej(w-1f»ejwndw
27r (21f) ~,
= _!_ f X(ejw)ei(w+1f)ndw
27r (21<)
= _!_ej1<n f X(ejw)eiwndw
27r (21<)
= ( -l)nx[nJ. All odd samples of x[n] = 0, because x[n] = -x[n]. Hence x[3] = 0.

(c) Yes, y[n] contains all even samples of x[n], and all odd samples of x[n] are 0.

x[nJ = {y[n/2], n even.

0, otherwise

4.10. Use x[nJ = xc(nT), and simplify: (a) x[nJ = cos(27rn/3).

(b) x[n] = sin(47rn/3) = - sin(21rn/3)

() [] _ sin(21<n(S) C x n - 1<n/SOOO

4.11. (a) Pick T such that

x[n] = xc(nT) = sin(101rnT) = sin(1rn/4)

===? T = 1/40

There are other choices. For example, by realizing that sin(1rn/4) = sin(91rn/4), we find T = 9/40. (b) Choose T = 1/20 to make x[nJ = xc(nT). This is unique.

4.12. (a) Notice first that H(eiw) = 10jw, -1r S W < 1r.

(i) After sampling,

x[nJ 31r
= cos( Sn),
y[n] .~ 31r .~
= IH(e's )1 cos( -n + LH(eJ s »
5
31r 1r
= 61r cos( Sn + 2")
. 31r
= -61rsm( Sn)
yc(t) = -61r sin(61rt). (ii) After sampling, x[n] = coscr;n) = cose;n), 50 again, yc(t) = -61r5in(6m).

(b) yc(t) is what you would expect from a differentiator in the first case hut not in the second case.

This is because aliasing has occurred in the second case.