15 Discrete-Time Modulation
Solutions to
Recommended Problems
S15.1
Recall that the Fourier transform of a train of impulses p(t) is P(w), as shown in
Figure S15.1-1.
P(w)
4T 27 0 27 4in
T T T T
Figure S15.1-1
Since x,(t) = x(t)p(t),
X,(w) = -2 KX(O)P(w - 0) dO
by the modulation property. Thus, X,(w) is composed of repeated versions of X(W)
centered at 27rk/T for an integer k and scaled by 1/T, as shown in Figure S15.1-2.
X,(W)
1
T
27r 7T
_7' 27T
T 2T 2T T
Figure S15.1-2
Since Xr(w) = X,(xw)H(w), it is as indicated in Figure S15.1-3.
Thus
1 1
X,(w) = X(wo) or x, = x(t)
S15-1
Signals and Systems
S15-2
S15.2
For go = ir/2, C(Q) is given as in Figure S15.2-1.
C(n)
-27r li31
T 7lI r 1 7
2
2 2 -~ I y f
Figure S15.2-1
By the modulation theorem,
9{x[n]c[n]) = 5{y[n]} = = - C(O) X(Q - 0) dO
Thus, Y(Q) is X(Q) centered on each impulse in Figure S15.2-1 and scaled by J, as
shown in Figure S15.2-2.
Y(R)
II I I I I
3If _T int 3nt 7T 3'r 21r
2 2 4 2 4 2
Figure S15.2-2
For go, = ir/4, C(Q) is given as in Figure S15.2-3.
C(92)
7T T 7IT
7_it it itf7
4 4 4 4
Figure S15.2-3
Thus, Y(Q) in this case is as shown in Figure S15.2-4.
Discrete-Time Modulation / Solutions
S15-3
Y(u)
1
2
2
-2-r 21T
4 4 4 2 4
Figure S15.2-4
S15.3
From the lecture we know that the system in Figure S15.3-1 is equivalent to a filter
with response centered at 9 = 1r, as shown in Figure S15.3-2.
x[n] H(9) yyn]
(-1)"
Figure S15.3-1
H'(92)
S1
-
4 r 14
Figure S15.3-2
Therefore, the total response is the sum of H'(Q) and H(Q), shown in Figure
S15.3-3.
I
I
a
I
II I
I
1' 3iw _i
4 4
Figure S15.3-3
Signals and Systems
S15-4
As an example, consider x[n] with Fourier transform X(Q) as in Figure S15.3-4.
X(92)
-7r if
Figure S15.3-4
Then, after multiplication by (-1)", the resulting signal has the Fourier transform
given in Figure S15.3-5.
F eIi
Figure S15.3-5
After filtering by H(Q), the resulting signal has the spectrum given in Figure
S15.3-6.
Figure S15.3-6
Finally, multiplying by (- 1)' again yields the spectrum in Figure S15.3-7.
-IT
A i
E
Figure S15.3-7
Discrete-Time Modulation / Solutions
S15-5
Thus, the spectrum of y[n] is given by the sum of the spectrum in Figure S15.3-8
and X(Q), as shown in Figure S15.3-8.
Y(n)
FI S
Figure S15.3-8
S15.4
(a) P(Q) is composed of impulses spaced at 21r/N, where N is the period of the
sequence. In this case N = 2. The amplitude is 2 1ra,:
1
ak 2 T p[n]e j2kn2
n=O
= 1[le -j(2wkO/2) + Oe-j(2 kl/ 2 )I 2
Thus, P(Q) is as shown in Figure S15.4-1.
We now perform the periodic convolution of X(Q) with P(Q) and scale by
1/(21r) to obtain the spectrum in Figure S15.4-2.
Signals and Systems
S15-6
(b) To recover x[n] from y[n], we can filter y[n] with H(Q) given as in Figure
S15.4-3.
(c) Using p[n] we can send only every other sample of x 1 [n]. Similarly, we can send
every other sample of x2[n] and interleave them over one channel. Note, how
ever, that we can do this only because X(Q) is bandlimited to less than 7r/2.
S15.5
We note that s(t) is a periodic signal. Therefore, S(w) is composed of impulses cen
tered at (21rk)/T for integer k. The impulse at w = 0 has area given by 2rao, where
ao is the zeroth Fourier series coefficient of s(t):
ao =
T r
s(t) dt = j 1dt
A/2
-sis
=
A
T
Thus, S(w) is as shown in Figure S15.5-1.
2na_1
2 irA
T
1 2ra
2 27
TrT
Figure S15.5-1
The Fourier transform of x(t)s(t), denoted by R(w), is given by
R(w) = f 1 X(O)S(w - 0) dO = T, a.X W - 27rn
Discrete-Time Modulation / Solutions
S15-7
If X(w) = 0 for I I > w/T, then R(w) will equal (A/T)X(w) in the region I co I < w/T.
Therefore, for H(co) as in Figure S15.5-2, the signal y(t) = x(t).
H(w)
T T
Figure S15.5-2
Solutions to
Optional Problems
S15.6
(a) Consider the labeling of the system in Figure S15.6.
x [n] X r [n] h [n] Iv [n] X
Xo y [n]
[n] r1 $2 [n]
Figure S15.6
r[n] = $1[n]x[n]
v[n] = E r[kjh[n - k] = $[kjx[kjh[n
0 - k]
k=- k=
y[n] = v[n] 2[n] 2[n] h[n - k]4 1[k]x[k]
k=-w
Suppose x1[n] = ax[n]. Then
yi[n] = 02[n] j h[n - k]# 1[k]ax[k] = ay[n]
k= -w
Now let x 2[n] = x1[n] + x 0 [n]. Then
y2[n] = 02[n] E h[n - k]41[k](x 1 [k] + xo[k]) = y 1[n] + yo[n]
k= -w
and the system is linear.
Signals and Systems
S15-8
If 4 1[n] = b[n], then
y[n] = 0 2 [n] 3 h[n - k]b[k]x[k] = 02[n]h[n]x[O]
k=-wo
If x[n] is shifted so x 1 [n] = x[n - 11, then
y1n] = 4 2[n]h[n]x1[0] = 4 2[n]h[n]x[-1] # y[n - 1]
and the system is not time-invariant.
(b) From part (a),
y[n] = Z ( h[n - k]z-x[k]
k=-wT
Let x[n - m] = xijn]. Then
y 1 [n] = z" h[n - k]z-kx1[k] = z- h[n - k]z-kx[k - m]
k= -w k= -w
Letp = k - m, k = p + m. Then
y1[n] = z" h[(n - m) - p]z-'-x[p]
= z m h[(n - m) - p]z-x[p]
p=-w0
= y[n - m]
Therefore, the system is time-invariant.
S15.7
In general, w(t) is recoverable from w,(t) if W,(w) contains repeated versions of
W(w) that do not overlap, i.e., that have no aliasing, as shown in Figure S15.7.
W,(C)
27r c _r 27r
T T T
W(W)
Figure S15.7
Since W(c) is repeated with period 27r/T, the largest frequency component of W(W),
Co,must be less than or equal to ir/T. From the modulation property,
1
W(W) = - X(W) * X 2(W)
27r
Discrete-Time Modulation / Solutions
S15-9
Thus, since the length of a convolution of two signals is the sum of the individual
lengths,
c= W1 ± W2
From the preceding observations,
ir
- > W1 + W2 or T<
T W1 + A2
S15.8
(a) If a, = -12/21r, then the portion of X(Q) around Qj will be modulated down to
about 0 = 0 and then filtered by H(Q). We now need to reshift the spectrum
back to its original position. Therefore, we need to modulate by e2'i', or # =
+Qj/27r.
(b) Consider i = 0, 1. Then the corresponding filters are as given in Figure S15.8.
-20 0 E20 2r . 2 2 + E2
0
N N N 0
Figure S15.8
For no overlap and complete coverage of the frequency band, we need
27r
0 - 0, or 90 =
N N
S15.9
(a) Since s(t) is periodic in T, S(w) will consist of impulses located at 27rk/T. See
Figure S15.9-1.
Signals and Systems
S15-10
If Jos(t) = 0, then the spectrum looks like Figure S15.9-2.
S(W)
_ ___ t wt
6n _4r _2n 0 27r 47r 67r
T T T T T T
Figure S15.9-2
Of course, other impulses may also be zero.
(b) Y(w) will be equal to a sum of the shifted and scaled versions of X(w). Spe
cifically,
S(27)x 2 9n
Y(w) = f X()S(w - 0) d = 1
27r -n 1 T T /(S15.9-1)
an X
where an is the nth Fourier series coefficient of one period of s(t). For some
region Y(w) to be zero, successive terms in the sum in eq. (S15.9-1) cannot over
lap. Thus, the maximum T is such that /irT= w,, or T = 7r/we.
(c) In general, we need to find some n such that an # 0. Then we use an ideal real
bandpass filter to isolate the nth term of the sum in eq. (S15.9-1). The resulting
signal r(t) has Fourier transform R(w) given by
R(c) = anX 2 Tr) + aX o + 2)
Let an = rne"". Then r(t) can be thought of as
r(t) = x(t) + On
2r COS
I,.o (T
(remember the effect of modulating by a cosine signal). Suppose we multiply
r(t) by
I cos + O
Then
q(t) = r(t) -cos 2T + 0i = x(t) 2 cos 2(27rT
T + O
= x(t) 1 + cos 42rnt+ 20.)
If we now use a lowpass filter with cutoff wi/T, we get x(t). If we had picked the
smallest n such that an # 0, we could have avoided the bandpass filtering
because higher harmonics are eliminated by the lowpass filter.
Discrete-Time Modulation / Solutions
S15-11
S15.10
(a) Y(Q) will consist of repeated versions of X(Q) centered at (21r/5) + 21rk and
scaled by 1. Thus, Y(Q) is as shown in Figure S15.10-1.
Y(n)
1
2
-iT 2nr_ 2 7
5 5
Figure S15.10-1
(b) Z(Q) will consist, in turn, of repeated versions of Y(Q), centered at (47r/5) +
21rk and scaled by 2, as shown in Figure S15.10-2.
Z(W)
-ote thatthe 2i 2 4
5 5 5 5
Figure S15.10-2
Note that the version of Y(Q) centered at 67r/5 contributes to the spectrum
between -3.7r/5 and ir.
(c) Two possible choices are given in Figures S15.10-3 and S15.10-4.
MIT OpenCourseWare
[Link]
Resource: Signals and Systems
Professor Alan V. Oppenheim
The following may not correspond to a particular course on MIT OpenCourseWare, but has been
provided by the author as an individual learning resource.
For information about citing these materials or our Terms of Use, visit: [Link]
