Homework 2 Solution

EE 140: Introduction to Communication Systems

Problem 1 (12 points)

For a lowpass signal with a bandwidth of 6000 Hz

(a) What is the minimum sampling frequency for perfect reconstruction of the signal?
(b) What is the minimum required sampling frequency if a guard band of 2000 Hz is required?
(c) What is the minimum required sampling frequency and the value of K for perfect recon-
struction if the reconstruction filter has the following frequency response



 K |f | < 7000
H(f ) = K− K |f |−7000
3000
7000 < |f | < 10000


 0 otherwise

Solution 1

(a) For no aliasing to occur we must sample at the Nyquist rate

fs = 2 · 6000 samples/sec = 12000 samples/sec

(b) With a guard band of 2000 Hz

fs − 2W = 2000 =⇒ fs = 14000

(c) The reconstruction filter should not pick-up frequencies of the images of the spectrum
X(f ). The nearest image spectrum is centered at fs and occupies the frequency band
[fs − W, fs + W ]. Thus the highest frequency of the reconstruction filter (= 10000) should
satisfy

1
 10000 ≤ fs − W =⇒ fs ≥ 16000

For the value fs = 16000, K should be such that

K · fs = 1 =⇒ K = (16000)−1

Problem 2 (13 points)

The lowpass signal x(t) with a bandwidth of W is sampled at the Nyquist rate and the
signal
∞
(−1)n x(nTs )δ(t − nTs )
P
x1 (t) =
n=−∞

is generated. Find the Fourier transform of x1 (t).

Solution 2
∞ ∞
(−1)n x(nTs )δ(t − nTs ) = x(t) (−1)n δ(t − nTs )
P P
x1 (t) =
n=−∞ n=−∞
∞
P ∞
P
= x(t)[ δ(t − 2lTs ) − δ(t − Ts − 2lTs )]
l=−∞ l=−∞
∞ ∞
X1 (f ) = X(f ) ∗ [ 2T1 s l 1 l
)e−j2πf Ts ]
P P
δ(f − 2Ts
) − 2Ts
δ(f − 2Ts
l=−∞ l=−∞
∞ ∞
1
P l 1
P l −j2π 2Tl Ts
= 2Ts
X(f − 2Ts
) − 2Ts
X(f − 2Ts
)e s
l=−∞ l=−∞
∞ ∞
1 l 1 l
)(−1)l
P P
= 2Ts
X(f − 2Ts
) − 2Ts
X(f − 2Ts
l=−∞ l=−∞
∞
1 1 l
P
= Ts
X(f − 2Ts
− Ts
)
l=−∞

Problem 3 (10 points)

Two binary random variables X and Y are distributed according to the joint distribution
p(X = Y = 0) = p(X = 0, Y = 1) = p(X = Y = 1) = 13 . Compute H(X), H(Y ), H(X|Y ), H(X, Y )
and I(X; Y ).

2
Solution 3

The marginal probabilities are given by

X 2
P (X = 0) = p(X = 0, Y = k) = p(X = 0, Y = 0) + p(X = 0, Y = 1) =
k
3

X 1
p(X = 1) = p(X = 1, Y = k) = p(X = 1, Y = 1) =
k
3

X 1
p(Y = 0) = (X = k, Y = 0) = p(X = 0, Y = 0) =
k
3

X 2
p(Y = 1) = (X = k, Y = 1) = p(X = 0, Y = 1) + p(X = 1, Y = 1) =
k
3

Hence,
X 1 1 2 2
H(X) = − pi log2 pi = −( log2 + log2 ) = .9183
3 3 3 3

X 1 1 2 2
H(Y ) = − pi log2 pi = −( log2 + log2 ) = .9183
3 3 3 3

X 1 1 1 1 1 1
H(X, Y ) = − pi log2 pi = −( log2 + log2 + log2 ) = 1.5850
3 3 3 3 3 3

H(X|Y ) = H(X, Y ) − H(Y ) = 1.5850 − 0.9183 = 0.6667

I(X; Y ) = H(X) − H(X|Y ) = 0.9183 − 0.6667 = 0.2516

Problem 4 (10 points)

A signal can be modeled as a lowpass stationary process X(t) whose PDF at any time t0

3
is given in Figure 1.

Figure 1

The bandwidth of this process is 5 KHz, and it is desired to transmit it using a PCM
system.
(a) If sampling is done at the Nyquist rate and a uniform quantizer with 32 levels is employed,
what is the resulting SQNR? What is the resulting bit rate?
(b) If the available bandwidth of the channel is 40 KHz, what is the highest achievable
SQNR?

Solution 4

(a)

0 Z 2
−x + 2
Z
2 x+2
2
E[X (t)] = x( )dx + x2 ( )dx
−2 4 0 4
1 1 2 1 1 2
= ( x4 + x3 )|0−2 + (− x4 + x3 )|20
4 4 3 4 4 3
2
=
3

Hence,
2 2
3 × 4ν × 3
3 × 45 × 3
SQNR = = = 512 = 27.093(db)
x2max 22
(b)
If the available bandwidth of the channel is 40 KHz, then the maximum rate of trans-

4
mission is ν = 40/5 = 8. In this case the highest achievable SQNR is

2
3 × 48 × 3
SQNR = = 32768 = 45.154(db)
22

Problem 5 (15 points)

Show that a pulse having the raised cosine spectrum given by Equation (1) satisfies the
Nyquist criterion given by Equation (2) for any value of the roll-off factor α.


 T , 0 ≤ |f | ≤ (1 − α)/2T

T πT 1−α 1−α 1+α

Xrc (f ) = [1 + cos( (|f | − ))] , ≤ |f | ≤ (1)
 2 α 2T 2T 2T
1+α


|f | ≥

0 ,
2T

∞
X m
X(f + )=T (2)
m=−∞
T

Solution 5

The pulse x(t) having the raised cosine spectrum is

cos(παt/T )
x(t) = sinc(t/T )
1 − 4α2 t2 /T 2

The function sinc(t/T ) is 1 when t = 0 and 0 when t = nT . On the other hand


cos(παt/T ) 1
 t=0
g(t) = 2 2 2
=
1 − 4α t /T 6 0
bounded t =

The function g(t) needs to be checked only for those values of t such that 4α2 t2 /T 2 = 1
or αt = T2 .

5
 However,
cos(παt/T ) cos( π2 x)
lim 2 2 2
= lim
αt→ T2 1 − 4α t /T x→1 1 − x

and by using L’Hospital’s rule

cos( π2 x) π π π
lim = lim sin( x) = < ∞
x→1 1 − x x→1 2 2 2

Hence, 
1 n=0
x(nT ) =
0 n 6= 0

meaning that the pulse x(t) satisfies the Nyquist criterion.

Problem 6 (15 points)

Binary PAM is used to transmit information over an unequalized linear filter channel.
When a = 1 is transmitted the noise-free output of the demodulator is



 0.3 , m=1


0.9

, m=0
xm = (3)
0.3

 , m = −1


0

, otherwise

(a) Design a three-tap zero-forcing linear equalizer so that the output is


1 , m =0
qm = (4)
0 , m=±1

(b) Determine qm for m = ±2, ±3, by convolving the impulse response of the equalizer with
the channel response

Solution 6

(a)

6
 The equivalent discrete-time impulse response of the channel is

1
X
h(t) = hn δ(t − nT ) = 0.3δ(t + T ) + 0.9δ(t) + 0.3δ(t − T )
n=−1

If by cn we denote the coefficients of the FIR equalizer, then the equalized signal is

1
X
qm = cn hm−n
n=−1

which in matrix notation is written as

    
0.9 0.3 0 c−1 0
    
 0.3 0.9 0.3   c0  =  1 
    
0 0.3 0.9 c1 0

The coefficients of the zero-force equalizer can be found by solving the previous matrix
equation.
Thus,
   
c−1 −0.4762
   
 c0  =  1.4286 
   
c1 −0.4762

(b)
The values of qm for m = ±2, ±3 are given by

1
X
q2 = cn h2−n = c1 h1 = −0.1429
n=−1

1
X
q−2 = cn h−2−n = c−1 h−1 = −0.1429
n=−1

1
X
q3 = cn h3−n = 0
n=−1

7
 1
X
q−3 = cn h−3−n = 0
n=−1

Problem 7 (10 points)

Find the capacity of an additive white Gaussian noise channel with a bandwidth of 1
N0
MHz, power of 10W, and noise power-spectral density of 2
= 10−9 W/Hz.

Solution 7

The SNR is
2P P 10
SN R = = = −9 6
= 104
N0 2W 2W 10 × 10
Thus the capacity of the channel is

P
C = W log2 (1 + ) = 106 log2 (1 + 10000) ≈ 13.2879 × 106 bits/sec
N0 W

Problem 8 (15 points)

For the channel shown in Figure 2, find the channel capacity and the input distribution that
achieves capacity.

Figure 2

8
Solution 8

The capacity of the channel of the channel is given by

C = max I(X; Y ) = max[H(Y ) − H(Y |X)]

p(x) p(x)

Let the probability of the inputs C, B and A be p, q and 1 − p − q respectively. From

the symmetry of the nodes B, C, we expect that the optimum distribution p(x) will satisfy
p(B) = p(C) = p. The entropy H(Y |X) is given by
X
H(Y |X) = p(x)H(Y |X = x) = (1 − 2p)H(Y |X = A) + 2pH(Y |X = B)
= 0 + 2ph(0.5) = 2p

The probability mass function of the output is

X
p(Y = 1) = p(x)p(Y = 1|X = x) = (1 − 2p) + p = 1 − p

X
p(Y = 2) = p(x)p(Y = 2|X = x) = 0.5p + 0.5p = p

Therefore,
C = max[H(Y ) − H(Y |X)] = max(h(p) − 2p)
p p

To find the optimum value of p that maximizes I(X; Y ), we set the derivative of C with
respect to p equal to zero. Thus,

∂C 1 −1
= 0 = − log2 (p) − p + log2 (1 − p) − (1 − p) −2
∂p p ln(2) (1 − p) ln(2)
= log2 (1 − p) − log2 (p) − 2

and therefore
1−p 1−p 1
log2 =2⇒ =4⇒p=
p p 5

9
The capacity of the channel is

1 2
C = h( ) − = 0.7219 − 0.4 = 0.3219bits/transmission
5 5

10