EE160: Analog and Digital Communications

SOLVED PROBLEMS

Copyright (c) 2005. Robert Morelos-Zaragoza. San Jos

e State University
1

Digital communication systems

1. With reference to Fig. 1.2 of the textbook, illustrating the basic elements of a digital communication system, answer the following questions:
(a) What is source coding?
(b) What is the purpose of the channel encoder and channel decoder?
(c) What is the purpose of the digital modulator and digital demodulator?
(d) Explain how is the performance of a digital communication system measured.
Solution:
(a) Source coding is the process of eciently converting the output of either an analog or a
digital source, with as little or no redundancy, into a sequence of binary digits.
(b) The channel encoder introduces, in a controlled (structured) manner, certain amount
of redundancy that can be used at the receiver to overcome the eects of noise and
interference encountered in the transmission of the signal through the channel. This
serves to increase the reliability of the received data and improves the quality of the
received signal. The channel decoder attempts to reconstruct the original information
sequence from knowledge of the code used by the channel encoder, the digital modulation
scheme and the redundancy contained in the received sequence.
(c) The digital modulator serves as the interface to the communications channel. Its primary purpose is to map the information sequence into signal waveforms. The digital
demodulator processes the corrupted transmitted waveform and reduces each waveform
to a single number that represents an estimate of the transmitted data symbol. If this
number is quantized into more levels that those used in the modulator, the demodulator
is said to produced a soft output. In this case, the channel decoder is known as a softdecision decoder. Otherwise, the demodulator produces hard outputs that are processed
by a hard-decision decoder.
(d) The performance of a digital communication system is typically measured by the frequency with which errors occur in the reconstructed information sequence. The probability of a symbol error is a function of the channel code and modulation characteristics,
the waveforms used, the transmitted signal power, the characteristics of the channel
e.g., noise power and the methods of demodulation and channel decoding.
2. What are the dominant sources of noise limiting performance of communication systems in
the VHF and UHF bands?
Solution: The dominant noise limiting performance of communication systems in the VHF
and UHF bands is thermal noise generated in the front end of the receiver.
3. Explain how storing data on a magnetic or optical disk is equivalent to transmitting a signal
over a radio channel.
Solution: The process of storing data on a magnetic tape, magnetic disk or optical disk is
equivalent to transmitting a signal over a wired or wireless channel. The readback process
and the signal processing used to recover the stored information is equivalent to the functions
performed by a communications system to recover the transmitted information sequence.

4. Discuss the advantages and disadvantages of digital processing versus analog processing. Do
a web search. An interesting, albeit non-technical, discussion was found at
http://www.usatoday.com/tech/bonus/2004-05-16-bonus-analog x.htm
Solution: A digital communications system does not accumulate errors. Analog signals are
prone to interference and noise. There is no equivalent in an analog system to the correction
of errors. However, a digital system degrades the quality of the original signal thorugh
quantization (analog-to-digital conversion). Also, a digital system requires more bandwidth
than an analog system and, in general, relatively complex synchronization circuitry is required
at the receiver. Analog systems are very sensitive to temperature and component value
variations. It should be noted that no digital technology is used today in the front end of
a transmitter and receiver (RF frequency bands of 1GHz and above), where mixers, channel
lters, ampliers and antennas are needed. The world today is still a mix of analog and digital
components and will continue to be so for a long time. A key feature of digital technology
is programmability, which has resulted in new concepts, such as software-dened radios and
cognitive radio communications systems.

Fourier analysis of signals and systems

5. Show that for a real and periodic signal x(t), we have




a0

n
+
an cos 2 t ,
xe (t) =
2
T0
n=1



n
bn sin 2 t ,
xo (t) =
T0
n=1

where xe (t) and xo (t) are the even and odd parts of x(t), dened as
x(t) + x(t)
,
2
x(t) x(t)
.
2

xe (t) =
xo (t) =

Solution: It follows directly from the uniqueness of the decomposition of a real signal in an
even and odd part. Nevertheless for a real periodic signal


n
n
a0

+
an cos(2 t) + bn sin(2 t)
x(t) =
2
T0
T0
n=1
The even part of x(t) is
xe (t) =
=

x(t) + x(t)
 2


n
n
1
a0 +
an (cos(2 t) + cos(2 t))
2
T0
T0
n=1

n
n
+bn (sin(2 t) + sin(2 t))
T0
T0

n
a0

+
an cos(2 t)
2
T0
n=1

The last is true since cos() is even so that cos() + cos() = 2 cos whereas the oddness of
sin() provides sin() + sin() = sin() sin() = 0.
Similarly, the odd part of x(t) is
x(t) x(t)
2

n
bn sin(2 t)
=
T0

xo (t) =

n=1

6. Determine the Fourier series expansion of the sawtooth waveform, shown below
x(t)
1

t
-3T

-2T

-T

2T

3T

-1

Solution: The signal is periodic with period 2T . Since the signal is odd we obtain x0 = 0.
For n
= 0
 T
 T
n
1
1
t j2 n t
j2 2T
t
2T dt
e
x(t)e
dt =
xn =
2T T
2T T T
 T
n
1
tej T t dt
=
2
2T T


T 2 j n t T
jT j n t
1
T +
T
te
e
=

2T 2 n
2 n2
T
 2

1 jT jn
T 2 jn jT 2 jn
T 2 jn
e
e
=
+ 2 2e
+
2 2e
2T 2 n
n
n
n
j
(1)n
=
n

7. By computing the Fourier series coecients for the periodic signal
n= (t nTs ), show
that

1
jn 2t
(t nTs ) =
e Ts .
T
s
n=
n=
Using this result, show that for any signal x(t) and any period Ts , the following identity holds
 

2t
1

n
x(t nTs ) =
X
ejn Ts .
Ts n=
Ts
n=
From this, conclude the following relation, known as Poissons sum formula:
 

1

n
x(nTs ) =
X
.
Ts n=
Ts
n=
4

Solution:

n
1
x(t) 
ej2 Ts t
Ts
n=
n=


1 1
n
F
(f )
X(f )
Ts
T
s
n=




n
n
1 1

F
X
(f )
Ts
Ts
Ts
n=



n
1

n
X
ej2 Ts t
Ts n=
Ts

x(t nTs ) = x(t) 

n=

=
=

(t nTs ) =

If we set t = 0 in the previous relation we obtain Poissons sum formula

 

1

n
x(nTs ) =
x(mTs ) =
X
T
T
s n=
s
n=
m=

8. Find the Fourier transform P1 (f ) of a pulse given by

 
t
,
p1 (t) = sin(8t)
2
where


1, |t| 12 ;
(t) =
0, otherwise.

and shown in the gure below:

p1(t)

t
-1

(Hint: Use the convolution theorem.)

Solution: Using the Fourier transform pair (t) sinc(f ) and the time scaling property
(from the table of Fourier transform properties), we have that
 
t
2 sinc(2f ).

2
1
[(f + f0 ) + (f f0 )] and the convolution property, we
From the pair sin(2f0 t) 2j
arrive to the result
P1 (f ) = j {sinc [2(f + 4))] sinc [2(f 4))]} .

9. Determine the Fourier series expansion of the periodic waveform given by

p(t) =

p1 (t 4n),

n=

and shown in the gure below:

p(t)

t
-3

-5

-1

(Hint: Use the Fourier transform P1 (f ) found in the previous problem, and the following
equation to nd the Fourier coecients: pn = T1 F1 ( Tn ).)
Solution: The signal p(t) is periodic with period T = 4. Consequently, the Fourier series
expansion of p(t) is




pn exp j t n ,
p(t) =
2
n=
where

 n

 n

n
1 
1
=
sinc 2( + 4)) sinc 2( 4)) .
pn = P1
4
4
4j
4
4

10. Classify each of the following signals as an energy signal or a power signal, by calculating the
energy E, or the power P (A, , and are real positive constants).
(a) x1 (t) = A | sin(t + )|.

(b) x2 (t) = A / + jt, j = 1.

(c) x3 (t) = At2 et/ u(t).
(d) x4 (t) = (t/ ) + (t/2 ).
Solution:
(a) Power. The signal is periodic, with period /, and

P1 =

(b) Neither:

A2 | sin(t + )|2 dt =

E2 = lim

T T

and
1
P2 = lim
T 2T
6

A2
.
2

(A )2

dt ,
+ jt jt

(A )2

dt = 0.
2 + t2

(c) Energy:


E3 =

(d) Energy:

A2 t4 exp(2t/ ) dt =


E4 = 2

/2

(2)2 dt +
0

3A2 5
.
4

(1)2 dt

= 5.

/2

11. Sketch or plot the following signals:

(a) x1 (t) = (2t + 5)
(b) x2 (t) = (2t + 8)
(c) x3 (t) = (t 12 ) sin(2t)
(d) x4 (t) = x3 (3t + 4)
(e) x5 (t) = ( 3t )
Solution:
x1(t)

x3(t)

1
1

-5/2

1/2

-1
x4(t)

x2(t)
1

1/2

4/3

-1

x5(t)
1

t
-3/2

3/2

12. Classify each of the signals in the previous problem into even or odd signals, and determine
the even and odd parts.
Solution:
The signal xi (t), for 1 i 4, is neither even nor odd. The signal x5 (t) is even symmetric.
7

For each signal xi (t), with 1 i 4, the gures below are sketches of the even part xi,e (t)
and the odd part xi,o (t). Evidently, x5,e = x5 (t) and x5,o (t) = 0.
x1,e(t)
1/2

t
-5/2

5/2

1/2

1/2

x1,o(t)

1/2

t
-5/2

5/2
-1/2

1/2
1/2

x2,e(t)
1/2
t
-4

1/2

1/2

x2,o(t)
1/2
-4

t
-1/2

4
1/2

1/2

x3,e(t)
1/2

t
1

-1
-1/2

x3,o(t)
1/2

t
-1

1
-1/2

x4,e(t)
1/2

t
-4/3

-1

4/3

-1/2

x4,o(t)
1/2

t
-4/3

-1

1
-1/2

4/3

13. Generalized Fourier series

(a) Given the set of orthogonal functions


4 [t (2n 1)T /8]
,
n (t) =
T

n = 1, 2, 3, 4,

sketch and dimension accurately these functions.

(b) Approximate the ramp signal
t
x(t) =
T

t T /2
T

by a generalized Fourier series using these functions.

(c) Do the same for the set


2 [t (2n 1)T /4]
,
n (t) =
T

n = 1, 2.

(d) Compare the integral-squared error (ISE) N for both parts (b) and (c). What can you
conclude about the dependency of N on N ?
Solution:
(a) These are unit-amplitude rectangular pulses of width T /4, centered at t = T /8, 3T /8, 5T /8,
and 7T /8. Since they are spaced by T /4, they are adjacent to each other and ll the
interval [0, T ].
(b) Using the expression for the generalized Fourier series coecients,

1
x(t)n (t)dt,
Xn =
cn T


where

|n (t)|2 dt =

cn =
T

we have that

1
3
, X2 = ,
8
8
Thus, the ramp signal is approximated by
X1 =

x4 (t) =

4


n=1

Xn n (t) =

T
,
4

5
X3 = ,
8

X4 =

7
.
8

1
3
5
7
1 (t) + 2 (t) + 3 (t) + 4 (t),
8
8
8
8

This is shown in the gure below:

1
x(t)
x4(t)
0.5

t
T/2

10

0 t T.

(c) These are unit-amplitude rectangular pulses of width T /2 and centered at t = T /4 and
3T /4. We nd that X1 = 1/4 and X2 = 3/4. The approximation is shown in the gure
below:
1
x(t)

x2(t)

0.5

t
T

T/2

(d) Use the relation


|x(t)| dt
2

N =
T

and note that

N

cn |Xn |2 ,

n=1

 2
T
t
dt = .
T
3

|x(t)| dt =
2

It follows that the ISE for part (b) is given by



T
1
9
25 49
T
+
+
+
= 5.208 103 T,
4 =
3
4 64 64 64 64
and for part (c)
T
T
2 =
3
2

1
9
+
16 16

= 2.083 102 T.

Evidently, increasing the value of N decreases the approximation error N .

14. Show that the time-average signal correlation

1
Rx ( ) = lim
T 2T

x(t)x(t + )dt
T

can be written in terms of a convolution as

R( ) = lim

Solution: Note that:

x(t)  x(t) =

1
[x(t)  x(t)]t= .
2T


x( )x(t ) d =
11

x(u)x(t + u) du,

where u = . Rename variables to obtain

1
R( ) = lim
T 2T

x()x( + ) d.
T

15. A lter has amplitude and phase responses as shown in the gure below:
|H(f)|
4

f
-100

-50

50

100

H(f)
/2
f
-100

50

-50

100

-/2

Find the output to each of the inputs given below. For which cases is the transmission
distortionless? For the other cases, indicate what type of distorsion in imposed.
(a) cos(48t) + 5 cos(126t)
(b) cos(126t) + 0.5 cos(170t)
(c) cos(126t) + 3 cos(144t)
(d) cos(10t) + 4 cos(50t)
Solution: Note that the four input signals are of the form xi (t) = a cos(2f1 t)+ b cos(2f2 t),
for i = 1, 2, 3, 4. Consequently, their Fourier transforms consist of four impulses:
Xi (f ) =

a
b
[(f + f1 ) + (f f1 )] + [(f + f2 ) + (f f2 )] ,
2
2

With this in mind, we have the following

(a) Amplitude distortion; no phase distortion.
(b) No amplitude distortion; phase distortion.
(c) No amplitude distortion; no phase distortion.
(d) No amplitude distortion; no phase distortion.
12

i = 1, 2, 3, 4.

16. Determine the Fourier series expansion of the following signals:

(a) x4 (t) = cos(t) + cos(2.5t)
(b) x8 (t) = | cos(2f0 t)|
(c) x9 (t) = cos(2f0 t) + | cos(2f0 t)|
Solution:
(a) The signal cos(t) is periodic with period T1 = 2 whereas cos(2.5t) is periodic with
period T2 = 0.8. The ratio T1 /T2 = 5/2 and LCM (2, 5) = 10. It follows then that
cos(t) + cos(2.5t) is periodic with period T = 2(2) = 5(0.8) = 4. The trigonometric
Fourier series of the even signal cos(t) + cos(2.5t) is
cos(t) + cos(2.5t) =
=

n=1

n=1

By equating the coecients of cos( n2 t)

unless n = 2, 5 in which case a2 = a5 =
other values of n.

n cos(2

n
t)
T0

n
n cos( t)
2

of both sides we observe that an = 0 for all n

1. Hence x4,2 = x4,5 = 12 and x4,n = 0 for all

(b) The signal x8 (t) is real, even symmetric, and periodic with period T0 =
x8,n = a8,n /2 or
 1
4f0
cos(2f0 t) cos(2n2f0 t)dt
x8,n = 2f0
4f1


= f0

1
4f0

4f1

cos(2f0 (1 + 2n)t)dt + f0

1
4f0

1
2f0 .

Hence,

cos(2f0 (1 2n)t)dt

4f1

 4f1
 4f1
1
1
sin(2f0 (1 + 2n)t) 10 +
sin(2f0 (1 2n)t) 10
2(1 + 2n)
2(1 2n)
4f
4f0

 0
n
1
1
(1)
+

(1 + 2n) (1 2n)

=
=

(c) The signal x9 (t) = cos(2f0 t) + | cos(2f0 t)| is even symmetric and periodic with period
T0 = 1/f0 . It is equal to 2 cos(2f0 t) in the interval [ 4f10 , 4f10 ] and zero in the interval
[ 4f10 , 4f30 ]. Thus

x9,n = 2f0

= f0

1
4f0

4f1

cos(2f0 t) cos(2nf0 t)dt

1
4f0

4f1

cos(2f0 (1 + n)t)dt + f0

=
=

1
4f0

4f1

cos(2f0 (1 n)t)dt

 4f1
 4f1
1
1
sin(2f0 (1 + n)t) 10 +
sin(2f0 (1 n)t) 10
2(1 + n)
2(1 n)
4f0
4f0

1
sin( (1 + n)) +
sin( (1 n))
(1 + n)
2
(1 n)
2
13

Thus x9,n is zero for odd values of n unless n = 1 in which case x9,1 = 12 . When n is
even (n = 2) then


(1)
1
1
+
x9,2 =

1 + 2 1 2
17. A triangular pulse can be specied by

t + 1,
1 t 0;
(t) =
t + 1, 0 t 1.
(a) Sketch the signal

x(t) =

(t + 3n).

n=

(b) Find the Fourier series coecients, xn , of x(t).

(c) Find the Fourier series coecients, yn , of the signal y(t) = x(t t0 ), in terms of xn .
Solution:
(a) Sketch:
x(t)
1

t
-4

-3

-2

-1

(b) The signal x(t) is periodic with T0 = 3. The Fourier series coecients are obtained from
the Fourier transform, XT0 (f ), of the truncated signal xT0 (t) as
xn =

1
XT0 (f )|f = Tn .
0
T0

In this case,
xT0 (t) = (t)

Consequently,
xn =

14

XT0 (f ) = sinc2 (f ).

n
1
sinc2
.
3
3

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0
10

10

(c) Using the time-shift property of the Fourier transform, we have

yT0 (t) = xT0 (t t0 ) = (t t0 )

YT0 (f ) = XT0 (f ) ej2f t0 = sinc2 (f ) ej2f t0 ,

and it follows that

n

yn = xn ej2( 3 )t0 =

n
n
1
sinc2
ej2( 3 )t0 .
3
3

18. For each case below, sketch the signal and nd its Fourier series coecients.
(a) x(t) = cos(2t) + cos(3t). (Hint: Find T0 . Use symmetry.)
(b) y(t) = | cos(2f0 t)|. (Full-wave rectier output.)
(c) z(t) = | cos(2f0 t)| + cos(2f0 t). (Half-wave rectier output.)
Solution:
(a) The signals cos(2t) and cos(3t) are periodic with periods T1 = 1 and T2 = 23 , respectively. The period T0 of x(t) is the least common multiple of T1 and T2 :


1
6
2
= lcm (3, 2) = = 2.
T0 = lcm 1,
3
3
3
Sketch:

15

2.5
2
1.5
1

x(t)

0.5
0
0.5
1
1.5
2
2.5
2

1.5

0.5

0
t

0.5

1.5

Using Eulers formula:


1  j2t
e
+ ej2t + ej3t + ej3t .
2
Comparing (1) with the Fourier series expansion of x(t), with T0 = 2:
x(t) =

x(t) =

xn ejnt ,

n=

we conclude that

1
x2 = x3 = ,
2
and xn = 0 for all other values of n.
(b) Sketch:

0.8

y(t)

0.6

0.4

0.2

0.6

0.4

0.2

0
t/T0

16

0.2

0.4

0.6

(1)

Note that y(t) is periodic with period T1 = T0 /2. A fortunate choice of a truncated
signal yT1 (t), over an interlval of length T1 seconds, is given by
 
2t
,
yT1 (t) = cos(2f0 t)
T0
with Fourier transform (modulation property)
YT1 (f ) =
=
It follows that (with f0 =
yn =
=



T0
1
T0
[(f + f0 ) + (f f0 )] 
sinc
f
2
2
2





T0
T0
T0
sinc
(f + f0 ) + sinc
(f f0 ) .
4
2
2
1
T0 )

1
YT (f )|f = n = 2n
T1
T0
T1 1





1
1
1
sinc
(2n + 1) + sinc
(2n 1) .
2
2
2

(2)

The above result can be further simplied by using the denition of the sinc function,
sinc(x) = sin(x)
x , noticing that



+1, n = 0, 2, 4,
(2n + 1)
=
sin
2
1, n = 1, 3, 5,
= (1)n ,
and using the odd symmetry of the sine function for negative values of n. This gives
(details omitted):


1
1
(1)n
+
.
(3)
yn =

1 + 2n 1 2n
You are invited to verify that both (2) and (3) yield the same result. For example, you
can do this using Matlab with the commands:
n=-9:1:9;
subplot(2,1,1)
stem(n,0.5*(sinc((2*n+1)/2)+sinc((2*n-1)/2)))
subplot(2,1,2)
stem(n,((-1).^n/pi) .* ( (1./(2*n+1)) + (1./(1-2*n)) ) )

17

Equation (2)
0.8
0.6

yn

0.4
0.2
0
0.2
10

0
n

10

10

Equation (3)
0.8
0.6

yn

0.4
0.2
0
0.2
10

0
n

(c) The sketch of z(t) is shown in the following page. Here the period is T0 . The truncated
signal is
 
2t
,
zT0 (t) = cos(2f0 t)
T0
with Fourier transform
ZT0 (f ) =






T0
T0
T0
sinc
(f + f0 ) + sinc
(f f0 ) .
4
2
2

(Remarkably, ZT0 (f ) = YT1 (f ).) Therefore,






1
1
1
1
n
sinc
(n + 1) + sinc
(n 1) .
ZT0 (f )|f = T =
zn =
0
T0
2
2
2
As before, there is a simplication possible (but not necessary!) using the denition of
the sinc function. This gives, z1 = 12 and


1
1
(1)
+
,
zn =

1 + 2 1 2

18

n = 2,

 integer.

2
1.8
1.6
1.4

z(t)

1.2
1
0.8
0.6
0.4
0.2
0
1

0.8

0.6

0.4

0.2

0
t/T0

0.2

0.4

0.6

0.8

0.7
0.6
0.5

0.4
0.3
0.2
0.1
0
0.1

0
n

19. Sketch the signal x(t) whose Fourier series coecients are given by

1,
n = 0;

n = 2, +2;
2,
1
xn = + 4 j, n = 4;

14 j, n = +4;

0,
elsewhere.
19

Solution: We are given the Fourier series coecients. Therefore,

j2

n
T0


1 j2
e
= 1+
2

2
T0

x(t) =

xn e

n=

= 1 + cos(4f0 t) +

j2

+e

2
T0


t


1 j2
+
e
4j

4
T0

j2

4
T0


t

1
sin(8f0 t).
2

2.5

x(t)

1.5

0.5

0.5
0.5

0.4

0.3

0.2

0.1

0
t/T0

0.1

0.2

0.3

0.4

0.5

20. Modify the Matlab script example1s05.m in the web site, to compute the Fourier series
coecients xn of an even-symmetric train of rectangular pulses of duty cycle equal to 0.12
over the range 50 n 50. Attach a printout of the resulting plot.
Solution: Using the Matlab script homework3s05.m (available in the web site) we obtain:

20

0.12

0.1

0.08

xn

0.06

0.04

0.02

0.02

0.04

0.06
50

40

30

20

10

0
n

10

20

30

40

50

21. Let xn and yn denote the Fourier series coecients of x(t) and y(t), respectively. Assuming
the period of x(t) is T0 , express yn in terms of xn in each od the following cases:
(a) y(t) = x(t t0 )
(b) y(t) = x(t)
Solution:
(a) The signal y(t) = x(t t0 ) is periodic with period T = T0 .
yn

 +T0
1
j2 Tn t
0 dt
=
x(t t0 )e
T0
 t0 +T0
1
j2 Tn
0 (v + t0 )dv
x(v)e
=
T0 t0
 t0 +T0
j2 Tn t0 1
j2 Tn v
0
0 dv
x(v)e
= e
T0 t0
j2 Tn t0

= xn e

where we used the change of variables v = t t0 .

(b) The signal y(t) is periodic with period T = T0 /.
yn =
=

1
T
1
T0

+T

n
j2 T
t

y(t)e

+T0

dt =
T0

j2 Tn v

x(v)e

where we used the change of variables v = t.

21

dv = xn

T0

j2 n
t
T

x(t)e

dt

22. Determine whether these signals are energy-type or power-type. In each case, nd the energy
or power spectral density abd also the energy or power content of the signal.
(a) x(t) = et u(t), > 0
(b) x(t) = sinc(t)

(t 2n)
(c) x(t) =
n=

(d) x(t) = u(t)

1
(e) x(t) =
t
Solution:
1
and the energy spectral
(a) x(t) = et u(t). The spectrum of the signal is X(f ) = +j2f
density
1
GX (f ) = |X(f )|2 = 2
+ 4 2 f 2

Thus,
RX ( ) = F 1 [GX (f )] =

1 | |
e
2

The energy content of the signal is

EX = RX (0) =

1
2

(b) x(t) = sinc(t). Clearly X(f ) = (f ) so that GX (f ) = |X(f )|2 = 2 (f ) = (f ). The

energy content of the signal is

EX =


GX (f )df =

(f )df =

1
2

12

(f )df = 1

(c) x(t) =
n= (t 2n). The signal is periodic and thus it is not of the energy type.
The power content of the signal is


 1
1 1
1 0
2
2
|x(t)| dt =
(t + 1) dt +
(t + 1)2 dt
Px =
2 1
2 1
0


 0
 1


1 1 3
1 1 3
2
2

t +t +t  +
t t + t 
=
2 3
2 3
1

1
3

The same result is obtain if we let

SX (f ) =

n=

22

|xn |2 (f

n
)
2

with x0 = 12 , x2l = 0 and x2l+1 =

PX

2
(2l+1)

(see Problem 2.2). Then

|xn |2

n=

1
1
8

8 2
1
1
=
+ 2
+
=
4
2
4
(2l + 1)
4 96
3

l=0

(d)


EX = lim

T
2

|u1 (t)| dt = lim

2

T T
2

T
2

T 0

T
=
T 2

dt = lim

Thus, the signal is not of the energy type.

1
T T

PX = lim

T
2

T2

|u1 (t)|2 dt = lim

1
1T
=
T 2
2

Hence, the signal is of the power type and its power content is
spectral density we nd rst the autocorrelation RX ( ).
1
RX ( ) = lim
T T

T
2

T2

1
2.

To nd the power

u1 (t)u1 (t )dt

 T
2
1
dt
= lim
T T
1
1 T
= lim ( ) =
T T 2
2
Thus, SX (f ) = F[RX ( )] = 12 (f ).
(e) Clearly |X(f )|2 = 2 sgn2 (f ) = 2 and EX = limT

T
2
T2

2 dt = . The signal is not

of the energy type for the energy content is not bounded. Consider now the signal
t
1
xT (t) = ( )
t T
Then,
XT (f ) = jsgn(f )  T sinc(f T )
and
 f
2



|XT (f )|2
= lim 2 T 
sinc(vT )dv
sinc(vT )dv 
T
T
T

SX (f ) = lim

However, the squared term on the right side is bounded away from zero so that SX (f )
is . The signal is not of the power type either.

23. Consider the periodic signal depicted in the gure below.

23

x(t)
0.5

-2.5

-1

2.5

-0.5

(a) Find its Fourier transform X(f ) and sketch it carefully.

(b) The signal x(t) is passed through an LTI system with impulse response h(t) = sinc(t/2).
Find the power of the output y(t).

Solution:
(a) T0 = 5/2 and
xT0 (t) =
As a result

 
 
1
2t
t

.
2
2
5

5
XT0 (f ) = 2 sinc (2f ) sinc
4
2

5f
2


.

Fourier series coecients:

xn =
=


 

2 5
1
n
2
2 4
n sinc (n)
XT0
= 2 sinc
T0
T0
5
5
5 4
 
4
4n
3
(n) + sinc2
.
10
5
5

Fourier transform:


  
  


4
2
3
8

2
3
2 4n
2 4n
(n) + sinc
f n =
(f )+
f n
sinc
X(f ) =
10
5
5
5
10
5
5
5
n=
n=1

X(f)
0.3

f
0.4 0.8

-3

-2

-1

(b) H(f ) = 2 (2f ). Therefore, Y (f ) =

6
10 (f )

and Py =


6 2
10

= 0.36.

24. Matlab problem. This problem needs the Matlab script homework1f04.m, available in the
class web site. The script uses the fast Fourier transform (FFT) to compute the discrete
amplitude spectrum of the periodic signal x(t) = 2sin(100t) + 0.5cos(200t) cos(300t).
24

(a) Run the script homework1f04.m. To do this, you must save the le to a local directory,
change the working directory in MATLAB to that location, and enter homework1f04 at
the prompt in the command window. You will be requested to enter your student ID
number. The script produces a gure that you are required to either print or sketch.
Also, record in your solution the value of the magic number that will appear in the
command window after execution of the script.
(b) Verify the results of part (a) by computing the Fourier series coecients of x(t).
Solution:
(a)
Signal
3
2
1
0
1
2
3
4

10

15

20

25
Time (ms)

30

35

40

45

50

10

Discrete amplitude spectrum

0.8

0.6

0.4

0.2

0
10

0
n

Magic number: 0.53490560606366733

(b) x(t) is a periodic signal. The signal sin(100t) has fundametal frequency f0 = 50,
while the signals cos(200t) and cos(300t) have fundamental frequencies 2f0 = 100 and
3f0 = 150, respectively. Consequently, f0 is the fundamental frequency of x(t). Expand
x(t) using Eulers formula:
x(t) = 2 sin(100t) + 0.5 cos(200t) cos(300t)
 1  j200t
 1  j300t


e
e
+ ej200t
+ ej300t .
= j ej100t ej100t +
4
2
It follows that |x1 | = 1, |x2 | = 0.25, and |x3 | = 0.5. The script gives correctly the
three nonzero components of the discrete spectrum of x(t). We note that the amplitude
values of the Fourier series coecients are not correct, although their ratios are close
to the correct values, that is |x1 |/|x3 | = |x3 |/|x2 | = 2. This is believed to be an
artifact that results from the use of the FFT.
25

25. Determine the Fourier transform of each of the following signals:

(a) (t 3) + (t + 3)
(b) sinc3 (t)
Solution:
(a)) Using the time-shifting property of the Fourier transform,
F[x(t)] = F[(t 3) + (t + 3)]
= sinc(f ) ej2f (3) + sinc(f ) ej2f (3)
= 2 cos(6f ) sinc(f )

(b) Using the convolution property of the Fourier transform,

T (f ) = F[sinc3 (t)] = F[sinc2 (t)sinc(t)] = (f )  (f ).
Note that


(f )  (f ) =


()(f )d =

1
2

21


(f )d =

f + 12
f 12

(v)dv,

From which it follows that

For
For
For

For

3
T (f ) = 0
f ,
2
f + 1
 f+ 1
 2
2
1
9
1 2
1
3
3
T (f ) =
(v + 1)dv = ( v + v)
= f2 + f +
<f ,
2
2
2
2
2
8
1
1
 0
 f+ 1
2
1
1
T (f ) =
(v + 1)dv +
(v + 1)dv
<f ,
2
2
f 12
0
0
f + 1

 2
1
3
1 2
2
+ ( v + v)
= f 2 +
= ( v + v)
2
2
4
f 12
0

 1
1
3
9
1
1
3
1
<f ,
T (f ) =
(v + 1)dv = ( v 2 + v)
= f2 f +
2
2
2
2
2
8
f 1
f 1
2

For
Thus,

3
< f,
2

T (f ) = 0

0,

1 2
3
9

2f + 2f + 8,


T (f ) = F sinc3 (t) = f 2 + 34 ,

1 2
3
9

2f 2f + 8,

f 32
32 < f 12
12 < f 12
1
3
2 <f 2
3
2 <f

A plot of T (f ) is shown in the following gure, and was produced with Matlab script
proakis salehi 2 10 4.m, available in the web site of the class.

26

Fourier transform of sinc (t)

0.8

0.7

0.6

Amplitude

0.5

0.4

0.3

0.2

0.1

0
2.5

1.5

0.5

0
0.5
Frequency (Hz)

1.5

2.5

26. Matlab problems. These two problems needs the following three Matlab scripts: homework2af04.m,
rectpulse.m and homework2bf04.m, available in the class web site.
(a) The scripts homework2af04.m and rectpulse.m plot the amplitude spectrum of the
Fourier transform X(f ) of the signal
 
t
.
x(t) =

Run the script homework2af04.m. You will be requested to enter the width of the
pulse. Use values of equal to 0.1 and 0.2. Print or sketch the corresponding gures.
Based on the scaling property, discuss the results.

(b) The scripts homework2bf04.m uses the inverse fast Fourier transform (IFFT) to compute
numerically the signal associated with a spectrum consisting of pair of impulses:
X(f ) =

1
1
(f + F c) + (f F c) .
2
2

Run the script homework2a.m and print or sketch the corresponding gures.
27

Solution:
(a) Pulse width = 0.1:
Rectangular pulse
1
0.8
0.6
0.4
0.2
0
1

0.8

0.6

0.4

0.2

0
Time (s)

0.2

0.4

0.6

0.8

Amplitude spectrum
0.1
0.08
0.06
0.04
0.02
0
60

50

40

30

20

10
0
10
Frequency (Hz)

20

30

40

50

60

Pulse width = 0.2:

Rectangular pulse
1
0.8
0.6
0.4
0.2
0
1

0.8

0.6

0.4

0.2

0
Time (s)

0.2

0.4

0.6

0.8

Amplitude spectrum
0.2

0.15

0.1

0.05

0
30

25

20

15

10

0
5
Frequency (Hz)

28

10

15

20

25

30

The plots agree with the theoretical expression:

  
t
= sinc( f ).
F

(b)

Spectrum Amplitude

0.5
0.4
0.3
0.2
0.1
0
0.1

60

40

20

0
Normalized frequency

20

40

60

Signal Amplitude

1
0.5
0
0.5
1
10

20

30
Time (samples)

40

50

60

27. Using the convolution theorem, show that

sinc(t)  sinc(t) =

1
sinc(t),

Solution: Note that, for ,

1
F {sinc(t)  sinc(t)} =

and
F

 
 

 
f
1
f
1 1
f

=

 
f

= sinc(t),

As a result,
sinc(t)  sinc(t) =

29

1
sinc(t).

28. Find the output y(t) of an LTI system with impulse response h(t) = et u(t) when driven by
the input x(t) = et u(t). Treat the special case = separately. Determine if y(t) is an
energy signal or a power signal by nding the energy E or the power P .
Solution: Using the convolution theorem we obtain
1
1
)(
)
+ j2f + j2f
1
1
1
1

( ) + j2f
( ) + j2f

Y (f ) = X(f )H(f ) = (
=
Thus

1
[et et ]u1 (t).
( )

y(t) = F 1 [Y (f )] =

If = then X(f ) = H(f ) =

1
+j2f .

In this case

y(t) = F 1 [Y (f )] = F 1 [(

1
)2 ] = tet u1 (t)
+ j2f

The signal is of the energy type with energy

 T
 T
2
2
2
|y(t)| dt = lim
Ey = lim

1
(et et )2 dt
T T
T 0 ( )2
2



T /2 

1 2t T /2
2
1
1 2t T /2
e
e(+)t 

+
e
= lim


2
T ( )
2
2
( + )
0
0
0

1
2
1
1
1
+

]=
[
2
( ) 2 2 +
2( + )

29. Can the response of an LTI system to the input x(t) = sinc(t) be y(t) = sinc2 (t)? Justify
your answer.
Solution: The answer is no. Let the response of the LTI system be h(t) with Fourier
transform H(f ). Then, from the convolution theorem we obtain
Y (f ) = H(f )X(f ) = (f ) = (f )H(f )
This is impossible since (f ) = 0 for |f | >

1
2

whereas (f )
= 0 for

1
2

< |f | 1.

30. Consider the periodic signals

(t 2n)
(a) x1 (t) =
n=

(b) x2 (t) = n= (t n)
Find the Fourier series coecients without any integrals, by using a table of Fourier transforms (such as Table 2.1 in the textbook) and the relation
 
1
n
XT
.
xn =
T0 0 T0
Solution:
30

(1) XT0 (f ) = sinc2 (f ), and T0 = 2. Therefore,

 
n
1
n
1
2
sinc
.
= sinc2
xn =
T0
T0
2
2
(2) Note that x2 (t) = 1, as shown in the gure below:
x2(t)
1

t
-3

-2

-1

It follows that X2 (f ) = (f ). The signal can also be consider as periodic with period
T0 = 1 and therefore xn = (n). In other words, x0 = 1 and xn = 0, n
= 0.
31. MATLAB problem.
Download and execute the Matlab script homework3f04.m from the web site of the class. The
script nds the 50% (or 3-dB) energy bandwidth, B3dB , and the 95% energy bandwidth,
B95 , of a rectangular pulse
x(t) = (t) ,
from its energy spectral density, G(f ) = sinc2 (f ). Give the values of B3dB and B95 , and
print or sketch G(f ) in dBm, where dBm is with reference to 103 Joule/Hz.
Solution: B3dB = 0.268311 Hz and B95 = 1.668457 Hz.
Energy spectral density of (t)
30

25

20

dBm

15

10

10
15

10

0
Frequency (Hz)

31

10

15

32. MATLAB problem

Based on the script homework3f04.m of the previous problem, write a Matlab script to nd
numerically the energy E1 contained in the rst lobe of the energy spectral density, that is,
 1
E1 =
G(f )df,
1

Solution:
E1 = 0.902823 Joules. This was produced by the following script:
% Name: homework3_2.m
% For the EE160 students of San Jose State University in Fall 2004
N = 4096;
f = -1:1/N:1;
G = sinc(f).^2;
E = sum(G)/N;
fprintf(The energy in the main lobe of G(g) is %8.6f Joules\n, E);

33. Sketch carefully the following signals and their Fourier transform
 
.
(a) x1 (t) = 3t
 12

(b) x2 (t) = 2 (t 3) .
Solution:
(a) X1 (f ) =

2
3

sinc

2 
3 f .
X1(f)
x1(t)
2/3

t
-1/3

f
-9/2

1/3

-2

-3/2

3/2

9/2

(b) X2 (f ) = 2 sinc2 (2f ) ej6f .

x2(t)

|X2(f)|
2

t
2

-3/2

32

-1

-1/2

1/2

3/2

34. MATLAB problem.

Download and execute the Matlab script homework4f04.m from the web site of the class. The
script illustrates two signals in the time domain and their corresponding Fourier transforms.
This serves to verify that the time variation is proportional to the bandwidth. Sketch or print
the plots.
Solution:
x1(t)

|X1(f)|

2
1
1

0.8
0.6

0.4
1
0.2
2
0.2

0
10

0.2

10

0
5
Frequency (Hz)

10

x2(t)

|X2(f)|

0.6
0.5

0.4
0

0.3
0.2

0.1
2
0.2

0
Time (s)

0
10

0.2

35. Determine the Fourier transform of the signals shown below.

x1(t)

x3(t)

x2(t)

2
1
1
t

-2

t
-2 -1

-2

-1

2
-1

33

Solution:
(a) Write x1 (t) = 2 ( 4t ) 2 ( 2t ). Then

 

 
t
t
F 2
= 8 sinc(4f ) 4 sinc2 (2f )
X1 (f ) = F 2
4
2
(b) Write x2 (t) = 2 ( 4t ) (t). Then
X2 (f ) = 8 sinc(4f ) sinc2 (f )
(d) Note that x3 (t) = (t + 1) (t 1). Then
X3 (f ) = sinc2 (f )ej2f sinc2 (f )ej2f = 2j sinc2 (f ) sin(2f )

36. Use the convolution theorem to show that

sinc(t)  sinc(t) = sinc(t)

Solution:
F[x(t)  y(t)] = F[x(t)] F[y(t)] = X(f ) Y (f )
Thus
sinc(t)  sinc(t) = F 1 [F[sinc(t)  sinc(t)]]
= F 1 [F[sinc(t)] F[sinc(t)]]
= F 1 [(f ) (f )] = F 1 [(f )]
= sinc(t)

37. Using the Fourier transform, evaluate the following integrals:


et sinc(t)
(a)
0
et sinc2 (t)
(b)
0
et cos(t)
(c)
0

Solution:
(a)

sinc(t)dt =

 1
2
1
1
(f )df =
df
21 + j2f
+ j2f
1/2
+ j
1
1

1
ln( + j2f )1/2 =
ln(
) = tan1
j2
j2
j


=
=

et u1 (t)sinc(t)dt

34

(b)


2

sinc (t)dt =


et u1 (t)sinc2 (t)dt

1
(f )df df
+ j2f
 1
 0
f +1
f + 1
df +
df
=
1 + jf
0 + jf
=

But




x
a+bx dx

a
b2

ln(a + bx) so that

0

f

sinc (t)dt = (
+ 2 ln( + j2f ))
j2 4
1
1
1



f
1

+ 2 ln( + j2f )) +
ln( + j2f )
(
j2 4
j2
0
1

1
2

tan1 ( ) + 2 ln(
=
)

2
2 + 4 2

(c)

x
b


cos(t)dt =

=
=

et u1 (t) cos(t)dt

1
1
((f
) + (f +
))dt
2 + j2f
2
2
1
1

1
[
+
]= 2
2 + j
j
+ 2

Sampling of lowpass signals

38. The signal x(t) = A sinc(1000t) be sampled with a sampling frequency of 2000 samples per
second. Determine the most general class of reconstruction lters for the perfect reconstruction of x(t) from its samples.
Solution:

f
A
(
)
1000 1000
Thus the bandwidth W of x(t) is 1000/2 = 500. Since we sample at fs = 2000 there is a gap
between the image spectra equal to
x(t) = A sinc(1000t) X(f ) =

2000 500 W = 1000

The reconstruction lter should have a bandwidth W  such that 500 < W  < 1500. A lter
that satisfy these conditions is




f
f
1

=
H(f ) = Ts
2W 
2000
2W 
and the more general reconstruction lters have the form
1
|f | < 500
2000
arbitrary 500 < |f | < 1500
H(f ) =

0
|f | > 1500
35

39. The lowpass signal x(t) with a bandwidth of W is sampled at intervals of Ts seconds, and the
signal

x(nTs )p(t nTs )
xp (t) =
n=

is generated, where p(t) is an arbitrary pulse (not necessarily limited to the interval [0, Ts ]).
(a) Find the Fourier transform of xp (t).
(b) Find the conditions for perfect reconstruction of x(t) from xp (t).
(c) Determine the required reconstruction lter.
Solution:
(a)
xp (t) =

x(nTs )p(t nTs )

n=

= p(t) 

x(nTs )(t nTs )

n=

= p(t)  x(t)

(t nTs )

n=

Thus

Xp (f ) = P (f ) F x(t)
= P (f )X(f )  F
= P (f )X(f ) 
=


(t nTs )

n=

1
Ts


(t nTs )

n=

(f

n=

n
)
Ts

n
1
P (f )
X(f )
Ts
Ts
n=

(b) In order to avoid aliasing T1s > 2W . Furthermore the spectrum P (f ) should be invertible
for |f | < W .
f
) with W < W < T1s W .
(c) X(f ) can be recovered using the reconstruction lter ( 2W

In this case
f
)
X(f ) = Xp (f )Ts P 1 (f )(
2W
40. Consider a signal s(t) whose Fourier transform is given below:
S(f)
1

f
-1

36

Sketch carefully the Fourier transform S (f ) of the sampled signal

s (t) = s(t)

(t nT )

n=

s(nT )(t nT )

n=

for (a) T = 2/3 and (b) T = 1/2. For each case, only if possible, speciy a lter characteristic
that allows a complete reconstruction of s(t) from s (t).
Solution:
S(f)

(a) T = 2/3:

3/2

f
-3

-2

-1

S(f)

(b) T = 1/2:

f
-3

-2

-1

H(f)
Reconstruction filter:
(only for T=1/2)

1/2

f
-1

41. A sinusoidal signal of frequency 1 Hz is to be sampled periodically.

(a) Find the maximum allowable time interval between samples.
(b) Samples are taken at 1/3 second intervals. Show graphically, to your satisfaction, that
no other sine waveform with bandwidth less that 1.5 Hz can be represented by these
samples.
(c) Samples are taken at 2/3 second intervals. Show graphically these samples may represent
another sine waveform of frequency less than 1.5 Hz.
Solution:
37

(a) Ts = 1/2, i.e., the inverse of the Nyquist rate which in this case is 2 Hz.
(b) At fs = 3 Hz, the sampled signal spectrum consists of nonoverlapping copies of the
signal:

1/2

1/2

1/2

1/2 1/2

1/2

f
-4

-3

-2

-1

0 +1

+2 +3 +4

(c) In this case, fs = 3/2 Hz which is less than the Nyquist rate. The copies of the signal
spectrum overlap and the samples can be those of a signal with lower frequency.
1/2

1/2

1/2

1/2

f
-1.5 -1

0 +1 +1.5

42. The signal x(t) = cos(2t) is ideally sampled with a train of impulses. Sketch the spectrum
(t), for the following values
X (f ) of the sampled signal, and nd the reconstructed signal x
of sampling period Ts and ideal lowpass reconstruction lter bandwidth W  :
(a) Ts = 1/4, W  = 2.
(b) Ts = 1, W  = 5/2.
(c) Ts = 2/3, W  = 2.
Solution: The spectra of the signal X(f ), and that of the sampled signal X (f ), for each
case are shown in the gure below:

38

X(f)
1/2

1/2

f
-1

1
X(f)

(a)
2

f
-9

-7

-5

-1

-3

X(f)

(b)
1

1 1

-5

-3

-1

X(f)

(c)
3/4

3/4

3/4

3/4

3/4

3/4

f
-1

The reconstructed signals for each value of sampling period Ts and ideal lowpass reconstruction lter bandwidth W  are:
(a) x
(t) = cos(2t).
(b) x
(t) = 1 + 2 cos(2t) + 2 cos(4t).
(c) x
(t) = cos(t) + cos(2t) + cos(4t).
43. The signal x(t) = sinc2 (t) is ideally sampled with a train of impulses. Sketch the spectrum
X (f ) of the sampled signal, for the sampling periods below. For those values of Ts for
which reconstruction is possible, specify the range of the cuto frequency W  of the ideal
reconstruction lter.
(a) Ts = 2/3.
(b) Ts = 1.
(c) Ts = 1/4.
Solution: The spectra of the signal X(f ), and that of the sampled signal X (f ), for each
case are shown in the gure below:
39

X(f)
1
f
-1

X(f)

(a)

1.5

-3

-1

-2

(b)

X(f)
1

-3

-1

-2

(c)

X(f)

f
-1

-4 -3

The value of Ts for which reconstruction is possible, is (c) Ts = 1/4. The range of the cuto
frequency W  of the ideal reconstruction lter is 1 < W  < 3.
44. A lowpass signal has spectrum as shown below.
X(f)
3/2

1/2

f
-2

-1

This signal is sampled at fs samples/second with impulses and reconstructed using an ideal
(t) denote the output of
lowpass lter (LPF) of bandwidth W = 2 and amplitude 1/fs . Let x
40

the LPF.
(a) Give an expression for x(t) and sketch the waveform.
(b) The sampling frequency is fs = 3. Sketch the spectra of the sampled signal, X (f ) and
). Also, sketch the reconstructed waveform x
that of the recovered signal X(f
(t).
(c) Repeat part (b) with fs = 4.

Solution:
(a) x(t) = 2 sinc(4t) + sinc2 (t).
4.5
4
3.5
3

x(t)

2.5
2
1.5
1
0.5
0
0.5
4

0
t (sec)

(b) x
(t) = sinc(2t) + sinc2 (t) + 2 sinc(t) cos(3t).
X(f)
3/2

1/2

f
-4

-3

-2

-1

41

4.5
4
3.5
3

x~(t)

2.5
2
1.5
1
0.5
0
0.5
4

0
t (sec)

(c) As shown by the sampled spectrum below, there is no overlap between the shifted copies
of X(f ). Therefore, x
(t) = x(t).
X(f)
3/2

1/2

f
-6

-5

-4

-3

-2
2

-1

45. A compact disc (CD) records audio signals digitally using PCM. Assume the audio signal
bandwidth to be 15 KHz.
(a) What is the Nyquist rate?
(b) If the Nyquist samples are quantized to L = 65, 536 levels and then binary coded,
determine the number of bits required to encode a sample.
(c) Assuming that the signal is sinusoidal and that the maximum signal amplitude is 1 volt,
determine the quatization step and the signal-to-quatization noise ratio.
(d) Determine the number of bits per second (bit/s) required to encode the audio signal.
(e) For practical reasons, signals are sampled at above the Nyquist rate, as discussed in class.
Practical CDs use 44,000 samples per second. For L = 65, 536 determine the number of
bits per second required to encode the signal and the minimum bandwidth required to
transmit the encoded signal.

42

Solution:
(a) fs = 30000 samples/s
(b) log2 (L) = 16 bits
(c) = 2/216 = 215 volts
(d) 16fs = 480000 bits/s
(e) 16 44000 = 704000 bits/s and BT = 704000/2 = 352000 Hz.

Bandpass signals
46. Consider a signal s(t) whose Fourier transform is given below:
S(f)
1/2

f
-4

-3

-2

-1

Sketch carefully the Fourier transform S (f ) of the sampled signal

s (t) = s(t)

(t nT )

n=

s(nT )(t nT )

n=

for (a) T = 1/4 and (b) T = 1/2. For each case, only if possible, speciy a lter characteristic
that allows a complete reconstruction of s(t) from s (t).
Solution:

43

(a) T=1/4:
S(f)
Original (scaled)

-1

-2

-1

+1

-4 -3 -2 -1

+2

+1

(b) T=1/2:
S(f)
Overlap of original
and third copy (-3)

Overlap of original
and third copy (+3)

f
-4 -3 -2 -1

Reconstrucion filter:
H(f)
1/4
f
-4 -3 -2 -1

47. Determine the range of permissible cuto frequencies for an ideal low pass lter used to
reconstruct the signal
x(t) = 10 cos(600t) cos 2 (1600t),
which is sampled at 4000 samples per second. Sketch X(f ) and X (f ). Find the minimum
allowable sampling frequency.
Solution:
The cuto frequency of the reconstruction lter can be in the range between W = 1900 Hz
and fs W = 2100 Hz.

44

X(f)

f
-1600

1600

1900

X(f)

f
-5600
-5900

-1600

-4000
-2100

1600

-1900

4000

5600

5900

1900 2100

The minimum (Nyquist) sampling frequency is 2W = 3800 Hz.

48. Given the bandpass signal spectrum shown in the gure below, sketch the spectra for the
following sampling rates fs and indicate which ones are suitable for the reconstruction of the
signal fom its samples: (a) 2B (b) 2.5B (c) 3B (d) 4B (e) 5B (f) 6B.
X(f)

f
-3

-2

-1

Solution:
For bandpass sampling and recovery, all but (b) and (e) will work theoretically, although an
ideal lter with bandwidth exactly equal to the unsampled signal bandwidth is necessary. For
lowpass sampling and recovery, only (f) will work.
In the gure next page, the case (a) fs = 2B, with B = 1 for convenience, is illustrated. The
terms used are n = 0, 1, 2, 3 in the expression of the sampled spectrum:
X (f ) = fs

X(f n fs ) = 2

n=

X(f 2n).

n=

Spectra for higher values of n do not overlap with the spectrum of the original signal and are
therefore not shown.

45

Case fs=2B:

(B=1)

X(f)

n=0

A
f
-4

-3

-2

-1

X(f-2), X(f+2):
-2

+2

+2

2A -2

n=1,-1
f

-9

-8

-7

-6

-5

-4

-3

-2

-1

X(f-4), X(f+4):
-4

-4

+4

2A

+4

n=2,-2
f

-9

-8

-7

-6

-5

-4

-3

-2

-1

X(f-6), X(f+6):
-6

-6

+6

2A

+6

n=3,-3

f
-9

-8

-7

-6

-5

-4

-3

-2

-1

2A
f
-9

-8

-7

-6

-5

-4

-3

-2

-1

46

49. (Downconversion by bandpass sampling) Consider the bandpass signal x(t) whose spectrum
is shown below.
X(f)
2

f
-8

-7

This signal is sampled at fs = 4 samples/second with ideal impulses.

(a) Sketch the signal x(t).
(b) Sketch the spectrum X (f ) of the sampled signal.
(c) Sketch the output x
(t) of an ideal bandpass lter (BPF) of bandwidth B = 1 and
amplitude 1/fs , centered at f0 = 7.5.
(d) The sampled signal is now passed through an ideal lowpass lter (LPF) of bandwidth
d (t) is also a bandpass signal,
W = 1 and amplitude 1/fs . Show that the output x
equivalent to the original signal, but with a lower value of center frequency f0 . Find the
d (t).
value of f0 and sketch the signal x
Solution:
(a) The signal is x(t) = 4 sinc2

t
2

cos(15t), which is plotted below via the Matlab script:

t=-4:0.01:4; plot(t,4*sinc(t./2).^2.*cos(15*pi.*t))

5
4
3
2
1
0
1
2
3
4
5
4

47

(b)
X(f)
8

f
-9 -8 -7

-5 -4 -3

-1

(c) There is no overlap between copies of X(f ) and therefore x

(t) = x(t).
(d) The spectrum at the output of the LPF is:
~

X(f)
2

f
-1

Therefore, the output signal is given by x(t) = 4 sinc2

via the Matlab script:

t
2

cos(t), which is plotted below

t=-4:0.01:4; plot(t,4*sinc(t./2).^2.*cos(pi.*t))

2
4

48

50. Assume that the Fourier transform of a signal x(t) is real and has the shape shown in the
gure below:
X(f)
A

f
-W

Determine and plot the spectrum of each of the following signals, where x
(t) denotes the
Hilbert transform of x(t),
(a) x1 (t) =
(b) x2 (t) =
(c) x3 (t) =
(d) x3 (t) =

3
4

x(t) + 14 j x
(t)

3
 43
 43
4


x(t) + 34 j x
(t) ej2f0 t , f0 W

x(t) + 14 j x
(t) ej2W t

x(t) 14 j x
(t) ejW t

Solution:
(a) Note that F{j x
(t)} = j[j sgn(f )]X(f ). As a result,
1
3
X(f ) + j[j sgn(f )]X(f )
4
4


3 1
+ sgn(f ) X(f )
=
4 4

1
2 X(f ), f < 0;
=
X(f ),
f 0.

X1 (f ) =

(b) In this case,

3
[1 + sgn(f f0 )] X(f f0 )
4

0,
f < f0 ;
=
3
2 X(f f0 ), f f0 .

X2 (f ) =

(c) The answer here is the same as in part (a), with the dierence of a shift to the right in
the frequency domain by W ,

1
2 X(f W ), f < W ;
X3 (f ) =
X(f ),
f W.
(d) Here,


W
W
3 1
sgn(f
) X(f
)
X4 (f ) =
4 4
2
2


49

(4)

X1(f)

X2(f)

2A
1.5A
A

f
-W

f
0

X3(f)

f0

F0-W

X4(f)

2A

2A

A
f
0

2W

-W/2

W/2

3W/2

51. Consider the signal

x(t) = 2W sinc(2W t) cos(2f0 t)
(t)
(a) Obtain and sketch the spectrum of the analytical signal xp (t) = x(t) + j x
(b) Obtain and sketch the spectrum of the complex envelope (or complex baseband representation) x
(t)
(c) Find the complex envelope x
(t)
Solution:
(a) The spectrum of the analytical signal is
Xp (f ) = X(f ) + j[j sgn(f )]X(f ) = [1 + sgn(f )] X(f ),
where X(f ) is the Fourier transform of x(t), given by
 



f + f0
f f0
1

+
.
X(f ) =
2
2W
2W
Consequently,


f f0
,
f0 > 2W,
2W
a rectangular pulse of width 2W centered at f = f0 .
(b) The complex envelope xp (t) is


Xp (f ) =

(t)ej2f0 t .
xp (t) = x
Therefore, x
(t) = xp (t)ej2f0 t , and

)=
F{
x(t)} = [Xp (f )]f f +f0 =
X(f

a rectangular pulse of width 2W centered at f = 0.

50

f
2W


,

(c) The complex envelope is given by

)} = 2W sinc(2W t).
x
(t) = F 1 {X(f

52. The signal

x(t) =

 
t
cos [2(f0 f )t] , f  f0

is applied at the input of a lter (LTI system) with impulse response

h(t) = et cos(2f0 t)u(t).
Find the output signal y(t) using complex envelope techniques.
Solution: For t < /2, the output is zero. For |t| /2, the result is
y(t) = !

/2
2 + (2f )2


cos[2(f0 + f )t ] e(t+ /2) cos[2(f0 + f )t + ] ,

and for t > /2, the output is

y(t) = !

(/2)et
2 + (2f )2



e /2 cos[2(f0 + f )t ] e /2 cos[2(f0 + f )t + ] .

53. The bandpass signal x(t) = sinc(t) cos(2f0 t) is passed through a bandpass lter with impulse
response h(t) = sinc2 (t) sin(2f0 t), Using the lowpass equivalents of both input and impulse
response, nd the lowpass equivalent of the output and from it nd the output y(t).
Solution:
x(t) = sinc(t) cos(2f0 t)

h(t) = sinc2 (t) sin(2f0 t)

1
1
X(f ) = (f + f0 )) + (f f0 ))
2
2
1
1
H(f ) = (f + f0 )) + (f f0 ))
2j
2j

The lowpass equivalents are

Xl (f ) = 2u(f + f0 )X(f + f0 ) = (f )
1
Hl (f ) = 2u(f + f0 )H(f + f0 ) = (f )
j
1
12 < f 0
2j (f + 1)
1
1
Xl (f )Hl (f ) =
(f + 1) 0 f < 12
Yl (f ) =
2j
2
0
otherwise
Taking the inverse Fourier transform of Yl (f ) we can nd the lowpass equivalent response of

51

the system. Thus,

yl (t) = F 1 [Yl (f )]
 0
 1
2
1
1
j2f t
(f + 1)e
df +
(f + 1)ej2f t df
=
2j 1
2j 0
2



1 j2f t 0
1 1 j2f t 0
1
1
j2f t
+ 2 2e
=
fe
 1 + 2j j2t e
 1
2j j2t
4 t
2
2
1

 1
2
1
1
1
1 1 j2f t  2
f ej2f t + 2 2 ej2f t  +
e


2j j2t
4 t
2j j2t
0
0


1
1
sin t + 2 2 (cos t 1)
= j
4t
4 t
The output of the system y(t) can now be found from y(t) = Re[yl (t)ej2f0 t ]. Thus


1
1
y(t) = Re (j[
sin t + 2 2 (cos t 1)])(cos 2f0 t + j sin 2f0 t)
4t
4 t
1
1
sin t] sin 2f0 t
= [ 2 2 (1 cos t) +
4 t
4t
Note: An alternative solution is covered in class.

54. A bandpass signal is given by

x(t) = sinc(2t) cos(3t).
(a) Is the signal narrowband or wideband? Justify your answer.
(b) Find the complex baseband equivalent x (t) and sketch carefully its spectrum.
(a) Give an expression for the Hilbert transform of x(t).
Solution:




f 3/2
1

+
. The
(a) The Fourier transform of the signal is X(f ) = 12 12 f +3/2
2
2
2
following sketch shows that the signal is wideband, as B = 2 and f0 = 3/2.
X(f)
1/4

f
-1.5

1.5

(b) From the quadrature modulator expression x(t) = xc (t) cos(2f0 t) xs (t) sin(2f0 t),
it follows that xs (t) = 0 and therefore x (t) = xc (t) = sinc(2t). The corresponding
spectrum is sketched below:

52

Xl(f)
1/2

f
-1

(c) Use the expression x

(t) = xc (t) sin(2f0 t) + xs (t) cos(2f0 t), from which it follows that
x
(t) = sinc(2t) sin(3t).

55. As shown in class, the in-phase and quadrature components, xc (t) and xs (t), respectively, of
the complex baseband (or lowpass) equivalent x (t) of a bandpass signal x(t) can be obtained
as
 



cos(2f0 t) sin(2f0 t) x(t)
xc (t)
=
,
(t)
xs (t)
sin(2f0 t) cos(2f0 t) x
where x
(t) is the Hilbert transform of x(t).
(a) Sketch a block diagram of a system using H to label the block that performs the
Hilbert transform that has as input x(t) and as outputs xc (t) and xs (t).

(b) (Amplitude modulation) Let x(t) = a(t) cos(2f0 t). Assume that the bandwidth W of
the signal a(t) is such that W  f0 . Show that a(t) can be recovered with the following
system

H(f)
2
a(t)

x(t)
f
-W
cos(2f0t)

Low-pass filter

Solution:
(a)

53

xc(t)

xs(t)

cos(2f0t)

x(t)

H
-1
H

(b) Use the modulation property of the Fourier transform. Let y(t) denote the mixer output.
The spectra are shown shown in the gure below.
X(f)
A0/2

f
-f0

f0

B=2W

B=2W

Y(f)
2

H(f)
A0/2

A0/4

f
-W

-2f0

2f0

B=2W

B=2W

A(f)
A0

f
-W

56. A lowpass signal x(t) has a Fourier transform shown in the gure (a) below.

54

X(f)
1

f
-W

-W/2

W/2

(a)

sin(2f0t)
x1(t)

x3(t)
H
2cos(2f0 t)
-

x(t)

+
sin(2f0t)
x2(t)

x5(t)

x6(t)

LPF
[-W,W]

x7(t)

x4(t)

H
(b)

The signal is applied to the system shown in gure (b). The blocks marked H represent
Hilbert transform blocks and it is assumed that W  f0 . Determine the signals xi (t) and
plot Xi (f ), for 1 le7.

0 t) = x(t) sin(2f0 t),

0 t) = x(t) cos(2f0 t) and x(t) cos(2f
(Hint: Use the fact that x(t) sin(2f
when the bandwidth W of x(t) is much smaller than f0 .)
Solution: This is an example of single sideband (SSB) amplitude modulation (AM).
x1 (t) = x(t) sin(2f0 t)
1
1
X1 (f ) = X(f + f0 ) + X(f f0 )
2j
2j
(t)
x2 (t) = x
X2 (f ) = jsgn(f )X(f )

0 t) = x(t) cos(2f0 t)
x3 (t) = x1 (t) = x(t) sin(2f
1
1
X3 (f ) = X(f + f0 ) X(f f0 )
2
2
(t) sin(2f0 t)
x4 (t) = x2 (t) sin(2f0 t) = x
1
1
X4 (f ) = X(f + f0 ) + X(f f0 )
2j
2j
1
1
[jsgn(f f0 )X(f f0 )]
= [jsgn(f + f0 )X(f + f0 )] +
2j
2j
1
1
sgn(f + f0 )X(f + f0 ) sgn(f f0 )X(f f0 )
=
2
2
55

x5 (t) = x
(t) sin(2f0 t) + x(t) cos(2f0 t)
1
1
X5 (f ) = X4 (f ) X3 (f ) = X(f + f0 )(sgn(f + f0 ) 1) X(f f0 )(sgn(f f0 ) + 1)
2
2
x(t) sin(2f0 t) + x(t) cos(2f0 t)]2 cos(2f0 t)
x6 (t) = [
X6 (f ) = X5 (f + f0 ) + X5 (f f0 )
1
1
X(f + 2f0 )(sgn(f + 2f0 ) 1) X(f )(sgn(f ) + 1)
=
2
2
1
1
+ X(f )(sgn(f ) 1) X(f 2f0 )(sgn(f 2f0 ) + 1)
2
2
1
1
= X(f ) + X(f + 2f0 )(sgn(f + 2f0 ) 1) X(f 2f0 )(sgn(f 2f0 ) + 1)
2
2
x7 (t) = x6 (t)  2W sinc(2W t) = x(t)
f
) = X(f )
X7 (f ) = X6 (f )(
2W
jX2 (f )

2jX1 (f )

1)

LL f0




L

2)
AA

AA

f0

3)
AA f0




A

A
f0



AA

5)

AA

f0

f0

X5 (f )
f0

LL
L

AA
A

2X4 (f )

AA
A

f0

2f0

AA

4)

2X3 (f )

6)
A
AA

7)
A
AA

X6 (f )

X7 (f )

56

2f0

Analog amplitude-modulation (AM) systems

57. The message signal m(t) = 2 cos(400t) + 4 sin(500t + 3 ) modulates the carrier signal c(t) =
A cos(8000t), using DSB amplitude modulation. Find the time domain and frequency domain representation of the modulated signal and plot the spectrum (Fourier transform) of
the modulated signal. What is the power content of the modulated signal?
Solution: The modulated signal is
u(t) = m(t)c(t) = Am(t) cos(24 103 t)


250

200
t) + 4 sin(2
t + ) cos(24 103 t)
= A 2 cos(2

3
200
200
)t) + A cos(2(4 103
)t)
= A cos(2(4 103 +

250
250
)t + ) 2A sin(2(4 103
)t )
+2A sin(2(4 103 +

3
Taking the Fourier transform of the previous relation, we obtain


200
200
2 j
2 j
250
250
3
3
U (f ) = A (f
) + (f +
) + e (f
) e
)
(f +

1
 [(f 4 103 ) + (f + 4 103 )]
2
200
200
A
(f 4 103
) + (f 4 103 +
)
=
2

250
250
) + 2ej 6 (f 4 103 +
)
+2ej 6 (f 4 103

200
200
) + (f + 4 103 +
)
+(f + 4 103


250
250
j 6
3
j 6
3
) + 2e (f + 4 10 +
)
(f + 4 10
+2e

The gure gure below shows the magnitude and the phase of the spectrum U (f ).
6

|U (f )| A . . . . . . . . . . . . . . . . . . . . . .
6

. . . . . . . . . . . . . . . . . . . . . . .A/2
6

200
200
250
fc 250
fc fc + fc +

6
200
fc 250
fc

6
250
fc + 200
fc +

U (f )

s . . . . . . . . . . . . . . . . . . . . . . .6. . . . . . s

s . . . . . . . . . .6. . . . . . . . . . . . . . . . . . . s

57

To nd the power content of the modulated signal we write u2 (t) as

200
200
)t) + A2 cos2 (2(4 103
)t)

250
250

)t + ) + 4A2 sin2 (2(4 103

)t )
+4A2 sin2 (2(4 103 +

3
+terms of cosine and sine functions in the rst power

u2 (t) = A2 cos2 (2(4 103 +

Hence,


P = lim

T
2

T T
2

u2 (t)dt =

A2 A2 4A2 4A2
+
+
+
= 5A2
2
2
2
2

58. In a DSB AM system, the carrier is c(t) = A cos(2fc t) and the message signal is given by
m(t) = sinc(t) + sinc2 (t). Find the frequency domain representation and the bandwidth of
the modulated signal.
Solution:
u(t) = m(t)c(t) = A(sinc(t) + sinc2 (t)) cos(2fc t)
Taking the Fourier transform of both sides, we obtain
U (f ) =
=

A
[(f ) + (f )]  ((f fc ) + (f + fc ))
2
A
[(f fc ) + (f fc ) + (f + fc ) + (f + fc )]
2

(f fc )
= 0 for |f fc | < 12 , whereas (f fc )
= 0 for |f fc | < 1. Hence, the bandwidth

of the modulated signal is 2.

59. A DSB-modulated signal u(t) = A m(t) cos(2fc t) is mixed (multiplied) with a local carrier
xL (t) = cos(2fc t+) and the output is passed through a lowpass lter (LPF) with bandwidth
equal to the bandwidth of the message signal m(t). Denote the power of the signal at the
,
output of the LPF by Pout and the power of the modulated signal by PU . Plot the ratio PPout
U
as a function of , for 0 .
Solution: The mixed signal y(t) is given by
y(t) = u(t) xL (t) = Am(t) cos(2fc t) cos(2fc t + )
A
m(t) [cos(22fc t + ) + cos()]
=
2
The lowpass lter will cut-o the frequencies above W , where W is the bandwidth of the
message signal m(t). Thus, the output of the lowpass lter is
z(t) =

A
m(t) cos()
2
2

If the power of m(t) is PM , then the power of the output signal z(t) is Pout = PM A4 cos2 ().
2
The power of the modulated signal u(t) = Am(t) cos(2fc t) is PU = A2 PM . Hence,
1
Pout
= cos2 ()
PU
2
58

A plot of

Pout
PU

for 0 is given in the next gure.

0.5
0.45
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0

0.5

1.5
2
Theta (rad)

2.5

3.5

60. The output signal from an AM modulator is

u(t) = 5 cos(1800t) + 20 cos(2000t) + 5 cos(2200t).
(a) Determine the message signal m(t) and the carrier c(t). (Hint: Look at the spectrum of
u(t).)
(b) Determine the modulation index.
(c) Determine the ratio of the power in the sidebands to the power in the carrier.
Solution:
(a)
u(t) = 5 cos(1800t) + 20 cos(2000t) + 5 cos(2200t)
1
= 20(1 + cos(200t)) cos(2000t)
2
The modulating signal is m(t) = cos(2100t) whereas the carrier signal is c(t) =
20 cos(21000t).
(b) Since 1 cos(2100t) 1, we immediately have that the modulation index is = 12 .
(c) The power of the carrier component is Pcarrier =
2
= 50. Hence,
sidebands is Psidebands = 400
2

400
2

= 200, whereas the power in the

1
50
Psidebands
=
=
Pcarrier
200
4

61. An SSB AM signal is generated by modulating an 800 kHz carrier by the message signal
m(t) = cos(2000t) + 2 sin(2000t). Assume that the amplitude of the carrier is Ac = 100.
(a) Determine the Hilbert transform of the message signal, m(t).

59

(b) Find the time-domain expression for the lower sideband SSB (LSSB) AM signal.
(c) Determine the spectrum of the LSSB AM signal.
Solution:
(a) The Hilbert transform of cos(21000t) is sin(21000t), whereas the Hilbert transform of
sin(21000t) is cos(21000t). Thus
m(t)

= sin(21000t) 2 cos(21000t)
(b) The expression for the LSSB AM signal is
sin(2fc t)
ul (t) = Ac m(t) cos(2fc t) + Ac m(t)

= sin(21000t)
Substituting Ac = 100, m(t) = cos(21000t) + 2 sin(21000t) and m(t)
2 cos(21000t) in the previous, we obtain
ul (t) = 100 [cos(21000t) + 2 sin(21000t)] cos(2fc t)
+ 100 [sin(21000t) 2 cos(21000t)] sin(2fc t)
= 100 [cos(21000t) cos(2fc t) + sin(21000t) sin(2fc t)]
+ 200 [cos(2fc t) sin(21000t) sin(2fc t) cos(21000t)]
= 100 cos(2(fc 1000)t) 200 sin(2(fc 1000)t)
(c) Taking the Fourier transform of the previous expression we obtain
Ul (f ) = 50 ((f fc + 1000) + (f + fc 1000))
+ 100j ((f fc + 1000) (f + fc 1000))
= (50 + 100j)(f fc + 1000) + (50 100j)(f + fc 1000)
Hence, the magnitude spectrum is given by
!
502 + 1002 ((f fc + 1000) + (f + fc 1000))
|Ul (f )| =

= 10 125 ((f fc + 1000) + (f + fc 1000))

62. The system shown in the gure below can be used to generate an AM signal.
x(t)

m(t)

Nonlinear
memoryless
system

y(t)
Filter

u(t)
AM signal

c(t)

The carrier is c(t) = cos(2f0 t) and the modulating signal m(t) has zero mean and its
maximum absolute value is Am = max |m(t)|. The nonlinear device has a quadratic inputoutput characteristic given by
y(t) = a x(t) + b x2 (t).
(a) Give an expression of y(t) in terms of m(t) and c(t).
60

(b) Specify the lter characteristics such that an AM signal is obtain at its output.
(c) What is the modulation index?
Solution:
(a)
y(t) = ax(t) + bx2 (t)
= a(m(t) + cos(2f0 t)) + b(m(t) + cos(2f0 t))2
= am(t) + bm2 (t) + a cos(2f0 t)
+b cos2 (2f0 t) + 2bm(t) cos(2f0 t)
(b) The lter should reject the low frequency components, the terms of double frequency
and pass only the signal with spectrum centered at f0 . Thus the lter should be a BPF
with center frequency f0 and bandwidth W such that f0 WM > f0 W
2 > 2WM where
WM is the bandwidth of the message signal m(t).
(c) The AM output signal can be written as
u(t) = a(1 +

2b
m(t)) cos(2f0 t)
a

Since Am = max[|m(t)|] we conclude that the modulation index is

=

2bAm
a

63. Consider a message signal m(t) = cos(2t). The carrier frequency is fc = 5.

(a) (Suppressed carrier AM) Plot the power Pu of the modulated signal as a function of the

phase dierence (between transmitter and receiver) = c r .

(b) (Conventional AM) With a modulation index a = 0.5 (or 50%), sketch carefully the
modulated signal u(t).

Solution:
(a) The demodulated signal power is given by Py = Pu cos2 (), which has maximum value
Py = Pu for = 0, and minimum value Py = 0 for = /2, 3/2. This is shown
below, which is a plot of Py /Pu as a function of .

61

Demodulated signal power

0.8

Py/Pu

0.6

0.4

0.2

3
4
Phase error in radians

(b) The modulated signal is

u(t) = Ac [1 + 0.5 cos(2t)] cos(10t),
and plotted in the gure below, together with the carrier cos(10t) (top graph). Both
plots are normalized with respect to the carrier amplitude Ac .
Carrier signal
1

0.5

0.5

1
0.4

0.3

0.2

0.1

0.1

0.2

0.3

0.4

0.1

0.2

0.3

0.4

Modulated signal
1.5
1
0.5
0
0.5
1
1.5
0.4

0.3

0.2

0.1

62

Probability and random signals

64. A (random) binary source produces S = 0 and S = 1 with probabilities 0.3 and 0.7, respectively. The output of the source S is transmitted over a noisy (binary symmetric) channel
with a probability of error (converting a 0 into a 1, or a 1 into a 0) of  = 0.2.

S=0

R=0

S=1

R=1

(a) Find the probability that R = 1. (Hint: Total probability theorem.)

(b) Find the (a-posteriori) probability that S = 1 was produced by the source given that
R = 1 is observed. (Hint: Bayes rule.)

Solution:
(a)
P (R = 1) = P (R = 1|S = 1)P (S = 1) + P (R = 1|S = 0)P (R = 0)
= 0.8 0.7 + 0.2 0.3 = 0.62
where we have used P (S = 1) = .7, P (S = 0) = .3, P (R = 1|S = 0) =  = 0.2 and
P (R = 1|S = 1) = 1  = 1 0.2 = .8
(b)
P (S = 1|R = 1) =

P (R = 1|S = 1)P (S = 1)
0.8 0.7
P (S = 1, R = 1)
=
=
= 0.9032
P (R = 1)
P (R = 1)
0.62

65. A random variable X has a PDF fX (x) = (x). Find the following:
(a) The CDF of X, FX (x).
(b) Pr{X > 12 }.

(c) Pr{X > 0|X < 12 }.

(d) The conditional PDF fX (x|X > 12 ).

Solution:
(a)
x < 1 FX (x) = 0

1 x 0 FX (x) =

x

1 2
1
1
(v + 1)dv = ( v + v) = x2 + x +
2
2
2
1
1
 0
 x
1
1
(v + 1)dv +
(v + 1)dv = x2 + x +
0 x 1 FX (x) =
2
2
1
0
1 x FX (x) = 1
x

63

(b)
7
1
1
1
p(X > ) = 1 FX ( ) = 1 =
2
2
8
8
(c)


FX ( 12 ) FX (0)
p(X > 0, X < 12 )
3
1
=
=
p(X > 0X < ) =
1
1
2
7
p(X < 2 )
1 p(X > 2 )

(d) We nd rst the CDF



p(X x, X > 12 )
1
1
FX (xX > ) = p(X xX > ) =
2
2
p(X > 12 )

If x 12 then p(X xX > 12 ) = 0since the events E1 = {X 12 } and E1 = {X > 12 }
are disjoint. If x > 12 then p(X xX > 12 ) = FX (x) FX ( 12 ) so that

FX (x) FX ( 12 )
1
FX (xX > ) =
2
1 FX ( 12 )
Dierentiating this equation with respect to x we obtain
 f (x)
X

x>
1
1FX ( 12 )
fX (xX > ) =
2
0
x

1
2
1
2

66. A random process is given by X(t) = A + Bt, where A and B are independent random variables uniformly distributed in the interval [1, 1]. Find:
(a) The mean function mx (t).
(b) The autocorrelation function Rx (t1 , t2 ).
(c) Is X(t) a stationary process?
Solution:
mX (t) = E[A + Bt] = E[A] + E[B]t = 0
where the last equality follows from the fact that A, B are uniformly distributed over [1 1]
so that E[A] = E[B] = 0.
RX (t1 , t2 ) = E[X(t1 )X(t2 )] = E[(A + Bt1 )(A + Bt2 )]
= E[A2 ] + E[AB]t2 + E[BA]t1 + E[B 2 ]t1 t2
The random variables A, B are independent so that E[AB] = E[A]E[B] = 0. Furthermore
 1
1 1
1
1
2
2
x2 dx = x3 1 =
E[A ] = E[B ] =
2
6
3
1
Thus
RX (t1 , t2 ) =
and the provess is not stationary.

64

1 1
+ t1 t2
3 3

67. A simple binary communication system model, with additive Gaussian noise, is the following:
S is a binary random variable, representing the message bit sent, taking values 1 and +1
with equal probability. Additive noise is represented by a Gaussian random variable N of
zero mean and variance 2 . The received value is a random variable R = S + N . This is
illustrated in the gure below:

N
Find the probability density function (pdf) of R.
(Hint: The pdf of R is equal to the convolution of the pdfs of S and N . Remember to
use the pdf of S, which consists of two impulses. The same result can be obtained by rst
conditioning on a value of S and then integrating over the pdf of S.)
Solution: Note that S is a discrete random variable. Therefore, it can be specied by its
probability mass function (PMF), P [S = +1] = P [S = 1] = 1/2, or by a PDF that consists
of two impulses, each of weight 1/2, centered at 1.
Given a value of S = s, the conditional PDF of R is



1
2
exp 2 (r s) .
pR|S (r) = pN (r s) =
2
2 2
1

Then, applying the total probability theorem, the PDF of R is obtained as

pR (r) = pR|S=+1 (r) P [S = +1] + pR|S=1 (r) P [S = 1]
=
=

1
1
pN (s 1) + pN (s + 1)
2
2





1
1
1
2
2

exp 2 (r 1) + exp 2 (r + 1)
.
2
2
2 2 2

68. The joint PDF of two random variables X and Y can be expressed as



1 x 2 y 2
c
+
pX,Y (x, y) = exp

2
3
2
(a) Are X and Y independent?
(b) Find the value of c.
(c) Compute the probability of the event {0 < X 3, 0 < Y 2}.
[Hint: Use the Gaussian Q-function.]
Solution:
(a) The given joint PDF can be factored as the product of two functions f (x) and g(y). As
a consequence, X and Y are independent.
65

(b) pX,Y (x, y) has the form of the joint PDF of two zero-mean independent Gaussian random
2 = 9 and 2 = 4:
variables (i.e., = 0) with variances X
Y


 
1
x 2 y 2
1
,
exp
+
pX,Y (x, y) =
2 3 2
2
3
2
from which it follows that c = 1/12.
(c) Let E = {0 < X 3, 0 < Y 2}. Then
2



 
 
1
3
2
2
Q(1) = 0.1165
P [E] = Q(0) Q
Q(0) Q
= (Q(0) Q(1)) =
X
Y
2

69. A 4-level quantizer is dened by the following input-output relation:

+3, x 2;

+1, 0 x < 2;
y = g(x) =

1, 2 x < 0;

3, x < 2.
Let the input be a Gaussian random variable X, of zero mean and unit variance, N (0; 1).
You are asked to compute the probability mass function (PMF) of the output Y = g(X), i.e.,
P [Y = y] for y {1, 3}. Express your result in terms of the Gaussian Q-function.
Solution: The PMF of Y is given by:
P [Y = +3] = P [Y = 3] = Q(2)
P [Y = +1] = P [Y = 1] = Q(0) Q(2)



= 2.275 102


1
= 2.275 102 = 0.47725
2

2 = E{(X m )2 }. Find the mean

70. Let X be an r.v. of mean mX = E{X} and variance X
X
2 , of the r.v. Y = X + a, where a is a constant.
mY and variance Y2 , in terms of mX and X
2 .
Solution: mY = E{Y } = E{X + a} = E{X} + a = mX + a, and Y2 = E{Y 2 } m2Y = X

71. Let G be a Gaussian r.v. of mean m = 0 and variance 2 = 1. Find the mean mN and
2 of the r.v. N = aG, where a is a constant. This transformation is used in
variance N
computer simulations to generate samples of an AWGN process. If it is desired to generate
2 = N /2, what is the value of a?
samples with mN = 0 and variance N
0
2
2
2
2
2 2
2
Solution: mN = E{N } = aE{G} = am = 0, and
! N = E{N } = a E{G } = a = a . If
2
2
N = N0 /2, then a = N0 /2 and therefore a = N0 /2.

66

72. Let X be a uniform r.v. over the unit interval [0, 1]. Show that P [X a] = a, where
0 < a 1. This fact is used in computer simulations to generate random bits.
Solution: The CDF of X is

0, x < 0,
FX (x) = x, 0 x < 1,

1, x 1.
Therefore P [X a] = FX (a) = a, for 0 < a 1.

Optimum receivers for

data transmission over AWGN channels
73. Verify that the output of a correlator and a matched lter are maximum at t = T , even
though they may dier at other times. Download the MATLAB script corr vs MF.m from
the web site. Execute the script and print or sketch the four graphs. Explain why this is the
case, by examining the output of the matched lter expressed as a convolution integral and
making a change of variables.
Solution:

convolution of the rectangular pulse

correlation of the rectangular pulse

120

120

100

100

80

80

60

60

40

40

20

20
50

100

150

200

250

20

convolution of the sine pulse

40

60

80

100

120

correlation of the sine pulse

80
60

60
40

50

20

40

30

20
20

40

10

60
80

50

100

150

200

250

20

40

60

80

100

120

Matched lter and correlator outputs for a rectangular pulse and a sine pulse.

67

To see why the outputs have the same value at t = T , write the convolution integral of the
matched lter:
v(T ) = h(t) s(t)|t=T


=
0 T


h( )s(T )dt =

s(T )s(T )dt

s(u)2 du,

=
0

with u = T . Therefore, the output of the matched lter sampled at t = T equals the
output of a correlator over the interval [0, T ], with s(t) as the reference signal.

74. The received signal in a binary communication system that employs antipodal signals is
r(t) = s(t) + n(t),
where s(t) is shown in the gure below and n(t) is AWGN with power spectral density N0 /2
W/Hz.
s(t)
A

t
0

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

Sketch carefully the impulse response of the lter matched to s(t)

Sketch carefully the output of the matched lter when the input is s(t)
Dtermine the variance of the noise at the output of the matched lter at t = 3
Determine the probability of error as a function of A and N0

Solution:
(a) The impulse response of the lter matched to s(t) is
h(t) = s(T t) = s(3 t) = s(t)
where we have used the fact that s(t) is even with respect to the t =
(b) The output of the matched lter is
 t
s( )s(t )d
y(t) = s(t)  s(t) =
0

0
t<0

2t

0t<1
A

A (2 t) 1 t < 2

2A2 (t 2) 2 t < 3
=
2A2 (4 t) 3 t < 4

A2 (t 4) 4 t < 5

A2 (6 t) 5 t < 6

0
6t
68

T
2

3
2

axis.

A scetch of y(t) is depicted in the next gure

2A2 . . . . . . . . . . . . . . . .

A

A
A2 . . . . . .
A


@
A
@

AA
@

@
@
@

(c) At the output of the matched lter and for t = T = 3 the noise is
 T
n( )h(T )d
nT =
0
 T
 T
=
n( )s(T (T ))d =
n( )s( )d
0

The variance of the noise is


n2 T

n( )n(v)s( )s(v)d dv
0

=
0

=
=

= E

0
 T

s( )s(v)E[n( )n(v)]d dv
 T
s( )s(v)( v)d dv

N0
2 0 0

N0 T 2
s ( )d = N0 A2
2 0

(d) For antipodal equiprobable signals the probability of error is

"  
S
P (e) = Q
N o
where

S
N o

is the output SNR from the matched lter. Since



S
N


=
o

4A4
y 2 (T )
=
N0 A2
E[n2T ]

"

we obtain

P (e) = Q

4A2
N0

75. Montecarlo simulation of a binary transmission system over an AWGN channel.

(a) Download and execute the Matlab script integrate and dump.m from the web site of the
class. Upon completion, a plot shows waveforms associated with the transmission of ten
random bits over an AWGN channel, with a rectangular pulse shape and an integrateand-dump receiver. Execute the script with your student ID, an amplitude a = 1 and
with N0 = 1. Sketch or print the plot.
69

(b) Download and execute the Matlab scripts intdmp simulation.m and Q.m from the web
site of the class. The rst script simulates the transmission of random bits over and
ideal AWGN channel and computes the bit error rate (BER) as a function of the signalSNR
to-noise ratio (SNR), Es /N0 in dB. The script will plot the simulated BER versus
!
as well as the theoretical expression for the bit error probability P [e] = Q( 2Es /N0 ).
Execute the script with your student ID and sketch or print the resulting plot.
NOTE: The script may take several minutes to nish.
Solution:
(a)
Original bit sequence

Received pulse sequence

2

1
0.5

0
1

2
20
40
60
80
Transmitted pulse sequence

100

20
40
60
80
100
Integrate and dump receiver output

1
20
40
60
80
Received pulse sequence

100

20

40
60
80
Estimated bit sequence

100

1
0

0.5

1
0

2
20

40
60
samples

80

100

20

70

40
60
samples

80

100

Error performance of binary transimission over an AWGN chan nel

10

Simulated
Theory

10

Bit error rate

10

10

10

10

5
Eb/N0 (dB)

10

(b)
76. In computer communications at 10 Mbps using the Ethernet standard (IEEE 802.3), Manchester
or bi-phase pulse formatting is used. Polar mapping is employed such that, in the interval
[0, T ], a 0 is sent as s0 (t) = a g(t), and a 1 is sent as s1 (t) = a g(t), where the pulse
shape gT (t) is sketched in the gure below:
gT(t)

1/T

t
T/2

1/T

(a) Find the impulse response h(t) of the matched lter (MF) for gT (t).
(b) Assume that a 0 is sent and no noise is present. The input of the MF is r(t) = s0 (t).
Find the corresponding response y(t) = r(t) h(t), sampled at t = T . i.e., Y = y(T ).
What is the value of Y if a 1 is sent?
(c) A key advantage of the Manchester format is its capability to detect collisions. To see
this, consider two bit sequences from two users. User 1 transmits the sequence 001101
71

and user 2 the sequence 011011. Sketch the corresponding transmitted sequences
s1 (t) and s2 (t), as well as the received sequence s1 (t) + s2 (t). Comment on your results.
Specically, how is a collision detected?

Solution:
(a)
gT(t)
h(t)

1/T
1/T

t
T/2

T
T/2

1/T

1/T

Pulse shape gT (t) and impulse response of matched lter.

(b) Y = +a. If a 1 is sent, then Y = a.

(c)
s1(t)
a

1/T

t/T

-a

1/T

s2(t)
a

1/T

t/T

-a

1/T

[s2(t)+s2(t)]/2
a

1/T

t/T
1

-a

1/T

Collisions are detected in the received sequence s1 (t)+ s2 (t) by the absence of transitions
at times nT + T /2, for n = 1, 3, 4.
72

77. Download the Matlab script rf pulse mfoutput.m from the class website. This script computes the output of a correlator for an RF pulse,
'
2
t [0, T ],
cos(2f0 t),
g(t) =
T
where f0 = n0 /T , and n0 is a positive integer. You are required to run the script and record
the resulting plots for two values of n0 , n0 = 2 and n0 = 20. What conclusion can you draw
from these plots?
Solution:

RF pulse: sqrt(2/T)*cos(2f0t), f0=1

1
0.5
0
0.5
1
0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

1.8

1.4

1.6

1.8

Correlator output
1
0.8
0.6
0.4
0.2
0

0.2

0.4

0.6

0.8

73

1.2

RF pulse: sqrt(2/T)*cos(2f0t), f0=10

1
0.5
0
0.5
1
0

0.2

0.4

0.6

0.8

1.2

1.4

1.6

1.8

1.4

1.6

1.8

Correlator output
1
0.8
0.6
0.4
0.2
0

0.2

0.4

0.6

0.8

1.2

The MF output approaches a ramp as the value of the center frequency f0 increases.

78. Compare NRZ and RZ signaling techniques in terms of probability of a bit error P [e]. The
bit rate R = 1/T and pulse amplitude a are xed.
(a) Plot or sketch the curve of energy-to-noise ratio E/N0 (dB) versus probability of P [e].
(Hint: The energy of an RZ pulse is one half that of an NRZ pulse.)
(b) How should the rate R be modied in the case of RZ signaling so that its P [e] is the
same as NRZ signaling?
Solution:
(a)

74

10

10

10

RZ
3

10

P[e]

10

10

NRZ

10

10

10

10

8
E/N0 (dB)

10

12

14

16

(b) To achieve the same P [e] as NRZ, for the same peak amplitude constraints, the rate
of RZ should be 50% smaller than that of NRZ. To see this, consider as an example
recangular pulse signaling with polar mapping. Then EN RZ = A2 T and ERZ = A2 T /2.
Therefore, the period T in RZ should be doubled.

79. Compare polar and unipolar bit mapping techniques in terms of P [e]. The peak pulse energy
is xed. Plot or sketch curve of energy-to-noise ratio E/N0 (dB) versus probability of abit
error for both techniques.
(Hint: The distance between the signal points d12 is d12 = 2 E

for polar and d12 = E for unipolar.)

Solution: If the peak pulse energy E is xed. Then for polar mapping

'
2E
,
P [e] = Q
N0
and for unipolar mapping

'
P [e] = Q

E
2N0


.

The gure below shows the plots of E/N0 in dB versus P [e].

75

10

10

10

Unipolar

10

P[e]

10

10

Polar

10

10

10

10

10
E/N (dB)

12

14

16

18

80. The bit stream {bn } = 0, 1, 1, 0, 0, 0, 1, 0 is to be sent through a channel (lowpass LTI system
with large bandwidth). Assume that rectangular pulses of amplitude A are used and the bit
rate is 1/T bps. In polar mapping, use the rule:
bn
0
1

an
A
+A

Sketch the transmitted signal for each of the following line coding schemes:
(a) Unipolar NRZ
(b) Unipolar RZ
(c) Polar NRZ
(d) Polar RZ
(e) AMI-NRZ (Assume that A is the inital state).
(f) AMI-RZ (Assume that A is the inital state).
(g) Manchester
Solution: (Line coding. Bit stream {bn } = 0, 1, 1, 0, 0, 0, 1, 0.) The transmitted signals are
shown in the gure below (amplitude A = 2), and can be reproduced using the Matlab script
homework6s05.m available in the class web site.

76

Bits

UNRZ

0
0

2
1
0

URZ

2
1
0

PNRZ

2
0
2

PRZ

2
0

AMIRZ

AMINRZ

2
2
0
2
2
0

Manchester

2
2
0
2

77

81. Compare the seven schemes in problem 1, in terms of average (DC) power and average (DC)
amplitude level.
Solution:
Technique

Pave

ak

U-NRZ

A2 T
2

A
2

U-RZ

A2 T
4

A
4

P-NRZ

A2 T

P-RZ

A2 T
2

AMI-NRZ

A2 T
2

AMI-RZ

A2 T
4

Manhester

A2 T

82. Manchester coding has the desirable feature that it is possible to detect the presence of errors
in the received signal. Explain how this is achieved. Sketch a block diagram of an error
detection circuit.
Solution:
Manchester coding has the desirable feature that it is possible to detect the presence of errors
in the received signal. This is done by checking that there is always a transition in the middle
of a bit period. A simple block diagram on to achieve this is shown below:
XOR
input

flag

+
t=T/2

Delay
T/2

t=T

The input is assumed to be sampled at a proper timing phase t = kT /2 + , so that the XOR
gate outputs a 1 at least every other sample (spaced by T /2). The second sampler checks
that there is a transition every bit. If this is the case, then the ag is always equal to 1;
otherwise, the ag is 0 and an error is detected.

78

A Simulink model based on this idea can be found in the web site of the class under the
name ETHERNET ERROR CHECK.mdl. After a transition period, the system always outputs a
1 whenever the input is Manchester coded, and a 0 whenever there is no transition in (at
least) the following bit.

83. Demodulation of binary antipodal signals

(
s1 (t) = s2 (t) =

0 t T;

Eb
T ,

0, otherwise.

can be accomplished by an integrate-and-dump receiver followed by a detector (threshold

device).
(a) Determine the SNR at the output of the integrator sampled at t = T . (Hint: Show that
the signal power at the output of the integrator, sampled at t = T , is Es = Eb T . The
noise power at the output of the integrator, sampled at t = T , is Pn = N20 T . The output
SNR is then the ratio Es /Pn .)
(b) Without any computation, what is the answer to part (a)? (See class notes.
The
integrator is the correlator of a rectangular pulse of width T and amplitude 1/ T .)
Solution:
(a) The output of the integrator is

[si ( ) + n( )]d
0

si ( )d +

=
0

At time t = T we have

y(T ) =

r( )d =

y(t) =

n( )d
0

si ( )d +

'
n( )d =

Eb
T+
T

n( )d
0

The signal energy at the output of the integrator at t = T is

2
 '
Eb
T
= Eb T.
Es =
T
The noise power can be computed as follows:
 T  T

n( )n(v)d dv
Pn = E


0
T

=
0

=
=
=

0
 T

E[n( )n(v)]d dv
 T
( v)d dv
0
 T
( v)d dv

N0
2 0

N0 T
2 0
0

N0
N0 T
T
1 dv =
2 0
2
79

(5)
(6)

Hence, the output SNR is

SNR =

2Eb
Es
=
Pn
N0

(b) Without any computation, what is the answer to part (a)?

The signals can be expressed as
!
si (t) = Eb (t),

i = 1, 2,

0 < t T,

where (t) is a rectangular pulse of unit energy, i.e., amplitude 1/ T and duration T .
The lter matched to (t) has impulse response h(t)
t). Its output y(t) when
= (T
s(t) is the input sampled at t = T is y(T ) = Y = Eb E = Eb , and has energy Eb .
Also, when the input is AWGN, the output is zero mean with variance N0 /2E = N0 /2.
Therefore,
2Eb
SNR =
N0
84. A binary communication system employs the signals
s0 = 0, 0 t T ;
s1 = A, 0 t T,
for transmission of the information. This is called on-o signaling. The demodulator crosscorrelates the received signal r(t) with s1 (t) and sampled the output of the correlator at
t = T.
(a) Determine the optimum detector for an AWGN channel and the optimum threshol,
assuming the the signals are equally probable
(b) Find the probability of error as a function of the SNR. How does on-o signaling compare
with antipodal signaling?
Solution:
(a) The received signal may be expressed as

n(t)
if s0 (t) was transmitted
r(t) =
A + n(t) if s1 (t) was transmitted
Assuming that s(t) has unit energy, then the sampled outputs of the correlators are
r = sm + n,

m = 0, 1

where s0 = 0, s1 = A T and the noise term n is a zero-mean Gaussian random variable

with variance


 T
 T
1
1
2
n(t)dt
n( )d
n = E
T 0
T 0
 
1 T T
E [n(t)n( )] dtd
=
T 0 0
 
N0
N0 T T
(t )dtd =
=
2T 0 0
2

80

The probability density function for the sampled output is

f (r|s0 ) =
f (r|s1 ) =

2
1
r
e N0
N0
2
T)
1
(rA
N
0

e
N0

Since the signals are equally probable, the optimal detector decides in favor of s0 if
PM(r, s0 ) = f (r|s0 ) > f (r|s1 ) = PM(r, s1 )
otherwise it decides in favor of s1 . The decision rule may be expressed as

(rA T )2 r 2
(2rA T )A
PM(r, s0 )

N0
N0
=e
=e
PM(r, s1 )

s0
>
<
s1

or equivalently
s1
1
r >
< 2A T
s0

The optimum threshold is 12 A T .

(b) The average probability of error is
P (e) =
=
=

=
=

1
1
P (e|s0 ) + P (e|s1 )
2
2

 1
1 2A T
1
f (r|s0 )dr +
f (r|s1 )dr
2 1 AT
2
2
2

 1
T)
r2
1 2A T
1
1
1
N
(rA
N
0 dr +
0

e
e
dr
2 1 AT N0
2
N0
2

 1  2 A T
2
2
N0
x
x2
1
1
1
1
e 2 dx +
e 2 dx


2 1 2 A T 2
2
2
2
N0
'




2
1
A T = Q SNR
Q
2 N0

where
SNR =

1 2
2A T

N0
Thus, the on-o signaling requires a factor of two more energy to achieve the same
probability of error as the antipodal signaling.

85. SPRING 2004 Homework 4

Multi-dimensional signals
81

86. In class, the rst three steps of the Gramm-Schmidt procedure for problem 7.6 in the textbook
were completed.

(a) Perform step 4 of the procedure and nd the 3-dimensional basis, = {1 (t), 2 (t), 3 (t)},
that contains the 4 signals sm (t), m = 1, 2, 3, 4. Represent the signals as points sm ,
m = 1, 2, 3, 4, in this three-dimensional signal space.
(b) Verify that the distance to the origin (or norm) of the points sm in the signal space
with basis of part (a) is the same as that of the signal points obtained with the basis

 = {1 (t), 2 (t), 3 (t)}, where n (t), n = 1, 2, 3, are illustrated in Fig. 1 below.
1(t)

2(t)

3(t)

t
1

t
1

Orthonormal waverforms in basis  .

Solution:
(a) In class it was shown that
1
, 0 < t 3,
3
(
2,
0 < t 1;
3
(
2 (t) =
1 2 , 1 < t 3,
2
3
3 (t) = 0.
1 (t) =

Compute the projections, c41 and c42 , of s4 (t) onto the signals 1 (t) and 2 (t), respectively,
 3
4
s4 (t)1 (t)dt = ,
c41 =
3
0
'
 3
2
.
s4 (t)2 (t)dt =
c42 =
3
0
Then form the dierence signal

0 < t 1;
0,
d4 (t) = s4 (t) c41 1 (t) c42 2 (t) = 1,
1 < t 2;

1, 2 < t 3.
The energy of d4 (t) is Ed4 = 2 and it follows that

0 < t 1;

0,
d4 (t)
1
1 < t 2;
= 2 ,
4 (t) = !

Ed4

1 , 2 < t 3.
2

82

The coordinates of the vector representation sm = (sm1 sm2 sm3 ) of the signal sm (t),
with respect to the basis = {1 (t), 2 (t), 4 (t)}, 1 m 4, are computed as


smn =

1 m 4,

sm (t)n (t),

1 n 3.

This results in


2 3 0 0


(
2
2

=
2 3 0
3


(
=
43 2 23 0


(
2
4
2
=
3

s1 =
s2
s3
s4

The energies (squared distances to the origin) of the signals are computed from their
vector representation as
Em =

3

|smn |2 ,

1 m 4.

n=1

It follows that
E1 = 12,

E2 =

4+8
= 4,
3

E3 =

16 + 8
= 8,
3

E4 =

16 + 2
+ 2 = 8.
3

(b) In this case, by inspection, the vector representation of the signals is easy to obtain:


2 2 2
s1 =


2 0 0
s2 =


0 2 2
s3 =


2 2 0
s4 =
Note that s3 (t) = s2 (t) s1 (t) and therefore the dimensionality of the signal space
spanned by  is 3. It is easy to check that the signal energies obtained from this vector
representation are the same as those obtained in part (a) above.

(
(
87. The signals 1 (t) = T2 cos(2fc t) and 2 (t) = T2 sin(2fc t), for (k 1)T < t kT , where
k is an integer and T is the symbol duration, are used as basis functions for bandpass digital
modulation.
Show that 1 (t) and 2 (t) are orthogonal over the interval [(k 1)T, kT ].
Solution: It is relatively easy to show that


kT
(k1)T

2
1 (t)2 (t) dt =
T

kT

cos(2fc t) sin(2fc t) dt = 0.
(k1)T

83

88. The gure below depicts two pulse shapes used in a binary orthogonal signaling scheme
know as binary PPM (or 2-PPM). Transmission takes place over an AWGN channel with
SN (f ) = N20 W/Hz.

s1(t)

s2(t)

t
T/2

T/2

(a) Find the value of the amplitude a in terms of the energy per bit Eb and the bit duration
T.
(b) Determine the orthonormal basis signals 1 (t) and 2 (t) for 2-PPM and sketch the
location of the signal points s1 and s2 , corresponding to signals s1 (t) and s2 (t), in the
two-dimensional signal space.
(c) Give an expression
 probability of a bit error. (Hint: Use the general expression
! for the
2
d12 /2N0 .)
P [error] = Q
Solution:
(
b
(a) a = 2E
T .
(b)
1(t)

2(t)

t
T/2

T/2

2(t)

1(t)
E

84

 (
(c) P [error] = Q

Eb
N0


.

89. The two pulses shown in the gure below are employed in a binary orthogonal signaling
scheme to be used in a digital communication system.

s1(t)

s2(t)

a
t

t
T

T/2
-a

-a

Pulses used in a binary orthogonal signaling scheme.

(a) Sketch carefully the impulse responses of a pair of matched lters for this system.
(b) Find the means and variances of the outputs of the matched lters, sampled at t = T ,
when s1 (t) is sent through an AWGN channel with noise energy N0 /2.
(c) A correlator receiver is employed. However, due to a timing error, the output is incorrectly sampled at t = 0.75T . What is the increase in the probability of a bit error, Pb ,
compared to that of an ideal system?

Solution:
(a) Matched lters:
h2(t)

h1(t)

a
t

T/2

T
-a

-a

(b) When s1 (t) is sent, Y1 is a Gaussian r.v. of mean m1 = E1 = a2 T /2 and variance

12 = N20 E1 = N20 (a2 T /2), and Y2 is a Gaussian r.v. of mean m2 = 0 and variance
22 = N20 E2 = N20 (a2 T /2).
Alternatively,
dening the orthonormal functions 1 (t) and 2 (t), we can write s1 (t) =
!
a T /2 1 (t). In this case, the outputs of the lters matched to the orthonormal!functions, given that s1 (t) is sent, are such that Y1 is a Gaussian r.v. of mean m1 = a T /2
and variance 12 = N20 , and Y2 is a Gaussian r.v. of mean m2 = 0 and variance 22 = N20 .

85

(c) Correlator outputs:

y1(t)

y2(t)

a2T/2

a2T/2

t
T/2

T/2

(a)

(b)

(a) First output when s1 (t) sent. (b) Second output when s2 (t) sent.

When sampled at t = 0.75T , the mean and variance of Y1 , given that s1 (t) is sent,
2
2
remain at m1 = a 2T and 12 = N20 a 2T . However, when s2 (t) is sent, the mean and
2 
2 
2 
2
variance of Y2 change to m2 = (0.5) a 2T = a 4T and 12 = (0.5) N20 a 2T = N20 a 4T .
From this observation, it follows that
"

' 
2T
E
a
, and
= Q
P [error|
s1 sent] = Q
N0
2N0
"

'

2
E
a T
= Q
.
P [error|
s2 sent] = Q
2 N0
4N0
As a result, the average probability of a bit error becomes
"

"

2
2
1
a T
a T
+ Q
.
Pb = Q
2
2N0
4N0
In conventional binary orthogonal signaling,
"
Pb = Q

a2 T
2N0

The increase in the probability of a bit error, due to the timing error, is
"

"

2
2
1
a T
a T
Q
.
Pb = Q
2
4N0
2N0

Pulse-amplitude modulation (PAM)

90. A binary PAM communication system is used to transmit data over an AWGN channel. The
prior probabilities for the bits are Pr{am = 1} = 1/3 and Pr{am = 1} = 2/3.
86

(a) Determine the optimum maximum-likelihood decision rule for the detector
(b) Find the average probability of a bit error
Solution:
(a) The optimum threshold is given by
N0
1p
N0
= ln 2
= ln
p
4 Eb
4 Eb
(b) The average probability of error is ( =


N0

4 Eb

ln 2)

2
1
e(r+ Eb ) /N0 dr
N0


2
1

e(r Eb ) /N0 dr
+p(am = 1)
N0


1
+ Eb
Eb
2
+ Q !
Q !
=
3
3
N0 /2
N0 /2

'


!
'
2N0 /Eb ln 2
2N0 /Eb ln 2
2Eb
2Eb
1
2
Q
+

+ Q
=
3
4
N0
3
N0
4

P (e) = p(am = 1)

91. Determine the average energy of a set of M PAM signals of the form
sm (t) = sm (t),
where
sm =

m = 1, 2, , M,

Eg Am ,

0 t T,

m = 1, 2, , M.

The signals are equally probable with amplitudes that are symmetric about zero and are
uniformly spaced with distance d between adjacent ampltiudes.
Solution: The amplitudes Am take the values
d
Am = (2m 1 M ) ,
2

m = 1, . . . M

Hence, the average energy is

Eave =

=
=
=

M
M


1
2
d2
Eg
sm =
(2m 1 M )2
M m=1
4M
m=1
M


d2
Eg
[4m2 + (M + 1)2 4m(M + 1)]
4M
m=1

 M
M




d2
2
2
Eg 4
m + M (M + 1) 4(M + 1)
m
4M
m=1
m=1


M (M + 1)
M (M + 1)(2M + 1)
d2
2
Eg 4
+ M (M + 1) 4(M + 1)
4M
6
2
2
2
M 1d
Eg .
3
4

87

Moreover, note that if a = d/2 and Eg = 1, as we did in class, then

Eave =

M2 1 2
a ,
3

'

and therefore
a=

3 Eave
.
M2 1

92. A speech signal is sampled at a rate of 8 KHz, logarithmically compressed and encoded into
a PCM format using 8 bits per sample. The PCM data is transmitted through an AWGN
baseband channelvia M -level PAM signaling. Determine the required transmission bandwidth
when (a) M = 4, (b) M = 8 and (c) M = 16. (Assume rectangular pulses and the zero-to-null
denition of bandwidth.)
Solution: The bandwidth required for transmission of an M -ary PAM signal is
W =

Rb
Hz
2 log2 M

The bit rate can be obtained from the data given as

Rb = 8 103
As a result,

bits
samples
bits
8
= 64 103
sec
sample
sec

M = 4;
16 KHz,
W =
10.667 KHz, M = 8;

8 KHz,
M = 16.

93. HOMEWORK 9, Spring 2005. Problems 1 and 3

Quadrature-amplitude modulation (QAM)

94. The 16-QAM signal constellation shown in the gure below is an international standard for
(analog) telephone-line modems known as V.29. Determine the decision boundaries for the
detector, assuming equiprobable signals and high SNR so that errors only occur between
adjacent points.

88

-5

-3

-1

-1

-3

-5

Solution: The optimum decision boundary of a point is determined by the perpendicular

bisectors of each line segment connecting the point with its neighbors. The decision regions
for the V.29 constellation are sketched in the next gure:

95. A digital communication system transmits data using QAM signaling over a voice-band telephone channel at a rate 2400 symbols/s (baud). The additive noise is assumed to be white
and Gaussian. You are asked to determine the energy-per-bit-to-noise ratio Eb /N0 required
to achieve an error probability of 105 for a bit rate equal to:
(a) 4800 bits/s
(b) 9600 bits/s
(c) 19200 bits/s
89

(d) 31200 bits/s

(e) What conclusions can you draw from parts (a) to (d)?
(Hint: Use the following expression for the error probability of an M -ary QAM system:
"
PM 4 Q

3Eb
(M 1)N0


,

where M = 2 and  is the number of bits per symbol. Also, if needed, use the following
approximation of the Gaussian Q-function:
Q(x)

1 x2
e 2 .)
2

Solution:
We assume that ideal Nyquist signaling is employed so that the bandwidth of the bandpass
signal is equal to the baud rate (symbols/second), i.e., 1/T Hz.
(a) The number of bits per symbol is
4800
4800
=
= 2.
W
2400
Thus, a 4-QAM constellation is used for transmission. Using the given approximation
of the probability of error for an M-ary QAM system, with PM = 105 and  = 2 we
obtain

'
Eb
2Eb
= 10.58 = 10.24 dB,
= 105 =
4Q
N0
N0
=

where Table 5.1, on page 220 of the textbook has been used.
(b) If the bit rate of transmission is 9600 bps, then
9600
= 4.
2400
In this case a 16-QAM constellation is used and

'
Eb
4 Eb
= 26.45 = 14.22 dB.
= 105 =
4Q
15 N0
N0
=

(c) If the bit rate of transmission is 19200 bps, then

19200
=8
2400
In this case a 256-QAM constellation is used and
k=

Eb
= 224.82 = 23.52 dB.
N0
(d) If the bit rate of transmission is 31200 bps, then
31200
= 13
2400
In this case an 8192-QAM constellation is used and
k=

Eb
= 4444.14 = 36.48 dB.
N0
90

(e) The following table gives the SNR per bit and the corresponding number of bits per
symbol for the constellations used in parts (a)-(d):

SNR per bit (dB)

2
10.24

4
14.22

8
23.52

13
36.48

As observed in the next plot, there is an increase in average transmitted power of approximately 2.5 dB per additional bit per symbol:
5

SNR per bit required to achieve P = 10

e

40

35

SNR per bit

30

25

20

15

10

10

12

14

Number of bits per symbol

96. Plots of the approximated probability of a bit error Pb are compared with the simulated BER,
for M -PSK and M -QAM, in Figs. 1 and 2, respectively, with Matlab scripts digmodMPSK.m
and digmodMQAM.m. The approximated Pb values were produced by the Matlab scripts
digmodMPSK approx.m and digmodMQAM approx.m posted in the web site.

91

Error performance of MPSK. SJSU. EE160 Spring 2004

10

BPSK
QPSK
8PSK
1

10

10

BER

10

10

10

10

10

10
E/N0 (dB)

12

14

16

18

20

Simulation (blue solid line) versus approximated Pb (red dotted line) for M -PSK.

92

Error performance of MQAM. SJSU Spring 2004

10

QPSK
16QAM
64QAM
1

10

BER

10

10

10

10

15

20

25

E/N0 (dB)

Simulation (blue solid line) versus approximated Pb (red dotted line) for M -QAM.

93

97. HOMEWORK 9, Spring 2005. Problems 2, 4c and 5

Phase-shift keying (PSK) modulation

98. Binary PSK (BPSK) is used for data transmission over an AWGN channel with power spectral
density N0 /2 = 1010 W/Hz. The transmitted signal energy is Eb = A2 T /2, where T is the
bit duration and A is the signal amplitude. Determine the value of A needed to achieve an
error probability of 106 , if the data rate is:
(a) 10 Kbit/s
(b) 100 Kbit/s
(c) 1 Mbit/s
Solution: For binary phase modulation, the error probability is
"

'

2T
2Eb
A

= Q
P2 = Q
N0
N0
With P2 = 106 we nd from tables that
"
A2 T
= 4.74 = A2 T = 44.9352 1010
N0
If the data rate is 10 Kbps, then the bit interval is T = 104 and therefore, the signal
amplitude is
!
A = 44.9352 1010 104 = 6.7034 103
Similarly we nd that when the rate is 105 bps and 106 bps, the required amplitude of the
signal is A = 2.12 102 and A = 6.703 102 respectively.

99. A QPSK modulated signal can be written as

'
2E
cos(2fc t + i ),
si (t) =
T
where

i = (i 1) ,
2

(k 1)T < t kT

i = 1, 2, 3, 4.

(a) Express si (t) in terms of the basis functions 1 (t) and 2 (t) dened in Problem 1. (Hint:
Use a well-known trigonometric identity.)
(b) Represent the signals si (t), i = 1, 2, 3, 4, as vectors in the 1 2 -plane (or IQ-plane).
(c) Sketch carefully the decision regions Zi , i = 1, 2, 3, 4, in the 1 2 -plane.
Solution:

94

(a) Using the trigonometric identity cos(A + B) = cos(A) cos(B) sin(A) sin(B), we have
that
'
2E
cos(2fc t + i )
si (t) =
T
'
2E
=
[cos(i ) cos(2fc t) sin(i ) sin(2fc t)]
T

E cos(i ) 1 (t) E sin(i ) 2 (t)

=

(b) Representation in the 1 2 -plane (note clockwise numbering of signals):

2
s4

s3

s1
E

s2

(c) Decision regions:

Z4
2
s4

s3

Z3
-

s1

Z1

s2

Z2

100. Using a computer environment (e.g., Matlab), compare the exact probability of a symbol
error, P [], with the union bound. What is the minimum value of E/N0 (dB) after which the
union bound is tight?
95

Solution: With a Matlab script, we may compare the exact expression

' 
' 2
E
E
,
Q
P [] = 2 Q
N0
N0
'

with the union bound

P [] = 2 Q

E
N0

'
+Q

2E
N0


.

The plots are shown in the graph below:

0

10

Union
bound
Exact

P[]

10

10

10

10

E/N0 (dB)

Note that the bound is tight after a minimum value of approximately (E/N0 )min = 9 dB.
The Matlab script used to compute the points in the graph is the following:
esno=[-3 -2 -1 0 1 2 3 4 5 6 7 8 9 10];
esnor=10.^(esno/10);
pexct=2*erfc(sqrt(esnor)/sqrt(2))/2 - (erfc(sqrt(esnor)/sqrt(2))/2).^2;
pebnd=2*erfc(sqrt(esnor)/sqrt(2))/2 + erfc(sqrt(2*esnor)/sqrt(2))/2;
semilogy(esno,pexct);
hold on;
semilogy(esno,pebnd);

96

101. HOMEWORK 7, Spring 2004. Problems 2 and 3

102. Consider a QPSK communication system over an AWGN channel with SN (f ) =

Bit values 0 and 1 are assumed to be equally likely.

N0
2

W/Hz.

(a) Using the fact that Q(x) < Q(y) for x > y, show that the probability of a bit error,
Pb with Gray-mapped QPSK modulation is approximately the same as that with BPSK
modulation,
as a function of the energy per bit-to-noise ratio Eb /N0 . (Hint: P [error]
!
E/N0 , and E = 2 Eb .)
Q
(b) Using the signals 1 (t) and 2 (t) from problem 2, sketch carefully the signals si (t)
corresponding to the points si , for i = 1, 2, 3, 4, in the QPSK constellation shown in the
gure below.
2(t)

s2

s1

E
1(t)

s3

s4

Solution:
(a) Since Q(x) < Q(y) for x > y, P [error] is dominated by the probability of transmitting a
signal point and making a decision favorable to one !
of two possible
nearest points. The

distance between two nearest signal points is d12 = 2 E/2 = 2E. From this it follows
that

'
E
.
P [error] 2 Q
N0
Since each QPSK signal carries two bits, E = 2Eb . In addition, with Gray mapping,
Pb = P [error]/2. As a result,
'

2Eb
.
Pb Q
N0
(b)

97

s1(t)

s2(t)

t
T/2

T/2

E
T

s3(t)

s4(t)

t
T/2

T/2

E
T

103. MATLAB experiment

Simulate the error performance of M-PSK (M = 2, 4, 8) and M-QAM (M = 4, 16, 64) digital
modulation schemes. The scripts digmodMQAM.m and digmodMPSK.m are available in the web
site of the course, in section MATLAB EXAMPLES, under the heading Simulation of M-PSK
and M-QAM modulation over AWGN channels.
(a) Run the script, sketch or plot the resulting curves and include them in your solution.
(b) Compare the simulated error performance of QPSK and 16-QAM against the approximated expressions given in the book. You will need to write a script to plot the approximated expressions and the simulation results in the same graph. Refer to the notes on
digital modulation in the web site for the expressions.

Solution: The gures below show plots of simulated (symbols) and theoretical (symbols and
line) BER for M -PSK and M -QAM, respetively. For BER values below 102 , the theoretical
expressions match the simulation results very closely.

98

Error performance of MPSK. SJSU Fall 2004.

10

BPSK
QPSK
8PSK

10

BER

10

10

10

10

10

10
Es/N0 (dB)

15

20

Simulated and theoretical BER performance of M -PSK modulation.

Error performance of MQAM. SJSU Fall 2004.

10

QPSK
16QAM
64QAM

10

BER

10

10

10

10

10

10

15
Es/N0 (dB)

20

25

30

Simulated and theoretical BER performance of M -QAM modulation.

99

104. A digital communication system is designed using bandpass QPSK modulation with Gray
mapping. Due to a DC oset and phase error in the modulator, the constellation is translated
and rotated as shown in the gure below.
2(t)

s1
a

2a

4a

1(t)

s4
s2

s3

A translated and rotated QPSK constellation.

(a) Compute the average energy of the signals and sketch carefully the decision regions.
(b) Find the probability of a bit error Pb of this system.
(c) If a demodulator for conventional QPSK is employed, what is the resulting value of Pb ?

Solution:
(a) Average energy:
1

Ei
E =
4
i=1

 
 

1 
(2a)2 + (a)2 + (4a)2 + (2a)2 + (a)2 + (4a)2 + (a)2 + (a)2
=
4

1
44a2 = 11 a2 .
=
4
!
From which it follows that a = E/11.
4

The decision regions are shown in the gure below:

100

2(t)

Z1
s1

Z4
a

2a

1(t)

4a

s4

Z2

s2

s3

Z3
Decision regions of a translated and rotated QPSK constellation.

(b) Note that

d212 = d223 = d234 = d241 = 4a2 + 9a2 = 13 a2 .
As a result,

"
P [error] 2 Q


'
d212
13E
,
= 2Q
2N0
22N0
'

and
Pb Q

26Eb
22N0


.

 
Compared to conventional QPSK, there is a loss of 10 log 10 44
26 = 2.28 dB in SNR.
This is due to the DC oset. That is, the average of the signal points is not the origin,
resulting in an increase of the average probability of a bit error.

(c) In conventional QPSK, the decision regions Zi , i = 1, 2, 3, 4, are the quadrants of the
Y1 /Y2 -plane. Consequently, the probability of error is given by
1

P [error|
si ],
4
4

P [error] =

i=1

where

'
/
P [error|
s1 ] = P [Y

Z1 |
s1 ]

= 1Q

101

8E
11N0

1Q

'

2E
11N0


,

'
P [error|
s2 ] = P [Y
/ Z2 |
s2 ] = 1 Q
'
/ Z3 |
s3 ] Q
P [error|
s3 ] = P [Y

2E
11N0

'
/
P [error|
s4 ] = P [Y

Z4 |
s4 ]

2Q

8E
11N0

1Q

2E
11N0

'

'
+Q

32E
11N0

32E
11N0


,


,


,

where the fact that Q(x) > Q(y), for x < y, has been used. It follows that

'
1
2E
2 + 3Q
P [error]
,
4
11N0
and therefore, for QPSK modulation (i.e., 2 bits per symbol), the desired result is

'
1
4Eb
2 + 3Q
.
Pb
8
11N0
Finally, note that at high SNR values, bit error events are dominated by error events
associated with the transmission of signal points s1 and s2 , which are located in the
wrong quadrants. Therefore,
Pb

1
,
4

for high values of Eb /N0 .

In this case, we say that there is an error oor. The value of Pb is never less than
0.25, regardless of how high the value of Eb /N0 is!!!

Frequency-shift keying (FSK) modulation

105. The transmitter of a BFSK communication system sends an RF rectangular pulse sm (t), for
m = 1, 2, in the interval 0 t T , and in correspondence to the value of a source bit
M {0, 1}, as follows:
M = 0 s1 (t) = a cos(2f1 t),
M = 1 s2 (t) = a cos(2f2 t),
1
. Source bits take
where a is the amplitude and the frequency separation is f2 f1 = 2T
values 0 and 1 with equal probability. Transmission takes place over an AWGN channel with
SN (f ) = N20 W/Hz.

(a) Find the probability of a bit error, P [error], in terms of the amplitude a and N0 .
(b) It is desired to transmit bits at a rate of 1 Mbps with P [error] = 105 . The receiver introduces AWGN with N0 = 1010 . Determine the minimum value amin of the amplitude
of the RF rectangular pulses.

102

(c) It is desired to increase the rate to 2 Mbps. The other conditions are the same as in part
(b). What is the minimum value amin of the amplitude of the RF rectangular pulses?
Express in dB the additional power required compared to part (b).

Solution:
(a) Note that si (t) =

2E
T

(
cos(2fi t), for i = 1, 2. In other words, a =

the energy of s1 (t) (or s2 (t)) is equal to E =

2E
T .

Alternatively,

a2 T

. It follows that

"

'
2
E
a T
.
= Q
P [error] = Q
N0
2N0
2

(b) We have that T = 106 and N0 = 1010 . Therefore

"


"
2
2 (106 )
a
T
a
=Q
.
P [error] = 105 = Q
2N0
2 (1010 )
5
From theMatlab
 function invQ.m given in the web site, we have that Q(x) = 10 , for
1
5
10
= 4.2649. Therefore,
x=Q

'

a2 (104 )
= 4.2649
2

and from this it follows that a2min = 3.64 103 V2 , or amin = 60.3 mV.

(c) The bit period is reduced by 1/2. Therefore amin = 2 (60.3) mV = 85.3 mV. Compared
to part (b), the additional signal power required to achieve a probability of a bit error
equal to 105 is 3 dB.

106. Find z = Eb /N0 required to give Pb = 105 for the following coherent digital modulation
schemes: (a) on-o keying (ASK), (b) BPSK, (c) BFSK and (d) BPSK with a phase error of
5 deg.
Solution:

z) = 105 gives z = 18.19 or z = 12.6 dB.

(a) For ASK, PE = Q(
(b) For BPSK, PE = Q( 2z) = 105 gives z = 9.1 or z = 9.59 dB.
(c) Binary FSK is the same as ASK.
(d) The degradation of BPSK with a phase error of 5 degrees is
Dconst = 20 log10 [cos(5o )] = 0.033 dB,
so the required SNR to give a bit error probability of 105 is 9.59 + 0.033 = 9.623 dB.

103

PAM over bandlimited channels

107. Binary data at a rate of 4 106 bits/s are transmitted using M -PAM over a bandlimited
channel, using the raised cosine roll-o characteristic shown in the gure below:
P(f)
T

T/2

f (MHz)
-1

-0.75 -0.5

-0.25

0.25

0.5

0.75

Raised cosine roll-o characteristic.

(a) Determine the roll-o factor and the value of M of this system.
(b) Signals are transmitted over a bandlimited AWGN channel. Determine the minimum
required bit energy-to-noise ratio, Eb /N0 , to achieve a probability of a bit error less than
103 . Express your result in decibels (dB).

Solution:
(a) Rb = 4 Mbits/s. Note that 1/2T = 0.5 MHz, the point of symmetry of the raised cosine.
As a result, 1/T = 1 MHz. Since Rb = 1/Tb = log2 M (1/T ), it follows that log2 M = 4
and therefore M = 16.
To nd the roll-o factor, note that 2(1/2T ) = 0.75 0.25 = 0.5 MHz (the region of
the cosine characteristic) and therefore, with 1/2T = 0.5 MHz, we obtain = 0.5.
(b) The probability of a bit error for M -PAM is
'

6 log2 M Eb
M 1
Q
.
P [] =
M
M 2 1 N0
For M = 16,
15
Q
P [] =
16

"

24
255

Eb
N0

< 1 103
min

From the table of the Q-function, we have Q(3.07) = 0.0011. Therefore,

"
8
85

Eb
N0


= 3.07

min

104

Eb
N0


=
min

85
(3.07)2 = 100.14 = 20 dB
8

108. Binary data are transmitted using 8-PAM over a bandlimited channel, using the raised cosine
roll-o characteristic shown in the gure below:
P(f)
T

T/2

f (MHz)
-4

-3.75 -2.5

-1.25

1.25

2.5

3.75

Raised cosine roll-o characteristic.

(a) Determine the roll-o factor and the bit rate, in bits/s, of this system.
(b) Signals are transmitted over a bandlimited AWGN channel. Determine the minimum
probability of a bit error if the bit energy-to-noise ratio is limited to a maximum of 20
dB.

Solution:
(a) M = 8, log2 M = 3. Note that 1/2T = 2.5 MHz, the point of symmetry of the raised
cosine. As a result, 1/T = 5 Mhz and the bit rate is therefore Rb = log2 M (1/T ) = 15
Mbits/s.
To nd the roll-o factor, note that 2(1/2T ) = 3.75 1.25 = 2.5 MHz (the region of
the cosine characteristic) and therefore, with 1/2T = 2.5 MHz, we obtain = 0.5.
(b) The probability of a bit error for M -PAM is
'

6 log2 M Eb
M 1
Q
.
P [] =
M
M 2 1 N0
For a bit energy-to-noise ratio equal to 20 dB, we have that

10 log10

Eb
N0


= 20

Eb
= 100.
N0

Consequently, the minimum bit error probability is

7
P [] = Q
8

'

6(3)
(100)
63


=

7
Q(5.345) = 3.96 108
8

109. HOMEWORK 4 Summer 2004, problems 1-4

The Nyquist criterion for ISI-free transmission

105

110. Plot the raised-cosine pulse for dierent values of rollo factor, . For this purpose, download
the script plot raised cosine pulse.m from the web site, execute it and sketch or print the
resulting waveforms.
Solution:
Raisedcosine pulses for various values of the rolloff fact or,
=1
=0.5
=0.25

0.8

pulse amplitude

0.6

0.4

0.2

0.2
4

0
t/T

Raised-cosine pulse for dierent rollo factors.

111. The eye diagram is useful in analyzing pulse shapes used in communication over bandlimited
channels. You will experiment with this tool using MATLAB. Download the script ayayay.m
from the web site.
(a) Run the script for the following pairs of values E/N0 (dB) and : [35, 1], [35, 0.5] and
[35, 0.25]. Print or sketch the resulting eye diagrams and comment on the eects of .
(b) Run the script for the following pairs of values of E/N0 (dB) and : [35, 1], [25, 1] and
[15, 1]. Print or sketch the resulting eye diagrams and comment on the eects of E/N0 .
Solution:

106

Eye diagram of raisedcosine pulse. E/N =35 dB, and =1

0

1.5

Amplitude

0.5

0.5

1.5
0.5

0.4

0.3

0.2

0.1
0
0.1
Normalized time, t/T

0.2

0.3

0.4

0.5

0.4

0.5

0.4

0.5

Eye diagram of raisedcosine pulse. E/N0 =35 dB, and =0.5

1.5

Amplitude

0.5

0.5

1.5

2
0.5

0.4

0.3

0.2

0.1
0
0.1
Normalized time, t/T

0.2

0.3

Eye diagram of raisedcosine pulse. E/N =35 dB, and =0.25

0

1.5

Amplitude

0.5

0.5

1.5

2
0.5

0.4

0.3

0.2

0.1
0
0.1
Normalized time, t/T

0.2

0.3

(a) It is observed from the eye diagrams that the timing error margin increases with the
value of . On the other hand, the occupied bandwidth grows linearly with the rollo
factor. Thus there is a tradeo between timing error margin and occupied bandwidth.

107

Eye diagram of raisedcosine pulse. E/N =35 dB, and =1

0

1.5

Amplitude

0.5

0.5

1.5
0.5

0.4

0.3

0.2

0.1
0
0.1
Normalized time, t/T

0.2

0.3

0.4

0.5

0.4

0.5

0.4

0.5

Eye diagram of raisedcosine pulse. E/N0 =25 dB, and =1

1.5

Amplitude

0.5

0.5

1.5
0.5

0.4

0.3

0.2

0.1
0
0.1
Normalized time, t/T

0.2

0.3

Eye diagram of raisedcosine pulse. E/N =15 dB, and =1

0

1.5

Amplitude

0.5

0.5

1.5

2
0.5

0.4

0.3

0.2

0.1
0
0.1
Normalized time, t/T

0.2

0.3

(b) The eye opening and timing error marging are reduced with increasing noise levels (decreasing SNR values).

108

PCM, quantization and delta modulation

112. A message signal m(t) is transmitted by binary PCM without compression. Let the signalto-quantization noise (SNRq) required be at least 47 dB. Determine the minimum number
of quantization levels L required, assuming that m(t) is sinusoidal. With this value of L,
determine the SNRq.
Solution: 1.8 + 6v 47. Therefore, v = 8 bits and L = 256. This gives SNRq = 49.8 dB.

113. In a delta modulation system, a sinusoidal signal m(t) = sin(6000t) is sampled at 30,000
samples/sec. Determine the minimum step size to avoid slope overload.


 d m(t) 

= 6000. As a result, 6000Ts = /5.
Solution: 
dt max

114. Find the maximum amplitude of a 1 KHz sinusoidal signal input to a delta modulator that
will prevent slope overload, when the sampling rate is 10,000 samples/sec and the step size
is = 0.1.
Solution:

1
= 10000. = 0.1. This implies that
Ts

Ts

2000 A

1000 2000 A
and consequently,
Amax =

1
.
2

115. Let m(t) = Am cos(2fm t) be the input to a delta modulator with parameters (step size)
and Ts (sampling period). Show that the minimum sampling frequency fs,min , needed to
avoid slope overload distorsion, is given by
fs,min =

2 fm Am
.

Solution: Note that



 d m(t) 


= |2fm Am sin(2fm t)|max = 2fm Am .
 dt 
rmmax
Therefore, to avoid slope overload:

2fm Am ,
Ts
109

from which it follows that

fs =

1
2fm Am

.
Ts

116. HOMEWORK 7 Spring 2004, problems 1 and 2

Error correcting coding

117. Data are encoded with a binary (7, 4) Hamming code and sent over a noisy channel. A
received vector is r = (1101111). Determine the syndrome s, the corresponding error vector
e and the estimated code vector v.
Solution: The syndrome is:

1
1


0 1 0 1 1 1 0 0
1

1 1 0 1 0 1 0 = 0 .
1
s = rT H =

1 0 1 1 1 0 0 1
1

1
1
Observe that the syndrome s is equal to the third column of H. The corresponding
error


vector is e = 0 0 1 0 0 0 0 , and the estimated code vector v = re = 1 1 1 1 1 1 1 .
118. Write a parity-check matrix of a binary (15, 11) Hamming code.
Solution: Any binary matrix with 15 columns consisting of all non-zero combinations of
m = 15 11 = 4 bits can be taken as a parity check matrix of a binary (15, 11) Hamming
code. An example is shown below:

0
0
H=
0
1

0
0
1
0

0
0
1
1

0
1
0
0

0
1
0
1

0
1
1
0

0
1
1
1

1
0
0
0

1
0
0
1

1
0
1
0

1
0
1
1

1
1
0
0

1
1
0
1

1
1
1
0

1
1

1
1

119. Consider the application of a (7,4) Hamming code in a digital communication system.
(a) If the received word is r = 0101101, what is the decoded message m
?
(b) This code is applied to a BPSK system operating at an SNR of 8 dB. Estimate the
probability of a bit error after decoding.
(c) (Bonus) Design the complete digital circuits, in terms of ip-ops and logic gates, of an
encoder and a decoder for this error correcting code.

110

Solution:
(a) Assuming systematic encoding and the generator

1 0 0
 0 1 0

G = I4 A =
0 0 1
0 0 0

matrix
0
0
0
1

0
1
1
1

1
1
0
1

1
0
,
1
1

the parity-check matrix is given by


H = AT

I3

0 1 1 1 1 0 0
= 1 1 0 1 0 1 0 .
1 0 1 1 0 0 1

The syndrome s corresponding to the received word

0
1



T
s = rH = 0 1 0 1 1 0 1
1
1

0
0

is computed as follows

1 1
1 0

0 1


1 1
=
1
0
0
.

0 0

1 0
0 1

The syndrome s is equal to the transpose of the fth column of the parity-check matrix
H. Consequently,
the most

 likely event is a single error in the fth position. That is,
e = 0 0 0 0 1 0 0 , and the estimated codeword is:

 
 

c = r e = 0 1 0 1 1 0 1 0 0 0 0 1 0 0 = 0 1 0 1 0 0 1 .

(b) We have BPSK modulation with 10 log10 NE0 = 8 dB or NE0 = 6.3. The probability of
error without coding is

'
2E
= Q(3.55) = 2.33 104 .
=Q
N0
With coding,

Pb 32 = 1.63 107 .

111

(c) Digital logic design of encoder and decoder of the binary (7,4) Hamming code.
1. Encoder Let m
represent the message bits. Then


c = c1 c2 c3 c4 c5 c6 c7

1 0

 0 1
= mG
= m1 m2 m3 m4
0 0
0 0

0
0
1
0

0
0
0
1

0
1
1
1

1
1
0
1

1
0
,
1
1

and it follows that the coded bits are

c1
c2
c3
c4
c5
c6
c7

=
=
=
=
=
=
=

m1
m2
m3
m4
m2 m3 m4
m1 m2 m4
m1 m3 m4

2. Decoder The syndrome s is computed as



s = s1 s2 s3


= rH T = r1 r2 r3 r4 r5 r6

0
1

 1
r7
1
1

0
0

1
1
0
1
0
1
0

1
0

1
,
0

0
1

and it follows that the syndrome bits are

s1 = r2 r3 r4 r5
s2 = r1 r2 r4 r6
s3 = r1 r3 r4 r7
The syndrome s associated with a single-error pattern e = e1 e2 e3 e4 e5 e6 e7 is
equal to the transpose of the column of the parity-check matrix H in the position where
the error occurs. Therefore,
e1
e2
e3
e4
e5
e6
e7

112

=
=
=
=
=
=
=

s1 s2 s3
s1 s2 s3
s1 s2 s3
s1 s2 s3
s1 s2 s3
s1 s2 s3
s1 s2 s3

The circuits for encoding and decoding the (7,4) Hamming code are shown below.

m1

c1

m2

c2

m3

c3

m4

c4
c5
c6
c7

Encoder of a (7,4) Hamming code.

c1
c2
c3
c4
c5

~
c

c6
c7

~
c

s1

s2

s3

e1

~
c

e2

~
c

e3

~
c

e4

~
c

e5

~
c

e6
e7

Decoder of a (7,4) Hamming code.

113

120. MATLAB experiment

Download the Matlab script homework10f04.m from the website. Record the output and
verify it. That is, verify that, for the given message and error vectors, the codeword, the
received word and the syndrome are correct. (Note that in Matlab the position of the identity
matrices in the generator and parity-check matrices of the code is dierent from that given
in class.)
Solution:

12/15/04 12:25 AM

MATLAB Command Window

To get started, select MATLAB Help or Demos from the Help menu.
>> homework10
Enter your student ID (tower card) number: 123456789
G =
1
0
1
1

1
1
1
0

0
1
1
1

1
0
0
0

0
1
0
0

0
0
1
0

0
0
0
1

1
0
0

0
1
0

0
0
1

1
1
0

0
1
1

1
1
1

1
0
1

0
0
0
0
1
0
0
0

0
0
1
0
0
0
0
0

0
1
0
0
0
0
0
0

0
0
0
0
0
0
1
0

0
0
0
1
0
0
0
0

0
0
0
0
0
0
0
1

0
0
0
0
0
1
0
0

H =

trt =

msg =

code =

err =

recd =
1

114

1 of 2

12/15/04 12:25 AM

MATLAB Command Window

Syndrome = 2 (decimal), 0

2 of 2

0 (binary)

errvect =
0

correctedcode =
1

121. Find the capacity of the cascade connection of n binary-symmetric channels with the same
crossover probability . What is the capacity when the number of channels goes to innity?
Solution: The overall channel is a binary symmetric channel with crossover probability p.
To nd p note that an error occurs if an odd number of channels produce an error. Thus,

 n 
k (1 )nk
p=
k
k=odd
which is equal to
p=


1
1 (1 2)2
2

and therefore,
C = 1 h(p)
If n , then (1 2)n 0 and p 12 . In this case
1
C = lim C(n) = 1 h( ) = 0
n
2

122. In the transmission and reception of signals to and from moving vehicles, the transmitted
signal frequency is shifted in direct proportion to the speed of the vehicle. The so-called
Doppler frequency shift imposed on a signal that is received in a vehicle traveling at a velocity
v relative to a (xed) transmitter is given by the formula
v
fD = ,

where is the wavelength and the sign depends on the direction (moving toward or moving
away) that the vehicle is traveling relative to the transmitter. Suppose that a vehicle is
traveling at a speed of 100 km/hr relative to a base station in a mobile cellular communication
system. The signal is narrowband and transmitted at a carrier frequency of 1 GHz.
(a) Determine the Doppler frequency shift.
(b) Suppose that the transmitted signal bandwidth is 1 MHz centered at 1 GHz. Determine
the Doppler frequency spread between the upper and lower frequencies in the signal.

115

Solution:
(a) The wavelength is
3
3 108
m
m =
9
10
10
Hence, the Doppler frequency shift is
=

fD =

100 103 10
100 Km/hr
u
=
=
Hz = 92.5926 Hz
3

3 3600
10 m

The plus sign holds when the vehicle travels towards the transmitter whereas the minus
sign holds when the vehicle moves away from the transmitter.
(b) The maximum Doppler frequency shift is obtain when f = 1 GHz + 1 MHz and the
vehicle moves towards the transmitter. In this case
min =

3 108
m = 0.2997 m
109 + 106

and therefore

100 103
= 92.6853 Hz
0.2997 3600
Thus, the Doppler frequency spread is Bd = 2fD max = 185.3706 Hz.
fD max =

123. A multipath fading channel has a multipath spread of Tm = 1 s and a Doppler spread
Bd = 0.01 Hz. The total channel bandwidth at bandpass available for signal transmission is
W = 5 Hz. To reduce the eect of intersymbol interference, the signal designer selects a pulse
duration of T = 10 s.
(a) Determine the coherence bandwidth and the coherence time.
(b) Is the channel frequency selective? Justify your answer.
(c) Is the channel fasding slowly or rapidly? Justify your answer.

Solution:
(a) Since Tm = 1 second, the coherence bandwidth
Bcb =

1
= 0.5
2Tm

Hz

and with Bd = 0.01 Hz, the coherence time is

Tct =

1
= 100/2 = 50
2Bd

seconds

(b) Since the channel bandwidth W Bcb , the channel is frequency selective.
(c) Since the signal duration T  Tct , the channel is slowly fading.

116

124. Demonstrate that a DS spread-spectrum signal (without coding) provides no improvement in

performance against AWGN.
Solution: The probability of error for DS spread spectrum with binary PSK may be expressed
as
"

2W/Rb
P2 = Q
PJ /PS
where W/R is the processing gain and PJ /PS is the jamming margin. If the jammer is a
broadband, WGN jammer, then
PJ
PS

= W J0
= Eb /Tb = Eb Rb

Therefore,

'
P2 = Q

2Eb
J0

which is identical to the performance obtained with a non-spread signal.

125. A DS spread-spectrum system is used to resolve the multipath signal component in a twopath radio signal propagation scenario. If the path length of the secondary path is 300 m
longer than that of the direct path, determine the minimum chip rate necessary to resolve
the multipath components.
Solution: The radio signal propagates at the speed of light, c = 3108 m/ sec . The dierence
in propagation delay for a distance of 300 meters is
Td =

300
= 1 sec
3 108

The minimum bandwidth of a DS spread spectrum signal required to resolve the propagation
paths is W = 1 M Hz. Hence, the minimum chip rate is 106 chips per second.

117

SAMPLE EXAMS

118

EE160. Spring 2005.

San Jos
e State University
Solution Midterm Exam # 1-Type B

Problem 1 (25 points) Consider the bandpass signal

x(t) = 2 sinc(t/5) sin(t).
(a) Sketch carefully the spectrum of this signal. What is its bandwidth?
The center frequency is f0 = 1/2. The spectrum is
X(f ) = 5j (5(f + 1/2)) + 5j (5(f 1/2)),
and shown in the gure below. The bandwidth is B = 1/5.
X(f)
5j

-1/2

f
0

1/2
B=1/5

-5j

Spectrum of a bandpass signal.

(b) Give an expression of the complex baseband equivalent x (t).
By inspection, we have xs (t) = 2 sinc(t/5), xc (t) = 0. Therefore x (t) = j xs (t) =
2j sinc(t/5).
(c) Sketch carefully the complex baseband equivalent spectrum X (f ).
X (f ) = j Xs (f ) = 2j 5 (5f ),
which is sketched below.
Xl(f)
f
0

-10j
1/5

119

Spectrum of the complex baseband signal.

Problem 2 (30 points) The spectrum of a lowpass signal is shown below:

X(f)
1

f
-1/3

1/3

Spectrum of a lowpass signal.

(a) Determine the value of the Nyquist sampling rate, fs,min .
W = 13 . Therefore the Nyquist rate is fs,min = 23 .
(b) Sketch the ideal sampled spectrum X (f ), for the sampling rate fs = 3fs,min .
The sampled spectrum X (f ) for fs = 2 is shown below.
X(f)
2

f
-6

-4

-2

2/3

Sampled spectrum X (f ).
(c) Sketch the output x
(t) of an ideal (rectangular) bandpass lter with f0 = 4, B = 0.7, and
gain 1/4.
) at the output of the lter is:
The spectrum X(f
~
X (f)
1/2
f
-4

2/3

2/3

).
Output spectrum X(f
We have
) =
X(f
=

 



3
3
1
(f + 4) + (f 4)
2
2
2





3
1
3
1 1
3 (f + 4) + 3 (f 4)
3 2
2
2
2
120

1
x
(t) = sinc
3

2t
3


cos(8t)

The amplitude envelope is a sinc function with zero crossings at multiples of 3/2. Also, the
sinusoidal waveform was period T0 = 1/4. A sketch is shown below:
~
x(t)
1/3

t
-3

-3/2

3/2

-1/3

T0=1/4

Sketch of output signal x

(t).

Problem 3 (15 points) Using the properties of the Fourier transform, nd the value of the integral

sinc2 (t) cos(6t) dt.
I=

Use the modulation property and

x(t) dt = [X(f )]f =0 .

This gives
I=

1
[(f + 3) + (f 3)]f =0 = 0.
2

Note (based on an idea of C. Igwebuike): The same result is obtained using Parservals theorem.
This follows from the evaluation property of the impulse function:

X(f )(f f0 )df = X(f0 ).

Problem 4 (30 points) The spectrum X(f ) of a signal is depicted below.

X(f)
2/3
2/3 sinc2(2f/3)

f
-4.5

-3

-1.5

1.5

0
3/8

121

4.5

Spectrum of a signal.
(a) Is this signal energy-type or power-type? Justify your answer. (Hint: You do not need to
compute the power or the energy.)
The signal spectrum,
X(f ) =

 



3n
2
2 3n
sinc2
f
,
3
3
8
8
n=

is that of a periodic signal. Consequently, the signal is of the power type.

(b) Sketch carefully the time-domain signal x(t).
Compare the given spectrum with that of a periodic signal:


n
X(f ) =
xn f
T0
n=
Then T0 = 8/3, and
2
xn = sinc2
3
From this it follows that
16
sinc2
XT0 (f ) =
9
and

2 3n
3 8

2
f
3

1
=
XT
T0 0

n
T0

8
xT0 (t nT0 ) =
x(t) =
3
n=

xT0 (t) =

3
8
( t),
3
2


 
8n
3
t
,

2
3
n=

which is sketched below:

x(t)
8/3

t
-16/3

-8/3

8/3

4/3

Sketch of the time signal x(t).

(c) Find the average value of x(t). (Hint: No integrals required.)
The average of the signal is
1
T0

+T0

2
x(t) dt = x0 = ,
3

the area of the impulse at the origin in X(f ).

122

16/3

EE160. Fall 2003.

San Jos
e State University
Solution Midterm Exam # 1 - B

1. (30 points) A periodic signal s(t) is depicted in Fig. 1.

s(t)
1

/2 /2

-2

Figure 1: Periodic signal s(t).

(a) Find the Fourier series expansion of s(t).
(b) Sketch carefully the spectrum, S(f ), of signal s(t) .
Solution:
(a) The Fourier transform of the signal sT (t), over one period T = 2 seconds, is
ST (f ) = F ((t)) = sinc( f ).
The Fourier series coecients sn can be computed from ST (f ) as
sn =

n 
n 1
1
ST
= sinc
.
T
T
2
2

The Fourier series expansion of s(t) is therefore

n 


1
sinc
ejnt
2
2
n=
n=

n 
+
cos(nt)
=
sinc
2
2
n=1

s(t) =

sn ej2nt/T =

(b) The signal is periodic with period T = 2. As a result, its spectrum is

n
sn f
2
n=

n 
n


f
sinc
=
2 n=
2
2

S(f ) =

and shown in the gure below:

123

S(f)
/2

f/2
-3

-2

-1

1/T

Figure 2: Spectrum of the periodic signal of problem 1.

2. (50 points) The spectrum of a bandlimited signal s(t) is shown below.
S(f)
2
1
f
-8

Spectrum of a bandlimited signal.

(a) Find an expresion of signal s(t). (Hint: The spectrum is a linear combination of a
rectangular pulse and a triangular pulse.)
(b) Sketch carefully the spectrum of the sampled signal using a sampling rate equal to
Rs = 1.5Rnyq , where Rnyq denotes the Nyquist rate.
(c) Specify a lowpass lter characteristic that allows reconstruction of s(t) from its samples.
Solution:
(a) The spectrum can be written as

 



f 8
f +8
1
2 [(f 8) + (f + 8)] + 2
+
S(f ) =
2
2
2
The inverse Fourier transform yields


s(t) = 2 sinc2 (t) + 2 sinc(2t) cos(16t).
(b) This is a bandpass signal. Nyquist rate is Rs = 2fu /[fu /2] = 4.5. With this rate, the
spectrum is shown in the gure below, where a = 4.5

124

S(f)
2a

-8 -6 -4 -2

Spectrum of a sampled bandpass signal.

However, sampling at 1.5Rs results in overlapping of spectra. Therefore, we regard the
signal as a lowpasss signal and sample at Rs = 1.5(2)(9) = 27.
S(f)
2a
f
-27

-8

27

Spectrum of a sampled bandpass signal regarded as a lowpass signal.

(c) The signal can be recovered by a reconstruction lter that has constant response 1/27
in the frequency band of the signal 7 < |f | 9 and zero for |f | 18.

3. (20 points) Sketch the spectrum of the complex baseband equivalent of the bandpass signal
whose spectrum is depicted in the following gure.
S(f)
1
f
-5

Spectrum of a bandpass signal.

Solution: The spectrum is below:
Sl(f)
2

f
-1

Spectrum of complex baseband signal.

125

EE160. Fall 2003.

San Jos
e State University
Solution Midterm Exam # 2

1. (35 points) A binary baseband communication system transmits one bit every T seconds,
with T = 4. The pulse used is shown in Fig. 3 below.
x(t)
1/2

-1/2

Figure 3: Pulse shape used in a binary baseband communication system.

(a) Find the impulse response h(t) of the matched lter for x(t).
(b) Determine the output of the matched lter at t = 4 when the input is x(t).
(c) For transmission over an AWGN channel, polar mapping is employed, i.e., a bit bk is
mapped onto a voltage level ak as follows:
bk
0
1

ak
-1
+1

After polar mapping the levels are input to a transmit lter with impulse response x(t)
every 4 seconds. The AWGN process has power spectral density equal to 1/2. Determine
the probability of a bit error of the optimal receiver. Express your answer in terms of
the Gaussian Q-function.
Solution:
(a) Matched lter:
h(t)
1/2

-1/2

(b) Sampled output of matched lter at t = T = 4:


y(4) = Ex =
0

 2
1
1
|x( )| d = 2
=
2
2

126

(c) Polar mapping:

'
P [] = Q

2Ex
N0

"

= Q

2( 12 )

(1)

= Q (1)

2. (30 points) Binary data at a rate of 4 106 bits/s are transmitted using M -PAM over a
bandlimited channel using a raised cosine roll-o characteristic as shown in Fig. 4 below.

P(f)
T

T/2

f (MHz)
-1

-0.75 -0.5

-0.25

0.25

0.5

0.75

Figure 4: Raised cosine roll-o characteristic.

(a) Determine the roll-o factor and the value of M of this system.
(b) Signals are transmitted over a bandlimited AWGN channel. Determine the minimum
required bit energy-to-noise ratio, Eb /N0 , to achieve a probability of a bit error less than
103 . Express your result in decibels (dB).
Solution:
(a) Rb = 4 Mbits/s. Note that 1/2T = 0.5 MHz, the point of symmetry of the raised cosine.
As a result, 1/T = 1 MHz. Since Rb = 1/Tb = log2 M (1/T ), it follows that log2 M = 4
and therefore M = 16.
To nd the roll-o factor, note that 2(1/2T ) = 0.75 0.25 = 0.5 MHz (the region of
the cosine characteristic) and therefore, with 1/2T = 0.5 MHz, we obtain = 0.5.
(b) The probability of a bit error for M -PAM is
'

6 log2 M Eb
M 1
Q
.
P [] =
M
M 2 1 N0
For M = 16,
15
Q
P [] =
16

"

24
255

Eb
N0

< 1 103
min

From the table of the Q-function, we have Q(3.07) = 0.0011. Therefore,

"
8
85

Eb
N0


= 3.07

min

127

Eb
N0


=
min

85
(3.07)2 = 100.14 = 20 dB
8

3. (35 points) A communication system is designed for binary data transmission over an ideal
AWGN channel, with RZ unipolar pulses of average energy equal to Es = A2 T /4. This
system is then applied to an optical link that has the feature that the variance of the AWGN
process varies with the signal level.
Specically, it is found that the sampled output of the matched lter (with kM F = 1) Y is a
Gaussian random variable with conditional PDF given by



y2
1

exp

22 Es ,
22 Es
pY (y|ak ) =

1
42 Es

ak = 0 (bk = 0);



2
s)
exp (yE
, ak = +1 (bk = 1).
42 Es

In other words, the variance of Y equals 2 Es is the transmitted level is 0 and 2 2 Es if the
transmitted level is +1.
(a) Find the probability of a bit error for the optical link in terms of the energy-to-noise
power ratio, Es / 2 . Express your result in terms of the Gaussian Q-function.
(b) Give an expression to determine the threshold of the decision device used in an optimal
receiver for the optical link.
Solution:
(a) The conditional PDFs of the matched lter output are sketched below:
pY(y|ak)

p pY(y|ak=0)

(1-p) pY(y|ak=+1)

y
0

Es/2

+Es

Probability of error:




Es /2 Es
Es /2 0

+ (1 p) 1 Q
P [] = p Q 2
Es
2 2 Es

'

'
Es
Es
+ (1 p) Q
= pQ
2
4
8 2


(b) The optimum threshold y = is obtained by equating the a-posteriori PDFs:

p pY (|ak = 0) = (1 p) pY (|ak = +1)




2
( Es )2
1
1

exp 2
exp
= (1 p)
p
2 Es
4 2 Es
2 2 Es
4 2 Es

128

2
2 2 Es



1p
2 + 2Es Es2 + 2 Es 2 ln
ln 2 = 0.
p

129

1p
( Es )2 1
ln 2 + ln
4 2 Es
2
p

EE160. Spring 2005.

San Jos
e State University
Solution Midterm Exam # 2-B

Problem 1 (25 points) A line coding scheme uses Manchester encoding with rectangular pulses.
(a) Sketch the signal corresponding to the bit sequence 110101
x(t)
1

t
T

2T

3T

4T

5T

6T

-A

(b) Sketch carefully, showing all relevant labels, the power spectral density of this scheme.
0.7
2

E=A T

0.6

S(fT)/E

0.5

0.4

0.3

0.2

0.1

0.5

1.5

2
fT

2.5

3.5

(c) The energy per bit is Eb = 1 109 Joules and the bit rate is Rb = 1 106 bits/sec. What
is the amplitude of the pulses?
2
= 106 . The energy
The bit
per bit is Eb = A T from which it follows that
! rate is R
b = 1/T !
9
6
3
A = Eb /T = Eb Rb = (10 )(10 ) = 10 = 31.62 mV.

(d) The noise at the receiver is AWGN with n2 = 1.25 1010 W/Hz. Determine the probability
of a bit error.
'

'


2Eb
2 109
8 = 2.34 103 .
=Q
=
Q
Pe = Q
N0
2 1.25 1010

130

Problem 2 (40 points) A binary communication system uses NRZ pulses and polar mapping:

2 (t),
bit = 0
s(t) =
2 (t), bit = 1,
where (t) is a unit-energy pulse. The bit duration is 3 seconds.
(a) Determine the energy per bit, Eb .

We have s(t) = Eb (t), and therefore Eb = 2 or Eb = 4.

(b) Sketch carefully a block diagram of the demodulator. In particular, you must give a detailed
specication of the impulse response of the matched lter, the sampling time, and the decision
rule.

r(t)

Y=y(3)

(3-t)

t=3

Decision rule:
=
M

0, Y > 0;
1, Y 0,

(c) Noise at the receiver is AWGN with n2 = 3. Determine the probability of a bit error, Pe .

' 
'

'
2Eb
24
4
=Q
= 1.24 101 .
=Q
Pe = Q
N0
23
3
(d) By mistake, the transmitter sends RZ pulses with polar mapping. Determine the resulting
probability of a bit error, Pe,mistake .
If RZ pulses of amplitude A are sent and the receiver has a lter (or correlator) matched to
NRZ pulses, then the energy is scaled by one half. Consequently,

' 
'

'
Eb
4
2
=Q
= 2.07 101 .
=Q
Pe,mistake = Q
N0
23
3

131

Problem 3 (35 points) The pulse waveform shown below is used for binary communication with
unipolar mapping (also known as on-o keying).
s(t)
3a
2a
a
t(sec)
0

(a) Determine the value of a in terms of the average energy per bit Eb .

!
The energy of the pulse is Es = (3a)2 (1)+(2a)2 (1)+a2 (1) = 14a2 , and therefore a = Es /14.
On the
! other hand, the average energy per bit with unipolar mapping is Eb = Es /2. Therefore,
a = Eb /7.

(b) Sketch the impulse response of the lter matched to s(t).

h(t)=s(3-t)
3a
2a
a
t(sec)
0

(c) It is desired that the bit rate Rb (bits/sec) be at least 1 106 . Using M -PAM modulation,
you are asked to determine the minimum value of M .
The pulse duration is T = 4 106 sec. The bit rata rate of M -PAM, with M = 2m , is
Rb = m/T . The requirement is

 
m
106 m 3 106 106 m 3.
T
Therefore, the minimum value is M = 23 = 8.
(d) With the value of M found in part (c), determine the new value of a in terms of the average
energy per symbol Eave .
For 8-PAM,
the average
energy is: Eave = (M 2 1)Es /3 = 21Es . As a result, the new value
!
!
is a = Es /7 = Eave /147.

132

EE160. Spring 2005.

San Jos
e State University
Solution Final Exam

Problem 1 (35 points) The raised-cosine spectrum X(f ) of a digital communication system is
shown in Fig. 5 below. The excess bandwidth is 25%.
X(f)
b
b/2
f
-f3 -f2 -f1

f1

f2

f3

Figure 5: Raised-cosine spectrum of a digital communication system.

(a) The system transmits information at a rate of 10 Mbps and utilizes 4-PAM modulation.
Determine the values of b, f1 , f2 and f3 .
We have m = 2 and Rb = 10106 bps. Since Rb = m/T , it follows that 1/T = Rb /2 = 5106
baud. Therefore,
b = T = 2 107 = 200 ns
1
= 2.5 106 = 2.5 MHz
2T


(1 + )
= (1.25) 2.5 106 = 3.125 106 = 3.125 MHz
2T


(1 )
= (0.75) 2.5 106 = 1.875 106 = 1.875 MHz
2T

f2 =
f3 =
f1 =

(Type B: b = 400 ns, f2 = 1.25 MHz, f3 = 1.6625 MHz, and f1 = 837.5 KHz.)
(b) RZ rectangular pulses are utilized. Determine the amplitude of the pulses if N0 = 1 108
and the probability of a bit error Pb is desired to be lower than 102 .
With RZ pulses, E = a2 T /2. For 4-PAM modulation, the requirement that Pb < 102
translates into
3
Pb Q
4

'

4Eb
5N0

'


< 102

4Eb
5N0


< 1.33 102 .

With the help of Table 1, we nd that

'

4Eb
> 2.3
5N0

Eb
5
> (2.3)2 = 6.6125 = 8.2 dB.
N0
4

With N0 = 1 108 ,

133

Eb = 6.6125 10

a2 T
=
2

'

a=

2(6.6125 108 )
= 0.813 V.
2 107

(Type B: Eb /N0 = 6.8 dB, and a = 1.095 V.)

(c) The reliability is improved by requiring that Pb < 103 ? What amount of additional power
will be needed? Express your answer in dB.
(
4Eb
With the aid of Table 1, the requirement Pb < 103 translates into 5N
> 3.1, from which
0
it follows that


Eb
N0


=
impr

5
(3.1)2 = 12.0125 = 10.8 dB.
4

Consequently, the amount of additional power is 10.8 8.2 = 2.6 dB.

(Type B: (Eb /N0 )impr = 9.6 dB, additional power is 2.8 dB.)

Problem 2 (35 points) A binary communication system transmits information at a rate of 1

Mbps. The mapping is polar and a correlation-type receiver is employed. The waveform used is
shown in Fig. 6 below.
s(t)
0.075

t(sec)
0

0.35

Figure 6: Waveform used in a binary communication system.

(a) The noise at the receiver is AWGN with n2 = 5 1010 . Determine the probability of a bit
error.
T = 1 s and mapping is polar. The energy per bit is
Eb = (0.075)2 (0.35 106 ) = 1.96875 109 .
Also, n2 = N0 /2 = 5 1010 or N0 = 1 109 . Therefore,
'
Pb = Q

2Eb
N0

'
=Q

2(1.96875 109 )
1 109

(Type B: a = 64.6 mV.)

134


= Q(1.984) 2 102 .

(b) Sketch the output of the correlator in the range 0 to 1 sec. (Hint: No integrals.)
y(t)
1.97x10-9

t(sec)
0

0.35

(Type B:
y(t)
1x10-9

t(sec)
0

0.25

0.5

)
(c) Sketch carefully, with as much detail as possible, a block diagram of this system, including
both transmitter and receiver. (The more detail is shown the higher the grade of this part.)

Bits-to
Signal
Mapping

(t)

n(t)

Decision Device

r(t)

t
Y=y(T)
M
t=T=1 s

0
(t)

0, Y <= 0
M =
1, Y > 0

The waveform (t) is a rectangular pulse of duration 0.35 s and amplitude

135

1/(0.35 106 ) .

(d) If the transmitter sends NRZ pulses at the same rate and with the same energy as the pulse
of Fig. 6, and the receiver is the same as that of part (c), determine the resulting probability
of a bit error.
6
An NRZ pulse with
the same energy and duration is a rectangular pulse of width 1 10
3
s and amplitude 1.96875 10 V. The correlator output is a waveform similar to that of
the previous gure with maximum amplitude (energy) a = (1.96875 103 )(0.35 106 ) =
6.89 1010 . Therefore,

'
Pb = Q

2(6.89 1010 )
1 109


= Q(1.174) 4 102 .

(Type B: Pb 3.59 102 .)

Problem 3 (35 points) A bandpass QPSK modulation system uses the following orthonormal
signals,




2 (t) = 110 sin 104 t .
1 (t) = 110 cos 104 t ,
Noise is AWGN with N0 = 1 and the target probability of a bit error is Pb = 105 .
(a) Determine the maximum bit rate (bps) of this system.
For QPSK modulation, m = 2 and Pb = 105 , we have that (Homework 9!) Eb /N0 9.12.
Therefore, with N0 = 1, E = 2Eb 18.24. Also, the amplitude of the signal is
'
1
(110)2
2E
= 110

= 331.7 baud.
T
T
2 18.24
As a result,
Rb,max =
(Type B: Rb,max =

4
T

2
= 663.38 bps.
T

= 4002 bps.)

(b) The channel bandwidth is 450 Hz and a raised-cosine spectrum is employed to eliminate
ISI. Determine the excess bandwidth and sketch carefully the spectrum, paying particular
attention to the frequency values.
We have B =

1
T (1

+ ) = 450. Therefore, =

450
331.7

1 = 0.3567, or 35.7%

X(f)
A
A/2
f (KHz)
4.775

4.8712 4.8974

136

5.107 5.1658 5.225

In the gure above, A = 2.12 103 .

(Type B: B = 1220, = 0.22 or 21.94%, and Rb /B = 3.28 bps/Hz. Also, A = 7.07 104 .)
(c) Sketch the constellation points used in the bits-to-signal mapper.
2(t)

10

00

E/2

E/2

- E/2

11

1(t)

01

- E/2

(Type B:
2(t)

-3

1011

1010

1001

1000

E/10

E/10

E/10

- E/10

1101

1100

1111

1110

0010

0011

0000

0001

E/10

- E/10

-3

E/10

137

E/10

0100

0101

0110

0111

1(t)

(d) A new RF technology produces a 5 dB improvement in SNR at the receiver. Discuss how this
can be used to enhance performance.
A larger SNR at the receievr can be used in three alternative ways:
(1) To achieve a reduced probability of a bit error or bit error rate (BER), for the same
transmitted power
(2) A lower transmitted power with the same target BER
(3) An increased transmission rate for the same BER. For example, the bit rate could be
doubled to Rb,new = 2 663.38 = 1266.8 bps by using 16-QAM modulation, which
requires 14.6 dB (5 dB more than QPSK modulation). See also the solution of Homework
9. (Type B: Can use 32-QAM and increase the rate to Rb,new = 54 4002 = 5002.5 bps.)

138

EE160. Fall 2003.

San Jos
e State University
Solutions to final exams

1A. The orthonormal pulses shown in Fig. 7 are used in a transmission system using quadrature
modulation.

1(t)

2(t)

t
1/2

1/2

Figure 7: Two orthonormal signals.

(a) Determine the pulses associated with the points in the 1 2 -plane shown in Fig. 8 (a).
(b) Determine the signal point s associated with the pulse s(t) shown in Fig. 8 (b).

2(t)
s(t)

s1

1/3

1(t)
1/2
-1/3
-2/3

s2

(b)

(a)

Figure 8: (a) Two signal points; (b) a pulse.

139

Solution:
(a)

s1(t)

s2(t)

2 2

t
1/2

2
1/2

Figure 9: Pulse signals associated with points s1 and s2 .

(b) The coordinates of the point are equal to the outputs of the lters matched to 1 (t) and
2 (t), respectively, and given by

 
 1/2
1
1
2

( 2) =
,
s(t)1 (t) dt =
s1 =
2
3
6
0

 
 1

2
1
2

( 2) =
,
s(t)2 (t) dt =
s2 =
2
3
3
1/2

2(t)
s
2/3

1(t)
-

2/6

Figure 10: Point in the 1 2 -plane associated with pulse s(t).

140

1B. The orthonormal pulses shown in Fig. 11 are used in a transmission system using quadrature
modulation.

1(t)

2(t)

t
1/2

1/2

Figure 11: Two orthonormal signals.

(a) Determine the pulses associated with the points in the 1 2 -plane shown in Fig. 12 (a).
(b) Determine the signal point s associated with the pulse s(t) shown in Fig. 12 (b).

2(t)
2

s(t)

s1

1/2

1(t)
1/2

s2

-1/2
-1

(b)

(a)

Figure 12: (a) Two signal points; (b) a pulse.

141

Solution:
(a) The signals are shown in Fig. 13.
s1(t)

s2(t)=-s1(t)

2 2

2 2

t
1/2

1/2

2
2

Figure 13: Pulse signals associated with points s1 and s2 .

(b) The coordinates of the point are equal to the outputs of the lters matched to 1 (t) and
2 (t), respectively, and given by

 
 
 3/4

1
1
1 1
2

( 2) +
( 2) =
,
s(t)1 (t) dt =
s1 =
4
2
4 2
4
1/4

s2 =

1/4
0

1
1
2
.
s(t)2 (t) dt +
s(t)2 (t) dt = (1)( 2) + (1)( 2) =
4
4
2
3/4


2(t)
s
2/2

1(t)
-

2/4

Figure 14: Point in the 1 2 -plane associated with pulse s(t).

142

2. A binary modulation system uses quadrature modulation with the signal points shown in Fig. 15.
Transmission of equally likely bits takes place over an AWGN channel with power spectral density
N0 /2.

2(t)
Z1

2
s1

Z2

1(t)

s2

Figure 15: Binary signal constellation.

(a) Find the value of the constant as a function of the average signal energy E.
(b) Sketch carefully the decision regions.
(c) Find the probability of a bit error P [], in terms of E/N0 and the Q-function, and compare
it with binary transmission using square pulses and polar mapping.

Solution:
(a) =

E
5.

(b) Shown in Fig. 15.

(c) The probability of a bit error is given by
"
P [] = Q

d212
2N0

=Q

'

9E
5N0


.

For polar mapping (regardless of the pulse shape),


'
2E
.
P [] = Q
N0
As a result, the constellation in question has a gain of


 
9E/5N0
9
= 0.46 dB
= 10 log10
10 log 10
2E/N0
10
with respect to polar mapping. That is, the constellation in Fig. 15 needs 0.46 dB more power
to achieve the same error performance.

143

3. This problem deals with a Hamming code with m = 2 redundant bits. This (3, 1) code is also
know as a repetition code of length n = 3.
(a) Specify a parity-check matrix H for this code.
(b) Construct a look-up table with syndrome s as input and error vector e as output.


(c) If the
 vector
 r = 1 0 0 is received, determine the estimated code vector v. (Type B:
r = 1 0 1 .)

Solution:
(a) A parity-check matrix is:


H=


1 1 0
.
1 0 1

(b) Look-up table:

sT
00
11
10
01

0
1
0
0

e
0
0
1
0

0
0
0
1

(c) The syndrome corresponding to the received vector is

  
1 
1 1 0
1

s= 0
=
.
1 0 1
1
0


Therefore, e = 1 0 0 and

 
 

v =re= 1 0 0 1 0 0 = 0 0 0 .

For nal exam type B,



Therefore, e = 0 1 0 and

  
1 
1 1 0
1
s = 0
=
.
1 0 1
0
1


 
 

v =re= 1 0 1 0 1 0 = 1 1 1 .

144

Table 1: The Gaussian Q-function

0.0
1.0
2.0
3.0
4.0
5.0

0.0
5.00e-01
1.59e-01
2.27e-02
1.35e-03
3.17e-05
2.87e-07

0.1
4.60e-01
1.36e-01
1.79e-02
9.68e-04
2.07e-05
1.70e-07

0.2
4.21e-01
1.15e-01
1.39e-02
6.87e-04
1.33e-05
9.96e-08

0.3
3.82e-01
9.68e-02
1.07e-02
4.83e-04
8.54e-06
5.79e-08

0.4
3.45e-01
8.08e-02
8.20e-03
3.37e-04
5.41e-06
3.33e-08

0.5
3.09e-01
6.68e-02
6.21e-03
2.33e-04
3.40e-06
1.90e-08

0.6
2.74e-01
5.48e-02
4.66e-03
1.59e-04
2.11e-06
1.07e-08

0.7
2.42e-01
4.46e-02
3.47e-03
1.08e-04
1.30e-06
5.99e-09

Example: Q(2.5) = 6.21e-03 = 6.21 103 .

Table 2: Selected values of the inverse Gaussian Q-function

Q(x)
101
102
103
104
105
106
107
108
109
1010
1011
1012
1013
1014

145

x
1.28
2.33
3.10
3.73
4.27
4.76
5.20
5.61
6.00
6.63
6.71
7.03
7.35
7.65

0.8
2.12e-01
3.59e-02
2.56e-03
7.23e-05
7.93e-07
3.32e-09

0.9
1.84e-01
2.87e-02
1.87e-03
4.81e-05
4.79e-07
1.82e-09