Solutions to Selected Problems

from Madhow: Fundamentals of Digital Communication

and from Johannesson & Zigangirov: Fundamentals of
Convolutional Coding

Saif K. Mohammed

Revision history:
2012-02-01: Tasks from Tutorial 1.

Department of Electrical Engineering

Linkopings universitet, SE-581 83 Linkoping, Sweden

Linkoping 2012
Note: This material is prepared for the 2012 version of the Master course TSKS04 Digital
Communication Continuation Course. Most of it is however borrowed from the 2011 version
prepared by Mikael Olofsson.
For almost every task planned for tutorials you find either hints and answers or complete
solutions. For the tasks where we give complete solutions, we have chosen to adjust the
notations to concur with the notation that has been used in the lectures.
This document will evolve during the course as the tutorials go by.
This version of the document was compiled February 1, 2012.

Solutions to Selected Problems from Madhow: Fundamentals of Digital Com-

munication
and from Johannesson & Zigangirov: Fundamentals of Convolutional Coding

c 20102012 Mikael Olofsson & Saif K. Mohammed

Department of Electrical Engineering

Linkopings universitet
SE-581 83 Linkoping
Sweden

This document was prepared using LATEX2 with the aid of TeXnicCenter on an Dell PC
running CentOS 5. The figures were produced using Xfig (from xfig.org). Finally, the plots
were produced using Matlab (from MathWorks, Inc.).
Contents
Tutorial 1 1
Task 2.7 (Solution) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Task 2.8 (Hints and answer) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Task 2.10 (Hints and answer) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Task 2.11 (Solution) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Task 2.12 (Hints) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Task 2.13 (Solution) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Task 2.14 (Hints) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

i
ii
 1

Tutorial 1

Task 2.7

Task
Let v(t) denote a real passband signal, with Fourier transform V (f ), specified as follows
for negative frequencies:
(
f + 101, 101 f 99,
V (f ) =
0, f < 101 and 99 < f 0.

a. Sketch V (f ) for all f .

b. Without explicitly taking the inverse Fourier transform, can you say whether
v(t) = v(t) holds or not.
c. Choosing f0 = 100 Hz, find real baseband waveforms vI (t) and vQ (t) such that


v(t) = 2 vI (t) cos(2f0 t) vQ (t) sin(2f0 t)

holds.
d. Repeat c for f0 = 101 Hz.

Solution
a. The passband signal is real. Therefore we have

V (f ) = V (f ) (conjugate symmetric)

And totally
(
101 |f |, 99 |f | 101,
V (f ) =
0, else.

Our spectrum then looks like this:

V (f )

101 99 99 101

b. We use the following property of the Fourier transform, which holds regardless of
if v(t) is real or complex:

v(t) = v(t) V (f ) = V (f ).

Obviously, the second relation holds (known from a.) in this case, and therefore
the first does as well.
2

c. We want the determine vI (t) and vQ (t). For these signals, we have

2v(t) cos(2f0 t) = vI (t) + vI (t) cos(4f0 t) vQ (t) sin(4f0 t),

2v(t) sin(2f0 t) = vQ (t) vI (t) sin(4f0 t) vQ (t) cos(4f0 t),
which corresponds to
1

V (f + f0 ) + V (f f0 ) = VI (f ) + passband signal around 2f0 ,
2
j

V (f + f0 ) V (f f0 ) = VQ (f ) + passband signal around 2f0
2
in the frequency domain. By filtering through ideal lowpass filters with cutoff
frequency 1, we get
1

VI (f ) = V (f f0 ) + V (f + f0 ) I{|f |1} (f ) = 2I{|f |1} (f ),
2
j

VQ (f ) = V (f f0 ) V (f + f0 ) I{|f |1} (f ) = j 2f I{|f |1} (f ) = jf VI (f ).
2
We inverse transform and get

vI (t) = 8 sinc(2t),
(
2 cos(2t)sinc(2t)
8 d , t 6= 0,
vQ (t) = sinc(2t) = t
2 dt 0, t = 0.
As a comment: Sometimes people define the function
(
cos(x)sinc(x)
d x , x 6= 0,
cosc(x) = sinc(x) =
dx 0, x = 0.
Here is a graph of the sinc and cosc functions for those interested:
1.5
sinc(x)
cosc(x)
1

0.5

0.5

1.5
10 8 6 4 2 0 2 4 6 8 10

Then we can write

vI (t) = 8 sinc(2t),

8
vQ (t) = cosc(2t).

 3

d. Here we have the carrier frequency 101 Hz instead of 100 Hz. Denote the baseband
representation w.r.t reference frequency of 100 as

v(t) = vI (t) + jvQ (t)

and w.r.t. reference frequency of 101 as

v (t) = vI (t) + jvQ

(t).

We have

v(t) = Re{ 2v (t)ej202t } = Re{ 2v(t)ej200t }.

Thus, we get

v (t) = v(t)ej(t) ,

where (t) = 2t. This gives us

n o
vI (t) = Re v(t)ej(t) = vI (t) cos(2t) vQ (t) sin(2t)

8
= 8 sinc(2t) cos(2t) cosc(2t) sin(2t)

= 8 sinc2 (2t),
n o

vQ (t) = Im v(t)ej(t) = vI (t) sin(2t) + vQ (t) cos(2t)

8
= 8 sinc(2t) sin(2t) + cosc(2t) cos(2t)

2
= [1 sinc(4t)].
t
Here we have use the trigonometric relations

sin(2a) = 2 sin(a) cos(a) and 1 = cos2 (a) + sin2 (a)

a few times.
4

Task 2.8

Hints
a. One of the signals is (purely real baseband representation), and the other one can
be interpreted using a time dependent phase, in the same way as in Task 2.7d. Alter-
natively, it is possible to take the problem to the frequency domain and to use known
transform relations and then inverse transform.

b. You should definitely solve the problem in the frequency domain. (Consider the
complex baseband signal u(t) and v(t). )

c. You can use Parsevals relation and determine the inner product in the frequency
domain.
Z +
< up , vp >= Re{< u, v >} = Re{ U (f )V (f )df }.

where U (f ) and V (f ) are Fourier transformation of complex baseband representation

u(t) and v(t), respectively.

d. Consider the complex baseband signal.

Y (f ) = (1/ 2)U (f )V (f ) = (1/ 8)V (f ).

Answer
a. Baseband representation: u(t) = sinc(2t), v(t) = sinc(t)ej(t1/4) .
b. up has bandwidth 2 Hz and vp has bandwidth 1 Hz.
c. The inner product is 1/ 8.
d. yp = 21 sinc(t) sin(101t + /4)

Task 2.10

Hints
The input to the baseband box are the I and Q components of up (t) w.r.t f1 (plus
double frequency terms that get filtered out by the baseband filters used in the box.)
The output of the box should be the I and Q components of yp (t) w.r.t f2 .

a. The baseband box is just the complex baseband filtering operation (see page 26-27
in the textbook).

b. Note that the difference between carrier frequencies can be interpreted as a time
dependent phase, in the same way as in Tasks 2.7d and 2.8a. This means that it is
not enough to filter. You need to modulate as well. There are several solutions to this
problem.
 5

Answer
a. yc = 1/ 2(uc hc us hs ) and ys = 1/ 2(uc hs + us hc ).
b. First modulate by ejt to convert the reference frequency from fc + 1/2 to fc . Then
filter as in a. Then modulate by ejt to convert the reference frequency from fc to
fc 1/2.

Task 2.11

Task
Consider a pure sinusoid s(t) = cos(2fc t), which is the simplest possible example of a
passband signal with finite power.

a. Find the time-averaged PSD Rs (f ) and ACF rs ( ) proceeding from the definitions.
b. Find the complex envelope s(t), and its time-averaged PSD and ACF. Check that
Equation 2.70 holds for the passband and baseband PSDs.

Solution
We consider the deterministic signal as a stochastic process S(t) with only one realiza-
tion, s(t).

a. We start with the ACF. First we have the ordinary ACF

rS (t1 , t2 ) = E{S(t1 )S(t2 )} = s(t1 )s(t2 ) = cos(2fc t1 ) cos(2fc t2 )

1 1
= cos(2fc (t1 t2 )) + cos(2fc (t1 + t2 )).
2 2
The time-averaged ACF is then given by
1/fc
1
Z
rS ( ) = fc rS (t + , t) dt = cos(2fc ).
0 2
The corresponding PSD is then given by
1

RS (f ) = F {rS ( )} = (f + fc ) + (f fc ) .
4

b. Using the obvious carrier frequency fc , the complex envelope S(t) has the only
realization
1
s(t) = .
2
This gives us the averaged ACF
1
rS ( ) = ,
2
6

and thus the corresponding PSD

 1
RS (f ) = F rS ( ) = (f ).
2
Equation (2.70) obviously holds, i.e.
1

RS (f ) = RS (f + fc ) + RS (f fc )
2
 7

Task 2.12

Hints
a. See Section 2.3.1.

b. Express the time-averaged ACF of n. You get an expression in the time-averaged

ACFs of nc and ns and their cross-correlations.

c. Recall that WSS means constant mean, and ACF being only a function of the time
difference.

Task 2.13

Task
Notice that the given PSD is for the passband signal. The formulation of this task in
the course book does not follow the otherwise used notation of the book.

We discuss passband white noise, an important

nnoise model used o extensively in Chap-
ter 3. A passband random process N (t) = Re 2N (t)e j2f c t with complex envelope
N (t) = NI (t) + jNQ (t) has PSD

N0
RN (f ) = I W W ,
2 {fc 2 |f |fc + 2 }
where NI (t) and NQ (t) are i.i.d. processes with mean zero.

a. Find the PSD RN (f ) for the complex envelope.

b. Find the PSDs RNI (f ) and RNQ (f ) if possible. If this is not possible from the
given information, say what further information is needed.

Solution
+
a. Let RN (f ) denote the PSD RN (f ) for positive frequencies, i.e.
+
RN (f ) = RN (f )I{f >0} (f ).

Then for the baseband process, we have

+
RN (f ) = 2RN (f + fc ) = N0 I{|f | W }
2

according to Madhow, page 41.

8

b. To determine the PSDs RNI (f ) and RNQ (f ), we want to relate them to RN (f ).

From problem statement, a the passband random process N (t) is white, which
means that its mean is zero and its ACF has infinite power at zero time shift, i.e.,

rN (t1 , t2 ) = C(t1 t2 ).

where C is a constant positive value.

This means that the ACF is only dependent on = t1 t2 .
Therefore, we study the ACF

rN ( ) = E{N (t + )N (t)}


= rNI (t + , t) + rNQ (t + , t) cos(2fc )


+ rNI ,NQ (t + , t) rNQ ,NI (t + , t) sin(2fc )


+ rNI (t + , t) rNQ (t + , t) cos(2fc (2t + ))


rNI ,NQ (t + , t) + rNQ ,NI (t + , t) sin(2fc (2t + ))

The two component processes are independent and identically distributed with
mean zero. Then we can immediately draw the conclusion that we have

rNI ,NQ (t1 , t2 ) = rNQ ,NI (t1 , t2 ) = 0 (i.i.d)

rNI (t1 , t2 ) = rNQ (t1 , t2 ) (identicaldistribution)

Plugging this into the expression for rN ( ), we get

rN ( ) = 2rNI (, t) cos(2fc ) = C( ),

This implies that rNI (, t) only depends on .

Therefore, we have rNI (, t) = rNI ( ) = rNQ (, t) = rNQ ( ).
Now consider the complex envelope. Using the relations above, we can write the
ACF for the random process of the complex envelope as

rN (, t) = E{N (t + )N (t)}


= rNI ( ) + rNQ ( ) + j rNQ ,NI ( ) rNI ,NQ ( )
= 2rNI ( ) = 2rNQ ( ),

With Fourier transformation we get PSD for RNI (f ) and RNQ (f )

1 N0
RNI (f ) = RNQ (f ) = RN (f ) = I W .
2 2 {|f | 2 }

Task 2.14

Hints
a. Follow the instructions given in the task.
 9

i) We can write D = (a + D) mod T . And it is clear that D has the same

distribution as D and is independent of s as well.
ii) Direct consequence of cyclostationarity of s.
iii) v(t) = s(t D) and v(t) = s(t D). They are statistically indistinguishable,
because of the results in i).

b. Follow the instructions given in the task.

RT
i) 1). mv (t) = E{v(t)} = E{s(t D)} = 1/T 0 ms (t )d. 2). change variables
= t . 3) s is cyclostationary, ms is periodic with period T . So, we can
change the integral interval to [0,T ].
ii) Similar strategies as in i).

c. Change the problem formulation in i) as follows: s(t)s (t ) has the same statistics
as s(t + T )s (t + T ) w.r.t mean function.

Follow the instructions given in the task.