1.

S IGNAL T HEORY
Exercise 1. Verify whether the three L2 (a, b) functions
x1 (t) = 1
x2 (t) = t 2
x3 (t) = 1 + 2t 2
are linearly independent and find a basis for the subspace they generate.

Exercise 2. Give a representation of the function

x(t) = (1 + t)2
using the basis
ψ1 (t) = 1 + t
ψ2 (t) = 1 − t
ψ3 (t) = t 2

Exercise 3 (Walsh signals). Show that the functions {ϕi (t)}4i=1 given below form an orthonor-
mal basis and then represent x(t) using this basis.

ϕ1(t) ϕ2(t) ϕ3(t) ϕ4(t)

1 1 1 1

1
1 t 1
1 t 1
1 t 1
1 t
2 2 2 2
−1 −1 −1 −1
x(t)
1

1
1 t
2
−1

Exercise 4. Find an orthonormal basis for the space generated by the following functions

x1(t) x2(t) x3(t)

1 1 1

1 2 3 4 t 1 2 3 4 t 1 2 3 4 t
−1 −1 −1

Exercise 5. Find an orthonormal basis for the space generated by the following functions
2πnt
xn (t) = cos
T n = 0, ±1, . . . , ±L t ∈ (0, T )
2πnt
yn (t) = sin
T
1
 2

Exercise 6. Find the best approximation (in norm) to y(t) in the space generated by {xi (t)}4i=1
and then determine the error.
x1(t) x2(t) x3(t) x4(t)
2

1 1 1

1 2 3 4 t 1 2 3 4 t 1 2 3 4 t 1 2 3 4 t
−1 −1 −1

−2 y(t)
1

1 2 3 4 t
−1

Exercise 7. Given the M signals

h πi
si (t) = cos 2π f0t + (2i − 1) i = 1, 2, . . . , M
M
defined for t ∈ (0, T ), with f0 T  1,
(1) Determine the dimension of the space generated by such signals, one of its orthonormal
bases, and a vectorial representation of the signals themselves.
(2) Show that if the phase of all signals is changed by the same quantity φ , the signals can
still be represented using the same basis.

Exercise 8. Given a zero-mean stationary stochastic process x(t),

(1) Show that its samples taken at a fixed step T (uniform sampling) are uncorrelated if and
only if the autocorrelation function Rx (τ) satisfies the following condition
(
σ 2 for k = 0
Rx (kT ) =
0 for k 6= 0
(2) Show that if x(t) is bandlimited to 1/2T , the previous condition on Rx (τ) means that the
power spectrum of the process has a rectangular shape
(
α for | f | < 1/2T
Sx ( f ) = F {Rx (τ)} =
0 otherwise
(3) Give an example of stochastic process with bandwidth greater than 1/2T whose samples
taken at a fixed step T are uncorrelated.
 3

2. D IGITAL T RANSMISSION OVER THE AWGN C HANNEL

Exercise 9. Find the structure of an optimum receiver for the following signals

s1(t) s2(t)
2 2

1 1

2 4 t 2 4 t

taking into account that P(m1 ) = 3/4 and P(m2 ) = 1/4, and compute its probability of error for
AWGN with power spectral density N0 /2.

Exercise 10. Compute the probability of error of the (non-optimum) receiver shown in the
figure for the equiprobable signals s1 (t) and s2 (t) in AWGN with power spectral density N0 /2.
Also determine the value of α that minimizes the probability of error. Using this receiver
structure, is it possibile to obtain a lower probability of error by correlating r(t) with something
different from s1 (t)?

s1(t)
(
A s2 RT m1 z>α
r(t) z m̂ =
s1 0
m2 z<α
A

Exercise 11. In an AWGN channel, the received vector has the form r = s + n, where the signal
vector s can assume the possible values
     
1 1 −1
s1 = , s2 = , s3 =
−2 2 2
and the noise vector has independent components with zero mean and variance σ 2 . Determine
the decision zones in the following cases:
(1) all messages are equally likely;
(2) P(m1 ) = 1/9, P(m2 ) = P(m3 ) = 4/9, σ 2 = 1/ln 2.

Exercise 12. In a digital communication system, the noise is AWGN with power spectral den-
sity N0 /2 and the signals are
s1 (t) = −3ϕ(t)
s2 (t) = −ϕ(t) kϕ(t)k = 1
s3 (t) = ϕ(t) t ∈ (0, T )
s4 (t) = 3ϕ(t)
Determine the probability of error for the receiver depicted in the following figure
 4

ϕ (t) 


m1 z < λ1
m λ1 < z < λ2
2
RT m̂ =

m3 λ2 < z < λ3
r(t) 0 z 

m4 z > λ3 and find the values of λi that

minimize it, taking into account that all messages are equally likely.

Exercise 13. In a digital communication system the transmitted signal is

x(t) = ∑ ak s(t − kT )
k

where (
A for t ∈ (0, T )
s(t) =
0 elsewhere
and the symbols {ak } are equiprobable and independent with values +1 and −1.
The receiver is optimum and in each signalling interval it works as follows:
(
RT 1 z>0
r(t) 0 z â =
−1 z < 0

Due to a synchronization problem, the receiver reference turns out to be delayed by τ with
respect to the start of each symbol time, such that it works as if the received signal is r(t) =
x(t + τ) + w(t), where w(t) is AWGN, and thus the decision variable z depends on two symbols
instead of a single one.
Find the probability of error.

Exercise 14. In a communication system, the signals

(
s(t) t ∈ (0, T /2)
s1 (t) = P(m1 ) = 1/2
0 t ∈ (T /2, T )
s2 (t) = −s1 (t) P(m2 ) = 1/2
can reach the receiver following two different paths characterized by the delays θ1 and θ2 with
0 < θi < T /2 (see figure below).

θ1
Tx Rx
θ2
Each of the two paths is chosen according to traffic issues and on average they are equally
likely. Design an optimum receiver assuming that the noise is AWGN with power spectral
density N0 /2.
 5

Exercise 15. In a communication system, the noise is Gaussian with zero mean and power
spectral density
N0 2
N( f ) = f
2
and the transmitted signals are
s1 (t) = cos πt t ∈ (0, 1)
s2 (t) = cos πt t ∈ (0, 2)
s3 (t) = cos πt t ∈ (1, 2)
with a priori probabilities P(m1 ) = P(m3 ) = 1/4, P(m2 ) = 1/2.
Design an optimum receiver.

3. D IGITAL T RANSMISSION OVER BANDLIMITED AWGN C HANNELS

Exercise 16. In a communication system, the sampled signal can be written as

x(tk ) = ak + 0.1ak−1 + 0.1ak+1 + n(tk )
where n(tk ) is a Gaussian random variable with zero mean and variance σ 2 . Assuming that the
symbols are independent and equiprobable with values A/2 and −A/2, compute the probability
of error when the decision strategy is
(
−A/2 x(tk ) < 0
âk =
A/2 otherwise

Exercise 17. In a PAM system, the symbols are independent and assume the values ±A with
probability 1/4 and the value 0 with probability 1/2. The sampled signal is
x(tk ) = ak + 0.1ak−2 + n(tk ),
where n(tk ) is a Gaussian random variable with zero mean and variance σ 2 . Compute the
probability of error of a receiver that decides according to the rule:


−A x(tk ) < −A/2
âk = 0 |x(tk )| < A/2

 A x(t ) > A/2
k

Exercise 18. Verify whether the function

(
1 + j sin π f T | f | < 1/T
G( f ) =
0 elsewhere
satisfies the Nyquist criterion at the times tk = kT .

Exercise 19. Verify whether the function

(
(1 − | f T |)e− jπ f T | f | < 1/T
G( f ) =
0 elsewhere
satisfies the Nyquist criterion at the times tk = t0 + kT for some t0 .
 6

Exercise 20. Verify whether the function

(
T cos(π f T /2) | f | < 1/T
G( f ) =
0 elsewhere
satisfies the Nyquist criterion at the times tk = t0 + kT for some t0 .

Exercise 21. Find the transmit HT ( f ) and receive HR ( f ) filters such that the overall transfer
function G( f ) = HT ( f )HC ( f )HR ( f ) satisfies the Nyquist criterion and the signal to noise ratio
at the receiver is maximum when the channel is such that
1
HC ( f ) =
1 + j2 f T
and the noise is AWGN.

Exercise 22. A PAM system was designed assuming that the noise is AWGN, the channel can
be approximated as an ideal low-pass filter, and that the overall transfer function is of raised-
cosine type with rolloff equal to 1. However, it turns out that the channel transfer function
actually is
1
HC ( f ) =
1+ jfT
and thus, for getting rid of the intersymbol interference, an equalizer with transfer function
E( f ) = 1 + j f T is employed.
Compute the probability of error under the hypothesis of equiprobable and independent sym-
bols with values ±A/2.

Exercise 23. Letting tk = t0 + kT , the overall transfer function of a PAM transmission system
is such that
g(t0 ) = 1
g(t2 ) = 0.2
g(t−2 ) = −0.2
g(ti ) = 0 i 6= 0, ±2
Design a transversal equalizer with four taps that minimizes the peak distortion at the times
tk+2 .

Exercise 24. Letting tk = t0 + kT , design a 2-tap equalizer that minimizes the quadratic distor-
tion at the times tk+1 when the overall transfer function is such that
g(t0 ) = 1
g(t−1 ) = 0.1
g(t1 ) = 0.1
g(ti ) = 0 i 6= 0, ±1
 7

Exercise 25. Letting tk = t0 + kT , design a 2-tap equalizer that minimizes the peak distortion at
the times tk+1 when the overall transfer function is such that (see previous exercise)
g(t0 ) = 1
g(t−1 ) = 0.1
g(t1 ) = 0.1
g(ti ) = 0 i 6= 0, ±1