EEM 306 Introduction to Communications

Lecture 2

Department of Electrical and Electronics Engineering

Anadolu University

March 11, 2014

Lecture 2 1/15
Last Time

I Frequency Domain Analysis

I Fourier Series
I Fourier Transforms
I Basic Properties of the Fourier Transform
I Fourier Transform for Periodic Signals
I Power and Energy
I Energy-Type Signals

Lecture 2 2/15
Power-Type Signals

The time-avg autocorrelation function of the power-type signal

Z T /2
1
Rx (τ ) = lim x(t)x∗ (t − τ )dt
T →∞ T −T /2

1
R T /2 2
If τ = 0, Rx (0) = lim
T →∞ T −T /2 |x(t)| dt = Px
Power spectral density of the signal x(t):

Sx (f ) = F {Rx (τ )}
Z ∞
Px = Sx (f )df
−∞

Lecture 2 3/15
Passing Power-Type Signals through LTI Systems

x(t ) y (t )
h(t )

The output
Z ∞
y(t) = x(t) ∗ h(t) = x(τ )h(t − τ )dτ
−∞
The time-avg autocorrelation function for the output
R T /2
Ry (τ ) = lim T1 −T /2 y(t)y ∗ (t − τ )dt
T →∞
y(t) y ∗ (t−τ )
z }| { zZ }| {
R T /2
Z ∞ ∞
∗ ∗
= lim 1 −T /2 h(u)x(t − u)du h (v)x (t − τ − v)dv dt
T →∞ T −∞ −∞
= Rx (τ ) ∗ h(τ ) ∗ h∗ (−τ )
Taking the FT of both sides:
Sy (f ) = Sx (f )H(f )H ∗ (f ) = Sx (f )|H(f )|2
Lecture 2 4/15
Periodic Signals

All nonzero periodic signals are power-type!

Let x(t) be a periodic signal with period T0
R T /2
Rx (τ ) = lim T1 −T /2 x(t)x∗ (t − τ )dt
T →∞
..
=.
R T /2
= T10 −T0 0 /2 x(t)x∗ (t − τ )dt finite integral
∞
j2π Tn τ
|xn |2 e
P
Substituting the FS expansion Rx (τ ) = 0
n=−∞
For periodic signals, Rx (τ ) is periodic with period T0
∞  
X
2 n
Sx (f ) = |xn | δ f −
n=−∞
T0
Z∞ ∞
X
Px = Sx (f )df = |xn |2
−∞ n=−∞

Lecture 2 5/15
Periodic Signals cont’d

If a periodic signal passes through an LTI system with frequency

response H(f ), the output will be periodic.
The power spectral density of the output:

Sy (f ) = |H(f )|2 Sx (f )
∞
= |H(f )|2 |xn |2 δ(f − Tn0 )
P
n=−∞
∞  2
|xn |2 H Tn0 δ(f − Tn0 )
P
=
n=−∞

The power content of the output:

∞   2
X 2 n
Py = |xn | H
n=−∞
T0

Lecture 2 6/15
Cross Correlation

Energy-type signals:
Z∞
R12 (τ ) = g1 (t)g2 ∗ (t − τ )dt
−∞
R∞
If g1 (t) and g2 (t) are orthogonal, R12 (0) = g1 (t)g2 ∗ (t)dt = 0
−∞
In general,
R12 (τ ) = R21 ∗ (−τ )
R12 (τ ) ⇔ G1 (f )G2 ∗ (f )
Power-type signals:
T
Z2
1
R12 (τ ) = lim g1 (t)g2 ∗ (t − τ )dt
T →∞ T
− T2

Lecture 2 7/15
Example: The signal x(t) = sinc(10t) is input to a system with
the following transfer function
 
f
H(f ) = 3rect e−j4πf
4

a. Find the output y(t).

b. Find the energy spectral densities Gx (f ) and Gy (f ).
c. Find the energies of x(t) and y(t).
Note: sinc(t) ⇔ rect(f )

Lecture 2 8/15
Sampling

In many applications it is useful to represent a signal in terms of

sample values taken at appropriately spaced intervals, Ts .
Two questions in connection with such sampling:
I What are the restrictions on x(t) and Ts to allow perfect
recovery of x(t) from the sampled values.
I How is x(t) recovered from the sampled values.
Both questions are answered by the sampling theorem

Lecture 2 9/15
Sampling Theorem

Let x(t) be a bandlimited signal with bandwidth W , i.e.,

X(f ) = 0 for |f | ≥ W
1
Let Ts be the sampling period with Ts ≤ 2W
Produce a sequence
{x(nTs )}∞
n=−∞

Then it is possible to reconstruct the original signal x(t) from

the sampled values by the reconstruction formula
∞
X
x(t) = 2W 0 Ts x(nTs )sinc(2W 0 (t − nTs ))
n=−∞

where W ≤ W 0 ≤ T1s − W
W 0 the reconstruction filter bandwidth
fs = 2W the min sampling rate at which no aliasing occurs
⇒Nyquist rate

Lecture 2 10/15
Sampling Theorem cont’d

Special case: If Ts = 1
2W , W0 = W
∞  
P t
x(t) = x(nTs )sinc Ts −n
n=−∞
∞
n n
P


= x 2W sinc 2W t − 2W
n=−∞

Lecture 2 11/15
Illustration:
∞
P
Step 1: Multiply x(t) with an impulse train δ(t − nTs )
n=−∞
∞
P ∞
P
xs (t) = x(t) δ(t − nTs ) = x(nTs )δ(t − nTs )
n=−∞ n=−∞
Step 2: Taking the FT of both sides
∞
P
Xs (f ) = X(f ) ∗ F{ δ(t − nTs )}
n=−∞
∞
1 P n
= X(f ) ∗ Ts δ(f − Ts )
n=−∞
∞
1 P n
= Ts X(f − Ts )
n=−∞

Lecture 2 12/15
Sampling Theorem cont’d
∞  
1 P n
Recall that Xs (f ) = Ts X f− Ts
n=−∞

X( f )

f
W

Xs( f )

1
f  2W
W
1
W
1 2 Ts
Ts Ts Ts

Xs( f )

W 1
f  2W
1 2 Ts
1
W Ts
Ts Ts

Xs( f )

1
1
f  2W
W W 1 2 Ts
Ts Ts Ts

Lecture 2 13/15
Reconstruction
X( f )

f
W
Xs( f )
Reconstruction
filter
A / Ts

1
W
f  2W
W
1
W
1 2 Ts
Ts Ts Ts

To obtain the original signal back,

I H(f ) = Ts for |f | < W
1
I H (f ) = 0 for |f | ≥ Ts −W
 
f
One choice: H(f ) = Ts Π 2W 0 where W ≤ W 0 ≤ 1
Ts −W
 
Guard band: T1s − W − W = fs − 2W

Lecture 2 14/15
Aliasing

Now if T1s < 2W , then the replicated spectrum of x(t) overlaps,

and reconstruction of the original signal is not possible.

X( f )

f
W

Xs( f )

1
1
f  2W
W W 1 2 Ts
Ts Ts Ts

Undersampling ⇒ Aliasing

Lecture 2 15/15