Ch.

2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Chapter 2: Deterministic and Random Signal
Analysis
Instructor: Jia-Chin Lin
Department of Communication Engineering,
National Central University
Email: [email protected]
Instructor: Jia-Chin Lin 1
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Basic Materials
Basic Denitions:
Rectangular Pulse
(t) =
_

_
1 [t[ <
1
2
1
2
t =
1
2
0 otherwise
Sinc Function
sinc(t) =
_
_
_
sin(t)
t
t ,= 0
1 t = 0
Instructor: Jia-Chin Lin 2
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Signum
sgn(t) =
_

_
1 t > 0
1 t < 0
0 t = 0
Triangular Pulse
(t) = (t) (t) =
_

_
t + 1 1 t < 0
t + 1 0 t < 1
0 otherwise
Unit Step
u
1
(t) =
_

_
1 t > 0
1
2
t = 0
0 t < 0
Instructor: Jia-Chin Lin 3
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Fourier Transform Properties
Property Signal Fourier Transform
Linearity x
1
(t) +x
2
(t) X
1
(f) +X
2
(f)
Duality X(t) x(f)
Conjugacy x

(t) X

(f)
Time-scaling (a ,= 0) x(at)
1
[a[
X
_
f
a
_
Time-shift x(t t
0
) e
j2ft
0
X(f)
Modulation e
j2f
0
t
x(t) X(f f
0
)
Convolution x(t) y(t) X(f)Y (f)
Multiplication x(t)y(t) X(f) Y (f)
Differentiation
d
n
dt
n
x(t) (j2f)
n
X(f)
Differentiation in freq. t
n
x(t)
_
j
2
_
n
d
n
df
n
X(f)
Instructor: Jia-Chin Lin 4
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Integration
_
t

x()d
X(f)
j2f
+
1
2
X(0)(f)
Parsevals theorem
_

x(t)y

(t)dt =
_

X(f)Y

(f)df
Rayleighs theorem
_

[x(t)[
2
dt =
_

[X(f)[
2
df
Fourier Transform Pairs
Time Domain Frequency Domain
(t) 1
1 (f)
(t t
0
) e
j2ft
0
e
j2f
0
t
(f f
0
)
cos(2f
0
t)
1
2
(f f
0
) +
1
2
(f +f
0
)
sin(2f
0
t)
1
2j
(f f
0
)
1
2j
(f +f
0
)
Instructor: Jia-Chin Lin 5
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
(t) sinc(f)
sinc(t) (f)
(t) sinc
2
(f)
sinc
2
(t) (f)
e
t
u
1
(t), > 0
1
+j2f
te
t
u
1
(t), > 0
1
( +j2f)
2
e
|t|
, > 0
2

2
+ (2f)
2
e
t
2
e
f
2
sgn(t)
1
jf
u
1
(t)
1
2
(f) +
1
j2f
Instructor: Jia-Chin Lin 6
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
1
2
(t) +j
1
2t
u
1
(f)

(t) j2f

(n)
(t) (j2f)
2
1
t
jsgn(f)

n=
(t nT
0
)
1
T
0

n=

_
f
n
T
0
_
Instructor: Jia-Chin Lin 7
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Bandpass and Lowpass Representation
Fourier transform of a real signal x(t) has Hermitian Symmetry:
X(f) = X

(f)
[X(f)[ = [X

(f)[ = [X(f)[: even magnitude.

X(f) =

X

(f) =

X(f): odd phase.

_
X(f) f > 0
1
2
X(0) f = 0
0 f < 0
Instructor: Jia-Chin Lin 8
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
X

(f) = X(f)u
1
(f)
X

(f) =
_

_
0 f > 0
1
2
X(0) f = 0
X(f) f < 0
X(f) = X
+
(f) +X

(f).
If x(t) is real, X(f) is Hermitian Symmetric, i.e., X

(f) = X

+
(f).
Analytic Signal (or Pre-Envelope):
x
+
(t) = F
1
X
+
(f)
= F
1
X(f)u
1
(f)
= x(t)
_
1
2
(t) +j
1
2t
_
=
1
2
(x(t) +j x(t))
Instructor: Jia-Chin Lin 9
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Hilbert Transform: x(t) =
1
t
x(t).
Hilbert Transform: F x(t) = jsgn(f)X(f).
Lowpass Equivalent or Complex Envelope of x(t):
X
l
(f) = 2X
+
(f +f
0
) = 2X (f +f
0
) u
1
(f +f
0
)
.
Instructor: Jia-Chin Lin 10
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Because X
l
(f) = 2X
+
(f +f
0
) = 2X (f +f
0
) u
1
(f +f
0
),
x
l
(t) = F
1
X
l
(f)
= 2x
+
(t)e
j2f
0
t
= (x(t) +j x(t)) e
j2f
0
t
= [x(t) cos (2f
0
t) + x(t) sin(2f
0
t)]
+j [ x(t) cos (2f
0
t) x(t) sin(2f
0
t)]
Instructor: Jia-Chin Lin 11
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
x
l
(t) = x
i
(t) +jx
q
(t):
x
i
(t) = x(t) cos (2f
0
t) + x(t) sin(2f
0
t)
x
q
(t) = x(t) cos (2f
0
t) x(t) sin(2f
0
t)
Solving above equation:
x(t) = x
i
(t) cos (2f
0
t) x
q
(t) sin(2f
0
t)
x(t) = x
q
(t) cos (2f
0
t) +x
i
(t) sin(2f
0
t)
Envelope and Phase of x(t), which is a real-valued and bandpass
signal
r
x
(t) =
_
x
2
i
(t) +x
2
q
(t)

x
(t) = arctan
x
q
(t)
x
i
(t)
Thus, x
l
(t) = r
x
(t)e

x
(t)
Instructor: Jia-Chin Lin 12
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
x(t) = 1
_
r
x
(t)e
j(2f
0
t+
x
(t))
_
and x(t) = r
x
(t) cos (2f
0
t +
x
(t))
Instructor: Jia-Chin Lin 13
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Recall Hilbert Transform: h(t) =
1
t
and H(f) = jsgn(f)
Instructor: Jia-Chin Lin 14
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Instructor: Jia-Chin Lin 15
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Energy Considerations
Energy of a signal x(t) is dened as E
x
=
_

[x(t)[
2
dt.
Rayleigh Relation: E
x
=
_

[x(t)[
2
dt =
_

[X(f)[
2
df.
No overlap between X
+
(f) and X

(f),
E
x
=
_

[X
+
(f) +X

(f)[
2
df
=
_

[X
+
(f)[
2
df +
_

[X

(f)[
2
df
= 2
_

[X
+
(f)[
2
df = 2E
x
+
Inner Product of two signals x(t) and y(t):
x(t), y(t) =
_

x(t)y

(t) dt =
_

X(f)Y

(f) df.
Low pass Equivalent of a Bandpass System:
Instructor: Jia-Chin Lin 16
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
h(t) = 1
_
h
l
(t)e
j2f
0
t
_
.
Originally, Y (f) = X(f)H(f) in passband.
Y
l
(f) = 2Y (f +f
0
) u
1
(f +f
0
)
= 2X (f +f
0
) H (f +f
0
) u
1
(f +f
0
)
=
1
2
[2X (f +f
0
) u
1
(f +f
0
)] [H (f +f
0
) u
1
(f +f
0
)]
=
1
2
X
l
(f)H
l
(f)
Based on the above, y
l
(t) =
1
2
x
l
(t) h
l
(t).
Instructor: Jia-Chin Lin 17
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Signal Space Representation
Vector Space Concepts
Inner Product of two n-dimensional vectors
v
1
= [v
11
v
12
v
1n
]
T
and v
2
= [v
21
v
22
v
2n
]
T
:
v
1
, v
2
= v
1
v
2
=
n

i=1
v
1i
v

2i
= v
2
H
v
1
v
1
, v
2
= v
2
, v
1

and thus,
v
1
, v
2
+ v
2
, v
1

= 21v
1
, v
2
.
Two vectors v
1
and v
2
are orthogonal if v
1
, v
2
= 0.
The norm of a vector |v| is dened as
|v| = (v, v)
1
2
=

_
n

i=1
[v
i
[
2
Instructor: Jia-Chin Lin 18
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Triangle Inequality
|v
1
+v
2
| |v
1
| + |v
2
|
Cauchy-Schwartz Inequality
[v
1
, v
2
[ |v
1
| |v
2
| ,
with equality if v
1
= av
2
for some complex scalar a.
Pythagorean Relation for two orthogonal n-dim vectors: if v
1
and v
2
are orthogonal, i.e., v
1
, v
2
= 0,
|v
1
+v
2
|
2
= (v
1
+v
2
) , (v
1
+v
2
) = (v
1
+v
2
)
H
(v
1
+v
2
)
= |v
1
|
2
+ |v
2
|
2
+ 21v
1
, v
2

= |v
1
|
2
+ |v
2
|
2
Linear Transformation: v

= Av
v

= A
nn
v
n1
= v
n1
Instructor: Jia-Chin Lin 19
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Engenvector of the transformation: v and is the
corresponding eigenvalue.
Gram-Schmidt Procedure
Constructing a set of orthonormal vectors from a set of
n-dimensional vectors, v
i
, 1 i m.
Procedure:
u
1
=
v
1
|v
1
|
u

2
= v
2
(v
2
, u
1
) u
1
u
2
=
u

2
|u

2
|
u

k
= v
k

k1

l=1
(v
k
, u
l
) u
l
u
k
=
u

k
|u

k
|
Instructor: Jia-Chin Lin 20
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Signal Space Concepts
Inner product: x
1
(t), x
2
(t) =
_

x
1
(t)x

2
(t) dt.
Norm: |x(t)| =
__

[x(t)[
2
dt
_
1
2
=
_
E
x
Triangle Inequality: |x
1
(t) +x
2
(t)| |x
1
(t)| + |x
2
(t)|
Cauchy-Schwartz Inequality:
[x
1
(t), x
2
(t)[ |x
1
(t)| |x
2
(t)| =
_
E
x
1
E
x
2
.

_

infty
x
1
(t)x

2
(t)dt

[x
1
(t)[
2
dt

1
2

[x
2
(t)[
2
dt

1
2
Orthogonal Expansions of Signals
A set of functions
n
(t), n = 1, 2, . . . , K are orthonormal:
<
n
(t),
m
(t) >=
_

n
(t)

m
(t) =
_
_
_
1 m = n
0 m ,= n
Instructor: Jia-Chin Lin 21
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
s(t) =
K

k=1
s
k

k
(t)
e(t) = s(t) s(t)
Select the coefcients s
k
, k = 1, 2, . . . , K minimize the energy E
e
of the approximation error:
E
e
=
_

[s(t) s(t)[
2
dt =
_

s(t)
K

k=1
s
k

k
(t)

2
dt
Minimizing mean square error criterion:
s
n
= s(t),
n
(t) =
_

s(t)

n
(t) dt, n = 1, 2, . . . , K
Instructor: Jia-Chin Lin 22
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Gram-Schmidt Procedure
A set of nite energy signal waveforms s
m
(t), m = 1, 2, . . . , M,

1
(t) =
s
1
(t)

E
1
, E
1
=
_

[s
1
(t)[
2
dt.

2
= s
2
(t) s
2
(t),
1
(t)
1
(t),

2
(t) =

2
(t)

E
2
,
E
2
=
_

2
(t)[
2
dt.

k
= s
k
(t)
k1

l=1
s
k
(t),
l
(t)
l
(t),

k
(t) =

k
(t)

E
k
,
E
k
=
_

k
(t)[
2
dt.
Instructor: Jia-Chin Lin 23
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Some Useful Random Variables
The Bernoulli Random Variable
The Binomial Random Variable
The Uniform Random Variable
Instructor: Jia-Chin Lin 24
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
The Gaussian (Normal) Random Variable
Instructor: Jia-Chin Lin 25
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
The Chi-Square (
2
) Random Variable
The Rayleigh Random Variable
Instructor: Jia-Chin Lin 26
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
The Ricean Random Variable
The Chi-Square (
2
) Random Variable
Instructor: Jia-Chin Lin 27
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
The Nakagami Random Variable
Instructor: Jia-Chin Lin 28
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
The Lognormal Random Variable
Instructor: Jia-Chin Lin 29
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Instructor: Jia-Chin Lin 30
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Bounds on Tail Probabilities
The Markov Inequality: for Non-negative random variable X,
P [X ]
E [X]

E[X] =
_

0
xp(x) dx
_

xp(x) dx
_

p(x)dx = P [X ]
Chebyshev inequality:
X is an arbitrary random variable with the mean

X and the
nite variance
2
X
.
for any > 0, P[X

X[

2

2
.

2
=
_

_
x

X
_
2
f
X
(x)dx
_
|x

X|
_
x

X
_
2
f
X
(x)dx

2
_
|x

X|
f
X
(x)dx =
2
P[[X

X[ ]
Chernoff bound: P[X a] e
at

X
(t),
Instructor: Jia-Chin Lin 31
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%

X
(t) = Ee
tX
=
_

e
tx
f
X
(x) dx
Instructor: Jia-Chin Lin 32
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Limit Theorems for Sums of Random
Variables
The (strong) law of large numbers (LLN): X
i
, i = 1, 2, . . . is a
sequence of IID random variables with E X
i
< , then
1
n
n

i=1
X
i
converges to EX
i
almost surely.
The central limit theorem (CLT): X
i
, i = 1, 2, . . . is a sequence of
IID random variables with m = E X
i
< and

2
= VARX
i
< , then
1
n
n

i=1
X
i
m

n
converges to N(0, 1) in
distribution.
Instructor: Jia-Chin Lin 33
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Complex Random Variables
Z = X+jY
C
X
= E
_
(X EX) (X EX)
T
_
C
Y
= E
_
(Y EY ) (Y EY )
T
_
C
XY
= E
_
(X EX) (Y EY )
T
_
C
YX
= E
_
(Y EY ) (X EX)
T
_

Z =
_
_
X
Y
_
_
Covariance matrix of

Z
C

Z
=
_
_
C
X
C
XY
C
YX
C
Y
_
_
Instructor: Jia-Chin Lin 34
Ch. 2: Deterministic and Random Signal Analysis Class: Digital Communications, Fall 2010
'
&
$
%
Covariance matrix of Z: C
Z
= E
_
(Z E Z) (Z E Z)
H
_
Hermitian and Nonnegative Denite.
A = A
H
; and a
H
C
Z
a 0, a
N1
,= 0.
Pseudocovariance matrix of Z:

C
Z
= E
_
(Z E Z) (Z E Z)
T
_
Skew-Hermiaian: A
H
= A.
Proper: A complex random vector Z is called proper if its
pseudocovariance is zero; i.e.,

C
Z
= 0.
Circularly Symmetric: A complex random vector Z is called
circularly symmetric if rotating the vector by any angle does not
change its PDF; i.e., Z and e
j
Z have the same PDF for all .
Instructor: Jia-Chin Lin 35