Wireless Communications Group

Representation of bandpass signals

Francois Horlin

1
 Outline

Introduction
Bandpass deterministic signals
Bandpass system
Bandpass random signals
Exercises

2
 References

Introduction in ELEC-H-311 - Signaux et systemes de

telecommunications
Digital Communications, Fourth Edition, J. G. Proakis

3
 Outline

Introduction
Bandpass deterministic signals
Bandpass system
Bandpass random signals
Exercises

4
 Bandpass signal

Most of the digitally modulated signals occupy a limited bandwidth

around a carrier frequency so that the spectrum resources can be
shared among multiple channels

Narrowband bandpass signals: the bandwidth is much smaller than

the carrier frequency

5
 Motivation

Goal: represent the narrowband bandpass signals to equivalent

baseband signals

Advantages:
Reduce significantly the necessary signal sampling rate (and
therefore reduce the computational complexity)
Develop modulation/demodulation techniques independently of
carrier frequency (and therefore support system flexibility)
Consequently all wireless communication systems are built based on
the signal baseband representation

6
 Outline

Introduction
Bandpass deterministic signals
Bandpass system
Bandpass random signals
Exercises

7
 Outline: Bandpass deterministic signals

Hilbert transform and Analytic signal

Complex envelope
Rice components

8
 Narrowband bandpass signal

X()
x(t) real and deterministic:

[X()] = [X()] 1

[X()] = [X()] 0 0

No information is lost when the

negative part of the spectrum is 0 0

removed 1

9
 Analytic signal

Xa() = X() + j X()

X() filtered with:
2

0 <0
U () :=
2 0
0 2 0

gives the analytic signal:

0 0
0 <0
Xa () :=
2X() 0

10
 Analytic signal

Filter:

u(t) := TF1 [U ()]

j
= (t) +
t

Signal:

xa (t) = x(t) u(t)

1
= x(t) + jx(t)
t
:= x(t) + j x(t)

where x(t) is the Hilbert transform of x(t)

11
 Hilbert transform

X() = j sgn() X()

x(t) = j (xa (t) x(t)) 1

X() = j (Xa () X()) 0 0

jX() <0 1

=
jX() 0
0 0

= jsgn()X() 1

12
 Complex envelope

Ex() = Xa( + 0)

Frequency translation to the 2

origin:

0 2 0

Ex () := Xa ( + 0 )

j0 t
ex (t) = e xa (t) 0 0

13
 Rice components

A narrowband signal can be seen as the addition of two carriers in

quadrature (cos/sin) modulated by the real signals x1 (t) and x2 (t):

ex (t) := x1 (t) jx2 (t)

x(t) = [xa (t)]

j0 t
 
= ex (t)e
j0 t
 
= (x1 (t) jx2 (t))e
= x1 (t) cos (0 t) + x2 (t) sin (0 t)

14
 Rice components

x1 (t)
cos (0 t)
x(t)
sin (0 t)
x2 (t)

15
 Rice component 1

Ex ()+Ex ()
X1() = 2

x1 (t) = [ex (t)] 2

 j0 t 
= e xa (t)

 j0 t 
0 0
= e (x(t) + j x(t)) 2

= x(t) cos (0 t) + x(t) sin (0 t)

0 0

Note: computation of x1 (t) diffi-

2
cult to implement

16
 Rice component 2

X2() = j Ex()E x ()

x2 (t) = [ex (t)]
0 0
 j0 t 
= e xa (t)
 j0 t  2

= e (x(t) + j x(t)) 2

= x(t) sin (0 t) x(t) cos (0 t)

0 0

Note: computation of x2 (t) diffi-

2
cult to implement

17
 Practical computation of Rice components

2
2x(t) cos (0 t) = 2x1 (t) (cos (0 t)) + 2x2 (t) sin (0 t) cos (0 t)
= x1 (t) + x1 (t) cos (20 t) + x2 (t) sin (20 t)
2
2x(t) sin (0 t) = 2x2 (t) (sin (0 t)) + 2x1 (t) sin (0 t) cos (0 t)
= x2 (t) x2 (t) cos (20 t) + x1 (t) sin (20 t)

LPF x1 (t)
cos (0 t)
x(t)
sin (0 t)

LPF x2 (t)

18
 Alternative

A narrowband signal can also be seen as a carrier modulated in

amplitude and in phase:

ex (t) := x (t)ejx (t)

x(t) = [xa (t)]

j0 t
 
= ex (t)e
h i
= x (t)ejx (t) ej0 t
= x (t) cos (0 t + x (t))

19
 Alternative

Link between Rice components and amplitude/phase modulations:

q
x (t) = x21 (t) + x22 (t)
x (t) = arctan (x2 (t)/x1 (t))

20
 Energy of a deterministic signal

Z +
E = x2 (t) dt

Z +
2
= x2 (t) (cos (0 t + x (t))) dt

Z +
1
= x2 (t) (1 + cos (20 t + 2x (t))) dt
2
+
1
Z
x2 (t) dt
2

Energy can be computed based on the amplitude of the signal

21
 Analog front-end diagram *

A direct conversion analog front-end implements the

bandpass/baseband translation

22
 Outline

Introduction
Bandpass deterministic signals
Bandpass system
Bandpass random signals
Exercises

23
 Bandpass system

The response y(t) to a bandpass system h(t) can be computed in the

baseband domain (factor 1/2 to not accumulate the factors):

1
ey (t) = ex (t) eh (t)
2

24
 Outline

Introduction
Bandpass deterministic signals
Bandpass system
Bandpass random signals
Exercises

25
 Outline: Bandpass random signals

Hilbert transform and Analytic signal

Complex envelope
Rice components

26
 Random signals

Assume the centered and stationary signal x(t)

Goal: compute the auto-correlation / power spectral density functions

of the analytic signal, complex envelope and Rice components

27
 Hilbert transform

X () = X ()

Wiener-Kintchine:
2 1
x () = |jsgn()| x ()

= x () 0 0

Inverse Fourier transform:

0 0
x ( ) = x ( )

28
 Hilbert transform

Wiener-Kintchine: XX () = X X () = j sgn() X ()

xx () = j sgn() x ()
xx () = j sgn() x ()

= xx () 0 0

1
Inverse Fourier transform:

0 0
xx ( ) = x ( )
1

= xx ( )

29
 Analytic signal

Wiener Kintchine: Xa () = 2X () + 2j XX ()

xa () = |U ()|2 x () 4

0 <0

= 0 0

4x () 0

Inverse Fourier transform (simi-
0 0
larly to deterministic signals):

xa ( ) = 2x ( ) + 2j x ( )

30
 Complex envelope

Ex() = Xa ( + 0)

4
Frequency translation:

ex () = xa ( + 0 )
0 0

Inverse Fourier transform:

ex ( ) = ej0 xa ( ) 0 0

31
 In terms of power

1
x ( ) = [xa ( )]
2
1  j0 
= e ex ( )
2

The power of a bandpass signal is equal to the power of its complex

envelope divided by 2:
2 1 2
x = ex
2

32
 Rice components

x1 ( ) = E [x1 (t + ) x1 (t)]
= E[(x(t + ) cos (0 (t + )) + x(t + ) sin (0 (t + )))
(x(t) cos (0 t) + x(t) sin (0 t))]
= x ( ) cos (0 (t + )) cos (0 t)
+ x ( ) sin (0 (t + )) sin (0 t)
+ xx ( ) cos (0 (t + )) sin (0 t)
+ xx ( ) sin (0 (t + )) cos (0 t)
= x ( ) cos (0 ) + xx ( ) sin (0 )
= x ( ) cos (0 ) + x ( ) sin (0 )

33
 Rice components

x2 ( ) = E [x2 (t + ) x2 (t)]
= E[(x(t + ) sin (0 (t + )) x(t + ) cos (0 (t + )))
(x(t) sin (0 t) x(t) cos (0 t))]
= x ( ) sin (0 (t + )) sin (0 t)
+ x ( ) cos (0 (t + )) cos (0 t)
xx ( ) sin (0 (t + )) cos (0 t)
xx ( ) cos (0 (t + )) sin (0 t)
= x ( ) cos (0 ) + xx ( ) sin (0 )
= x ( ) cos (0 ) + x ( ) sin (0 )
= x1 ( )

34
 Rice components

x1 x2 ( ) = E [x1 (t + ) x2 (t)]
= E[(x(t + ) cos (0 (t + )) + x(t + ) sin (0 (t + )))
(x(t) sin (0 t) x(t) cos (0 t))]
= x ( ) cos (0 (t + )) sin (0 t)
x ( ) sin (0 (t + )) cos (0 t)
xx ( ) cos (0 (t + )) cos (0 t)
+ xx ( ) sin (0 (t + )) sin (0 t)
= x ( ) sin (0 ) + xx ( ) cos (0 )
= x ( ) sin (0 ) + x ( ) cos (0 )

35
 Rice components

x1 () = x2 ()

( ) + ( + ) || < B
x 0 x 0
=
0 else

j ( ( ) ( + )) || < B
x 0 x 0
x1 x2 () =
0 else

36
 On the other hand

ex ( ) = x1 ( ) + x2 ( ) + jx1 x2 ( ) jx2 x1 ( )
= 2x1 ( ) + 2jx1 x2 ( )

37
 Rice components

Ex ()E x ()
Ex ()+E x () X1X2 () = X2X1 () = j
X1 () = X2 () = 4
4

0 0 0 2 0

0 0 0 0

38
 In terms of power

The power of each Rice component is equal to the power of the

bandpass signal:
x21 = x22 = x1 (0) = x2

The Rice components are uncorrelated at the same instant:

x1 x2 (0) = x ( = 0)
Z +
1
= [jsng()] x () d
2
= 0

39
 Outline

Introduction
Bandpass deterministic signals
Bandpass system
Bandpass random signals
Exercises

40
 Exercise 1

Characterize the complex envelope of the white Gaussian noise n(t) of

power spectral density n () = N0 /2.

41
 Exercise 2

Describe how a wireless communication system operating at a carrier

frequency f0 with a bandwidth B can be efficiently simulated with the
Matlab software.
The performance is often assessed by evaluating the bit error
probability (bit error rate) as a function of the ratio bit energy on
noise power spectral density.

42