You are on page 1of 35

Wireless Communications Group

Transceiver Design

Francois Horlin

1
Outline

Introduction
Transceiver block diagram
Spectral characteristics
Nyquist criterion
Exercises

2
References

Digital Communications, Fourth Edition, J. G. Proakis

3
Outline

Introduction
Transceiver block diagram
Spectral characteristics
Nyquist criterion
Exercises

4
Motivation

Instead of considering only one transmitted symbol, extend the signal


description to a sequence of symbols
Design shaping functions spectrally more efficient than the rectangular
window

Discussion applicable to PAM, PSK and QAM signals, but not valid
for FSK signals

5
Outline

Introduction
Transceiver block diagram
Spectral characteristics
Nyquist criterion
Exercises

6
Transceiver block diagram

<
I[n] es (t) s(t)
b[n] g(t) cos(2fc t)

mapping = n(t)
sin(2fc t)

nT
y[n] y(t) er (t) r(t)
b[n] g( t) cos(2fc t)
j
demapping

sin(2fc t)

7
Baseband equivalent model

I[n] es (t)
b[n] g(t)

mapping en (t)

nT
y[n] y(t) er (t)
b[n] g( t)

demapping

8
Baseband equivalent model

The binary sequence b[n] is mapped on the sequence of complex


symbol I[n]
The baseband equivalent transmitted signal is formed by shaping the
symbols with a rectangular window g(t):
X
es (t) = I[n]g(t nT )
n

Additive white Gaussian noise corrupts the received signal:

er (t) = es (t) + en (t)

9
Received sequence

The received signal is matched filtered and sampled at symbol rate:

y[n] = (er (t) g(t))|t=nT


X Z
= I[m]h(nT mT ) + en (t)g(t nT )dt
m t
Z
= I[n] + en (t)g(t nT )dt
t

where h(t) = g(t) g(t)

The received sequence is the symbol sequence corrupted by noise

10
Outline

Introduction
Transceiver block diagram
Spectral characteristics
Nyquist criterion

11
Power spectra of linearly modulated signals

To compute the power spectral density (PSD) of the modulated


signal, it is sufficient to compute the PSD of the corresponding
complex envelope:

1  jc

s ( ) = < es ( ) e
2
1 jc jc

= es ( ) e + es ( ) e
4
1
s () = (es ( c ) + es ( c ))
4

12
Expression of the complex envelope

The baseband equivalent transmitted signal is given by:


X
es (t) = I[n] g(t nT T0 )
n

where:
I[n] are the complex symbols
g(t) is the shaping pulse
T0 is the time origin
The time origin T0 is a random variable uniformly distributed on
[0, T ] introduced to make the complex envelope es (t) stationary (else
the auto-correlation es ( ) is periodic of period T )

13
Computation of the auto-correlation

The auto-correlation of the complex envelope is given by:

es ( ) := EI,T0 [es (t + ) es (t)]


XX
= EI [I[n] I [n0 ]]
n n0
T
1
Z
g(t + nT T0 ) g (t n0 T T0 ) dT0
T T0 =0

14
Computation of the auto-correlation

It reduces to:
Z +
1 X
es ( ) = I (m) g(t + u) g (t + mT u) du
T m u=

1 X
= I (m) Cg ( mT )
T m

by defining:
the symbol auto-correlation I (m) := EI [I[n]I [n m]]
R +
the pulse time auto-correlation Cg ( ) := u= g(u + )g (u)du

15
Power spectral density

After Fourier transform, the PSD of the complex envelope is given by:
" #
1 X 2
es () = I [m] ejmT |G()|
T m

The spectral characteristics of the signal can be controlled by:


the design of the pulse shape g(t)
the design of the information sequence correlation I [n]
When the information symbols are independent of variance I2 , the
PSD reduces to:
I2 2
es () = |G()|
T

16
Rectangular pulse shape

10

15

|G(f)|2
20

25

30

35

40
4 3 2 1 0 1 2 3 4
f T []

The rectangular pulse shape causes a high level of the side lobes,
leading to interference on the adjacent channels
Alternative pulses will be investigated

17
Outline

Introduction
Transceiver block diagram
Spectral characteristics
Nyquist criterion
Exercises

18
Towards a better pulse shape...

The shaping pulse g(t) should ideally be such that:


The spectrum of the transmitted signal is limited to a well-defined
bandwidth
There is no inter-symbol interference in the received sequence (as
it is the case for a rectangular pulse)

19
Expression of the inter-symbol interference

In the absence of noise, the received sequence is equal to:


X
y[n] = I[m]h(nT mT )
m
X
= I[n]h(0) + I[m]h(nT mT )
m6=n

The first term is the term of interest, the second term is the
inter-symbol interference (ISI)

20
Nyquist criterion

To cancel the inter-symbol interference, the pulse should be such that:



1 (n = 0)
h(nT ) =
0 (else)

Equivalently in the frequency domain:



X m
H(f + ) = T
m=
T

Demonstration in the next slides...

21
Demonstration
Inverse Fourier transform and sampling:
Z
h(nT ) = H(f )ej2f nT df
f =

Break the integral into integrals covering a finite range 1/T :


X Z (2m+1)/2T
h(nT ) = H(f )ej2f nT df
m= f =(2m1)/2T
Z 1/2T
X m j2f nT
= H(f + )e df
m= f =1/2T
T

Z 1/2T !
X m
= H(f + ) ej2f nT df
f =1/2T m=
T

22
Demonstration

Therefore:
Z 1/2T
h(nT ) = B(f )ej2f nT df
f =1/2T

by defining:

X m
B(f ) := H(f + )
m=
T

23
Demonstration

The function B(f ) is periodic and can be expanded in terms of its


Fourier series coefficients b(n):

X
B(f ) = b[n]ej2f nT
n=

with:
Z 1/2T
b[n] = T B(f )ej2f nT df
f =1/2T

24
Demonstration

By comparing the expressions: b[n] = T h(nT )


No inter-symbol interference if:

T (n = 0)
b[n] =
0 (else)

or equivalently if:
B(f ) = T
P m
Making use of the definition of B(f ): m= H(f + T ) = T

25
Case 1

When the pulse bandwidth W is such that W < 1/2T , B(f ) consists of
non-overlapping replicas and there is no choice for H(f ) to cancel ISI

26
Case 2

When W = 1/2T , there is only one solution to cancel ISI:



T (|f | < W ) sin (t/T )
H(f ) = h(t) =
0 (else) t/T

27
Case 3

When W > 1/2T , B(f ) consists of overlapping replicas of H(f ) and


there exists an infinity of solutions to cancel ISI

28
Raised cosine or Nyquist pulse

29
Raised cosine or Nyquist pulse

 
1


T 0 |f | < 2T
 h i  
1
H(f ) = T
2 1 + cos T
(|f | 2T ) ( 1
2T |f |
1+
2T

 
0 |f | > 1+

2T

sin (t/T ) cos t/T


h(t) =
t/T 1 42 t2 /T 2

30
Raised cosine or Nyquist pulse

The overall raised cosine pulse h(t) is split evenly between the shaping
pulse g(t) implemented at the transmitter and its matched filter g(t)
implemented at the receiver

Therefore:
p
G(f ) = H(f )

31
Outline

Introduction
Transceiver block diagram
Spectral characteristics
Nyquist criterion
Exercises

32
Exercise 1

Demonstrate that the transceiver block diagram implements the


optimal receiver composed of:
A correlator with the basis functions (or bank of matched filters)
The maximum likelihood symbol detector

33
Exercise 2

Demonstrate that there is no inter-symbol interference in the received


sequence when the symbols are shaped with a rectangular window

34
Exercise 3

Demonstrate that a linearly modulated signal generated by shaping a


stream of independent symbols of variance I2 and duration T with a
low-pass filter of transmittance G() has a PSD equal to:
I2
() = |G()|2 .
T

35

You might also like