P i i l off Communications

Principles C i ti

Meixia Tao
Shanghai Jiao Tong University

Chapter 6: Signal Space Representation

Textbook: Chapter 8
8.1
1

2009/2010 Meixia Tao @ SJTU

 Signal Space Concepts
ƒ The key to analyzing and understanding the
performance of digital transmission is the
realization that
ƒ signals used in communications can be
expressed and visualized graphically

ƒ Thus, we need to understand signal space

p as applied
concepts pp to digital
g communications

2009/2010 Meixia Tao @ SJTU 2

 Traditional Bandpass Signal
R
Representations
t ti
ƒ Baseband signals
g are the message
g signal
g g
generated at the
source
ƒ Passband signals (also called bandpass signals) refer to
the signals after modulating with a carrier. The bandwidth
of these signals are usually small compared to the carrier
frequency fc
ƒ Passband signals can be represented in three forms

9 Magnitude and Phase representation

9 Quadrature representation
ƒ Complex Envelop representation

2009/2010 Meixia Tao @ SJTU 3

Magnitude and Phase Representation

Where a(t) is the magnitude of s(t)

and is the phase of s(t)

2009/2010 Meixia Tao @ SJTU 4

 Quadrature or I/Q Representation

where x(t) and y(t) are real-valued

real valued
baseband signals called the in-phase
and quadrature components of s(t).

Signal space is a more convenient way than

I/Q representation to study modulation scheme
2009/2010 Meixia Tao @ SJTU 5
 Vectors and Space
ƒ Consider an n-dimensional space with unity basis
vectors {e1, e2, …, en} (think of the x-y-z axis in a
coordinate system)
ƒ Any vector a in the space can be written as

Dimension = Minimum number of vectors that is

necessary and sufficient for
representation of any vector in space

2009/2010 Meixia Tao @ SJTU 6

ƒ Definitions:
ƒ Inner Product

ƒ a and b are Orthogonal if

ƒ = Norm of a

ƒ A set of vectors are orthonormal if they are mutually

orthogonal and all have unity norm

2009/2010 Meixia Tao @ SJTU 7

 Basis Vectors
ƒ The set of basis vectors {e1, e2, …, en} of a space are
chosen such that:
ƒ Should be complete
p or span
p the vector space:
p anyy vector a
can be expressed as a linear combination of these vectors.
ƒ Each basis vector should be orthogonal to all others

ƒ Each basis vector should be normalized:

ƒ A set of basis vectors satisfying these properties is also

said to be a complete orthonormal basis
ƒ In an n-dim space, we can have at most n basis vectors

2009/2010 Meixia Tao @ SJTU 8

 Signal Space
ƒ Basic Idea: If a signal
g can be represented
p byy n-
tuple, then it can be treated in much the same
way as a n-dim vector.
ƒ Let φ1(t), φ2(t),…., φn(t) be n signals
• Consider a signal x(t) and suppose that

ƒ If every signal can be written as above ⇒

~ ~ basis functions and we have a
n-dim signal
g space
p
2009/2010 Meixia Tao @ SJTU 9
 Orthonormal Basis
ƒ Signal set {φk(t)}n is an orthogonal set if
∞
⎧0 j≠k
∫−∞φ j ( t )φk ( t ) dt = ⎨⎩c j j=k

ƒ If cj≡1 ∀j ⇒ {φk(t)} is an orthonormal set.

set
ƒ In this case, ∞
xk = ∫ x(t )φk (t )dt
−∞
n
x(t ) = ∑ xiφi (t )
i =1

x = ( x1, x2 ,..., xn )
2009/2010 Meixia Tao @ SJTU 10
 Key Property
ƒ Given the set of the orthonormal basis

ƒ Let and be represented as

,

with ,

¾ Then the inner p

product of and is

2009/2010 Meixia Tao @ SJTU 11

ƒ Proof

Since ∞
⎧0 i ≠ j
∫−∞φi (t )φ j (t )dt = ⎨⎩1 i = j

¾ Ex = Energy off x(t)

()=

2009/2010 Meixia Tao @ SJTU 12

 Basis Functions for a Signal Set
ƒ Consider a set of M signals (M-ary
(M ary symbol) {si(t),
i=1,2,…,M } with finite energy. That is

ƒ Then, we can express each of these waveforms as

weighted linear combination of orthonormal signals

where N ≤ M is the dimension of the signal space

and are called the orthonormal basis functions

2009/2010 Meixia Tao @ SJTU 13

 Example 1

Consider the following signal set:

2009/2010 Meixia Tao @ SJTU 14

 Example 1 (Cont
(Cont’d)
d)
ƒ By inspection, the signals can be expressed in terms of
the following two basis functions:

+1 +1

1 2 1 2

ƒ Not that the basis is orthogonal

ƒ Also note that each these functions have unit energy

ƒ We say that they form an orthonormal basis

2009/2010 Meixia Tao @ SJTU 15
 Example 1 (Cont
(Cont’d)
d)
Constellation diagram:
ƒ A representation of a digital 1
modulation scheme in the
signal space -1 1

ƒ Axes are labeled with φ1(t)

-1
and φ2(t)
ƒ Possible signals are plotted
as points, called
constellation points

2009/2010 Meixia Tao @ SJTU 16

 Example 2
ƒ Suppose our signal set can be represented in the
following form

with
ith and
d

ƒ We can choose the basis functions as follows

2009/2010 Meixia Tao @ SJTU 17

 Example 2 (Cont
(Cont’d)
d)
ƒ Since

and

ƒ The basis functions are thus orthogonal and they are also
normalized

2009/2010 Meixia Tao @ SJTU 18

 Example 2 (Cont
(Cont’d)
d)
ƒ These basis functions are quite common and can
describe various modulation schemes
ƒ Example 2 is QPSK modulation
modulation. Its constellation
diagram is identical to Example 1

2009/2010 Meixia Tao @ SJTU 19

 Notes on Signal Space
ƒ Two entirely different signal sets can have the same
geometric representation.
ƒ The underlying geometry will determine the
performance and the receiver structure for a signal set
ƒ IIn previous
i examples,
l we were able
bl to guess the
h
correct basic functions
ƒ In general, is there any method which allows us to find
a complete orthonormal basis for an arbitrary singal
set?
ƒ Gram
Gram--Schmidt Orthogonalization (GSO) Procedure

2009/2010 Meixia Tao @ SJTU 20

 Gram Schmidt Orthogonalization (GSO)
P
Procedure
d
ƒ Suppose we are given a signal set

ƒ Find the orthogonal basis functions for this signal

set

2009/2010 Meixia Tao @ SJTU 21

 Step 1: Construct the First Basis Function

ƒ Compute the energy in signal 1:

ƒ The first basis function is just a normalized version

of s1(t)

2009/2010 Meixia Tao @ SJTU 22

 Step 2: Construct the Second Basis
F
Function
ti
ƒ Compute correlation between signal 2 and basic function 1

ƒ Subtract
S bt t off
ff the
th correlation
l ti portion
ti

ƒ Compute the energy in the remaining portion

ƒ Normalize the remaining portion

2009/2010 Meixia Tao @ SJTU 23

 Step 3: Construct Successive Basis
F
Functions
ti
ƒ For signal , compute

ƒ Define

ƒ Energy of :

ƒ k-th basis function:

ƒ In general

2009/2010 Meixia Tao @ SJTU 24

 Summary of GSO Procedure
ƒ 1st basis function is normalized version of the first
signal
ƒ Successive basis functions are found by removing
portions of signals that are correlated to previous
basis functions and normalizing the result
ƒ This procedure is repeated until all basis functions are
found
ƒ If , no new basis functions is added
ƒ The order in which signals are considered is arbitrary

2009/2010 Meixia Tao @ SJTU 25

 Example: GSO
1) Use the Gram-Schmidt
Gram Schmidt procedure to find a set
orthonormal basis functions corresponding to the signals
show below

2) Express x1, x2, and x3 in terms of the orthonormal basis

functions found in part 1)
3) Draw the constellation diagram for this signal set

2009/2010 Meixia Tao @ SJTU 26

 Solution: 1)
Step 1:

1 2

Step 2:

and

1 2
2009/2010 Meixia Tao @ SJTU 27
 Solution: 1) (Cont
(Cont’d)
d)
ƒ Step 3:

=> No more new basis functions

Procedure completes

2009/2010 Meixia Tao @ SJTU 28

 Solution: 2) and 3)
ƒ Express x1, x2, x3 in basis functions
,

ƒ Constellation diagram

2009/2010 Meixia Tao @ SJTU 29

 Exercise
Given a set of signals (8PSK modulation)

ƒ Find the orthonormal basis functions using Gram

Schmidt procedure
ƒ What is the dimension of the resulting signal space ?
ƒ Draw the constellation diagram of this signal set

2009/2010 Meixia Tao @ SJTU 30

 Notes on GSO Procedure
ƒ A signal set may have many different sets of basis
functions
ƒ A change of basis functions is essentially a rotation
of the signal points around the origin.
ƒ The order in which signal are used in the GSO
procedure affect the resulting basis functions
ƒ Th
The choice
h i off b
basis
i ffunctions
ti
does not affect the performance
of the modulation scheme

2009/2010 Meixia Tao @ SJTU 31