You are on page 1of 20

Digital Communications

Instructor: Dr. Phan Van Ca Lecture 04 : Signal-Space Approach to Modulation and Gram-Schmidt Procedure

Block Diagram of Digital Communications System


Analog Input Signal Sample Quantize Digital Input Data Source Encoder Encryption Channel Encoder Modulator Channel Digital Output Data Analog Output Signal D/A Converter Source Decoder Decryption Channel Decoder Equalization Demodulator

Modulation Principles

Almost all communication systems transmit digital data using a sinusoidal carrier waveform.
z Electromagnetic signals propagate well z Choice of carrier frequency allows placement of signal in

arbitrary part of spectrum

Physical system implements modulation by:


z Processing digital information at baseband z Pulse shaping and filtering of digital waveform z Baseband signal is mixed with signal from oscillator z RF signal is filtered, amplified and coupled with antenna

Representation of Modulation Signals


We can modify amplitude, phase or frequency. Amplitude Shift Keying (ASK) or On/Off Keying (OOK):
1 A cos(2 f c t ),0 0

Frequency Shift Keying (FSK): 1 A cos(2 f1t ),0 A cos(2 f 0 t ) Phase Shift Keying (PSK):

1 A cos(2 f ct )

0 A cos(2 f ct + ) = A cos(2 f ct )

Representation of Bandpass Signals


Bandpass signals (signals with small bandwidth compared to carrier frequency) can be represented in any of three standard formats:

Quadrature Notation

s( t ) = x ( t ) cos(2 f ct ) y ( t ) sin(2 f ct )
where x(t) and y(t) are real-valued baseband signals called the in-phase and quadrature components of s(t)

Representation of Bandpass Signals (Contd)

Complex Envelope Notation

s( t ) = Re ( x ( t ) + jy ( t ) )e j 2 f ct = Re sl ( t )e j 2 f ct

] [

where sl (t ) is the complex envelope of s(t).

Magnitude and Phase

s( t ) = a ( t ) cos(2 f ct + ( t ))
where a ( t ) = x 2 ( t ) + y 2 ( t ) and
y (t ) ( t ) = tan 1 x (t )

is the magnitude of s(t), is the phase of s(t).

Key Ideas from I/Q Representation of Signals

We can represent bandpass signals independent of carrier frequency. The idea of quadrature sets up a coordinate system for looking at common modulation types. The coordinate system is sometimes called a signal constellation diagram.

Example of Signal Constellation Diagram: BPSK

x ( t ) { 1}, y ( t ) = 0

y( t )

x (t )

Example of Signal Constellation Diagram: QPSK

x ( t ) { 1}, y ( t ) { 1}

y( t )
X X

x (t )
X X

Example of Signal Constellation Diagram: QAM

x ( t ) { 3,1,+1,+3}, y ( t ) { 3,1,+1,+3}

y( t )
X X X X X X X X X X X X X X X X
x(t )

Interpretation of Signal Constellation Diagram


Axis are labeled with x(t) and y(t) Possible signals are plotted as points Signal power is proportional to distance from origin Probability of mistaking one signal for another is related to the distance between signal points Decisions are made on the received signal based on the distance to signal points in constellation

A New Way of Viewing Modulation

The I/Q representation of modulation is very convenient for some modulation types. We will examine an even more general way of looking at modulation using signal spaces. By choosing an appropriate set of axis for our signal constellation, we will be able to:
z Design modulation types which have desirable properties z Construct optimal receivers for a given type of modulation z Analyze the performance of modulation types using very

general techniques.

Vector Spaces

An n-dimensional vector v = [v1, v2 , , vn ] consists of n scalar components {v1, v2 , , vn } The norm (length) of a vector v is given by:
v =
i =1

vi 2

The inner product of two vectors v1 = [v11, v12 , , v1n ] and v 2 = [v21, v22 , , v2n ] is given by: n v1 v 2 = v1i v2i i =1

Basis Vectors

A vector v may be expressed as a linear combination of its basis vectors {e1, e2 , , en } :


v = vi ei
i =1 n

where vi = ei v

Think of the basis vectors as a coordinate system (x-y-z... axis) for describing the vector v What makes a good choice of coordinate system?

A Complete Orthornomal Basis

The set of basis vectors {e1, e2 , , en } should be complete or span the n vector space {vi } . Any vector can be n expressed as v = vi ei for some
i =1

Each basis vector should be orthogonal to all others:


ei e j = 0, i j

Each basis vector should be normalized :

ei = 1, i

A set of basis vectors which satisfies these three properties is said to be a complete orthonormal basis.

Signal Spaces
Signals can be treated in much the same way as vectors. The norm of a signal x ( t ), t [a , b] is given by:
b x( t ) = x(t ) 2 dt a
1/ 2

= Ex

The inner product of signals


b a

x1( t )

and

x2 ( t )

is:

x1( t ), x2 (t ) = x1( t ) x2 ( t ) dt

Signals can be represented as the sum of basis functions:


x ( t ) = xk f k ( t ),
i =1 n

xk = x ( t ), f k ( t )

Basis Functions for a Signal Set


One of M signals is transmitted: {s1(t ), , s M (t )} The functions { f1( t ), , f K ( t )} ( K M ) form a complete orthonormal basis for the signal set if

z Any signal can be described by a linear combination:


si (t ) = si , k f k ( t ), i = 1, , M
k =1 K

z The basis functions are orthogonal to each other:


b a f i ( t ) f j ( t ) dt = 0, i j b a

z The basis functions are normalized: f k ( t ) 2 dt = 1, k

Example of Signal Space


Consider the following signal signal set:

s1(t )
+1

s 2 (t )
+1

t
-1 1 2 -1 1 2

s 3 (t )
+1

s 4 (t )
+1

t
-1 1 2 -1

Example of Signal Space (Contd)

We can express each of the signals in terms of the following basis functions:

f 1(t )
+1

f 2(t )
+1

t
2 1 -1 s1( t ) = 1 f1( t ) + 1 f 2 ( t ) s3 ( t ) = 1 f1 ( t ) + 1 f 2 ( t )

t
2 1 -1 s2 ( t ) = 1 f1 ( t ) 1 f 2 ( t ) s4 ( t ) = 1 f1 ( t ) 1 f 2 ( t )

Therefore the basis is complete

Example of Signal Space (Contd)

The basis is orthogonal:


f1( t ) f 2 ( t ) dt = 0

The basis is normalized:


2 f1(t ) dt = f 2 ( t ) dt = 1 2

Signal Constellation for Example

Weve seen this signal constellation before

f2 (t )
X s (t ) 3 X s (t ) 1

f1(t )
X s (t ) 4 X

s2 (t )

Another Example

Suppose our signal set can be represented in I/Q form:

s( t ) = x ( t ) cos(2 f ct ) y ( t ) sin(2 f ct ) 0

Then the functions:

where x (t ) and y(t ) are constants for t [0, T ]


T T

form a complete orthonormal basis

2 2 f1 ( t ) = cos(2 f ct ) , f 2 ( t ) = sin(2 f ct ) T T 0 0

Proof

All I/Q signals can be represented by the linear combination of these basis functions. These basis functions are orthogonal:
T T 2 2 cos(2 f ct ) sin(2 f ct )dt f1(t ) f 2 ( t ) dt = T 0 0 T 2 T1 = [sin(0) + sin(4 f ct )]dt T 02 1 T cos(4 f ct )] 0, for f cT >> 1 = [ 0 4 f cT

Proof (Contd)

These basis functions are normalized:


T

cos(2 f ct ) dt f1(t ) dt = f 2 ( t ) dt = 0 0 0 T
2 2

T 2

2 T1 1 = [ cos(0) + cos(4 f ct )]dt [1] T =1 T 02 T 0

Signal Constellation for QPSK


2 2 x ( t ) , y ( t ) T T

f2 (t )
X X

f1(t )
X X

Notes on Signal Spaces

Two entirely different signal sets can have the same geometric representation. The underlying geometry will determine the performance and the receiver structure for a signal set. In both of these cases we were fortunate enough to guess the correct basis functions. Is there a general method to find a complete orthonormal basis for an arbitrary signal set?
z Gram-Schmidt Procedure

Gram-Schmidt Procedure

Suppose we are given a signal set: {s1( t ), , s M ( t )} Find a complete orthonormal basis { f1( t ), , f K ( t )} K M for this signal set.

Step 1: Construct the First Basis Function

Compute the energy in signal 1:


E1 = s1( t ) 2 dt

The first basis function is just a normalized version of s1( t ) : s (t ) f1 ( t ) = 1


E1

Step 2: Construct the Second Basis Function

Compute correlation with 1st basis: c12 = s2 (t ) f1(t )dt

Subtract off correlated portion: f 2 ( t ) = s2 (t ) c12 f1( t )

Compute energy in remaining portion: E 2 = f 2 (t ) dt

f 2 (t ) Normalize basis function: f 2 ( t ) = E 2

Procedure for Successive Signals


cik = sk ( t ) f i ( t ) dt , i = 1,..., k 1

f k ( t ) = sk ( t ) cik f i ( t )
i =1
2

k 1

E k = f k ( t ) dt

f k (t ) f k (t ) = Ek

Summary of Gram-Schmidt Procedure


1st basis function is normalized version of 1st signal. Successive basis functions are found by removing portions of signals which are correlated to previous basis functions, and normalizing the result. This procedure is repeated until all basis functions are exhausted. If f k (t ) = 0 , then no new basis function is added. The order in which signals are considered is arbitrary.

Example of Gram-Schmidt Procedure

Use the signal set from last time:

s1(t )
+1

s 2 (t )
+1

t
-1 1 2 -1 1 2

s 3 (t )
+1

s 4 (t )
+1

t
-1 1 2 -1

Step 1
E1 = s1( t ) 2 dt = 2

s (t ) s (t ) f1( t ) = 1 = 1 E1 2

1 2 1 2

f 1(t )
t
1 2

Step 2
c12 = f1( t ) s2 ( t )dt = 0

f 2 ( t ) = s2 ( t ) c12 f1( t ) = s2 ( t )

E 2 = s2 ( t ) 2 dt = 2

1 2 1 2

f 2(t )
t
1 2

s (t ) s (t ) f 2 (t ) = 2 = 2 E2 2

Step 3
c13 = f1( t ) s3( t )dt = 0 c23 = f 2 ( t ) s3 ( t )dt = 2

f 3 ( t ) = s3( t ) c13 f1( t ) c23 f 2 (t ) = s3 ( t ) + 2 f 2 ( t ) = 0

No new basis function

Step 4
c14 = f1( t ) s4 ( t )dt = 2 c24 = f 2 ( t ) s4 ( t )dt = 0

f 4 ( t ) = s4 ( t ) c14 f1( t ) c24 f 2 ( t ) = s4 ( t ) + 2 f1( t ) = 0

No new basis function. Procedure Complete

Signal Constellation Diagram

s2 (t ) f2 (t ) 2 X s4 (t )
2 2 X
X

s1(t )
X

f1(t )

s3 (t )

Second Example of Gram-Schmidt Procedure


T M-ary PSK : si ( t ) = cos 2f ct + i , i = 0,,7 4 0
T 2 2 sin(2 f ct ) cos(2 f ct ) , f 2 ( t ) = f 1( t ) = T T 0 0

i c = sin i c = cos , 1i 1i 4 4

si (t ) = c1i f1(t ) + c2i f2(t ) = x(t )cos(2fct ) y(t )sin(2fct )

8-ary PSK

Notes on Gram-Schmidt Procedure

A signal set may have many different sets of basis functions. A change of basis functions is equivalent to rotating coordinates. The order in which signals are used in the Gram-Schmidt procedure will affect the resulting basis functions. The choice of basis functions does not effect performance.