Module

4
Signal Representation
and Baseband
Processing

Version 2 ECE IIT, Kharagpur

 Lesson
15

Orthogonality

Version 2 ECE IIT, Kharagpur

After reading this lesson, you will learn about

¾ Basic concept of orthogonality and orthonormality;

¾ Strum - Lion;
¾ Slope overload distortion;
¾ Granular Noise;
¾ Condition for avoiding slope overloading;

The Issue of Orthogonality

Let fm(x) and fn(x) be two real valued functions defined over the interval a ≤ x ≤ b. If the
product [fm(x)×fn(x)] exists over the interval, the two functions are called orthogonal to
each other in the interval a ≤ x ≤ b when the following condition holds:
a

∫f
b
m ( x) f n ( x) dx = 0 , m≠ n 4.15.1

A set of real valued functions f1(x), f2(x)…fN(x) is called an orthogonal set over an
interval a ≤ x ≤ b if
(i) all the functions exist in that interval and
(ii) all distinct pairs of the functions are orthogonal to each other over the interval,
i.e.
a

∫ f ( x) f
b
i j ( x)dx = 0 , i = 1,2,…; j = 1,2,… and i ≠ j 4.15.2

The norm f m ( x) of the function fm(x) is defined as,

a
f m ( x) = ∫f
2
m ( x)dx 4.15.3
b

An orthogonal set of functions f1(x), f2(x)… fN(x) is called an orthonormal set if,
a
⎧0, m ≠ n
∫b f m ( x). f n ( x) = ⎨⎩1, m = n 4.15.4

An orthonormal set can be obtained from a corresponding orthogonal set of functions by

dividing each function by its norm. Now, let us consider a set of real functions f1(x),
f2(x)… fN(x) such that, for some non-negative weight function w(x) over the interval a ≤
x≤b
a

∫f
b
m ( x). f n ( x).w( x)dx = 0 , m≠ n 4.15.5

Do fi -s form an orthogonal set? We say that the fi-s form an orthogonal set with respect
to the weight function w(x) over the interval a ≤ x ≤ b by defining the norm as,

Version 2 ECE IIT, Kharagpur

 a
f m ( x) = ∫f
2
m ( x).w( x)dx . 4.15.6
b

The set of fi-s is orthonormal with respect to w(x) if the norm of each function is 1.
The above extension of the idea of orthogonal set makes perfect sense. To see this, let

g m ( x ) = w( x) f m ( x ) , where w(x) is a non-negative function. 4.15.7

It is now easy to verify that,

a a

∫
b
f m ( x). f n ( x).w( x) dx = ∫ g m ( x).g n ( x)dx = 0 .
b
4.15.8

This implies that if we have orthogonal fi-s over a ≤ x ≤ b, with respect to a non-
negative weight function w(x), then we can form an usual orthogonal set of fi –s over the
same interval a ≤ x ≤ b by using the substitution,
g m = w( x) f m ( x )

Alternatively, an orthogonal set of gi-s can be used to get an orthogonal set of fi-s
with respect to a specific non-negative weight function w(x) over a ≤ x ≤ b by the
following substitution (provided w( x) ≠ 0, a ≤ x ≤ b):
gm ( x )
fm ( x ) = . 4.15.9
w( x)

A real orthogonal set can be generated by using the concepts of Strum-Liouville

(S-L) equation. The S-L problem is a boundary value problem in the form of a second
order differential equation with boundary conditions. The differential equation is of the
following form:
d ⎡ dy ⎤
⎢ p ( x ) ⎥ + ⎡⎣ q ( x ) + λ .ω ( x ) ⎤⎦ y = 0 , for a ≤ x ≤ b; 4.15.10
dx ⎣ dx ⎦

It satisfies the following boundary conditions:

dy
i) c2 + c1 y = 0 ; at x = a;
dx
dy
ii) d 2 + d1 y = 0 ; at x = b;
dx
Here c1, c2, d1 and d2 are real constants such that at least one of c1 and c2 is non zero and
at least one of d1 and d2 is non zero.

The solution y = 0 is a trivial solution. All other solutions of the above equation
subject to specific boundary conditions are known as characteristic functions or eigen-
functions of the S-L problem. The values of the parameter ‘λ’ for the non trivial
solutions are known as characteristic values or eigen values. A very important property of
the eigen-functions is that they are orthogonal.

Version 2 ECE IIT, Kharagpur

Orthogonality Theorem:
Let the functions p(x), q(x) and ω(x) in the S-L equation (4.15.10) be real valued and
continuous in the interval a ≤ x ≤ b. Let ym(x) and yn(x) be eigen functions of the S-L
problem corresponding to distinct eigenvalues λm and λn respectively. Then, ym(x) and
yn(x) are orthogonal over a ≤ x ≤ b with respect to the weight function w(x).

Further, if p(x = a) = 0, then the boundary condition (i) may be omitted and if p(x = b) =
0, then boundary condition (ii) may be omitted from the problem. If p(x = a) = p(x = b),
then the boundary condition can be simplified as,
dy dy
y ( a ) = y ( b ) and x=a = y ( a ) = y ( b ) =
' '
x =b
dx dx

Another useful feature is that, the eigenvalues in the S-L problem, which in
general may be complex based on the forms of p(x), q(x) and w(x), are real valued when
the weight function ω(x) is positive in the interval a ≤ x ≤ b or always negative in the
interval a ≤ x ≤ b

Examples of orthogonal sets:

Ex#1: We know that, for integer ‘m’ and ‘n’,

π
⎧0, m ≠ n
∫−π cos mx.cos nxdx = ⎨⎩π , m = n E4.15.1

π
⎧0, m ≠ n
∫
−π
sin mx.sin nxdx = ⎨
⎩π , m = n
E4.15.2

π
and ∫ cos mx.sin nxdx = 0
−π
E4.15.3

Let us consider equation E4.15.1 and rewrite it as:

⎧0, m ≠ n
1/ 2 f

∫
−1/ 2 f
(cos 2π mft ).(cos 2π nft )dt = ⎨
⎩π , m = n
E4.15.4

by substituting x = 2πft = ω t and dx = 2πfdt = ω dt

Note that the functions ‘cosmx’ and ‘cosnx’ are orthogonal over the range 2π of
the independent variable x and its integral multiple, i.e. M. 2π, in general, where ‘M’ is
an integer. This implies that equation (E4.15.4) is orthogonal in terms of the independent

Version 2 ECE IIT, Kharagpur

 1 1
variable ‘t’ over the fundamental range and, in general, over M = M T0, where ‘T0’
f f
indicates the fundamental time interval over which cos2πmft and cos2πnft are orthogonal
to each other. Now ‘m’ and ‘n’ can have a minimum difference ‘1’ if
T0

∫ (cos 2π mft ).(cos 2π nft )dt = 0

−T0
E4.15.5

1
i.e., mf –nf = f =
T0
So, if two cosine signals have a frequency difference ‘f’, then we may say,
1/ 2 f
f f
∫
−1/ 2 f
cos 2π ( f c + )t.cos 2π ( f c − )t.dt = 0
2 2
E4.15.6

Re-writing equation (E4.15.6)

T0
f f 1
∫−T cos 2π ( fc + 2 )t.cos 2π ( fc − 2 )t.dt = 0 where, T0 =
f
0

Looking back at equation E4.15.5, we may write a general form for equation (E4.15.6):
T0 / 2
f f
∫
−T0 / 2
cos 2π ( f c + p )t.cos 2π ( f c − p )t.dt = 0
2 2
E 4.15.7

where mf = (n+p)f and ‘p’ is an integer.

Following similar observations on equation E4.15.2, one can say,

T0 / 2
f f
∫
−T0 / 2
sin 2π ( f c + p
2
)t.sin 2π ( f c − p )t.dt = 0
2
E4.15.8

Equation E4.15.3 may also be expressed as,

T0 / 2
f f
∫
−T0 / 2
cos 2π ( f c + p
2
)t.sin 2π ( f c − p )t.dt
2
T0 / 2
f f
= ∫
−T0 / 2
sin 2π ( f c + p
2
)t.cos 2π ( f c − p )t.dt =0
2
E4.15.9

⎛ f ⎞ ⎛ f ⎞ ⎛ f ⎞
Let us define s1 = cos 2π ⎜ f c + p ⎟ t , s2 = cos 2π ⎜ f c − p ⎟ t , s3 = sin 2π ⎜ f c + p ⎟ t
⎝ 2⎠ ⎝ 2⎠ ⎝ 2⎠
⎛ f ⎞
and s4 = sin 2π ⎜ f c − p ⎟ t . Can we use the above observations on orthogonality to
⎝ 2⎠
1
distinguish among ‘si-s’ over a decision interval of T5 = T0 = ?
f
■
Ex#2: x1(t) = 1.0 for 0 ≤ t ≤ T/2 and zero elsewhere,
x2(t) = 1.0 for T/2 ≤ t ≤ T and zero elsewhere, ■

Version 2 ECE IIT, Kharagpur

Ex#3: x1(t) = 1.0 for 0 ≤ t ≤ T/2 and x1(t) = -1.0 for T/2 < t ≤ T, while
x2(t) = -1.0 for 0 ≤ t ≤ T ■

Importance of the concepts of Orthogonality in Digital

Communications
a. In the design and selection of information bearing pulses, orthogonality over a
symbol duration may be used to advantage for deriving efficient symbol-by-
symbol demodulation scheme.
b. Performance analysis of several modulation demodulation schemes can be carried
out if the information-bearing signal waveforms are known to be orthogonal to
each other.
c. The concepts of orthogonality can be used to advantage in the design and
selection of single and multiple carriers for modulation, transmission and
reception.

Orthogonality in a complex domain

Let, z1 ( t ) = x1 ( t ) + jy1 ( t ) and z2 ( t ) = x2 ( t ) + jy2 ( t )
z1 ( t ) + z1∗ ( t ) z 2 ( t ) + z 2∗ ( t )
Now, x1 ( t ) = and x2 ( t ) =
2 2
If x1 and x2 are orthogonal to each other over a ≤ t ≤ b,
b

∫ x ( t ) .x ( t ) dt = 0
a
1 2

b
i.e., ∫ ⎣⎡ z1 ( t ) + z1∗ ( t ) ⎦⎤ ⎣⎡ z2 ( t ) + z2∗ ( t ) ⎦⎤ dt = 0
a
b

∫ ⎡⎣ z ( t ) .z (t ) + z (t ).z (t ) + z1∗ (t ).z2 (t ) + z1∗ (t ).z2∗ (t ) ⎤⎦ dt = 0

∗
or, 1 2 1 2
a

Let us consider a complex function

z1 ( t ) = x ( t ) + jy ( t ) , a ≤ t ≤ b
= r ( t ) ⎡⎣ cos Φ ( t ) + j sin Φ ( t ) ⎤⎦
where, r ( t ) = z (t ) , a non – negative function of ‘t’.

∴ x ( t ) = r ( t ) cos Φ ( t ) and y ( t ) = r ( t ) sin Φ ( t )

b b

∫ x ( t ) . y ( t ) dt = ∫ r (t ).cos Φ (t ).sin Φ (t )dt

2
Now,
a a

We know that cosθ & sinθ are orthogonal to each other over -π ≤ θ < π , i.e.,

Version 2 ECE IIT, Kharagpur

π

∫ cosθ .sin θ dθ = 0
−π

So, using a constant weight function w = r, which is non-negative, we may say

π

∫π r cos θ .sin θ dθ = 0
2

Now, x = r cos θ and y = r sin θ are also orthogonal over -π ≤ θ < π.

Now, let ‘θ’ be a continuous function of ‘t’ over -π ≤ θ < π. And,
θ t = a = θ a = −π and θ t =b = θ b = π

Assuming a linear relationship, let, θ (t ) = 2π ft

∴ dθ (t ) = 2π fdt

Under these conditions, we see,

b π
1
∫a r (t ).cos Φ(t ).sin Φ(t )dt = 2π f ∫π r (t ).cos Φ (t ).sin Φ (t ) d Φ
2 2

−
π
1
= ∫π r .cos Φ(t ).sin Φ(t )d Φ = 0
2

2π f −

1 1 T T
i.e., x(t) and y(t) are orthogonal over the interval − ≤t ≤ or ≤ t ≤
2f 2f 2 2

So, if z (t ) = x (t ) + jy (t ) represents a phasor in the complex plane rotating at a

uniform frequency of ‘f’ , then x(t) and y(t) are orthogonal to each other over the interval
T /2
1 1 T T 1
− ≤t ≤ or, equivalently − ≤ t ≤ where T = . i.e., ∫ x(t ). y (t )dt = 0
2f 2f 2 2 f −T / 2

Now, let us consider two complex functions:

z 1 (t ) = x1 (t ) + jy1 (t ) = z 1 (t ) e jΦ1 (t )
and z 2 (t ) = x2 (t ) + jy2 (t ) = z 2 (t ) e jΦ 2 (t )
T T
[x1(t),y1(t)] and [x2(t),y2(t)] are orthogonal pairs over the interval − ≤ t ≤ . So, z 1 (t )
2 2
and z 2 (t ) may be viewed as two phasors rotating with equal speed.

Now, two static phasors are orthogonal to each other if their dot or scalar product is zero,
i.e.,
A.B = A B cos γ = Ax .Bx + Ay .By = 0 , where ‘γ is the angle between A and B

In general, two complex functions z 1 (t ) and z 2 (t ) with finite energy are said to be
orthogonal to each other over an interval a ≤ t ≤ b, if

Version 2 ECE IIT, Kharagpur

b
*
∫ z (t ).z
a
1 2 (t )dt = 0

Problems
Q4.15.1) Verify whether two signals are orthogonal over one time period of the signal
with smallest frequency signal.

i) X1(t) = Cos 2πft and X2(t) = Sin 2πft

π
ii) X1(t) = Cos 2πft and X2(t) = Cos (2πft + )
3
π
iii) X1(t) = Cos 2πft and X2(t) = Cos (4πft + )
4
π
iv) X1(t) = Sin 4πft and X2(t) = - Cos (πft - )
6

Version 2 ECE IIT, Kharagpur