Representation of Signals & Systems

Reference: Chapter 2,Communication Systems, Simon Haykin

1. Hilbert Transform

 Fourier transform  frequency content of a signal

(frequency selectivity – designing frequency-selective
filters for the separation of signals on the basis of
frequency contents)

Here we use phase selectivity  uses phase shifts between

signals under consideration to achieve separation.

 Simplest phase shift is 1800  polarity reversal in the case

of a sinusoidal signal. Requires the use of an ideal
transformer in general.

 Another shift of interest is  900  the resulting function

of time is known as the Hilbert transform of the signal.

Consider a signal g(t) with Fourier transform G(f).

The Hilbert transform of g(t) is defined by


1 g ( ) 1
  t  
gˆ (t )  d = g(t)  , - convolution (1)
t

This is a linear operation. The inverse Hilbert transform is

defined by


gˆ ( )
1 1
  t  
g (t )   d = ˆ
g (t )  - (2)
t
The functions g(t) and gˆ (t ) are said to constitute a Hilbert-
transform pair.

1
In terms of Fourier transform;

1
 -j sign(f)
t

Thus Gˆ ( f )   j sgn( f )G( f )

So for a given signal g(t) we may obtain its Hilbert transform

gˆ (t ) through a two port device with transfer function -j sign(f). It
may be considered as giving -900 phase shift for all positive
frequencies and +900 for all negative frequencies.
The amplitudes of all frequency components in the signal are
unaffected. Such a device is called as a Hilbert transformer.

+900 Phase of H(f)

-900

Hilbert transform is used to:

 realize phase selectivity in SSB modulation.
 provide the mathematical basis for the representation of
band-pass signals.
The Hilbert transform as defined above applies to any signal that
is Fourier transformable. Thus it may be applied to energy
signals as well as power signals.
e.g. Sinusoidal functions

2
Consider g(t)= cos(2fct)
G(f)=1/2. [(f-fc) + (f+fc)]

Therefore Gˆ ( f ) = -j sgn(f).G(f)
= -j/2. [(f-fc) + (f+fc)]sgn(f)
= 1/2j. [(f-fc) - (f+fc)]

This is the Fourier transform of sin(2fct). So the H.T. of the

cosine function is sin(2fct). Similarly sin(2fct) has a H.T.
equal to - cos(2fct).

Properties of the Hilbert Transform

Hilbert transform differs from the Fourier transform as it
operates only in time domain. The signal is usually assumed to
be real valued.

1. A signal g(t) and its Hilbert transform have the same

amplitude spectrum. | -j sgn(f) | = 1  f.
2. If gˆ (t ) is the H.T. of g(t), then the H.T. of gˆ (t ) is –g(t).
(check !!)
3. A signal g(t) and its H.T. are orthogonal.
Here we use multiplication theorem in F.T.
 
 g (t ) gˆ (t )dt   G( f )Gˆ ( f )df
 


=j  G ( f )Gˆ ( f )df


3
 
=j  sgn( f )G ( f )G ( f )df


=j  sgn( f )G( f )G * ( f )df


 sgn( f ) G( f ) df
2
=j

where G(-f) = G*(f) for g(t) real. The integrand is odd.


Thus  g (t ) gˆ (t )dt 0 .

Similarly we may show that a power signal g(t) and its Hilbert
transform gˆ (t ) are orthogonal over one period;

1 T
2T T
lim g (t ) gˆ (t )dt 0 .
T 

(check the earlier example)

4
2. Pre-Envelope

Consider a real-valued signal g(t). We define the pre-envelope

of the signal g(t) as the complex-valued function

g+(t)= g(t) + j gˆ (t )

where gˆ (t ) is the H.T. of g(t).

We note that the given signal g(t) is the real part of the pre-
envelope g+(t) and the H.T. of the signal is the imaginary part of
the pre-envelope.

Pre-envelope makes the handling of band-pass signals and

systems easier.

The Fourier transform:

G+(f) = G(f) + j [-jsgn(f)]G(f)

Thus

2G ( f ), f  0

G+(f) =  G (0), f  0
 0, f  0


Where G(0) is the value of G(f) at frequency f=0.


Therefore g  (t )  2  G ( f ) exp( j 2ft)df
0

|G(f)|
G(0)

f
-W W

5
 |G+(f)| 2G(0)

f
W

In the above case we used a low-pass signal but pre-envelope

can be defined for any signal with a spectrum.

We may define the pre-envelope for negative frequencies as

g-(t)= g(t) - j gˆ (t )

The two pre-envelopes g+(t) and g-(t) are complex conjugates of

each other.

 0, f  0

G-(f) =  G (0), f  0
2G ( f ), f  0


6
3. Canonical Representation of Band-Pass Signals

We say that a signal g(t) is a band-pass signal if its Fourier

transform G(f) is non-negligible only in a band of frequencies of
total extent 2W, e.g, centered about some frequency  fc.

We refer to fc as carrier frequency. In most cases 2W is small

compared with fc  narrow-band signal.

Let the pre-envelope of such a signal be expressed as

g+(t)= g~(t ) exp(j2fct)

We refer to g~(t ) as the complex envelope of the signal. The

spectrum of g+(t) is limited to the band fc-W  f  fc+W.

We find that the spectrum of g~(t ) is therefore limited to the

band -W  f  W and centered at the origin.

 The complex envelope g~(t ) of a band-pass signal g(t) is a

low-pass signal.

|G(f)|

f
-fc fc
~
| G( f ) |
2W 2|G(fc)| 2W

-W W f

7
 g(t) = Re[g+(t)]=Re [ g~(t ) exp(j2fct)]

In general, g~(t ) is complex. To emphasize this we usually

express it as

g~(t ) = gI(t) + j gQ(t)

where gI(t) and gQ(t) are both real-valued low-pass functions.

The low-pass property is inherited from the complex envelope

g~(t ) .

Therefore we can express the original band-pass signal in the

canonical form as

g(t) = gI(t)cos(2fct) - gQ(t) sin(2fct)

We refer to gI(t) as the in-phase component of the band-pass

signal g(t) and to gQ(t) as the quadrature component of the
signal.

We can now give the corresponding phasor diagrams.

Also the following schemes can be given to obtain gI(t) and

gQ(t) from g(t) and vice-versa.

The multiplication of low pass in phase component gI(t) by

cos(2fct) and the multiplication of low pass quadrature
component gQ(t) by sin(2fct) represent linear forms of
modulation.

Given that the carrier frequency fc is sufficiently large, g(t) is

referred to as a pass-band signaling waveform.

Correspondingly the mapping from gI(t) and gQ(t) into g(t) is

known as passband modulation.

8
We can express g~(t ) in polar form as

g~(t ) = a(t) exp [j(t)]

where a(t) and (t) are both real-valued low-pass functions.

Based on this polar representation, the original band-pass signal

g(t) is defined by

g(t) = a(t)cos [2fct + (t)]

We refer to a(t) as the natural envelope or just envelope of

bandpass signal g(t) and to (t) as the phase of the signal.

The above equation represents a hybrid form of amplitude

modulation and angle modulation.

 The information content of the signal g(t) is completely

represented by the complex envelope g~(t ) .

9
4 Band-Pass Systems

Already know how to represent band-pass signals.

 analyze band-pass systems based on the relation between

low-pass and band-pass systems. (Hilbert transform)

Consider a narrow-band signal x(t) with Fourier transform X(f).

Assume that the X(f) is limited to W Hz of the carrier frequency

fc. Also assume that W < fc.

Let x(t) = xI(t)cos(2fct) - xQ(t) sin(2fct)

Where xI(t) in-phase component, xQ(t) quadrature component.

The complex envelope of x(t),

~
x (t ) = xI(t) + jxQ(t)

Let x(t) be applied to a LTI (linear time-invariant) band-pass

system with impulse response h(t) and transfer function
H(f).H(f) is limited to B Hz of the carrier frequency fc.

The system bandwidth 2B is either narrower or equal to the i/p

signal bandwidth 2W.

We want to express h(t) in terms of hI(t) and hQ(t) its in-phase

and quadrature components.

Thus, h(t) = hI(t)cos(2fct) - hQ(t) sin(2fct)

Define the complex impulse response of the band-pass system as

~
h (t ) = hI(t) + jhQ(t)

10
 ~
So h(t) = Re[ h (t ) exp(j2fct)]
~
Note that hI(t), hQ(t)and h (t ) are all low-pass functions limited
to –B  f  B.
~ ~
2h(t) = h (t ) exp(j2fct) + h * (t ) exp(-j2fct)
~ ~
h * (t ) is the complex conjugate of h (t ) .

Taking Fourier transform,

~ ~
2H(f) = H ( f  f c )  H * ( f  f c )

Now for a real impulse response h(t), H*(f) = H(-f).

~
As H ( f ) represents a low pass function limited to | f |  B with
B < fc we can obtain from the above relation:
~
H ( f  f c ) = 2H(f) f>0

~
So we can find H ( f ) by taking the positive frequency part of
H(f) and shifting it to origin scaled by 2.
~
Then taking the inverse Fourier transform of H ( f ) we can find
~
the complex impulse response h (t ) .

Without loss of generality we assume that X(f) and H(f) are both
centered around fc. Let y(t) be the output signal. It is also a band-
pass signal. Hence,

y (t ) exp(j2fct)]
y(t)= Re [ ~


Also y(t) =  h( ) x(t   )d


11
In terms of pre-envelopes, we have

h(t)= Re [h+(t)] x(t) = Re [x+(t)]


Therefore y(t) =  Re[ h ( )] Re[ x (t   )]d


It can be shown that


1  
         
*
Re[ h ( )] Re[ x  ( )]d Re h ( ) x  ( ) d 

2 
  

As we use x(-) instead of x() we can avoid the conjugate. Thus

1  
y(t)= Re   h ( ) x (t   )d 
2   

1  ~ 
= Re   h ( ) exp( j 2f c ) ~
x (t   ) exp( j 2f c (t   )) d 
2   


1  ~ 
= Re exp( j 2f c t )  h ( ) ~
x (t   )d 
2   

Therefore,


~ ~
2~
y (t )   h ( ) ~x (t   )d = h (t )  ~x (t )


= [ hI(t) + j hQ(t)]  [ xI(t) + j xQ(t) ]

12
If ~
y (t ) = yI(t) + j yQ(t)
Then 2yI(t) = hI(t) xI(t) – hQ(t)  xQ(t)

2yQ(t) = hQ(t) xI(t) + hI(t)  xQ(t)

Therefore, for the purpose of obtaining the in-phase and

quadrature components of the complex envelope ~ y (t ) of the
system output, we can use the low-pass equivalent model shown
below.

xI(t)
hI(t)
+
 2yI(t)
-
hQ(t)

hQ(t)
+
 2yQ(t)
+
hI(t)
xQ(t)

13