Scribd
Upload
156 views

Chapter 3: The Uncertainty Principle & Time-Bandwidth Product

This document discusses the uncertainty principle and time-bandwidth product. It begins by introducing the Fourier transform and how it relates signals in the time and frequency domains. It then explains that there is a limit to how well signals can be localized in both time and frequency due to their inverse relationship. This fundamental limit is called the uncertainty principle. The time-bandwidth product measures the spread of a signal across time and frequency - the lower the product, the more localized the signal is. The document proves that the minimum time-bandwidth product is 0.25, and that a Gaussian signal achieves this minimum.

Uploaded by

Deepak Kumar
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
156 views

Chapter 3: The Uncertainty Principle & Time-Bandwidth Product

This document discusses the uncertainty principle and time-bandwidth product. It begins by introducing the Fourier transform and how it relates signals in the time and frequency domains. It then explains that there is a limit to how well signals can be localized in both time and frequency due to their inverse relationship. This fundamental limit is called the uncertainty principle. The time-bandwidth product measures the spread of a signal across time and frequency - the lower the product, the more localized the signal is. The document proves that the minimum time-bandwidth product is 0.25, and that a Gaussian signal achieves this minimum.

Uploaded by

Deepak Kumar
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

Chapter 3: The Uncertainty Principle & Time-Bandwidth

Product
Vineet Kumar (01D07001), Kashinath Murmu (01D07038)

Introduction
For a long time Fourier transforms have been an invaluable tool for signal processing and
its applications. The Fourier transform is essentially required to get the information of the
signal in the frequency domain.
To get a high resolution of the frequency data of a signal, one needs to observe the signal
in time domain for a long interval. If one observes the signal in the time domain for only
a short interval, then the (frequency data) Fourier transform of the signal is more spreader
out. E.g.: In the DFT computation if one wishes to have closely spaced frequency
components i.e. high resolution of samples (n) in frequency domain, then one would
require an equal number of samples (n) in the time domain for analysis. This is equivalent
to saying that one should observe the signal for a longer interval.
However this was not a problem as long as one was restricted to analysis in a single
domain (i.e. time domain or frequency domain). But a problem arises when one starts to
analyze signals in Composite domains. Composite domain analysis is studying the given
data in 2 or more domains simultaneously. The need for such analysis arises when one
wants to extract the frequency components at different instants of time. For e.g. if one
wants to analyze the composition of the notes played by a solo instrumental performance,
then composite domain analysis is required.
Composite domains in general do not pose a difficulty in analysis. This happens only
when the 2 domains are inversely related to each other. E.g.: Time and frequency
domains have a relation (t = 1/f). In this case it is seen that if a signal has compact
support in time then its frequency equivalent does not have it, and vice versa.
What is Uncertain?
If we analyze a signal in time and frequency together, then more we zoom in into time,
the equivalent amount we zoom out in frequency, and vice versa.
1) A sine wave from
to has FT as impulse at frequency w. That is we are certain
in frequency domain but interval in time is large.
2) An impulse signal has FT as constant 1 over frequency
to . That is we are
certain in time but in frequency the spread is large.
Hence if one wishes to zoom into time and frequency simultaneously, one comes across a
fundamental limit. This limit to the highest resolution that one can achieve in the timefrequency composite domain is the uncertainty principle. This is a limit imposed by
nature.
Since compact support in any one-domain dictates that representation in the other domain
is of non-compact support, one cannot use support as measure in composite domain
analysis. It would anyway be of little use. For analysis in composite domain it is therefore
the spread or variance (i.e. where approx 98 % of the energy concentration lies) that is
measured.

Time Bandwidth Product


Time-bandwidth product is the product of the duration of a signal and its spectral width.
Generally, signals of short duration have wide spectral width and vice versa. The timebandwidth product for actual signals will vary but there will always be a minimum timebandwidth product for a certain desired effect. In communications over a channel,
transmitting a certain amount of data over a given bandwidth requires a certain time.
The time-bandwidth product measures how well we use the available bandwidth for a
given channel.
Considering the predefined formulae for variances in time and frequency of the signal
x(t) the uncertainty principle derivation leads to a lower limit of 0.25. In other words the
smallest area covered in the composite domain by the product of variance in frequency
and variance in time is 0.25.
If one performs the calculation for a triangular wave then the time-bandwidth product is
observed to be 0.3. However to reach the fundamental limit, one can work out the
expression for the optimal waveform from the derivation itself. It turns out that the
optimal waveform is the Gaussian Waveform. The time and frequency variance is 0.5
each, and hence the time-bandwidth product is = 0.25.
Uncertainty Principle:
For a finite energy function x(t), the following relation holds for its time-bandwidth
product,
2
t

1
4

I.e. 0.25 is the lower bound for the time-bandwidth product of a finite energy signal.
Making the use of famous Schwarz inequality derives this inequality.
The equality holds for a Gaussian distributed signal, whose time and frequency variances
Are 0.5 each. So a Gaussian waveform is an optimal waveform.
Proof:
Let us consider a finite energy signal x(t) with centre in time and frequency,
t0 0
0

2
t

t 2 x(t ) dt
x
2

2
w

(Time Variance)

2
2

x( ) d

x( )

(Frequency Variance)

The frequency variance can also be written as (using duality property of Fourier
transform),

d
dt

x(t ) dt
x

2
2

2
t

t 2 x(t ) dt

d
dt

x(t ) dt

4
2

f (.), g (.)

Schwarz Inequality says:

Thus, 1st integral in the numerator,


2

t 2 x(t ) dt
2

dx(t )
dt
dt

tx(t ) dt

dx(t )
dt

tx(t )

2
2

From Schwarz Inequality,


Numerator of

2
t

dx(t )
tx(t )
dt
dt

dx(t )
Re tx(t )
dt
dt

1
(z z* )
2
dx(t )
Re x(t )
dt
dt

Re( x)

Now, Re( z )
Thus,

1
dx(t )
{x(t )
2
dt

x(t )

dx(t )
}
dt

1 d
2
x(t )
2 dt

Using this in (a),


Numerator of

2
t

1 d
2
t
x(t )
2 dt

1
d
2
t
x(t )
4 dt
Integrating by parts we get,
t x(t )

x(t ) dt

Consider the integral,


2

t 2 x(t ) dt , is finite,

t 2 x(t )

0 as t

2
t

is finite

(b)

x)

-(a)

And hence,
2
t x(t )
0
Therefore, 1st part of the integral (d) goes to zero.
2
t

Numerator of

1
4

x(t ) dt

1
x
4

1
x
4

2 2
2

4
2

2
t

2
t

1
4
x 2
4
4
x 2
1
4

Optimal function (Gaussian)


When would this equality be reached, i.e.
corresponding optimal" function?

2
t

2
w

1 / 4 , and if so then what is the

dx(t )
dt
where 0 is a constant i.e. the two functions are linearly dependant on each other
dx(t ) 1
tdt
x(t )
0
From Schwartz Inequality we can see that equality is achieved when tx(t )

Integrating on both sides, we get


1 t2
ln( x)
C0
0 2
t2

~
~
x(t ) C0 .e 2 0 Where C0 eC 0
Since we want finite energy, 0 must be negative real

~
C0 Is a constant, leaving it does not matter since it just scales the function. Hence we get
the Optimal function as the Gaussian function, given by
t2

x(t )
Where

2
0

, is its standard deviation.

This document was created with Win2PDF available at http://www.daneprairie.com.


The unregistered version of Win2PDF is for evaluation or non-commercial use only.

You might also like