Signal & Networks Autnm 2019-20: Course coverage on signals

by Tapas Kumar Bhattacharya

Referred books (for signal part):

1. Signals and Systems by P. Ramesh Babu & R. Ananda Natarajan

Publisher; SCITECH PUBLICATIONS (INDIA), 3rd edition

2. Signals & Systems by A. V. Oppenheim, A. S. Willsky and S. H. Nawab,

Prentice-Hall India, 2015.

3. Networks and Systems by D. Roy Choudhury

New AGE International Publishers

Lecture 1 : on 15/07/2019 Monday

Given brief outline of signal portion of the course. Signals classification: Continuous
time signal x (t ) and discrete time signal. Continuous and discrete time systems;
Characterisation & properties of signals and systems. Classical ways of solving
linear constant coefficient differential and difference equations. Fourier series for
periodic signals - its importance pointed out with an example of R−L circuit excited
by a square, periodic alternating voltage source. Fourier series to Fourier transform
for non-periodic signal. Discrete time Fourier transform.

The terms used during delivering the lecture: odd & even signals, enrgy & power
signals; causal/non-causal systems, sytems with memory and without
memory(memory less), linear/non-linear system, time variant & time invariant
system, stable/unstable system, invertible/not invertible system

Defined an unit impulse signal δ (t ) .

Lectures 2 & 3 : on 17/07/2019 Wednesday

Time shifting [ to get x (t−a) from x (t ) ] , time scaling [ to get x (at) from
x (t ) ] and folding operation [to get x (−t) from x (t ) ] on signal x (t ) were
discussed. In general when all the three operations is carried out on a given signal
x (t ) i.e., to get x (−at−b) from given x (t ) it is suggested that (i) first do
shifting, (ii) then scaling followed finally by (iii) folding operation so as to get the
result quickly and correctly.
Even & odd signals defined. It was shown that any signal x (t ) can be wxpressed as
[ x (t)+ x (−t )] [x (t)−x(−t)]
sum of x even(t)+ x odd (t) where x even(t)= and x odd (t)=
2 2
Unit step function u(t) and unit ramp r (t) signals were defined and sketched.
Use of these functions to describe a class of signals x (t ) (involving rectangles and
lines), in a neat way explained. A couple of examples were solved with students
d r ( t)
partcipation. It was also shown that u(t)= .
dt

The unit impulse function δ (t ) , its properties were explained in detail. For this
purpose we considered a rectangular pulse around the origin of width h and height
1
and allowed h→0 to get what is known as δ (t ) . under this condition area
h
under δ (t ) is unity.
The functional value of δ (t ) is zero if t ≠0 and at t =0 , δ (t ) races to an
+
∞ 0

infinitely large value but keeping this integral ∫ δ (t) dt=∫ δ (t) dt 1=1 finite.
−∞ -
0

The multiplication of a signal x (t ) with an impulse function means that

+
∞ 0

x (t) δ( t)=x (0)δ (t) . Which means that , ∫ x (t) δ(t )dt =∫ x (0) δ(t )dt =x (0) .
−∞ -
0

Similarly when x (t ) is multiplied with a shifted delta function δ (t−a) and

+
∞ a

integrated from t = -∞ to t = +∞ , ∫ x (t) δ(t−a)dt=∫ x (a)δ(t )dt=x (a) .

−∞ -
a

t t
It was also shown that ∫ δ (t) dt=0 for t< 0 and ∫ δ (t) dt=1 for t > 0 .
−∞ −∞
t
d u(t)
Therefore we conclude that ∫ δ (t) dt=u(t) for any t. This implies that δ (t )=
−∞ dt
.

Lecture 4 : on 18/07/2019 Thursday

If a signal x(t) has vertical sharp rise or fall at some ponints of time, then dx at
dt
those discontinuity ponts can be shown as impulses of appropriate strength. Time
scaling an impulse function results into an impulse of different strength and it was
δ(t)
shown that δ (at)= where a could be either +ve or -ve. Conditions for a
|a|
system to be linear and time invariant discussed.
 If the impulse response of a linear, time invariant system h(t ) is known , the
output of the same system for any input signal x (t ) , can be found out.

Recall for a linear system, x (t)→ y (t) where x(t) is input and y(t) is output of the
system, then following are true:

a x (t)→a y (t ) where scaling up factor.

a1 x 1(t)+a2 x 2(t )→a1 y 1(t)+a2 y 2 (t) This is scaling and superposition

Also if the system is time-invariant:

x (t)→ y (t) then x (t−a)→ y (t−a)

Using these facts about linear, time-invariat system, it was shown that

if δ(t )→h (t) then

∞
x (t)→ ∫ x (τ)h(t −τ)d τ= y (t)
−∞

This particular integral is called convolution operation on two signals x(t) and h(t)
and in short hand is written as:
∞
x (t)∗h(t)= ∫ x( τ)h(t−τ)d τ
−∞

It can be shown that the order of appearance of x and h inside

the integral can be interchanged i.e.,
∞ ∞
x (t)∗h(t)= ∫ x (τ)h(t−τ)d τ= ∫ h (τ) x (t−τ)d τ=h(t)∗x (t)
−∞ −∞
Convolution is commutative.

Lecture 5 : on 22/07/2019 Monday

Convolution between two signals is commutative, distributive and associative –

applied this knowledge to find out the overall impulse response of cascaded and
parallel LTI systems. Remember convolution can be applied to LTI system.
Discussed about memory less / memory system. Impulse response of a memory less
system must be an impulse i.e., h(t )=k δ ( t) .
Discussed about the causal system in general and how to test for it. For LTI system, it
was shown that for the system to be causal h(t )=0 for t < 0 . You need not always
try to find out h(t ) then decide about memory less system or causal system. If input
out relationship is known, we can easily decide about the system types as well. To
continue.

Lecture 6 & 7 : on 24/07/2019 Wednesday

Causality in linear system alternatively means , zero output for zero input which also
happens to be the homogeneity condition of the linear system. Discussed how to test
for time invariancy of a given system in general. Also discussed about invertible
system. Two linear systems will be invertibe with each other if the impulse response
of the overall system happens to be an impulse i.e., h1 (t)∗h2 (t )=δ (t ) where h1 (t)
and h2 (t) are the individual responses of the two systems.
Two intersting properties of convolution explained. If y (t)=x(t)∗h (t) then
(i) Area under y(t) will be the product of areas under x(t) and h(t) calculated
separately.
(ii) If x (t ) exists in the range a x < t<b x and h(t) exists in the range ah <t <b h ,then
support value of x (t ) is said to be (b x −a x ) and support value of h(t) = (bh −ah ) .
The range over which y (t) will be supported is (a x + ah )<t<(b x + bh ). Hence support
value y(t) will be [(b x + bh )−(a x + ah )] .

Started the topic solution of linear, constant coefficient differential equation

classically. For a first order equation : dy +a y =x (t) , it was shown that the solution
dt
will have atwo parts (i) natural response y n (t ) which does not depend on the
forcing function x(t) and (ii) solution y f (t) due to forcing function x (t ) . It was
further shown that y n (t)=C e−at where −a is the root of the characteristic equation.
The solution due to forcing function was shown to be the linear combination of
x (t ) and its higher order derivatives i.e.,
y f (t)=k 1 x (t)+k 2 ẋ +k 3 ẍ+ k 4 ⃛x + . . up to infinity
So the total solution : y (t)=C e−at + k 1 x (t)+k 2 ẋ +k 3 ẍ+ k 4 ⃛x + . .
Exception: if forcing function x (t )= A e−at then y f (t)= At e−at and total solution
will be of the form y (t)=( At +C) e−at .
Aloso it was pointed out that x (t ) excludes δ (t ) .
Discussion will be continued.

Lecture 8 : on 25/07/2019 Thursday

Reviewed last lecture and how to findout solution for higher order differential
equation.

It is interesting to note that if the input signal x (t) happens to be an exponentioal

function then the solution due to the forcing function y f (t) can be very easily
obtained. In fact the solution due to the exponential x (t ) will be once again
exponential itself as any order of derivative on an exponential function retains the
same exponential form.

Consider the following differential equation

d2 y dy
2
+a + b y= x(t ) and x(t )= A e j s t , where s is a number (real or even
dt dt
complex)

i.e., (D 2 +a D+b) y f (t)= A e s t

It was shown that, response y f (t) will be:

A es t
For y f (t)= 2
(D +a D+b)⋮D=s

Also note that for constant (dc) input x (t)= A= A e 0 t and

A e0 t A
y f (t)= 2
=
( D + a D+ b)⋮D=0 b

How this can help me? Consider x (t ) to be sinusoidally varying excitation, say
sin ω t . This type of functions can be written in exponential forms as follows

1 ( jω t ) (− j ω t)
sin ω t= [e –e ]
2j
 1
cos ω t= [e( j ω t)+e (− j ω t)]
2

Therefore y f (t) can be easily obtained by doing a little bit of algebra on the two
exponential terms.

The solution of linear differential equation with forcing function as an impulse as

δ (t ) has no finite functional value at t =0 .

How to solve differential equation when the excitation or input x(t )=δ(t) ?

1. Consider a first order differential equation with forcing function δ(t ) .

dy
+2 y=δ (t) with the boundary condition y (0- )=1
dt

Looking at the equation we know, to balance the equation LHS must produce an
impulse to balnce the equation. This is possible if y jumps from y (0 -)=1 to
dy
y (0+ ) , then can be an impulse and everything will be in place.
dt

Now multiply both sides of the equation by dt and integrate from 0- to 0+ to

get:

+ + +
0 0 0

∫ dy+2∫ y dt=∫ δ(t )dt

- - -
0 0 0

or y (0 + )=1+1=2

Since δ(t )=0 for t> 0+ , the problem then boils down to solving the following
differential equation with new B.C as shown below.

dy
+2 y=0 With B.C y (0 + )=2
dt

Characteristic equation m+ 2=0 gives m=−2 hence solution is :

y (t)=C e−2t
From B.C we get C=2 . Hence the solution is y (t)=2 e−2 t .

What about 2nd or higher order differential equations? Consider the following
equation.

d2 y dy
2
+a +b y=δ(t)
dt dt

dy d2 y
We immediately know that must have a step jump so that
gives the
dt dt 2
necessary impulse to balance the right hand side. So conclusion is y(t) will be
dy
continuous and will have a step jump; whch means:
dt

dy + dy -
y (0 -)= y (0+ ) and ( 0 )≠ (0 ) .
dt dt

d n−1 y
For n th order differential equation will have a step jump and lower order
dt n−1
differentiation terms will be continuous.

Why impulse response of a system (described by a linear differential equation) is

important to know? We shall show later that if the impulse response of a system is
known, we can find out the response of the system for any arbitrary input signal
x (t) .

Lecture 9 : on 29/07/2019 Monday

Reviewed last lecture and took a deeper look to inductor and capacitor. Both these
elements have memory.
Inductor L :
relationship between voltage across the inductor v (t ) and current flowing through it
i(t) are related by:
t
di 1
v =L or, i(t)= ∫ v (t )dt
dt L −∞

Therefore to inject current through an inductor you have to apply voltage across it
and present value of the current i(t) , depends on all past values of the voltage applied
including the current value of the voltage. We can break up the RHS of the above
equation into three pieces:
 - + +
(0 ) (0 ) t (0 ) t
1 1 1 1 1
i(t)= ∫ v (t )dt + ∫ v (t )dt + ∫ v (t )dt =i( 0- )+ ∫ v (t)dt + ∫ v (t) dt
L −∞ L (0 ) L (0 ) - L (0 ) L (0 )+ - +

The first term i (0- ) is nothing but the consequence of the past applied voltage till
time t =0-

At t =0 , new voltage is applied. If the new applied voltage excludes impulse, the
+
(0 )
1
second term ∫ v (t) dt must be zero. Therefore, current at any time t will be
L (0 ) -

given by
t
1
i(t)=i(0 )+ ∫ v (t) dt and
-
L (0 ) +

+
0
1
i(0 )=i( 0 )+ ∫ v (t )dt=i(0- )
+ -
if v (t ) excludes impulse.
L (0 ) -

+
0
1 1
i(0 )=i( 0 )+ ∫ δ (t) dt=i(0 -)+
+ -
And if v (t )=δ (t) .
L (0 ) L +

Capacitor C :

To build up voltage across a capacitor one has to pass current for accumulation of
charges, hence voltage.

relationship between voltage across the capacitor v (t ) and current flowing through
it i(t) are related by:
t
dv 1
i=C or, v (t )= ∫ i(t)dt
dt C −∞

Therefore to have voltage across a capacitor pass current through it it and present
value of the voltage v (t ) , depends on all past values of the current passed including
the present value of the current. We can break up the RHS of the above equation into
three pieces:

- + +
(0 ) (0 ) t (0 ) t
1 1 1 1 1
v (t )= ∫ i(t) dt + ∫ i(t )dt+ ∫ i (t) dt=v (0- )+ ∫ i(t )dt + ∫ i(t )dt
C −∞ C (0 ) C (0 ) -C (0 ) C (0 ) + - +
The first term v (0- ) is nothing but the consequence of the past currents till time
t =0-

At t =0 , new current flows. If the new current excludes impulse, the second term
+
(0 )
1
∫ i(t) dt must be zero. Therefore, voltage at any time t will be given by
C (0 )
-

t
1
v (t )=v (0- )+ ∫ i(t )dt and
C (0 )+

+
0
1
v (0 )=v (0 )+ ∫ i(t )dt=v (0- )
+ -
if i(t) excludes impulse.
C (0 ) -

+
0
1 1
v (0 )=v (0 )+ ∫ δ (t) dt=v ( 0- )+
+ -
And if i(t)= δ (t) .
C (0 ) - C

It was also explained that an inductor with zero initial current i(0- )=0 and a
capacitor with zero initial voltage v (0- )=0 can be treated as linear elements because
superposition hold. However, with i(0- )≠0 and v (0- )≠0 and inductor and
capacitor no longer remain linear elements.
It was shown that an inductor with initially charged current of i (0- ) can be
equivalently represented as an uncharged inductor in parallel with a current source of
value i(0- ) .
Similarly it was shown that an capacitor with initially charged voltage of v (0- ) can
be equivalently represented as an uncharged capacitor in series with a voltage source
of value v (0- ) .

Thus to solve a circuit involving energy storage elements, we can do it in two ways.
1. Write down the differntial equation and get the solution, then find out the constants
applying boundary conditions at t=0 .

2. Redraw the circuit showing the initial conditions as separate sources as described
in the last section so that so that energy storing elements can be thought of having
initial conditions relaxed.

Lecture 10 & 11 : on 31/07/2019 Wednesday

Briefly reviewed the last lecture.
Discussed how to solve for capacitor voltage v (t ) and circuit current i(t) , in a
relaxed series R−L−C circuit when excited by a step voltage Vu(t ) . From KVL
equation we wrote the following differential equation:
 d2 v dv dv
LC 2
+ R C +v=V noting that i=C with initial conditions v (0)=0 and
dt dt dt

i(0)=0 .

d 2 v R dv 1 1
Dividing both sides by LC , 2
+ + v= V
dt L dt LC LC

R 1 R C 1
Assumed
L
=2 ζ ωn and
LC
=ω2n which means ζ =
2 L
and ωn =
√
√ LC
.

We then dealt with the same differnential equation with changed values of coefficients as follows:

d2 v dv
2
+2 ζ ωn +ω2n v=ω2n V
dt dt

The characteristic roots are −ζ ωn±√ ζ 2−1 ωn and the roots will be complex
conjugates if 0< ζ <1 . Then roots are −ζ ωn± j √ 1− ζ2 ωn=−ζ ωn± j ω d , where
ωd =√ 1− ζ2 ωn . Here we are considering the case when the roots are complex
conjigate. After solving the final result will be:
2
V −jω t −1 ωd V − jω t −1 √ 1− ζ
v (t )=V – e n
sin( ω d t+ tan )=V – e sin
n
( ω d t +tan )
√1−ζ 2 ζ ωn √ 1−ζ 2 ζ
Expression for i(t) will be obtained from:
dv
i(t)=C .
dt
It was explained response of such a system will be oscillatory with over shoots and
undershoots and finally settling to the steady value. The factor ζ is called damping
coefficients. Lower the value of ζ , more oscillatory the system will become with
higher values of over shoots and undershoots . %omega_d is called the damped
angular frequency with which the oscillation will continue. ωn is called the natural
angular frequency with which the oscillation will continue for ever if ζ =0 which
means resistance R=0 .
Student were asked to solve a numerical problem and sketch variation of v (t )
against time for a system having complex conjugate roots.

Students were also asked to find out the response of the same circuit when the
(i) roots are real and distinct and roots are real and equal and (iii) when rotts are
purely imaginary.

Lecture 12 : on 01/08/2019 Thursday

An alternative way to obtain impulse response h(t ) :
If two LTI systems S1 & S2 are connected in cascade, the over all output and input
relation will not be affected then position of S1 and S2 can be interchanged. Using
this idea it was shown that, if the unit step response of a system is known to be s (t )
d s( t)
, then the unit impulse response of the system will h(t )= . Students were
dt
asked to verify this for an R-L series circuit.

New topic on Fourier series started.

Fourier series introduced. If x (t ) is periodic with a fundamental time period T and
if x(t) satisfies Dirichlet’s conditions, then x(t) can be represented as sum of some
constant value and infinite number of sinusoidal terms of different frequencies k ω ,
2π
where ω= is called the fundamental frequency.
T
∞ ∞
x(t )=c o + ∑ ak cos k ω t +∑ b k sin k ω t
k=1 k=1

The values of the constants were derived to be :

T /2 T
1 1
c o = ∫ x(t )dt= ∫ x(t )
T −T /2 T 0
T/2 T
2 2
a k = ∫ x(t )cos k ω t dt= ∫ x(t)cos k ω t dt
T −T /2 T 0

T /2 T
2 2
b k = ∫ x(t )sin k ω t dt = ∫ x (t )sin k ω t dt
T −T /2 T 0

It may be noted, all the above integrations are to be carried out over a time period
T . The results of the constants is independent of the lower and upper limits of the
integration so long the difference of the upper and the lower limit is the time period
T .

Lecture 13 : on 05/08/2019 Monday

Half wave and Quarter wave symmetry of periodic signal discussed.

For a periodic signal x (t ) if x (t)=−x (t± T ) then x(t) is said to have half wave
2
symmetry. The implication of this are: C o=0 (i.e., no dc value) and only the odd
harmonics will be present. Sine & cosine terms may be present for k = odd integers.
A periodic signal x (t ) is said to be quarter wave symmetric when x (t ) is either
even or odd signal and is also having half wave symmetry. The implication of this is
that for even x(t) , C o=0 with only cosine terms (a k ≠0) k = odd
and for odd x(t) , C o=0 with only sine terms (b k ≠0) k = odd

From a mere visual look at the periodic signal , it is sometimes very easy to figure out
whether the signal is half wave symmetric or quarter wave symmetric. It is needless
to say that if x(t) is neither even nor odd and not has half wave symmetry, then all the
terms and both odd and even harmonics may be present.

Alternative way of writing the Fourier series

∞ ∞ ∞
x(t )=c o + ∑ a k cos k ω t +∑ b k sin k ω t =c o +∑ C k sin (k ω t+θk )
k=1 k=1 k=1

ak
where C k =√ a2k +b 2k and θk =tan −1( )
bk

The plot of |C k| versus k ω is called line spectrum and tells us about the strength of
each harmonic present in the signal. Similarly θk can be plotted versus k ω to get
the phase information of the signal at a particular harmonic.

We have seen that for periodic x (t ) ,

∞ ∞
x(t )=c o + ∑ a k cos k ω t +∑ b k sin k ω t
k=1 k =1

The RHS of the above equation was manipulated and x(t) was written in a much more
compact way as below.
∞
j kω t
x(t )= ∑ Ck e This is called Fourier series in complex form.
k=−∞

where,

T /2
1
C k = ∫ x(t)e− jk ω t dt
T −T /2
 (a k − jb k ) (ak + jb k ) *
Also to note that : C k = and C−k = =Ck
2 2

Students were asked to find out the Fourier series coefficients of periodic impulses
having time period T .
It is suggested to represent a periodic function in terms of complex Fourier series.
After all, if we know C k then a k and bk can be easily obtained by equating real
and imaginary part.

Lecture 14 & 15 : on 07/08/2019 Wednesday

Recall that the plot of |C k| versus k ω is called lineor frequency spectrum and tells
us about the strength of each harmonic present in the signal. Similarly θk can be
plotted versus k ω to get the phase information of the signal at a particular
harmonic. While |C k| versus k ω will be an even function , θk versus k ω plot
will be odd. It is needles to say these plots will be discrete in nature.

Average power of a periodic signal in terms of Fourier coefficients was shown to be:
∞ ∞
Pav = ∑ C k C *k = ∑ |C k|2
k=−∞ k=−∞

It was shown that if x (t ) is an even periodic function, C k will have only real part
and if x (t ) is an odd periodic function, C k will have only imaginary part.
Fourier coefficients C k (hence a k and bk ) were found out for a even square
periodic function and an odd square periodic function with active participation of the
students.

dx
If x (t ) is periodic then and higher order derivatives too will be periodic of
dt
same fundamental period of x (t ) . If C k is Fourier coefficients of x (t ) , then it
dx
can be easily shown that Fourier coefficients of will be jk ω C k .
dt
Therefore for a halfwave symmetric square, the Fourier series coefficients ( which
was solved earlier), can obtained by this alternative means. First differentiate the
original square wave which results into a periodic impulse function – get its Fourier
coefficients C*k . The integration to be carried out will only involve delta or
impulse functions . Then get C k by dividing C /k with jk ω . The bottom line is ,
to get computational advantage, it is better you reduce the given periodic x (t ) with
succesive differentiation till you get periodic funtion described by impulses only -
find out its coefficients which will be much easier. Finally do the approprite number
of times divisions with jk ω to get Fourier coefficients of the original x (t ) .

Students applied this method to periodic square wave and to perodic triangular wave
to get coefficients correctly and efficiently

Lecture 16 : on 08/08/2019 Thursday

Briefly reviewed the last lecture.
The right hand side of the Fourier series will have various sine, cosine terms of
different frequencies along with perhaps a d.c term. RHS therefore is sum continuous
functions. If x (t ) has discontinuity at some time (as in square wave), it will never
be possible to reconsruct the the discontinuity from the RHS. If we calculate the
RHS at time (t= τ ) where discontinuity occurs, we will always get the average
value of x ( τ(-) ) and x ( τ(+) ) independent of the number of terms (N) we have
turncated the RHS. Also around the point of discontinuity there will be an overshoot
and under shoot present which can not be supressed even by considering higher and
higher number of terms. This is known as Gibb’s phenomenon.

General practise is to consider a finite number of terms of Fourier series. More the
number of terms you consider. less will be the error in matching LHS i.e., x (t ) .
Considering the physical system user will choose the value of N . Since in the
truncated F.S we use the same coefficients C k of the infinite terms F.S, the question
is, can there be a different choice Dk instead of C k which may result minimum
error in least sqare sense.
∞
j kω t
Let, x(t )= ∑ Ck e , the original F.S with no terms neglected.
k=−∞

The RHS is truncated to N terms and we approximate the given function as

∞
j k ωt
x N (t )= ∑ Dk e thinking that D_k will be a better choice instead of C_k in
k=− N
terms of sum error squared.

Therefore error

∞ N N N
j kω t jk ωt j k ωt j kω t
e(t )= ∑ C k e − ∑ Dk e = ∑ Cke + ∑ Ck e − ∑ Dk e j k ωt
k=−∞ k=− N |k|> N k=−N k=− N
or,
N
jk ω t
e(t )= ∑ Ck e + ∑ (Ck −D k )e j k ω t
|k|>N k=− N

This error signal is of course periodic. Therefore sum of the squared error will be
∞ N
2 2
∫|e( t)| = ∑ |Ck| + ∑ |(C k−Dk )|2
k=−∞ k=−N

Since the first term on RHS is constant, error will be minimum if Dk =C k .

So In the truncated F.S best choice of coefficients still will be C_k .

Lecture 17 & 18 : on 14/08/2019 Wednesday

Briefly reviewed the last lecture.
New topic.
Introduced Fourier Transform FT which give us information about frequency content
of a signal which is not periodic.
In such cases, it was shown that Fourier coefficient Density per Hz, X (ω) will
be meaningful. Starting from Fourier series concepts, following two important
results were derived.

Fourier Transform:
∞
− j ωt
X (ω)= ∫ x (t)e dt called the expansion equation.
t =−∞

∞
1
x (t)= ∫ X (ω)e j ω t d ω called the synthesis equation.
2 π ω=−∞

x (t ) and X (ω) form a Fourier transform pair.. If x (t ) is known we can get X (ω) from
the first equation and X (ω) is known, second equation can be used to get .
∞
− j ωt
Fourier transform of x(t) is: F {x (t)}= X (ω)= ∫ x (t )e dt
t =−∞

∞
1
Fourier inverse of X (ω) is : F {X (ω)}=x (t)=
−1
∫ X (ω)e j ωt d ω
2 π ω=−∞
Tansform pair is indicated by:
x (t)⇔ X (ω)

Two interesting observations:

∞
Area under the curve x (t ) is A time domain = ∫ x (t)dt =X (0) (put ω=0 ) in
t =−∞
first equation). In other words, area under the time domain curve is nothing but the
value of X (ω) at ω=0 .

Similarly by putting t =0 on both sides of the second equation, we get,

∞
1 1
x (0)= ∫ X (ω)d ω= × A frequency domain
2 π ω=−∞ 2π

or Area under the frequency domain curve (Fourier transform curve) will be
∞
A frequency domain =2 π x (0)= ∫ X (ω) d ω
ω=−∞

It was shown that if x (t ) is a real even function, X (ω) too will also be even
having real part only. Similarly, if x(t) is a real odd function, X (ω) too will be also
odd having imaginary part part only.

Fourier transform of single rectangular pulse of amplitude A , width d and

centered around t =0 is obtained as:

sin( ω d /2)
X (ω)=A d = A d sinc(ω d /2) , As expected this is an even function of ω
(ω d /2)
having no imaginary part.

Using the ideas of calculating area under x (t ) or X (ω) , it was shown that the
following integral has a value equal to π
+∞
sin θ
∫ d θ=π i.e., are under the sinc curve is π .
−∞ θ

This result to be used later.

We then tried to find out the FT of some standard non-periodic signals

Fourier transform of x (t)=δ (t) :
∞ ∞
− j ωt
F {δ(t )}= ∫ δ (t)e dt = ∫ δ(t )dt=1
t=−∞ t =−∞

Fourier transform of x (t)=1 :

To find this out quickly we use the inverse formula, i.e.,

∞
1
x (t)= ∫ X (ω)e j ω t d ω Now, x (t)=1
2 π ω=−∞
∞
1
1= ∫ X (ω)e j ωt d ω
2 π ω=−∞

Can we guess X (ω) so that integration of RHS will return us 1? The answer is yes
and gussing X (ω)=2 π δ(ω) will indeed make RHS to be 1.

Therefore, 1⇔ 2 π δ(ω) .

Some properties of Fourier Tranasform were explained.

If x(t )⇔ X (ω) then

1. Time shifting property

x(t −a)⇔ e(− j ω a) X (ω) shifting in time causes a modulation in frequency

domain.

jω t
2. x(t )e c ⇔ X (ω−ω c ) modulation in time domain causes shifting in
frequency domain.

dX
3. t x (t )⇔ j
dω
 d x (t)
4. ⇔ j ω X (ω )
dt

Fourier transform of unit step function was shown to be

1
F {u(t )}= +π δ(ω) the result was obtained by decomposing u(t) in its even
jω
and odd parts and by using results obtained and property got earlier.

Several other interesting properties of FT will discussed in the next class.

Lecture 19 : on 19/08/2019 Monday

Briefly reviewed the last lectures. Duality property established and its use
highlighted.
The differentiation property and FT of δ (t−a) was used to find out the FT of the
square pulse of duration d and of amplitude A , the result was the same sinc
function . Obtained the the FT of a triangular pulse using the same properties –
computational effort needed becomes less.

x(t )⇔ X (ω) , then it was shown that X (−ω)=X * (ω) .

Then it was shown that if x (t ) is an even real function X ( ω) too will be real and
even function.
Students were asked show that if x (t ) is an odd real function X ( ω) too will be
odd and will be purely imaginary.
+∞ +∞
It was shown earlier that ∫ θ d θ=π . It was further shown that ∫ sinθa θ d θ=π
sin θ
−∞ −∞
where a ia +ve real number. Then it was shown that
+∞
sin a θ
as a→∞ , then the integral ∫ θ d θ approaches a delta function of strength
−∞
π i.e., π δ ( θ ) .

Lecture 20 & 21 : on 21/08/2019 Wednesday

Earlier we guessed the the FT of x (t)=1 from the inverse formula
∞
1
1= ∫ X (ω)e j ωt d ω and decided by intuition that X (ω)=2 π δ(ω) .
2 π ω=−∞
∞
− jωt
Now we used the direct formula to evaluate F {1 }= ∫ 1e dt . The integral
t=−∞

on the RHS indeed was shown to be 2 π δ(ω) based on the fact that
 +∞
sin t ω
as t →∞ , then the integral ∫ ω d ω approaches a delta function of strength
−∞
π i.e., π δ ( ω ) .

Fourier transform of x (−t) and x (at ) were found out.

Energy of a signal was shown to be

∞ 2 1 ∞ 2
∫−∞ |x (t )| dt = 2 π ∫−∞ |X (ω )| d ω

It was pointed out that the sufficient condition for FT to exist , the signal x (t )
∞
should be an energy signal i.e., W =∫−∞ |x (t )|2 dt must be finite. Although it is not a
necessary condition.
1
Students were asked to show that F {e−at u(t)}= if a> 0 . They also
j ω+a
calculated energy of the signal both in time and frequency domain.

If F {x1 (t )}=X 1 (ω) and F {x2 (t )}=X 2 (ω) the it was shown that

F {x1 (t )∗x 2 (t)}=X 1 (ω) X 2(ω)

This result is interesting as well as important which states that FT of convolution of
signals in time domain means multiplication in frequency domain. Students were
asked to to find out FT of product of two signals x 1( t)x 2 (t) in time domain.

Then it was shown that

t X (ω) X (ω)
F {∫−∞ x (τ)d τ }= +π X (ω)δ (ω)= +π X (0)δ (ω)
jω jω

FT method was applied to R_L series circuit to solve for the differential in frequency
domain. The time domain differential equation gets converted to algebraic equation.
It was shown that that FT of i(t) is given by

X ( ω)
I ( ω )=
(R+ j ω L)

Finally to obtain i(t) one has to take inverse transformation of I(%omega) i.e.,
−1
i(t )=I (ω)
The circuit was solved when x (t)=δ (t) and when x (t)=u (t) .

Lecture 22 : on 22/08/2019 Thursday

FT of cosine or sine functions of time: If x (t)=cos ωc t it was shown that its FT was
shown to be
X ( ω)=π δ ( ω−ω c )+ π δ ( ω + ω c )
The result is not surprising, after all x(t) has a single frequency of ωc , so in
frequency domain we expect impulses at ωc . Knowing this we solved R-L series
circuit problem when x (t)=cos ωc t .in frequency domain i.e.,

π δ ( ω −ω c )+ π δ ( ω + ωc )
I ( ω )=
( R+ j ω L)

Taking inverse transform we got

1 ωc L
i(t)= cos ( ωc t−θ ) where tan θ =
2
√ R +ω c
2
L 2 R

Point to be noted is that the steady state current was obtained since it looks like, the
circuit was switched on long time before and if any transient was born at that time
must have died down also long before.
Students were asked to find out FT of x (t)={cos ωc t }u (t) & sketch its magnitude
spectra and finally solve the same R-L circuit problem when x (t)={cos ωc t }u (t) . In
this case it means, the voltage is applied at t =0 , so i(t) is expected to have both
steady state part and the transient part.

Two problems on FT were solved with student’s participation applying intelligently

the some properties of FT.
Lecture 23 : on 26/08/2019 Monday

FT of e−at where a> 0 , was shown to be of same nature.

Fourier Transform to Laplace Transform

We know that
∞
−jω t
F {x (t)}=X (ω)= ∫ x (t )e dt
t =−∞

Many function may not have FT when it is not a energy

signal (or fails to satisfy Perseval’condition) as
mentioned earlier. Now such a function x (t ) if
multiplied by an exponentially decaying function e−σ t
then the FT of x (t )e−σ t will exist. Where σ is a chosen
constant suitable constant to make x (t )e−σ t exponentially
decaying or energy signal.

So,
∞ ∞
−σ t −σ t − j ωt −(σ + j ω)t
F {x (t) e }= ∫ x (t)e e dt = ∫ x (t)e dt
t =−∞ t =−∞

Defining s=σ + j ω
∞ ∞
−σ t −σt − j ωt −st
F {x (t) e }= ∫ x (t)e e dt = ∫ x (t)e dt
t =−∞ t =−∞

Obviously, the integral on the RHS will be a function of s=σ + j ω t

. and we say,
∞
−st
∫ x (t )e dt =X (s) to be the Laplace transform of x (t ) .
t =−∞

Note with σ=0 , X (s)= X (w) will be true for energy functions.

We shall here discuss about one sided Laplace transform.

What is it?
We shall always put a restriction in x (t ) as follows:
x (t )=0 for t< 0 and x (t ) exist for t > 0 or in compact
form the signal is written as x (t )u(t) . This type of
signal is called causal signal. So for a causal signal
x(t), its Laplace transform will be:
∞
−s t
X (s)= ∫ x (t )e dt and this is called one sided Laplace
t=0
transform.

To accommodate delta ( impulse function) δ(t ) , the lower

limit should be changed to 0−¿ with out any loss of
generality. So,
∞ ∞
−s t −st
X (s)= ∫ x (t)e dt= ∫ x(t)e dt
-
t=0 t=0

It may be noted that the value of σ (for one sided L.T) must a
positive number and greater than some critical value of σ c so
−σ t
that x (t)e
becomes a decaying function and LT integral
converges. Fortunately σ will be behind the seen, we need
not bother about σ while finding LT of a given function. The
condition σ> σc is called the region of convergence (ROC).

The inverse formula : Knowing X (s) , how to get x (t ) ?

−σ t
Recall, Fourier transform of x (t)e } is given by:
∞
−σ t −(σ + j ω )t
F {x (t) e }= ∫ x (t)e dt
t =−∞

Therefore,
∞
−σ t 1
x (t)e = ∫ X (σ + jω t)e j ω t d ω
2 π t =−∞
Multiplying both sides by e^{%sigma t}
 ∞
1
or, x (t)= ∫ X (σ+ j ω t)e(σ + j ω)t d ω
2 π t =−∞
Now put, σ+ j ω=s

Thus, j d ω=ds and limits of integration will be σ− j ∞ to

σ+ j∞

σ+ j ∞
1 st −1
x (t)= ∫
2 π j t =σ − j ∞
X (s)e ds=L {X (s)} is Laplace inverse of X(s).

Although x(t) can be obtained by evaluating the above integral,

general practise is to use Laplace transform table where

x (t )⇔ X (s) are available for most of the useful x (t ) used in

linear time invariant system analysis. It was shown that:

1. L {δ(t)}=1 which also means L−1 {1 }=δ(t)

1 1
2. L {u(t )}= and L−1 { }=u(t )
s s
1 1
3. L {e at u(t )}= and L−1 =eat u(t )
s−a s−a

There are many more.

Students are advised to consult some standard text book where
where LT and the corresponding inverse are tabulated for more
functions. The above functions are most commonly used in circuit
analysis and may be memorised.
Lecture 24 & 25 : on 28/08/2019 Wednessday

Laplace transforms of more useful functions obtained such as:

s s
L {cos ω t u(t )}= L−1 { }=cos ω t u(t )
s2 +ω2 2
s +ω 2

L {sin ω t u(t )}= ω L−1 { 2 ω 2 }=sin ω t u(t )

s +ω2
2
s +ω
It was also shown if x(t )⇔ X (s) then,

then 1. L{ x(t−a)}=e−as X (s)

2. If x(t )⇔ X (s)
then L{e−at x (t )}= X (s +a)

3. If x(t )⇔ X (s)
−dX (s)
then L{t x (t)}=
ds

1
Also L{t u(t )}=L {r ( t)}=
s2

Derivative property:

If x(t )⇔ X (s)
dx
then L{ ]=s X (s)−x(0)
dt

d2 x dx
Also L{ ]=s 2 X (s )−s x(0)− ẋ (0) where ẋ means
dt 2
dt

d3 x
Also L{ 3 ]=s 3 X ( s)−s2 x(0)−s ẋ(0)− ẍ (0) and so on.
dt

If x (0) and all its derivatives are 0, then

 dx d2 x d3 x
L{ ]=s X (s) , L{ 2 ]=s 2 X (s ) , L{ 3 ]=s 3 X ( s)
dt dt dt

Integration property:
If x(t )⇔ X (s)
t
X (s)
then L{∫ x( τ)d τ }=
0 s
Since LT of an impulse is simply 1, it might be
advantageous to reduce the given x (t ) to delta function
by sucesive differentiation snd find out LT of the the
differentiated function easily from which LT of original
x (t ) can be easily obtained.

1. A pure inductance L is connected across a voltage

source v (t )u(t) . Let us assume i(0)=0 i.e., no initial
current. We wish to find out current i(t). KVL in time
domain is

di
L =v (t )u( t)
dt

Going to s-domain by taking LT of both the sides:

LsI (s )=V ( s) This algebraic equation is KVL in s-domain.

V ( s) V (s )
Hence I (s)= , therefore i(t)=L−1 { }
sL sL

In the s-domain the circuit can be redrawn with a V (s)

voltage and an impedance sL connected across the
source. The conclusion is that an inductance in time
domain can be replaced by sL in s-domain. But remember
this can be done if and only if the initial current
through the inductor is zero.
If the inductor had some initial current i(0- ) , then the
inductor can be replaced by a parallel combination of
sL and i (0- )/ s in s-domain.

2. A pure capacitor C has voltage current relationship

in time domain as:
dv
C =i(t)u (t)
dt
Going to s-domain by taking LT of both the sides:

CsV (s )=I (s ) This algebraic equation is KVL in s-domain.

In the s-domain the a pure capacitor C can be redrawn

with a V (s) voltage and an impedance 1/sC connected
across the source. The conclusion is that a capacitor in
time domain can be replaced by 1/sC in s-domain. But
remember this can be done if and only if the initial
voltage across the capacitor is zero.
If the capacitor had some initial voltage v (0- ) , then
the capacitor can be replaced by a series combination of
1/sc and voltage source of v ( 0- )/s in s-domain.

Lecture 26: on 29/08/2019 Thursday

LT of a convoluted signals x 1( t)∗x 2 (t) was shown to be

X 1 ( s) X 2 (s ) .So convolution in time domain means
multiplication in s-domain and vice versa.

LT of periodic function can be obtained by obtaining the

LT of the first period x 1( t)=x (t) {u( t)−u( t−T )} . It was shown
that:
X 1 ( s)
LT of a periodic function x (t ) = where T is
(1−e−sT )
the period and X 1 ( s) is the LT of the first period.

Revisited second order system in the light of Laplace

Transform
Discussed how to solve for capacitor voltage v (t ) and circuit current i(t) , in a
relaxed series R−L−C circuit when excited by a step voltage Vu(t ) . From KVL
equation we wrote the following differential equation:

d2 v dv
LC 2
+ R C +v=V noting that i=C dv with initial conditions v (0)=0 and
dt dt dt

i(0)=0 .
 d 2 v R dv 1 1
Dividing both sides by LC , 2
+ + v= V
dt L dt LC LC

R 1 R C 1
Assumed
L
=2 ζ ωn and
LC
=ω2n which means ζ =
2 L
and ωn =
√ LC
.
√
We then dealt with the same differnential equation with changed values of coefficients as follows:

d2 v dv
2
+2 ζ ωn +ω2n v=ω2n V u(t )
dt dt

Taking LT of the above equation:

V V
(s 2 +2 ζ ω n s+ ω 2n)V (s)=ω2n or, V (s)=ω 2n
s s ( s +2 ζ ω n s+ ω 2n)
2

After doing partial fraction expansion of the RHS:

V 1 ( s+2 ζ ωn ) 1 (s+ ζ ωn ) ( ζ ω n)
V (s)=ω 2n = − 2 = − 2 − 2
s ( s +2 ζ ω n s+ ω n) s (s +2 ζ ω n s+ ω n) s (s +2 ζ ωn s+ ω n) ( s +2 ζ ωn s+ ω2n)
2 2 2 2

( s+ ζ ω n)
or, V (s)= 1 −
ζ ωn
2 2
− 2 2
s (s + ζ ωn ) −ω d (s + ζ ωn) + ω d

The characteristic roots are −ζ ωn±√ ζ 2−1 ωn and the roots will be complex
conjugates if 0< ζ <1 . Then roots are −ζ ωn± j √ 1− ζ2 ωn=−ζ ωn± j ω d , where
ωd =√ 1− ζ2 ωn . Here we are considering the case when the roots are complex
conjigate. After taking inverse LT:
2
V −jω t −1 ωd V − jω t −1 √ 1− ζ
v (t )=V – e sin(
n
ω d t+ tan )=V – e sin ( ω d tn
+tan )
√1−ζ 2 ζ ωn √ 1−ζ 2 ζ
Expression for i(t) will be obtained from:
dv
i(t)=C .
dt
Use of LT in solving involved circuits with or without initial conditions in the
energy storing elements can be efficiently handled . You can even avoid writing down
the time domain differential equations (KVL equations) by drawing circuit in s-
domain etc.
Lecture 27: on 2/09/2019 Monday
Representation of discrete time signals – the x-axis is discrtized and denoted by n
where n is an integer. x [n] is the value of the function at a given n and this is
denoted by a stick with a buttet at the top.

Finite sequence of DT signal can be either plotted or can

be represented as a sequence of numbers as shown

x [n]={ x [−3] , x [−2] , x [−1], x [0 ], x [1], x [2], x [3], x [ 4 ]}={−2,1, 4, 3,−5, 2,−1, 7 }

The arrow mark indicates n=0 .

Unit impulse is defined as δ [n]=1 for n=0 and δ [n]=0 otherwise.

Unit step sequence is defined as

u[n]=1 for n≥0 and u[n]=0 otherwise. Note that u[0]=1 is well
defined at n=0 .

It is easy to see that

d u(t)
δ [n]=u(n)−u (n−1) . Recall in continuous time system δ (t )=
dt
Also
n
u[n]=∑ δ (n−k)
k=0

Any sequence can be broken into a sum of odd and even and odd sequence.

x [n]=x even [n]+ x odd [n]

where

1
x even [n]= (x [n]+ x [−n]) and x odd [n]= 1 ( x [n ]−x [−n])
2 2

Any sequence x [n] can be expreesed interms of impulses as follows:

∞
x [n]= ∑ x [k ] δ [n−k ] an important result.
k=−∞

If x [n]=x [n+ N ] , then the sequence is said to be periodic with the fundamental
period N where N is smallest integer for whcih x [n]=x [n+ N ] ,
Lecture 28 & 29: on 4/09/2019 Wednesday
A sequence x [n]=sin (Ω n) will be periodic if Ω can be expressed in the form

k
Ω=2 π where k and N are integers. And the fundamental period will be N .
N
For a multitone signals having several periodic sinusoidal sequences of periods
N 1, N 2, N 3 etc, the period of the overall signal will be LCM ( N 1 N 2 N 3 ).
Demonstrated this with simple examples.

Defined power and energy of discrete time signals and described how to calculate the
power and energy associated with a given sequence.

N
1 2
Power = P=lim N →∞ ∑ |x [n]| if this is finite , then power signal.
( 2 N +1) −N

N
Energy = W =lim N →∞ ∑ |x [n]|2 if this is finite , then energy signal.
−N

Shift and time scaling operation on discrete time signals. It was pointed out that

b b
δ [a n]=δ [n] and δ [a n+ b]= δ [a(n+ )]= δ [(n+ )]
a a
Recall this is different compared to continuous time case.

Discrete time systems can be classified as memory/memory less, linear/non-linear,

time (or shift) invariant / time (or shift) variant , causal/non-causal, stable/unstable
and invertible/non-invertible system in the same way as continuous time systems.

Discrete time Fourier Series (DTFS)

If a sequence x [n] is periodic with a fundamental period N , then the Discrete
Fourier series coefficients X k was derived to be:

2π
X k =∑ x [n]e− jk Ω n where fundamental frequency Ω= .Also note that unit of
N N
Ω is radian and not radian/sec

Note that X k is periodic with period N i.e., X k +N =X k

The reverse formula is x [n]=∑ X k e jk Ω n

N
DTFS coefficients were found out for a given periodic sequence with students
participation.
Convolution sum

For an LSI (loosely LTI) system, if the impulse response h[n] is known then
output of the system for any arbitrary sequence x [n] was shown to be the
convolution sum between x [n] and h[n] i.e.,
∞ ∞
y [n]=x [n]∗h [n ]=h [n ]∗x [n]= ∑ x [k ] h [n−k ]= ∑ h[ k ] x [n−k ]
k=−∞ k=−∞

∞
We have already seen that x [n]= ∑ x [k ] δ [n−k ] then x [n]=x [n]∗δ [n] .
k=−∞

If the sequences x [n] and h[n] are of finite lengths, then convolution sum can be
easily obtained based on some simple multiplication rules. Two rules were discussed.
The importance of knowledge of support values of x [n] and h[n] are highlighted.
One can avoid rather tedious steps generally involved in graphical method. Sum
problems were solved to explain this.

Lecture 30: on 5/09/2019 Thursday

If the sequences x [n] and h[n] are not of finite lengths then the convolution sum
can be obtained by using the formula
∞
y [n]=x [n]∗h [n ]=h [n ]∗x [n]= ∑ x [k ]h [n−k ]
k=−∞

In this case one has to carefully identify overlap regions and decide about range of
values of k over which sum is to be carried out. This was explained by considering
a problem.
Solution of linear difference equation
Difference equation is used to describe input-output relations in discrete time system.
(It is the counter part of differential equation used to describle input-output relations
in continuous time). Introduced shift operator E . If the initial values are known ,
one can of course get a solution recursively.

A linear difference equation will have a solution comprising of natural part and part
due to the forcing function.
The natural response is obtained by making the RHS = 0. Characteristic roots of the
difference equation will decide the natural response.
We shall try to seek total solution in closed form, classicaly of a given difference
equation.
Classical way to solve a linear constant coefficient difference equation was
introduced. If the excitation happens to be of the form x [n]=C bn u [n] , the solution
due to forcing function will be of same nature. Also to get the natural response,
characteristic equation is solved for roots and the total solution is sum of these two.
The constants are to be determined from the given boundary conditions applied to the
total solution. Discussion to be continued.

Lecture 31: on 9/09/2019 Monday

If it is a second order linear difference equation then two boundary conditions will be
necessaray namely y [−2] and y [−1] when excitation exists x [n] for n≥0
causal input signal.
If the equation is written in the form f (E) y [ n]=g[ E] x [n] , the solution due to
forcing function x [n]=C bn u [n] is:

g[ E]
y [n]= ⋮ x [n]
f [E ] E=b

Then total solution will be sum of natural response and forced response. Since this
solution is valid for n≥0 , we have to translate the given boundary condition
y [−2] and y [−1] to y [0] & y [1] recursively and use them to find out the
constants for getting complete solution.

The steps involved were explained by solving second order difference equation with
students’ participation when x [n]=u [n ] and then x [n]=δ [n]

Lecture 32 & 33: on 11/09/2019 Wednesday

The solution of difference equation when excitation is sinusoidal i.e.,
x [n]=cos Ω nu [n ] can be easily handled as
1 1
cos Ω n= e j Ωn + e− j Ωn .
2 2

Eigen function
For a linear shift invariant system whose impulse response is h[n] we know
∞ ∞
y [n]=x [n]∗h [n ]=h [n ]∗x [n]= ∑ x [k ] h [n−k ]= ∑ h[ k ] x [n−k ]
k=−∞ k=−∞

It is interesting to note that if x [n]=an then out put will be

N
∞ ∞
1 2
y [n]= ∑ h[k ]an−k =( ∑ h [k ]a−k ) an P=lim N →∞ ∑ |x [n]|
k=−∞ k=−∞ (2 N +1) −N
The bracketed term on the RHS is merely a serie sum whose value is decided by the
value of a and may be called eigen function – thus if x [n]=an out put will be
multiplication between the input and the eigen function. Note we don’t have to carry
out tedious convolution sum when x [n]=an .

As a special case if x [n]=e jΩ n=(e j Ω )n , then a=e j Ω and output will be

∞
y [n]=( ∑ h[k ] e− j Ωk )e j Ω n=[H (e jΩ )]e j Ω n
k=−∞

∞
where H (e j Ω )= ∑ h [k ]e− j Ωk is the eigen function of the system.
k=−∞

We now say that for any given sequence h[n] , we can find out the frequency
content of the sequence by computing H (e j Ω ) and this may be called the Discrete
Time Fourier Transform (DTFT) of h[n] .

The idea ∞can be extended to any other sequence x [n] and its DTFT is
X (e jΩ )= ∑ x [k ]e− j Ωk .
k=−∞

Is it possible to get back x [n] from X (e jΩ ) ? The answer is yes. It was shown that
π
1
x [n]= ∫ X (e j Ω )e j Ωn d Ω This is called inverse discrete time Fourier transform.
2 π −π

Considering a linear shift invariant system where input is any sequence x [n] and
the output is y [n]=h[n]∗x [n ] , it was shown that

Y (e j Ω )=H (e j Ω ) X (e j Ω )

So in frequency domain, convolution sum becomes mere multiplication of DTFTs.

Nyquist sampling theorem

To take advantages of of discretew time system one has to sample a coninuous time
signal x (t ) at regular interval of time say T s called sampling period or sampling
2π
frequency ω s= . If x (t ) has CTFT X ( ω) which is band limited between
Ts
− ωb to ωb then sampling frequency ω s≥2 ωb so that no information will be lost
about x (t ) even after sampling.
To arrive at the above result we assumed X ( ω) of x (t ) is known.
Then we used unit impulse sampler to convert x (t ) to x s (t ) a sampled version of
x (t ) . After this we got CTFT X s ( ω ) of x s (t ) . The idea is to examine whether
X s ( ω )=X ( ω ) ? Now it is established that the condition under which these two will
be equal is ω s≥2 ωb and this is known as Nyquist sampling theorem.
In general we say that find out the highest frequency content of x (t ) , then the
sampling frequency should be at least twice this highest frequency. While
establishing the above result we remained in continuous time domain and used
various properties of CTFT.

P.S: In the remaing time students were asked to establish the initial and final value
theorem of Laplace transform using the result of LT of dx =sX (s)−x (0)
dt

Intial value theorem : x (0)=lim s →∞ s X (s)

Final value theorem : x (∞)=lim s →0 s X (s)