Periodic and aperiodic signals

Periodic signals have the property that x(t + T ) = x(t), for all t
Period: The smallest value of T that satisfies the definition

Causal and anti-causal signals

A causal signal is zero for t < 0 and an anti-causal signal is zero
for t > 0

Right- and left-sided signals

A right-sided signal is zero for t < T and a left-sided signal is
zero for t > T where T can be positive or negative.

Even and odd signals

Even signals xe (t) and odd signals xo (t) are defined as
xe (t) = xe (−t) and xo (t) = −xo (−t)

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 1 / 61

Bounded and unbounded signals

Building-block signals
We will represent signals as sums of building-block signals.
Important families of building-block signals are the eternal,
complex exponentials and the unit impulse functions.

Eternal, complex exponentials

These signals have the form
x(t) = Xe st for all t and x[n] = Xz n for all n,

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 2 / 61

Eternal complex exponentials — why are they
important?

Almost any signal of practical interest can be represented as a

superposition (sum) of eternal complex exponentials.

The output of a linear time-invariant (LTI) system is simple to

compute if the input is a sum of eternal complex exponentials.

Eternal complex exponentials are the eigen functions or

characteristic (unforced, homogeneous) responses of LTI systems

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 3 / 61

Main fact: Complicated time waveforms can be expressed as a sum

of sine and cosine waves.
Why bother? Sine/cosine are the only bounded waves that stay the
same when differentiated.
Any electronic circuit: sine wave in ⇒ sine wave out (same
frequency). Hard problem: Complicated waveform → electronic
circuit → output = ?
Easier problem: Complicated waveform → sum of sine waves →
linear electronic circuit (⇒ obeys superposition) → add sine wave
outputs → output = ?

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 4 / 61

Sum of sine waves → sum of sine waves

In these lectures we will use

T0 for the fundamental period and
f0 = T10 for the fundamental frequency.

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 5 / 61

Joseph Fourier
1768-1830

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 6 / 61

Fourier series

1
For a periodic time-domain signal with a period T0 =
f0

gp (t) = a0 + 2 an cos(2πnf0 t) + bn sin(2πnf0 t) (1)

n=1
∞ h
X i

gp (t) = a0 + 2 an cos(nω0 t) + bn sin(nω0 t) (2)

n=1
∞ h
X i

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 7 / 61

ref: Dr. Ali J. Abboud, University of Diyala, College of Engineering, Department of Computer Engineering

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 8 / 61

FS representation as complex exponentials

Substituting
cos(x) = exp(jx) + exp(−jx)
2
1h i

sin(x) =
2j 2
exp(jx) − exp(−jx) = − j exp(jx) + j exp(−jx)
1h i 1h i

in equation (2)
we get

gp (t) = a0 + (an − jbn ) exp(j2πnf0 t) + (an + jbn ) exp(−j2πnf0 t)

n=1
∞ h
X i

(3)

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 9 / 61

let us define a new constant

cn = a 0 for n = 0 (4)



(an + jbn ) for n, 0

(an − jbn ) for n > 0


Substitute in the above equation



∞ ∞
gp (t) = [cn exp(j2πnf0 t)] = [cn e (j2πnf0 t) ] (5)
n=−∞ n=−∞
X X

In this course we will consider the

oscillation frequency f0 (Hz)- instead of
angular frequeny ω0 = 2πf0 (radians/second)

Compact form/ complex form

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 10 / 61

 Another form
Let the above equation be written as
gp (t) = a0 + ∞ n=1 [An cos(2πnf0 t + ϕn )]
P

gp (t) = a0 + n=1 [An cos(ϕn )cos(2πnf0 t) − An sin(ϕn )sin(2πnf0 t)]

Compare the coefficients

P∞

An cos(ϕn ) = an and −An sin(ϕn ) = bn then

n
the amplitude An = an2 + bn2 and the phase angle ϕn = −tan −1
an
hb i

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 11 / 61

Also,
and
−bn
2
a n + bn2
an
sin(ϕn ) = p

an2 + bn2
cos(ϕn ) = p

(consider a right triangle with sides −bn , an , and (an2 + bn2 )

∞
∴ gp (t) = a0 + [An cos(2πnf0 t + ϕn )] (6)
n=1
X

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 12 / 61

Evaluating a0, an and bn of the Fourier series

First the a0
Integrate both the sides of Equation 1 over a period T0
Choose the period depending on the convenience of evaluating the
the integral

T gp (t)dt =
− 20
R T20

T0 a0 + 2 n=1 an cos(2πnf0 t) + bn sin(2πnf0 t) dt

2
−
R T20  P∞ h i

= T0 a0 dt
2
−
R T20

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 13 / 61

= a0 T0

∴ a0 = T gp (t)dt
T0 − 20
1 R T20

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 14 / 61

Now evaluating the an and the bn
We evaluate the coefficients an and the bn by the use of property of
othgonality,

T0
2 T0 for n = m
2 cos(2πnf0 t)cos(2πmf0 t) dt = (7)
T 0 for n =
(

− 20
Z

̸ m
h i

T0
2 0 for n = m
2 cos(2πnf0 t)sin(2πmf0 t) dt = (8)
T
(

− 20
Z

0 for n ̸= m
h i

T0
2 T0 for n = m
2 sin(2πnf0 t)sin(2πmf0 t) dt = (9)
T 0 for n =
(

− 20
Z

̸ m
h i

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 15 / 61

∴ an = T gp (t)cos(2πnf0 t)dt
T0 − 20
and
1 R T20

bn = T gp (t)sin(2πnf0 t)dt
T0 − 20
1 R T20

Whereas the cn ’s will be

cn = T gp (t)exp(j2πnf0 t)dt
T0 − 20
1 R T20

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 16 / 61

Example
Periodic Rectangular Pulse Train

Finding the coefficients a0 , an and bn of the Fourier series

T gp (t)dt =
− 20
R T20

T0 a0 + 2 n=1 an cos(2πnf0 t) + bn sin(2πnf0 t) dt

2
−
R T20  P∞ h i

Adt = a0 T0 dt + 0 + 0 = a0 T0
2
− τ2 −
R τ2 R T20

or A τ = a0 T 0
τ
implies a0 = A
T0

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 17 / 61

Now
an = A cos(2πnf0 t)dt
T0 −τ
2 R +τ

2A sin(πnf0 t) +τ
=
T0 πnf0 −τ

A
= [sin(πnf0 τ ) − sin(−πnf0 τ )]
πnf0 T0

2A
= [sin(πnf0 τ )]
πn
an ’s are all constants

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 18 / 61

Similarly,

2
bn = T0 −τ
Asin(2πnf0 t)dt
2A cos(πnf0 t) +τ
R +τ

=
T0 πnf0 −τ
A
= [cos(πnf0 τ ) − cos(−πnf0 τ )] = 0
πnf0 T0

Substituting the constants an ’s and the bn ’calculated above

τ 2A
gp (t) = A + ∞ n=1 sin(πnf0 τ )cos(2πnf0 t)
T0 πn
P

∞
τ
gp (t) = A + an cos(2πnf0 t) (10)
T0 n=1
X

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 19 / 61

Fourier series coefficients an for the rectangular pulse train of

different time periods

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 20 / 61

Spectrum

Periodic cosinusoidal (cos/sin) waves will be called frequencies

A spectrum is a range of frequencies

Electromagnetic spectrum
All frequencies of electromagnetic radiation - visible light, radio,
infrared, etc.

speed of light
All travel at speed of light, so wavelength =
frequency

Transmission through atmosphere varies by wavelength

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 21 / 61

Problems: Fourier Series/ Periodic Signals

1. For the following signals,
(i) determine analytically which are periodic (if periodic, give the
period) and
(ii) sketch them
(Scale your time axis so that a sufficient amount of the signal is
being plotted)
(a) x(t) = 4cos(5πt) (b) x(t) = 4cos(5πt − π4 )
1
(c) x(t) = 4u(t) + 2sin(3t) (d) x(t) = u(t) − 2

(e) x(t) = cos(4t) + 2sin(8t) (f) x(t) = 3 cos(4t) + sin(πt)

(g) x(t) = cos(3πt) + 2cos(4πt)

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 22 / 61

2. Determine if the following discrete signals are periodic; if periodic,
give the period
(a) x[n] = 4cos(πn) (b) x[n] = 4cos(πn − 2)
(c) x[n] = 2 sin(3n)

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 23 / 61

3. Give an expression for the signal:

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 24 / 61

 Tarun Varma 22ECT104 Signals and Systems April 8, 2024 25 / 61

4. Are the following periodic? If so, give the period.

(a) x(t) = 4 cos(3πt + π4 ) + u(t)
(b) x(t) = 4 cos(3πt + π4 ) + 2 cos(4πt)
(c) x(t) = cos (2ω1 t) + cos (3ω1 t)
where ω1 is a specific frequency
(d) x(t) = 4 cos(3πt + π2 ) + 2 cos(8πt + π2 )
(e) x(t) = 2 cos(3πt + π2 ) + 4 cos(10t − π/2)

(f) x[n] = 4 cos(0.5πn + π4 )

(g) x[n] = 12 cos(20n)
(h) x[n] = 10 cos(2π(8)n)
(i) x[n] = 10 cos(8n)

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 26 / 61

5. Plotting Signals

(a)

for t < −4
(11)


x(t) = t + 2 for −4 ≤ t < 3


0

t − 2 for t ≥ 3



(b) y (t) = x(t − 1) where x(t) is defined in part (a)

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 27 / 61

(c)

for n < 2
(12)


x[n] = 2n − 4 for 2 ≤ n < 4


0

4 − n for n ≥ 4



(d) y[n] = x[n+1] where x[n] is defined in part c)

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 28 / 61

ref: Helsinky University of Technology, Communication Laboratory, Timo O. Corhonen

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 29 / 61

Fourier Transform

A non-periodic signal can be considered as a periodic signal with a

period = ∞
g (t) = gp (t)|T0 →∞
= limT0 →∞ n=−∞ [cn exp(j2πnf0 t)]
1
= [cn T0 exp(j2πnf0 t)]
P∞

limT0 →∞ n=−∞
T0
P∞

Let us define
1
cn T0 = G (fn ) and = ∆f0
T0
Substituting in the equation
= limT0 →∞ n=−∞ [G (fn ) exp(j2πnf0 t)] ∆f0
P∞

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 30 / 61

g (t) = −∞
G (f )exp(+j2πft)df
R +∞

= G (ω)exp(+jωt) dω
2π −∞
1 R +∞

In exactly the same manner,

G (f ) = −∞ g (t)exp(−j2πft)dt
R +∞

Fourier pair

g (t) ⇋ G (f )

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 31 / 61

Operation on Signals

Need

1 Linearity
2 Time shifting or delay
3 Frequency shifting
4 Time scaling or Time reversal
5 Duality
6 Differentiation
7 Integration
8 Convolution

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 32 / 61

If g (t) ⇋ G (f )
g1 (t) ⇋ G1 (f ) and
g2 (t) ⇋ G2 (f )
a, b, a1 , a2 , b1 , b2 , t0 , t1 , fc , f1 , f2 , f0 etc. are constants,

Linearity [g1 (t) + g2 (t)] ↔ [G1 (f ) + G2 (f )]

Proof:
Fourier transform of a1 g1 (t) + a2 g2 (t) will be
= −∞ [a1 g1 (t) + a2 g2 (t)]exp(−j2πft)dt
R +∞

= a1 −∞ g1 (t)exp(−j2πft)dt + a2 −∞ g2 (t)exp(−j2πft)dt
R +∞ R +∞

= a1 G1 (f ) + a2 G2 (f )
∴ [a1 g1 (t) + a2 g2 (t)] ⇋ [a1 G1 (f ) + a2 G2 (f )]

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 33 / 61

Time shifting or delay g (t − t0 ) ⇋ exp(−j2πft0 )G (f )

Proof:
Fourier transform of g (t − t0 ) will be
= −∞ g (t − t0 )exp(−j2πft)dt
R +∞

Now let (t − t0 ) = τ
∴ dt = dτ

Equation
= −∞ g (τ )exp(−j2πf (t0 + τ )dτ
R +∞ above becomes

= exp(−j2πft0 ) −∞
g (τ )exp(−j2πf τ )dτ
R +∞

= exp(−j2πft0 ) G (f )

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 34 / 61

 Frequency shifting exp(+j2πfc t)g (t) ⇋ G (f − fc )

Proof:

F [exp(+j2πfc t)g (t)]

= −∞
[exp(−j2πfc t)g (t)] exp(−j2πft)dt
R +∞

= −∞
exp(−j2π(f − fc )t) g (t)dt
R +∞

= G (f − fc )

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 35 / 61

Time scaling g (at) ⇋ G

|a| a
1 f 

Proof:

F [g (at)]

= −∞
g (at)exp(−j2πft) dt
R +∞

Now, let at = τ
∴ a dt = dτ
Substituting

dτ f
= g (τ )exp(−j2πf τ /a) = g (τ )exp(−j2π τ ) dτ
−∞
a a −∞ a
R +∞ 1 R +∞

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 36 / 61

Duality G (t) ⇋ g (−f )

Looking up into the equations of the Fourier pair

g (t) = −∞
G (f )exp(+j2πft)df
R +∞

G (f ) = −∞
g (t)exp(−j2πft)dt
R +∞

we change the variable f with y first in the equation for g(t)

g (t) = −∞
G (y )exp(+j2πyt)dy
R +∞

and then change the variable t with -f

g (−f ) = −∞
G (y )exp(−j2πyf )dy
R +∞

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 37 / 61

d g (t)
Differentiation ⇋ j2πf G (f )
dt
Proof:
The first equation of the Fourier pair is
g (t) = −∞ G (f )exp(+j2πft)df
R +∞

Differentiate both the sides of this equation

dg (t)
dt dt
= [ −∞ G (f )exp(+j2πft)df ]
d R +∞

Interchange the order of differentiation and integration on the RHS

[G (f )exp(+j2πft)df ] =
dt dt
= −∞
exp(+j2πft)(+j2πf )G (f )df
dg (t) R +∞ d

−∞
R +∞

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 38 / 61

 1 1
Integration g (τ )dτ ⇋ G (f ) + G (0)δ(f )
−∞
j2πf 2
Rt

Proof:
As it is the inverse of differentiation, the result is obviously as given
above

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 39 / 61

Multiplication in time domain

g1 (t) g2 (t) ⇋ −∞ G1 (λ) G2 (f − λ)dλ

R +−∞

Written as, g1 (t) g2 (t) ⇋ G1 (f ) ⊗ G2 (f )

Convolution

−∞
g1 (τ ) g2 (t − τ )dτ ⇋ G1 (f ) G2 (f )
Rt

Written as, g1 (t) ⊗ g2 (t) ⇋ G1 (f )G2 (f )

Demo
https://pages.jh.edu/signals/

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 40 / 61

Example
Rectangular Pulse

Now let us evaluate the Fourier Transform of a rectangular pulse

t
g (t) = A rect( )
T

Also known an ON-OFF function or the Gate function ga(t)

Plot it

By definition,
G (f ) = −∞ g (t)exp(j2πft)]dt
R +∞

= − T2
A exp(j2πft)]dt
R + T2

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 41 / 61

G (f ) =
j2πf
exp(j2πfT /2) − exp(−j2πfT /2)
A n o

or G (f ) = sin(πfT )
πf
An o

Write the above equation as

AT sin(πfT )
G (f ) = sin(πfT ) = AT
πfT πfT

sin(πλ)
Now, define = sinc(λ)
πλ

∴ G (f ) = AT sinc(fT )

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 42 / 61

Normalized sinc function
Read
Equations That Changed the World by Ian Stewart
sin(πx)
sinc(x) =
πx

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 43 / 61

Normalized sinc function

sin(πx) 1
sinc(x) = = sin(πx)
πx πx

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 44 / 61

sinc x plotted w.r.t. ω

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 45 / 61

90.3% of the power is contained in the main lobe

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 46 / 61

 Tarun Varma 22ECT104 Signals and Systems April 8, 2024 47 / 61

Example
Delta function δ(t)

Now let us evaluate its Fourier Transform

F [g (t)] = F [δ(t)] = −∞
δ(t)exp(j2πft)]dt
R +∞

This can be considered as the limiting form of the rectangular pulse

discussed in the previous slides with a unit area

1
Implies A =
T

∴ F [g (t)] = F [δ(t)] = 1

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 48 / 61

F [g (t)] = F [δ(t)] = −∞
δ(t)exp(j2πft)]dt = 1
R +∞

Writing
−∞
δ(t)exp(j2πft)]dt = −∞ δ(t)[cos(2πft) + j sin(2πft)]dt = 1
R +∞ way as
R +∞ this equation in another

We see that it picks up one value of the exp/cos/sin at t=0

i.e. at the point it exists

δ(t) exists at t=0

Sifting property of this Delta function

in general
−∞
δ(t)g (t)dt = g (0) or −∞
δ(t − t0 )g (t)dt = g (t0 )
R +∞ R +∞

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 49 / 61

Sifting property of this Delta function

in general

−∞
δ(t)g (t)dt = g (0)
R +∞

or −∞
δ(t − t0 )g (t)dt = g (t0 )
R +∞

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 50 / 61

Fourier Pairs

1. δ(t) ↔ 1
2. u(t) = −∞ δ(τ )dτ
using integration property,
Rt

1 1
u(t) ⇋ + δ(f )
j2πf 2

3. sgn(t) = 2u(t) − 1
2 2 1
∴ sgn(t) ⇋
j2πf 2 jπf
+ δ(f ) − δ(f ) =

Write the other important FT pairs

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 51 / 61

Fourier tranform pairs

Time function Fourier Transform

t
rect( ) T sinc(fT )
T
1 f
sinc(2Wt) rect( )
2W 2W
1
exp(−at)u(t), a > 0
a + j2πf
2a
exp(−a|t|), a > 0
a2 + (j2πf )2
exp(−πt 2 ) exp(−πf 2 )

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 52 / 61

Fourier tranform pairs

Time function Fourier Transform

δ(t) 1
1 δ(f )
δ(t − t0 ) exp(−j2πft0 )
exp(−j2πfc t) δ(f − fc )
1
cos(2πfc t)
2
δ(f − fc ) + δ(f + fc )
1
sin(2πfc t)
2j
δ(f − fc ) − δ(f + fc )

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 53 / 61

Fourier pairs

δ(t) → → u(t) → → unit ramp

R R

Sometimes delta function is written as

u0 (t) = δ(t)

where u(t) = u1 (t) = 0

u0 (t) dt = 0
δ(τ )dτ = −ϵ
δ(τ )dτ = 1
Rt Rt R +ϵ

un−1 (t) = −∞
un (t) dt
Rt

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 54 / 61

Why are engineers obsessed with sine waves?
Answer: Because ...

1. A sine wave remains a sine wave of the same frequency when you
multiply by a constant
add onto to another sine wave of the same frequency
differentiate or integrate or shift in time

2. Almost any function can be expressed as a sum of sine waves

� Periodic functions → Fourier Series
� Aperiodic functions → Fourier Transform

3. Many physical and electronic systems are

composed entirely of constant-multiply/add/differentiate
linear: u(t) → x(t) and v (t) → y (t) means that
u(t) + v (t) → x(t) + y (t)
Tarun Varma 22ECT104 Signals and Systems April 8, 2024 55 / 61

Lord Kelvin on Fourier’s theorem

Fourier’s theorem is not only one of the most beautiful results of

modern analysis, but it may be said to furnish an indispensable
instrument in the treatment of nearly every recondite question in
modern physics.
Lord Kelvin

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 56 / 61

Fourier Transform - Meaning

Fourier pair

g (t) = −∞
G (f )exp(+j2πft)df
R +∞

G (f ) = −∞
g (t)exp(−j2πft)dt
R +∞

g (t) ⇋ G (f )
Block 1 Block 2
Convolution Correlation
y (t) = −∞ x(τ )h(t − τ ) dτ y (t) = −∞ x(τ )h(t + τ ) dτ
Rt Rt

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 57 / 61

We can write the equation of the Fourier Transform alternatively as,

G (f ) = −∞ g (t) [cos(2πft) − j sin(2πft)] dt
R +∞

G (f ) = −∞
g (t) cos(2πft) dt − j −∞
g (t) sin(2πft)] dt
R +∞ R +∞

The Fourier Transform is the correlation of the given signal g(t) with
cos and sin waveforms with different values of f.

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 58 / 61

Significance of Negative Frequencies
Each real sinusoid in the general Fourier Series representation may be
written as a sum of complex exponentials,
an
an cos(2πnf0 t + ϕn ) = [exp(+j2πnf0 t + ϕn ) + exp(−j2πnf0 t + ϕn )]
2
Thus each real cosine frequency nf0 may be viewed in the domain of
complex exponentials as being composed of a positive frequency nf0
and a negative frequency −nf0 , each with an amplitude which is
one-half of an (and corresponding phase angles ϕn and −ϕn )

We may therefore think of the Fourier spectrum for real periodic

signals (or individual real pulses) as being symmetrically placed
around the origin, with one-half the amplitudes An at both negative
and positive values of each frequency.
Tarun Varma 22ECT104 Signals and Systems April 8, 2024 59 / 61

Fast Fourier Transforms (FFTs)/1963

To find the Fourier Transform of the signal captured by the

seismometer recording earthquakes and also the nuclear explosions
Physicist Dr. Richard Garwin and mathematician Dr. John W. Tukey
Physicist Dr. Richard Garwin computed the Fourier Transorm of the
seismometer signal at CERN.
Paper (published with IBM researcher Dr. James W. Cooley)
An algorithm for the machine calculation of Complex Fourier Series
For N = 106 samples will require N 2 = 1012 computations.

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 60 / 61

At the speed of the computers available in the 1960s, it would require
about two years to calculate th Fourier Tranform of one such signal
Complexity reduced from N 2 to N log2 N
1
A signal with N=1000 samples will require fewer calculations
100
1
and for N = 106 will require fewer calculations
50, 000
Computation time reduced from two years to almost less than 30
minutes

Youtube video on
The Remarkable Story Behind The Most Important Algorithm Of All
Time

Tarun Varma 22ECT104 Signals and Systems April 8, 2024 61 / 61