Handouts PDF

E1.
10 Fourier Series and Transforms
Mike Brookes
E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 1 / 14

Syllabus
⊲ Syllabus Main fact: Complicated time waveforms can be

Optical Fourier
Transform expressed as a sum of sine and cosine waves.
Organization
1: Sums and Why bother? Sine/cosine are the only bounded

Averages
waves that stay the same when differentiated.
Any electronic circuit:
sine wave in ⇒ sine wave out (same frequency).
Joseph Fourier
1768-1830
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 = ?
Syllabus: Preliminary maths (1 lecture)

Fourier series for periodic waveforms (4 lectures)
Fourier transform for aperiodic waveforms (3 lectures)

Optical Fourier Transform
Syllabus
Optical Fourier
A pair of prisms can split light up into its component frequencies (colours).
⊲ Transform This is called Fourier Analysis.
Organization
1: Sums and
A second pair can re-combine the frequencies.
Averages This is called Fourier Synthesis.
Fourier Analysis Fourier Synthesis
We want to do the same thing with mathematical signals instead of light.

Organization
Syllabus
Optical Fourier
• 8 lectures: feel free to ask questions
Transform
⊲ Organization • Textbook: Riley, Hobson & Bence “Mathematical Methods for Physics
1: Sums and
Averages
and Engineering”, ISBN:978052167971-8, Chapters [4], 12 & 13
• Lecture slides (including animations) and problem sheets + answers
available via Blackboard or from my website:
http://www.ee.ic.ac.uk/hp/staff/dmb/courses/E1Fourier/E1Fourier.htm
• Email me with any errors in slides or problems and if answers are

wrong or unclear

Syllabus
Optical Fourier
Transform
Organization
1: Sums and
⊲ Averages
Geometric Series
Infinite Geometric
Series
Dummy Variables
Dummy Variable
Substitution
Averages
1: Sums and Averages
Average Properties
Periodic Waveforms
Averaging Sin and
Cos
Summary

Geometric Series
Syllabus
Optical Fourier
A geometric series is a sum of terms that increase or decrease by a constant
Transform factor, x:
Organization
1: Sums and
S = a + ax + ax2 + . . . + axn
Averages The sequence of terms themselves is called a geometric progression.
⊲ Geometric Series
Infinite Geometric
Series We use a trick to get rid of most of the terms:
Dummy Variables
Dummy Variable
Substitution
S = a + ax + ax2 + . . . + axn−1 + axn
Averages xS = ax + ax2 + ax3 + . . . + axn + axn+1
Average Properties
Periodic Waveforms
Averaging Sin and Now subtract the lines to get: S − xS = (1 − x) S = a − axn+1
Cos
Summary Divide by 1 − x to get: a = first term n + 1 = number of terms
1−xn+1
S =a× 1−x
Example:
S = 3 + 6 + 12 + 24 [a = 3, x = 2, n + 1 = 4]
1−24 −15
=3× 1−2 =3× −1 = 45

Infinite Geometric Series
Syllabus 2 n 1−xn+1
Optical Fourier A finite geometric series: Sn = a + ax + ax + · · · + ax = a 1−x
Transform
Organization
What is the limit as n → ∞?
1: Sums and
Averages
Geometric Series
If |x| < 1 then xn+1 −→ 0 which gives a = first term
n→∞
Infinite Geometric
⊲ Series
1 a
Dummy Variables S∞ = a + ax + ax2 + · · · = a 1−x = 1−x
Dummy Variable
Substitution x = factor
Averages
Average Properties Example 1:
Periodic Waveforms
0.4
Averaging Sin and
Cos
0.4 + 0.04 + 0.004 + . . . = 1−0.1 = 0.4̇ [a = 0.4, x = 0.1]
Summary
Example 2: (alternating signs)

2
2 − 1.2 + 0.72 − 0.432 + . . . = 1−(−0.6) = 1.25 [a = 2, x = −0.6]
Example 3:
1 1
1 + 2 + 4 + . . . 6= 1−2 = −1 = −1 [a = 1, x = 2]
a
The formula S = a + ax + ax2 + . . . = 1−x is only valid for |x| < 1

Dummy Variables
Syllabus
P
Optical Fourier
Using a sign, we can write the geometric series more compactly:
Transform
Sn = a + ax + ax2 + . . . + axn = nr=0 axr
P
Organization
1: Sums and
Averages
[Note: x0 , 1 in this context even when x = 0]
Geometric Series
Infinite Geometric Here r is a dummy variable: you can replace it with anything else
Series
⊲ Dummy Variables Pn r
Pn k
Pn α
Dummy Variable r=0 ax = k=0 ax = α=0 ax
Substitution
Averages
Average Properties
Dummy variables are undefined outside the summation so they sometimes
Periodic Waveforms get re-used elsewhere in an expression:
Averaging Sin and
Cos

1−24 1−32
P3 r
P2 r
Summary
r=0 2 + r=1 3 = 1 × 1−2 + 3 × 1−3 = 15 + 12 = 27
The two dummy variables are both called r but they have no connection
with each other at all (or with any other variable called r anywhere else).

Dummy Variable Substitution
Syllabus
P
Optical Fourier
We can derive the formula for the geometric series using notation:
Transform
Sn = nr=0 axr and xSn = nr=0 axr+1
P P
Organization
1: Sums and
Averages We need to manipulate the second sum to involve xr .
Geometric Series
Infinite Geometric
Series Use the substitution s = r + 1 ⇔ r = s − 1.
Dummy Variables
Dummy Variable
Substitute for r everywhere it occurs (including both limits)
⊲ Substitution Pn+1 s Pn+1 r
Averages
Average Properties
xSn = s=1 ax = r=1 ax
Periodic Waveforms
Averaging Sin and It is essential to sum over exactly the same set of values when substituting
Cos
Summary for dummy variables.
Pn r
Pn+1 r
Subtracting gives (1 − x)Sn = Sn − xSn = r=0 ax − r=1 ax
r ∈ [1, n] is common to both sums, so extract the remaining terms:
0 n+1
Pn Pn
(1 − x)Sn = ax − ax r
+ r=1 ax − r=1 axr
0 n+1 n+1

= ax − ax =a 1−x
1−xn+1
Hence: Sn = a 1−x

Averages
Syllabus
Optical Fourier
If a signal varies with time, we can plot its waveform, x(t).
Transform
Organization The average value of x(t) in the range T1 ≤ t ≤ T2 is
1: Sums and R T2
Averages 1
Geometric Series
hxi[T1 ,T2 ] = T2 −T1 t=T1 x(t)dt
Infinite Geometric
Series
Dummy Variables
Dummy Variable
Substitution x(t) x(t)
⊲ Averages
<x>[T1,T2] <x>[T1,T2]
Average Properties
Periodic Waveforms
Averaging Sin and
T1 T2 T1 T2
Cos
Summary
The area under the curve x(t) is equal to the area of the rectangle
defined by 0 and hxi[T1 ,T2 ] .
Angle brackets alone, hxi, denotes the average value over all time
hx(t)i = limA,B→∞ hx(t)i[−A,+B]

Average Properties
Syllabus
Optical Fourier
The properties of averages follow from the properties of integrals:
Transform
Organization
Addition: hx(t) + y(t)i = hx(t)i + hy(t)i
1: Sums and
Averages Add a constant: hx(t) + ci = hx(t)i + c
Geometric Series
Infinite Geometric
Series
Constant multiple: ha × x(t)i = a × hx(t)i
Dummy Variables
Dummy Variable
Substitution where the constants a and c do not depend on time.
Averages
⊲ Average Properties
Periodic Waveforms
Averaging Sin and
Cos For example:
Summary R T2
1
hx(t) + y(t)i[T1 ,T2 ] = T2 −T1 t=T1
(x(t) + y(t)) dt
1
R T2 1
R T2
= T2 −T1 t=T1
x(t)dt + T2 −T1 t=T1
y(t)dt
= hx(t)i[T1 ,T2 ] + hy(t)i[T1 ,T2 ]
But beware: hx(t) × y(t)i =

6 hx(t)i × hy(t)i.

Periodic Waveforms
Syllabus
Optical Fourier
A periodic waveform with period T repeats itself at intervals of T :
Transform x(t + T ) = x(t) ⇒ x(t ± kT ) = x(t) for any integer k.
Organization
1: Sums and
Averages
The smallest T > 0 for which x(t + T ) = x(t) ∀t is the fundamental
Geometric Series period. The fundamental frequency is F = T1 .
Infinite Geometric
Series
Dummy Variables
Dummy Variable <x>
Substitution
Averages
T
Average Properties
Periodic
⊲ Waveforms For a periodic waveform, hx(t)i equals the average over one period.
Averaging Sin and
Cos
Summary
It doesn’t make any difference where in a period you start or how many
whole periods you take the average over.
Example:
x(t) = |sin t|
1 π
R 1
Rπ
hxi = π t=0 |sin t| dt = π t=0
sin t dt
1 π 1 2
= π [− cos t]0 = π (1 + 1) = π ≈ 0.637

[proof that x(t ± kT ) = x(t)]
Proof that x(t + T ) = x(t) ∀t ⇒ x(t ± kT ) = x(t) ∀t, ∀k ∈ Z

We use induction. Let Hk be the hypothesis that x(t + kT ) = x(t) ∀t. Under the assumption that
x(t + T ) = x(t) ∀t, we will show that if Hk is true, then so are Hk+1 and Hk−1 . Since we know that
H0 is definitely true, this implies that Hk is true for all integers k, i.e. for all k ∈ Z.
Suppose Hk is true, i.e. x(τ + kT ) = x(τ ) ∀τ . Now set τ = t + T . This gives x(t + T + kT ) =
x(t + T ) ∀t. But, we assume that x(t + T ) = x(t), so x (t + (k + 1)T ) = x(t + T + kT ) =
x(t + T ) = x(T ) ∀t. Hence Hk+1 is true.
Now suppose Hk is true as before but this time set τ = t − T . Substituting this into u(τ + kT ) =
u(τ ) gives u(t−T +kT ) = u(t−T ). Substituting it also into u(τ +T ) = u(τ ) gives u(t) = u(t−T ).
Finally, combining these two identities gives u (t + (k − 1)T ) = u(t) which is Hk−1 .
E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – note 1 of slide 12
Averaging Sin and Cos
Syllabus
Optical Fourier
A sine wave, x(t) = sin 2πF t, has a frequency F and a period T = F1
Transform 1

Organization so that, sin 2πF t + F = sin (2πF t + 2π) = sin 2πF t.
1: Sums and
1
Averages RT F=1kHz
1
Geometric Series hsin 2πF ti = sin (2πF t) dt
x(t)
0
Infinite Geometric T
t=0
Series
Dummy Variables
=0 -1
0 0.5 1 1.5 2
Time (ms)
Dummy Variable
Substitution
Averages Also, hcos 2πF ti = 0 except for the case F = 0 since cos 2π0t ≡ 1.
Average Properties
Periodic Waveforms
Averaging Sin and
(
⊲ Cos 0 F 6= 0
Summary Hence: hsin 2πF ti = 0 and hcos 2πF ti =
1 F =0
i2πF t

Also: e = hcos 2πF t + i sin 2πF ti
= hcos 2πF ti + i hsin 2πF ti
(
0 F 6= 0
=
1 F =0

Summary
Syllabus
Optical Fourier
• Sum of geometric series (see RHB Chapter 4)
◦ Finite series: S = a × 1−x
n+1
Transform
Organization 1−x
a
1: Sums and
◦ Infinite series: S = 1−x but only if |x| < 1
Averages
Geometric Series
Infinite Geometric
• Dummy variables
Series ◦ Commonly re-used elsewhere in expressions
Dummy Variables
Dummy Variable ◦ Substitutions must cover exactly the same set of values
Substitution
Averages 1
R T2
Average Properties
• Averages: hxi[T1 ,T2 ] = T2 −T1 t=T1 x(t)dt
Periodic Waveforms
Averaging Sin and
Cos
• Periodic waveforms: x(t ± kT ) = x(t) for any integer k
⊲ Summary ◦ Fundamental period is the smallest T
◦ Fundamental frequency is F = T1
◦ For periodic waveforms, hxi is the average over any integer
number of periods
◦ hsin 2πF ti = 0 (

i2πF t 0 F 6= 0
◦ hcos 2πF ti = e =
1 F =0

⊲ 2: Fourier Series
Periodic Functions
Fourier Series
Why Sin and Cos
Waves?
Dirichlet Conditions
Fourier Analysis
Trigonometric
Products
Fourier Analysis
Fourier Analysis
Example
Linearity 2: Fourier Series
Summary
E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – 1 / 11

Periodic Functions
2: Fourier Series
⊲ Periodic Functions
A function, u(t), is periodic with period T if u(t + T ) = u(t) ∀t
Fourier Series • Periodic with period T ⇒ Periodic with period kT ∀k ∈ Z+
Why Sin and Cos
Waves?
The fundamental period is the smallest T > 0 for which u(t + T ) = u(t)
Fourier Analysis T
Trigonometric
Products
Fourier Analysis
Fourier Analysis
Example
Linearity
Summary
If you add together functions with different periods the fundamental period
is the lowest common multiple (LCM) of the individual fundamental
periods.
Example:
2π
• u(t) = cos 4πt ⇒ Tu = 4π = 0.5
• v(t) = cos 5πt ⇒ Tv = 2π5π = 0.4
• w(t) = u(t) + 0.1v(t) ⇒ Tw = lcm(0.5, 0.4) = 2.0
Tu = 0.5 Tv = 0.4 Tw = 2.0

Fourier Series
2: Fourier Series
Periodic Functions
If u(t) has fundamental period T and fundamental frequency F = T1 then,
⊲ Fourier Series in most cases, we can express it as a (possibly infinite) sum of sine and
Why Sin and Cos
Waves? cosine waves with frequencies 0, F , 2F , 3F , · · · .
Fourier Analysis T
Trigonometric
Products
Fourier Analysis
u(t) =
Fourier Analysis T
Example
Linearity sin 2πF t [b1 = 1]
Summary
T/2
−0.4 sin 2π2F t [b2 = −0.4]

T/3
+0.4 sin 2π3F t [b3 = 0.4]

T/4
−0.2 cos 2π4F t [a4 = −0.2]

The Fourier series for u(t) is
a0 P∞
u(t) = 2 + n=1 (an cos 2πnF t + bn sin 2πnF t)
The Fourier coefficients of u(t) are a0 , a1 , · · · and b1 , b2 , · · · .
The nth harmonic of the fundamental is the component at a frequency nF .
Why Sin and Cos Waves?
2: Fourier Series
Periodic Functions
Why are engineers obsessed with sine waves?
Fourier Series Answer: Because ...
Why Sin and Cos
⊲ Waves?
Dirichlet Conditions 1. A sine wave remains a sine wave of the same frequency when you
Fourier Analysis
Trigonometric
(a) multiply by a constant,
Products
Fourier Analysis
(b) add onto to another sine wave of the same frequency,
Fourier Analysis (c) differentiate or integrate or shift in time
Example
Linearity
Summary
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
(a) composed entirely of constant-multiply/add/differentiate
(b) linear: u(t) → x(t) and v(t) → y(t)
means that u(t) + v(t) → x(t) + y(t)
⇒ sum of sine waves → sum of sine waves
1
In these lectures we will use T for the fundamental period and F = T for
the fundamental frequency.

2: Fourier Series
Periodic Functions
Not all u(t) can be expressed as a Fourier Series.
Fourier Series
Why Sin and Cos Peter Dirichlet derived a set of sufficient conditions.
Waves?
Dirichlet The function u(t) must satisfy:
⊲ Conditions
Fourier Analysis • periodic and single-valued
Trigonometric RT
Products • 0
|u(t)| dt < ∞
Fourier Analysis
Fourier Analysis • finite number of maxima/minima per period
Example Peter Dirichlet
Linearity • finite number of finite discontinuities per
Summary period 1805-1859
Good:

sin(t) t2 quantized
Bad:

1

tan (t) sin t ∞ halving steps

Fourier Analysis
2: Fourier Series
Periodic Functions
Suppose that u(t) satisfies the Dirichlet conditions so that
Fourier Series
u(t) = a20 + ∞
P
n=1 (an cos 2πnF t + bn sin 2πnF t)
Why Sin and Cos
Waves?
⊲ Fourier Analysis Question: How do we find an and bn ?
Trigonometric
Products
Fourier Analysis
Answer: We use a clever trick that relies on taking averages.
Fourier Analysis
Example hx(t)i equals the average of x(t) over any integer number of periods:
Linearity
Summary 1 T
R
hx(t)i = T t=0 x(t)dt
Remember, for any integer n, hsin 2πnF ti = 0 (

0 n 6= 0
hcos 2πnF ti =
1 n=0
Finding an and bn is called Fourier analysis.

Trigonometric Products
2: Fourier Series
Periodic Functions
sin(x ± y) = sin x cos y ± cos x sin y
1 1
Fourier Series ⇒ sin x cos y = 2 sin(x + y) + 2 sin(x − y)
Why Sin and Cos
Waves?
cos(x ± y) = cos x cos y ∓ sin x sin y
Fourier Analysis ⇒ cos x cos y = 12 cos(x + y) + 21 cos(x − y)
Trigonometric
⊲ Products sin x sin y = 12 cos(x − y) − 12 cos(x + y)
Fourier Analysis
Fourier Analysis
Example
Set x = 2πmF t, y = 2πnF t (with m + n 6= 0) and take time-averages:
Linearity
Summary
• hsin (2πmF

1 t) cos (2πnF t)i
1

= 2 sin(2π (m + n) F t) + 2 sin(2π (m − n) F t) = 0
• hcos (2πmF t) cos (2πnF t)i (

1
1 0 m 6= n
= 2
cos(2π (m + n) F t) + 2
cos(2π (m − n) F t) = 1
2
m=n
• hsin (2πmF t) sin (2πnF t)i (

1
1 0 m 6= n
= 2
cos(2π (m − n) F t) − 2
cos(2π (m + n) F t) = 1
2
m=n
Summary: hsin cosi = 0 [provided that m + n 6= 0]

1
hsin sini = hcos cosi = 2 if m = n or otherwise = 0.
[Trigonometric Products Proofs]
1 1
Proof that cos x cos y = 2
cos(x + y) + 2
cos(x − y)
We know that
cos(x + y) = cos x cos y − sin x sin y
cos(x − y) = cos x cos y + sin x sin y
Adding these two gives
cos(x + y) + cos(x − y) = 2 cos x cos y
From which: cos x cos y = 12 cos(x + y) + 21 cos(x − y)
Subtracting instead of adding gives: sin x sin y = 12 cos(x − y) − 1
cos(x + y)
2
(
0 m 6= n
Proof that 21 cos(2π (m + n) F t) + 12 cos(2π (m − n) F t) = 1

2
m=n
We are assuming that m and n are integers with m + n 6= 0 and we use the result that hcos 2πf ti is
zero unless f = 0 in which case hcos 2π0ti = 1 . The frequency of the first term, cos(2π (m + n) F t),
is (m + n) F which is definitely non-zero (because of our assumption that m + n 6= 0) and so the
average of this cosine wave is zero. The frequency of the second term is (m − n) F and this is zero
only if m = n. So it follows that the entire expression is zero unless m = n in which case the second
term gives a value of 12 . Since m and n are integers, we can take the averages over a time interval T
and be sure that this includes an integer number of periods for both terms.
E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – note 1 of slide 7
Fourier Analysis
2: Fourier Series a0 P∞
Periodic Functions
Find an and bn in u(t) = 2 + (an cos 2πnF t + bn sin 2πnF t)
n=1
Fourier Series
2 T
Answer: an = 2 hu(t) cos (2πnF t)i , T 0 u(t) cos (2πnF t) dt
Why Sin and Cos
R
Waves?
Dirichlet Conditions 2 T
bn = 2 hu(t) sin (2πnF t)i , T 0 u(t) sin (2πnF t) dt
R
Fourier Analysis
Trigonometric
Products a0
⊲ Fourier Analysis
Proof [a0 ]: 2 hu(t) cos (2π0F t)i = 2 hu(t)i = 2 × 2 = a0
Fourier Analysis
Example
Proof [an , n > 0]:
Linearity 2 hu(t)
cos (2πnF t)i P
Summary
= 2 a20 cos (2πnF t) + ∞ r=1 2 har cos (2πrF t) cos (2πnF t)i
+ ∞
P
r=1 2 hbr sin (2πrF t) cos (2πnF t)i

a
Term 1: 2 2 cos (2πnF t) = 0
0
(
an r = n
Term 2: 2 har cos (2πrF t) cos (2πnF t)i =
0 r 6= n
P∞
⇒ r=1 2 har cos (2πrF t) cos (2πnF t)i = an
Term 3: 2 hbr sin 2πrF t cos (2πnF t)i = 0
Proof [bn , n > 0]: same method as for an
Fourier Analysis Example
2: Fourier Series
Periodic Functions
Truncated Series:
uN (t) = a20 + N
P
n=1 (an cos 2πnF t + bn sin 2πnF t)
Fourier Series
Why Sin and Cos
Waves?
Pulse: T = 20, width W = T4 , height A = 8
10
5
Fourier Analysis
0
Trigonometric
2 T
an = T 0 u(t) cos 2πnt
Products
R 0 5 10 15 20 25
Fourier Analysis T dt 10
N=0
W
= T2 0 A cos 2πnt
Fourier Analysis
R 5
⊲ Example T dt 0
2πnt W
0 5 10 15 20 25
Linearity 2AT

Summary = 2πnT sin T 0 10
N=1
5
A A nπ
= nπ sin 2πnW
T = nπ sin 2 0
0 5 10 15 20 25
2 T 10
bn = T 0 u(t) sin 2πnt

R
T dt
N=2
5
2πnt W
0
2AT

= 2πnT − cos T 0 0 5 10 15 20 25
10
A nπ
N=5
= nπ 1 − cos 2 5
0
0 5 10 15 20 25
n 0 1 2 3 4 5 6 10
N=10
5
8 −8 8
an 4 π 0 3π 0 5π 0 0
0 5 10 15 20 25
10
8 16 8 8 16 N=20
bn π 2π 3π 0 5π 6π
5
0
0 5 10 15 20 25

[Small Angle Approximation]
In the previous example, we can obtain a0 by noting that a20 = hu(t)i, the average value of the
waveform which must be AW T
= 2. From this, a0 = 4. We can, however, also derive this value from
the general expression.
A
The expression for am is am = nπ sin nπ
2
. For the case, n = 0, this is difficult to evaluate because both
the numerator and denominator are zero. The general way of dealing with this situation is L’Hôpital’s
rule (see section 4.7 of RHB) but here we can use a simpler and very useful technique that is often
referred to as the “small angle approximation”. For small values of θ we can approximate the standard
trigonometrical functions as: sin θ ≈ θ, cos θ ≈ 1 − 0.5θ 2 and tan θ ≈ θ. These approximations are
obtained by taking the first three terms of the Taylor series; they are accurate to O(θ 3 ) and are exactly
correct when θ = 0. When m = 0 we can therefore make an exact approximation to an by writing
A
an = nπ sin nπ
2
≈ nπA
× nπ2
= A2
= 4. What we have implicitly done here is to assume that n is a
real number (instead of an integer) and then taken the limit of an as n → 0.
E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – note 1 of slide 9
Linearity
2: Fourier Series
Periodic Functions
Fourier analysis maps a function of time onto a set of Fourier coefficients:
Fourier Series u(t) → {an , bn }
Why Sin and Cos
Waves?
Dirichlet Conditions This mapping is linear which means:
Fourier Analysis
Trigonometric
(1) For any γ, if u(t) → {an , bn } then γu(t) → {γan , γbn }
Products
Fourier Analysis
(2) If u(t) → {an , bn } and u′ (t) → {a′n , b′n } then
Fourier Analysis (u(t) + u′ (t)) → {an + a′n , bn + b′n }
Example
⊲ Linearity
Proof for an : (proof for bn is similar)
Summary
2 T
R
(1) If T 0 u(t) cos (2πnF t) dt = an , then
2 T
R
T 0 (γu(t)) cos (2πnF t) dt
T
= γ T2 0 u(t) cos (2πnF t) dt = γan
R
2 T
R
(2) If T 0 u(t) cos (2πnF t) dt = an and
2 T ′
R ′
T 0 u (t) cos (2πnF t) dt = a n then
2 T
R
T 0 (u(t) + u′ (t)) cos (2πnF t) dt
2 T 2 T ′
R R
= T 0 u(t) cos (2πnF t) dt + T 0 u (t) cos (2πnF t) dt
= an + a′n

Summary
2: Fourier Series
Periodic Functions
• Fourier Series: P
Fourier Series u(t) = a20 + ∞n=1 (an cos 2πnF t + bn sin 2πnF t)
Why Sin and Cos
Waves?
Dirichlet Conditions • Dirichlet Conditions: sufficient conditions for FS to exist
Fourier Analysis
Trigonometric
◦ Periodic, Single-valued, Bounded absolute integral
Products
Fourier Analysis
◦ Finite number of (a) max/min and (b) finite discontinuities
Fourier Analysis
Example • Fourier Analysis = “finding an and bn ”
Linearity
⊲ Summary
◦ an = 2 hu(t) cos (2πnF t)i
2 T
, T 0 u(t) cos (2πnF t) dt
R
◦ bn = 2 hu(t) sin (2πnF t)i

2 T
, T 0 u(t) sin (2πnF t) dt
R
• The mapping u(t) → {an , bn } is linear

For further details see RHB 12.1 and 12.2.

3: Complex
⊲ Fourier Series
Euler’s Equation
Complex Fourier
Series
Averaging Complex
Exponentials
Complex Fourier
Analysis
Fourier Series ↔
Complex Fourier
Series
Complex Fourier
Analysis Example 3: Complex Fourier Series
Time Shifting
Even/Odd Symmetry
Antiperiodic ⇒ Odd
Harmonics Only
Symmetry Examples
Summary
E1.10 Fourier Series and Transforms (2014-5543) Complex Fourier Series: 3 – 1 / 12

Euler’s Equation
3: Complex Fourier
Series Euler’s Equation: eiθ = cos θ + i sin θ [see RHB 3.3]
⊲ Euler’s Equation
eiθ +e−iθ
Complex Fourier
Series
Hence: cos θ = 2 = 12 eiθ + 21 e−iθ
Averaging Complex iθ
e −e−iθ
Exponentials sin θ = 2i = − 12 ieiθ + 21 ie−iθ
Complex Fourier
Analysis
Fourier Series ↔
Complex Fourier
Series
Most maths becomes simpler if you use eiθ instead of cos θ and sin θ
Complex Fourier
Analysis Example
Time Shifting
Even/Odd Symmetry
The Complex Fourier Series is the Fourier Series but written using eiθ
Harmonics Only
Symmetry Examples Examples where using eiθ makes things simpler:
Summary
Using eiθ Using cos θ and sin θ

ei(θ+φ) = eiθ eiφ cos (θ + φ) = cos θ cos φ − sin θ sin φ
1 1
eiθ eiφ = ei(θ+φ) cos θ cos φ = 2 cos (θ + φ) + 2 cos (θ − φ)
d iθ d
dθ e = ieiθ dθ cos θ = − sin θ

Complex Fourier Series
3: Complex Fourier a0 P∞
Series Fourier Series: u(t) = 2 + n=1 (an cos 2πnF t + bn sin 2πnF t)
Euler’s Equation
Complex Fourier Substitute: cos θ = 21 eiθ + 21 e−iθ and sin θ = − 12 ieiθ + 12 ie−iθ
⊲ Series
Averaging Complex
Exponentials P∞ 1 iθ 1 −iθ
a0
bn − 21 ieiθ 1 −iθ

Complex Fourier u(t) = 2 + n=1 an 2e + 2e + + 2 ie
Analysis
Fourier Series ↔ P∞
a0 1 1 i2πnF t

Complex Fourier = 2 +
n=1 2 a n − 2 ibn e [θ = 2πnF t]
Series
Complex Fourier
P∞ 1 1
−i2πnF t
Analysis Example + n=1 2 an + 2 ibn e
Time Shifting P∞ i2πnF t
Even/Odd Symmetry = n=−∞ Un e
Harmonics Only
Symmetry Examples where [b0 , 0]
Summary 
1 1
 2 an − 2 ibn
 n≥1
1
Un = 12 a0

n=0 ⇔ U±n = 2 a|n| ∓ ib|n|
1
 1
2 a |n| + 2 ib|n| n ≤ −1
∗
The Un are normally complex except for U0 and satisfy Un = U−n
P∞ i2πnF t
Complex Fourier Series: u(t) = n=−∞ Un e [simpler ,]
Averaging Complex Exponentials
3: Complex Fourier T n
Series If x(t) has period n for some integer n (i.e. frequency T = nF ):
Euler’s Equation
Complex Fourier 1 T
hx(t)i , T t=0 x(t)dt
R
Series
Averaging Complex
⊲ Exponentials
This is the average over an integer number of cycles.
Complex Fourier
Analysis
Fourier Series ↔
Complex Fourier
For a complex exponential:
Series
i2πnF t
Complex Fourier e = hcos (2πnF t) + i sin (2πnF t)i
Analysis Example
Time Shifting = hcos (2πnF t)i + i hsin (2πnF t)i
Even/Odd Symmetry (
Harmonics Only 1 + 0i n = 0
Symmetry Examples =
Summary 0 + 0i n 6= 0
Hence:
(

i2πnF t 1 n=0
e = [,]
0 n=
6 0

Complex Fourier Analysis
P∞
3: Complex Fourier
Series Complex Fourier Series: u(t) = n=−∞ Un ei2πnF t
Euler’s Equation
Complex Fourier
Series
To find the coefficient, Un , we multiply by something that makes all the
Averaging Complex terms involving the other coefficients average to zero.
Exponentials
Complex Fourier

−i2πnF t

P∞ i2πrF t −i2πnF t

⊲ Analysis u(t)e = r=−∞ Ur e e
Fourier Series ↔
Complex Fourier
P∞ i2π(r−n)F t

Series = r=−∞ Ur e
Complex Fourier
Analysis Example P∞
i2π(r−n)F t
Time Shifting = r=−∞ Ur e
Even/Odd Symmetry
Harmonics Only
All terms in the sum are zero, except for the one when n = r which equals
Symmetry Examples Un :
Summary
−i2πnF t

Un = u(t)e [,]
This shows that the Fourier series coefficients are unique: you cannot have
two different sets of coefficients that result in the same function u(t).
Note the sign of the exponent: “+” in the Fourier Series but “−” for
Fourier Analysis (in order to cancel out the “+”).

Fourier Series ↔ Complex Fourier Series
3: Complex Fourier a0 P∞
Series u(t) = + n=1 (an cos 2πnF t + bn sin 2πnF t)
2
= ∞
Euler’s Equation P i2πnF t
Complex Fourier n=−∞ Un e
Series
Averaging Complex
Exponentials
We can easily convert between the two forms.
Complex Fourier
Analysis
Fourier Series ↔
Complex Fourier
Fourier Coefficients → Complex Fourier Coefficients:
⊲ Series
1 ∗

Complex Fourier
Analysis Example
U±n = 2 a|n| ∓ ib|n| [Un = U−n ]
Time Shifting
Even/Odd Symmetry
Antiperiodic ⇒ Odd Complex Fourier Coefficients → Fourier Coefficients:
Harmonics Only
Symmetry Examples
Summary
an = Un + U−n = 2ℜ (Un ) [ℜ = “real part”]
bn = i (Un − U−n ) = −2ℑ (Un ) [ℑ = “imaginary part”]
The formula for an works even for n = 0.

[Complex functions of time]
In these lectures, we are assuming that u(t) is a periodic real-valued function of time. In this case we
can represent u(t) using either the Fourier Series or the Complex Fourier Series:
a0 P∞ P∞
u(t) = 2
+ n=1 (an cos 2πnF t + bn sin 2πnF t) = n=−∞ Un ei2πnF t
We have seen that the Un coefficients are complex-valued and that Un and U−n are complex conjugates
so that we can write U−n = Un∗ .
In fact, the complex Fourier series can also be used when u(t) is a complex-valued function of time
(this is sometimes
Puseful in the fields of communications and signal processing). In this case, it is still
true that u(t) = n=−∞ Un ei2πnF t , but now Un and U−n are completely independent and normally
∞
U−n 6= Un∗ .
E1.10 Fourier Series and Transforms (2014-5543) Complex Fourier Series: 3 – note 1 of slide 6
Complex Fourier Analysis Example
3: Complex Fourier T 10
Series T = 20, width W = 4, height A = 8 5

W
A
Euler’s Equation 0
Complex Fourier
Series
Method 1: 0 5 10 15 20 25
n an bn Un
Averaging Complex
Exponentials U±n = 21 an ∓ i 12 bn
8
Complex Fourier −6 i 6π
Analysis Method 2: 4 4
Fourier Series ↔ −5 5π + i 5π
Complex Fourier −i2πnF t

Series Un = u(t)e
Complex Fourier
−4 0
⊲ Analysis Example 1 T
= T 0 u(t)e−i2πnF t dt
R
−4 4
Time Shifting −3 3π + i 3π
Even/Odd Symmetry 8
1 W −2
= T 0 Ae−i2πnF t dt i 2π
R
Harmonics Only 4 4
Symmetry Examples A
−i2πnF t W −1 + i
Summary
= −i2πnF T e 0
π π
A −i2πnF W
0 4 2
= i2πn 1 − e 8 8 4 −4
1 π π π +i π
Ae−iπnF W −iπnF W
eiπnF W

= i2πn −e 2 0 16
i −8
2π 2π
Ae−iπnF W 3 −8 8 −4
+ i −4
= nπ sin (nπF W ) 3π 3π 3π 3π
8 nπ
−i nπ
= nπ sin 4 e
4 4 0 0 0
8 8 4 −4
5 5π 5π 5π + i 5π
16 −8
6 0 6π i 6π
Time Shifting
P∞
3: Complex Fourier
Series Complex Fourier Series: u(t) = n=−∞ Un ei2πnF t
Euler’s Equation
Complex Fourier
Series
If v(t) is the same as u(t) but delayed by a time τ : v(t) = u(t − τ )
Averaging Complex P∞ P∞
i2πnF (t−τ ) −i2πnF τ
i2πnF t
Exponentials v(t) = n=−∞ Un e = n=−∞ Un e e
Complex Fourier P∞
Analysis
Fourier Series ↔
= n=−∞ Vn ei2πnF t
Complex Fourier
Series where Vn = Un e−i2πnF τ
Complex Fourier
Analysis Example
⊲ Time Shifting
Example:
5
Even/Odd Symmetry u(t) = 6 cos (2πF t) 0
u(t)
Harmonics Only Fourier: a1 = 6, b1 = 0 -5
0 0.5 1 1.5 2
Symmetry Examples
Summary Complex: U±1 = 21 a1 ∓ 12 ib1 = 3
v(t) = 6 sin (2πF t) = u(t − τ ) 5
v(t)
0
Time delay: τ = T4 ⇒ F τ = 14 -5
0 0.5 1 1.5 2
−i π
Complex: V1 = U1 e 2 = −3i
π
V−1 = U−1 ei 2 = +3i
Note: If u(t) is a sine wave, U1 equals half the corresponding phasor.

Even/Odd Symmetry
3: Complex Fourier ∗
Series
(1) u(t) real-valued ⇔ Un conjugate symmetric [Un = U−n ]
Euler’s Equation (2) u(t) even [u(t) = u(−t)] ⇔ Un even [Un = U−n ]
Complex Fourier
Series (3) u(t) odd [u(t) = −u(−t)] ⇔ Un odd [Un = −U−n ]
Averaging Complex ∗
Exponentials (1)+(2) u(t) real & even ⇔ Un real & even [Un = U−n = U−n ]
Complex Fourier ∗
Analysis (1)+(3) u(t) real & odd ⇔ Un imaginary & odd [Un = U−n = −U−n ]
Fourier Series ↔
Complex Fourier
Series
Complex Fourier
Proof of (2): u(t) even ⇒ Un even
1 T
U−n = T 0 u(t)e−i2π(−n)F t dt
Analysis Example
R
Time Shifting
Even/Odd
1 −T
= T x=0 u(−x)e−i2πnF x (−dx)
R
⊲ Symmetry [substitute x = −t]
Harmonics Only 1 0
= T x=−T u(−x)e−i2πnF x dx
R
Symmetry Examples [reverse the limits]
Summary
1 0
= T x=−T u(x)e−i2πnF x dx = Un
R
[even: u(−x) = u(x)]
Proof of (3): u(t) odd ⇒ Un odd

Same as before, except for the last line:
1 0
= T x=−T −u(x)e−i2πnF x dx = −Un
R
[odd: u(−x) = −u(x)]

Antiperiodic ⇒ Odd Harmonics Only
3: Complex Fourier
Series A waveform, u(t), is anti-periodic if u(t + 21 T ) = −u(t).
Euler’s Equation
Complex Fourier
If u(t) is anti-periodic then Un = 0 for n even.
Series
Averaging Complex Proof:
Exponentials
T
Complex Fourier
Analysis
Define v(t) = u(t + 2 ), then
Fourier Series ↔
Complex Fourier
Series
(1) v(t) = −u(t) ⇒ Vn = −Un
Complex Fourier
Analysis Example (2) v(t) equals u(t) but delayed by − T2
Time Shifting
Even/Odd Symmetry
(
Antiperiodic ⇒ T Un n even
Odd Harmonics ⇒ Vn = Un ei2πnF 2 = Un einπ =
⊲ Only −Un n odd
Symmetry Examples
Summary
Hence for n even: Vn = −Un = Un ⇒ Un = 0
Example:
U0:5 = [0, 3 + 2i, 0, i, 0, 1] U[0:5]=[0, 3+2j, 0, j, 0, 1]
Odd harmonics only ⇔ 5

0
Second half of each period is the -5
negative of the first half. -1 -0.5 0 0.5 1

Symmetry Examples
3: Complex Fourier ∗
Series
All these examples assume that u(t) is real-valued ⇔ U−n = U+n .
Euler’s Equation
Complex Fourier
Series (1) Even u(t) ⇔ real Un U[0:2]=[0, 2, -1]
Averaging Complex 2
Exponentials 0
Complex Fourier
U0:2 = [0, 2, −1] -2
-4
Analysis
-6
Fourier Series ↔ -1 -0.5 0 0.5 1
Complex Fourier
Series
Complex Fourier
(2) Odd u(t) ⇔ imaginary Un U[0:3]=[0, -2j, j, j]
5
Analysis Example
Time Shifting U0:3 = [0, −2i, i, i] 0
Even/Odd Symmetry
-5
-1 -0.5 0 0.5 1
Harmonics Only
Symmetry (3) Anti-periodic u(t) U[0:1]=[0, -j]
⊲ Examples
2
Summary ⇔ odd harmonics only 0
U0:1 = [0, −i] -2

-1 -0.5 0 0.5 1
(4) Even harmonics only 8

U[0:4]=[2, 0, 2, 0, 1]
⇔ period is really 12 T 6
4
2
U0:4 = [2, 0, 2, 0, 1] 0
-1 -0.5 0 0.5 1

Summary
3: Complex Fourier
Series
• Fourier Series: P
Euler’s Equation u(t) = a20 + ∞ n=1 (an cos 2πnF t + bn sin 2πnF t)
Complex Fourier
Series P∞
Averaging Complex • Complex Fourier Series: u(t) = n=−∞ Un ei2πnF t
Exponentials
−i2πnF t 1 T
, T 0 u(t)e−i2πnF t dt
Complex Fourier

R
Analysis ◦ Un = u(t)e
Fourier Series ↔ ∗
Complex Fourier ◦ Since u(t) is real-valued, Un = U−n
Series
Complex Fourier
Analysis Example
◦ FS→CFS: U±n = 21 a|n| ∓ i 21 b|n|
Time Shifting
Even/Odd Symmetry
◦ CFS→FS: an = Un + U−n
Harmonics Only bn = i (Un − U−n )
Symmetry Examples
⊲ Summary • u(t) real and even ⇔ u(−t) = u(t)
⇔ Un is real-valued and even ⇔ bn = 0 ∀n
• u(t) real and odd ⇔ u(−t) = −u(t)
⇔ Un is purely imaginary and odd ⇔ an = 0 ∀n
• u(t) anti-periodic ⇔ u(t + T2 ) = −u(t)
⇔ odd harmonics only ⇔ a2n = b2n = U2n = 0 ∀n
For further details see RHB 12.3 and 12.7.
4: Parseval’s
Theorem and
⊲ Convolution
Parseval’s Theorem
(a.k.a. Plancherel’s
Theorem)
Power Conservation
Magnitude Spectrum
and Power Spectrum
Product of Signals
Convolution
Properties
Convolution Example
Convolution and
4: Parseval’s Theorem and Convolution
Polynomial
Multiplication
Summary
E1.10 Fourier Series and Transforms (2014-5543) Parseval and Convolution: 4 – 1 / 9

Parseval’s Theorem (a.k.a. Plancherel’s Theorem)
4: Parseval’s
Theorem and Suppose we have two signals with the same period, T = F1 ,
Convolution
u(t) = ∞ i2πnF t
P
n=−∞ Un e
Parseval’s
Theorem (a.k.a.
Plancherel’s
P∞
⊲ Theorem) ⇒ u (t) = n=−∞ Un∗ e−i2πnF t
∗
[u(t) = u∗ (t) if real]
Power Conservation P∞
Magnitude Spectrum
and Power Spectrum
v(t) = n=−∞ Vn ei2πnF t
Product of Signals
Convolution Now multiply u∗ (t) and v(t) together and take the average over [0, T ].
Properties
Convolution Example
[Use different “dummy variables”, n and m, so they don’t get mixed up]
Convolution and
P∞ P∞
∗ ∗ −i2πnF t i2πmF t

Polynomial
Multiplication
hu (t)v(t)i = n=−∞ Un e m=−∞ Vm e
Summary
P∞ ∗
P∞
−i2πnF t i2πmF t
= n=−∞ Un m=−∞ Vm e e
P∞ ∗
P∞
i2π(m−n)F t
= n=−∞ Un m=−∞ Vm e
The quantity h· · · i equals 1 if m = n and 0 otherwise, so the only non-zero
element in the second sum is when m = n, so the second sum equals Vn .
P∞
∗
Hence Parseval’s theorem: hu (t)v(t)i = n=−∞ Un∗ Vn
D E P
2 ∞
|u(t)| = n=−∞ Un Un = n=−∞ |Un |2
∗
P∞
If v(t) = u(t) we get:

[Manipulating sums]
If you have a multiplicative expression involving two or more sums, then you must use different dummy
variables for each of the sums:
P P
n af (n) m bg(m)
(1) You can always move any quantities to the right

P P P P
n af (n) m bg(m) = n a m bf (n)g(m)
P P
= n m abf (n)g(m)
(2) You can move quantities to the left past a summation provided that they do not involve the dummy
variable of the summation:
P P P P
n m abf (n)g(m) = n af (n) m bg(m)
P P
6= n af (n)g(m) m b
P
The last expression doesn’t make sense in any case since m is undefined outside m
(3) You can swap the summation order if the sum converges absolutely
P P P P P P
n m h(n, m) = m n h(n, m) provided that n m |h(n, m)| < ∞
The equality on the left is not necessarily true if the sum does not converge absolutely. Of course,
if the sum has only a finite number of terms, it is bound to converge absolutely.
E1.10 Fourier Series and Transforms (2014-5543) Parseval and Convolution: 4 – note 1 of slide 2
Power Conservation
4: Parseval’s
E D
2
Theorem and
Convolution
The average power of a periodic signal is given by Pu , |u(t)| .
This is the average electrical power that would be dissipated if the
Theorem)
Power
signal represents the voltage across a 1 Ω resistor.
⊲ Conservation D E P
Magnitude Spectrum 2 ∞ 2
and Power Spectrum Parseval’s Theorem: Pu = |u(t)| = n=−∞ |Un |
Product of Signals
2
= |U0 | + 2 n=1 |Un |2
Convolution
P∞
Properties
[assume u(t) real]
P∞
Convolution Example
= 14 a20 + 12 n=1 a2n + b2n [U+n = an −ib

Convolution and 2
n
]
Polynomial
Multiplication Parseval’s theorem ⇒ the average power in u(t) is equal to the sum of the
Summary
average powers in each of its Fourier components.
Example: u(t) = 2 + 2 cos 2πF t + 4 sin 2πF t − 2 sin 6πF t
D E
2 1 2 2 2

|u(t)| = 4 + 2 2 + 4 + (−2) = 16
U[0:3]=[2, 1-2j, 0, j] U[0:3]=[2, 1-2j, 0, j]
8 60
6
4 40
u2(t)
u(t)
2 Pu=<u2>=16
0 20
-2
-4 0
-1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1
Time (s) Time (s)
|U0 |2 + 2
P∞ 2
U0:3 = [2, 1 − 2i, 0, i] ⇒ n=1 |Un | = 16
Magnitude Spectrum and Power Spectrum
4: Parseval’s
Theorem and
The spectrum of a periodic signal is the values of {Un } versus nF .
Convolution n q o
The magnitude spectrum is the values of {|Un |} = 21 a2|n| + b2|n| .
Theorem) n o n o
Power Conservation 2
Magnitude The power spectrum is the values of |Un | = 41 a2|n| + b2|n| .
Spectrum and
⊲ Power Spectrum
Product of Signals Example:
Convolution
Properties u(t) = 2 + 2 cos 2πF t + 4 sin 2πF t − 2 sin 6πF t
Convolution Example
Convolution and Fourier Coefficients: a0:3 = [4, 2, 0, 0] b1:3 = [4, 0, −2]
Polynomial
Multiplication Spectrum: U−3:3 = [−i, 0, 1 + 2i, 2, 1 − 2i, 0, i]
Summary √ √
Magnitude Spectrum: |U−3:3 | = 1, 0, 5, 2, 5, 0, 1
2 P
2
Power Spectrum: U−3:3 = [1, 0, 5, 4, 5, 0, 1]
[ = u (t) ]
5
2 Σ=16
|Un|
|U2n|
1
0 0
-3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3
Frequency (Hz) Frequency (Hz)
The magnitude and power spectra of a real periodic signal are symmetrical.
2
A one-sided power power spectrum shows U0 and then 2 |Un | for n ≥ 1.
Product of Signals
4: Parseval’s
Theorem and Suppose we have two signals with the same period, T = F1 ,
Convolution
u(t) = ∞ i2πnF t
P
n=−∞ Un e
Theorem)
P∞
Power Conservation
v(t) = m=−∞ Vn ei2πmF t
Magnitude Spectrum P∞
and Power Spectrum If w(t) = u(t)v(t) then Wr = m=−∞ Ur−m Vm , Ur ∗ Vr
⊲ Product of Signals
Convolution
Properties Proof: P∞ P∞
Convolution Example i2πnF t i2πmF t
Convolution and
w(t) = u(t)v(t) = n=−∞ Un e m=−∞ Vm e
Polynomial P∞ P∞
Multiplication = n=−∞ m=−∞ Un Vm ei2π(m+n)F t
Summary
Now we change the summation variable to use r instead of n:
r =m+n⇒n=r−m
This is a one-to-one mapping: every pair (m, n) in the range ±∞
corresponds to exactly one pair (m, r) in the same range.
P∞ P∞ i2πrF t
P∞
w(t) = r=−∞ m=−∞ Ur−m Vm e = r=−∞ Wr ei2πrF t
P∞
where Wr = m=−∞ Ur−m Vm , Ur ∗ Vr .
Wr is the sum of all products Un Vm for which m + n = r.
The spectrum Wr = Ur ∗ Vr is called the convolution of Ur and Vr .
Convolution Properties
4: Parseval’s
Theorem and
Convolution behaves algebraically like multiplication:
Convolution
Parseval’s Theorem 1) Commutative: Ur ∗ Vr = Vr ∗ Ur
Theorem) 2) Associative: Ur ∗ Vr ∗ Wr = (Ur ∗ Vr ) ∗ Wr = Ur ∗ (Vr ∗ Wr )
Power Conservation
Magnitude Spectrum
3) Distributive over addition: Wr ∗((Ur + Vr ) = Wr ∗ Ur + Wr ∗ Vr
and Power Spectrum
Product of Signals 1 r=0
Convolution
4) Identity Element or “1”: If Ir = , then Ir ∗ Ur = Ur
⊲ Properties 0 r 6= 0
Convolution Example
Convolution and
Polynomial Proofs: (all sums are over ±∞)
Multiplication
Summary 1) Substitute
P for m:Pn = r − m ⇔ m = r − n [1 ↔ 1 for any r]
m Ur−m Vm = n Un Vr−n
2) Substitute for n: k = r + m − n ⇔ n = r + m − k [1 ↔ 1]
P P P P
n (( m Un−m Vm ) Wr−n ) = k (( m Ur−k Vm ) Wk−m )
P P P P
= k m Ur−k Vm Wk−m = k (Ur−k ( m Vm Wk−m ))
P P P
3) m Wr−m (Um + Vm ) = m Wr−m Um + m Wr−m Vm
P
4) Ir−m Um = 0 unless m = r. Hence m Ir−m Um = Ur .

Convolution Example
4: Parseval’s
Theorem and
u(t) = 10 + 8 sin 2πt v(t) = 4 cos 6πt
Convolution
U−1:1 = [4i, 10, −4i] V−3:3 = [2, 0, 0, 0, 0, 0, 2] [0 = V0 ]
Theorem)
4
Power Conservation 2
50
Magnitude Spectrum
w(t)
u(t)
v(t)
10 0 0
and Power Spectrum -2
-50
0 -4
Product of Signals -2 -1 0 1 2 -2 -1 0 1 2 -2 -1 0 1 2
Time (s) Time (s) Time (s)
Convolution 10 2 20
Properties
|Wn|
|Un|
|Vn|
Convolution 5 1 10
⊲ Example 0
-1 0 1
0
-3 -2 -1 0 1 2 3
0
-4 -3 -2 -1 0 1 2 3 4
Convolution and Frequency (Hz) Frequency (Hz) Frequency (Hz)
Polynomial
Multiplication
Summary w(t) = u(t)v(t) = (10 + 8 sin 2πt) 4 cos 6πt
= 40 cos 6πt + 32 sin 2πt cos 6πt
= 40 cos 6πt + 16 sin 8πt − 16 sin 4πt
W−4:4 = [8i, 20, −8i, 0, 0, 0, 8i, 20, −8i]
To convolve Un and Vn :
Replace each harmonic in Vn by a scaled copy of the entire {Un }
(or vice versa) and sum the complex-valued coefficients of any
overlapping harmonics.
Convolution and Polynomial Multiplication
4: Parseval’s
Theorem and Two polynomials: u(x) = U3 x3 + U2 x2 + U1 x + U0
Convolution
Parseval’s Theorem v(x) = V2 x2 + V1 x + V0
Theorem)
Power Conservation Now multiply the two polynomials together:
Magnitude Spectrum
and Power Spectrum w(x) = u(x)v(x)
Product of Signals
Convolution
= U3 V2 x5 + (U3 V1 + U2 V2 ) x4 + (U3 V0 + U2 V1 + U1 V2 ) x3
Properties
Convolution Example
+ (U2 V0 + U1 V1 + U0 V2 ) x2 + (U1 V0 + U0 V1 ) x +U0 V0
Convolution and
⊲
Polynomial
Multiplication
The coefficient of xr consists of all the coefficient pair from U and V where
Summary the subscripts add up to r. For example, for r = 3:
P2
W3 = U3 V0 + U2 V1 + U1 V2 = m=0 U3−m Vm
If all the missing coefficients are assumed to be zero, we can write
P∞
Wr = m=−∞ Ur−m Vm , Ur ∗ Vr
So, to multiply two polynomials, you convolve their coefficient sequences.
Actually, the complex Fourier Series is iust a polynomial:
i2πF t n
P∞ i2πnF t
P∞
u(t) = n=−∞ Un e = n=−∞ Un e

Summary
P∞
4: Parseval’s
Theorem and
∗
• Parseval’s Theorem: hu (t)v(t)i = n=−∞ Un∗ Vn
Convolution D E P
2
◦ Power Conservation: |u(t)| = ∞ n=−∞ |U n | 2
Theorem)
Power Conservation
Magnitude Spectrum
◦ or in terms of an and bn :
and Power Spectrum
D E
2 P∞
|u(t)| = 14 a20 + 1 2
b2n

Product of Signals
2 n=1 an +
Convolution
Properties
Convolution Example • Linearity: w(t) = au(t) + bv(t) ⇔ Wn = aUn + bVn
Convolution and
Polynomial
Multiplication • Product of signals ⇔ Convolution of complex
P∞ Fourier coefficients:
⊲ Summary
w(t) = u(t)v(t) ⇔ Wn = Un ∗ Vn , m=−∞ Un−m Vm
• Convolution acts like multiplication:
◦ Commutative: U ∗ V = V ∗ U
◦ Associative: U ∗ V ∗ W is unambiguous
◦ Distributes over addition: U ∗ (V + W ) = U ∗ V + U ∗ W
◦ Has an identity: Ir = 1 if r = 0 and = 0 otherwise
• Polynomial multiplication ⇔ convolution of coefficients
For further details see RHB Chapter 12.8.

5: Gibbs
⊲ Phenomenon
Discontinuities
Discontinuous
Waveform
Gibbs Phenomenon
Integration
Rate at which
coefficients decrease
with m
Differentiation
Periodic Extension
t2 Periodic
Extension: Method
5: Gibbs Phenomenon
(a)
t2 Periodic
Extension: Method
(b)
Summary
E1.10 Fourier Series and Transforms (2014-5559) Gibbs Phenomenon: 5 – 1 / 11

Discontinuities
5: Gibbs Phenomenon
⊲ Discontinuities
A function, v(t), has a discontinuity of amplitude b at t = a if
Discontinuous
Waveform lime→0 (v(a + e) − v(a − e)) = b 6= 0
Gibbs Phenomenon
Integration
Rate at which Conversely, v(t), is continuous at t = a if the limit, b, equals zero.
with m
Differentiation
Periodic Extension
t2 Periodic
Examples:
Extension: Method b b
(a)
u(t)
v(t)
t2 Periodic
0 0
Extension: Method
a–e a a+e a–e a a+e
(b) Time (t) Time (t)
Summary
Continuous Discontinuous
We will see that if a periodic function, v(t), is discontinuous, then its

Fourier series behaves in a strange way.

Discontinuous Waveform
1
5: Gibbs Phenomenon
Discontinuities
Pulse: T = = 20, width= 12 T , height A = 1
F
1
0.5
Discontinuous 0
⊲ Waveform 1 0.5T
Ae−i2πmF t dt
R 0 5 10 15 20
Gibbs Phenomenon Um = T 0 1
max(u0)=0.500
Integration −i2πmF t 0.5T 0.5
i
Rate at which
= 2πmF T e 0
0
0
N=0
5 10 15 20
with m
i −iπm
((−1)m −1)i
Differentiation = 2πm e −1 = 2πm
1
0.5
max(u1)=1.137
Periodic Extension  0
N=1
t2 Periodic
0 m 6= 0, even 0 5 10 15 20
Extension: Method

1
(a) max(u3)=1.100
t2 Periodic
= 0.5 m = 0 0.5
N=3
0
Extension: Method  −i

(b)
mπ m odd 0 5 10 15 20
1
Summary max(u5)=1.094
0.5
1
So, u(t) = 2 + π2 sin 2πF t + 31 sin 6πF t 0
0
N=5
5 10 15 20
+ 15 sin 10πF t + . . . 1
max(u41)=1.089
0.5
N=41
0
PN i2πmF t 0 5 10 15 20
Define: uN (t) = m=−N Um e 1
max(u41)=1.089
0.5
uN (0) = 0.5 ∀N 0
-1 -0.5 0 0.5 1
1 1 π sin t
R
maxt uN (t) −→ 2 + π 0 t dt ≈ 1.0895 [Enlarged View: u41 (t)]
N →∞

[Powers of −1 and i]
Expressions involving (−1)m or, less commonly, im arise quite frequently and it is worth becoming
familiar with them. They can arise in several guises:
m
e−iπm = eiπm = eiπ = cos (πm) = (−1)m
1 πm
1 m
i2
e = ei 2 π = im
1 πm
m
−i 1

e−i 2
= e 2 π
= (−i)m
As m increases these expressions repeat with periods of 2 or 4. Simple expressions involving these
quantities make useful sequences.
m −4 −3 −2 −1 0 1 2 3 4
(−1)m = cos πm = eiπm 1 −1 1 −1 1 −1 1 −1 1
im = ei0.5πm 1 i −1 −i 1 i −1 −i 1
(−i)m = e−i0.5πm 1 −i −1 i 1 −i −1 i 1
1
2
(1 + (−1)m ) 1 0 1 0 1 0 1 0 1
1
2
(1 − (−1)m ) 0 1 0 1 0 1 0 1 0
1 m
2
(i + (−i)m ) = cos 0.5πm 1 0 −1 0 1 0 −1 0 1
1
4
(1 + (−1)m + im + (−i)m ) 1 0 0 0 1 0 0 0 1
E1.10 Fourier Series and Transforms (2014-5559) Gibbs Phenomenon: 5 – note 1 of slide 3
Gibbs Phenomenon
5: Gibbs Phenomenon PN
Discontinuities Truncated Fourier Series: uN (t) = m=−N Um ei2πmF t
Discontinuous
Waveform If u(t) has a discontinuity of height b at t = a then:
⊲ Gibbs Phenomenon
u(a−e)+u(a+e)
Integration
Rate at which
(1) uN (a) −→ lime→0 2
N →∞
with m
Differentiation
(2) uN (t) has an overshoot of about 9% of b at the discontinuity. For
Periodic Extension large N the overshoot moves closer to the discontinuity but does
t2 Periodic
Extension: Method not get smaller (Gibbs phenomenon). In the limit the overshoot
π
(a)
equals − 21 + π1 0 sint t dt b ≈ 0.0895b.
R
t2 Periodic
Extension: Method −1
(b) (3) For large m, the coefficients, Um decrease no faster than |m| .
Summary
Example:
1
uN (0) −→ 0.5 0.5
N →∞ 0
0 5 10 15 20
maxt uN (t) −→ 1.0895 . . . 1

N →∞ 0.5
max(u41)=1.089
 N=41
0
0
 m 6= 0, even 0 5 10 15 20
1
Um = 0.5 m = 0 0.5
max(u41)=1.089
0
 −i

mπ m odd -1 -0.5 0 0.5 1

[Origin of Gibbs Phenomenon]
This topic is included for interest but is not examinable.
Our first goal is to express the partial Fourier series, uN (t), in terms of the original signal, u(t). We
begin by substituting the integral expression for Un in the partial Fourier series
R
P+N i2πnF t
P+N 1 T τ
uN (t) = n=−N Un e = n=−N T 0 u(τ )e −i2πnF dτ ei2πnF t
Now we swap the order of the integration and the finite summation (OK if the integral converges ∀n)
P
1
RT +N i2πnF (t−τ )
uN (t) = T 0 u(τ ) n=−N e dτ
Now apply the formula for the sum of a geometric progression with z = ei2πF (t−τ ) :
P+N n = z
−N
−z N +1 z −(N +0.5) −z N +0.5
n=−N z 1−z
= z −0.5 −z 0.5
i2π(N +0.5)F (τ −t)
−e−i2π(N +0.5)F (τ −t)
uN (t) = T1 0T u(τ ) e
R
dτ
ei2π0.5F (τ −t) −e−i2π0.5F (τ −t)
sin π(2N +1)F (τ −t)
= T1 0T u(τ )
R
sin πF (τ −t)
dτ
sin((N +0.5)x)
So if we define the Dirichlet Kernel to be DN (x) = sin 0.5x
, and set x = 2πF (τ − t), we obtain
uN (t) = T1 0T u(τ )DN (2πF (τ − t)) dτ
R
So what we have shown is that uN (t) can be obtained by multiplying u(τ ) by a time-shifted Dirichlet
Kernel and then integrating over one period. Next we will look at the properties of the Dirichlet Kernel.
[Dirichlet Kernel]
This topic is included for interest but is not examinable.

sin((N +0.5)x)
Dirichlet Kernel definition: DN (x) = +N
P inx = 1 + 2
PN
n=−N e n=1 cos nx = sin 0.5x
DN (x) is plotted below for N = {2, 5, 10, 21}. The vertical red lines at ±π mark one period.
D2(x) 10 D5(x) 20 D10(x) 40 D21(x)
4
D10(x)
D21(x)
D2(x)
D5(x)
2 5 10 20
0 0 0 0
-6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6 -6 -4 -2 0 2 4 6
x x x x
• Periodic: with period 2π

1
R +π
• Average value: hDN (x)i = 2π −π DN (x)dx = 1
π π π
• First Zeros: DN (x) = 0 at x = ± N +0.5 define the main lobe as − N +0.5 <x< N +0.5
• Peak value: 2N + 1 at x = 0. The main lobe gets narrower but higher as N increases.
• Main Lobe semi-integral:
π π
R N +0.5 R N +0.5 sin((N +0.5)x) 1
Rπ sin y
x=0 DN (x)dx = x=0 sin 0.5x
dx = N +0.5 y=0 sin y dy[y = (N + 0.5)x]
2N +1
where we substituted y = (N + 0.5)x. Now, for large N , we can approximate sin 2Ny+1 ≈ 2Ny+1 :
π
R N +0.5 1
Rπ sin y R π sin y
x=0 D N (x)dx ≈ N +0.5 y=0 y dy = 2 y=0 y dy ≈ 3.7038741 ≈ 2π × 0.58949
2N +1
We see that, for large enough N , the main lobe semi-integral is independent of N .
[In MATLAB DN (x) = (2N + 1) × diric(x, 2N + 1)]
[Gibbs Phenomenon Overshoot]
This topic is included for interest but is not examinable. 1

u41(t)
0.5 T=20
The partial Fourier Series, uN (t), can be obtained by multiplying u(t) 0

-1 -0.5 0 0.5 1 1.5 2
by a shifted Dirichlet Kernel and integrating over one period:
1
RT
uN (t) = T 0 u(τ )DN (2πF (τ − t)) dτ 80 t=0
60
40
For the special case when u(t) is a pulse of height 1 and width 0.5T : 20
0
-20
uN (t) = T1 00.5T DN (2πF (τ − t)) dτ

R -1 -0.5 0 0.5 1 1.5 2
80 t=0.24
60
40
Substitute x = 2πF (τ − t) 20
0
R πF T −2πF t R π−2πF t -20
1 1
uN (t) = 2πF T −2πF t
DN (x) dx = 2π −2πF t DN (x) dx -1 -0.5 0 0.5 1 1.5 2
80 t=0.96
60
• For t = 0 (theRblue integral and the blue circle on the upper graph): 40
1 π 20
uN (0) = 2π 0 DN (x) dx = 0.5
0
-20
-1 -0.5 0 0.5 1 1.5 2
T
• For t = (the red integral and the red circle on the upper graph):
2N +1
π
R π− N +0.5 R0 π
R π− N +0.5
T 1 1 1
uN 2N +1 = 2π − π DN (x) dx= 2π − π DN (x) dx + 2π 0 DN (x) dx
N +0.5 N +0.5
For large N , we replace the first term by a constant (since it is the semi-integral of the main lobe)
and replace the upper limit of theR second term by π:
1 π
≈ 0.58949 + 2π 0 DN (x) dx = 1.08949
• For 0 ≪ t ≪ 0.5T , uN (t) ≈ 1 (the green integral and the green circle on the upper graph).
Integration
P∞
5: Gibbs Phenomenon
Discontinuities
Suppose u(t) = m=−∞ Um ei2πmF t
Discontinuous
Waveform Define v(t) to be the integral of u(t) [boundedness requires U0 = 0]
Gibbs Phenomenon Rt R t P∞
⊲ Integration v(t) = u(τ )dτ = m=−∞ Um e
i2πmF τ
dτ
Rate at which
coefficients decrease P∞ R t i2πmF τ R P
with m = m=−∞ Um e dτ [assume OK to swap and ]
Differentiation P∞ 1
Periodic Extension = c + m=−∞ Um i2πmF ei2πmF t
t2 Periodic P∞
Extension: Method
(a) = c + m=−∞ Vm ei2πmF t where c is an integration constant
t2 Periodic −i
Extension: Method Hence Vm = 2πmF Um except for V0 = c (arbitrary constant)
(b)
Summary
Example:
Square wave: Um = −2i
mπ for odd m (0 for even m)
−i
Triangle wave: Vm = 2πmF × −2i
mπ = −1
π m2 F for odd m (0 for even m)
2
1 5
u7(t) v7(t)
0 0
-1 -5
0 5 10 15 20 0 5 10 15 20
1
Convergence: v(t) always converges if u(t) does since Vm ∝ m Um
vN (t) is a good approximation even for small N

Rate at which coefficients decrease with m
5: Gibbs Phenomenon
Discontinuities
Square wave: Um = −2i
π m −1
for odd m (0 for even m)
Discontinuous
Waveform
Triangle wave: Vm = π−1
2F m
−2
for odd m (0 for even m)
Gibbs Phenomenon 1 5
u7(t) v7(t)
Integration
0 0
Rate at which
coefficients -1 -5
⊲ decrease with m 0 5 10 15 20 0 5 10 15 20 Integrating
Differentiation −i
Periodic Extension u(t) multiplies the Um by 2πF × m−1 ⇒ they decrease faster.
t2 Periodic
Extension: Method
(a)
The rate at which the coefficients, Um , decrease with m depends on the
t2 Periodic
Extension: Method lowest derivative that has a discontinuity:
(b)
Summary • Discontinuity in u(t) itself (e.g. square wave)
−1
For large |m|, Um decreases as |m|
• Discontinuity in u′ (t) (e.g. triangle wave)
For large |m|, Um decreases as |m|−2
• Discontinuity in u(n) (t)
−(n+1)
For large |m|, Um decreases as |m|
• No discontinuous derivatives
For large |m|, Um decreases faster than any power (e.g. e−|m| )

Differentiation
5: Gibbs Phenomenon −i
Discontinuities
Integration multiplies Um by 2πmF .
Discontinuous
Waveform 2πmF
Gibbs Phenomenon
Hence differentiation multiplies Um by −i = i2πmF
Integration
Rate at which
If u(t) is a continuous differentiable function and w(t) = du(t)
dt then,
with m
⊲ Differentiation
provided that w(t) satisfies the Dirichlet conditions, its Fourier coefficients
Periodic Extension are: (
t2 Periodic
Extension: Method 0 m=0
(a) Wm = .
t2 Periodic i2πmF Um m 6= 0
Extension: Method
(b)
Summary Since we are multiplying Um by m the coefficients Wm decrease more
slowly with m and so the Fourier series for w(t) may not converge (i.e.
w(t) may not satisfy the Dirichlet conditions).
d d
dt dt
−→ −→
−2 −1 −0
Um ∝ |m| Um ∝ |m| Um ∝ |m|
Differentiation makes waveforms spikier and less smooth.

Periodic Extension
5: Gibbs Phenomenon
Discontinuities
Suppose y(t) is only defined over a finite interval (a, b).
Discontinuous
Waveform You have two reasonable choices to make a periodic version:
Gibbs Phenomenon
Integration
Rate at which (a) T = b − a, u(t) = y(t) for a ≤ t < b
coefficients decrease (
with m
Differentiation y(t) a≤t≤b
⊲ Periodic Extension (b) T = 2(b − a), u(t) =
2
t Periodic y(2b − t) b ≤ t ≤ 2b − a
Extension: Method
(a)
t2 Periodic
Example:
Extension: Method
(b) y(t) = t2 for 0 ≤ t < 2
Summary
4 4 4
2 2 2
0 0 0
-2 0 2 4 -2 0 2 4 -2 0 2 4
y(t) (a) T = 2 (b) T = 4

Option (b) has twice the period, no discontinuities, no Gibbs phenomenon
overshoots and if y(t) is continuous the coefficients decrease at least as fast
−2
as |m| .

t2 Periodic Extension: Method (a)
5: Gibbs Phenomenon
Discontinuities
y(t) = t2 for 0 ≤ t < 2
Discontinuous
Waveform
Gibbs Phenomenon
Method (a): T = F1 = 2
Integration
1 T 2 −i2πmF t 1
RT 4
t2 dt =
R
Rate at which
Um = T 0 t e dt U0 = T 0 3
with m h iT
Differentiation 1 t2 e−i2πmF t 2te−i2πmF t 2e−i2πmF t
Periodic Extension = T −i2πmF − (−i2πmF )2
+ (−i2πmF )3 0
t2 Periodic
Extension: Method
⊲ (a) Substitute e−i2πmF 0 = e−i2πmF T = 1 [for integer m]
t2 Periodic
Extension: Method
h 2
i
(b) = T1 −i2πmF
T 2T
− (−i2πmF )2
Summary
2i 2
= πm + π 2 m2
4 4 4
K=1 K=3 K=6
2 2 2
0 0 0
-2 0 2 4 -2 0 2 4 -2 0 2 4
U0:3 = [1.333, 0.203 + 0.637i, 0.051 + 0.318i, 0.023 + 0.212i]

t2 Periodic Extension: Method (b)
5: Gibbs Phenomenon
Discontinuities
y(t) = t2 for 0 ≤ t < 2
Discontinuous
Waveform
Gibbs Phenomenon
Method (b): T = F1 = 4
Integration
1 0.5T 1
R 0.5T 4
Um = T −0.5T t2 e−i2πmF t dt t2 dt =
R
Rate at which
U0 = T −0.5T 3
with m
Differentiation
h 2 −i2πmF t i0.5T
1 t e 2te−i2πmF t −i2πmF t
2e
Periodic Extension = T −i2πmF − (−i2πmF )2 + (−i2πmF )3 −0.5T
t2 Periodic
Extension: Method
m
(a)
Substitute e±iπmF T = e±iπm = (−1) [for integer m]
t2 Periodic
Extension: Method h i
⊲ (b) (−1)m −2T
Summary
= T (−i2πmF )2
[all even powers of t cancel out]
(−1)m T 2 (−1)m 8
= 2π 2 m2 = π 2 m2
4 4 4
K=1 K=3 K=6
2 2 2
0 0 0
-2 0 2 4 -2 0 2 4 -2 0 2 4
U0:3 = [1.333, −0.811, 0.203, −0.090] [u(t) real+even ⇒ Um real]

Convergence is noticeably faster than for method (a)
Summary
5: Gibbs Phenomenon
Discontinuities
• Discontinuity at t = a
Discontinuous
Waveform
◦ Gibbs phenomenon: uN (t) overshoots by 9% of iump
Gibbs Phenomenon ◦ uN (a) → mid point of iump
Integration Rt −i
Rate at which
• Integration: If v(t) = u(τ )dτ , then Vm = 2πmF Um
with m and V0 = c, an arbitrary constant. U0 must be zero.
Differentiation
Periodic Extension
t2 Periodic
• Differentiation: If w(t) = du(t)
dt , then Wm = i2πmF Um provided
Extension: Method
(a)
w(t) satisfies Dirichlet conditions (it might not)
t2 Periodic
Extension: Method
• Rate of decay:
(b) −(k+1) dk u(t)
⊲ Summary
◦ For large n, Un decreases at a rate |n| where dtk
is
the first discontinuous
D derivative
E P
2
◦ Error power: (u(t) − uN (t)) = |n|>N |Un |2
• Periodic Extension of finite domain signal of length L
◦ (a) Repeat indefinitely with period T = L
◦ (b) Reflect alternate repetitions for period T = 2L
no discontinuities or Gibbs phenomenon
For further details see RHB Chapter 12.4, 12.5, 12.6

6: Fourier
⊲ Transform
Fourier Series as
T →∞
Fourier Transform
Fourier Transform
Examples
Dirac Delta Function
Dirac Delta Function:
Scaling and
Translation
Products and
Integrals
Periodic Signals
6: Fourier Transform
Duality
Time Shifting and
Scaling
Gaussian Pulse
Summary
E1.10 Fourier Series and Transforms (2014-5559) Fourier Transform: 6 – 1 / 12

Fourier Series as T → ∞
P∞
Fourier Series as Fourier Series: u(t) = n=−∞ Un ei2πnF t
⊲ T → ∞
Fourier Transform 1
Fourier Transform The harmonic frequencies are nF ∀n and are spaced F = T apart.
Examples
As T gets larger, the harmonic spacing becomes smaller.
Scaling and
Translation
e.g. T = 1 s ⇒ F = 1 Hz
Dirac Delta Function: T = 1 day ⇒ F = 11.57 µHz
Products and
Integrals
Periodic Signals If T → ∞ then the harmonic spacing becomes zero, the sum becomes an
Duality
Time Shifting and
integral and we get the Fourier Transform:
Scaling R +∞
Gaussian Pulse u(t) = f =−∞ U (f )ei2πf t df
Summary
Here, U (f ), is the spectral density of u(t).

• U (f ) is a continuous function of f .
• U (f ) is complex-valued.
• u(t) real ⇒ U (f ) is conjugate symmetric ⇔ U (−f ) = U (f )∗ .
• Units: if u(t) is in volts, then U (f )df must also be in volts
⇒ U (f ) is in volts/Hz (hence “spectral density”).

Fourier Transform
P∞
Fourier Series as Fourier Series: u(t) = n=−∞ Un ei2πnF t
T →∞
⊲ Fourier Transform The summation is over a set of equally spaced frequencies
Fourier Transform
Examples fn = nF where the spacing between them is ∆f = F = T1 .
Dirac Delta Function R 0.5T
−i2πnF t
= ∆f t=−0.5T u(t)e−i2πnF t dt

Scaling and
Un = u(t)e
Translation
Products and
Spectral Density: If u(t) has finite energy, Un → 0 as ∆f → 0. So we
Un
Integrals
Periodic Signals
define a spectral density, U (fn ) = ∆f , on the set of frequencies {fn }:
Duality R 0.5T
Un
Time Shifting and U (fn ) = ∆f = t=−0.5T u(t)e−i2πfn t dt
Scaling
Gaussian Pulse we can writeP [Substitute Un = U (fn )∆f ]
Summary
u(t) = ∞ n=−∞ U (fn )e
i2πfn t
∆f
Fourier Transform: Now if we take the limit as ∆f → 0, we get
R∞
u(t) = −∞ U (f )ei2πf t df [Fourier Synthesis]
R∞
U (f ) = t=−∞ u(t)e−i2πf t dt [Fourier Analysis]
Un
For non-periodic signals Un → 0 as ∆f → 0 and U (fn ) = ∆f remains
finite. However, if u(t) contains an exactly periodic component, then the
corresponding U (fn ) will become infinite as ∆f → 0. We will deal with it.
Fourier Transform Examples
Fourier Series as
Example(1: 1
a=2
u(t)
0.5
T →∞
Fourier Transform
e−at t ≥ 0 0
-5 0 5
Fourier Transform
u(t) = Time (s)
⊲ Examples 0 t<0 0.5

0.4
|U(f)|
0.3
Dirac Delta Function 0.2
R∞ 0.1
Scaling and U (f ) = −∞ u(t)e−i2πf t dt -5 0
Frequency (Hz)
5
<U(f) (rad/pi)
Translation 0.5
R ∞ −at −i2πf t
Products and = 0 e e dt 0
-0.5
Integrals R ∞ (−a−i2πf )t -5 0 5
Periodic Signals = 0 e dt Frequence (Hz)
Duality (−a−i2πf )t ∞
−1 1
Time Shifting and
Scaling
= a+i2πf e 0
= a+i2πf
Gaussian Pulse 1
a=2
v(t)
Summary 0.5
Example 2: 0
-5 0 5
−a|t|
v(t) = e 1
Time (s)
|V(f)|
R∞ 0.5
V (f ) = −∞ v(t)e−i2πf t dt 0
-5 0 5
Frequency (Hz)
R0 at −i2πf t
R ∞ −at −i2πf t
= −∞ e e dt + 0 e e dt
1
(a−i2πf )t 0 −1
(−a−i2πf )t ∞
= a−i2πf e −∞
+ a+i2πf e 0
1 1 2a
= a−i2πf + a+i2πf = a2 +4π2 f 2 [v(t) real+symmetric
⇒ V (f ) real+symmetric]
Fourier Series as
We define a unit area pulse of width w as 4
δ 0.2(x)
T →∞
(
1 δ 0.5(x)
Fourier Transform
w −0.5w ≤ x ≤ 0.5w 2
δ 3(x)
Fourier Transform dw (x) = 0
Examples 0 otherwise -3 -2 -1 0
x
1 2 3
Dirac Delta
⊲ Function 1
Scaling and
This pulse has the property that its integral equals δ (x)
0.5
Translation 1 for all values of w:
Dirac Delta Function: R∞ 0
Products and
Integrals x=−∞ w
d (x)dx = 1 -3 -2 -1 0
x
1 2 3
Periodic Signals
Duality If we make w smaller, the pulse height increases to preserve unit area.
Time Shifting and
Scaling We define the Dirac delta function as δ(x) = limw→0 dw (x)
Gaussian Pulse
Summary • δ(x) equals zero everywhere except at x = 0 where it is infinite.
R∞
• However its area still equals 1 ⇒ −∞ δ(x)dx = 1
• We plot the height of δ(x) as its area rather than its true height of ∞.
δ(x) is not quite a proper function: it is called a generalized function.

Dirac Delta Function: Scaling and Translation
Fourier Series as
Translation: δ(x − a) 1 δ (x) δ (x-2)
T →∞
Fourier Transform δ(x) is a pulse at x = 0 0
Fourier Transform
Examples δ(x − a) is a pulse at x = a -0.5δ(x+2)
-1
Dirac Delta Function -3 -2 -1 0 1 2 3
Dirac Delta
Function: Scaling
Amplitude Scaling: bδ(x) x
⊲ and Translation
R∞ 1
Dirac Delta Function: δ(x) has an area of 1 ⇔ −∞ δ(x)dx = 1
Products and δ (4x) = 0.25δ (x)
0
Integrals
Periodic Signals
bδ(x)R has an area of b since -3δ(-4x-8) = -0.75δ(x+2)
∞ R∞ -1
Duality
−∞
(bδ(x)) dx = b −∞ δ(x)dx = b -3 -2 -1 0
x
1 2 3
Time Shifting and
Scaling
Gaussian Pulse b can be a complex number (on a graph, we then plot only its magnitude)
Summary
Time Scaling: δ(cx)
R∞ R∞
c > 0: x=−∞ δ(cx)dx = y=−∞ δ(y) dy
c [sub y = cx]
1 ∞
= c y=−∞ δ(y)dy = 1c = |c|
1
R
R∞ R −∞
c < 0: x=−∞ δ(cx)dx = y=+∞ δ(y) dy
c [sub y = cx]
+∞
= −1 δ(y)dy = −1 1
R
c y=−∞ c = |c|
1
In general, δ(cx) = |c| δ(x) for c 6= 0

Dirac Delta Function: Products and Integrals
Fourier Series as
If we multiply δ(x − a) by a function of x: 6
y=x2
T →∞
Fourier Transform
y = x2 × δ(x − 2) 4
Fourier Transform
Examples
The product is 0 everywhere except at x = 2. 2
0
So δ(x − 2) is multiplied by the value taken by -1 0
x
1 2
Scaling and
Translation
x2 at x = 2: 6
y=δ(x-2)
Dirac Delta 2
2 4
Function: x × δ(x − 2) = x x=2 × δ(x − 2) 2
Products and
⊲ Integrals = 4 × δ(x − 2) 0
Periodic Signals -1 0 1 2
x
Duality
Time Shifting and In general for any function, f (x), that is 6
y=22×δ(x-2) = 4δ(x-2)
Scaling
Gaussian Pulse continuous at x = a, 4
Summary 2
f (x)δ(x − a) = f (a)δ(x − a) 0
-1 0 1 2
x
Integrals:
R∞ R∞
−∞
f (x)δ(x − a)dx = f (a)δ(x − a)dx
−∞ R∞
= f (a) −∞ δ(x − a)dx
= f (a) [if f (x) continuous at a]
R∞ 2
2
Example: −∞
3x − 2x δ(x − 2)dx = 3x − 2x x=2 = 8

Periodic Signals
R∞
Fourier Series as Fourier Transform: u(t) = −∞ U (f )ei2πf t df [Fourier Synthesis]
T →∞ R∞
Fourier Transform U (f ) = t=−∞ u(t)e−i2πf t dt [Fourier Analysis]
Fourier Transform
Examples 1.5

Example: U (f ) = 1.5δ(f + 2) + 1.5δ(f − 2) 1
1.5δ(f+2) 1.5δ(f-2)

R∞
Scaling and u(t) = −∞ U (f )ei2πf t df 0.5
0
Translation R∞
Dirac Delta Function: = −∞ 1.5δ(f + 2)ei2πf t df -3 -2 -1 0 1
Frequency (Hz)
2 3
Products and R∞
Integrals + −∞ 1.5δ(f − 2)ei2πf t df 4 3cos(4πt)
⊲ Periodic Signals i2πf t i2πf t 2
Duality = 1.5 e + 1.5 e

f =−2 f =+2
0
Time Shifting and -2
Scaling
Gaussian Pulse
= 1.5 ei4πt + e−i4πt = 3 cos 4πt 0 1 2
Time (s)
3 4 5
Summary
If u(t) is periodic then U (f ) is a sum of Dirac delta functions:
P∞ i2πnF t
P∞
u(t) = n=−∞ Un e ⇒ U (f ) = n=−∞ Un δ (f − nF )
R∞
Proof: u(t) = −∞ U (f )ei2πf t df
R ∞ P∞
= −∞ n=−∞ Un δ (f − nF ) ei2πf t df
P∞ R∞
= n=−∞ Un −∞ δ (f − nF ) ei2πf t df
P∞
= n=−∞ Un ei2πnF t

Duality
R∞
T →∞ R∞
Fourier Transform
Examples
Dirac Delta Function: Dual transform:
Scaling and
Translation Suppose v(t) = U (t), then
Dirac Delta Function: R∞
Products and
Integrals
V (f ) = t=−∞ v(t)e−i2πf t dτ
R∞
Periodic Signals V (g) = t=−∞ U (t)e−i2πgt dt [substitute f = g, v(t) = U (t)]
⊲ Duality
R∞
Time Shifting and
Scaling = f =−∞ U (f )e−i2πgf df [substitute t = f ]
Gaussian Pulse
Summary = u(−g)
So: v(t) = U (t) ⇒ V (f ) = u(−f )
Example:
2
u(t) = e−|t| ⇒ U (f ) = 1+4π 2 f 2 [from earlier]
2
v(t) = 1+4π 2 t2 ⇒ V (f ) = e−|−f | = e−|f |

Time Shifting and Scaling
R∞
T →∞ R∞
Fourier Transform
Examples
Dirac Delta Function: Time Shifting and Scaling:
Scaling and
Translation
Suppose v(t) = u(at + b), then
R∞
Products and
Integrals V (f ) = t=−∞ u(at + b)e−i2πf t dt [now sub τ = at + b]
Periodic Signals
−i2πf ( τ −b
a ) 1 dτ
R∞
Duality
Time Shifting and
= sgn(a) τ =−∞ u(τ )e a
⊲ Scaling (
Gaussian Pulse 1 a>0
Summary note that ±∞ limits swap if a < 0 hence sgn(a) =
−1 a < 0
1 i 2πf b R∞ −i2π fa τ
= |a| e
a
τ =−∞
u(τ )e dτ
b

1 i 2πf f
= |a| e
a U a
b

1 i 2πf f
So: v(t) = u(at + b) ⇒ V (f ) = |a| e
a U a

Gaussian Pulse
6: Fourier Transform t2
√ 1 − 2σ
Fourier Series as
T →∞
Gaussian Pulse: u(t) = 2πσ 2
e 2
Fourier Transform
R∞
Fourier Transform
This is a Normal (or Gaussian) probability distribution, so −∞ u(t)dt = 1.
Examples R∞ R ∞ − t2 −i2πf t
−i2πf t 1
U (f ) = −∞ u(t)e dt = √2πσ2 −∞ e 2σ2 e dt
Scaling and
1
R ∞ − 1 (t2 +i4πσ2 f t)
Translation =√ e 2σ2
2πσ 2 −∞
dt
Products and 2 2
R∞ − 2σ1 2 t2 +i4πσ 2 f t+(i2πσ 2 f ) −(i2πσ 2 f )
Integrals = √ 1 e dt
Periodic Signals 2πσ 2 −∞
2 R ∞ − 1 (t+i2πσ2 f )2
Duality
= e 2σ
1
2 ( i2πσ 2
f ) √ 1
e 2σ 2 dt
Time Shifting and 2 −∞
Scaling 2πσ
(i) 2
⊲ Gaussian Pulse
=e
1
2σ 2
(i2πσ2 f )
=e − 21 (2πσf )2
Summary
[(i) uses a result from complex analysis theory that:
1
R ∞ − 1 (t+i2πσ2 f )2 1
R ∞ − 1 t2
√
2πσ 2 −∞
e 2σ 2
dt = √
2πσ 2 −∞
e 2σ 2 dt = 1]
0.4 σ=1 1 1/(2πσ)=0.159
|U(f)|
u(t)
0.2 0.5
0 0
-4 -2 0 2 4 -0.6 -0.4 -0.2 0 0.2 0.4 0.6
Time (s) Frequency (Hz)
Uniquely, the Fourier Transform of a Gaussian pulse is a Gaussian pulse.

Summary
Fourier Series as
• Fourier Transform: R∞
T →∞
Fourier Transform
◦ Inverse transform (synthesis): u(t) = −∞ U (f )ei2πf t df
Fourier Transform R∞
Examples ◦ Forward transform (analysis): U (f ) = t=−∞ u(t)e−i2πf t dt
Scaling and
⊲ U (f ) is the spectral density function (e.g. Volts/Hz)
Translation
• Dirac Delta Function: R∞
Products and
Integrals
◦ δ(t) is a zero-width infinite-height pulse with −∞ δ(t)dt = 1
Periodic Signals R∞
Duality ◦ Integral: −∞ f (t)δ(t − a) = f (a)
Time Shifting and
1
Scaling
Gaussian Pulse
◦ Scaling: δ(ct) = |c| δ(t)
⊲ Summary
u(t) = ∞ Un ei2πnF t
P
• Periodic Signals: n=−∞
P∞
⇒ U (f ) = n=−∞ Un δ (f − nF )
• Fourier Transform Properties:
◦ v(t) = U (t) ⇒ V (f ) = u(−f )
b
1 i 2πf
◦ v(t) = u(at + b) ⇒ V (f ) = |a| e a U fa
2
√ 1 − 21 ( σt ) 1 2
◦ v(t) = 2πσ 2
e ⇒ V (f ) = e− 2 (2πσf )
For further details see RHB Chapter 13.1 (uses ω instead of f )
7: Fourier
Transforms:
Convolution and
Parseval’s
⊲ Theorem
Multiplication of
Signals
Multiplication
Example
Convolution Theorem
Convolution Example
Convolution
Properties 7: Fourier Transforms: Convolution and
Energy Conservation
Energy Spectrum
Summary
E1.10 Fourier Series and Transforms (2014-5559) Fourier Transform - Parseval and Convolution: 7 – 1 / 10
Multiplication of Signals
7: Fourier
Transforms: Question: What is the Fourier transform of w(t) = u(t)v(t) ?
Convolution and R +∞ R +∞
i2πht
Multiplication of
Let u(t) = h=−∞ U (h)e dh and v(t) = g=−∞ V (g)ei2πgt dg
⊲ Signals
Multiplication [Note use of different dummy variables]
Example
Convolution Theorem w(t) = u(t)v(t)
Convolution Example R +∞ R +∞
i2πht
Convolution
Properties
= h=−∞ U (h)e dh g=−∞ V (g)ei2πgt dg
Parseval’s Theorem R +∞ R +∞
Energy Conservation = h=−∞ U (h) g=−∞ V (g)ei2π(h+g)t dg dh [merge e(··· ) ]
Energy Spectrum
Summary Now we make a change of variable in the second integral: g = f − h
R +∞ R +∞
= h=−∞ U (h) f =−∞ V (f − h)ei2πf t df dh
R∞ R +∞ i2πf t
R
= f =−∞ h=−∞ U (h)V (f − h)e dh df [swap ]
R +∞
= f =−∞ W (f )ei2πf t df
R +∞ R +∞
where W (f ) = h=−∞ U (h)V (f − h)dh h=−∞ U (h)V (f − h)dh ,
U (f ) ∗ V (f )
This is the convolution of the two spectra U (f ) and V (f ).
w(t) = u(t)v(t) ⇔ W (f ) = U (f ) ∗ V (f )
Multiplication Example
7: Fourier
( 1
Transforms: e−at t≥0 a=2
u(t)
0.5
Convolution and
u(t) = 0
Multiplication of
0 t<0 -5 0
Time (s)
5
0.5
Signals 0.4
|U(f)|
0.3
Multiplication 1
⊲ Example U (f ) = a+i2πf [from before] 0.2
0.1
-5 0 5
Convolution Theorem Frequency (Hz)
<U(f) (rad/pi)
0.5
Convolution Example
Convolution v(t) = cos 2πF t 0
-0.5
Properties -5 0 5
Parseval’s Theorem V (f ) = 0.5 (δ(f + F ) + δ(f − F )) 1
Frequence (Hz)
Energy Conservation
v(t)
0
Energy Spectrum
-1
Summary
w(t) = u(t)v(t) -5 0
Time (s)
5
0.4 F=2 0.5 0.5

W (f ) = U (f ) ∗ V (f )
|V(f)|
0.2
0.5 0.5 0
= a+i2π(f +F ) + a+i2π(f −F )
-5 0
Frequency (Hz)
5
1
a=2, F=2
0.5
w(t)
0
If V (f ) consists entirely of Dirac impulses -0.5

-5 0
Time (s)
5
then U (f ) ∗ V (f ) iust replaces each impulse 0.2
|W(f)|
with a complete copy of U (f ) scaled by the 0.1
0
-5 0 5
area of the impulse and shifted so that 0 Hz Frequency (Hz)
<W(f) (rad/pi)
0.5
lies on the impulse. Then add the 0

-0.5
overlapping complex spectra. -5 0
Frequence (Hz)
5
Convolution Theorem
7: Fourier
Transforms: Convolution Theorem:
Convolution and
Parseval’s Theorem w(t) = u(t)v(t) ⇔ W (f ) = U (f ) ∗ V (f )
Multiplication of
Signals
w(t) = u(t) ∗ v(t) ⇔ W (f ) = U (f )V (f )
Multiplication
Example Convolution in the time domain is equivalent to multiplication in the
Convolution
⊲ Theorem frequency domain and vice versa.
Convolution Example
Convolution
Properties
Proof of second line:
Parseval’s Theorem Given u(t), v(t) and w(t) satisfying
Energy Conservation
Energy Spectrum w(t) = u(t)v(t) ⇔ W (f ) = U (f ) ∗ V (f )
Summary
define dual waveforms x(t), y(t) and z(t) as follows:
x(t) = U (t) ⇔ X(f ) = u(−f ) [duality]
y(t) = V (t) ⇔ Y (f ) = v(−f )
z(t) = W (t) ⇔ Z(f ) = w(−f )
Now the convolution property becomes:
w(−f ) = u(−f )v(−f ) ⇔ W (t) = U (t) ∗ V (t) [sub t ↔ ±f ]
Z(f ) = X(f )Y (f ) ⇔ z(t) = x(t) ∗ y(t) [duality]
Convolution Example
7: Fourier
( 1
Transforms: 1−t 0≤t<1
u(t)
0.5
Convolution and
u(t) = 0
Multiplication of
0 otherwise -2 0 2
Time t (s)
4 6
Signals
( 1
Multiplication
e−t t ≥ 0
v(t)
0.5
Example
Convolution Theorem v(t) = 0
-2 0 2 4 6
Convolution 0 t<0 Time t (s)
⊲ Example 0.3
Convolution 0.2
w(t)
Properties 0.1
Parseval’s Theorem w(t) = u(t) ∗ v(t) 0
-2 0 2 4 6
Energy Conservation R∞ Time t (s)
Energy Spectrum = −∞ u(τ )v(t − τ )dτ 1
∫ = 0.307 t=0.7
u(τ) v(0.7-τ)
Summary 0.5
R min(t,1)
= 0 (1 − τ )eτ −t dτ 0
-2 0 2 4 6
Time τ (s)
min(t,1) 1
= [(2 − τ ) eτ −t ]τ =0 0.5
∫ = 0.16
u(τ) v(1.5-τ)
t=1.5
 0
0 t<0 -2 0 2 4 6
Time τ (s)

= 2 − t − 2e−t 0 ≤ t < 1 1
0.5
∫ = 0.059
u(τ) v(2.5-τ)
t=2.5
(e − 2) e−t

t≥1 0

-2 0 2 4 6
Time τ (s)
Note how v(t − τ ) is time-reversed (because of the −τ ) and time-shifted to

put the time origin at τ = t.
Convolution Properties
7: Fourier
R∞
Transforms: Convloution: w(t) = u(t) ∗ v(t) , −∞
u(τ )v(t − τ )dτ
Convolution and
Multiplication of Convolution behaves algebraically like multiplication:
Signals
Multiplication 1) Commutative: u(t) ∗ v(t) = v(t) ∗ u(t)
Example
Convolution Theorem 2) Associative:
Convolution Example
Convolution u(t) ∗ v(t) ∗ w(t) = (u(t) ∗ v(t)) ∗ w(t) = u(t) ∗ (v(t) ∗ w(t))
⊲ Properties
Parseval’s Theorem 3) Distributive over addition:
Energy Conservation
Energy Spectrum w(t) ∗ (u(t) + v(t)) = w(t) ∗ u(t) + w(t) ∗ v(t)
Summary
4) Identity Element or “1”: u(t) ∗ δ(t) = δ(t) ∗ u(t) = u(t)
5) Bilinear: (au(t)) ∗ (bv(t)) = ab (u(t) ∗ v(t))
Proof: In the frequency domain, convolution is multiplication.
Also, if u(t) ∗ v(t) = w(t), then
6) Time Shifting: u(t + a) ∗ v(t + b) = w(t + a + b)
1
7) Time Scaling: u(at) ∗ v(at) = |a| w(at)
How to recognise a convolution integral:
the arguments of u(· · · ) and v(· · · ) sum to a constant.
7: Fourier
Transforms:
Lemma:
Convolution and
X(f ) = δ(f − g) ⇒ x(t) = δ(f − g)ei2πf t df = ei2πgt
R
Multiplication of R i2πgt −i2πf t R i2π(g−f )t
Signals
Multiplication
⇒ X(f ) = e e dt = e dt = δ(g − f )
Example
Convolution Theorem R∞ R +∞
∗ ∗
Convolution Example Parseval’s Theorem: t=−∞
u (t)v(t)dt = f =−∞
U (f )V (f )df
Convolution
Properties
Parseval’s
⊲ Theorem Proof: R +∞ R +∞
Energy Conservation Let u(t) = f =−∞ U (f )ei2πf t df and v(t) = g=−∞
V (g)ei2πgt
dg
Energy Spectrum
Summary
[Note use of different dummy variables]
Now multiply u∗ (t) = u(t) and v(t) together and integrate over time:
R∞ ∗
t=−∞
u (t)v(t)dt
R∞ R +∞ ∗ −i2πf t
R +∞
= t=−∞ f =−∞ U (f )e df g=−∞ V (g)ei2πgt dgdt
R +∞ R +∞ R∞
= f =−∞ U (f ) g=−∞ V (g) t=−∞ ei2π(g−f )t dtdgdf
∗
R +∞ ∗
R +∞
= f =−∞ U (f ) g=−∞ V (g)δ(g − f )dgdf [lemma]
R +∞
= f =−∞ U ∗ (f )V (f )df
Energy Conservation
7: Fourier R∞ ∗
R +∞ ∗
Transforms: Parseval’s Theorem: t=−∞
u (t)v(t)dt = f =−∞
U (f )V (f )df
Convolution and
Multiplication of For the special case v(t) = u(t), Parseval’s theorem becomes:
Signals R∞ R +∞
∗ ∗
Multiplication
t=−∞
u (t)u(t)dt = f =−∞
U (f )U (f )df
Example
Convolution Theorem R∞ R +∞
Convolution Example ⇒ Eu = t=−∞ |u(t)| dt = f =−∞ |U (f )|2 df
2
Convolution
Properties
Parseval’s Theorem Energy Conservation: The energy in u(t) equals the energy in U (f ).
Energy
⊲ Conservation
Example: (
Energy Spectrum
Summary
e−at t≥0 R 2
h
−e−2at
i∞
1
u(t) = ⇒ Eu = |u(t)| dt = 2a = 2a
0 t<0 0
1
U (f ) = a+i2πf [from before] 1
a=2
u(t)
0.5
0
R 2 R df -5 0 5
⇒ |U (f )| df = a2 +4π 2f 2
Time (s)
0.5
−1 2πf ∞ 0.4
|U(f)|
0.3
tan ( a ) π 1
0.2
0.1
= 2πa = 2πa = 2a
-5 0
Frequency (Hz)
5
−∞
Energy Spectrum
7: Fourier
Transforms: Example from before: 1
0.5
a=2, F=2
w(t)
0
Convolution and ( -0.5
Parseval’s Theorem e−at cos 2πF t t ≥ 0 -5 0
Time (s)
5
Multiplication of
Signals
w(t) = 0.25
0 t<0 0.2
|W(f)|
0.15
Multiplication 0.1
0.05
Example -5 0 5
Convolution Theorem 0.5 0.5 Frequency (Hz)
W (f ) = +
<W(f) (rad/pi)
0.5
Convolution Example a+i2π(f +F ) a+i2π(f −F )
0
Convolution
Properties a+i2πf -0.5
= a2 +i4πaf −4π 2 (f 2 −F 2 )
-5 0
Frequence (Hz)
5
0.06
Energy Conservation
|W(f)| 2
0.04
⊲ Energy Spectrum
|W (f )|2 =
2
a +4π f 2 2 0.02
0
Summary (a2 −4π 2 (f 2 −F 2 ))2 +16π 2 a2 f 2 -5 0
Frequency (Hz)
5
Energy Spectrum
2
• The units
R ∞of |W (f )|2 are “energy per Hz” so that its integral,
Ew = −∞ |W (f )| df , has units of energy.
2
• The quantity |W (f )| is called the energy spectral density of w(t) at
frequency f and its graph is the energy spectrum of w(t). It shows
how the energy of w(t) is distributed over frequencies.
2
• If you divide |W (f )| by the total energy, Ew , the result is non-negative
and integrates to unity like a probability distribution.
Summary
7: Fourier
Transforms: • Convolution: R∞
Convolution and
Parseval’s Theorem ◦ u(t) ∗ v(t) , −∞ u(τ )v(t − τ )dτ
Multiplication of
Signals
⊲ Arguments of u(· · · ) and v(· · · ) sum to t
Multiplication
Example
◦ Acts like multiplication + time scaling/shifting formulae
Convolution Theorem
Convolution Example
• Convolution Theorem: multiplication ↔ convolution
Convolution ◦ w(t) = u(t)v(t) ⇔ W (f ) = U (f ) ∗ V (f )
Properties
Parseval’s Theorem ◦ w(t) = u(t) ∗ v(t) ⇔ W (f ) = U (f )V (f )
Energy Conservation
R∞ R +∞
Energy Spectrum
⊲ Summary
• Parseval’s Theorem: t=−∞ u (t)v(t)dt = f =−∞ U ∗ (f )V (f )df
∗
• Energy Spectrum:
◦ Energy spectral Rdensity: |U (f )|R2 (energy/Hz)
2 2
◦ Parseval: Eu = |u(t)| dt = |U (f )| df
For further details see RHB Chapter 13.1
⊲ 8: Correlation
Cross-Correlation
Signal Matching
Cross-corr as
Convolution
Normalized Cross-corr
Autocorrelation
Autocorrelation
example
Fourier Transform
Variants
Scale Factors
Summary 8: Correlation
Spectrogram
E1.10 Fourier Series and Transforms (2015-5585) Fourier Transform - Correlation: 8 – 1 / 11

Cross-Correlation
8: Correlation
⊲ Cross-Correlation
The cross-correlation between two signals u(t) and v(t) is
Signal Matching R∞ ∗
Cross-corr as w(t) = u(t) ⊗ v(t) , −∞ u (τ )v(τ + t)dτ
Convolution
R∞ ∗
Autocorrelation
= −∞ u (τ − t)v(τ )dτ [sub: τ → τ − t]
Autocorrelation
example The complex conjugate, u∗ (τ ) makes no difference if u(t) is real-valued
Fourier Transform
Variants but makes the definition work even if u(t) is complex-valued.
Scale Factors
Summary
Correlation versus Convolution:
Spectrogram
R∞ ∗
u(t) ⊗ v(t) = −∞ u (τ )v(τ + t)dτ [correlation]
R∞
u(t) ∗ v(t) = −∞ u(τ )v(t − τ )dτ [convolution]
Unlike convolution, the integration variable, τ , has the same sign in the
arguments of u(· · · ) and v(· · · ) so the arguments have a constant
difference instead of a constant sum (i.e. v(t) is not time-flipped).
Notes: (a) The argument of w(t) is called the “lag” (= delay of u versus v).
(b) Some people write u(t) ⋆ v(t) instead of u(t) ⊗ v(t).
(c) Some swap u and v and/or negate t in the integral.
It is all rather inconsistent /.

Signal Matching
8: Correlation
Cross-Correlation
Cross correlation is used to find where two 1
0.5
u(t)
⊲ Signal Matching signals match: u(t) is the test waveform. 0
Cross-corr as 0 200 400 600
Convolution
Normalized Cross-corr Example 1: 1
Autocorrelation
v(t) contains u(t) with an unknown delay 0.5
v(t)
Autocorrelation 0
example and added noise. 0 200 400 600 800
Fourier Transform
Variants w(t) = u(t)
R ∗⊗ v(t)
20
w(t)
Scale Factors 10
Summary = u (τ − t)v(τ )dt gives a peak 0
Spectrogram
at the time lag where u(τ − t) best 0 200 400 600 800
matches v(τ ); in this case at t = 450 2
y(t)
0
-2
Example 2: 0 200 400 600 800
y(t) is the same as v(t) with more noise 20

10
z(t)
z(t) = u(t) ⊗ y(t) can still detect the 0
-10
0 200 400 600 800
correct time delay (hard for humans)
1
Example 3:
p(t)
0
-1
p(t) contains −u(t) so that 0 200 400 600 800
q(t) = u(t) ⊗ p(t) has a negative peak 0
q(t)
-10
-20
0 200 400 600 800

Cross-correlation as Convolution
8: Correlation
R∞ ∗
Cross-Correlation
Correlation: w(t) = u(t) ⊗ v(t) = −∞
u (τ − t)v(τ )dτ
Signal Matching
Cross-corr as If we define x(t) = u∗ (−t) then
⊲ Convolution R∞ R∞ ∗
Autocorrelation
x(t) ∗ v(t) , −∞ x(t − τ )v(τ )dτ = −∞ u (τ − t)v(τ )dτ
Autocorrelation
example = u(t) ⊗ v(t)
Fourier Transform
Variants Fourier Transform of x(t):
Scale Factors R∞ −i2πf t
R∞ ∗
Summary X(f ) = −∞ x(t)e dt = −∞ u (−t)e−i2πf t dt
Spectrogram
R∞ ∗ R ∗
∞
= −∞ u (t)ei2πf t dt = −∞ u(t)e−i2πf t dt
= U ∗ (f )
So w(t) = x(t) ∗ v(t) ⇒ W (f ) = X(f )V (f ) = U ∗ (f )V (f )
Hence the Cross-correlation theorem:
w(t) = u(t) ⊗ v(t) ⇔ W (f ) = U ∗ (f )V (f )
= u∗ (−t) ∗ v(t)
Note that, unlike convolution, correlation is not associative or commutative:
v(t) ⊗ u(t) = v ∗ (−t) ∗ u(t) = u(t) ∗ v ∗ (−t) = w∗ (−t)

Normalized Cross-correlation
8: Correlation
R∞ ∗
Cross-Correlation
Correlation: w(t) = u(t) ⊗ v(t) = −∞
u (τ − t)v(τ )dτ
Signal Matching
Cross-corr as If we define y(t) = u(t − t0 ) for some fixed t0 , then Ey = Eu :
Convolution
Normalized
R∞ 2 R∞ 2
⊲ Cross-corr Ey = −∞ |y(t)| dt = −∞ |u(t − t0 )| dt
Autocorrelation R∞
Autocorrelation
example
= −∞ |u(τ )|2 dτ = Eu [t → τ + t0 ]
Fourier Transform R 2
Variants ∞ ∗
Scale Factors Cauchy-Schwarz inequality: −∞ y (τ )v(τ )dτ ≤ Ey Ev
Summary
Spectrogram R 2
2 ∞ ∗
⇒ |w(t0 )| = −∞ u (τ − t0 )v(τ )dτ ≤ Ey Ev = Eu Ev
√
but t0 was arbitrary, so we must have |w(t)| ≤ Eu Ev for all t
We can define the normalized cross-correlation

u(t)⊗v(t)
z(t) = √
Eu Ev
with properties: (1) |z(t)| ≤ 1 for all t

(2) |z(t0 )| = 1 ⇔ v(τ ) = αu(τ − t0 ) with α constant

[Cauchy-Schwarz Inequality Proof]
You do not need to memorize this proof

R 2
∞ ∗
We want to prove the Cauchy-Schwarz Inequality: −∞ u (t)v(t)dt ≤ Eu Ev

R∞
where Eu , −∞ |u(t)|2 dt.
R∞ ∗
Suppose we define w , −∞ u (t)v(t)dt. Then,
0 ≤ |Ev u(t) − w∗ v(t)|2 dt [|· · · |2 always ≥ 0]
R
= (Ev u∗ (t) − wv ∗ (t)) (Ev u(t) − w∗ v(t)) dt [|z|2 = z ∗ z]

R
= Ev2 u∗ (t)u(t)dt + |w|2 v ∗ (t)v(t)dt − w∗ Ev u∗ (t)v(t)dt − wEv u(t)v ∗ (t)dt

R R R R
2 2R
2 |v(t)|2 dt − Ev w∗ w − Ev ww∗
R
= Ev |u(t)| dt + |w| [definition of w]

= Ev2 Eu + |w|2 Ev − 2 |w|2 Ev = Ev Eu Ev − |w|2 [|z|2 = z ∗ z]
Unless Ev = 0 (in which case, v(t) ≡ 0 and the C-S inequality is true), we must have |w|2 ≤ Eu Ev
which proves the C-S inequality.
Also, Eu Ev = |w|2 only if we have equality in the first line,
that is, |Ev u(t) − w∗ v(t)|2 dt = 0 which implies that the integrand is zero for all t.
R
w∗
This implies that u(t) = Ev
v(t).
So we have shown that Eu Ev = |w|2 if and only if u(t) and v(t) are proportional to each other.
E1.10 Fourier Series and Transforms (2015-5585) Fourier Transform - Correlation: 8 – note 1 of slide 5
Autocorrelation
8: Correlation
The correlation of a signal with itself is its autocorrelation:
Cross-Correlation
Signal Matching
R∞ ∗
Cross-corr as
w(t) = u(t) ⊗ u(t) = −∞ u (τ − t)u(τ )dτ
Convolution
Normalized Cross-corr The autocorrelation at zero lag:
⊲ Autocorrelation R∞ ∗
Autocorrelation
example
w(0) = −∞ u (τ − 0)u(τ )dτ
Fourier Transform R∞ ∗
Variants = −∞ u (τ )u(τ )dτ
Scale Factors R∞
Summary
Spectrogram
= −∞ |u(τ )|2 dτ = Eu
The autocorrelation at zero lag, w(0), is the energy of the signal.
The normalized autocorrelation: z(t) = u(t)⊗u(t)

Eu
satisfies z(0) = 1 and |z(t)| ≤ 1 for any t.
Wiener-Khinchin Theorem: [Cross-correlation theorem when v(t) = u(t)]

w(t) = u(t) ⊗ u(t) ⇔ W (f ) = U ∗ (f )U (f ) = |U (f )|2
The Fourier transform of the autocorrelation is the energy spectrum.

Autocorrelation example
8: Correlation
Cross-Correlation
Cross-correlation is used to find when two different signals are similar.
Signal Matching Autocorrelation is used to find when a signal is similar to itself delayed.
Cross-corr as
Convolution
Normalized Cross-corr First graph shows s(t) a segment of the microphone signal from the initial
Autocorrelation
Autocorrelation
vowel of “early” spoken by me. The waveform is “quasi-periodic” =
⊲ example “almost periodic but not quite”.
Fourier Transform
Variants
Scale Factors Second graph shows normalized autocorrelation, z(t) = s(t)⊗s(t)
Es .
Summary
Spectrogram z(0) = 1 for t = 0 since a signal always matches itself exactly.
z(t) = 0.82 for t = 6.2 ms = one period lag (not an exact match).
z(t) = 0.53 for t = 12.4 ms = two periods lag (even worse match).
1 1
Norm Autocorr, z(t)

0.82 Lag = 6.2 ms (161 Hz)
0.5
Speech s(t)
0.5
0
-0.5 0
-1
10 20 30 40 50 0 5 10 15 20
Time (ms) Lag: t (ms)

Fourier Transform Variants
8: Correlation
Cross-Correlation
There are three different versions of the Fourier Transform in current use.
Signal Matching
Cross-corr as (1) Frequency version (we have used this in lectures)
Convolution R∞ −i2πf t
R∞
Normalized Cross-corr U (f ) = −∞ u(t)e dt u(t) = −∞ U (f )ei2πf t df
Autocorrelation
Autocorrelation
example
• Used in the communications/broadcasting industry and textbooks.
Fourier Transform • The formulae do not need scale factors of 2π anywhere. ,,,
⊲ Variants
Scale Factors
Summary (2) Angular frequency version
Spectrogram R∞ R∞
e
U (ω) = u(t)e−iωt dt u(t) = 1
Ue (ω)eiωt dω
−∞ 2π −∞
Continuous spectra are unchanged: Ue (ω) = U (f ) = U ( ω )
2π
However δ-function spectral components are multiplied by 2π so that
U (f ) = δ(f − f0 ) ⇒ U e (ω) = 2π × δ(ω − 2πf0 )
• Used in most signal processing and control theory textbooks.
(3) Angular frequency + symmetrical scale factor
R∞ R∞
b 1
U (ω) = √2π −∞ u(t)e −iωt
dt 1
u(t) = √2π −∞ U b (ω)eiωt dω
b (ω) = √1 U
In all cases U e (ω)
2π
• Used in many Maths textbooks (mathematicians like symmetry)
Scale Factors
8: Correlation
Fourier Transform using Angular Frequency:
Cross-Correlation
R∞ R∞
Signal Matching e
U (ω) = −∞ u(t)e −iωt
dt 1 e (ω)eiωt dω
u(t) = 2π −∞ U
Cross-corr as
Convolution R 1
R
Autocorrelation
Any formula involving df will change to 2π dω [since dω = 2πdf ]
Autocorrelation
example
Fourier Transform Parseval’s Theorem:
Variants R ∗ R ∗
⊲ Scale Factors
1
u (t)v(t)dt = 2π U e (ω)Ve (ω)dω
R 2
Summary
R
Spectrogram 2 1
Eu = |u(t)| dt = 2π U e (ω) dω
Waveform Multiplication: (convolution implicitly involves integration)

w(t) = u(t)v(t) ⇒ W f (ω) = 1 U e e
2π (ω) ∗ V (ω)
Spectrum Multiplication: (multiplication ; integration)

w(t) = u(t) ∗ v(t) ⇒ W f (ω) = U e (ω)Ve (ω)
To obtain formulae for version (3) of the Fourier b

√ Transform, U (ω),
substitute into the above formulae: Ue (ω) = 2π U b (ω).
Summary
8: Correlation
R∞
Cross-Correlation
• Cross-Correlation: w(t) = u(t) ⊗ v(t) = −∞ u∗ (τ − t)v(τ )dτ
Signal Matching
Cross-corr as ◦ Used to find similarities between v(t) and a delayed u(t)
Convolution
Normalized Cross-corr ◦ Cross-correlation theorem: W (f ) = U ∗ (f )V (f )
Autocorrelation √
Autocorrelation ◦ Cauchy-Schwarz Inequality: |u(t) ⊗ v(t)| ≤ Eu Ev
example
Fourier Transform u(t)⊗v(t)
Variants ⊲ Normalized cross-correlation: √E E ≤ 1
u v
Scale Factors
⊲ Summary R∞ ∗
Spectrogram • Autocorrelation: x(t) = u(t) ⊗ u(t) = −∞ u (τ − t)u(τ )dτ ≤ Eu
◦ Wiener-Khinchin: X(f ) = energy spectral density, |U (f )|2
◦ Used to find periodicity in u(t)
• Fourier Transform using ω:
◦ Continuous spectra unchanged; spectral impulses multiplied by 2π
R 1
R
◦ In formulae: df → 2π dω; ω-convolution involves an integral
For further details see RHB Chapter 13.1

Spectrogram
8: Correlation
Cross-Correlation
Spectrogram of “Merry Christmas” spoken by Mike Brookes
Signal Matching
Cross-corr as
Convolution
Autocorrelation
Autocorrelation
example
Fourier Transform
Variants
Scale Factors
Summary
⊲ Spectrogram

[Complex Fourier Series]
All waveforms have period T = 1. δcondition is 1 whenever “condition” is true and otherwise 0.
Waveform x(t) for |t| < 0.5 Xn

2 sin 0.5πn
Square wave 2δ|t|<0.25 − 1 πn
× δn6=0
sin πdn
Pulse of width d δ|t|<0.5d πn
i(−1)n
Sawtooth wave 2t πn
× δn6=0
2(1−(−1)n )
Triangle wave 1 − 4 |t| π 2 n2
[Fourier Transform Properties A]
R∞
You need not memorize these properties. All integrals are −∞
Property x(t) Xf )
x(t)e−i2πf t dt
R
Forward x(t)
X(f )ei2πf t df
R
Inverse X(f )
R
Spectral Zero x(t)dt = X(0)
R
Temporal Zero x(0) = X(f )df
Duality X(t) x(−f )
Reversal x(−t) X(−f )
conjugate x∗ (t) X ∗ (−f )
dn
Temporal Derivative dtn
x(t) (i2πf )n X(f )
dn
Spectral Derivative (−i2πt)n x(t) df n
X(f )
Rt 1
Integral −∞ x(τ )dτ i2πf
X(f ) + 21 X(0)δ(f )
2iπf β
1 f
Scaling x(αt + β) |α|
e α X( α )
Time Shift x(t − T ) X(f )e−i2πf T
Frequency Shift x(t)ei2πF t X(f − F )
[Fourier Transform Properties B]
R∞
You need not memorize these properties. All integrals are −∞
Property x(t) Xf )
Linearity αx(t) + βy(t) αX(f ) + βY (f )
Multiplication x(t)y(t) X(f ) ∗ Y (f )
Convolution x(t) ∗ y(t) X(f )Y (f )
Correlation x(t) ⊗ y(t) X ∗ (f )Y (f )
Autocorrelation x(t) ⊗ x(t) |X(f )|2
x∗ (t)y(t)dt = X ∗ (f )Y (f )df
R R
Parseval or
Ex = |x(t)|2 dt = |X(f )|2 df
R R
Plancherel
1 P
P k k
Repetition n x(t − nT )

T kX T δ f − T
1 P
P k
Sampling n x(nT )δ(t − nT )

T kX f − T
1
Modulation x(t) cos(2πF t) 2
X(f − F ) + 12 X(f + F )
R
Convolution: x(t) ∗ y(t) = x(τ )y(t − τ )dτ
R ∗ R ∗
Cross-correlation: x(t) ⊗ y(t) = x (τ )y(τ + t)dτ = x (τ − t)y(τ )dτ
[Fourier Transform Pairs]
You need not memorize these pairs.

x(t) X(f ) x(t) X(f )
δ(t) 1 1 δ(f )
sin(πf ) sin(t)
rect(t) πf t
πrect(πf )
sin2 (πf ) sin2 (t)
tri(t) π2 f 2 t2
πtri(πf )
1 1 i
cos(2παt) 2
δ (f + α) + 2
δ (f − α) sin(2παt) 2
δ (f + α) − 2i δ (f − α)
1 1
e−αt u(t) α+2πif
te−αt u(t) (α+2πif )2
2α 2 2
e−α|t| α2 +4π 2 f 2
e−πt e−πf
1 1 1
sgn(t) u(t) 2
δ(f ) + 2πif
P∞ 1 P∞ iπf k

n=−∞ δ(t − nT )
T k=−∞ δ f − T
Elementary Functions:
( (
1, |t| < 0.5 1 − |t|, |t| < 1
rect(t) = tri(t) =
0, elsewhere 0, elsewhere
 
−1, t < 0
 0,
 x<0
1
sgn(t) = 0, t=0 u(t) = 2
(1 + sgn(t)) = 0.5, x=0
 
1, t>0 1, x>0
 

Handouts PDF

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Handouts PDF

Hochgeladen von

Copyright:

Verfügbare Formate

E1.

10 Fourier Series and Transforms

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 1 / 14

⊲ Syllabus Main fact: Complicated time waveforms can be

1: Sums and Why bother? Sine/cosine are the only bounded

Hard problem: Complicated waveform → electronic circuit→ output = ?

Syllabus: Preliminary maths (1 lecture)

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 2 / 14

Fourier Analysis Fourier Synthesis

We want to do the same thing with mathematical signals instead of light.

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 3 / 14

• Email me with any errors in slides or problems and if answers are

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 4 / 14

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 5 / 14

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 6 / 14

Example 2: (alternating signs)

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 7 / 14

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 8 / 14

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 9 / 14

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 10 / 14

But beware: hx(t) × y(t)i =

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 11 / 14

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 12 / 14

Proof that x(t + T ) = x(t) ∀t ⇒ x(t ± kT ) = x(t) ∀t, ∀k ∈ Z

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 13 / 14

E1.10 Fourier Series and Transforms (2014-5509) Sums and Averages: 1 – 14 / 14

E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – 1 / 11

E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – 2 / 11

−0.4 sin 2π2F t [b2 = −0.4]

+0.4 sin 2π3F t [b3 = 0.4]

−0.2 cos 2π4F t [a4 = −0.2]

E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – 4 / 11

E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – 5 / 11

Remember, for any integer n, hsin 2πnF ti = 0 (

Finding an and bn is called Fourier analysis.

E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – 6 / 11

Summary: hsin cosi = 0 [provided that m + n 6= 0]

bn = T 0 u(t) sin 2πnt

E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – 9 / 11

E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – 10 / 11

◦ bn = 2 hu(t) sin (2πnF t)i

• The mapping u(t) → {an , bn } is linear

E1.10 Fourier Series and Transforms (2014-5379) Fourier Series: 2 – 11 / 11

E1.10 Fourier Series and Transforms (2014-5543) Complex Fourier Series: 3 – 1 / 12

Using eiθ Using cos θ and sin θ

E1.10 Fourier Series and Transforms (2014-5543) Complex Fourier Series: 3 – 2 / 12

E1.10 Fourier Series and Transforms (2014-5543) Complex Fourier Series: 3 – 4 / 12

E1.10 Fourier Series and Transforms (2014-5543) Complex Fourier Series: 3 – 5 / 12

E1.10 Fourier Series and Transforms (2014-5543) Complex Fourier Series: 3 – 6 / 12

Series T = 20, width W = 4, height A = 8 5

Note: If u(t) is a sine wave, U1 equals half the corresponding phasor.

Proof of (3): u(t) odd ⇒ Un odd

E1.10 Fourier Series and Transforms (2014-5543) Complex Fourier Series: 3 – 9 / 12

Odd harmonics only ⇔ 5

E1.10 Fourier Series and Transforms (2014-5543) Complex Fourier Series: 3 – 10 / 12

Summary ⇔ odd harmonics only 0

U0:1 = [0, −i] -2

(4) Even harmonics only 8

E1.10 Fourier Series and Transforms (2014-5543) Complex Fourier Series: 3 – 11 / 12

E1.10 Fourier Series and Transforms (2014-5543) Parseval and Convolution: 4 – 1 / 9

E1.10 Fourier Series and Transforms (2014-5543) Parseval and Convolution: 4 – 2 / 9

(1) You can always move any quantities to the right