Week 2, Lecture A: CT Fourier Series (Trig Form) · 1=2 1=2 0 1 x(t) 9% You can decrease (and even...
Transcript of Week 2, Lecture A: CT Fourier Series (Trig Form) · 1=2 1=2 0 1 x(t) 9% You can decrease (and even...
Transforms
Signals are functions that convey information. However, thestraightforward way of representing a signal may not exposeimportant properties of the signal.
It is often useful to have multiple different ways of lookingat a signal.
6.003 Signal Processing Week 2 Lecture A (slide 2) 12 Feb 2019
Transforms
For example, consider the following two representations ofa speech signal:
6.003 Signal Processing Week 2 Lecture A (slide 3) 12 Feb 2019
Transforms
We call such an alternative view of a signal a transform.
In 6.003, we will primarily focus on a family of transformsreferred to as Fourier transforms, where signals arerepresented as sums of harmonically-related sinusoids, forexample:
f(t) = c0 +
∞∑k=1
(ck cos(kωot) + dk sin(kωot)
)
PSet 1 contained an example of this. Given a bunch ofcoefficients associated with harmonically-related sinusoids,find the waveform that results from combining themtogether.
Today: how can we go in the other direction?Under what conditions can f(t) be expressed in this form?How can we compute ck and dk?
6.003 Signal Processing Week 2 Lecture A (slide 4) 12 Feb 2019
Preliminaries: Jargon
A periodic signal is one that repeats after some amount oftime T , such that, for all t,
x(t) = x(t+ T )
A function that is periodic in T is also periodic in2T, 3T, 4T, . . .. For a given signal, the smallest value T forwhich the above holds is called the signal’s fundamentalperiod.
The associated frequency, ω0 = 2πT, is the signal’s
fundamental frequency.
Frequencies are harmonically related if they are integermultiples of some fundamental frequency.
6.003 Signal Processing Week 2 Lecture A (slide 5) 12 Feb 2019
Preliminaries: Sinusoids
A cos(ωt) +B sin(ωt) =(√
A2 +B2)cos
(ωt+ tan−1
(−b
a
))
6.003 Signal Processing Week 2 Lecture A (slide 6) 12 Feb 2019
Preliminaries: Sinusoids
Where k is a positive integer:∫ t0+T
t0
sin
(2πk
Tt
)dt = 0
∫ t0+T
t0
cos
(2πk
Tt
)dt = 0
6.003 Signal Processing Week 2 Lecture A (slide 7) 12 Feb 2019
Preliminaries: Sinusoids
Where k and m are positive integers:∫ t0+T
t0
sin
(2πk
Tt
)cos
(2πm
Tt
)dt = 0
6.003 Signal Processing Week 2 Lecture A (slide 8) 12 Feb 2019
Preliminaries: Sinusoids
Where k and m are positive integers:
∫ t0+T
t0
cos
(2πk
Tt
)cos
(2πm
Tt
)dt =
{T/2 if k = m,
0 otherwise
6.003 Signal Processing Week 2 Lecture A (slide 9) 12 Feb 2019
Preliminaries: Sinusoids
Where k and m are positive integers:
∫ t0+T
t0
sin
(2πk
Tt
)sin
(2πm
Tt
)dt =
{T/2 if k = m,
0 otherwise
6.003 Signal Processing Week 2 Lecture A (slide 10) 12 Feb 2019
Preliminaries
Fourier series are sums of harmonically related sinusoids:
f(t) = c0 +
∞∑k=1
(ck cos(kωot) + dk sin(kωot)
)where ωo represents the “fundamental frequency.”
Basis functions:
2πωo
t
cos0t
2πωo
t
sin0t
2πωo
t
cosωot
2πωo
tsinωot
2πωo
t
cos2ω
ot
...2πωo
t
sin2ωot
...
6.003 Signal Processing Week 2 Lecture A (slide 11) 12 Feb 2019
Fourier Representations of Signals
What signals can be represented by sums of harmoniccomponents?
ωωo 2ωo 3ωo 4ωo 5ωo 6ωo
T= 2πωo
t
T= 2πωo
t
Only periodic signals: all harmonics of ωo are periodic inT = 2π/ωo.
6.003 Signal Processing Week 2 Lecture A (slide 12) 12 Feb 2019
Fourier Representations of Signals
Assume f(t) is periodic in T and is composed of a weightedsum of harmonics of w0 = 2π/T :
f(t) = c0 +
∞∑k=1
(ck cos
(2πk
Tt
)+ dk sin
(2πk
Tt
))How do we find the weights?
6.003 Signal Processing Week 2 Lecture A (slide 13) 12 Feb 2019
c0: The “DC” Term
The constant c0 is referred to as the “DC” term: itrepresents the constant offset of a signal (if any). Statedanother way, it represents the average value of a signal overone period.
c0 =1
T
∫ t0+T
t0
f(t)dt
6.003 Signal Processing Week 2 Lecture A (slide 14) 12 Feb 2019
Computing the other ck terms
We are assuming that our signal is comprised ofharmonically-related sines and cosines. Here is an examplesignal:
f(t) = c1 cos(ω0t) + d1 sin(ω0t) + c3 cos(3ω0t) + d5 sin(5ω0t)
Let’s imagine we wanted to determine the coefficient c3,associated with cos(3ω0t).
Our strategy will be to multiply f(t) by cos(3ω0t) andintegrate the result over one period.
Check Yourself: How does this help us find c3?
6.003 Signal Processing Week 2 Lecture A (slide 15) 12 Feb 2019
Computing the other terms
In general, we can compute the coefficients:
ck =2
T
∫ t0+T
t0
f(t) cos
(2πkt
T
)dt
We can use similar logic to find the coefficients associatedwith the sine waves:
dk =2
T
∫ t0+T
t0
f(t) sin
(2πkt
T
)dt
6.003 Signal Processing Week 2 Lecture A (slide 16) 12 Feb 2019
Fourier
Complicated waveforms represented as sums ofharmonically-related sinusoids. Two “views”:
Frequency
Time
Magnitude
6.003 Signal Processing Week 2 Lecture A (slide 17) 12 Feb 2019
Moving Between Representations
Synthesis:
f(t) = c0 +
∞∑k=1
(ck cos
(2πk
Tt
)+ dk sin
(2πk
Tt
))
Analysis:
c0 =1
T
∫ t0+T
t0
f(t)dt
ck =2
T
∫ t0+T
t0
f(t) cos
(2πkt
T
)dt
dk =2
T
∫ t0+T
t0
f(t) sin
(2πkt
T
)dt
6.003 Signal Processing Week 2 Lecture A (slide 18) 12 Feb 2019
Check Yourself!
What are the Fourier series coefficients associated with thefollowing signal?
f(t) = 0.8 sin(6πt)− 0.3 cos(6πt) + 0.75 cos(12πt)
6.003 Signal Processing Week 2 Lecture A (slide 19) 12 Feb 2019
Check Yourself!
What are the Fourier series coefficients associated with thefollowing signal?
f(t) = cos(2πt− π/4)
6.003 Signal Processing Week 2 Lecture A (slide 20) 12 Feb 2019
Check Yourself!
Let ck and dk be the Fourier series coefficients associatedwith the following signal:
t
1/2
−1/2
0 1
x(t)
Which of the following statements are true?
• ck = 0 for all k
• dk = 0 if k is even
• |dk| decreases with k2
• there are an infinite number of non-zero dk
6.003 Signal Processing Week 2 Lecture A (slide 21) 12 Feb 2019
Gibbs Phenomenon
Partial sums of Fourier series of discontinuous functions“ring” near discontinuities: Gibbs’ phenomenon.
t
1/2
−1/2
0 1
x(t) 9%
You can decrease (and even eliminate the ringing) bydecreasing the magnitudes of the Fourier coefficients athigher frequencies.
6.003 Signal Processing Week 2 Lecture A (slide 22) 12 Feb 2019
Summary
Today, we learned about the CT Fourier series, which is away of representing signals as sums of sinusoidalcomponents.
Recitation 2A: Practice with finding Fourier series
Lecture 2B: CT Fourier Series, Complex Exponential Form
6.003 Signal Processing Week 2 Lecture A (slide 23) 12 Feb 2019