CH 03
CH 03
Representation of Signals
1
The Electromagnetic Spectrum
2
MOTIVATION
3
SPECTRUM of SINE
Sine = sum of 2 complex exponentials:
A sin(7t ) = 2Aj e j 7t − 2Aj e − j 7 t
= 12 Ae − j 0.5π e j 7t + 12 Ae j 0.5π e − j 7t
−1 =
j
j = e j 0.5π
Positive freq. has phase = -0.5π
Negative freq. has phase = +0.5π
GRAPHICAL SPECTRUM
EXAMPLE of SINE
A sin(7t ) = 1
2
Ae − j 0.5π e j 7t + 12 Ae j 0.5π e − j 7t
-7 0 7 ω
5
• the real part of a complex term is equal to one
half the sum of that term and its complex conjugate
• each sinusoid in the sum decomposes into two
terms, one with positive frequency fk and the other
with negative frequency −fk
• When we want to describe the frequency content
of this signal, we need to specify all the complex am-
plitudes and frequencies.
• How many different complex amplitudes/frequencies
do we have in the above signal?
6
ex:
π π
x(t) = 10 + 14 cos 200πt − + 8 cos 500πt +
3 2
Using the first one of inverse Euler formulas (for the
cosine), we can expand each cosine term into complex
exponential terms.
14 h j(200πt− π ) −j(200πt− π)
i 8h
j(500πt+ π) −j(500πt+ π)
i
x(t) = 10 + e 3 +e 3 + e 2 +e 2
2h i 2
h i
j200πt −j π3 −j200πt j π3 j500πt j π2 −j500πt −j π2
= 10 + 7 e e +e e +4 e e +e e
π π π π
= 10 + 7ej200πte−j 3 + 7e−j200πtej 3 + 4ej500πtej 2 + 4e−j500πte−j 2
−j 3 j200πt π π π π
= |{z}
10 + |7e{z }e 7ej 3 e−j200πt + |{z}
+ |{z} 4ej 2 ej500πt + 4e −j
| {z }2 e
−j500πt
Xo X1 X1∗ X2 X2∗
2 2 2 2
pairwise description of the two-sided frequency spectrum:
−j π3 j π3 j π2 −j π2
(0, 10), (100, 7e ), (−100, 7e ), (250, 4e ), (−250, 4e )
We would prefer to have a graphical representation:
FREQUENCY DIAGRAM
10
7e jπ / 3
7e − jπ / 3
4e − jπ / 2 4e jπ / 2
7
Vertical lines are called spectral lines and they in-
dicate magnitude. Because of this, they should never
go below the horizontal axis.
Graphical frequency-spectrum representation makes
it easy to see:
• relative location of the frequencies
• relative amplitudes of the sinusoidal components
8
Example: Synthetic Vowel
Vowel Waveform
(sum of all 5 components)
9
SPECTRUM of VOWEL
Note: Spectrum has 0.5Xk (except XDC)
Conjugates in negative frequency
To = 1/fo or fo = 1/To
10
Periodic signals can be synthesized by adding two
or more sinusoids having harmonically related fre-
quencies. In other words, all frequencies are integer
multiples of the fundamental frequency.
earlier model (can cover periodic/aperiodic signals):
XN
x(t) = Ao + Ak cos(2πfk t + φk )
k=1
for periodicity, fk = kfo where k is an integer:
N
X
x(t) = Ao + Ak cos(2πkfot + φk )
k=1
12
Periodicity is tied to harmonic frequencies.
What happens if the frequencies have no relation to
one another?
If the frequencies fk are not integer multiples of a
common fundamental frequency, the signal will be
aperiodic (non-periodic).
IRRATIONAL SPECTRUM
SPECIAL RELATIONSHIP
to get a PERIODIC SIGNAL
13
Harmonic Signal (3 Freqs)
T=0.1
NON-Harmonic Signal
NOT
PERIODIC
14
Fourier Series Representation of Periodic Signals
synthesis formula:
∞
X j ( T2π )kt 2π
x(t) = ak e o ωo =
To
k=−∞
15
How do we derive the Fourier series coefficients ak
for the harmonic sum? In other words, how do we
go from x(t) to ak ? Analyzing only one period is
sufficient for this purpose:
1 To
Z
−j ( T2π )kt
ak = x(t)e o dt
To 0
A special and simpler case occurs when k = 0:
Z To
1
ao = x(t) dt
To 0
• the integral gives the area under the function over
one period.
• ao is simply the average value (d.c. value) of the
signal over one period.
• The ao coefficient is an additive constant. There-
fore, a change in its value will move the plot of the
signal up or down vertically (in amplitude). There
will be no change to the shape and the nature of the
repeating pattern; therefore, all the other ak ’s (for
k 6= 0) remain the same.
• The above integrals would be useful if we have
a formula that defines x(t) over one period. If not,
numerical methods would be needed.
16
Some background material before deriving the FS:
∆ j ( 2π )kt
Let v (t) = e To
k
(complex exponential for the kth harmonic).
The FS derivation relies on the simple property of
the complex exponential signal that the integral of a
complex exp. signal over an integral number of peri-
ods is zero. We will show this in two different ways:
first way: (for k 6= 0)
To
j ( T2π )kt
Z To Z To
j ( T2π )kt e o
vk (t) dt = e o dt =
0 0 j 2π To k
0
j T2π kTo j T2π k0
e( ) o −e ( ) o ej2πk − 1
= =
j 2π
To k j 2πTo k
1k − 1 1−1 0
= = = =0
j 2π
To k j 2π
To k j 2π
To k
j ( T2π )k(t+nTo )
vk (t + nTo) = e o
18
orthogonality principle:
Z To (
0 if k =6 `
vk (t)v`∗(t) dt =
0 To if k = `
• inner product of vk (t) and v`(t)
proof:
j ( T2π )`t ∗ −j ( T2π )`t
If v`(t) = e o , then v` (t) = e o .
Z To Z To
∗ j ( T2π )kt −j ( T2π )`t
vk (t)v` (t) dt = e o e o dt
0
Z0 To
2π
= ej ( To )(k−`)t dt
Z0 To Z To
2π
if k = ` : = ej ( To )0t dt = j0
|{z} dt
e
0 0 1
Z To To
= 1 dt = t = To
0 0
Z To
2π
if k 6= ` : = ej ( To )(k−`)t dt
Z0 To
j ( T2π )mt ∆
= e o dt where m = k − ` 6= 0
0
= 0
(We have shown in two different ways on p.17 that
this integral is zero.)
19
Now, we are going to start with the synthesis equa-
tion and derive the analysis equation. Assuming the
synthesis equation is valid:
∞
j ( T2π )kt
X
x(t) = ak e o
k=−∞
2π
Multiply both sides of the equality by v`∗(t) = e−j ( To )`t
and integrate both sides from 0 to To:
∞
Z To Z To X !
∗ j ( T2π )kt
x(t)v` (t) dt = ak e o v`∗(t) dt
0 0 k=−∞
∞
!
Z To Z To
−j ( T2π )`t j ( T2π )kt −j ( T2π )`t
X
x(t)e o dt = ak e o e o dt
0 0 k=−∞
∞ Z To
2π
ej ( To )(k−`)t dt
X
= ak
k=−∞ | 0 {z }
0 if k 6= `
T
o if k = `
= a`To
Leaving a` on the left-hand side, it follows that:
Z To
1 −j ( T2π )`t
a` = x(t)e o dt
To 0
This is the analysis equation. You can always re-
name the dummy index ` to call it k as before.
20
• It is also possible to do the derivation in the re-
verse direction, that is, starting with the analysis
equation to get the synthesis equation. The steps
of the derivation are not obvious in both cases.
• Thus, we have two key equations for the Fourier
series representation of periodic signals:
∞
j ( T2π )kt
X
x(t) = ak e o
k=−∞
(synthesis equation)
1 To
Z
−j ( T2π )kt
ak = x(t)e o dt
To 0
(analysis equation)
21
• It is not always necessary to evaluate the analy-
sis integral to obtain the {ak } coefficients. In some
cases, it may be possible to directly expand the peri-
odic signal into a sum of complex exponentials. This
is the case if the periodic signal is already expressed
as a sum of (one or more of) constants, complex ex-
ponentials, and sinusoids.
π
ex: x(t) = −4 + 6 cos 20πt + − 2e−j(30πt−2)
j(20πt+ π ) 6 −j(20πt+ π )
e 6 +e 6
= −4 + 6 − 2e−j(30πt−2)
2
π −j(20πt+ π6 )
= −4 + 3e j(20πt+ 6 )
+ 3e − 2e−j(30πt−2)
j π6 j20πt −j π6 −j20πt 2j −j30πt
−4 + |{z}
= |{z} 3e e | {z } e
+ 3e −2e
| {z } e
ao a2 a−2 a−3
x(t) = sin3(3πt)
−j3πt 3
j3πt
e −e
=
2j
1 j9πt j6πt −j3πt j3πt −j6πt −j9πt
= e − 3e e + 3e e −e
−8j
j −3j j3πt 3j −j3πt −j −j9πt
= ej9πt + e + e + e
8 8 8 8
We see that the signal contains the four different fre-
quencies: ω = ±3π and ω = ±9π rad/s.
23
Since g.c.d.{3π, 9π} = 3π, the fundamental frequency
is ωo = 3π rad/s. The FS coefficients are indexed in
terms of the fundamental frequency, therefore:
0 for k = 0
3
∓j 8 for k = ±1
ak = 0 for k = ±2
1
±j 8 for k = ±3
0 for k = ±4, ±5, ±6, . . .
24
Fourier Series Representation of the Square Waveform:
(
1 for 0 < t ≤ T2o
x(t) =
0 for T2o < t ≤ To
analysis:
1 To
Z
−j ( T2π )kt
ak = x(t)e o dt
To (0
Z To Z To )
1 2 2π 2π
= 1.e−j ( To )kt dt + 0.e−j ( To )kt dt
To 0 To
2
Z To
1 2
−j ( T2π )kt
= 1.e o dt
To 0
To
2
−j ( T2π )kt −j ( T2π )k T2o −j ( T2π )k0
1 e o 1 e o −e o
= = (k 6= 0)
To −j 2π To −j To k 2π
To k
0
25
−jπk −jk0 −jπk
1 e −e e −1
= = (k 6= 0)
To −j2πk
2π
−j To k
k k
(−1) − 1 1 − (−1)
= = (k 6= 0)
−j2πk j2πk
( 1−(−1)
2 1 −j
j2πk = j2πk = jπk = πk for k odd
= 1−1
j2πk = 0 for k even
The result is not valid for k = 0 because k is in the
denominator.
Evaluate ao separately (the average value of x(t)):
1 To
Z
ao = x(t) dt
To "0
Z To Z To #
1 2
= 1 dt + 0 dt
To 0 To
2
Z To To
1 2 1 2 1 To 1
= 1 dt = t = =
To 0 To 0 To 2 2
26
• the square waveform has only odd harmonics
in its frequency spectrum.
• even harmonics are non-existent
• the formula for ak does not depend on To
• the average value ao can be easily changed
• |ak | ↓ as |k| ↑
27
Fourier Synthesis:
recall the synthesis formula:
∞
j ( T2π )kt
X
x(t) = ak e o
k=−∞
synthesis:
Suppose we choose N = 0:
Then, the truncated summation contains only a sin-
gle term (the central one) which is x0(t) = ao (the
average value).
We try to approximate the periodic waveform using
its average value which is an extremely crude approx-
imation.
28
Now, let N = 1:
−j ( T2π )t j ( T2π )t
x1(t) = a−1e o +ao+a1e o
= a−1e−jωot+ao+a1ejωot
= ao+a−1e−jωot + a1ejωot
= ao+a1ejωot + a−1e−jωot
1 −j jωot j −jωot
= + e + e
2 π π
1 1 −j π jωot 1 j π −jωot
= + e 2e + e 2e
2 π π
1 1 j(ωot− π ) 1 −j(ωot− π )
= + e 2 + e 2
2 π π
1 2 π
= + cos ωot −
2 π 2
1 2
= + sin(ωot)
2 π
29
Next, let N = 3:
−j3( T2π )t −j ( T2π )t j ( T2π )t j3( T2π )t
x3(t) = a−3e o +a−1 e o + ao + a1e o +a3e o
= a−3e−j3ωot+ |a−1e−jωot +{zao + a1ejωo}t +a3ej3ωot
x1 (t)
= a−3e−j3ωot+x1(t)+a3ej3ωot
= x1(t)+a−3e−j3ωot + a3ej3ωot
= x1(t)+a3ej3ωot + a−3e−j3ωot
−j j3ωot j −j3ωot
= x1(t)+ e + e
3π 3π
1 −j π j3ωot 1 π
= x1(t)+ e 2e + ej 2 e−j3ωot
3π 3π
1 j(3ωot− π ) 1 −j(3ωot− π )
= x1(t)+ e 2 + e 2
3π 3π
2 π
= x1(t)+ cos 3ωot −
3π 2
1 2 2
= + sin(ωot) + sin(3ωot)
2 π 3π
31
At this point, we can clearly see the pattern:
1 2 2 2
xN (t) = + sin(ωot) + sin(3ωot) + . . . + sin(N ωot)
2 π 3π Nπ
N
1 X 2
= + sin(kωot) (k odd integer)
2 kπ
k=1
32
derivation:
Assuming that x(t) is a real and periodic signal, its
Fourier series representation in terms of harmonically
related complex exponentials is given by:
2π 2π 2π
x(t) = . . . + a−3e−j3( To )t + a−2e−j2( To )t + a−1e−j ( To )t
2π 2π 2π
+ao + a1ej ( To )t + a2ej2( To )t + a3ej3( To )t + . . .
= . . . + a−3e−j3ωot + a−2e−j2ωot + a−1e−jωot
+ao + a1ejωot + a2ej2ωot + a3ej3ωot + . . .
= ao + a1ejωot + a−1e−jωot + a2ej2ωot + a−2e−j2ωot
j3ωo t −j3ωo t
+ a3 e + a−3e + ...
= ao + a1ejωot + a∗1 e−jωot + a2ej2ωot + a∗2 e−j2ωot
33
Synthesis: 1st & 3rd Harmonics
1 2 π 2 π
y (t )= + cos(2π (25 )t− 2 )+ cos (2π (75 )t− 2 )
2 π 3π
34
Fourier Synthesis
1 2 2
x N (t )= + sin(ω 0 t )+ sin (3ω0 t )+…
2 π 3π
Gibbs’ Phenomenon
Convergence at DISCONTINUITY of x(t)
There is always an overshoot
9% for the Square Wave case
35
Error Magnitude When Approximating a Square Wave:
fseriesdemo GUI
36
Fourier Series Representation of the Triangular Waveform:
∫ sin ( 2π
1 − j(2π /T )kt
a k= T t) e 0
dt
0
0 T 0
T /2 j (2π /T 0 )t − j(2π / T 0 )t
0
1 e −e − j(2π / T 0 )kt
=T
0
∫ 2j
e dt
0
T /2 T /2
0 0
1 − j(2π /T 0 )(k −1)t 1 − j(2π / T 0 )( k+1 )t
= j2T
0
∫ e dt−
j2T 0
∫ e dt
0 0
− j(2π /T 0 )( k−1) t − j (2π /T 0 )(k +1)t
e T /2 e T /2
= ∣0 0 − ∣0 0
j2T 0 (− j(2π /T 0 )(k −1)) j2T 0 (− j(2π /T 0 )(k +1))
38