Upload
others
View
10
Download
0
Embed Size (px)
Citation preview
Chapter 3Fourier Series Representation of Periodic Signals
If an arbitrary signal x(t) or x[n] is expressed as a linear combination of some basic signals, the response of an LTI system becomes the sum of the individual responses of those basic signals.Such basic signal must:
• be capable of representing a large number of signals,• have system responses simple enough for computational convenience.Complex exponentials, given below, are such basic signals/functions:
est for CT systems,zn for DT systems, where s, z are complex numbers.
Both have the property that:• Response to an LTI system has the same form as the input with a change in the amplitude only. •A function with this property is called Eigen function and the corresponding amplitude ratio is called Eigen value.
H(s)
x(t) y(t)
est H(s) est
H[z]
x[n] y[n]
zn H[z] zn
1
• est and zn Eigen functions,
• H(s) and H[z] Eigen values.
Consider an input x(t) = est in a CT system. The corresponding output:
... (1)
Similarly, for a DT system:Input: x[n] = zn ,Output:
→
→
d e )(h)s(H
where,e )s(H d e )(he
d e )(h d )t(x )(h)t(h)t(x)t(y
s
stsst
)t(s
∫
∫
∫∫
∞
∞−
τ−
∞
∞−
τ−
∞
∞−
τ−∞
∞−
ττ=
=ττ=
ττ=ττ−τ=∗=
,z ]z[H z].k[hz
z].k[h]kn[x].k[h]n[h]n[x]n[y
n
k
kn
k
kn
k
==
=−=∗=
∑
∑∑∞
−∞=
−
∞
−∞=
−∞
−∞=
2
where
... (2)
CT systems:LTI system. Thus, with an input
the output :
In general: If ... (3)
then ... (4)
DT systems:
If , ... (5)
then, ... (6)
•If the input to an LTI system is a linear combination of complexexponentials, its output is also the linear combination of the same exponentials.
∑∞
−∞=
−=k
kz ]k[h]z[H
,eaeaea)t(x ts3
ts2
ts1
321 ++=
. e )s(Hae )s(Hae )s(Ha)t(y ts33
ts22
ts11
321 ++=
,e a)t(xk
tsk
k∑=
.e )s(Ha)t(yk
tskk
k∑=
∑=k
nkk z a]n[x
∑=k
nkkk z ]z[H a]n[y
3
In Fourier series, such complex exponentials are:s = jω, for CT systems
z = e jω, for DT systems.
Fourier Series of CT Periodic Signals:
A signal x(t) with a fundamental frequency of ω0 is expressed as a linear combination of complex exponential and its harmonics
as:
... (7)
k = 1 , fundamental frequency= 2, second harmonics
= 3, third harmonics, so on.
R.H.S. Is called the ‘Fourier series’ and ak is called the k-th harmonic component.
Fourier Series Coefficient ak :From eqn. (7), we may write:
∑∞
−∞=
ω=k
tjkk
0e a)t(x
∑∑∞
−∞=
ω−∞
−∞=
ω−ωω− ==k
t)nk(j k
k
ntjtjk k
ntj 0000 eae eae )t(x
4
.. ,e .e t3jt2j 00 ω±ω±
tj 0e ω
Integrating over an interval 0 to T (T is the fundamental period):
... (8)
Using Euler’s formula:
Thus,
... (9)
From eqn. (8):
∑ ∫∫ ∑∫∞
−∞=
ω−∞
−∞=
ω−ω− ==k
T
0
t)nk(jk
T
0 k
t)nk(jk
T
0
ntj )dt e( adt e adt e )t(x 000
n .k for ,T nkfor 0,
dt t)nksin(jdt t)nkcos(dt eT
00
T
0
T
00
t)nk(j 0
==≠=
ω−+ω−= ∫∫ ∫ω−
n kfor ,0
nk for ,T dt eT
0
t)nk(j 0
≠=
==∫ ω−
. Tadt e )t(x n
T
0
tj 0 =∫ ω−
5
or
... (10a)
As x(t) is periodic, eqn. (9) is applicable to any time interval T. Hence,
.... (10 b)
•Eqn. (7) is called Fourier synthesis equation and eqn. (10b) as Fourier analysis equation.
If x(t) is a real periodic function, then , consequently
... (11)
Comparing eqn. (7) with eqn. (11):
∫ ω−=T
0
tjnn dt e )t(x
T1a 0
∫ ω−=T
tjnn dt e )t(x
T1a 0
)t(x)t(x ∗=
ktj
kk
ktj
kk
ktj
k kk k
0
00kt0j
e a
e ae a)t(xea)t(x
ω∞
−∞=
∗−
ω−∞
∞=
∗−
ω−∞
−∞=
∞
−∞=
∗∗
∑
∑∑ ∑
=
==== ω
. aa kk −∗ =
6
Problem 3.3 p. 251 of text.Find fundamental frequency and Fourier co-efficient ak of:
Solution:Using Euler’s formula:
Hence, the fundamental frequency is ω0 = π/3. It consists of the fundamental, 2nd and 5th harmonics only.Thus,
Problem 3.22 p. 256 of text.d) Find the Fourier series of:
).t3
5( sin4)t3
2( cos2)t(x π+
π+=
∑∞
∞=
ω
π−
ππ−
π
=
−+++=
-k
ktjk
t3
5jt3
5jt3
2jt3
2j
0e a
ej2 e
j2 e
21 e
212)t(x
.a ,j2a ,
21a ,
21 a ,2a
giving ,e ae ae aeaa)t(x
j2
5-52-20
t5j5
t5j5
t2j2
t2j 20
0000
−=====
++++= ω−−
ωω−−
ω
x(t)
1 11
-2
t0 1-1-2
-2 -2
2 3
7
Solution:
Here, T =2, ω0 = (2π/T)= π.
Problem 3.5 p. 193 of text. [Very important result]:
Find the Fourier series of:
Solution:
∑∞
−∞=
ω=k
ktjk
0e a)t(x
.)1(21)(cos
21
]) (e21[21] e21[
21 dt] e2. dt1[
21
]dt e )1t( 2dt e )t([21dt e )t(x
21a
kk
k-jπ]jk2
1jk1
0
2
1ktj1
0ktj2
0ktj
k0
−−=π−=
−=−=−=
−δ−+δ==
π−π−
π−π−ω−
∫∫
∫∫∫
x(t)
-T1 t
1
0 T1 -T/2 T/2 T -T
.k
)kTsin(ω )kTsin(ω.kTω2
]ee [kTjω
1ee.
T1dt e
T1a
1010
0
kTjωkTjω
0
T
Tkj
ktjT
T
ktjk
1010
1
10
01
1
0
π==
−−
=== −
−ω−
ω−
−
ω−∫
8
Special case:
Convergence of the Fourier Series:
For a function to have Fourier series, the series is to be convergent. The convergence conditions for x(t) is given by Dirichlet conditions as follows:
Over any period, x(t):i) must be absolutely integrable over any period, i.e.,
Example:
The function has discontinuities and hence not convergent.
ii) does not have more than a finite number of discontinuities over any period.
iii) may have at most a finite number of finite discontinuities.
.TT2
k
kT.T2
kkTa Lima 1
110
k0k0 =π
π
=π
ω== →
∞<∫ dt)t(x
T
x(t)
t1 0
9
Gibbs Phenomenon:• As the number of terms N is increased, Fourier series represents the function more accurately. •At discontinuities, the series has a value ½ the sum of values just before and just after the discontinuity.•The series shows ripples at discontinuity.•As N increases, the ripples get compressed toward the discontinuity, but its peak amplitude remained unchanged.
Properties of Fourier Series:• Linearity:
•Time Shift:kk
k
k
B b A aB y(t) A x(t) Then
b )t(y a )t(x If
+→+
→→
ktjk
0
k
a . e )tt(x
a )t(x00ω−→−
→
10
Time Reversal:
Time Scaling:
Multiplication:
Conjugate:
Conjugate Symmetry:For real x(t):
k
k
a )t(xa )t(x
−→−→
∑∞
−∞=
αω=⇒→α
→
k
t)(jkkk
k
0e a a )t(x
a )t(x
. b . a )t(y).t(x
b )t(ya )t(x
llkl
k
k
∑∞
−∞=−→
→→
∗−
∗ →
→
k
k
a )t(x
a )t(x
kk
kk
a a
a a
−∗
∗−
=
=
11
Derivative:
Parseval’s Relation:
• Consult Table 3.1 p. 206 of text. • Go through example 3.8 p. 208 of text.
Examples:Prob. 3.6 p.251 of text.a)Which of the following are real valued?b) Which are even?
k0ajk dt
)t(dxω→
∑∫∞
−∞=
=k
2kT
2 adt )t(xT1
∑
∑
∑
−=
π
−=
π
=
π
π=
π=
=
100
100k
t502jk
3
100
100k
t502jk
2
100
0k
t502jkk
1
e )2
ksin(j)t(x
e )kcos()t(x
e )21()t(x
12
Solution:a)
. x1(t) is not real-valued.
x
∑∞
∞−
ω= tj k1
0ea)t(x
else. 0,
100 k 0 ,)21( a a
else. 0.
100 k 0 ,)21( a )t(x
kkk
kk1
=
≤≤==
=
≤≤=⇒
−∗
∗−≠ kk a a
real. is (t) x . aa
else 0, 100k100 ),kcos( a
else. 0, 100k100 ),kcos( a :)t(x
2kk
k
k2
∗−
−
=
=≤≤−π=
=≤≤−π=
real. is )t(x . aa
else. 0.
100k100 ,2
ksinja
else. 0.
100k100 ,2
ksinja
else. 0.
100k100 ,2
ksinja :)t(x
3kk
k
k
k3
∗−
∗−
−
=
=
≤≤−π
=
=
≤≤−π
−=
=
≤≤−π
=
13
Problem 3.22(a) p.255 of text,Find the Fourier series of x(t):
Solution:
where
∑∑∞
−∞=
π∞
−∞=
ω ==
π=π
=ω=≤≤−=
k
ktjk
k
ktjk
0
eaeax(t)
hence ,2
2 ,2T and ;1t1 for ,t)t(x
0
.)1( kj)kcos(
kj
]ee [k21]ee [
kπ2j
} dt kj
e .1kj
e .t {21dt e t
21a
k
jkjk22
jkjk
1
1
ktj1
1
ktj1
1
ktjk
−π
=ππ
=
−π
++=
π−−
π−==
ππ−π−π
−
π−
−
π−
−
π− ∫∫
14
-1
-1
1
1 2 3 -2 0 4 t
x(t)
Fourier Series Representation of DT Signals
DT Fourier Series:As with CT signals, a DT signal can be expressed as a series combination of complex exponentials. The series, however, is finite.Let x[n] be periodic with a fundamental period N, i.e.,
x[n]=x[n+N].
Consider a complex function
For an integer r ,
or ... (12)
• φk[n] repeats itself with a period of N. Thus,
• a set of N values of φk[n] is enough to represent a periodic DT signal.
15
n)N2(jknjk
k ee]n[ 0
πω ==φ
].n[ee .e
e .ee]n[
kn)
N2(jkrN2jn)
N2(jk
n)N2.(jrNn)
N2(jkn)
N2)(rNk(j
rNk
φ===
==φπ
ππ
πππ+
+
]n[]n[ rNkk +φ=φ
Therefore, a periodic DT signal may be expressed as
... (13)
where <N> means any arbitrary consecutive N exponentials.
Eqn. (13) is the DT Fourier series, ak is called Fourier coefficient.
Determination of ak:
Consider
Now:
With k = r,
.N2 ,e a]n[ a]n[x 0
Nk
njkk
Nkkk
0 π=ω=φ= ∑∑
>=<
ω
>=<
∑>=<
ω−−ω− =Nk
n)rk(jk
njr 00 eae]n[x
else. 0,
.. 2N. N, 0, r-k if N,
e a
e ae a e]n[x
Nk Nn
n)N2)(rk(j
k
Nk Nn
n)rk(jk
Nn Nk
n)rk(jk
Nn
njr 000
=±±==
=
==
∑ ∑
∑ ∑∑ ∑∑
>=< >=<
π−−
>=< >=<
ω−−
>=< >=<
ω−−
>=<
ω−
.N ae]n[x rNn
njr 0 =∑>=<
ω−
16
Replacing r by k ,
... (14)
Eqn. (13) is called synthesis equation and eqn. (14) the analysis equation.
As ak in an interval N repeats itself in any other periods,
• Since the series is finite having N terms, it is always convergent.
Example:Example 3.12 p.218 of text. Find the Fourier series of:
Here, N = 9.
. N2 ,e ]n[x
N1a 0
njk
Nnk
0 π=ω= ω−
>=<∑
... aaa N 2 kN kk === ±±
-2 -1 0 1 2 n
x[n]
.... -N
.... N
.
2k sin
211N2k sin
N1
e e N1e
N1e
N1a
0
0
1N2
0m
mjk1Njk1N2
0m
)1Nm(jk1N
1Nn
njkk
0000
ω
+ω
=
=== ∑∑∑=
ω−ω
=
−ω−
−=
ω−
17
For k = 0,
With N=9, there are 9 terms, e.g.,a-4, a-3, a-2, a-1, a0, a1, a2, a3, and a4.
If the R.H.S. Of eqn. (14) is truncated to less than 9 terms, distortion occurs. See Fig. 3.18 p.220 of text.
Note that Gibbs phenomenon does not occur in DT signal approxi-mation.
Problem 3.9 p. 252 of text.Find the Fourier series of
Solution:
N12N1 1)(2N1
N1
2k
)11N2(k N1a
0
00
+=+=
ω+ω
=
}4m]-1-[n84m]-[n{4]n[x m
∑∞
∞−=
δ+δ=
-4 -3 -2 -1 0 1 2 3 4 5 n
x[n]
. . 4
8 8 8
4 4
. . . . . .
18
Here,
givinga0=3, a1 = 1-j2, a2= -1, a3= 1+j2.a4 = a0 =a-4 = .... = 3;a5 = a1 = a-3 = a9 = ... = 1-j2, etc.
Problem 3.30 p. 258 of text.Find the Fourier coefficients of x[n], y[n] and z[n] given by:
Solution:N=6 for x[n], y[n] and hence for z[n].
e21]e84[41]e8e4[
41
e]}1n[8]n[4{N1e]n[x
N1a
. 24
2 ,4N
k2j
kj1.kj0.knj
3
0n
knj
Nn
knjk
0
000
00
π−ω−ω−ω−
=
ω−
>=<
ω−
+=+=+=
−δ+δ==
π=
π=ω=
∑∑
].n[y . ]n[x]n[z
)4
n6
2sin(]n[y
)n6
2cos(1]n[x
=
π+
π=
π+=
.36
20
π=
π=ω
3620π=π=ω
19
a)
Hence,
b)
(c)
.ea
)ee(211)ee(
211n)
62cos(1]n[x
Nkk
jnjn3
jn3
j
kn0j
n0j0
∑>=<
−ωπ
−π
ω
ω
=
++=++=π
+=
.21a ,
21a ,1a 110 === −
.j2
e - b ,j2
eb ,0b
giving ,e j2
ee j2
e
]ee[j2
1)4
nsin()4
n3
sin(]n[y
4j-
14
j
10
nj4j-
nj4j
)4
n(j)4
n(j0
00
00
π
−
π
ω−
π
ω
π
π+ω−
π+ω
===
−=
−=π
+ω=π
+π
=
.0cc , cj4
ec ,cj2
ec
,4
sin21
j2e
210
j2e
21abababc
giving ,0ababababc
.0bbb and ,0aaa ,a and b ercov to N gsinChoo
.abc ]n[y].n[x
3324
j
214
j
1
4j
4j
1100110
1k1k01k1
4
1lklk
432432
1-1-
Nllklk
======
π=−+=++=
+++==
======><
=→
∗−
∗−
π
∗−
π
π−
π
−−
−+−−
−
>=<−
∑
∑
20
Fourier Series and LTI SystemsFor a system with an impulse response H(jω) or H[ejω] :
For CT: ... (15a)
For DT:... (15b)
CT: With Fourier series of x(t):
... (16)
DT:... (17)
To find the system output y(t) or y[n]: 1.Find H(jω) or H(ejw) using eqn. 15a or 15b,2. Find F. S. of x(t) or x[n].3.Find y(t) or y[n] using eqn. 16 or eqn.17.
h(t) h[n]
x(t) y(t)
x[n] y[n] H(jω) H(ejω) ejωn y[n]
ejωτ y(t)
∫∞
∞−
ωτ− ττ=ω d e )(h)j(H j
∑∞
−∞=
ω−ω =k
kjj e ]k[h)e(H
∑∞
−∞=
ω=k
ktjk , ea)t(x 0
tjk
k0k
0e )jk(H a)t(y ω∞
−∞=∑ ω=
∑>=<
ωω=Nk
nkjkjk
00 e )e(H a]n[y
21
Prob. 3.34 p. 260 of text.Given h(t)=e-4|t| , find y(t) with
(a) ,
(c) x(t) given as:
Solution:
(a)
Here, T=1, ω0 =2π.
∑∞
−∞=
−δ=n
)nt()t(x
x(t)
-1 -1/4 0 1/4 1 t
1 1 1
.j4
1j-4
1
d e ed e ed e )(h)j(H0
j4-0
j4j
ω++
ω=
τ+τ=ττ=ω ∫∫∫∞
ωτ−τ
∞−
ωτ−τ∞
∞−
ωτ−
x(t)
1
-2 -1 0 1 2 t
... ...∑∞
−∞=
ω=k
ktjk . ea)t(x 0
∫
∫ ∑∫ ∑
==
δ=−δ= ω−∞
−∞=
ω−
1
0
T
1
0
0.kjT
n
ktjk
1dt.111
dt e )t(T1dt e )nt(
T1a 00
22
Hence,
(c)
where,
Prob. 3.38 p. 261:Given:
.e ]k2j4
1k2j4
1[
e )k2j(He )kj(H a)t(y
k
kt2j
kt2j
k
ktj
k0k
0
∑
∑∑∞
−∞=
π
π∞
−∞=
ω∞
−∞=
π++
π−=
π=ω=
∑∑
∑∑
∞
−∞=
ππ∞
−∞=
π∞
−∞=
ω∞
−∞=
=π+
+π−π
π
=
π
π
=π
ω=
k
kt2jk
kt2j
k
kt2j
k
ktj
k
10
e be ]k2j4
1k2j4
1[ k
)2ksin(
)t(y
.e k
)2ksin(
e k
)kTsin()t(x 0
odd. k for ],k2j4
1k2j4
1[k
)k2
sin(
0 k for ,41
even k for ,0bk
→π+
+π−π
π
=
==
→=
∑∞
−∞=
−δ=
=−≤≤−−=
≤≤=
k
].k4n[]n[x
.else .0 1n2 ,1
2n0 ,1]n[h
23
Find the Fourier coefficients of output y[n].
Where,
h[k]
-2 -1 0 2 k
1
-1
ωωωω=
ω−−
−=
ω−
∞
−∞=
ω−ω
−++−=
+=
=
∑∑
∑
j2j2-jj-
2
0k
kj1
2k
kj
k
kjj
ee1ee
e .1e -1)(
e ]k[h)e(H
.41e
41e ]n[
41e ]n[x
N1a
.24
2 4,N here ,e a]n[x
0.k0j00
0
3
0n
knj
Nn
knjk
0Nk
njk
==δ==
π=
π=ω==
ω−
∑∑
∑
=
ω−
>=<
ω−
>=<
ω
∑
∑
∑
=
ω
ωωω−ω
=
ω−
=
ωω
=
−++−=
=
3
0k
knjk
knjk2jk2jkj3
0k
kj
3
0k
knjkjk
0
00000
00
e b41
e]ee1ee[41
e )e(H a]n[y
].ee1[41
]ee1ee[41b
k2
jk2
j
kjkjk2
jk2
jk
ππ−
ππ−ππ
−
−+=
−++−=
x[n] 1
-4 -3 -2 -1 0 1 2 3 4 n . . . . . .
24
Frequency-Shaping or Frequency-Selective Filters:Ideal Filters:Low-pass:
High-pass:
Band-pass:
Band-stop:
Non-ideal Filters:CT Filter:
25
0,
,1)j(H
c
c
ω>ω=
ω<ω=ω
|H(jω)|
-ωc 0 ωc ω
1
1,
,0)j(H
c
c
ω>ω=
ω<ω=ω
|H(jω)|
-ω2 −ω1 0 ω1 ω2 ω
1 1 else. 0
,1)j(H 21
=
ω<ω<ω=ω
|H(jω)|
-ω2 −ω1 0 ω1 ω2 ω
1 1 else. 1,
,0)j(H 21
=
ω<ω<ω=ω
R
C y(t) = v0(t) x(t) = vi(t)
|H(jω)|
-ωc 0 ωc ω
1 1
System differential equation:
The response plot:
This is a low-pass filter.
26
)t(x)t(ydt
)t(dyRC =+
0
1-
)(1
1
0
0
tjtjtj
-tan)H(j ,)H(j
RC1 ,
j1
1RCj1
1)X(j)Y(j) H(j
giving 1, )RC)H(jj(1 or
e)j(Hjdt
)t(dy ,e )H(j y(t),e)t(x
2
0
ωω
=ω∠=ω
=ω
ωω
+=
ω+=
ωω
=ω
=ωω+
ωω=ω==
ωω
+
ωωω
-ω0 0 ω0
|H(jω)|
ω
21
)j(H ω∠
ω
4π
4π
−
−ω0 ω0
DT FiltersConsider a first-order filter given by the difference equation:
y[n] – a y[n-1] = x[n], |a| < 1.
With x[n] = ejωn,
The magnitude |H(ejω)| and the phase angle for a=0.6 and a=-0.6 are shown in Fig, 3,34 p. 246 of the text.
• For 0 < a < 1 , it is low-pass filter.
• For 0> a > -1, it is high-pass filter.
27
. e a1
1)e(H
giving ,ee ]ae1)[e(H
or ,ee )e(H ae ).e(H
jj
njnjjj
nj)1n(jjnjj
ω−ω
ωωω−ω
ω−ωωωω
−=
=−
=−
)e(H jω∠
Problem 3.16 p. 253 of text.With the given impulse response H(ejω) , find the filter output with the following inputs:
Solution:
Here, N=2 and ω0=π.
None is passed.
28
)48
3sin(1]n[x )b
)1(]n[x )a
2
n1
π+
π+=
−=
n1 )1(]n[x )a −=
|H(ejω)|
π −5π/12 −π/3 0 π/3 5π/12 π ω
1 1
x1[n] 1
-2 -1 0 1 2 n
-1
. . . . . .
0). )e(H ce(sin .0
e ).e(H ae )e(H a]n[y
.1a ,0a
giving ),e1(21e)1(
21a
j
njj1
1
0kk
10
1
0n
kjknjnk
knjkj
==
==
==
−=−=
π
ππ
=
=
π−π−
π−π−
∑
∑
b)
Hence,
ω= 1 is blocked, ω= 3π/8 is passed.
29
0. others ,ej2
1a ,ej2
1a ,1a
e )e j2
1(e )e j2
1(1]ee[j2
11]n[x
.816
2 16,N )48
3sin(1]n[x
4j
34
j30
83j
4j
83j
4j)
483(j)
483(j
2
02
=−===⇒
−+=−+=
π=
π=ω=⇒
π+
π+=
π−
−
π
π−π
−πππ
+π
−π
+π
)4
n8
3sin(]ee[j2
1
e.1.e2j1e.1.e
2j1
e )e(H ae )e(H a0
e )e(H ae )e(H ae )e(H a
e )e(H a]n[y
)4
n8
3(j)4
n8
3(j
n83j
4jn
83j
4j
n8
3j8
3j3
n8
3j8
3j3
n3j3j3
n3j3j3
0j0j0
Nk
knjkjk
0000
00
π+
π=−=
−=
++=
++=
=
π+
π−
π+
π
π−
π−
ππ
π−
π−
−
ππ
ω−ω−−
ωω
>=<
ωω∑