Professional Documents
Culture Documents
Ch03 - Fourier Representations of Signals LTI Systems PDF
Ch03 - Fourier Representations of Signals LTI Systems PDF
Ch03 - Fourier Representations of Signals LTI Systems PDF
3.1 Introduction
1. A signal can be represented as a weighted superposition of complex
sinusoids.
x(t) or x[n] y(t) or y[n]
2. LTI system: LTI System
Output = A weighted superposition
of the system response to each
complex sinusoid.
3. Four distinct Fourier representations
3.2 Complex Sinusoids and Frequency Response of LTI
Systems
Frequency response The response of an LTI system to a sinusoidal input.
Discrete-time LTI system
1. Impulse response of discrete-time LTI system = h[n], input = x[n] = e j n
2. Output:
yn hk xn k hk e j n k
k k
k
Complex scaling factor
3. Frequency response:
A function of frequency
j
H (e ) h[k ]e
k
j
(3.1)
3. Frequency response:
H ( j ) h( )e j d (3.3)
Polar form complex number c = a + jb:
1
1 1 j Figure 3.2 (p. 197)
e RC
RC circuit for Example 3.1.
RC 1
j Magnitude response:
RC 0
1 1
1
0 1
RC 1 H j RC
j 2
RC 1
2
1 RC
RC Phase response:
argH j arctanRC
1
j
RC
Fig. 3.6 (a) and (b)
(t ) (t ) e j t
H ( j )e j t
e jn H e j e jn
H H H
[n] [n]
Figure 3.4 (p. 198)
Illustration of the eigenfunction property of linear systems. The action of
the system on an eigenfunction input is multiplication by the corresponding
eigenvalue. (a) General eigenfunction (t) or [n] and eigenvalue .
(b) Complex sinusoidal eigenfunction e jt and eigenvalue H(j).
(c) Complex sinusoidal eigenfunction e jn and eigenvalue H(ej).
Eigenfunction: (t ) e j t
Eigenvalue: H ( j )
2. If ek is an eigenvector of a matrix A with eigenvalue k, then
Ae k k e k
Arbitrary input = weighted superpositions of eigenfunctions
Convolution operation Multiplication
Ex. Input: Output: M
y t a k H j e
M
xt a k e jk t
j t k
k 1 k 1
Signals and Systems_Simon 6
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
x[n] ak e j k n
k 1 k 1
x[n] A[k ]e jk0n (3.4)
k
o = 2/N Fundamental
frequency of x[n]
2. x(t) = continuous-time signal with fundamental period T. FS of x(t) is
x(t ) A[k ]e jk0t (3.5)
o = 2/T Fundamental
k frequency of x(t)
“^” denotes approximate value. A[k] = the weight applied to the kth harmonic.
e jk0t is the k th harmonic of e j0t .
3. The complex sinusoids exp(jkon) are N-periodics in the frequency index k.
<pf.> e j N K 0n e jN0n e jk0n e j 2n e jN0n e jk0n
Mean-square error (MSE) between the signal and its series representation:
Discrete-time case:
1 N 1
MSE x[n] x[n] dt
2
(3.8)
N n 0
Continuous-time case:
1 T 2
MSE x(t ) x(t ) dt (3.9)
T 0
3.3.2 Nonperiodic Signals: Fourier-Transform Representations
1. FT of continuous-time signal:
1
X(j)/(2) = the weight applied to a sinusoid
x t X j e d
j t
of frequency in the FT representation.
2
Notation: x n
DTFS ; 0
X k
Example 3.2 Determining DTFS Coefficients
Find the frequency domain representation of the signal depicted in Fig. 3.5.
<Sol.>
1. Period: N = 5 o = 2/5
Signals and Systems_Simon 10
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1
X k x n e jk 2 n / 5
5 n 2
x 2e jk 4 / 5 x 1e jk 2 / 5 x0e j 0 x1e jk 2 / 5 x2e jk 4 / 5
1
5
1 1 1
X [k ] {1 e jk 2 / 5 e jk 2 / 5 }
5 2 2
(3.12)
1
{1 j sin(k 2 / 5)}
5
Signals and Systems_Simon 11
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
One period of the DTFS coefficients X[k], k = 2 to k = 2:
sin 4 / 5
X 2
1
j 0.232e j 0.531
5 5
sin 2 / 5
X 1 j
1
0.276e j 0.760
5 5
X 0 0.2e j 0
1
Fig. 3.6.
5
sin 2 / 5
X 1 j
1
0.276e j 0.760
5 5
sin 4 / 5
X 2 j
1
0.232e j 0.531
5 5 Eq. (3.11)
3. Calculate X[k] using n = 0 to n = 4:
X k
1
5
x0e j 0 x1e jk 2 / 5 x2e jk 4 / 5 x3e jk 6 / 5 x4e jk 8 / 5
1 1 1
1 e jk 2 / 5 e jk 8 / 5 e jk 8 / 5 e jk 2 e jk 2 / 5 e jk 2 / 5
5 2 2
Signals and Systems_Simon 12
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
X [k ] Magnitude spectrum of x[n]
X [k ] Even function
arg X [k ]
arg X [k ] Phase spectrum of x[n]
Odd function
Signals and Systems_Simon 15
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
Example 3.4 DTFS Representation of An Impulse Train
Find the DTFS coefficients of the N-periodic impulse train
x n n lN
l
/3
1, M n M
xn
0, M n N M
That is, each period contains 2M + 1 consecutive ones and the remaining
N (2M +1) values are zero, as depicted in Fig. 3.11. Note that this definition
requires that N > 2M +1.
<Sol.> Figure 3.11 (p. 209)
Square wave for Example 3.6.
k N / 2 1
X me
N / 2 1
xn X 0 X N / 2e jk0 n / 2
jk 0 n
X me jk0n
m 1
X k , k 0, N / 2
B k
2 X k , k 1, 2, , N / 2 1
and write the DTFS for the square wave in terms of a series of harmonically
related cosines as
N /2
x[n] B[k ]cos(k 0 n) (3.17)
k 0
<Sol.>
1. Fig. 3.14 depicts the J th term in the sum, B[J] cos(Jon), and one period of
x J [n] for the specified values of J.
2. Only odd values for J are considered, because the even-indexed coefficients
B[k] are zero when N = 25 and M = 12.
3. The approximation improves as J increases, with x[n] represented exactly
when J = N/2 = 25.
4. The coefficients B[k] associated with values of k near zero represent the low-
frequency or slowly varying features in the signal, while the coefficients
associated with values of k near N/2 represent the high frequency or
rapidly varying features in the signal.
Example 3.8 Numerical Analysis of the ECG
Evaluate the DTFS representations of the two electrocardiogram (ECG)
waveforms depicted in Figs. 3.15 (a) and (b). Fig. 3.15 (a) depicts a normal ECG,
while Fig. 3.15 (b) depicts a heart experiencing ventricular tachycardia. The
discrete-time signals are drawn as continuous functions, due to the difficulty of
depicting all 2000 values in each case. Ventricular tachycardia is a series
cardiac rhythm disturbance (i.e., an arrhythmia) that can result in death.
Signals and Systems_Simon 27
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
Figure 3.15
(p. 214)
Electrocardiograms
for two different
heartbeats and the
first 60 coefficients
of their magnitude
spectra.
(c) Magnitude
spectrum for the
normal heartbeat.
(d) Magnitude
spectrum for
ventricular
tachycardia.
xt xt dt 0
1 T
T 0
2
4
1
X k
2 2 jk
e
2 jk t
1
4 jk 2
1 e4e jk 2
1 e
4 jk 2
0
Fig. 3.17.
e jk2 = 1
The Magnitude of X[k] the magnitude spectrum of x(t);
the phase of X[k] the phase spectrum of x(t).
Figure 3.17
(p. 217)
Magnitude
and phase
spectra for
Example 3.9.
<Sol.>
1. Fundamental period of x(t) is T = 4, each period contains an impulse.
2. By integrating over a period that is symmetric about the origin 2 < t 2, to
obtain X[k]:
1 2 1
X k jk / 2 t
4 2
t e dt
4
3. The magnitude spectrum is constant and the phase spectrum is zero.
Inspection method for finding X[k]: Whenever x(t) is expressed in terms of
sinusoid, it is easier to obtain X[k] by inspection. The method of inspection is
based on expanding all real sinusoids in terms of complex sinusoids and
comparing each term in the resulting expansion to the corresponding terms
of Eq. (3.19).
j t / 2 / 4 j t / 2 / 4
e e 3 3
x t 3 e j / 4e j t / 2 e j / 4e j t / 2
2 2 2
2. Equating each term in this expression to the terms in Eq. (3.21) gives the FS
coefficients: 3 j / 4
2 e , k 1
Magnitude and
3
X [k ] e j / 4 , k 1 (3.22) phase spectrum:
2 Fig. 3.18.
0, otherwise
Signals and Systems_Simon 36
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
/4
/4
Figure 3.18 (p. 219)
Magnitude and phase spectra for Example 3.11.
Time-domain representation obtained from FS coefficients
Example 3.12 Inverse FS
Find the time-domain signal x(t) corresponding to the FS coefficients
X k 1 / 2 e jk / 20
k
Assume that the fundamental period is T = 2.
<Sol.>
1. Fundamental frequency: o = 2/T = . From Eq. (3.19), we obtain
x t 1/ 2 e 1/ 2 e jk / 20e jk t
k jk / 20 jk t k
e
k 0 k 1
xt
1 1
1
1 1 / 2e j t / 20
1 1 / 2e j t / 20
xt
3
5 4 cost / 20
Example 3.13 FS for A Square Wave
Determine the FS representation of the square wave depicted in Fig. 3.21.
Figure 3.21 (p. 221)
Square wave for Example 3.13.
<Sol.>
Signals and Systems_Simon 38
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
T0
1 jk0t
e , k 0
Tjk0 T0
2 e jk0T0 e jk0T0
, k 0
Tk0 2j
2sin k 0T0
, k 0
Tk0
For k = 0, we have
X 0 dt 0
1 T0 2T
T T0 T
3. By means of L’Hôpital’s rule, we have
Signals and Systems_Simon 39
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
Figure 3.22a&b
(p. 222)
The FS
coefficients,
X[k], –50 k 50,
for three square
waves. (see Fig.
3.21.) (a) To/T =
1/4 . (b) To/T =
1/16. (c) To/T =
1/64.
Figure 3.22c
(p. 222)
FS coefficients:
B[0] = X[0] represents the time-
1 T
B[0]
T 0
x(t )dt averaged value of the signal.
2 T
B[k ] x(t ) cos(k0t )dt (3.26)
T 0
2 T
A[k ] x(t )sin(k0t )dt
T 0
6. Substituting for H(jko) with RC = 0.1 s and o = 2, and To/T = ¼, gives
sin k / 2
Y k
10
j 2k 10 k
7. We determine y(t) using the approximation
100 The magnitude and phase spectra for the
y(t )
k 100
Y [k ]e jk0t
(3.30) range 25 k 25: Fig. 3.26 (a) and (b).
Waveform of y(t): Fig. 3.26(c).
Signals and Systems_Simon 50
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
x n X e j
3. Notation: DTFT
If x[n] is not absolutely summable, but does satisfy (i.e., if x[n] has finite
energy),
It can be shown that the sum in Eq. (3.32)
xn
2
converges in a mean-square error sense,
n but does not converge pointwise.
Example 3.17 DTFT of An Exponential Sequence
Find the DTFT of the sequence x[n] = nu[n].
<Sol.>
1. DTFT of x[n]: This sum diverges for 1
une
X e j n jn
n e jn
n n 0
X e j
1 Euler’s Formula
1 cos j sin
Signals and Systems_Simon 57
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
sin
arg X e j arctan Odd
1 cos function
Fig. 3.29 for = 0.5 and = 0.9.
Example 3.18 DTFT of A Rectangular Pulse
Let
1, n M
x n
0, n M
as depicted in Fig. 3.30 (a). Find the DTFT of x[n].
<Sol.>
1. DTFT of x[n]:
X e 1e
M
j jn
.
n M
Figure 3.29
(p.232)
The DTFT of an
exponential signal
x[n] = ()nu[n]. (a)
Magnitude
spectrum for =
0.5. (b) Phase
spectrum for =
0.5. (c) Magnitude
spectrum for
= 0.9. (d) Phase
spectrum for =
0.9.
Figure 3.30
(p. 233)
Example 3.18.
(a) Rectangular
pulse in the time
domain. (b)
DTFT in the
frequency
domain.
p
X e e
2M 2M
Change of variable
j
m0
j ( m M )
e j M
e jm
m0
m=n+M
jM 1 e j 2( M 1)
e , 0, 2 , 4 ,
1 e j
2 M 1, =0, 2 , 4 ,
2. The expression for X(e j ) when 0, 2 , 4 ,
j
X (e ) e j M
e j 2 M 1 / 2 e j 2 M 1 / 2 e j 2 M 1 / 2
e j / 2 e j / 2 e j / 2
e
j 2 M 1 / 2
e
j 2 M 1 / 2
e j / 2 e j / 2
sin 2 M 1 / 2 L’Hôptital’s Rule
lim 2 M 1;
0,2 ,4 , , sin 2
sin 2M 1 / 2 With understanding that X(e j ) for
j
X (e ) 0, 2 , 4 , is obtained as
sin 2 limit.
Signals and Systems_Simon 61
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1 jn W
e W , n 0
2 nj
1
sin Wn , n 0.
n
3. For n = 0, the integrand is unity and we have x[0] = W/. Using L’Hôpital’s
rule, we easily show that
1 W
lim sin Wn ,
n 0 n
and thus we usually write
1
x n sin Wn
n
as the inverse DTFT of X(e j ), with the understanding that the value at n =
0 is obtained as limit.
4. We may also write
W
x n sin c Wn / , Fig. 3.31 (b).
Signals and Systems_Simon 63
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
Example 3.20 DTFT of The Unit Impulse
Find the DTFT of x[n] = [n].
<Sol.>
1. DTFT of x[n]:
X e j
n e jn
1 Figure 3.32 (p. 235)
n
Example 3.20. (a) Unit impulse
in the time domain. (b) DTFT of
n
DTFT
1. unit impulse in the frequency
2. This DTFT pair is depicted in Fig. 3. 32. domain.
1
x n e jn d . Sifting property of impulse function
2
1
DTFT
, .
2
We can define X(e j ) over all by writing it as an infinite sum of delta
functions shifted by integer multiples of 2.
p
2. This DTFT pair is depicted in Fig. 3. 33.
y1 n
1
2
x n x n 1 and y2 n
1
2
x n x n 1
The first system averages successive inputs, while the second forms the
difference. The impulse responses are
1 1 1 1
h1 n n n 1 and h2 n n n 1
2 2 2 2
Find the frequency response of each system and plot the magnitude responses.
<Sol.>
1. The frequency response is the DTFT of the impulse response.
2. For system # 1 h1[n] :
Frequency response:
1 1 j
e
j j
H1 (e j ) e j j
e
2 2
j e 2
cos .
2 2 H1 (e ) e 2
2
2
Signals and Systems_Simon 66
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
H2 e j
sin ,
2 , 0
2 arg H 2 ( e j ) 2
p
H e j 1 a e
j arga
. Apply Euler’s formula
1 a 2 a cos arg a
2 12
,
H inv e j
1 Expressing a in polar form
j
, a 1
1 ae
1 2
1 a 2 a cos arg a
2
,
The frequency response of the inverse system is the inverse of the frequency
response of the original system.
5. Magnitude response of the inverse system:
H inv e j
1
.
1 a 2 a cos arg a
2 12
6. Fig. 3.37 depicts the magnitude response of H(e j ) for both a = 0.5 e j/3 and
a = 0.9 e j2/3 on the interval < .
7. The magnitude response approaches a maximum of 1 + a when = arg{a}
and a minimum of 1 a when = arg{a} .
8. The magnitude response of the corresponding inverse system: Fig. 3.38.
Signals and Systems_Simon 70
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
Frequency-domain representation of
X ( j ) x(t )e j t
dt (3.36) the signal x(t)
1 1
e
a j t
a j a j
0
a 2
2 2
Fig. 3.39 (b) and (c).
arg X j arctan a ,
arg{X(j)}
/4
/4
/4
/2
Signals and Systems_Simon 76
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
Example 3.25 FT of A Rectangular Pulse
Consider the rectangular pulse depicted in Fig. 3.40 (a) and defined as
1, T0 t T0
x t
0, t T0
Find the FT of x(t).
<Sol.>
1. The rectangular pulse x(t) is absolutely integrable, provided that To < .
2. FT of x(t):
X j x t e j t dt e j t dt
T0
T0
With understanding that
the value at = 0 is
1 j t 2
e T0
T0 sin T0 , 0 obtained by evaluating
j a limit.
3. For = 0, the integral simplifies to 2To.
2 2
lim sin T0 2T0 . X j sin T0 ,
0
4. Magnitude spectrum:
Signals and Systems_Simon 77
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
sin T0
X j 2 , Fig. 3. 40 (b).
5. Phase spectrum: 6. X(j) in terms of sinc function:
0, sin T0 0 X j 2T0 sinc T0 .
arg X j
, sin T0 0
As To increases, the nonzero time
extent of x(t) increases, while X(j) v ; p
becomes more concentrated about
the frequency origin.
v ; p
1. Inverse FT of x(t):
1 1 j t 1
x t
W
e j t d sin(Wt ), t 0
W
e W
2 W j t t
2. When t = 0, the integral simplifies to W/, i.e.,
Signals and Systems_Simon 79
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1
lim sin Wt W , With understanding that the value
t 0 t
at t = 0 is obtained as a limit.
3. Inverse FT is usually written as
1 W Wt
x t sin Wt , or x t sinc , Fig. 3. 42 (b).
t
As W increases, the frequency-domain reprsentation becomes less
concentrated about = 0, while the time-domain representation X(j)
becomes more concentrated about t = 0.
Duality between Example 3.25 and 3.26.
Example 3.27 FT of The Unit Impulse
Find the FT of x(t) = (t).
<Sol.>
1. x(t) does not satisfy the Dirichlet conditions, since the discontinuity at the
origin is infinite.
2. FT of x(t):
X j t e
j t
dt 1
A, t T0 / 2
xr t
0, t T0 / 2
and a raised-cosine pulse defined as Figure 3.44 (p. 249)
Pulse shapes used in
Ac / 2 2cos 2 t / T0 ,
t T0 / 2
xc t
BPSK communications.
0, t T0 / 2 (a) Rectangular pulse.
(b) Raised cosine pulse.
The transmitted BPSK signals for communicating a sequence of bits using each
pulse shape are illustrated in Fig. 3.45. Note that each pulse is To seconds long,
Signals and Systems_Simon 82
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
so this scheme has a transmission rate of 1/To bits per second. Each user’s
signal is transmitted within an assigned frequency band, as depicted in Chapter
5. In order to prevent interference with users of other frequency bands,
governmental agencies place limits on the energy of a signal that any user
transmits into adjacent frequency bands. Suppose the frequency band
assigned to each user is 20 kHz wide. Then, to prevent interference with
adjacent channels, we assume that the peak value of the magnitude spectrum
of the transmitted signal outside the 20-kHz band is required to be 30 dB
below the peak in-band magnitude spectrum. Choose the constant Ar and Ac so
that both BPSK signals have unit power. Use the FT to determine the maximum
number of bits per second that can be transmitted when the rectangular and
raised-cosine pulse shapes are utilized.
<Sol.>
1. BPSK signals are not periodic, but their magnitude squared is To periodic.
2. Powers in rectangular pulse and raised-cosine pulse:
1 T0 / 2 2
Pr Ar dt Ar2
T0 T0 / 2
and
Signals and Systems_Simon 83
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1
A 4 1 2cos 2 t 2
T0 2
2
Pc 2
c dt
T0 T0 2
Ac2
1 2cos 2 t T0 1 2 1 2cos 4 t T0 dt
T0 2
4T0 T0 2
3 Ac2
8
Hence, unity transmission power is obtained by choosing Ar = 1 and Ac = 8 / 3 .
3. FT of the rectangular pulse xr(t):
sin T0 2
X r j 2
In terms of Hz rather than rad/sec ( = 2f): X r ( j ) X r ( jf )
sin fT0
X r jf 2
f
Normalized Spectrum of 20log10 X r ( jf ) / To dB: Fig. 3.46.
Figure 3.46
(p. 250)
Spectrum of
rectangular
pulse in dB,
normalized by
T0.
The normalized
by To removed
the
dependence of
the magnitude
on To.
X c j
1 8 T0 2
2 3 T0 2
1 2cos 2 t T0 e j t
dt
Euler’s formula
2 T0 2 j t 1 2 T0 2 j 2 T0 t 1 2 T0 2 j 2 T0 t
X c j e dt e dt e dt
3 0 T 2
2 3 0 T 2
2 3 0 T 2
Figure 3.48
(p. 252)
Spectrum of
the raised-
cosine pulse in
dB, normalized
by T0.
X k 1 k sin k 4
From Example 3.13.
x t
FS ;2
y t
FS ;2
Y k 1 k sin k 2
Linearity
3. FS of z(t):
z t
FS ;2
Z k 3 2k sin k 2 1 2k sin k 2
3.9.1 Symmetry Properties: Real and Imaginary Signals
Symmetry property for real-valued signal x(t):
1. FT of x(t):
*
X ( j ) x(t )e dt x* (t )e j t dt
* j t
(3.37)
X * ( j ) X ( j ) (3.38)
Re X ( j ) ReX ( j ) and ImX ( j ) ImX ( j )
Signals and Systems_Simon 94
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
x t A / 2 e j t A / 2 e j t
2. Real-valued impulse response of LTI system is denoted by h(t).
3. Output signal of LTI system:
y(t ) H ( j ) ( A / 2)e j (t argH ( j )) H ( j ) ( A / 2)e j (t argH ( j ))
Exploiting the symmetry conditions:
Applying Eq. (3.2)
H ( j ) H ( j ) and linearity
argH ( j ) argH ( j )
y t H j A cos t argH j
Signals and Systems_Simon 96
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
The LTI system modifies the amplitude of the input sinusoid by H(j)and
the phase by arg{H(j)}.
v ; f
Fig. 3.50.
y n H e j A cos n arg H e j
The LTI system modifies the amplitude of the input sinusoid by H(e j)and
the phase by arg{H(e j)}.
x(t) is purely imaginary:
1. x(t) = x(t).
2. Eq.(3.37) becomes
X *
j x t e j ( )t dt
X * ( j ) X ( j ) (3.39)
Re X ( j ) Re X ( j ) and ImX ( j ) ImX ( j )
3.9.2 Symmetry Properties: Even and Odd Signals
1. x(t) is real valued and has even symmetry.
x(t) = x(t) and x ( t) = x(t) x(t) = x( t)
Signals and Systems_Simon 98
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
2. Eq.(3.37) becomes
Change of variable = t
X *
j x t e j ( t )
dt
X * j x e j d X j
3. Conclusion:
1) The imaginary part of X(j) = 0: X ( j ) X ( j )
If x(t) is real and even, then X(j) is real.
2) If x(t) is real and odd, then X(j) = X(j) and X(j) is imaginary.
3.10 Convolution Properties
The convolution property is a consequence of complex sinusoids being
eigenfunctions of LTI system.
3.10.1 Convolution of Nonperiodic Signals — Continuous-time case
1. Convolution of two nonperiodic continuous-time signals x(t) and h(t):
y t h t x t h x t d
2. FT of x(t ):
Signals and Systems_Simon 99
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1
x t X j e j ( t )d
2
0, 0, 2
Since y t h t x t
FT
Y j X j H j
1,
Y j y(t) = (1/(t))sin(t)
0,
Example 3.32 Finding Inverse FT’s by Means of The Convolution Property
Use the convolution property to find x(t), where
4
x t
FT
X j sin 2
2
<Sol.>
2
1. Write X(j) = Z(j) Z(j), where Z j sin
2. Convolution property:
z t z t
FT
Z j Z j
Fig. 3.52 (a)
x t z t z t
1, t 1
3. Using the result of Example 3.25: z
t
FT
Z j
0, t 1
4. Performing the convolution with itself gives the triangular waveform depicted
in Fig. 3.52 (b) as the solution for x(t).
Figure 3.52 (p. 261)
Signals for
Example 3.32. (a)
Rectangular pulse
z(t). (b)
Convolution of z(t)
with itself gives x(t).
Signals and Systems_Simon 102
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
v ; p
6. The edge of the passband is usually defined by the frequencies for which
the response is 3 dB, corresponding to a magnitude response of (1/ 2 ).
Unity gain = 0 dB
7. Energy spectrum of filter: Y j H j X j
2 2 2
Plot the magnitude responses of both systems on a linear scale and in dB, and
characterize the filtering properties of the systems.
<Sol.>
1
1. Frequency response corresponding to hC(t): H C j
i RC 1
i RC
2. Frequency response corresponding to hR(t): H R j
i RC 1
3. Magnitude response in linear scale: Fig. 3-55 (a) and (b).
Magnitude response in dB scale: Fig. 3-55 (c) and (d).
4. System corresponding to yC(t):
Low-pass filter H C j and H R j
Cutoff frequency = c = 1/(RC)
System corresponding to yR(t):
High-pass filter
Cutoff frequency = c = 1/(RC)
v
j Y (e j )
For DT system: H (e ) (3.43)
X (e j )
Example 3.34 Identifying a System, Given Its Input and Output
The output of an LTI system in response to an input x(t) = e 2 t u(t) is y(t) =
e t u(t) . Find the frequency response and the impulse response of this system.
<Sol.>
1 1
1. FT of x(t) and y(t): X j and Y j
j 2 j 1
2. Frequency response:
Y j j 2
H j H j
X j j 1
3. Impulse response:
j 1 1 1
H j 1 h t t et u t
j 1 j 1 j 1
Recover the input of the system from the output:
CT case DT case
DTFT
H inv e j H e j 1
Signals and Systems_Simon 111
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
H inv e j
1 Frequency response of inverse system
H e j
3. DTFT of h[n]:
h n
DTFT
H e j 1 ae j
H inv e j
1
1 ae j
The frequency response of the inverse system is then obtained as
hinv n a u n
n
1. Define the periodic convolution of two CT signals x(t) and z(t), each having
period T, as
y t x t # z t x z t d
T
where the symbol denotes that integration is performed over a single period
of the signals involved.
2
FS ;
y (t ) x(t ) # z(t ) Y [k ] TX [k ]Z [k ]
T (3.44)
3. FS Coefficients of y(t):
1/ k 1
Y k X k Z k y t 2 cos 2 t
0 otherwise
Explanation for the origin of the Gibbs phenomenon:
Example 3.14:
1. A partial-sum approximation to the FS representation for x(t) may be obtained
by using the FS coefficients
Xˆ J k X k W k
T=1
where
1 J k J
W k
0 otherwise
2. In the time domain, xˆ j (t ) is the periodic convolution of x(t) and
sin t 2 J 1 T=1
w t
sin t
(3.45)
2
DTFS ;
y[n] x[n] # z[n]
Y [k ] NX [k ]Z [k ]
N (3.46)
d
x(t )
FT
j X ( j )
dt
◆ Differentiation of x(t) in Time-Domain (j) X(j) in Frequency-Domain
d
F x (t ) ( j ) X ( j ) 0 0
dt 0
d at
dt
e u t aeat u t e at t ae at u t t
a j Y j b j X j
k k
a k k
y t bk x t k k
k 0 dt k 0 dt k both sides k 0 k 0
M
bk j
k
Y j
k 0
X j
M
bk j
N
a j
k
k
H j
k
k 0
k 0 (3.47)
N
a j
k
2. Frequency response of the system: k
k 0
Signals and Systems_Simon 119
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
H j
1
j 2 n j n2
Q
2. Magnitude response for (a) Q = 2/5, (b) Q = 1, and (c) Q = 200: Fig. 3.58.
Discussion: Refer to textbook, p. 273.
n = 10,000 rads/s Resonant condition!
Signals and Systems_Simon 120
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
v
d
F x (t ) ( jk0 ) X [k ] k 0 0
dt k 0
Example 3.39 Differentiation Property
Use the differentiation property to find the FS representation of the triangular
wave depicted in Fig. 3. 59 (a).
<Sol.> Fig. 3. 59 (b)
d
1. Define a waveform: z ( t ) y (t )
dt
Signals and Systems_Simon 122
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
g (t ) 1/ 2 e t 2 / 2
D.E. of g(t)
<Sol.> p
1. The derivative of g(t) with respect to t:
d
g (t ) (t / 2 )e t 2 / 2
tg (t ) (3.48)
dt
2. Differentiation-in-time property:
d
g t
FT
jG j
dt
(3.48) tg (t )
FT
jG( j ) Figure 3.60 (p. 275)
(3.49)
Gaussian pulse g(t).
d
3. Differentiation-in-time property: jtg t
FT
G j
d
Signals and Systems_Simon 125
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1 d
tg (t )
FT
G( j ) (3.50)
j d
Since the left-hand sides of Eqs. (3.49) and (3.50) are equal, then we have
D.E. of G(j)
G j G j
d
d
G j ce 2 2
1/ 2 e dt 1
G j0 t 2 / 2
The FT of a Gaussian pulse is
also a Gaussian pulse!
1/ 2 e e
t 2 / 2 2
FT /2
3.11.3 Integration
The operation of integration applies only to continuous dependent variables.
In both FT and FS, we may integrate with respect to time.
In both FT and DTFT, we may integrate with respect to frequency.
Signals and Systems_Simon 126
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
t 1
x( )d X ( j ) X ( j 0) ()
FT
(3.53)
j
Ex. Demonstration for the integration property by deriving the FT of the unit step
1. Unit step signal:
u t d
t
2. FT of (t): (t )
FT
1
1
3. Eq. (3.53) suggests that: u t
FT
U j
j
Check:
1. Unit step: 1 1
u (t ) sgn(t ) (3.54)
2 2
Fig. 3.61
2. Signum function: 1, t 0
sgn t 0, t 0
1, t 0
x t t0
FT
e jt0 X j
x t t0
FT ; 0
e jk0t0 X k
x n n0
DTFT
e jn0 X e j
x n n0
DTFS ; 0
e jk0n0 X k
a yn k b xn k
k 0
k
k 0
k
e jk Z e j
Taking DTFTof both
z n k
DTFT
Y e b e X e
N M
ak e
k 0
j k j
k 0
k
j k j
M
bk e j
k
Y e j
k 0
a e
j N
X e j k
k
k 0
Signals and Systems_Simon 133
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
M
bk e j
k
2. Frequency response:
H e j k 0
(3.55)
a e
N
j k
k
k 0
e j t x t
FT
X j
e jk00t x t
FS ; 0
x k k0
e jn x[n]
DTFT
X e j
e jk00n x[n]
FS ; 0
X k k0
Example 3.42 Finding an FT by Using the Frequency-Shift Property
Use the frequency-shift property to determine the FT of the complex sinusoidal
pulse:
e ,
j10t
t
z t
0, t
<Sol.>
1. Express z(t) as the product of a complex sinusoid ej10t and a rectangular pulse
Signals and Systems_Simon 135
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1, t
xt
0, t
2. Use the results of Example 3.25, we have
2
x t
FT
X j sin
2
Frequency-shift property x t
FT
X j sin
3. FT of z(t):
X j 10 sin 10
2
e j10t x t
FT
z t
FT
10
Example 3.43 Using multiple Properties to Find an FT
Find the FT of the signal
xt
d
dt
e ut * e u t 2
3t t
<Sol.>
1. To solve this problem, it needs to use three properties: differentiation in time,
convolution, and time shifting.
Signals and Systems_Simon 136
Haykin & Barry Van Veen
X j jW j V j
CHAPTER
Fourier Representations of Signals & LTI Systems
X j jW j V j
Convolution and differentiation property
W j
1 1
e at u t
FT
a j 3 j
4. Use the same transform pair and the time-shift property to find V(j) by first
writing j 2
V j e 2
e
vt e 2 e t 2 ut 2
1 j
5. FT of x(t):
je j 2
X j e 2
1 j 3 j
X j
j N a N 1 j N 1 a1 j a0 A j
Assume that M < N. If M N, then we may use long division to express X(j)
in the form
Partial-fraction expansion
B j
A j
M N
X j f j k
k
is applied to this term
k 0
Applying the differentiation property
2. Let the roots of the denominator
and the pair (t ) FT
1 to these
A(j) be dk, k = 1, 2, …, N. These
terms
roots are found by replacing j
with a generic variable v and determining the roots of the polynomial
Signals and Systems_Simon 138
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
v N a N 1v N 1 a1v a0 0
5. For M < N, we may the write
M
bk j
k
X j k 0
N
Ck, k = 1, 2, …, N are determined
j d
k 1
k
by the method of residues
Assuming that all the roots dk, k = 1, 2, …, N, are distinct, we may write
N
Ck
X j Appendix B
k 1 j d k
In Example 3.24, we derived the transform pair
1
edt u t
FT This pair is valid even if d is complex,
j d provided that Re{d} < 0.
N N
Ck
x t Ck e u t
dk t FT
X j
k 1 k 1 j d k
Example 3.44 MEMS Accelerometer: Impulse Response
Find the impulse response for the MEMS accelerometer introduced in Sec. 1.10,
assuming that n = 10,000 rads/s, and (a) Q = 2/5, (b) Q = 1, and (c) Q = 200.
<Sol.>
Frequency response for the MEMS accelerometer:
1
H j
n
j j n2
2
Q
Case (a): n = 10,000 rads/s and Q = 2/5
1. Frequency response:
H j
1
j 2 25000 j 100002
Signals and Systems_Simon 140
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
2. Partial-fraction expansion:
1 C1 C2
j 2 25000 j 100002 j 20000 j 5000
The roots of the denominator polynomial are d1 = 20,000 and d2 = 5,0000.
Coefficients C1 and C2:
1
C1 j 20,000
j 25000 j 10000
2 2
j 20,000
1
1/15000
j 5,000 j 20,000
1
C2 j 5000
j 25000 j 10000
2 2
j 5,000
1
1/15,000
j 20,000 j 5,000
H j
1
j 2 10000 j 100002
2. Partial-fraction expansion: d1 5000 j5000 3
The roots of the denominator polynomial are
d 2 5000 j5000 3
H j
j / 10000 3
j / 10000 3
j 5000 j5000 3 j 5000 j5000 3
3. Impulse response:
h t j 10000 3 e 5000t e j 5000 3t
e 5000t e j 5000 3t
1/ 5000 3 e 5000t sin 5000 3t u t
Case (c): n = 10,000 rads/s and Q = 200
1. Frequency response:
The roots of the denominator
H j
1 polynomial are
j 2 50 j 100002 d1 25 j10,000
2. Impulse response: d 2 25 j10,000
ht 1 / 10000e 25t sin10000t ut
Impulse response for (a) Q = 2/5, (b) Q = 1, and (c) Q = 200: Fig. 3.64 (a) ~ (c).
1. For both Q = 2/5 and Q = 1, the impulse response is approximately zero for
t > 1 ms.
2. When Q = 200, the accelerometer exhibits resonant nature.
A sinusoidal oscillation of n = 10,000 rads/sec.
N
Ne jN
N 1e j N 1
1e j
1 1 d k e j
k 1
3. Partial-fraction expansion:
Assuming that M < N and all the dk are distinct, we may express X(e j )as
C
N
j k
X e j
1 d ek 1 k
Since
Signals and Systems_Simon 145
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1 Expansions for
d k u n
n DTFT
repeated roots are
1 d k e j
treated in Appendix B.
The linearity property implies that
N
xn C k d k un
n
k 1
5
e j 5
6 C1 C2
1 1 1 1
1 e j e j 2 1 e j 1 e j
6 6 2 3
Coefficients C1 and C2
5 5
e j 5 e j 5
1 6
C1 1 e j 6 4
2 1 1 e j 1 e j 2 1 j
1 e
6 6 e j 2 3 e j 2
5 5
e j 5 e j 5
1 6
C2 1 e j 6 1
3 1 1 e j 1 e j 2 1 j
1 e
6 6 e j
3 2 e j 3
yt X jv Z j e j v t ddv
1
2 2
Change variable: = v Inner Part: Z(j) X(j)
1 jt
1
y t 2 X jv Z j v dv e d
2
3. FT of y(t): Outer Part: FT of y(t)
1
y (t ) x(t ) z(t )
FT
Y ( j) X ( j) * Z ( j) (3.56)
2
Scaled by 1/2
Signals and Systems_Simon 148
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
where X j * Z j X jv Z j v dv
1
y[n] x[n]z[n]
DTFT
Y (e j ) X (e j ) # Z (e j ) (3.57)
2
where the symbol denotes periodic convolution.
Here, X(e j ) and X(e j ) are 2-periodic, so we evaluate the convolution over
a 2 interval:
X j # Z j X e j Z e j d
1. FT of y(t):
1
y t
FT
Y j X j *W j
2
2. If w(t) is the rectangular window depicted
in Fig. 3. 65 (b), we have p;v
W j sin T0
2
Smoothing the details
in X(j) and introducing
oscillation near
discontinuities in X(j).
v , u
v , u
4. Since
1, /2 On the basis of
He j
Example 3.18
0, / 2
sin 2 M 1 / 2
and
W e j sin / 2
/2 Discussion: refer to
Ht e j
1
2 F d
p. 295 in textbook.
/2
Figure 3.69
(p. 298)
FS magnitude
spectrum for
0 k 1000.
The result is
depicted as a
continuous
curve, due to
the difficulty
of displaying
1000 stems
◆ Fundamental period = N.
◆ The multiplication properties of all four Fourier representations are
summarized in Table 3.9.
Table 3.9 Multiplication Properties of Fourier Representations
1
x ( t ) z (t ) FT
X ( j ) Z ( j )
2
FS ; o
x ( t ) z (t ) X [k ] Z [k ]
1
x[n ]z[n ]
DTFT
X ( e j ) # Z ( e j )
2
DTFT ; o
x[n]z[n ] X [k ] # Z [k ]
Signals and Systems_Simon 160
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
(1/ a ) x( )e j ( / a ) d , a 0
Z ( j )
(1/ a ) x( )e j ( / a ) d , a 0
Z ( j) (1/ a ) x( )e j ( / a ) d ,
v
Figure 3.71
(p. 301)
Application of the
FT scaling
property in
Example 3.48. (a)
Original time
signal. (b)
Original FT. (c)
Scaled time
signal y(t) = x(t/2).
(d) Scaled FT
Y(j) = 2X(j2).
v
<Sol.>
1. Differentiation in frequency, time shifting, and scaling property to be used to
solve the problem.
2. Transform pair: 1
s(t ) e u(t ) S ( j )
t FT
1 j
X ( j ) j
d
d
e j 2 S ( j / 3)
3. Treatment procedure: scaling time-shift differentiation
Define Y(j) = S(j/3).
y(t ) 3s(3t ) 3e3t u(3t ) 3e3t u(t )
Define W(j) = e j2 Y(j/3).
Differentiation
w(t ) y(t 2) 3e3(t 2)u(t 2) property
Since d
X ( j ) j W ( j ) x(t ) tw(t ) 3te3(t 2)u(t 2)
d
FS ; a0
z(t ) x(at ) 2Z [k ] X [k ], a0 (3.61)
1
Wx x(t ) X ( j )e jt d dt
2
Wx
1
2
X ( j )
x
(t )e jt
dt d
1
Wx
2
X ( j ) X ( j )d
Signals and Systems_Simon 167
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1
x(t ) dt X ( j ) d
2 2
(3.62)
2
Normalization!
◆ Energy in Time-Domain Representation
Energy in Frequency-Domain Representation (1/2)
◆ The Parseval Relationships of all four Fourier representations are
summarized in Table 3.10.
Table 3.10 Parseval Relationships for the Four Fourier Representations
Representations Parseval Relationships
1
x(t ) dt d
2 2
FT X ( j )
2
1 T
k
T 0
2 2
FS x (t ) dt X [ k ]
1
n
2 j 2
DTFT x [ n ] X ( e ) d
2
1
k 0
N 1 2 N 1 2
DTFS x[ n ] X [ k ]
N n 0
Signals and Systems_Simon 168
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
2. Since W
j
1,
x[n] DTFT
X (e )
0, W
1 W
2 W
1d W /
v , p
4. The product of the time extent T0 and mainlobe width 2/T0 is a constant.
5. Compressing a signal in time leads expansion in the frequency domain and
vice versa.
Signla’s time-bandwidth product !
6. Bandwidth: The extent of the signal’s significant frequency content.
1) A mainlobe bounded by nulls.
Ex. Lowpass filter One half the width of mainlobe.
2) The frequency at which the magnitude spectrum is 1/2 times its peak values.
7. Effective duration of signal x(t):
1/ 2
t 2 x(t ) 2 dt
Td (3.63)
dt
2
x (t )
8. Effective bandwidth of signal x(t):
1/ 2
2 X ( j ) 2 d
Bw (3.64)
2
X ( j ) d
Signals and Systems_Simon 171
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
Td Too
T
To dt
Signals and Systems_Simon 172
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1/ 2
To t 2dt
T
Td Too [(1/(2To ))(1/ 3)t 3 To 1/ 2
] Td / 3
To
To
dt
1
x(t )
2
X ( j )e jt d and X ( j ) x(t )e jt dt
2. General equation:
1
y (v )
2
z ( )e jv dv (3.66)
v ; p
1
y (t )
2
z ( )e jt d
y(t )
FT
z() (3.67)
3. Interchange the role of time and frequency by letting = and = t, then
Eq. (3.66) implies that
1
y ( )
2
z (t )e jt dt
z(t )
FT
2 y( ) (3.68)
Fig. 3.74
4. If we are given by an FT pair
z(t )
FT
2 y( ) (3.69) f (t )
FT
F ( j) (3.70)
Example 3.52 Applying Duality
Find the FT of 1
x(t )
1 jt
Signals and Systems_Simon 175
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
v ; p
<Sol.> 1
1. Note that: f (t ) e t u(t )
FT
F ( j )
1 j
2. Replacing by t, we obtain
1
F ( jt ) x(t) has been expressed as F(jt).
1 jt
3. Duality property:
2
DTFS ; N = time index; k =
x[n] N
X [k ] (3.71)
frequency index
2
DTFS ; 1
X [n] N
x[k ] (3.72)
N
3.18.3 The Duality Property of the DTFT and FS
1. FS of a periodic continuous time signal z(t):
z (t ) Z[k ]e
k
jkot
Require
3. Duality relationship between z(t) and X(e j) T = 2
In the DTFT t in the FS
Assumption: 0 = 1
n In the DTFT k in the FS
Signals and Systems_Simon 178
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
4. FS coefficients Z[k]:
1 In the DTFT t in the FS
z(t )e
jkt
Z [k ] dt n In the DTFT k in the FS
2
x[n]
DTFS
X (e j ) (3.73)
X (e jt )
FS ;1
x[k ] (3.74)
◆ The Duality Properties of Fourier representations are summarized in
Table 3.11.
Example 3.53 FS-DTFT Duality
Use the duality property and the results of Example 3.39 to determine the
inverse DTFT of the triangular spectrum X(e j) depicted in Fig. 3. 75 (a).
<Sol.>
Signals and Systems_Simon 179
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
FT f (t )
FT
F ( j ) F ( jt )
FT
2 f ( )
DTFS ; 2 / N DTFS ; 2 / N
DTFS x[n] X [k ] X [n ] (1/ N ) x[ k ]
FS-DTFT x[n ]
DTFS
X ( e j ) X (e jt )
FS ;1
x[k ]
p
z(t )
FS ;1
Z [k ] x[n] Z [n]
Hence, we seek the FS coefficients Z[k] associated with z(t).
3. Assuming that T = 2, z(t) is a time-shifted version of the triangular wave y(t),
i.e.
z(t ) y(t / 2)
, k 0
Z [k ] e jk / 2Y [k ] 4 j k 1 sin(k / 2)
, k 0
k 2
h[n]* e jn , lb n kh lv
H (e j )e jn , lb n kh lv
Signals and Systems_Simon 182
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
1) Magnitude: y[n] H (e j ) , lh n kh lv
Amplitude
(p. 314)
Sinusoidal 0
steady-state
response
computed with -0.5
the use of 0 10 20 30 40 50 60
MATLAB. The Time
values at times 1 imag(y1)
0.5
through 50
represent the
Amplitude
sinusoidal steady-
state response. 0
-0.5
0 10 20 30 40 50 60
Time
Signals and Systems_Simon 185
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
2) MATLAB commands for obtaining the magnitude and phase components from
any element of the vectors of y1 and y2:
Ex. Find the fifth element:
>> H1mag=abs(y1(5)) The phase response
H1mag = is measured in radians.
0.3827
>> H2mag=abs(y2(5))
H2mag =
0.9239
>> H1phs=angle(y1(5))-omega1*5
H1phs =
-5.8905
>> H2phs=angle(y2(5))-omega2*5
H2phs =
-14.5299
The angle command always returns a value between and radians.
Command: angle(y1(n))-omega1*n may result in answers that differ by integer
multiples of 2 when different values of n are used.
Signals and Systems_Simon 186
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
2. MATLAB commands fft and ifft may be used to evaluate the DTFS.
1) Given a length-N vector x representing one period of an N periodic signal x[n],
the command
MATLAB run from 0 to N 1
>> X=fft(x)/N
produces a length-N vector X containing the DTFS coefficients X[k].
◆ The division by N is necessary because fft evaluates the sum in Eq. (3.11) without
dividing by N.
2) Given DTFS coefficients in a vector X, the command
>> x=ifft(X)*N
produces a length-N vector x that represents one period of the time-domain
waveform.
Signals and Systems_Simon 187
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
Example
Consider using MATLAB to solve Problem 3.3 (a) for the DTFS coefficients.
The signal is
3
x[n] 1 sin n /12
8
<Sol.>
>> x=ones(1,24)+sin([0:23]*pi/12+3*pi/8);
1. Period = 24. 2. MATLAB command:
>> X=fft(x)/24
Output:
X=
Columns 1 through 5
1.0000 0.4619 - 0.1913i 0.0000 + 0.0000i -0.0000 + 0.0000i -0.0000 + 0.0000i
Columns 6 through 10
-0.0000 - 0.0000i 0.0000 - 0.0000i 0.0000 - 0.0000i 0.0000 + 0.0000i -0.0000 - 0.0000i
Columns 11 through 15
-0.0000 - 0.0000i -0.0000 - 0.0000i 0 -0.0000 + 0.0000i -0.0000 + 0.0000i
Columns 16 through 20
-0.0000 + 0.0000i 0.0000 - 0.0000i 0.0000 + 0.0000i 0.0000 + 0.0000i -0.0000 + 0.0000i
Columns 21 through 24
-0.0000 - 0.0000i -0.0000 - 0.0000i 0.0000 - 0.0000i 0.4619 + 0.1913i
ans =
1.9239-0.0000i 1.9914+0.0000i 1.9914+0.0000i 1.9239-0.0000i
Example
The partial-sum approximation used in Example 3.7 is easily evaluated in MATLAB:
>> k=1:24;
>> n=-24:25;
>> B(1)=25/50; % coeff for k=0
>> B(2:25)=2*sin(k*pi*25/50)./(50*sin(k*pi/50));
>> B(26)=sin(25*pi*25/50)./(50*sin(25*pi/50)); % coeff for k =N/2
>> xjhat(1,:)=B(1)*cos(n*0*pi/25);
% term in sum for k=0
%accumulate partial sums
>> for k=2:26
xjhat(k,:)=xjhat(k-1,:)+B(k)*cos(n*(k-1)*pi/25);
end
This set of commands produces a matrix xjhat whose (J + 1)st row corresponds to
xˆ J [n ]
Signals and Systems_Simon 190
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
3.19.3 The FS
Trigonometric FS expression of x(t) in Example 3.14:
x(t ) B[k ]cos(k0t ) (3.29)
k 0
where
B[0] 2T0 / T
2sin(k 2 T0 / T )
B[k ] , k 0 (3.28)
k
Let the partial-sum approximation to the FS in Eq. (3.29), be given by
J
x J t B k cos k0t
k 0
bk j
k
H j k 0
N (3.47)
a j
k
k
k 0
Example
The frequency response of the MEMS accelerometer depicted in Fig. 3.58 for Q =1
is obtained via the MATLAB commands:
-155
>> w=[0:100:100000];
>> b=1; -160
>> a=[1 10000 10000*10000]; -165
>> H=freqs(b,a,w);
>> plot(w,20*log10(abs(H))) -170
-175
3. MATLAB command:
h=freqz(b,a,w) returns the -180
b e
M
j k
k
H e j
k 0
(3.55)
a e
N
j k
k
DT periodic
k 0 signals
3.19.5 Time-Bandwidth Product
1. Command fft can be used to evaluate DTFS and explore the time-bandwidth
product property.
2. Example 3.6
1, n M
1) Time-domain signal: x[n]
0, M n N M
1, n M
2) DTFS coefficients: x[n]
0, M n N M
3) Parameter definitions:
Td = 2M +1 =the nonzero portion of one period of x[n]
Signals and Systems_Simon 195
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
Bw N/(2M +1) = Bandwidth = the “frequency” of the first null of X[k]
The time-bandwidth product for the square wave, Td Bw N, is independent of M.
4) MATLAB command:
>> N=50; Fig. 3. 12 (b)
>> M=12;
>> x=[ones(1,M+1),zeros(1,N-2*M-1),ones(1,M)];
0.5
>> X=fft(x)/N;
>> k=[0:N-1]; %frequency index 0.4
>> stem(k,real(fftshift(X)))
0.3
fft find DTFS coefficients
stem display the results 0.2
centered on k = 0
-0.2
0 10 20 30 40 50
>> N=50;
>> M=4; Fig. 3. 12 (a)
>> x=[ones(1,M+1),zeros(1,N-2*M-1),ones(1,M)];
>> X=fft(x)/N;
>> k=[0:N-1]; %frequency index 0.2
>> stem(k,real(fftshift(X)))
0.15
0.1
0.05
-0.05
0 10 20 30 40 50
2 2
n x[ n ] k X [ k ]
Td ( N 1) / 2 Bw ( N 1) / 2
n ( N 1) / 2 k ( N 1) / 2
(3.75) (3.76)
2 2
x[ n ] X [ k ]
n ( N 1) / 2 n ( N 1) / 2
MATLAB function for >> function TBP=TdBw(x)
computing the product >> %Compute the Time-Bandwidth product using the DTFS
TdBw: >> %One period must be less than 1025 points
>> N=1025;
1) The length of input vector >> M=(N-max(size(x)))/2;
x = odd and centers on >> xc=[zeros(1,M),x,zeros(1,M)];
middle >> % center pulse within a period
2) .* = element-by-element >> n=[-(N-1)/2:(N-1)/2];
>> n2=n.*n;
product
>> Td=sqrt((xc.*xc)*n2'/(xc*xc'));
3) * = inner product >> X=fftshift(fft(xc)/N); % evaluate DTFS and center
4) ' = complex-conjugate >> Bw=sqrt(real((X.*conj(X))*n2'/(X*X')));
transpose >> TBP=Td*Bw;
Signals and Systems_Simon 198
Haykin & Barry Van Veen
CHAPTER
Fourier Representations of Signals & LTI Systems
4. Use the function TdBw to evaluate the time-bandwidth product for two
rectangular, raised cosine, and Gaussian pulse train as follows:
>> x=ones(1,101); % 101 point rectangular pulse
>> TdBw(x)
ans =
788.0303
>> x=ones(1,301); % 301 point rectangular pulse
>> TdBw(x)
ans =
1.3604e+003
>> x=0.5*ones(1,101)+cos(2*pi*[-50:50]/101); % 101 point rectangular pulse
>> TdBw(x)
ans =
277.7327
>> x=0.5*ones(1,301)+cos(2*pi*[-150:150]/301); % 301 point rectangular pulse
>> TdBw(x)
ans =
443.0992
>> n=[-500:500];
>> x=exp(-0.001*(n.*n)); % Narrow Gaussian pulse
>> TdBw(x)
ans =
81.5669
>> x=exp(-0.0001*(n.*n)); % Broad Gaussian pulse
>> TdBw(x)
ans =
81.5669
Note that the Gaussian pulse trains have the smallest time-bandwidth product.
Time-bandwidth product is identical for both the narrow and broad Gaussian pulse
trains .