Signals and Systems I-EEE 203: Lecture Notes 2b
Signals and Systems I-EEE 203: Lecture Notes 2b
Signals and Systems I-EEE 203: Lecture Notes 2b
Papandreou-Suppappola
Lecture Notes 2b
Importance of processing signals in the frequency domain
frequency domain
y ptq “ xptq ‹ hptq ÐÝÝÝÝÝÝÝÝÝÑ Y pjΩq “ X pjΩq H pjΩq
1
A. Fourier transform definition (Sections 4.1-4.7)
– Used Fourier series for periodic signals; use continuous-time Fourier transform (FT) for aperiodic signals
– Inverse FT (IFT)
ż8
1
x p tq “ X pjΩq ejΩt dΩ
2π ´8
2
B. Convergence of Fourier transform
ż8
a) xptq must be absolutely integrable over a period ñ |xptq| dt ă 8
´8
b) xptq must have a finite number of maxima and minima (extrema) over a finite interval
c) xptq must have a finite number of discontinuities over a finite interval and discontinuities must be finite
ż8 ż8 ż8 8
|xptq| dt “ |e2 t uptq| dt “ e2 t dt “ 0.5 e2 t “ 8
´8 ´8 0 0
ż8 ż8 ż8 8
|xptq| dt “ |e´2 t uptq| dt “ e´2 t dt “ ´0.5 e´2 t “ 0.5
0
´8 ´8 0
3
C. Examples of FTs
´ ¯
(a) Rectangular pulse: xptq = 1.5 upt ` 4q ´ upt ´ 4q (not periodic)
x p tq
2
t
´4 ´2 2 4
To compute FT
ż8 ż4 4
´jΩt ´jΩt ´1.5 ´jΩt ´1.5 ´ ´jΩ4 jΩ4
¯
X pjΩq “ x p tq e dt “ 1.5 e dt “ e “ e ´e
´4 jΩ jΩ
´8 ´4
ejΩ4 ´ e´jΩ4
ˆ ˙
3 3 sin p4 Ωq
“ “ sin p4 Ωq “ 12 “ 12 sincp4 Ωq, @Ω
Ω 2j Ω 4Ω
sin pΩq
Note: Continuous-time sinc function: sincpΩq =
Ω
4
FT pair (equation)
´ ¯
FT sin p4 Ωq
xptq “ 1.5 upt ` 4q ´ upt ´ 4q ÐÑ X pjΩq “ 12 “ 12 sincp4 Ωq, @Ω
4Ω
&'(%) *$
!"#$%&'()%*#+,- $!
"
%
$! ! ! $!
# # "
" " "
5
FT pair (graph)
&'(%) *$
x p tq
2
!"#$%&'()%*#+,- $!
"
1
%
t #
$!
#
! ! $!
" " " "
´4 ´2 2 4
FT
Ð
Ñ
Observations:
where bandwidth the range of frequencies for which X pjΩq exists (ending frequency - starting frequency)
6
FT pair (general)
´ ¯
FT sin pT Ωq
x p tq “ A u p t ` T q ´ u p t ´ T q Ð
Ñ X pjΩq “ 2 A “ 2 A T sincpT Ωq, @Ω
Ω
'()&* $"%
,(+*
$! ! ! $! &
#" " # #
+ " " " "
2π
Duration is time: 2 T Mainlobe width in frequency:
T
As T increases, duration in time increases but mainlobe bandwith decreases (inverse relationship)
7
(b) Exponential signal: xptq = e´at uptq (also see Ex. 4.1 in book)
FT 1
xptq “ e´at uptq, a ą 0 Ð
Ñ X pjΩq “
a ` jΩ
To show FT computation
ż8 ż8 8
´1 1
X pjΩq “ e´at uptqe´jΩt dt “ e ´tpjΩ`aq
dt “ e ´tpjΩ`aq
“ , aą0
0 a ` jΩ a ` jΩ
0
´8
In general, FT X pjΩq = RpjΩq ejΦpjΩq is complex ñ plot magnitude RpjΩq and phase ΦpjΩq
ˇ ˇ ˇ ˇ ˇ ˇ dˆ ˙2 ˆ ˙2
ˇ 1 ˇ ˇ a ´ jΩ ˇ ˇ a ´ jΩ ˇ a ´ Ω
RpjΩq “ ˇˇ ˇ“ˇ ˇ“ˇ ˇ“ `
a ` jΩ ˇ ˇ pa ` jΩqpa ´ jΩq ˇ ˇ a2 ` Ω2 ˇ a2 ` Ω 2 a2 ` Ω 2
d
a2 ` Ω 2 1
“ “ ?
p a2 ` Ω 2 q 2 a2 ` Ω 2
´Ω { a 2 ` Ω 2
ˆ ˙
ΦpjΩq “ arctan “ ´arctanpΩ{aq
a {a 2 ` Ω 2
8
Plot magnitude and phase
1
X pjΩq “ RpjΩq ejΦpjΩq, RpjΩq “ ? , ΦpjΩq “ ´arctanpΩ{aq
a2 ` Ω 2
9
(c) Impulse function: xptq = δ ptq
ż8 ż8
1
δ ptqe´jΩt dt e´jΩ0
✿
✘
✘
X pjΩq “ “✘✘✘ δ ptq dt “ 1
´8 ´8
FT
ñ δ p tq Ð
Ñ 1, @Ω
ż8 ż8
X pjΩq “ δ pt ´ t0 qe´jΩt dt “ e´jΩ t0 δ pt ´ t0 q dt “ e´jΩ t0
´8 ´8
FT
ñ δ p t ´ t0 q Ð
Ñ e´jΩ t0 , @Ω
FT
(d) Complex exponential: ejΩ0t Ð
Ñ 2π δ pΩ ´ Ω0 q (derivation when cover FT properties)
10
D. FT of periodic signals
Recall Fourier series (FS) of a periodic signal xptq with fundamental frequency Ω0
ÿ
x p tq “ αk ejkΩ0 t
k
FT
ÿ ÿ
x p tq “ αk ejkΩ0t Ð
Ñ X pjΩ0q “ 2π αk δ pΩ ´ Ω0 kq
k k
FT
using FT pair ejΩ0 t Ð
Ñ 2π δ pΩ ´ Ω0 q and also that FT is a linear transformation
11
Example: xptq = cos pΩ0tq with FS xptq = 0.5 ejΩ0 t ` 0.5 e´jΩ0t
FT
For Ω = π {4, FT pair is cos ppπ {4qtq Ð
Ñ π δ pΩ ´ π {4q ` π δ pΩ ` π {4q, and plot shown below
π π
X pjΩq
0
´π {2 ´π {4 0 π {4 π {2
Ω
12
E. Fourier transform properties (see Table 4.1)
FT FT
Considering TF pairs x1 ptq Ð
Ñ X1 pjΩq and x2 ptq Ð
Ñ X2 pjΩq
[P1] Linearity
FT
a x 1 p tq ` b x 2 p tq Ð
Ñ a X1 pjΩq ` b X2 pjΩq
ż
satisfied since X pjΩq = xptq e´jΩt dt and integration is linear
FT
Ñ e´jΩ t0 , then
Example: since δ pt ´ t0 q Ð
FT
Ñ 5 ` 3 ej2 Ω ´ e´j3 Ω
5 δ p tq ` 3 δ p t ` 2 q ´ 2 δ p t ´ 3 q Ð
13
[P2] Time-shift : shift in time results in multiplication by complex sinusoid in frequency
FT
Ñ Y pjΩq “ e´jΩ t0 X pjΩq
y p tq “ x p t ´ t0 q Ð
Letting τ = t ´ t0 , ñ t = τ ` t0 and dt = dτ
ż8 ż8
Y pjΩq “ xpτ q e´jΩpτ `t0q dτ “ e´jΩt0 xpτ q e´jΩτ dτ “ e´jΩt0 X pjΩq
´8 ´8
FT 1
Note that xptq = e´2t uptq Ð
Ñ X pjΩq =
2 ` jΩ
and from unit step upt ` 2q, y ptq can be expressed in terms of xpt ` 2q ñ must also have shift in e´2 t term
14
[P3] Modulation: multiplication by complex sinusoid in time results in shift in frequency
FT
y ptq “ ejΩ0 t xptq Ð
Ñ Y pjΩq “ X pj pΩ ´ Ω0 qq
FT
Ñ e´jΩ t0 X pjΩq (duality)
Note similarity in structure with time shift property: xpt ´ t0 q Ð
[P4] Scaling time axis: scale in time by a results in scale in frequency by 1{a
FT 1 ´ Ω¯
y p tq “ x p a tq Ð
Ñ Y pjΩq “ X j
| a| a
FT 1 1 1 1
y ptq “ e´6 t uptq “ e´3p2 tq uptq “ xp3 tq Ð
Ñ Y pjΩq “ X pjΩ{3q “ “
3 3 2 ` j p Ω {3 q 6 ` jΩ
15
[P5] Scaling frequency axis (dual)
FT
y p tq “ | a | x p t{ a q Ð
Ñ Y pjΩq “ X pjaΩq
FT
y ptq “ xp´tq Ð
Ñ Y pjΩq “ X p´Ωq
FT 1
y ptq “ e2 t up´tq “ xp´tq Ð
Ñ Y pjΩq “ X p´jΩq “
2 ´ jΩ
16
[P7] Duality
FT
Using duality, if known one FT pair, xptq Ð
Ñ X pjΩq, then can obtain a second FT pair
FT
x p tq Ð
Ñ z pΩq “ X pjΩq
FT
y p tq “ z p tq Ð
Ñ Y pjΩq “ 2πxp´Ωq
1 FT
g p tq “ z p´tq Ð
Ñ GpjΩq “ xpΩq
2π
where z ptq “ z pΩq and x p Ω q “ x p tq
Ω Ñ t t Ñ Ω
Example
FT FT sin pΩ T1 q
box Ð
Ñ sinc x p tq “ u p t ` T 1 q ´ u p t ´ T 1 q Ð
Ñ z pΩq “ X pjΩq “ 2
Ω
FT sin pt Ω1q FT ´ ¯
sinc Ð
Ñ box z p tq “ 2 Ð
Ñ 2πxp´Ωq “ 2π upΩ ` Ω1 q ´ upΩ ´ Ω1 q
t
17
1
Example: Find the FT of rptq =
3´jt
FT 1
Using duality and the FT pair xptq = e´3t uptq Ð
Ñ X pjΩq “
3`jΩ
FT
Ñ RpjΩq “ 2π xpΩq “ 2π e´3Ω upΩq
rptq “ z p´tq Ð
1 FT
Resulting FT pair: Ñ 2πe´3Ω upΩq
Ð
3´jt
FT
Ñ z pΩq “ X pjΩq “ e´jΩt0 , then
Using xptq = δ pt ´ t0 q Ð
FT
rptq “ z p´tq “ ejΩ0 t Ð
Ñ RpjΩq “ 2πxpΩq “ 2πδ pΩ ´ Ω0q
FT
Resulting FT pair: ejΩ0 t Ð
Ñ 2πδ pΩ ´ Ω0 q
18
[P8] Differentiation in time results in multiplication by frequency variable in frequency
d FT
y p tq “ x p tq Ð
Ñ Y pjΩq “ jΩ X pjΩq
dt
FT d
[P9] Dual property of P8: differentiation in frequency y ptq “ t xptq Ð
Ñ Y pjΩq “ j X pjΩq
dΩ
FT 1
Example: Find the FT of y ptq = t e´2t uptq using xptq = e´2t uptq Ð
Ñ X pjΩq = ,
2 ` jΩ
´2t FT d d ´ 1 ¯ 1 1 1
y p tq “ t e u p tq “ t x p tq Ð
Ñ Y pjΩq “ j X pjΩq “ j “ ´j 2 “ “
dΩ dΩ 2 ` jΩ p2 ` jΩq2 p2 ` jΩq2 4 ´ Ω2
FT 1
t e´at uptq Ð
Ñ , aą0
pa ` jΩq2
19
[P10] Integration in time
żt
FT 1
y p tq “ xpτ q dτ Ð
Ñ Y pjΩq “ X pjΩq ` πX p0qδ pΩq
´8 jΩ
FT 1
u p tq Ð
Ñ ` πδ pΩq
jΩ
20
[P12] Multiplication of two signals in time
FT 1
g p tq “ x p tq r p tq Ð
Ñ GpjΩq “ X pjΩq ‹ RpjΩq
2π
amplitude A, transmit signal xptq, carrier frequency Ωc (to modulate xptq to appropriate frequency band)
¸
FT A A´
´ ´ ¯ ¯
A xptq cos pΩc tq Ð
Ñ X pjΩq ‹ π δ pΩ ´ Ωc q ` δ pΩ ´ Ωc q “ X pj pΩ ´ Ωc qq ` X pj pΩ ` Ωc qq “
2π 2
21
[P13] Convolution in time (dual to P11)
FT
y ptq “ xptq ‹ hptq Ð
Ñ Y pjΩq “ X pjΩq RpjΩq
FT 2
Recall: FT of a box bptq = upt ` a{2q ´ upt ´ a{2q Ð
Ñ B pjΩq “ Ω
sin pΩ a{2q
Using property
FT 4
xptq “ bptq ‹ bptq “ |t| pupt ` aq ´ upt ´ aqq Ð
Ñ X pjΩq “ B pjΩq B pjΩq “ 2
sin2 pΩ a{2q
Ω
22
Example: Linear time-invariant (LTI) with impulse response hptq and frequency response H pjΩq (Sect. 4.5)
ż
FT
y ptq “ xptq ‹ hptq “ xpt ´ τ q hpτ q dτ Ð
Ñ Y pjΩq “ X pjΩq H pjΩq
!"#$ & #
%"#$
*"()$ +"()$
'"()$
23
Filtering: processing to remove frequencies that are not of interest
24
Example: Signal xptq = 2 sin p2π50 tq ` sin p2π250 tq duration 1 s and frequencies 50 Hz and 250 Hz
Design lowpass filter (LPF) with frequency response H pjΩq) and cuttoff 150 Hz
0.9
0.8
0.7
0.6
Magnitude
0.5
0.4
0.3
0.2
0.1
0
-500 -400 -300 -200 -100 0 100 200 300 400 500
Frequency (Hz)
25
FT
xptq “ 2 sin p2π50 tq ` sin p2π250 tq Ð
Ñ
´ ¯ ´ ¯
X pjΩq “ j π δ pΩ ` 2π50q ´ δ pΩ ´ 2π50q ` j π δ pΩ ` 2π250q ´ δ pΩ ´ 2π250q `
$"# !"
Sum of 50 and 250 Hz sinusoids in time in frequency
1000
900
800
700
600
magnitude
500
400
300
200
100
0
-500 -400 -300 -200 -100 0 100 200 300 400 500
!"
frequency, Hz
"# !"#
26
Time and frequency domain signal representations, before and after LPF
Sum of 50 and 250 Hz sinusoids in time After lowpass filtering with cutoff 150 Hz in time
3 1
2
0.5
1
amplitude
amplitude
0 0
-1
-0.5
-2
-3 -1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
time, s time, s
Sum of 50 and 250 Hz sinusoids in time in frequency After lowpass filtering with cutoff 150 Hz in frequeny
1000 500
800 400
magnitude
magnitude
600 300
400 200
200 100
0 0
-500 0 500 -500 0 500
frequency, Hz frequency, Hz
27
Systems characterized by linear constant coefficient differential equations (DE) (Sect. 4.7)
d d
LTI system whose input and output satisfy DE a0 y ptq ` a1 y ptq = b0 xptq ` b1 xptq (stable, initially at rest)
dt dt
Step 1: take FT of both sides of the DE (if differentiate in time then multiply with Ω in frequency)
d d FT
a 0 y p tq ` a 1 y ptq` “ b0 xptq ` b1 xptq Ð
Ñ a0 Y pΩq ` a1 j Ω Y pjΩq “ b0 X pΩq ` b1 j Ω X pjΩq
dt dt
FT Y pjΩq
Step 2: From y ptq = xptq ‹ hptq Ð
Ñ Y pjΩq = X pjΩq H pjΩq, solve for frequency response: H pjΩq =
X pjΩq
` ˘ ` ˘ Y pjΩq b0 ` jb1 Ω
from Step 1, Y pjΩq a0 ` ja1 Ωq = X pjΩq b0 ` jb1 Ωq ñ H pjΩq = =
X pjΩq a0 ` ja1Ω
b0 ` jb1 Ω IFT
H pjΩq “ ÐÑ hptq
a0 ` ja1 Ω
28
d d
Example: LTI stable system as DE 2 y ptq = 3 xptq ` xptq ´ y ptq; find impulse response hptq
dt dt
Using FT properties
d d FT
2 y p tq “ 3 x p tq ` x p tq ´ y p tq Ð
Ñ 2Y pjΩq “ 3X pjΩq ` jΩX pjΩq ´ jΩY pjΩq
dt dt
Y pjΩq 3 ` jΩ 3 1
ñ H pjΩq “ “ “ ` jΩ
X pjΩq 2 ` jΩ 2 ` jΩ 2 ` jΩ
´d ¯
´2t ´2t
ñ hptq “ 3 e u p tq ` e uptq “ 3 e´2t uptq ´ 2 e´2tuptq “ e´2t uptq
dt
29
Some additional FT properties
FT
x ˚ p tq Ð
Ñ X ˚ p´jΩq
FT
Also, recall xp´tq Ð
Ñ X p´jΩq
FT
real signal : xptq “ x˚ ptq Ð
Ñ X pjΩq “ X ˚ p´jΩq conjugate symmetric in frequency
FT
imaginary signal : xptq “ ´x˚ ptq Ð
Ñ X pjΩq “ ´X ˚ p´jΩq conjugate antisymmetric
FT
even signal : xptq “ xp´tq Ð
Ñ X pjΩq “ X p´jΩq even in frequency
FT
odd signal : xptq “ ´xp´tq Ð
Ñ X pjΩq “ ´X p´jΩq odd in frequency
FT
real and even signal : xptq “ xp´tq “ x˚ ptq Ð
Ñ X pjΩq “ X p´jΩq “ X ˚ p´jΩq real and even
ñ X pjΩq “ X ˚pjΩq
where conjugate symmetric implies that real part is even and imaginary part is odd
30