chapter 2 discrete fourier transform
TRANSCRIPT
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
1/75
03/04/16 1
Chapter 2Chapter 2
DiscreteDiscrete Fourier TransformFourier Transform
Instructor: TedInstructor: Ted
Email:[email protected]:[email protected]
Phone:13836034068Phone:13836034068
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
2/75
03/04/16 2
Three Questions aboutThree Questions about DiscreteDiscrete Fourier TransformFourier Transform
Q1:Q1: WWHAT is DFT?HAT is DFT?
Q2:Q2: WWHY is DFT?HY is DFT?
Q3: HOQ3: HO WW to DFT?to DFT?
W !T is relationshi" #et$een %&T and other 'inds o( &ourierW !T is relationshi" #et$een %&T and other 'inds o( &ourier
Trans(orm)Trans(orm)
W * $e need %&T)W * $e need %&T)
+W to realize %&T) o$ to use %&T to sol,e the "ractical+W to realize %&T) o$ to use %&T to sol,e the "ractical
"ro#lems)"ro#lems)
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
3/75
03/04/16 3
Basic contents of this chapterBasic contents of this chapter -.1 e,ie$ o( &ourier Trans(orm-.1 e,ie$ o( &ourier Trans(orm
-.- %iscrete &ourier /eries-.- %iscrete &ourier /eries -.3-.3 %iscrete%iscrete &ourier Trans(orm&ourier Trans(orm
-.4-.4 elationshi" #et$een %&T z Trans(orm and se2uence selationshi" #et$een %&T z Trans(orm and se2uence s
&ourier Trans(orm&ourier Trans(orm
-.-. &re2uency sam"lin5 theorem&re2uency sam"lin5 theorem
-.6-.6 om"ute se2uence s linear con,olution usin5om"ute se2uence s linear con,olution usin5 %&T%&T
-.7-.7 /"ectrum analysis #ased on %&T/"ectrum analysis #ased on %&T
-.8 e,ie$-.8 e,ie$
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
4/75
03/04/16 4
2 12 1 Fourier TransformFourier Transform
( ) ( )sin 2 x t f t π = × × ( )2 randn+ ×In some situation si5nal s (re2uency s"ectrum can re"resent its characteristics
more clearly.
( ) x t in (re2uency domain( ) x t in time domain
&ourier
Trans(orm
( ) ( )sin 2 500 x t t π = × ×
/i5nal !nalysis and Processin5/i5nal !nalysis and Processin5
11 Time %omain !nalysis: t !Time %omain !nalysis: t !
-- &re2uency %omain !nalysis: ( !&re2uency %omain !nalysis: ( !
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
5/75
03/04/16 5
2 12 1 Fourier TransformFourier Transform
/i5nal !nalysis and Processin5:/i5nal !nalysis and Processin5:
11 Time %omain !nalysisTime %omain !nalysis -- &re2uency %omain !nalysis&re2uency %omain !nalysis
&ourier Trans(orm is a #rid5e (rom time domain to (re2uency domain.&ourier Trans(orm is a #rid5e (rom time domain to (re2uency domain.
haracteristic: continuous discrete "eriodic haracteristic: continuous discrete "eriodic non"eriodicnon"eriodic ..
ontinuous "eriodic si5nalsontinuous "eriodic si5nals
ontinuous non"eriodic si5nalsontinuous non"eriodic si5nals
%iscrete "eriodic si5nals%iscrete "eriodic si5nals
%iscrete non"eriodic%iscrete non"eriodic si5nalssi5nals
Ty"e:Ty"e:
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
6/75
03/04/16 6
1! "ontinuous perio#ic si$na%&&Fourier 'eries1! "ontinuous perio#ic si$na%&&Fourier 'eriesIt is "ro,ed that continuous time "eriodic si5nal can #e re"resented #y aIt is "ro,ed that continuous time "eriodic si5nal can #e re"resented #y a
&ourier /eries corres"ondin5 to a sum o( harmonically related com"le9&ourier /eries corres"ondin5 to a sum o( harmonically related com"le9
e9"onential si5nal. To a "eriodic (unction $ith "eriod e9"onential si5nal. To a "eriodic (unction $ith "eriod
( ) ( )0 0202
2 1,
T jn t jn t
T n nn
f t F e F f t e dt T T
ω ω π
ω
+∞−
−=−∞= = ⇔ =∑ ∫
( ) f t T
Conclusion:Conclusion: Continuous periodicContinuous periodic function— function—
Nonperiodic discrete Nonperiodic discrete frequency impulse sequence frequency impulse sequence
Time-domain Frequency-domain
( ) f t A
t
T 2τ 2τ −0
T n
n F
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
7/75
03/04/16 7
2!2! "ontinuous"ontinuous nonperio#icnonperio#ic function(s Fourier Transformfunction(s Fourier Transform
∫ ∞
∞
Ω= dt et x j X t j ;;
∫ ∞
∞Ω Ωd e j X t x t j :;
-1
:;
Conclusion :Conclusion : Continuous nonperiodicContinuous nonperiodic function— function— Nonperiodic continuous Nonperiodic continuous function function
Time-domain Frequency-domain
t
( ) x t A
2τ 2τ −0 Ω
( ) X jΩ
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
8/75
03/04/16 8
3!3! Discrete&timeDiscrete&time nonperio#icnonperio#ic se)uence(s Fourier Transformse)uence(s Fourier Transform
Conclusion :Conclusion : Discrete nonperiodic Discrete nonperiodic function— function—Continuous-time periodicContinuous-time periodic function function
; :
1; :
-
j j n
n
j j n
X e x n e
x n X e e d
∞
= ∞
=
=∑∫
; - ; -
-
; j M j M n
n
j n j nM j n
n n
X e x n e
x n e e x n e
∞
= ∞ ∞ ∞
= ∞ = ∞
=
= =
∑
∑ ∑
Time-domain Frequency-domain
ω
( ) j X e ω
π − π 02 π − 2 π
n
( ) x n A
2 N 2 N −0
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
9/75
03/04/16 9
*! "onc%usion*! "onc%usion 11 /am"lin5 in time domain #rin5s "eriodicity in (re2uency/am"lin5 in time domain #rin5s "eriodicity in (re2uency
domain.domain.-- /am"lin5 in (re2uency/am"lin5 in (re2uency domain #rin5s "eriodicity in timedomain #rin5s "eriodicity in time
domain.domain. 33 elationshi" #et$een (re2uency domain and time domainelationshi" #et$een (re2uency domain and time domain
Time domain &re2uency domain Trans(ormTime domain &re2uency domain Trans(orm
ontinuous "eriodicontinuous "eriodic %iscrete non"eriodic &ourier series%iscrete non"eriodic &ourier series
ontinuous non"eriodic ontinuous non"eriodicontinuous non"eriodic ontinuous non"eriodic &ourier Trans(orm&ourier Trans(orm
%iscrete non"eriodic ontinuous "eriodic%iscrete non"eriodic ontinuous "eriodic /e2uence s &ourier/e2uence s &ourier
Trans(ormTrans(orm
%iscrete "eriodic %iscrete "eriodic %iscrete &ourier /eries%iscrete "eriodic %iscrete "eriodic %iscrete &ourier /eries
PeriodicPeriodic %iscrete< =on"eriodic%iscrete< =on"eriodic ontinuousontinuous
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
10/75
03/04/16 10
+! Basic i#ea of Discrete Fourier Transform+! Basic i#ea of Discrete Fourier Transformn !rac"ica# a!!#ica"ion, si$na# !rocessed %y com!u"er &as "'o main c&arac"eris"icsn !rac"ica# a!!#ica"ion, si$na# !rocessed %y com!u"er &as "'o main c&arac"eris"ics
;1 %iscrete;1 %iscrete ;- &inite len5th;- &inite len5thimi#ar#y, si$na#*s +requency mus" a#so &a e "'o main c&arac"eris"icsimi#ar#y, si$na#*s +requency mus" a#so &a e "'o main c&arac"eris"ics
;1 %iscrete;1 %iscrete ;- &inite len5th;- &inite len5th
IdeaIdea ::
.!and.!and +ini"e-#en$"& sequence "o !eriodic sequence+ini"e-#en$"& sequence "o !eriodic sequence , com!u"e i"s iscre"e, com!u"e i"s iscre"e
Fourier eries, so "&a" 'e can $e" "&e discre"e s!ec"rum in +requency domainFourier eries, so "&a" 'e can $e" "&e discre"e s!ec"rum in +requency domain
u" non!eriodic sequence*s Fourier Trans+orm is a con"inuous +unc"ion o+u" non!eriodic sequence*s Fourier Trans+orm is a con"inuous +unc"ion o+ ωω, and i", and i"is a !eriodic +unc"ion inis a !eriodic +unc"ion in ωω 'i"& a !eriod 2'i"& a !eriod 2 ππ o i" is no" sui"a%#e "o so# e !rac"ica# o i" is no" sui"a%#e "o so# e !rac"ica#di$i"a# si$na# !rocessin$di$i"a# si$na# !rocessin$
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
11/75
03/04/16 11
2 2 Discrete Fourier 'eries1 %iscrete &ourier /eries Trans(orm Pair
/imilar $ith continuous time "eriodic si5nals a "eriodic se2uence $ith"eriod = can #e re"resented #y a &ourier /eries corres"ondin5 to a sum o(
harmonically related com"le9 e9"onential se2uences such as:
( ) ( )1
0
1k
N j n
k
x n X k e
N
ω −
=
= ∑ %%
Attention: &ourier /eries (or discrete time si5nal $ith "eriod = re2uiresonly = harmonically related com"le9 e9"onentials.
( )
( )
12 , s +undamen"a# an$u#ar +requency
2, s &armonic an$u#ar +requencyk
x n N
k x n kth N
π ω
π ω
=
=
%
%
2-1
( ) x n%
$here
2 2( )
"&a" is j kn j k n N
N N k N k e e
π π
ω ω +
+= =:Q
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
12/75
03/04/16 12
( )( )
( )
221
20
11 1 1
01
j r k N N N j r k n
N
j r k n N
r k mN ee
otherwise N N e
π
π
π
−− −
−=
= +−= × =−∑
2-
+"er mu#"i!#yin$ %o"& sides o+ q (2-1) 'i"& , and summin$
+rom 0 -1, 'e o%"ain
j kn N e
n to N
π
=
( ) ( )
( ) ( ) ( ) ( )
2 2 21 1 1
0 0 0
2 21 1 1 1
0 0 0 0
1
1 1
N N N j kn j rn j kn N N N
n n r
N N N N j r k n j r k n N N
n r r n
x n e X r e e N
X r e X r e N N
π π π
π π
− − −− −
= = =
− − − −− −
= = = =
=
= = ÷
∑ ∑∑
∑∑ ∑ ∑
%%
% %
com!u"a"ion( ) X k %
( ) ( ) ( )21
0
N j kn N
n
X r X k x n eπ − −
=⇒ = =∑% % %
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
13/75
03/04/16 13
Attention:
:;>
:;> :;> :;> -1
0
;-1
0k X en x en x mN k X
kn N
j N
n
nmN k N
j N
n=
=
=
so ( ) is a#so a !eriodic sequence 'i"& !eriod
( ) and ( ) is a !eriod sequence !air in +requency domain and "ime domain
X k
X k x n
%
% %
Discrete Fourier Series for periodic sequence:
( ) ( ) ( ) ( )
( ) ( ) ( ) ( )
21 1
0 0
21 1
0 0
1 1
N N j kn kn N N
n n
N N j kn kn N N
k n
X k DFS x n x n e x n W
x n IDFS X k X k e X k W N N
π
π
− −−
= =− −
−
= =
= = =
= = =∑ ∑
∑ ∑
% % % %
% % %%
N j
N eW
-
=
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
14/75
03/04/16 14
- Pro"erties o( %&/u!!ose ( ) ( ) ( ) ( ) DFS x n X k DFS y n Y k = = % %% %
1 ,inear,inear :;> :;> :?;> :;> @ k Y bk X n ybn x D!"
2 'e)uence 'hift'e)uence 'hift( ) ( )
( ) ( )
km N
nl N
DFS x n m W X k
IDFS X k l W x n
−+ =+ =
%%
% %
( )
1
0
1 1
1
0
( ) ( ) ,
( ) ( )
( ) ( ) ( )
N kn
N n
N m N mk r m km kr
N N N r m r m
N km kr km
N N N r
Proof DFS x n m x n m W let r n m
Then x r W W x r W
W x r W DFS x n m W X k
−
=+ − + −
− −
= =−
− −
=
+ = + = +
= =
= ⇒ + =
∑
∑ ∑
∑
% %
% %
%% %
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
15/75
03/04/16 15
- Pro"erties o( %&/u!!ose ( ) ( ) ( ) ( ) DFS x n X k DFS y n Y k = =% %% %
3 -erio#ic "on.o%ution-erio#ic "on.o%ution
1 1
0 0
+ ( ) ( ) ( ) "&en
( ) ( ) ( ) ( ) ( ) ( ) N N
m m
F k X k Y k
f n IDFS F k x m y n m y m x n m− −
= =
=
= = − = −∑ ∑ % % %
% % % % % %
om!ared 'i"& #inear con o#u"ion, !eriodic con o#u"ion*s maindi++erence is
T&e sum is o er "&e +ini"e in"er a# m 0 -1
:eriodic con o#u"ion
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
16/75
03/04/16 16
1
0
(1) ( ) (1 ) 0 N
m
f x m y m−
== − =∑% % %
1
0
(0) ( ) (0 ) 1 N
m
f x m y m−
== − =∑% % %
1
0
(2) ( ) (2 ) 1 N
m
f x m y m−
=
= − =∑% % %
1
0
( ) ( ) ( ) N
m
f n x m y n m−
=
= −∑% % %-erio#ic con.o%ution-erio#ic con.o%ution
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
17/75
03/04/16 17
'/mmetr/'/mmetr/ :
1 1
0 0
+ ( ) ( ) ( ) "&en1 1
( ) ( ) ( ) ( ) ( ) ( ) N N
l l
f n x n y n
F k DFS f n X l Y k l Y l X k l N N
− −
= =
== = − = −∑ ∑
%
% %
%% % % % %
0u%tip%ication of perio#ic se)uence in timeomain is0u%tip%ication of perio#ic se)uence in timeomain iscorrespon# to con.o%ution of perio#ic se)uence incorrespon# to con.o%ution of perio#ic se)uence infre)uenc/ #omainfre)uenc/ #omain
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
18/75
03/04/16 18
Periodic se2uence and its %&/
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
19/75
03/04/16 19
2 3 Discrete Fourier Transform&DFT 2 3 Discrete Fourier Transform&DFT
Periodic se2uence and its %&/
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
20/75
03/04/16 20
HINTS
-erio#ic se)uence is infinite %en$th
but on%/ se)uence .a%ues contain information
-erio#ic se)uence finite %en$th se)uence
e%ationship bet een these se)uences ?
4nfinite4nfinite FiniteFinite
-erio#ic-erio#ic NonperiodicNonperiodic
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
21/75
03/04/16 21
2 3 Discrete Fourier Transform&DFT 2 3 Discrete Fourier Transform&DFT e%ationship bet een perio#ic se)uence an# finite&%en$th se)uence
-erio#ic se)uence can be seen as perio#ica%%/ copies of finite&%en$th se)uenceFinite&%en$th se)uence can be seen as e5tractin$ one perio# fromperio#ic se)uence
Main period
Finiteuration 'e)uenceFiniteuration 'e)uence -erio#ic 'e)uence-erio#ic 'e)uence
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
22/75
03/04/16 22
2 3 Discrete Fourier Transform&DFT 2 3 Discrete Fourier Transform&DFT u!!ose ( ) is a +ini"e-#en$"& sequence; ( ) is a !eriodic sequence; x n x n%
( ) ( ) ( ) N x n x n R n= %1, 0 1
( ) is a square-'a e sequence ( )0, o"&er'ise N N
n N R n R n
≤ ≤ −=
'e use (( )) "o deno"e (n modu#o ) N n
( )( ) ( )12
5 5 x x=
(0) x(1) x
(2) x
(3) x
(4) x
(6) x(7) x
(8) x
(11) x(10) x
(9) x
(5) x
12 N =
( )( ) ( )12
20 8 x x=
( )( ) ( )12
1 11 x x− =
( )( )( ) ( )r
N r x n x n rN x n
=∞
=−∞= + =∑%
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
23/75
03/04/16 23
( ) x n
2 3 Discrete Fourier Transform2 3 Discrete Fourier Transform
Get DFT by extracting one period of DFS
DFS of periodic sequence
( ) N R k
om"utation o( %&T #y e9tractin5 one "eriod o( %&/
To a (inite len5th se2uence :
Periodicalcopies
( )( ) N
x n( ) X k ⇔ DFS
Attention DFT is ac)uire# b/ e5tractin$ one perio# of
DF'6 it is not a ne 7in# of Fourier Transform
DFT
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
24/75
03/04/16 24
1
0
1
0
1
0
+ini"e-#en$"& sequence can %e e.!ressed in "&is 'ay
( ) ( ) ( )
( ) ( ) ( ) (( )) ( )
"&en ( ) ( ) 0 1
1"&en ( ) ( ) 0 1
N
N kn
N N N N n
N kn
N n
N kn
N k
x n x n R n
X k X k R k x n W R k
X k x n W k N
x n X k W n N N
−
=−
=−
−
=
=
∴ = =
= ≤ ≤ −
= ≤ ≤ −
∑∑
∑
%
%
DFT TransforPair
=
=
10;1
;
10;;
1
0
1
0
N nW k X N n x
N k W n x k X
N
k kn N
N
n
kn N
( ) ( ) ( ) ( ) X k DFT x n x n IDFT X k = =abbr .
4n.erse Transform4n.erse Transform
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
25/75
03/04/16 25
Pro"erty o( %&Tu!!ose ( ) ( ) ( ) ( ) X k DFT x n Y k DFT y n= =
;1 Ainearity( ) ( ) ( ) ( ) , are coe++icien" DFT ax n by n aX k bY k a b+ = +
;- ircular /hi(t
ircular shi(t o( x#n$ can #e de(ined:( ) (( )) ( )m N N x n x n m R n= +
( ) ( ) ( ) x n x n x n m+% %
( )m x n
Periodi
!o"ies
Shift #xtra t
$ain "eriod (( )) N x n m= +
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
26/75
03/04/16 26
2x (n)
"ircu%ar shift of se)uence"ircu%ar shift of se)uence ,inear shift of se)uence,inear shift of se)uence
N - %
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
27/75
03/04/16 27
( ) ( ) (( )) ( ) ( )mk m m N N N X k DFT x n DFT x n m R n W X k −= = + =
(( )) ( ) ( )nl N N N IDFT X k l R k W x n+ =
'/mmetric bet een DFT an# 4DFT '/mmetric bet een DFT an# 4DFT
(( )) ( ) ( ) ( ) N N N DFT x n m R n DFT x n m R n+ = +%( ) ( ) N DFS x n m R k = +%
( ) ( )
( )
mk
N N mk
N
W X k R k
W X k
−
−
==
%
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
28/75
03/04/16 28
3 -arse.a%(s Theorem-arse.a%(s Theorem1 1
0 01 1
2 2
0 0
1( ) ( ) ( ) ( )
1
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
29/75
03/04/16 29
4 ircular con,olution
1
0
1
0
u!!ose ( ) ( ) ( )
"&en ( ) ( ) ( ) (( )) ( )
or ( ) ( ) (( )) ( )
N
N N m
N
N N m
F k X k Y k
f n IDFT F k x m y n m R n
f n y m x n m R n
−
=−
=
=
= = −
= −
∑∑
Periodic con,olution is con,olution o( t$o se2uences $ith "eriod= in one "eriod so it is also a "eriodic se2uence $ith "eriod =.
ircular con,olution is ac2uired #y e9tractin5 one "eriod o(
"eriodic con,olution e9"ressed #y ⊗.
ircularcon,olution
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
30/75
03/04/16 30
f#n$
1 N y m & n%
- N y m & n%
0 N y m & n%
"ircu%ar con.o%ution"ircu%ar con.o%ution -erio#ic con.o%ution-erio#ic con.o%ution
0 f
1 f
- f
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
31/75
03/04/16 31
( ) ( )
( ) ( )( ) ( ) ( ) ( )
DFT x n X k
DFT y n Y k IDFT X k Y k x n y n
==
= ⊗
Circular convolution can be used to
compute two sequence !s linear
convolution.
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
32/75
03/04/16 32
onBu5ate symmetric "ro"erties
a %&T o( conBu5ate se2uence
2
1 1( )
0 0
2
u!!ose ( ) is com!#e. con>u$a"e sequence o+ ( ) "&en
( ) ( ),0 1
:roo+ ( ) ( ) ( ) ,0 1
1
(
N
N N nk n k
N N n n
j nN nN j n N
x n x n
DFT x n X N k k N
DFT x n x n W x n W k N
W e e
DFT x n
π π
∗
∗ ∗
− −∗ ∗ − ∗
= =−
−
∗
= − ≤ ≤ −
= = ≤ ≤ −
= = =
∑ ∑
Q1 1
( ) ( )
0 0
) ( ) ( )
( ),0 1
N N nN n k n N k
N N N n n
x n W W x n W
X N k k N
− −− ∗ − ∗
= =∗
= =
= − ≤ ≤ −
∑ ∑
Attention C;' has only ' ,alid ,alues 0≤' ≤= 1 %u" '&en 0 (( 0)) (0) no" ( ),
so ( ) (( )) ( ) in a s"ric" 'ay N
N N
k X N X X N
DFT x n X N k R k
∗ ∗ ∗
∗ ∗
= − == −
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
33/75
03/04/16 33
# %&T o( se2uence s real and ima5inary "art
u!!ose ( ) s rea# and ima$inary !ar" are ( ) and ( )
"&en ( ) ( ) ( )1 1
( ) ( ) ( ) , ( ) ( ) ( )2 2
( ) and ( ) s /FT are ( ) and ( ), "&en
1( ) ( ) ( )2
r i
r i
r i
r i e o
e r
x n x n jx n
x n x n jx n
x n x n x n jx n x n x n
x n jx n X k X k
X k DFT x n DFT x n
∗ ∗= +
= + = −
= =
'
1( ) ( ) ( )2
1 1( ) ( ) ( ) ( ) ( ) ( )
2 2o i
x n X k X N k
X k DFT jx n DFT x n x n X k X N k
∗ ∗
∗ ∗
+ = + −
= = − = − −
( ) ( ) ( )e o X k X k X k = +
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
34/75
03/04/16 34
na#ysis o+ ( ) and ( ) s symme"ric
1( ) ( ) ( )
21 1
( ) ( ) ( ) ( ) ( )2 2
"&en ( ) ( )
e o
e
e
e e
X k X k
X k X k X N k
X N k X N k X N N k X N k X k
X k X N k
∗
∗ ∗ ∗ ∗
∗
= + −
− = − + − + = − +
= −
Q
CC ee;' is e,en com"onents o( C;' C;' is e,en com"onents o( C;' C ee;' is conBu5ate symmetric
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
35/75
03/04/16 35
C e;' conBu5ate e,en "art
conBu5ate symmetric<
real "art is e2ual ima5inary
"art is o""osite.
C e;' s real "art
C e;' s ima5inary "art
C o;' con"ugate odd part con"ugate asy etric#real part is opposite$ i aginarypart is equal%
C o;' s real "art
C o;' s ima5inary "art
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
36/75
03/04/16 36
"onc%usion1 %&T o( se2uence s real "art is corres"ondin5 to C;' s conBu5ate symmetric "art.
- %&T o( se2uence s ima5inary "art is corres"ondin5 to C;' s conBu5ate asymmetric
"art.
3 /u""ose 9;n is a real se2uence that is 9;n D9 r ;n
then C;' only has conBu5ate symmetric "art that is C;' DC e;'
sa"is+y ( ) ( )
(0) (0) ( ) ( )2 2
X k X N k N N
X X X X
∗
∗ ∗
= −= =:
/o: I( $e 5et hal( X#k$' $e can ac2uire all X#k$ usin5 symmetric"ro"erties.
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
37/75
03/04/16 37
01
0 1 1
1
( ( (
N N N ( ) N-)
N N N
( N-) #N-)$#N-)$ N N N
W W W X
W W W X
x x x N
W W W
X N
÷÷ ÷÷ ÷=÷ ÷÷ ÷÷÷÷
K K M O M
M K
1
0( ) ( ) 0 1
N kn
N n X k x n W k N
−
== ≤ ≤ −
∑ k N →0 1
n N →0 1
DFT -ro$rammin$ 85amp%e
DFT 0atri5
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
38/75
03/04/16 38
function &'()*dft+xn,
N*lengt +xn,# .lengt of sequencen*/:N-1# . ti e sa ple(*/:N-1#
0N*exp+-" 2 pi N,#n(*n3 (#0Nn(*0N%4n(# .calculate t e DFT 5atrix
'(*xn 0Nn(# .co pute DFT
5ore effecti6e et od%
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
39/75
03/04/16 39
Fs * 7//# . Get t e analy8ed signal
T * 1 Fs#9 * 1///#t * +/:9-1, T#x * /% sin+2 pi ;/ t,#
plot+1/// t+1:2//,$x+1:2//,,#
< * dft+x, 9# . Discrete Fourier Transfor
f * Fs 2 linspace+/$1$9 2=1,#ste +f$2 abs+
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
40/75
03/04/16 40
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
41/75
03/04/16 41
2 * DFT6 'e)uence(s Fourier Transform an# 9&transform
; : x t X j Ω
; : x n x n* j+ X e
%&/
/am"lin5
N
x n
Periodic o"ies
X k %
x n
E9tract +ne "eriod E9tract +ne "eriod
X k %&T
/e2uence s &ourierTrans(orm
&ourier Trans(ormontinuoustime
%iscretetime
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
42/75
03/04/16 42
Three #ifferent fre)uenc/omain representations of a finite&%en$th #iscrete&time se)uence
2 'e)uence(s Fourier Transform2 'e)uence(s Fourier Transform
3 Discrete Fourier Transform3 Discrete Fourier Transform DFT!DFT!
1 9&Transform1 9&Transform
-1
0
-1
0
; : ; : 0 1
1; : ; : 0 1
N j kn kn N
N n
N j kn N
k
X k x n e W k N
x n X k e n N N
=
=
= ≤ ≤
= ≤ ≤
∑
∑
11
0
1; ; ; ;
-
N n n
C n
X , x n , x n X , , d, =
= = ∫Ñ
1
0
1; ; ; ;
-
N j+ j+n j+ j+n
n
X e x n e x n X e e d+
=
= = ∫
j+ , e}
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
43/75
03/04/16 43
j I m5?6
Re5?6
C;' and C;z )C;' and C;z )
C;' and C;eC;' and C;e j+ j+ ) )
0 N W
-
N W
1 N W
$ N #
N W -
$ N # N W
1
k N
j k
N
W ,
eW ,
, X k X k N
-
E:;:;
=
=
=
:;:;-
k N
j e X k X
=
101
0
∑= N k W $n# x $k # X N
n
kn N
1
0
( ) ( ) N
n
n
X % x n % −
−
==∑
1
0
( ) ( ) N
jw jwn
n
X e x n e−
−
=
=∑
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
44/75
03/04/16 44
; : ; : j X k X e andelationshi" #et$een
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
45/75
03/04/16 45
( ) ( ) or ( ) j X k X e X % ω ⇒ 2 + Fre)uenc/ samp%in$ theorem2 + Fre)uenc/ samp%in$ theorem
o$ to realize) Prere2uisite (or im"lementation)o$ to realize) Prere2uisite (or im"lementation)What is inter"olation (ormula)What is inter"olation (ormula)
11 /am"lin5/am"lin5 9;n s z trans(orm:9;n s z trans(orm:
∑ ∞ =
n
n
, n x , X :;:;
2
( ) ( ) = ( )k N
kn N N % W
n
j k k N N
X k X % x n W
% W eπ
−
∞
==−∞
−
= =
= =
∑e5ular inter,al sam"lin5 on unit circle:e5ular inter,al sam"lin5 on unit circle:
,oss after samp%in$?,oss after samp%in$?
!(t "li 5 i ( 2 d i $ 2 i 2!(ter sam"lin5 in (re2uency domain can $e ac2uire se2uence
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
46/75
03/04/16 46
!(ter sam"lin5 in (re2uency domain can $e ac2uire se2uence!(ter sam"lin5 in (re2uency domain can $e ac2uire se2uencere"resentin5re"resentin5 x x ;; nn #y in,erse trans(ormin5 (rom C #y in,erse trans(ormin5 (rom C == ;; k k ))
is "eriodical co"ies o(is "eriodical co"ies o( x x ;; nn that is sam"lin5 in (re2uency that is sam"lin5 in (re2uency
domain causes "eriodical co"ies o( se2uence in time domain.domain causes "eriodical co"ies o( se2uence in time domain.:;> n x N
I( $e $ant to reco,er the (inite len5th se2uenceI( $e $ant to reco,er the (inite len5th se2uence x x ;; nn $ith no $ith no
loss a(ter sam"lin5 in (re2uency domain then it must #e satis(ied:loss a(ter sam"lin5 in (re2uency domain then it must #e satis(ied:
/u""ose:/u""ose: FF is num#er o( "oints in time domain
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
47/75
03/04/16 47
-- Inter"olation (ormulaInter"olation (ormula
( ) N X k ( ), 1 x n n $ =
( ) N X k 1
0
( ) ( ) $
n
n
X % x n % −
−
=
= ∑
( )1
0
( ) $
j jn
n
X e x n eω ω −
==∑
1
0
2( ) ( ) ,
N j
N k
X e X k k N
ω π
ω
−
=
= Φ − ÷ ∑ ω
ω
ω ω
−−
=Φ 21
)2/sin()2/sin(1
)( N
j
e N
N
1
0
( ) ( ) ( ), N
N k k
X % X k % −
== Φ∑ 11
11)( −−
−
−−=Φ
% W %
N % k
N
N
k
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
48/75
03/04/16 48
Ob;ecti.e DFT or 4DFT can be use# to compute t o se)uence(s circu%arcon.o%ution6 an# DFT6 4DFT ha.e their fast a%$orithm 'o if ecan bui%# the re%ationship bet een t o se)uences( circu%ar
con.o%ution an# %inear con.o%ution6 e can impro.e computationspee# of %inear con.o%ution b/ fast Fourier Transform a%$orithm
2 < "omputin$ se)uence(s %inear con.o%ution ith DFT
( ) ( )i $ & @ d i $ &h
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
49/75
03/04/16 49
Circular Convolution
Linear Convolution
( ) ( )u!!ose is a #en$"& @ sequence and is a #en$"& sequence,"&eir #inear con o#u"ion is
x n h n
( ) ( )1
0
( ) ( ) ( ) $
l m
y n x n h n x m h n m−
=
= ∗ = −∑
What relationship between and ?( )l y n ( ) y n
[ ]Aero !addin$ ( ) and ( ) "o "&e same #en$"& , ma. , , "&en x n h n & & N $ ≥( ), 0 1 ( ),0 1
( ) ( )0, 1 0, 1
x n n $ h n n N
x n and h n $ n & N n &
≤ ≤ − ≤ ≤ − = = ≤ ≤ − ≤ ≤ −
( ) ( ) ( ) ( ) ( ) ( )( ) DFT x n X k DFT h n ' k Y k X k ' k = = =
( )0
( ) ( ) ( ) ( ) (( )) ( ) &
N &m
y n IDFT Y k x n h n x m h n m R n=
′ ′ ′ ′= = = −
∑
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
50/75
03/04/16 50
" can %e !ro ed "&a" "&a" "&e re#a"ions&i! %e"'een ( ) and ( ) isl y n y n
( )( ) ( ) l &(
y n y n (& R n∞
=−∞
= +∑'&ic& means !oin"s circu#ar con o#u"ion ( ) is !eriodic co!ies
o+ ( ) %y e."rac"in$ "&e main !eriodl
& y n
y n
nd ( ) s #en$"& is 1l y n N $ + −
so '&en -1, "'o sequence s !oin"s circu#ar con o#u"ion
( ) is e.ac"#y "&eir #inear con o#u"ion ( ) l
& N $ &
y n y n
∴ ≥ +
( ) ( ) ( ) y n x n h n′ ′= ( ) ( )( )l y n x n h n= ∗=
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
51/75
03/04/16 51
( ), 3h n N =
( ), 5 x n $ =
( )l y n
5 & =
6 & =
7 & =
8 & =
rocess
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
52/75
03/04/16 52
-rocess( ), 0 1
#e" ( )0, 1
( ), 0 1( ) 0, 1
x n n $ x n
$ n &
h n n N
h n N n &
≤ ≤ −= ≤ ≤ −
≤ ≤ −= ≤ ≤ −
"onc%usion:We can compute %inear con.o%ution usin$ circu%ar con.o%ution if%en$th of DFTs satisf/ -1 & N $ ≥ +
x #n$
. #n$
Gero "addin5
Gero "addin5 ⊗
C;'
;'
C;' ;' x
#n$ .
#n$ x#n$ .#n$
FT
FT
FT∗
Pro#lems:
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
53/75
03/04/16 53
After FFT a%$orithm6 o.er%ap&a## metho# an# o.er&%ap sa.e
metho# i%% be %earne#
Pro#lems:4n practica% app%ication: / n!=5 n!>h n!6suppose 5 n!(s %en$th is 0 h n!( %en$th is
sua%%/6 0@@ 6 4f ,= 0&16 then:For short se)uence: man/ 9eros pa##e# into h n!
For %on$ se)uence: compute after a%% se)uence inputDifficu%ties ,ar$e memor/6 %on$ computation time6
so rea%&time propert/ can not be satisfie#'o%ution: #ecomposition computation on %on$ se)uence
Di.i#e# an# "on)uer Di.i#e# an# "on)uer
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
54/75
03/04/16 54
2 'pectrum ana%/sis usin$ DFT 2 'pectrum ana%/sis usin$ DFT
1! Appro5imation process1! Appro5imation process
; : x t
X j Ω
'amp%e ; : x n x n* ⇒
1
0
N nk
N n
X k x n W
=
= ∑
1! -rocess of spectrum ana%/sis usin$ DFT 1! -rocess of spectrum ana%/sis usin$ DFT
%&T
2! 8rror ana%/sis2! 8rror ana%/sis3! 4mportant parameters3! 4mportant parameters
/"ectrum analysis/"ectrum analysis %&T om"utation%&T om"utation%iscretization in time and%iscretization in time and
(re2uency domain(re2uency domain
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
55/75
03/04/16 55
Basic theor/ of Fourier TransformB &inite duration si5nal H In(inite $idth (re2uency s"ectrum
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
56/75
03/04/16 56
-rocess of spectrum ana%/sis usin$ DFT -rocess of spectrum ana%/sis usin$ DFT
22 8rrors of spectrum ana%/sis usin$ DFT8rrors of spectrum ana%/sis usin$ DFT
3! Fence effect3! Fence effect
; : x t
X j Ω
'amp%in$ x n
j X e
⇒ ⇒
- j
X k
X e k N
=
′ =
⇒
x n x n + n=
K j j j X e X e W e ′ ="on.o%ution
1
3
2
1!1! A%iasin$A%iasin$
2! "utoff effect2! "utoff effect
Win#o in$
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
57/75
03/04/16 57
1
; : x t
X j Ω
'amp%in$ ; : x n x n* ⇒
1
0
N nk
N n
X k x n W
=
=
∑
%&T
22 8rrors of spectrum ana%/sis usin$ DFT8rrors of spectrum ana%/sis usin$ DFT
-rocess of spectrum ana%/sis usin$ DFT -rocess of spectrum ana%/sis usin$ DFT
1! A%iasin$1! A%iasin$ I( condition is not met: there $ill #e s"ectrum distortion at (sJ-
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
58/75
03/04/16 58
22 "utoff effect of DFT
22 8rrors of spectrum ana%/sis usin$ DFT8rrors of spectrum ana%/sis usin$ DFT
; : x t
X j Ω
x n
j
X e
⇒ ⇒
- j
X k
X e k N
=
′ =
⇒
x n x n + n=
K j j j
X e X e W e
′ =
"on.o%ution
1 2
Win#o in$
-rocess of spectrum ana%/sis usin$ DFT -rocess of spectrum ana%/sis usin$ DFT
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
59/75
03/04/16 59
"utoff effect of DFT
!m"litude o( s2uare $a,e (unctions
s s"ectrum #e(ore and a(ter$indo$in5 #y s2uare $a,e (unction.cos; :4 n
Aea'a5e%istur#ance
'o%ution: increase 'amp%in$ points 6 or usin$ other 7in# of'o%ution: increase 'amp%in$ points 6 or usin$ other 7in# of in#o functionin#o function
2 f $8 f %/ i i $ DFT
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
60/75
03/04/16 60
3
; : x t
X j Ω
x n
j X e
⇒
- j
X k
X e k
N
=
=
%&T
22 8rrors of spectrum ana%/sis usin$ DFT8rrors of spectrum ana%/sis usin$ DFT
-rocess of spectrum ana%/sis usin$ DFT -rocess of spectrum ana%/sis usin$ DFT
3! Fence effect3! Fence effect = %&TH= e2ual inter,al sam"lin5 o( &T.= %&TH= e2ual inter,al sam"lin5 o( &T. /"ectrum (unction ,alue is omitted #et$een sam"lin5 "oints = inter,als./"ectrum (unction ,alue is omitted #et$een sam"lin5 "oints = inter,als.
/olution: Gero "addin5 or chan5e se2uence s len5th increase =./olution: Gero "addin5 or chan5e se2uence s len5th increase =.
3 DFT important parameters
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
61/75
03/04/16 61
p*
L−L
( ) BMC e
M
1 1 s
"
f F N NT T = = =
3 DFT important parametersTime-domain +s, T, , T!; +requency domainTime-domain +s, T, , T!; +requency domain FF
n
2L
:a$e 46 ana#o$ +requency +s/2
di$i"a# +requency L
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
62/75
03/04/16 62
3 4mportant parameter of DFT
ome im!or"an" conc#usionome im!or"an" conc#usion1 1 s
"
f F
N NT T = = =
1 1(1)
"
F T NT
= =
;- I( N unc. n/ed & incensement can only #e ac2uired #y
lo$erin5 f s 0/o s"ectrum analysis sco"e $ill #e small.
;3 f s unchan5ed & incensement can only #e ac2uired #y
increase N *p1N* that is increase sam"lin5 len5th.
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
63/75
03/04/16 63
3 4mportant parameter of DFT
ome im!or"an" conc#usionome im!or"an" conc#usion1 1 s
"
f F
N NT T = = =
1 1(1)
"
F T NT
= =
;- I( N unc. n/ed & incensement can only #e ac2uired #y
lo$erin5 f s 0/o s"ectrum analysis sco"e $ill #e small.
;3 f s unchan5ed & incensement can only #e ac2uired #y
increase N *p1N* that is increase sam"lin5 len5th.
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
64/75
03/04/16 64
%etermine sam"lin5 rate #y si5nal s hi5hest%etermine sam"lin5 rate #y si5nal s hi5hest(re2uency .(re2uency .
c f s f
-roce#ure of spectrum ana%/sis usin$ DFT
!dBust "arameters #y %&T results.!dBust "arameters #y %&T results.
%etermine e9tractin5 len5th = #y (re2uency%etermine e9tractin5 len5th = #y (re2uency
resolution.resolution.
DFT ro$rammin$ 85amp%e
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
65/75
03/04/16 65
01
0 1 1
1
( ( ( N N N
( ) N-) N N N
( N-) #N-)$#N-)$ N N N
W W W X W W W X
x x x N
W W W X N
÷÷ ÷÷ ÷=÷ ÷÷ ÷÷ ÷÷
K K M O M
M K
1
0
( ) ( ) 0 1 N
kn
N n
X k x n W k N −
== ≤ ≤ −
∑ k N →0 1
n N →0 1
DFT -ro$rammin$ 85amp%e
DFT 0atri5
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
66/75
03/04/16 66
function &'()*dft+xn,
N*lengt +xn,# . lengt of sequencen*/:N-1# . ti e sa ple(*/:N-1#
0N*exp+-" 2 pi N,#
n(*n3 (#0Nn(*0N%4n(# .calculate t e DFT 5atrix
'(*xn 0Nn(# .co pute DFT
0ore effecti.e metho#?
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
67/75
03/04/16 67
Fs * 7//# . Get t e analy8ed signal
T * 1 Fs#9 * 1///#t * +/:9-1, T#x * /% sin+2 pi ;/ t,#
plot+1/// t+1:2//,$x+1:2//,,#
< * dft+x, 9# . Discrete Fourier Transfor
f * Fs 2 linspace+/$1$9 2=1,#ste +f$2 abs+
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
68/75
03/04/16 68
l b E l #2l b E l #2
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
69/75
03/04/16 69
atlab Example #2atlab Example #2
f f 11 2C?2C? f f 22 2 05C?2 05C?
f f ss 10C?10C?
x)n*+sin), x)n*+sin), π π f f -- n.f n.f s s */ sin), */ sin), π π f f ,, n.f n.f s s * * !"!" #$%&'()*+,#$%&'()*+,
-. /0-. /0
2 1 0 05 f f f '% ∆ = − =
0 05 F '% ≤
min 200 s f N N
F = ⇒ =
(1)(1) 128 128 11 .(n).(n) 2323 D(E) D(E)
(2)(2) 256 256 11 .(n).(n) 2323 D(E) D(E)
(3)(3) 512 512 11 .(n).(n) 2323 D(E) D(E)
1 1 s "
f F
N NT T = = =
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
70/75
03/04/16 70
45-
9;n 0NDnN 6
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
71/75
03/04/16 71
9;n 0NDnN- 6
0
100
1 0
)*-
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
72/75
03/04/16 72
)*
-
9;n 0NDnN10 4
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
73/75
03/04/16 73
9;n 0NDnN10-4
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
74/75
03/04/16 74
Three Questions aboutThree Questions about DiscreteDiscrete Fourier TransformFourier Transform
Q1:Q1:
WW
HAT is DFT?HAT is DFT?
Q2:Q2: WWHY is DFT?HY is DFT?
Q3: HOQ3: HO WW to DFT?to DFT?
W !T is relationshi" #et$een %&T and other 'inds o( &ourierW !T is relationshi" #et$een %&T and other 'inds o( &ourier
Trans(orm)Trans(orm)
W * $e need %&T)W * $e need %&T)
+W to realize %&T) o$ to use %&T to sol,e the "ractical+W to realize %&T) o$ to use %&T to sol,e the "ractical
"ro#lems)"ro#lems)
-
8/20/2019 CHAPTER 2 Discrete Fourier Transform
75/75
HO08WO CHO08WO C
6771 +1 + !! 11 "#"#
1