ELEN E4810: Digital Signal Processing

Topic 3: Fourier domain

1. The Fourier domain
2. Discrete-Time Fourier Transform (DTFT)
3. Discrete Fourier Transform (DFT)
4. Convolution with the DFT

Dan Ellis 2013-09-23 1

 1. The Fourier Transform
! Basic observation (continuous time):
A periodic signal can be decomposed
into sinusoids at integer multiples of the
fundamental frequency
! i.e. if x÷ (t) = x÷ (t +T )
we can approach x÷ with
Harmonics
M " #
! 2πk of the
x̃(t) ≈ ak cos t + φk fundamental
T
k=0
Dan Ellis 2013-09-23 2
 M " #
2πk
Fourier Series
!
ak cos t + φk
T
k=0
! For a square wave,
! k−1
(−1) 2 k1 k = 1, 3, 5, . . .
φk = 0; ak =
0 otherwise
! " ! " ! "
2π 1 2π 1 2π
i.e. x(t) = cos
T
t − cos
3 T
3t + cos
5 T
5t − . . .
"

%#$

! %#$

! "
! "#$ ! " ! %#$ % %#$ " "#$

Dan Ellis 2013-09-23 3

M
 Fourier Domain
! x is equivalently described
by its Fourier Series ak 1.0
parameters:
k−1 1 1 2 3 4 5 6 7 k
ak = (−1) 2 k = 1, 3, 5, . . .
k φk
!

Negative ak is
equivalent to phase of π 1 2 3 4 5 6 7 k
!
M
j 2πk
T t
! Complex form: x̃(t) ≈ ck e
k=−M

Dan Ellis 2013-09-23 4

 !
M

Fourier Analysis x̃(t) ≈

k=−M
j 2πk
ck e T t

! How to find {|ck |}, {arg{ck }} ?

Inner product with
(conjugate) complex sinusoids:
! T /2
1 −j 2πk
ck = x(t)e T t dt
T −T /2

Dan Ellis 2013-09-23 5

 Fourier Series Analysis
! T /2
1 −j 2πk
T t
! Thus, ck = T x(t)e dt
−T /2
−j 2πk
T t
because complex sinusoids e
pick out the corresponding sinusoidal
components linearly combined in
!
M
j 2πk
T t
x(t) = ck e
k=−M

Dan Ellis 2013-09-23 7

 Fourier Transform
Fourier series for periodic signals
!

extends naturally to Fourier Transform

for any (CT) signal (not just periodic):
! ∞
Fourier
X(jΩ) = x(t)e−jΩt dt Transform (FT)
∞
!
1 ∞
jΩt Inverse Fourier
x(t) = X(jΩ)e dΩ Transform (IFT)
2π ∞
! Discrete index k " continuous freq. #
Dan Ellis 2013-09-23 8
 Fourier Transform
! Mapping between two continuous
functions: ##!1"# !
/*0*/++
,+12 3#!

3$!

!"!#
!!""
3%!

3&!
!"!4 ! #!!! $!!! %!!! &!!!
56*7+,+89

!
$%&'#!1"(
/
3!"!4
! !"!!# !"!!$ !"!!% !"!!& /!"
'()*+,+-*.
!

</!"

</
! #!!! $!!! %!!! &!!!
2# ambiguity 56*7+,+89

Dan Ellis 2013-09-23 9

 Fourier Transform of a sine
j! 0 t
! Assume x(t) = e ∞
1
Z
jΩt
Now, since x(t) = X(Ω)e dΩ
2π −∞
...we know X(Ω) = 2πδ(Ω − Ω0)
...where δ(x) is the Dirac delta function
$(x-x0)
(continuous time) i.e.
# ! ( x " x0 ) f ( x) dx = f ( x0 ) x0
f(x)
x
j! 0 t
! " x(t) = Ae X(!) = A" (! # ! 0 )
Dan Ellis 2013-09-23 10
 Fourier Transforms
Time Frequency
Fourier Series Continuous Discrete
(FS) periodic ~x(t) infinite ck
Fourier Continuous Continuous
Transform (FT) infinite x(t) infinite X(Ω)
Discrete-Time Discrete Continuous
FT (DTFT) infinite x[n] periodic X(ejω)
Discrete FT Discrete Discrete
(DFT) finite/pdc ~x[n] finite/pdc X[k]
Dan Ellis 2013-09-23 11
 2. Discrete Time FT (DTFT)
! FT defined for discrete sequences:
#
j!
X(e ) = $ x[n]e " j!n
DTFT
n="#
! Summation (not integral)
! Discrete (normalized)
frequency variable ω
! Argument is ejω, not jω

Dan Ellis 2013-09-23 12

 DTFT example
! e.g. x[n] = αnáμ[n], |α| < 1 -1 1 2 3 4 5 6 7 n
$
" X(e ) = %n=#$ " µ[n]e
j! n # j!n

$ # j! n
= %n=0 ("e ) ! !
S = " c # cS = " c
n n
!" !"#"_$#&t 1
#$ = # j!
n =0 n =1

%
1 # "e # S $ cS = c = 1 0
!!""#t#! $%&'!""#t#(
# /
1 ( |c | < 1 )
"
/ "/ #/ $/ %/
t
#S =
&

! / "/ #/ $/ %/
t
</
1$c
Dan Ellis 2013-09-23 13
 Periodicity of X(ejω)
! X(ejω) has periodicity 2π in ω :
X(e j (! +2 " )
) = # x[n]e $ j (! +2 " )n

= # x[n]e $ j!n
e $ j 2 "n
= X(e ) j!

!!""#t#! $%&'!""#t#(
# /

" t
/ "/ #/ $/ %/
&
t
</
! / "/ #/ $/ %/

! Phase ambiguity of ejω makes it implicit

Dan Ellis 2013-09-23 14
 Inverse DTFT (IDTFT)
! Same basic ÒFourier SynthesisÓ form:
1 !
x[n] =
2!
$# ! X(e j"
)e j"n
d" IDTFT

! Note: continuous, periodic X(ejω)

" discrete, infinite x[n] ...
! IDTFT is actually Fourier Series analysis
(except for sign of ω)

Dan Ellis 2013-09-23 15

 DTFTs of simple sequences
#
! x[n] = δ[n] j!
X(e ) = $ x[n]e " j!n

n="#
" j! 0
=e =1 (for all ω)

! i.e. x[n]" " X(ejω)

" " " δ[n]" " 1
x[n] X(ej&)

-3 -2 -1 1 2 3 n -π π ω

Dan Ellis 2013-09-23 19

 DTFTs of simple sequences
1 ! IDTFT
j! n
! x[n] = e 0 : x[n] =
2!
$ #!
X(e j"
)e j"n
d"
j!
" ""X(e ") = 2"" # $ (!" % ! 0 ) over -π < ω < π
but X(ejω) must be periodic in ω
e j! 0 n
&k 2" # $ (! % ! 0 % 2"k)
! If ω0 = 0 then x[n] = 1 # n
X(ejω)
so 1 &k 2! " # ($ % 2!k)
0 2π 4π

Dan Ellis 2013-09-23 20

 DTFTs of simple sequences
! From before:
n 1
! µ[n] " j# ( |α | < 1)
1 " !e
! μ[n] tricky - not finite
1
µ[n] ! j"
+ %k #$ (" + 2#k)
1! e
DTFT of 1/2
Dan Ellis 2013-09-23 21
 DTFT properties
! Linear:
j$ j$
!g[n] + "h[n] # !G(e ) + "H (e )
! Time shift:
! j#n 0 j#
g[n ! n0 ] " e G(e )
! Frequency shift: ÔdelayÕ
j (! #! 0 )
e j! 0 n
g[n] " G (e ) in
frequency

Dan Ellis 2013-09-23 22

 DTFT example
! x[n] = δ[n] + αn μ[n-1] ?
" = δ[n] + α(αn-1 μ[n-1])
j! % # j! $1 1 (
" X(e ) = 1+ " ' e $ *
& 1 # "e )
# j!

# j! # j! # j!
"e 1 # "e + "e
= 1+ # j!
= # j!
1 # "e 1 # "e
1
= # j! x[n] = αn μ[n]
1 # "e
Dan Ellis 2013-09-23 23
 #

DTFT symmetry X(e j! ) = $ x[n]e" j!n

n="#

!If x[n] " X(ejω) then...

" x[-n]" X(e-jω) from summation
" x*[n] " X*(e-jω) (e-jω)* = ejω
1
Re{x[n]} " XCS(e ) = [ X (e j! ) + X * (e" j! )]
jω
2
conjugate symmetry cancels Im parts on IDTFT
1
jIm{x[n]} " XCA (ejω)= [ X (e ) " X (e )]
j! * " j!

2
" xcs[n] " Re{X(ejω)}
" xca[n] " jIm{X(ejω)}"
Dan Ellis 2013-09-23 24
 DTFT of real x[n]
! When x[n] is pure real, X(ejω) = X*(e-jω)
XCS
xcs[n] $ xev[n] = xev[-n] XR(ejω) = XR(e-jω)
xca[n] $ xod[n] = -xod[-n] XI(ejω) = -XI(e-jω)

&'#(
!%&!$"

x[n] real, even !"#$

X(ejω) even, real !"#!$"

Dan Ellis 2013-09-23 25

 DTFT and convolution
! Convolution: x[n] = g[n]! h[n]
j! $ # j!n
X(e ) = %n =#$ ( g[n] " h[n]) e
= %n ( %k g[k]h[n # k])e # j!n

= %k ( g[k]e # j!k
%n h[n # k]e # j! (n #k )
)
j! j!
= G(e ) & H (e )
Convolution
j# j#
g[n]! h[n] " G(e )H (e ) becomes
multiplication
Dan Ellis 2013-09-23 26
 Convolution with DTFT
j# j#
! Since g[n]! h[n] " G(e )H (e )
we can calculate a convolution by:
! finding DTFTs of g, h " G, H
! multiply them: G!H
! IDTFT of product is result, g[n]! h[n]
G(e j! )
g[n] DTFT
Y (e j! ) y[n]
IDTFT
h[n] DTFT
H (e j! )

Dan Ellis 2013-09-23 27

M
 DTFT convolution example
1
j!
! x[n] = αnáμ[n] X(e ) = " j!
1 " #e
! h[n] = δ[n] - αδ[n-1]
" " " " H (e ) = 1 " # (e
j! " j! $1
) $1
! y[n] = x[n] % h[n]
j! j! j!
" " Y (e ) = H (e )X(e )
1
! j# (
= $ 1 ! "e ) = 1
! j#

1 ! "e
" y[n] = δ[n] i.e. ...
Dan Ellis 2013-09-23 28
 ParsevalÕs relation
! ÒEnergyÓ in time and frequency domains
are equal:
1 #
" g[n]h [n] = 2# &% # G(e )H (e )d$
* j$ * j$

!n

! If g = h, then gág* = |g|2 = energy...

Dan Ellis 2013-09-23 30

 Energy density spectrum
2
! Energy of sequence ! g = # g[n]
"n

1 "
%$ " G(e j# 2
! By Parseval !g = ) d#
2"
! Define Energy Density Spectrum (EDS)
j! j! 2
Sgg (e ) = G(e )

Dan Ellis 2013-09-23 31

 EDS and autocorrelation
! Autocorrelation of g[n]:
"
rgg [!] = # g[n]g[n ! !] = g[n]$ g[!n]
n=!"
j! " j!
DTFT {rgg [!]} = G(e )G(e )
! If g[n] is real, G(e-jω) = G*(ejω), so
j! 2 j! no phase
DTFT {rgg [!]} = G(e ) = Sgg (e ) info.
! Mag-sq of spectrum is DTFT of autoco
Dan Ellis 2013-09-23 32
 3. Discrete FT (DFT)
Discrete FT Discrete Discrete
(DFT) finite/pdc x[n] finite/pdc X[k]
! A finite or periodic sequence has only
N unique values, x[n] for 0 $ n < N
! Spectrum is completely defined by N
distinct frequency samples
! Divide 0..2π into N equal steps,
2πk
{ωk } =
N
Dan Ellis 2013-09-23 33
 DFT and IDFT
! Uniform sampling of DTFT spectrum:
N #1 2 "k
#j n
X[k] = X(e ) ! = 2 "k = $ x[n]e
j! N

N n=0

N !1 2π/N
! DFT: X[k] = " x[n]W N
kn
1
n=0 WN

2"
!j
where WN = e N
i.e. -1/Nth of a revolution

Dan Ellis 2013-09-23 34

 DFT examples !
1 n=0
! Finite impulse x[n] =
0 n = 1 . . . N − 1
N!−1
" ⇒ X[k] = kn 0
x[n]WN = WN = 1 ∀k
n=0
! Periodic
! sinusoid:
"
2πrn (r ∈ Z) = 1 !W −rn + W rn "
x[n] = cos N N
N 2
1
!N −1 rn kn
⇒ X[k] = 2 n=0 (W −rn
N + W N )W N
"
N
(0 ! k < N) k = r, k = N − r
2
=
0 otherwise
Dan Ellis 2013-09-23 36
 DFT: Matrix form
N "1
! X[k] = #n=0 x[n] ! W N
kn
as a matrix multiply:
 

X[0]
 1 1 1 ··· 1 x[0]

(N −1) 
 X[1]  1 WN1 WN2 ··· WN  x[1]

   2(N −1)   
 X[2]  
=1 WN2 WN4 ··· WN  x[2]

.. ..
  
   .. .. .. .. .. 
 
 . .
  . . . .  .
X[N − 1] (N −1) 2(N −1) (N −1)2 x[N − 1]
1 WN WN ··· WN

! i.e. X = DN ! x
Dan Ellis 2013-09-23 37
 Matrix IDFT
! If X = DN ! x
!1
then x = DN " X
! i.e. inverse DFT is also just a matrix,
 
1 1 1 ··· 1
−(N −1) 
1
 WN−1 WN−2 ··· WN 
1 
1 WN−2 WN−4
−2(N −1) 
WN
D−1
N = ··· 
N .. .. .. .. .. 
. . . . .


−(N −1) −2(N −1) −(N −1)2
1 WN WN ··· WN
=1/NDN*
Dan Ellis 2013-09-23 38
 DFT and MATLAB
! MATLAB is concerned with sequences
not continuous functions like X(ejω)
! Instead, we use the DFT to sample
X(ejω) on an (arbitrarily-fine) grid:
! X = freqz(x,1,w); samples the DTFT
of sequence x at angular frequencies in w
! X = fft(x); calculates the N-point DFT
of an N-point sequence x

Dan Ellis 2013-09-23 39

M
 DFT and DTFT
#
¥ continuous freq ω
DTFT
j!
X(e ) = $ x[n]e " j!n
¥ infinite x[n], -'<n<'
n="#
N !1
¥ discrete freq k=Nω/2π
DFT X[k] = " x[n]W N
kn
¥ finite x[n], 0$n<N
n=0

! DFT ÔsamplesÕ DTFT at discrete freqs:

j! X[k]
X[k] = X(e ) ! = 2 "k X(ejω)
N ω
k=1...
Dan Ellis 2013-09-23 40
 !

" ! # ! ! $ %& ' ('

 Properties: Circular time shift
! DFT properties mirror DTFT, with twists:
! Time shift must stay within N-pt ÔwindowÕ
kn0
g[!n − n0 #N ] ↔ WN G[k]
! Modulo-N indexing keeps index between
0 and N-1: !
g[n − n0 ] n ≥ n0
g[!n − n0 #N ] =
g[N + n − n0 ] n < n0
0 $ n0 < N

Dan Ellis 2013-09-23 48

 Circular time shift
! Points shifted out to the right donÕt
disappear Ð they come in from the left
g[n] g[<n-2>5]
ÔdelayÕ by 2

1 2 3 4
n 1 2 3 4
n
5-pt sequence

! Like a Ôbarrel shifterÕ:

!"#$#%&'!#%()"

Dan Ellis 2013-09-23 49

 Circular time reversal
! Time reversal is tricky in Ômodulo-NÕ
indexing - not reversing the sequence:
x÷ [n]
5-pt sequence
made periodic
-7 -6 -5 -4 -3 -2 -1 1 2 3 4 5 6 7 8 9 10 11
n
x÷ [ !n N ]
Time-reversed
periodic sequence
-7 -6 -5 -4 -3 -2 -1 1 2 3 4 5 6 7 8 9 10 11
n
! Zero point stays fixed; remainder flips
Dan Ellis 2013-09-23 50
 Circular convolution
! Can also do entire convolution with
modulo-N indexing
! Hence, Circular Convolution:
N !1

" g[m ]h[ n ! m N ] # G[k]H[k]

m=0

! Written as g[n] !
N h[n]

Dan Ellis 2013-09-23 53

 Circular convolution example
! 4 pt sequences: g[n]={1 2 0 1} h[n]={2 2 1 0}
N !1

" g[m ]h[ n ! m N ] 1 2 3 n 1 2 3 n

m=0

g[n] 4 h[n]={4 7 5 4}
h[<n - 0>4] n !1
1 2 3

h[<n - 1>4] n !2
1 2 3

h[<n - 2>4]
1 2 3 n
n !0
1 2 3

h[<n - 3>4] check: g[n] * h[n]

n !1
1 2 3 ={2 6 5 4 2 1 0}
Dan Ellis 2013-09-23 54
 DFT properties summary
! Circular convolution
N !1
"m=0 g[m ]h[ n ! m N ] # G[k]H[k]
! Modulation
N #1
g[n] ! h[n] " 1
N $m=0 G[m]H[ k # m N ]
! Duality
G [n] ! N " g[ #k N ]
! Parseval N !1 2 N !1 2
"n=0 x[n] = 1
N "k=0 X [k ]
Dan Ellis 2013-09-23 55
 Overlap-Add convolution
! Very long g[n] " break up into
segments, convolve piecewise, overlap
" bound size!of DFT, processing delay
g[n] i · N ≤ n < (i + 1) · N
! Make gi [n] =
0 otherwise
!
⇒ g[n] = i gi [n]
!
⇒ h[n] ⊛ g[n] = i h[n] ⊛ gi [n]
! Called Overlap-Add (OLA) convolution...
Dan Ellis 2013-09-23 58
 Overlap-Add convolution
g[n] h[n]
M
n n
g0[n] g0[n] * h[n]
N=
L+M-1
n n
g1[n] g1[n] * h[n]
n n
g2[n] g2[n] * h[n]
n n
L 2L 3L valid OLA sum
h[n] * g[n]
n
ML 2L 3L
Dan Ellis 2013-09-23 59