DTFT

2: Three Different Fourier
Transforms
• Fourier Transforms
• Convergence of DTFT
• DTFT Properties
• DFT Properties
• Symmetries
• Parseval’s Theorem
• Convolution
• Sampling Process
• Zero-Padding
• Phase Unwrapping
• Uncertainty principle
• Summary
• MATLAB routines Transforms
DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 1 / 14

Fourier Transforms

Transforms
Three different Fourier Transforms:
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Fourier Transforms

Transforms
• Convergence of DTFT • CTFT (Continuous-Time Fourier Transform): x(t) → X(jΩ)
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Fourier Transforms

Transforms
• DTFT (Discrete-Time Fourier Transform): x[n] → X(ejω )
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Fourier Transforms

Transforms
• DTFT Properties
• DFT Properties
• Symmetries
• DFT a.k.a. FFT (Discrete Fourier Transform): x[n] → X[k]
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Fourier Transforms

Transforms
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Sampling Process Forward Transform Inverse Transform
• Zero-Padding
R∞ −jΩt 1
R∞ jΩt
• Phase Unwrapping CTFT X(jΩ) = −∞
x(t)e dt x(t) = 2π −∞
X(jΩ)e dΩ
• Summary
• MATLAB routines

Fourier Transforms

Transforms
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
R∞ −jΩt 1
R∞ jΩt
• Phase Unwrapping CTFT X(jΩ) = −∞ x(t)e dt x(t) = 2π −∞
X(jΩ)e dΩ
P∞ 1
Rπ
X(e ) = −∞ x[n]e−jωn
jω jω jωn
• Summary DTFT x[n] = 2π −π
X(e )e dω
• MATLAB routines

Fourier Transforms

Transforms
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
R∞ −jΩt 1
R∞ jΩt
X(jΩ)e dΩ
P∞ 1
Rπ
jω jω jωn
X(e )e dω
• MATLAB routines
We use Ω for “real” and ω = ΩT for “normalized” angular frequency.

f
Nyquist frequency is at ΩNyq = 2π 2s = Tπ and ωNyq = π .

Fourier Transforms

Transforms
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
R∞ −jΩt 1
R∞ jΩt
X(jΩ)e dΩ
P∞ 1
Rπ
jω jω jωn
X(e )e dω
• MATLAB routines PN −1 kn 1
PN −1 j2π kn
DFT X[k] = 0 x[n]e−j2π N x[n] = N 0 X[k]e N

f

Fourier Transforms

Transforms
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
R∞ −jΩt 1
R∞ jΩt
X(jΩ)e dΩ
P∞ 1
Rπ
jω jω jωn
X(e )e dω
• MATLAB routines PN −1 kn 1
PN −1 j2π kn
DFT X[k] = 0 x[n]e−j2π N x[n] = N 0 X[k]e N

f
For “power signals” (energy ∝ duration), CTFT & DTFT are unbounded.
Fix this by normalizing:
1
RA −jΩt
X(jΩ) = limA→∞ 2A −A
x(t)e dt
1
PA −jωn
X(ejω ) = limA→∞ 2A+1 −A x[n]e

Convergence of DTFT
P∞
DTFT: X(ejω ) = x[n]e−jωn does not converge for all x[n].
Transforms −∞
jω
PK −jωn
• Convergence of DTFT Consider the finite sum: XK (e ) = −K x[n]e
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Convergence of DTFT
P∞
Transforms −∞
jω
PK −jωn
• DTFT Properties
• DFT Properties Strong Convergence:
• Symmetries
• Parseval’s Theorem x[n] absolutely summable ⇒ X(ejω ) converges uniformly
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Convergence of DTFT
P∞
Transforms −∞
jω
PK −jωn
• DTFT Properties
• Symmetries
• Parseval’s Theorem x[n] absolutely summable ⇒X(ejω ) converges uniformly
P∞
• Convolution
• Sampling Process −∞ |x[n]| < ∞ ⇒ supω X(ejω ) − XK (ejω ) −−−−→ 0
K→∞
• Zero-Padding
• Summary
• MATLAB routines

Convergence of DTFT
P∞
Transforms −∞
jω
PK −jωn
• DTFT Properties
• Symmetries
P∞
• Convolution
K→∞
• Zero-Padding
Weaker convergence:
• Summary x[n] finite energy ⇒ X(ejω ) converges in mean square
• MATLAB routines

Convergence of DTFT
P∞
Transforms −∞
jω
PK −jωn
• DTFT Properties
• Symmetries
P∞
• Convolution
K→∞
• Zero-Padding
Weaker convergence:
X(ejω ) − XK (ejω )2 dω −−−−→ 0
• MATLAB routines P∞ 2 1
Rπ
−∞ |x[n]| < ∞ ⇒ 2π −π K→∞

Convergence of DTFT
P∞
Transforms −∞
jω
PK −jωn
• DTFT Properties
• Symmetries
P∞
• Convolution
K→∞
• Zero-Padding
Weaker convergence:
X(ejω ) − XK (ejω )2 dω −−−−→ 0
Rπ
−∞ |x[n]| < ∞ ⇒ 2π −π K→∞
Example: x[n] = sin πn

0.5πn
1 K=5 1 K=20 1 K=50

0.8 0.8 0.8
0.6 0.6 0.6
jω
jω
jω
K
K
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0
-0.2 -0.2 -0.2
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
ω/2π (rad/sample) ω/2π (rad/sample) ω/2π (rad/sample)

Convergence of DTFT
P∞
Transforms −∞
jω
PK −jωn
• DTFT Properties
• Symmetries
P∞
• Convolution
K→∞
• Zero-Padding
Weaker convergence:
X(ejω ) − XK (ejω )2 dω −−−−→ 0
Rπ
−∞ |x[n]| < ∞ ⇒ 2π −π K→∞
Example: x[n] = sin πn

0.5πn
1 K=5 1 K=20 1 K=50

0.8 0.8 0.8
0.6 0.6 0.6
jω
jω
jω
K
K
0.4 0.4 0.4
0.2 0.2 0.2
0 0 0
-0.2 -0.2 -0.2
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
ω/2π (rad/sample) ω/2π (rad/sample) ω/2π (rad/sample)
Gibbs phenomenon:
Converges at each ω as K → ∞ but peak error does not get smaller.
DTFT Properties
P∞
DTFT: X(ejω ) = x[n]e−jωn
Transforms −∞
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

DTFT Properties
P∞
Transforms −∞
• DTFT is periodic in ω : X(ej(ω+2mπ) ) = X(ejω ) for integer m.
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

DTFT Properties
P∞
Transforms −∞
• DTFT Properties
• DFT Properties • DTFT is the z -Transform evaluated at the point ejω :
• Symmetries P∞
• Parseval’s Theorem X(z) = −∞ x[n]z −n
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

DTFT Properties
P∞
Transforms −∞
• DTFT Properties
• Symmetries P∞
• Convolution
DTFT converges iff the ROC includes |z| = 1.
• Zero-Padding
• Summary
• MATLAB routines

DTFT Properties
P∞
Transforms −∞
• DTFT Properties
• Symmetries P∞
• Convolution
• Zero-Padding
• Phase Unwrapping • DTFT is the same as the CTFT of a signal comprising impulses at
• Summary
the sample times (Dirac δ functions) of appropriate heights:
P
• MATLAB routines xδ (t) = x[n]δ(t − nT )

DTFT Properties
P∞
Transforms −∞
• DTFT Properties
• Symmetries P∞
• Convolution
• Zero-Padding
• Summary
P
• MATLAB routines xδ (t) = x[n]δ(t − nT )
Equivalent to multiplying a continuous x(t) by an impulse train.

DTFT Properties
P∞
Transforms −∞
• DTFT Properties
• Symmetries P∞
• Convolution
• Zero-Padding
• Summary
P P∞
• MATLAB routines xδ (t) = x[n]δ(t − nT )= x(t) × −∞ δ(t − nT )

DTFT Properties
P∞
Transforms −∞
• DTFT Properties
• Symmetries P∞
• Convolution
• Zero-Padding
• Summary
P P∞
jω
P∞ −jωn
Proof: X(e ) = −∞ x[n]e

DTFT Properties
P∞
Transforms −∞
• DTFT Properties
• Symmetries P∞
• Convolution
• Zero-Padding
• Summary
P P∞
jω
P∞ −jωn
Proof: X(e ) = −∞ x[n]eR
P∞ ∞ −jω Tt
n=−∞ x[n] −∞
δ(t − nT )e dt

DTFT Properties
P∞
Transforms −∞
• DTFT Properties
• Symmetries P∞
• Convolution
• Zero-Padding
• Summary
P P∞
jω
P∞ −jωn
P∞ ∞ −jω Tt
n=−∞ x[n] −∞
δ(t − nT )e dt
(i) R ∞ P∞ t
= −∞ n=−∞ x[n]δ(t − nT )e−jω T dt
P∞
(i) OK if −∞ |x[n]| < ∞.

DTFT Properties
P∞
Transforms −∞
• DTFT Properties
• Symmetries P∞
• Convolution
• Zero-Padding
• Summary
P P∞
jω
P∞ −jωn
P∞ ∞ −jω Tt
n=−∞ x[n] −∞
δ(t − nT )e dt
(i) R ∞ P∞ t
= −∞ n=−∞ x[n]δ(t − nT )e−jω T dt
(ii) R ∞
= −∞ xδ (t)e−jΩt dt
P∞
(i) OK if −∞ |x[n]| < ∞. (ii) use ω = ΩT .

DFT Properties
2: Three Different Fourier PN −1 kn

Transforms DFT: X[k] = 0 x[n]e−j2π N
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

DFT Properties

• DTFT Properties
• DFT Properties
• Symmetries
• Parseval’s Theorem Case 1: x[n] = 0 for n ∈
/ [0, N − 1]
• Convolution
• Sampling Process DFT is the same as DTFT at ωk = 2π
N k.
• Zero-Padding
• Summary
• MATLAB routines

DFT Properties

jω
P∞ −jωn
• DTFT Properties
DTFT: X(e )= −∞ x[n]e
• DFT Properties
• Symmetries
/ [0, N − 1]
• Convolution
N k.
• Zero-Padding
• Summary
• MATLAB routines

DFT Properties

jω
P∞ −jωn
• DTFT Properties
• DFT Properties
• Symmetries
/ [0, N − 1]
• Convolution
N k.
• Zero-Padding
• Phase Unwrapping −1
• Uncertainty principle The {ωk } are uniformly spaced from ω = 0 to ω = 2π NN .
• Summary
• MATLAB routines

DFT Properties

jω
P∞ −jωn
• DTFT Properties
• DFT Properties
• Symmetries
/ [0, N − 1]
• Convolution
N k.
• Zero-Padding
• Summary
• MATLAB routines
DFT is the z -Transform evaluated at N equally spaced points
around the unit circle beginning at z = 1.

DFT Properties

jω
P∞ −jωn
• DTFT Properties
• DFT Properties
• Symmetries
/ [0, N − 1]
• Convolution
N k.
• Zero-Padding
• Summary
• MATLAB routines
Case 2: x[n] is periodic with period N
DFT equals the normalized DTFT

DFT Properties

jω
P∞ −jωn
• DTFT Properties
• DFT Properties
• Symmetries
/ [0, N − 1]
• Convolution
N k.
• Zero-Padding
• Summary
• MATLAB routines

N
X[k] = limK→∞ 2K+1 × XK (ejωk )

DFT Properties

jω
P∞ −jωn
• DTFT Properties
• DFT Properties
• Symmetries
/ [0, N − 1]
• Convolution
N k.
• Zero-Padding
• Summary
• MATLAB routines

N
X[k] = limK→∞ 2K+1 × XK (ejωk )
jω
PK −jωn
where XK (e ) = −K x[n]e

Symmetries
If x[n] has a special property then X(ejω )and X[k] will have
Transforms
corresponding properties as shown in the table (and vice versa):
• DTFT Properties
• DFT Properties
• Symmetries One domain Other domain
• Convolution Discrete Periodic
• Zero-Padding Symmetric Symmetric
Antisymmetric Antisymmetric
• Summary Real Conjugate Symmetric
• MATLAB routines
Imaginary Conjugate Antisymmetric
Real + Symmetric Real + Symmetric
Real + Antisymmetric Imaginary + Antisymmetric

Symmetries
Transforms
• DTFT Properties
• DFT Properties
• MATLAB routines
Symmetric: x[n] = x[−n]

Symmetries
Transforms
• DTFT Properties
• DFT Properties
• MATLAB routines

X(ejω ) = X(e−jω )

Symmetries
Transforms
• DTFT Properties
• DFT Properties
• MATLAB routines

X[k] = X[(−k)mod N ] = X[N − k] for k > 0

Symmetries
Transforms
• DTFT Properties
• DFT Properties
• MATLAB routines

Conjugate Symmetric: x[n] = x∗ [−n]

Symmetries
Transforms
• DTFT Properties
• DFT Properties
• MATLAB routines

Conjugate Symmetric: x[n] = x∗ [−n]
Conjugate Antisymmetric: x[n] = −x∗ [−n]

Parseval’s Theorem

Transforms
Fourier transforms preserve “energy”
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines


Transforms
2 1
|X(jΩ)|2 dΩ
R R
• Convergence of DTFT CTFT |x(t)| dt = 2π
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines


Transforms
2
|x(t)| dt = 2π |X(jΩ)|2 dΩ
1
R R
• Convergence of DTFT CTFT
• DTFT Properties
jω 2
P∞ 2 1
Rπ
DTFT −∞ |x[n]| = 2π −π X(e ) dω
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines


Transforms
2
|x(t)| dt = 2π |X(jΩ)|2 dΩ
1
R R
• DTFT Properties
jω 2
P∞ 2 1
Rπ
DTFT −∞ |x[n]| = 2π −π X(e ) dω
• DFT Properties
• Symmetries
PN −1 2 1
PN −1 2
• Convolution
DFT 0 |x[n]| = N 0 |X[k]|
• Zero-Padding
• Summary
• MATLAB routines


Transforms
2
|x(t)| dt = 2π |X(jΩ)|2 dΩ
1
R R
• DTFT Properties
jω 2
P∞ 2 1
Rπ
DTFT −∞ |x[n]| = 2π −π X(e ) dω
• DFT Properties
• Symmetries
PN −1 2 1
PN −1 2
• Convolution
DFT 0 |x[n]| = N 0 |X[k]|
• Zero-Padding More generally, they actually preserve complex inner products:
• Uncertainty principle PN −1 PN −1
∗ 1
• Summary
• MATLAB routines 0 x[n]y [n] = N 0 X[k]Y ∗ [k]


Transforms
2
|x(t)| dt = 2π |X(jΩ)|2 dΩ
1
R R
• DTFT Properties
jω 2
P∞ 2 1
Rπ
DTFT −∞ |x[n]| = 2π −π X(e ) dω
• DFT Properties
• Symmetries
PN −1 2 1
PN −1 2
• Convolution
DFT 0 |x[n]| = N 0 |X[k]|
∗ 1
• Summary
Unitary matrix viewpoint for DFT:
If we regard x and X as vectors, then X = Fx where F is

kn
a symmetric matrix defined by fk+1,n+1 = e−j2π N .


Transforms
2
|x(t)| dt = 2π |X(jΩ)|2 dΩ
1
R R
• DTFT Properties
jω 2
P∞ 2 1
Rπ
DTFT −∞ |x[n]| = 2π −π X(e ) dω
• DFT Properties
• Symmetries
PN −1 2 1
PN −1 2
• Convolution
DFT 0 |x[n]| = N 0 |X[k]|
∗ 1
• Summary
Unitary matrix viewpoint for DFT:
If we regard x and X as vectors, then X = Fx where F is

kn
a symmetric matrix defined by fk+1,n+1 = e−j2π N .
1 H
The inverse DFT matrix is F−1 = N F
equivalently, G = √1 F is a unitary matrix with GH G = I.
N

Convolution

Transforms
DTFT: Convolution → Product
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Convolution

Transforms
P∞
• Fourier Transforms x[n] = g[n] ∗ h[n]= k=−∞ g[k]h[n − k]
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Convolution

Transforms
x[n] = g[n] ∗ h[n]= ∞
P
k=−∞ g[k]h[n − k]
• DTFT Properties ⇒ X(ejω ) = G(ejω )H(ejω )
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Convolution

Transforms
x[n] = g[n] ∗ h[n]= ∞
P
k=−∞ g[k]h[n − k]
• DFT Properties
• Symmetries
DFT: Circular convolution→ Product
PN −1
• Convolution x[n] = g[n] ⊛N h[n]= k=0 g[k]h[(n − k)modN ]
• Zero-Padding
• Summary
• MATLAB routines

Convolution

Transforms
x[n] = g[n] ∗ h[n]= ∞
P
k=−∞ g[k]h[n − k]
• DFT Properties
• Symmetries
PN −1
• Zero-Padding ⇒ X[k] = G[k]H[k]
• Summary
• MATLAB routines

Convolution

Transforms
x[n] = g[n] ∗ h[n]= ∞
P
k=−∞ g[k]h[n − k]
• DFT Properties
• Symmetries
PN −1
• Summary
• MATLAB routines
g[n] : h[n] :

Convolution

Transforms
x[n] = g[n] ∗ h[n]= ∞
P
k=−∞ g[k]h[n − k]
• DFT Properties
• Symmetries
PN −1
• Summary
• MATLAB routines
g[n] : h[n] : g[n] ∗ h[n] :

Convolution

Transforms
x[n] = g[n] ∗ h[n]= ∞
P
k=−∞ g[k]h[n − k]
• DFT Properties
• Symmetries
PN −1
• Summary
• MATLAB routines
g[n] : h[n] : g[n] ∗ h[n] : g[n] ⊛3 h[n]

Convolution

Transforms
x[n] = g[n] ∗ h[n]= ∞
P
k=−∞ g[k]h[n − k]
• DFT Properties
• Symmetries
PN −1
• Uncertainty principle DTFT: Product→ Circular Convolution ÷2π
• Summary
• MATLAB routines y[n] = g[n]h[n]
g[n] : h[n] : g[n] ∗ h[n] : g[n] ⊛3 h[n]

Convolution

Transforms
x[n] = g[n] ∗ h[n]= ∞
P
k=−∞ g[k]h[n − k]
• DFT Properties
• Symmetries
PN −1
• Summary
1
⇒ Y (ejω ) = 2π G(ejω ) ⊛π H(ejω ) =
1
Rπ
2π −π
G(ejθ )H(ej(ω−θ) )dθ
g[n] : h[n] : g[n] ∗ h[n] : g[n] ⊛3 h[n]

Convolution

Transforms
x[n] = g[n] ∗ h[n]= ∞
P
k=−∞ g[k]h[n − k]
• DFT Properties
• Symmetries
PN −1
• Summary
1
⇒ Y (ejω ) = 2π G(ejω ) ⊛π H(ejω ) =
1
Rπ
2π −π
DFT: Product→ Circular Convolution ÷N
y[n] = g[n]h[n]
g[n] : h[n] : g[n] ∗ h[n] : g[n] ⊛3 h[n]

Convolution

Transforms
x[n] = g[n] ∗ h[n]= ∞
P
k=−∞ g[k]h[n − k]
• DFT Properties
• Symmetries
PN −1
• Summary
1
⇒ Y (ejω ) = 2π G(ejω ) ⊛π H(ejω ) =
1
Rπ
2π −π
DFT: Product→ Circular Convolution ÷N
y[n] = g[n]h[n]
⇒ Y [k] = N1 G[k] ⊛N H[k]
g[n] : h[n] : g[n] ∗ h[n] : g[n] ⊛3 h[n]

Sampling Process
Time Time Frequency

CTFT
Analog −→

Sampling Process
Time Time Frequency

CTFT
Analog −→
Low Pass CTFT

* → −→
Filter

Sampling Process
Time Time Frequency

CTFT
Analog −→
Low Pass CTFT

* → −→
Filter
DTFT
Sample × → −→

Sampling Process
Time Time Frequency

CTFT
Analog −→
Low Pass CTFT

* → −→
Filter
DTFT
DTFT
Window × → −→

Sampling Process
Time Time Frequency

CTFT
Analog −→
Low Pass CTFT

* → −→
Filter
DTFT
DTFT
Window × → −→
DFT
DFT −→

Zero-Padding
Zero padding means added extra zeros onto the end of x[n] before
Transforms
performing the DFT.
• DTFT Properties
• DFT Properties
• Symmetries Time x[n]
• Convolution
• Sampling Process Windowed
• Zero-Padding
Signal
• Summary
• MATLAB routines

Zero-Padding
Transforms
performing the DFT.
• DTFT Properties
• DFT Properties
• Symmetries Time x[n]
• Convolution
• Zero-Padding
Signal
• Summary
• MATLAB routines
With zero-
padding

Zero-Padding
Transforms
performing the DFT.
• DTFT Properties
• DFT Properties
• Symmetries Time x[n] Frequency |X[k]|
• Convolution
• Zero-Padding
Signal
• Summary
• MATLAB routines
With zero-
padding

Zero-Padding
Transforms
performing the DFT.
• DTFT Properties
• DFT Properties
• Convolution
• Zero-Padding
Signal
• Summary
• MATLAB routines
With zero-
padding

Zero-Padding
Transforms
performing the DFT.
• DTFT Properties
• DFT Properties
• Convolution
• Zero-Padding
Signal
• Summary
• MATLAB routines
With zero-
padding
• Zero-padding causes the DFT to evaluate the DTFT at more values

of ωk . Denser frequency samples.

Zero-Padding
Transforms
performing the DFT.
• DTFT Properties
• DFT Properties
• Convolution
• Zero-Padding
Signal
• Summary
• MATLAB routines
With zero-
padding

• Width of the peaks remains constant: determined by the length and
shape of the window.

Zero-Padding
Transforms
performing the DFT.
• DTFT Properties
• DFT Properties
• Convolution
• Zero-Padding
Signal
• Summary
• MATLAB routines
With zero-
padding

• Width of the peaks remains constant: determined by the length and
shape of the window.
• Smoother graph but increased frequency resolution is an illusion.

Phase Unwrapping

Transforms
Phase of a DTFT is only defined to within an integer multiple of 2π .
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Uncertainty principle x[n] |X[k]|
• Summary
• MATLAB routines

Phase Unwrapping

Transforms
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines
∠X[k]

Phase Unwrapping

Transforms
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines
∠X[k]
Phase unwrapping adds multiples of 2π onto each ∠X[k] to make the

phase as continuous as possible.

Phase Unwrapping

Transforms
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines
∠X[k] ∠X[k] unwrapped
Phase unwrapping adds multiples of 2π onto each ∠X[k] to make the

phase as continuous as possible.

Uncertainty principle
2: Three Different Fourier R 21 R 12

Transforms t2 |x(t)|2 dt ω 2 |X(jω)|2 dω 1
CTFT uncertainty principle: R R ≥
|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
The first term measures the “width” of x(t) around t = 0.
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
It is like σ if |x(t)| was a zero-mean probability distribution.
• Convolution
• Zero-Padding
• Summary
• MATLAB routines


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
• Convolution The second term is similarly the “width” of X(jω) in frequency.
• Zero-Padding
• Summary
• MATLAB routines


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
• Zero-Padding A signal cannot be concentrated in both time and frequency.
• Summary
• MATLAB routines


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
• Uncertainty principle Proof Outline:
• Summary R 2 R 2
• MATLAB routines Assume |x(t)| dt = 1⇒ |X(jω)| dω = 2π [Parseval]


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
• Summary R 2 R 2
Set v(t) = dxdt ⇒ V (jω) = jωX(jω) [by parts]


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
• Summary R 2 R 2
R dx 1 2
∞ R 1 2 1
Now tx dt dt= 2 tx (t)t=−∞ − 2 x dt =0− 2 [by parts]


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
• Summary R 2 R 2
∞
Now tx dt dt= 2 tx (t) t=−∞ − 2 x dt = 0 − 21 [by parts]
1
R dx 2
R 1 2

2
R 2
So 41 = tx dx dx dt [Schwartz]
R R 2 2

dt dt ≤ t x dt dt


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
• Summary R 2 R 2
∞
1
R dx 2
R 1 2

2
R 2
R R 2 2

dt dt ≤ t x dt dt
R 2

t2 x2 dt
R
= |v(t)| dt


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
• Summary R 2 R 2
∞
1
R dx 2
R 1 2

2
R 2
R R 2 2

dt dt ≤ t x dt dt
R 2
R 2 2 1 R 2

t2 x2 dt
R
= |v(t)| dt = t x dt 2π |V (jω)| dω


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
• Summary R 2 R 2
∞
1
R dx 2
R 1 2

2
R 2
R R 2 2

dt dt ≤ t x dt dt
R
2
R
1 2

t2 x2 dt t2 x2 dt 2π |V (jω)| dω
R R
= |v(t)| dt =
R 2 2 1 R 2 2

= t x dt 2π ω |X(jω)| dω


|x(t)|2 dt |X(jω)|2 dω 2
• DTFT Properties
• DFT Properties
2
• Symmetries
• Summary R 2 R 2
∞
1
R dx 2
R 1 2

2
R 2
R R 2 2

dt dt ≤ t x dt dt
R
2
R
1 2

t2 x2 dt t2 x2 dt 2π |V (jω)| dω
R R
= |v(t)| dt =
R 2 2 1 R 2 2

= t x dt 2π ω |X(jω)| dω
No exact equivalent for DTFT/DFT but a similar effect is true

Summary

Transforms • Three types: CTFT, DTFT, DFT
• DTFT Properties
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Summary

• DTFT Properties ◦ DTFT = CTFT of continuous signal × impulse train
• DFT Properties
• Symmetries
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Summary

• DFT Properties
• Symmetries ◦ DFT = DTFT of periodic or finite support signal
• Convolution
• Zero-Padding
• Summary
• MATLAB routines

Summary

• DFT Properties
• Convolution • DFT is a scaled unitary transform
• Zero-Padding
• Summary
• MATLAB routines

Summary

• DFT Properties
• Zero-Padding
• DTFT: Convolution → Product; Product → Circular Convolution
• Summary
• MATLAB routines

Summary

• DFT Properties
• Zero-Padding
• Summary
• MATLAB routines
• DFT: Product ↔ Circular Convolution

Summary

• DFT Properties
• Zero-Padding
• Summary
• MATLAB routines
• DFT: Zero Padding → Denser freq sampling but same resolution

Summary

• DFT Properties
• Zero-Padding
• Summary
• MATLAB routines
• Phase is only defined to within a multiple of 2π .

Summary

• DFT Properties
• Zero-Padding
• Summary
• MATLAB routines
• Whenever you integrate over frequency you need a scale factor
1 1
◦ 2π for CTFT and DTFT or N for DFT
◦ e.g. Inverse transform, Parseval, frequency domain convolution

Summary

• DFT Properties
• Zero-Padding
• Summary
• MATLAB routines
• Whenever you integrate over frequency you need a scale factor
1 1
◦ 2π for CTFT and DTFT or N for DFT
◦ e.g. Inverse transform, Parseval, frequency domain convolution
For further details see Mitra: 3 & 5.

MATLAB routines

Transforms fft, ifft DFT with optional zero-padding
fftshift swap the two halves of a vector
• DTFT Properties conv convolution or polynomial multiplication (not circular)
• DFT Properties
• Symmetries x[n]⊛y[n] real(ifft(fft(x).*fft(y)))
• Convolution unwrap remove 2π jumps from phase spectrum
• Zero-Padding
• Summary
• MATLAB routines

DTFT

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DTFT

Uploaded by

Copyright:

Available Formats

2: Three Different Fourier

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 1 / 14

2: Three Different Fourier

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 2 / 14

2: Three Different Fourier

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 2 / 14

2: Three Different Fourier

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 2 / 14

2: Three Different Fourier

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 2 / 14

2: Three Different Fourier

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 2 / 14

2: Three Different Fourier

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 2 / 14

2: Three Different Fourier

We use Ω for “real” and ω = ΩT for “normalized” angular frequency.

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 2 / 14

2: Three Different Fourier

We use Ω for “real” and ω = ΩT for “normalized” angular frequency.

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 2 / 14

2: Three Different Fourier

We use Ω for “real” and ω = ΩT for “normalized” angular frequency.

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 2 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 3 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 3 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 3 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 3 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 3 / 14

Example: x[n] = sin πn

1 K=5 1 K=20 1 K=50

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 3 / 14

Example: x[n] = sin πn

1 K=5 1 K=20 1 K=50

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 4 / 14

2: Three Different Fourier PN −1 kn

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 5 / 14

2: Three Different Fourier PN −1 kn

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 5 / 14

2: Three Different Fourier PN −1 kn

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 5 / 14

2: Three Different Fourier PN −1 kn

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 5 / 14

2: Three Different Fourier PN −1 kn

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 5 / 14

2: Three Different Fourier PN −1 kn

Case 2: x[n] is periodic with period N

DFT equals the normalized DTFT

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 5 / 14

2: Three Different Fourier PN −1 kn

Case 2: x[n] is periodic with period N

DFT equals the normalized DTFT

DSP and Digital Filters (2017-10159) Fourier Transforms: 2 – 5 / 14

2: Three Different Fourier PN −1 kn