DTFT Recall: A discrete-time 1-D signal x is a map from Z to C (i.e., x : Z → C).

DTFT

Chirayu Athalye Chirayu Athalye

DTFT
Basics of DTFT Basics of DTFT
DTFT
DTFT Properties ∞
X DFS
iω −iωn
DTFT Pairs Analysis Equation: X (e ) := x[n] e ; DTFT Properties
Frequency Response of n=−∞ DTFT Pairs
Discrete Time Fourier Transform (DTFT) LTI Systems Z
1 Frequency Response of
References
Synthesis Equation: x[n] = X (e iω ) e iωn dω. LTI Systems
2π 2π
References

Chirayu D. Athalye F DT F T
−→ X (e iω ) or x[n] ←−−−→ X (e iω ).
DTFT Pair: x[n] ←−
R
Notation: 2π denotes the integral over a period of length 2π.
Department of Electrical & Electronics Engineering,
BITS Pilani, K K Birla Goa Campus.
Remark
2025 1 Notice that DTFT analysis equation is similar to CFS synthesis equation.
∞
X
CFS Synthesis Equation: x(t) = ak e ikω0 t .
k=−∞

2 DTFT X : R → C is periodic with period 2π. (Why?)

3 When is the DTFT analysis equation well-defined?

M
X DTFT DTFT

Let XM (e iω ) := x[n] e −iωn . Chirayu Athalye Remark Chirayu Athalye

n=−M
Basics of DTFT 1 If x[·] is absolutely summable, then the analysis equation of DTFT converges uniformly. Basics of DTFT
DTFT DTFT
Theorem (DTFT: Uniform Convergence) DFS 2 If x[·] is square summable (finite energy), then the analysis equation of DTFT converges in DFS

∞
X DTFT Properties the mean-square sense. DTFT Properties

If x : Z → C is absolutely summable (i.e., |x[n]| < ∞), then DTFT Pairs 3 Uniform convergence ⇒ pointwise convergence. DTFT Pairs

n=−∞ Frequency Response of Frequency Response of

LTI Systems 4 Over a finite interval, uniform convergence ⇒ mean-square convergence. LTI Systems

lim XM (e iω ) = X (e iω ) ∈ C ∀ω ∈ R; References 5 Absolute summability ⇒ square summability; the converse is not true (Counterexample?). References
M→∞

i.e., XM → X (pointwise convergence). In fact, XM ⇒ X (uniform convergence). Remark

Using Dirac delta (impulse) function, it is possible to find DTFT of certain bounded
Theorem (DTFT: Mean-Square Convergence) sequences which are not square summable.
∞
X In this case, DTFT X : R → Ce , where Ce is the extended complex plane.
If x : Z → C is square summable (i.e., |x[n]|2 < ∞). Then,
n=−∞
Z Necessary Conditions for DTFT
iω iω 2
lim |X (e ) − XM (e )| dω = 0; 1 If DTFT exists for x : Z → C, then x[·] is a bounded sequence.
M→∞ 2π
2 Moreover, if DTFT is finite valued, then lim x[n] = lim x[n] = 0.
n→∞ n→−∞
i.e., XM converges to X in the mean-square sense.
 2π DTFT DTFT
Let x̃ : Z → C be periodic with the fundamental period N; thus, ω0 = N
. DTFT of Periodic Signals (Relation Between DTFT and DFS)
Chirayu Athalye Chirayu Athalye
2π
DFS (Representation-1) [Signals & Systems (Oppenheim et al.)] Basics of DTFT
Let x̃ : Z → C be periodic with the fundamental period N; thus, ω0 = N
. Basics of DTFT
DTFT DTFT
X 1 X
iω 0 kn
X̃ [k] WN−kn .
DFS DFS
N−1 x̃[n] = ak e or x̃[n] =
1 X 1 X DTFT Properties N DTFT Properties
Analysis Equation: ak := x̃[n] e −iω0 kn = x̃[n] e −iω 0 kn
; DTFT Pairs
k=hNi k=hNi
DTFT Pairs
N N n=0
n=hNi Frequency Response of ∞ Frequency Response of
LTI Systems iω
X LTI Systems
X N−1
X X̃ (e ) = 2π ak δ(ω − kω0 ).
iω 0 kn iω 0 kn References References
Synthesis Equation: x̃[n] = ak e = ak e . k=−∞
k=hNi k=0

DFS (Representation-2) [Discrete-Time Signal Processing (Oppenheim et al.)]

X i 2π
Analysis Equation: X̃ [k] := x̃[n] WNkn , where WN := e − N ;
n=hNi

1 X
Synthesis Equation: x̃[n] = X̃ [k] WN−kn .
N
k=hNi

DF S
DFS Pair: x̃[n] ←−
−→ X̃ [k]. Notice that X̃ [k] = Nak .
Figure: Fourier Series and Fourier Transform

DTFT DTFT

F F
Linearity Chirayu Athalye
−→ X (e iω )
x[n] ←− and −→ Y (e iω ).
y [n] ←− Chirayu Athalye

F F Basics of DTFT Basics of DTFT

−→ X1 (e iω ) and x2 [n] ←−
If x1 [n] ←− −→ X2 (e iω ), then DTFT Properties DTFT Properties

F iω iω DTFT Pairs Difference and Accumulation DTFT Pairs

a x1 [n] + b x2 [n] ←−
−→ a X1 (e ) + b X2 (e ),
Frequency Response of F Frequency Response of
−iω iω
where a, b ∈ C. LTI Systems Backward Difference: x[n] − x[n − 1] ←−
−→ (1 − e ) X (e ). LTI Systems

References References
F
−→ (e iω − 1) X (e iω ).
Forward Difference: x[n + 1] − x[n] ←−
Time Shifting and Frequency Shifting
F F n
−→ e −iωm X (e iω ), where m ∈ Z.
−→ X (e iω ), then x[n − m] ←−
Time Shifting: If x[n] ←−
X
Accumulation: If y [n] = x[m] ∀n ∈ Z, then
m=−∞
F F
−→ X (e iω ), then e iω1 n x[n] ←−
−→ X e i(ω−ω1 ) , where ω1 ∈ R.


Frequency Shifting: If x[n] ←− ∞
X (e iω ) X
Y (e iω ) = −iω
+ πX (e i0 ) δ(ω − 2πk),
1−e
Periodicity and Parseval’s Relation k=−∞

Periodicity: X e i(ω+2π)

iω
= X (e ). where the impulse train reflects the dc or average value that can result from summation.

∞
Differentiation in Frequency Domain
Z
X 1
Parseval’s Relation: |x[n]|2 = |X (e iω )|2 dω.
n=−∞
2π 2π F dX iω
nx[n] ←−
−→ i (e )
dω
 DTFT DTFT

F iω
x[n] ←−
−→ X (e ). Chirayu Athalye Chirayu Athalye

Basics of DTFT Basics of DTFT

Time Reversal DTFT Properties DTFT Properties

F −iω DTFT Pairs DTFT Pairs

x[−n] ←−
−→ X (e ). Frequency Response of Frequency Response of
LTI Systems LTI Systems

References References
Time Expansion
(  
x kn , if n is a multiple of k
Let k ∈ N and x(k) [n] :=
0, otherwise.

F
−→ X (e ikω ).
x(k) [n] ←−

Remark: Time-stretched sequence x(k) [n] is obtained from x[n] by placing (k − 1) zeros
between values of x[n].
Recall: Time-compressed sequence x[kn] is obtained by taking k th samples of x[n].

Homework
Let y [n] := x[2n]. Find Y (e iω ) in terms of X (e iω ).
Figure: Inverse Scaling between the Time and Frequency Domains.

DTFT DTFT

Chirayu Athalye Chirayu Athalye

F F
Conjugation Basics of DTFT −→ X1 (e iω )
x1 [n] ←− and −→ X2 (e iω ).
x2 [n] ←− Basics of DTFT

DTFT Properties DTFT Properties

F iω ∗ F ∗ −iω
If x[n] ←−
−→ X (e ), then x [n] ←−
−→ X (e ). DTFT Pairs DTFT Pairs

Frequency Response of Convolution Property Frequency Response of

LTI Systems LTI Systems
Conjugate Symmetry for Real-Valued Sequences F iω iω
References (x1 ∗ x2 )[n] ←−
−→ X1 (e ) X2 (e ). References

1 If x[n] is a real-valued sequence, then X (e iω ) = X ∗ (e −iω ). Note that

X (e iω ) = X ∗ (e −iω ) ⇐⇒ X (e −iω ) = X ∗ (e iω ). Multiplication/Modulation Property
Therefore,
Z
Re X (e iω ) and X (e iω ) are even functions; whereas, 1


a F
−→ X (e iω ) = X1 e iθ X2 e i(ω−θ) dθ.



x[n] = x1 [n] x2 [n] ←− (Periodic Convolution)
b Im X (e iω ) and ]X (e iω ) are odd functions.

2π 2π

F
F (
−→ Re X (e iω ) and Od x[n] ←−
−→ i Im X (e iω ) . X1 (e iω ), if ω ∈ [−π, π]



Furthermore, Ev x[n] ←− iω
Let X̂1 (e ) := . Note that
0, otherwise
2 If x[n] is real and even, then X (e iω ) is real and even.
Z ∞
If x[n] is real and odd, then X (e iω ) is purely imaginary and odd. 1
X (e iω ) = X̂1 e iθ X2 e i(ω−θ) dθ.
3



2π −∞
 DTFT DTFT

Signal DTFT Chirayu Athalye

2π
Chirayu Athalye
For periodic sequences (with the fundamental period N), ω0 = N
.
Basics of DTFT Basics of DTFT
1. δ[n] 1 DTFT Properties DTFT Properties
Signal DTFT DFS {a0 , . . . , aN−1 }
2. δ[n − m] e −iωm DTFT Pairs DTFT Pairs

Frequency Response of X ∞
X Frequency Response of
∞
1 X LTI Systems 1. ak e iω0 kn 2π ak δ(ω − kω0 ) {a0 , . . . , aN−1 } LTI Systems
3. u[n] +π δ(ω − 2πk) References k=hNi k=−∞ References
1 − e −iω
k=−∞
∞
1 ∀n ∈ Z
X
δ(ω − 2πk) a0 = 1 (N = 1)
4. an u[n], |a| < 1 2. x̃[n] = 1 2π
1 − ae −iω k=−∞
1
5. a−n u[−n], |a| < 1 ∞ ∞
1 For k ∈ {0, . . . , N − 1},
1 − ae iω
X X
3. δ[n − kN] 2π δ(ω − kω0 )
N ak = N1 .
ae −iω k=−∞ k=−∞
6. n an u[n], |a| < 1
(1 − ae −iω )2 ω
Periodic iff 2π1 = m 1
N ∈ Q.
∞
X For k(∈ {0, . . . , N − 1},
4. e iω1 n 2π δ(ω − ω1 − 2πk)
1, if k = m
Remark (Refer to the Difference and Accumulation Property of DTFT) k=−∞ ak =
0, otherwise.
Let y [n] − y [n − 1] = x[n]. Using the difference property, we can always obtain X (e iω ) from Y (e iω ); in
fact, (1 − e −iω )Y (e iω ) = X (e iω ). However, Y (e iω ) can be obtained from X (e iω ) provided the series in
Analysis Equation of y [n] converges (or X (e i0 ) = 0).
1
m and N are coprime and ω0 = ω1 /m. WLOG, ω1 ∈ [0, 2π); thus, m < N.

DTFT DTFT

Signal DTFT DFS {a0 , . . . , aN−1 } Chirayu Athalye

Homework Chirayu Athalye

∞ ω
Let |r1 | < 1. Show that
2 Basics of DTFT Basics of DTFT
X Periodic iff 2π1 = m N ∈ Q. For
π δ(ω − ω1 − 2πk) + k ∈ {0, . . . , N − 1}, F 1 − (r1 cos ω1 ) e −iω
r1n cos(ω1 n) u[n]
DTFT Properties DTFT Properties


←−
−→
5. cos(ω1 n)
k=−∞
∞
(
1
, if k ∈ {m, N − m}
DTFT Pairs 1 − (2r1 cos ω1 ) e −iω + r12 e −2iω DTFT Pairs
X ak = 2 Frequency Response of Frequency Response of
π δ(ω + ω1 − 2πk) 0, otherwise. −iω
LTI Systems
F (r1 sin ω1 ) e LTI Systems
r1n sin(ω1 n) u[n]


k=−∞ ←−
−→
References
1 − (2r1 cos ω1 ) e −iω + r12 e −2iω References
ω
∞ Periodic iff 2π1 = mN ∈ Q. For
π X k ∈ {0, . . . , N − 1},
δ(ω − ω1 − 2πk) −
i k=−∞ 
1
Homework
6. sin(ω1 n)  2i ,
 if k = m
∞
π X ak = −1
2i , if k = N − m 1 Show that
δ(ω + ω1 − 2πk) 
0, otherwise.

i k=−∞ if ω ∈ {2πk | k ∈ Z}
2m + 1,
( 
1, if |n| ≤ m F iω
x[n] = ←−
−→ X (e ) = sin

ω(2m+1)

0, otherwise  2
, otherwise.
sin( ω
2 )

Homework
Show that 2 Let x̃ : Z → R be periodic with the fundamental period N s.t. N2 ∈ N; ω0 = 2π . Show that
 N
1,
 if n > 0  2m+1
F 1 1  N , if k = 0
(
S(e iω ) =


←−
−→ − = −i cot ω/2 .

s[n] = 0, if n = 0 1, if |n| ≤ m DF S
 1 − e −iω 1 − e iω x̃[n] = N
←− −→ ak = sin kω0 (2m+1) 
−1, if n < 0 0, if m < |n| ≤ 2
  2  , if k ∈ {1, . . . , N − 1}.
 kω0
N sin 2
2
m and N are coprime and ω0 = ω1 /m. WLOG, ω1 ∈ [0, 2π); thus, m < N.
 DTFT DTFT

Chirayu Athalye Chirayu Athalye

Example (Rectangular Sequence) Example (Ideal Lowpass Filter)
( Basics of DTFT Basics of DTFT

1, if 0 ≤ n ≤ M DTFT Properties
ωc DTFT Properties
Find the DTFT of x[n] = h[n] = sinc ωc n


.
0, otherwise. DTFT Pairs
π π DTFT Pairs

Frequency Response of Frequency Response of

( LTI Systems LTI Systems
M
X M + 1, if ω ∈ {2πk | k ∈ Z} 1 Ideal LPF is non-causal.
iω −iωn References References
X (e ) = e = 1−e −iω(M+1)
, otherwise. 2 Since h[·] is not absolutely
n=0 1−e −iω
summable, it’s unstable.
Notice that for all ω ∈ R \ {2πk | k ∈ Z}, (
1, if |ω| < ωc ,
iω(M+1)
 iω(M+1) iω(M+1)
 H(e iω ) =
e− 2 e 2 − e− 2 sin ω(M+1) 0, if ωc < |ω| ≤ π; (a) Impulse Response


iωM
iω
X (e ) = iω iω iω
= 2
ω

e− 2 .
− 2 − 2 sin


e e −e
2 2 and it is periodic with period 2π.
iω
What is lim X (e )? How to get HPF from LPF?
ω→2π

(b) Frequency Response

sin(2ω) sin(2ω) sin(3ω) F
Remark: Note that lim = −2, lim = 2, and lim = 3. −→ H(e iω )?
In what sense, h[n] ←−
Figure: Discrete-time Ideal Lowpass Filter.
ω→π sin(ω) ω→2π sin(ω) ω→π sin(ω)

DTFT DTFT

Chirayu Athalye x[n] h[n] y [n] X (e iω ) H(e iω ) Y (e iω ) Chirayu Athalye

Example (Ideal Lowpass Filter) Basics of DTFT (a) Time Domain (b) Frequency Domain Basics of DTFT

DTFT Properties DTFT Properties

M Figure: DT-LTI (Discrete LSI) System
iω
X −iωn DTFT Pairs DTFT Pairs
HM (e ) := h[n] e . Frequency Response of Frequency Response of
n=−M LTI Systems Frequency Response and Eigenfunction Property of DT-LTI Systems LTI Systems

References References
F
1 −→ H(e iω ).
Consider a DT-LTI system with the impulse response h : Z → C. Let h[n] ←−
Mean-Square Convergence:
Z 2 The output of this system to an input signal x : R → C is given by
lim |H(e iω ) − HM (e iω )|2 dω = 0. y [n] = (h ∗ x)[n] ∀n ∈ Z.
M→∞ 2π
3 Therefore, using the convolution property of DTFT, we have
Pointwise Convergence: Y (e iω ) = H(e iω ) X (e iω ),
HM (e iω ) → H(e iω ) ∀ω ∈ [−π, π]\{±ωc }. where H(e iω ) is the Frequency Response of the DT-LTI system.

Figure: Convergence of DTFT? (Gibbs Phenomenon) 4 For the input signal x[n] := e iω1 n , where ω1 ∈ R, the output is given by
HM ⇒ H (Uniform Convergence). X∞
y [n] = h[k] e iω1 (n−k) = H(e iω1 ) e iω1 n ∀n ∈ Z.
k=−∞

Therefore, x[n] := e iω1 n is an eigenfunction of DT-LTI Systems.