EE 301 Signals and Systems I

Homework 4
(due Dec. 15, 2019)

4 questions.

Question 1
𝑒 𝑗Ω
A signal 𝑥[𝑛] has F.T. 𝑋(𝑒 𝑗Ω ) = 2+𝑒 −𝑗Ω .

a) What is the total sum of the signal 𝑥[𝑛]?

∞ ∞
1
∑ 𝑥[𝑛] = ( ∑ 𝑥[𝑛] 𝑒 −𝑗Ω𝑛 )| = (𝑋(𝑒 𝑗Ω ))|Ω=0 = 𝑋(𝑒 𝑗0 ) =
3
𝑛=−∞ 𝑛=−∞ Ω=0

b) Let 𝑦[𝑛] be the accumulation of 𝑥[𝑛], i.e, 𝑦[𝑛] = ∑𝑛𝑘=−∞ 𝑥[𝑘]. What is the total sum of the
signal 𝑦[𝑛]?
∞
𝑗Ω
𝑋(𝑒 𝑗Ω )
𝑌(𝑒 ) = + 𝜋𝑋(𝑒 𝑗0 ) ∑ 𝛿(Ω − 2𝜋𝑘)
1 − 𝑒 −𝑗Ω
𝑘=−∞
𝑒 𝑗Ω ∞ ∞
𝑗Ω 2+𝑒 −𝑗Ω 1 𝑒 𝑗Ω 1
𝑌(𝑒 ) = + 𝜋 ∑ 𝛿(Ω − 2𝜋𝑘) = + 𝜋 ∑ 𝛿(Ω − 2𝜋𝑘)
1 − 𝑒 −𝑗Ω 3 2 − 𝑒 −𝑗Ω − 𝑒 −𝑗2Ω 3
𝑘=−∞ 𝑘=−∞

∞
1 1
∑ 𝑦[𝑛] = 𝑌(𝑒 𝑗0 ) = + 𝜋 𝛿(0)
2−1−1 3
𝑛=−∞
It is infinity.

𝜋
c) What is the numerical value of |𝑋(𝑒 𝑗Ω )| at Ω = 2 , Ω = 𝜋? If 𝑋(𝑒 𝑗Ω ) is the frequency response
of an LTI system, what would you say about its characteristics?
π

𝑗
π 𝑒 𝑗2 𝑗 |𝑗| 1
|𝑋 (𝑒 )| = |
2
π |=| |= =
2+𝑒 −𝑗
2
2−𝑗 |2 − 𝑗| √5
𝑒 𝑗𝜋 −1
|𝑋(𝑒 𝑗𝜋 )| = | −𝑗𝜋
|=| |=1
2+𝑒 2−1
It has high-pass characteristics because
𝜋
|𝑋(𝑒 𝑗0 )| < |𝑋 (𝑒 𝑗 2 )| < |𝑋(𝑒 𝑗𝜋 )|

d) As a last step, determine 𝑥[𝑛].

𝑒 𝑗Ω 1 𝑗Ω 1 1 𝑗Ω 1 𝑛
𝑋(𝑒 𝑗Ω ) = = 𝑒 ( 1 ) = 𝑒 𝐹 {(− ) 𝑢[𝑛]}
2 + 𝑒 −𝑗Ω 2 1 − (− )𝑒 −𝑗Ω 2 2
2
1 1 𝑛+1 1 1 𝑛+1
= 𝐹 {(− ) 𝑢[𝑛 + 1]} = 𝐹 { (− ) 𝑢[𝑛 + 1]}
2 2 2 2

1 1 𝑛+1
𝑥[𝑛] = (− ) 𝑢[𝑛 + 1]
2 2
1 𝑛+2
𝑥[𝑛] = − (− ) 𝑢[𝑛 + 1]
2
Question 2

We defined periodic convolution in class and found the DTFS coefficients for the periodically
convoluted signal. Alternatively, one can use DTFT properties and find the resultant DTFS
coefficients.

Let the periodic signals 𝑥𝑝 [𝑛] and 𝑦𝑝 [𝑛] have period 𝑁 and DTFS coefficients 𝑎𝑘 and 𝑏𝑘 ,
respectively.

a) Define 𝑥𝑠𝑝 [𝑛] = 𝑥𝑝 [𝑛] for 0 ≤ 𝑛 < 𝑁 and zero otherwise. Write down the DTFS coefficients
𝑎𝑘 of 𝑥𝑝 [𝑛] in terms of the DTFT of 𝑥𝑠𝑝 [𝑛]?

∞ 𝑁−1

𝑋𝑠𝑝 (𝑒 ) = ∑ 𝑥𝑠𝑝 [𝑛] 𝑒 −𝑗Ω𝑛 = ∑ 𝑥𝑠𝑝 [𝑛] 𝑒 −𝑗Ω𝑛

𝑗Ω

𝑛=−∞ 𝑛=0
𝑁−1 𝑁−1
1 2π
−𝑗 𝑘𝑛 1 2π 1
𝑎𝑘 = ∑ 𝑥𝑝 [𝑛] 𝑒 N = ∑ 𝑥𝑠𝑝 [𝑛] 𝑒 −𝑗 N 𝑘𝑛 = (𝑋𝑠𝑝 (𝑒 𝑗Ω ))| 2π
𝑁 𝑁 𝑁 Ω= 𝑘
N
𝑛=0 𝑛=0

b) Show that the periodic convolution of 𝑥𝑝 [𝑛] and 𝑦𝑝 [𝑛] equals 𝑥𝑠𝑝 [𝑛] ∗ 𝑦𝑝 [𝑛] and is periodic
with 𝑁.

𝑧[𝑛] ≜ 𝑥𝑠𝑝 [𝑛] ∗ 𝑦𝑝 [𝑛]

∞ 𝑁−1 𝑁−1

𝑧[𝑛] = ∑ 𝑥𝑠𝑝 [𝑘]𝑦𝑝 [𝑛 − 𝑘] = ∑ 𝑥𝑠𝑝 [𝑘]𝑦𝑝 [𝑛 − 𝑘] = ∑ 𝑥𝑝 [𝑘]𝑦𝑝 [𝑛 − 𝑘]

𝑘=−∞ 𝑘=0 𝑘=0
𝑁−1 𝑁−1

𝑧[𝑛 + 𝑘𝑁] = ∑ 𝑥𝑝 [𝑘]𝑦𝑝 [𝑛 + 𝑘𝑁 − 𝑘] = ∑ 𝑥𝑝 [𝑘]𝑦𝑝 [𝑛 − 𝑘] = 𝑧[𝑛], ∀𝑘 ∈ 𝑍

𝑘=0 𝑘=0

c) What is the DTFT of 𝑥𝑠𝑝 [𝑛] ∗ 𝑦𝑝 [𝑛]?

𝑍(𝑒 𝑗Ω ) = 𝑋𝑠𝑝 (𝑒 𝑗Ω ) 𝑌𝑝 (𝑒 𝑗Ω )
∞
𝑗Ω
2𝜋
𝑌𝑝 (𝑒 ) = 2𝜋 ∑ 𝑏𝑘 𝛿 (Ω − 𝑘)
𝑁
𝑘=−∞
∞ ∞
2𝜋 2𝜋 2𝜋
𝑗Ω 𝑗Ω
𝑍(𝑒 ) = 𝑋𝑠𝑝 (𝑒 ) 2𝜋 ∑ 𝑏𝑘 𝛿 (Ω − 𝑘) = 2𝜋 ∑ 𝑋𝑠𝑝 (𝑒 𝑗 𝑁 𝑘 ) 𝑏𝑘 𝛿 (Ω − 𝑘)
𝑁 𝑁
𝑘=−∞ 𝑘=−∞

d) Based on part c, determine the DTFS coefficients of the periodically convoluted signal in
terms of 𝑎𝑘 and 𝑏𝑘 .
1
𝑎𝑘 = (𝑋 (𝑒 𝑗Ω ))| 2π , (𝑝𝑎𝑟𝑡 𝑎)
𝑁 𝑠𝑝 Ω= 𝑘
N

∞
2𝜋 2𝜋
𝑍(𝑒 ) = 2𝜋 ∑ 𝑋𝑠𝑝 (𝑒 𝑗 𝑁 𝑘 ) 𝑏𝑘 𝛿 (Ω −
𝑗Ω
𝑘) , (𝑝𝑎𝑟𝑡 𝑐)
𝑁
𝑘=−∞
∞
𝑗Ω
2𝜋
𝑍(𝑒 ) = 2𝜋 ∑ 𝑁𝑎𝑘 𝑏𝑘 𝛿 (Ω − 𝑘)
𝑁
𝑘=−∞

𝑐𝑘 = 𝑁𝑎𝑘 𝑏𝑘

where 𝑐𝑘 are DTFS coefficients of 𝑧[𝑛].

Question 3

We will investigate the conversion of a DT signal into CT and then back to DT in this question as
complementary to our coverage in class.

𝑥𝑐 (𝑡) 𝑦𝑐 (𝑡) 𝑦[𝑛]

𝑥[𝑛]
Physical
D/C ℎ1 (𝑡) ℎ2 (𝑡) C/D
link

π, −π ≤ Ω < 0
Consider a DT signal 𝑥[𝑛] with 𝑋(𝑒 𝑗Ω ) = { . The two CT filters have frequency
π − Ω, 0≤Ω<π
responses
|𝜔|𝑇1 2π 2π
1− , |𝜔| ≤ 1, |𝜔| ≤
𝐻1 (𝑗𝜔) = { 2π 𝑇1 and 𝐻2 (𝑗𝜔) = { 𝑇1 .
0, o. w. 0, o. w.

a) Define a signal 𝑥𝑝 (𝑡) = ∑∞ 𝑗Ω

𝑛=−∞ 𝑥[𝑛]δ(t − n𝑇1 ). Evaluate 𝑋𝑝 (𝑗𝜔) in terms of 𝑋(𝑒 ) and
sketch over 𝜔 ∈ [−3𝜋/𝑇1 , 3𝜋/𝑇1 ].

∞ ∞ ∞

𝑋𝑝 (𝑗𝜔) = ∫ 𝑥𝑝 (𝑡)𝑒 −𝑗𝜔𝑡 𝑑𝑡 = ∫ ∑ 𝑥[𝑛]δ(t − n𝑇1 ) 𝑒 −𝑗𝜔𝑡 𝑑𝑡

−∞ − ∞ 𝑛=−∞
∞ ∞ ∞
−𝑗𝜔𝑡
= ∑ 𝑥[𝑛] ∫ δ(t − n𝑇1 )𝑒 𝑑𝑡 = ∑ 𝑥[𝑛] 𝑒 −𝑗𝜔𝑛𝑇1
𝑛=−∞ −∞ 𝑛=−∞
∞

𝑋(𝑒 𝑗Ω ) = ∑ 𝑥[𝑛] 𝑒 −𝑗Ω𝑛

𝑛=−∞
𝑋𝑝 (𝑗𝜔) = (𝑋(𝑒 𝑗Ω ))|
Ω=ωT1
𝑗Ω 𝜋 −π ≤ Ω < 0 𝑗(Ω+k2π)
𝑋(𝑒 ) = { , 𝑎𝑛𝑑 𝑋(𝑒 ) = 𝑋(𝑒 𝑗Ω ), ∀𝑘 ∈ 𝑍
𝜋−Ω 0≤Ω<π
𝜋 −π ≤ ωT1 < 0 Ω + k2π Ω
𝑋𝑝 (𝑗𝜔) = { , 𝑎𝑛𝑑 𝑋𝑝 (𝑗 ) = 𝑋 (𝑗 ) , ∀𝑘 ∈ 𝑍
𝜋 − 𝜔𝑇1 0 ≤ ωT1 < π 𝑇1 𝑇1
π
𝜋 − ≤ω<0
T1 k2π
𝑋𝑝 (𝑗𝜔) = { π , 𝑎𝑛𝑑 𝑋𝑝 (𝑗 (ω + )) = 𝑋(𝑗𝜔), ∀𝑘 ∈ 𝑍
𝜋 − 𝜔𝑇1 0 ≤ ω < 𝑇1
T1
b) Find the FT of 𝑥𝑐 (𝑡) = ∑∞
𝑛=−∞ 𝑥[𝑛]ℎ1 (t − n𝑇1 ). Sketch it over 𝜔 ∈ [−3𝜋/𝑇1 , 3𝜋/𝑇1 ].

δ(t) → ℎ1 (𝑡) → ℎ1 (t)

δ(t − n𝑇1 ) → ℎ1 (𝑡) → ℎ1 (t − n𝑇1 )
𝑥𝑝 (𝑡) → ℎ1 (𝑡) → 𝑥𝑐 (𝑡)
𝑋𝑐 (𝑗𝜔) = 𝑋𝑝 (𝑗𝜔)𝐻1 (𝑗𝜔)
|𝜔|𝑇1 2π π
(𝜋 − 𝜔𝑇1 − 2𝜋) (1 − ) − ≤ω<−
2π T1 T1
|𝜔|𝑇1 π
𝜋 (1 − ) − ≤ω<0
2π T1
𝑋𝑐 (𝑗𝜔) = |𝜔|𝑇1 π
(𝜋 − 𝜔𝑇1 ) (1 − ) 0≤ω<
2π T1
|𝜔|𝑇1 π 2π
𝜋 (1 − ) ≤ω<
2π T1 T1
{ 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

c) The physical link is also an LTI and have frequency response 𝐻𝑙 (𝑗𝜔) = 3. Find and sketch
𝑌𝑐 (𝑗𝜔).
𝑌𝑐 (𝑗𝜔) = 𝑋𝑐 (𝑗𝜔)𝐻𝑙 (𝑗𝜔)𝐻2 (𝑗𝜔) = 3𝑋𝑐 (𝑗𝜔)𝐻2 (𝑗𝜔) = 3𝑋𝑐 (𝑗𝜔)
d) What is the Nyquist rate for the sampling of 𝑦𝑐 (𝑡) and the corresponding sampling period in
terms of 𝑇1 ?
2π
𝑌𝑐 (𝑗𝜔) = 0, 𝑓𝑜𝑟 |𝜔| >
T1
2𝜋 2π
>2
𝑇𝑠 T1
𝑇1
𝑇𝑠 <
2
e) We will now perform undersampling, that is, we will pick an insufficient period. Define a
signal 𝑦𝑐𝑝 (𝑡) = ∑∞
𝑛=−∞ 𝑦𝑐 (n𝑇2 )δ(t − n𝑇2 ) for 𝑇2 = 𝑇1 . Evaluate 𝑌𝑐𝑝 (𝑗𝜔) in terms of 𝑌𝑐 (𝑗𝜔).
∞

𝑦𝑐𝑝 (𝑡) = 𝑦𝑐 (𝑡) ∑ δ(t − n𝑇2 )

𝑛=−∞
∞
1
𝑌𝑐𝑝 (𝑗𝜔) = 𝑌 (𝑗𝜔) ⋆ 𝐹 { ∑ δ(t − n𝑇2 )}
2𝜋 𝑐
𝑛=−∞
∞ ∞
2𝜋 2π
𝐹 { ∑ δ(t − n𝑇2 )} = ∑ δ (j (ω − k ))
𝑇2 T2
𝑛=−∞ 𝑘=−∞
∞
1 2π
𝑌𝑐𝑝 (𝑗𝜔) = ∑ 𝑌𝑐 (j (ω − k ))
𝑇2 T2
𝑘=−∞

f) Write the FT of 𝑦[𝑛] = 𝑦𝑐 (n𝑇2 ). Simplify it until you have a summation only in terms of
𝑋(𝑒 𝑗Ω ) and 𝐻1 (𝑗𝜔). Check this summation in Ω ∈ [−𝜋, 𝜋] and draw 𝑌(𝑒 𝑗Ω ).
∞ ∞ ∞ ∞
𝑡
𝑗Ω −𝑗Ω𝑛 −𝑗Ω
𝑌(𝑒 ) = ∑ 𝑦[𝑛]𝑒 = ∑ 𝑦𝑐 (𝑛𝑇2 )𝑒 −𝑗Ω𝑛 = ∫ 𝑦𝑐 (𝑡)𝑒 𝑇2 ∑ δ(t − n𝑇2 ) 𝑑𝑡
𝑛=−∞ 𝑛=−∞ −∞ 𝑛=−∞
∞ ∞
𝑡
−𝑗Ω
= ∫ 𝑦𝑐 (𝑡)𝑒 𝑇2 ∑ δ(t − n𝑇2 ) 𝑑𝑡
−∞ 𝑛=−∞
∞
𝑡
−𝑗Ω
= ∫ 𝑦𝑐𝑝 (𝑡)𝑒 𝑇2 𝑑𝑡 = (𝑌𝑐𝑝 (𝑗𝜔))| Ω
𝜔=
𝑇2
−∞

Ω
𝑌(𝑒 𝑗Ω ) = (𝑌𝑐𝑝 (𝑗𝜔))| Ω = 𝑌𝑐𝑝 (𝑗 )
𝜔=
𝑇2
𝑇2

∞
1 2π
𝑌𝑐𝑝 (𝑗𝜔) = ∑ 𝑌𝑐 (j (ω − k ))
𝑇2 T2
𝑘=−∞
∞
1 Ω 2π
𝑌(𝑒 𝑗Ω ) = ∑ 𝑌𝑐 (j ( − k ))
𝑇2 𝑇2 T2
𝑘=−∞

𝑌𝑐 (𝑗𝜔) = 3𝑋𝑐 (𝑗𝜔)

∞
𝑗Ω
3 Ω 2π
𝑌(𝑒 ) = ∑ 𝑋𝑐 (j ( − k ))
𝑇2 𝑇2 T2
𝑘=−∞

𝑋𝑐 (𝑗𝜔) = 𝑋𝑝 (𝑗𝜔)𝐻1 (𝑗𝜔)

∞
3 Ω 2π Ω 2π
𝑌(𝑒 𝑗Ω ) = ∑ 𝑋𝑝 (j ( − k )) 𝐻1 (j ( − k ))
𝑇2 𝑇2 T2 𝑇2 T2
𝑘=−∞
 𝑋𝑝 (𝑗𝜔) = (𝑋(𝑒 𝑗Ω ))|
Ω=ωT1
Ω
𝑋(𝑒 𝑗Ω ) = 𝑋𝑝 (𝑗 )
T1
∞
𝑗Ω
3 Ω 2π
𝑌(𝑒 ) = ∑ 𝑋(𝑒 𝑗(Ω−k2π)T1 /𝑇2 )𝐻1 (j ( − k ))
𝑇2 𝑇2 T2
𝑘=−∞
𝑇1 = 𝑇2
∞
3 Ω 2π
𝑌(𝑒 ) = 𝑋(𝑒 𝑗Ω ) ∑ 𝐻1 (j ( − k ))
𝑗Ω
𝑇1 𝑇1 T1
𝑘=−∞
Check this summation in Ω ∈ [−𝜋, 𝜋] :
2π
𝐻1 (𝑗𝜔) = 0, 𝑓𝑜𝑟 |𝜔| >
𝑇1
Ω 2π
𝐻1 (j ( − k )) = 0, 𝑓𝑜𝑟 Ω > k2π + 2π 𝑎𝑛𝑑 Ω < k2π − 2π
𝑇1 T1
∞
3 Ω 2π
𝑌(𝑒 ) = 𝑋(𝑒 𝑗Ω ) ∑ 𝐻1 (j ( − k )) ,
𝑗Ω
Ω ∈ [−𝜋, 𝜋]
𝑇1 𝑇1 T1
𝑘=−∞
1
3 Ω 2π
𝑌(𝑒 𝑗Ω ) = 𝑋(𝑒 𝑗Ω ) ∑ 𝐻1 (j ( − k )) , Ω ∈ [−𝜋, 𝜋]
𝑇1 𝑇1 T1
𝑘=−1
|𝜔|𝑇1 2π
1−, |𝜔| ≤
𝐻1 (𝑗𝜔) = { 2π 𝑇1
0, o. w.
Ω
Ω 1− , 0<Ω≤π
𝐻1 (𝑗 ) = { 2π
𝑇1 Ω
1+ , −π ≤ Ω ≤ 0
2π

Ω 2π Ω − 2π Ω
𝐻1 (𝑗 ( − )) = {1 − , 0 < Ω ≤ π = {2π , 0 < Ω ≤ π
𝑇1 T1 2π
0, −π ≤ Ω ≤ 0 0, −π ≤ Ω ≤ 0

Ω 2π 0, 0<Ω≤π 0, 0<Ω≤π
𝐻1 (𝑗 ( + )) = { Ω + 2π ={ Ω
𝑇1 T1 1− , −π ≤ Ω ≤ 0 − , −π ≤ Ω ≤ 0
2π 2π

1
Ω 2π
∑ 𝐻1 (j ( − k )) = 1, Ω ∈ [−𝜋, 𝜋]
𝑇1 T1
𝑘=−1
∞
Ω 2π
𝐼𝑛 𝑓𝑎𝑐𝑡: ∑ 𝐻1 (j ( − k )) = 1, ∀Ω
𝑇1 T1
𝑘=−∞

3
𝑌(𝑒 𝑗Ω ) =
𝑋(𝑒 𝑗Ω )
𝑇1
g) Based on part f, what is the equivalent filter 𝐻(𝑒 𝑗Ω ) = 𝑌(𝑒 𝑗Ω )/𝑋(𝑒 𝑗Ω ) corresponding to all
the D/C, C/D, and CT filtering operations?
3
𝐻(𝑒 𝑗Ω ) =
𝑇1
The output is a scaled version of the input. That is, the input is transmitted from one side to
the other side (through a physical link) without any distortion. It can be recovered at the
output.
h) Although sampling rate is lower than the Nyquist rate, it does not harm the application in this
case as we observe in part g. This is a common trick used in communications. Please check
π
whether the same outcome holds if the cutoff frequency of the second filter is set to 𝑇 .
1

2π π π
When it was 𝑇1
, clipped parts of 𝑌𝑐 (𝑗𝜔) in [− 𝑇 , + 𝑇 ) were being compensated with the parts
1 1
at both sides (see the figure in part c). It is equivalent to the fact that
∞
Ω 2π
∑ 𝐻1 (j ( − k )) = 1, ∀Ω
𝑇1 T1
𝑘=−∞

from part f. In this expression, 𝐻2 (𝑗𝜔) is omitted because it was ineffective in the nonzero
region of 𝐻1 (𝑗𝜔). Actually, the true expression to check is
∞
Ω 2π Ω 2π
∑ 𝐻1 (j ( − k )) 𝐻2 (j ( − k )) = 1, ∀Ω
𝑇1 T1 𝑇1 T1
𝑘=−∞
π
When the cutoff frequency of the second filter is set to 𝑇 , this expression is not valid
1
anymore. The parts at the both sides of the spectrum of 𝑋𝑐 (𝑗𝜔) are cancelled by the second
π π
filter, and clipped parts of 𝑌𝑐 (𝑗𝜔) in [− 𝑇 , + 𝑇 ) cannot be compensated. Therefore, the output
1 1
is not a scaled version of the input anymore, it is distorted now. That is, the input cannot be
recovered at the output and transmission failed.
Question 4

Define the following DT signal in Matlab.

W=pi/4;
N=32;
n=0:(N-1);
x_n=W/pi*sinc(W*(n-N/2)/pi);

a) Plot x_n vs n.

2𝜋
b) Calculate the FT of x_n for Ω = k ,𝑘 = 0, … ,127, with the following code fragment. Plot
128
|𝑋(𝑒 𝑗Ω )| for the defined Ω.

N2=4*N;
Omega_seq=(0:(N2-1))*(2*pi/N2);
X_Omega=0*Omega_seq; % allocate memory for DTFT
for kk=1:length(Omega_seq)
Omega=Omega_seq(kk);
X_Omega(kk)=exp(-j*Omega*(0:(N-1)))*(x_n).'; % summation
end
figure,plot(Omega_seq,abs(X_Omega),'+-')
c) DFT of a signal can be calculated as we derived in class by the fft function in Matlab. Read
the description of the fft function in Matlab. Use it to find the DFT of x_n.

Matlab fft( ):
𝑁
2𝜋
(𝑛−1)(𝑘−1)
𝑓𝑓𝑡(𝑥[𝑛]) = 𝑋𝐹𝐹𝑇 [𝑘] = ∑ 𝑥[𝑛]𝑒 −𝑗 𝑁 , 𝑘 = 1, … , 𝑁
𝑛=1
𝑁
2𝜋
𝑓𝑓𝑡(𝑥[𝑛 − 1]) = 𝑋𝐹𝐹𝑇−1 [𝑘] = ∑ 𝑥[𝑛 − 1]𝑒 −𝑗 𝑁 (𝑛−1)(𝑘−1)
𝑛=1

𝑁−1 𝑁
2𝜋 2𝜋
−𝑗 𝑘𝑛
𝑋𝐷𝐹𝑇 [k] = ∑ 𝑥[𝑛]𝑒 𝑁 = ∑ 𝑥[𝑛 − 1]𝑒 −𝑗 𝑁 𝑘(𝑛−1)
𝑛=0 𝑛=1

𝑋𝐷𝐹𝑇 [k] = 𝑋𝐹𝐹𝑇−1 [𝑘 + 1], 𝑘 = 0, … , 𝑁 − 1

d) When the DFT analysis formula is considered, DFT’s kth element corresponds to what
frequency?
𝑁−1
2𝜋
𝑋𝐷𝐹𝑇 [k] = ∑ 𝑥[𝑛]𝑒 −𝑗 𝑁 𝑘𝑛
𝑛=0
2𝜋
Ωk = 𝑘
𝑁
e) Draw the plot in part b and the absolute of DFT in part d vs Ω ∈ [−𝜋, 𝜋). Verify the same
values where you expect them to see.
X_fft_1=fft(x_n); %signal length is 32.
plot((0:31)*(2*pi/32),abs(X_fft_1),'o');
 f) The fft function can also give us a highly sampled DTFT. This is done by padding with
trailing zeros to the length given to the fft function as an argument when the argument is larger
than the DT signal’s original length. Plot fft(x_n,2*N), fft(x_n,4*N), fft(x_n,8*N) with the
corresponding Ω values.

g) There are many helpful functions in Matlab. One of them is the upsample function that inserts
zeros in between signal samples. Use the following code fragment and draw the resultant DFT
vs Ω ∈ [−𝜋, 𝜋). Is this related to any of the properties discussed in class?

usr=3;
x_n_up=upsample(x_n,usr);