INS 3144 Digital Signal Processing

Lecture 07: Discrete Fourier

Transform (Topic 5)

Lecturer: Dr Ngoc Thanh Pham.

Faculty of Engineering and Technology, International School, Vietnam National.
University Hanoi.
DISCRETE FOURIER TRANSFORM
❖ For time domain “t”, digital signals: the
signal amplitude and the sampling time
instant (sample number).
❖ Signal frequency content: very useful
than the digital signal samples.
❖ Digital signal in terms of its frequency
component in a frequency domain
(signal spectrum), needs to be developed.
❖ Top: signal samples of a continuous
1000 Hz at a sampling rate of 8000 Hz
❖ Bottom: magnitude of signal spectrum.
Amplitude peak at 1000 Hz Digital signal and its amplitude spectrum.

Dr Ngoc Thanh Pham VNUIS 2

DISCRETE FOURIER TRANSFORM
❖ Discrete Fourier transform: algorithm -
the time domain signal samples to the
frequency domain components.
❖ establishes a relationship between the
time domain representation and the
frequency domain representation.
❖ DFT: to perform frequency analysis of a
time domain sequence
❖ DFT: spectral analysis, acoustics,
imaging/video, audio, instrumentation,
and communications systems.
❖ To develop DFT, study the spectrum of
periodic digital signals using the Fourier
Digital signal and its amplitude spectrum.
series.

Dr Ngoc Thanh Pham VNUIS 3

FOURIER SERIES COEFFICIENTS OF PERIODIC DIGITAL SIGNALS

❖ Estimate the spectrum of a

periodic digital signal 𝑥 𝑛 ,
sampled at a rate of 𝑓𝑠 Hz ,
fundamental period 𝑇0 = 𝑁𝑇.
❖ 𝑁 samples within the duration
of the fundamental period and
𝑇 = 1/𝑓𝑠 is the sampling
period
❖ Assume : all harmonic
frequencies < the folding
frequency (𝑓𝑠 /2) → aliasing
does not occur
Periodic digital signal.

Dr Ngoc Thanh Pham VNUIS 4

FOURIER SERIES (LAST YEAR)
❖ A periodic signal x(t) with a period of T, ❖ Complex exponential form is
∞
❖ 𝝎𝟎 = 𝟐𝝅/𝑻𝟎 is the fundamental angular
𝑥(𝑡) = ෍ 𝑐𝑘 𝑒 𝑗𝑘𝜔0 𝑡 ,
frequency (radians / second).
𝑘=−∞
❖ The fundamental frequency: 𝑓0 = where 𝑐𝑘 represents the complex Fourier
1/𝑇0 (Hz). coefficients, which can be found from
❖ The Fourier series expansion
1
x 𝑡 = 𝑎0 + σ∞ 𝑘=1 𝑎𝑘 cos 𝑘𝜔0 𝑡 𝑐𝑘 = න 𝑥(𝑡)𝑒 −𝑗𝑘𝜔0 𝑡 𝑑𝑡
𝑇0 𝑇0
+ σ∞ 𝑘=1 𝑏𝑘 sin 𝑘𝜔0 𝑡
1
𝑎0 = ‫׬‬ 𝑥 𝑡 𝑑𝑡. Euler's formula 𝑒 ±𝑗𝑥 = cos(𝑥) ± 𝑗sin(𝑥),
𝑇0 𝑇0
2
𝑎𝑘 = ‫׬‬ 𝑥 𝑡 cos 𝑘𝜔0 𝑡 𝑑𝑡. 𝑒 𝑗𝑥 +𝑒 −𝑗𝑥 𝑒 𝑗𝑥 −𝑒 −𝑗𝑥
𝑇0 𝑇0 cos(𝑥) = , sin(𝑥) =
2 2𝑗
2
𝑏𝑘 = ‫׬‬ 𝑥(𝑡)sin 𝑘𝜔0 𝑡 𝑑𝑡.
𝑇0 𝑇0

Dr Ngoc Thanh Pham VNUIS 5

FOURIER SERIES COEFFICIENTS OF PERIODIC DIGITAL SIGNALS

❖ Estimate the spectrum of a

periodic digital signal 𝑥 𝑛 ,
sampled at a rate of 𝑓𝑠 Hz ,
fundamental period 𝑇0 = 𝑁𝑇.
❖ 𝑁 samples within the duration
of the fundamental period and
𝑇 = 1/𝑓𝑠 is the sampling
period
❖ Assume : all harmonic
frequencies < the folding
frequency (𝑓𝑠 /2) → aliasing
does not occur
Periodic digital signal.

Dr Ngoc Thanh Pham VNUIS 6

FOURIER SERIES (LAST YEAR)
❖ coefficients of the Fourier series expansion of the ❖ Complex exponential form
periodic signal 𝑥(𝑡) in a complex form is (remember)
∞
1 𝑥(𝑡) = ෍ 𝑐𝑘 𝑒 𝑗𝑘𝜔0 𝑡 ,
𝑐𝑘 = න 𝑥(𝑡)𝑒 −𝑗𝑘𝜔0 𝑡 𝑑𝑡, −∞ < 𝑘 < ∞,
𝑇0 𝑇0 𝑘=−∞
1
❖ we substitute 𝑇0 = 𝑁𝑇, 𝜔0 = 2𝜋/𝑇0 and 𝑐𝑘 = න 𝑥(𝑡)𝑒 −𝑗𝑘𝜔0 𝑡 𝑑𝑡
𝑇0 𝑇0
approximate the integration over one period
using a summation by substituting 𝑑𝑡 = 𝑇 ❖ Our example: digital signal 𝑥 𝑛 ,
and 𝑡 = 𝑛𝑇. sampled at a rate of 𝑓𝑠 Hz ,
𝑁−1 fundamental period 𝑇0 = 𝑁𝑇.
1
𝑐𝑘 = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁 , − ∞ < 𝑘 < ∞
𝑁
𝑛=0

Dr Ngoc Thanh Pham VNUIS 7

FOURIER SERIES COEFFICIENTS OF PERIODIC DIGITAL SIGNALS
𝑁−1
1 𝑒 −𝑗2𝜋𝑛 = cos(2𝜋𝑛) − 𝑗sin(2𝜋𝑛) = 1, it
𝑐𝑘 = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁 , − ∞ < 𝑘 < ∞
𝑁 follows that 𝒄𝒌+𝑵 = 𝒄𝒌 (periodic of 𝑵)
𝑛=0
❖ Since the coefficients 𝑐𝑘 are obtained the two-sided line amplitude spectrum
from the Fourier series expansion in the 𝒄𝒌 is periodic
complex form, the resultant spectrum 𝒄𝒌
will have two sides.
❖ 𝒄𝒌 : periodic of 𝑵
𝑁−1
1
𝑐𝑘+𝑁 = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋(𝒌+𝑵)𝑛/𝑁
𝑁
𝑛=0
𝑁−1
1
𝑐𝑘+𝑁 = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁 𝑒 −𝒋𝟐𝝅𝒏 Amplitude spectrum of the periodic digital
𝑁
𝑛=0 signal.
Dr Ngoc Thanh Pham VNUIS 8
FOURIER SERIES COEFFICIENTS OF PERIODIC DIGITAL SIGNALS

❖ only the line spectral portion between the

frequencies 𝑓𝑠 /2 and 𝑓𝑠 /2 (folding
frequency) represents frequency
information of the periodic signal.
❖ Spectral portion from 𝒇𝒔 /𝟐 to 𝒇𝒔 is a
copy of the spectrum in the negative
frequency range from −𝑓𝑠 /2 to 0 Hz due
to the spectrum being periodic for
every 𝑵𝒇𝟎 𝐇𝐳. Amplitude spectrum of periodic digital signal.
❖ Again, the amplitude spectral match the amplitude spectral components
components indexed from 𝑓𝑠 /2 to 𝑓𝑠 can indexed from 0 to 𝑓𝑠 /2 in terms of 𝑓𝑠 − 𝑓 Hz,
be folded at the folding frequency 𝑓𝑠 /2 to where 𝑓 is in the range from 𝑓𝑠 /2 to 𝑓𝑠 .

Dr Ngoc Thanh Pham VNUIS 9

FOURIER SERIES COEFFICIENTS OF PERIODIC DIGITAL SIGNALS

❖ For convenience, we compute the

spectrum over the range from 0 to 𝑓𝑠 Hz
with nonnegative indices, that is,
𝑁−1
1
𝑐𝑘 = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁 , 𝑘
𝑁
𝑛=0
= 0,1, … , 𝑁 − 1.
❖ We can apply 𝒄𝒌+𝑵 = 𝒄𝒌 to find the Amplitude spectrum of periodic digital signal.
negative indexed spectral values if they
❖ The frequency spacing between the
are required.
consecutive spectral lines, called the
❖ For the 𝑘 th harmonic, the frequency is frequency resolution, is 𝑓0 Hz.
𝑓 = 𝑘𝑓0 Hz

Dr Ngoc Thanh Pham VNUIS 10

FOURIER SERIES COEFFICIENTS OF PERIODIC DIGITAL SIGNALS

Example 𝟓. 𝟏. The periodic signal x t = ❖ sampling interval T = 1/𝑓𝑠 = 0.25 s.

sin 2πt is sampled using the sampling ❖ sampled signal as
𝑥(𝑛) = 𝑥(𝑛𝑇) = sin(2πn𝑇)
rate 𝑓𝑠 = 4 Hz.
= 𝑠𝑖𝑛(0.5πn).
❖ Compute the spectrum 𝑐𝑘 using the
samples in one period.
❖ Plot the two-sided amplitude spectrum
𝑐𝑘 over the range from -2 to 2 Hz.

Solution 5.1: the fundamental frequency

ω 2π
ω0 = 2π rad/s and 𝑓0 = 0 = = 1 Hz.
2π 2π

❖ fundamental period 𝑻𝟎 = 1 s

Dr Ngoc Thanh Pham VNUIS 11

FOURIER SERIES COEFFICIENTS OF PERIODIC DIGITAL SIGNALS

𝑁−1
❖ Choosing the duration of one period, 𝑁 = 1
4, we have the sample values: 𝑥 0 = 0, 𝑐𝑘 = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁 , 𝑘
𝑁
𝑥(1) = 1; 𝑥(2) = 0; 𝑥(3) = −1. 𝑛=0
𝟏 𝟑
= 0,1, … , 𝑁 − 1.
𝒄𝟎 = σ 𝒙(𝒏) . 𝒆−𝒋𝟐𝛑 ×𝟎𝒏/𝟒
𝟒 𝒏=𝟎 Euler's formula 𝑒 ±𝑗𝑥 = cos(𝑥) ± 𝑗sin(𝑥),
1
= 𝑥 0 + 𝑥 1 + 𝑥 2 + 𝑥 3 𝑒 𝑗𝑥 +𝑒 −𝑗𝑥
4 𝑒 𝑗𝑥 −𝑒 −𝑗𝑥
cos(𝑥) = , sin(𝑥) =
1 2 2𝑗
= 0 + 1 + 0 − 1 = 0.
4
𝟏 𝟑 1
❖ 𝒄𝟏 = σ 𝒙(𝒏) . 𝒆−𝒋𝟐𝛑 × 𝟏𝒏/𝟒 = (x(0) − jx (1) − x(2) + jx(3))
𝟒 𝒏=𝟎 4
1 1
(𝑥(0) + 𝑥 (1) . 𝑒 −𝑗π/2 + 𝑥(2) . 𝑒 −𝑗π + = (0 – j(1)−0 +j(−1))=−0.5j.
4
4
𝑥(3) . 𝑒 −3𝑗π/2 ) 𝟏 𝟑
❖ 𝒄𝟐 = σ𝒏 = 𝟎 𝒙(𝒏) . 𝒆−𝒋𝟐𝛑× 𝟐𝒏/𝟒
𝟒

Dr Ngoc Thanh Pham VNUIS 12

FOURIER SERIES COEFFICIENTS OF PERIODIC DIGITAL SIGNALS
1ቀ 𝑗𝜋
−2
= 𝑥 0 + 𝑥 1 .𝑒 + 𝑥 2 . 𝑒 −2𝑗𝜋
4
6𝑗𝜋
− 2 ቁ
+ 𝑥 3 .𝑒 =𝟎
𝟏 𝟑
❖ 𝒄𝟑 = σ 𝒙(𝒏) . 𝒆−𝒋𝟐𝛑 𝒙 𝟒𝒏/𝟒 = 𝟎. 𝟓𝒋
𝟒 𝒏=𝟎
Using periodicity, it follows that:
𝑐−1 = 𝑐3 = 𝑗0.5 𝑎𝑛𝑑 𝑐−2 = 𝑐2 = 0.
Euler's formula 𝑒 ±𝑗𝑥 = cos(𝑥) ± 𝑗sin(𝑥),
The spectrum in the range of –2 to 2Hz
𝑒 𝑗𝑥 +𝑒 −𝑗𝑥 𝑒 𝑗𝑥 −𝑒 −𝑗𝑥
presents the information of the sinusoid with cos(𝑥) = , sin(𝑥) =
2 2𝑗
a frequency of 1Hz and a peak value of
|𝑐1 | = 1. which is converted from two sided Note that we do not double the direct-current
to one sided by doubling the spectral value. (DC) component, that is, 𝑐0 .

Dr Ngoc Thanh Pham VNUIS 13

DISCRETE FOURIER TRANSFORM FORMULAS

❖ Assume that the process acquires data

samples from digitizing the interested
continuous signal for a duration of 𝑻𝟎
seconds.
❖Assume that a periodic signal 𝑥(𝑛) is
obtained by copying the acquired 𝑵
data samples with the duration of 𝑇0 to
itself
❖ Note that we assume continuity between
the N data sample frames. This is not
true in practice

Dr Ngoc Thanh Pham VNUIS 14

DISCRETE FOURIER TRANSFORM FORMULAS

❖ Determine the Fourier series coefficients using

one-period 𝑁 data samples, then multiply the
Fourier series coefficients by a factor of 𝑁
𝑁−1

𝑋(𝑘) = 𝑁𝑐𝑘 = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁 , for 𝑘

𝑛=0
= 0,1, … , 𝑁 − 1,
❖ 𝑋(𝑘) constitutes the DFT coefficients.
❖ the factor of 𝑁 is a constant and does not
affect the relative magnitudes of the DFT
coefficients 𝑋(𝑘).

Dr Ngoc Thanh Pham VNUIS 15

DISCRETE FOURIER TRANSFORM FORMULAS

❖ applying DFT with 𝑁 data samples of 𝑥(𝑛)

sampled at a sampling rate of 𝑓𝑠 (sampling
period is 𝑇 = 1/𝑓𝑠 ) produces 𝑁 complex DFT
coefficients 𝑋(𝑘).
❖ The index 𝑛 is the time index representing the
sample number of the digital sequence,
❖ 𝑘 is the frequency index indicating each
calculated DFT coefficient, and can be further
mapped to the corresponding signal frequency
in terms of Hz.

Dr Ngoc Thanh Pham VNUIS 16

DISCRETE FOURIER TRANSFORM FORMULAS
DISCRETE FOURIER TRANSFORM
𝑋 𝑘 = 𝑥 0 𝑊𝑁𝑘0 + 𝑥 1 𝑊𝑁𝑘1 + 𝑥 2 𝑊𝑁𝑘2
Given a sequence 𝑥(𝑛), 0 ≤ 𝑛 ≤ 𝑁 − 1, its 𝑘(𝑁−1)
DFT is defined as + ⋯ + 𝑥(𝑁 − 1)𝑊𝑁 , for 𝑘
𝑁−1 = 0,1, ⋯ , 𝑁 − 1,
𝑋(𝑘) = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁 .
𝑛=0
The inverse of DFT is given by
𝑁−1 𝑁−1
= ෍ 𝑥(𝑛)𝑊𝑁𝑘𝑛 , for 𝑘 = 0,1, ⋯ , 𝑁 − 1. 1
𝑥(𝑛) = ෍ 𝑋(𝑘)𝑒 𝑗2𝜋𝑘𝑛/𝑁
𝑛=0 𝑁
𝑘=0
where the factor 𝑊𝑁 is defined as
𝑁−1
1
−𝑗2𝜋/𝑁
2𝜋 2𝜋 = ෍ 𝑋(𝑘)𝑊𝑁−𝑘𝑛 , for 𝑛
𝑊𝑁 = 𝑒 = cos − 𝑗sin . 𝑁
𝑁 𝑁 𝑘=0
= 0,1, ⋯ , 𝑁 − 1
𝑋(𝑘)can be expanded as
Dr Ngoc Thanh Pham VNUIS 17
DISCRETE FOURIER TRANSFORM FORMULAS
Example 5.2. Given a sequence x(n) for Solution to Example 5.2
0 ≤ 𝑛 ≤ 3, where x(0) = 1, x(1) = 2, x(2) = ❖ one-period 4 data samples N = 4
π
−𝑗 2
3, and x(3) =4, evaluate its DFT X(k) ❖ DFT factor 𝑊4 = 𝑒
3 3
π𝑘𝑛
−𝑗
𝑁−1 𝑋𝑘 = ෍ 𝑥(𝑛) . 𝑊4𝑘𝑛 = ෍ 𝑥(𝑛) . 𝑒 2
𝑋(𝑘) = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁 𝑛=0 𝑛=0
𝑛=0
2𝜋 2𝜋 for k =0, 𝑿𝟎 = σ𝟑𝒏 = 𝟎 𝒙(𝒏) . 𝒆−𝒋𝟎
𝑊𝑁 = 𝑒 −𝑗2𝜋/𝑁 = cos − 𝑗sin .
𝑁 𝑁 = 𝑥 0 . 𝑒 −𝑗0 + 𝑥 1 . 𝑒 −𝑗π0 + 𝑥 2 . 𝑒 −𝑗0 +
𝑥(3) . 𝑒 −𝑗0 = 𝑥(0) + 𝑥(1) + 𝑥(2) + 𝑥(3)
Euler's formula 𝑒 ±𝑗𝑥 = cos(𝑥) ± 𝑗sin(𝑥),
= 1 + 2 + 3 + 4 = 10.
𝛑𝒏
𝑒 𝑗𝑥 +𝑒 −𝑗𝑥 𝑒 𝑗𝑥 −𝑒 −𝑗𝑥 −𝒋 𝟐
cos(𝑥) = , sin(𝑥) = ❖for k = 1, 𝑿𝟏 = σ𝟑𝒏 = 𝟎 𝒙(𝒏) .𝒆
2 2𝑗

Dr Ngoc Thanh Pham VNUIS 18

DISCRETE FOURIER TRANSFORM FORMULAS
Solution to Example 5.2
π
−𝑗 2 𝑁−1
−𝑗0 −𝑗π
= x (0) .𝑒 + x(1) .𝑒 + x(2) .𝑒 + x(3)
3π 𝑋(𝑘) = ෍ 𝑥(𝑛)𝑒 −𝑗2𝜋𝑘𝑛/𝑁
−𝑗
.𝑒 = 𝑥(0) − 𝑗𝑥(1) − 𝑥(2) + 𝑗𝑥(3)
2 𝑛=0
= 1 − 𝑗2 + 3 + 𝑗4 = −2 + 2𝑗. Euler's formula 𝑒 ±𝑗𝑥 = cos(𝑥) ± 𝑗sin(𝑥),
for k = 2: 𝑿𝟐 = σ𝟑𝒏 = 𝟎 𝒙(𝒏) . 𝒆−𝒋𝝅𝒏
𝑒 𝑗𝑥 +𝑒 −𝑗𝑥 𝑒 𝑗𝑥 −𝑒 −𝑗𝑥
= x (0) .𝑒 −𝑗0 + x(1) .𝑒 −𝑗π + x(2) .𝑒 −𝑗2π + cos(𝑥) = , sin(𝑥) =
2 2𝑗
x(3) .𝑒 −𝑗3π
= 1 + j2 - 3 - j4 = -2 - 2j.
= x(0) − x(1) + x(2) − x(3)
= 1 − 2 + 3 − 4 = −2 Let us verify the result using the MATLAB
𝟑𝛑𝒏
function fft(): >> X=fft([1 2 3 4])
for k = 3: 𝑿𝟑 = σ𝟑𝒏 = 𝟎 𝒙(𝒏) . 𝒆−𝒋 𝟐
3π X = 10.0000-2.0000 + 2.0000i – 2.0000 –
= x (0) .𝑒 −𝑗0 + x(1) .𝑒 −𝑗 2 + x(2) .𝑒 −𝑗3π +
9π 2.0000 -2.0000i
x(3) .𝑒 −𝑗 2 = x(0) + jx(1) - x(2) - jx(3)
Dr Ngoc Thanh Pham VNUIS 19
DISCRETE FOURIER TRANSFORM FORMULAS
Example 5.3. Using the DFT coefficients 𝑁−1
1
X(k) for 0 ≤ 𝑘 ≤ 3 computed in Example 𝑥(𝑛) = ෍ 𝑋(𝑘)𝑒 𝑗2𝜋𝑘𝑛/𝑁
𝑁
4.2, evaluate its inverse DFT to determine 𝑘=0
𝑁−1
the time domain sequence x(n). 1
= ෍ 𝑋(𝑘)𝑊𝑁−𝑘𝑛 , for 𝑛 = 0,1, ⋯ , 𝑁 − 1
𝑁
Solution to Example 5.3. 𝑘=0
𝜋
Since N = 4 and 𝑊4−1 = 𝑒𝑗 2 . we achieve a 1 3
σ𝑘 = 0 𝑋(𝑘) . 𝑒 −𝑗0
for n = 0, x(0) =
4
simplified formula: 1
1 3
x(0) = (X(0) 𝑒 𝑗0 + X (1). 𝑒 𝑗0 + X(2). 𝑒 𝑗0
4
𝑋𝑘 = σ 𝑋(𝑘) . 𝑊𝑁−𝑘𝑛
4 𝑛=0 1
+ X(3) 𝑒 𝑗0 ) =
10 + (−2 + j2 − 2 +
π𝑘𝑛 4
1 𝑗 2
𝑋𝑘 = σ3𝑛 = 0 𝑋(𝑘) . 𝑒 (−2 − 𝑗2) = 1.
4

Dr Ngoc Thanh Pham VNUIS 20

DISCRETE FOURIER TRANSFORM FORMULAS
Solution to Example 5.3. (cont) Applying the MATLAB function ifft() >> x
𝑘𝜋
1
for n = 1,x(1) = σ3𝑘 = 0 𝑋(𝑘) . 𝑒 𝑗 2 = ifft ( 10 − 2 + 2𝑗 − 2 − 2 − 2𝑗
4
1൫ 𝜋 x=1234
𝑗
= 𝑋 0 𝑒 𝑗0 + 𝑋 1 .𝑒 2 + 𝑋 2 . 𝑒 𝑗𝜋 +
4 1
𝑗
3𝜋 = (10 +(-2 +j2) + (-2) – (-2 -j2)) = 3
𝑋 3 𝑒 ൯ 2 4
1
= 𝑋(0) + 𝑗𝑋(1) − 𝑋(2) − 𝑗𝑋(3) = 2.
for n =3,
4 3𝑘𝜋
1 1 3 1
for n =2,x(2) = σ3𝑘 = 0 𝑋(𝑘) . 𝑒 𝑗𝑘𝜋 x(3) = σ𝑘 = 0 𝑋(𝑘) . 𝑒𝑗 2 = (X(0)
4 4 4
3𝜋 9𝜋
𝑗2
1
= ൫𝑋 0 𝑒 𝑗0 + 𝑋 1 . 𝑒 𝑗𝜋 + 𝑋 2 . 𝑒 𝑗2𝜋
𝑗0
𝑒 + X (1). 𝑒 + X(2). 𝑒 𝑗3𝜋
+ X(3) 𝑒 𝑗 2 )
4 1
+ 𝑋 3 𝑒 𝑗3𝜋 ൯ = 𝑋 0 − 𝑗𝑋 1 − 𝑋 2 + 𝑗𝑋 3
4
1 1
= 𝑋 0 −𝑋 1 +𝑋 2 −𝑋 3 = (10 - j(-2 +j2) - (-2) + j(-2 -j2)) =4.
4 4

Dr Ngoc Thanh Pham VNUIS 21

DISCRETE FOURIER TRANSFORM FORMULAS
Example 5.4. In Example 6.2, given a (b) Determine the frequency resolution,
sequence x(n) for 0 ≤ 𝑛 ≤ 3, where 𝑥(0) = frequency bin, and mapped frequencies
1, 𝑥(1) = 2, 𝑥(2) = 3, 𝑎𝑛𝑑 𝑥(3) = 4, for each of the DFT coefficients X(1) and
❖ we have computed four DFT coefficients X(3) in frequency domain
X(k) for 0 ≤ 𝑘 ≤ 3 as
Solution to 5.4.
𝑋 0 = 10, 𝑋 1 = −2 + 𝑗2,
(a)In time domain, we have the sampling
𝑋(2) = − 2, 𝑎𝑛𝑑 𝑋(3) = −2 − 𝑗2.
period calculated as:
❖ If the sampling rate is 10 Hz,
1 1
(a) Determine the sampling period, time T= = = 0.1𝑠
𝑓𝑠 10
index, and sampling time instant for a digital
sample x(3) in time domain.

Dr Ngoc Thanh Pham VNUIS 22

DISCRETE FOURIER TRANSFORM FORMULAS
Solution to 5.4. (cont) ❖ The frequency bin for X(1) should be
❖ For data x(3), the time index is n =3 and 𝑘 = 1 and its corresponding frequency
the sampling time instant is determined by is determined by:
𝑘𝑓𝑠 1 𝑥 10
𝑡 = 𝑛𝑇 = 3 × 0.1 = 0.3𝑠. f= = = 2.5𝐻𝑧
𝑁 4
❖ In frequency domain, since the total ❖ Similarly, for X(3) and k =3:
number of DFT coefficients is four, the 𝑘𝑓𝑠 3 𝑥 10
frequency resolution is determined by: f= = = 7.5𝐻𝑧
𝑁 4
𝑓𝑠 10
Δf = = = 2.5 𝐻𝑧
𝑁 4

Dr Ngoc Thanh Pham VNUIS 23