Fast Fourier Transform

Advanced DSP

Dr. Muhammad Younis 2023-2024

FFT DSP-PhD 23-24

Fast Fourier Transform

Why?

+ +

FFT
FFT DSP-PhD 23-24

There are two formats of FFT which are :

1- The decimation in frequency algorithm.
2- The decimation in time algorithm.
They are referred to as the radix-2 FFT algorithms
N pt DFT

pt pt Base radix

pt pt pt pt
• Without loss of generality, we consider the sequence x(n)
consisting of samples, where m is a positive integer—the
number of samples of the digital sequence x(n) is a power of 2,
N = 2, 4, 8,16, etc. If x(n) does not contain samples, then we
simply append it with zeros until the number of the appended
sequence is equal to an integer of a power of 2 data points.
FFT DSP-PhD 23-24

The decimation in frequency algorithm

𝟐𝝅
𝑵 the twiddle factor N pt DFT

+ +
pt pt
FFT DSP-PhD 23-24

The decimation in frequency algorithm

𝟎
𝑵
𝟏
𝑵
𝟐
𝑵
𝟑
𝑵
FFT DSP-PhD 23-24

The decimation in frequency algorithm

Bit reversal

𝟎
𝟖
𝟐
𝟖
𝟎
𝟖
𝟏
𝟖
𝟐 𝟎
𝟖 𝟖
𝟑 𝟐
𝟖 𝟖
FFT DSP-PhD 23-24

Complex Multiplications
Complex Multiplications of DFT =
Complex Multiplications of FFT =
Example: What is the effectiveness of FFT, with a sequence of
1024 data point?
Solution:
Appling DFT will required
𝟐 complex multiplications
BUT
Appling FFT will required

𝟐 𝟐

𝟐
𝟏𝟎
𝟐 complex multiplications
FFT DSP-PhD 23-24

Example: Given x(n)={1, 2, 3, 4},

find X(k) using DIF-FFT algorithm and the complex multiplication?
Solution:
DIF-FFT

𝟎
𝑵

𝟏
𝑵

𝑾𝟑𝟒 = 𝒋
Complex multiplication
𝑾𝟐𝟒 = −𝟏 𝑾𝟎𝟒 = 𝟏
𝟐
𝟐 𝟐 𝟐 𝟐
𝑾𝟏𝟒 = −𝒋
FFT DSP-PhD 23-24

Example: Given for

find X(k) using DIF-FFT algorithm?
Solution:
DIF-FFT

𝑾𝟎𝟖 −𝟏
−𝟏 𝑾𝟐𝟖

𝑾𝟎𝟖 −𝟏 −𝟏
−𝟐. 𝟖𝟐𝟖
−𝟏 𝑾𝟏𝟖 + 𝟐. 𝟖𝟐𝟖𝒋

−𝟏 𝑾𝟐𝟖 𝑾𝟎𝟖 −𝟏

−𝟏 𝑾𝟑𝟖 −𝟏 𝑾𝟐𝟖
𝟐. 𝟖𝟐𝟖 −𝟏
−𝟏 + 𝟐. 𝟖𝟐𝟖𝒋 −𝟏
 FFT DSP-PhD 23-24

The decimation in time algorithm

𝑬𝒗𝒆𝒏

𝟎
𝑵
𝟏
𝑵
𝒐𝒅𝒅

𝟐
𝑵
𝟑
𝑵
FFT DSP-PhD 23-24

Example: Given x(n)={1, 2, 3, 4},

find X(k) using TIF-FFT algorithm and the complex multiplication?
Solution:

TIF-FFT

−𝟏
𝑾𝟎𝟒 = 𝟏

−𝟏
𝑾𝟏𝟒 = −𝒋

−𝟏 −𝟏
FFT DSP-PhD 23-24

Example: Given for

find X(k) using DIT-FFT algorithm?
𝑾𝟎𝟖 = 𝟏
Solution: 𝑾𝟒𝟖 = −𝟎. 𝟕𝟎7
−𝒋. 𝟎𝟕𝟎𝟕
DIF-FFT 𝑾𝟐𝟖 = −𝒋

𝑾𝟎𝟖

−𝟏

𝑾𝟐𝟖 −𝟏

−𝟏 −𝟏 𝑾𝟎𝟖

𝑾𝟎𝟖 𝑾𝟏𝟖 −𝟏

−𝟏 𝑾𝟐𝟖 −𝟏

𝑾𝟐𝟖 −𝟏 𝑾𝟑𝟖 −𝟏

−𝟏 −𝟏
−𝟏
FFT DSP-PhD 23-24

Inverse FFT

Comparing the above IDFT Equation with DFT Equation, we notice the
difference as follows:
• The twiddle factor 𝑵 is changed to be 𝑵 𝑵
𝟏 pt pt
• The sum is multiplied by a factor of / .
Hence, by modifying the FFT block diagram as shown in previous Figure, we
achieve the IFFT block diagram shown in Figure below 𝟏
𝟖
𝟏
𝟖
𝟏
𝑾𝟎𝑵 −𝟏 𝟖
−𝟏 𝑾𝟐 𝟏
𝑵 𝟖
−𝟏 𝟏
𝑾𝟎𝑵 −𝟏 𝟖
−𝟏 𝟏
𝑾𝟏𝑵 𝟖
−𝟏 𝟏
−𝟏 𝑾𝟐𝑵 𝑾𝟎𝑵 𝟖
−𝟏 𝑾𝟑𝑵 −𝟏 𝑾𝟐 𝟏
𝑵 𝟖
−𝟏 −𝟏 −𝟏
FFT DSP-PhD 23-24

Example: Given
, find x(n)
Solution:

−𝟏
𝑾𝟎𝟒

−𝟏
𝑾𝟏𝟒
−𝟏
−𝟏
FFT DSP-PhD 23-24

Zero Padding
• Most implementations of the FFT require that the length of x(n) be an
integer power of 2 (i.e., 4, 8, 16, 32, …).
• What if x(n) is not an integer power of 2 in length? No problem! Just append
zeros to x(n) until a power of 2 is reached.
• This is called zero padding, and can be used to increase the FFT frequency
resolution even though the number of nonzero samples is limited
• It is evident that increasing the data length via zero padding to compute
the signal spectrum does not add basic information and does not change the
spectral shape but gives the ‘‘interpolated spectrum’’ with the reduced
frequency spacing . We can get a better view of the two spectral peaks
described in this case as shown: