https://github.com/klafyvel/avr-fft

3 implementations of the FFT for Arduino AVR platforms!
https://github.com/klafyvel/avr-fft

arduino arduino-uno avr avr-assembly avr-microcontroller cpp diy diy-electronics fft fft-algorithm

Last synced: 2 months ago
JSON representation

3 implementations of the FFT for Arduino AVR platforms!

• Host: GitHub
• URL: https://github.com/klafyvel/avr-fft
• Owner: Klafyvel
• License: mit
• Created: 2022-03-12T16:42:37.000Z (almost 3 years ago)
• Default Branch: main
• Last Pushed: 2022-11-16T12:45:15.000Z (about 2 years ago)
• Last Synced: 2024-10-13T04:45:48.343Z (3 months ago)
• Topics: arduino, arduino-uno, avr, avr-assembly, avr-microcontroller, cpp, diy, diy-electronics, fft, fft-algorithm
• Language: C++
• Homepage: https://klafyvel.me/blog/articles/fft-arduino/
• Size: 4.15 MB
• Stars: 141
• Watchers: 9
• Forks: 7
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Implementation of the FFT on Arduino

This is the companion repository for my blog post [here](http://klafyvel.me/blog/articles/fft-arduino). You'll find plenty of details there.

This directory contains several trials of implementing efficiently the Fast

Fourier Transform (FFT) on Arduino. As a comparison, I took [this implementation](https://www.instructables.com/ApproxFFT-Fastest-FFT-Function-for-Arduino/).

If you want to understand the basis of how to implement the FFT, I wrote a small

tutorial

[here](https://zestedesavoir.com/tutoriels/3939/jouons-a-implementer-une-transformee-de-fourier-rapide/)

(English version is [here](https://klafyvel.me/blog/articles/fft-julia/)).

Currently, here are the benchmarks on the following test signal, with a varying

number of points.

![Test signal](./test_signal.png)

![Benchmarks](./execution_time_comparison.svg)

![Benchmarks](./error_comparison.svg)

Below is a list of the different implementation.

## `ExactFFT`

This is a direct translation to C++ of the implementation in my tutorial.

![Result of FFT for different input sizes](result_plots/results_ExactFFT.svg)

## `ApproxFFT`

This is an implementation of the FFT by user

[abhilash_patel](https://www.instructables.com/member/abhilash_patel/) on

[Instructables](https://www.instructables.com/ApproxFFT-Fastest-FFT-Function-for-Arduino/).

It is there for comparison purpose only, as it is not very space-efficient.

![Result of FFT for different input sizes](result_plots/results_ApproxFFT.svg)

## `FloatFFT`

A floating-point implementation of the FFT, using some rewriting of

multiplications in `ExactFFT`, and some bit magic to accelerate float operations.

Featuring, but not limited to :

* A table of sines to remove any call to `sin()`;

* [A bit-magic implementation of the square root](https://en.wikipedia.org/wiki/Methods_of_computing_square_roots#Approximations_that_depend_on_the_floating_point_representation);

* A homemade implementation of the square operation for floats that I detailed

  [here](https://zestedesavoir.com/billets/4153/approximer-rapidement-le-carre-dun-nombre-flottant/)

  (in French);

* Some rewritings of complex multiplications of ExactFFT using [this method](https://en.wikipedia.org/wiki/Multiplication_algorithm#Complex_number_multiplication);

 

![Result of FFT for different input sizes](result_plots/results_FloatFFT.svg)

## `Fixed16FFT`

A 16-bits fixed-point implementation of the FFT on AVR.

![Result of FFT for different input sizes](result_plots/results_Fixed16FFT.svg)

## `Fixed8FFT`

An 8-bits fixed-point implementation of the FFT on AVR.

![Result of FFT for different input sizes](result_plots/results_Fixed8FFT.svg)