Crate microfft

source ·
Expand description

Library for computing fast fourier transforms on embedded systems.

microfft provides an in-place implementation of the Radix-2 FFT algorithm. All computations are performed directly on the input buffer and require no additional allocations. This makes microfft suitable for no_std environments.

This crate provides three FFT implementations:

  • complex: FFT on Complex32 input values (CFFT).
  • real: FFT on real (f32) input values (RFFT). An N-point RFFT internally computes an N/2-point CFFT, making it roughly twice as fast as the complex variant.
  • inverse: Inverse FFT (IFFT), implemented in terms of a CFFT.

§Example

use std::convert::TryInto;
use std::f32::consts::PI;

// generate 16 samples of a sine wave at frequency 3
let sample_count = 16;
let signal_freq = 3.;
let sample_interval = 1. / sample_count as f32;
let mut samples: Vec<_> = (0..sample_count)
    .map(|i| (2. * PI * signal_freq * sample_interval * i as f32).sin())
    .collect();

// compute the RFFT of the samples
let mut samples: [_; 16] = samples.try_into().unwrap();
let spectrum = microfft::real::rfft_16(&mut samples);
// since the real-valued coefficient at the Nyquist frequency is packed into the
// imaginary part of the DC bin, it must be cleared before computing the amplitudes
spectrum[0].im = 0.0;

// the spectrum has a spike at index `signal_freq`
let amplitudes: Vec<_> = spectrum.iter().map(|c| c.norm() as u32).collect();
assert_eq!(&amplitudes, &[0, 0, 0, 8, 0, 0, 0, 0]);

Modules§

Type Aliases§