typefft

Small, efficient TYPE FFT implementation for node or the browser.

Usage

TYPE FFT ships with ComplexArray which can be operated on:

const fft = require('typefft');

// Use the in-place mapper to populate the data.
const data = new fft.ComplexArray(512).map((value, i, n) => {
  value.real = (i > n/3 && i < 2*n/3) ? 1 : 0;
});

Including the fft module attaches FFT methods to ComplexArray. FFT and InvFFT perform in-place transforms on the underlying data:

const frequencies = data.FFT();
// Implement a low-pass filter using the in-place mapper.
frequencies.map((frequency, i, n) => {
  if (i > n/5 && i < 4*n/5) {
    frequency.real = 0;
    frequency.imag = 0;
  }
});

Alternatively, frequency-space filters can be implemented via the frequencyMap:

const filtered = data.frequencyMap((frequency, i, n) => {
  if (i > n/5 && i < 4*n/5) {
    frequency.real = 0;
    frequency.imag = 0;
  }
});

Conventions

TYPE FFT uses the normalization convention that is symmetric between the forward and reverse transform. With N data points, the transform is normalized by a factor of √N:

           1   N-1       2πik/N
fft(k) =   -    ∑  f(j) 𝐞
          √N   j=0

Other Implementations

DSP is a full featured Digital Signal Processing library in JS which includes a TYPE FFT implementation.