numra-fft

FFT and spectral analysis for the Numra workspace — complex / real FFT, IFFT, 2D FFT, convolution, PSD, Welch, STFT, and windowing — built on rustfft.

Generic over Scalar (f32 / f64), with FFT computation internally in f64 for numerical stability. Covers the spectrum from raw FFT/IFFT through windowed power-spectral-density and short-time Fourier transforms.

Example

use numra_fft::{fft, ifft, Complex};

let signal: Vec<Complex<f64>> = vec![
    Complex::new(1.0_f64, 0.0),
    Complex::new(0.0, 0.0),
    Complex::new(0.0, 0.0),
    Complex::new(0.0, 0.0),
];
let spectrum = fft(&signal);
let recovered = ifft(&spectrum);

for (a, b) in signal.iter().zip(recovered.iter()) {
    assert!((a.re - b.re).abs() < 1e-12);
}

What's in this crate

Core FFT: fft, ifft, fft2 (complex 1D / 2D)
Real FFT: rfft, irfft (efficient real-to-complex)
Complex type: Complex, ComplexF64
Spectral analysis: psd (power spectral density), welch (Welch's method), stft / StftResult (short-time Fourier transform)
Convolution: fftconvolve, fftcorrelate
Utilities: fftfreq, fftshift, ifftshift, window_func, Window (Hann, Hamming, Blackman, Bartlett, …)

Composes with

numra-ode — spectral analysis of time-series ODE solutions
numra-signal — FFT backbone for resample, hilbert, and FIR windowed-sinc design (filtfilt itself is SOS-based IIR and stays in the time domain)
numra-stats — PSD treated as a frequency-domain statistic
numra-interp — resampling onto uniform grids before FFT

See interop workflows for the verified ODE → FFT → signal-processing → peak-detection workflow.

Install

[dependencies]
numra-fft = "0.1"

Or via the umbrella crate:

[dependencies]
numra = "0.1"

Documentation

API: https://docs.rs/numra-fft
Book: FFT and spectral
Source: https://github.com/moussaoutlook/numra-rs/tree/main/numra-fft

License

Numra Academic & Research License (Non-Commercial). Academic and research use is free; commercial use requires a separate license — contact contact@spectralautomata.com. See LICENSE.