Window Functions API

Signal window functions for spectral analysis.

Overview

Window functions reduce spectral leakage when performing FFT on finite-length signals.

All generator functions require size to be a positive integer. A size of 1 returns [1] so single-sample inputs remain finite and unchanged.

Window Function Generators

All window functions accept a size parameter and return a Float32Array.

hannWindow()

function hannWindow(size: number): Float32Array;

The Hann window provides good frequency resolution with moderate sidelobe suppression.

hammingWindow()

function hammingWindow(size: number): Float32Array;

The Hamming window offers better sidelobe suppression than Hann at the cost of wider mainlobe.

blackmanWindow()

function blackmanWindow(size: number): Float32Array;

The Blackman window provides excellent sidelobe suppression for high-dynamic-range applications.

flatTopWindow()

function flatTopWindow(size: number): Float32Array;

The Flat Top window minimizes amplitude error for precise amplitude measurements.

rectangularWindow()

function rectangularWindow(size: number): Float32Array;

No windowing (equivalent to no window). Best for transient signals.

applyWindow()

Applies a window function to real-valued data.

function applyWindow(
  data: Float32Array,
  window: Float32Array
): Float32Array;

data.length and window.length must match. A mismatch throws RangeError instead of silently dropping or zeroing samples.

Example

import { hannWindow, applyWindow } from 'webgpu-fft';

const signal = new Float32Array(1024);
// ... fill signal ...

const window = hannWindow(1024);
const windowed = applyWindow(signal, window);

applyWindowComplex()

Applies a window function to interleaved complex data.

function applyWindowComplex(
  data: Float32Array,
  window: Float32Array
): Float32Array;

data must contain complete interleaved real/imaginary pairs, and window.length must equal data.length / 2. Invalid shapes throw RangeError.

Comparison

Window	Mainlobe Width	Sidelobe Level	Best For
Rectangular	Narrowest	-13 dB	Transients
Hann	2 bins	-31 dB	General purpose
Hamming	2 bins	-41 dB	Narrowband signals
Blackman	3 bins	-58 dB	High dynamic range
Flat Top	4 bins	-90 dB	Amplitude accuracy

Related

• Spectrum Analyzer — Uses window functions internally
• Spectrum Analysis Tutorial — Practical usage