autofft - A time-frequency analyser for Matlab

autofft is a package of MATLAB® functions for automated frequency analysis of discrete-time signals, inspired by the operation of Brüel & Kjær FFT analysers. It automatically filters, segments, weights and averages input signals, returning spectra in a wide range of engineering units (dB, RMS, peak-to-peak, PSD, etc.). It works without the Signal Processing Toolbox™ and is optimised for performance.

Introduction

Signal Processing Toolbox™ provides several functions for power spectrum estimation, including pspectrum, pwelch and stft. While these functions are high-quality and well-documented, they can be cumbersome for certain engineering applications. These applications include estimating magnitudes of vibration, noise and other discrete-time signals in engineering units or comparing theoretical results with measurements.

The autofft package is designed to address these needs by mimicking the operation of Brüel & Kjær FFT analysers. Based on user input, autofft automatically filters the signal, divides it into segments, applies window functions, and performs spectral averaging. The resulting averaged spectrum, also called modified periodogram, can be returned in various engineering spectral units, including decibels, root mean square (RMS) and peak-to-peak magnitudes, and power spectral density (PSD). In addition, autofft can compute spectral derivatives and integrals, as well as perform short-time Fourier transform (STFT) analysis.

Key features

• No toolbox dependency: works without Signal Processing Toolbox™.
• High performance: performs significantly better than pwelch and pspectrum in computationally intensive problems.
• Works with engineering units: works not only with relative units (dB) but also with engineering units (EU), such as RMS, peak-to-peak or power spectral density.
• Configurable analyser setup: provides greater control over the setup of the frequency analyser than built-in functions.
• Built-in filtering: can apply high-pass filters in the time domain.
• Advanced post-processing: can apply frequency weighting filters per IEC 61672-1:2013, spectral derivation or spectral integration.

Figure: CPU times required to run autoFFT and similar Matlab functions measured on a Code Ocean using timeit function, see this reproducible capsule. Each run includes a predefined number of segments weighted using a 4096-point Hann window.

What's new?

• v1.5.5: Bug fix: Uncommented text preventing use of custom filters has been removed.
• v1.5.5: Bug fix: Erroneous coefficient that caused bandstop filters to be generated incorrectly using autoButter function has been corrected.
• v1.5.5: Code optimisation: Generation of windows with a very large number of samples has been optimised.
• v1.5.4: Bug fix: Automatic plotter no longer uses white backgroud color in R2025a dark mode.
• v1.5.4: Bug fix: Plotting error when the user selected tiled layout for time-frequency analysis results from only one channel has been fixed.
• v1.5.4: Code optimisation: Code optimisation: Evaluation of magnitude and spectral unit optimised. Other minor optimisations implemented and code refactored. Implemented changes reduced CPU time by up to 35 %.
• v1.5.4: Code optimisation: Error handling during filtering has been improved.
• v1.5.4: Documentation: Introduction has been revised.

Getting started

1. Download the latest release from the Releases page, extract it and add the folder to your MATLAB path: addpath('path_to_autofft_folder'); savepath;

2. By clicking this link, you can install the package directly from MATLAB File Exchange.

For further information read user manual.