Understanding FFT Size in Audacity Filter Curve EQ

This article explains the Fast Fourier Transform (FFT) size setting in Audacity’s Filter Curve EQ effect and how it directly impacts audio precision. FFT size determines the mathematical resolution used to convert audio from the time domain into the frequency domain, directly affecting how accurately the equalizer matches your drawn equalization curve—especially at lower frequencies.

What is FFT Size?

Fast Fourier Transform (FFT) is an algorithm that converts an audio signal into its individual frequency components so that Audacity can manipulate them. In the Filter Curve EQ, the FFT size setting defines the number of discrete audio samples processed in each analysis window. Audacity allows you to select FFT sizes typically ranging from 1024 up to 8192 or higher.

How FFT Size Affects Precision

The primary function of the FFT size setting is balancing frequency precision against time resolution:

• Frequency Precision: A higher FFT size (such as 4096 or 8192) provides more data points per frequency band. This yields greater frequency precision, allowing Audacity to execute sharp, surgical cuts and boosts precisely where you place them on the curve.
• Bass and Low-End Accuracy: Low frequencies have longer wavelengths, requiring a wider window of samples to be identified and filtered accurately. A small FFT size fails to capture enough cycles of low frequencies, causing the actual equalization curve in the bass region to become rounded, inaccurate, or ineffective. Increasing the FFT size is essential for precise low-end control.
• Time Resolution Trade-Off: While higher FFT sizes increase frequency accuracy, they decrease temporal (time) accuracy. Processing a larger block of samples can introduce subtle pre-ringing, phase anomalies, or "time-smearing," where fast transients (like drum hits or plosive consonants) lose their sharpness.

Choosing the Right Setting

• Default (4096): Suitable for most general equalization tasks, providing a solid balance between low-frequency accuracy and transient preservation.
• High (8192 or above): Best for steep cuts, high-pass filtering in the extreme low end, notch filtering, or when high precision is required to remove specific hums or resonances.
• Low (1024 to 2048): Best for broad, gentle equalization shapes across mid-to-high frequencies where sharp transient response is critical and deep bass filtering is not required.