Audio Dithering & Noise Shaping: Why and When to Dither in DAWs

In digital audio, dither is the intentional addition of a microscopic amount of randomized noise (less than the least significant bit) prior to word-length reduction (quantization). Counter-intuitively, adding noise dramatically increases audio fidelity.

Truncation Distortion vs Dither

When downsampling a 24-bit or 32-bit floating-point master to 16-bit for CD release, the simplest operation is chopping off the bottom 8 bits (truncation). However, low-level signals (such as reverberation tails fading into silence) become rounded to fixed stair-step voltages. This generates correlated harmonic distortion known as quantization error distortion.

By adding Triangular Probability Density Function (TPDF) dither, quantization errors are randomized and decoupled from the musical signal, turning harsh digital distortion into a benign, steady hiss sitting at -96 dBFS.

Noise Shaping Curves

Advanced dithering algorithms apply noise shaping filters that shift the dither noise energy into frequency bands where human hearing is least sensitive (above 14 kHz), lowering the perceived noise floor in the sensitive 2 kHz–5 kHz midrange to around -115 dBFS.