Dither and noise shaping explained
Dither is a tiny amount of deliberately added noise used when reducing digital-audio bit depth. It trades signal-related quantisation distortion for a very low, more benign noise floor. Noise shaping can then move much of that noise into frequency regions where hearing is less sensitive.
What is dither?
When a high-precision signal is reduced to fewer bits, many sample values can no longer be represented exactly. If the extra precision is simply truncated or rounded, the resulting quantisation error can become correlated with the audio and behave like distortion.
Dither adds a carefully controlled random signal before quantisation so that this error is decorrelated from the music.
Why would adding noise improve audio?
Because random low-level noise is often less objectionable than deterministic distortion. Dither does not remove quantisation error; it changes its behaviour so low-level signals, fades and reverberation tails can be represented more naturally after bit-depth reduction.
Common misconception: dither is not an enhancement filter that restores lost bits. It is a controlled way to make the unavoidable error from reducing bit depth less correlated and less objectionable.
What is noise shaping?
Noise shaping uses feedback filtering to redistribute the quantisation-noise energy across frequency. The total error is not magically removed; more of it is moved away from frequency regions where human hearing is most sensitive.
This is why some 16-bit mastering systems can achieve a perceived noise floor that is better in critical parts of the audible band than a simple unshaped 16-bit quantiser would suggest.
When is dither normally used?
- When reducing a 24-bit or higher-precision master to 16-bit CD audio.
- When exporting to a lower fixed-point bit depth than the processing chain used.
- As the final stage after level, EQ, compression and other processing that changes sample values.
If the final output keeps the same or greater fixed-point precision, an additional dither stage may not be required simply for the sake of it.
What dither and noise shaping do not do
- They do not turn 16-bit audio into true 20-bit or 24-bit source material.
- They do not recover information lost by lossy compression.
- They do not improve a poor recording or mastering decision.
- They do not change the official bit depth of the delivery format.
Why this matters for “20-bit” and “24-bit” remasters
Mastering can be carried out at higher precision even when the final CD must remain 16-bit. A carefully designed dither/noise-shaping stage is the bridge between those two facts.
Sony's 1990s Super Bit Mapping is a well-known example: 20-bit material was quantised to the standard 16-bit CD format while the quantisation error was shaped psychoacoustically.
Frequently asked questions
Can you hear dither?
At normal listening levels it is generally extremely low-level. Its purpose is to replace more objectionable correlated quantisation distortion with a benign noise floor.
Should dither be applied repeatedly?
Normally no. It is most useful at the final bit-depth reduction stage. Repeated unnecessary dithering can accumulate extra noise.
Is noise shaping the same as dither?
No. Dither randomises quantisation error. Noise shaping changes the frequency distribution of that error/noise. They are often used together.
Related pages
20-bit remaster · 24-bit remaster · Audio bit depth · 24-bit vs 16-bit audio