Dithering in mastering: when it matters and when it does not

Dithering is a small but important part of digital audio mastering. It becomes relevant when a recording is reduced from a higher bit depth to a lower one, such as when a 24-bit master is converted to a 16-bit file for a CD or another fixed-point delivery format. The process adds a carefully controlled layer of noise so that quantisation errors are less noticeable and low-level detail remains more natural.

For artists in Australia, the subject often appears when preparing masters for streaming, digital distribution, radio, CD manufacture, video, or an audiobook platform. A release may pass through several sample rates, file types and loudness checks, so knowing exactly when to apply dither helps prevent unnecessary processing. The basic rule is simple: dither belongs at the final bit-depth reduction, not automatically on every export.

How digital audio creates quantisation error

Digital audio represents the amplitude of a waveform with a finite number of values. Bit depth determines how many values are available. A 16-bit file provides 65,536 possible amplitude steps, while a 24-bit file provides many millions. At normal listening levels, this difference is difficult to hear as a general loss of fidelity, but it matters when a signal becomes very quiet.

When a 24-bit signal is converted to 16-bit, samples have to be rounded to fit the smaller set of available values. That rounding is called quantisation. With a strong signal, the resulting error is usually masked by the music. During a fade, a reverb tail, or a quiet acoustic passage, however, the error can become correlated with the audio waveform. Instead of sounding like a consistent background hiss, it may produce graininess, modulation or low-level distortion.

Dither adds a very quiet, randomised noise signal before the word-length conversion. This makes the rounding error less related to the music, replacing potentially objectionable distortion with a more consistent noise floor. The added noise is extremely quiet, but its effect can be useful when preserving the final details of a recording.

Dither does not increase resolution or restore information that has already been removed. It is a way of making an unavoidable reduction in numerical precision behave more gracefully. That distinction is useful when evaluating claims about “higher resolution” files: dither can improve the character of a 16-bit conversion, but it cannot turn a 16-bit source into a true 24-bit recording.

The correct point in the mastering chain

The usual place for dither is the last significant processing stage, immediately before exporting to the required fixed-point bit depth. If a mix arrives at 24-bit and the final master must be 16-bit, the mastering engineer completes EQ, compression, limiting, clipping, fades and metering first. Dither is then applied as part of the final conversion.

Any process that changes the audio after dithering can alter or partially obscure the noise, and another bit-depth conversion may require a fresh decision. This is why a dithered 16-bit master should not be reopened, processed, normalised and exported again without care. The safest workflow is to keep a high-resolution production master, then create each delivery file from that source.

A modern DAW may use 32-bit floating-point processing internally. This gives plugins and automation considerable headroom during editing and mixing, but it does not automatically mean that the final file is 32-bit float. Export settings still determine whether the file becomes 24-bit integer, 16-bit integer or another format. Dither is generally associated with the transition to a lower fixed-point word length.

A final limiter may include a dither option, while other systems provide a dedicated dither plugin or an export checkbox. Only one stage should normally be responsible for it. If both the limiter and the export process add dither, the result may contain more noise than intended. The exact controls vary between DAWs, so the engineer should verify what the software is doing rather than selecting every available dither setting.

Choosing between types and noise shaping

Different dither algorithms use different noise distributions. Triangular probability density function, often called TPDF, is a widely trusted general-purpose choice because it removes tonal quantisation patterns without introducing unusual modulation. It is a sensible option when a transparent, dependable conversion is more important than squeezing every last fraction of perceived noise performance from the file.

Noise-shaped dither moves some of the added noise away from the most sensitive part of human hearing. A suitable curve can make the noise floor seem quieter in normal listening, particularly in music with fades or sparse arrangements. However, noise shaping places more energy at higher frequencies. The choice therefore needs to suit the delivery format, the programme material and the monitoring environment.

For a pop master intended for Australian streaming services, a moderate noise-shaped option may be appropriate when the final output is 16-bit. For a spoken-word project, aggressive high-frequency shaping deserves more caution because listeners may hear the top end clearly through headphones. In an audiobook, room tone, breaths and fade-outs can expose processing that is masked by a dense musical arrangement.

A dither setting is not a substitute for careful gain staging or a good master. There is no universal “best” algorithm for every release, and the difference is often subtle when compared with choices involving balance, dynamics, clipping and true-peak control. Engineers should audition the beginning and end of the programme, quiet sections and any long fades at a realistic monitoring level.

Streaming, downloads and Australian releases

Many streaming platforms accept 24-bit files, and a mastering studio may deliver 24-bit WAV masters at 44.1 or 48 kHz, depending on the release requirements. If the file remains 24-bit from the final processing stage through delivery, there is normally no reason to add dither simply because it will be uploaded. The platform may transcode the file internally, but that is a separate process outside the artist’s control.

A 16-bit master can still be perfectly valid for digital distribution, especially when a distributor or client specifically requests it. In that case, dither should be applied once during the final reduction. It should not be added merely because a file is being uploaded to a distributor, nor should it be applied to a 24-bit master as a ritual step.

Australian artists often prepare several related assets for one campaign: a streaming master, a video master for YouTube, a radio edit, an instrumental version and perhaps a CD-ready file. A Sydney electronic act, a Melbourne band or a Brisbane singer-songwriter may all receive different technical requests from labels, distributors, venues or broadcasters. Maintaining one clean high-resolution premaster and making controlled derivatives prevents confusion between these versions.

Listening habits also matter. A track may be checked on earbuds during a Melbourne tram commute, in a car between suburbs, through a phone speaker or on studio monitors. Dither itself is unlikely to be obvious in those everyday situations. The more audible risks usually come from excessive limiting, clipped transients, harsh high frequencies or an unsuitable tonal balance. Even so, a well-managed final conversion protects the quietest musical details when listeners use revealing headphones.

A professional facility can also help keep delivery decisions consistent across projects. Reviewing a studio’s client roster can show the range of artists and spoken-word creators it supports, while a mastering discussion can establish whether the release needs 16-bit, 24-bit, broadcast or video deliverables.

Situations where dither is unnecessary

If the final file is exported at the same bit depth used by the mastering process, dither is usually unnecessary. For example, a 24-bit master created and delivered as 24-bit integer does not need dither solely because it is being saved. The same applies when a 32-bit floating-point file is retained for further production, provided no fixed-point reduction has taken place.

Dither is also not a cure for sample-rate conversion. Changing 48 kHz audio to 44.1 kHz involves interpolation and filtering, while changing bit depth involves quantisation. Some export tools perform both operations together, but they remain separate technical decisions. A mastering engineer should use a high-quality sample-rate converter and apply dither only if the word length is also being reduced.

Repeated dither can accumulate unnecessary noise. It is especially easy to do this when a file is exported from a DAW, opened in a video application, processed by a distributor, and then converted again. Not every stage exposes its settings clearly. Keeping a 24-bit or 32-bit working master, documenting the intended delivery format and avoiding avoidable conversions provides a reliable safeguard.

Some engineers choose not to dither a 16-bit export when the source material already contains substantial analogue noise, tape hiss, room sound or dense ambience. That decision should be made deliberately rather than assumed to make dither irrelevant. The added noise may still help decorrelate quantisation distortion, even when it is unlikely to be heard in isolation.

The final file should be checked after export. Confirm its bit depth, sample rate, start and end points, fades, peaks, metadata and playback behaviour. If a 16-bit file was required, verify that the conversion actually happened and that no later application silently changed the format. A technical delivery sheet can be valuable when several people are handling the release.

Applying the process in a real mastering workflow

A practical workflow starts with a high-resolution mix or premaster. The engineer confirms the requested destination, checks whether the client needs a streaming master, video file, CD image, broadcast WAV or spoken-word delivery, and completes all creative processing. The mix is then monitored for tonal balance, dynamics, distortion, noise and fade behaviour before the final export settings are selected.

Suppose a 24-bit mix is mastered for a 16-bit CD production file. The engineer finishes the tonal and dynamic work, sets the final ceiling, chooses a suitable dither algorithm, reduces the word length and exports once. The resulting file is then auditioned from start to finish, with particular attention to fade-outs, sparse passages and reverberation tails. The 24-bit premaster remains available for other versions.

For a 24-bit streaming master, the engineer may leave dither disabled and export directly at 24-bit. If a separate 16-bit download or CD file is needed, it should be created from the same high-resolution master with its own final conversion. The two files can share the same musical master while using different technical endings.

Studios working with Australian voice-over, podcast and audiobook clients need the same discipline. Spoken-word content often has long pauses and low-level consonants where noise and distortion are easier to notice. The file may also be used for a broadcaster, a publisher or an online platform with its own specifications. Clear communication about sample rate, bit depth, mono or stereo format, loudness and file naming is as important as the dither choice.

An engineer may draw on different microphones, analogue equipment and monitoring paths before reaching the final export. Facilities such as LnL Recording list their available studio equipment, which can be useful context when discussing how a project will be recorded, edited, mixed and mastered. Dither is only one small step, but it should fit into that wider chain rather than being treated as an isolated quality setting.

Australian copyright law, including the Copyright Act 1968, does not prescribe a particular dither algorithm or bit depth. It does make clear why artists should retain properly identified masters and records of who created and supplied them. Technical preparation does not replace ownership documentation, cue information or licensing. Keeping an unaltered high-resolution master also makes future formats easier to produce if a label, publisher or distributor changes its requirements.

For a release aimed at Australian listeners, there is no special local dither setting for Sydney, Perth, Adelaide or any other city. The relevant variables are the actual delivery format, the word-length conversion and the quality of the final verification. Apply dither once when reducing bit depth, avoid it when no reduction is taking place, and treat the final export as a controlled technical decision within the complete mastering process.