Mastering & delivery
When to use dither, and where to put it in your export chain
A simple decision process for bit-depth reduction, final mastering and avoiding accidental double dithering.
Dither adds a small noise signal before quantisation to a fixed integer bit depth, preventing signal-related quantisation distortion. A final 16-bit PCM export from a floating-point mix is a clear use case. A floating-point file for more production work does not need a 16-bit dither merely because the original recording was 16-bit.
By Spectraforge Audio3 min read
Published
Guide review: Captum Records and AS Mastering. AI-assisted writing and source checks. How these guides are made.
Use one dither stage at the final integer quantisation
First choose the file you are making. Then place the appropriate dither after the final EQ, dynamics, gain and sample-rate changes. MP3 and AAC do not have a fixed PCM bit depth; use the encoder’s requested source format instead of looking for a “16-bit MP3” setting.
Choose one of two verifiable signal paths
Host route: keep plugin dither off, complete the mix processing, and let the host dither its final integer export. Check whether export normalisation or rate conversion occurs before that step. Plugin route: use the final processor’s dither/quantisation at the target depth, then follow its documented export instructions with host dither off and no later audio changes.
FabFilter specifically advises placing Pro-L 2 after gain changes when using its dither. A master fader after the plugin can invalidate that intended order. Turning on both the limiter’s dither and the host’s dither does not make the file more precise; it adds another noise source.
Worked chain check: a level change after the limiter
Imagine you have enabled 16-bit dither in a limiter, then lowered a downstream master fader to meet the delivery level. The dither is no longer at the last audio change. Move the level decision before final quantisation, or disable plugin dither and use the host’s final export stage. Re-export from the higher-precision session rather than editing the already finished 16-bit file.
Noise-shaped choices move the added noise around the spectrum. Choose a developer-supported mode for the sample rate and destination. Keep a higher-precision archive for future revisions; a collaborator who will EQ or resample the file needs that production version, not repeated processing of your final low-bit-depth master.
Check a decay without chasing an audible dither effect
Suggested check: export a sparse piano or reverb ending with the intended final settings and listen for an uninterrupted decay. If you magnify the quiet tail’s level to investigate it, turn the monitor level back down before playing the song. Do not change the master because a tiny effect becomes dramatic only under extreme gain.
Dither may be inaudible in ordinary listening, particularly when the programme already contains noise. Hearing no obvious difference does not diagnose a broken export. Confirm the actual file depth, the single dither location and the absence of later processing. Record those settings with the final file; do not add dither again to make it “more finished”.
Technical references
- FabFilter Pro-L 2: dithering, noise shaping and final-stage placement
- Audacity: dither and bit-depth conversion
Listening exercises are proposed practice routines. Worked examples illustrate the stated setup; they are not reports of our listening tests. Editorial process and corrections.