Dithering for 16-Bit Delivery: When, Why, and How

What Dither Actually Does

When you reduce bit depth—from 24-bit or 32-bit float to 16-bit—quantization error becomes correlated with the signal and can sound like grainy distortion on low-level detail (decays, reverb tails, soft room tone). Dither adds a tiny amount of optimized noise so that error becomes uncorrelated, preserving perceived low-level linearity.^[1]

Dither is not a loudness tool, not a replacement for good gain staging, and not required when you stay at 24-bit or float for the entire pipeline. It matters at the mathematical truncation/rounding step into fewer bits.

If you bounce 16-bit without dither, quiet material can sound slightly crunchier or less smooth—sometimes subtle, sometimes obvious on long fades. If you dither multiple times across generations, you stack noise. Once at the end is the rule.

When You Need Dither in 2026

Streaming platforms typically accept 24-bit masters and then transcode. Delivering 16-bit to a platform that accepts 24-bit only discards resolution early. Check your distributor’s current upload specs rather than assuming “CD settings for everything.” Spotify for Artists and other DSP help centers document delivery preferences and loudness normalization separately from bit depth.^[2]

TPDF vs Noise-Shaped Dither

• TPDF (triangular) Industry default for transparent word-length reduction. Flat-ish noise, predictable, widely recommended for general music delivery.
• Noise-shaped (POW-r, UV22-style, etc.) Pushes quantization noise toward less audible bands. Can sound “blacker” on some material but may be less ideal if further processing will occur.
• No dither / truncate Avoid for final 16-bit music masters unless a rare technical constraint forbids noise addition.

If the 16-bit file might still be EQ’d or limited later, prefer plain TPDF and keep a 24-bit undithered master as the real source of truth. Noise shaping assumes you are done processing.

Export Workflow That Avoids Double Dither

Sample-rate conversion and dither are related only in that both can be final-stage operations. If you must go 96 kHz → 44.1 kHz and 24 → 16, use high-quality SRC and dither at bit reduction—order depends on your tool, but never dither before a destructive edit chain you still plan to change.^[3]

True Peak, Limiting, and 16-Bit Delivery

Dither does not fix overs. Limit and leave true-peak headroom appropriate to the delivery path before word-length reduction. Inter-sample peaks can still cause trouble after codecs even if sample peaks look under 0 dBFS—true-peak meters help you see that risk.

On CD-style 16-bit/44.1, a ceiling slightly under 0 dBFS is still wise. For streaming, follow current platform true-peak recommendations (commonly around −1 dBTP for delivery masters) and keep the high-resolution master as your revision file.

Myths

Myth: “Always dither to 16-bit for Spotify.” Prefer 24-bit uploads when the distributor allows. Myth: “Dither makes it louder.” No. Myth: “32-bit float needs dither when bouncing float.” Float bounces are not the same as fixed-point 16-bit truncation. Myth: “More noise shaping is always more pro.” Not if further processing or SRC follows.

DAW Export Checklists (Practical)

Ableton Live: freeze/flatten only when needed; export Audio/Video with high-quality SRC if changing rate; enable dither only when bit depth is 16. Keep a separate 24-bit export preset named clearly so you never reuse a 16-bit preset by habit. For Return tracks and master chain, confirm no second dither utility sits after the export stage.

FL Studio: use the export dialog’s bit depth carefully—select 24-bit PCM for archives, 16-bit with dither for CD-style clients. Avoid “export as mp3 only” as your only master. Keep the project sample rate consistent with the session; convert once at the end with highest quality mode rather than bouncing through multiple intermediate rates.

Logic Pro and other DAWs follow the same philosophy: one high-resolution master, one optional dithered 16-bit derivative, filenames that encode bit depth and sample rate, and a text note in the delivery email stating which file is archival. If a client later asks for a louder version, reopen the undithered master—never re-limit a dithered 16-bit file and re-export 16-bit again without understanding you may be stacking quantization noise.

For game audio or sample pack 16-bit assets, dither per final asset when reducing from 24-bit design sessions. Do not dither the design session itself. Keep a 24-bit pack archive for future re-encoding if a platform changes requirements.