Quick answer: keep the working master at its native high resolution, finish editing, EQ, repair, level changes, limiting, and sample-rate conversion, then open Process → Bit Depth → Bit-Depth Converter. Reduce to the delivery depth and apply dither once in that final reduction. Don’t add more gain, fades, EQ, or other DSP to the dithered file.
Dither is not a loudness effect and not noise reduction. It deliberately adds a very small noise signal so quantization error no longer follows low-level audio as correlated distortion. The practical decision is therefore not “Can I hear hiss?” but “Am I throwing away precision, and is this the final time I will do it?”
Bit Depth Answers a Resolution Question
| Situation | Recommended decision | Why |
|---|---|---|
| Recording or active processing | Keep 24-bit or higher working precision | More room for low-level calculations and gain changes |
| 24-bit delivery requested | Deliver 24-bit; do not reduce to 16-bit first | No 16-bit reduction is required |
| 16-bit WAV/CD-style delivery | Reduce once with appropriate dither | Avoid correlated quantization distortion |
| 16-bit source opened in 24-bit | Useful for processing, not restoration | New empty bits do not recover lost detail |
| Lossy output such as MP3/AAC | Follow the encoder and delivery specification | Codec conversion and PCM bit depth are different decisions |
A nominal 24-bit file can hold much lower-level values than a 16-bit file, but real converter noise and the source environment limit usable dynamic range. The benefit in production is robust processing headroom and fewer rounding problems—not a promise that every recorded file contains 144 dB of meaningful acoustic detail.
What Goes Wrong Without Dither?
Reducing bit depth maps a finer set of possible sample values onto a coarser set. Without dither, the rounding or truncation error is related to the signal. At normal listening levels this may be masked, but on quiet fades, reverb tails, sparse acoustic material, or a boosted null test it can appear as grain, modulation, or harmonic distortion.
Dither decorrelates that error. The file still has the target number of bits, and the added noise is real, but low-level behavior is more linear. This is why judging dither on silence alone is misleading: the relevant comparison is a very quiet signal or fade converted with and without dither at the same level.
How to Convert 24-Bit to 16-Bit
- Open the approved high-resolution master, not a previously dithered derivative.
- Finish all edits, restoration, gain, EQ, dynamics, fades, and limiting.
- If sample rate must change, perform and verify that conversion before final dither.
- Confirm peaks and the required delivery loudness.
- Open Process → Bit Depth → Bit-Depth Converter.
- Choose 16-bit and a suitable dither/noise-shaping option.
- Preview exposed quiet material, then apply once.
- Save as a new 16-bit lossless delivery master and verify its format.
The current Sound Forge 2026 Bit-Depth Converter documentation confirms the menu path, available dither families, and noise-shaping controls. It also notes that increasing bit depth does not improve existing audio quality, although higher precision can be useful for later processing.
Set Level Before Bit-Depth Reduction
Sound Forge’s help recommends maximizing useful signal level before reducing bit depth because the lower-depth file has a poorer signal-to-noise ratio. Interpret that as an ordering rule, not permission to peak-normalize every master to 0 dBFS. Complete the intended loudness and peak strategy before dither, leaving any true-peak or codec headroom required by the delivery.
If normalization is appropriate, use the Sound Forge normalization guide before the Bit-Depth Converter. If the programme already meets its level specification, do not make it louder merely to “use all the bits.” A well-mastered quiet work can remain intentionally quiet.
Choose Dither and Noise Shaping
Sound Forge lists Half Rectangular, Rectangular, Triangular, Highpass Triangular, Gaussian, and POW-r choices. The current help says Highpass Triangular with noise shaping is generally favorable, while also making clear that material matters. A safe comparison uses identical 16-bit renders from the same untouched high-resolution master.
- Triangular: a conservative general-purpose choice that avoids signal-dependent noise-floor modulation.
- Highpass Triangular: shifts more dither energy upward; useful when the delivery sample rate and material support it.
- POW-r: offers three shaped modes with increasingly specific noise distribution and should be chosen by programme and delivery, not by “higher number is better.”
- Noise shaping off: keeps the noise distribution simpler and can be safer if the file may be processed again—though the better solution is to process the undithered master.
Sound Forge warns against high-pass noise shaping below 44.1 kHz because moving noise near a low Nyquist frequency can place it in a more audible region. That is a version-specific, documented reason to check sample rate instead of applying one favorite preset to every file.
POW-r #1, #2, and #3 Are Not Quality Grades
The official descriptions distinguish different shaping behavior: POW-r #1 uses a special dither curve; #2 adds broad noise shaping; #3 applies stronger optimized shaping in a sensitive mid-frequency region. The numbers do not mean that #3 is universally superior. Dense music, exposed speech, classical dynamics, and later codec encoding can respond differently.
When the delivery matters, render the same quiet ending with candidate settings, level-match, and listen at normal and raised monitor levels. Don’t pick a mode from an isolated silent section; shaped dither is designed around audibility in the presence of programme material.
Why Dither Must Be Last
Any later gain change, fade, EQ, dynamics process, or sample-rate conversion creates new sample values that must be quantized again. The original dither has already served its purpose at the earlier boundary; processing it and then dithering again adds another noise layer and makes the chain harder to reason about.
If a client requests a revised fade after the 16-bit master is made, reopen the high-resolution master, make the fade there, and create a new 16-bit delivery with one final dither pass. Don’t edit the existing 16-bit dithered file unless the high-resolution source is genuinely unavailable.
Keep One High-Resolution Source of Truth
Archive the approved 24-bit or higher-resolution master before making delivery variants. From it, create 24-bit WAV, 16-bit WAV, CD, broadcast, or encoded outputs as separate branches. This avoids a ladder in which 24-bit becomes 16-bit, then the 16-bit file becomes MP3, then that MP3 is used for another edit.
Use filenames that record role and format: Project_Master_24b48k.wav, Project_Delivery_16b44k1.wav. The sample-rate conversion guide covers the separate rate decision; the file conversion workflow covers container and codec output.
How to QA a Dithered Master
- Verify the file reports the requested bit depth and sample rate.
- Check the beginning, quietest passage, fade-out, reverb tail, and final silence.
- Compare at matched loudness with the high-resolution master.
- Confirm peak and loudness targets were set before dither.
- Look for accidental clipping introduced before conversion.
- Confirm no plugin or gain process ran after the dither step.
- Keep a written record of dither type and noise-shaping mode.
Do not expect a dramatic audible difference on every master. Dither is a controlled technical safeguard at the resolution boundary. Its value is preventing avoidable low-level distortion, not producing an obvious “processed” sound.
Test Dither Without Fooling Yourself
Make candidate renders from the same high-resolution master with identical level, sample rate, and bit depth. Change only the dither or noise-shaping setting. Compare normal playback first; then, if needed, raise monitor gain on a quiet tail while protecting speakers and hearing. Switching between files that differ in level or limiter behavior is not a dither comparison.
A null test can reveal the conversion residue, but the residue is not meant to sound like programme audio. Invert one aligned render and sum it with another only after confirming that both files start at the same sample and contain no metadata-induced offset. Use the test to understand error behavior, not to rank whichever difference signal sounds quieter in isolation.
Noise-shaped dither may measure more energy in one frequency range while sounding less intrusive in context. An unweighted RMS number therefore does not settle the listening decision. Check the destination too: aggressive shaping that is benign in a 16-bit PCM master may interact differently with later lossy encoding, sample-rate changes, or broadcast processing.
Delivery Examples
24-Bit Streaming or Archive Master
If the specification asks for 24-bit PCM and the working master is already 24-bit or higher precision, do not create an intermediate 16-bit file. Save the required 24-bit master and keep the project source. Whether a workstation internally calculates at floating point does not change the delivery bit-depth requirement.
16-Bit WAV or Audio CD Master
Complete loudness, peaks, sample-rate conversion to 44.1 kHz when required, then reduce to 16-bit with one dither pass. Verify that the saved file truly reports 16-bit PCM and that no export stage silently processed it again.
MP3 or AAC Delivery
Lossy encoders do not preserve PCM samples by simply “keeping 16 bits.” Feed the encoder the approved master format recommended by the delivery workflow and avoid an unnecessary 16-bit intermediate unless the specification requires one. Retain the PCM master separately from encoded distribution copies.
Common Bit-Depth Mistakes
- Dithering a 24-bit master by habit: only dither when an actual precision reduction requires it.
- Upconverting 16-bit and claiming restoration: the container changes; lost information does not return.
- Normalizing after dither: return to the high-resolution master, set level, then dither again once.
- Dithering every save: keep processing at high precision and dither only the final reduced-depth delivery.
- Choosing POW-r #3 because the number is largest: compare modes on the actual programme.
- Using high-pass shaping at a low sample rate: follow the documented sample-rate caution.
Bit Depth and Dither FAQ
Do I need dither when converting 24-bit to 16-bit?
Yes as the normal final mastering practice. Apply it once while reducing the approved high-resolution master to the 16-bit delivery.
Should I dither a 24-bit export?
Not merely because it is an export. Dither is tied to reducing precision. Keep a 24-bit delivery at its required depth unless the processing chain or specification creates a real reduction.
Does converting 16-bit audio to 24-bit improve quality?
No. It can provide a higher-precision container for further processing, but it cannot recreate detail already lost at 16-bit.
Which POW-r setting is best?
No mode is best for every source. The modes use different noise shaping, so compare the actual programme—especially quiet passages and fades—against the delivery requirement.
Can I normalize after dithering?
Do not do so in a planned workflow. Normalize or limit the high-resolution master first, then perform the final bit-depth reduction and dither once.
The Practical Rule
Finish the master at high resolution, set level and sample rate, reduce bit depth once with appropriate dither, and never process that dithered delivery again. Revisions begin from the high-resolution source of truth.
Last fact-checked August 5, 2026 against the current Sound Forge 2026 online help.