Delivery Sample Rate
Delivery rate follows the receiver; convert once, and never after the limiter.
Related
48 kHz is a compatibility choice, not a quality grade
Converting a finished 44.1 kHz track to 48 kHz does not make it clearer; it changes the time base so a video project can accept it. Go the other way, and 48 kHz material headed for CD must return to 44.1. At delivery, sample rate is first a question of compatibility and only second a question of sound.
Match the rate to the receiver
| Receiver | Rate | Note |
|---|---|---|
| Digital release and streaming | 44.1 or 48 | Both accepted; send the master as cut |
| Video, advertising, broadcast | 48 | Shares the picture time base and saves a conversion |
| CD master | 44.1 | With 16-bit depth and dither |
| Film post | 48 or 96 | Follow their session; ask before converting |
| Game middleware | 48 | Recheck loop points and loudness after conversion |
| High-resolution release | 96 or 192 | Meaningful only if made at that rate throughout |
Do not mix rates inside one delivery batch. Editing software reads each file at its own rate, and a mixed batch drifts out of sync with picture.
Convert once, before the limiter
Resampling is filtering, and it alters transients and high-frequency phase; two conversions filter twice and stack the loss. Order matters as well: convert the rate, then limit, then measure true peak. Convert after limiting and filter overshoot can push a file that sat at -1 dBTP past 0. If a finished file must be converted, leave at least 1 dB of margin and measure again afterwards.
The ratio between 44.1 and 48 is not an integer, so the converter performs asynchronous resampling, where quality differences show more clearly than in integer ratios. Use an offline converter with an anti-aliasing filter rather than exporting with real-time preview settings.
Keep 96 kHz in production
The point of 96 kHz is margin for pitch shifting, time stretching and nonlinear plugins, not a delivery format. Files double in size, plugins work harder, and the receiver often converts back to 48 anyway. Deliver 96 only when the whole project was made there and the publisher asks for it; otherwise work at 96, deliver at 48, and archive the high-rate session.
How it goes wrong
- Changing only the header or extension, so speed and pitch shift
- A file labelled 48 kHz that actually holds 44.1, drifting as soon as it is imported
- Resampling after the limiter, with no true-peak recheck
- A batch that mixes 44.1 and 48, impossible to align in the edit
- Upconverting 44.1 to 96 for supposed clarity, gaining size but no detail
Listen in the library
All tracks from midicn-lib — inline badge is the license tier (C1 commercial / C2 non-commercial / C3 study only).
Along the same line
Sources
- Audio Engineering Society, AES5-2018，专业音频应用的首选采样频率（标准）
- ITU-R BS.1770-4，节目响度与真峰值测量（标准）
- Avid Pro Tools Reference Guide，会话采样率与采样率转换的行为说明（产品文档）