Sample-Rate Conversion
Converting between 44.1 and 48 kHz is not a renumbering exercise; the filter decides how much detail survives.
Related
The problem
Material arrives at 44.1 kHz from records, 48 kHz from video projects and 96 kHz from location recordings, and all of it has to meet in one session. Simply relabelling the rate changes speed and pitch, so the audio must be resampled. Whether it goes well is audible in how much high end survives and whether transients smear.
How it works
Resampling interpolates a continuous waveform at the old rate, then samples it again at the new one. Ideal interpolation needs an infinitely long sinc kernel; real converters approximate it with finite filters whose length and roll-off set the quality. Before downsampling, everything above the new Nyquist frequency must be filtered out, or it folds back into the audible band as aliasing. The ratio between 44.1 and 48 is 147 to 160, a non-integer, so the filter must be longer; 48 to 96 is an integer ratio and much simpler. Real-time conversion shortens filters to keep latency low, while offline conversion can take its time.
Settings and practice
Keep one sample rate per project and convert as few times as possible. For picture work, record at 48 kHz and skip the last conversion. Keep the master at its recording rate and convert once at delivery. Use high-quality settings for offline export and low-latency settings only for live monitoring. Bit depth and sample rate are separate issues, and resampling does not add dither by itself. Rates above 96 kHz buy little on playback, because microphones and rooms usually limit first.
What to listen for
On the funk kit, listen for grit on the cymbals; on the Chopin ballade, for thinning overtones in the tail; on the organ hymn, for hum in the low pedal. Bad conversion rarely sounds obviously broken, it just sounds less clean.
Listen in the library
All tracks from midicn-lib — inline badge is the license tier (C1 commercial / C2 non-commercial / C3 study only).
Sources
- 采样定理与重建滤波的表述，依数字音频通识教材核对，未抄原文
- 常见采样率之间的整数与非整数比关系，依公开技术资料核对，未抄文本
Used for fact-checking only; all prose is original.