Sample Rate and Nyquist
Sample rate sets the highest frequency a system can capture; Nyquist gives that ceiling.
Related
The problem
A sampler stores only a limited number of points each second, yet the result sounds continuous. Why does some high-frequency content seem to become a different sound instead of disappearing?
How it works
Sampling reads a continuous waveform at fixed intervals. With fs points per second, the highest representable frequency is fs/2, the Nyquist frequency. At 44.1 kHz the ceiling is about 22.05 kHz, covering most listeners. Content above the ceiling does not vanish; it folds back into the audible band as unrelated tones. An anti-aliasing filter before conversion removes that content, usually inside the converter chip. A 48 kHz rate leaves room for video work, while 96 kHz mainly gives processing and pitch shifting more headroom, not more hearing.
Practice and settings
Deliver music at 44.1 or 48 kHz. Record at 88.2 or 96 kHz when material will be shifted or stretched far, giving the algorithm room. When downsampling, filter first and drop points second; the order matters. Inspect the spectrum near 20 kHz for a sudden cliff, the filter's footprint.
What to listen for
Start with celesta or bell high notes and notice the fine layer of partials above them. Play the same passage at 22.05 kHz: the top turns dull and some bright sparks become midrange hiss. Hi-hats lose their edge first. Hear the bass, which barely changes, and do not mistake a transposition artifact for a rate problem.
Common misconceptions
That a higher rate always sounds better; that ultrasonic content is useless; that downsampling just turns down the treble.
Listen in the library
All tracks from midicn-lib — inline badge is the license tier (C1 commercial / C2 non-commercial / C3 study only).
Sources
- Harry Nyquist, Certain Topics in Telegraph Transmission Theory — 采样上限的原始论述核对，未抄原文
- Ken C. Pohlmann, Principles of Digital Audio — 采样率与转换滤波的通行说法核对，未抄原文
Used for fact-checking only; all prose is original.