IT lexicon › Programming › Sample rate conversion

Sample rate conversion

Programming På svenska → Updated: 2026-10-27

Recomputing audio from one sample rate to another — trivial on paper, and a classic source of audible junk in practice.

Going from 44.1 to 48 kHz is a ratio of 160:147, so you cannot simply drop or duplicate samples. The principle is to interpolate new values between the old ones and low-pass away everything above the new Nyquist frequency, or it folds back down as aliasing. The ideal filter is a sinc function of infinite length; real converters differ mainly in how much of it they can afford to compute.

Cheap converters leave aliasing products and dull top end, and for years you could measure exactly that in operating system mixers. Libraries like libsamplerate and SoX made high quality free. The same machinery sits inside oversampling, and inside every true peak meter, which resamples to find the peaks hiding between samples.

← Back to the lexicon