Sample rate
How many times per second a continuous signal is measured on its way into digital form. 44.1 kHz is the CD standard, 48 kHz the video and pro-audio one, and 96 or 192 kHz the numbers sold as "high resolution."
The Nyquist–Shannon result sets the floor: reconstructing a band-limited signal needs sampling at
twice its highest frequency, so ~40 kHz covers the 20 Hz–20 kHz range of human hearing. CD's extra
4.1 kHz is headroom for the anti-alias filter to roll off before the limit rather than fall off a
cliff — the practical reason the number is not 40. (The theorem itself is
../mathematics-wiki's subject; see fourier-analysis there for sampling theory proper. This page
holds the claim about audio.)
Two questions that keep getting merged
Does a higher rate sound better on playback? The corpus's answer is that nobody has demonstrated it does. meyer-moran-adda-loop (AES, double-blind, a year of testing) inserted a whole 16-bit/44.1 kHz conversion loop into high-resolution playback and found it undetectable by any subject on any system. headphonesty-192khz-vs-441khz adds Mark Waldrep reporting the same outcome from his own blind studies, and practitioner statements to match — Bob Katz and Dan Worrall both put the audible difference between 44.1 and 96 kHz at negligible, and both redirect the question to mixing and mastering decisions.
Is a higher rate useful in production? Here the same sources say yes, and the distinction is the one consumer coverage keeps losing. Oversampling headroom above the audio band is what keeps nonlinear processes from folding aliases back into it — pitch shifting, time stretching, extreme EQ, and saturation or vintage-modelled plugins that generate harmonics above Nyquist by design. Worrall's line is that these are "engineering conveniences, not playback enhancements."
So the working position: high rates are a tool for the studio and a specification for the store. A record mastered at 48 kHz/24 bit and distributed at 44.1 is not making a compromise anyone has shown to be audible.
The claim that it is actively worse
Distinct from "pointless," and it rests on a mechanism rather than a listening result. Montgomery argues ultrasonic content is a liability: amplifiers and tweeters not designed to reproduce 30 kHz respond nonlinearly, and two ultrasonic tones can intermodulate down into the audible band. Dan Lavry, who designs converters, argues 192 kHz is "about 3 times faster than optimal" and compromises accuracy in the converter itself.
Both are plausible and neither is a measurement of audible harm from real program material on real consumer equipment — see the reading of the evidence on the source page. A null result from a listening test says no difference; it cannot be run backwards into a difference in the other direction. The corpus records the mechanism as a mechanism.
What is not settled
digital-filter-audibility (Stuart & Craven) is the standing complication: double-blind, AES, and it finds listeners are sensitive to the filters inside digital audio, with signals a CD cannot carry transparently. It and Meyer & Moran are not testing the same thing — a whole loop in ordinary material versus an isolated filter stage under conditions chosen to expose it — and this wiki keeps both. The honest summary stays what analog-vs-digital already records: transparent in practice, not provably transparent.
Untouched by any source here: bit depth, which is a different axis and gets conflated with this one constantly. Katz's own quoted preference is for 24-bit, not for a higher sample rate.
Related
analog-vs-digital · meyer-moran-adda-loop · digital-filter-audibility · headphonesty-192khz-vs-441khz · monty-montgomery · loudness-normalization · daw · synthesis