Spokes.wiki Search About
Scholarly Article source ↗ source url updated Sat Aug 08 2026 00:00:00 GMT+0000 (Coordinated Universal Time)

The Audibility of Typical Digital Audio Filters in a High-Fidelity Playback System

An AES convention paper by Stuart and Craven testing whether the filters used inside digital audio — anti-alias filters in A/D converters, and filters applied in mastering — are audible in high-resolution wideband playback.

Method is double-blind psychophysical discrimination between filtered and unfiltered signals. The conclusion: listeners are sensitive to the small signal alterations these filters and quantization introduce, and there exist audible signals that a standard CD cannot encode transparently.

The contradiction, kept

That is the opposite verdict to meyer-moran-adda-loop, which found a full CD-standard A/D/A round trip undetectable by any subject on any system. Two peer-reviewed AES papers, both double-blind, disagreeing.

They are not testing quite the same thing, which is the honest reading and also the interesting one. Meyer and Moran inserted a whole conversion loop into ordinary program material and asked whether anyone noticed. Stuart and Craven isolated the filter stage and asked whether its specific alteration is discriminable. A system can be transparent in normal use while one of its components is discriminable under conditions chosen to expose it — those results coexist without either being wrong.

What that means for this wiki: the honest position is “transparent in practice, not provably transparent in principle,” not “digital is transparent, case closed.” The spoke’s oldest edge asked for measurement instead of listening impressions. It now has measurement, and the measurement is contested.

A caveat this wiki should carry

Stuart and Craven are not disinterested parties on this question — the argument that standard CD encoding is insufficient is the argument behind high-resolution and later MQA advocacy. That does not make a peer-reviewed double-blind result wrong, and it is not grounds to downgrade the tier. It is grounds to want a third party, which the corpus does not have.

meyer-moran-adda-loop · analog-vs-digital · analog-vs-digital-pedals · synthesis