A properly dithered 16-bit digital audio storage system with accurate analog-to-digital (A/D) and digital-to-analog (D/A) converters will outperform any analog storage medium in existence. With reasonable accurate converters, such a system performs and measures well at high levels. Its low-level behavior, however, is vitally dependent upon the small-signal differential linearity of the A/D and D/A converters used. Only if nonlinearities are held to tight tolerances can the performance come close to theoretical perfection, and the system not display low-level distortions. This problem is exacerbated by the fact that most converters display their worse differential nonlinearity at zero level, in the center of their range. We discuss here some of the problems in the D/A converter system design, including the digital filter of interpolating D/A systems (so-called "oversampling" D/As), and display measurements made on compact disc players of assorted vintages. We find that the small-signal accuracy of the D/A systems of many currently available CD players is worse than that of models of a few years ago, and this applies even to some of the most expensive machines. The arithmetic design of many of the digital filter circuits in use is also suboptimal. We make the case that some of these defects are of sufficient magnitude to audibly degrade musical programs, and that proper low-level measurements including the use of dithered test tones should be used in the assessment of D/A converter systems.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.