Direct Stream Digital

Audio Codecs

The 1-bit high-sampling-rate digital audio format promoted by Sony and Philips is based on DSD64, which is recorded at 2.8224 MHz and serves as the audio codec for the high-density layer of SACD.

Explanation

Direct Stream Digital (DSD, 直接比特流数字编码) is a high-sampling-rate, 1-bit digital audio format promoted by Sony and Philips, originally used for archival conversion and Super Audio CDs. SACD uses a base rate of 2.8224 MHz, which is 64 times that of 44.1 kHz, and came to be commonly known as DSD64.

DSD represents changes in the signal at adjacent time points through a high-speed 1-bit sequence. Since a single sample has only two states, the system relies on oversampling and noise shaping to push a large amount of quantization noise above the audible frequency band; the playback device must use a low-pass filter to suppress high-frequency noise. It is a different digital representation method from multi-bit linear PCM, but both require an analog front end, clocking, filtering, and conversion processes; the use of 1-bit does not imply that the entire recording chain involves only a single simple comparison.

Names such as DSD64, DSD128, and DSD256 distinguish data rates based on multiples of the base 44.1 kHz frequency. The SACD physical specification uses DSD64; higher-speed DSD is primarily found in file distribution, recording, and workstation processing; playback devices that support DSD256 files cannot, by that fact alone, read a “DSD256 SACD.”

DSD is a coding method, while SACD is a disc format that includes a physical layer, file structure, program area, and protection mechanisms. DSD files are often saved as DSF or DFF/DSDIFF: DSF facilitates tag embedding, while DSDIFF originates from a professional exchange format; neither is an SACD image. DoP (DSD over PCM), on the other hand, packages DSD bits into transmission frames that appear to be PCM, in order to utilize existing digital audio interfaces; the receiving end identifies the markers and extracts the original DSD. This is not the same as converting DSD values to PCM.

Editing, equalization, dynamic processing, and mixing often require multi-bit operations. Actual production systems may process signals in high-precision PCM, low-bit multi-level formats, or dedicated DSD processing domains before outputting DSD. The DSD identifier in media metadata only indicates the final encoding format; it does not prove that all stages from recording to mastering were free of PCM. High-frequency noise, level specifications, and differences in equipment filtering also mean that comparisons between DSD and PCM cannot rely solely on sampling figures.