What is DSD (Direct Stream Digital) Audio Format

DSD (Direct Stream Digital) is a method of digital audio signal storage based on the principle of pulse density modulation (PDM). DSD is also a registered trademark of Sony Corporation. In principle, a DSD data stream can be stored on any digital storage medium of sufficient size. Corresponding files can be obtained from music download portals.

Read about various audio file formats.

In DSD, the 1-bit data stream obtained by delta-sigma modulation is recorded directly, instead of being decimated internally – as is common in classic analog-to-digital converters – and output it at a lower rate as a data word with a width of 16 to 24 bits than PCM. Due to the oversampling, the stored audio information is technically more precise than data downscaled to PCM at 44.1 or 88.2 kHz, as it is not yet discretized to this sample rate. Digital-to-analog conversion does not require steep-sided interpolation or anti-aliasing filters, which are used in PCM to isolate frequencies above 20 kHz.

Instead, the DSD format can be output directly, because the resulting harmonic spectrum is far into the inaudible range. Due to the fundamentally low dynamic range of a sigma-delta analog-to-digital converter with only one bit as a quantization stage, an enormous conversion noise is created, but this is shifted into the high-frequency range by the effect of noise shaping. Visually, the signal can be understood in such a way that the course of the sound signal is ultimately reproduced by rapidly changing plus and minus on a loudspeaker with forward-backward phases of different lengths. Due to the real but always present low-pass behavior of the playback system and especially that of the loudspeaker diaphragms, there is a band limit. Furthermore, the inertia of the air and the hearing also play a role, which increasingly attenuate the waves, so that ultimately only the desired audio spectrum is perceptible.

Image source and further reading reference : https://zhuanlan.zhihu.com/p/217733414

Criticism of DSD (Direct Stream Digital) Audio Format

The advantages and disadvantages of DSD were controversially discussed in the professional world at the beginning of the millennium. Some scientists criticized it as “unsuitable for high-quality applications” or “completely unsuitable for high-resolution audio”, while others defended the technology. So far, no consensus has been formed in the scientific community.

It is debatable whether the sound improvement of DSD over PCM claimed by developers and users actually exists. The authors of different studies conclude that “even with the highest quality equipment under optimal listening conditions and a wide variety of listening focuses or listening experiences of the test subjects, as a rule, no significant differences between DSD and High Resolution PCM (24 bit/176.4 kHz) are audible, so that the hypothesis could be put forward that none of the tested systems stands out due to sound characteristics”, and refer to “the high degree of frustration that many subjects, the majority of whom were accustomed to professional and critical-analytical listening, felt during the performance of the tests, and which they attributed to tonal differences that were not even remotely recognizable to them”.

Since the development of SACD in the 1990s, AD delta-sigma converters have been used in recording technology, from which the DSD signal can be tapped directly. In the recording studio, digital audio workstations are used for digital processing, including effects and other signal processing. End-to-end digital DSD processing from recording to finished mixing is not possible, as today’s computers use a defined word width, e.g. 32 bit, which only has a PCM format. With the exception of one manufacturer, 1-bit files cannot be processed by them at all, the DSD signal of the converter is then not used, or the output on the chip is partially no longer available. When the native DSD recordings are processed, they are always converted first to PCM and later back to DSD, for which the only DSD DAW used to date, the extremely high-resolution PCM format DXD, is used. Conversion from DSD to PCM is possible without loss, but when converting from PCM to DSD, increased quantization noise is generated.

In DSD64 format, the frequency response is extended to about 100 kHz. However, the frequency range above 20 kHz does not contain a useful signal, but only the noise shifted into this range by noise shaping, which already starts at frequencies from 15 kHz. As a result, the theoretically possible signal-to-noise ratio of 150 dB cannot be achieved in practice. In fact, only about 20 dB is possible with modern converters between 20 and 120,000 Hz, which is less than with 24-bit PCM (144 dB). DSD therefore requires significantly larger amounts of data than PCM for a comparable signal-to-noise ratio and frequency range, and many times the data that is possible with PCM through lossless compression, e.g. in FLAC.