Direct Stream Digital
DSD (Direct Stream Digital) is a method of high-resolution audio signal storage based on the principle of delta-sigma modulation .
The DSD method is mainly used for Super Audio CD (SACD) . The direct data stream of a delta-sigma modulator is saved (this means the output signal of the feedback loop of the modulator ), which operates at 2.8224 MHz; this corresponds to 64 times (DSD64) the sampling rate of 44.1 kHz, which is also used in the audio CD ( Red Book CDDA), which works with linear 16-bit pulse code modulation (PCM) . Newer, higher-resolution DSD versions go far beyond this, e.g. B. up to 512 times the 44.1 kHz rate (DSD512), but DSD128 and DSD256 are also common. Hard drives, SSDs, USB sticks and flash memory cards are used as storage media today.
function
The high-resolution one-bit data stream obtained in this way is recorded directly instead of being internally decimated - as is customary in classic analog-digital converters - and output at a lower rate as a data word with a width of 16 or 24 bits as PCM. Due to the oversampling, the stored audio information is technically more precise than data recorded according to PCM, since there is no quantization to the sampling rate. In terms of reproduction, there is an advantage here, since steep-edged interpolation and anti-aliasing filters are superfluous, which were previously used on conventional CDs to separate frequencies above 20 kHz in order to reduce the sample frequency of e.g. B. to suppress 44.1 kHz.
Instead, the DSD format can be output directly because the resulting harmonic spectrum is far in the inaudible range. Due to the fundamentally low dynamics of a sigma-delta analog-digital converter with only one bit as the quantization stage, there is indeed enormous conversion noise , but this is shifted into the high-frequency range due to the effect of noise shaping . The signal can be understood visually in such a way that by rapidly alternating plus and minus on a loudspeaker with forward-backward phases of different lengths, the course of the audio signal is ultimately reproduced. Due to the inertia of the air and hearing, only the audio wave can be heard. In practice, however, there is already a band limitation in the playback system - but at the latest by the loudspeakers.
In the DSD64 format, the frequency response is extended to around 100 kHz, with a dynamic range of around 120 dB in the audible frequency range.
criticism
The data volumes to be stored in the DSD format are generally larger than in the commonly used PCM format, but this does not necessarily lead to noticeable improvements in sound. So far there is no practical evidence that the sound improvement claimed by developers and users of DSD over PCM actually exists. In a study by the Hochschule für Musik Detmold , the participants could not hear any statistically relevant differences between the data formats in corresponding blind tests. The authors of the study came to the conclusion that “even with the highest quality equipment under optimal listening conditions and with the most varied listening focussing or listening experiences of the test subjects, no significant differences between DSD and High Resolution PCM (24 bit / 176.4 kHz) are usually audible consequently the thesis could be made that none of the systems tested stand out due to their sound properties ”and refer to“ the high level of frustration experienced by many test persons, who were mostly used to professional and critical-analytical listening, during the execution of the tests felt and that they attributed to tonal differences that were nowhere near recognizable ”.
The reason can be assumed that the methods used in modern audio converters such as oversampling and phase modulation ("dithering") are sufficiently good for both recording and playback to avoid the potentially disadvantageous effects of early decimation of the data rate on a low sampling frequency to be heard. The filter technology in the chips in particular has made great strides in recent years. As a result, the amplitude and phase errors generated by the reconstruction filters are predominantly in the high frequency range, which is only slightly occupied by typical audio signals.
Today, sampling rates of 192 kHz with resolutions of 24 bit are used, which are not only considerably more accurate than the original DSD64 format, but also enable less steep filters with fewer errors to be used. It can be shown mathematically that a relatively smoothly tapering reconstruction filter, which is controlled with a 192 kHz sampling rate and therefore only has to block fully at 96 kHz, can easily transmit fully linearly up to well over 20 kHz and thus only theoretical improvements result if it operated at even higher rates. The pre-filtering of data integrated in the chips means that there are no significant losses.
From a technical point of view, however, DSD offers an advantage in terms of simplicity when processing different sampling rates, since it does not require the above-mentioned auxiliary filtering methods of PCM: Sample rate converters use the DSD format as an intermediate level by initially starting from a low sampling frequency go up to then filter the DSD data similar to analog data and finally resample or decimate it at the target sampling rate. If a finished DSD data stream is available, it is therefore much easier and sometimes technically more precise to decimate to any PCM sampling frequency than is possible from existing PCM data with a different sampling frequency, since upsampling then does not have to take place. Therefore, such DSD data are sometimes used in samplers for music production.
literature
- Thomas Görne: Sound engineering . 1st edition, Carl Hanser Verlag, Leipzig 2006, ISBN 3-446-40198-9
- Hubert Henle: The recording studio manual . 5th edition, GC Carstensen Verlag, Munich 2001, ISBN 3-910098-19-3
Individual evidence
- ↑ Dominik Blech, Min-Chi Yang: Aural comparison between DSD and high-resolution PCM . Diploma thesis at the Erich Thienhaus Institute of the Detmold University of Music. Ed .: Erich Thienhaus Institute of the Detmold University of Music. 2003 ( http://www.eti.hfm-detmold.de/lehraktiv/diplomarbeiten/untersprüfung-zur-differenzierbarkeit-digitaler-aufnahmverfahren-hoervergleich-direct-stream-digital-dsd-und-high-resolution-pcm-24bit- 176-4khz access = 2016-01-28).