Digital Music Purity: PCM vs DSD
The audio technology used for recording music has been improving for decades. Now we are close to fully transparent playback in the home.
Digital music today comes in different file formats – notably distinguished by the file extension (or file format) such as MP3, MP4, WAV or others. The files contain digital data which represents the original analog music encoded in a digital form. The digital encoding is performed at different bitrates using ‘bps or bits-per-second’ nomenclature like 128k, 320kbps which refers to the level of compression used. The ultimate bitrate is called ‘lossless’ and preserves all the detail of the music. Additional to bitrate are the terms ‘sampling frequency’ and ‘sample size’ – which all together give music its digital basis in resolution and fidelity.
Most music is sampled through a mechanism called PCM or Pulse Coded Modulation. This is where the analog waveform is converted to a series of numbers representing the varying voltage level. The rate of this sampling (sampling frequency) and the resolution of the number used at each sample (sample size) determine the fidelity of the encoding and has increased through the years. In the 1970s, the CD format was invented with a sampling frequency of 44.1kHz and a sample size of 16bits – shorthanded as 16/44.1. DVD Audio in 1990 supported 24/96. Bluray supported 24/192. Today, a sample size of 32bits and a sample frequency of 384kHz is considered the maximum needed to capture all the detail and nuance of the original recording.
Sony Corporation launched SACD in the late 1990s and also developed an entirely new audio encoding scheme based on a single bit encoding at very high frequencies. They showed that it was possible to accurately encode an analog waveform using a sampling frequency of 2.8224 MHz – which is 64 times the CD sample rate of 44.1kHz. This format was called Direct Stream Digital (DSD) and became part of the SACD standard. DSD makes use of noise shaping quantization techniques to push noise up to inaudible ultrasonic frequencies. This gives the DSD format a greater dynamic range and wider frequency response than CD.
The DSD format file has become an industry standard with the extension DSD (and also DFF and DSF) and increased sampling rates of 5.644 Mhz (DSDx2), 11.289 Mhz (DSDx4) and 22.579 MHz (DSDx8). These extreme resolution formats also have a commensurate large storage requirements: DSD files are regularly gigabytes in size making them impractical for streaming. Also, the nature of DSD makes it difficult to use during the music editing process: most DSD content is mixed/edited and mastered initially in high resolution PCM then converted to DSD.
There is certainly much debate in the industry over the benefit and need of the DSD file format – with supporters and detractors arguing which is best as a mastering format. From a theoretical perspective, both DSD and PCM can be said to provide all the information needed to recreate the original analog waveform. Both formats have wide industry support in software (media players, recording workflow) and hardware (A/D, D/A, DACs, chips). There is, however, some recent evidence that technical implementation details favour PCM over DVD to deliver maximum fidelity.
One of the more significant products launches of the past few years was the Chord Electronics DAVE DAC in 2015. This product departed from the existing path of D/A design emphasis on tone fidelity toward one that concentrated on preserving transient response and eliminating noise. The Chord DAVE is an all-PCM device and the products’ designer, Rob Watts, makes a compelling technical argument that only PCM encoding is suitable when reconstructing the original waveform. He argues that DSD has an upper limit on fidelity during this waveform reconstruction and is inherently most susceptible to noise degradation. Moreover, he shows through human hearing trials that while DSD can sound smoother and is comparable in overall detail, it displays weaker depth perception and a less audible sense of reality than PCM.
Some of Rob’s arguments suggest that the human ear/brain are capable of resolving noise floor modulation below 350dB and waveform jitter smaller than 88ns. These are numbers at the limit of machine measurement and are cited as being incredulously small – with many in the industry refusing to believe that our hearing is that sensitive. Yet, his designs are highly successful in the market with many hundreds of thousands of listeners agreeing that his approach to digital decoding is currently the best – and that places a vote for PCM.
In the search for Digital Music Purity – the ultimate in resolving the original analog recording in the home - the battle may be won by a combination of technical details and customer acceptance. Recent development have shown we may be closing in on what even a seasoned audiophile considers playback which is fully transparent to the original. We can only hope that affordable products with such a level of quality may be available to the average consumer within a few years.