The ABCs of DSD Downloads

A Primer on DSD Hardware and Software

Equipment report
Music servers and computer audio,
The ABCs of DSD Downloads

It seems like there’s always some hot-button issue for audiophiles to debate. For the last couple of years, it was the ability to play high-resolution (176.4 or 192kHz/24-bit) computer-audio music files. Now it’s almost impossible to find a DAC that won’t play files with at least 192/24 resolution, and many will play 384kHz/32-bit files, even though such files aren’t yet commercially available. This year, the hot-button issue is whether a digital audio system (server and DAC) will play Direct Stream Digital (DSD) files in their native format, without converting them to pulse-coded modulation (PCM) files first. A consortium of industry gurus devised a way to do that, and both server and DAC manufacturers have labored long and hard to produce DSD-capable playback gear. What has been missing until recently is a significant number of commercially available DSD music files to play on the hardware. We thought it might be useful to survey the field to see which sites currently offer DSD and what’s coming. We’ll also review what equipment is available to play DSD recordings without first converting them to PCM. But first, in case you haven’t been following this issue, let’s review a few basics.

What Is DSD?
Direct Stream Digital, or DSD, is a recording system used to master Super Audio CDs (SACDs). Although for a variety of reasons, the SACD wasn’t as successful as its developers would have liked, many recording engineers liked the sound produced by the DSD recording process. But until recently, playing back DSD files directly wasn’t easy unless you had professional equipment. Sony offered playback of DSD files on two of its SACD players and on its VAIO computers. SACDs are copy-protected, so (except for the aforementioned Sonys) they can’t be read by any other consumer equipment that can play pure DSD files.

A PCM (Pulse Code Modulation) encoding system samples an analog waveform from, say, a microphone many times each second. Each sample is a “snapshot” of the analog waveform’s amplitude at the time the sample is taken. The frequency at which the waveform is sampled is called the sampling rate. The waveform’s amplitude is encoded as a binary number (“word”) which in the case of CD is 16 bits long. Each 16-bit sample can encode one of 65,536 discrete amplitude levels. For high-resolution PCM files, the sampling rate may be as high as 352.8kHz, with a word length of 24 bits, or 16,777,216 discrete amplitude levels. With today’s computing power, even higher sampling rates and longer word lengths are possible, but so far these higher rates and longer word lengths have not been used (the limitations are the analog-to-digital and digital-to-analog converters). The resultant bitstream can be stored as files on a computer in a variety of formats, including WAV, FLAC, AIF, or M4A files. There are several other uncompressed PCM file formats, but the first three formats (WAV, FLAC, and AIF) can be played by all of the high-end music players I’ve tried.

DSD, on the other hand, uses a much higher sampling rate, 2.8224MHz, or 2,822,400 samples per second. That’s 64 times as fast as the sampling rate for CDs. But the word length is only one bit. If that bit is a “1,” the amplitude of the signal is increasing; if it’s a “0,” the signal amplitude is decreasing. The actual waveform is encoded by the frequency or density of the “1s” and “0s.” An actual sample of a musical waveform takes the form of a series of pulses of varying density (those “1s” and “0s” clumped together) and actually looks a bit like an analog waveform. When stored on a computer, DSD files have the file extensions DFF or DSF. If you’d like to delve deeper into how digital audio works, I’d highly recommend Robert Harley’s indispensable book The Complete Guide to High-End Audio. As a former recording and CD mastering engineer, Robert’s first-hand experience adds a valuable layer of practicality to the sometime dry theory.

In addition to the original 2.8224MHz sampling rate, current computing power allows sampling at twice that rate, or 5.6448MHz, and some recordings have been made at that rate. Sometimes these rates are referred to as DSD64 and DSD128, denoting multiples of the CD sampling rate. One DAC I know of, the exaSound Audio e20 Mk III, is capable of playing DSD files sampled at 12.288MHz, which they refer to as DSD256+. I don’t know of any files recorded at that rate, but someone will sooner or later push the envelope that far.

Why Would You Want To Play DSD Files?
The audio industry has figured out how to record and play back very high-resolution computer PCM audio files, and they sound pretty doggone good; so why do we need a yet another type? Why is DSD such a big deal, anyhow? There’s only one thing that would justify the trouble and expense of playing DSD files: if they sounded better than PCM files. Why would they sound better? The answer may lie in the playback

requirements. The CD needs a steep brickwall filter to rapidly attenuate frequencies above 20kHz. Raising the sampling rate, as high-resolution files do, reduces the steepness of the filter requirement (although it’s still relatively steep). But DSD files need a much simpler filter—basically just a low-pass filter like you find in many crossover networks. Such a filter should damage the sound less than a steeper one. Like many issues in high-end audio, there are many pros and cons about each of the recording systems, but what really matters to me is: How do they sound? After all, my basic audio philosophy is: If it sounds good, it is good. So while I’d encourage reading about different theoretical advantages of PCM and DSD, I’d really recommend listening to both to determine whether you think DSD files played back natively sound better than PCM files. If you do, you may want to consider upgrading your digital playback system.