The advantages of 1-bit recording
Take it from a reliable source

You may find yourself thinking, "if we release music on 16-bit CDs and 24-bit/96kHz sounds pretty darn good, why do I need 1-bit? Aside from the fact that 1-bit/5.6 can play back at twice the frequency range of analog tape (which is what we’ve all been after in the digital world, the resolution of analog), let’s talk about renowned mastering engineer Bob Katz’s "reliable source rule." As Bob so accurately states, while bouncing digital audio is touted as lossless, that’s not really the case. Each time a digital word passes through a converter or processing, such as digital EQ or compression, digital words become longer, and as such, numbers are crunched and bits are dropped in order for them to fit the selected multibit format. This dropping of digital audio information becomes most noticeable in the decay of long reverb tails. Instead of a smooth decay, you hear little disjointed bits of sound decreasing in volume. With the 1-bit high-resolution format, reverb tails fade smoothly, since no digital information is lost.

According to Bob, when saving a stereo mix to a digital format for mastering, it’s essential to start out with the best possible sound—a reliable source—thus ensuring the best quality final product. He also recommends that you do as little digital processing (number crunching) as possible prior to mastering (again, less math is more). If you’ve ever been in the position of remastering a 16-bit recording to 24-bit (with no discernable improvement), or dealt with magnetic tape degraded in storage, you’ll understand how mastering engineers feel about reliable source material. With the MR-1 and MR-1000 Mobile Recorders, Korg has built the most reliable source to date—and one that will be viable long after we’ve moved to higher resolutions for release.

Chomping at the bit

Most high-quality 24-bit A/D converters start out with a high-speed, 1-bit data stream to begin with. The1-bit stream is then converted to a multi-bit format by a decimation filter, which drops unusable samples. In converting to 24-bits, the decimation filter has 16,777,216 possible chances to screw up, whereas a 1-bit system only has two choices, 1 or 0. Naturally, the 1-bit system is going to be more accurate.

After the decimation filter does its job, an additional filter that limits frequency range to half of the sampling rate (Nyquist Theory) must be used to eliminate unwanted digital audio artifacts (noise) known as aliases. For example, at a sampling rate of 44.1kHz, a 22.05kHz filter is employed. Meanwhile, it’s been proven in the lab that while we do not hear above 20kHz, there is ultrahigh-frequency information that we do perceive and miss by its absence. With digital systems limiting the frequency range to 22.05kHz, we lose that information and all the "air" and nuances that come with it. Beyond that, the design of this filter does have a direct effect on audio quality in terms of phase, linearity, and transient response based on the math being done, which in turn is dependent on code written for the decimation filter. This can be likened to the way the same mic will sound different from one preamp to another.

In a 1-bit system, the audio is recorded at super-high sample rates, commonly 2.8224MHz (MR-1), and 5.6448MHz for the MR-1000, which is twice the DSD (Direct Stream Digital) standard. At this high rate the MR-1000 is able to reproduce frequencies up to 100kHz (magnetic tape can only reproduce up to 50kHz) and provide superior transient response (see fig. 1). The ultra-high sample rate also eliminates the need for filters that stonewall the frequency range. Another advantage provided by frequency content above the range of human hearing is the elimination of phase issues in the audible frequency range, resulting in more realistic sound reproduction. Better yet, the decimation filter is also eliminated during the recording process, as are interpolation and oversampling filters, which use estimation processes to reconstruct the data for playback. With all of these filters and processes that affect sound quality and accuracy removed, what comes in the MR-1 and MR-1000 Mobile Recorders, goes out—with no extra math in-between.

Figure 1:

For even more information see: Future Proof Recording Explained by Korg (pdf file format)

---