musicandaudioguide
  • Making the DSD Sausage: Different Recipes

    My recent discussion of Sound Liaison and their DSD releases got me thinking about all of the ways that DSD recordings are produced. A subsequent email from the company, indicated that they take their original 96 kHz/24-bit PCM recordings and convert them to DSD 64 through an analog process, "We do a D/A/D transfer using high end converters" was their response.
    As we already know, any conversion degrades the fidelity of the original sound. The method used by many companies is to take perfectly wonderful high-resolution PCM recordings and play them back using the best digital to analog converter available. The analog signal is then immediately converted back to a digital format, except this time it's a PDM (Pulse Density Modulation format) 1-bit DSD file running at 2.8224 MHz (64 times the sample rate of a standard CD...which means absolutely nothing when it comes to "resolution"). This newly converted file is then prepared as both DSD-dff and DSD-dsf files for download sale...usually with a premium price tag.

    However, this is not the only way to produce and release DSD recordings. Most of the SACDs and DSD files available for downloading started life as analog recordings. The last time I looked at the catalog of DSD recordings, over 85% of the available albums were transfers from older standard definition masters. So recipe number 2 for creating a DSD file is an archive of a preexisting recording. And the fidelity of the new "more analog like" DSD master? It's the same as the second or third generation tape from which it was made.

    The DSD vs. high-resolution PCM "shoot out" that I was involved in at Snow Ghost studios in Montana a couple years ago (you can read my post by <a href="http://www.realhd-audio.com/?p=629" target="_blank">clicking here</a>) demonstrated another method of recording a DSD project. The source microphones were routed through an analog SSL console to a 24-track DSD 64 Sonoma system by way of Meitner analog to digital conversion. When it came time to mix the multichannel DSD tracks to stereo, these same tracks were converted back to analog and blended in the SSL console (which was a special version with high-frequency low pass filters included to help with the ultrasonic noise inherent in the 1-bit encoding process). Finally, the stereo bus outputs of the SSL analog console were digitized once again to DSD.

    Along the way, each of the analog tracks was subjected to some additional signal processing. During a mixdown session, it is common to add a little EQ or compression. Even if neither of these were used, a mixer will always add a little artificial reverberation (studios are relatively dead acoustically). Would you be surprised to learn that ALL of the outboard reverberation units are PCM based? It shouldn't because we know that DSD cannot be used to create reverberation. In fact, on the Merging Technologies website, they acknowledge, "DXD is a 352.8 kHz/24bit audio signal, which is used to be able to seamlessly move a DSD signal into PCM (<strong>standard method of digitizing audio</strong>) world so EQ, dynamics and other effects processing can occur." One of those other effects is reverberation. Many labels take this route to produce DSD releases. It's crazy!

    I can't figure out why any producer would subject their recordings to so many conversions and modifications. Each conversion step degrades the fidelity of the final product. Maybe this is why people like the "sound" of DSD.

    The methods discussed above are not native DSD recordings although some high-end labels would have you believe they are and they charge a premium for these converted DSD files. Native DSD recordings do exist...but they are extremely rare. My best guess is that there are less than 1000.

    We'll explore this further tomorrow.