DSD Best Practices
After reviewing the technical and business history of Sony’s 1-bit PCM or DSD 64 format, it’s possible to describe the “best practices” associated with the format. And what is best for one type of music and production flow might vary as we contemplate other types of music and methods of recording. Each individual engineer, producer, and label has to sort that out for themselves, but they can’t deny the essential facts of the DSD format…although plenty try.
The original intended use of DSD 64 was for the archiving of the vast catalog of Sony Music. The engineers behind the Delta Sigma conversion technique never imagined that their innovation would find its way into the consumer realm as a delivery format. And remember the archives that they wanted to preserve are all analog tapes…not new high-resolution masters. This means that the fidelity “box” associated with DSD 64 (which turns out to be about 110 dB from 20-23 kHz) only has to be larger than analog tape…and third generation analog tapes at that. CDs can do that and DSD 64 can do that as well. I have no argument with the original use of the DSD 64 format. It’s an ideal archiving format for existing analog tapes. As far as I’m concerned it’s not high-resolution or high-definition but in the specialized case of archiving analog masters it doesn’t have to be. The preceding method is “pure DSD”.
A second methodology can be described as a real time analog mixdown to DSD 64. In the specialized world of classical and jazz record production, this method can work very well. There are at least a couple of companies that I’m aware of that have chosen to make recordings of live musicians (all playing and singing during a single pass) through an analog mixing console to a DSD 64 recorder or PCM high-resolution deck. Channel Classics and Bill Schnee’s Bravura Records both work this way (although I believe Bill avoids DSD and captures to high-resolution PCM).
During the Snow Ghost sessions that I was part of in Montana a few of years ago, we recorded Wayne Horvitz’s trio using this method as well. The microphones were sent through a big SSL Series 9000 analog console, eq’d, balanced, panned and adjusted during the live performance, and the stereo output bus recorded using DSD 64 (as well as analog tape and 96 kHz/24-bit PCM). It is critical that a “final master” sound is output from the analog mixing desk because you can’t go back and adjust levels, panning, reverb, or anything else after the fact. Simple edits are possible but nothing else. This method simply can’t work in most commercial recording situations. The result is a 99% pure DSD recording.
Both of the above methods maximize the sound quality of DSD 64. At that rate the specs are only slightly better than compact discs but if you prefer the “sound” of DSD, then fine. If you raise the rate to DSD 128 or even 256 the problems of excessive ultrasonic noise that can damage your electronics (amplifiers and other processors don’t like to get large amounts of ultrasonic frequencies at their inputs) goes away or is filtered out. Recording made using either of the previous methods can be spectacular but the limitations of these schemes make them very impractical. I know Michael Bishop makes his recordings using quad DSD. I’ve exchanged a couple of emails with him about doing an interview for a future post.
The final technique and the most common is to “cheat” and use very high sampling rate PCM to do all of the production work that can’t be done in native DSD. Some engineers do their recording at 352.8 kHz 24-bit PCM (deceptively called DXD to avoid any association with PCM), do all of their post production in PCM/DXD, and simply downconvert to whatever flavor they make available to their customers. I know Morten Lindberg at 2L does things this way. He prefers the sound of high rate PCM/DXD over DSD.
According to the guys at Grimm Audio, the best release format would be a PCM recording at 88.2, 176.4, 96, 192, 352.8, or 384 kHz and 24-bits…just like the source recording. In other words, don’t do any conversions. Converting to 1-bit at 128 or 256 would be second best with DSD 1fs or DSD 64 the last best option…the standard for SA-CD. While I disagree, they relegate Redbook CDs as the least desirable format to convert to. In reality, a well-done CD captures virtually everything that a DSD 64 can.
So there you have it…the unvarnished truth behind the DSD format. I’m not quite finished though. Tomorrow, I’ll explore the varied world of alternative production techniques and DSD delivery. It’s all about maintaining the myth…and maximizing money.
12 thoughts on “DSD Best Practices”
No question you know your stuff re DSD/PCM.
One only has to compare a quality SACD with a quality DVD-Audio disc to hear the different format characteristics. SACD has a silky, smooth, slightly warm character, thus the comparisons to analog. High-rate PCM simply sounds like music w/o the digital nasties. It’s back to the old saw; truthful reproduction and easy listening are not the same thing, and many audiophiles don’t want to know this. While I like the sound of sacd, 24/96 etc. done right is more faithful to the original sound. As we have shared previously, the hang-up w/ this tight definition of hi-res is that 80% (at least) of the music that folks would like to hear at master tape level was or is recorded on analog tape, and hearing a download that pretty much cloned the master is a thrill that the public should not be kept from having due to semantic issues. It’s like trying to put a foot in a shoe that is too big. It’s not the foot’s fault; shouldn’t have used such a big shoe. Shouldn’t have called it hi-res either. “First Gen Sound” would have covered all bases, everybody gets what they want. I have said it before; try and look at the big picture Mark, not just what the ocean looks like from one dock with one ship attached. Thank You.
The studies that have tried to establish a sonic difference between DVD-Audio and SA-CD have found not the “silky, smooth, slightly warm character” present in either format. In fact, the test subjects were unable to tell any difference. With regards to the percentage of real high-resolution audio available to consumers, you’re right that 80% of the masters (especially the older masters) are never going to be high-resolution. And that’s as it should be. Nothing wrong with hearing the studio masters in their original fidelity. No one, least of all me, is trying to deny audio enthusiasts from enjoying their favorite music at the best possible fidelity…I just prefer to be informed about the actual resolution of the transfers. It’s hard to spin things any other way.
I appreciate very much your clear words about the ‘myth’ of SACD!
Could you – maybe – make a comments on the ability and quality of the multichannel part of the SACD.
This might (?) be a reason for choosing a SACD version over a CD version.
I’ll try to get a post up about the multichannel aspects of SA-CD. I don’t believe the MCH capability has much to do with the format’s success or failure.
I for one would have zero interest in SA-CD if not for multichannel. I’ve done hours and hours of A/B testing to see if I could discriminate between SA-CD stereo and the stereo RBCD layer that’s standard on almost every SA-CD, and I couldn’t reliably hear any difference.
The multichannel aspect is a very positive one…just wish they chose PCM instead of DSD.
Dear Dr. AIX,
“In reality, a well-done CD captures virtually everything that a DSD 64 can.”
OK. But, how about a multi-channel SACD?
What are the options for a MC SACD lover like me?
All the best,
(from Salvador, Brazil)
I’m a multichannel advocate as well. I prefer to get DVD-Audio or Blu-ray discs for my surround music. Any SA-CD surround will most likely have come from analog originals sources. Just give the mixes in PCM.
Just to check things out, I downloaded an album from Channel Classics in DSD64. The recording was captured straight from the analog mixing board in DSD. No PCM processing or I wouldn’t have bothered. MY DAC does not do DSD but Audirvanna converts it to 24/176.4 for my DAC to process. Yes, it is a warm sound but lacking in HF detail and air. I compared it to a 24/192 captured recording of similar material which has more spacial detail and a less rolled off sound, which to these ears sounds more lifelike. I have been an audiophile kind of guy since the 80’s and I can see why some audiophiles love DSD as being older guys, their setups are too bright and the DSD sound fits right in to their systems. To me, the good PCM stuff sounds more like the live sound I experience at venues, especially small classical and jazz venues. On this one trial their is no doubt that the DSD recording (of very fine music I might add) sounds very pleasing, just not as accurate as PCM. The good news out of all of this is at least, with computer based music, we have choices to pick the particular flavor that we enjoy. Some like warm sound. I want as accurate to the live event that I can get. But at least we have choices. Keep up the good work Mark.
Jared and Channel Classics do a very good job at optimizing their DSD recordings by avoiding the messy intervening steps that most other labels suffer through with DSD/DXD etc. I’m with you in wanting accuracy, clarity and refinement in my recordings and the ones that I prefer listening to.
I think if one was to take an analogue master tape, give it to the same mastering engineer and have the final master laid to both PCM and DSD machines of equal build quality, anyone would be hard pressed to tell them apart. No reason for DSD to lack air or roll off the top end, it’s just as capable at capturing high frequencies.
Also, I don’t understand the rational of not wanting to buy a song in the DSD format that has first gone trough conversion from PCM but then think that letting the software on your PC do a conversion from DSD to PCM is better? I mean a studio grade converter will be better than consumer software I would assume, or at least equivalent. If you only have a PCM DAC, probably better to buy PCM recordings?
DSD over PCM frames I think results in bit perfect playback though right? May be this is what your software was doing?
There was an AES study that tested the perceptibility of DSD 64 vs PCM 192 and the results showed no one could. You hardware…that’s another issue. People advocate for DSD because it’s “warmer and more analog like”. It’s not…so why not stick to PCM, which we know works and avoids all of the problems of DSD?