Peeling Back the Curtain on DSD…Again!
“I got to ‘hear’ the Sony HD audio player today, feeding their $2K stereo integrated amp and a pair of Monitor Audio speakers. The demo material was a very old Beach Boys track, first played in glorious DSD, which sounded…um…ok…then from a ‘red book CD rip’, which sounded like it had a single-ended noise reduction system applied, softened transients, gated noise around short sounds, etc. The demonstrator claimed the red book ‘threw a veil over the sound’ and how the DSD version ‘opened everything up and improved clarity’. I just cringed.”
A reader sent the paragraph above to me a few days ago. I’m not surprised at the tactics employed by the demonstrator…I understand that everyone wants their stuff to sound the best. But if someone went to the time and trouble to juice the DSD and downgrade the PCM for a demo, then there’s something more sinister going on. It’s true that DSD is a “hot” topic and will be the savior of high-end audio if you listen to the writers at some of the print and online magazines. My very good friend Jim Merod of Blue Port Jazz is all about DSD. They’re uninformed or choosing to avoid the facts.
SoundMagic has introduced an inexpensive DXD/DSD High Resolution Music Workstation that is “affordable and portable”…less than $1000. I’ve heard from a few readers that think the editing tools that have been absent from DSD are now available using Serenade or Pyramix. It’s not true. These systems convert everything to high rate PCM to do things like reverberation, fades, EQ etc and then convert back to DSD for delivery…and raise the price to the extreme because the perception is that DSD is the ultimate “resolution format. It’s all a myth.
The article “DSD Myth” by the guys at Grimm Audio in the Netherlands might not be the most well written piece but it does a very good job of slicing through the clutter of misinformation that pervades the whole area of high-end audio with regards to DSD. I thought I would highlight a few of the most important facts that are contained in the article.
DSD is seductive. As they and others have rightly pointed out having a single bit at a very high sampling rate capturing and reproducing an audio signal that is very close to the analog original sounds wonderful. But only according to very limited conditions…the ones that Sony imaged for the format as it was first used to archive their analog tape collection. It was never intended to be a consumer format. That only happened after DSD was dismissed by the DVD Forum.
If you read the article…and you should…you’ll find a lot of relevant information in the first few of paragraphs.
“The problem is that the format alone does not tell the whole story. Pure, original ‘native’ DSD using truly 1-bit DSD converters for A-to-D AND D-to-A can sound stunning. If the converters are done right technically of course (implementation, layout, components, etc.) and if the subsequent amplification isn’t bothered by the relatively high level of high frequency noise specific to DSD [NOTE: the ultrasonic noise bothers EVERY amplifier].
But most SACD players and ‘DoP” capable DACs are not equipped with 1-bit DAC chips. And of the recordings on SACDs and DSD Internet downloads, the vast majority are not pure [NOTE: less than 15% in fact] native DSD but have seen many different phases, starting perhaps as 5-bits inside the A-to-D converter, being converted to 24-bit PCM for editing and ending as 1-bit DSD after mastering. Meaning they all have undergone one or more conversions including going back to 1-bit DSD, which not a lossless process in itself. If you want maximum sound quality once a recording is in PCM format, you’re better off listening to it in PCM than to a flawed DSD transformation of the latter.”
We talked about format conversions recently. Moving between formats does not improve the sound of recording. Analog tapes to digital transfers work pretty well because the fidelity of the original analog tape is less than that of the digital format. This was the thinking of Sony when they chose DSD for their archiving. DSD 64 does a very good job when used in the specific case of capturing the output of an older analog tape…there are no edits, no processing required etc. Perfect. In every other case, DSD suffers from conversion loss and bandwidth limitations due to the high frequency noise.
We’ll go through the rest of the article over the next couple of days. It’s really important that audiophiles realize what studio owners already know…DSD as a replacement for PCM is just not going to happen.
34 thoughts on “Peeling Back the Curtain on DSD…Again!”
I have been following your posts for several months now, and I feel very happy that there are still some sane, rational people left in the audio industry!
This DSD vs. PCM thing seems to just be the next marketing BS being pushed down the uninformed throats of audiophiles…
I am an engineer myself (computer science), as well an audio enthusiast and I’m really baffled by the second coming of vinyl and analog in general, as those formats are very much inferior to CDs in technical potential.
The only reason for vinyl sounding “better” than a CD (as I’m sure you know) is that because of the limitations of vinyl (how deep/wide the grooves can be) it cannot be made with such compressed dynamics, as CDs can.
That however is no reason to go back to a format with so many drawbacks and inconveniences – vinyl wears out quickly (10-15 playback sessions are enough to make a detectable difference), requires regular cleaning and maintenance, even then it pops and crackles, etc. I won’t go into magnetic tape as I’ve digressed enough.
My point is there must be a better way to let the studios know that the way they master records is crap…
On the topic of DSD – I really see no technical reason why it should better preserve the “analog” quality of a master.
Ultrasonic noise aside, all that matters in digital audio is dynamic range and the highest frequency that can be accurately digitized and in that regard even Redbook CDs are more than many peoples ability to hear and their playback systems…
If anything it’s exactly that ultrasonic noise and the need to filter it out that make DSD inferior to high-resolution PCM.
Ultimately there is good reason why DVD-Audio and SACD failed to become the next CD – most people don’t care enough or can’t afford (my case) the equipment necessary to hear the difference (if any) between a standard CD and “high-resolution” format (which is most often sourced from the same analog master, making any differences too small in the first place)…
I’m not saying that HR audio doesn’t have a future, but it will most likely stay a niche as you already suspect…
PS: I really enjoyed the free HR tracks that you shared although I’m not really into those genres of music.
I so much wish more labels would produce records like yours!
Best of luck in your endeavor!
Then Boris…it seems to me that you understand what I’ve been saying. Music production could be so much better than it is these days but labels don’t care. The public probably doesn’t care but I think giving people choices might see the return of better fidelity.
I have read a few of the comments here which are based upon a variety of experiences. I am not an audiophile, just someone who did the research on the web and combined with my aerospace engineering, physics, and math background and 3 decades in defense technology R&D (34 yrs performing with symphonies and quartets as a violinist) and what i ended up with was a 7.1 KEF Q series speakers combined with an Onkyo 705 amplifier, plus a lenovo yoga pro i7 laptop and the Teac ud501 DAC in a dedicated mini theater – all of this equipment is not even high-end. When taking CD audio and converting it to lossless FLAC and in real-time upsampling via Signalyst HQPlayer (software filters) to DSD64 (~2.8MHz/1bit) then to analog stereo input to the amplifier, all i can say is the music was Glorious, e.g., Engelbert Humperdinck, Frank Sinatra, Jazz, Classical genres, etc. The entire family including the dog stayed in the room listening to music for 3 hours, the first time ever seeing the family drawn to the music and music a few of us dont listen too at all (me included). I didnt spend years tweaking anything, just ttok my time putting things together since 2001 – in fact i hardly listened to music in the theater, we primarily watch movies (i also have SACD player Denon ud4010 which i probably only used a dozen times and paid too much for $1799 at the time – sounded really good via 6 analog outputs but too lazy to go get the discs and play them)! For some reason this new basic chain of technologies (doesnt include the SACD player at all) just blows away anything i have heard before. My only experience that was equivalent to this technology chain (complete 2 weeks ago) was at Goodwins Audio in Waltham, MA in 2004, where they had a pair of Wilson Alexander speakers $150K and played a classical cello piece and jazz CD in a specialized room made of wood panels, diffuser panels, etc., the music was astounding. I dont have any measurement equipment, i dont have time for that, and i dont have an audio engineering degree, but i do know that between the Goodwin Audio room that played CDs and the “technology chain” at my house playing CD upsampled to DSD then to analog out, that they both evoke the same sort of musical involvement in listening that you ask yourself if this sort of music reproduction is even possible and you are hearing it like at a live venue. The ages for our family ranged from 26yrs to 89yrs and everyone was glued to listening to music from the new “technology chain”. You get the feeling from the experience as though you had stolen something, next thing you know, im calling my brother about it, telling my friends, and coworkers and they get this puzzled look, what is SACD, what is DSD, do you mean DAC ro be digital analog converter…..and this is at the laboratory.
I would never argue that your personal tastes can prefer the sound of DSD over other formats. As long as you can appreciate and understand the DSD is inherently limited as a format for those of use that record and produce records. The sound that you achieve by converted to DSD is nothing more than a tonal modification and not a benefit of the format itself. You could easily EQ the output of your CDs to achieve the same result.
DSD has a certain amount of buzz…although it does seem to have dried up somewhat. I noticed a new album by Sound Liaison today in my email. They record to 96/24 and then convert to DSD…and charge 40% more. That doesn’t seem to me to be appropriate.
Thank you for retaining my post which of course was very subjective.
You are right absolutely right about DSD, i read your response! I’ve been learning from your website as well as Archimago’s blog on basic tests and graphs, fortunately with the same DAC.
I have been immersed reading significant numbers of papers on the web to include a few from Lavery Engineering. I’m finding their is a lot of misinformation out there.
This was posted at hd-audio http://www.realhd-audio.com/?p=3768#comment-283787
I forgot to mention that as a result of reading information on your website, i made a purchase decision to buy the DXD (PCM 384Khz/24bit) instead of the DSD128 Brahms Violin Concerto from High Definition Tape Transfers (HDTT). I saved a few dollars and because of DXD, i am sure was edited/corrected, i ended up very likely with a better recording. I liked the disclaimer note HDTT put with the various download selections in that the DSD recording was not altered and may retain the audio “blemishes” from the recorded tape transfer and format.
Found this link for Dr Waldrep’s ebook and bluray disc of samples! http://musicandaudioguide.com
Looking forward to buying this soon to add to analyzing my basic system a bit.
I came across this post on the web with an AES 131 paper from John Bailey…is it possible that the recording systems are flawed? And unusual discrepancies an indicator for further research? Click here.
John Bailey at post #58
I read the paper from the dropbox link and wont pretend i understand the paper at first read since I am unfamiliar with these DAW machines, but would guess AIX has similar production workflows like the rest of the recording and mastering industry.
I read the paper and would disagree with John’s assessment. His comparison and nulling approach depends on DSD (at whatever rate), which makes the whole exercise meaningless.
Thank you for your feedback on the paper. As i continue on my literature search (hobby) i am becoming more enlightened to why PCM is really the way to go for how.
I found this dated (1995) paper from Meridian Audio, although without equations, i do prefer seeing the math, but had some difficulty in wrapping my mind around all their explanations – of course by no means a Digital Signal Processing expert or an Electrical Engineer, my studies were BS AE and MS EM where we didnt dive that deep into circuit theory (1980s).
If there are any recommended audio processing related technical papers or a useful link i could start at, i would be glad to read them. At work leading a team doing OR analytic modeling and simulation-based studies for the lab enables me to dive into equations and parameters a bit – my at home hobby of dabbling with tools like Maple, Mathematica, ISEE Stella, Netica, Matlab and Simulink, Nutonian Eureqa (symbolic regression), Megaputer Find Laws (GMDH), Risk Simulator, and TK Solver 5 (not an expert in any one tool).
I read papers more than i use tools, in contrast to the younger generation who typically graduate out of universities with exposure to some of these computational tools where back in the 80s the lucky ones had an HP 28c calculator (me), Apple or IBM PC, or some late night access to a VAX mainframe with Fortran 77!
So any recommendations would be great! I wont bother this topic anymore since i am already quite a bit off topic.
Let me put my thinking cap on and see what I can come up with.
Why 1-Bit Sigma-Delta Conversion is Unsuitable for High-Quality Applications by Stanley P. Lipshitz and John Vanderkooy Audio Research Group, University of Waterloo Waterloo, Ontario N2L 3G1, Canada
Why 1bit has problems – a short proof by math and simulation
Thank you!!! A wealth of info, i am now reading your qualitative and quantitative posts, i will ask less minor questions, my apology for posting so much and taking up time..
Your interview here is very helpful in my literature search so i am reading the state of the art rather than old papers
Lab breakthru….microphone could capture ultrasonics by 32x? http://www.eurekalert.org/pub_releases/2015-11/iop-gmo112515.php
Hi! What’s your take on devices that take CDs (PCM) and convert them to DSD, ie the PS Audio Directstream DAC? Some commenters say that the conversion results in a stripping away of veils extant in PCM and recovers musical information theretofore inaudible. True? Sales talk?
Unless you turn it off the Sony HD players “upsample” everything to DSD. After reading your post about many DACs not really being 1 bit DACs I am.left wondering what is going on.
You’re right. The Sony engineers engineered their high-end player to convert everything to DSD. They included a John Mayer track on the players but he refused to allow them to convert his track to DSD…he doesn’t like it. So they rerouted the signal for his track around the DSD conversion. The whole 1-bit DAC thing is a myth…a sales pitch that makes for good spin. It’s all PCM in disguise.
When transferring from format to format in the analog domain, or making a copy in the analog domain, in order to avoid any noise increase the target system must have a noise floor that is 20dB below the noise floor of the source, assuming similar noise spectra in both. Worst case, of the source and target have identical noise floors the noise in the copy will build by 6dB. That’s true for every generation. For an analog tape to be digitized without noise buildup the noise in the digital system must be 20dB below the analog noise floor. That can be a bit of a task if the analog source is Dolby SR encoded/decoded, and the digital system is only 16 bits deep!
To avoid any spectral alteration in the copy the target system must have flat frequency response to beyond the maximum and minimum frequency ability of the source. For analog recordings, there isn’t a single frequency above or below which the system just stops working, there’s a roll-off at the top and bottom. However, once the response has dropped below -10dB (a perceptual halving of volume), the system could be considered unresponsive. That means that 15ips tape has to be digitized by something sampling at at least 48KHz or higher.
Clearly none of that is a real challenge for any HD audio system, PCM or DSD. However, nobody ever just copies an analog tape to digits, there’s always some kind of “post” that has to happen, even it it’s just a bit of trim editing. Can’t do that in DSD, so you’re back to PCM, after which, what’s the point?
No up-sampling ever improves anything, in fact the process can actually increase noise if not done right. Up sampling PCM to DSD doesn’t improve anything. How could it? Other than the (rather powerful) psychological effects of not understanding the differences and similarities between inter-sample interpolation and a good ol’ reconstruction filter.
Sony’s HD audio player stores everything “side-sampled” (my term) to DSD because they can, and want to transmit the “suggestion” that you’re hearing extremely high-rate audio (ignoring the production path completely, of course).
Note that the Sony HAP-Z1ES “Hi-Res Music Player” has industrial design that returns to an analog era of finely brushed aluminum panels with minimalist controls, and the matching (in size, shape, color and cost) TA-A1ES “Hi-Res Integrated 2-Channel Amplifier” carries the theme forward. It’s beautiful design, and no doubt has a huge impact on the perceived sound of DSD.
I’ll appaud Sony for making the Hi-Res line an inroad into HD audio, but it looks (and sounds) like way too much mirrors and magic. HD audio is an opportunity to build on integrity, mics to ears. No idea why Sony would miss that, when they do it so will with their 4K video products, going “lens to screen” with true 4K.
So does Pono rhyme with “Oh-no” or “Ono”?
“DSD is seductive. As they and others have rightly pointed out having a single bit at a very high sampling rate capturing and reproducing an audio signal that is very close to the analog original sounds wonderful.”
Yes, and the other thing that makes DSD seductive to audiophiles is the notion that it can be implemented with no ‘digital DAC’. That is, apply a simple analog filter to the digital playback stream and voila, there’s your analog signal. (DACs are generally considered evil because they are digital — you know the drill).
Well…not really. DSD still requires conversion back to analog and they aren’t using analog filters. Stay tuned…I’m going to continue talking about the article today.
That reader comment reminded me of what was discussed at the TC-HRA meeting at AES, back in October— what was that comment about ‘might as well clean up an Edison Wax Cylinder, up-convert it to 24-bit/96kHz, and call it HiRes’?
What the reader seems to be describing is what Bill Whitlock would refer to as the demonstrator spreading fairy dust all around, and then proclaiming it’s HiRes. Not a good demo if you go by the reader’s comment,… nope, not a very good demo at all. Disappointing of Sony to tolerate that.
Somewhere, somehow, there has to be something in the HiRes standard to distinguish and designate the difference between a HiRes recording and a “transfer” to HiRes. And maybe even the reverse for going from HiRes to “Red Book”.
I’ve heard you talk about taking a Red Book rip and up-converting to HiRes. And there does seem to be a level of bogus-ness about that if it’s not made clear that it’s a convert from a “Red Book rip”.
Any HiRes designation needs to be clear and un-a ambiguous.
So Mark, keep pushing the envelope. Thanks.
What is the benefit of capturing and archiving analogue tape as DSD 64 versus 24/192 PCM?
Can mixing consoles work completely in PCM and have the filters etc. producers want or do they have to go analogue for some effects?
I would imagine for a classical release, it could be mixed entirely PCM since no self-respecting record producer is going to apply effects (apart from levels perhaps) to a say orchestral piece.
Here’s the simple truth. Archiving to DSD 64 as Sony imagined for their analog masters makes sense (it’s the only case that does) for DSD 64 because there is no subsequent processing or “remastering”. The point was simply archive a deteriorating catalog of 2-channel analog tapes. It is preferable to archive to 192/24 PCM for a couple of compelling reasons. If you ever plan to have additional processing or “remastering” done to the tunes you can do it in the digital domain (DSD requires conversion or an analog mixing desk). The plug ins used in DAWs today are virtually identical to the rare and expensive analog outboard processors. And secondly, the PCM format can capture higher frequencies than DSD 64.
As far a classical releases, you’d be amazed at the amount of digital or analog signal processing that goes on. They use compression, reverberation, and restoration processing in classical recordings all the time.
I had no idea that classical music was processed so much.
What about the recordings you make (I like the Zephyr Voices disc I have). How much post processing do you do and how did you mix that?
I don’t process my recordings at all…no EQ, no artificial processing, no dynamics processing. The Zephyr recording consisted of 12 mono microphones with ambient mics in the hall. I mixed in stereo and surround.
…and long may you rule, Mark. How many new recordings have you made (and released) this year?
I’ve put together a sampler or two…but I don’t have the resources to produce any new titles right now. I have a couple of things that were never finished and will get to those soon. It’s pretty busy around here.
John Stronczer, chief designer at Bel Canto Design, has chosen an interesting compromise in marketing to all digital formats. In a paper he wrote, available on Bel Canto’s website, he does a nice job of explaining of how his DAC’s handle DSD. He doesn’t condemn DSD, but does a great job in explaining it’s place in the competitive digital market. Sometimes compromise wins the war. http://www.belcantodesign.com/pdfs/Optimal_DSD_Playback.pdf
Thanks Matt…I’ve seen this paper previously and applaud John for his approach. The makers of DACs understand that DSD advocates, reviewers, and publications are making a big deal out of the format…it gives them something to write about. Bel Canto Berkeley Audio Designs, Benchmark, and others would be jeopardizing their sales if they ignored the hype about DSD…even if they know it is a flawed format and contributes nothing meaningful to audio fidelity. Therefore they do the best they can to incorporate DSD into their designs so that they are “Swiss” with regards to formats. It makes sense. But it doesn’t make sense for those of us on the production site of the music recording chain…unless you simply want to ride the latest hot topic and charge more for your recordings or downloads. I’m not willing to compromise the fidelity of my recordings under any circumstances…they will remain high-resolution PCM for source to delivery.
Unfortunately, once certain audiophiles are told of the “superiority” of DSD, they just have to have it. They just refuse to listen and believe what their ears tell them. I received an email from Blue Coast Music today. One of the recordings they offer is from the SF Symphony, Masterpieces in Miniture. It is a native 24/192 PCM recording. Even they note that the original wav 24/192 files sounds best, which is what I downloaded, followed of course by the DSD file over the 24/192 FLAC. I will stick with PCM every time when the recording was originally captured in PCM. Frankly, the DSD tracks I have didn’t do much for me, especially when one considers the file sizes which are huge. I have a larger external La Cie drive, but the DSD file sizes are ridiculous.
Labels are businesses and have to follow the trends if they see a financial advantage. DSD is a mistake but it will persist for a while until the next hot thing comes along.
I attended a similar demo at TAVES. I believe the same PCM source was played with the DSD conversion and without. The DSD sounded better. I wondered if Sony had purposely bullt the player with a high quality DSD DAC and a very poor quality PCM DAC. As we know, you can discuss theory all you like but the actual implementation of the DAC can make a huge difference.
Wow…I’m surprised. It’s true that DSD can change the color of the sound…whether you think for the better or not is a personal choice. But it’s certainly not more accurate. I would have loved to checked this out.