Stephen Mejias

The News about DSD

Stephen Mejias  |  Nov 13, 2013

For many, the current hot topic in the world of high-end audio is Direct-Stream Digital (DSD), a method, developed by Sony and Philips, of digitally encoding an analog signal. The irony is that DSD is nothing new. The basis of the technology dates to 1946. Stereophile described it in “Industry Update,” as early as Vol.19 Nos.1 and 5, and again in Vol.20 No.9. And, almost exactly 14 years ago, in November 1999, John Atkinson went into greater detail, contrasting DSD with the more common Pulse-Code-Modulated (PCM) encoding used on CD:

The most straightforward way of encoding an analog signal as a Pulse-Code-Modulated (PCM) digital datastream is to use an A/D converter operating at the sampling frequency that puts out digital words of the desired length. If you are talking about the CD's 16-bit/44.1kHz data, an ADC samples the analog signal 44,100 times every second, each time describing the instantaneous signal amplitude to the nearest one of 65,536 voltage levels (2 to the 16th power). This is, in fact, how digital audio recordings were made up to the mid-'80s. But the complexity of the ADC increases almost exponentially with the number of bits required, and the critical demands made on the analog antialiasing filter needed to eliminate every trace of signal above half the sample rate are extreme. A different A/D paradigm was required to achieve resolution greater than 16 bits and to achieve more accurate 16-bit resolution at lower cost and circuit complexity. . .

Instead of trying to attain higher resolution by increasing the number of bits, it was thought: why not increase the sample rate instead? In the limit, if you increase the sample rate to a sufficiently high frequency, you can use a 1-bit quantizer: a simple voltage comparator that outputs a "1" if the analog signal level is higher than it was at the previous sampling moment, or a "0" if it is lower. Because this "delta modulation" technique uses a sample rate very much higher than the baseband audio signal, the requirements for a "brickwall" analog antialiasing filter on the ADC's input can be relaxed. You can then either feed the high-rate pulse stream to a simple low-pass filter to reconstruct the analog original, or you can use a low-pass digital filter to "decimate" the low-resolution, high-sample-rate data to derive the desired multi-bit, low-sample-rate data. . .

The elegance of the idea behind DSD is that this decimation filter can be eliminated. Why not, Sony's engineers thought, just store the output of a 7th-order noise-shaped delta-sigma modulator running at a very high frequency (in DSD's case, 2.8224MHz, or 64 x 44.1kHz) on an appropriate medium. For playback, this datastream could be fed, in theory at least, to a D/A converter consisting of just a simple low-pass filter.

The use of such a high sampling frequency would mean the ADC's analog antialiasing filter needn't be a brickwall type but could instead be a sonically benign low-order type; linearity would inherently be excellent; there would be no digital decimation filter, with its necessary mathematical approximations on either the A/D or D/A conversions reintroducing PCM quantization noise or time-domain dispersion problems; there would be no multi-bit DAC, with its possible performance compromises—this would be the closest thing to a digital topology with analog-like properties.

DSD encoding became commercially available with the SACD format, which, for various reasons best discussed over Buffalo wings and beers, never reached its full potential. At least for hi-fi enthusiasts, then, the news is that DSD has finally become practical: Mastering engineers have the necessary tools, record labels are beginning to release the music, and hardware manufacturers are responding with DSD-capable digital-to-analog converters that most enthusiasts can actually afford.

Elsewhere, in that same November 1999 issue, JA wrote: "I can't help wondering if sounding better is, on its own, sufficient reason for SACD to become established. To become dominant, a new medium needs to be different in kind, not just offer more of the same—even if that more is much better."

JA told it like it was then, while, at the same time, telling it like it is now.

As I see it, the problem with audiophiles is that they care too much about sound. Good sound is not enough; has never been, and will never be, enough.

CD succeeded, JA argues (and I agree), not only because it offered digital encoding, but because it provided random access, portability, greater longevity, and lack of surface noise—conveniences, mostly, and things largely unrelated to good sound.

DSD's day has finally come. The encoding is freed from the disc; things are now as they should have always been.

Why such fine news should be at all controversial is beyond me, but, hey, I suppose there is something in human nature that enjoys a good waste of time. And who am I to argue with human nature?

Interestingly, back in 1999, JA concluded his DSD discussion with these words:

The proof of any audio pudding is in the hearing, and in that respect DSD-encoding would seem to be beyond reproach. Every Stereophile writer who has auditioned DSD under critical conditions has found it both very much better than 16/44.1k CD and much closer to the analog experience.

Again, nothing much has changed.

In case you missed it, last week, AudioStream.com’s Michael Lavorgna interviewed Andreas Koch of Playback Designs, a proponent of DSD encoding and one of the early developers of SACD. ML and Andreas Koch aim for the heart of the matter. It’s an excellent read: informative, entertaining, and clear. Check it out.

COMMENTS
Music_Guy's picture
Submitted by Music_Guy on November 13, 2013 - 1:12pm

I truly want to enjoy the benefits of digital sound with quality better than 44.1/16 CD/PCM sources can provide.  I really do.  But, I am put off by the cost/complexity of added componentry (Cables, outboard DACs Music server computers, networks) beyond what I have now, that are required to get that sound improvement.   This is compunded by the variety of standards which threaten to obsolete whatever component I buy today.

You and others have captured it here and elsewhere that, in almost equal measures, the improvement in sound has to be coupled with more convenience before a new format takes over.  While the former has happened, the latter has not.

These next few months will be interesting....

Jason Victor Serinus's picture
Submitted by Jason Victor Serinus on November 13, 2013 - 4:09pm

Hi-res and DSD music file playback are not prohibitively complex. You already have a computer. All that is necessary to play hi-resolution audio, be it PCM or DSD, is after market music playback software, a USB cable, a DSD-capable DAC, and cables to connect the DAC to self-powered loudspeakers. If you currently listen to audio on your computer, you may already have those. A music server helps with file storage and retrieval, but is not essential for music playback. Although I'd love to have one to store the 20,000 CDs that now clog up our home, I've made do without one.

Audiostream.com is filled with how-to artices on how to get started with computer audio. There's one on the homepage now: http://www.audiostream.com/content/getting-started-computer-audio. And Michael Lavorgna has authored many other primers on the subject, all easily accessible on the site. 

As for the "variety of standards which threaten to obsolete whatever component I buy today," there is one commonly accepted standard for DSD file playback, DoP. You can read a very recent interview about the development of DoP in the following review: here: http://www.audiostream.com/content/dcs-puccini-plays-dsd. This interview with David J Steven of dCS complements the interview with Andreas Koch discussed above: http://www.audiostream.com/content/qa-andreas-koch

With the advent of Sony's four new file playback devices - the top level one is mentioned in the review I've cited above - Sony's release of a large number of master files in DSD format, and the recent addition of nativedsd.com and acousticsounds.com/superhirez to a list of major download sites that includes HDTracks, AIX Records, Linn Records, and Blue Coast Records, it is clear that major labels are embracing both hi-res PCM and DSD. The path to sound that betters the physical disc has been made clear. The time has come.

Music_Guy's picture
Submitted by Music_Guy on November 13, 2013 - 6:29pm

Cost aside, the Sony, non-amplified offering looks good.  Let's see...transfer existing medium-res and bit-perfect CD files to the box and add DSD and other new truly Hi-res music going forward.  Hmmm, just one box and some cables to the pre-amp. That ain't so bad.   Bye-bye CD player!

I had not given the latest Hi-res audio players a serious look before your post.  (I was stuck in the music-on-computer-player software-network-streaming-distribution-to-DAC-RCA cables-to-pre/power amp paradigm)

Things are moving faster than I keep up!  Thanks for the info, Jason!

- Cheers!

2_channel_ears's picture
Submitted by 2_channel_ears on November 17, 2013 - 11:45am

While one can certainly dip their toes with a minalist configuration, it's not free of problems or so simple if one desires greater detail.  I bought a Dragonfly and went through several pieces of software, and CODECs, before I got "hi-rez" out of it.  The software that "worked" crashes on occasion sporadically and leaves the file "skipping" in place over and over, sounding atrociously worse than any analog.  If I plug my monitor in to my laptop when the DAC is connected to my system, there's a hum I can't get rid of.  On top of all this, issues with metadata, and even with the hi-rez DL sites, there is a dearth of the music I want.

But to the point of complexity that Music Guy cites, I don't think it's so easy to wade through all the literature or figure out all the connections and options.  Audiostream is a good resource.  Local dealers have not been.  I'm getting there, slowly, but it's several monmths and I still haven't decided on a purchase beyond the winged marvel.

michael123's picture
Submitted by michael123 on November 14, 2013 - 11:08am

DSD by large is inferior technically to PCM, the level of noise and distortion rise after 20KHz..

DSD does not allow any kind of processing (like room correction), and most DACs internally convert DSD to PCM anyway (except 2-3 pieces)

 

Moreover, why do you push for DSD while 99.9% of the contents is PCM?

 

Both Playback and EMM, which I know quite good, actually do not reproduce real high-rez very well.. probably due to their upsampling algorithms, the sound is always 'soft'.

 

 

tnargs's picture
Submitted by tnargs on November 20, 2013 - 7:02pm

Correct on all counts.

DSD should be ditched from the marketplace as soon as possible.

The fact that Stereophile continue to talk it up, shows a remarkable long-term ability to promote the ripping off of the consumer, and the profit and other interests of the industry.

music or sound's picture
Submitted by music or sound on November 14, 2013 - 5:59pm

The only toy I presently have which can play DSD files is the Astell&Kern AK120. The few samples I got sound good but these DSD files are first converted to 88.2kHz/24 PCM. I assume most DSD capable players do not directly convert it to analog but first to PCM. 88.2/24 has only around 70% of the filesize of DSD. That some DSD downloads sound good may be more related to better recording and mastering than to the format. 

So why we need further fragmentation of that small niche market - I see that mostly some producer of hardware are promoting DSD (with the claim of superior technology). I am more interested in DSP which is not possible in DSD.

That some DSD sound good may have more to do that sound processing and mixing is restricted in DSD. Less processing and simple but expert recording technology can produce better recordings (e.g. MA recordings which some are recorded in DSD and other in PCM which I am mostly prefering).

Louis Motek's picture
Submitted by Louis Motek on November 14, 2013 - 9:18pm

 

DSD can only be directly converted on 1-bit DAC chips, and PCM can only be directly  converted on parallel resistor DAC chips, so it is not really possible to "compare the sound of DSD to PCM" directly. A vastly larger diffference in sound quality is due to the Jitter amount during conversion. So it is easy to set up a test where:

 

(1) DSD is made to sound like crap.

 

(2) PCM is made to sound heavenly in "direct" comparison.

 

(3) DSD is made to sound great.

 

(4) PCM is made to sound like crap in "direct" comparison. 

 

But if you are using the same converter for the test, you are never comparing DSD to PCM directly. You are only comparing one data conversion process to another.

 

If you compare these 2 formats on a universal player, you had better first figure out what DAC chip the thing uses, because if it uses a 1-bit DAC, then it will "make" DSD sound better and PCM worse. If it uses a ladder DAC (Parallel resistor) then it will "make" PCM sound better and DSD sound worse. For some reason, I am not finding reviewers talking about this crucial differentiation. Could be because they believe themselves that they are listening to PCM on a 1-bit DAC chip, or DSD on a parallel resistor DAC chip. But that's impossible. 

 

 

The digital format which undergoes the least processing prior to conversion will always sound best.

 

 

PCM through Ladder DAC = little processing = sounds good!

 

DSD through Ladder DAC = much processing = sounds as good as the processing doesn't corrupt the sound.

 

PCM through 1-bit chip = much processing = sound as good as the processing doesn't corrupt the sound.

 

DSD through 1-bit chip = little processing = sounds good!

 

 

So really it is most important to set up the test correctly by announcing the conversion intricacies instead of thinking "I am comparing DSD to PCM because I downloaded both versions". You almost never are comparing DSD to PCM "sound" directly. Because most of the time you are using the same DAC for the conversion. And if you DO go the extra step to make sure you listen to PCM on parallel resistor DAC and DSD on 1-bit DAC, then you have the added difficulty to explain to your audience that both DACs already sound different to begin with. 

 

=====

 

Also, if due to high sampling rate conversion any ultrasound gets through to the analogue outputs, chances are very high that the sound system in general is adding inharmonic distortion through intermodulation of the ultrasound with the signal which is in the audible bandwidth.

 

 

Louis Motek

music or sound's picture
Submitted by music or sound on November 15, 2013 - 10:40am

Quote: "The digital format which undergoes the least processing prior to conversion will always sound best."

That should also apply to the recording process. Pure DSD recording (no intermittent PCM mastering) played via DSD and PCM recording via ladder PCM DAC. Is there even any recordings available which have done that in parallel?

Geoff's picture
Submitted by Geoff on November 15, 2013 - 4:16pm

Jason - 20,000 CD's?

Music_Guy - Bye-Bye CD player OK when EZ to rip CD's. Those who've invested much in SACD purchases are SOOL unless they have the correct PS3, software and lots of time to figure out the work around. Not so ideal for many older audiophiles who'd rather listen to music than jump into rabbit holes.

Louis - Nice post. Thanks for the Hi-Res music freebies on your website. I have enjoyed the music after just a few clicks to download the files.

Listen to more music and get off the merry-go-round. Time really speeds-up after fifty. Einstein had no theory to support that (c; but just ask anyone over fifty.

Louis Motek's picture
Submitted by Louis Motek on November 15, 2013 - 7:59pm

Quote: "... any recordings available which have done that in parallel?"

All A/D sampling processes are 1-bit. I know of only one person on Earth who is working on an A/D sampling process using discreet resistors. There are no recordings done in "true" PCM directly. All are translated to PCM while being recorded. 

It is not the translation of the sampled data into PCM which is harmful, and it is not the translation of the PCM back to 1-bit DSD which is harmful. The only thing harmful is the sound of 1-bit DAC chips when using these to "evaluate" the sound of PCM. They will always do a real-time translation back to 1-bit before playing the stream.

Actually, vastly more harm is done by tasteless engineering (digital compression, etc. etc. etc.). So, agreed, just listen to music and be happy. 

I'm perfectly happy with 16 bit 44.1 kHz PCM. Quality does not lie in the numbers! When we made those recordings, we cheated. We recorded at 24 bits / 96 kHz PCM. Then did sampling rate conversion and dithering down to 16 / 44.1. But I have experiences under my belt which show me that actually recording directly at 24 bit / 44.1 kHz and then just dithering (noise shaping) to 16 bit is actually superior than any sampling rate conversion. No smart averaging algorithm can do what a real signal and a real clock "would have done" in the past. It can get close, but we don't like that now, do we?  ;)

 

Louis Motek

Jason Victor Serinus's picture
Submitted by Jason Victor Serinus on November 16, 2013 - 2:43pm

After reading the first of Louis' two posts to this discussion, I wrote Andreas Koch of Playback Designs to see if had any additional insight to share. Since Andreas usually steers clear of forums, he has given me the okay to post his reply:

Yes, ideally we would have analog converters for DSD and PCM that do not first convert the digital signal to PCM or DSD. In other words we would need a DSD converter and a PCM converter with identical power supply and analog output stages. That is utopic.


But we can also do 2 comparisons: 1 with a PCM converter that also accepts DSD, 1 with a DSD converter that also accepts PCM. Play both formats on both converters. If you then end up with a consistent story that either PCM or DSD sounds better then chances are that you are hearing the differences between the 2 formats. If the story is inconsistent then the converters are messing too much with the formats. It still would not be exact science, but probably the closest we can get to.

But this is only the easy part of the comparison test. Getting the same recording in 2 formats (PCM and DSD) is the tricky part, because the same story that applies to the DAC also applies to the ADC. Taking the PCM output from a ADC and converting that to DSD or vice versa won't cut it. So you would need a parallel recording PCM and DSD on 2 different converters - you can see the same problem here too. When you then take into consideration that most ADC's are delta-sigma, then the whole thing becomes questionable.

The whole reason why delta-sigma even got applied in audio is to make the ADC more predictable, easier and better. With that most digital audio recordings today spent at least a short moment in the DSD world - maybe not 1 bit, but something noise shaped. Removing the PCM downsample stage will give the advantage to DSD, but then again that was the reason why DSD (or delta-sigma) is used in ADC to begin with, because straight PCM converters couldn't achieve the same performance.

So setting up a scientifically exact comparison test between the 2 formats is impossible. If you include the ADC in the studio my experience is that DSD will most often win, just because it requires less processing from microphone to speaker. However, if the recording requires a lot of post production then there is a limitation in DSD tools and producers are then forced to use PCM. For those recordings a release in DSD is an extra step that may or may not give the advantage to PCM.

tnargs's picture
Submitted by tnargs on November 20, 2013 - 6:56pm

Koch has a major vested interest in one side of this debate, so his opinions should be dismissed out of hand.

Ignore him.

Louis Motek's picture
Submitted by Louis Motek on November 16, 2013 - 5:13pm

Quote: "[...] that was the reason why DSD (or delta-sigma) is used in ADC to begin with, because straight PCM converters couldn't achieve the same performance."

 

Ending this sentence with the added words "at the same production price" may provide a more encompassing overview. 

 

Certainly in the D to A converter market, the whole 1-bit revolution was feuled by low price. Certain parameters on spec sheets gave superior results to parallel resistor DAC chips. The enthusiasts knew it all along: discreet resistor DACs sounded better to those who had taken the trouble to compare. 'Utopian' comparison devices had indeed been built with both DACs onboard. Thus the huge dismay that flared up when discreet resistor DACs became harder and harder to obtain (Burr-Brown's PCM1704 being the famous example of a precious rarity for a time, especially the pre-selected ones marked PCM1704U-J and PCM1704U-K which not only measured better, but had substantially higher prices). When you consider what it takes to build a DAC with an expensive mono parallel resistor DAC chip (you need two for stereo), plus the needed oversampling chip (you need one), plus the associated parts on the circuit board, and you can substitute all that with a single stereo 1-bit chip for a lot less money, who's going to stop the trend? The engineers choosing parts looked at the spec sheets and then the price, and it was a done deal. This is the same trend that was set by the CD in lieu of vinyl: it's easier, cheaper and it measures better. Sound?? That's just an opinion. Can't build a business on that!

 

I still maintain that Jitter is the real evil here, and is the elusive beast of all digital.

 

If I ever get my hands on a discreet parallel resistor Analogue to Digital PCM-direct converter, I will be the first to publish some resulting files for public scrutiny. Although my hunch is not yet substantiated by experience in the A to D realm, for lack of such a device, the D to A realm already has me convinced on the merits of the (yes, expensive!) discreet PCM process.  

 

Having said all that, don't forget: just admit noise from the power line to either 1-bit or parallel resistor DACs and you can easily make either one sound worse than the other. 

 

Louis Motek

pablolie's picture
Submitted by pablolie on November 18, 2013 - 12:47pm

The advantages and disadvantages of any modulation technology are mathematical and theoretical. I think either can be made to sound perfect in the real world. As perfect as can be.

when audio discussions get into the realm of "I need perfect measurements from 5hz to 100khz" it's like... Don't even bother going to live concerts. Just *imagine* that Kant ideal of Das Ding An Sich that you shall never hear anywhere. It is an illusion.

the gatekeeper will always be the quality of the recording. It won't get better because of what tech you use in the backend. Both methods can be made to sound as close to perfect as you can get, and whoever claims they can hear a difference... Well, good for them...

oh. And turntables are so inherently flawed no matter what you plonk down on them that it is in no way ever a golden standard for accuracy. One thing we have to be honest about is the fact that while we debate technologies' accuracy in the end our personal preferences reflect a degree of non-linearity and artificiality we ought to stand by. Just like musicians will never be able to tune their instrument to produce every note mathematically perfect, we will not and should not claim audio enjoyment is about getting everything bit- and analog-perfect to the billionth behind the dot.

deckeda's picture
Submitted by deckeda on November 21, 2013 - 11:06am

1) Many new recordings are done 24/44.1 (or if done "better," are only released that way.) Just browse hdtracks.com to see what's available in new recordings.

2) Sony doesn't appear to be archiving analog recordings to DSD anymore. (eg, Kind of Blue's recent 24/192 conversion.)

Has anyone at Stereophile asked why, for either?

X