Remaining Considerations on DSD

[Saidera]

DSD proponents tend to only see aspects of the theoretical benefits and infer the superiority of the format. And subjectively, at least if you record to DSD using well designed hardware and play it back, it is apparently quite good but who has that hardware!

A full comparison of DSD and PCM is not really attempted by any book or article, even if it is narrowed down to a basic comparison of characteristics of the formats. Most explanations proffered to consumers tend to only grapple with the surface of the engineering concepts. Mr Nishio’s brief article from 2016 also was not permitted the amount of depth it should have had.

There is DSD recorded and kept like that, DSD converted from PCM, and DSD that is then converted to PCM. This gives rise to questions: is pure DSD superior and why? Is DSD playback of PCM (Sony still included the DSD Remastering Engine via FPGA on their products as well as HQPlayer) sources superior and why? Is PCM sourced from DSD better than PCM that stays as PCM?

There are also the questions: how does Cirrus, AKM, ESS, TI, BB do DSD? How does CXD9872, ALC889DSD, S-Master HX all three of which involved Sony’s designs, play DSD? It can be guessed that at that time, a 1 bit DAC may have been incorporated in VAIO. How do custom DACs work?

Although many assume that DSD is simple, and uses just an analog FIR filter followed by a 1 bit DAC, were there not jitter issues? There are simply too many misconceptions surrounding DSD audio. And if PCM will end up with high frequency noise after passing through the DAC conversion process, are they the same or is DSD arguably superior? Although DSD tends to bypass the interpolator (8 times oversampling and filtering), in Cirrus chips there is the DSD Processor and then multibit delta sigma modulator and then a DAC to go through. All in all, what really is going on? Is anything objectively comparable at all?

I remember the shock many years ago when I learned that actually, there are no 1 bit DACs except in very old chips. And that these modern chips do a ton of things to DSD. So DSD suddenly became close to pointless.

Also, where are the latest efforts to get DSD editing finally possible? Waseda University (which created the format container .wsd to accompany dsf and dff) has been quiet for years now.

Ultimately, having forgotten the reasons why DSD wasn’t all that big a deal, a complete weighing of factors regarding the formats and playback pathways is much desired to come to the absolute final conclusion that utilising any of the three (or four) forms of DSD has any effect on sound quality. And because it’s near impossible to do this subjectively for most people, it has to be done on ASR. And the plain answer in general is: no effect on sound to warrant writing about.

Consider:
https://www.audiosciencereview.com/forum/index.php?threads/choosing-a-dac-for-dsd-playback.7701/
https://www.audiosciencereview.com/...ements-of-holo-audio-cyan-dsd-dac.6992/page-2
https://www.audiosciencereview.com/...ing-d90-balanced-usb-dac-review.10519/page-41
https://www.audiosciencereview.com/forum/index.php?threads/the-sound-quality-of-dsd.14773/page-2
https://www.audiosciencereview.com/.../asr-myth-regarding-discrete-circuitry.12064/
https://www.audiosciencereview.com/...dsd-pins-unused-or-dsd-mode-turned-off.23444/
https://www.audiosciencereview.com/...-based-dac-adc-which-uses-realtek-asio.23443/
https://www.audiosciencereview.com/...s-music-jazz-at-the-pawnshop-naxos-dsd.23137/
https://www.audiosciencereview.com/...und-liaison-high-res-format-comparison.23232/
https://www.audiosciencereview.com/...cm-with-room-correction-or-dsd-without.20023/
https://www.audiosciencereview.com/...read-dedicated-to-cirrus-logic-cs43131.23138/
https://www.audiosciencereview.com/forum/index.php?threads/digital-filter-game.23795/#post-810355/
https://www.audiosciencereview.com/...of-redbook-16-44-pcm.22102/page-4#post-778996
https://www.audiosciencereview.com/forum/index.php?threads/dsd-is-it-of-any-value.10682/
https://www.audiosciencereview.com/...-dacs-have-low-specs.20867/page-2#post-691547
https://www.audiosciencereview.com/forum/index.php?threads/dsd-playback.11476/page-2
https://www.audiosciencereview.com/...iew-and-measurements-of-korg-ds-dac-100.6134/
https://www.audiosciencereview.com/...-the-mighty-pcm-d100-one-bit-ambitious.24506/
https://www.audiosciencereview.com/...f-information-on-dsd-playback-pathways.24505/
https://www.audiosciencereview.com/...m-dsd64-via-64-bit-fir-filter-32k-taps.24504/
https://www.audiosciencereview.com/forum/index.php?threads/why-isn’t-mr-ayataka-nishio-credited-with-creating-dsd-audio.24503/

 

[Saidera]

But after seemingly consolidating the DSD material on ASR, and posting in quick succession, I have reread the old DSD thread I initiated, and I can confidently say that it is the most comprehensive, most technical and well-written thread on DSD there has ever been on the Internet. Of course, it doesn’t address every issue, and even if it does, it is not arranged in a highly user-friendly way. But its quality is so high with what it manages to cover. Yet there are many concepts a layperson would not understand completely, and other areas of interest which have not been more deeply explored yet.

Now, rather than general concerns on DSD and the format and its evaluation, I turn towards DSD as a directly recorded file, as a playback format like on HQPlayer, and as a source from which PCM is derived, most commonly in archives. It is possibly impractical to determine whether recording in DSD is better. When the inability to edit, and the sound quality which subjectively is no different is factored in, it renders any objective differences moot. But in fact some degree of editing is possible, and many find the sound to be different, though obviously less so compared with high rate PCM. The way an ADC encodes PCM and DSD should be different, and is there any evidence suggesting that DSD recorders result in slightly different sounding files? Playback in DSD through HQPlayer, which might be useful, but why is it? What would make SBM Direct (PCM derived from DSD) superior to a pure PCM file? Does this relate to the different way a DSD recorder encodes sound?

In all three questions there is a major issue. None of them can be answered properly while they are framed in such broad and general terms. The answers will vary depending on specifics, and at different times throughout the history of digital audio the answers would vary. The rewording of the third question could be like this: Is SBM Direct CD WAV file superior to a non-SBM Direct but still DSD derived CD WAV file but not any CD (for that would bring in mastering differences and too many unknown variables). But then is this still a DSD question or a question about the quality of SBM Direct dithering versus other dithering methods?

What is definite however is the rejection of any claims that there is a major benefit in DSD in general. Indeed, given high rate DSD and PCM, the similarities are greater and a few additional factors will point towards PCM being better. Objectively, there is too much evidence to favour PCM in general. Subjectively, the differences are unlikely to be massive or conclusive enough for the mainstream adoption of DSD.

As a matter of preference there is definitely a bias that DSD equates to higher quality sound, not so much due to the format though. It could also be the misconception that perhaps classical music would benefit from the DSD format. Because I entered the world of audio due to DSD, SBM Direct, and DSEE HX, it is difficult to allow any aspect of the three trademarks remain unclear. HQPlayer’s DSD, the DSD pathways in audio chips, and Sony’s SBM Direct are three areas where the information is not clear. Consumers who have blindly adopted trademarked technologies and are interested, need to be able to know what these words really mean, from their origins right up to the detailed inner workings of software and semiconductors, as far as is practicable. Then an informed decision will be easier to make. And perhaps something better can be created for the future.

 

[Head_Unit]

I once asked famed classical recording engineer John Eargle about DSD versus PCM. His answer was that he felt 96/24 PCM was quite capable of carrying all the sound in a recording (he didn't say anything against DSD, he just appeared to feel it was no needed because the PCM was sufficient). On another occasion at a surround sound conference, I asked guys from Mobile Fidelity: they answered that they had set up parallel pure 96/24 and DSD chains, and the felt the DSD a bit closer to the microphone feed. However, in light of all the items mentioned above, and the fact that a lot of SACD or DVD-Audio titles are actually being scraped off ancient analog tape, I think all the argument about format is silly. Personally I wish stuff would come out on Blu-ray or DVD-Audio (assuming DVD-video layer) purely due to compatibility-now and especially into the future I think it will be harder and harder to find universal disc players.

 

[DimitryZ]

I once asked famed classical recording engineer John Eargle about DSD versus PCM. His answer was that he felt 96/24 PCM was quite capable of carrying all the sound in a recording (he didn't say anything against DSD, he just appeared to feel it was no needed because the PCM was sufficient). On another occasion at a surround sound conference, I asked guys from Mobile Fidelity: they answered that they had set up parallel pure 96/24 and DSD chains, and the felt the DSD a bit closer to the microphone feed. However, in light of all the items mentioned above, and the fact that a lot of SACD or DVD-Audio titles are actually being scraped off ancient analog tape, I think all the argument about format is silly. Personally I wish stuff would come out on Blu-ray or DVD-Audio (assuming DVD-video layer) purely due to compatibility-now and especially into the future I think it will be harder and harder to find universal disc players.

Quick question.

Is hires LPCM noise shaping similar to DXD which pushes the noise to 50KHz-150KHz band and a high level?

With audiophiles demanding DXD masters as the "ultimate fidelity" it seems that level of noise shaping is a bad idea for an actual distribution format (which DXD was not meant to be, of course)..

 

[tmtomh]

Quick question.

Is hires LPCM noise shaping similar to DXD which pushes the noise to 50KHz-150KHz band and a high level?

With audiophiles demanding DXD masters as the "ultimate fidelity" it seems that level of noise shaping is a bad idea for an actual distribution format (which DXD was not meant to be, of course)..

DXD is high-res LPCM.

 

[Blumlein 88]

I've written my opinion before on hires and various formats. If there was a substantial difference this far down the road (40 years) we would have long before now determined it definitively. I've my doubts about the paper a few years ago which was a meta-study of various tests of redbook vs other formats. I'm told by those doing such things that well done meta-analysis is one of the stronger ways to find something out. The conclusion was there is something to higher rate audio vs redbook. However, such a metaanalysis would only have been needed if the difference is very hard to tease out.

https://qmro.qmul.ac.uk/xmlui/bitst...High Resolution 2016 Published.pdf?sequence=1

So my opinion is maybe, maybe there is small benefit at the highest levels if comparing 16/44 to 24/96, but it is a damn small difference. Redbook must get at least 99% of what is audible. The chances any more than 24/96 benefits anyone has to be vanishing small, and likely non-existent. Add to it virtually any other aspect of the recording and playback will swamp such differences, and one is left thinking this is even worse than fretting over DACs with 120 db Sinad vs 123 db Sinad values.

DSD is a little different thing as it isn't PCM. It is however a pox on audio, as recording and playing back in DSD all the way is effectively not happening for consumer music. If it goes thru a PCM stage and sounds different, if it sounds different if you convert from PCM to DSD then it can only be because it has a detectable coloration of its own. So it isn't better.

 

[DimitryZ]

DXD is high-res LPCM.

I didn't know this - I thought that it was a particular flavor of DSD. I see that you are correct.

Does this mean that hires LPCM is similarly noise shaped - as shown in @amirm ASR"s hires explainer video?

 

[dadregga]

This forum (and some other resources) have made it pretty clear to me that there is nothing DSD can do that > 44.1k/16 PCM can't do as well or better, and PCM has a huge advantage in terms of market support. As mentioned the theoretical simplicity of DSD DAC design is moot in the face of decades of iteration on excellent PCM DAC design. All the rest is mastering differences.


I do still personally prefer to buy high-res/multichannel music on physical discs, and I wonder why boutique companies still manufacture SACDs - surely it can't be cheaper or more cost effective than mastering and manufacturing a Blu-ray audio disc, even if you factor in the obscurity-hound factor.

Perhaps there are extra licensing fees involved in publishing Blurays? Surely not.

 

[xaviescacs]

But aren't some recording studios using DSD and SACD simply because it's the only way available to physically ship the 24/96 content?

 

[dadregga]

But aren't some recording studios using DSD and SACD simply because it's the only way available to physically ship the 24/96 content?

How on earth could it be the only way when your garden variety Bluray disc will happily support any number of 24/96 or 24/192 PCM streams?

Especially now, it has to be much easier and cheaper to burn and master Blurays than SACDs.

 

[MakeMineVinyl]

DSD looks good on paper, but the reality is that even Paul McGowen admitted in a recent video that when they do any editing (even simple gain changes) on their DSD masters, it has to be converted to PCM to do the edits and then re-converted to DSD. Kind of makes DSD a moot point.

Speaking from my own experience, the simple variances in adjusting the bias on an analog tape machine routinely done before recording sessions makes far more real and measurable difference in sound than there ever could be between DSD and PCM - yet nobody ever screamed about that.

 

[xaviescacs]

How on earth could it be the only way when your garden variety Bluray disc will happily support any number of 24/96 or 24/192 PCM streams?

Especially now, it has to be much easier and cheaper to burn and master Blurays than SACDs.

Ok. Right. But an audiophile ins't going to buy a Bluray. I mean, if you go to s store and ask for music, they show you CDs and SACDs, not Blurays.

 

[Saidera]

No, my questions simply cannot get an answer at this point in time,
since DSD as a directly recorded file compared to directly recording PCM, isn't really prevalent out there, and the theory of the ADCs don't appear to be comparable.

And DSD as a playback format like on HQPlayer may have measurable differences for certain DACs, but on the whole, there is nothing to suggest that DACs can process DSD better than it does PCM at this point in history.

DSD as a source from which PCM is derived, most commonly from archived DSD64, may be a fine option, as my SBM Direct thread will show. But it hinges on the question of whether 24/96 converters are comparable with DSD to CD converters. It is common sense that 24/96 can be converted to CD without any degradation in quality. But why are some still leaning towards DSD to CD conversion - is it mathematically superior?

Finally, DSD is now prevalent because so many chips support it now. And it is cheap and affordable to get such a DAC chip device, like the now rare Meizu Hifi, whereas in 2018, you'd still have to pay around $100 for a DSD capable DAC. Although there were DIY boards from LG phones, and other random stuff.

 

[tmtomh]

I didn't know this - I thought that it was a particular flavor of DSD. I see that you are correct.

Does this mean that hires LPCM is similarly noise shaped - as shown in @amirm ASR"s hires explainer video?

What’s the issue/problem with noise-shaping? It’s just a form of dither that increases the effective dynamic range (aka reduces the noise floor) of the source.

 

[MakeMineVinyl]

- is it mathematically superior?

Does it really matter to anybody other than a mathematician? And what does it have to music?

 

[Kal Rubinson]

No, my questions simply cannot get an answer at this point in time,
since DSD as a directly recorded file compared to directly recording PCM, isn't really prevalent out there, and the theory of the ADCs don't appear to be comparable.

And DSD as a playback format like on HQPlayer may have measurable differences for certain DACs, but on the whole, there is nothing to suggest that DACs can process DSD better than it does PCM at this point in history.

DSD as a source from which PCM is derived, most commonly from archived DSD64, may be a fine option, as my SBM Direct thread will show. But it hinges on the question of whether 24/96 converters are comparable with DSD to CD converters. It is common sense that 24/96 can be converted to CD without any degradation in quality. But why are some still leaning towards DSD to CD conversion - is it mathematically superior?

Finally, DSD is now prevalent because so many chips support it now. And it is cheap and affordable to get such a DAC chip device, like the now rare Meizu Hifi, whereas in 2018, you'd still have to pay around $100 for a DSD capable DAC. Although there were DIY boards from LG phones, and other random stuff.

OK. But why are we wasting attention on these, imho, settled issues? The marketing of DSD files has not compensated for the withering of the SACD business, the acknowledgement that all new recordings are made in some form of PCM (although labeled as DXD) and the near ubiquitous ability of modern DACs to handle both DSD and PCM confirms this.

There will always be those who are influenced by market talk and trends to make irrational purchase decisions and there will always be those who are happy to take advantage of such people.

Given that, the only real matters are the skill of producers/engineers and honesty/transparency of the entire supply chain with regard to provenance.

 

[Vincent Kars]

I didn't know this - I thought that it was a particular flavor of DSD. I see that you are correct.

The problem with DSD is that you can't edit it.
This makes any post-production impossible.
Hence they invented DXD, 24 bit / 352.8 kHz PCM.
Convert you DSD to DXD, do your edits, convert back to DSD, sell it as DSD.

 

[DimitryZ]

What’s the issue/problem with noise-shaping? It’s just a form of dither that increases the effective dynamic range (aka reduces the noise floor) of the source.

Usually nothing, as the distribution format will leave the high ultrasonics behind.

However, audiophiles now demand native DXD as the the "ultimate fidelity," which creates a potentially unfriendly environment for their equipment.

 

[tmtomh]

Usually nothing, as the distribution format will leave the high ultrasonics behind.

However, audiophiles now demand native DXD as the the "ultimate fidelity," which creates a potentially unfriendly environment for their equipment.

I keep wondering where this allegedly large community of audiophiles demanding DXD is.

 

[DimitryZ]

I keep wondering where this allegedly large community of audiophiles demanding DXD is.

Apparently enough to keep companies like this (and many others) in business,:

https://www.nativedsd.com/