• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Remaining Considerations on DSD

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,348
Likes
3,462
Location
San Diego
Apparently enough to keep companies like this (and many others) in business,:

https://www.nativedsd.com/
I would be interested in some of these recordings because I have some old direct to disc LPs that sound good and the performances can be inspired as there are no re-dos or dubs or mixing. Unfortuneately if you read carefully that is not what going on here...it seems all the recordings are still post processed so the whole "native" thing is suspect.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
OP
Saidera

Saidera

Senior Member
Joined
Jul 18, 2020
Messages
388
Likes
309
Location
Adelaide, South Australia, Australia
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.

Not sure about settled for me.

I would really like to know how DSD recorders work as opposed to PCM recorders.

I would really like to know what processes differ in the DAC between PCM and DSD paths.

I would really like to know what differences DSD to PCM methods have with each other, and also compared to plain PCM.

Was there any, even the slightest evidence behind the market talk and trends? And as my discussion isn't really interested in the DSD being sold out there, but rather is about simply using DSD as a format to record nature, speeches, local concerts etc and to playback for example through HQPlayer, or to downconvert DSD to PCM to save space and perhaps keep its quality, the skill of producers/engineers isn't the only matter.

But I acknowledge that even PCM-D100 is like a toy and isn't real DSD. It seems we can use phones to record notes in AAC but the time has not come when DSD can be treated the same way.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
Not sure about settled for me.

I would really like to know how DSD recorders work as opposed to PCM recorders.

I would really like to know what processes differ in the DAC between PCM and DSD paths.

I would really like to know what differences DSD to PCM methods have with each other, and also compared to plain PCM.

Was there any, even the slightest evidence behind the market talk and trends? And as my discussion isn't really interested in the DSD being sold out there, but rather is about simply using DSD as a format to record nature, speeches, local concerts etc and to playback for example through HQPlayer, or to downconvert DSD to PCM to save space and perhaps keep its quality, the skill of producers/engineers isn't the only matter.

But I acknowledge that even PCM-D100 is like a toy and isn't real DSD. It seems we can use phones to record notes in AAC but the time has not come when DSD can be treated the same way.
Worth reading this Lipschitz and Vanderkooy paper on 1 bit sigma-delta encoding (DSD).

https://timbreluces.com/assets/sacd.pdf
 

danadam

Addicted to Fun and Learning
Joined
Jan 20, 2017
Messages
956
Likes
1,496
Apparently enough to keep companies like this (and many others) in business,:

https://www.nativedsd.com/
Is that why recently they started offering "normal" 96k FLACs? (sadly, no 44.1k) https://www.nativedsd.com/catalogue/?filter_quality=flac-96khz
(Don't get me wrong, I fully support it. I'm actually surprised that it took them so long.)
They offer DSD512
Looks like they starting with 1024 too: https://www.nativedsd.com/catalogue/?filter_quality=dsd-1024-fs
:rolleyes:
 

DimitryZ

Addicted to Fun and Learning
Forum Donor
Joined
May 30, 2021
Messages
667
Likes
342
Location
Waltham, MA, USA

Head_Unit

Major Contributor
Forum Donor
Joined
Aug 27, 2018
Messages
1,340
Likes
688
Is hires LPCM noise shaping similar to DXD
I think you meant DSD? To oversimplify, a one bit system only has a 6 dB signal-to-noise ratio. That would suck so DSD shapes the noise shoving it above the hearing range, where it is either harmless or horrible depending on your system and your beliefs and the phase of the moon. (I don't recall if the amplitudes become enough to cause problems with sensitive equipment...ah there's analog output filtering? Blanking out mentally...).

LPCM-if you had a pure PCM system at 24 bits then in theory the S/N would be 144 dB which only exists in the digital domain-only a couple hardware units claim anywhere near that noise performance and it far exceeds any real recording or playback room's noise floor.

There aren't too many pure ADC nor DACs I don't think. Most seem to be hybrids between 1-bit and pure PCM. I recall a boss of mine being very bothered when I kept saying "so this new DAC is like 3 bits inside instead of 1 bit, right?" OMG he was SO annoyed by that but actually I was correct. Anyway the noise shaping can vary according to the DAC design.

Corrections welcome, it's been a long time since I worked with the engineering details of this stuff.
 

DimitryZ

Addicted to Fun and Learning
Forum Donor
Joined
May 30, 2021
Messages
667
Likes
342
Location
Waltham, MA, USA
I think you meant DSD? To oversimplify, a one bit system only has a 6 dB signal-to-noise ratio. That would suck so DSD shapes the noise shoving it above the hearing range, where it is either harmless or horrible depending on your system and your beliefs and the phase of the moon. (I don't recall if the amplitudes become enough to cause problems with sensitive equipment...ah there's analog output filtering? Blanking out mentally...).

LPCM-if you had a pure PCM system at 24 bits then in theory the S/N would be 144 dB which only exists in the digital domain-only a couple hardware units claim anywhere near that noise performance and it far exceeds any real recording or playback room's noise floor.

There aren't too many pure ADC nor DACs I don't think. Most seem to be hybrids between 1-bit and pure PCM. I recall a boss of mine being very bothered when I kept saying "so this new DAC is like 3 bits inside instead of 1 bit, right?" OMG he was SO annoyed by that but actually I was correct. Anyway the noise shaping can vary according to the DAC design.

Corrections welcome, it's been a long time since I worked with the engineering details of this stuff.
You are correct. The ultra hires file in amir's explainer video was likely a DSD. I was surprised how noisy the 50-150KHz region was.
 

Head_Unit

Major Contributor
Forum Donor
Joined
Aug 27, 2018
Messages
1,340
Likes
688
an audiophile ins't going to buy a Bluray.
Er, many of us must beg to differ. I have maybe 3 dozen audio-only, non-concert Blu-rays (surround remixes of studio albums). Now, is there a special SACD market? Seems like it, to my annoyance since I believe it will get harder and harder to find players, I'd far rather those titles came out on Blu-ray (or even DVD-Audio with a DVD-Video layer).
 

Kal Rubinson

Master Contributor
Industry Insider
Forum Donor
Joined
Mar 23, 2016
Messages
5,270
Likes
9,770
Location
NYC
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.
How many "audiophile" buyers go to a real store these days? You buy on the web and choose from the available formats/media. Odd's are, more and more are not buying physical media anyway.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
CD sales were 4% of music sales in 2020. Currently at 3% of the peak year in 2000. SACD and Bluray were approximately 5% as much as CD sales or less than .2% of total sales by revenue. Physical media isn't selling much, and going down by more than 20% per year (except for LP).
 

xaviescacs

Major Contributor
Forum Donor
Joined
Mar 23, 2021
Messages
1,494
Likes
1,971
Location
La Garriga, Barcelona
CD sales were 4% of music sales in 2020. Currently at 3% of the peak year in 2000. SACD and Bluray were approximately 5% as much as CD sales or less than .2% of total sales by revenue. Physical media isn't selling much, and going down by more than 20% per year (except for LP).

Thanks for the data. Do you have an idea of the percentage of hi-res produced music can't be found in a physical support? I mean, are there many recording brands that only publish their materials online? I mean, they ship the music phisically but it doesn't get sold, but they are afraid to go only for the online market.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,521
Likes
37,050
Thanks for the data. Do you have an idea of the percentage of hi-res produced music can't be found in a physical support? I mean, are there many recording brands that only publish their materials online? I mean, they ship the music phisically but it doesn't get sold, but they are afraid to go only for the online market.
I know of a couple places that only sale physical media. But they are truly one horse operations and aren't even a blip on the sales chart.
 
OP
Saidera

Saidera

Senior Member
Joined
Jul 18, 2020
Messages
388
Likes
309
Location
Adelaide, South Australia, Australia
Go ahead and defend DSD

On the Wikipedia page there should be Lipshitz and those who argued against his paper, but what remains is that by the time DSD got around to people, multibit converters had been made. Dsd is a relic. So what if it bypasses decimation? It still has to be processed. If it’s edited even a little, it has already lost the benefit of avoiding PCM’s weaknesses. The issue is whether DSD processing is somewhat better.

Perhaps I had expected DSD would be more popular
 
OP
Saidera

Saidera

Senior Member
Joined
Jul 18, 2020
Messages
388
Likes
309
Location
Adelaide, South Australia, Australia
Usually my interest in dsd is a fleeting one, insufficient to last long enough for me to revise foundational concepts to fully comprehend the various detailed literature on this topic.

At least, it’s just too bad that I entered audio with a strong preference for DSD audio. Indeed, it was indignance at cheaply implemented Realtek 24/48 laptop audio chips that led me to strongly prefer CXD9872 with its DSD and great noiseless sound (implemented well no doubt). Having had enough of the subjective commentary that has buoyed DSD until now, I sought some scientific or engineering knowledge. I got that in the 2020 ASR DSD thread. But when I am reading any DSD research, my purpose is to find anything to vindicate DSD audio. If something negative about DSD is written, I will try to find any alternative where the problem would not occur or concede defeat but then change the premise and leave the issue of DSD open to future consideration again.

Still it’s not clear whether 768kHz is pretty much DSD, without the negative features of DSD. One could relegate DSD as a time filler, only necessary to bide enough time for the arrival of 24/96 to the masses.

Still another way to look at it is to consider how the really good DSD recorders and editors were created very early on and took years to develop and perfect, but then became ultra-rare relics. I can not view modern recorders in the same way with the same respect despite capabilities like DSD256 recording. Certainly the modern recorders are mostly used in-house and in Japan and are not commercially viable, so they, and the DSD format, might just as well be deemed to not exist at all. A format which consumers can not record in is no format at all. Although chip based DSD ADCs do not do justice to the DSD format, it would be nice to have them. All this will require an understanding of what potential or detrimental flaws this much maligned format has. DSD purists would want the purest DSD, but what is the position of PCM derived from DSD? It has been degraded and should no longer contain ‘benefits of DSD’. What difference if any, exists between that and normally recorded PCM?

Oh no, I also think the reason for DSD’s success in a select group is the long history of the prevalence of bad PCM-only designs such as those in laptops. Very few exceptions can be made. DSD was and is like a trustworthy badge of confirmation of audio quality and attention to design, with more power than the all too common hires logo that is slapped around everywhere.

Whenever one gets carried away with such minute details, it has to be remembered that MIDI can be played back with the minimum of resources and its sound can surpass any recorded sound by a mile. It seems like madness that somehow PCM and DSD strives to achieve the perfect MIDI sound with such large amounts of data.

But then again, there’s also the worry that somehow chip based people are missing out on even better sound than a CMOS chip can provide. As if the CMOS chip has been hiding from all of us the true potential and actual real sound represented by the data. And that DSD might be part of the solution.

Such is the endless nature of silly considerations in audio.

571 words
 
Last edited:
OP
Saidera

Saidera

Senior Member
Joined
Jul 18, 2020
Messages
388
Likes
309
Location
Adelaide, South Australia, Australia
The momentary boom of DSD audio probably died out around 2018. It started around 2015 or thereabouts. The first boom of DSD would be around the turn of the century leading up to 2009. Then it died down until digital downloads came up, and then some niche streaming of DSD256 maybe.

I came upon old threads of ASR from the early days and reading them made me feel that what I am doing is repeating history, bringing up such useless topics as HQPlayer upconversion, DIY DSD implementations and the piecemeal partial knowledge hidden in all four corners of the Internet.

Really, for laypersons, the conclusion was already clear then. Although hires has a slight difference which could be an improvement, over CD, as the AES meta-analysis might have shown, you have to be highly trained. Even lossy compression gives us a hard time, so what on earth is possibly discernible with DSD? At least nobody has heard obvious problems with DSD. One should be thankful and not require that DSD live up to its marketed image. I will hazard a wild guess that DIY DSD implementations often aren’t so good, and hence DSD might sound better than PCM on them. But that is only in some instances.

Meanwhile, DIY DSD implementation really gets people to have a thorough understanding of their equipment, and that obviously is a source of great satisfaction. For me, being a complete chip person ever since I became interested due to VAIO DSD with its chip CXD9872 7mm square chip by Sigmatel and later IDT, such DIY DSD has little relevance. Incidentally, it was nice to see that many Australians do DIY audio as well. So rooted in CXD9872 am I that I also reject the dual (mono mode) dac uptake as well as balanced 2.5 or 4.4 mm connections, preferring single chip 3.5 mm instead.

But because my preferences are rooted in CXD9872, I tend to view the DSD circuits in chips as being superior to the PCM side. In terms of CXD9872, Sigmatel’s normal STAC92xx lineup had an SNR of around 105, whereas Sony’s DSD circuit upped it to 107. Of course this is merely a difference of technology and has no bearing on modern chips designed by the same manufacturer, but it has coloured my perception.

Given that the majority of DSD users will be using non-DIY off the shelf Cirrus/ESS/AKM/Qualcomm WHS/Realtek, with some having access to TI/BB/S-Master HX (via 4.4 mm balanced only for DSD Native) and other less prevalent chips, my only practical concern was whether DSD in these black boxes is processed better than PCM. Although Realtek cooperated with Sony on ALC889DSD, their current DSD mode is not likely to be as well made. Qualcomm is likely to have integrated Realtek technology into their WHS chips, and there is some evidence to suggest that their DSD is non-Native playback i.e. converted to PCM. That leaves Cirrus/ESS/AKM, whose DSD pathways have been considered by other interested people.

As Cirrus’ CS4398/43131 has likely been the most affordable and high-performing DSD chip (excluding Realtek) from 2015-present, it is of most interest to me. Many devices have used CS4398/43131. CS4398 was in Cyberdrive Aura/Feather with its overheating XMOS chip and very affordable price when competitors were much more expensive back in 2015. It was simply not so widely known, unfortunately. CS43131 with CS46L41 bridge achieved a finger-sized DAC with Meizu and subsequent copies. CS43131 with Savitech bridges enabled matchbox sized Sonata HD Pro and subsequent copies including the more pricey DC03/04.

The DSD issue was easily settled when I had tried out the PCM and DSD on CS43131. Volume matched by the chip, switching back and forth in HQPlayer, using similar filters and modulators, you’d be hard pressed to ‘hear’ differences. As it was not blind, I’d try higher volumes and harsher sequences and find no difference. It turns out of course that Cirrus subjects both to similar processing at the DSM and its DAC.

Only with CS4398 or CS43198 direct DSD mode would there be slight differences, and then volume matching would be impossible. One might wonder if direct DSD mode would measure more superior if it could be made louder by the chip. One may find that AKM’s DSD does sound different to its PCM, but that is not affordable. On the whole, there are very few statements suggesting that DSD is better, as it is clear that preferences are divided. For CS43131 purposes, either PCM or DSD would sound the same, except that PCM can achieve better measurements when driven loud. But at lower volumes, one wonders whether there might just be 10-20% less processing done on DSD relative to PCM. I might as well stick to DSD just for that inaudible difference.

There is the claim that a truly well designed recorder simply capturing and storing DSD, and then another playback device simply playing it back would achieve good sound. It seems so deceptively simple. And yet, DSD cannot be played back using a single resistor and capacitor. It appears that plenty of processing is necessary. It is very complicated. It is a way of thinking perhaps. If current DACs are 6 bit, one would like to get rid of as many stages as possible and just store that 6 bit audio for playback. Recorders would also have to record in 6 bits. Taking a step back from this, it would appear that PCM serves consumer needs sufficiently well. As long as it is done well, there is no need to eliminate stages. One should be pragmatic and focus on clear subtle differences one can actually hear.

Overall, due to my superficial chip-based view of audio, DSD and PCM are no different. Although Direct DSD mode sounds alluring, the benefits do not as yet outweigh the disadvantages. Truly good, real non-chip based DSD Recorders are far too expensive and difficult to design and basically all studios take no interest in them, save the few in Japan and especially the KORG, TEAC, SONY DSD trio affiliated firms. SBM Direct may carry over the DSD source qualities into PCM, but they are unlikely to be due to the inherent qualities of DSD as a format. More likely the qualities of the supreme DSD capable recorder with a rubidium clock, the ADA-7000R that weighs dozens of kilos. That was merely the opinion of a misguided Sony engineer and has nothing to do with common users. It may be correct of course. And upconversion to DSD to take advantage of the DSD pathway really depends on the DAC. For Cirrus chip users, either choice is fine.

The preferred outcome would have been to find proof that Cirrus’s DSD is the best audio playback that exists on this planet.

Inconsequential stresses in life made me kind of binge on DSD audio superficial knowledge, and it truly has been a useful diversion of my concentration.

Word count: 1149
 
Last edited:
OP
Saidera

Saidera

Senior Member
Joined
Jul 18, 2020
Messages
388
Likes
309
Location
Adelaide, South Australia, Australia
After seeking some concrete direct DSD facts in chip based audio, I found plenty of sites.

https://www.audiosciencereview.com/forum/index.php?threads/1-bit-dsd-dac-implementations.129/
https://www.audiosciencereview.com/forum/index.php?threads/what-is-audio-dither.141/
https://www.audiosciencereview.com/forum/index.php?threads/dsd-genuine-improvement-or-marketing-tool.130/
https://www.audiosciencereview.com/...s/aes-paper-digest-dvd-audio-versus-sacd.150/
https://www.audiosciencereview.com/...-high-sampling-digital-recording-formats.151/
https://audiophilestyle.com/forums/topic/15783-to-dsd-or-not-to-dsd/
https://audiophilestyle.com/forums/topic/17882-the-multibit-dsd-debate/
https://audiophilestyle.com/forums/...d-provides-a-direct-stream-from-ad-to-daquot/
https://audiophilestyle.com/forums/topic/28275-mark-waldrep-and-more-silliness/
https://audiophilestyle.com/forums/topic/28250-direct-stream-digital-and-the-power-of-imagination/

There is very little real knowledge in those threads. Although very old threads in computer audiophile and some in diyaudio do crop up in searches, they aren’t in the end very useful. DSD discussion threads actually date back a long way and in some cases popular engineers or over enthusiastic DSD artists were there to basically support the format.

It was certainly interesting to read Miska, Hiro, bunpei on one side, and mansr on the other. Some questioned the focus on chips if the only use they could be put towards was the final DAC stage referred to as ‘switched capacitor filter or array’. That form of DAC was used in CS4398, the long-living 2005 DAC chip used in countless products for a decade, including the KORG DAC measured on ASR and the cheap XMOS USB products (although it was an idea floated on diyaudio, those using XMOS with analog outputs possibly had the ability to choose direct DSD, whereas the simpler headphone 3.5 mm jack versions are not customisable but I will check – it should be apparent that the volume control is unusable or the volume should not be matched with PCM). After Wolfson was integrated, changes must have been made. The DAC is now 512 ‘single-cells’ lined up in series where the idea is that the signal gets better by the end. It is easily misinterpreted to mean 512 single bit DACs. Or is the switched capacitor filter still used?

It is possible to view the PCM path as having the most steps, as does DSD, but only direct DSD jumps straight to the end. After going through all this surface knowledge one has to conclude that CXD9872 with its ability to control the volume, must have used a DSD Processor type pathway, or at worst, a DSD to PCM path like Wolfson offered. CXD9872 emerged in 2005, but Cirrus had released chips capable of DSD in 2004 as well. CXD9872 was by no means the first. And Soundblaster had used CS4398 in 2006 or prior. DSD Playback was out there for interested persons. What’s more, since 2003 (Deuwer et al. ), the DSD Processor way has remained the same. Its internals may have changed. But this is some 18 years on and nothing has changed! Indeed, Deuwer’s paper shows the use of DSD Processor WITHOUT the multibit DSM. One has to wonder how DSD Processor works to change DSD volume.

Meanwhile AKM’s direct DSD is similar to Cirrus’ direct DSD way, but ESS and its processing of DSD is similar to the Cirrus’ DSD Processor way. One wonders when AKM and ESS started to support DSD playback. One has to question the use of DSD and DSD-wide compared to PCM in the fully unknown territory of multibit ADCs. And in the end, it boils down to, what does Cirrus’ direct DSD sound like? As it’s a straight analog out, each implementation might change the sound slightly. A random Bluray player isn’t going to prove direct DSD either way. Ruling out such subjective comparisons, it seems the fact that direct mode measures slightly worse as shown on the datasheets will have to do. Since I cannot reach the grapes I will be content with labelling them as sour and not worth my effort. DSD Processor is for the PCM-like-convenience and headphone jack output, and that is important if we’re to rely on the chip only. As the rest of the path is shared with PCM, they will end up sounding the same. Most likely the PCM and DSD, regardless of their original formats, will become more or less the same format in the DSM.

There are 48kHz files which have to be converted to 384kHz PCM and 44.1 kHz files which may be converted to DSD. Both have to coexist in this environment.

Incidentally, Sony’s PCM-D series are all temporarily halted since Feb 2021 due to supply issues. I didn’t realise they were in such demand. Their SNR is just around 100 and PCM-D100 is the only fake multibit chip based DSD recorder (unless they're using old chips).

https://www.aes.org/e-lib/browse.cfm?elib=12392
SESSION Z4: POSTERS: SIGNAL PROCESSING, PART 1 Saturday, October 11 2:00 PM— 3:30 PM
Z4-6: A Multibit Delta-Sigma DAC with Mismatch Shaping in the Feedback Loop—Bruce Duewer, John Melanson, Heling Yi, Steve Green, Cirrus Logic, Inc., Austin, TX, USA

It was cited in a textbook.

Expanded on here: https://www.audiosciencereview.com/...cirrus-logic-cs43131.23138/page-2#post-833668
 
Last edited:
Top Bottom