The sound quality of DSD?

[Pluto]

The SACD spec permits brief excursions a few dB beyond this (I don't recall the details)

If I recall, the actual maxima relate to the number of permitted consecutive 1s or 0s cut onto the optical disc; much like the eight-to-fourteen modulation process on a Red Book disc.

 

[mansr]

If I recall, the actual maxima relate to the number of permitted consecutive 1s or 0s cut onto the optical disc; much like the eight-to-fourteen modulation process on a Red Book disc.

There is probably such a restriction (it would make sense), but it's for very different reasons than the EFM rules for CD. The latter are necessary for the tracking PLL to work, whereas the audio data on an SACD is stored in files just like on a DVD. There is nothing in the design of the disc that prevents storing a track with all ones. The term modulation appears in both contexts, but the meaning is quite different.

 

[Pluto]

So why does the Scarlet Book expressly prohibit more than 24 (or fewer than 4) consecutive 1s within any 28 bit ‘window’ within the bitstream? That seems very like satisfying the demands of the laser optic system to me, yet I cannot see why such a limitation would be imposed were it not for the needs of the optical sensor.

 

[mansr]

So why does the Scarlet Book expressly prohibit more than 24 (or fewer than 4) consecutive 1s within any 28 bit ‘window’ within the bitstream? That seems very like satisfying the demands of the laser optic system to me, yet I cannot see why such a limitation would be imposed were it not for the needs of the optical sensor.

Consider that the DSD bitstream is often compressed (DST), and multiple channels are interleaved. That limitation can't possibly have anything to do with the optical readout. What motivation does the spec give? I don't have a copy, so I can't check.

 

[analogSurviver]

I do not know the exact definitions - and, TBH, I do not really care what 0dBFS "should" be for DSD.

What I can - and DO - care about, is how real hardware that can record native DSD reacts to input levels. Korg MR-1 and MR-1000 ( which I own, both in stock and highly modified versions ), can "survive" up to +6dB ( or even a fraction of dB above ) without any audible protest. Both the recording and playback, as long you are not driving the headphone out into clipping ( on MR-1, headphone and line out are "shared" through the same op-amp ) - line out on MR-1000 is OK no matter what under these conditions.
Any export of such "hot" DSD file to PCM can be - and has to be - normalized not to exceed 0dBFS using Korg Audiogate 3 or 4 software. Since many DACs do not like being pushed above -1dBFS, it is better to normalize to say -1.5dBFS; any further transcoding (MP3...) can add another +0.3dB or so - and can combined definitely exceed the limits of much playback gear out there.

TASCAM DA-3000 is a totally different animal. Here, a fraction over 0dBFS creates hard clipping - it can not handle not even above 0dBFS DSD recording made on Korg MR Series. The clipping produced really is obnoxious sounding - and audible instantly.
So, whatever the max level still permissible on Korg, has to be set at least 6dB lower on TASCAM unit.

For live recordings, those approx 6dB "safety exit curves" in Korg MR Series is a gift from above. After all, there are PCM recorders that - additionally to the main "0dB" recording - record another at -20dB. For a reason...

Does anyone have experience with DSD maximum clean sounding recording level(s) on RME ADI-2 Pro FS (R), Merging Anubis or Hapi ? I am eyeing "something" DSD256 capable - but the possibility of using the RME as measuring instrument in PCM also looks alluring.

 

[AnalogSteph]

You need direct recording to DSD about as much as a hole in the head. Use PCM and convert if necessary. It's good enough for Audio Precision, too.

With close to 120 dB of instantaneous dynamic range on tap in the ADI-2 Pro, you can let things peak at -20 or even -30 dBFS if you think you need that much headroom. (You can also further shift this range up and down by choosing input 0 dBFS level, for a total coverage of 139 dB.) "Safety tracks" are a safeguard for situations where only a minimal soundcheck is possible at best and the one shot you've got has to work out whatever happens.

Compared to the days of tape recording, things are so much more relaxed nowadays. Yes, you need to stay away from clipping, but there is so much dynamic range available. I bet those who did classical recordings back in the 1960s or '70s would have been jumping for joy had they been shown our modern-day ADCs.

 

[EB1000]

Native DSD1024 may sound as good as CD... I'd stay away from anything else...

 

[Blumlein 88]

You need direct recording to DSD about as much as a hole in the head. Use PCM and convert if necessary. It's good enough for Audio Precision, too.

With close to 120 dB of instantaneous dynamic range on tap in the ADI-2 Pro, you can let things peak at -20 or even -30 dBFS if you think you need that much headroom. (You can also further shift this range up and down by choosing input 0 dBFS level, for a total coverage of 139 dB.) "Safety tracks" are a safeguard for situations where only a minimal soundcheck is possible at best and the one shot you've got has to work out whatever happens.

Compared to the days of tape recording, things are so much more relaxed nowadays. Yes, you need to stay away from clipping, but there is so much dynamic range available. I bet those who did classical recordings back in the 1960s or '70s would have been jumping for joy had they been shown our modern-day ADCs.

You tell 'em Steph.

Besides everybody knows you don't have the real deal unless you can do DSD1024. It gets you well over that billion bits per second spec.
1.445 bbps. I mean we are talking audio here. Can hardly work with so few bits, and less than 3/4 ghz response.

Dang: Ninja'd by EB1000

 

[Pluto]

I do not really care what 0dBFS "should" be for DSD

You would do if you had to send SACD masters to the pressing plant.

 

[julian_hughes]

On the subject of whether people can hear a difference between DSD & PCM in a double blind test: https://www.researchgate.net/public...ution_recordings_in_PCM_and_DSD_audio_formats
The link is to an AES paper which is free to read and/or download.

People can discern a difference, and they express a preference too.

Personally I listen to PCM because it's supported by all my playback devices whereas DSD is supported on most but not all. I have no idea if I could hear a difference or subsequently prefer one or the other; I don't have the facilities to perform a proper DSD vs PCM blind or abx test and I'm not much bothered either way. I think well presented 16-bit 44.1 KHz sounds really fine.

Some persistent themes throughout this thread:

Nobody makes SACD discs, it's a dead format. Not true. I mostly listen to classical/orchestral/choral music. Several (many?) "Classical" labels sell hybrid SACD/CD discs by default so without seeking out SACD I found I had accumulated a lot of these hybrid discs. In fact one arrived in the post today thanks to the magic of ecommerce. This leads onto the next point:

SACD is hard to rip and requires special hardware and software. Not really. I bought a used Sony BDP-S390 Blu-ray player which supports SACD. It cost me £22 (about $28 US). By reading the hideously badly organised thread https://audiophilestyle.com/forums/topic/28569-sacd-ripping-using-an-oppo-or-pioneer-yes-its-true I found I could write a simple script to a USB stick and then use a free application, sacd_extract, on my PC to extract the DSD (stereo, multichannel or both) to my PC. This gives unencrypted DSF or DFF (as you prefer) files to use or convert as you wish.

I can't claim to tell the difference between my DSF files, the derived 24-bit 88.2 KHz flac, or the original red book layer on the discs, except for the slightly lower level of the flacs converted from DSD! But people can tell the difference and so claiming that this is impossible is not useful or objective or sceptical, it's just silly. The difference in formats makes home abx testing DSD vs PCM impractical or impossible for most people, and the paucity of such tests doesn't mean there is no difference. The AES paper referenced above is a good read and very thorough, so far as I can tell.

Anyway, I'll now sign off and return to listening to my new hybrid SACD...or the red book rip....or the DSFs...

 

[Pluto]

The core of the debate is not whether the DSD stream on Scarlet Book CDs sounds any better/different to 16/44 PCM; I suspect few would deny that the former has the potential to be superior.

The question, for me, is simple: is it worth bothering with DSD or SACD at all in a world with excellent 24/96 PCM.

 

[julian_hughes]

The core of the debate is not whether the DSD stream on Scarlet Book CDs sounds any better/different to 16/44 PCM; I suspect few would deny that the former has the potential to be superior.

The question, for me, is simple: is it worth bothering with DSD or SACD at all in a world with excellent 24/96 PCM.

I agree the more pertinent issues are "will I notice?" and "if so, do I care?" but it gets lost in a sea of erroneous negative claims and the quite unjustified dismissal (imo) of people who say they hear a difference and have a preference for DSD. The people presenting as a supposedly conclusive argument the claim that DSD is indistinguishable from PCM are really just muddying the waters. Or you might say they are making a lot of high frequency noise for no good purpose.....OMG

 

[mansr]

It may be possible to hear a difference between a particular DSD implementation and a particular PCM implementation, just as it may be possible to hear a difference between two implementations of the same format. Studies suggesting audible differences between better-than-CD formats are invariably set up such that it's impossible to say with any certainty that the differences heard were not actually caused by something other than what is claimed.

 

[julian_hughes]

It may be possible to hear a difference between a particular DSD implementation and a particular PCM implementation, just as it may be possible to hear a difference between two implementations of the same format. Studies suggesting audible differences between better-than-CD formats are invariably set up such that it's impossible to say with any certainty that the differences heard were not actually caused by something other than what is claimed.

If you're unable to accept the results of the perfectly valid above linked AES paper describing in detail how the differences were *proven* to be audible then rational conversation with you on this subject is not possible because you're viewing it from an immovable position that is unaffected by facts. That position is by definition entirely subjective.

 

[mansr]

If you're unable to accept the results of the perfectly valid above linked AES paper describing in detail how the differences were *proven* to be audible then rational conversation with you on this subject is not possible because you're viewing it from an immovable position that is unaffected by facts. That position is by definition entirely subjective.

That paper doesn't even mention level matching the playback volumes. It does, however, show quite different frequency responses between the recording formats (1 dB shift for graph legibility only):

*plonk*

 

[SIY]

If you're unable to accept the results of the perfectly valid above linked AES paper describing in detail how the differences were *proven* to be audible then rational conversation with you on this subject is not possible because you're viewing it from an immovable position that is unaffected by facts. That position is by definition entirely subjective.

It’s a conference preprint, not a full refereed paper. And several things jump out at me as potential problems, so I think it’s a stretch to summarily declare it “perfectly valid.”

edit: mansr got there first.

 

[BDWoody]

I saw one study that is often tortured to try to show the slightest of preference, once listeners receive some training, but that was certainly not conclusive.

On the subject of whether people can hear a difference between DSD & PCM in a double blind test: https://www.researchgate.net/public...ution_recordings_in_PCM_and_DSD_audio_formats
The link is to an AES paper which is free to read and/or download.

yep...that's the one.

Already jumped on earlier...but I found this part of the paper interesting (bold emphasis mine):

Although the authors’ expectation was that
brightness can be used to discriminate between the
formats, opposite result was obtained. “Sharpness,”
a synonymous attribute to brightness, is the at-
tribute related to spectral centroid with weight on
high frequency [7, 8]. The results suggest that par-
ticipants were not able to hear the differences be-
tween the spectral differences in high frequency re-
gions. On the other hand, although very subtle,
difference in frequency curves of DSD and PCM is
larger below 31.5 Hz compared to that of high fre-
quency ranges above 16 kHz disregarding the spec-
tral noise above 32 kHz in DSD (2.8 MHz) (Figure 1).
Participants may have relied on the low frequency
contents to discriminate the formats, and it is sup-
ported by Triangle stimulus which has less low fre-
quency contents was not being highly significant in
the comparison

Ya think?

So, they went in thinking it would all be in the highs, and the low's wouldn't matter, didn't bother MAKING RECORDINGS WITH THE SAME FREQUENCY RESPONSE, and thought that was good?

As I said...it keeps getting trotted out and tortured to try to show what hasn't yet been shown.

 

[krabapple]

On the subject of whether people can hear a difference between DSD & PCM in a double blind test: https://www.researchgate.net/public...ution_recordings_in_PCM_and_DSD_audio_formats
The link is to an AES paper which is free to read and/or download.

People can discern a difference, and they express a preference too.

Nope. Seriously, this ground has been covered a lot already. Including this non peer reviewed convention presentation.

The fact is, after decades no, there are exactly *zero* papers that definitively confirm audible difference due to DSD vs PCM.

And amon those that claim to offer such evidence. there aren't (and IMO never will be) any papers that support the sort of 'OMG what a difference' claims fans make for DSD vs PCM. Any such 'effect' simply isn't that strong.

Indeed it took a meta analysis of data from multiple papers, to even *hint* that such difference might be audible, to some people. That paper also proved controversial, because meta analysis always involves picking and choosing which data to include and exclude.

 

[PaulD]

On the subject of whether people can hear a difference between DSD & PCM in a double blind test: https://www.researchgate.net/public...ution_recordings_in_PCM_and_DSD_audio_formats
The link is to an AES paper which is free to read and/or download.

People can discern a difference, and they express a preference too.

Personally I listen to PCM because it's supported by all my playback devices whereas DSD is supported on most but not all. I have no idea if I could hear a difference or subsequently prefer one or the other; I don't have the facilities to perform a proper DSD vs PCM blind or abx test and I'm not much bothered either way. I think well presented 16-bit 44.1 KHz sounds really fine.

Some persistent themes throughout this thread:

Nobody makes SACD discs, it's a dead format. Not true. I mostly listen to classical/orchestral/choral music. Several (many?) "Classical" labels sell hybrid SACD/CD discs by default so without seeking out SACD I found I had accumulated a lot of these hybrid discs. In fact one arrived in the post today thanks to the magic of ecommerce. This leads onto the next point:

SACD is hard to rip and requires special hardware and software. Not really. I bought a used Sony BDP-S390 Blu-ray player which supports SACD. It cost me £22 (about $28 US). By reading the hideously badly organised thread https://audiophilestyle.com/forums/topic/28569-sacd-ripping-using-an-oppo-or-pioneer-yes-its-true I found I could write a simple script to a USB stick and then use a free application, sacd_extract, on my PC to extract the DSD (stereo, multichannel or both) to my PC. This gives unencrypted DSF or DFF (as you prefer) files to use or convert as you wish.

I can't claim to tell the difference between my DSF files, the derived 24-bit 88.2 KHz flac, or the original red book layer on the discs, except for the slightly lower level of the flacs converted from DSD! But people can tell the difference and so claiming that this is impossible is not useful or objective or sceptical, it's just silly. The difference in formats makes home abx testing DSD vs PCM impractical or impossible for most people, and the paucity of such tests doesn't mean there is no difference. The AES paper referenced above is a good read and very thorough, so far as I can tell.

Anyway, I'll now sign off and return to listening to my new hybrid SACD...or the red book rip....or the DSFs...

And to add to krabapple's post above, in post 2 I mention the blinded level-matched test I did that showed no audible difference between DSD on a Korg MR-1000 and 24/96 PCM. Here is the excerpt from post 2.

"I did do a blind test in a studio with the Korg MR-1000 DSD recorder when it came out, signal split after the mic preamps and sent to the Korg recorder and also to Avid Protools PCM converters at 24bit 96KHz. Level matched on playback, there was no audible difference between the DSD and PCM version."

That fact that DSD did not sound worse was the surprising thing to me.

 

[BDWoody]

If you're unable to accept the results of the perfectly valid above linked AES paper describing in detail how the differences were *proven* to be audible

Did you post the wrong paper?

There should be a special place for AES papers like that... The AES-Hole.

That fact that DSD did not sound worse was the surprising thing to me.

Well, if anything, it showed a preference for a more accurate full range frequency response. What a surprise.
Beyond that...