The sound quality of DSD?

[mansr]

I'd put money on it.

The entire DSD debacle goes back the best part of 20 years. Back then it undoubtedly was better than the PCM then available. Sony and the others who ran “big music” saw DSD and its relationship with SACD as a solution to the problem of CD copying. But PCM got better and better; as soon as bit 18+ became fairly reliable (rather than part of random noise), DSD became obsolescent.

A 1-bit sigma-delta converter (ADC or DAC) is better than any (reasonable) 16-bit converter for reasons including component matching tolerance and settling times. Once in the digital domain, converting to a 24-bit format (at a lower sample rate) loses nothing and never did. I don't know what it was like 30 years ago, but with today's IC technology, the best results are obtained from multi-level sigma-delta converters. 1-bit DSD is a relic that belongs on the scrapheap.

 

[Thomas savage]

It is perfectly possible that SACDs sound better on their PCM layer - and it is usually the hardware used for the playback. If the actual DAC used is optimized for PCM and has DSD more as an afterthought or "must check box", that is rather common. If the DAC used is about equally at home both in PCM and DSD (not to mention if it is optimized primarily for DSD ), then DSD will sound better.
Why would you think that the mastering for the PCM and DSD layer has to be different ( other than format ) ? And why would crappy mastering have to be "audiophile" - and furthermore assigned to DSD layer only ?

I do NOT "get" identical masters - I have to actually make them. Microphone feed and parallel recording both to DSD and PCM. Anything else is one step further removed from the original sound.
And, to my expereience, DSD128 and PCM96/24 are NOT indistinguishable.
Very close, yes - indistinguishable, no.

If you come back with some controlled listening comparisons that show what you say ( and have repeatedly asserted in this thread ) is actually possible we have something to discuss.

It could just well be you feel DSD is more representative of the analogue feed, in laymans terms and this informs how you assess it .


It might not be that , who knows .

 

[Pluto]

And, to my expereience, DSD128 and PCM96/24 are NOT indistinguishable

How very true. When I worked full time in “the biz” my favoured test methodology was to operate the system some 50dB or more below its intended level so as to really emphasize the flaws.

And it was entirely apparent that 96kHz+ PCM was superior in every important respect.

 

[analogSurviver]

How very true. When I worked full time in “the biz” my favoured test methodology was to operate the system some 50dB or more below its intended level so as to really emphasize the flaws.

And it was entirely apparent that 96kHz+ PCM was superior in every important respect.

I concur - with DSD, any recording level "schemes" have to be tossed right away; for the DSD to be good and/or better, it has to be operated very close to 0dBFS. The reason is ultrasonic noise of DSD above 20 kHz.
That goes down my 12 dB with each doubling of sampling frequency - or DSD256 has 24dB better SNR than DSD64.

However, if that is known and recording peaks close to 0dBFS, , DSD will outperform PCM in time domain. Not only in signal shape ( pulse, square wave, anything fast ) - but also in - synchronization. There is no way to ascertain that PCM can retain that both channels start at the same time - not across all the hardware/software that can play that PCM file. Things can get as bad as square wave recorded to both channels absolutely simoultaneously ( mono signal using Y splitter to both inputs of ADC ) one channel lagging the entire rise time at chosen sampling rate behind the another. This alone is a strong case for higher sampling rates; although the shape/speed/frequency response in both channels remain exactly the same ( say 48k DAC ), by feeding it a higher sampling rate signal, this interchanell lag gets proportionally lower. In audible terms, soundstage DOES get wider.

DSD has inherently "synchronized" channels. Impossible to induce such an error.

 

[Pluto]

or the DSD to be good and/or better, it has to be operated very close to 0dBFS

Which is something nobody would do unless a piece is very accurately rehearsed (and good editing is available). The single biggest advantage of the digital world is the fact that it can be operated with at least 10dB headroom and less consequent need for dynamic range control during the session, yet nonetheless remain so far clear of quantizing noise that ventilation is likely to be the dominant noise source, by a significant margin.

I'm afraid that your concept of inter-channel lag has me beat at a number of levels. Given that the phase accuracy of 20kHz tones on each channel is so high that it is possible, in inverse phase, to cancel them to better than -100dB, I really do not understand the point you are trying to make.

 

[mansr]

I concur - with DSD, any recording level "schemes" have to be tossed right away; for the DSD to be good and/or better, it has to be operated very close to 0dBFS. The reason is ultrasonic noise of DSD above 20 kHz.
That goes down my 12 dB with each doubling of sampling frequency - or DSD256 has 24dB better SNR than DSD64.

However, if that is known and recording peaks close to 0dBFS, , DSD will outperform PCM in time domain. Not only in signal shape ( pulse, square wave, anything fast ) - but also in - synchronization. There is no way to ascertain that PCM can retain that both channels start at the same time - not across all the hardware/software that can play that PCM file. Things can get as bad as square wave recorded to both channels absolutely simoultaneously ( mono signal using Y splitter to both inputs of ADC ) one channel lagging the entire rise time at chosen sampling rate behind the another. This alone is a strong case for higher sampling rates; although the shape/speed/frequency response in both channels remain exactly the same ( say 48k DAC ), by feeding it a higher sampling rate signal, this interchanell lag gets proportionally lower. In audible terms, soundstage DOES get wider.

DSD has inherently "synchronized" channels. Impossible to induce such an error.

Uh-huh.

 

[analogSurviver]

Which is something nobody would do unless a piece is very accurately rehearsed (and good editing is available). The single biggest advantage of the digital world is the fact that it can be operated with at least 10dB headroom and less consequent need for dynamic range control during the session, yet nonetheless remain so far clear of quantizing noise that ventilation is likely to be the dominant noise source, by a significant margin.

I'm afraid that your concept of inter-channel lag has me beat at a number of levels. Given that the phase accuracy of 20kHz tones on each channel is so high that it is possible, in inverse phase, to cancel them to better than -100dB, I really do not understand the point you are trying to make.

Oh, the safetry net of the PCM ...
Well, that is WHY I always insist being on rehearsal - in the same venue the recording will take place. Otherwise, I do not even bother.
Furthermore, "rehearsal/performance" ratio - for EACH "act"- be it solo voice or instrument - or Mahler's 8th apparatus - has to be "learned". The hard way... Most musicians tend to be rather shy in rehearsals - and will usually pull out all the stops in say finale of a symphony. Can be up to plus 5-6 dB difference. And there are those who will "roar" at the rehearsal ... - just to get an idea how to "purr" in the concert. It can be up to minus 4 dB - or even more.
Given a known venue and possibillity to attend rehearsals, there have been quite a few cases of "guesstimate" accurate to within - 0.2dBFS.
Given absolutely no rehearsal and having gain on the preamp set to min from previous extremely loud gig , there were 2 or 3 cases I had to normalize the result... - by plus 30-40dB. Of course, that means only PCM up to 48kHz sampling could be salvaged. I try to avoid such situations as pleague - but sometimes it has to be done.

Regarding the lag that can occur with PCM; to be blunt, one channel has to reach full SPL before the lagging one can start ramping up ... - and the same goes for the decay. Despite the original signal has been completely simoultaneous and equal.

 

[analogSurviver]

Uh-huh.

Which part you think I have not written correctly ?

 

[Pluto]

one channel has to reach full SPL before the lagging one can start ramping up

I'm sorry, but that sounds like grade 1 doodoo.

But I'm happy to be proven wrong. Stories like that were from the days of the Sony digital audio > video converters (PCM-F1) that shared a single ADC on a time-division multiplexed basis and, even then, the leading channel was deliberately delayed at playback to correct the error introduced at the recording.

 

[mansr]

I'm sorry, but that sounds like grade 1 doodoo.

It's not even wrong.

 

[analogSurviver]

I'm sorry, but that sounds like grade 1 doodoo.

But I'm happy to be proven wrong. Stories like that were from the days of the Sony digital audio > video converters (PCM-F1) that shared a single ADC on a time-division multiplexed basis and, even then, the leading channel was deliberately delayed at playback to correct the error introduced at the recording.

Given how much did the digital audio spread since PCM-F1, similar effects can still occur in 2020. No one can test all hardware with all software available. PCM is NOT specified in such a way that this could not occur.

 

[mansr]

We need something stronger than uh-huh.

 

[Jimbob54]

We need something stronger than uh-huh.

 

[BDWoody]

.

We need something stronger than uh-huh.

 

[raif71]

https://darko.audio/2020/08/ps-audio-preps-perfectwave-sacd-transport/

 

[krabapple]

Iv always felt disappointed by DSD , all my 2ch SACDS sound better to me on the PCM layer , however that may be due to the crappy ' audiophile' mastering on the DSD layer .

? That's odd. Typically where CD layer and 2channel DSD layer mastering differ on a hybrid SACD, it's due to compression being added to the CD layer (to make it more car/discman friendly, one presumes...) . The most notorious case being the Dark Side of the Moon SACD.

Maybe you just like added compression?

I'm certain if you get identical masters , 128 DSD and 96 PCM will be indistinguishable.

I am too. I am similarly certain that if you convert the 128 DSD to 96 PCM, the same applies.

Why is this still something we talk about ?

Zombies are notoriously hard to kill.

 

[Jimbob54]

https://darko.audio/2020/08/ps-audio-preps-perfectwave-sacd-transport/

The last thing the world needs is another Darko and/or BS Audio discussion. But f*ck it, its 6 grand of CD/ SACD transport that STILL NEEDS A DAC CONNECTED to make any electrical impulses to convert to sound. I dug my £150 blu ray player out of the loft to play some old DVDs. I overpaid , I know . But really, how much better or worse could this be? A farce.

 

[RayDunzl]

However, if that is known and recording peaks close to 0dBFS, , DSD will outperform PCM...

How is 0dBFS for DSD determined?

No-brainer for PCM.

But what sets the limit (full scale) for DSD?

 

[danadam]

How is 0dBFS for DSD determined?

No-brainer for PCM.

But what sets the limit (full scale) for DSD?

That would be defined in Scarlet Book, I guess. The only thing I could find is SACD - A Technical Overview of the Format and Production Workflow and from page 27:

Since the reference level of SACD (0dBSACD) corresponds to a sine wave with a peak amplitude of 50% of the theoretical maximum DSD signal level in ANNEX D.2, there is 6dB more headroom over 0dB for a DSD signal. However, the allowed maximum level on an SA-CD disc is highly restricted by ANNEX D.3.

Peak signal levels above +3.10 dB SA-CD in quasi DC-50kHz bandwidth determined in ANNEX D.3 (MaxPeak/MP) are not allowed.

The accumulated RMS signal + noise level of the DSD signal in 40kHz-100kHz bandwidth determined in ANNEX D.4 is maximally equal to the RMS level of an input sinewave with a peak amplitude of -20 dB SA-CD.

 

[mansr]

How is 0dBFS for DSD determined?

No-brainer for PCM.

But what sets the limit (full scale) for DSD?

The 0 dBFS level is nominally set at half the quantisation level. The SACD spec permits brief excursions a few dB beyond this (I don't recall the details). With a higher signal level, the modulator easily destabilises, but there's no hard limit. The 50% value was merely a reasonable choice with a simple number.