DSD comparison to PCM.

Ragnar

New Member
Feb 24, 2013
34
0
0
Near Seattle
DSD-Wide (2.8MHz/8-bit) is an interesting system, because it does not require noise-shaping, unlike DSD-Narrow (2.8MHz/1-bit). Although dither is still desirable (as it is in any digital system), it's not required for the system to operate, again, unlike DSD-Narrow. Although still awkward to level-adjust, equalize, or synthesize reverb, the conversion to the PCM domain is a lot more transparent, since high levels of dither are not required at each stage of conversion.

I am not sure how DSD-Wide is different from 8-bit PCM or a 6-bit ADC running at 11.2896MHz, such as those from TI. In either case, both dither and noise shaping are required if one wishes to get any more than 14-bit performance, and, I believe, the latter converter is easier to fabricate.

Further, how does one convert DSD-Wide to single bit DSD for commercial release without the supposed "accursed" conversion to some form of PCM!

Please all read Lynn's excellent articles in "Positive Feedback" plus all his links and then tell me it's resolution we're all after.
 

opus111

Banned
Feb 10, 2012
1,286
3
0
Hangzhou, China
Part Two of the Positive Feedback Online article is out:
http://www.positive-feedback.com/Issue66/dsd.htm

I note you've included Andreas' FFT comparing the noise floor of DSD with PCM @24bits. Those comparisons are utterly false because the audio band noise floor isn't shown on an FFT. What an FFT shows is a series of datapoints representing the noise power in each of the FFT output bins, a much, much smaller bandwidth than the total. Its appearing to show that the noise of DSD is below that of 24bit PCM which is total nonsense.

I'm left pondering if Andreas is really so ignorant about what's called the 'processing gain' of the FFT or whether this is a conscious attempt to mislead over the 'advantage' of DSD?

http://www.analog.com/static/imported-files/seminars_webcasts/MixedSignal_Sect5.pdf Page 5.19 and 5.20 along with figure 5.25 explain.
 

LL21

Well-Known Member
Dec 26, 2010
14,411
2,509
1,448
I note you've included Andreas' FFT comparing the noise floor of DSD with PCM @24bits. Those comparisons are utterly false because the audio band noise floor isn't shown on an FFT. What an FFT shows is a series of datapoints representing the noise power in each of the FFT output bins, a much, much smaller bandwidth than the total. Its appearing to show that the noise of DSD is below that of 24bit PCM which is total nonsense.

I'm left pondering if Andreas is really so ignorant about what's called the 'processing gain' of the FFT or whether this is a conscious attempt to mislead over the 'advantage' of DSD?

http://www.analog.com/static/imported-files/seminars_webcasts/MixedSignal_Sect5.pdf Page 5.19 and 5.20 along with figure 5.25 explain.

Interesting...too many facts and figures for non-techies like me. It does appear in this thread that many of the 'leading experts' here lean away from SD DAC architecture, with the exception perhaps of pure DSD all the way through from start to finish.

However, are you personally suggesting you lean away from DSD in all instances and that you find superiority with PCM at 24bit?
 

opus111

Banned
Feb 10, 2012
1,286
3
0
Hangzhou, China
Interesting...too many facts and figures for non-techies like me.

For the non-techies : the line showing the noise floor on 24bit PCM is drawn in the wrong place. It should be considerably below the line showing the noise of DSD. At the top end of the frequency range (getting close to 20kHz) the SNR of DSD is actually worse than for 16bit PCM. 24bit PCM is 48dB quieter than 16, so I reckon the line should be drawn just above the -200dB level for 24bits.

However, are you personally suggesting you lean away from DSD in all instances and that you find superiority with PCM at 24bit?

I have no experience of PCM at 24bits, for me 20bits is more than adequate in theory. Almost of my material is 16bits, that's satisfying and I have no inclination towards hires. That is at least until I've developed my own hires NOS DAC...
 

LL21

Well-Known Member
Dec 26, 2010
14,411
2,509
1,448
For the non-techies : the line showing the noise floor on 24bit PCM is drawn in the wrong place. It should be considerably below the line showing the noise of DSD. At the top end of the frequency range (getting close to 20kHz) the SNR of DSD is actually worse than for 16bit PCM. 24bit PCM is 48dB quieter than 16, so I reckon the line should be drawn just above the -200dB level for 24bits.



I have no experience of PCM at 24bits, for me 20bits is more than adequate in theory. Almost of my material is 16bits, that's satisfying and I have no inclination towards hires. That is at least until I've developed my own hires NOS DAC...

Thanks, Opus. Obviously, as a non-techie who has 100% redbook (incl maybe 15% hybrid SACD where i play the redbook layer only)...i have made my choice based on more practical requirements...best way (to me) to play my existing music collection. Still good to try to understand a bit more about what i am hearing and why.
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
Read Bruno's Q&A and have to disagree in one aspect. I'd much rather work with DSD files on the Sonoma system than my Pyramix system. The Sonoma is my main archival system for tapes.
 

LynnOlson

New Member
Feb 22, 2013
74
0
0
Read Bruno's Q&A and have to disagree in one aspect. I'd much rather work with DSD files on the Sonoma system than my Pyramix system. The Sonoma is my main archival system for tapes.

Hi Bruce, your real-world comments are much appreciated! What are the internals of the Sonoma and Pyramix systems; in other words, what is the internal format? DSD-Narrow throughout, DSD-Wide, DXD, or something different?

From my quite limited understanding of DSD, it is almost impossible to do anything to the DSD bitstream, including operations as simple as changing level, never mind more complex signal processing such as compression, equalization or adding reverb.

One of the things that mystifies me about the Playback Designs DAC is the claim that it upsamples 64fs DSD to 128fs DSD. How is this done, without conversion into PCM? The only thing I can imagine where the signal stays in the DSD domain is simply doubling the number of samples; for example, a sequence of 1010 would be doubled to 11001100.
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
Hi Bruce, your real-world comments are much appreciated! What are the internals of the Sonoma and Pyramix systems; in other words, what is the internal format? DSD-Narrow throughout, DSD-Wide, DXD, or something different?

From my quite limited understanding of DSD, it is almost impossible to do anything to the DSD bitstream, including operations as simple as changing level, never mind more complex signal processing such as compression, equalization or adding reverb.

One of the things that mystifies me about the Playback Designs DAC is the claim that it upsamples 64fs DSD to 128fs DSD. How is this done, without conversion into PCM? The only thing I can imagine where the signal stays in the DSD domain is simply doubling the number of samples; for example, a sequence of 1010 would be doubled to 11001100.

The Sonoma system works at DSD-wide only for the edit points as Pyramix works at 32/352.8 at the edit points. You can do basic copy/cut/paste without it doing any conversions. If you want to do gain changes, fades or splices, then the conversion is done at that samplerate. So if you splice 2 pieces together, those 2-3ms will be calculated at the coversion and then rendered back to DSD in both workstations.

I have the MPS-5 but don't know much about the innards. I did have to pop open the chassis a couple of times though.
 

LynnOlson

New Member
Feb 22, 2013
74
0
0
So for a multitrack master, with all the tracks recorded in native DSD format, when you want to adjust individual track levels for the final 2-channel mixdown, it goes through DSD-wide on the Sonoma workstation and 32/352.8 (DXD) on the Pyramix?
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
So for a multitrack master, with all the tracks recorded in native DSD format, when you want to adjust individual track levels for the final 2-channel mixdown, it goes through DSD-wide on the Sonoma workstation and 32/352.8 (DXD) on the Pyramix?

Correct....
 

LynnOlson

New Member
Feb 22, 2013
74
0
0
If I understand correctly, in the real world where multitrack recordings are the norm, nearly all SACD and DSD-download releases have passed through:

(A) DSD-wide (Sonoma)
(B) 32/352.8 PCM/DXD (Pyramix)
(C) are converted from PCM sources varying from 20/88.2 to 24/192
(D) are sourced from two-track analog mastertapes

If that's so, then "pure" DSD recordings hardly exist at all, in the narrow audiophile sense of 2 microphones -> preamp -> ADC -> DSD -> DAC -> consumer playback. What we're hearing are either all-analog sources, dating from the Eighties or earlier, or more modern recordings which are sourced and mixed in studio-grade PCM, DXD/PCM, or DSD-wide.

Bruce, you're the professional, and know the answers here. What about recordings from the mid-Eighties to the early 2000's, before Pyramix and Sonoma? The 24-track tape machines and EMT reverbs were gradually replaced by all-digital systems, at fairly low PCM resolutions (20/44.1?) at first, then 24/88.2 and 24/96 by the mid-Nineties.

I'm guessing these recordings from this era might be partially analog, but have passed through at least one stage of PCM of varying quality. If the 24-track analog master still exists, the recording could always be re-mixed in a purist Sonoma environment, but if the multitrack master is PCM (on disk or tape), there has to be a transcoding step into the world of DSD-wide, then a reduction to DSD-narrow.

Is this a correct picture? You have one of the few studios that specialize in DSD, so you must know more about the origins of SACD and DSD-download recordings than most of us.
 

tailspn

Member
Jun 28, 2011
169
0
16
Lynn, your questions would be best addressed to Andreas Koch, who I'm sure will reply. Both the EMM Labs and PD DAC's upsample to 128fs DSD for D/A conversion, regardless of the format and sample rate presented. They're both single sampling rate 128fs D/A converter back ends.

According to Andreas, the algorithms for any signal processing are the same, whether DSD or PCM. Again, he can provide the explanation details. Sonoma was originally available with a multi-channel mixer module, and software, that performed many of the more common post production processing functions, including level changing, EQ, etc. It contained multiple E-Chips, just like the fewer E-Chips contained on a Sonoma card. These functions were performed in DSD-Wide, which are 8-bit stand alone words, like PCM, but is derived without decimation. It's the decimation (low-pass filtering to avoid the Nyquist conflict at the lower new PCM sampling rate, that does the sonic harm., as opposed to the conversion from a derivative like series of samples (DSD) to separate stand-alone discrete level words (PCM).

Genex also made a 48 channel DSD mixer, which it sold a number of copies to Clear Channel, for their Hard Rock Cafe "instant" concert performance productions. It's also been used on several SACD pure DSD releases, and sounds stunning. The functions, particularly the track level changing performed in the E-Chip contained in Sonoma were performed in multiple FGPA's, contained on each mixer channel.
 

tailspn

Member
Jun 28, 2011
169
0
16
If I understand correctly, in the real world where multitrack recordings are the norm, nearly all SACD and DSD-download releases have passed through:

(A) DSD-wide (Sonoma)
(B) 32/352.8 PCM/DXD (Pyramix)
(C) are converted from PCM sources varying from 20/88.2 to 24/192
(D) are sourced from two-track analog mastertapes

A fifth category is Channel Classics, among possible others. That is doing an stereo analog mix at the session, setting all the individual mic balances, and archiving the analog output in DSD. The necessary edits then can be preformed with either a Pyramix, or Sonoma as Bruce explained, where only the edit interval is affected.
 

Bruce B

WBF Founding Member, Pro Audio Production Member
Apr 25, 2010
7,006
512
1,740
Snohomish, WA
www.pugetsoundstudios.com
If I understand correctly, in the real world where multitrack recordings are the norm, nearly all SACD and DSD-download releases have passed through:

(A) DSD-wide (Sonoma)
(B) 32/352.8 PCM/DXD (Pyramix)
(C) are converted from PCM sources varying from 20/88.2 to 24/192
(D) are sourced from two-track analog mastertapes

This is correct. If you have a setup like Channel and can capture in DSD and have no edits or need no processing, then this would be the ideal situation. Remember, I said just the edits are calculated in the conversion. The rest of the material is untouched.
When someone brings us a DSD file for mastering, I play directly from that DSD file into an analog console/processing and back into DSD. That's the purest form of DSD mastering I can think of.
Most recording/mix studios that I know of pretty much go the same rout, hopefully. I have never head a digital plugin that I liked. The only plugins that I use are for restoration.
There were 2 other workstations that could handle DSD. SADiE used the Sony E-chip and the other was Sonic. Don't remember how they implemented the files.

As far as Genex.... that equipment was best used as a boat anchor. I had the 9048, AD8 and DA8 and it was a POS!
 

LynnOlson

New Member
Feb 22, 2013
74
0
0
With the indulgence of the readers of this forum. I'd like to make a few subjective comments about the sound of DSD vs PCM.

I suspect most, or maybe all, of the edgy, flat, and tonally impaired "PCM" sound is the result of the wrong choice of electronics for the I/V converter and active lowpass filter. Nearly all opamps, and many transistor circuits (like the Marantz/Philips HDAM circuits), are too slow for PCM - by a factor of 100 or more. The slewing interval is very short, which means that it hardly shows at all in a long-duration FFT measurement. But it impairs Nyquist reconstruction of the waveform, and the departure from ideal reconstruction is greatest at high frequencies.

DSD, by contrast, has a spread spectrum in the 1~10MHz region, which results in very different IM products down in the audio band. In effect, all it does is raise the noise floor, instead of create a myriad of sum-and-difference tones.

What's behind this half-baked hypothesis? Well, what I hear from DACs with current-output ladder converters and non-slewing analog electronics is a sound a lot like professional-grade DSD. What underlined this was a direct comparison, of several DSD tracks, recorded at a very high quality level, played with two very different converters:

The first was the Invicta, which is a pretty decent ESS 9018-based DAC (and designed in close collaboration with the ESS team) with direct DSD playback. It might not be in the exalted class of the Playback Designs, but it's not that far off, either.

The second method was a lot clunkier and at first blush, most of us would expect a lot worse sound. Pure Music converted the 128fs DSD track to 24/88.2, which went to the PCM-1704-based Monarchy through the Resonessence Concero USB -> S/PDIF bridge. Based on the signal path, I expected pretty funky sound.

But ... in practice, both sounded about equally good. Neither sounded like "PCM", and the sound quality was actually excellent from both. In some ways, the Monarchy was better ... more spacious, more open, which is a good quality for symphonic material. The Invicta was a bit more "focused" and a bit more in-your-face, but certainly not obnoxious. Both were some of the best digital I've ever heard, on any system.

The Invicta and certainly the Monarchy are probably one or two steps below the "ultimate" DSD and PCM converters. Everyone mentions the Playback Designs as one of the best DSD machines out there. I have not heard the Phasure NOS1 or the TotalDAC, but these seem to be the pinnacle of ladder conversion technology, and both are free of the usual analog slew-limiting problems of conventional PCM DACs.

All three are unfortunately far out of my price range, so I won't be buying a Playback Designs, Phasure NOS1, or TotalDAC any time soon. (Sigh!) But I will encourage David Robinson to audition them in his MBL-based system, and try the same comparison that I've done in my own more modest system.
 
Last edited:

Andre Marc

Member Sponsor
Mar 14, 2012
3,970
7
0
San Diego
www.avrev.com
With the indulgence of the readers of this forum. I'd like to make a few subjective comments about the sound of DSD vs PCM.

I now suspect most, or maybe all, of the edgy, flat, and tonally impaired "PCM" sound is the result of the wrong choice of electronics for the I/V converter and active lowpass filter. Nearly all opamps, and many transistor circuits (like the Marantz/Philips HDAM circuits), are too slow for PCM - by a factor of 100 or more. The slewing interval is very short, which means that it hardly shows at all in a long-duration FFT measurement. But it impairs Nyquist reconstruction of the waveform, and the departure from ideal reconstruction is greatest at high frequencies.

DSD, by contrast, has a spread spectrum in the 1~10MHz region, which results in very different IM products down in the audio band. In effect, all it does is raise the noise floor, instead of create a myriad of sum-and-difference tones.

What's behind this half-baked hypothesis? Well, what I hear from DACs with current-output ladder converters and non-slewing analog electronics is a sound a lot like professional-grade DSD. What underlined this was a direct comparison, using the same tracks, of a DSD track, recorded at a very high quality level, played with two very different converters:

The first was the Invicta, which is a pretty decent ESS 9018-based DAC (and designed in close collaboration with the ESS team) with direct DSD playback. It might not be in the exalted class of the Playback Designs, but it's not that far off, either.

The second method was a lot clunkier and at first blush, most of us would expect a lot worse sound. Pure Music converted the 128fs DSD track to 24/88.2, which went to the PCM-1704-based Monarchy through the Resonessence Concero USB -> S/PDIF bridge. Based on the signal path, I expected pretty funky sound.

But ... in practice, both sounded about equally good. Neither sounded like "PCM", and the sound quality was actually excellent from both. In some ways, the Monarchy was better ... more spacious, more open, which is a good quality for symphonic material. The Invicta was a bit more "focused" and a bit more in-your-face, but certainly not obnoxious. Both were some of the best digital I've ever heard, from any system.

The Invicta and certainly the Monarchy are probably one or two steps below the "ultimate" DSD and PCM converters. Everyone mentions the Playback Designs as one of the best DSD machines out there. I have not heard the Phasure NOS1 or the TotalDAC, but these seem to be the pinnacle of ladder conversion technology, and both are free of the usual analog slew-limiting problems of conventional PCM DACs.

All three are unfortunately far out of my price range, so I won't be buying a Playback Designs, Phasure NOS1, or TotalDAC any time soon. (Sigh!) But I will encourage David Robinson to audition them in his MBL-based system, and try the same comparison that I've done in my own more modest system.

I've reviewed a bunch of Marantz digital and found them to be nothing short of excellent.

The new SA-11S3 SACD player and PCM DAC I just reviewed was absolutely stellar both as a disc player and DAC.

http://www.avrev.com/home-theater-audio-sources/dvd-audio-sacd-players/marantz-sa-11s3-sacd-player-dac.html
 

LynnOlson

New Member
Feb 22, 2013
74
0
0
What was your experience when you compared DSD to high-res PCM on the Marantz SA-11S3? Did SACD or DSD-download sound "smoother" and more analog-like than high-res PCM, or did they sound fairly similar?

If Marantz has improved the HDAM modules, that's good news all around. I'm also curious what converter the SA-11S3 is using ... does Marantz disclose this?

P.S. My prejudice against delta-sigma (or sigma-delta) converters seems to be a minority thing amongst most audiophiles, just like preferences for different kinds of power amplifiers.

P.P.S. I know that Marantz uses delta-sigma converters in the AV-8003 pre-pro that I use for my HT setup; of course, the AV-8003, aimed at the HT market, is a long way from Marantz's premium two-channel products.

(Back when I was shopping for a half-decent HT setup, the Marantz products were the only ones I could tolerate for more than 5 minutes. Most of the HT receivers were really gruesome sounding, worse than Pioneer, Sansui, and Kenwood receivers from the Seventies. To their credit, the AV-8003 and MM-8003 combo can actually play music, as well as the usual movie sound tracks with lots of explosions and SFX.)
 
Last edited:

Andre Marc

Member Sponsor
Mar 14, 2012
3,970
7
0
San Diego
www.avrev.com
What was your experience when you compared DSD to high-res PCM on the Marantz SA-11S3? Did SACD or DSD-download sound "smoother" and more analog-like than high-res PCM, or did they sound fairly similar?

If Marantz has improved the HDAM modules, that's good news all around. I'm also curious what converter the SA-11S3 is using ... does Marantz disclose this?

The SA-11S3 has USB, Toslink, and Coax digital inputs. All do 192/24.

I am a big fan of SACD, and when done correctly, I believe it is the best digital there is. No way to compare DSD
downloads to PCM on the unit.

Disclosure. I have zero interest in DSD downloads. I actually have a few since I am reviewing the MyTek DAc with Andy Schaub for PFO.

Yes, there is a certain "smoothness" to DSD/SACD, but the organic ease and analog like quality are worth the trade off in most cases.

Technical info here:
http://us.marantz.com/us/products/pages/ProductDetails.aspx?CatId=ReferenceSeries&ProductId=SA11S3

BTW, Ken Ishiwata, who designs much of the reference series is in my book brilliant.
 

NorthStar

Member
Feb 8, 2011
24,305
1,323
435
Vancouver Island, B.C. Canada
...
P.P.S. I know that Marantz uses delta-sigma converters in the AV-8003 pre-pro that I use for my HT setup; of course, the AV-8003, aimed at the HT market, is a long way from Marantz's premium two-channel products.

(Back when I was shopping for a half-decent HT setup, the Marantz products were the only ones I could tolerate for more than 5 minutes. Most of the HT receivers were really gruesome sounding, worse than Pioneer, Sansui, and Kenwood receivers from the Seventies. To their credit, the AV-8003 and MM-8003 combo can actually play music, as well as the usual movie sound tracks with lots of explosions and SFX.)

---- Hi Lynn; are those Marantz components still in your possession? ...And playing?
Are they kind of a Reference to you?

* The AV8003 uses indeed a delta-sigma Cirrus Logic CS4382A 192-kilohertz/24-bit DAC, which is an eight-channel DAC on a single chip. ...And the Cirrus Logic CS5361 ADCs.
It also has HDCD decoding; the 'Microsoft' Pacific Microsonics PMD-200 decoder/digital filter. ...And XLR balanced inputs and outputs.

As for Audyssey; only Audyssey MultEQ, and with some restrictions, because of limited DSP horsepower. ...Totally inadequate IMO.
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing