DSD Battle Royale!

[dallasjustice]

I agree that this is fun stuff to discuss in the abstract. I personally enjoy Julf's technical banter on this forum. But it's important to apply these things to real system playback scenarios.

PCM playback can utilize its advantages to surpass the best analog volume controls available, IME.

If done right, PCM playback can also can significantly and exclusively contribute to much improved in-room bass response in almost every small to medium size room.

Now, there are Jussi's claims concerning something he calls multi-bit SDM. He says that he can basically do the above in the 1 bit DSD domain, sort of. Even if that's really true, there's no widely accepted standard there and his HQplayer is no Jriver, in terms of usability. I recognize the "purist" advantages to DSD. But having heard the format for myself in it's natural habitat, I would still easily trade that advantage for the above PCM advantages.

The fundamental problem for PCM is there are no good native PCM A/D converters available today, outside of the few remaining Pacific Microsonics PM1 and PM2, and virtually no labels using those that are available. The available ADC's today are all multi or single bit PDM, that must be decimate filtered from their sampling rate to whatever PCM rate is used for the recording. That filtering process is quite discernible. IMO, that effect is far greater than the perceived problems often associated with DSD.

 

[tailspn]

I recognize the "purist" advantages to DSD. But having heard the format for myself in it's natural habitat, I would still easily trade that advantage for the above PCM advantages.

My view is different, since my interests are probably different than most. I'm interested in recreating, as closely as possible, an acoustic music event. To that end, I'm associated with the recording production business, and hear projects at all the stages. DSD recording, with just editing and no format conversions or post process sweetening, in multi-channel, allows me to come closest to my objective. You are most certainly correct that PCM allows far greater flexibility in not only the recording side, but also the consumer/listener side. But for me, it's too high a sound quality price to pay.

 

[tailspn]

How many recordings do you have that you know for sure are really native DSD from ADC-recording-DAW-mixing-mastering-to consumer digital file without any conversions?

Better than 80% of the Channel Classics catalog, the vast majority of the Telarc Catalog, most projects from the Super Audio Center, including a large percentage of their Acoustic Sounds projects, much of the analog remasters done by Bruce at Puget Sound, Blue Coast's projects, and MANY small volume projects done by small labels requiring only editing, and no post process sweetening

The filtering process in your context should only be discernible once you drop below a certain bit depth and sampling rate (Bruno Putzeys has done an article based upon research when at Philips Research lab where it is transparent transcoding between them); 16bit/44khz is not applicable to any of these discussions IMO as it is flawed in terms of constraints applied to digital filters, that is one reason PCM hirez can be better than CD quality with implementation of ideal reconstruction filters that do have notable slow rolloff.

You can theorize all you want, and cite all the papers you choose, but in a studio environment, you can easily distinguish between DSD and PCM formats, and the results of format conversions. That's not to say that you may not prefer the processed/converted product. It may in fact actually sound better, or at least more pleasing. After all, that's what sweetening's all about. It's just not what the original mic/mix feed were, which is what I'm after.

 

[LL21]

It is academic tbh.
DSD is converted for editing, and as you say we have Delta-Sigma ADC influencing PCM.
What we can focus on though is the behaviour of the signals in the real world studio and the associated processes, and implementation at the DAC including architecture.
How many recordings do you have that you know for sure are really native DSD from ADC-recording-DAW-mixing-mastering-to consumer digital file without any conversions? Orb

Better than 80% of the Channel Classics catalog, the vast majority of the Telarc Catalog, most projects from the Super Audio center, including a large percentage of their Acoustic Sounds projects, much of the analog remasters done by Bruce at Puget Sound, Blue Coast's projects, and MANY small volume projects done by small labels requiring only editing, and no post process sweetening...

Very interesting read...thanks for posting. Unfortunately, I could not go for the best of both DSD and PCM, as I understand it is rare to find a player that does both as well as one could hope for the money...and the above labels would represent less than 25% of my music (though still a few hundred albums). Much of my music remains plain old redbook, and I have focused on redbook playback. I am intrigued by players that actually provide 2 D/A conversion methods in one unit (Light Harmonics, I believe)...seeking to provide best of both. For now, I am happy to enjoy my redbook until more of my music becomes available on true hi-res.

 

[microstrip]

Better than 80% of the Channel Classics catalog, the vast majority of the Telarc Catalog, most projects from the Super Audio center, including a large percentage of their Acoustic Sounds projects, much of the analog remasters done by Bruce at Puget Sound, Blue Coast's projects, and MANY small volume projects done by small labels requiring only editing, and no post process sweetening
(...)

Nice to know. Can you point me a few of what you consider the best Channel Classics chamber music recordings? How can we know what are the 20% of the Channel Classics that are not "pure" DSD?

 

[tailspn]

Nice to know. Can you point me a few of what you consider the best Channel Classics chamber music recordings? How can we know what are the 20% of the Channel Classics that are not "pure" DSD?

I personally enjoy recordings by the Holland Baroque Society, Amsterdam Sinfonietta, and Ebony Band on Channel. Here's a link to their chamber music catalog:

http://www.channelclassics.com/chamber-music.html

But Jared is the best person by far to advise on this subject.

Large orchestra scale recordings, with many spot microphones, are those that occasionally need re-balancing in sections, and very rarely the smaller groups. For instance, the recent BFO Wagner Gotterdammerung Immolation scene was re-balanced for the soprano, Petra Lang, needed to be brought forward.

 

[Julf]

Better than 80% of the Channel Classics catalog, the vast majority of the Telarc Catalog, most projects from the Super Audio Center, including a large percentage of their Acoustic Sounds projects, much of the analog remasters done by Bruce at Puget Sound, Blue Coast's projects, and MANY small volume projects done by small labels requiring only editing, and no post process sweetening

So yes, the case for DSD seems to be either direct transfers of analog masters (the purpose it was originally designed for, back in the days of steam railways) or direct-to-disk recordings of the output of an analog mixing chain - cases where no editing or processing is needed. Unfortunately that doesn't cover much of modern music being produced today.

 

[Julf]

PCM playback can utilize its advantages to surpass the best analog volume controls available, IME.

If done right, PCM playback can also can significantly and exclusively contribute to much improved in-room bass response in almost every small to medium size room.

Both are valid points.

Now, there are Jussi's claims concerning something he calls multi-bit SDM. He says that he can basically do the above in the 1 bit DSD domain, sort of.

Multi-bit SDM is nothing new or special - and that is precisely what "DSD-wide" is - and as the name says, it is not 1-bit. There has to be a conversion from 1-bit PDM to multi-bit PDM (where the delta or "change" from one sample to the next is represented with more than one bit), and another conversion back to 1-bit. That is the compromise with 1-bit PDM - while it is easy to convert from analog to 1-bit PDM and back, any manipulation of the data (to change gain/volume/level, to do EQ or filtering, or whatever) requires conversion into a "processable" format.

 

[Orb]

Tailspn,
just to clarify when I mention transcoding can be transparent, this does not mean DSD/DXD-PCM sound the same (I agree they do not as I mentioned earlier that I am glad we have both "formats" as a consumer option).

Regarding the technicalities of the DAC chipset/architectures I am happy to let this drop as we already have a massive thread on the subject with DSD and PCM, although it should be noted not all chipsets and their architecture/functionality are made equal and this complicates the real world from recording to consumer (their DAC) for both PCM and DSD.

Cheers
Orb

 

[tailspn]

Multi-bit SDM is nothing new or special - and that is precisely what "DSD-wide" is - and as the name says, it is not 1-bit. There has to be a conversion from 1-bit PDM to multi-bit PDM (where the delta or "change" from one sample to the next is represented with more than one bit), and another conversion back to 1-bit. That is the compromise with 1-bit PDM - while it is easy to convert from analog to 1-bit PDM and back, any manipulation of the data (to change gain/volume/level, to do EQ or filtering, or whatever) requires conversion into a "processable" format.

Think of multi-bit PDM as multiple 1-bit in parallel. Then the transcoding is not so ominous as you think. In any case, it's the results that count. Going in and out of a Sonoma (1-bit to multi-bit and back to 1-bit PDM) has MUCH less actual sonic impact on the resulting audio than converting from any digital format to analog, then back for processing.

 

[tailspn]

Tailspn,
just to clarify when I mention transcoding can be transparent, this does not mean DSD/DXD-PCM sound the same (I agree they do not as I mentioned earlier that I am glad we have both "formats" as a consumer option).

But isn't the issue really to avoid changing the "sound" of the resulting audio? That is what transparent means, isn't it?

 

[Julf]

Think of multi-bit PDM as multiple 1-bit in parallel.

But it isn't. The 1-bit stream doesn't magically split into parallel streams. There has to be a conversion from 1-bit to multi-bit.

 

[microstrip]

Think of multi-bit PDM as multiple 1-bit in parallel. Then the transcoding is not so ominous as you think. In any case, it's the results that count. Going in and out of a Sonoma (1-bit to multi-bit and back to 1-bit PDM) has MUCH less actual sonic impact on the resulting audio than converting from any digital format to analog, then back for processing.

Sorry but I can not - in multi-bit the different bits have different weights. IMHO, the essence of PDM is that the bit has always the same weight. Perhaps we can consider that the errors associated to of a small number of bits in a multibit are less offending subjectively, but it is not easy to accept!

Can you explain what you mean by "back for processing"?

BTW, thanks for the chamber music references.

 

[Orb]

But isn't the issue really to avoid changing the "sound" of the resulting audio? That is what transparent means, isn't it?

TBH tailspn all DACs (not talking within the DAW at the studio) change the sound subtly due to their implementation of noise shaping/filters/coefficients; I understand the ESS Sabre DACs do not even keep the data in native PCM or DSD but changes both to an internal format for their hyperstream.
If all of them did not we would not see various preferences as we did with Bruce's recent downloads (big thanks for doing this for us hobbyists).

Anyway.
To quote directly from a paper at Grimm Audio:

It was found that it was possible to convert a DSD signal to 352.8kHz/32 bit and back without incurring any audible quality loss, as long as good care was taken with the filtering and remodulation stages.
It follows that using the same software it is also possible to convert from DXD to DSD and back without perceivable quality loss.
If a converter or DAW cannot convert from DXD to DSD and vice versa transparently, this means the processing is incorrectly implemented.

The above is specifically about transcoding between DXD and DSD (it is still possible once in PCM to then downsample this further).
However it means that by being transparent the music will not have artifacts or loss of "information", but as mentioned before the end result will sound different due to chip-dac architecture/filter-coefficients/etc.

Tailspn post also on Hydrogenaudio and inform them that 96kHz/24bits PCM is not completely capturing the music/soundwaves and let me know the conclusion
Sure they will get more enthusiastic than me as I am more interested in the studio process/application of dither-digital processing-etc/native sampling rate-bit depth/DAC filter all being done correctly.
Cheers
Orb

 

[Julf]

Tailspn post also on Hydrogenaudio and inform them that 96kHz/24bits PCM is not completely capturing the music/soundwaves and let me know the conclusion

That's cruel!

 

[tailspn]

TBH tailspn all DACs (not talking within the DAW at the studio) change the sound subtly due to their implementation of noise shaping/filters/coefficients; I understand the ESS Sabre DACs do not even keep the data in native PCM or DSD but changes both to an internal format for their hyperstream.
If all of them did not we would not see various preferences as we did with Bruce's recent downloads (big thanks for doing this for us hobbyists).

Anyway.
To quote directly from a paper at Grimm Audio:

The above is specifically about transcoding between DXD and DSD (it is still possible once in PCM to then downsample this further).
However it means that by being transparent the music will not have artifacts or loss of "information", but as mentioned before the end result will sound different due to chip-dac architecture/filter-coefficients/etc.

I agree with your first statement, and I'm limiting my comments to the recording/post processing side of the process only. The issue to me is how to capture digitally a mic/analog feed accurately, and have them be perceivably identical when compared. In my experience, we're not there yet, but getting tantalizing close with the best ADC's, including the Grimm and Merging Hours in DSD.

Jared and I talked with Bruno and Eelco for hours over the past few months as to why the conversion of DSD>DXD>DSD is not transparent, and what can be done to improve it. Bruno insists it's in the filter implementation. Whatever its cause, it's clearly demonstrable. In the meantime, several industry heavyweights are pushing for updating/expanding the multi-bit PDM tools for post processing approach.

96KHz/24 bit PCM? Get real

 

[Orb]

Yes it is clearly demonstrable with the filter implementation (that has always been one of my caveat), but did you also ask Bruno if with the right filter-processing it can be transparent?
You should point out to him they have done a paper on their website clearly stating that it can be transparent.
Yeah 24/96 guaranteed to annoy everyone (those who want it higher and those that want it lower)

Cheers
Orb

 

[tailspn]

Yes it is clearly demonstrable with the filter implementation (that has always been one of my caveat), but did you also ask Bruno if with the right filter-processing it can be transparent?

Sure, and he adamantly insists it can be. But please understand that there's a practical disconnect between the experiences of the people who do the recording and hear all the time the live, the analog mix, the resulting digital, and those who design hardware and theorize, listening to content from god knows where. It's hard to design when the reference content has already had the finest details stripped out from previous release post processing. Since Jared Sacks, Eelco and Bruno all live within miles of one another, they're embarking on several projects to optimize the conversion processes, as well as optimize all the 64fs DSD content already recorded, but wanting to be played back at higher DSD sampling rates by customers.

 

[tailspn]

Can you explain what you mean by "back for processing"?

Sorry, probably poorly worded. There are many mastering houses and studios, Bruce's Puget Sound for one, preferring to do some amount of post processing using analog components, rather than digital workstation plug-ins. When the content is in a digital format form, it needs to be converted back to analog, processed in analog, then re-digitized to a digital format for release.

 

[tailspn]

But it isn't. The 1-bit stream doesn't magically split into parallel streams. There has to be a conversion from 1-bit to multi-bit.

Hi Jud,

In the interest of my limited understanding, there's a discussion about multi-bit PDM going on now at Computer Audiophile:

http://www.computeraudiophile.com/f...version/multibit-dsd-debate-18437/#post276649

that talks to our points. It's in Miska's post 38. Hopefully you and others will find it interesting.

I hope it isn't against forum rules to reference other sites.