Multi-bit DSD versus PCM

[NorthStar]

Bob Stuart had, and still does have have a choice in the design of his gear and in supporting DSD or not. He rejected DSD for his own technical, sonic and business reasons, hence his gear does not support it. It is not the other way 'round, as you would have it. He is not alone in the engineering and technical community. Many prefer PCM. Others prefer DSD. Anecdote: it is interesting to note that the late, great recording/mastering engineer, Doug Sax, of one-time Sheffield Direct-to-disc analog recording fame, tried DSD, then became a strong advocate of 192k PCM before he died a few years ago.

I try to be open minded, myself. My system supports DSD up to 4x on playback, and also PCM up to 384k. So far, I have yet to hear the compelling sonic difference afforded by DSD. I have also heard other's systems set up specifically for DSD, one at a DSD-advocate recording engineer's home studio, with outstanding components. Again, for me, there is no compelling sonic reason for me to switch to pure DSD playback, given DSD's disadvantages in playback processing, such as the lack of tools for bass management, speaker distance correction and most Room EQ.

...One of the reasons I posted this link: http://www.mojo-audio.com/blog/dsd-vs-pcm-myth-vs-truth/

 

[Blizzard]

...One of the reasons I posted this link: http://www.mojo-audio.com/blog/dsd-vs-pcm-myth-vs-truth/

The problem with that article is it's way outdated information. It talks about the shortcoming's mainly with the single rate SACD format. There's an brief mention of double DSD, but fails to explain that it deals with most of the shortcomings. Quad DSD or higher, which does deal with all of the shortcomings isn't explained at all, and rather brushed off as a "studio only" format.

It's pure marketing literature to justify their design philosophy of their product. It's funny they don't mention all the artifacts that come with outdated R2R technology?

 

[Al M.]

The problem with how most people think it's done is they are wrong. Berkeley is among this crowd. It's best to talk with Dustin Foreman of ESS or Jussi at signalist for a more accurate explanation.

Here is a quick reference you can read: http://www.aes.org/e-lib/browse.cfm?elib=9903

"We demonstrate these effects, and using coherent averaging techniques, are able to display the consequent profusion of nonlinear artefacts which are usually hidden in the noise floor. Recording, editing, storage, or conversion systems using single-stage, 1-bit sigma-delta modulators, are thus inimical to audio of the highest quality. In contrast, multi-bit sigma-delta converters, which output linear PCM code, are in principle infinitely perfectible. "

Berkeley is also of the opinion that multi-bit sigma-delta converters are PCM, and they have explained the Why in detail in two lengthy posts on Computer Audiophile. Miska from HQPLayer has another opinion. So the experts in the field disagree. Anything new under the sun?

 

[NorthStar]

The problem with that article is it's way outdated information. It talks about the shortcoming's mainly with the single rate SACD format. There's an brief mention of double DSD, but fails to explain that it deals with most of the shortcomings. Quad DSD, which does deal with all of the shortcomings, isn't mentioned at all. It's pure marketing literature to justify their design philosophy of their product. It's funny they don't mention all the artifacts that come with outdated R2R technology?

Yes, you're right Mike...this is old stuff...and so it lost the new today's approach...absolutely.
And that's the problem with most of us; we are stuck in the past with old extinct audio values. :b

Some stuff is still relevant today, but a bunch of it needs to be up-dated with the latest and best.

I agree that the audio world (in particular the ultra high-end) is a deceptive world full of crack jack pack pots. ...Like the caramel cracker jack substitute for popcorn...way of speech...with a surprise inside the box...a plastic gun or whatever. It sounds weird that comparison but reality is weird too in our audio world.
People cherish immensely their audio investment, and believe firmly in their 60+ set of ears. And they also believe what they read from the master audio salesmen.

Me I think you can obtain musical nirvana for less than $500,000

And your Quad DSD should be tested by more PCM people. And also by analog tube people (TT and R2R). But it is also sublime the vintage youth. Perfection, the best, is often the simplest, the lesser expensive, and the most rewarding in a less than perfect world.

I highly respect hi-end audiophile people who worked all their life to perfect and improve their audio rigs...turntables and all.
It's fun to contemplate all the cadillac of turntables, they are a beauty to see spinning their platter. Some platines can lend you north of $850,000!

Anyway, TTs don't do DSD and PCM. Off topic, just quickly mentioning the deep love affair I have with platines.

____________

I need to get Quad DSD before I die. I figure within the next few months...maybe a year or two max if I'm lucky.

And yes, there is a little bit of good humor in what I've just said; it'll make me live a little longer.

 

[Blizzard]

Berkeley is also of the opinion that multi-bit sigma-delta converters are PCM, and they have explained the Why in detail in two lengthy posts on Computer Audiophile. Miska from HQPLayer has another opinion. So the experts in the field disagree. Anything new under the sun?

All of the experts who actually make SDM DAC chips, along with Jussi are in the same camp. This is because they actually know how the products they created work. The actual engineers behind the chip wouldn't dare come on a public forum in a million years. I don't even know if they are allowed to by contractural agreement.

Some DAC manufacturers who buy chips from 3rd party chip manufacturers, come up to their own conclusions of how the products they didn't engineer work. Me personally will take the advice from the chip and software engineers who make SDM/SRC software, over the folks who simply solder the chips onto a board, and assume they know how they work based on looking at a block diagram provided by the chip manufacturer.

 

[Bruce B]

Here's the CA thread on the topic...

multibit DSD

 

[Blizzard]

Yes, you're right Mike...this is old stuff...and so it lost the new today's approach...absolutely.
And that's the problem with most of us; we are stuck in the past with old extinct audio values. :b

Some stuff is still relevant today, but a bunch of it needs to be up-dated with the latest and best.

I agree that the audio world (in particular the ultra high-end) is a deceptive world full of crack jack pack pots. ...Like the caramel cracker jack substitute for popcorn...way of speech...with a surprise inside the box...a plastic gun or whatever. It sounds weird that comparison but reality is weird too in our audio world.
People cherish immensely their audio investment, and believe firmly in their 60+ set of ears. And they also believe what they read from the master audio salesmen.

Me I think you can obtain musical nirvana for less than $500,000

And your Quad DSD should be tested by more PCM people. And also by analog tube people (TT and R2R). But it is also sublime the vintage youth. Perfection, the best, is often the simplest, the lesser expensive, and the most rewarding in a less than perfect world.

I highly respect hi-end audiophile people who worked all their life to perfect and improve their audio rigs...turntables and all.
It's fun to contemplate all the cadillac of turntables, they are a beauty to see spinning their platter. Some platines can lend you north of $850,000!

Anyway, TTs don't do DSD and PCM. Off topic, just quickly mentioning the deep love affair I have with platines.

____________

I need to get Quad DSD before I die. I figure within the next few months...maybe a year or two max if I'm lucky.

And yes, there is a little bit of good humor in what I've just said; it'll make me live a little longer.

Yes all true. Let me know where you're ready for a quad DSD solution because I know of a few low cost solutions that will blow your mind. Combine with HQplayer upsampling all to quad DSD, is pure bliss with any format around

 

[Al M.]

All of the experts who actually make SDM DAC chips, along with Jussi are in the same camp. This is because they actually know how the products they created work. The actual engineers behind the chip wouldn't dare come on a public forum in a million years. I don't even know if they are allowed to by contractural agreement.

Some DAC manufacturers who buy chips from 3rd party chip manufacturers, come up to their own conclusions of how the products they didn't engineer work. Me personally will take the advice from the chip and software engineers who make SDM/SRC software, over the folks who simply solder the chips onto a board, and assume they know how they work based on looking at a block diagram provided by the chip manufacturer.

As the folks at Berkeley have explained, they have designed converter chips themselves. So they obviously should know how it works. Again, a simple disagreement between experts -- just like in any field of engineering and science.

 

[Blizzard]

Here's the CA thread on the topic...

multibit DSD

We won't beat that thread here, because we don't have the experts here, like were on that thread. Mainly Jussi (Miska). Theres a few more on there as well with some solid information. Takes a while to sort the gold from the trash though. Best thing to do is to only read what Miska has to say, and who he responds to

 

[amirm]

I want to make sure my original point is not lost. I was saying that if a convolution kernel works by resampling the DSD to a multi-bit format, then whatever DSD was is lost at that point. We now have fixed or floating point representation of audio samples exactly as we do with PCM. To then convert this to DSD on the way out incurs another conversion. That then goes to some DAC which in turn may resample the bits to its favorite internal format. To call all of this some kind of pure end-to-end "DSD" makes zero sense to me. It is messy all around and each stage stands the chance to change the high frequency response making it slightly brighter or muted. And what to do with all the ultrasonic information in DSD in these format conversions. It is like taking four pre-amps of different designs and cascading them together.

In such a system, PCM stands far more superior. It starts and (almost) finishes as PCM.

At this point DSD is just a marketing name; it does not translate into an understandable technical term.

 

[Blizzard]

As the folks at Berkeley have explained, they have designed converter chips themselves. So they obviously should know how it works. Again, a simple disagreement between experts -- just like in any field of engineering and science.

Their philosophy is also a few years old now as well. At the time they designed their Alpha DAC's, it was before the time of better ways being fully recognized. Their recommendation of using Jriver to get the best DSD to PCM conversion possible proves this.

 

[Al M.]

We won't beat that thread here, because we don't have the experts here, like were on that thread. Mainly Jussi (Miska). Theres a few more on there as well with some solid information. Takes a while to sort the gold from the trash though. Best thing to do is to only read what Miska has to say, and who he responds to

And what the folks at Berkeley have to say whom he responds to

 

[Al M.]

Their philosophy is also a few years old now as well. At the time they designed their Alpha DAC's, it was before the time of better ways being fully recognized. Their recommendation of using Jriver to get the best DSD to PCM conversion possible proves this.

Their recommendation only proves that they are not very interested in DSD. And they are not the only ones who aren't.

Yet from the reports about the NADAC it seems clear that Quad-DSD is very good to excellent. I will soon hear for myself.

 

[amirm]

We won't beat that thread here, because we don't have the experts here, like were on that thread. Mainly Jussi (Miska).

Come again? What are we, pot of plant?

I just read the first few posts and he says this:

"PCM has the ugly limitation that you are much more bounded between transport and conversion. (you would have more freedom and less wasted bandwidth, by using delta-encoded PCM where you only transfer difference between two adjacent samples)"

Who says I can't use PCM as a transport just the same? Plenty of systems and DACs resample PCM and then convert that to analog much like how DSD is resampled. And they don't have to figure out what to do with the ultrasonic noise in DSD that comes for the ride.

Theres a few more on there as well with some solid information. Takes a while to sort the gold from the trash though. Best thing to do is to only read what Miska has to say, and who he responds to

If you are using him to make your point, you need to quote the part you want us to read. We can't be expected to believe things just as appeal to his highness.

 

[NorthStar]

Very very old articles:

• From Ayre: http://www.ayre.com/insights_dsdvspcm.htm

• Octuple-rate DSD: https://en.wikipedia.org/wiki/Direct_Stream_Digital

 

[Blizzard]

I want to make sure my original point is not lost. I was saying that if a convolution kernel works by resampling the DSD to a multi-bit format, then whatever DSD was is lost at that point. We now have fixed or floating point representation of audio samples exactly as we do with PCM. To then convert this to DSD on the way out incurs another conversion. That then goes to some DAC which in turn may resample the bits to its favorite internal format. To call all of this some kind of pure end-to-end "DSD" makes zero sense to me. It is messy all around and each stage stands the chance to change the high frequency response making it slightly brighter or muted. And what to do with all the ultrasonic information in DSD in these format conversions. It is like taking four pre-amps of different designs and cascading them together.

In such a system, PCM stands far more superior. It starts and (almost) finishes as PCM.

At this point DSD is just a marketing name; it does not translate into an understandable technical term.

The problem is, that's not how it does it. It converts it to "multibit SDM" like is done internally inside the DAC chip anyways, and then when it gets into the chip, the internal resource constrained SDM algorithms are bypassed since it's already been done by superior algorithms.

 

[amirm]

The problem is, that's not how it does it. It converts it to "multibit SDM" like is done internally inside the DAC chip anyways, and then when it gets into the chip, the internal resource constrained SDM algorithms are bypassed since it's already been done by superior algorithms.

Is there documentation somewhere I can read?

 

[Blizzard]

Their recommendation only proves that they are not very interested in DSD. And they are not the only ones who aren't.

Yet from the reports about the NADAC it seems clear that Quad-DSD is very good to excellent. I will soon hear for myself.

Yes some are slower to catch on than others I don't think theres any dispute from anyone that quad DSD isn't superior. What their end of the story is the most popular formats can be done better (PCM) using their approach. But they are also not using third party software based SDM/SRC when making this claim either. This is assuming a regular native format is sent to the DAC as is. And once again this is based chips from yesteryear as well.

 

[Blizzard]

Come again? What are we, pot of plant?

I just read the first few posts and he says this:

"PCM has the ugly limitation that you are much more bounded between transport and conversion. (you would have more freedom and less wasted bandwidth, by using delta-encoded PCM where you only transfer difference between two adjacent samples)"

Who says I can't use PCM as a transport just the same? Plenty of systems and DACs resample PCM and then convert that to analog much like how DSD is resampled. And they don't have to figure out what to do with the ultrasonic noise in DSD that comes for the ride.


If you are using him to make your point, you need to quote the part you want us to read. We can't be expected to believe things just as appeal to his highness.

Well we need 2 sides to have a proper debate, and I'm not going to claim to have the knowledge of Jussi.

This argument is quite outdated anyways because it's all old tech either way the opposition was referring to. This was a 2013 thread. Pretty much everyone on CA actually tried HQplayer after this and the arguments stopped. Now they only ask questions, and listen.

The bottom line is the modulators/filters built into DAC chips suck compared to what can be done on a powerful Intel processor. So if they are preformed in software, with an Intel processor, what is the best way to handle the data after that point?

This is what needs to be focused on.

 

[amirm]

Well we need 2 sides to have a proper debate, and I'm not going to claim to have the knowledge of Jussi.

This argument is quite outdated anyways because it's all old tech either way the opposition was referring to. This was a 2013 thread.

The bottom line is the modulators/filters built into DAC chips suck compared to what can be done on a powerful Intel processor. So if there are preformed in software, with an Intel processor, what is the best way to handle the data after that point?

This is what needs to be focused on.

Well, my focus was on what happens when you add digital processing in the middle as every system should. If the bits are taken out of DSD domain, then calling the stream still DSD is just wrong.