DSD comparison to PCM.

[LL21]

That if the recording engineering community at large believed "PCM done right" converters were a significant contributor to better sound, there would be numerous manufactures building them. The word gets around, and except for a few evangelists, there is no push in that direction that I'm aware of. Just the opposite. The move is towards native DSD recording for acoustic music recorded in a natural environment. It just sounds more like the analog mic feeds.

Not necessarily...look at the Beta vs VHS fight...everyone knows Beta is better and it was used for years after in the commercial space. But it died in the consumer market and for 10 years we all lived with the inferior product. i am not convinced that decisions are purely on quality...they are sometimes also on marketability, ease of use of formatting, transferability, who knows.

As for which is actually better, i am reading about 2 different camps in this thread...cost set aside...some prefer pure DSD and the pure PCM guys feel like few have heard their format properly. I am here to learn about the quality aspect...though i admit that i ultimately, too, have made my decision based on commercial reality...i am focused on 16/44 because that is 99% of what is available in music i wish to buy.

 

[PeterSt]

That if the recording engineering community at large believed "PCM done right" converters were a significant contributor to better sound, there would be numerous manufactures building them. The word gets around, and except for a few evangelists, there is no push in that direction that I'm aware of. Just the opposite. The move is towards native DSD recording for acoustic music recorded in a natural environment. It just sounds more like the analog mic feeds.

Famous quote eh ?
Ok, in the same realm I was told I never could have created anything good because when it was (PCM1704) the big guys would have chosen that too and the chip cookers would have cooked more of this breed.
But it really doesn't work like this.

Watch the new future, where all will be based on less power consumption as the sheer reason for something new. Remember this text when it happened ...
(and we will be stuck to *that*)

 

[tailspn]

But even that is only half the story because I am sure you will find it is also being pushed that DSD is processed in the Studio/DAW as DXD-PCM 352.8/32bit and from this to whatever is required whether 192/24 or DSD.

It's true that labels/production companies that track each mic at full level in DSD, then do the balance and mix in post production in DXD (Pyramix DAW's), are actually producing a DXD product, not a native DSD. But there are labels like Channel Classics and Blue Coast, among others that do the balance and mix in analog at the session, archive that in DSD, and only do editing in DXD with no other post production. Pyramix supports butt splices in DSD, and crossfades in DXD for the crossfade interval. That's tens of milliseconds typically.

As I've posted before, when you compare the DSD original take to a DXD processed copy of the same file, you can easily detect which is which. The low level information that enhances our ability to resolve space is reduced when transferred to DXD, even with no post processing like level shifting. PeterSt is correct. DXD is the most ordinary 24/352.8 (PCM)

Recording engineers and producers have an advantage that the rest of us do not. If they do an analog mix at the session, they can compare the actual mic feeds to the recorded, regardless of the converters and formats used. As Lee and other engineers have experienced, the DSD archival sounds the closest to the mic feeds.

 

[mep]

Recording engineers and producers have an advantage that the rest of us do not. If they do an analog mix at the session, they can compare the actual mic feeds to the recorded, regardless of the converters and formats used. As Lee and other engineers have experienced, the DSD archival sounds the closest to the mic feeds.

And yet people who weren't there want to argue that the recording engineers don't hear what they think they heard.

 

[PeterSt]

Recording engineers and producers have an advantage that the rest of us do not. If they do an analog mix at the session, they can compare the actual mic feeds to the recorded

And yet people who weren't there want to argue that the recording engineers don't hear what they think they heard.

This is all a bit dangerous;
I have been talking to the recording engineer of Turtle Records (really the best out there IMO), and I was told that during the session all compared well. After the session they went out for dinner and some coffee, came back and listened again, and ... wondered how they ever could have judged this to be OK before.

This is similar to why I personally don't like (and can't) A-B. Once you heard a thing, you will keep on hearing it. It just depends on which (A or B) you heard first.

This is how I dare argue both above quotes.
0c of course. But think about it.

 

[mep]

This is all a bit dangerous;
I have been talking to the recording engineer of Turtle Records (really the best out there IMO), and I was told that during the session all compared well. After the session they went out for dinner and some coffee, came back and listened again, and ... wondered how they ever could have judged this to be OK before.

This is similar to why I personally don't like (and can't) A-B. Once you heard a thing, you will keep on hearing it. It just depends on which (A or B) you heard first.

This is how I dare argue both above quotes.
0c of course. But think about it.

Maybe that had something more in their coffee than just coffee.

 

[Orb]

It's true that labels/production companies that track each mic at full level in DSD, then do the balance and mix in post production in DXD (Pyramix DAW's), are actually producing a DXD product, not a native DSD. But there are labels like Channel Classics and Blue Coast, among others that do the balance and mix in analog at the session, archive that in DSD, and only do editing in DXD with no other post production. Pyramix supports butt splices in DSD, and crossfades in DXD for the crossfade interval. That's tens of milliseconds typically.

As I've posted before, when you compare the DSD original take to a DXD processed copy of the same file, you can easily detect which is which. The low level information that enhances our ability to resolve space is reduced when transferred to DXD, even with no post processing like level shifting. PeterSt is correct. DXD is the most ordinary 24/352.8 (PCM)

Recording engineers and producers have an advantage that the rest of us do not. If they do an analog mix at the session, they can compare the actual mic feeds to the recorded, regardless of the converters and formats used. As Lee and other engineers have experienced, the DSD archival sounds the closest to the mic feeds.

I agree DXD is just another PCM format as I said in my own post.
It has come out earlier in this thread where I mentioned one of the 1st ones to study DSD to PCM transcoding was Philips (and Bruno Putzey) who reported you needed 352.8khz (I thought 32bit rather than 24 in their research), and from this probably why we see DXD today.
In my edit I clarify that DXD is only 24bit, which is less than what a few of the DAWs can do these days at 32bits.

Regarding those natively DSD, only problem though is as I mentioned you are talking about niche solution for a niche music market, now that is pretty niche if I must say

If transcoding to 352.8/32 is totally transparent as Philips labs and Bruno report, then it it marginalises the case for native DSD even if there are those who say DXD is not transparent (which is not 32bits).
I think Bruce has mentioned they take DSD (Grimm AD1 to DAW) and can process this as PCM 352.8/32 in the DAW, and from what I remember in another thread without any negative consequences.

Just wondering; is the DXD 24 bits specific to when ADC to the DAW, or limit 24bits for all aspects including internal DAW processing?
Curious as Pyramix supports 352.8/32 PCM for DSD transcoding, which is what should be used if going by Bruno and Philips.
Cheers
Orb

 

[Orb]

OK looking I can see that DXD in the DAW is 32bits minimum, 24bits relates to complete chain if using DXD end-to-end.

CHeers
Orb

 

[opus111]

That if the recording engineering community at large believed "PCM done right" converters were a significant contributor to better sound, there would be numerous manufactures building them.

Oh so now your argument is about the beliefs of the recording community. OK, I agree - they don't have that belief.

The word gets around, and except for a few evangelists, there is no push in that direction that I'm aware of.

I agree again, there's no push in the direction of PCM done right. There's no need for push, that's only used when there's an agenda (like a marketing one for example).

Just the opposite. The move is towards native DSD recording for acoustic music recorded in a natural environment. It just sounds more like the analog mic feeds.

Let me enunciate this clearly - the push is in the direction of DSD, and again I agree.

 

[opus111]

Can anybody here clearly state just what is wrong with redbook CD? ie, a summings up?

Its not 'wrong' but the time domain response is sub-optimal by the sampling frequency being too close to the top of the wanted band. The rest is implementation.

 

[bbb]

That if the recording engineering community at large believed "PCM done right" converters were a significant contributor to better sound, there would be numerous manufactures building them. The word gets around, and except for a few evangelists, there is no push in that direction that I'm aware of. Just the opposite. The move is towards native DSD recording for acoustic music recorded in a natural environment. It just sounds more like the analog mic feeds.

I think it is a bit naive to think the recording industry at large is really bothered about what sound best or take consumers' interests to heart. Of course there are a few who care but these are generally very small player who do not have any influence at all. History tells us that the industry will only move to what's cost effective and if they can charge you more for it, all the better. Of course they will come up with all kinds of marketing speak to convince you that you are getting better quality. Perfect sound forever? Ha Ha.

This extend to just about everything that used to be high end. Now you pay top dollar for Italian designer suits that's fused together in a factory in Asia. Patek's (and nearly all others) that's 95% machine made (oh but they'll tell you it's handmade as the last 5% of polishing work is by hand in Swiss) that cost way much more than those 30 year old handmade one's. Tiffany, whose jewelery used to be heirloom stuff but is now mass market masquerading as luxury. I could go on and on. The only exception is probably mass consumer electronics/computers where you consistently get lower prices and better products.

We as consumer will never win this race. In the end, there will always be a small handful of real high end, a lot more pretending to be high end and then there's the rest. And unlike some members here who have access to unobtainium, the rest of us will make the best out of whatever that's imposed on us. I chose to drop out of the race and hope like hell there's still cheap CD to be had till the day I am gone. Oh and a reasonable and good condition 993 tt somewhere down the road.

 

[PeterSt]

[ Please notice this is a bit difficult for me to explain in English ]

If transcoding to 352.8/32 is totally transparent as Philips labs and Bruno report, then it it marginalises the case for native DSD even if there are those who say DXD is not transparent (which is not 32bits).

Maybe we must be careful not to see Gods where Gods do not exist. IOW, no way I'd trust anyone for his ears just like that and the text from above certainly is not to be leading his own life (or start to be). On that matter we could just as well say that SACD *had* to be the better format because Philips (/Sony) developed it, Bruno Putzeys was part of it ... and such ? Haha, no. It could be the best format, but not for these reasons.

Fact is that a relative strange many Dutch individuals come up with more or less "extreme" products for audio and it won't be a coincidence they ever back worked at Philips. This, btw, also includes "Grimm".

When I read something like above quote it can only tell me something when the technical merits of the person('s background) are well known, and well, we can't be negative about our Bruno here (who btw is Belgian I think). On the other hand, the parameters for audio are so broad that mistakes are made easily, and will be made. And apart from that, when time proceeds audio progresses (vastly). Read : Above quote is obsolete before you know it.

Ok, so you lost me by now.
What I wanted to say is this :

No person is able to "state" that a format is completely transparant as long as 100% (audio) reality is not achieved. But his ears accepted something, apparently.
No person is able to "state" that a format is completely transparant as long as it is technically not right. And this is digital audio, always - NOT right.

I too claim that with native 24/352.8 recordings the big bang suddenly happens (remember, played through pure PCM). So, where hires in general is flawed (for me) always, the way 2L records it in 24/352.8 suddenly everything works. However, keep this in mind :

What's special about "DXD" recordings is that they are not to be filtered. That is, this is how 2L is doing it explicitly because they felt it is not needed. So, the format allows for 176.4KHz of frequency and since it is assumed that nothing exists from normal instruments and voices above that, indeed no filtering would be needed and the result would be technically right (no ringing, no aliasing and such). This is not the case for e.g. 24/192 because 96KHz frequencies might exist, and so that *is* filtered (by the ADC).

If we see this in the context of DSD, then what to say. It's more filter than audio so to speak. IOW, mangled all over and we *still* end up with HF noise all over (but which can be filtered analoguely to some (sufficient ?) degree.

Back to my lead-in ... If you now read Bruno's "statement" about transparency, he can be right *if* it is to be seen in the context of no filtering needed. But I really wonder whether he saw it in this context and whether he not merely used his ears. 2L did it for the technical aspects (2006 or earlier) and it really really works (their recordings).

Are we now getting somewhere with DSD vs PCM ?

 

[PeterSt]

Where bits are not bits

Please skip when you are sensitive to something which could be taken as blah, and I only pose it for the good cause of understanding how we compare audio (without being able to listen together).

My objective with the playback software (XXHighEnd) is to achieve the most 1:1 playback from the file. If you understood my previous post about the merits of 24/352.8 (some call that DXD) then you will understand that for this format this is possible. No further filtering of any sort is needed, and I indeed can play back the digital data 1:1. Electrically I proceed with this (NOS1) by having the DAC as fast as possible, and here the 1:1 exhibits as a best impulse response, or more frunctionally for "music" : transient response.
Easy.

Now watch out; for the further story it is best to see the 1:1 as described above, while this can not be true for Redbook CD data. It *has* to be filtered, and so I do. But, now it is not 1:1 anymore. However, the filter I created is lossless hence "consistent" and it can be regarded 1:1. Volume the same thing (the attenuated file can recreate the original). So, for the below the 1:1 is to be taken as the base. Here goes :


The software plays "bit perfect" in all circumstances. Thus, what goes into the D/A was in the file. No single bit is changed.
Numerous dials and settings exist in the software that alter the sound quality. And no, still no single bit is changed. Never forget this.

Especially recording engineers and loudspeaker designers will know how easy it is to make ladies spit audibly in microphones or how tongues can annoyingly click. Right ? So, by software this can easily be brought forward by turning the right dial. Or moved out. No-ho, no bits change anywhere !!
Bass can be made more tight or more deep (and not good). Same story. Standing waves can be eliminated completely (I think 6moons is planning an article about that).

Sound can change so drastically that it is beyond even my own belief that all is still but perfect. Still it is. And as I put it forward some times : "this new version makes your amps worth 20K more".

The moral ?
There is no single way that "we" can talk eachother into any format sounding better when not first this is taken into account and taken out of the equation. The playback software just matters too much. This is nothing which can be explained in a single post, and not even in a 1000 when you see that a complete forum (phasure) is working on this for almost 7 years now, and STILL we can create improvement with all our ears and finding better settings and tweaks. Worse : the improvements go faster the more close we are to reality (this is logic because judgement becomes less and less subjective because we can compare to live-reality).

Peter

 

[Orb]

Heya Peter,
Well I am mentioning Bruno (who worked within research and acknowledged as an excellent engineer) and Philips Research because their labs did evaluate what is required for transcoding DSD, and after all Philips has a pretty good idea of digital music coding-playback wouldn't you agree considering they were there at the very beginning and involved in its creation.
BTW Bruno also has the advantage of being heavily involved with the development and manufacturing of the the Grimm AD1 and clock they sell for studios, now ask Bruce and others what they think of that DSD unit.
So if he with a small team can make one of the best DSD pro ADC (not just Bruce who recommends this unit and also the clock is well reviewed by pro mags), then it is worth taking note what he says IMO.

Not suggesting he is a god, but same way I have a lot of respect for Walt Jung and his work, even though many complain about op amps (too broad a statement I agree) it is worth looking at what he says in papers-lectures and his op amp cook book.
Cheers
Orb

 

[PeterSt]

Channel Classics

Here's an easy test. Go get a copy of any Reference Recording production (recorded with the PM ADC), and compare it to any recent (Grimm ADC) Channel Classics recording. You be the judge.

Encouraged by this, last night I did something I never did before :
I looked for as many as possible various CDs which were taken from DSD masters and did the same for PM masters (all from RR). Then I compared.

Maybe I was a kind of shocked that indeed the DSD deriviates were sounding more analog.
But as I have said earlier in the thread (I think), maybe "analog" must be redefined into something which just doesn't carry/exhibit all the dynamics digital can. So how does this compare to the PM recordings ?
More dull.

I say "more dull" (and not "dull") because the DSD's just lack the sprinkle and sparkle the PM recordings show while the DSDs are not bad at all.

And now it becomes difficult, because how easy will it be for most of the world to perceive the PM recordings as "digital" and harsh (better : more digital, more harsh), just because it needs a better playback chain to pass that on undistorted.

So, at comparing the both the PM wins. With HDCD encoding even more so (highs get more refined).
When not comparing (as I always just "listened") no complaints anywhere.

The PM recordings are more detailed, have more air around the instruments, and maybe I noticed a bit farther sound stage (which btw seems to be a property of more detail, as we found at Phasure the other day).

The DSD recordings exhibit a more analog "Moog" (= analog synth) sound on the squares. A better roar maybe.

If you play a few DSD recordings and after that go to the PM recordings, it feels like a (not all that thick) blanket is removed, which makes you say "finally !".

All 'n all there sure is a difference.
With one exception ...

Here's an easy test. Go get a copy of any Reference Recording production (recorded with the PM ADC), and compare it to any recent (Grimm ADC) Channel Classics recording. You be the judge.

Same quote, because I tried to do that too;
I don't own any Channel Classics recordings so I had to do it with the free sample download. This is a quartet of Dutch violins, viola and cello (provided in 16/44.1, 24/96 and 24/176.4 + DSD). I tried to find a RR (violin chamber music) representative and came up with The Hotclub of San Francisco which is 24/96. And well, there you go again. "Rubbish" in my view. Flat, cold but well detailed and the typicle exhibit of hires under 24/352.8 and especially 24/96. Just doesn't work. But, violins were a pleasure to listen to.
Not so the Channel Classics sample. Violins were harsh and with too many (thus false) overtones. 16/44.1 vs 24/176.4 made no difference.

I actually started out with this, and because it didn't work out much I went the general DSD vs PM/RR route (all 16/44.1) which also allowed me to find better comparable music. The strange thing is that the Channel Classics showed the opposite of "too dull" or blanketed. It was the other way around, but not in good fashion. Just not right somehow.
I must add that this is a close mic recording (as far as I can tell) and it even might be so that the mics were on the instruments, where the hall of the church was captured by another set of microphones. Not the best IMHO, if setup like this at all.

Anyway Tailspn, it has been an eye opener for me to compare albums this way, and let's say Thank you for the question. I learned a few things.
Peter

 

[PeterSt]

Hi Orb,

So if he with a small team can make one of the best DSD pro ADC (not just Bruce who recommends this unit and also the clock is well reviewed by pro mags), then it is worth taking note what he says IMO.

As I said, it is all a bit hard to express in English what I wanted to say. But let me add this for better balance :

I don't think Bruno really worked on SACD development (but don't take my word for it), BUT, he ever back created a first DSD DAC which was based on SACD rips. Well, something like that. This couldn't go without the knowledge (of SACD) of course but anyway he did that for himself, and this was something like over 15 years ago. I think he couldn't make that a commercial product because it was not allowed.
I am sorry if this story is all over wrong, but when I saw this it was a deep hat off.

More clear is the nCore amplifier. The best measuring amp ever. Also from his hand and ... with DSD knowledge of course (Class D).

So it is the other way around, and indeed guys like him (and Eelco Grimm, Guido Tent, Peter van Willenswaard) all in this same "Grimm" realm, are to be highly respected and we can learn a lot from them, when we listen.
Still, and I hinted to that earlier, we must be careful because when time proceeds, audio progresses behind anyone's back. For example, you referred to those clocks, but while I used exactly those in the NOS1 at first, since more than a year that is history. Better became available elsewehere and the (SQ) difference is huge. This is only one example of which I think it is allowed to say en public, but there is more.
Everybody has to stay awake and proceed with new technology available, and I can tell you, the subject is HUGE. Especially because it also includes software these days.

Peter

 

[Orb]

Well the Class-D IC/UcD is more clear as he was the lead engineer on the project at the Philips Applied Technologies lab, but he has involvment with regards to audio digital aspects also researched there, but agreed his involvement may be small or just following said research (internally to Philips research labs) on DSD transcoding.
Like how Dr. Kees A. Schouhamer Immink is the one remembered at Philips (appreciate also involved Sony) for the definition and creation of Compact Disc even though the team was pretty large from the Philips research.
Cheers
Orb

 

[Orb]

Just to add,
what also complicates this is how different DAW systems handle the transcoding between DSD and PCM; this was more diverse in the past as Pyramix used the approach outlined by Philips while some of the others argued against this solution.
Also important is how DAW can process internally at a minimum of 32-bits and in theory this is the ideal-minimum criteria, so any tests done using 24-bits transcoding and also possibly which DAW/plug-ins would provide a compromised result.

32-bits is also realistically the minimum for processing music internally without dither in a studio environment I would say (which also adds complexities when too much dither is used or type implemented).

Cheers
Orb

 

[PeterSt]

32-bits is also realistically the minimum for processing music internally without dither in a studio environment I would say (which also adds complexities when too much dither is used or type implemented).

Well, think that any single processing step (like normalizing is one, equalizing is one, a mix of channels is one (I don't dare mentioning compression )) takes out one bit. When 24 bits is to be the output, that consumes fast.

But then I'd say that any self-respecting DAW software would use way more than 32 bits ?

 

[Orb]

Well, think that any single processing step (like normalizing is one, equalizing is one, a mix of channels is one (I don't dare mentioning compression )) takes out one bit. When 24 bits is to be the output, that consumes fast.

But then I'd say that any self-respecting DAW software would use way more than 32 bits ?

In a thread chatting to Bruce about dither and his experience with 32bit and higher in the studio, 32bit was enough although it was a brief discussion and completely focused on dither so scope pretty restricted.
Although as you say and what I was hinting the more dither applied the more it becomes a mess and then the hassle of what type used (which Bruce mentions a fair few studio engineers it seems do not pay enough attention to) - freaking nightmare
Same may be said about noise shaping and going from 1-bit DSD to 8-bit for editing.

Cheers
Orb