DSD mastering and quality

Octave Records
82

But is DSD still more transparent than DXD in an analog to DSD to DXD (mastering) to DSD (or directly keeping DXD) playback format to analog, as Octave now processes (and many others)? It’s hard to understand, that even 1xDSD is more transparent than DXD, but converting 4xDSD to DXD mastering format shouldn’t have the same losses.

I hope my conclusion out of this, that a back-conversion from the DXD master to DSD can rather be disadvantageous than advantageous (but probably hardly audible) isn’t too wrong :wink:

Edit: when I read this it seems to me, with DSD we might have a similar interest in format consistence as in the analog Mofi topic :wink: Could be interesting to know which DSD releases went through which analog, Sonoma or DXD steps and what this would mean for sound quality if compared. Best would probably be something like direct2disc for LP’s: a direct to DSD recording without editing (if that’s possible) :wink:

83

Could you explain this last logic gap I have, Ted?

Is it a fact that the higher transparency level captured by DSD vs. PCM you described continues to stay present even at a later conversion stage into DXD/PCM? And that as you say the same transparency isn’t given when the initial recording format is DXD from start without any conversion?

I hope you understand that this one way transparency needs a bit of explanation after first recognition :wink:

84

I don’t know where the quoted statement comes from and I’m afraid that I don’t understand the rest of your questions without more context. In particular I don’t have any context for “I hope you understand that this one way transparency needs a bit of explanation after first recognition.” I never said anything of the like.

You keep throwing the terms DSD, DXD and PCM around without qualifications. I’ve been trying to explain how the sample rate matters a lot and changes very nature of the conversation.

Given that my earlier statements which I thought were reasonably clear aren’t helping with your questions which at the time I thought were also reasonably clear, I’m not sure how to continue with questions that have even more supposition and/or context missing.

Even so, perhaps these will help clarify:

Converting from 44.1/16 to 24/96 doing a lot of work and then converting the final product to 44.1/16 shouldn’t be much of a problem with transparency, etc. but converting from 44.1/16 to 24/96, doing something then converting back to 44.1/16, then converting to 24/96 doing something else and then converting back to 44.1/16, then… is a different thing.

Converting single rate DSD to some other format, doing a lot of work and then converting back should be relatively transparent (but not as clean as not doing it.) But repeated conversions to single rate DSD and back from any format isn’t a good idea. The repeated conversions each add a little more noise near the audio band and over time that noise adds up to something audible, just at repeated dithering back to 16 bits isn’t great.

This isn’t really the case with double or quad rate DSD, repeated conversion to and from wide DSD or PCM which is at at least the same sample rates (NOT 352.8 or 705.6k) at those same sample rates isn’t really a problem.

Editing single rate DSD is is similar to the difference of editing 16 bit material in higher resolution, say 24 or 32 bits, then dithering back to 16 bits one time vs many times, BUT editing 24 bit material in 24 or 32 bits and converting back to 24 bits multiple times is reasonably transparent.

Some people I respect have suggested that 352.8k at 32 bits PCM (DXD) is sufficient to capture everything in DSD, but in the past most of the time the conversions used PCM reconstruction filters that weren’t up to the task. I don’t have any direct experience to prove them wrong, but I strongly suspect that even if using much better filters improves matters for conversions to and from DXD that doing the same work with double rate DSD to and from PCM at 5.6448MHz (the sample rate of double rate DSD) with good filters would be superior.

9 Likes
85

Thanks again Ted! I’m sorry that I’m so hard to understand. Several words you understood as something you had said, were just my conclusions out of your explanations, not your words. I think I have to give up for the moment and later make a better final post, just using cites of you and what I understood from them, as well as putting one or two hopefully simple questions.

One question regarding your last post:

Why then is DXD (which is 352.8, which you mentioned with „NOT“) used as editing/mastering format for DSD if the math suggests it’s not an optimal format to convert to or from it in connection with DSD?

86

It’s arguable whether the math says anything about it. Many people think the math says you don’t need anything higher than 48k, others claim that it says 96k is sufficient… What’s really going on is 1) what do people assume is audible? and 2) what can we do and is that sufficient?

DXD was “invented” at a time when 24/192 was available and commonly considered near state of the art. Merging Tech had good editing systems and wanted to also edit DSD. DXD was their compromise that allow them to use their hardware/software to also edit DSD. I don’t think anyone there would have chosen that sample rate if they could have supported a higher one.

They now support higher sample rates and have much nicer hardware.

5 Likes
87

I am not sure there is another technical niche that sees the perceived SOTA goal posts move as frequently as do several of the HI-Fi kit sub genres…

88

Good point, and an interesting metaphor, as the perception tends to be that the Latest 2x-Whatever means that the Goal Posts are now another 100 yards away. But the movement for all practical purposes is on the other side of the decimal point.

1 Like
89

For the moment I realized that at least for the first two of my 3 initial questions (the 3 abstracts above) , the answers got obvious by several explanations, although partly not directly answered.

My conclusion regarding the first one is:
no it makes no sense to use a reconverted DSD format if the DXD mastering forma of a DSD recordingt is available. Even if differences wouldn’t be audible, the reconverted DSD format is at least theoretically more lossy or less clean.

My conclusion regarding the second one is:
Yes, DXD can at least theoretically keep the , DSD information due to its high resolution and sampling rate,

So later I try to put the third and most important question in an understandable way. The quick recap in different words would be:

Why is DXD able to inherit the more or less full DSD information after a conversion, but is not able to record the analog signal as completely as DSD does? It’s still a contradiction for me that DXD should be more or less transparent in keeping the DSD informatio after a conversionn, but unable to record on DSD level. It’s possibly because I don’t get the difference between a recording and conversion process yet.

As I said I’ll work on making this last question understandable.

Edit: a different way to express would be: how different would a DXD recording sound compared to a 4xDSD recording converted to DXD?

90

I didn’t mean to imply that DXD can keep all DSD information - I said the opposite. DXD isn’t bad, it’s pretty good, but it doesn’t have the sample rate to preserve all info from DSD. If differences with high sample rates didn’t matter double rate DSD wouldn’t be different than single rate DSD or quad DSD.

There are good analog to digital converters for DXD output out there and there are good analog to digital converters for DSD. At their hearts (with a very few exceptions) those converters neither record DSD or DXD, instead they are usually a small number of bits per sample at a high sample rate and then get converted to whatever. DSD has a higher sample rate and arguably it is simpler to simply record one bit audio. In practice these days things are a lot more complicated and many different engineering tradeoffs have to be made.

Again, I say that in my experience recording single rate DSD sounds better than DXD, but it’s a small difference. Double and quad rate DSD are better than both in my opinion. Which of the three would sound better after doing a conversion depends a lot on the conversion process and on the DAC you listen to the result on. In a simple world, higher sample rates beat wider samples.

6 Likes
91

I just wonder how this all matters. The music industry has no standard whatsoever and information to this level about mastering is never shared with the consumer.

Therefore as a consumer you will never tell the difference.

92

I respectfully disagree with that statement.

3 Likes
93

I am referring to the SUPER audio CDs that claim to provide DSD where as a matter of fact it are upsampled or converted PCM files. FOOLING customers paying good money.

Or that high res media is not even available in most music stores.

Or that Gus’s Skinnas DSD remastering of Nils Lofgrens Acoustic Live would not be sold by the online shop as copy rights prohibited them taking my money (not pity money) and providing me the DSD files as an alien, non US citizen.

Or that Cookie Marenco claims on her WEB site that the file format does not matter. I acknowledged above, the recording does.

The only standards that the industry has, do date from the previous millennium and are still sold globally today: CD’s and vinyl. How that music was processed to end up on that media is typically not mentioned, because there is no standard that requires this information to be made transparent and comprehensive.

I call all of this LACK OF STANDARD!

94

Thanks Ted!
I may have made a certain step today. I have to continue tomorrow, as I just have access to a mobile at the moment. More is clear but still questions :wink:

95

Does this thread relate to Mofi gate?

Mofi gate confirms: it doesn’t matter! Many people paid extra money for the all analog principle that was advertised by Mofi and never heard that in fact the process included 2 conversions, analog => DSD and DSD => analog.

I will happily buy Mofi records. Based on the knowledge gained at PS Audio, thanks @Paul, thanks @Ted, we can be certain that DSD converted master tapes are a very smart move to not be depending on the de-magnetizing aging and mechanical wear of analog tapes.

While Mofi is to blame for false advertising, I don’t care and never cared about the mastering / storage process, PCM or DSD as long as the recording was good and the mastering engineer knew how to avoid to screw it up.

I have a Gus Skinnas DSD re-mastered vinyl album of The Rolling Stones. That label did a very smart move and published that fact as a positive statement on the Album. Whether it is better than the Analog master only the engineer and people in the mixing room listening to the same monitors can tell. But the result sounded very good on my system at home. Better than many other vinyl records.

Another fact about sound quality of medium:
CD’s and vinyl editions of the same recording sound different because guess what:

Most record companies produce different digital master files for each medium addressing the most common sonic characteristics for each different medium.

I quite enjoy the differences as each medium performs better in particular areas which are also heavily influenced by my mood (psychology). Even the streaming version sounds very good in cases I decide to settle for speed and convenience of changing titles when listening.

96

Another thought on Mofi gate:

Is DSD really digital, or is it analog?

Why do we categorize class D amplifiers analog and DSD information digital?

The signal in both cases is based on energy density.

With a DSD signal you could theoretically drive an electronic switching device hook the secondary side of that switching device between a power supply and speakers and get music.

If I am not mistaken, this is basically what Lyngdorf’s “digital” amplifiers do. To me the digital part in the Lyngdorf is the room correction DSP in the PCM domain and conversion of PCM to DSD. The DSD signal by itself is not necessarily digital. As we can read in above thread DSD very hard / impossible to edit. It can be upsampled with higher clock rates and filtered in the analog domain to keep any noise out of the audible signal but that is basically it.

97

Rudolph is correct. There are no industry standards but there are personal standards. Know the company you are dealing with so you can trust the results.

Working with DSD is always going to be a challenge with compromises, but so too is analog. In fact, the only analog process I know of without compromise is a direct-to-disc (or tape) live recording. Everything else is to some degree compromised.

DSD can too be direct (but rarely is). One could record live in single rate DSD and then transfer that recorded file to an SACD without any conversions. That’s rare (as are all single take live recordings). I would estimate 99% of all DSD recordings have conversions back and forth in the process of becoming finished. (And this does not include fake DSD where the original recording was PCM).

Take the process we at Octave Records (in our new studio) use. We record everything in 4XDSD (11.2mHz). Those files are then low pass filtered to 352.8PCM (DXD), mixed and mastered, then converted back to 4XDSD through the Pyramix DSD rendering process.

Compare this with how we started out working when we used the Sonoma system.

In the Sonoma 1XDSD system we recorded everything in 1XDSD, then ran each track though a DAC to produce analog. Mixed in on an old analog mix board (the Studer), then A/D converted back to 1XDSD. From there we low pass filtered the DSD master to get the PCM versions. The results were excellent. Better than most any analog recordings that used tape.

That entire process was a real milestone of quality that we at Octave Records are proud of. It set a new standard that blows away most analog and PCM recordings.

But now, it’s time to take the next step upwards on the journey.

Octave’s new Pyramix DSD system has raised the bar yet again.

Comparing the sound quality output of both methods—the original 1XDSD/analog/1XDSD to the new 4XDSD/DXD/4XDSD chain—the new all digital Pyramix recordings are “light years” better. I can’t wait for people to make their own judgements.

It is true there are no industry standards. But, there are certainly standards by which people like us, and Gus, (and Cookie at Blue Coast), adhere to because we hear the difference in poor choices of technique and abhor compromise for the sake of saying something is this way or that. Each of our techniques may be different, but at the end of the day it’s the results that matter.

Openness of process is critical. We want you to know how the “sausages are made” because when you hear how extraordinary the results are you’ll want to seek out more of it.

4 Likes
98

i sometimes think of it as analogue recording using digital techniques.

99

Thank you, Paul

2 Likes
100

The problem with all this is that the consumer hasn’t got a clue what’s going on and so won’t buy into it. I think @jazznut really just highlights the confusion, and he’s far more knowledgable than most.

We had some friends round for dinner on Saturday night and one happens to collect vinyl and was explaining the MoFi issue to me and someone else who is a music lover who knows nothing about hifi or vinyl. The very simple point was that if you sell something on the premise that it is analog start to finish and it turns out to have gone through a digital conversion, the customer feels cheated and all trust and goodwill evaporates. Doesn’t matter how good the record sounds, it’s the principle that counts.

We were then listening to Black Acid Soul by Lady Blackbird, which is a very good recording that has been issued on vinyl and, according to Amazon reviews, the pressing has been completely messed up. Because it was only $25, my friend would buy it, but I would not.

Companies that issue vinyl remasters and reissues are increasingly transparent and unambiguous about how their products have been produced. If not, people generally won’t buy them.

The problem with DXD and DSD is that everyone knows you can’t easily edit DSD and DXD is actually PCM anyway. So even if it says DSD on the label, there is little or no trust. There is a presumption that is has been converted to PCM at some point. Once that happens, the consumer’s idea of pure DSD disappears.

Whilst a handful of professionals may get to make comparisons in a studio, you get to the position where the consumer will never be able to make a valid comparison, even with the time or inclination to do so. It’s not like the 1980s when the difference between analogue and CD was clear on a modest system, whether or not you liked it.

Anyway, it’s water under the bridge. Us simple people understand it’s either DSD or it isn’t. If it isn’t, then why should I care?

Finally, you need customers to have audio equipment that will play the high rate DSD files in their native format. I doubt many do, I don’t and my hifi system is not exactly budget. It can play a DSD64 file, I think Innuos converts it to DoP (repackaging 1-bit as 16-bit, but the same data) and then the A/D-D/A converts it to 40/384 or 40/352. I’m not sure, I but it sounds fine.

I would certainly agree that if you can use technology (I don’t care what) to make better PCM music files, then that is a good thing.

For a digital file, this consumer does not care how the sausage is made, because I have no idea about the relative merits of the ingredients. If you don’t know the difference between meat and fat, how can you form an opinion on what makes a good sausage? All these options in digital processing, with analogue mixed in, mean nothing to me at all.

All I care about is how it sounds and really I don’t care how it got there.

1 Like
101

Just to isolate this one thing….does this mean, there is no „DSD“ recording, but DSD recorders use a proprietary format (which? PCM? other DSD kind?) with more than one bit but still a very high sampling rate and then convert this to e.g. DSD or DXD or whatever PCM?