• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

KeithPhantom

Addicted to Fun and Learning
Forum Donor
Joined
May 8, 2020
Messages
641
Likes
658
Isn't it possible to record something that does not need any kind of mixing or mastering, record both in PCM and DSD, and then with those files do a double-blind?
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
Isn't it possible to record something that does not need any kind of mixing or mastering, record both in PCM and DSD, and then with those files do a double-blind?
Of course it is.

You will probably have to change level to match each. And since you can't do much with DSD you'll be changing level of the PCM.
 

KeithPhantom

Addicted to Fun and Learning
Forum Donor
Joined
May 8, 2020
Messages
641
Likes
658
The point is to reduce the conversion between encoding methods and purely compare both formats without the excuse of using the PCM file to create the DSD equivalent or the other way around. Also, I want to not have the argument of "you have to convert DSD into PCM if you want to do any kind of editing".

About level-matching PCM, will this affect the blind test?
 

PaulD

Senior Member
Joined
Jul 22, 2018
Messages
453
Likes
1,341
Location
Other
The point is to reduce the conversion between encoding methods and purely compare both formats without the excuse of using the PCM file to create the DSD equivalent or the other way around. Also, I want to not have the argument of "you have to convert DSD into PCM if you want to do any kind of editing".

About level-matching PCM, will this affect the blind test?
I reported on exactly that in post 2...

Here is the excerpt from post 2 of this thread.
"I did do a blind test in a studio with the Korg MR-1000 DSD recorder when it came out, signal split after the mic preamps and sent to the Korg recorder and also to Avid Protools PCM converters at 24bit 96KHz. Level matched on playback, there was no audible difference between the DSD and PCM version."

The level matching was done in the analogue path after conversion back to analogue.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
The point is to reduce the conversion between encoding methods and purely compare both formats without the excuse of using the PCM file to create the DSD equivalent or the other way around. Also, I want to not have the argument of "you have to convert DSD into PCM if you want to do any kind of editing".

About level-matching PCM, will this affect the blind test?
With most volume or level changes being done in software the answer should be no. Or if done by analog gear correctly the answer is no.
 

KeithPhantom

Addicted to Fun and Learning
Forum Donor
Joined
May 8, 2020
Messages
641
Likes
658
With most volume or level changes being done in software the answer should be no. Or if done by analog gear correctly the answer is no.
We want to minimize all possible changes to the actual signal, so I would say software level-matching.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
Been said before. If there was a substantial difference in the sound of DSD and PCM it would have been confirmed a long time ago. If there are any differences they'll be small, fleeting, audible to only some people, on some particular material. In other words if there are real differences almost anything will cause a bigger difference than the sound quality of DSD vs PCM. And those differences simply won't be of large consequence. They'll be inconsequential. Considering all the inconvenience of DSD anything it simply represents a poor decision.
 

krabapple

Major Contributor
Forum Donor
Joined
Apr 15, 2016
Messages
3,169
Likes
3,717
Fixed it. The good papers are all at IEEE or similar places. AES exists only to promote the industry agenda under a thin guise of being scientific.

I don't agree, unless that's a recent development. Case in point JAES published Robert Stuart's cheerleading for 'hi rez', and the *also* published rebuttals to it. Not to mention papers by Toole and Olive that we all reference here.
 

krabapple

Major Contributor
Forum Donor
Joined
Apr 15, 2016
Messages
3,169
Likes
3,717
The point is to reduce the conversion between encoding methods and purely compare both formats without the excuse of using the PCM file to create the DSD equivalent or the other way around. Also, I want to not have the argument of "you have to convert DSD into PCM if you want to do any kind of editing".

And I want peace on earth and good will among men (and women).

About level-matching PCM, will this affect the blind test?


Basic psychoacoustics says: yes, level matching is critical (depending on the particulars of the level difference).
 

krabapple

Major Contributor
Forum Donor
Joined
Apr 15, 2016
Messages
3,169
Likes
3,717
Been said before. If there was a substantial difference in the sound of DSD and PCM it would have been confirmed a long time ago. If there are any differences they'll be small, fleeting, audible to only some people, on some particular material. In other words if there are real differences almost anything will cause a bigger difference than the sound quality of DSD vs PCM. And those differences simply won't be of large consequence. They'll be inconsequential. Considering all the inconvenience of DSD anything it simply represents a poor decision.


And I can't fail to reiterate: the original intent of DSD was for it to be an archival medium that would always be seamlessly converted to PCM (at multiples of 44.1 kHz) for consumer products. DSD is a ridiculously unwieldy and unnecessary format for consumer use. But SACD proved to be (until relatively recently) a very secure copy-protected medium. Priorities, right?
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
Isn't it possible to record something that does not need any kind of mixing or mastering, record both in PCM and DSD, and then with those files do a double-blind?
That's needlessly complicated, and it introduces several unwanted variables. There's a much better, and simpler, way to do it. If the contention is that DSD is superior and PCM detracts something, start with a DSD recording of your choosing. Convert it to 24-bit 48 kHz PCM and back to DSD using good software. Compare the resulting DSD file to the original using the same playback chain. If PCM is as horrid as some say, the difference should be readily audible.
 

Pluto

Addicted to Fun and Learning
Forum Donor
Joined
Sep 2, 2018
Messages
990
Likes
1,631
Location
Harrow, UK
SACD proved to be (until relatively recently) a very secure copy-protected medium
And remains so, to this day. Yes, I know there are workarounds but the copying process is far too tedious for the average Joe. For all intents and purposes, SACD remains tightly locked.
DSD is a ridiculously unwieldy and unnecessary format for consumer use
Not only for consumer use. Most of the DSD production hardware rushed to market in the first 10 years of the millennium is no longer made or even supported. There are one or two highly specialist solutions available which keep the workflow within the DSD space but you would have to be very confident of your DSD sales to make such an investment a sound proposition.

As a matter of interest, I wonder if anyone could determine, with reliability, the provenance of four anonymous files: i) pure DSD ii) pure PCM iii) DSD converted to PCM and iv) PCM converted to DSD.

I wonder if anyone could reliably distinguish i from iv and ii from iii
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
I don't agree, unless that's a recent development. Case in point JAES published Robert Stuart's cheerleading for 'hi rez', and the *also* published rebuttals to it. Not to mention papers by Toole and Olive that we all reference here.
Sure, a few good papers have made their way there. Still, the bulk of AES contributions come from people in the hi-fi industry whose livelihoods depend on the myth being perpetuated that some nebulous differences exist outside the ordinarily audible range. Even a solid result in favour of one inaudible thing over another could ruin business for the half of the industry touting the losing option. It is in the collective interest of the industry as a whole to keep cranking out poorly executed studies that hint at one thing or another without ever reaching a firm conclusion.
 

KeithPhantom

Addicted to Fun and Learning
Forum Donor
Joined
May 8, 2020
Messages
641
Likes
658
unwanted variables
I don't get it, which variables?. I think splitting the output to an ADC that records both in PCM and DSD will eliminate the conversion between formats. Why do we have to start with one format and then convert? I want to see the maximum theoretical potential both methods have, and I don't see a better way to do this.
 

KeithPhantom

Addicted to Fun and Learning
Forum Donor
Joined
May 8, 2020
Messages
641
Likes
658
Basic psychoacoustics says: yes, level matching is critical (depending on the particulars of the level difference).
I meant level-matching the PCM file to have its full scale at the same value as the DSD file.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
I don't get it, which variables?. I think splitting the output to an ADC that records both in PCM and DSD will eliminate the conversion between formats. Why do we have to start with one format and then convert? I want to see the maximum theoretical potential both methods have, and I don't see a better way to do this.

You don't have to, but it would be a good method.

It would be similar to us arguing over one needs bandwidth to 200 khz or 20 mhz to be audibly transparent. If you have a recording with 20 mhz and we put it thru a filter of 200 khz, then do these sound audibly the same or not. If the answer is yes, we don't need 20 mhz. If the answer is no, then we may need 20 mhz or something in between. Or maybe we can see some residual difference that matters.
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
I don't get it, which variables?
Different ADCs and DACs with associated analogue circuitry, for starters. Then there's whatever signal path additions are needed to match levels. On top of that, there's the possibility of downstream side effects of the residual ultrasonic noise after the DAC, which will differ. If you hear a difference, you have no way of knowing what actually caused it.

I think splitting the output to an ADC that records both in PCM and DSD will eliminate the conversion between formats.
ADC chips, at least the good ones, use multi-level sigma-delta and convert to PCM or 1-bit DSD digitally. You can't escape that conversion.

If there is something audible that only one of the formats can capture, converting to the other and back has to reveal it. It's by far the simplest way to go about it, and if everything else is kept exactly the same, you can be sure any audible difference really does stem from the conversion. Of course, it could still be the result of a poorly executed conversion rather than an actual deficiency in the alternate format.
 

PaulD

Senior Member
Joined
Jul 22, 2018
Messages
453
Likes
1,341
Location
Other
Different ADCs and DACs with associated analogue circuitry, for starters. Then there's whatever signal path additions are needed to match levels. On top of that, there's the possibility of downstream side effects of the residual ultrasonic noise after the DAC, which will differ. If you hear a difference, you have no way of knowing what actually caused it.


ADC chips, at least the good ones, use multi-level sigma-delta and convert to PCM or 1-bit DSD digitally. You can't escape that conversion.

If there is something audible that only one of the formats can capture, converting to the other and back has to reveal it. It's by far the simplest way to go about it, and if everything else is kept exactly the same, you can be sure any audible difference really does stem from the conversion. Of course, it could still be the result of a poorly executed conversion rather than an actual deficiency in the alternate format.
Indeed mansr, internally most DAC chips like the ESS ADC and DAC chips, and from other manufacturers, are a combination of PCM and DSD - the 9018 is a 6-bit PCM engine oversampled in a DSD way, with a PCM output. Here is an explanation:
https://benchmarkmedia.com/blogs/application_notes/inside-the-dac2-part-2-digital-processing

So the whole argument about DSD vs. PCM is moot, the chips internally use the best of both techniques to get the best results they can, and both formats are used to their best advantage, inside the chip.

There is still no reason for DSD outside of the chip - someone much earlier in the thread said it was a very "nasty format" outside of the chip, and that is true.
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
4,685
Likes
10,700
Location
Hampshire
Indeed mansr, internally most DAC chips like the ESS ADC and DAC chips, and from other manufacturers, are a combination of PCM and DSD - the 9018 is a 6-bit PCM engine oversampled in a DSD way, with a PCM output. Here is an explanation:
https://benchmarkmedia.com/blogs/application_notes/inside-the-dac2-part-2-digital-processing

So the whole argument about DSD vs. PCM is moot, the chips internally use the best of both techniques to get the best results they can, and both formats are used to their best advantage, inside the chip.

There is still no reason for DSD outside of the chip - someone much earlier in the thread said it was a very "nasty format" outside of the chip, and that is true.
Some DAC chips have a "pure" DSD path. In fact, I think most except ESS do if they accept DSD input at all. Whether the implementation makes it available is a different story. PCM is invariably processed with multi-level sigma-delta. That's all a bit tangential, though. When DSD is claimed to be superior as distribution format, it is implicit that the goodness can be reproduced by some DAC or other, and a listening test would presumably be performed using one of these.
 
Top Bottom