Oops! They Did It Again!

[sphinxsix]

@Ralf11 If you can hear the difference between redbook (16-44) and 24-96 or 24-192 and it matters to you, you will very likely hear the redbook - SACD/DSD difference (and it will matter to you too).

I wrote about it before on CA - in my system I find in particular the DSD top end very 'attractive' (sweet, smooth, which is important to me). I heard some time ago that 'Stereophile' has proved that DSD resolution above 10kHz is comparable with 16-768, I didn't manage to find their article on this topic though (even if it's so I'm lucky enough to be a subjectivist so I may not care ).

@mansr Have you by any chance explored this matter (DSD high frequencies resolution).?

Beside that I believe that in many (maybe even most!) cases the SACD remasters are better done (eg label discussed here but not only) than 24 bit ones. 

But you know it's my system and my taste, why don't you check it out for yourself?

[Ralf11]

actually, I've never heard 24-96 or 24-192

 

I do have ~~20 SACDs and they sounded fine (when I had an SACD player, which I don't at the moment).  I think I'll just ignore SACD releases for now, while waiting for Oppo to get their new player out.

[mansr]

10 hours ago, sphinxsix said:

@Ralf11 If you can hear the difference between redbook (16-44) and 24-96 or 24-192 and it matters to you, you will very likely hear the redbook - SACD/DSD difference (and it will matter to you too).

I wrote about it before on CA - in my system I find in particular the DSD top end very 'attractive' (sweet, smooth, which is important to me). I heard some time ago that 'Stereophile' has proved that DSD resolution above 10kHz is comparable with 16-768, I didn't manage to find their article on this topic though (even if it's so I'm lucky enough to be a subjectivist so I may not care ).

@mansr Have you by any chance explored this matter (DSD high frequencies resolution).?

Beside that I believe that in many (maybe even most!) cases the SACD remasters are better done (eg label discussed here but not only) than 24 bit ones. 

But you know it's my system and my taste, why don't you check it out for yourself?

The capabilities of DSD depend very much on the sigma-delta modulator used. A perfect modulator would provide a low flat noise floor up to some chosen frequency F, and beyond this a rising noise level, eventually reaching its highest level and staying there for the remainder of the spectrum (up to half the sample rate). For the flat region (up to F) this would be comparable to 24-bit (or better) PCM with a sample rate of 2F. As the modulator noise level rises, the available dynamic range is gradually reduced until at some frequency it becomes useless (the exact value depends on the input signal level and on what noise level is acceptable in the specific application). PCM with twice the sample rate of this frequency can represent the original signal better than DSD since the noise level is constant and low up to the Nyquist cut-off.

 

Looking at real systems, we find that for DSD64 the noise level typically starts rising sharply around 25 kHz, sometimes as low as 20 kHz. It usually drowns out what remains of the music before 50 kHz (and that's being generous), and by 80 kHz the noise level is high enough to cause problems in analogue circuitry. The usable frequency range thus somewhat exceeds CD but falls short of 24/96 PCM. Below 20 kHz the noise level should be flat, but this tends not to be the case in practice, although the peak noise levels are still lower than 16-bit PCM. Many modulators also exhibit some amount of noise modulation whereby the noise floor rises with increased signal level. Finally, there is usually a small amount of harmonic distortion. These imperfections are reduced or eliminated with multi-bit sigma-delta designs used in modern equipment.

 

As for the claim that "DSD resolution above 10kHz is comparable with 16-768," I'm not sure what to make of it. PCM resolution below the Nyquist frequency depends on the bit depth, not the sample rate, and well implemented DSD certainly exceeds 16-bit PCM resolution at low frequencies. PCM with a high sample rate of course allows for noise shaping, so it's possible to lower the noise level at low frequencies. Perhaps the comparison to 16/768 had such a scheme in mind. Of course, 24-bit PCM at any sample rate would be better still.

[sphinxsix]

40 minutes ago, mansr said:

As for the claim that "DSD resolution above 10kHz is comparable with 16-768,"

The claim that Stereophile has proved it may be false. It came from an untrustworthy source. I searched for the article on that quite hard and didn't find it. To me DSD high frequencies sound very (!) differnent from 16 bit (at least from 16-44)! Thanks for your explanation!

[mansr]

6 minutes ago, sphinxsix said:

The claim that Stereophile has proved it may be false. It came from an untrustworthy source. I searched for the article on that quite hard and didn't find it. To me DSD high frequencies sound very (!) differnent from 16 bit (at least from 16-44)! Thanks for your explanation!

Forget about proof. It's not even clear what it's supposed to mean.

[sphinxsix]

14 hours ago, Ralf11 said:

Are you aware of any other tests?

Accidentally just read this. I think you can call it 'other test'. Quote:

 

'Interestingly, back in 1999, JA concluded his DSD discussion with these words:

 

The proof of any audio pudding is in the hearing, and in that respect DSD-encoding would seem to be beyond reproach. Every Stereophile writer who has auditioned DSD under critical conditions has found it both very much better than 16/44.1k CD and much closer to the analog experience.'

The whole article:

http://www.stereophile.com/content/news-about-dsd

 

Edit : Doing the (probbably not existing) Stereophile article search again, just saw this too:

http://www.audiostream.com/content/dsd-v-pcm-file-comparison-16441-2496-24192-64x-dsd-128x-dsd

[Ralf11]

A meta-analysis published in JAES in 2016 found that listeners, if trained, could distinguish HiRes from Redbook CD... 52% correct, which is not much.

 

 

[sphinxsix]

3 hours ago, Ralf11 said:

A meta-analysis published in JAES in 2016 found that listeners, if trained, could distinguish HiRes from Redbook CD... 52% correct, which is not much.

 

 

52%, hmmm.. too many correct answers IMO. I'd rather be in more elitary club

[mmerrill99]

4 hours ago, Ralf11 said:

A meta-analysis published in JAES in 2016 found that listeners, if trained, could distinguish HiRes from Redbook CD... 52% correct, which is not much.

 

 

Yes training is NECESSARY to successfully participate in blind testing - it doesn't mean that in normal listening differences are not perceived.

Surprising that someone who claims to have a background in such perceptual testing doesn't know this & has to emphasise it  in italics

It's also very surprising that you don't understand the statistical significance of 52% claiming it's "not much"

[Ralf11]

mmmerl, I don't know what you do when not mucking about on the internet, but almost every clause and sentence in your post is incorrect.

 

I'd like to see some definitive tests done, but right now it appears that SACD per se is not a huge SQ increment over Redbook CD, and that large differences are due to mastering or recording quality.

 

Unfortunately, there also appears no way to ensure that attention to SQ is done, except by getting impressions from those who have compared various releases before buying.

[mmerrill99]

5 hours ago, Ralf11 said:

mmmerl, I don't know what you do when not mucking about on the internet, but almost every clause and sentence in your post is incorrect.

Well, I've a pretty good idea that your handwaving denial without any specific details in you reply is proof of just what I said

 

13 hours ago, mmerrill99 said:

Surprising that someone who claims to have a background in such perceptual testing doesn't know this & has to emphasise it  in italics

It's also very surprising that you don't understand the statistical significance of 52% claiming it's "not much"

 

[Don Hills]

16 hours ago, mmerrill99 said:

...

It's also very surprising that you don't understand the statistical significance of 52% claiming it's "not much"

 

He's right though, it's not much. (Statistical significance,)

[mmerrill99]

6 hours ago, Don Hills said:

 

He's right though, it's not much. (Statistical significance,)

No, he's not right & his post shows that he doesn't work in the area despite his many claims that he lectures in this field.

 

How many trials were in this Meta analysis? 12,645 is it? What is the number of correct answers needed for a statistical significant result? 

 

To just cross the 0.05 significance threshold for cumulative p, one needs 6419/12645 correct (50.8%), for a p = 0.043

 

Furthermore, the results were categorised into 2 groups - trained & untrained. The untrained groups got an aggregate mean of 51% correct; the trained group 62% right on average ( a range of means of 56.9-74.7% ).

 

So ralf11's statement "

On 20/05/2017 at 7:07 AM, Ralf11 said:

that listeners, if trained, could distinguish HiRes from Redbook CD... 52% correct, which is not much.

is both factually incorrect, trained listeners achieved 62% correct (or as high as 74% in some studies) & shows he doesn't comprehend statistics 101

 

Reiss, in the paper, also comments on the bias evident in most of the test studies towards Type II error, a bias toward not hearing a difference where there was one - so, in fact, with this bias towards a null in many such tests, the meta study signifies more ability to discriminate than the results suggest.

[Ralf11]

What field do you think I claimed I lectured in?

 

Also, mmril - you should read the actual article before commenting.  You will not look like such a fool that way.

 

Let me know if you are unable to use Google to find it.

[mmerrill99]

47 minutes ago, Ralf11 said:

What field do you think I claimed I lectured in?

 

Also, mmril - you should read the actual article before commenting.  You will not look like such a fool that way.

 

Let me know if you are unable to use Google to find it.

You teach biology, AFAIR & you claim some knowledge of psychoacoustics (surely you must know stats 101)

Here's your chance to make a fool of me rather than pretend you know more than you do - go on then argue specific errors in what I posted & let's see you back up your generalised barb!!

[Ralf11]

No,

training is not NECESSARY to successfully participate in blind testing - whatever that means

 

it doesn't mean that in normal listening differences are not perceived. -- What?

 

mmerril  doesn't understand the statistical significance of the 52% finding in the article

 

but no, it's "not much"

 

For the record, I am retired from being a Biology Professor and no longer teach 

 

Now, I don't know what you do, but I hope it is not anything important.

[mmerrill99]

On 21/05/2017 at 8:19 PM, Ralf11 said:

No,

training is not NECESSARY to successfully participate in blind testing - whatever that means

 

it doesn't mean that in normal listening differences are not perceived. -- What?

 

mmerril  doesn't understand the statistical significance of the 52% finding in the article

 

but no, it's "not much"

 

For the record, I am retired from being a Biology Professor and no longer teach 

 

Now, I don't know what you do, but I hope it is not anything important.

You obviously don't know the first thing about perceptual testing but that was obvious from the start

Instead of handwaving & dealing in generalities, try to address just one specific point I made - that might show what you know. At the moment you are demonstrating a dearth of knowledge in the area - generalized claims with nothing to back them up

[Ralf11]

2 hours ago, mmerrill99 said:

I, mmerriill99, am making the following generalized claims with nothing to back them up:

 

You obviously don't know the first thing about perceptual testing but that was obvious from the start

Instead of handwaving & dealing in generalities, try to address just one specific point I made - that might show what you know. At the moment you are demonstrating a dearth of knowledge in the area

 

 

...and personal attacks too.  Is today troll day or something?

 

I also suggest that you write to the editors of JAES regarding your criticisms of the methodology used by these experts*.  I feel certain that the editors would just love to hear you from you.

 

 

* their bios are on the site