Right. The only problem are "sampled points" and there 16-bit distance . All the rest is fine .

If they think that louder is better, why would they make CD louder if they want SACD to sound better ??



Filtering at 28 kHz was audible for this person : http://www.hydrogenaudio.org/forums/index....ndpost&p=198806



What about this one ?

Yoshikawa, Shokichiro; Noge, Satoru; Yamamoto, Takeo; Saito, Keishi
Does High Sampling Frequency Improve Perceptual Time-Axis Resolution of Digital Audio Signal?
Preprint 4562 in
http://www.aes.org/publications/preprints/search.cfm

Oops, sorry, it doesn't deal with musical program

Excellent question. Now, who has an answer?





I should have been more specific. 1st, I did ammend the 16kHz part of my statement in this thread already, because an unusual sample(with a strong >16kHz constant component at high amplitude, which I suspect is basicly as audible as a sine wave, in the confined circumstances) was easily ABXed, by basicly anyone, that was provided by a previous poster. Fear of this type of sample was probably one of the reasons that the designers of RBCD decided to go ahead and leave room for error, and have a bandwidth >20kHz. As for the person who ABXed a sample with 28kHz filtering, you realize this is a questionable result, especially since this person stated they could not even hear an 18kHz sine wave. I already participated in the thread linked, and we(you and I) already discussed the requirements for a test to have high credibility/reliability. I don't see a need for us to rehash that in this current thread.



1. I have not reviewed the paper, so I can't comment on the test specifically(do you have a copy you could e-mail me? It would be appreciated if you could. ).

2. As you already stated in the last reply, the test did not use musical program material.

-Chris

I don't understand the distinction you are making between "real sound" and "digital sound". When an analog audio signal is converted into PCM (CD and DVD) it is sampled (limiting it's bandwidth) and quantized (limiting it's signal to noise ratio). If we do the sampling process at 96kHz (with correct filtering) the sampled signal can be used to perfectly reconstruct the original signal up to 48kHz. The quantization step reduces the signal to noise ratio, effectively adding wide-band quantization noise to the signal. For 24bit, the quantization noise is at -144dB (this can be improved in the audible band by noise shaping).

Converting back from PCM to analog gives us our input signal, bandlimited and with a small amount of added wide band noise. The output of the DAC is "real sound" - it's not been tainted by it's trip to the world of PCM in any way other than the bandlimiting and noise (both of which should be inaudible for 24/96). Real-world DACs and ADCs do mangle the sound, but decent ones hardly mangle it at all. What do you base your distinction between "real sound" and "digital sound" on?

What makes you believe that DSD is any more "analog" than L-PCM? I suggest you read up on some of the research into the properties of DSD, starting with "Why 1-Bit Sigma-Delta Conversion is Unsuitable for High-Quality Applications" by Lipshitz and Vanderkooy (AES 110th convention). Please go back and read some of the previous discussions of this topic here on HydrogenAudio. The search function will guide you on your way.

my thought of real sound is sound pre-recording; sound that has yet to be imprinted on a medium. Digital sound is sound that has been put into a digital form; 1s and 0s that when processed acordingly gives us American Pie.

... If digital sound and the sound prior to being digitized is, as you say, a "[perfect] reconstruct [of] the original signal", then why is there a debate at all, and why are different formats still being tested and released? If there was a way to reconstruct the original signal 100%, why are we even discussing anything else?

Yes. I was rather answering 2bdecided. From my point of view, this result is not a proof in itself, but it justifies taking some time to test the same sample.



No. I'd like to read it too, but I don't know if I shall spend 20 $ for it.

I could try to do it if needed. I know that I'm sensitive to 16KHz lowpassing with some instruments (harpsichord, cymbals...), which appeared as dull, desaturated with ~16Khz lowpass.
I seriously don't think that 16KHz could be fully transparent, even with 'usual music'. At least not with headphones.

The same with me. While being a student, I tested my hearing in a lab and was able to hear up to 19kHz generated noise. Although, do not know what I can now.
Besides, ears is not the only part of our body we percept sounds with. If I can not hear frequencies above 20kHz it does not mean that I do not percept them and that they do not influence me.

There are several reasons new digital formats are being developed. First, with high rate PCM (DVD-A) there are some technical advantages - 24/96 is better than 16/44.1. Whether that difference is audible is still something that is under discussion. Next, it is easier to build DACs and filters which exhibit closer to ideal behaviour in the audio band when given high-rate data. This makes equipment cheaper and better. DSD (as used in SACD for example) is also very easy to build good DACs for. I believe the industry push towards digital path amplifiers (S-Master and friends) has been a major factor in some company's backing of the format.

Lastly, an most important in my mind, is that the new formats offer DRM to music lables. Whether you think this is a good idea or not, it is in the interests of music lables to limit your freedom as much as possible when it comes to your use of their recordings. In my opinion, the next generation of digital formats has been driven by two factors - 5% technical considerations and 95% business considerations.

oh of course. I disagree with the premise of their MO, but I understand it completely... and thats been the case since tinpan ally, only getting worse.

oh well.

I've added to my web page ( http://www.kikeg.arrakis.es/lowpass/ ) that other sample where a 16 KHz lowpass is audible. It's the Androgyny sample. The difference is not as obvious as in the Chenoa sample, but for me it's still pretty easy to abx (16/16 in 45 sec.)

I have a friend who has an AES membership(he can download PDFs for free). I'll ask him if he'll let me have a copy. If I can get him to do it, I'll forward you a copy if you give me an e-mail address.

-Chris

It says on the site it'll still cost him $5.



I have a copy of the document but it says that you need "direct permission from the Journal of the Audio Engineering Society" to distribute it.

Edit: I can provide a description of the test without infringing copyright.

The sample material used for the test was recorded on a 96 kHz 16 bit DAT tape. The reference sample (sample R) had a FIR lowpass filter at 40 kHz while the test sample (sample X) had a lowpass at 20 kHz. 11 adult males, 22-24 years of age, were asked to identify which sample was X and which was R. The result was statistically significant (pval < 0.05), and the folllowing was concluded.

I haven't read far enough ahead in the thread to see if this was addressed. The above statement is not correct. For more read the paper on the consequences of traditional peak meters here:

http://www.cadenzarecording.com/papers

It is very possible (in fact likely) that signals wherein the samples approach FS will clip the output of even a good D/A converter. The digital reconstruction filters on those D/As simply can't reconstruct above FS even if the analog can.

Cheers!
Nika

Erich,

Your question implies that the changes to formats have been made because they inherently provide an improvement. This is an optimistic view of the marketing and multinational corporations today. Sony has proved inside their building that 44.1kS/s is better than DSD and can sound as good for any recording/mixing/delivery situation as higher L-PCM sample rates such as 96kS/s and 192kS/s. There is a reason that you don't hear about this.

96kS/s was originally proposed back in the early 90s or late 80s as a solution to a problem the industry was having with filter design. That problem was fixed long ago, but the push for higher fs had started and it was viewed as potentially profitable as well, so despite the lack of need for it, the industry pushed on.

SACD (DSD) was similarly created as a patch for a problem that the industry had. But while the industry fixed that problem (that being integral and differential non-linearity in converters yielding high distortion) the DSD marketing machine had already gotten started. And therefore, despite the fact that modern converters designed for L-PCM use inherently have a leg up on DSD, Sony can't back down now and DSD continues as a major marketing push.

In each of these situations the industry had a choice: fix what was wrong or develop a new format altogether. Since there is a lot of profitable money tied with new formats we can't be surprised which way the labels and consumer gear manufacturers went, while the R&D guys at the converter manufacturers were busy fixing the problems...

Nika

But, did the test designers take any measures to avoid speaker distortion products falling into audible band, when feed with signals very rich in ultrasonic content, such as the pulse train they make mention of?

1. Was this test using linear phase filtering? That will throw off everything immediately, as non-linear phase filtering will create audible phase problems in the audible range, despite the presence of HF content.

2. Were the speakers linear above 20kHz? Non-linear speakers above 20kHz can create audible bandwidth distortion when presented with HF content. Again, however, this isn't an indication of audibility above 20kHz but rather audibility below.

3. Despite the persistent notion that we can hear with organs other than our ears, absolutely nothing has been presented on the market that substantiates that we can do this for HF content at musical program listening levels. We can "feel" HF content at several hundred dB above normal listening levels (you would certainly "feel" it if someone started sonar-welding your skin, for example) but even studies that have been done on this subject have indicated no "perceptibility" of HF content. Nobody asked the "listeners" in such tests to only indicate the HF content that they could "hear" and to ignore the content that they "felt in other ways." U

Nika

This type of stuff is really annoying. How in the world did they draw that conclusion from that test?

Nika

That is for the old standard membership model. There is an online subscription model that allows for unlimited downloads of papers, which was recently introduced within this year.



Then, consider me an infringer(of the trade of AES research papers).

As for the general description; that's not exactly what I need. I need to review a paper in full before I come to any conclusion(s). This is not a published paper, for one thing, so the quality of paper could be anything.

-Chris

There is a good explanation of the behaviour of DACs when in the presence of levels nearing 0dBfs and the relation to peak output distortion(s), in this paper:

OdBFS+ Levels in Digital Mastering
Sgren H. Nielsen and Thomas Lund
AES Preprint 5251

This paper(unlike most) is available for free download from the authors:
http://www.tcelectronic.com/media/nielsen_...00_0dbfs_le.pdf

-Chris

Good question. While this might have little effect for musical content, the difference can be very big with such signals. I remember, back in school, having played with a generator with a very cheap speaker plugged in (the kind of speaker for mono TV set or FM-Alarm-clock, if you see what I mean), and the 10 kHz sine, square and triangle had a completely different sound because of distortion.

If they did it wasn't mentioned in the preprint.

Hi,
Just for correction, the engineer for the new Dark Side Of The Moon SACD was James Guthrie, and the ORIGINAL engineer was Alan Parsons, so that was NOT engineered by the original engineer.

Hi Nika,
since reconstruction filter "headroom" isn't part of any DAC specification I know of, it's hard to know the performance in that respect. After I read your post I called Prism (my DAC brand) to ask about their specs and although they couldn't give numbers ("would have to look that up") they assured me that their reconstruction filter has been designed with "quite some" margin for overloads.
Do you know of any test or (AES?) specification that can help to quantify DAC overload performance ? I'd love to know the limits of my DAC.

Kees,

I don't know of any specification that is recognized regarding reconstruction filter headroom. As for a test, you would have to be pretty smart about it, but you could certainly create an artificial signal for testing purposes. The easiest one would be a 1/4 sampling rate sine wave wherein you create it such that the samples end up at .707, .707, -.707, -.707 FS repeatedly throughout (sampling the sine wave halfway between zero and peak throughout). If you normalize the samples then the signal will exceed full scale by about 3dB, and if the signal is clipping the filters then this would be easy to identify on test equipment on the back end.

So: create a digital signal of FS, FS, -FS, -FS (repeat ad nauseum) and this represents a 11.025kHz sine wave with the samples normalized at full scale. How does this look coming out of the DAC? Is it an 11.025kHz sine wave or is it a mess? Now start turning that down. What you see on your scope will probably change in other ways than just amplitude. By the time you get down to 3dB of attenuation you should see a pure sine wave.

My guess is that the folks at Prism will tell you that they have "tons of headroom" on their filters, but they are only speaking about the analog filters. The digital filters that are built on to the converter chips they have almost no control over. I doubt they have any more headroom there than the chips have inherently - which you'd have to speak to the chip designer about (good luck).

I hope this helps?
Nika

Aha, I think you've helped pinning down the reason for my confusion.
During my 20 years with digital audio, I've never allowed the digital signal of my (cd-)masters to reach FS completely. Overload indicators are set to 1 sample at 16 bit.
I'm still trying to find out how DAC overload can occur with signals that have never (nowhere in the signal-path, not even inside the dsp) reached FS.
Lowering overall level by say 0.01dB to avoid overload indication is a nono IMO since it doesn't remove clipping.
From what I understood, a DAC can still overload "between samples" with some signals (probably with lots of high freq. content), even when the signal never reaches FS.
In your FS, FS, -FS, -FS example, you're proposing to create a distorted (by definition, at least in my vocabulary) signal to start with. I have no doubt that this might result in audible distortion.

ps: if this is getting off topic too much I wouldn't mind a new thread.

Doesn't matter. Set it as I said except .01dB lower. You now have FS(-1 LSB) FS(-1LSB), -FS(+1 LSB), -FS(+1LSB) etc. yet your waveform is still exceeding full scale by 3dB, and this is a "legal, legitimate" waveform - a sine wave at 11.025kHz.



Go to http://www.cadenzarecording.com/papers and read the paper on "traditional peak meters." There are graphics and whatnot that explain.

Cheers!
Nika

I've tried the test and didn't hear strange things. But then, to my not so young ears there is no difference between a 11.025 kHz sine wave and a ditto square wave. So if there's only harmonic distortion, I won't hear it. When I have some time I should take a look at the spectrum of the analog output of the DAC. Perhaps there's something visable but inaudible.

Your example is *useless* for addressing that issue. The two layers opf Dark Side have *clearly* been remastered *very* differently. This is obvious from visible inspection of the waveforms, as well as objective measurement of same. I've done this at home, and so has John Atkinson of Stereophile. This means that using two different A/D transfer chains for the analog output -- Atkinson's being *far* more high-end than mine -- we were both able to see that the two layers are mastered differently. My own experience in doing the same comparison on other hybrid SACDs, btw, is that some CD layers *do* look extremely like the SACD layer (e.g., the Rollling Stones hybrids). DSotM's isn't one of them.

http://www.stereophile.com/news/11649/

With the mastering of the two layers being this different, you can't separate that effect, from any supposed difference between the *formats*.

Scientific evidence for *audible* difference of SACD/DVD-A/Redbook CD distribution formats *qua formats* simply has not been forthcoming, either from the industry or anywhere else.

From what I read on prosound forums, it's more a problem of moron producers/musicians demanding
entry into the loudness wars...and mastering engineers (who are hired hands, in the end) doing the job they're paid to do.

That Krall CD waveform doesn't look that bad, really , compared to lots of the stuff out there that's close to solid-brick green.

Great question. I'd say it's because it's not easy to queue up SACD and CD tracks togther quickly. But it's quite easy to queue up multiple CD tracks (changers have been around for decades, and now of course we have lossy format players in abundance). Also, SACD's afaik don't tend to get much radio play. So it's the 'competition' between CDs, not between CD and SACD tracks, that has producers and musicians worrying that their mixes aren't 'punchy;' enough.

There's another reason for the SACD being quiter...

The DSD signal on SACD is always at digital full scale - that's the nature of DSD.

However, you can't put information in the audio band at a level comparable to digital full scale because it would make the whole DSD conversion process go unstable. You might just get a ~1MHz square wave with absolutely no audio information modulated onto it at all, or something equally unpleasant and unrelated to the original audio. So you keep well away from digital full scale.

How "loud" SACDs appear at the analogue outputs relative to a standard CD depends on how the DAC is calibrated, but for DSD encoding, Sony suggest using 75% modulation as the limit, and not one that you should be aiming to hit.

For these reasons, brick wall limiting is pointless and useless for SACD, unless you want that "sound". Plus (I think) Sony have some control over what is release on SACD, and in an attempt to make the format "sound good" in the ears of consumers, they try to ensure only high quality recordings are released on SACD.

Cheers,
David.

Actually, before each release, all the DSD track are verified in Japan. The dynamic of the high frequencies being limited, they want to be sure that it won't be distorted.

.
Hands down, flat-out, the defacto study of DVD-A vs. SACD...

http://www.hfm-detmold.de/eti/projekte/dip..._paper_6086.pdf

Andrew D.
cdnav.com
.

This study does not seem to reach any conclusion that either is "better", only that under rare conditions they can be perceptively different. It would have been a lot more interesting to me if it had compared either one to 44.1/16.