[Tomcat]

I just want to add some confusion by mentioning a thread from our learned colleagues over at the newsgroup alt.rec.audio.high-end:
http://groups.google.com/groups?hl=d...fe099,159&ic=1
Basically it is a discussion about this paper: "Why 1-Bit Sigma-Delta Conversion is Unsuitable for High-Quality Applications" by Stanley P. Lip****z and John Vanderkooy Audio Research Group, University of Waterloo Waterloo, Ontario N2L 3G1, Canada. This thread is 159 posts long and you will probably need some stamina for reading it. But there is some healthy dose of Sony-bashing in it. Since I have not yet forgiven them for introducing CDPs - I enjoyed it.

 

[ai0tron]

Ok the first 40 posts are these guys just bitching at each other about who knows what and who doesn't etc. Should I keep reading?

 

[DanG]

Well, I snuck into a Barnes and Noble on the way home from work to read July's Stereophile... they were out, so I picked up this month's copy of The Absolute Sound instead. Wow, what a good read... this month's issue is primarily devoted to the debate of DVD-A vs. SACD. (Articles on Hi-Rez audio make Dan happy

) Of course, I was too cheap to pay the $7.95 cover price, so I read it in-store

And, of course, I was too foolish to think that it might be available online (here, in fact).

Anyway, the main article discussing the pros and cons of DSD (high sampling rate, 1-bit depth) vs. PCM (lower sampling rate, 16- and 20-bit depth). The mainstay of the writer's argument against DSD was, actually, the paper mentioned by Tomcat, written by Lipschitz and Vanderkooy. The reason they dislike DSD, apparently, is that it requires far too much information to eliminate audible noise from a recording algorithm that has such a high sampling rate; in fact, eliminating noise from audio above 50 kHz DSD audio is almost impossible given the amount of space it would take up. Apparently, DVD-A's PCM technology can eliminate noise up to twice as efficiently (in the audible spectrum, as I understood it).

Of course, as the author himself states, few adults can hear frequencies above 20 kHz, and no human can hear above 25 kHz... the reason he thinks it so important that DSD has difficulty eliminating noise above 50 kHz is that he believes it's just another piece of evidence pointing at the inherent and central deficiency of DSD -- no matter how "perfect" you make it, it can never be perfect; aside from that, it takes up too much space.

I wholeheartedly agree that if you use higher and higher sampling rates for DSD, it will still never become perfect. I also agree, as should anyone, that given higher and higher bit-depths and sampling rates for PCM audio, PCM will come closer and closer to "perfect," or life-like, sound. I have always agreed with this.

However, it is important to note that while it's important to judge in terms of future potential of technological paradigms, you can only judge different recording formats by their own merits; in the case of DVD-A, its highest potential of 24/192 audio is still (debatably) worse than SACD's standard DSD audio. And furthermore, the DSD technology used today, in my opinion, is more sophisticated and closer to real, live sound than the best commercially-available PCM audio available in DVD-A form.

In fact, recording engineer Michael Bishop whose views are also featured in the magazine (and within the page whose URL I gave above), when listening to DSD, DVD-A PCM, and standard redbook (regular CD-style) PCM, decided along with the other engineers present that DSD was clearly more realistic and thus better than DVD-A's PCM.

So, to sum up my argument: while I agree that in theory PCM has more potential than DSD, I think the technology available today makes SACD's DSD superior in sound to DVD-A's PCM because it is a format fundamentally geared towards media that still don't have the data-storing capacity needed for perfect PCM sound.

Dan

 

[thomas]

yes, in theory, PCM is definately better than DSD

The reason that SACD seems to be better than DVD-A right now is the simple fact that there are NO R2R 192khz DACs. In all currently available 192khz DACs, the PCM is converted into DSD anyways by the delta-sigma modulators. So comparasons between DACD and DVD-A are both comparasons of DSD.

And in which format the data is stored is meaningless. The only time where there is a difference is in the DAC stage. and once high quality r2r dacs for DVd-A are developed, then i am sure it will out-perform SACD. This has nothing to do with data storage capacity, DVD-A already storesa lot more data than SACD.

 

[DanG]

Thomas, apparently you didn't understand me. What I was saying is that when we have media that can hold far more information than DVDs or SACDs (and believe me, there are many in development that promise to do so), it may be possible to store near-perfect sound in PCM format. Until then, SACDs sound better -- not because of the way that they're recorded, since direct PCM recording is possible (read the link I supplied), but because DSD is a better solution for the amount of data we can deal with at this time.

 

[Holy Zoo]

To throw another variable in the mix...

DVD-A has an *audible* watermark imprinted (for lack of a better word) in the audio stream.

I read on red.audio.high-end (so take this with a grain of salt) that double blind tests have already been done between watermarked and unwatermarked DVD-A, the results showed that the watermarking *was* indeed audible and degraded the sound.

Last... there have been extensive test done in recording studios with top notch recording engineers that tested a live jazz band coming into a mixing board and then split into:
44KHZ PCM
196KHZ PCM
DSD

They switched back and forth between the live feed and the three digital, and *ALL* of the engineers preferred DSD, saying it was nearly indistinguishable from the live feed.

If anyone's interested, I can try and find the article.

 

[ai0tron]

thanks DanG I learned mroe from your summery than I did reading those ridiculous posts form supposed sound engineers.

 

[aos]

I'd personally prefer opinion of 8 year olds who can still hear up to 20kHz instead of some audio engineers who abused their ears for 30 years or so...

Not that it matters anyway. Which format will win will be decided on some exclusive golf course or a formal ball or something.

 

[thomas]

there is no such thing as perfect sound.

24bit/sample is already capable of resolving signals weaker than resistor noise.

0.00000011 Volts to be exact

And 192khz sampling can reproduce frequencies up to 96khz. That leaves plenty of room for filtering and should be able to "perfectly" reproduce EVERYTHING between 20-20000hz.

How much more data do you want?

So DSD uses a unique method of noise shaping (not so unique, its been around for more than 10 years) to make it a little for efficient than PCM at the same bitrate. (concentrates all the noise into the high frequencies) However, it DOES not offer twice the efficiency of PCM.

The point i'm saying is that current comparasons must be taken with a grain of salt. DSD vs PCM does not matter at all during the storage stage, it only matters in the actual digital to analog conversion. And every DVD-A player converts the PCM to DSD in the process.

i'll quote thorsten loch here:
Quote:

The big advantage of the Timeslicing DAC [ie DSD] is that it is an almost pure digital circuit. All and everything can be done in the digital domain, so much so the output from earlier Timeslicing DAC’s is a pure pulsewidth modulated squarewave. And unlike precision analogue Chip’s, even very complex digital chip’s can be made dirt-cheap. Even the integrated analogue circuits are usually of the “5 pence the piece if you buy 1000” type, making not much of an impact on production cost. And let’s not forget that many Multibit DAC’s require four to six separate supply-voltages. The Timeslicing DAC usually requires only one and that of course costs a lot less again. And the production cost is exactly the reason for the introduction of Timeslicing DAC’s. Any claims for superior performance or any better technology and such, can be safely relegated into the realm of fairytales spun by marketing futzies. Neither sonically or measurable are there ANY advantages for the current Timeslicing DAC’s over classic Multibit Designs. On the contrary, they are technologically much less capable of high fidelity reproduction than Multibit DAC’s and as they operate on the basis of clock cycle modulation any instabilities of the master clock (jitter) have a much larger impact on sound as with the old Multibit Designs.

and i think the biggest limitation to digital audio today is not the resoltution or sampling rate, but jitter. Jitter is somthing that can't be solved by using more and more data, but must be solved by design. And PCM is inheritantly much more resistant to jitter than DSD.

 

[ai0tron]

I think jitter filters really make a difference with PCM as well. Considering that it appears that DSD would almost require antijitter filters. I currently have one hooked up to my dcm 370 cdp and it enhances the sound in a million and one different ways. Smoother more extended top end, fuller deeper bass, better imaging, wider soundstage, incredible instrument seperation etc. of course part of the improvement is my link dac.

 

[DanG]

Thomas, you again interpret my statements in ways not in accordance with their (explicit) meanings.

First of all, perfect sound can exist when we're talking about the digital representation of sound. No, there is no such thing right now, but as I have said many times before, I'm talking about the potential of digital technology. I'm not talking about the problems of not having enough channels, enough high-frequency extension in tweeters, or anything of the sort; I'm talking about the ability to record all the information of real, live sound.

Just because we have the ability to record frequencies from 10 Hz to 96000 Hz doesn't mean we have the ability to record them perfectly. Anyone who has listened to crappy Sony headphones (like the V700DJ) knows that they can reproduce rather low frequencies and yet so horribly that it's obviously unrealistic and unfaithful to its source. With digital audio, especially PCM, we have to consider the issue of resolution. Although to record frequencies up to 25 kHz we only need to have a sampling rate of 50 thousand samples per second, to record perfectly all the sound we hear we probably need a sampling rate nearer to 50 million samples per second with a bit-depth greater than 24 bits per sample.

A crude analogy for the PCM algorithm is a computer monitor or television screen. Many people are familiar with the term "resolution" for computer monitors. If you have more pixels on your screen, pictures displayed on the screen have greater resolution. You have more pixels on your screen when you have more lines of resolution, both vertical and horizontal.

Although sound is a longitudinal wave, that is, a wave with one dimension, we can represent sound as a two-dimensional wave (or, more precisely, a complex two-dimensional wave). Imagine that you can draw the "picture" of sound on a piece of paper by drawing out a wave. But when you draw this picture, you have to do so by drawing it dot-by-dot. If your dots are few and far between (horizontally), you won't have a picture that's very realistic. If your dots are never close to each other vertically, you still won't have a realistic picture.

That's the kind of resolution problems present when recording using PCM. If you can get the vertical and horizontal distance between these so-called dots to be infinitessimally small, then you have perfect sound.

Of course, this is simply not possible. You cannot have zero distance between dots. However, the ear will not be able to perceive the difference at a certain point, and beyond this certain point, there will be no difference between actual sound and its digital PCM representation.

Thus, if you think that there's no reduction of sound quality at the stage of putting the 24/192 signal onto a DVD disc, you're wrong. Just because a sampling rate of 192 kHz allows for frequency reproduction up to 96 kHz, what's in between 0 Hz and 96 kHz is not necessarily perfect. There are problems with jitter. These problems may be as large (or even larger) than the problems of digital representation. But as we are talking about the theoretical potential of each format per se, that's all that has to enter into the discussion. I agreed that there are problems with the extreme sampling rate used for DSD.

 

[Trap]

Well mathematically speaking every bandwidthlimited signal is perfectly represented when sampled at at least twice the highest signal frequency at an infinite resolution.
The limited sampling resolution introduces a noise, the quantatisation(sp?) noise. Thats it.
44 kHz/24 Bit is a perfect representation of the original signal for the human ear, because it is bandwidthlimited at around 20 kHz and 24 bits is more than the hearing range of 130 dB, so the noise is outside of our hearing range.

Practically speaking it is far easier to construct a DAC for higher resolution formats (especially for SACD), because the DAC may then introduce errors and still produce no hearable differences. With the minimalistic representation the DAC has to be perfect.

DanG, first of all the bitrate in a PCM format only dictates the dynamic range of the format between full scale and noisefloor. 130 dB of it is enough, the ear has no bigger range.
Secondly, if you choose to use 44.2 kHz as your sampling frequency you have to completely remove all frequencies above 22.1 kHz from your signal before observing anything about the behavior of the system. The system is not working above 22.1 kHz and you don't have to check this, it's already known.
If you have done that there are no changes between two of your dots that are not already known. You can easily compute every value for any time using a discrete fourier transformation.

The ability to record all the information of a real, live sound? All information of a sound that the ear can notice ought be enough. And this is not limited by the digital format as i have shown above. It is limited by recording and playback. You have to reproduce the physical sensation of sound, so you have to use speakers. Speaker playback only works perfectly, when there is no room to reflect the emited sound and there is a seperate perfect speaker for every soundsource. Recording is even more complicated.

Your analogy with a monitor is flawed because the digital format is already better than our ears, the monitor is not better than our eyes.

 

[DanG]

If what both of you (Trap and Thomas) say is so, then why is it that all recording engineers say that 24/192 recordings sound better than 24/96 recordings? And why do 24/96 recordings sound so much better than 16/44.1 recordings?

 

[Holy Zoo]

Quote:

Originally posted by DanG If what both of you (Trap and Thomas) say is so, then why is it that all recording engineers say that 24/192 recordings sound better than 24/96 recordings? And why do 24/96 recordings sound so much better than 16/44.1 recordings?

It is conceivable that if the listening tests are being done "sighted" - meaning the recording engineers know which flavor of digital recording they are listening to - that they are just imagining that one sounds better than the other. Meaning: they WANT one to sound better.

I don't think we'll really know if these guy's can truly tell the difference till someone does a double blind test between CD and SACD and/or DVD-A.

Oh... and remember the article I mentioned earlier about the tests done with a live jazz band and several top recording engineers? Here's the article:

http://www.theabsolutesound.com/tas130_sacd.htm

A snippet from it:

In 1998, we had a live, in-concert session planned with Oscar Petersen and his quartet in Munich, Germany. This would be a prime opportunity to record in different high-resolution formats, both stereo and multi-channel surround, and have immediate comparison to the live source on-stage. By that time, Data Conversion Systems had added DSD capabilities to its high-resolution converters. Philips also had developed a prototype 8-channel DSD recording system. With fortunate timing of schedules, and the wonderful cooperation of the people involved, we were able to have a stereo + 6-channel dCS/Genex 24-bit/96-kHz system, a stereo dCS/Sony DSD system, and a stereo + 6-channel Philips DSD system present at the session. Jack Renner and I were engineering — Renner was taking care of the stereo mix and I was doing the 6-channel surround. Gus Skinas of the Sony Super Audio Group coordinated the set-up of the Sony and dCS DSD equipment, with Mike Hatch of Floating Earth Audio assisting. Andrew Demery and Walter Verhoef of Royal Philips Electronics handled the set-up and operation of the Philips 8-channel prototype DSD recording system. Recording engineer Martin Wieland coordinated our work with the Munich Philharmonic Hall crew and set-up.

We were able to instantly compare the 24/96 PCM and DSD stereo returns with the live feed from the console repeatedly during the in-concert recording. Remember, both versions were from the dCS converters, so all was equal between the converters except for the digital outputs, one being PCM and the other DSD. All in attendance agreed the DSD return was so close to sounding like the live feed that it was hard to tell which was which. In switching between the live feed and the DSD, we had to keep double-checking to see if it was the correct source — it was that close. By contrast, switching to the 24/96 return always called attention to itself. It sounded like a "recording." The image collapsed slightly and high-frequency detail would cluster somewhat compared to the live feed. The most noticeable difference between the live feed and the 24/96 was the change in piano harmonics, particularly the definition of the low end. In addition, the high-end of the piano became somewhat glassy in the 24/96 feed. The definition of the acoustic bass smeared slightly. These were not dramatic differences — they were fairly subtle. Detecting them required close scrutiny by all listening. If the 24/96 recording had to stand by itself with no direct comparison, I'm sure it would be considered an outstanding recording. After a few comparisons, most wanted to return to the DSD feed, preferring that chain to the PCM. With further comparison to the live feed, we were able to determine that the dCS DSD converters seemed to be accentuating the high frequencies to a small degree, causing the drum-kit cymbals to be a bit more "pointy" than on the live feed. Most importantly, however, the piano harmonics and depth remained intact on the DSD return. The recording's width did not change from live to DSD.

 

[thomas]

Quote:

Thomas, you again interpret my statements in ways not in accordance with their (explicit) meanings. First of all, perfect sound can exist when we're talking about the digital representation of sound. No, there is no such thing right now, but as I have said many times before, I'm talking about the potential of digital technology. I'm not talking about the problems of not having enough channels, enough high-frequency extension in tweeters, or anything of the sort; I'm talking about the ability to record all the information of real, live sound.

I never said anything about the number of channels, high frequency extension of tweeters, or anything of the sort. I've been talking about digital technology the whole time. Perhaps you are not interpreting my statements correctly.

Quote:

If what both of you (Trap and Thomas) say is so, then why is it that all recording engineers say that 24/192 recordings sound better than 24/96 recordings? And why do 24/96 recordings sound so much better than 16/44.1 recordings

What i were saying is that there are limits to how effectively natural sound can be recorded or heard.

Humans hear sound based on frequency vs time, not amplitude vs time. So the exact shape of the sound wave is not what matters, its the frequencies it contains. There are an infinate number of different wave shapes that can sound identical to humans. Humans can hear up to approx. 20 khz, some people a little more, some people a little less. 192khz PCM is rounding off the soundwave, but much less than our own ears are rounding off the signal.

There are also limits to differences in amplitude that out ears can detect. However, an even smaller limit is the resolution of the electronics in the signal path. Everything with a resistance will add noise to the signal, and the adding more bits to the DACs resolution will only capture more noise. No new information will be captured. And human ears are even less sensitive than the electronics.

So why does 96/24 sound better than 44.1/16?

Thats simple- 44.1 sampling rates can only record a maximum audio frequency of 22.05 khz. That means that everything about 22.05khz must be COMPLETELY blocked out. Try designing a filter that can block out EVERYTHING above22.05 without affecting anything below 20khz! Its impossible. Its this brickwall filter that destroys the sound of 44.1 PCM. And then there's the 16 bit resolution. That is not enough data to store everything that is in the signal, there is some rounding off of the signal. 24 bits is enough.

Why does 192khz sound better than 96khz? personally, i've never seen a direct comparason, and i doubt that you can make a blanket statement that "all recording enginners" feel this way.
How many recording engineers are even using these new formats?

If this is indeed true, then the higher sampling rates of 192khz allow higher frequencies to be captured, allowing the use of filters with a more gentle roll off. This will have virtually no effect on audiable frequencies. Also, its possible that some frequencies about 20khz can have an effect on the sound. See apheared's post in the music forum. Again, if that article is accurate, then maybe we should record frequencies up to 100khz (ie 192khz pcm). But saying that we need to go beyond this is crazy- no musical instrument can possibly put out a lot of energy above 100khz, and even if it did, it would have no effect on us.

And if that is true, then we should definately use DVD-A instead of SACD. SACD concentrates its performance on 20-20khz, and is slightly more efficient becuase it shifts all the noise up ultrasound frequencies.

So if you claim that ultrasound is so important then DVD-A would definately be the best choice.

Finally, the last and most probable explanation for 192khz sounding better is that all electronics will make slight errors and add some distortion. By using a DAC that uses more bits and faster speeds, then the effect of slight errors are minimized.