24/192 Music Downloads and why they make no sense

So the CD player is on it's way out...what are you going to be listening to, live music?

This is very interesting. I read the whole article, and it sounded very convincing. I guess it's time for me to stop wasting money on audiophile 192kHz/24bit tracks... More money left to buy good old 44/16 ones.

This hi-res business reminds me a bit of what happened with point-and shoot digital cameras, when the Mpixel number just kept creeping up, as though it was the sole indicator of image quality.

Thanks for sharing, Jinjuku. I had been wondering whether or not there was much benefit from the higher sampling and bit rates. I think I'll be quite happy sticking with 44/16 files and--based on that other recent discussion--my internal sound card that has noise and distortion well below reasonable listening range of human hearing

If you want to hear brick wall filter ringing instead of music, this is your choice. The key to high-resolution audio is not very high fequency content, which you can't hear (and I agree with it). But the way you process analog signal. Brick wall filter with very sharp cutoff influences the sound in audible range. That is why people can hear differences in digital filters. The higher sampling frequency, the less sharp filter is needed. That is what improves sound quality.

I'm aware of extensive listening tests that were done to listen for this kind of fault.

Since people rarely if ever hear the imposition of a 22.05 KHz brick wall filter on so-called hi-rez recorded music, it is hard to get excited about this problem with good modern converters.

The problem is that the ringing takes place close to or at the Nyquist frequency where it is generally not heard unless very intense, and also masked by musical program material.


Thing is, processing the analog signal transparently is commonly done, except in really cheap equipment such as the headphone jack driver of a CD-ROM drive.




It is now possible to produce this kind of filter in such a way that it is linear phase. IOW, it adds very little phase shift right up to the cutoff point.


That's one of those things that it takes a sighted evaluation to do! ;-)

I've done experiements with sliding brick wall filters down to lower and lower frequencies. Usually, they start being barely audible around 16 KHz.


As I keep telling people around here, you can't improve on sonically transparent, if your goal is high fidelity. And, a good set of 44/16 converters, which are now almost free. are sonically transparent.

Prove me wrong with a properly done listening test. Pleaase!

But this is nothing new at all. We have known for a while now that we don't get audible improvement beyond CD.

I don't think anyone is arguing the value of oversampling. Rather, I believe the article discusses the utility 24/192 native recordings with ultrasonic content.

Was double blind testing used? Personally I can hear a difference, that is it is subjective not objective, between 16 bit, 20 bit (HDCD), and 24 bit but no differences between sampling rates above 48kHz. It might be my ears, I don't know. I do buy the 96kHz and some 192kHz recordings from HDTracks and have compared some music, with the same mastering, at both 48kHz and 96kHz and 192kHz, and noticed no difference which of course is totally without scientific value.

Obviously. You should try it sometime.


No, you really can't. No one can.

I will give that the credence that it obviously deserves.

I hate the way that HDTracks charges more for the higher resolution. Absurd. $11.98 is high enough for 44.1/16. $24.99 for 192/24 is just ridiculous. Whenever I have any interest in a title there, one of the first things I check is whether it is available at 44.1/16 and most of what I have purchased there is 44.1/16. I'll purchase a higher resolution if 44.1/16 is not available and it is something I must have, but even then, I purchase the lowest resolution available. Saxophone Colossus, recorded in 1956, is just not worth having at 192/24.

Yes I agree Everything streaming should be at the high rez

from an old hydrogenaudio test most could't discern a 17kHz brickwall from the original, few could distinguish 18,5kHz, one individual could barely, but statistically significantly distinguish 19,5?kHz filter. This was with actual music, headphones. With pure sines and loud enough levels this could get higher. There were of-course many component misbehaviors, like someone not hearing 17kHz, but suddenly 20kHz became audible etc - these are all equipment faults (resampling/aliasing artifacts due poor hardware etc.)
No matter how someone spins it, 16bit 44kHz can store information beyond human physiological threshold. Someone care to argue about this? Hell, pop recordings with a pathetic dynamic range can be stored with 6bit resolution

Monty? in the above link explained it well.

Under what circumstances do you hear these differences? How are you doing your DBTs?

As I said it was preference and not scientifically valid in any way. Read my post again.

Here's what you posted:


There was other stuff in the post, but the word preference did not appear there.

The goals for setting a standard here shouldn't be what is adequate but what has some safety margin as to give us high confidence of inaudibility. In that regard, we need to also allow for less than optimal implementations. To that end, Bob Stuart has published a much more authoritative version of this report at AES. Here is an online copy: http://www.meridian-audio.com/w_paper/Coding2.PDF. These are his recommendations:

"This article has reviewed the issues surrounding the transmission of high-resolution digital audio. It is
suggested that a channel that attains audible transparency will be equivalent to a PCM channel that
uses:
· 58kHz sampling rate, and
· 14-bit representation with appropriate noise shaping, or
· 20-bit representation in a flat noise floor, i.e. a ‘rectangular’ channel"


So as we see, the CD standard somewhat misses the mark on sampling rate. And depending on whether you trust the guy reducing the sample depth from 24-bit to 16 bits, we may be missing the right spec there too.

Ultimately, I think to the extent bandwidth and storage have become immaterial for music, it is best to get access to the same bits the talent approved when the content was produced. For a high-end enthusiast, there is no need for them to shrink down what they recorded before delivery. Let the customer have the same bits and then there is no argument one way or the other .

Do you like M&M?

"personally I can hear" was meant to indicate a subjective difference, not an objective one, therefore I've changed it.

Interestingly enough this paper is neither an AES conference paper or a JAES article." At least I can't find it published that way. It appears to be a rewrite of a 1988 (24 year old!) article in the now-long-departed Audio magazine. It's a corporate white paper that has no standing as an industry standard or recommendation.

This paper is arguably part of the support for SACD and DVD-A which are now known to be failed technical initiatives that failed to make it in the mainstream consumer marketplace.


The paper in question is full of unsupported assertions. Probably the most honest statement it contains is:

"Different listeners bring different prejudices."

Managing those prejudices is a well-known and widely-practiced art (among professionals, particularly those doing R&D) which the paper in question thoroughly ignores.

The sensible statement above is followed by the following which is easy to misinterpret:

"As digital audio has progressed, we have also evolved the capability to record and play back with resolution that exceeds that of Red Book CD and current studio practise recognises this Red Book channel as a bottleneck'"

The above is a true statement because we do have the ready ability to record and play back with word lengths and sample rates that vastly exceed that defined by the Red Book. Good quality audio recorders and computer interfaces running at 24/192 and delivering approximately 20 bit actual performance are readily and fairly cheaply available. I personally own several.

However current studio (and live recording) practice includes a lot more than just recording and playing back line level audio signals. It includes the sonic purity of the performance space, playback room, microphone, speakers, and human ears. It is well known that these are the actual technical bottlenecks, not the use of Red Book CDs and equivalent computer audio files for distribution.

"There is very little hard evidence to suggest that it is important to reproduce sounds above 25kHz. Instead there tends to be a general impression that a wider bandwidth can give rise to fewer in-band problems."

The above is a mixture of technical truth and opinon stated as fact (OSAF). The technical truth is that "There is very little hard evidence to suggest that it is important to reproduce sounds above 25kHz." The OSAF is that "Instead there tends to be a general impression that a wider bandwidth can give rise to fewer in-band problems." In fact many careful workers have encountered serious problems with audible artifacts due to excessive bandwidth.

http://people.xiph.org/~xiphmont/demo/neil-young.html:


The key sentence above is: "... unnecessary reproduction of ultrasonic content diminishes performance."

Enough said, eh? ;-)

I like the dark chocolate ones.

The online version of the paper is a revised an updated version of what was presented at AES Conference: http://www.aes.org/e-lib/browse.cfm?elib=7140 [ from 1997 not 1988]

"Coding Methods for High Resolution Recording Systems
This paper reviews the recording and reproduction chain from the viewpoints of digital audio engineering and psychoacoustics. It also attempts to define the audio requirements of a transparent digital audio channel. The theory and practice of selecting high sample rates such as 96kHz and word lengths of up to 24 bits are examined. The relative importance of sampling rate and word size at various points in the recording, mastering, transmission, and replay chain is discussed. The paper then examines types of coding that are capable of attaining the target performance, describing the advantages of schemes such as lossless coding, near-lossless coding, and matched noise shaping with pre-emphasis.

Author: Stuart, J. Robert
Affiliation: Meridian Audio Ltd., Stukeley Meadows, Huntingdon, UK"

In contrast, we are discussing an online article by Xiph that has none of credentials you require and has its own built-in biases.


Bob Stuart company's solution for lossless compression, MLP, was selected as the mandatory lossless audio compression for that for DVD audio and eventually (optionally) in Blu-ray in the form of Dolby TrueHD. As such, Bob had little love for SACD. If you read the report, or even if you don't and go by the conclusion of what I post, no one who is a die-hard DVD-A fanboi would say that anything above 14-bits dithered and 58 KHz is a wasted of bandwidth. So no, your characterization is almost the opposite of the spirit of the paper.

Bob's position is that of a moderate here that simply says CD is just a bit too low of a spec and he goes on for a whopping 38 pages making his case, far, far more detail than the article being discussed in this thread.

FYI, I usually cite this paper when someone demands to highest sampling rate, bit depth, not the other way around. It makes for very poor evidence in support of the formats you mention due the conclusions it reaches, especially regarding bit depth.


As is your statement Arny . The paper has 27 references (none of the forum arguments as cited in Xiph article). It has ~17 pages of graphs, simulations and diagrams to back what it describes. You read two reports that more or less agree with each other yet opt to back the one with smaller numbers better. I call that prejudice .


The paper is a scientific look at the issues. You make it sound like he listened to two tracks and declared high resolution audio the way to go. Look at statements like this from Bob:

"Although there is a small lobby that suggests even higher sample rates should be used – like 192kHz– the author disputes this; preferring to point out that when 96kHz channels have been correctly designed in terms of transmission, filtering, etc, that higher rates simply will not offer any benefit.

I realise that by expressing the requirement of transparent audio transmission – I am nailing a flag to the mast and lay myself open to all kinds of attack! However, this analysis has been based on the best understanding to date on this question and we should exceed this requirement only when there is no detrimental cost to doing so."

So as you see and I noted before, his stance is against "biggest numbers are better" and agrees far more with the Xiph article than disagrees.


As I noted regarding bias, the rest of us also express opinions all the time. So putting aside the authoritative position Bob holds with 30+ years of design and research in digital audio, your attempts to position us us as saints who never post opinions, and folks who have published such papers at AES do is non sequitur. Yes, as with everyone he expresses his opinions at times based on wisdom he has gathered in this field. That alone doesn’t make what he says wrong unless we can demonstrate his science is wrong.

On the topic of opinion, how about this statement from Monty in his Xiph article:

“It's true enough that a properly encoded Ogg file (or MP3, or AAC file) will be indistinguishable from the original at a moderate bitrate.”

Indistinguishable at moderate bit rate? What is a moderate bit rate and where is the data to prove that?


Not at all. Indeed, he didn’t say enough. This was of the most disappointing parts of Xiph article for me. Here is what he says:

"The effect is very slight, but listening tests have confirmed that both effects can be audible."

I can't find any reference he provides for that. Maybe there are such tests and if so, I love to read them but for now, you don’t say there are listening tests and not at least mention them by name.

As to the graph above, he assumes the 26.5 Khz ultrasonic tone to have the same amplitude as the music signal at 7 KHz (green and red bars respectively). This can happen with SACD with its noise shaping but probably very rare with naturally recorded PCM music. If it does happen in real music, are we supposed to chop it off as to take it easy on our equipment?

Look at it this way. Are we to believe that we can design audio systems that have inaudible distortion if the frequency response up to 22.5 Khz of CD, but somehow fall flat on our face if we bump that up just 16% to 26.5 KHz? You won't find anecdotal type data points like this in Bob's paper.

Again, let me repeat that Bob's paper is much in support of this article in grand scheme of things rather than the other way around. It is just that if you are going to read something like this, read Bob’s paper which is from someone who has designed such equipment and has far more credentials in this field than Monty. For a web article, Monty's article is very good but let's not be so biased as to put his effort forward as A+, and people whose shoulders he is standing on, much lower, just because they dare setting a more solid standard for audio.

Bottom line is what I said: music product in the studio is not 44.1 Khz/16-bits. If folks are not troubled by the bandwidth and storage costs, let's let them have what is produced there. Folks like yourself can chop it down to the resolution you like. And others can leave them as is. The notion that in this day and age we should stick to the CD spec as if that exact sampling rate was magic, when we don't use that medium anymore in digital distribution, is antiquated in my opinion .

Amir:

Can you hear the difference between an equally mastered SACD and CD?

When the format came out, I did a bunch of comparisons and would often find a small preference for SACD. It is a format I don't like technically (find it silly that it advocates wider bandwidth for audio and then stuffs a bunch of noise in ultrasonics) but my listening results pointed that way. I suspect though that those were high-resolution PCM recordings converted to DSD and then to CD. What dithering/resampling (transformations) was used in each or additional manipulations of CD, I don't know. And of course equipment design varies.

The biggest issue with SACD wasn't its performance but the fact that they thought they should outlaw PC playback and ripping. When it comes to music, I almost always put convenience ahead of performance. If I can't listen to some music, I don't care how good it can sound .

the actual benefit of sacd comes due to better mastering. dynamic compression is the real evil, not the bandwidth.

There is also the MCH aspect--I consider that the most important benefit of SACD (and thus my SACD purchases are, with rare exceptions, MCH).

That's because he didn't feel the need to given the number of other papers that mention this effect and even go way out of their way to address is.


Their name is legion.

Here is one example:

"Most of the conventional audio systems that have been used to present sound for determining sound quality were found to be unsuitable for this particular study. In the conventional systems, sounds containing HFCs are presented as unfiltered source signals through an all-pass circuit and sounds without HFCs are produced by passing the source signals through a low-pass filter Thus the audible low-frequency components (LFCs) are presented through different pathways that may have different transmission characteristics, including frequency response and group delay. In addition, inter-modulation distortion may differentially affect LFCs. Therefore it is difficult to exclude the possibility that any observed differences between the two different sounds, those with and those without HFCs, may result from differences in the audible LFCs rather than from the existence of HFCs. To overcome this problem, we developed a bi-channel sound presentation system that enabled us to present the audible LFCs and the nonaudible HFCs either separately or simultaneously. First, the source signals from the D/A converter high-speed, one-bit coding signal processor were divided in two. Then, LFCs and HFCs were produced by passing these signals through programmable low-pass and high-pass filters respectively, with a crossover frequency of 26 or 22 kHz and a cutoff attenuation of 170 or 80 dB/octave, depending on the type of test. Then, LFCs and HFCs were separately amplified with power amplifiers, respectively, and presented through a speaker system consisting of twin cone-type woofers and a horn-type tweeter for the LFCs and a dome-type super tweeter with a diamond diaphragm for the HFCs. This sound reproduction system had a flat frequency response of over 100 kHz. The level of the presented sound pressure was individually adjusted so that each subject felt comfortable; thus the maximum level was approximately 80–90 dB sound pressure level (SPL) at the listening position."



Amir, the paper you object to so strenuously full of factoids that are well known to people who are well read.



Again Amir if you were well-read you'd know that musical program material with this kind of spectral content is commonly used by advocates of very high sample rates. One word: Gamalon.

It is highly unlikely that there would be a SACD with this kind of spectral content due to noise shaping. Again, a little reading of even recent AVS forum threads would set you straight about that.

http://www.lindberg.no/english/collection/004.pdf

Figure 3 shows that the shaped noise above 20 KHz is 55 dB or more below FS. It is highly unlikely to be the source of audible IM products with even marginal equipment. About the only amplifier nonlinear enough to potentially cause problems from this source might be a SET.

Also:




Depends on how important it is to you that listeners have the most sonically accurate possible reproduction. In this case chopping off the ultrasonics provides more accurate (as in less audible distortion) sound for a wider audience of listeners.

Remember, people who produce mainstream music recordings don't make money by selling recordings that go out of their way to make their customer's audio systems sound bad.



Irrelevant argument. This discussion is not about recording formats that support frequencies limited to just 26.5 KHz. It is about recording formats like DSD (SACD) that claim response up to 90 KHz and beyond.

That's potentially a trick question since we generally don't know all of the details of how recordings are mastered.

There's only one way to know for sure that a high sample rate source and a lower sample rate source are comparable, and that's to do the downsampling yourself. This is BTW no big deal to do at home.