Focus on 24/192 Misguided?.....


As I've upgraded by digital front end over the last few years, like most people I've been focused on 24/192 and related 'hi rez' digital playback and music to get the most from my system. However, I read this pretty thought provoking article on why this may be a very bad idea:
http://people.xiph.org/~xiphmont/demo/neil-young.html

Maybe it's best to just focus on as good a redbook solution as you can, although there seem to be some merits to SACD, if for nothing else the attention to recording quality.
128x128outlier
04-23-12: Audiofreak32
To summarize....

I guess we have concluded just the opposite - that 24/192 is a VERY GOOD idea actually,....
I don't know how you concluded this? I concluded that there is no need for 24/192 - 24/96 suffices as it solves the issue of steep skirt analog filter for 16/44.1 & that once this issue is resolved & we have more DR w/ 24 bits there is no real need for 24/192.
To summarize....

I guess we have concluded just the opposite - that 24/192 is a VERY GOOD idea actually, as it SOUNDS GREAT and hard drive space is fairly inexpensive.

Using dbPoweramp (very cheap program) to rip your CD collection into 24/96 or 24/192 FLAC/WAV is easy to do and once the up-front work is done, you can sit back and enjoy music at least at or above the quality level of your current CD playback system all in the comfort of your listening chair.

Also, we have concluded that all the specs, tests, charts and measurements will not tell you how good a component will sound (especially how it will sound to a given person), as that is 100% personal opinion which cannot be measured at all.

Note: For what I would consider a "budget friendly" choice for a DAC solution that will play up to 24/192 files - the "Wyred 4 Sound" model DAC-2 is a very, very good piece at $1,499 and has many connection choices and built-in volume control.
Al and Kijanki, not sure this is relevant about extended bandwidth, but when I play the Four Seasons/Jersey Boys, I hear neighborhood dogs howling when Frankie Valli hits the high ones. LOL
Thank you Al. If there are any harmonics within 50kHz amplifiers should be still linear (modulation can only happen on nonlinear element). Any problem would already show with vinyl gear that has similar bandwidth. I don't really see any source of audible harmonics above it. On one hand studio engineers would clean it up but on the other microphones already do it. Most of microphones go only to 30kHz and some extended response go to 50kHz. One of the most popular Neumann U87 ($3600) goes only to 20kHz while the most expensive I could find Sony C-800G ($8000) is only 18kHz. At the concerts as well as in the studio nothing goes directly - everything comes thru microphones to PA system or studio console. Further more, I suspect that studio equipment bandwidth does not extend any higher providing natural filters as well.
Excellent post by Kijanki. I agree completely.
04-22-12: Bifwynne
... do you know what the power bandwidth is on ARC amps, particularly the VS-115. Freq. response is approx. 100K, but I don't know if that is the same as power bandwidth.
My understanding is that unless otherwise stated frequency response and bandwidth are usually specified under "small signal" conditions. I believe that for power amplifiers "small signal" is commonly defined to mean 1 watt or 2.83 volts (2.83 volts corresponds to 1 watt into 8 ohms). Full power bandwidth will sometimes be considerably less, in part because in some designs it will be limited by what is called slew rate, which isn't a factor under small signal conditions.

The specifications at the ARC website do not appear to indicate either full power bandwidth or slew rate for your VS-115, so there isn't enough information to answer your question. As Kijanki indicated, though, high power levels are not required at ultrasonic frequencies, so small signal bandwidth is a more meaningful number than full power bandwidth.

Returning to the question of ultrasonic intermodulation distortion, I'm not sure that bandwidth limitations are directly relevant to the issue, although they might play a role. What is relevant is non-linearity. As long as the amp's output amplitude is linearly proportional to input amplitude, at each of the frequencies for which an ultrasonic spectral component is present, there won't be a problem. Perhaps there will often be a tendency for linearity to degrade at frequencies where the amplifier's small signal frequency response is rolling off, in which case bandwidth would have some relevance to the issue. Or perhaps not; I have no particular knowledge on that question.

Best regards,
-- Al
Kijanki, u could have answered your own question about ultrasonics from 192K sampling now that you have answered the question of power bandwidths of power amps. Now you can see why a power amp would be nonlinear in the ultrasonic range & why those ultrasonic intermod distortion products folded down to the 20-20K. Now , the amplitude of these ultrasonics might not be large enough but better not to have them at all.....
Bombaywalla, Bifwynne - No, I don't know the numbers, but suspect it is at about a half or less. Delivering full power at high frequencies is not really important since very little power goes to tweeter. In my Rowland 102 max power at high frequencies would damage amplifier (burn out output choke). Icepower module 200ASC used in my amp is specified at 200W at 10Hz-20kHz but it is only momentary power. FTC rated power is specified as 55W but only up to 8kHz, with warning about damage to the choke at higher frequencies. It is not really important because average power when playing music is only few percent of peak power. The reason for that is that if on average music has half of peak loudness, it means 1/10 of power (logarithmic scale) and then music also has gaps (unless one listens to sinewaves). 55W of power at any frequency above 8kHz would most likely damage any tweeter, not to mention hearing. What worries me a little is 22 deg phase shift at 20kHz (-3dB bandwidth is 65kHz). It would weaken upper harmonics summing. My amp would benefit from a little more "air" but it might also be my hearing (not getting any younger) or the fact that speaker has warm character and is never bright - even on worst CDs. It has, in the system, very clean, pronounced natural sibilants. I don't want to change it and before I audition another amp (like Rowland 625) I need to fix room acoustics.

I understand Rowland's idea behind 350kHz bandwidth in model 625 - no phase shift at 20kHz and perfect step response but 1MHz bandwidth of Soulution 710 is perhaps too much. According to reviews it is excellent amplifier with very little negative feedback but in general the easiest method to improve most of amplifier's spects (like THD, IMD, DF, Bandwidth) is to use deep negative feedback that also enhances odd harmonics (overshoot in time domain caused by amps signal delay and thus late feedback summing - known as TIM) making unpleasant bright sound that SS amps are famous for. Class D has small advantage here, being practically one stage (little delay). Less than perfect design of such 1MHz amp can cause problems including instability followed by oscillations that can damage speakers or sensitivity to RFI. I would tend to agree with Audiofreek32 that numbers are not that important and often amplifier with the best spects has the worst sound. Selecting gear for audition by company reputation or designer's name makes more sense to me.
Kijanki thanx for that info on power amp freq bandwidth. I suppose that many more amps are higher freq bandwidth than I imagined but I don't know how many of these have the same number for power bandwidth? Any idea?
Kijanki, do you know what the power bandwidth is on ARC amps, particularly the VS-115. Freq. response is approx. 100K, but I don't know if that is the same as power bandwidth.
The same recording transfered to 16/44 does not sound as good as the same recording transfered to a higher resolution whether it is SACD or 24/96 (or 24/192). Can we move on and try to get recordings put out on something higher than 256k?
According to all the engineers and tests and specs, vinyl does not sound as good as a CD right? But so many people would prefer the sound of vinyl and/or tubes over SS amps and a CD player. So I guess all this talk about specs and cutoff freq., etc. is just a numbers thing. Will it tell you if one amp sounds better to a person than another?
Bombaywalla, Making amplifiers cutoff frequency at 20kHz means that phase shift at this frequency will be in order of 45 degree causing bad summing of harmonics. My small Rowland 102 amp has 65kHz bandwidth with about 22deg phase shift at 20kHz. New Rowland amp model 625 ($15k) has bandwidth of 350kHz. Mr. Jeff Rowland knows what he's doing (I'd like to think). As for 50kHz - a lot of power amps have -3dB bandwidth of 50kHz:

All Atmasphere amps: >100kHz
All Rowland Amps: >65kHz
All Cambridge Audio Amps: >50kHz
All Krell Amps >95kHz
All Classe Delta Series: >100kHz
All Classe CT series: >80kHz
All Luxman Amps: >100kHz
All Parasound Halo Amps >100kHz
All Parasound NewClassic Amps >50kHz
etc.

But if you won't to spend 2nd mortgage you'll find amps like:

MBL Reference 9011: 320kHz
Goldmund Mimesis 8: 800kHz +/-1dB
Soulution 710: 1MHz
Onhwy61, yes most of the electronics today is still well below 100KHz bandwidth. Hi-end does not mean hi bandwidth; it means better sonics in the 20-20K band. It might be easier for preemie to extend to 10s of KHz above 20K but for power amps to have a power bandwidth of 100KHz will cost you very close to a 2nd mortgage. Don't believe me, do some research yourself & find out just how many power amps have a power bandwidth that even touches 50KHz. Find out what your gears' bandwidths are. Almost all audio gear was never meant to amplify ultrasonics. The Pro studio gear might be a different ball of wax. Thanx.
hi-res may not be necessary according to theory, however in practice modern D/A chips behave better with it (digital filtering), and as a result it sounds better.

Steve N.
Empirical Audio
Al, My DAC and perhaps a lot of them cuts at 45kHz. AFAIK vinyl recordings also extend to about 50kHz. It is considered to be an advantage of vinyl.
It no his arguments, but the conclusions he draws that takes him over the top.

Much of his argument is premised upon amplifier circuits not being able to handle ultra sonic signals cleanly. Do they still make high end electronics that can't handle a 100kHz signal?
I guess if you are happy with 16/44 then that is all that matters, but there is a whole lot more to be had.
Outlier - thanks for posting the liink. Very interesting and helpful article - no matter what side you come down on in the hi res debate. Thanks also to Bombaywalla for the "beginners guide" link. Am I missing something or isn't the convenience of computer-based audio going to apply whether or not music files are hi res?
04-20-12: Kijanki
Al, I wonder if 24/192 contains any ultrasonic frequency at all. Why would they leave it preparing hi-rez files? Where this ultrasonic frequency comes from? Again, notion that 192kHz sampling is harmful is a little farfetched.
Hi Kijanki,

What he is referring to is the ultrasonic output of the musical instruments themselves. Yes it would be at very low levels, and with a lot of instruments it would probably not be present to a significant extent at all. But his point, debatable though it may be, is that leaving it in can't do any good, and MIGHT do some harm, depending on the non-linearities that may be present in the playback system.

It would be left in the hi rez recording to avoid introducing a sharp cutoff filter into the signal path, which as you realize is one of the fundamental benefits of high rez.

Along the lines of my earlier comments, I'm skeptical and/or uncertain about a lot of his points, and how they would trade off in terms of significance against the presumable benefits of high rez. But I don't consider his arguments to be outlandish or unreasonable.

Best regards,
-- Al
Audiofreak32, I enjoy my system all the time. ALAC is lossless while wireless transmission is bit perfect. Benchmark is as clean as it gets on the verge of being sterile but it fits perfectly with my warm sounding Hyperion HPS-938 speakers. In addition my Benchmark is modified with better sounding op-amps. Sure I could be doing much better but for much more $$$$.

Linn looks very impressive but it is $7500 while Benchmark + AE were $1100 total. I'm perfectly happy with 16/44 limitation of AE, having over 1500 redbook CDs on HD. Computer costs me nothing since I already have one. My setup also requires only one pair of ICs.
Kijanki.... put aside the technical and go listen and enjoy your system. But using ALAC (compressed) and a Benchmark DAC1, you could be doing much better.
A LINN Akurate DS is around $7,500 new. Only other things you need are a NAS and an iPad. So, less than $10k retail, easily. Sure that setup you described was expensive, but you do not need a "dedicated computer" at all. I am using a $350 NAS with a 3TB HDD loaded with 24/96k and 24/192k WAV and FLAC files. The only IC's I need are a pair of RCAs into my amps. So, I am talking about a DAC (no moving parts) and yes, a HDD in a dedicated NAS, but compared to the alternative? Really?
Al, I wonder if 24/192 contains any ultrasonic frequency at all. Why would they leave it preparing hi-rez files? Where this ultrasonic frequency comes from? Again, notion that 192kHz sampling is harmful is a little farfetched. Do we have any studio sound engineers on our forum that could explain it to us?

Bombaywalla, Thanks for the info on filters. I'm dealing mostly with 4-tap lowpass FIR filters at work but 500-tap filter is really something. One graph shows interesting step response typical to most of CDPs with ringing appearing before and after the pulse. That might affect the sound since our ears are very sensitive to it. Stereophile posted similar test results comparing apodizing and non-apodizing filters. In comparison there is no antialias filters used in SACD creation making better, more natural step response (transients).
I do not have golden ears and like the sound of my system very much but just believe that processing back and forth 24/192=>16/44=>24/192 is not likely to improve anything. Higher sample rates are not to extend bandwidth but rather improve filter response reducing pre-echo effect. Apodizing (windowing) filters, available in few CDPs like Meridian, allow to eliminate pre-cho completely but AFAIK are not suitable for 44.1kHz because there is not enough space between 20kHz passband and first alias to fit filter's windowing function. DSP processing is not my field of expertise but even if everything looks peachy in frequency domain there is a lot to be improved, possibly by higher sampling rate, in time domain (transient response).

Audiofreak32, technical articles are to understand better what is happening but you're right, that at the end what counts is listening experience. At the level of 20/96 or 24/192 placebo (or negative placebo) effect might be a dominating factor. Just the fact that I feel good about my gear can make it sound better to me than to others.

Oldears, Choice or audibility of different formats might depend on setup. In my setup, for instance data is wirelessly delivered ALAC compressed to Airport Express and contains no timing. It is also bit perfect. Lack of timing is important because it eliminates any influence of computer processing or playback program, computer noise, etc. At this point timing is recreated in AE and data is streamed to Benchmark DAC1 with low 258ps jitter further suppressed by Benchmark processing. I could also save data in other formats but it would eat up some processing power of my computer that I use for other chores (like typing this).
put it into a device that could only add noise and jitter and eventually will fail (drive, optics), ....
hey Audiofreak32, don't talk about hardware failures! With your being sold on computer playback you don't have a leg to stand on when it comes to hardware failures! How often does computer hardware fail compared to CD drive & its optics? Even the cheapo $40 DVD players from Walmart outlast almost all HDDs & other computer hardware.....

Yeah, the convenience of HDD playback is immense, I have to agree.

If you have not heard a properly setup DAC with hi-res files, you owe it to yourself to do so...
I have - dedicated computer for music playback, going into a dCS upsampler, going into a dCS DAC. Both dCS upsampler & DAC were clocked by a dCS Verona Master CLock. All interconnects were some very expensive WireWorld stuff. The total $ outlay on this whole setup made my knees weak - I could never afford anything like this for a long time! The sonics were easily beaten by my 1-box CDP.....There was no body or soul to playback MUSIC but the SOUND was stellar.
04-20-12: Bombaywalla
Al, Kijanki: I *think* that I might know what the author is intending to say here: To do an A/B comparison, the author would like to level the playing field.... Does this make sense guys?

Hi Bombaywalla,

I think that what he is referring to in note 21 and in the "Clipping" paragraph is a comparison between a 192 kHz hi rez signal, and that same signal downsampled to 44.1 or 48 kHz and then upsampled back to 192 kHz. Both 192 kHz signals would be played back through the same DAC and the same downstream components. If they were to sound different in any way it would presumably mean that the lower sample rate, and/or the downsampling and upsampling processes, degraded the signal. Which signal sounds subjectively better would be irrelevant.

As I indicated earlier, though, it seems to me that the flaw in that methodology is that it does not take into account the sonic effects of the anti-alias and reconstruction filters that would be used if the recording and playback processes were done at the lower sample rate.

Best regards,
-- Al
I found this nice article called "A BeginnerÂ’s Guide to High Resolution Downloads of Music". here is the link:
http://audaud.com/2012/03/a-beginners-guide-to-high-resolution-downloads-of-music/

In this article is a para called "How hi-res should you go?" towards the bottom (scroll almost to the end). That para cut & paste:
'Unless you have extremely youthful hearing ability plus the highest-end speakers and audio gear, many of us feel that the improvement of 192K over 96K is inaudible. The word length expansion from 16 bits to 24 bits makes a much greater enhancement in the sound. 24/96 or 24/88.2 is fine for nearly everything. Also, remember that 192K and 176.4K files take up much more memory on hard drives, for little audible improvement.'

Looks like many people think alike: there's a case for 88.2K or 96K sampling but not beyond......
As for filters - look at typical response of 2and 8 pole 20kHz Bessel filter in dB:

2pole 8pole
20kHz -3 -3
22kHz -3.63 -3.67
40kHz -9.82 -13.68
80kHz -20.32 -51.81

Kijanki, you provided us w/ the freq resp of ANALOG Bessel filters. I agree with you & I did write this in my prev post - analog filters cannot creat a sharp cutoff like what the author has shown - large attenuation between 20K-24K.
But, how about digital FIR filters? Can they create such a sharp roll-off?
Yeah, sure they can! Did you bother to read any of the links I referenced in my post? The paper from Dan Lavry shows 1 example & then there is that AES paper by Julian Dunn that also shows 4 filters that have 100dB atten & only a modest # of taps. All FIRs have flat group delay in-band.
if you're not sold on hi-res.....fine. just don't tell me what i hear isn't for real and i'm "wrong".
Levy03
if you are sold on hi-rez, fine. I don't know what you are listening to - true hi-rez (which would mean the analog masters sampled at 96KHz or 192KHz & made available for purchase) OR bogus hi-rez (whcih would mean taking the 16-b CD data, resampling it at 96K or 192K & providing it for sale to the unsuspecting public).
There have been sooooooo many scams re. hi-rez (recently read something about HDTracks were the offenders. Here is the link:
http://www.computeraudiophile.com/content/Metallica-Black-Album-HDTracks-Download)
that it's really very difficult to tell what the manuf has provided for sale.
You are probably listening to some digital filter thinking it's hi-rez & you are in 7th Heaven.
Maybe we should let you be - ignorance would be a bliss for you.....
04-20-12: Kijanki
Al, On one hand he has whole chapter titled "192kHz considered harmful" describing harm that 192kHz can do to amplifiers and speakers to later say this about oversampling:......
Kijanki, all that the author is saying is that when sampling at higher freq like 96KHz or 192KHz, you get intermodulation products that fold down into the 20-20K audio band due to typical preamp, power amp bandwidth limitations of not being able to reproduce higher freq products distortion-free i.e. due to the non-linearities of the electronics. And, systems having smaller bandwidths have the situation worse in that the probability that they'll amplify the high freq signals is much higher. So, the point is that if you do not sample at 96K or 192K you won't have these higher freq intermod products, they won't fold down into 20-20K & your preamp/power amp will not amplify them due to its non-linearity.

it's clear to see that if a 96K or 192K sampled signal is downsampled to 48KHz then the anti-aliasing filter will cut off all these high freq intermod products. So, according to the author, since this signal is free of any ultrasonic content, it's safer to playback with the idea that distortion products due to ultrasonics are not being played back.

I do not think that it's unreasonable to say that ultrasonics created due to higher freq sampling can create in-band intermod products that can be amplified by the non-linearities of the playback electronics & that they are harmful to the playback listening pleasure.

I don't think that the author should have labeled the paragraph as "192KHz considered harmful". People like Kijanki have read this literally thinking that the very act of sampling at 192KHz is harmful. No, I don't think that the very act is harmful; it's those ultrasonics folded down & amplified that are harmful.....

Suppose we want to compare the fidelity of 48kHz sampling to a 192kHz source sample. A typical way is to downsample from 192kHz to 48kHz, upsample it back to 192kHz, and then compare it to the original 192kHz sample in an ABX test [21].
Al, Kijanki: I *think* that I might know what the author is intending to say here: To do an A/B comparison, the author would like to level the playing field. Thus, he does not want to use the original 192KHz signal as-is. What he wants to do is downsample on-the-fly the 192KHz signal to 48KHz & create signal A. Then, upsample this 48KHz signal on-the-fly back upto 192KHz & create signal B. Now, the playing field is level because the same machine downsampled & upsampled the signal & the same filters have screwed up the A & B signals. The signal X is the original 192KHz signal. If you were to use the original 192KHz which was created on some different machine against the 48KHz created on your CDP, you would have the effect of 2 different digital filters & you could not do a true A/B comparison. Does this make sense guys?
Ok, I see what you are referring to. Note this statement:
Oversampling is simple and clever. You may recall from my A Digital Media Primer for Geeks that high sampling rates provide a great deal more space between the highest frequency audio we care about (20kHz) and the Nyquist frequency (half the sampling rate). This allows for simpler, smoother, more reliable analog anti-aliasing filters, and thus higher fidelity. This extra space between is [sic] 20kHz and the Nyquist frequency is essentially just spectral padding for the analog filter.

Because digital filters have few of the practical limitations of an analog filter, we can complete the anti-aliasing process with greater efficiency and precision digitally. The very high rate raw digital signal passes through a digital anti-aliasing filter, which has no trouble fitting a transition band into a tight space. After this further digital anti-aliasing, the extra padding samples are simply thrown away. Oversampled playback approximately works in reverse.
So what distinguishes the two situations you are referring to is that the 192 kHz hi rez format will presumably include a significant amount of ultrasonic audio information, which is what he is saying might have harmful consequences as a result of intermodulation effects in downstream components, while the oversampled redbook data will not include that information, and therefore those effects will not occur.

Best regards,
-- Al
Al, On one hand he has whole chapter titled "192kHz considered harmful" describing harm that 192kHz can do to amplifiers and speakers to later say this about oversampling:

"This means we can use low rate 44.1kHz or 48kHz audio with all the fidelity benefits of 192kHz or higher sampling (smooth frequency response, low aliasing) and none of the drawbacks (ultrasonics that cause intermodulation distortion, wasted space)."

According to above when DAC is exposed to 192kHz sample rate from CD it is harmful to all analog circuitry afterwards (including power amp and speaker) but when DAC upsamples redbook CD (24/192 downmixed to 16/44), the same 24/192 becomes benign. In either case DAC outputs samples at 192kHz rate but in one case it is less harmful to analog circuitry?
Hi Kijanki,

The only reference to downsampling + upsampling that I recall seeing was in the paragraph headed "clipping" in the lower third of the page, and in footnote 21. He was saying that by taking 192 kHz source material, downsampling it, and then upsampling back to 192 kHz, a sonic comparison could be made between the two 192 kHz signals that would be indicative of the adequacy of the lower sample rate.

Not sure if that is what you are referring to. But in any event the methodology he is describing doesn't make sense to me, because the comparison would not reflect the effects of the sharper anti-aliasing and reconstruction filters that would be required for recording and playback at the lower sample rate.

Best regards,
-- Al
Al, What sounds inconceivable to me is that 24/192 recording supposed to gain sound quality by downsampling it to 16/44 to be upsampled again, perhaps to the same 24/192. Am I reading it right? Is downsampling + upsampling somehow improving sound by replacing real samples with artificial interpolated samples and recreating same harmful 192kHz?
On the question of continuous vs. non-continuous waveforms, I think that part of the reason for the disagreement is that the word "continuous" is misleading in this context. No waveform is truly "continuous." Regardless of the nature of the waveform, the Sampling Theorem will only be perfectly accurate (i.e., to 100.00000000...%) when an infinitely long sample record is available, covering the period from the beginning of the universe to the end of time. :-)

Any real-world waveform, whether sinusoidal or not, and "continuous" or not, will not meet that criterion. As a result there will always be some non-zero loss of information, at and near the times when the waveform begins, when it ends, and when it changes character. In theory the spectral content of those transitions extends out to infinity Hertz, although as a practical matter much of the high frequency spectral content of those transitions will be at amplitudes that are utterly negligible.

The information that is lost in those transitions will correspond to the spectral components that lie above the cutoff point of the anti-aliasing filter. The lower the cutoff point of the anti-aliasing filter, and the more abrupt the transitions are in the waveform that is being sampled, the greater the amount of information that will be lost.

Will any of that particular form of information loss be audibly significant when a music waveform is sampled at 44.1 kHz? It's hard to say, and I doubt that empirical assessment (by listening) can yield a meaningful answer considering how many other variables and unknowns are involved in the recording and playback processes. My guess is that it probably has some significance, especially on high frequency transients such as cymbal crashes, but only to a relatively small degree.

Is oversampling plus noise shaping an essentially perfect means of overcoming the problems inherent in sampling just slightly above the Nyquist rate, as the article seems to suggest? It's probably fair to say that it can work pretty well, but IMO it would be hard to argue that it is "essentially perfect." Can the ultrasonic frequency content that is retained by hi rez formats have adverse consequences, as claimed in the article, as a result of intermodulation effects within the system's electronics, or things like crosstalk effects for that matter? It certainly seems conceivable, to a greater or lesser extent depending on the particular components that are in the system. Will sampling at a higher rate result in sampling that is less accurate, assuming equal cost and comparable design quality? That would seem to be a reasonable expectation. But complex and sophisticated digital signal processing does not come for free either.

What does it all add up to? I would have to say that the paper referenced by the OP, and also the Lavry paper, make better cases against hi rez than I would have anticipated, but they are certainly not conclusive as I see it. And given the many tradeoffs and dependencies that are involved, my suspicion is that there will ultimately be no one answer that is inarguably correct.

Best regards,
-- Al
All this talk about tests and links to this study and all this technical stuff....

REALITY: All that means nothing when you have people that like and dislike things based on their own opinion of what "sounds good". Many feel tubes sound better than SS.... HOW CAN YOU TEST THAT? YOU CANNOT. The listener has their opinion of what sounds good. So posting links to these different things is POINTLESS.
What a great thread! Civil discussion. Ruffled feathers, for sure. But, civil and respectful. Keep going guys! I am still on the fence about hi-rez and the OP really brought up a great article and it sparked a great discussion. It will ultimately come down to objectivists vs subjectivists, but who cares? Those who dismiss A/B/X testing believe that there is more to listening than can be quantified or measured, and those who believe in A/B/X think the rest of us are fools if we don't follow the science. Thanks to all for an entertaining discussion!
Trust your ears and decide. What I have in 24/96 or 24/192 blows away the 16/44.1 versions most of the time. It's all about the mastering.
The article seems right, as far as the author goes. He admits to the problem of brick wall filters on Redbook CD, but he forgets to mention timing errors. This is why a cirrectly clocked computer regenerated waveform seems to improve on the CD, thru the same DAC. I would propose a much reduced timing error as part of the improvement of the higher res sampling frequenies. There is also a lot of talk on WAV beng better than FLAC (even though "bits are bits").
Stretching, I would hypothesize that the regeneration of the waveform by the additional complication of the FLAC decompression "bothers" our sensibilities in some way. IF so, then I would also propose that the decompression of MLP on a DVD-A might be similar.
My hypothetical ranking of sound which seems to agree with what I hear is:
CD<=FLAC<=WAV (for 16/44.1)<=24/96MLP<=24/96PCM<=24/192MLP (for DVD/A <= DSD, and just for fun <= LP. At some points the minor improvements may not be worth the additional storage requirements.
Remember, just speculating!
Tubes better than SS, vinyl better than CD, Redbook better than HiRez with all the theory to prove it.

The pattern persists.

And MP3 will kill us all.
"nonsense! The Nyquist criteria applies to any signal that needs to be quantized. The Nyquist criteria only gives the minimum requirement; it does not say that one is forced to have only 2 samples per highest frequency."

Yes, you can have more samples (for instance 192kHz) but he claims that 44.1kHz (two samples) is all you need.

Again, Nyquist applies to continuous waves ONLY.

from Wikipedia:
"The theorem assumes an idealization of any real-world situation, as it only applies to signals that are sampled for infinite time; any time-limited x(t) cannot be perfectly bandlimited."

Perfect reconstruction of continuous signals close to Nyquist frequency (for instance 15-20kHz) is possible but when signals become very short, reconstruction is much less than perfect.

As for filters - look at typical response of 2and 8 pole 20kHz Bessel filter in dB:

2pole 8pole
20kHz -3 -3
22kHz -3.63 -3.67
40kHz -9.82 -13.68
80kHz -20.32 -51.81

As you can see there is very little attenuation difference at 44.1kHz/2=22kHz with 4x higher number of poles. You would perhaps need hundreds of poles and still not get -96dB. Dramatic difference shows at higher frequencies beyond the "knee" of the filter (160dB vs 40dB per decade). Whole purpose of converting analog to digital at higher rate is to represent bandwidth of 20kHz more accurately and not to extend bandwidth. Downsampling 24/192 master tapes to 16/44 removes some information, (audible or not) but to claim that 24/192 is inferior to 16/44 is complete nonsense.

As for dynamic range, again the point is resolution of the signals above noise floor. According to this article if I listen at 85dB peak and have 35dB ambient noise at home I should not be able to tell the difference between 16 and 8 bit recording (corresponding to about 50dB range). That's nonsense as well.

What about 192kHz being harmful? It doesn't get more silly than that.
OK guys.... here is a link to this study, no wait.... here is a link to this paper, no, i mean, her a link to this test a dude did a while back that says you are wrong... etc. etc.....

SO WHAT!!!!!!!! How is all this arguing about studies and technical data and technical specs going to tell you how ANOTHER PERSON will interpret how a certain CD player playing a std. redbook CD sounds vs. a DAC playing a digital file !!!!!!!!!!!
If you have not heard a properly setup DAC with hi-res files, you owe it to yourself to do so... before you spend any money on a CD player... you will be happy that you did. Not to mention the benefit seeing your entire music collection on an iPad and being able to make playlists, etc. Why would anyone want to search through a bunch of CDs, put it into a device that could only add noise and jitter and eventually will fail (drive, optics), and hear a few songs that you may like by one artist on that CD. Instead, if you like Chris Issak for example, have every song he every recorded in AT LEAST the same quality, and most times better quality sound than the std. CD, all at your finger tips. How can you beat that?
04-20-12: Kijanki
............
Second nonsense is that digital filter can suppress 96dB within 4kHz without any problem. There is no filter in the world, digital or analog, that can do that (no matter how many poles) with even group delays (or linear phase if you prefer).
.........
not true! Here is a link to paper written by Dan Lavry of Lavry Engineering who wrote this paper in 1997 that shows a 500-tap FIR filter that has a passband of 15KHz & a transition band of 1KHz & stopband starting at 16KHz. The attenution achieved in the 1KHz transition is a whopping 100dB!! See page 3 of 7:
http://www.lavryengineering.com/white_papers/fir.pdf
yeah, it came at a price: 88.2 million operations per second using a dedicated DSP. Very high # of MOPS but do-able.
If one opens the transition band to 4KHz like the paper referenced by the OP then I'm sure that the # of taps will come down.
The paper also goes not to say that the group delay of the FIR filter is flat all the way out to 15KHz.

here is another digital filter paper (from the AES) that shows brickwall digital FIR filters:
http://www.nanophon.com/audio/antialia.pdf.
it's possible to have these brickwall FIR filters with reasonable DSP capacity.
i disagree with the article and the premiss behind it. nonsense sounds about right. i have nothing to offer in dispute of it except three years of listening to hi res (burned dvd's and streaming). the improved sound quality of hi res has been obvious to me in my set-up (given well recorded music from the start).

if you're not sold on hi-res.....fine. just don't tell me what i hear isn't for real and i'm "wrong".
Lots of technical info (on paper), but what really matters is how it sounds to each person. If you cannot hear that 24/192 WAV or FLAC sounds better than redbook CD, you may want to start selling off that expensive gear and get a boombox to listen to. My LINN Akurate DS playing 24/192 WAV or FLAC will sound better (everything else in the system being the same), than almost any CD player you put next to it.
Short high frequency bursts like cymbals will suffer the most of distortion.
Kijanki, this one is for you: here is a wonderful thread showing frequency spectrum of various brand of cymbals: http://www.drummerworld.com/forums/showthread.php?t=66957

The top quarter of the thread shows some really very good spectra of various cymbals. You can see that by 40KHz the spectrum has died down to 30-40dB SPL. The major part of a cymbal crash freq content is in the 20-20K range & the content falling off rapidly thereafter. I agree there is content beyong 20KHz but atleast 30dB by the time you hit 30KHz.
So, one could make the case for a 96KHz sampling rate wherein all the freq content upto 48KHz would be included. This sounds reasonable. At 48KHz the analog filter spec becomes reasonable too. Looks like it's a win-win situation....
04-20-12: Onhwy61
And what's wrong with mathematically perfect response out to 16kHz? Most people, and certainly most middle age and older audiophiles' hearing doesn't go beyond 16kHz.
while it might be true that older ears do not have the 20-20K respone, the music is prepared for everyone. Like the article says there is a 100 yrs worth data that shows that 20-20K is the human hearing limit. So, when preparing digital music might as well keep the audio spectrum to its max limits. Younger people certainly can hear this range & so can many other older folks.

FM doesn't go beyond 15kHz and at its best it's pretty damn good sounding.
FM has (air) spectrum bandwidth limitations that force it to curtail bandwidth. If they could help it, they would have also transmitted in the 20-20K range. Air spectrum is very expensive so this compromise seems reasonable.

FM doesn't go beyond 15kHz and at its best it's pretty damn good sounding.
Like I wrote in my prev post & I'll write it again: if you start off w/ 12-b you'll end up with 9-10 bits after the mixing & mastering processes. If you start off w/ 16-b, you'll probably end up w/ 12-13 bits. The section "The dynamic range of 16 bits" explains quite well the DR of 16 bits & also how it might be possible to encode fainter signals using 16-b.
Since a lot of data already shows that sounds at absolute levels of +120dB, +130dB permanently damage ears, my understanding is that it might not be worth encoding sounds on a disk that cover the enitre 140dB dynamic range of human hearing. It appears that covering 120dB of dynamic range is sufficient. If one uses 12-b only & one attempts to encode very faint sounds my understanding is that 72dB could be a limiting factor trying to cover the entire 120 DR. 16-b & 96dB is adequate & the article shows a plot of a -105dB signal at 1KHz using clever dithering techniques.

Nyquist criteria applies only to continuous waves.
nonsense! The Nyquist criteria applies to any signal that needs to be quantized. The Nyquist criteria only gives the minimum requirement; it does not say that one is forced to have only 2 samples per highest frequency.

There is no filter in the world, digital or analog, that can do that (no matter how many poles) with even group delays (or linear phase if you prefer).
yeah, I know what you mean for analog filters & I agree w/ you in that respect but for digital FIR filters (linear phase) I'm not sure I totally agree with you. My understanding is that if you had a, say, 64-tap FIR you could have a very steep skirt digital filter that would have flat group delay & group delay distortion. I would have find some evidence of this before I contend this issue w/ you but for right now I'm skeptical that it cannot be done. I'll leave it that....

I see the case for upto 24/96 as it seems to alleviate most of the pressing issues such as noise creeping into the music signal during mixing/mastering, analog filters having too steep a skirt at 44.1KHz. I'm not sure that I buy the case for 24/192, etc.

If anyone is interested in looking at some signals look at the Powerpoint presentation at reference #17 in the article. SLides 20, 21, 24-28 show spectrum of instruments & spectra of music from commercial CDs. Look at the freq where the content dies off even for SACDs.
I agree with Onhwy61 - article is a nonsense. First, motion that 16/44 is perfect if meets Nyquist criteria is first nonsense. Nyquist criteria applies only to continuous waves. Short high frequency bursts like cymbals will suffer the most of distortion. Second nonsense is that digital filter can suppress 96dB within 4kHz without any problem. There is no filter in the world, digital or analog, that can do that (no matter how many poles) with even group delays (or linear phase if you prefer). Uneven group delays will cause poor summing of harmonics (delayed differently) and change in sound. Reducing suppression won't help since low level signals above 24kHz will "fold" into audible band starting at 0Hz. Next nonsense is that ultrasonic frequency is harmful to the ear and modulate tweeter. Not only that 192kHz is WAY easier to completely filter out than 44kHz but also modulation can only happen on nonlinear element and for this to happen membrane has to move - not likely at 192kHz (even if your amp and CDP have such bandwidth). Then he claims that higher resolution does not increase dynamic range because of ambient noise floor forgetting that it is still improving resolution for louder signals. He claims that oversampling can increase resolution and sampling rate - true, but it is done with interpolated samples while 24/192 contains real samples. I agree that we might have hard time to hear better above certain resolution/rate, for instance 20/96 but claim that 24/192 is harmful is complete nonsense.