24/96 or 24/192

[Julf]

So many errors and suspiciously misleading claims and even scare mongering in that link.

 

Well, claiming that without any specific examples our counterproof doesn't really help.

 

Give me the audio in the format it was mastered in.

 

Well, even if you still had a proper reel-to-reel tape deck, and somehow managed to get the record company to make you a copy of the original master tape, that process of analog copying would result in much more loss and distortion than the resampling from, let's say, a 24-192 to a 24-96 digital format.

 

If its 24-192 then give me that.

 

Why would one more step of processing, after the hundreds of steps done by modern digital audio workstations used in contemporary mastering make any real difference?

 

You may not hear to 96 kHz but if you downsample any data you cannot avoid round off errors and artefacts IN THE AUDIBLE range.

 

You started by criticizing "errors and misleading claims". Can you substantiate your claim that downsampling from 192 to 96 kHz causes audible artefacts in the audible range?

[Julf]

And if you are developing your own audio codec and want to make a new hires version, but find yourself snookered by the big boys patents without room to move, maybe sour grapes will push you to putting up an anti hires propaganda web page to try and convince people that your 16-44 codec is good enough. A few facts covering up some fundamental lies will do it.

 

Woah! Not sure who has the sour grapes here. Could you please present your claimed "facts"?

[Jud]

Well, even if you still had a proper reel-to-reel tape deck, and somehow managed to get the record company to make you a copy of the original master tape, that process of analog copying would result in much more loss and distortion than the resampling from, let's say, a 24-192 to a 24-96 digital format.

 

I would say that's quite possible. But, at least for me personally, there is a great deal to learn about the audible effects (or lack thereof) of the compromises inherent in sample rate conversion.

[bdiament]

Hi Jud,

 

I would say that's quite possible. But, at least for me personally, there is a great deal to learn about the audible effects (or lack thereof) of the compromises inherent in sample rate conversion.

 

My experience has been that it depends very much on the sample rate conversion algorithm (and how it is used).

Having beta tested several, allowing me the opportunity to test and hear them in detail, I find most tend to harden and brighten instrumental timbres. (Personally, I think this is MUCH more damaging than making say, a 15 ips copy of an analog master tape... and even mores than making a 30 ips copy.)

 

The best algorithms I've heard, however, don't suffer from the artifacts I hear in the great majority of the others and create results that sound, to my ears, very much like the unconverted original.

 

I would add that even with my favorite algorithm, Alexey Lukin's wonderful 64-bit SRC marketed by isotope (easily the most transparent in my experience), the magic isn't there when used in real-time, i.e., when the music is playing. (I'm no fan of on-the-fly sample rate conversion.) It is when the algorithm is used off-line (i.e. NOT while the music is playing) that it performs its magic.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[bdiament]

Hi audiophile65,

 

...given the fact that the difference between 96 and 192 are very subtle...

 

I would say, if the difference between the 96 and the 192 is "very subtle" either the recording or the playback converters (and/or perhaps the rest of the playback setup) are not revealing all there is to hear.

 

I've said elsewhere that in my experience, a great many converters with "192" on their spec sheet actually sound *worse* to my ears, at this rate than they do at lower rates like 96k. This, I attribute to the significantly increased demands on accurate clocking that come with the higher rates and the increased demands on analog stage performance at the wider bandwidths.

 

With the best converters running at 192, for the first time in my experience, I have not been able to discern the output from the direct input signal. At recording sessions, the playback and the direct mic feed are, to my ears, indistinguishable. This is something I've never experienced from any recording device, regardless of format or price, in my decades as an audio engineer and audio enthusiast.

 

The same converters, which provide the most transparent 96k I've ever heard, are, at that rate, reduced to sounding like "great digital". To me, that is a BIG step down as I find the 192k performance to be a bigger step up from 96k than the latter is from ordinary CD 44.1k. It is that threshold that is crossed, to where the result is no longer "great digital" (or "great analog") but is, at long last, truly transparent to the input signal.

 

Subtle? Like an earthquake! ;-}

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[Jud]

I would add that even with my favorite algorithm, Alexey Lukin's wonderful 64-bit SRC marketed by isotope (easily the most transparent in my experience), the magic isn't there when used in real-time, i.e., when the music is playing. (I'm no fan of on-the-fly sample rate conversion.) It is when the algorithm is used off-line (i.e. NOT while the music is playing) that it performs its magic.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

 

Except we tend to forget (admittedly there is way too much to remember) that very nearly all DACs do on-the-fly sample rate conversion. Those that don't either have it done for them in the computer or convert to analog whatever resolution they are fed. Since there is far more material at RedBook resolutions than at 24/192 or higher, most such conversions are subject to well known artifacts. I also wonder what the meaning of "on the fly" really is when the results of such a sample rate conversion can be held in a buffer for some merest tick of time that is an eternity to a CPU performing a few billion operations per second.

 

Nevertheless I have some ideas about why it might make perfect sense for you to like "offline" conversion but dislike it "on the fly." (I'm sure you're relieved.) Would you be willing to describe for me what you listened to or compared in each of these two situations?

[bdiament]

Hi Jud,

 

Except we tend to forget (admittedly there is way too much to remember) that very nearly all DACs do on-the-fly sample rate conversion. Those that don't either have it done for them in the computer or convert to analog whatever resolution they are fed. Since there is far more material at RedBook resolutions than at 24/192 or higher, most such conversions are subject to well known artifacts. I also wonder what the meaning of "on the fly" really is when the results of such a sample rate conversion can be held in a buffer for some merest tick of time that is an eternity to a CPU performing a few billion operations per second.

 

Nevertheless I have some ideas about why it might make perfect sense for you to like "offline" conversion but dislike it "on the fly." (I'm sure you're relieved.) Would you be willing to describe for me what you listened to or compared in each of these two situations?

 

I'm not so sure oversampling filters in DACs are doing exactly the same thing. Perhaps they are. If so, the effects don't seem to be the same, which is why I suspect it is not exactly the same. I think of the DAC as looking at each sample multiple times, as opposed to changing the sample rate of the file (increasing the number of samples) -- two very different processes. Again, perhaps my understanding of this is not what it will be. (I would always hope to know more tomorrow than I did yesterday. ;-})

 

A particular DAC comes to mind at the moment, which *does* convert the sample rate of everything that passes through it. It has gotten great reviews in the audio press. To me, it is an ear ripper. It has "that sound" - the brightening and hardening common (to varying degrees) to all the not-so-good SRC algorithms and to what I've come to think of as the sound of on-the-fly SRC.

 

To answer your question about what I've listened to/compared, the sampling (no pun intended) has been very wide and over the course of several years. I've used things in my collection (all sorts of music and recordings), things I've mastered for other folks (again, all sorts of music and recordings) and most telling to me, my own recordings. Converting a file off-line and comparing the result to converting the original again while listening don't sound the same. The off-line conversion sounds, to my ears, more like the unconverted original; the on-line conversion starts to have hints of the hardening and brightening I hear with lesser SRC algorithms (even when they are used off-line).

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[eternaloptimist]

So many errors and suspiciously misleading claims and even scare mongering in that link.

 

Give me the audio in the format it was mastered in. If that's 16-44 then that's what I want.

If its 24-192 then give me that. You may not hear to 96 kHz but if you downsample any data you cannot avoid round off errors and artefacts IN THE AUDIBLE range. And every different down convert and dither I've tried adds its own flavor. An no single setting provides the best results for all material.

 

So just avoid downsampling altogether and just give me the audio in the depth and sample rate it was mastered in.

 

Fair comment. I plan to do some testing with 96 vs 192 kHz downloads of identical albums to see if I can hear the difference - even if theoretical, if I can't hear it, doesn't matter to me.

[spdif-usb]

I plan to do some testing with 96 vs 192 kHz downloads of identical albums to see if I can hear the difference - even if theoretical, if I can't hear it, doesn't matter to me.

The question is not IF people can hear the difference, but WHEN they can. With DAC prices coming down nearly as fast as computer hardware prices, it won't take very long until practically EVERYONE will start to hear the difference (it happened with CD and 24/96, it will happen again with 24/96 and 24/192). Stop reading FUD articles like the "192 kHz are very silly indeed". All that's written in them has been mathematically proven wrong (for crying out loud, the author of that article knows less about the Nyquist-Shannon theorem than the goldfish mentioned in my signature knows about quantum physics).

[Julf]

The question is not IF people can hear the difference, but WHEN they can.

 

You are entitled to your opinion. Others might disagree.

 

With DAC prices coming down nearly as fast as computer hardware prices, it won't take very long until practically EVERYONE will start to hear the difference

 

Ah, it was a question of price, not ears - I see!

 

Stop reading FUD articles like the "192 kHz are very silly indeed". All that's written in them has been mathematically proven wrong (for crying out loud, the author of that article knows less about the Nyquist-Shannon theorem than the goldfish mentioned in my signature knows about quantum physics).

 

Would you mind pointing out which actual points have been proven wrong, and would you happen to have pointers to the proof?

[Aerocraft67]

Woah! Not sure who has the sour grapes here. Could you please present your claimed "facts"?

 

The Monty paper is obviously not the final word on 24/192, even though it purports to be. A paper needn't be perfect to be credible, but nor does criticism of said paper. I think the Monty paper makes a contribution, but has shortcomings with a motive sufficiently questionable to warrant some criticism, if not outright condemnation.

[Blu]

I think it was Barry Diament from Soudkeeper Recordings who found that when making digital copies from the mastertape that he could not distinguish between the sound of the master tape and the digital copy until the digital copy got to 192/24.

 

Barry please forgive me if I misquoted, if it was not you, it was someone else in the recording industry.

 

Cheers from Australia.

 

Sorry Barry on re-reading my comment it does read the opposite of what I was intending.

 

I was intending to say that it wasn't until the copy got to 24/192 that it was indistinguishable to the master tape. But as you pointed out, you were talking about comparison of 24/192 recording to the microphone feed and not master tape, which is an even better judgement on hires digital recording.

 

My apologies again, I just have to listen to the, you are in the recording venue sound, I get on my system from the 24/192 WAV's of your 'Equinox' recording, for verification.

[bdiament]

Hi Blu,

 

Sorry Barry on re-reading my comment it does read the opposite of what I was intending.

 

I was intending to say that it wasn't until the copy got to 24/192 that it was indistinguishable to the master tape. But as you pointed out, you were talking about comparison of 24/192 recording to the microphone feed and not master tape, which is an even better judgement on hires digital recording.

 

My apologies again, I just have to listen to the, you are in the recording venue sound, I get on my system from the 24/192 WAV's of your 'Equinox' recording, for verification.

 

I had a feeling I understood your intention but that the wording somehow didn't match. ;-}

Thank you.

 

Best regards,

Barry

Soundkeeper Recordings

Barry Diament Audio

[Julf]

The Monty paper is obviously not the final word on 24/192, even though it purports to be.

 

Not sure it purports to be that. It is an opinion piece, backed by facts, from someone with a heck of a lot of experience.

 

A paper needn't be perfect to be credible, but nor does criticism of said paper. I think the Monty paper makes a contribution, but has shortcomings with a motive sufficiently questionable to warrant some criticism, if not outright condemnation.

 

Warrant some criticism? Possibly, but it would be better if backed by actual factual arguments instead of vague insinuations of ""errors and misleading claims" and "sour grapes". Condemnation? That would require even more factual support, and so far I have not seen you provide any.

[Paul R]

Now you are making sense! Listen for yourself and experience trumps theory every single time, regardless of what the results are.

This is supposed to be fun, not agony.

 

-Paul

 

 

Fair comment. I plan to do some testing with 96 vs 192 kHz downloads of identical albums to see if I can hear the difference - even if theoretical, if I can't hear it, doesn't matter to me.

[spdif-usb]

Would you mind pointing out which actual points have been proven wrong, and would you happen to have pointers to the proof?

All of the points. I don't need to provide pointers to the proof, for the exact same reason the author of the article has refuted to provide any. For starters, rather than confirming it, he contradicts what's written in the Wikipedia article on the Sampling Theorem that he linked. He claims stuff about digital filters having "no trouble", but without backing up his claim in any way. So, if you want pointers, go and ask him instead. Take my word for it, he won't be able to give you any. That's because it doesn't take very long for anyone interested enough to find out those filters all do have ringing artifacts in them. Bob Stuart's "Coding High Quality Digital Audio", which he linked, completely contradicts him also, and so do AES papers written by dCS founder Mike Storey during the early 90s. Oh and, don't even get me started about those Meyer & Moran tests. I can hear the difference between 16/44.1 and proper Hi Res from the next room. Too bad if you can't.

[Julf]

I can hear the difference between 16/44.1 and proper Hi Res from the next room. Too bad if you can't.

 

And that's of course what it all boils down to in the end. We can keep arguing in circles until the cows come home to hear the fat lady sing, but the whole issue is so easy to resolve - all it takes is some proper double-blind ABX (or ABC/HR) listening.

[Aerocraft67]

Julf, you have mistaken me for someone that has a stake in the outcome of this debate beyond simply increasing my knowledge of computer audio and making personal decisions about buying high resolution music. So I will not be providing any facts to support any particular position. But I will elaborate on my observations and address the issues you raise here.

 

Not sure it purports to be that. It is an opinion piece, backed by facts, from someone with a heck of a lot of experience.

 

A passage from the article, "OK, so 192kHz music files make no sense. Covered, done," could quite reasonably be interpreted as purporting a final word. Actually, you need read no further than the subtitle, "and why they make no sense," to glean an authoritative tone. O.K., maybe that counts as evidence, and I've already reneged on my promise not to provide you any.

 

As for experience, I'm comfortable asserting that there are many contributors to this debate, including on this thread, that have equal or superior experience to Monty. I do not purport to be one of them. But I will be so bold as to offer some results of my critical reading of the thread.

 

Warrant some criticism? Possibly, but it would be better if backed by actual factual arguments instead of vague insinuations of ""errors and misleading claims" and "sour grapes". Condemnation? That would require even more factual support, and so far I have not seen you provide any.

 

It's certainly fair to point out vague insinuations for what they are. But what I observe in your initial post about facts and evidence that I commented on, and your subsequent mandate for that here, is something I see often on discussion forums. A counterpoint or criticism is dismissed for lack of facts and evidence. This tactic is very effective, because, who doesn't respect facts and evidence? You can't argue against facts and evidence!

 

But I'll make another assertion, that not every contribution to a discussion forum requires a litany of facts and evidence to be credible and productive, particularly an incremental one to a well established thread. Lacking a litany of facts and evidence does not render a post a counterproductive violation of the forum's terms of service. Strident calls for facts and evidence in response to opposing points of view can have the effect of squelching debate rather than strengthening it. And it just gets tedious after a while.

 

To fully disclose my bias here, I'm taking out a frustration I have with otherwise reasoned and intelligent discussion on the internet with you here. I certainly read HumanMedia's commentary as critical as anybody's, and I don't accept it as gospel any more than I do Monty's. But I thought it was an interesting point, and bristled a bit at the "facts & evidence" routine with which you responded. Beyond venting that frustration, which is dubiously productive to this thread and forum for sure, I guess what I'm trying to do is move the debate forward past that. Probably not very successfully by picking a fight with you about it. For what it's worth, that's truly not my intention, and I offer my criticism most respectfully, even though I probably can't have it both ways.

 

Perhaps a slightly more productive venting of my frustration and disclosure of my stake in this discussion has to do with a lack of resolution to my previous post to this thread (which I'd phrase somewhat differently now that my system has evolved to include a 192 kHz-capable DAC, among other things). I'm genuinely torn between making the investment in 192 kHz downloads for the sake of once-and-for-all posterity, putting aside the ultimate question of whether there will ever be a practical difference in performance for my use, and putting that investment premium instead toward more music at lower sampling rates. For me, it's less about the ultimate difference between 192 kHz and 96 kHz, and more about whether that difference is worth the extra investment. Right now the pendulum is favoring 192 kHz, even though I'm under no illusions that I'll be basking in glorious sonic superiority over even 44 kHz, not to mention 96 kHz.

 

Part of the answer to my question is that it depends on the value I ascribe to the music. I'm still not above buying a one-off 256 kbps pop song from iTunes, or buying a used CD to rip an album into my collection. But for incremental investments in music that warrants once-and-for-all premium investment and treatment, it's a tougher to ascribe value, as the stakes are higher and the distinctions more difficult to parse out—as aptly evidenced by this thread.

[wgscott]

The other cool thing is HDtracks provides us often with a triple blind test of sorts. If you hear a difference between high-res and redbook, and then have the audacity to look at the spectra and find it is upsampled, then you can safely conclude the power of suggestion is at play.

[Aerocraft67]

And that's of course what it all boils down to in the end. We can keep arguing in circles until the cows come home to hear the fat lady sing, but the whole issue is so easy to resolve - all it takes is some proper double-blind ABX (or ABC/HR) listening.

 

And just to double down on my previous post before there's been time to respond, you can use "double-blind ABX" interchangeably with "facts & evidence" accordingly.

[Paul R]

Aw com'on Bill - DACs sound different when you enable or disable up sampling, why do you think up-sampled audio files would not? The why, I agree, is in question. But that they sound different? Much less questionable.

 

-Paul

 

 

 

The other cool thing is HDtracks provides us often with a triple blind test of sorts. If you hear a difference between high-res and redbook, and then have the audacity to look at the spectra and find it is upsampled, then you can safely conclude the power of suggestion is at play.

[Julf]

But I'll make another assertion, that not every contribution to a discussion forum requires a litany of facts and evidence to be credible and productive, particularly an incremental one to a well established thread. Lacking a litany of facts and evidence does not render a post a counterproductive violation of the forum's terms of service. Strident calls for facts and evidence in response to opposing points of view can have the effect of squelching debate rather than strengthening it. And it just gets tedious after a while.

 

Point taken - but *not* presenting facts and evidence also easily leads down the path of endless (and tedious) "I disagree, just because I say so". While it sounds very clever and mature to say "we should just all take the middle road, and accept each other's views as equally valid as our own", it really falls down when one participant claims 2 + 2 = 5 (for suitably large values of 2), while another maintains 2 + 2 = 4. The solution is *not* to agree to 2 + 2 = 4.5. The only way to resolve a debate like that is by facts and proof.

 

Perhaps a slightly more productive venting of my frustration and disclosure of my stake in this discussion has to do with a lack of resolution to my previous post to this thread (which I'd phrase somewhat differently now that my system has evolved to include a 192 kHz-capable DAC, among other things). I'm genuinely torn between making the investment in 192 kHz downloads for the sake of once-and-for-all posterity, putting aside the ultimate question of whether there will ever be a practical difference in performance for my use, and putting that investment premium instead toward more music at lower sampling rates. For me, it's less about the ultimate difference between 192 kHz and 96 kHz, and more about whether that difference is worth the extra investment. Right now the pendulum is favoring 192 kHz, even though I'm under no illusions that I'll be basking in glorious sonic superiority over even 44 kHz, not to mention 96 kHz.

 

My simple answer (kind of anticipated by you) would have been "just do a proper double-blind ABX. If you can hear a difference, then go for the higher resolution". But I understand your question is more complex than that, based on the concern that even if you don't hear a difference now, maybe you might if you upgrade your system at some future point.

 

So yes, as you say, it comes down to the value you ascribe to music. If you feel it is worth the extra cost, go for 192, and compare the 192 k files with versions that you downsample to 96, as some DACs and streaming players do better at 96. So keep the 192k versions "just in case", and use the versions that sound better (if there is a difference) on your current gear. How's that for a non-committal?

[Aerocraft67]

While it sounds very clever and mature to say "we should just all take the middle road, and accept each other's views as equally valid as our own", it really falls down

 

Absolutely agree with you here and hope that my posts did not imply otherwise.

 

Thank you for taking a shot at my 192 evaluation (the question remains open to others, by the way!).

 

On a related matter, I am in the process of evaluating two DACs, and I need to choose soon, so this would certainly be an opportune time to engage in some proper testing. What holds me back from this are time and resource constraints and lack of dedication, never mind the remaining limits to the conclusions of even a scientifically robust approach. Maybe I just have more talent for critical reading than executing the scientific method, and pursue my hobby accordingly. So it will be a much more casual evaluation that most people do, overwhelmed by expectation bias and all the rest, ultimately offering little to satisfy anything objectively for the record, although that's not really the goal in choosing my personal DAC beyond sharing the experience with others. The next best thing is to understand these conceded limitations and moot conclusions accordingly. Do feel free to challenge any falsely absolute conclusions you hear me make, but hopefully I won't step out of the bounds of the evaluation I've done!

[spdif-usb]

And that's of course what it all boils down to in the end. We can keep arguing in circles until the cows come home to hear the fat lady sing, but the whole issue is so easy to resolve - all it takes is some proper double-blind ABX (or ABC/HR) listening.

Double-blind listening tests are fundamentally flawed because we have memory. If you listen to A first and B next, what you hear when listening to B will be biased by things you can remember about what A sounded like. Likewise, if you listen to B first and A next, you're going to be informed by B. And if you listen to both A and B (or both B and A) to practice before doing the actual test, you're going to be informed by the both of them, no matter if blinded, all simply because you can't erase memory (if you had the memory of a goldfish, maybe it would work).

 

At the AES show of 2009 held in New York City, psychoacoustics expert Poppy Crum, who holds a Ph.D in auditory neuroscience, presented undisputed objective evidence of the fact human hearing can very easily be tricked into expectation bias even while blinded. This is simply because we are biased by what we can remember about how human speech (for example) should sound like. The human speech used as the test sound in the experiment can be replaced with music (in fact, alot of music contains vocals, so the experiment is already totally relevant even without using music as the test sound).

 

Even the fact we can remember that we're doing a blinded listening test causes expectation bias during our listening test. Arguably, so can the fact we can remember that we're NOT doing a blinded listening test during our normal listening sessions, but that's my whole point to begin with, as most people buy audio gear for the purpose of normal musical enjoyment rather than for blinded listening tests, which leads to the logical conclusion NOT doing a blinded listening test comes closest to the actual music listening experience which we're so desperately trying to compare........

[Julf]

What holds me back from this are time and resource constraints and lack of dedication, never mind the remaining limits to the conclusions of even a scientifically robust approach. Maybe I just have more talent for critical reading than executing the scientific method, and pursue my hobby accordingly.

 

Indeed, and I guess that applies to most of us. It is the road I take most of the time too - but we both also realize the dangers of extrapolating that to statements about factual reality