High-Resolution: Does it matter? Part IV Finale
I read a recent post from a fellow member of an audiophile FB group looking for any reasonable explanation as to why high-resolution audio might be better than CD spec. As someone that advocates for 96 kHz/24 bit PCM as fully capable of capturing real world frequencies and dynamic range — high fidelity, I also recognize that CD specifications do actually meet or exceed the needs of “normal” music capture and reproduction. But we’re not dealing with “normal” audio when we enter the world of audiophiles. Right away Ethan Winer (author of The Audio Expert) chimed in with his strongly held position that 44.1 kHz/16-bit PCM CDs are all we need in the world of music reproduction. He added the link to his comparison challenge (downloadable files that purport to be high-res vs. standard-res) and asked people to take his test. I was somewhat surprised that he persists in linking to his “challenge page” as I showed in a couple of recent posts that his “hi-res music” examples lack the requisite qualities that make a high-resolution file a high-resolution file. Music files that measure as 96 kHz/24-bits don’t necessarily deliver 96/24 fidelity. In fact, I’ve shown over the years of posting on this site, that virtually ALL “so-called” hi-res music available from the well known high-resolution download sites are no better then the vinyl LPs we enjoyed 30-40 years ago. A standard-resolution, analog master transferred to a 96 kHz/24 bit PCM bit bucket remains at standard-res music track.
The misinformation continues on both sides of the issue. I would argue that no satisfactory test has been done comparing hi-res vs. standard-res because the content used in these studies failed in the same way Ethan’s failed. You certainly can’t use any major label commercial SACDs or DVD-Audio discs — they originate from older analog masters. There continue to be new efforts to resolve the issue. I know Bob Katz is working on a new approach.
Let’s say I got funding for a rigorous, double blind ABX study comparing real high-resolution audio files (like mine) to downconverted versions of the same original masters. Inquiring minds really want to know if humans can “perceive” any differences between recordings made using 96 kHz/24-bit PCM and the same recording made at CD spec. I want to know if it really matters. I would assemble a state-of-the-art system in an ultra quiet studio environment and bring a sufficient number of listeners to get a good sample — maybe 100 or 1000 listeners at all skill levels. The test subjects could take as long as they want to make up their minds. They would be allowed to switch back and forth between examples as much as they want and an unbiased computer would tally the results. Would that convince those interested in the outcome one way or the other? Maybe or maybe not. Both camps are dug in pretty deep. I do believe that a survey done in this way would put the issue to bed once and for all.
But even a rigorous study that unambiguously proves people can perceive a real high-resolution recording vs. a standard-resolution track WON’T MATTER! Why? Because there simply aren’t enough artists, engineers, producers, and record labels interested in making and releasing real high-resolution recordings. Who really cares? The music industry would like us all to think fidelity matters but in all honesty they don’t. The hardware designers can brag about new digital systems using 192 and 384 kHz sample rates and 32-bits but without content made to take advantage of ultra high-resolution hardware, they are just pushing bigger numbers for marketing purposes. Ethan Winer is right. High-resolution audio doesn’t matter in the real world of music recording and distribution. The music industry is satisfied with “normal” fidelity. And I’m right too. A real high-resolution audio track has more fidelity than a downconverted version of the same track — and it doesn’t matter whether people can perceive a difference or not. A 96 kHz/24-bit recording captures and preserves everything in the original acoustic environment where it was made. That matters to me.
As audiophiles, we need to get over it. The debate about high-resolution is pointless because it won’t change anything. The music industry is promoting high-resolution audio but they know that it’s meaningless. I’ve been in those discussions. When I was on the CEA High End Audio board, it was obvious that they were interested in “commerce” over quality. The music business wants to maximize profits not fidelity. The real focus needs to be convincing the music industry that elevating fidelity would reap additional financial rewards. I seriously doubt that’s going to happen.
Nice wrap up, thank you. I have read and agree with much of Ethan’s stuff and also have read and agree with much of yours. Ha, what a conundrum! I certainly appreciate the fidelity offered by your tracks (the BD I got with the book is great) but certainly will never give up the music of my youth or historical music done when the technology was not there to provide true high resolution. In the end, while I know the differences, I just am comfortable with the compromises, and simply enjoy the music no matter what medium or file type, or bitrate :-]. Happy listening!
Thanks Larry. That’s exactly the point. It is possible to produce recordings with amazing fidelity but the music that we cherish can’t be rerecorded, remixed, or remastered to high-resolution. We have to accept the best version of those masters. And going forward no one seems interested in high-resolution music…just louder.
Perhaps you could provide Ethan with a sample that meets your definition of 2-ch high-res sound, plus a 16/44 downsize of that same sample, and suggest he puts it up on his comparison challenge page.
Plus, regarding your ideal blind test, I don’t see much wrong with people using headphones at home. Headphones are proven to be more discerning of detectable difference than any room-speaker setup.
cheers
I offered my files. Scott Wilkinson and AVSforum did use a variety of my files several years ago for a similar non-rigorous challenges and the results were better than random.
Hi Grant
That would be totally useless and is the basis for why there is such contradiction in peoples opinions on this matter. People would be listening on a wide variety of playback equipment, it has to be done in a totally enclosed consistent environment with the same gear to have any value.
John, if home ABX testers have a decent pair of headphones or ear speakers that go 20-20k* with a reasonably balanced frequency response, then they should be able to detect an audible difference.
I agree that for a population study the listeners should all be subject to the same test environment (and unlike Marks’s ‘ideal’ test, they should not be allowed to ‘take as long as they want’, because that introduces a variable, as you noted. But if you or I want to know whether oneself can detect the difference, it is fine to do a home ABX test on decent headphones.
Note that I stopped at 20kHz. If somebody wants to claim humans can hear above 20k, then they should design a test of maximum detectable frequency, not the test being discussed here. These tests have of course been done by general science and the result is 20k, so I will go with that. I strongly suggest that everyone here accept that.
“I would argue that no satisfactory test has been done comparing hi-res vs. standard-res because the content used in these studies failed in the same way Ethan’s failed. You certainly can’t use any major label commercial SACDs or DVD-Audio discs — they originate from older analog masters.”
I disagree. There were quite a few commercially released classical hybrid SACDs released by Polygram that contain both a high-resolution (2.8224 Mb/second 1-bit DSD) layer and a Red Book CD (44.1 kHz 16-bit PCM) layer with a downconverted copy of the hi-res master.
While Philips was promoting the high-resolution quality of SACDs, they were fully aware that this aspect would not be its primary attribute to consumers, most of whom were perfectly happy with Red Book CDs. It was multichannel that would be the reason for consumers to upgrade.
Regrettably, both the format war between SACD and DVD-Audio, which caused many to hold off purchasing initially, and then the advent of the iPod, which soaked up most of the mass-market spending on audio, eliminated any chance of wide acceptance of high-resolution audio at that time.
Now, as you note, most of what is purported to be high-resolution originates from masters that are not high resolution. It’s a shame.
Disclaimer: I was a consultant to Philips on SACD, but my comments are my own and do not necessarily reflect their view.
Phillips and Sony did heavily promote SACDs at the dawn of the “high-resolution”, era in the early 2000s. But most of those SACDs were made from analog masters. I calculated the numbers and fully 85% of the “so-called” high-res releases weren’t.
Hi Mark,
How about a crowd funding strategy to get such a research project underway? – I’m up for it and anyone seriously interested or trying to promote HD should be to, particularly those with a financial interest, frankly would it not spark of hypocrisy for them not to do so? let them put their money where there mouth is! Regardless of the fact that as you say it won’t make any difference to the recording industry, if we care about music and the art of capturing it and playing it back we owe it to do this and answer this question once and for all.
A rigorous test should be done. I actually wrote a proposal to the CEA some years ago. The cost would have been about $50,000 but they weren’t interested. However, in the end it won’t matter to the music industry.
Mark, this is even more disingenuous than your previous posts. I know you’re usually on the right side of science, but on this issue you’re just wrong. As I’ve explained before, you don’t need a bunch of crowd-funding money to prove your case. You just need to do a proper level-matched blind test with a handful of people. If you can’t get 20 people, then do 20 trials. I’m not a statistics expert, but I do know this could be done with the members of a typical audiophile club. But it has to be done properly. Once you do that you’ll find that high quality audio is all about a flat response and low distortion, and has nothing to do with the presence of ultrasonics or needing a noise floor 30 dB softer than anyone could possibly hear. It’s also wrong that my test somehow “”failed” because the 24/96 files weren’t “high-def” enough. The only failures here are people’s ability to identify HD versus CD when tested properly. So until you do a proper test yourself, you really should stop bashing other people’s attempts. At least we’re trying to find and report the truth.
As for your earlier claim that Meyer & Moran “used only standard-resolution sources” that is simply untrue. A list of the tracks they used is at the bottom of this page:
http://bostonaudiosociety.org/explanation.htm
And this page (I hope the link works) contains a reprint of M&M’s related letter in the AES Journal:
https://onedrive.live.com/?authkey=%21ABJzaXp0NzGhQV4&cid=E080D27D7BCE4A94&id=E080D27D7BCE4A94%21915&parId=E080D27D7BCE4A94%21102&o=OneUp
I stand by my post. The information may run counter to your own assessment and opinion but in no way is it disingenuous. In fact, I agree with you that virtually all commercial music releases would not benefit from a transfer to high-resolution specifications. You insist that your files (one with barely any ultrasonic content and another with 4 dB of dynamic range and “phantom” ultrasonics) qualify as “high-resolution”. I disagree. If your participants (or those that participated in the Meyer and Moran survey) had actually been presented with bona fide high-resolution recordings, perhaps the results would have been different. I don’t know. They were in the AVS Forum test that mirrored your own challenge. Ultimately, the truth may be determined by a proper test. As I concluded in my piece, in the end it doesn’t really matter. The industry isn’t interested in whether music enthusiasts can hear — and would prefer and purchase — real high-resolution content. The CEA group that I was part was involved in the crafting the message of “hi-res music” and they didn’t care about any improvements — real or imagined. Any recording — including an Edison cylinder — transferred to a 48 kHz/20 bit PCM digital file would qualify as “hi-res music” according to their published definition. Hi-res music was — and remains — a hoax to increase sales as practiced by the major labels.
But that doesn’t mean that it’s not worth doing for those few audiophile labels that want to advance the art of recording. Why not use 96 kHz/24-bits instead of CD specs? It doesn’t cost any more and does allow for additional dynamic range and extended frequency response — perceptible or not! My interpretation of “fidelity” means that everything in the acoustic environment of the performance should be captured and reproduced. There are ultrasonics coming from the instruments, therefore they should be reproduced IMHO.
Perhaps we should collaborate on another version of your survey with files that I provide as has been suggested by some readers. I would be happy to be involved. It certainly would not provide a definitive answer but it might open a few doors and recast the discussion.
The M&M results were — and remain — completely invalid. That hasn’t changed. Does anyone really believe that just because the “manufacturers” said the content was high-resolution that it actually was? That was the fatal flaw of their study. Those label, equipment manufacturers and organizations are the same people promoting the myth of high-res. Of course, they’re going to say they qualify. I looked at the list that M&M provided (and the link that you included in your comment) years ago and again just now. My original assessment remains the same. ANY SACD release, whether newly recorded or transferred from analog tape, should not be considered “hi-res music” because they ALL lack ultrasonic content. The required noise shaping algorithm (and LPF filtering on playback) required when using a DSD 64 soundfile produces large amounts of noise just above the audio band. I’m confident that you would agree with that fact statement (maybe newer DSD files at 128 or 512 avoid this problem). All but one of the discs used in the M&M test were SACDs making them unsuitable for the test. The only DVD-Audio album (the Steely Dan album) was a transfer of a third generation analog master — also not hi-res. So it’s no wonder that no one was able to perceive any differences.
I’m reporting what I know to be factually correct.
According to the article I linked, the M&M tests used 20 disks that all claim to be true HD audio. Think about this: If every single one of those 20 samples is not really adequate to be considered true HD, that proves the entire “HD Audio” industry is crooked and should be exposed.
As for your offer, I’ll be glad to do another test using source files you prepare, so we can tap into my 4,000 friends and followers on Facebook. But you need to devise a way to keep people from identifying the files by peeking inside rather than just listening. Email me to arrange this when you get a chance.
Ethan, thanks for the comment. Yes, I was part of the “HD Audio” industry as an insider on the committees, organizations, and panels and I can confirm that the entire thing is a fraud! I’ve been saying this for many years. The only motivation of the DEG, NARAS, CEA/CTA, MQA, and labels is to spin the myth of high-resolution audio and hi-res music as the next thing consumers should spend money on. I still have the files I prepared for the AVS Forum survey, which I’ll send along. The qualification that Scott Wilkinson and I set for that experiment was that everyone had to qualify that they indeed had a high-res capable system. A set of ear buds and an iPhone wouldn’t cut it. And I disagree that we have to modify the tracks to keep people from peeking. I trust people to do the right thing. This is not a rigorous test and the results will be no more valid than the AVS Forum test so I don’t think adding any “phantom ultrasonic” content is called for. I’ll be in touch. Thanks.
We might be able to add some data points to the argument. But in the end, it doesn’t really matter to me. As I’ve said, I’ll keep doing my work at 96 kHz/24-bits because it’s too easy, offers potentially more fidelity, and has no adverse risks.
Mark, I’ve always respected your opinions on this matter very much after I first came across you in that Youtube video called “High resolution audio demystified”. I very much liked that you said that nobody had been able to tell a difference between hi-rez and CD quality, but you just wanted to include everything from the original signal in the recording. I can respect such an opinion, rather than the people who scream that the difference is “so fucking obvious!” without having done any blind testing.
I usually enjoy Ethan Winer’s writings very much, and have been in touch with him several times (always a pleasure), but I think your proposal is fair enough: Provide 24/96 recordings with “sufficient” content above 20 kHz, and then do a blind test with these files vs. a downsampled version (and the volume levels should be matched, and the tracks should be perfectly aligned, which went wrong in the test with your Laurence Juber recording). Although I actually essentially take Ethan’s side in this matter, it shouldn’t matter to him if we replace his files with your files – the results will be the same if people really can or can’t hear a difference.
I would be happy to participate in a test if I can (but I live in Spain). In fact, I bought your release of Laurence Juber’s “Guitar Noir”, and I’ve done a blind test of the hi-rez files vs. a downsampled version. While doing it I thought “this seems pretty easy, actually”, but when I saw the result it was no better than flipping a coin, ha ha ha!
Anyway, I have passed and failed many blind tests. My proudest moment (for testing at least) was when I passed with 15 out of 16 correct for identifying a 0.2 dB difference in volume level. I’ve also identified very small EQ differences that I’ve done myself. I don’t consider myself a golden ear nor a super well trained listener at all, but apparently I’m not too shabby at these things. So, if I can help with the testing I would be very happy to. And as you’ve seen, I’ve admitted my biases upfront: I don’t think I would be able to hear a difference between hi-rez and the downsampled version, so I usually don’t bother with hi-rez, but I would be very happy to participate, and I’ve surprised myself several times in blind tests, thinking I was imagining things, only to find out that I’ve passed with 15 or 16 correct out of 16.
Thanks Anders. I have a great deal of respect for Ethan and have learned a lot from his book. We’re discussing using some of my files as you and others have suggested. My challenge to his examples was not meant to diminish his reputation or be a personal attack. I simply want any survey of audio resolution to use the very best and most dramatic audio examples. I believe my catalog contains many good examples in a variety of genres. I’ve listened to my original 96 kHz/24-bit samples and downconverted versions and cannot tell the difference in a quick A|B comparison. I doubt that anyone using a typical system can.
Yes, let’s do a test if possible. I have a very good system as well, but I also doubt that anybody, no matter their system, could hear a difference between hi-res and CD quality, but let’s all try it out with your recordings :-).
We also just have to remember that if many people participate, a statistic prediction would be that someone will get, say, 14 out of 16 correct, so they should be asked to do the test at least once more to weed out any false positives (and again, volume level and alignment should be perfect, and we should try to eliminate cheating).
I also admire that you’ve always kept a respectable tone in all your debates. As much as I like Ethan, he doesn’t always keep as calm as you, but perhaps he just been arguing with the powercord-loving placebophiles for too long (those discussions upset me too) :-).
On a sidenote: I know that both you and Ethan really like the Revel Saloon speakers, so if possible, I heartily recommend that both of you give the Golden Ear Triton One a listen. I just bought a pair. I also listened to the Revel ones, and I liked the Golden Ear ones better. They are among less than a handful of speakers that I’ve auditioned that really sounded “right” to me. On top of this they also measure exceptionally well.
Thanks Anders…I’m working on the materials for the proposed evaluation. I’ll try to get it out today or tomorrow. And thanks for the comments regarding keeping a positive attitude. I see no point in engaging in abusive conversations although at times I’ve been tempted. We’re talking about audio fidelity after all. I know Ethan and have a great deal of respect for his experience and knowledge but I disagree with his bottom line assessment of high-resolution audio. We’re going to try and push forward a little.
Dear Mark,
Just came across your site for the first time today and am greatly enjoying your detailed comparisons of various sources using spectral analysis. One thing drives me crazy, though – why do you compare different parts of the stereo tracks?! Why show a lo-res spectrum on the first half of the track and the high resolution on the last half? Why not just pick the most interesting and instructive track segment and use that same segment for both the lo- and hi-res comparisons?? It would be so much easier to see the differences. Readers could even overlay the two in photoshop and highlight differences to their heart’s content.
I actually never considered doing what you suggest. The similarities from section to section were sufficient to do it the way that I have. The purpose is to show the hard line at 22 kHz. Thanks for the suggestion. I get your point.