Why I think HD audio is irrelevant

High Definition (HD) audio.Those words sound great! You hear about HD and you immediately picture the huge improvement those two letters have brought to our TV sets. If you love music, you can’t help thinking about the promising wonders HD has to offer to our ears… If that’s the case, I have really bad news to tell you. The only thing HD audio guarantees is a tremendous waste of storage space in your hard disk drives and portable players. At worst, it could even degrade the quality of sound. For technical details, I refer you to this excellent article on the subject, but my purpose here is to show you a very simple experiment that proves what I say in the title.

Here is the idea: you take a digital audio signal, make changes to it, then “subtract” the original and modified waves, and listen to what is left. That, in fact, is quite easy to do with audio editing software.

The signal we start with is an HD audio file of the highest possible quality, sampled at 192 kHz and 24 bit (see note at the end). Next we convert it to 44.1 kHz and 16 bit, which downgrades the original to CD quality. (This is somehow similar to reducing the number of pixels of a digital image.) In the process, we shrink the original file size by a factor of 6, that is, we discard more than 80% of the original data!

Now comes the tricky part. We have to superimpose the two signals to see how similar they are (the alignment must be perfect!), but differences will be too faint to be appreciated unless we find a way for the signals to cancel out. This is accomplished by inverting one of the waves (that is, changing the sign) before we mix them. For two identical signals, this process would produce a perfect flat line (which means absolute silence).

One more thing. In order to visualize the changes we make to the original file, we use spectrograms, which are graphical representations of a sound wave that show its frequency content (the whole spectrum of pitches, from lowest to highest). Colors depict intensity, with red being the loudest and violet, the softest; black means absence of signal.

So, let’s do it!

Here is the spectrogram of the HD audio clip I used for the experiment:

And this is what it sounds like:

(This is an HD audio .wav file. Portable devices may not be able to play it.)

The same file, converted to CD quality looks like this:

The picture above shows the portion of data that has been removed in the downgrade process.

The converted CD version sounds like this:

Don’t strive to hear the differences, nobody’s ears are suited for the challenge…

… and you are just about to see why.

This is what you get when you mix the signals. The lower black band represents the region of the spectrum where the signals cancel out: 

What does the removed data sound like?

(This is an HD audio .wav file. Portable devices may not be able to play it.)

No, your sound card has not stopped working, nor is there any problem with the sound file. What you are hearing (not hearing, actually) is what you are missing when you “downgrade” from HD to CD quality.

Let us crank up the volume and listen to it one more time:

(This is an HD audio .wav file. Portable devices may not be able to play it.)

Again, this is what that 80% of data we cut through conversion amounts to. By now, I hope you realize that data loss and degraded sound quality are two very different things.

The result should not surprise us, if we remember that our ears perceive frequencies from 20Hz to 20 kHz. Removing the part of the spectrum that is above 22 kHz (44.1 kHz sample rate) does not degrade the audible signal.


As Mr. Waldrep points out in his comment, it is clear that the track I used is not of the “highest possible quality” (You can clearly see it contains quite a bit of high frequency noise, sorry about that). I think, however, that this fact is irrelevant, and it does not affect the outcome of my experiment. In the spectrogram, one can clearly see that there is frequency content that comes from sampling at a higher rate than 44.1 kHz, and that this content belongs to the signal being recorded. It is not just generated noise. In any case, there is enough high frequency signal for anybody who claims they can hear the differences to detect. My point is, of course, that nobody would be able to, as these frequencies are not audible.

 

Advertisements
Tagged with: , , , , ,
Posted in Uncategorized
21 comments on “Why I think HD audio is irrelevant
  1. Mark Waldrep says:

    Gabriel,

    I’m an advocate for high-resolution recordings, recording engineer, blogger, and instructor of audio and do believe that there are audible differences between recordings that were made originally at 96 kHz/24-bits (and left unmastered) and those that are heavily compressed and released on CDs and as files.

    I write a daily blog about Hi-Res audio at (realHD-audio.com). One of my readers provided me a link to your blind test and I was looking forward to taking the test. But from the spectra and information in the article above, it was clear that the original file in the example above is not at all an HD-Audio file “of the highest possible quality”. It obviously came from a DSD recording and while it may have been transferred to a 192 kHz/24-bit PCM container, it’s no better than a CD in terms of fidelity.

    So I downloaded the test files and took a look at whether these are actually limited to CD resolution vs. HD resolution…in every case, they are not! The A vs. B are virtually identical in terms of their spectra…making it difficult if not impossible to tell them apart. The first section of the Vivaldi clip for example has a small dip in the frequencies between 20 kHz-25 kHz(about 10-15 dB at -90 dB) but is otherwise identical all the way out to 36 kHz! How can you consider this a fair test?

    Other examples are equally skewed…some have small frequency ranges attenuated and then they continue to extend beyond 22 kHz…the max for a CD. I would be happy to provide spectrographs for you and your readers. This test is seriously flaw because of the samples and the process by which you delivered them.

    I provide my listeners the ability to download FREE Real HD-Audio files made at 96 kHz/24-bits and then a downconverted example at 44.1/16-bits…and while it is very challenging to heard the difference (and impossible by playing them through marginal equipment and with such short clip segments), many do detect a difference.

    • cdvsmp3 says:

      Thank you very much for your detailed comment. Where can I get those free HD-Audio files you mention in the last paragraph? I’ll be happy to replace the samples in the HD test with better-quality HD-Audio files. It is sad to see that many tracks that are advertised as HD-audio look like upsampled CD versions. I made sure, however, that the ones I chose showed frequencies over the 22 kHz limit.

    • miceblue says:

      “that there are audible differences between recordings that were made originally at 96 kHz/24-bits (and left unmastered) and those that are heavily compressed and released on CDs and as files.”
      Yes, but only because the two recordings are of different masters to begin with, which makes the comparison invalid and unfair.

      Take this free, true high-resolution recording with legitimate frequencies above the limits of human hearing and downsample it to 24/44.1 and I can guarantee you you won’t hear any differences.
      http://www.lessloss.com/drums-drums-drums-p-203.html

      I’ve done my own tests with them and I’ve failed 11 of the 12 ABX tests I’ve done (10+ trials each) with a DAC, amplifier, and headphones that can all reproduce the ultrasonic frequencies.

    • pieroxy says:

      You say “I’m an advocate for high-resolution recordings, recording engineer, blogger, and instructor of audio and do believe that there are audible differences between recordings that were made originally at 96 kHz/24-bits (and left unmastered) and those that are heavily compressed and released on CDs and as files.” and I agree with you, but the culprit isn’t CD’s technical abilities, it’s in the processing made while preparing the data to be pressed into CDs. I’ve seen that time and again, where I would buy a CD of a vinyl I had and it sounded worse, far worse. I then recorded my vinyl on my Soundblaster (back in the days) and then burned it onto a CD-R. The CD-R sounded just like the vinyl did. No audible difference. But it sounded much better than the new CD I just bought. So the technical medium that is the CD is is perfectly able to sound like a vinyl. If only it was mastered to…

      In other words, if you take the HD file, resample it to 44.1/16 and burn it onto a CD without further processing, you will have no audible difference.

  2. […] the link if you want to visit the site for […]

  3. Ran Perry says:

    Having data above 22kHz does not make the file Hi-Res. Transferring an old master with tape hiss that goes above 22kHz may register in a spectrograph but it is just noise.

    • cdvsmp3 says:

      True! I don’t understand how this affects my analysis, though. If you claim you can hear frequencies from this part of the spectrum, you should be able to spot the differences, whether they are due to noise or to a valid part of the musical signal.

  4. Phi says:

    While he didn’t say so explicitly, Mr. Waldrep’s name at the top of his post is a link to his site, where you’ll see a box on the right-hand side offering the free samples he was talking about. I’ll be interested in seeing what you come up with, working with those.

    There’s much misinformation spread about this subject by those with hidden agendas – like the analog holdovers who prefer the “warmth” of a recording shorn of high frequencies, who flock to the DSD process, which brickwalls high frequencies much as CDs do, unlike high sample rate PCM, which preserves the high frequencies.

    But once DSD has removed them, however, PCM can’t add them back in!

    • cdvsmp3 says:

      Unfortunatlely, Mr. Waldrep has not granted me permission to use clips from those tracks in my test. I’ll see what I can do to get better samples. I’ll do my best, though I am afraid some people will never consider it fair enough. I’ll be grateful if you can suggest any other sources.

  5. martinga13 says:

    It seems to me that this is a case of “not seeing the wood for the trees”. Whilst I admit that for the majority of people (listening on portable players and low grade headphones)
    HD will be irrelevant however for does of us who have highly resolving system that are capable of reproducing anything above 22kHz is is highly relevant. I also concede that there is to much misinformation out there which does not help the cause.

    Ever since CD was introduced in 1982 there have been proponents of the LP format that have said CD just does not sound right and in the beginning everybody said it is just does diehard analoge fans who don’t want to change. In the in the last couple of years there has actually been a resurgence of vinyl sales with 65% being purchased by the latest generation !
    So why is this ? Analog does not have this 22kHz restriction.

    Did you know that the upper harmonics of a violin can go beyond 40kHz and that of a french horn up to 90kHz !

    My next argument is that of Super-tweeters (these are tweeters that play in the supersonic range. i.e above 20k – 100kHz range) For years it has been claimed by the science community that these devices simply can’t work because human hearing does not go
    beyond 22kHz, yet experienced reviewers say they can tell the difference. Here a recent comment by reviewer Stuart Smith :

    ” The bass being produced from the 6.5″ drivers in these speakers was really something but it’s the super tweeters that really do it for me. Turn them off and everything sounds fine. Turn them on and theirs a feeling of much more air around the instruments and an indefinable quality to the music.”

    And here is the proof that it does work.

    As a more recent submission to the International Tinnitus Journal [Vol. 13, No.1, 3-10 (2007)] suggested, our eyes may serve as ‘fenestrations to the ears’.Having measured the physical frequency response of the eye up to 60kHz, it is postulated that our eyeballs may act as acoustic lenses for the ears, ‘passing frequencies beyond the impedance-matching capacity of the eardrum.’ No listening tests here but electroencephalograms that illustrate physiological activity in the brainstem and thalamus ‘when ultrasonic musical frequencies are combined with the musical spectrum below 22kHz.

    Here the link to the research paper : http://www.tinnitusjournal.com/imagebank/pdf/v13n1a02.pdf Just read the CONCLUSIONS

    • miceblue says:

      “So why is this ? Analog does not have this 22kHz restriction.”
      Analog LPs are also infamous for having distortions, which digital-to-analog converters lack if well-built. Furthermore, vinyl pressings often have a different mastering process compared to the CD counterparts, so the comparison between the two mediums is invalid to begin with since you’re comparing apples to oranges (sound format) in addition to pineapples and bananas (master formats).

      “Just read the CONCLUSIONS”
      The point of reading through scientific papers is not only just to read the conclusions, but also the researchers’ methodologies. If you read through the paper, the tests were done on a mere 5 subjects, which is hardly enough to for any test to have a solid conclusion. It was a nice proof-of-concept, but you would need much more evidence to claim this is significant. Secondly, they blasted patients’ eyes with high-frequency sounds at a whopping 96 decibels, way above the standard listening volume levels for everyday use, and considering they were blasting the eyes at 10 mm away is already enough to say that this is just a proof-of-concept. No one in their right mind listens to speakers 10 mm away from their heads. Thirdly, the paper fails to make any attempt to go through any conclusive data from their proposed controlled tests, so we have absolutely no idea if these results are significant at all compared to the controlled tests. Fourthly, this study was only cited by 5 other scientific papers, which suggests that this study has a low impact-factor. Fifthly, that paper takes 3 pages of the paper for discussion about the ultrasonic content (not even related to the results from this particular paper) and only spends 1 page discussing the results of the actual study, a good indication that there wasn’t much novel material obtained resulting from their experiments. Still, the paper was interesting to read nonetheless, but I would have liked to see some real-world tests to see if the subjects could actually detect the ultrasonic sounds rather than measuring them with external transducers and making the assumption that the subjects could detect them.

      I agree that recording audio at 96 kHz sampling rates is good to capture most ultrasonic content that’s present in the real world, but I don’t think playback of that audio is necessary.

      Take these recordings at their native 24/96 formats and down-sample them to 24/44.1 and I can guarantee you you won’t hear a difference between the two tracks.
      http://www.lessloss.com/drums-drums-drums-p-203.html

  6. Frank says:

    it is something like is multi-ampling; is it better than single ampling? Yes also if “YOU” do not ear any difference.
    The fact is that is what is called “noise’ is not only “noise” but “harmonics” that most people cannot ear directly, but they create ambient and stage as they beat and interfere with the other sounds in your room.
    Many audiophile systems response range extend well beyond 20 kHz up to 250 kHz in some cases; loudspeakers systems have the capacity to arrive around 40kHz or can be completed with a “supertweeter” to extend the basic range.
    All this not to perceive directly the sound above 20kHz (16KHz or lesser for most of us if aged) but the interactions with audible range that in the 20 to 40 kHz range keep still some effects.
    Therefore the instrumental analysis reveals in the last track and make audible those harmonics that has been called noise, but only partially are.
    Finally is better to cut all frequencies recorded above 20kHz or not? If you do not cut them the sound is the same or better? The answer is: it is surely better also if you cannot ear any audible difference.

  7. Ignacio says:

    While it its not “full HD” I can provide a recording made at 96khz 24 bit with RME converters. Write if you are interested.
    Best.

    • cdvsmp3 says:

      Thank you, Ignacio. I have decided to interrupt my HD vs AAC test, so I won’t be needing any extra HD files. All the null testing I have done with HD vs CD always gives the same result: silent tracks, as in the example in the article. I don’t see the point of using HD tracks if the tests are about “hearing” the differences.

  8. This “vinyl is better than a CD” is starting to tire me a bit, although admittedly I also partake in the discussion (and more than what is good for my mental health). It is, however, true that almost all CDs from the 80s sound worse than the vinyl counterpart. Those CDs were cold, thin and shrill. Very unpleasant. I think almost anybody would agree with that in a listening test. Of all the CDs I have heard from the 80s, only two albums were downright better on CD, and one of them was a reissue of an album from 1977 (the other was “Disintegration” by The Cure).
    However, this became a lot better around 1992-1993, and from 1994 onwards, the CD is almost always better, or then at least it’s a matter of taste (and sometimes I definitely prefer the vinyl record). However, reissues are a lottery. Some CD reissues are an enormous improvement over the record, some are significantly worse.
    All this aside, I simply don’t understand why so many people insist on arguing a point based on theory and ideology (“Vinyl is always better than CRAP CDs”) instead of using their ears.

    Anybody who claims that vinyl goes to much higher frequencies should actually sit down and compare their entire record collection to CDs and then hear the difference. I have been/am still in the process of doing that (and I have started to work on my upcoming website http://www.VinylVsCd.com about my findings), as I decided that I wanted the best sounding media – not the vinyl record in all cases. For 15 years I insisted on vinyl only. Now I have compared more than 600 albums.
    What I have found in most cases is perfectly illustrated with the following quote from http://wiki.hydrogenaud.io/index.php?title=Myths_%28Vinyl%29:

    “Myth: Vinyl is better than CD because it reproduces higher frequencies than CD and avoids anti-aliasing filter issues at the frequencies CDs can reproduce

    The recording/tracking ability of vinyl is easily at least 50 kHz and perhaps as high as 100 kHz. The most notable proof of this is the CD4 quadraphonic system which relied on a 45 kHz bandwidth to be accurately reproduced. That said, the high-frequency response accuracy of vinyl varies tremendously. Amplitude deviations of 5-10 dB or greater are not uncommon in the 20 kHz range for many records.

    Playback of ultrasound frequencies is still not guaranteed. Many MM cartridges have resonant peaks defined by the preamp loading, or stylus tip resonances defined by the cantilever, that attenuate high-frequency content.

    When groove wear does occur, it occurs much faster at high frequencies than at low frequencies. For modern styli this is not as much of a concern, though.

    There are rarely, if ever, any ultrasonic frequencies for vinyl to preserve. In audio recordings, such frequencies, when present, are normally low-energy noise imparted by electrical equipment and storage media used during recording, mixing, and mastering. Although some musical instruments can produce low-energy overtones in the ultrasonic range, they could only be on the vinyl if every piece of equipment and storage medium in the recording, mixing, and mastering stages was able to preserve them—which is unlikely even in modern recordings, since the average microphone or mixing console is designed only with audible frequencies in mind. Even if the overtones were preserved all the way to the mastering stage, mono and stereo lacquer cutting equipment typically includes a lowpass filter to avoid overheating the cutting head with ultrasonic frequencies.

    Finally, on top of all of these issues, there is simply no scientific evidence that frequencies beyond the 22 kHz limit of CD audio are audible to any known group of people, or that such frequencies affect anyone’s perception of the audible range. There is no evidence that reconstruction and anti-aliasing issues are audible.”

    What my test (on a stereo system that cost around $16,000) has shown is simply that the most common difference between vinyl and CD is that the highest frequencies are cut off on vinyl. I have literally only heard a handful of CDs (out of 600 albums) where the record played higher frequencies than the CD, and that was simply with CDs that sounded completely “wrong”.
    And honestly, the ubiquitous “reason” vinyl lovers give is that vinyl has a “warm” sound, and that CDs are too shrill. In other words: The vinyl lovers prefer that records play LESS high frequencies than CDs.
    Moreover: I have used the program Wavelab to also look at audio files with a spectrometer – but a different one than the one used here. I used one that looked like one of those equalizers (like this: http://fc09.deviantart.net/fs70/f/2010/102/0/8/Equalizer_by_speedster2000.jpg).
    I looked at both recordings of vinyl and CD-rips, and I looked at heavy metal, jazz, grunge, orchestral soundtrack music, progressive rock and pop music, and this was music ranging from 1963 until today.
    To my own surprise, hardly ANY of them went above 10-15 kHz! I also got my hands on some 24 bit/96 kHz recordings done with a turntable that including arm and needle cost $19,500 (it was a Dr. Feickert Woodpecker turntable with an Inertia platter, Reed 3Q 10,5″ tonearm and Dynavector DV DRT XV-1t cartridge). I had recorded the same album with my own Rega turntable, and despite a price of one tenth of the other turntable, the spectrometer showed that my Rega played back higher frequencies than the Dr. Feickert one. Moreover, it simply sounded much better. Even my wife said that. And those recordings with the Rega one hardly went over 10 kHz either! And ask yourself: Would you really enjoy listening to sounds that high-pitched?
    I will include the mentioned screenshots from the spectrometer when I launch my website, so keep an eye out for http://www.VinylVsCd.com

  9. By the way, Gabriel, although your discussion of HD audio versus CDs seems to be over, would you consider replacing the old audio files with the ones the other commentators offered, just so that newly arrived people can “hear” the difference and perhaps make the HD lovers shut up ;-)? I have been told by various people that 24 bit is supposed to be more “airy” than 16 bit, so I would be interested in hearing this.

    • cdvsmp3 says:

      Hello, Anders. Unfortunately the site where I host my test will be retiring the service soon, so I will not be able to offer any of my tests online within a month or so. I am working on a downloadable version that visitors could try on their own. Anyway, you need not wait for it. Do your own trials! It is easy. Just get free 24/192 samples from HDtracks, for example, make CD or lossy AAC versions using iTunes and use an ABX tool (Lacinato is a good free tool for all platforms) to make sure you test them blind, because our brains are truly amazing at deceiving us!

  10. Okay, I’ll try that. I have actually done something slightly similar once, but I will try again. I hear you can also do it in Foobar, so I will try that or otherwise get Lacinato.

  11. paulonpaper says:

    What a debate in the comments! When it comes to Hd audio, I do the simple thing: I just listen to my bank account which then tells me to avoid it altogether.

  12. Ricardo says:

    Very good experiment, the true question in the experiment is, DSD music has more audio (more audio signals) in the audible range?

    It’s like your second part of the experiment when you show the spectrogram of the CD.

    The question is if you cut the audio of the DSD in the same frequencies that it has a CD (0Hrz to 22 or 23 Khrz), DSD and lossless format has the same audio?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: