Some technical debates are so polarised that it is hard to believe there can be sane people on both sides. One such is that over whether high resolution audio is audibly superior to CD.
First, a little orientation. A standard redbook CD stores music encoded as 16-bit PCM (Pulse Code Modulation) sampled at 44,100 Hz. This standard was first published in 1980 and the first production CD player appeared in 1982.
Technology has moved on a long way since then, and around 10 years ago there was an industry format war over what higher resolution disc should replace CD. Sony championed SACD, which uses Direct Stream Digital, a 1-bit system with samples at 2822.4 kHz. Meridian and the DVD forum supported DVD Audio, offering a variety of possible formats up to 24-bit PCM sampled at 192 kHz.
Bigger numbers, better quality, right?
The 16/44 transparency test
Maybe, maybe not. Some audio engineers maintained that even 16/44 was more than adequate to convey audio with an accuracy greater than the limits of human hearing, at normal listening levels. In 2007, Brad Meyer and David Moran put this theory to the test. They conducted a series of listening experiments using high-resolution sources, testing one simple proposition: if you convert the audio signal to 16/44 digital and back during playback, is the difference in quality detectable?
Meyer and Moran wrote up their results in a paper published by the Journal of the Audio Engineering Society. Unfortunately the full paper is only available to AES members, but the results are well known and widely discussed. Their tests, which used double-blind testing techniques where neither listener nor tester knew which signal was being played, showed that no listener could reliably detect when the additional 16/44 conversion was inserted into the signal path.
On the face of it, this shows that no matter how good a stereo SACD or DVDA disc sounds, it could also be encoded onto a CD and sound the same to human ears. There was an unimportant caveat. If the volume was whacked up to very high levels, you could hear a difference in the noise floor.
What the transparency test does not prove
It’s worth noting that the Meyer/Moran test only covers one point: that 16/44 is (or is not) acoustically transparent. It does not prove that high-resolution audio is pointless; in fact, it has obvious value in production (as opposed to delivery), since production means digital processing, which degrades the sound; it makes sense to work in a resolution much better than you need for final output. I guess the same argument can be applied even to the distribution format, if the player is processing the sound to apply equalisation or bass management, for example.
Even if the test result is correct, it may still be that SACDs sound better than the same music on CDs, maybe because it is mastered better, or the player performs better with SACD, or maybe better source tapes were used for the SACD. In fact, Brad Meyer says:
Those who have read the JAES paper written by me and David Moran may remember that we too thought that the high-bit recordings we heard sounded, as a class, really exceptionally good. Our experiment, however, made a very good case for the theory that the reason for this lies not in the extra bits but in the market niche these recordings occupy.
Your rant against what you call square-wave recordings (i.e. ones in which the dynamic range is very heavily compressed to make the average level higher, which is a common mastering practice) is one I quite agree with, but it too has nothing to do with the number of bits in the recording. Our experiment showed that those awful-sounding things could just as easily have been issued as SACDs — and conversely that the excellent sounds we heard from our test material could have been issued in 16/44.1 without audible degradation.
Here’s how I think it works. SACDs are issued to a tiny niche market that is known to use good to excellent equipment, and to be fanatically devoted to realistic timbres and dynamics. Because the big guys in the record companies don’t care at all about such a tiny niche and are financing these SACDs because it’s the modern thing and sort of prestigious (and the other companies are doing it), they leave the engineers and producers alone, and the latter just make the stuff sound good on their own studio monitors and good home systems, and send ’em on out there.
And guess what? If a skilled engineer has as his only goal making something sound good enough to show off to his colleagues, you’re gonna think it’s pretty damn good too.
It is also worth mentioning that both SACD and DVDA are not solely about high-resolution stereo. They are also for surround sound, something which a CD does not support.
In a nutshell: you could agree with the test result and still want to invest in SACD discs and hardware.
Does it matter?
Does it matter? Well, a considerable part of the marketing behind SACD and DVDA is that the sound quality is better than ordinary CDs. Unfortunately Sony has back-pedalled so much on SACD that I can no longer find its marketing material online; but it emphasised that Direct Stream Digital is more analogue-like than PCM as well as higher resolution. I am not focusing here on the benefits of DSD over PCM, which are also contested, but will mention that the transparency test casts doubt on this as well as on high resolution sound itself. This is from a leaflet tucked into early SACD releases:
The SACD format is about to change the way you listen to music … thanks to an amazing 2,822,400 samples per second, you get audio performance that no other format can deliver. Where CD frequency response extends to 20,000 Hz, DSD technology can theoretically reach 100,000 Hz. Where CD has dynamic range of 96 dB, DSD recording can achieve 120 dB across the entire audible range. Thanks to DSD technology, the SACD difference is breathtaking. If you care passionately about music, then SACD will inflame that passion as never before.
Unlike multi-bit PCM recording, SACD’s DSD technology uses a one-bit pulse that is analogous to the music waveform.
You can also get a flavour of it here, an unofficial FAQ, including this statement:
…the difference between regular audio CD and the high-density layer of SA-CD can be quite easily perceived, even to untrained ears
There is still an SACD market, though the format has disappointed in an absolute sense, and Sony (and others) still sell high-end SACD players. There are also audiophile sites where you can purchase high-resolution downloads, such as HDtracks and Linn Downloads. The high resolution downloads cost more; so if nothing else, Meyer and Moran might save you some money even if you want the best no-compromise sound.
Industry and community reaction
Meyer and Moran’s paper has been widely debated, though considering the implications for high-end audio perhaps not as much as you would expect. Many audiophiles simply disbelieve the result. They are happy that earlier papers such as this one [pdf] by J Robert Stuart of Meridian present technical arguments why 16/44 is not enough:
The CD channel with 44.1kHz 16-bit coding (even with noise shaping to extend the resolution) is inadequate. Even 48kHz sampling is not quite high enough.
Hi-fi critic Robert Harley says that the conclusions of the test prove that double-blind testing does not work:
Every few years, the results of some blind listening test are announced that purportedly “prove” an absurd conclusion. These tests, ironically, say more about the flaws inherent in blind listening tests than about the phenomena in question.
The latest in this long history is a double-blind test that, the authors conclude, demonstrates that 44.1kHz/16-bit digital audio is indistinguishable from high-resolution digital. Note the word “indistinguishable.” The authors aren’t saying that high-res digital might sound a little different from Red Book CD but is no better. Or that high-res digital is only slightly better and not worth the additional cost. Rather, they reached the rather startling conclusion that CD-quality audio sounds exactly the same as 96kHz/24-bit PCM and DSD, the encoding scheme used in SACD. That is, under double-blind test conditions, 60 expert listeners over 554 trials couldn’t hear any differences between CD, SACD, and 96/24. The study was published in the September, 2007 Journal of the Audio Engineering Society.
I contend that such tests are an indictment of blind listening tests in general because of the patently absurd conclusions to which they lead.
Reactions like this are to be expected from parties that are heavily invested in the merits of some facet of technology. Hi-fi is particularly problematic for several reasons, for example:
- We have difficulty in separating our enjoyment of the music from the quality of the technology.
- There are known placebo effects which make sighted tests suspect. Our brain tells us something should sound better, and therefore it does sound better; even if science or blind testing later proves that it was not for technical reasons.
- Differences such as a slight increase in volume may be perceived as an improvement in quality; hi-fi demonstrators are well aware of this. Exact matching of volume is difficult to do at home.
Others have advanced different reasons for the Meyer/Moran results. Perhaps the equipment was not good enough. Perhaps the testing was not rigorous enough, or they used the wrong source material. The test authors posted some more information so that others could form an opinion about this. The equipment looks reasonable to me, though it is not at the ultra high end of hi-fi.
Still, what is surprising is that (as far as I’m aware) nobody has repeated the tests correcting such flaws and obtaining a different result. Of course that may yet happen; but the lack of such counter-tests so far strikes me as equally as significant as the test itself.
I asked Harley about this and he replied:
I’m not aware of any formal DBT of standard-resoution digital audio with high-resolution digital audio. The difference between 44.1kHz/16-bit digital audio and 176.4kHz/24-bit is obvious, in my experience. It is, in fact, so obvious that no one (no one that is a disinterested experimenter, that is) has bothered to organize and conduct it.
Your last sentence reminds me of the cartoon by B. Kliban showing a professor at a blackboard full of mathematical equations in front of a classroom in which all the students are fish. The caption is “Proving the Existence of Fish
Is it obvious? The resources section below offers some samples in both 16/44 and 24/96 that you can try and I’ve also done my own experiments, though comparing such samples is not quite the same test that Meyer/Moran conducted. I’d contend though that the difference is not so obvious that it needs no argument and I would not put money on my ability to tell them apart under blind conditions.
Mastering matters more
The Meyer/Moran test is persuasive, but I am open to the idea that some further research may prove it to be not the last word on the transparency of 16/44. Still, I’d expect the audible differences between 16/44 and higher resolutions to be subtle at best, and I am sceptical of any claims otherwise (like Harley’s above).
A bigger concern is that the focus on high-resolution formats, or minutiae like whether an SACD player converts DSD to PCM during its output processing, distracts from what matters more: the way the music is recorded and mastered.
Thanks to the ongoing loudness wars some audiophiles seek out early CDs from the eighties, or even vinyl records, to get the best-sounding versions of favourite albums. Some of those old CDs do indeed sound stunningly good; not only because of greater dynamic range, but also because a light touch was applied to mastering in the early days so that a CD was close in sound to the master tape. Not all old CDs sound good of course; sometimes bad tapes were used or damaging noise reduction applied; but the fact that some twenty-year old CDs can out-perform modern pressings shows that mastering decisions count for more than digital technology or resolution.
HDtracks offers a free sample of various styles of music in both 16/44 and 24/96
Sample from Barry Diament’s Soundkeeper Records comparing 16/44 to 24/96