The Guardian Looks at High-Res Music, and Gets a Lot Wrong

An article in the Guardian today, How much difference is there between MP3, CD and 24-bit audio?, examines a question I’ve looked at here often: can you here a difference between high-resolution music and CD-quality music?

They discuss “24-bit audio,” without talking about sample rates. Most “high-res” music is 24/96, or 24-bit, 96 kHz, but they treat the subject as though the 24-bit depth is all that matters. [1]

They go on to test high-res music, and to do so, enlist a purveyor of such music, Linn Records. Now, one could argue that it’s best to examine a product by a company that sells it, but Linn Records is interested in proving a certain outcome in any such tests. This company set up tests, using their own “high-quality system and speakers.”

The problem with the test is, according to the way it was described, it wasn’t blind. One of the test subjects says:

…listening a very high-quality set of Linn speakers, going from low quality MP3, to high-quality MP3 and finally the high resolution studio master,

If they know what they’re listening to, their confirmation bias enters into the equation. But the same person goes on to say:

My impression at the end of our listening session was that yes, there’s a distinct quality difference between the kind of compressed, middling MP3 commonly downloaded from the major platforms and the 24-bit high-res studio master. But mostly I found the CD-quality track on Linn’s superb hi-fi equipment to be the overall best listen for my particular ears.

When vendors of these products run tests, they generally start with a crappy MP3 rip; the kind that no one listens to any more. I don’t know what the “middling MP3 commonly downloaded from the major platforms” is, because the test claimed to also include a 320 kbps MP3 file, which is not “middling.”

But the article does point out one thing that I’ve often mentioned:

But that difference wasn’t always a good thing. It was disappointing to hear a recording of Pavarotti’s Nessun Dorma sound worse in studio master, as it exposed the fact that the orchestra and the tenor’s tracks were recorded separately in different environments. They sounded disconnected — something that is masked in the CD version.

A bad master can sound worse in a higher resolution; however, this is likely only to be noticeable on a very high-end stereo. And what the person heard is likely not a different environment, but different miking. It’s possible, however, that either the singer or the orchestral parts had a slightly different amount of reverb applied to them.

So the Guardian articles did three things wrong: they got someone interested in proving that high-res music is better to run a test (who chose which recordings were tested); that test was done on high-end equipment, that none of them could probably afford; and the test wasn’t blind. At least their comments weren’t astounding praise of the music quality, but they’ve misled readers by falling into a test using the conditions that a retailer wanted, rather than truly objective conditions.

  1. I wasn’t going to niggle about the “44 kHz sample rate” they discuss, but after an email from a friend, I decided to add a comment. It’s not hard to get a number right: it’s 44.1 kHz, and it’s easy enough to find out. But it’s typical of The Guardian’s tech reporting that they get the details wrong too. (I’m reminded of an article about Android phones they ran a couple of months ago, illustrated by a photo of an iPhone.)  ↩