Can your ears tell the difference between a lossless and lossy song? I’m talking about the difference between audio that’s been compressed in a way that preserves the original audio data perfectly when it’s decompressed and played back (lossless, as in “nothing lost in translation”), and audio that’s been compressed in a way that actually changes the audio by either slightly or more noticeably attenuating it (lossy, or “some things lost in translation”).
When you hear numbers like 128 Kbps or 256 Kbps tossed around in relation to MP3 files or songs you’ve purchased through services like Amazon or iTunes, they’re referring to the bit rate of an audio file. With lossless audio, that bit rate (for songs of average sonic complexity) typically lives in the vicinity of 700 to 1,000 Kbps — a range well above the human ear’s ability to discern differences. With lossy audio, however, the standard these days is 256 Kbps, a rate that produces audio well within the human ear’s ability to notice differences when compared to the lossless analogue.
The sole reason for compressing an audio file to low-level bit rates is to control size. The smaller the audio file, the easier it is to fit on a mobile playback device with limited space. My 28,000-song music library is fully lossless (Apple’s codec), but it’s also nearly 600 GB. It’ll be awhile before we’re toting around 1TB phones or tablets (Apple’s crème de la crème iPod for storage, the so-called iPod Classic, tops out at 160GB). If I encoded all my music at the going rate of 256 Kbps, my library would clock in well below half its current size, and I’d probably triple or quadruple the number of songs I could cram on my 16GB iPhone 4.
The question is, can most people really hear the difference between lossless and lossy audio — say 256 Kbps, since it’s the current standard? The answer’s a bit more complex, but before we delve into that sometimes touchy subject, let’s talk about what’s called an “ABX test.” That’s a double-blind (that is, eliminating subjective bias) way to compare an A and B source with a random X source selected from A or B that you then try to match to the original source. In an audio ABX test, you first listen to the lossless and lossy versions of a song, then attempt to match random playbacks (X) of one or the other with the correct version (A or B).
You can find several free utilities that offer the test for both Mac and Windows. My favorite on the Mac is ABXTester, and for Windows, there’s a plugin for the free audio player foobar2000 called ABX Comparator that’ll let you do the same (feel free to suggest others in the comments below). You pick the audio files for the A and B sources, then the app plays (or lets you play) either sample at random and asks you to match it with its source. ABXTester give you five randomly selected ‘X’ samples, for instance, then lets you check your results, returning a percent correct score (from 0% to 100%).
A crucial point before testing: What you listen to the audio samples on matters just as much as things like the sample’s bit rate. Your laptop speakers aren’t going to cut it. Neither are your computer’s desktop speakers if they’re your garden variety Labtec or Bose or Insignia. You’re going to want a pair of reasonably high-end, over-the-ear headphones, ideally. That’s for the audiophile listening test.
But as any audiophile has to admit, most people don’t sit around listening to audio through high-end, head-mounted earphones. They’re more likely listening in noisy ambient environments (say your living room’s stereo system, while making dinner, kids playing, the dog barking, etc.), or while out for a walk or run as cars go by, or in a car with vehicle noise, wind noise, passing traffic noise, etc. Our listening lives are sonically cluttered, compromising our ability to pick up subtle sonic differences as incoming frequencies overlap. It’s thus worth plugging your working pair of earphones, e.g. earbuds, those sports headphones with over-the-ear hooks, etc., and running an ABX test somewhere with ambient noise. Even if you can easily tell the difference between lossless and lossy audio in the audiophile test, chances are excellent you won’t be able to while jogging past the construction guy who’s jack-hammering the pavement across the street.
I won’t go into the differences between encoders here (that is, the mechanism/algorithm for converting an audio source into another format). Suffice to say there’s an often noticeable difference between different encoding schemes at the same bit rate. AAC, for instance, is generally considered a little better, bit rate for bit rate, than MP3 (Apple’s iTunes Store uses 256 Kbps AAC). Bear that in mind as you’re testing.
I’ll be back shortly to talk about my experience grappling with this over the past several years. My contention, since I’m usually able to hear the difference between 256 Kbps and lossless audio in the audiophile version of these ABX tests, is that merchants like Apple and Amazon ought to offer lossless audio and let us choose how to encode it. I’d rather have the option to encode down, and that it be my choice, than be stuck with a compromised audio source that — however indistinguishable in a noisy ambient environment — isn’t being reproduced (assuming reasonably high-end audio reproduction equipment) as the original audio engineering team intended.