This, or That?

Dirty word, three letters: ABX.

Dirty, that is, to an audio subjectivist; holy to an objectivist. There are those who like to divide us into these warring clans: the subjectivists who believe that only long-term, close music listening can reveal the quality of a playback system or a component; the objectivists who depend on direct, controlled experimental comparison (and measurement, where applicable).

Now, I am not so foolish as to declare an allegiance to one tribe or the other, right here in front of God and everybody; not by a long chalk. I am, however, Missourian enough (the “Show Me” state) to want to experience both modes for myself.

The subjective pleasures and challenges of long-term listening are easily obtained: we all do it, most every day. The other thing can be considerably harder to come by, since the hardware to make direct, rapid, and correctly level-matched comparisons of speakers, amplifiers, or even cables is not trivial, nor cheap, and is always open to criticism for audible or otherwise detectable influences.

There is one arena, however, where precisely such comparisons are easily available, free of charge, to one and all: data-compressed audio codecs. We constantly read how musically execrable low-bit rate encoding is, how it presents “only 5 percent of the music” (thank you, Neil Young), has slain the compact disc and thus the music industry as it was, and how MP3 is ruining music-listening and despoiling our youth.

All of which very well may be so. But shouldn’t we weigh it for ourselves? Should we not let our listening ears, and not our reading eyes, make our evaluations? We can, and we shall.

The term “ABX” refers to a comparison between samples A and B, be they audio components or formats or cola drinks, in which the subject is asked only to identify each trial, “blind,” as one or the other, having previously had liberal opportunity to sample the pair “sighted.” A proper test is furthermore “double-blind”: not only is the subject ignorant of which one is “X”on every trial, but the operator or administrator of the test is equally so. In such tests, multiple trials are required to achieve statistical meaning. Without going into the mathematics of the question, we can accept that, in the case of a single subject, that is to say, you or me, 10 to 20 trials is the usual accepted range.

And as it happens, there’s a double-blind software ABX codec-comparator available to us for free—at least to the better-looking, more intelligent and socially developed among us, by which I mean Mac users. ABXTester is a small, essentially anonymous OSX freeware app available for download from Apple’s Macintosh App Store ( (Okay, okay: the rest of you, if you can tear yourselves away from Minesweeper and Visicalc, can head over to and download the eponymous freeware player, and its extra-component ABX plug-in. I have not played with these myself, however, nor with any of the several other free/shareware Windows ABX programs available around the web.)

ABXTester couldn’t be simpler, and it’s entirely self-explanatory. After pointing the program to your A and B files via familiar “Open” dialog-boxes, you’re given five trials of “X,” which you can play, pause, and advance/rewind at will, and are asked each time to identify it as “A” or “B” (you can hit the button to “Reset X” for five more trials).

For A and B, of course, you must first rip two copies of a music selection, which you can freely audition, named, as often as you like before or during your “X” run. I would recommend a well-recorded, “open,” not-too-complex track with a variety of sounds, clean transients, loud and soft elements, and silences. (A track like “Lil’ Liza Jane” from Chesky’s excellent Dixieland recording Dreams of New Orleans by trombonist Wycliffe Gordon is near ideal). I’d further advise making your first comparison between an uncompressed or Apple Lossless rip (the app doesn’t accept FLACs), and an MP3 coded at 64 kilobits per second or even lower: you want to start yourself out with an “easy” one. (From what I can tell, Foobar2000’s plug-in works similarly.)

Use your favorite headphones, or your best system; use an outboard headphone amp and/or DAC if you like (ABXTester will follow playback device chosen in OSX’ preferences), and settle in, uninterrupted, in a quiet room. After you complete your five trials, a button returns your score: 100% for five correct picks; 60% for three out of five, and so on. If you can’t score at least 80% (four for five) on two consecutive five-trial runs, you probably cannot tell the difference between A and B. Once you have 64K MP3 cooled (if you do) you can try your luck with a 128K file, a 256K file, or even a 320K file. If you find it difficult to nail a 64K vs. lossless/uncompressed comparison, try, 48K or even lower. (If you can’t discriminate a 32K MP3 every single time, we may have to suspend your subscription and repossess your S&V secret de-codec ring.)

What will you find? Only your ears can answer that, but I’m betting that many may be surprised and intrigued—especially as you work your way up to higher bit-rates, or try comparing same-rate MP3 with MP4/AAC files. (You can even compare Lossless to uncompressed, but take my word for it: it’s a total waste of time.) And you will almost certainly learn something. You may confirm your self image as a true golden-ear, or discover a dark secret you will keep from your audio-geek friends to the grave. Either way you will have experienced ABX testing for yourself.