Compression 101

We were taking our morning constitutional around the Interwebs one day last week when we happened upon an article on Timesonline titled "Why Music Really Is Getting Louder". Oh boy, we thought, a mainstream outlet is catching on to the whole issue of dynamic compression—a subject we have inveighed against repeatedly over the years. (JA first preached that particular sermon back in 1999.)

But when we actually read the article by Adam Sherwin, "media consultant," we happened upon a sentence so ill-informed that our eyebrows nearly zoomed through our hairline (well, if we had a hairline, that is). After a discussion of the effect of dynamic compression on the sound of recordings, Sherwin wrote: "Downloading has exacerbated the effect. Songs are compressed once again into digital files before being sold on iTunes and similar sites. The reduction in quality is so marked that EMI has introduced higher-quality digital tracks, albeit at a premium price, in response to consumer demand."

This is, of course, poppycock. We are sure that Mr. Sherwin is a fine and princely fellow, undoubtedly kind to his mother and generous to small children and tiny animals, but what he most certainly is not is a writer who knows anything about audio. Why, oh why do newspapers seem to think that "media" writers—writers who have spent years familiarizing themselves with computers or appliances can write intelligently about hi-fi matters. Real estate correspondents need to know real estate, business correspondents presumably need to know about, well, business. Why are the standards lower for audio?

For the record then, here's what got Mr. Sherwin in trouble: "compression" means different things depending on how it is used. There's compression that is used as a studio tool—allowing, say, drums to "sound louder" and have more impact. Then there's dynamic compression, which "louderizes" softer sounds until they are as loud as the loudest portions of the recording. This makes the track "sound louder," but at the expense of true dynamic range.

Finally, there's data compression (also known as source coding), which encodes information using fewer bits. This can be done by discarding redundant bits—called "lossless" compression, because the original data can be completely restored—and by eliminating information that a psychoacoustic model suggests might not be audible anyway—"lossy" compression. MP3 is a form of lossy compression; as well as discarding redundant bits, the coding eliminates information on the theory that space is more important than accuracy. (When accuracy is valued, but space also matters, there's lossless compression.)

Digital recording, which allows engineers to work closer to the 0dB "ceiling," is frequently cited as the reason the practice of dynamic compression has become so ubiquitous, but its roots lie back in the 1950s, when record executives wanted singles mastered so that they sounded louder than the competition when played on the radio—on the theory that louder sounded better. Eventually, everybody caught on to that trick—and then radio stations started compressing their signals so that they sounded louder than their competition. Now the practice is so common that there's little point in owning a high-end tuner. Almost every station's signal sounds like crap.

Of course, digital mastering has helped the "loudness wars" reach their current state, where some recordings have as little as 4dB dynamic range—maybe less. How can there be less? 'Splaining that is hard to do, but fortunately, this page shows you what that means.

Ouch, that hurts to look at.

Confusing? Yes, it can be—if you're just visiting from the gadget pages. Next time, perhaps Timesonline will hire a hi-fi expert. They could call me—I know a few.