Audio compression may not be so bad

Our tests with music professionals found that they had a hard time distinguishing between compressed and uncompressed songs

We all listen to compressed music on our computers and portable players and we all know that compression is supposed to hurt sound quality. But does it hurt it enough to affect our listening pleasure?

I tested the effect of compression with a jury of music professionals and the results surprised me. Although the type of compression had a definite effect-my judges preferred .wma to .m4a and .mp3--the level of compression had little effect on how they rated sound quality. Most importantly, none of the compressed music sounded really bad to their well-trained ears.

The most common consumer audio formats--.mp3, .wma, and .m4a--all use lossy compression that throws away musical nuances to cut download times and fit more songs on our players. For instance, if you ripped Dire Straits' album Brothers in Arms to your hard drive as uncompressed .wav files, they would take up 556MB of space. You couldn't fit 15 albums of equal length on an 8GB iPod Touch. But as .wma files, compressed to 160kbps, the whole album is less than 64MB, leaving room for 127 similarly-squeezed albums.

Many people say they can hear the difference between music compressed at different levels, but their experiences could be shaped by their expectations. They know they're listening to a highly compressed song, so they blame the compression for every real or imagined flaw.

To get a less biased judgment, I set up blind audio quality tests. A jury of musicians rated various compressed samples on how closely they resembled the original.

Sound Methodology

You should consider my results to be anecdotal, not definitive. I only used two recordings and four jurors--hardly a scientific sample size. Besides, this type of test is inherently subjective; people respond to what they like.

I ripped two 30-second clips from CDs for tests. One came from the first movement of Tchaikovsky's sixth symphony, performed by the London Symphony Orchestra, with Igor Markevitch conducting. The other came from Joan Osborne's cover of Bob Dylan's "Man in the Long Black Coat."

I saved seven versions of each clip. To test music formats, I saved the clips with 128kbps compression in the three most popular formats: .m4a (which uses aac compression and is what you get from iTunes), .wma (Microsoft's format), and the old standard, generic .mp3.

To find out how compression levels (as opposed to formats) affected audio, I also saved .mp3s at 192, 256, and 320kbps. The larger number means less compression and, in theory, better sound. I also saved the clips as uncompressed .wav files.

[Want to try a portion of this test? We've set up a page with versions of both clips at different compression rates. (You'll need to have Quicktime installed.) You can listen and tell us what you think the quality is of each clip. SPOILER ALERT! Hold down the Control key as you click the link, then wait a couple of minutes before you go the new tab. The less compressed the file, the longer it takes to load, so if you see the page load, you'll get an idea of which files are high quality. Click here for the test.]

I played the uncompressed .wav file, then a compressed version, and asked the jurors to grade the second sample compared to the original. For one test, I played the uncompressed original twice, without the jurors knowing they were listening to the same file twice to see if they could detect the high quality .wav version.

I played the files using my Lenovo X60 laptop, connected to my home theater sound system with a Yamaha RX-V465 receiver and satellite speakers and a subwoofer from Cambridge Soundworks.

I instructed the jurors to grade the audio from 5 ("sounds identical to the uncompressed original") through 1 ("unacceptable"). Decimal numbers were allowed, and comments encouraged.

Join the newsletter!

Error: Please check your email address.

Tags audio compressionMP3audio codecs

More about LenovoMicrosoftStanford UniversityYamaha

Show Comments
[]