MP3, AIFF, AAC, and ME

A while back (somewhere in the posts below, or in one of Jackson's posts, or in a dream I had) Ted and I were going about our usual "We hate MP3's" routine, when Hazmat suggested that AAC (aka m4a) was the real format of choice for digitized music.

I did a bit of research and found this, on Apple's website:

AAC compressed audio at 128 Kbps (stereo) has been judged by expert listeners to be “indistinguishable” from the original uncompressed audio source.*

* Information provided by Dolby Labs

As I said at the time, this is a bold statement.

Ken (aka Hazmat) recommended that we do a blind taste (sound?) test, comparing original AIFF's with AAC's. He was interested to see if I could tell the difference.

Now, I don't consider myself to be an "expert listener." I'm still really an amateur, and I suspect it'll take my whole life to really become an expert, by which time I will likely be deaf. But I decided to Take The Challenge.

So Ted and I took the new Brain Shivers record and converted each track to an AAC (at 128Kbps).

I sat at the back of the room as Ted played me bits of ten songs, each song twice, either the AAC or the AIFF first. I marked down my first impression, my gut feeling. Was I hearing the original, uncompressed audio, or the "indistinguishable" AAC?

I listened to no more than a minute of each, and in some cases made my choice after ten seconds or so. This was through my monitors, at "regular" listening volumes.

The differences were damned subtle, and in every case I made my prediction based more on an overall feel than anything else.

One thing's for certain. AAC is good stuff, and at 128Kbps it's damned impressive.

So after ten tracks Ted checked my results. I got 80% correct.

According to statistician/rocker Eric from Geek Farm, these results are statistically significant. I could hear the difference.

Good thing I don't work for Dolby.

Anyhow, it was good fun. The main difference between the AAC's and the AIFF's was a sense of space and localization, which makes perfect sense since this information is all coded in the high frequencies. These require more data to represent, and are the first things compromised when data is compressed.

As I said, the differences are subtle. And at a higher bit rate (192Kbps) I don't know if I could repeat my performance.