Note: This page has been thrown together quickly, and only has the simple HTML mark-up needed to work with Safari, Netscape, and Mozilla. This page will not work with Internet Explorer. I've only tested this page on my Mac, so Windows performance is unknown.
The sample material is 30 seconds from R.E.M.'s Drive. This first sample uses 128Kbps AAC encoding, as performed by iTunes 4.0.1:
(I'm using (roughly at least) the sample 30-second cut that Apple makes available as a sample in the iTunes Music Store, so hopefully I'm staying within the bounds of fair use as far as copyright is concerned!)
I've taken the above sample, decompressed it back into AIFF, then recompressed the AIFF result back into AAC at 128Kbps. Normally you'd want to avoid doing this, because there's definitely a loss in sound quality from piling one generation of lossy compression on top of another. The interesting question, however, is: Just how audible is the loss?
The little test I've put together below is far from definitive. I'm only presenting one sample passage rather than a variety of sound and music. This passage has not been chosen as something specifically known to be difficult to process with AAC. Nevertheless, I'm still curious how many people will be able to pick out one generation of compression from two.
In the table below, two of the samples are one-generation AAC, two have undergone two generations of processing. Your challenge is to guess which is which.
| A | |
| B | |
| C | |
| D |
For best results, please wait until all of the samples have been buffered so you don't have rebuffering glitches interfering with your listening. Going from A to D, the possible choices of generation counts are: 1-1-2-2, 1-2-1-2, 1-2-2-1, 2-1-1-2, 2-1-2-1, and 2-2-1-1. You have a one out of six chance of choosing correctly just by luck.
Please note: It's possible to cheat just by comparing the files sizes of these samples. Please resist that urge, or if you can't resist it, please don't answer the AppleInsider poll after cheating to get the answer.
For anyone interested, I also have an uncompressed 5-megabyte AIFF sample file available for comparison, identical to the original CD data except for the half-second fade-in and half-second fade-out at the beginning and end of the sample.