Compressed Music Test Ready

As discussed in an earlier post, I've prepared a number of samples of tracks ripped from CDs at different bit rates. They range from 96 kbps to lossless, covering the full range of possibilities that one would use for classical music.

There are six tracks, covering the following instrumentations: solo harpsichord, solo piano, voice (baritone) and piano, string quartet, voice (soprano) and (small) orchestra, and orchestra. I've stuck with familiar works, from Bach to Mahler, and I think most listeners will know all the works in question. I've tried to cover a wide range of instrumentations so you'll have ample opportunities to hear the differences (if there are any) between uncompressed and compressed tracks with different types of sounds.

(Note: one way of seeing how much the music is really compressed is to look at the resulting bit rate when ripping tracks to lossless. Lossless compression works the same was as ZIP compression for text files; nothing is lost, but the data is compressed due to redundancies. The bit rates of the lossless files I created range from 464 kbps to 729 kbps; this shows you that the _real_ compression that _could_ result in audible differences is only from those lossless bit rates to the resulting compressed bit rate. Even the lowest level of "natural" compression removes half of the data with no quality loss. For example, an uncompressed CD is 1411 kpbs. If it is ripped to lossless, and results in a 500 kbps file, that means that it was reduced in size by 2/3 with _no loss in audible quality_. If it is compressed to an MP3 file at 256 kbps, only half of the original data is lost, and much or most of this data is inaudible (in theory).

I ripped the music in the following bit rates: 96 kbps, 128 kbps, 160 kpbs, 320 kpbs and lossless. I did this with iTunes, in AAC format (which is generally said to be better than MP3, especially at lower bit rates) then converted the tracks to WAV, which you can listen to in just about any music program on your computer, or which you can burn to CD. (You can set up the exceprts so all the samples of 3 pieces will go on each CD, though perhaps not in order.) Converting the files to WAV files hides the bit rate at which they were ripped, since WAV files are 1411 kbps. In other words, when they are "de-compressed" to WAV, "empty" data is used to fill up the spaces. After compression, any musical data "lost" cannot be retrieved, but the file format of WAV requires that the files be at 1411 kbps.

There is no indication in the WAV files what bit rate was used to create them. HOWEVER, you could cheat if you wanted to, performing an operation that would give you a rough idea of how much each file was compressed (or at least tell you which was the lowest bit rate and which was the highest). If you feel you need to do that, fine; but I'm pretty sure that I'll spot whether anyone did that from the results. But I think the interest here is to _not_ know, to try and gauge what you think sounds best, then to compare that with the actual information.

I've uploaded these files to a web server. There are two ZIP files, the first is 522 MB and the second 841 MB. This will take you a while to download, unless you have a very fast connection.

To download the files, go to http://public.me.com/kirkmc with any web browser. You'll need to enter a user name (public) and a password (musicweb), both lower-case, after which you'll access a page displaying the files available. Click the arrow icons to the right of the file names to download them.

Uncompress the ZIP archives, then either add them to a music program, such as iTunes, or use a disc burning program to burn them to CD. You'll note that the files are labeled Sample 1A, Sample 1B, etc. After listening, for each sample (1, 2, etc.), provide me with a list rated from 1 to 5, 1 being the highest quality you think you hear, and 5 being the lowest. Any comments you want to add are welcome (should you hear something specific in a file which, perhaps, gives you a clue as to whether it is highly compressed or not).

When you've tallied up your results, send them to me by e-mail (kirkmc@mac.com), with a subject line of "MusicWeb Test" (so I can make sure they don't get spam-filtered). I'll collate the results, and provide an enlightening report telling the true bit rates of the different files and showing how well you all did in hearing the difference in quality. Note that this is a group of people familiar with music, and who should be more likely to spot any degradation of sound that exists. The results should therefore be better than placebo (or, more correctly, better than if the ratings were given at random).

Have fun! I hope you all find this challenging!