Wednesday, December 7th 2005

Testing 128kbps Audio is No Piece of Cake

Over at Hydrogenaudio.org a public listening test of audio encoders at 128 kbps (VBR) has been launched. The goal of the test is simply (?) to try to establish which of the actively developed encoders gives the best audio quality at the selected quality range.

They need as much help as they can get, so go help them out if this kind of thing interests you. The contestants are…

The test goes like this: You download the test package which includes the program you use for the blind test (ABC/HR) and a bunch of configuration files for that program. Then you download some or all of the samples.

There are 18 samples in all, and they come in packs which include one encode for each encoder, as well as the original. The encodes are inflated into wave files using batch files included in the main test package. After that, just get to it. Using ABC/HR, load up the test configuration file corresponding to the sample you wish to start with. Now comes the difficult part: Identify the encodes (as opposed to the originals) and rate them on a scale from 1.0 to 5.0. The scale corresponds to the values very annoying, annoying, slightly annoying, perceptible but not annoying and imperceptible.

This is pretty difficult (or impossible, if you lack the patience) because the quality of modern audio encoders such as Ogg Vorbis and AAC is a lot higher than say your average 128 kbps CBR MP3. One big difference is that Ogg Vorbis & co by default vary the bitrate along the song as the complexity of the song varies. This feature is also available for MP3 encoders such as LAME.

It’ll be truly interesting to see the outcome of this test, which ends on Christmas Day, since I’m using Ogg Vorbis @ Q4 (~128 kbps) for my rips at the moment. Maybe I’ll have to* switch default encoder?

*) Have to since I’m a DAN.

Comments are closed.