MP3 is considered one of the best and most widely used methods of compressing music. Although it is lossy, it is often said that the difference between compressed and uncompressed music is not imperceptible for humans. Even professionals do not seem to be able to distinguish MP3 from uncompressed music. Also because that, MP3 has become the standard on the market, eliminating most competitors. Almost all DRM-free portals offer downloadable music using the MP3 format. Other formats are more likely to be found in special stores or on physical media.

MP3 is not as good as expected, because of the psychoaccustic model.

To develop MP3, it was necessary to create a psychoacoustic model of human hearing. Listening experiments were held where the participants had to decide what they hear or not hear. The results were incorporated in the programming of the MP3 algorithm and told the algorithm how human ears works. So the algorithm could filter all things humans do not hear. By dropping these parts, the compressing is achived.

But MP3 has a small error, which is hidden in the development of the psychoacoustic model. But to understand this, you have to deal first with speaker technology.

The most commonly used speaker is the cone speaker. It has a permanent magnet and a coil through which the current flows. The coil moves in the magnetic field and causes the membrane to vibrate. The coil then moves the air to create the sound. The design has its disadvantages, the coil has a mass, is therefore sluggish and takes time until it begins to swing. As a result, music played on the speaker loses percussiveness. A bang is no longer a bang, but slowly swings in and out. The largest mass must be moved at the depths, the least at the heights. At a blast you hear the highs first and then they are gone for a long time before the lows are created.

Because of the mass of the coil the high frequencys are produced first, and are not in time with the deep frequencys.

This picture illustrates the delays caused by the mass of the cone speakers.

This shortcoming does not exist with all speakers. For example, the Manger Sound Transducer works almost without any moving mass. Thus, all the tones of a bang have the same timing, without the delay described. The speaker does this by simulating the human ear and spreading the waves on the membrane, which deforms itself and creates the sound this way. The quality that such speakers can deliver is different than that produced by cone speakers.

And this is the problem with MP3. In order to produce the psychoacoustic model, listening tests were made on cone speakers. The tested person does not hear a certain amount of music because of the construction of the loudspeaker. And this fraction is filtered and sorted out thanks to the measurements made by the MP3 algorithm - Frequencies that humans are able to hear.

You are unable to notice that as long as you hear music using cone speakers. Listening to MP3 on a Manger Sound Transducer, even the inexperienced ear immediately hears a difference to the original. So humans can hear the difference between compressed MP3 files and original music files.

This criticism is of course far away from daily listening environments, but it shows that MP3 does not quite live up to all of its promise, and there is a need for uncompressed music that is not part of musical production work.

Zimmermann, Dr. Christian, VL Human-Factors Engineering WS 2017/18 and SS 2018, LMU Munich
Hußmann, Prof. Dr. Heinrich, VL Digital Media 2016/17, LMU Munich

Comment this page



Can you guess the instrument?*


All Fields marked with * are required. Your data will be displayd together with your comment. You can find more information on Privacy


Before your comment is published we will first proof that it uses appropriate language. Please be patient to any delays of publishing.