The Idiots Guide to ATRAC

    This guide is aimed at those who want to get a very basic idea of what ATRAC is about. More technical based information can be found at the Minidisc Community Pages.

What is ATRAC and why is it needed?

    ATRAC is the name given to the compression system used on a minidisc and is an abbreviation for "Adaptive Transform Acoustic Coding". Because Minidiscs only hold one fifth of the information that a compact disc does, the compression is needed so that a full 74 minutes of music can be recorded without sacrificing much in the way of sound quality. If compression wasn't used, the maximum amount of music you would be able to store at CD quality, would be just under 15 minutes. Despite only one fifth of the information being stored on a minidisc, the difference in sound quality is amazingly small, and this is all down to the clever stratagies involved in the ATRAC system.

How does ATRAC Work?

    The basis of the ATRAC system lies within the assumption that a large amount of the audio information can be thrown away without it making any noticable difference to the listener. ATRAC employs many methods when deciding what to throw away, which will now be explained as simply as possible.

Part 1 - Frequencies.

    The ear is more sensitive to some frequencies than others. If a tone was played at a fixed volume level, say 40 decibels, and at different frequencies, it would seem to vary in volume, and there would be points at which it could not be heard. ATRAC takes this into consideration, and will concentrate more on areas where human hearing is more sensitive. This principle is often referred to as the "Psychoacoustic Principle". Precise measurements of the sensitivity of the human ear to different frequencies can be found by looking up "Equi-loudness Curves". To go into either of these in detail is far beyond the scope of this guide.

Part 2 - Masking (mice and dynamite).

    Masking is a very simple principle. Imagine you are in a room with some mice. If the room was completely silent, and one of the mice was to fart, you may just about hear it. Now imagine there is a stick of dynamite in the room and it explodes just as the mouse farts. The chances of you still hearing the fart are practically zero, owing to the fact that it has been drowned out by the mighty great explosion. This effect is called "masking". The explosion "masked" the sound of the fart. There is a little more to the system that just that though. Imagine the mouse farts just a fraction before the dynamite explodes. You may still hear a tiny bit of the fart, and may notice if it were to stop dead when the explosion started. The same may apply if the explosion finishes slightly after the mouse farts - you may hear the end of the fart. Get the idea? There's also other considerations such as frequency. Remember we said the ear is more sensitive at some frequencies? What if the quiet sound were at one of the most sensitive frequencies, and the loud sound were near the least sensitive frequency. There would be a chance that you may still hear the quiet sound over the loud one, because to your ears, the quiet sound would appear to be louder. By adopting these principles, ATRAC decides what will be masked, and throws that away too.

Part 3 - Compressing what is left

    Despite the clever way in which some of the sound information has been discarded, there is no difference in the actual physical storage requirements of the sound at this stage. The reason for making the various adjustments to the sound in the two methods described above is simple, it makes whats left compress much more efficiently. If you were to just compress the incoming sound without making some pre-judgements as to what is best discarded, the resulting quality would be much poorer, because the compression system would be storing information about sounds that cannot be heard at the expense of the sounds that can be. If there was a dog whistle on the recording, a large amount of the storage available would be taken up by recording the dog whistle, reducing the space left for any other sounds. By tailoring what is to be compressed to the human ear, you can assure that the compression system puts all its efforts into recording the most important sound elements.

    The compression system now embarks on a very complicated set of analysis routines, which I will outline very very simplisticly here. The compression system splits the incoming sound into various frequency bands. It them analysis the content of those bands, and splits them down again. Based on its analysis of those bands, it builds up a block of information which can be one of two pre-determined sizes, that will tell the decompressor how to recreate the original sound. This block of information is then written onto the minidisc.

Part 4 - Decompressing what is on the minidisc

    When the data on a minidisc is decoded, it takes the blocks of data from the disc, and uses the information contained in them to reconstruct the sound. Although most of the sound is reconstructed very accurately, the system cannot put back what was removed during parts 1 and 2 of the encoding process. Because the audible differences between the music at the start of the coding, and just before part 3 are so carefully calculated, the actual difference in the sound is incredibly small. Even though the data ended up only 1/5th of it's original size, it did not lose very much information at all.

What is the real impact of ATRAC on the sound?

    To most people, it's just about nothing. The sound difference is very very small indeed, and only people who sit there listening for the slight differences will even notice it. These are the type of people who claim they could still here the mouse fart during the explosion. Far more differences will be noticed between different brands of audio equipment than any differences to the sound as a result of ATRAC compression. It can be very difficult to tell the difference between a CD recording and a recording on MD. Note that even though a minidisc only holds 1/5th of the information compared to a CD, this does not mean that 4/5ths of the sound are lost, only that the actual physical amount of data stored on the disc is much less.