Why do data compression processes (mp3, for example) strongly affect loud masters with no headroom?
All standard aural data compression processes use the masking effect which is part of human hearing. This involves eliminating data after loud sound events which the ear would filter out anyway, making it possible to redistribute energy in the data flow.
Take a recording from a CD with 0.5dB headroom to full scale that also has a very high loudness level (DR4 for example) and convert this to mp3. If you reconvert this mp3 file back into a Wave file (PCM), you will notice that the file has overs above 1dB (provided you have the appropriate measuring instruments). This reconversion is accomplished by any mp3 player when a file goes through the Digital-to-Analog converter for the headphone output. However, these converters cannot process signals over 0dB and will generate audible distortion as a result of the overs.
A rule of thumb is that the higher the degree of data compression (the lower the bit rate), the higher the potential for overs which produce distortion. Overs of up to 6dB are no rarity nowadays.
If you wish to have good sound quality with standard data compression formats compressed from loud releases, then lower the level of the file you are encoding by up to 6dB (the lower the bit rate, the more you should reduce the level).
Unfortunately nobody does this because songs are usually purchased in compressed formats. Even the iTunes software provides no setting which allows you reduce levels when converting a CD to mp3. Radio stations often ignore this, leading to a deterioration of sound quality even when archiving music for later broadcasting.
The Dynamic Range process of the Pleasurize Music Foundation eliminates this problem completely so that distortion and artifacts no longer occur when files are encoded into standard data compression formats.