How does MLP work?
The first concept MLP uses is this: audio that is of interest to the human listener contains some redundancy. Second, in music signals, the information content varies with time, and each input channel’s bandwidth is rarely fully used. The MLP encoder uses a variable data rate on most audio content, reducing both the average data rate (i.e. compressed file size) and the instantaneous peak data rate (for high sampling rates such as 96kHz or 192kHz). MLP tackles this by forcing the audio data into two substreams and 6 channels and by maximizing compression at all times using three basic techniques: • Looking for “dead air” – channels that do not exercise all the available word size, and channels that do not use all the available bandwidth. • Removing inter-channel correlations (similar data on more than one channel), and efficiently coding the remaining information. • Smoothing the transfer of high-bandwidth information with buffering. MLP accomplishes these tasks by using a series of dat
