MLP Decoder / Encoder

Meridian Lossless Packing (MLP) is an encoding system designed to compress high-quality digital audio data with bit-for-bit accuracy. Unlike perceptual, or lossy data reduction, MLP guarantees not to alter the final decoded signal in any way, but merely packs the audio data more efficiently into a smaller data rate.

MLP encoding provides two main benefits

It minimises the size of the compressed data, allowing a larger amount of data to be stored in a given capacity
It can reduce the maximum instantaneous peak data rate

This second feature is important for DVD-Audio which has an upper limit of 9.6Mbps. Six channels of 96kHz 24-bit Linear PCM audio have a data rate of 13.824Mbps, which is well in excess of the capability of DVD-Audio. Also at this data rate the data capacity of the disk would limit the playing time to approximately 45 minutes. MLP encoding can reduce the worse case data rate to 9.6Mbps, in addition to extending the playing time to the industry norm of 74 minutes.

The following list gives examples of the typical compression that can be achieved with different audio sample rates and wordsizes:

Format Minimum Typical

48kHz, 16-bit 0% 50%
96kHz, 20-bit 40% 55%
96kHz, 24-bit 38% 52%
192kHz, 24-bit 43% 50%

The following list gives examples of playing times on DVD-Audio that can be obtained with different channel, sample rate, and wordsize combinations:

Format Playing time

6 channels, 96kHz, 24-bit 86 minutes
5.1 channels, 96kHz, 24-bit 100 minutes
2 channels, 192kHz, 24-bit 2 hours
2 channels, 96kHz, 24-bit 4 hours
2 channels, 44.1kHz, 16-bit 12 hours
1 channel, 44.1kHz, 16-bit 25 hours (talking book)

How it works

Because MLP was originally designed with consumer applications in mind, one of its design aims is that any complexity must be in the encoder rather than the decoder. The design of the decoder ensures that it will remain lossless irrespective of the hardware platform or processor it is implemented on. In addition, the decoder is driven by information in the bitstream, allowing improved versions of the encoder to be developed without becoming incompatible with the installed base of decoders.

MLP encoding uses a combination of three methods to reduce the data rate:

Lossless matrixing is used to reduce the correlation between channels
Lossless waveform prediction is used to reduce the inter-sample correlation
Entropy coding is used to reduce the data rate by efficiently encoding the most likely occurring values in the audio data.

MLP does not make any assumptions about the assignment of channels, or the correlation between channels, but takes advantage of whatever
redundancy is present in the overall signal to encode the data using the smallest possible bandwidth. In addition to these procedures for reducing the data rate, MLP uses stream buffering to reduce the variations in the transmitted data rates, and absorb transients that are hard to compress, in order to ensure a maximum instantaneous peak data rate. The buffer allows the peak data rate to be minimised for virtually all practical audio data.
If the audio data cannot be compressed within the specified peak data rate the MLP Encoder will signal an error. The producer can then use one or
more options for reducing the data rate, or reducing the total space used by the recording. These include

Reducing the bit width of one or more channels, such as from 24-bit to 22-bit
Filtering one channel to LFE
Reducing the audio bandwidth

All of these options will increase the amount of compression that MLP can achieve, thus increasing the playing time or reducing the peak data rate.

Two-channel Mixdown

Content providers will often want to make a two-channel version of a multi-channel audio stream available on a DVD-Audio disk, for consumers
who only have a two-channel playback device. One option is to create separate multi-channel and two channel streams, and write these
separately to the disk. However, this requires two separate mastering and authoring processes, and uses disk capacity. MLP provides an elegant and simple solution to providing a two channel mixdown. The encoder includes lossless matrixing, which can encode a twochannel mixdown as a linear combination of the multi-channel mix and encode this alongside the multi-channel version on the DVD-Audio disk. The advantage of this approach is that the producer can listen to the mixdown at the encoding stage in the knowledge that it will be delivered bit-for-bit to the end user at the decoding stage. Another advantage is that a two channel only playback device does not need to decode the multi-channel stream, and need only decode the stereo.

Other features

In addition to audio the MLP stream can carry hierarchical metadata, which can include

Dynamic range control data
Ownership and copy protection fields
Time codes
Descriptive text fields.

In addition MLP has powerful built-in error detection that allows rapid recovery from bit-stream errors, and prevents any erroneous noises, clicks,
or bangs following data errors.