Audio Definition

Although it's easy to visualize what spatial and temporal resolution look like for video, understanding audio resolution is a little different. Temporal resolution is relatively easy to understand, as it's represented in samples per second or Hertz (Hz). Generally kHz (kiloHertz are used since tens or hundreds of thousands of samples are taken each second. Besides samplerate, another factor called bit depth determines the accuracy of measurement or precision of reproduction. With a higher bit depth, a wider variety of sounds can be reproduced accurately.

Standard Definition Audio

SD audio consists of CD-Audio quality, sampled 44,100 times per second at 16 bits. Other SD audio sources include VCD and SVCD, which use the same sampling as CD-A, but also apply lossy compression. Standard DVD-Video (DVD-V) also uses SD audio, although it's sampled at a slightly higher rate of 48,000 samples per second - still at 16 bits. Lossy compression is normally applied to this audio as well.

Compression

There are essentially two types of audio compression, each of which is appropriate for different purposes. In order to get a basic understanding of them, it's first necessary to look at standard uncompressed audio, also known as LPCM or Linear Pulse Code Modulation. Pulse Code Modulation means that the magnitude of an analog sound wave is sampled at regular intervals. Linear refers to a linear scale being used for magnitude of the sample, which means the difference between 0 and 1 is the same as the difference between 127 and 128, which is the same as the difference between 254 and 255.

Lossy Compression

The easiest way to reduce the size of audio data is to apply psychoaccoustic compression - that is compression which removes information about sounds the human ear can't hear. This is a type of lossy compression, meaning you will always lose at least a little bit of quality. The advantage is a massive reduction in storage size. When any type of lossy compression is applied, audio is no longer considered high definition.

Lossless Compression

Lossless compression, the variety used for day to day computer file archives, is a very complex process. The key to lossless compression is something called RLE (Run-length Encoding), which involves finding repeated patterns of 1's and 0's and replacing them with an individual pattern of 1's and 0's that take significantly less space. Although LPCM encoded audio typically is somewhat more random than most computer files, the amount of compression that can be obtained is less than something like Zipping a text file. On the other hand, the more channels there are, the more compression can generally be achieved.

High Definition Audio

High definition audio typically has both a larger Data Word (sample) size, and a higher samplerate. The most common combination is probably 24/96, although Data Word sizes also include 20 and occasionally 32 bits. Common samplerates are 48kHz, 96kHz, and 192kHz, although some applications also use multiples of CD-A's 44.1kHz, including 88.2kHz and 176.4kHz.

DVD-Audio

Both DVD-V and its audio companion DVD-Audio (DVD-A) are capable of high definition audio, but since its not mandatory to support playback above 16 bits/48kHz in DVD-Video players, most downsample to that resolution and bit depth. Better players, particularly those that also support DVD-A playback, can reproduce 24/96 audio from DVD-V. DVD-A also includes 24/192 in one, two or three channels or 24/96 in up to 6 channels. It's also capable of CD-Audio samplerates, as well as 88.2kHz and 192kHz.

MLP

Meridian Lossless Packing is used on DVD-A to losslessly compress LPCM audio. Besides reducing storage requirements, it also reduces the audio bitrate, which is necessary to fit 5.1 lossless audio in a bitrate of 9,600kbps (DVD-A's limit).

Blu-ray and HD DVD

In addition to audio encoded with lossy compression, both HD DVD and Blu-ray support Dolby TrueHD, a more advanced version of MLP, and DTS Lossless, as well as uncompressed LPCM.

Acknowledgement

I'd like to thank Neil Wilkes for volunteering his professional knowledge to answer questions and point me toward some excellent reference sources.


Version History

v1.0 2007.10.31 Original version by Rich Fiscus