Newton Developer Technical Information: Newton Programmer's Guide: 2.1 OS Addendum /

Chapter 7 - Sound / About Sound

Sound Compression

Under previous versions of the Newton OS, sound was not compressed. The Newton 2.1 OS now supports recording and playback of compressed sounds. The following types of sound compression are supported:

• muLaw. This is a standard compression technique that is often used to compress voice data. It compresses each 16-bit sample into 8-bits. Rather than simply throwing away the least significant 8 bits, it instead saves fewer bits from each sample, but it shifts them to save the important ones. This tends to preserve the dynamic range better than truncation.
• IMA (Interactive Multimedia Association). This is another popular compression technique that is especially good for voice data. It converts a frame of 64 16-bit samples to a 34-byte frame. It compresses each 16-bit sample to 4 bits, and uses 2 additional bytes as decompression information. This technique works better for music than muLaw or GSM.
• GSM (Global System for Mobile-Communications). This is a standard compression used for cellular phone data in Europe. It converts a frame of 160 16-bit samples to a 33-byte frame. It is a mathematically intensive compression technique that works effectively only on devices containing a fast CPU, such as the MessagePad 2000. GSM compressed data doesn't sound quite as good as IMA, but it is less than half the size.

These types of sound compression are implemented by built-in codecs (compressor-decompressors). Using a codec, the system plays a compressed sound using a two-step process. The sound data is decompressed into an intermediate buffer by the codec, then the buffer is scheduled and played like an uncompressed sound. Similarly, during recording, sound sample data is recorded into a temporary buffer first and then compressed into the binary object where it is to be stored.

With codecs, intermediate buffers are used to avoid skips in the sound that might be caused by the system trying to access the binary object, which is typically a virtual binary object (VBO). This is in contrast to uncompressed sounds which are played directly from the sound object, or recorded directly to the binary object, with no intermediate buffering.

For details on how to compress and uncompress sounds, see "Compressing Sound".