Downloadable Sounds

In the past, most computer music has been produced in one of two fundamentally different ways, each with its advantages and disadvantages:

• Waveforms are reproduced from digital samples, typically stored in a .wav file or, in the case of Red Book audio, on a standard CD track. Digital samples can reproduce any sound, and the output is very similar on all sound cards. However, they require large amounts of storage and resources for streaming.
• Instrument sounds are synthesized, usually in hardware, in response to messages, typically from a MIDI file. MIDI files are compact and require few streaming resources, but the output is limited to the number of instruments available in the General MIDI set and in the synthesizer, and can sound very different on different systems.

One way to combine the advantages of digital sampling with the compactness and flexibility of MIDI is wave-table synthesis—the synthesis of instrument sounds from digital samples. These samples are obtained from recordings of real instruments, and then stored on the hardware. The samples are looped and adjusted in such a way as to produce sounds of any length at various pitches and volumes.

Wave-table synthesis produces more realistic timbres than algorithmic FM synthesis, but is still limited to a fixed set of instruments. Moreover, a particular instrument can sound different on different pieces of hardware, depending on the manufacturer's implementation of that instrument.

To overcome these limitations, the downloadable sounds (DLS) standard has been published by the MIDI Manufacturers Association. DLS is a way of allowing wave-table synthesis to be based on samples provided at run time, rather than hard-wired into the system. The data describing an instrument is downloaded to the synthesizer, and then the instrument can be played like any other MIDI instrument. Because DLS data can be distributed as part of an application, developers can be sure that their soundtracks are delivered uniformly on all systems. Moreover, they are not limited in their choice of instruments.

A DLS instrument is created from one or more digital samples, typically representing single pitches, which are then modified by the synthesizer to create other pitches. Multiple samples are used to make the instrument sound realistic over a wide range of pitches. When a DLS instrument is downloaded, each sample is assigned to a certain range of pitches, called a region. Usually, there are no more than 16 regions.

In addition, samples can be given an articulation, which defines things like attack (how quickly a note reaches full volume), decay (how quickly it falls away from full volume), and other characteristics that make the sound more like that produced by a real instrument.

Downloadable sounds are stored in instrument collections, from which they are downloaded to the synthesizer.

DLS instruments are assigned patch numbers and respond to MIDI messages as do other MIDI instruments. However, a DLS instrument does not have to belong to the General MIDI set. In fact, it does not have to represent a musical instrument at all. Any sound, even a fragment of speech or a fully composed measure of music, can be turned into a DLS instrument.

For more information on DLS collections and how instruments are created, see the documentation for DirectMusic Producer. For a guide to incorporating DLS in your applications, see Using Downloadable Sounds.