img CanFly

The Science of JavaTM Sound Technology

发表于2001/11/29 9:30:00  2035人阅读

分类: java



The Science of JavaTM Sound Technology

By Michael D. Meloan
(June 1999)

Sound creates mood, triggers memories, and in conjunction with visual imagery weaves whole worlds of fantasy. Sound is the cornerstone of multimedia content. That's why Sun's JavaTM Media team is busy readying the Java Sound 1.0 API for inclusion with the next release of the Java Development Kit (JDKTM).

Prior to the advent of the Java 2 platform, the language handled only telephone quality sound, in the form of µ-law AU files recorded in mono at an 8 kHz sampling rate. The Java 2 platform adds support for AIFF, WAV and three types of MIDI formats. The supported MIDI formats are Type 0 MIDI, Type 1 MIDI, and RMF.

The 1.0 API will offer a comprehensive toolkit for programmers to access the capabilities of the underlying synthesis and rendering engine, and to expand the applications of Java Sound. The two primary areas of emphasis are digital audio and MIDI. Extensive low level support functions will be available so that programmers can input and output sound, control MIDI instruments, and query system functions.

A Sound File Primer

The major sound file types are defined as follows:

AU - (.AU or .SND) short for AUdio—a common format for sound files on Solaris and Next machines, and also the standard audio format for the Java platform. The three audio formats typically used for AU files are: 8-bit µ-law (usually sampled at 8kHz), 8-bit linear, and 16-bit linear.

WAV - (.WAV) — developed jointly by Microsoft and IBM, support for WAV was built into Windows 95 and is carried over to Windows 98. WAV files can store a variety of formats including µ-law, a-law and PCM (linear) data. They can be played by nearly all Windows applications that support sound.

AIFF - (.AIF or .IEF) Audio Interchange File Format is the standard audio file format for both Macintosh computers and Silicon Graphics (SGI) machines. AIFF and AIFF-C are almost identical, except the latter supports compression such as µ-law and IMA ADPCM.

MIDI - (.MID) Musical Instrument Digital Interface is the standard adopted by the music industry for controlling devices such as synthesizers and sound cards. MIDI files do not contain digital audio samples, instead they consist of instructions for synthesizing songs from a sequence of notes coming from various instruments. Some MIDI files contain additional instructions for programming a variety of synthesizer settings.

Most synthesizers support the MIDI standard, so music created on one synthesizer can be played and manipulated on another. Computers that have a MIDI interface can manipulate MIDI data to produce new music or sound effects. For example, a whole musical composition can be transposed to a new key with one software driven command.

The Java Sound engine supports two MIDI> file types:

  • MIDI Type 0 files—contain only one sequence where all instrumental parts are contained on the same logical "track."

  • MIDI Type 1 files—contain multiple "tracks" so that various instruments are logically separated for easier manipulation and reproduction.

RMF - (.RMF) Rich Music Format is a hybrid file type developed by Beatnik that encapsulates MIDI and audio samples along with interactive settings. RMF acts as a container for all music related files. Detailed documentation of copyright information is also facilitated by RMF. An RMF file might contain MIDI compositions and audio samples by different artists, each with associated copyright information.

The Search for a Sound Engine

In 1997, Sun's Java Media team was looking for way to upgrade sound capabilities while providing a solid platform for future growth. According to Java Media manager Michael Bundschuh,

"We wanted a very high quality playback engine handling everything from eight bit µ-law up to CD quality sound. We wanted portability across platforms such as the SolarisTM platform, Windows, Mac and others. We also wanted a highly developed MIDI capability across the various platforms. Using those criteria, the logical choice was the Beatnik Audio Engine (formerly Headspace.)"
Sun licensed the Beatnik Audio Engine as the basis of the sound engine used by the Java Sound API.

Thomas Dolby Robertson—Beatnik

Before the early 1990s, Thomas Dolby Robertson's career was strictly musical. His 1982 hit "She Blinded Me with Science," was an early MTV blockbuster. He continued to compose and record throughout the 80s, using a variety of off-the-shelf music generation software. But in 1990, while collaborating on a virtual reality display for the Guggenheim Museum, he began to think in terms of improving what was available.

"I was leaning over a C programmer's shoulder, and I suddenly realized that there was great stuff out there for making records, but nothing for run-time interactivity." With that idea in mind, Robertson founded Headspace in 1992, hiring moonlighting coders to help realize his vision. Headspace incorporated in 1996, and is now known as Beatnik.

The Java Sound Audio Engine

The Java Sound engine was created for multimedia with game design and web content in mind. Using standard MIDI files, RMF files, and/or samples from any source, the engine will play music or sound effects with minimal CPU impact. It offers complete playback control with the ability to mix sounds and respond in real time to user input.

The Java Sound engine is a software MIDI synthesizer, sample playback device, and 16-bit stereo mixer. It supports mixing of up to 64 stereo MIDI voices and audio samples. MIDI Type 0 and Type 1 files are directly supported and wavetable synthesis from 8-bit and 16-bit instruments is provided. The engine supports all General MIDI controllers and includes features like reverb, LFO (for controlling sweeping filters or stereo placement), and ADSR envelopes (for shaping samples as they are played.)

Even with all functions enabled, the Java Sound engine will stay below 30% CPU usage on a 90 MHz Pentium computer. It can be made even more efficient by selectively disabling unneeded features. In addition, it delivers rich content stored in compact RMF music files. Thomas Dolby Robertson's "She Blinded Me with Science," a 7 minute 21 second song, would take roughly 76 MB to store as a CD quality sound file. Storing this song in RMF format reduces the file size to approximately 636 KB, a 120:1 reduction, with impeccable playback quality.

For more Beatnik information see, go to the Beatnik web site.

A Brief History of Java Platform Sound

Under JDK 1.0.x and JDK 1.1.x, the AudioClip interface offered the following functionalities.

  • AudioClip interface
    • play
    • loop
    • stop

The simplest way to retrieve and play a sound is through the play() method of the Applet class. The play() method takes one of two forms:

  • play()—with one argument, a URL object, loads and plays the audio clip stored at the URL

  • play()—with two arguments, a base URL and a folder pathname, loads and plays that audio file. The first argument most often will be a call to getCodeBase() or getDocumentBase().

The following code snippet illustrates a direct way to play hello.au . The AU file resides in the same folder or directory as the applet.

  play(getCodeBase(), "hello.au");

The play() method retrieves and plays the sound as soon as possible after it is called. If the sound file cannot be found, there will be no error message, only silence.

To start and stop a sound file, or to play it as a loop, you must load it into an AudioClip object by using the applet's getAudioClip method. The getAudioClip method takes one or two arguments, as indicated above for play(). The first, or only, argument is a URL argument identifying the sound file, and the second argument is a folder path reference.

The following code line illustrates loading a sound file into a clip object:

  AudioClip co = 
    getAudioClip(getCodeBase(), "horns.wav");

The getAudioClip() method can only be called from within an applet. With the introduction of Java 2, applications can load sound files by using the newAudioClip method of the Applet class. The previous example can be rewritten for use in an application as follows:

  AudioClip co = newAudioClip("horns.wav");

After you have created an AudioClip object, you can call the play(), loop(), and stop() methods using the object. If the getAudioClip or newAudioClip methods can't find the indicated sound file, the value of the AudioClip object will be null. Trying to play a null object results in an error, so testing for this condition should be standard procedure.

The complete programming sample below generates an applet which will play a flute+hrn+mrmba.au music sample when the mouse is pressed inside the applet. The AU sample file is in the same directory or folder as the applet.

  import java.applet.*;
  import java.awt.event.*;
  public class PlayAudio extends Applet
          implements MouseListener {
          AudioClip audio;
  public void init() {
          audio = getAudioClip(getDocumentBase(),
  public void mousePressed(MouseEvent evt) {
          if (audio != null) audio.play();
  public void mouseEntered (MouseEvent me) {
  public void mouseExited  (MouseEvent me) {
  public void mouseClicked (MouseEvent me) {
  public void mouseReleased(MouseEvent me) {

Note: The mouseDown() method included in many Java 2 technology books is actually part of the Java 1.0 event model. Using this method will result in warnings that the method is deprecated, meaning that it may not be supported in the future. Use of MouseListener in conjunction with the mousePressed method is preferred under the Java 2 platform.

API 1.0—Quantum Leap

Note: The following comments are based on the 0.86 early access version of the Java Sound API. While most of the objects and concepts discussed here will remain the same, it is possible that changes may occur as the API is finalized.

The Java Sound 1.0 API defines a comprehensive set of basic low-level audio capabilities for the Java platform. It provides interfaces for:

  • Audio capture and rendering
  • MIDI synthesis and sequencing

These two major modules of functionality are provided in separate packages:

  • javax.media.sound.sampled—This package specifies interfaces for capture, mixing, and playback of digital (sampled) audio.

  • javax.media.sound.MIDI—This package provides interfaces for MIDI synthesis, sequencing, and event transport.

The 1.0 API will feature the following capabilities:

Digital Audio

  • Audio capture - data capture from input sources such as microphones.
  • Mixing and playback - mix and playback sound from various sources
  • Controls and codecs - adjust the gain, pan, reverb, etc., and apply format conversions
  • Status and notification - receive events when playback starts and stops, when a device opens or closes, and other related events

MIDI Support

  • MIDI messaging - exchange messages (note on, note off, etc.)
  • Synthesis - load instruments and generate sound from MIDI data
  • Sequencing - load a MIDI sequence, start and stop playback, and adjust tempo


  • File I/O - read and write common audio file types such as WAV, AIFF and MIDI
  • Configuration - query system for information about components and devices; install and remove codecs, file parsers, and devices

Digital Audio


"Channel is the basic functional unit in the audio pipeline," says Kara Kytle, lead engineer and architect for the Java Sound API. A class that implements the Channel interface represents an element of this "pipeline," such as a hardware device, an audio mixer, or a single audio stream.

InputChannel and OutputChannel extend Channel with interfaces for reading captured data or writing data for playback, respectively. The Clip sub-interface supports looping and repositioning within its pre-loaded audio data. Device represents any hardware or software device for capturing, playing, or mixing audio.

This interface hierarchy is represented in the diagram below. Package reference: javax.media.sound.sampled

Interface Inheritance Diagram
Interface Inheritance Diagram

When a Channel opens, it reserves system resources for itself, and when it closes, those resources are freed for other objects or applications. The isOpen() method queries whether a Channel is open or closed. Processing of data is typically initiated by subinterface methods such as read(), which is described in the Interface InputChannel spec (see Java Sound API Spec.)

Processing methods engage the Channel in active input or output of audio data. A Channel in this state can be identified through the isActive method. The Channel can be paused by invoking pause(), and relevant Channel status can be queried using the isPaused() method. When a Channel is paused, there are three options: retaining the data (the default), discarding the rest of the data in the internal buffer using flush(), or causing the rest of the buffer to be processed immediately using drain().

An object may register to receive notifications whenever the Channel's state changes. The object must implement the Channel.Listener interface, which consists of the single method update(). This method will be invoked when a Channel opens, closes, starts, and stops. A Channel produces start and stop events when it begins or ceases active presentation or capture of data.


InputChannel is a source of captured audio data. The interface provides a method for reading the captured data from the InputChannel buffer and for determining how much data is currently available for reading. If an application attempts to read more data than is available, the read method blocks until the requested amount of data is available.


OutputChannel accepts audio data for playback. This interface provides a method for writing data to the OutputChannel buffer for playback, and for determining how much data the channel is prepared to receive without blocking. If an application attempts to write more data than is available, the read method blocks until the requested amount of data can be written.


The Clip interface represents a special channel into which audio data may be loaded prior to playback. Because the data is pre-loaded rather than streamed, clips support duration queries, looping, and repositioning.


The Device interface provides methods for classes that represent audio devices. An audio device is a shareable or exclusive-use system resource, one that can be based in hardware, software, or a combination of the two. It can be opened and closed repeatedly, often reporting its inherent latency and supporting a set of audio formats. It also provides an info object that describes the Device.

The Java Sound API further describes three device sub-interfaces:


The InputDevice interface provides a method (getInputChannel) for obtaining an InputChannel from which captured audio data may be read.


The OutputDevice interface provides a method (getOutputChannel) for obtaining an OutputChannel to which audio data may be written for playback.


A Mixer supports multiple InputChannels and/or Clips. In addition, it supports methods for querying the total number of channels it supports and for pausing and resuming playback of multiple channels synchronously.


Channels and audio ports (such as the speaker and the microphone) commonly support a set of controls such as gain and pan. The Java Sound API's channel objects and port objects let you obtain a particular control object by passing its class as the argument to a getControl() method.


Codecs encode and decode audio data, allowing translations between different formats and encodings. The Java Sound API provides high-level interfaces for these translations through methods in the AudioSystem class. Given a particular audio stream, an application may query the audio system for supported translations, and obtain a translated stream of a particular format.

Files and Streams

An audio stream is an input stream with an associated audio data format and data length. A file stream is an input stream with an associated file type and data length. The Java Sound API provides interfaces for translating between audio files and audio streams in the AudioSystem class.

Querying and Accessing Installed Components

The AudioSystem class acts as the entry point to sampled audio system resources. This class allows programmers to query and access input devices, output devices, and installed mixers. In addition, AudioSystem includes a number of methods for converting audio data between different formats. It also provides methods for obtaining an input channel or output channel directly from AudioSystem without dealing explicitly with devices.

System Configuration - Service Provider Interface (SPI)

Configuration of the sampled audio system is handled in the javax.media.sound.sampled.spi package. Through the AudioConfig class methods, devices may be installed or removed from the system, and defaults may be established. Service providers may wish to provide and install their own codecs and file parsers. Mechanisms for accomplishing this are provided in the package.

The diagrams below describe functional flows for audio input and output.

Typical Audio Input System

Typical Audio Input System

Typical Audio Output System

Typical Audio Output System


Interfaces describing MIDI event transport, synthesis, and sequencing are defined in the javax.media.sound.MIDI package. The major concepts used in the package are described below.


The basic MIDI transport interface is MidiDevice. All devices provide methods for listing the set of supported modes and for querying and setting the current mode. Devices support listeners for events such as open and close, and an info object describes the device.

Generally, devices are either transmitters or receivers of MIDI events. The Transmitter interface provides methods for setting and querying the receiver to which they currently direct MIDI events. Receiver supports a method for receiving MIDI events.

The basic MidiEvent object specifies the type, data length, and status byte associated with a message. It also provides a tick value used by devices involved in MIDI timing, such as sequencers.


The Synthesizer interface is a special type of Receiver that generates sound. It provides methods for manipulating soundbanks and instruments. In addition, a synthesizer may support a set of global non-MIDI controls such as gain and pan. It also provides access to a set of MIDI channels through which sound is actually produced.

The MidiChannel interface supports methods representing the common MIDI voice messages such as noteON, noteOff, and controlChange. Query of the current channel state is also supported.


The Sequencer interface extends MidiDevice with methods for basic MIDI sequencing operations. A sequencer may load and play back a sequence, query and set the tempo, and control the master and slave sync modes. An application may also register for notification when the sequencer processes meta and controller events.

Files and Sequences

The Sequence object represents a MIDI sequence as one or more tracks and associated timing information. A Track object contains a list of time-stamped MIDI events.

The Java Sound API provides high-level interfaces for converting between MIDI files and Sequence objects in the MidiSystem class.

Querying and Accessing Installed Components

The MidiSystem class acts as the entry point to the MIDI music system. It provides information and access relating to the set of installed devices, including transmitters, receivers, synthesizers, and sequencers. It also provides access to SoundBank objects.

System Configuration - Service Provider Interface (SPI)

Configuration of the MIDI system is handled in the javax.media.sound.midi.spi package. Through the MidiConfig class methods, devices may be installed or removed from the system, and defaults may be established. Service providers may wish to provide and install their own file and soundbank parsers. Mechanisms for accomplishing this are provided in this SPI package.

Now let's see what can be done with all these classes and methods. Check it out!

ToySynth Application

ToySynth Application

The ToySynth application exercises the Early Access Java Sound API by offering a variety of instrument settings, MIDI channel selection, volume control, stereo pan, reverb, and other options. Clicking on the keyboard plays notes using the selected instrument rendered using the Java Sound engine.

Complete ToySynth.java code example.

Download the Early Access Java Sound API with sample applications and code listings.

Target Markets for the Java Sound API

The Java Sound API provides audio support for a wide variety of applications. Some of the possibilities are enumerated below.

Communication Frameworks
  • Conferencing
  • Telephony

Content Delivery Systems

  • Music
  • Streamed content
  • Media Players

Interactive Applications

  • Games
  • Web sites
  • Dynamic content generation

Tools and Toolkits

  • Content generation
  • Utilities

Where Do We Go From Here?

Java 2 offers access to the Java Sound audio engine via the AudioClip interface. With the release of the 1.0 API, the world of possibilities we've just explored will be available to developers everywhere.

"Currently, we are enabling multimedia in the computer desktop market by adding true sound support in the Java 2 platform. In the future, we would like to see our Java Sound API technology used in professional, consumer and Internet audio applications," says Michael Bundschuh.

Movie and record companies need professional quality audio to use the Java Sound API technology. Implementing support for 24-bit audio and multi-channel configurations in the Java Sound Engine will encourage the development of professional editing and playback applications.

According to Kara Kytle, extending support for a wider variety of audio data types such as MP3 is on the agenda. "Another agenda item is MIDI data capture from an external device. That will come soon." she says.

"Java Sound API technology is already well suited for audio distribution over the web," says Bundschuh. "But we're always tracking new and developing technologies such as MP3 and secure delivery of music files. We plan to support these technologies in future Java Sound API releases."

Recent marketplace developments, like the competing strategies for delivering web-based musical content, have put digital sound at center stage. As new media technologies continue to evolve, sound will play a key role. The Java Sound 1.0 API will arrive on the scene at the perfect time to participate in the development of new killer apps.

Code is available for ToySynth.java listing

This application brings up a graphical representation of a typical synthesizer keyboard. Clicking on the keys with the mouse, plays the notes. You can adjust voice settings and other options to vary the sounds produced. Have fun!

Reference URLs

Java Sound API - Spec, Early Access Download Info, Data Sheet, FAQs


Java Media Framework API (JMF)

Java 3D API

Reference Texts

Java 2 Platform by Laura Lemay and Rogers Cadenhead

Using Java 2 Platform by Joseph L. Weber



0 0



取 消