Documentation      Speaker Positions       Examples      Extensions

This is the new native format for Windows2000, the Kernel Audio services from which have been incorporated into Windows98 Second Edition.
WAVE-EX was defined to meet a number of objectives:

• eliminate ambiguity regarding sample types and container sizes in the standard WAVE format
• support surround sound and general multi-channel streaming by defining speaker positions beyond mono and stereo
• create a core format that can be freely extended  without the need to register the extensions with Microsoft
• create a format that can be handled efficiently by 64bit platforms

The format supports both integer (WAVEFORMATPCMEX) and floating-point (WAVEFORMATIEEEFLOATEX) samples.
 

Documentation

Full details of the specification can be read  here.

A paper (Word 97 format)  by Dave Rossum (Creative Labs), 'An Integrated Approach to Multi-Channel Audio', presented at WinHEC 99, is well worth reading, for a broad perspective on the many multi-channel paradigms, argued from the perspective of a 'virtual' recording studio, as might be implemented in a modern Audio Workstation. It includes a reference to WAVE-EX.
 

Speaker Positions.

Currently 18 positions are defined. The first twelve correspond to those defined by the USB 1.0 protocol. The remainder provide a reasonably complete set of  high and overhead positions. In particular, a standard B-Format (Ambisonic) cube speaker array can be directly represented. Future versions of the format may define further speaker positions.
 
 

WAVE-EX Speaker Positions

FRONT_LEFT

FRONT_RIGHT

FRONT_CENTER

LOW_FREQUENCY

BACK_LEFT

BACK_RIGHT

FRONT_LEFT_OF_CENTER

FRONT_RIGHT_OF_CENTER

BACK_CENTER

SIDE_LEFT

SIDE_RIGHT

TOP_CENTER

TOP_FRONT_LEFT

TOP_FRONT_CENTER

TOP_FRONT_RIGHT

TOP_BACK_LEFT

TOP_BACK_CENTER

TOP_BACK_RIGHT

The new format is closely associated with the audio rendering subsystem in Windows2000 and Windows98 Second Edition.
The following facilities are available to all applications, independently of the specification of the soundard(s):

• The channels sent to the renderer (generated internally, or taken directly from a source WAVE-EX file) are mixed down to the number of channels supported by the hardware device.
• floating-point data can be sent directly to the hardware; it is not necessary to convert to integer first.
• if required, samples will be dithered prior to rendering, using a Triangular Probability Density Function (TPDF). This serves to preserve signal to noise ratio which can be degraded by quantization noise when, for example, a 24bit sample is reduced to 16bits.

Note that plain multi-channel files (WAVE and AIFF) can be played directly (e.g in Windows95, using MediaPlayer) if the soundcard's driver offers a true multi-channel 'device'. See Richard Dobson's Soundcard Attrition Page for more information.

Examples.

Here are two quad soundfiles in the new format. We hope to post more examples soon, and also that other sites will appear, as other developers complete the software required to handle the new format. CDP has published a free Multi-Channel Toolkit (WIN32 console), which includes a program to play these files using a multi-channel MME device, and a program to create the quadraphonic pan illustrated below.
 
 

Example 1: from Microsoft
Speech idents, in LCRS surround format(44100Hz,16bit, 3.83 secs)
Wave_ex1.zip (291KB)
 
 

example 2: created using the CDP Multi-Channel Toolkit (illustrated below)
Drum, panned ambisonically in a circle in quad format (LF,RF,LR,RR) (22050Hz,16bit, 4.78 secs)
(also contains the PEAK chunk)
Drmapan.zip (660KB)

Example Extensions
 

• B-Format (Ambisonic)
• Phase vocoder data (PVOC-EX)

(Return to NOS-DREAM home page)