Multi Format Audio Working Group (MFWG)

Status 11/15/99

The MFWG was formed to address the problem of taking audio authored in a variety of channel formats and playing it back on all the possible types of speaker configurations out there. We started out with with a chart developed at BBQ '98, in which several input formats of audio were matrixed to several output formats, with appropriate channel mapping instructions at each point in the grid. We began by examining and evaluating this chart, as well as some general comments that accompanied it. In the course of discussion, we explored some of the principles and commonalities inherent in the problem being addressed by the chart. As a result, the chart itself was distilled into a set of much more readable and understandable rules and procedures. Some other key points were also identified, such as that low frequency content would automatically be routed to the subwoofer by the speaker system, and that total signal power needed to be conserved during format conversion. The result is that the guidelines are now much simpler and more useful, without having lost any of the original problem solving value of the chart.

Assumptions:

• At any given time, there may be many input formats in use, but only one output format.
• The output format is selected by the user based on his/her speaker configuration. The channel format of the primary mix buffer is set to the same format as the selected speaker configuration.
• Sound card outputs are discrete channels, either analog or digital, and in the same channel format as the primary mix buffer and speaker configuration, with one exception: the L,C,R,S speaker array implies a post mix encode to Dolby Pro Logic stereo.
• At this time, it seems unlikely to use AC3 as an output format from the sound card due to the high cost of AC3 encoding.
• We assume that the LFE channel is the only source for the subwoofer in discrete 5.1 systems.

A Pitfall to Avoid

Mono 3D object --> 5.1 mix --> virtualized to stereo speakers.

To see why this is bad, imagine a gunshot positioned directly behind the listener. Mixed to 5.1 it would be present equally in LS and RS. If this were then virtualized there would be a separate HRTF performed on each of the two instances and the two HRTF’s would interfere with each other.

The moral is that if the ultimate destination is stereo speakers, the main mix should be done to stereo. Users should be instructed that even if they have stereo speakers with virtual 5.1, they should still select stereo speakers as their output format. By extension, there is no role in interactive applications for virtualization within the speakers.

Audio I/O Format Matrix