Next Generation Audio for Gaming
November 13, 2013, AMD Developer Summit, San Jose, CA—Jerry Mahbub from Gen Audio
Sound has evolved from monophonic, to stereo, to various surround, mostly now at 5.1 audio channels. Some new experimental systems are going to 11.1, 13.1, and even 22.3, and use very complex mixing tools and hardware to get to their spatial sound management. One big issue is how to get these very complex systems into the hands of consumers without creating overwhelming scin=ence projects for the installation.
One key is to change the focus from just physics to psychoacoustics. This change calls for a head relative tracking function and the ear-brain response transfer function. This brain inclusion makes most measurements subjective without advanced tools like fMRI and EEG to confirm hypotheses that all people hear in the same way.
Research for over 20 years created sets of fMRI images and time dependent data to identify the full set of responses. these data can be simulated with computational physics and numerical analysis to provide localized sound sources. The initial implementation used finite impulse filters (FIR) which were replaced with infinite impulse response (IIR) filters to reduce the memory requirements.
One result is a real-time interface to gaming. This true 3-D audio meets the low audio-video latency requirements and is dynamic, capable of panning and real-time relative positioning. The solution is easily integrated into existing systems and requires no special setup nor calibration.
Configurations include mono in to spatialized stereo out, multi-channel in to 5.1 or 7.1 stereo out, and stereo in to spatialized stereo out. The spatial interpolation can handle over 100 filter points and provides audio positioning in XYZ or R?? coordinate systems. A distance control adjusts reverberations to level out sounds.
The tools are compatible with the major sound development tools for games like WWise and Fmod. The sound sources and XYZ coordinates are transformed to provide a full 3-D audio image that can track the player’s position and heading. Developers can set up sources to identify their location to the system, and automatically map source effects.
Distance cues are indicated through variable reverberation through a scale of 1-100 in the game. The system is fairly efficient, 10 voices use less than 10 percent of a quad core i7, and with the AMD true audio co-processor this drops to zero. The ACP uses less than 40 percent of the capacity for 10 voices. Developers use a vitualized mix-down to virtual speakers to add the location information through a WWise API for a true 3-D stereo audio image.
