A chat with Aureal

I had the opportunity to speak with Aureal's Toni Schneider, who was kind enough to relate to me the details of the Vortex 2 and why they think that it's the Right Thing to do. I also talked to the folks at Creative Labs at Comdex, and Micah Stroud was able to fill me in on the details concerning EAX and the Sound Blaster Live!.

3D Positional Sound
Both the SBLive and the MX300 support 3D positional sound via 4 speakers. They also support positional audio with headphones or stereo speakers, the MX300 via A3D, and the SB Live through E-Mu environmental modeling. What each of these basically amount to is virtualized 3D through only two speakers. Aureal has always advocated the use of Head Related Transfer Functions, or HRTFs, and in speaking with Micah Stroud from Creative Labs at Comdex, he mentioned that the SBLive also uses HRTFs while in two-speaker mode. However, using the Speaker utility from Creative's AudioHQ, you can only audition 3D positioning while using 4-speaker mode; it was definitely a not a focus for Creative, and they pushed hard for the "desktop theatre" 4-speaker setup for surround.

The first question I wanted to know was: does virtualized 3D really work? This tends to be a very subjective issue, as HRTF algorithms are derived from tiny microphones placed inside the ear to analyze the effect on sounds coming from known locations. Since every ear is different, everyone perceives sound slightly differently, and to be completely accurate, HRTFs must be individually measured per individual. However, that not being possible, an HRTF from an "average" person should be accurate enough for most other people to perceive. So does it work? In short…sort of. On a pair of stereo speakers, it's slightly possible to perceive a sound travelling around your head at ear level, at about the same level of being slightly possible to fool yourself into thinking that's what you hear!

On a pair of decent headphones, it's a lot more clear, but at least to me, still pretty iffy. Toni from Aureal mentioned that in some cases, a pair of headphones can convey 3D positions more accurately than any speaker array. I was pretty skeptical. Accurately reproducing where a sound came from by sending it to a speaker in the approximate location was one thing, trying to trick the ear with a generalized, compressed mathematical algorithm is completely different. Either way, I tried it out. I used the included graphical applet and dragged a sound around my virtual head, with a little effort I actually could make out front and rear positions, but only when I was familiar with the sound, and if it stays in motion.

Vertical HRTFs need some work

From what I could make out (and this is a simplification), the HRTFs generally used phase-shifting (playing sounds slightly sooner from one speaker to simulate sounds closer to one side of the head) to expand the stereo image, which works great. For sounds originating from behind and on the vertical axis, various muffling filters are used to simulate sound passing by the outer ear (ever wonder why it's such an irregular-cone shape?), past the head, etc. So a steady sound source would appear to be occluded (aha!) by parts of the listener's own body parts. It's tricky, and it works, but it's not "real," as in it tends to be a learned response. I eventually learned to equate such filtered effects as sounds originating from "behind" rather than instinctively believing the sound was to the rear. That's not necessarily bad; if a game were to take advantage of the HRTFs, players would eventually be able to learn the filters and use their output, but it's more analogous to playing a different sound for something happening from behind rather than fooling the listener into thinking it's actually in the rear. Vertical sounds are even more dubious - I think seeing the graphical representation of the sound moving above or below my head was convincing me more than the sound filters itself (which sounded nothing more than very muffled from "below" as the sound supposedly passes through your body, shoulders, or earlobes). What this does mean is that there's still work to be done with 2-driver 3D.