ALive! - features - Dave Rossum

3D Audio - Who's That Behind Me?

In the January issue of Maximum PC, you stated that Creative has adopted the perceptual approach towards 3D audio while Aureal prefers to go the actual physics-modeling route. Yet recent feedback on the Web and magazines has indicated that Aureal's A3D 2.0 gives better positional audio, so is the approach wrong?

Realize first that "positional audio" is the phenomenon which Aureal has recently been favorably compared. One of the big points of Environmental Audio is that experience of spatial audio goes far beyond simply identifying the coordinates of objects. In particular, allowing artistry, rather than being mechanistic about sound production, was a very basic driving factor in leading us to the perceptual approach we took. Our approach has not only diverged from Aureal’s in that we adopted a perceptual paradigm, but also that we chose to prioritize establishing the environment over perfecting our positional coordinate model. Consequently, when compared on "positional audio" without the environmental effects in use, our early releases don’t always sound as compelling. Through the Live!Ware program, we are introducing improved positional algorithms that address these issues. Fortunately, since the API for positional audio is fixed through DirectSound 3D, SB Live! users can gain benefit from these changes when using products supporting EAX1.0 or even just DS3D.

To answer your question directly, no, we are quite certain the perceptual approach we have taken is superior, and that the marketplace will ultimately agree. However, it’s important to distinguish between the positional audio implementation and the benefits of the fundamental approach to the environmental API.

Is the weakness of the 3D audio, in particular z-axis positioning, on the SB Live! attributable to the DS3D API or the rendering engine? Can we see better performance with both headphones and speakers in the near future or are we limited to what DS3D can do?

The DS3D API has no problem in Z-axis positioning; any weakness must be due to the rendering engine, the physical configuration, or the listener. As I said above, our priorities have been on environmental effects rather than ultimate positional audio, so we have many improvements now and in future releases for positional audio effects like elevation. But I feel it’s always necessary to set user’s expectations. Speaker implementations have substantial limitations, and even headphones without head-trackers are quite limited in their positional audio capabilities. So while DS3D won’t limit us, some users may be disappointed in the ultimate performance given their speaker configuration.

E-mu is a renowned music company that drives the Sound Blaster product line with their chips while Aureal had been stressing that they have 10 years of 3D Audio research and development behind them (NASA being a frequent example of "ultra-high" end application of their techonology). The question is, does E-mu have the necessary research and expertise to challenge Aureal in 3D audio technology?

When Aureal acquired Crystal River, they inherited the expertise of Scott Foster, one of several early pioneers in 3D audio. Scott has been a friend of mine for years, and NASA did indeed sponsor development of his "Convolvotron," which was a first generation, brute force, experimental virtual reality engine which could support at most four sources. While this was cutting edge stuff at the time, it’s horribly obsolete by today’s standards.

While I doubt there’s much in common between the Convolvotron and Vortex technology and algorithms, I think Scott’s early "brute force" approaches may have had too strong an influence on Aureal’s outlook. In the A3D2.0 demos I’ve seen, the acoustical modeling seems to always use a separate source for each reflection. This completely fails to reproduce any late tail in the decaying sound, which I feel is the most perceptually important part. By trying to leverage Scott’s work on 3D positional audio into the rendering of the environment, they’ve revealed how little they understand of actual acoustical modeling.

While it’s true that E-mu has 25 years of experience supplying equipment for sound designers in Hollywood, we really didn’t develop much classical 3D audio technology as a result. Instead, we’ve brought in research experts like Jean-Marc Jot who have been more directly involved with 3D audio and built what I feel is a highly capable 3D audio team. And our background in developing tools for the audio entertainment industry serves us very well in integrating and supporting this team.

I’m much less concerned with our research expertise than I am with the sheer momentum of establishing working standards, enticing developers to support them (and support them well), and gaining market recognition. The technology’s the easy part. We’ve had lots of opportunity to acquire additional 3D audio research expertise, but it’s always been my judgment that the limiting factor is deployment.

Will Creative implement HRTF routines on multiple speakers (4/6/8) in the near future?

Yes. The upcoming driver enhancements available through the Live!Ware program will provide up/down discrimination and overall improved 3-D positional audio over headphones, two loudspeakers or multiple loudspeakers. These improvements involve binaural techniques (based on HRTFs) and other proprietary methods developed at the Tech Center. We will also make available in the near future some new driver enhancements to handle a wide variety of loudspeaker configurations involving up to 8 output channels.

DS3D models all sounds as point sources, that is the size of sound source don't matter. Do you have any plans to address that?

This is certainly one of several problems with the DS3D model. We are working with Microsoft to resolve these issues, and we’re also working on how to implement the sonic effects. It is too early now to say when we will address this particular problem in released drivers, but we are not neglecting it. It will require extensions of the positional audio part of the API, which might be handled either in EAX or in DS3D.

contents


Back To Top	h t t p : // a l i v e . s i n g n e t . c o m . s g	Copyright © 1998-2003 ALive! and its writers.