Sound Blaster Audio—The Trojan Horse for Motherboard Integration of DSPs

The evolution of hardware/software solutions from innovative, expensive add-on functions to indispensable, low cost motherboard implementations is the ultimate goal of most new PC technologies. Disk controllers and video/graphics controllers are prime examples of core ICs for PC architectures. As multimedia continues to drive PC demand, new silicon implementations, such as codecs, sound controllers, and scalable signal processing accelerators will be added—each using Digital Signal Processing (DSP) technology. From a business perspective, this is how that extension will most likely occur.

First came the widespread adoption of hard wired, fixed-function DSP chips embedded in popular add-in cards, such as Creative Labs Sound Blaster music synthesizer chips, Compaq Business Audio codecs, and fax/modem chipsets. These ICs processed real time, multimedia signals for games, CD-quality music, and data communications applications much more cost-effectively than host-based or embedded RISC chip designs. As the sales of fixed-function cards soared, a second trend was soon underway: to compete against single function ICs by offering multi-function, programmable DSP chips.

Programmability offers major advantages throughout the fast-growing, highly-competitive PC industry. By converting music synthesis, wavetable or fax/modem functions to software, a programmable DSP chip can be reconfigured—by the OEM, the reseller, or even the OEM's or the reseller's customer—to act as a sound card, a modem, and so on, at greatly reduced system costs. The demand for Sound Blaster functionality, and the ability of IC manufacturers to integrate board-level products into a few chips or even a single chip highlighted the benefits and accelerated the appeal of DSP.

Today, industry analysts predict that 75% of PC motherboards shipped next year will have Sound Blaster audio capability (through fixed-function DSP technology) on the motherboard. Add in audio codecs and modem chipsets and those same analysts project nearly every PC motherboard will include some form of chips with DSP technology.

Application Forces
The demand for interactive applications, like teleconferencing, telegaming, and teletraining is just beginning. In order to enable these platform extensions, codec technology must advance beyond sigma-delta toward technologies with the ability to convert and synchronize many different signal types and sample rates on one codec. This technology must be incorporated into system architectures to cost-effectively manage the "multi" in multimedia. New ICs that provide full duplex capabilities (simultaneous play and record or send and receive) are already here. So too are signal processing front ends such as the Analog Devices AD1843 SoundComm codec, which combines audio, speech, PSTN, and video data conversion and synchronization into a single component. Such multipurpose devices point out the limitations of fixed-function DSPs and improve the cost-effectiveness of programmable approaches.

Cost Forces
There are practical limits—in terms of system complexity and cost—to the number of single purpose signal processors that PC OEMs are willing to implement. Few envision a system with individual processors for Sound Blaster games, wavetable music synthesis, fax/data modem functions, and various MPEG audio/video compression standards. While such dedicated signal processors will initially deliver many of these single capabilities to motherboards, it is far more practical for OEMs to collapse those pre-existing MIPS into a single programmable DSP chip.

Operating Environment
The industry-wide adoption of an open architecture for signal processing solutions is almost complete. Most DSP manufacturers have converged on the SPOX operating system as a standard real time OS. Now Intel has based its IA-SPOX Native Signal Processing (NSP) initiative on the Spectron technology. This will ignite algorithm software providers. Intel is investing in its NSP initiative and will use IA-SPOX to link signal processing hardware with the Microsoft Application Programmer's Interfaces (APIs) for integrating audio, telephony, and communications into software applications. This will accelerate application software development and spawn a range of compatible hardware solutions for the market to choose from—from Pentium CPU with a full-duplex codec through full-featured Signal Processing Accelerators that do for multimedia what graphics accelerators do for graphics. Signal Processing Accelerators will off-load the host CPU of math-intensive signal processing tasks such as echo cancellation in speakerphones and high-speed data pumps in modems. The result will be a totally level playing field. While proprietary, fixed-function chips will continue to compete primarily on IC price, programmable signal processing implementations will provide innovation and differentiation while maintaining significant system level cost and performance advantages.

How about the ability to add a high speed fax/modem to your DSP-based audio solution for less than $10? Or the ability to then add wavetable music synthesis, MPEG playback, teleconferencing, telegaming, and telephony features for just a few dollars more?

Because they enable virtually every level of the PC market infrastructure to add value, differentiate and profit, programmable DSP chips seem destined to become integral silicon building blocks for future PC architectures. And the Trojan Horse effect is starting right now with the industry-wide adoption of Sound Blaster music synthesizer chips into mainstream PC architectural designs.