As strange as it might sound, one of the pleasures of the DSP market is that system designers continue to surprise us. When we develop a new digital signal processor chip, we generally have specific markets in mind-but with each new generation of devices, system designers usually find a new set of very inventive and constructive uses for our parts. They take the processing power we put in their hands and their creative minds to work to develop some rather phenomenal applications-engineering at its finest. This is good not only for the semiconductor industry, but for the electronics market and our economy as a whole.
From the outset, the development of high-performance, inexpensive DSP chips has been instrumental in bringing signal-processing capabilities to new applications, including commercial and consumer products. From the time we released the first single-chip DSP more than 20 years ago, our customers have continued to raise the bar with innovative applications designed around our products. This trend will certainly continue unabated. So, each time we show our customers more performance, as we have this year with the demonstration of a 1 GHz DSP, we look forward to the groundbreaking applications they will create.
Of course, not all applications catch us off guard. Most of the time we have an idea of what's coming prior to an application's development. For instance, we promoted DSPs as an enabler for digital cell phones because we knew that the technology had reached the level where digitized voice transmissions were not only feasible, but also economical. We also knew that the rush of audio/video and communications standards in the mid-1990s would require DSPs for compression, videoconferencing, DSL, cable modems and the many wireless telephony standards.
Still, we are unable to predict everything. With our very first DSP, the TMS32010, we thought the primary target market was speech I/O, such as voice synthesis, recognition and verification. As it turns out, not many people used the chip with those intentions. Instead, we soon found the chips being used in modems, hard-disk controllers and 3-D graphics. We were indeed surprised, and we shortly revised our marketing strategy to align with our customers and started to design the next-generation device to better handle these applications.
Consider another interesting example. One day several years ago, one of my colleagues pulled out a car-racing magazine and pointed to a story that mentioned how Lotus was doing something with DSPs. We checked into it and found that it was using our TMS320 DSP family to build an active suspension to replace conventional passive shock absorber/spring suspensions. Active suspension detects when a race car is going over a bump and lifts the tire a bit to compensate, which in turn reduces tire wear and improves performance (e.g., fuel efficiency). In addition, when the car takes a sharp corner, the suspension system adjusts the hydraulics to compensate for the centrifugal and centripetal forces, thus allowing the car to take corners a bit faster. As you can imagine, being able to corner just a few miles per hour faster, or getting better mileage and avoiding an extra pit stop for gas or tires, can make a huge difference in a Grand Prix event. But our original marketing plan certainly didn't include Grand Prix racers.
In a similar case, a customer surprised us by demonstrating some of our DSPs in the first working system that implemented VoIP (voice-over-Internet Protocol) some years ago. Using that protocol as a carrier for the transmission of DSP-encoded voice was a brand-new application, and again one we had not anticipated. Today, VoIP has become commonplace in enterprise networks, and the growth of the market for this application continues to rise. In fact, without digitized, packetized voice transmission, converged voice and data networks would not exist.
Over the years, we've detected what seems to be a correlation between advances in DSP performance and levels of technological innovation. When modem manufacturers began designing DSPs into their systems, transmission speeds in a succession of products methodically jumped from 1200 to 56,000 bits/second-enabled by advances in theory and a rapid increase in DSP performance. The active suspensions and VoIP just mentioned also developed alongside surges in DSP performance. Likewise, advances in DSP architectures drove the development of cell phones and digital subscriber lines (DSL). These advances lowered power consumption in devices for portable systems while they introduced the massive parallel throughput of data for multichannel systems.
The progression to 1-GHz DSPs represents the latest move to a higher level of performance. We predict 1-GHz DSPs will enable designers to pack many more channels into basestations, DSL line cards, routers, switches, video servers and other communication equipment. Digital still cameras will emerge with sensing arrays of greater than 10 Megapixels, providing professional-quality digital photos. In hospitals, MRIs and CAT scans will provide detailed medical imaging so that doctors can prepare a more precise diagnosis and treatment (maybe someday in the home rather than in the hospital). Greater accuracy in computer simulations will aid aerodynamic design, fluid and stress analysis, weather and environmental forecasts as well as other computation-intensive applications. Navigation, guidance and recognition systems will become more precise and thus safer and more reliable.
We can anticipate these and many other uses of 1-GHz DSPs because they're extensions of existing applications. What we can't predict are the new uses creative designers will unleash. For instance, we know prosthetic design is starting to give some level of vision to those with a degree of impairment-and we hope before long even to the blind. But this field is in its infancy, and we don't know exactly what capabilities these artificial eyes might have. When developments in artificial vision combine with increasingly sophisticated computer-controlled motion, robotics, unlike anything seen to date, could well emerge.
Another factor contributing to the situation of predicting applications is that the DSP market has matured-considerably. Today, not just one, but many markets and industries are using the powerful processing found in DSPs. The first devices were generic in that they met the needs of multiple applications. Now we find that many applications need different versions on this base capability. The DSP needed for a wireless handset is different from the one needed for a wireless basestation. The DSP needed for audio differs from the one needed in a digital camera. Each application demands a unique core architecture, a different power budget, a different performance level and a different set of I/O. Just as we're excited about the 1-GHz DSP and how it will drive new applications, we're equally excited about what other performance aspects such as lower power dissipation and lower cost will do to drive unexpected applications.
A major lesson we've learned is that both we as vendors and our users should not let ourselves get locked into preconceived notions about where DSPs can make a difference. We must eschew traditional thinking if we are to be successful and stay at the leading edge. Experience shows that this will certainly happen; we know for certain that we will be blindsided again. There's no way we can predict by whom or in what way-but, then, that's one of the particularly fascinating aspects of this industry.
Gene A. Frantz is principal fellow and business development manager for the Digital Signal Processing Semiconductor Group at Texas Instruments Inc. (Dallas).
See related chart