Evolution is the Revolution

Electronic News, Nov 13, 2000 by Jim Turley

The diversity of embedded apps. requires adaptable products

We know from insects trapped in prehistoric amber that some ants have scarcely evolved since the Cretaceous period 100 million years ago. They haven't evolved because they haven't needed to ... they're perfectly adapted to their environment. Since that same period, however, larger creatures have evolved in numerous ways to cope with the fiercer competition among vertebrates. This is a good analogy for what's happening today in the embedded systems and DSP/mixed signal markets.

As the 21st century unfolds, we are witnessing an explosion of new applications for embedded processors, many of them unanticipated. Semiconductor vendors cannot possibly offer general-purpose processors and other standard parts to meet every contingency. Nor can off-the-shelf chips fit the exact requirements of every application. The demand for higher integration and greater product differentiation, coupled with time-to-market pressures, is forcing embedded system developers toward more adaptable solutions. Those solutions include ASICs, synthesizable DSP cores, configurable processors, and licensable intellectual property (IP).

According to San Jose-based GartnerGroup Inc.'s Dataquest unit, application-specific processors are the fastest-growing segment of the embedded processor market. Dataquest pegs the worldwide market for application-specific processors at $12.5 billion in 1999, forecast to grow to $32.6 billion in 2003. Dataquest also estimates there are 4,000 design groups starting 15,000 new system-on-a-chip designs each year.

In essence, embedded developers are obliterating the historical difference between hardware and software. The old model is that hardware is "hard" (fixed) and software is "soft" (adaptable to hardware). The new model--enabled by modem EDA tools, synthesizable cores, and other digital-design technology--is that hardware is becoming as malleable as software. Only when the whole solution is flexible can developers adapt to a rapidly changing environment.

Consider how fast a consumer-product market can change. Until recently the Sony Walkman was the portable music player of choice for teenagers. Suddenly, Internet-audio MP3 players have become the hot products. The impact is startling. Consumers have found a whole new way to obtain music (downloading MP3 files from the Web), consumer electronics manufacturers have stumbled onto an entirely new product category (portable MP3 players), and semiconductor vendors have discovered a lucrative new market for their embedded processors, cores, DSPs, and flash memory chips. Meanwhile, the music industry has been shaken to its foundations and the legal system is struggling with the implications for copyright law.

That's just the beginning. Expect more meteors like MP3 to hit, altering the environment in favor of some companies and threatening the extinction of others. Of course, no one can truly expect the unexpected. Even the best industry analysts and pundits are little more than weekend weathermen in this age, unable to predict the long-term future with any certainty. However, it is possible to identify some trends and potentially disruptive technologies:

Communications--Everything and everybody will communicate with everything and everybody else. That presents enormous opportunities for companies oriented around networking, telephony, wireless, and the components required to make all those things work together. Because data is digital and signals are analog, mixed signal devices and DSPs will be in great demand. Almost any application involving voice communication needs echo cancellation and other DSP filtering. But price pressure will be tremendous. Highly integrated solutions that can quickly adapt to different manufacturing processes will enjoy an advantage, especially in the mixed signal communication realm, where chips might be fabricated on anything from CMOS to silicon germanium.

Privacy -- If you think this is a big issue now, just wait until rampant identity theft, electronic espionage, and recreational hacking threaten the integrity of financial assets, stock markets, and national economies. Encryption will be built into almost every embedded product that handles digital data in any form, whether it's your dollars, your DNA, or your dance music. This will create countless opportunities for customizable DSPs and processor cores. Executing an encryption algorithm in hardware is much faster than executing it in software, and it's tougher to crack.

Integration -- When a new product category emerges, the industry usually responds in three stages. In stage 1, product designers build a system out of readily available, off-the-shelf parts. It works, but it's usually not optimal, because it's relatively costly, consumes too much power, isn't fast enough, isn't small enough, etc. (Think of the first-generation scanners, MP3 players, cell phones, PDAs, digital cameras.) In stage 2, engineers create a more integrated, specific solution with ASICs. That yields big improvements along all dimensions. In stage 3, if the product category really takes off, a semiconductor company may design an application-specific standard processor/product that combines the integration and specificity of an ASIC with the immediate availability of a standard part. That allows many more competitors to join the race without designing ASICs. But it also leads to the next point.