At last month's Texas Instruments Developer Conference, I attended a panel on the future of signal-processing applications. As has been the case at most industry events I've attended, the panelists focused on ultrahigh-volume applications like DVD players and the iPod.
On the one hand this focus makes sense, because that is where the big money is being made. On the other hand, these markets are of interest to a relatively small number of system designers. Only a few types of products attain huge volumes, and these markets are typically dominated by a few very large players. In many cases, it is very difficult for new players to enter these markets.
In contrast, there are thousands of low- to medium-volume markets. These markets are often highly competitive, with many players targeting the same application. In many cases, it is relatively easy for new players to enter these markets. With so many markets and so many competitors, it is safe to say that most system designers are working on low- to medium-volume products. So a panel discussion focused on the needs of high-volume markets isn't useful for most system designers.
What's interesting is that the focus on high-volume markets isn't just an artifact of conference panels. This focus is also apparent in processor vendors' strategies. In March, for example, LSI Logic announced plans to dump its DSP processor division so it can focus on storage and consumer-electronics applications. Similarly, TI is churning out application-specific chips at an impressive clip, while its releases of general-purpose DSPs have slowed to a trickle.
That creates challenges for the many system designers targeting low-to-medium-volume markets. One particular challenge is finding a DSP that offers the right kinds of integration for these markets. DSP families typically offer only a dozen or so variants. You may find that none of the variants offer quite the right on-chip integration for your application. So what should you do?
One answer may be to look to general-purpose processors (GPPs). For a given GPP architecture, there are often dozens or even hundreds of chips to choose from. This may make it easier to find a GPP that offers the right kinds of integration for your application. Another possibility is to use an FPGA. With an FPGA, you can design-in exactly the peripherals you need.
If I were a DSP vendor, I would worry about GPPs and FPGAs taking over low- to medium-volume markets. Sure, none of these markets are big by themselves, but taken together they add up to significant revenue and profit. Perhaps more importantly, most high-volume markets don't appear overnight. Instead, most high-volume markets start out with low volumes and grow over time. If DSP vendors cede today's low- to medium-volume markets to GPPs and FPGAs, they may lose the opportunity to participate when these markets grow to higher volumes.
Jeff Bier (www.BDTI.com) is the president of Berkeley Design Technology Inc., a consulting firm providing analysis and advice on DSP technology. Kenton Williston of BDTI contributed to this column.