Considering how ubiquitous and indispensible they are today, it’s amazing to think that digital cellular phones were a novelty 15 years ago. In the mid-1990s, digital cellular handsets were just beginning to be deployed in large volumes. At that time, these handsets were becoming the “killer app” for digital signal processors. Chip suppliers such as Texas Instruments (TI) and AT&T Microelectronics ramped up their DSP processor development efforts to compete for design wins in this growing market.
The size of the opportunity in digital cellular handsets made them the defining application for DSP processors in the 1990s. They inspired a new level of investment and intense competition among DSP processor vendors, which in turn spurred a substantial burst of innovation. Fifteen years later, we’re still using many of the processor architectures developed during that era.
It appears that history is now repeating itself, in that a wireless application is once again becoming the “killer app” for digital signal processors. Only this time, it isn’t handsets (which these days tend to rely on highly specialized application-specific chips). Instead, these days it’s cellular base stations that are driving the development of DSP processors.
I know, it sounds absurd. After all, it’s obvious that base stations are not a high-volume application, right? No, actually—wrong. As it turns out, a single base station can consume numerous DSPs, and there are so many base stations being sold these days (150,000-200,000 per year, according to market research firm Forward Concepts) that base stations have become a high-volume application for DSPs. TI claims to be shipping nearly 10 million DSPs a year for base stations. A significant minority of these are big, expensive chips: Boasting multiple DSP cores, multiple co-processors, and sophisticated memory and bus systems, these chips can sell for up to $200. Do the math: Millions of DSPs a year at hundreds of dollars per chip is big money. Yes, this is a high-volume application.
It is a high-volume application—and a demanding one. Cellular standards continue to evolve, with ever-more-complex algorithms and ever-higher data rates creating a need for more and more processing power. Meanwhile, wireless network operators are locked in intense competition, which results in escalating demands for more energy efficient and cost-effective processors. So, we’ve got a large market with heavy processing requirements and intense energy and cost pressures. This sounds like a perfect environment for intense competition and rapid innovation—and that’s exactly what we’re seeing.
Not too long ago, TI had this market virtually to itself. But a growing, half-billion-dollar-a-year market is apt to attract competitors, and that’s what happened here. The first real contender to step into the ring with TI was Freescale, newly focused and energized by a new CEO and executive management team. Freescale’s made significant inroads into the base station DSP market, and undoubtedly that has spurred TI to innovate faster. Last week we saw some of the fruits of TI’s efforts, as TI introduced a new high-performance DSP architecture, and upped the ante for competitors by bringing high-performance floating-point capabilities to base stations for the first time.
Last week we also saw a new family of high-performance base station DSPs from Freescale, and a new base-station-oriented licensable DSP core from CEVA. Core licensors like CEVA typically focus only on high-volume applications, where they stand to earn substantial unit royalties. So when CEVA rolls out a chip for base stations, that’s another good indication that base stations have become a high-volume application.
Another interesting indicator of the size of the base station opportunity is Intel’s interest. When a company is as big as Intel, it can only afford to pursue big markets. At the SDR ’09 technical conference a year ago, Intel showed a demo of wireless base station baseband processing running on off-the-shelf Intel CPUs. Earlier this year, Intel released a video commercial touting the benefits of running all layers of base station processing (including but not limited to the physical layer processing typically implemented on DSPs) on Intel processors.
I’m really pleased to see the intensified competition in base station DSPs, and the associated intensified innovation. Big investments—in money and intellect—are needed to address the big challenges facing the wireless industry, such as making multi-core DSPs easy to use, and integrating digital signal processing with higher layer processing in ways that yield real benefits to network operators. And now that these investments are being made, they’ll benefit many demanding applications beyond base stations—just as the investments and innovations spurred by digital handsets in the 1990s did.
Add new comment