Language is a lot like food. Too many servings of a popular expression can turn yesterday's filet mignon into tomorrow's hamburger. You don't have to search very hard in the electronics industry to discover terms that have turned bland from overuse. Now, I'm not referring to those terms that have well-defined or mathematical definitions, but rather the more-vague conceptual phrases that get bandied about in promotional literature, news stories, and even many technical papers.
One that comes immediately to mind is "single-chip." It expresses the idea of circuit integration in a succinct way, and when a design is reduced from multiple ICs down to just one, that's definitely a worthy accomplishment. But isn't it a little odd to speak of a "single-chip" solution, when the chip in question requires dozens of external passive parts to function?
Consider, for example, single-chip radios. Not long ago, I read a news story about a newly introduced one-chip radio for GSM. The IC was being lauded for reducing the external parts count down to a mere 50 components.
A close cousin to the "single-chip" description is the "system-on-a-chip" (SoC). For several years, the SoC industry has been focusing on ways to exploit the vast amount of silicon real estate afforded by submicron chip design. Completely monolithic ASIC solutions, SoCs represent the ultimate in circuit integration and cost reduction for high-volume applications.
In many cases, though, SoCs haven't been practical for various reasons. Different circuit functions can be implemented more efficiently in different semiconductor processes; development cycles may be too short; and blocks of silicon IP from different sources aren't as easy to mix and match as board-level components. These obstacles have led to frequent "poo-pooing" of the SoC idea.
But as sure as Father Time makes way for Baby New Year, decreasingly popular expressions must bow to their successors. With SoCs losing some of their cachet, several segments of the industry are promoting SoPs and SiPs, which are systems-on- or in-a-package. We're hearing a lot about these from packaging vendors. They point out that it's often easier to build a multicomponent module using the latest thin- and thick-film techniques than to chase the holy grail of an SoC.
Their case is bolstered not only by demands for miniaturization, but also by performance requirements. SoPs can provide the shorter interconnects necessary to route high-speed signals with fast edge rates.
It's very likely that SoP approaches to design will gain increasing popularity. But the term suffers from the same flaw that dooms SoCs. Like their SoC and single-chip cousins, SoPs will be assembled onto pc boards along with a variety of other components. In time, we will start hearing more about what functions can and can't be integrated using SoP-style devices.
It shouldn't take too long for engineers to realize that one board-mountable component rarely comprises a system. The SoPs will be subsystems within some larger system-level designs. Perhaps someone will be tempted to coin new phrases, like subsystems-on-a-package or minisystems-on-a-package. But it's bad enough now that we have to keep our SoPs straight from our SOPs (small outline packages). Acronyms such as MSoP and SSoP would only confuse matters further.
Despite the discrepancies that exist between their concepts and real-world implementations, terms like single-chip, SoC, and SoP can't be dismissed as mere buzzwords. They signify the trends and goals that are driving innovation. But a more judicious use of these terms might help us to separate engineering fact from fiction.