viewpoint

System Design Goes beyond SOCs True system design isn't about what you can cram onto a chip. More accurately, it's about what you put into the entire electronic product. by Will Herman

Many people would argue that "system design" really refers to system-on-a-chip design. Though it's true that ICs are becoming larger systems, SOC design certainly doesn't have the same concepts behind it as does the system-level design performed by most electronics companies. These companies don't worry only about what's on the chip--they're also concerned with packaging, software, buses, interconnect, power supplies, fans, and the like. Real system design produces not an IC, but a marketable end product.

Let's face it--an SOC doesn't differ substantially from the chips that designers have been working on for a long time. SOC design isn't a truly distinct methodology, but simply the practice of integrating more functions onto a chip. In addition, the trend toward SOCs merely results from utilizing the density available in today's silicon. True system design concerns the actual end product, and will always require designers to add functionality beyond what any one IC or chipset can offer. Consequently, the industry is renewing its focus on product design and differentiation at the full system level.

For the last ten years the EDA community has emphasized semiconductor design. As a result, the quantity and quality of tools produced for true system design have evaporated. What's worse, semiconductor advances that brought about faster, larger, and denser devices have thrown a lot of design problems up to the board level. But for the most part, product design engineers lack the tools and often the knowledge they need to create successful systems.

Today's ICs, for example, require data to be pumped on and off the chip at blazing speeds, causing signal-integrity problems at the board and backplane levels. Or consider that an SOC's density demands hundreds of package pins, which is likely to cause timing or manufacturing problems. The tools available for the system designer don't measure up to the complexity of design needed for today's electronic products, and designers often lack the training necessary to accomodate all of the electrical effects of the complex silicon.

Moreover, the overwhelming consumerization of the electronics marketplace has exacerbated the dilemma. A great number of electronic products must reach the shelves by September because companies rely so heavily on the holiday season sales--a familiar source of stress on the toy or simple electronic product markets. Increasingly complex devices, like cellular phones and PCs, must bow to the tight holiday market window that drives the entire product development process.

Today's ICs, however, have become so complex that their development time is longer than the design cycle available for the entire product, which is typically less than a year. Consider that if an ASIC requires 12 months to develop, the resulting system design will never make the market deadline. It doesn't matter how complex, intelligent, or dense an ASIC is, or how much of the system appears on a chip--the bottom line is that the end product missed the market window.

Consequently, shrinking design cycle times are putting pressure on system designers to use existing chips and differentiate their products at the system level. As a result, the number of ASIC design starts will decrease. Some analysts say that the number is decreasing already, with more of the ASICs under development designed for multiple products. Again, the strategy demands that the product designer build a unique system around existing chips. The design of the electronic product then becomes the challenge of designing the system, not the chip.

The personal computer may be the ultimate example of such a strategy. Most PC manufacturers use the same Intel chipset, which they combine with a large number of standard parts to create a unique end product. PC makers don't dare put the entire computer on one chip, because they wouldn't have the means to differentiate their product by changing system components and software.

In a nutshell, SOC design is following the same path that designers have been traveling all along. The current objective, though, isn't to condense an entire end product onto a single piece of silicon--rather, it's to create the next level of standard parts for system designers.

Will Herman is the president and CEO and a cofounder of Viewlogic Systems, Inc. in Marlboro, Mass. He's also served as president and CEO of Silerity and president of Scopus Technology.

To voice an opinion on this or any Integrated System Design article, please email your message to jeff@isdmag.com.

integrated system design  June 1999