SAN JOSE, Calif. In a fast-moving keynote address at the first IEEE International Symposium on Quality Electronic Design (ISQED), John East, president and chief executive officer of Actel Corp., highlighted a number of widely-known design failures from the last several decades and offered a primer on the complexities of design quality.
East first outlined the wide field infrared explorer project that was intended to gather information, through satellite-based sensors, on infrared signals generated by the primordial Big Bang. Those sensors failed miserably when shields protecting the equipment self- destructed prematurely and ruined any opportunity to collect data. "If we'd just done a little better job at the outset of the design, we could have avoided paying a big price" in lost opportunities at the end, he said.
In his second example, East cited the recently cancelled Iridium communication satellite project. Iridium never succeeded in launching the full number of satellites envisioned by its organizers. Only 66 of the planned 77 satellites were eventually launched, and they failed in large numbers due to an erroneous decision to avoid costly space-grade radiation-hardened circuitry on board. The message: "You can't ever sacrifice on quality in design," East said.
Finally, East mentioned Intel's design win to provide the 8088 processor in the original IBM PC hardware platform in 1980. The Motorola 6800 and the Zilog Z8000 undoubtedly offered superior design options, East said, but "they weren't quite there" when IBM needed a here-and-now chip upon which to base its PC product line. At least $350 billion of Intel's current $400 billion market valuation can be viewed as directly attributable to the crucial 1980 design win, East said.
Damned if you do, damned if you don't
The moral of these stories is troubling, according to East. "You're damned if you do. You're damned if you don't," he lamented. If designers work to optimize quality, they risk being late to market with their product, generating unacceptable design costs and lost profit opportunities. If, however, designers work to meet time-to-market demands, they usually end up compromising on product quality.
To specifically address the design dilemmas facing his audience, East outlined the trends in IC design which offer potentially successful strategies. First, reusable intellectual property though difficult to design and verify will help alleviate time-to-market pressures, he said. Second, though EDA tools today are good, they're not as good as they need to be, and that must change going forward. Finally, and not surprisingly as an executive of a programmable product company, East suggested that building additional programmability into products will permit tweaking on behalf of design quality throughout the process of product development.
East ended his keynote with a laundry list of specific suggestions to hasten and enhance IC design: build more speed into product specs than are really needed; never lose track of power considerations; don't allow EDA tools to force design strategies; design for observability in the era of deep-submicron geometries; design for fixability; work toward hybrids that combine standard-cell designs and programmable blocks "ASSPs bolted to FPGA hard cores," East said.
In conclusion, East offered a corollary to the rhetorical question, "If a tree falls in the forest and no one is there, does is make a sound?" The IC design version: "Is a great product truly great if, when it's released, there's no one there to buy it because it's late to market and over cost?"
Peggy Aycinena is Senior Editor of ISD Magazine, a sister publication to EE Times.