SAN JOSE, Calif. The costs of poor design quality as well as the costs and trade-offs of maintaining good quality came under scrutiny at a panel discussion entitled, "The hidden costs of design quality," which was presented as part of an EE Times reception at the International Symposium on the Quality of Electronic Design (ISQED).
While panel participants agreed to the compelling need for design quality, several unanswered questions became apparent: How much cost and effort should go into quality? What constitutes "acceptable" quality? How is quality ultimately defined?
The developing "global Internet machine" has made quality a much higher priority than it was in the past, according to Gadi Singer, vice president and general manager of Intel Corp.'s Microprocessor Products Group. Due to the use of the Internet for commerce, education, and communications, Singer said, the general population now has a "dependence on highly automated and interacting machines."
What's more, Singer noted, one no longer has to be a "certified geek" to use a computer. So at the very time when the knowledge level of the user is declining, computer equipment and the processors that drive it are becoming much more complex.
John East, president and chief executive officer of Actel Corp., reiterated some of the design quality horror stories he had cited in his earlier ISQED keynote. These include the failure of the Iridium satellite communications system, and the inability of Motorola and Zilog to provide reliable processors to power the first generation of PCs, giving Intel a huge victory. "Several hundred billion dollars have been lost, some because quality was bad," he said.
One example cited by Keith Lobo, president of the Quickturn subsidiary of Cadence Design Systems Inc., speaks well to the cost of poor design quality. Lobo said that NASA skipped a $100,000 test that should have been run on the Hubble Space Telescope, then had to spend several billion dollars including a space shuttle mission to fix a resulting problem.
These and other examples are "only the tip of the iceberg," Lobo said. "We are living on the edge in terms of design quality."
IP surprise
Howard Sachs, president of the Virtual Socket Interface Alliance (VSIA), noted that VSIA member companies surprised that organization's leaders last year when they identified "quality of intellectual property" as their biggest concern. As a result, VSIA has convened a group to examine the issue, although it's not yet set up as an official VSIA development working group, Sachs said.
The costs of poor design quality include missed revenue, huge legal costs, and the dangers of not having a product at all, Sachs said. But what are the costs of maintaining good quality? "You have to design quality in," he said. "It doesn't just magically get there. It takes longer to do it right and the product costs more."
During a question-and-answer period, an audience member picked up on that theme and asked when one should stop "fighting for quality" and decide to ship the product.
"I don't know what the formula is," Sachs responded. "You do the best you can, but ultimately you have to ship the product. You can't make it perfect."
"On any digital logic design, it's only a matter of time before you find flaws in it. No complex piece of hardware is totally flawless," said Lobo.
He argued strongly, however, for a larger design verification effort. In the $1 trillion market for electronic products in 1999, only about $500 million was spent on verification tools, Lobo said. That's around 0.05 percent of the cost of the products shipped to end customers.
"If you want quality, you need to have a verification strategy and make the appropriate investment," Lobo said.
When asked to define quality, panelists made it clear that it involves more than reliability. A product also needs to meet its original specs, and be on time. "A product delivered late is not a quality design," said Singer.
Sachs, who serves as VSIA president while remaining employed at Fujitsu Microelectronics, said that quality might be defined by a Japanese company as "whatever the customer wants." But Singer said that might not be good enough, if the customer is going to install something in a larger system that might not work. "The immediate customer is not always your best authority," he said.
Lobo noted that customer viewpoints about "acceptable" quality are highly variable. People put up with long waits to get onto Web servers, he noted, but no one would tolerate such delays when making telephone calls. And for a digital avionics system going into an aircraft, he noted, no quality defect is acceptable.