[From iSource Business, December 2001] It tells that little red light on your car's dashboard to blink "Check Gauges." It tells your telephone to store numbers and direct calls. It connects you to a world beyond the one you experience on a daily basis, the world that glows on your computer screen. However, these few examples don't even begin to cover the vast amount of things we use in our daily life that are powered by chips - an amount that is in direct disproportion to our feeble amount of working knowledge about them.
So, before we go any further, here's a quick lesson in Silicon 101. Semiconductors, which initially start out as 8" wafers but are eventually cut into chips after processing, are fabricated in foundries, not manufactured in factories. A facility that has a foundry to make semiconductors is called a fab. If a company does not have a foundry, it is said to be fabless, and it designs, develops and markets its products; and it outsources its production. Oh, and there is just one more thing you should note about this industry: it has a supply chain burdened with out of date, useless inventory and enormous volumes of information.
The San Diego-based Applied Micro Circuits Corp. (AMCC) has been in the semiconductor business since 1979 as a key supplier of integrated circuit (IC) technology for fiber optic networks, consumer electronics and defense systems. Like other chip makers industry-wide, AMCC is having to come to terms with the industry's supply chain, among other issues. First, the proliferation of fabless companies has loosened the semiconductor supply chain by creating far-flung networks of engineers, foundries and suppliers that complicate the routing of information in time to act before market opportunities disappear. The result has only added to a time-to-market frenzy that would blow a widget-maker's mind. As chip speeds double every 18 months, in accordance with Moore's law, new products become obsolete almost as quickly as they're released. Meanwhile, since chips quietly power just about everything we consider essential to daily living, they have to work right. When they don't, those 8"wafers can "yield" too many deficient chips, thus leaving the companies that make them with useless, unsold products.
Such inefficiency is illustrated in graphic detail when you look at AMCC's FY 2001 net revenues, which rose 153 percent to $435 million while its profits, in turn, tanked. This was due in large part to a $3 billion write-off for excess and obsolete inventory sitting stagnant in its supply chain and aided by the savaging its telecommunications customers have been taking for over a year. However, the long view among analysts and industry folks is that AMCC's innovations will continue to sharpen the "intelligent optical edge."
And, fortunately, there are solutions that have been created especially for the purpose of disintegrating cumbersome and inefficient supply chains. The one implemented by AMCC for the purpose of improving its yield management is dataConductor, a collaborative engineering software offered by San Diego-based Syntricity Inc. that provides engineers with faster, easier access to centralized test data. A hosted application, dataConductor sits on an Oracle platform that centralizes raw data and provides a standard user format that enables engineers to input data from their desktops, retrieve data and automated reports over the Web, and share the information online. Jim Icuss, AMCC's director of semiconductor development, says that among the company's 1,200 employees is a worldwide cadre of engineers, and their ability to input data using a standard format from wherever they are is dataConductor's chief asset.
This is collaborative technology, a concept that's still being defined. What makes the Internet better at collaborative technology than, say, a telephone? "From a supply chain perspective [collaboration] means being able to see far enough upstream to plan and manage inventory better," says Amy Meyers, principal of Tigris Consulting. Cutting the time it takes engineers to get test information from the production floor has improved AMCC's yields by enabling them to spot problems early. Materials costs are reduced through better planning, but there's also less waste resulting from product deficiencies that are identified during the production process.
Echoing Meyers' comments, Syntricity's Director of Client Services Dean Barnat says, "Engineers can predict the behavioral characteristics of their materials and make decisions about what to do with that material in the future." dataConductor collects and stores "characterization" data (detailed description of a chip), which provides engineers with a history that helps them compare the properties of specific materials that are apparent in different tests but can elude them in the sheer volume of test data. Barnat, a former product engineer and one of dataConductor's earliest users, says the application enables him to do a full characterization and analysis of product in one day, whereas it used to take one or two weeks.
To describe the ROI that dataConductor provides Syntricity's clients, Jami Fry, Syntricity's marketing manager, relates the story of a customer that tested dataConductor on one product line during a six-week evaluation period. The company realized a 75 percent reduction in the time engineers normally spent in device characterization; and it showed a 5 percent increase in yields, which equaled $2 million in increased profits.
Bolstering Fry's example, Icuss says that before implementing dataConductor it was difficult to get rapid feedback on yield issues from wafer foundries. "The amount of manpower required to get wafer maps and packages of data has gone way down. We can feed information to our wafer fabs almost daily, and there's been a tremendous yield improvement as a result."
Although the exact numbers for ROI are not available, Icuss notes that the adoption rate is one of the best informal metrics. "At AMCC the yields are higher, the engineers are routinely using the dataConductor package, and daily reports are including histograms and detailed data. For me, that is a successful implementation and the best form of ROI," Icuss says.
While Icuss has nothing but praise for AMCC's working relationship with Syntricity, in terms of its user focus and the product's performance, he believes there is at least one issue about dataConductor that could stand improving. dataConductor's closed architecture and proprietary record structure prevents seamless integration with some of AMCC's existing software, such as Access and Statistica, two of Icuss's primary tools for analyzing data and generating additional reports. Syntricity will customize dataConductor's automated reports to the user's specifications, but engineers often like to manipulate data in myriad ways on the fly. Syntricity's solution requires capturing raw data from dataConductor, converting it into a universal format and inputting it into the user's preferred software.
Fry responds that Syntricity works closely with its customers to resolve these types of issues, however Barnat adds that a large portion of the problem occurs when trying to migrate processes to a Web-based environment. Another obstacle is that HTML, the common language of the Internet on which dataConductor is based, has reached its limit in handling complex functions. Says Barnat, "Our product is going to be ported over to Java in the near future, and we are taking steps to enhance its reporting ability." (Java is a more flexible Internet language.)
The semiconductor supply chain isn't disintegrating so much as evolving, and this is only a flaky moment in time during which products like dataConductor are being used to add new links to an increasingly global supply chain. The result is streamlined data gathering, better knowledge management, faster research and development, enhanced communications, improved quality control, less waste, reduced material costs, faster time-to-market, and higher profitability. It's the end-to-end supply chain being born.
Stephanie Pretzer is a freelance writer based in Tucson, Arizona.