Install the Intelligence First
John Baliga, Associate Editor -- Semiconductor International, 7/1/1998
The requirements of 300 mm processing will force companies to employ more efficient and automated manufacturing techniques. A pervasive level of automation is coming to the semiconductor industry, and for those who think it is already here, the word pervasive will be redefined.
Those who have seen the diagram that SEMATECH has been using to encourage work on equipment and factory effectiveness (see Fig. 1 in "MES and CIM: At the Center of Productivity," p. 104) are aware of how wafer size increases, die shrinks and yield management efforts are expected to contribute to the industry's goals for improving productivity. There is a widening gap in that diagram labeled "other equipment effectiveness" that will have to be bridged while 300 mm wafers and 0.18 µm (180 nm) processing are also made real.
There are a lot of solutions in the works to realize that improvement in equipment and factory effectiveness. They are all occurring at different levels in the manufacturing scheme, but they all operate on the same mantra: "Install the intelligence first."
Two very important goals will have to be dealt with at every level: continuous processing and avoiding unscheduled downtime. Meeting them both sufficiently will require more decision support from a computer integrated manufacturing (CIM) system than many people can conceive.
Wafer processing has become a data driven activity. Huge amounts of data have to be collected, processed, communicated and used to make a fab work. It is stretching the state-of-the-art in computer and software technology. Very few people think of the CIM system in their fab the same way they think of a piece of equipment, but it is now just as important to install and qualify a CIM system as it is for any tool on a fab floor. In fact, it should be the first.
With the levels of decision support needed from a CIM system, a tools-first approach to building a fab would be similar to the construction of Frankenstein's monster. All the pieces are the best, but do they make the best factory system? Also, using the tools-first approach would be a rush to gain the wafer and feature size advantages without a thought to the largest gap in the productivity graph. The largest gap should be addressed first.
The components within a tool must be able to control and monitor themselves to the greatest degree possible. The sensor buses, their drivers and the control systems for the tools must be highly effective and streamlined. The manufacturing execution system (MES) must also make sure that the exponentially growing set of data is collected, processed, communicated and used appropriately.
Though the installation of intelligence involves a lot of software, software is merely a medium for the intelligence, and the mantra applies even more to the construction of software. If we can expect the functionality of semiconductor devices to double with only a 30% increase in cost, we should expect software to become twice as functional with only a 30% increase in size and cost. CIM software is no longer just glue, it is the blueprint for how a fab will operate.
The 300 mm era will put a great deal of pressure on fabs to be efficient as well as flexible. This cannot be accomplished unless the way the tools work together is very well established before tools are even installed.
