Intelligence in the Connected Fab
By David Lammers
Before we consider how connected fab tools are changing the semiconductor industry, let’s think about the larger trend: the Internet of Things that is slowly but surely connecting just about everything. Supported by smart sensors and faster wireless protocols, the Internet of Things builds on three previous waves in computing: PCs, the Internet, and mobile systems.
With some 15 billion smart devices soon to be deployed worldwide, the general manager of Intel Corp.’s intelligent solutions division, Kumar Balasubramanian, speaking at IDC’s recent SmartTech conference on intelligent systems, said analyzing the data created by these so-called edge devices is driving “new insights” and business opportunities valued at more than $3 trillion.
Many of these opportunities are already in the consumer space, ranging from Internet-connected televisions to crowd-sourced traffic navigation systems. Exercise buffs, for example, are among those engaging in “quantification of the self,” keeping track of workout sessions with body monitors that seamlessly connect to a smartphone and then to the cloud.
As important as smartphones are, the larger Internet of Things is based on what Mario Morales, semiconductor research manager at IDC, refers to as “intelligent systems.” These are Internet-connected devices which analyze data collected in real time. And there are a lot of them out there, some 25 billion units in the next decade, larger than the combined number of PCs, phones, and tablets, Morales said.
Many of these intelligent systems will be in factories and other industrial sites, “accurately monitoring conditions in real time,” said Gareth Noyes, chief strategy and technology officer at Wind River. Noyes said intelligent monitoring will support two overarching capabilities: predictive maintenance, and the “adaptive analysis of historical data, providing nuanced understanding” of what is going on inside the fab.
Maciej Kranz, vice president of Cisco’s connected industries group, notes that deploying intelligent systems will result in several forms of efficiency, including better asset utilization, faster reaction times, and better workforce optimization.
The Fab’s Internet of Things
In the semiconductor industry, much attention has been on the transitions to new transistors (finFETs), lithography tools (EUV), and wafer sizes (450). But the Internet of Things is on the chipmakers agenda as well. Connectivity is providing data from tools so that problems can be solved more easily. Predictive maintenance, intelligent scheduling, advanced process controls - all are making progress. Another equally important challenge exists: how to deal with Big Data - the analysis of huge streams of data coming from sensor-laden and increasingly connected fabs.
“Deploying intelligent systems effectively is management’s top challenge,” said Tom Sonderman, a consultant who was a pioneer in advanced process control while at AMD and GLOBALFOUNDRIES. “Equipment companies have had their silos: hardware and software, and between the different product groups working on deposition, etch, and so on. Managers such as (Applied CEO) Gary Dickerson get it: they have to break down the walls.”
Sonderman lists “management mindset” among the attributes needed for a winning Big Data solution. Technology must be joined with a service-minded organization able to turn data into usable information. Finally, companies must emphasize fast response times and have an awareness of total cost of ownership (TCO).
A leading-edge fab might contain >1,000 tools, with several dozen key sensors on each tool, and each wafer may see as many as ~1,400 process steps. Huge streams of information are coming from fab tools, wafers, GDS (design) files, and facilities.
“All of this information is like a sunken treasure; you have to figure out how to get it into the boat,” Sonderman said.
Too Much of a Good Thing?
The data coming from fabs is expected to triple, from 50 terabytes (TB) at the 45nm node to 140 TB at 20nm design rules, according to John Scoville, senior director of application engineering in the Applied Global Services (AGS) group. A typical 300mm tool might have 500 system variable identifiers (SVIDs) running at 5 Hz, while a 200mm tool has about 160 SVIDs operating at 1 Hz. Making sense of all that data is part of an advanced technology-enabled services push at AGS, where Applied engineers work in close partnerships with the engineering teams at semiconductor manufacturers. These productivity-enhancement projects focus on excursion control, predictive maintenance, scrap reduction, chamber matching, and reduction of particles/defects, among others.
Alex Schwarm, senior product manager in the AGS group, said technology-enabled services go beyond the maintenance-driven relationships Applied has established with scores of chipmakers. Most of the large chip makers, Schwarm said, are already collaborating with Applied on data-driven analysis projects, drawing upon Applied’s E3 performance-tracking and data mining software, as well as libraries and models. “We have developed a method of working with the customer to help them determine how and what to analyze. It is not a closed box, it is more interactive and integrated than a typical maintenance relationship. You are now an extension of the customer’s team. Rather than silo things, which you can easily do with a maintenance service relationship, for these device- and film-oriented problems we need to be more integrated into the customer’s team.”
Most semiconductor manufacturers have worked with Applied on process challenges. “They want to know the long-term strategy,” Schwarm said. “There is a change in what customers expect from Applied, from how best to replace parts or fix electromechanical issues, to a different type of value, based on a specialized skill set and data analysis and management capabilities we bring to the table.”
For every dollar invested in the collaboration, the customer looks for yields and productivity to improve measurably.
“The expectation is to double your money. Whether it is scrap or parts consumption or whatever metric we’ve agreed upon as the focus, we figure out how we will measure it. Then we publish the data every quarter, and say, ‘Here Mr. Customer, is what we’ve done so far.’ It is very transparent, very clear,” Schwarm said.
At the end of the project, the Applied-customer team often moves on to a different set of tools, or a new device, or a new technology node. “It is beholden on us to demonstrate the value to the customer. Otherwise, we wouldn’t expect them to continue,” he said.
Big Data – Big Challenge
Sonderman said fabless companies face a particularly big challenge. In the age of the Internet of Things, closer partnerships between foundries and design teams are needed to turn a raw data into useful information. “All the barriers, all of the data partitioning between a design team and the fab, has to come down. That’s the conclusion people are coming to as they face the question of what it is going to take to make Big Data truly useful.”
For a further glimpse at the future of intelligent connectivity and the Internet of Things in the fab, watch for Dave Lammers’ continuing report in the December 2013 edition of Nanochip Fab Solutions.
For more information contact firstname.lastname@example.org