Interface A deployment, automated decision-making tools essential for full fab automation

HARVEY WOHLWEND (manager, e-manufacturing, International Sematech Manufacturing Initiative [ISMI]): As IC makers optimize 300-mm fabs, the pressure to significantly improve factory productivity remains intense. Process complexity is huge, and with two- to three-year technology-node cycles, it is imperative that things be done in a new way. Innovation is required to get the right data and to have the right tools to meet the complex needs of future technology. Getting more and better data from factory equipment is part of this innovation, and the path to that capability is called Interface A, or equipment data acquisition (EDA).

Interface A uses a second data port on factory equipment to focus expert resources on solving issues, rather than searching for data on how the process or equipment is performing. With this new capability, the time to ramp (install, configure, and qualify) a new piece of fab equipment is reduced, saving time and money for both tool suppliers and chipmakers.

ISMI recently repeated its study of required dates for standards imple-mentation. This Factory Automation Standards Tracking (FAST) II Roadmap rolled out at the recent AEC/APC North America Symposium shows semiconductor companies requesting Interface A production-ready software on equipment between summer 2005 and first-quarter 2006. Momentum is gathering for Interface A, with nearly 30 companies, including OEMs and software suppliers, on board so far.

For the past three years, ISMI's e-manufacturing program has advocated standardizing EDA. As these standards were being developed and approved, a parallel effort used prototypes as a proof of concept to accelerate industry learning and improve tools. These prototypes allowed for early identification and refinement of the
requirements as well as the structure and content of the standards.

The critical starting point for e-manufacturing is getting access to the manufacturing equipment data feeding these applications. Interface A data will be leveraged for automated decision making. A wide variety of applications will use the data, including e-diagnostics, advanced equipment control/advanced process control (AEC/APC), statistical process control (SPC), equipment performance tracking, off-line engineering analysis, predictive maintenance, and spare parts management. The ability to obtain near-real-time equipment, process, health monitoring, metrology, and yield data is absolutely essential for these e-manufacturing applications to fully function. APC and e-diagnostics will be among the early uses, although new applications will ultimately be needed.

Since this new, second equipment data port uses mainstream XML-based computing technology, the performance of these new technologies must satisfy the various applications' data requirements. The results of several recent performance studies all suggest that speed on properly architected solutions is not a major issue. Equipment internals supporting Interface A must be designed to provide dedicated, high-throughput data acquisition while maintaining equipment run rates.

Interface A implementation schemes that plan to layer on top of old SECS-II communication methods will not meet performance and reliability needs. Layering and SECS message translation also would not provide the additional data needed to feed the applications mentioned above. Initial rates are expected to be comparable with the current SECS rates, or about 50–100 total scalar variables at 3 Hz. The rate should ramp to 50 variables per chamber over time,
at a maximum on-tool sample rate of 10 Hz. Rates must be matched to the specific process dynamics implemented by each tool type.

As productivity improvements using data-driven, real-time, automated decision making accelerate, data quality must be ensured. This includes a reduction in sampling-to-reporting latency and accurate time stamps, with factorywide time synchronization based on a factory-provided master clock and internal time synchronization supporting a <10-millisecond sampling rate.

GILLES HURON (FDC product group leader, Si Automation): Advanced process control has been identified as a significant contributor to fab productivity improvement and yield enhancement. APC includes fault detection and classification (FDC), run-to-run control, integrated metrology, and e-diagnostics. E-diagnostics is the gate to e-manufacturing and is expected to enable efficient production control. Every component of APC requires a significant amount of quality data acquired at high speed.

The industry has identified a second data port—called the EDA port or Interface A—as the port that will enable efficient production control. There has been and continues to be a large effort to define the standards, some of which are partially available but not yet fully implemented. Many toolmakers are looking at the impact of implementing an EDA-type specification.

The major OEMs already deliver tools equipped with a second port. When available, the second port is either proprietary or compatible with high-speed messaging service (HSMS). Although a proprietary port is not an open port and consequently not accessible, an HSMS port is a step in the right direction that would allow additional data to be more generally available. Since second ports have started to become available on many new tools, applications such as FDC will quickly benefit from this development, since a dedicated equipment controller provides better and faster data and relieves the SECS/GEM port bottleneck. However, a migration path of these second ports to Interface A, which is the gateway to e-manufacturing, must be established.

In order to take full advantage of the second HSMS port, a few improvements could be quickly implemented. Data quality can always be improved by optimizing detection, acquisition, and delivery. The existence of process steady-state equipment events with appropriate logistics data could help a data collector (internal or external to the equipment) resolve just-in-time data collection plans (in contents or in frequency) without missing a critical part of the signal. Finally, the automation and equipment interfaces need to improve their reporting to FDC about the real nature (ID, characteristics, etc.) of the product being processed. With the implementation of these improvements in the near term, everyone would benefit from them while they wait for Interface A to become available.

JAMES MOYNE (director of APC technology, software solutions group, Brooks Automation): As e-manufacturing capabilities are brought into the fab integration equation, problems arise because many e-capabilities, such as APC, were developed from the ground up with little thought to fabwide integration. These capabilities need to be integrated at all levels, from intratool through the station controller, manufacturing execution system (MES) backbone, and even up and down the supply chain. Not only do the integration and automation of e-manufacturing capabilities have to be provided but, more importantly, "closed-loop automation" to basically close the loop around the factory is necessary. For example, layers of control in an ideal fab could include real-time equipment control; wafer-to-wafer process control; coordinated interprocess control (e.g., litho to etch to control CD); fabwide control targeting electrical characteristics, yield, and throughput; and enterprise control coordinating the manufacturing and business domains.

There are a number of requirements that must be addressed to make this a cost-effective reality. The adherence to e-manufacturing standards is at the top because the complete e-manufacturing solution cannot be provided by a single party or developed internally by a single user. SEMI standards for process control systems, data quality, and equipment data acquisition will lead the way.

A second key requirement is a flexible software approach for configuring e-manufacturing strategies. E-manufacturing solutions are achieved by coordinating the capabilities of a number of software modules into strategies that are specific to the manufacturing goals and environments. An example of such a strategy would be invoking fault detection on a process after completion and, if a specific fault is classified, calling a maintenance management system to request a maintenance event. These strategies must be highly configurable, hopefully not requiring any coding or system downtime. Strategy-builder solutions that should be effective here (and have been proven effective in other industries) use the event-condition-action paradigm.

A final important requirement of future e-manufacturing systems is the use of a dashboard approach to user interfaces (UIs). One major drawback of e-manufacturing systems is that they tend to be complex, and that complexity presents a barrier to acceptance to the process or equipment engineers who still have to do their day jobs. The key here is understanding that, in order for e-manufacturing capabilities to be embraced, the UIs must be customized to the user-class requirements. For example, this means that the UIs must present high-level data from a number of applications simultaneously, just like a dashboard.

If a comprehensive base of standards is employed along with a flexible system for maintaining e-manufacturing software strategies and a dashboard approach to UIs, we should be able to cost-effectively close the loop around the entire fab with e-manufacturing capabilities.

CHRISTOPHER A. BODE (member, automated precision manufacturing group, Advanced Micro Devices): Effective, holistic fab automation is predicated on making sound, automated decisions. The implementation of robust production control systems requires comprehension of fab-level objectives and the data and capability to optimize them while making any and all manufacturing decisions. Advances in semiconductor manufacturing capability and efficiency will be gated by how well manufacturers understand and implement fab automation along these lines.

The first step toward achieving improvements in automated manufacturing control systems is the integration of isolated systems to better comprehend the interactions between them. Improved automation will follow from richer data streams, as well as from a better understanding of how each decision affects the overall manufacturing flow. The consideration of well-characterized fab-level objectives when making lower-level decisions allows for the optimization of the former objectives at each moment within the manufacturing flow.

Automated yield management systems may find a signal that requires a controlled response. These systems may then direct supervisory process control systems to alter and control in-line process control objectives to alleviate the signal. Scheduling systems may, in turn, support process control through sampling lots that most benefit control performance. While any of these systems may adequately manage manufacturing within its given role, the interoperability of the systems allows for more-optimal performance.

Improved objectives go hand-in-hand with the need for greater capabilities within fab automation. Many manufacturing systems are limited to a lot-level scope, allowing only for lot-level optimization. An increase in granularity to the wafer level, however, facilitates control down to the wafer level. Wafer-level tracking is one capability that needs widespread implementation in order to support wafer-level decisions. Without wafer-level information, there is no effective way to optimize individual-wafer processing.

The other necessary component is the ability to make per-wafer processing decisions. Recipe modifications must be supported down to the wafer level, either through individual wafer recipes or the ability to make wafer-to-wafer adjustments to a single process recipe. To support this control, integrated metrology and embedded sensor technology implementation should proliferate to provide complete and timely characterization of process performance. In terms of improving control performance, tool throughput, and fault detection, these systems will be vital in the migration to wafer-level control.

The integration of control, scheduling, and analysis systems, coupled with wafer-level tracking and control, enables the implementation of effective automated decision making in fab automation.