AVS 52nd International Symposium
    Exhibitor Workshop Tuesday Sessions
       Session EW-TuL

Paper EW-TuL3
Robust System Identification and Optimized Tuning for Control of Evaporation Processes: Benchmark Study Results of Manufacturing Performance

Tuesday, November 1, 2005, 12:40 pm, Room Exhibit Hall C&D

Session: Vacuum Components and Measurement Optimization
Presenter: M. Gevelber, Cyber Materials, LLC
Authors: G. Reimann, Cyber Materials, LLC
B. Vattiat, Cyber Materials, LLC
M. Gevelber, Cyber Materials, LLC
J. Hildebrand, Maxtek, Inc.
C. Hildebrand, Maxtek, Inc.
Correspondent: Click to Email
Crystal monitors have been used for over 30 years to provide real-time control of evaporation sources in order to maintain a desired deposition rate. However, the performance of these systems is dependent on proper choice of controller gains. Review of a number of commercial operations has revealed that many controllers are mistuned and fail to compensate for large deposition rate variation. In many cases, poor controller tuning actually magnifies rate variations. These significant variations adversely impact coating quality, reduce yield, and limit throughput. In an evaporation system, the controller must be tuned to react to the dynamic response characteristics and disturbances typical of evaporation processes. The tuner will need to robustly deal with the process nonlinearities and variations that occur during a run. While a number of tuning approaches have been developed or suggested for controlling evaporation processes, none are optimized for the specific conditions observed in the processes, nor are they designed to handle the variety of conditions that typically occur. We present our work on a robust and automatic method for obtaining optimized controller tuning. The performance of the proposed tuning is evaluated under manufacturing conditions. We propose a two step process that first robustly identifies the system characteristics, and then applies an appropriate optimization scheme that selects controller gains based on the identified process characteristics. By robust, we mean that the identification process works despite all variations observed in practice including arcing, nonlinearities due to operating point dependency, and variations in system characteristics. In addition to describing the new robust and optimized controller tuning scheme, this paper reports our initial benchmark performance results of the new controller tuning system, as well as our analysis of system drift in several manufacturing systems.