Monitoring a deposition process using an extended adaptive Kalman filter

Multilayer infra-red filters are produced by heating alternate materials of differing refractive index in an evacuated chamber to form a vapour which condenses onto product substrates. To observe this deposition a monochromatic infra-red beam is reflected from a monitor substrate and the reflected beam is combined and transduced to form a reflected energy curve (a repeating interference fringe) which is related to the thickness of deposit in a highly nonlinear manner. The present method of manufacture needs a highly skilled operator to interpret this curve in order to ascertain the thickness of deposit and the rate of deposition. This paper describes the application of a Kalman filter to this process to provide estimates of the thickness of deposit and the rate of deposition. The resulting estimates are checked using an exact theory and demonstrated to be of sufficient quality to be used in the manufacture of infra-red filters. Limitations in the present application are also noted and ways of overcoming them presented. At present the Kalman filter estimates are used by the operator as a `navigational¿¿ aid to manufacture but it may be possible to use them in a feedback loop to control the process automatically.