Share Email Print

Proceedings Paper

Design of multi-layer dielectric thin film filters using error back-propagation
Author(s): Hyuek-Jae Lee
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

Although multi-layer dielectric thin films are widely used for frequency filters, its design process requires tedious parameter optimizations. Recently an efficient parameter optimization algorithm known as error back-propagation (EBP) has been used for adaptive training of feed-forward neural networks, i.e. multi-layer Perceptron (MLP). In this paper we have investigated analogy between light transmission in multi-layer thin films and signal feed-forward in the MLP, and developed a new design procedure for the dielectric thin film frequency filters. An error is defined as sum of differences between desired frequency response and actual output response, and minimized by gradient descent algorithm with EBP. In this algorithm sensitivities of the error to each film thickness are efficiently calculated by back-propagation of the output error through the layers, and readily available during optimization. Special consideration on frequency response such as important frequency range and low thickness sensitivities are easily incorporated into the algorithm. Adaptive increase of layer number may also be incorporated to achieve required frequency response and allowable error. Efficiency of this algorithm is demonstrated by designing low pass, high pass, and band pass filters.

Paper Details

Date Published: 26 July 1993
PDF: 2 pages
Proc. SPIE 1983, 16th Congress of the International Commission for Optics: Optics as a Key to High Technology, 198330 (26 July 1993); doi: 10.1117/12.2308527
Show Author Affiliations
Hyuek-Jae Lee, KAIST (Korea, Republic of)

Published in SPIE Proceedings Vol. 1983:
16th Congress of the International Commission for Optics: Optics as a Key to High Technology
Gyorgy Akos; Tivadar Lippenyi; Gabor Lupkovics; Andras Podmaniczky, Editor(s)

© SPIE. Terms of Use
Back to Top