Abstract

Among the various factors that slow lens optimization—insufficient performance targets, the absence of a unique solution, false local minima, a poorly scaled change vector, failure to find the optimum damping number, and failure to equalize the parameter gradients—the importance of parameter gradient equalization has been insufficiently recognized. Gradients can be approximately equalized by scaling the lens to a suitable size while it is being optimized. For best results, the size of the damping number should also be optimized during each iteration. If these two procedures are followed, scaling the change vector is usually not crucial. To illustrate the importance of parameter gradient equalization, a lens optimization is analyzed.

© 1981 Optical Society of America

Accelerating convergence in automatic lens design

Paul N. Robb
Appl. Opt. 18(24) 4191-4194 (1979)

Lens design and local minima

Berlyn Brixner
Appl. Opt. 20(3) 384-387 (1981)

Automatic Lens Design for Nonexperts

Berlyn Brixner
Appl. Opt. 2(12) 1281-1286 (1963)

Automatic lens design with pseudo-second-derivative matrix: a contribution

Andrea Faggiano
Appl. Opt. 19(24) 4226-4229 (1980)

Damped Least-Squares Method for Automatic Lens Design

Joseph Meiron
J. Opt. Soc. Am. 55(9) 1105-1109 (1965)