Critical view of three lens design methods: damped least squares, Spencer's and Glatzel's

Abstract

A lens system can be represented as a point in the N-dimensional parameter space. There is a set of quantities called merit functions associated with points in this parameter space that define the performance of the lens system. Ideally this set can be reduced to a single quantity called the merit function. Lens design iterations consist of transferring from one point in parameter space to another where the merit function indicates an improvement. By repeating this process the designer may eventually reach a satisfactory solution. Even when the functions involved are theoretically continuous, the transferring in parameter space may never be strictly continuous because it is a numerical process. Yet, it should be possible to take arbitrary short steps on going from one point to another. Long steps, or leaps, can abruptly lead to undesirable regions from where it may be difficult to return to the previous, more desirable region. There are three common methods to set up the merit function: Levenberg’s damped least squares (DLS), G. Spencer’s constrained damped least squares (CDLS) and E. Glatzel’s automatic adaptive correction (AAC). This paper reviews the pros and cons of these methods. A discontinuity found in Spencer’s method will be discussed in detail.