To properly account for imperfections in fabrication of lenses and mechanics as well as optics alignment limitations, it is necessary to run a tolerance analysis during the optical design process. But it is often unclear what level of detail the tolerance model necessitates to accurately predict the variation in performance, and there are consequences of time and money for both overly optimistic and overly pessimistic designs. In this work, we compare the assembly results of a precision microscope objective with an initially overly pessimistic tolerance analysis and develop an improved tolerance model in agreement with the performance of as-built systems. The system is analyzed in two industry-standard optical design software packages, Zemax and CODE V, and the results are compared with each other and with assembly data. The Zemax Monte Carlo analysis agrees with as-built data by including manufacturing and assembly tolerance statistics as well as by modelling compensation procedures used in the lab. Analyses in CODE V’s TOR and Monte Carlo yield consistently similar results to each other but are more pessimistic than the Zemax analysis. All three methods are found in agreement by reducing the irregularity tolerance on CODE V analyses to reflect the software’s differing tolerance definitions. Although much faster, CODE V’s TOR is unable to model complex compensation procedures, suggesting that a preliminary analysis in TOR should be followed by a more thorough Monte Carlo analysis to avoid overspecification of components. Several aspects of the tolerance model developed here are generalizable to other systems.
|