Choosing tolerances for parameters that affect optical performance presents difficulty. Cost of the finished product is proposed as the only selection criterion and evaluation of cost requires an estimate of yield. A methodology is outlined for interpreting tolerances statistically for use in a monte carlo ray tracing program in order to estimate performance variability and hence yield. The method has been used to achieve high production yield but the loop through cost has not yet been completed. Feedback from production has been used with the aim of refining both the tolerance modeling techniques and the production processes.
David P. Forse,
"Statistical tolerancing for optics", Proc. SPIE 2775, Specification, Production, and Testing of Optical Components and Systems, (19 August 1996); doi: 10.1117/12.246746; https://doi.org/10.1117/12.246746