Abstract
<jats:p>Optical monitoring stops each deposited layer using an in-situ photometric signal, and a well-chosen termination rule makes the stack compensate its own thickness errors. We analyze one such rule, here called quasi-swing (QS), which terminates each layer at a fixed fraction of the running fringe's swing past its last fitted turning point, tracking a phase position rather than a signal level. From the linearized deposition-error problem and the monochromatic monitor's single observable, we show that (i) the deposition hands the controller a rank-one actuator and the monitor a rank-one sensor; (ii) QS executes exactly the admittance-norm-minimizing cut, robust to gain and calibration drift; (iii) its self-compensation can be scored against the broadband merit as a quadratic form; and (iv) that score becomes a pre-run criterion for choosing the monitoring wavelength, computable from the design alone: no Monte Carlo, no measured noise. On three filter designs of 20–75 layers built from dispersive materials, from a monitoring-wavelength study [Opt. Express 33, 42028 (2025)], the criterion matches a nonlinear deposition simulator's wavelength ranking (Spearman ρ ≥ 0.98) and optimum. Only the cut law and noise model are rule-specific: the construction is a template for wavelength selection under any termination rule.</jats:p>