A coefficient fabrication error model for sampled data filters

A coefficient fabrication error model is derived for sampled data filters that requires the first two moments of the random perturbed filter coefficients as inputs. Two examples of model usage are given for an electrooptical processor (EOP) implementation of a finite impulse response (FIR) filter. Both types of EOP coefficient fabrication errors (mask aperture misalignment and mask aperture dimension error) were characterized and then evaluated. The model results were found to be in close agreement with a corresponding Model Carlo simulation predicting highest sidelobe increases of 5 and 6 dB, respectively, for misalignment and dimension tolerances of 6 μm.