August_EDFA_Digital

edfas.org 7 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 3 on its σ w2w offset component (Fig. 5a). The probability of flagging that wafer at WET will vary correspondingly. As wafers experience larger σ w2w offsets that approach and eventually exceed the limits, the probability of them failing at WET rises quickly from 0 to 1 (Fig. 5b). However, the probability of a wafer experiencing a given σ w2w offset diminishes the further that offset lies fromthemean value (Fig. 5c). The overall expected probability of flagging wafers with a non-zero w2w variance component is then determined by integrating over the product of probability of rejection curve times probability of occurrence curve (Fig. 5d = 5b * 5c). By computing the area under the product of the bino- mial probabilities andwafer component probability densi- ties, we cannowproperly project the expecteddisposition rate reflecting the impact of w2w variationonour sampling results (Fig. 6). When we extract the expected disposition rates for our L2L MC (tallying the fraction of wafers exceeding a given threshold for non- compliant sites within the MC population itself ), we find that it aligns with those from the W2W MC (Fig. 7). Subdividing our MC into separate 25% w2w and 25% l2l components does not alter disposition rates over prior projections based on 50% w2w . For wafer- based sampling schemes monitoring randomly dis- t r ibuted de fect modes throughout the population, we only need to account for the total combined variance components at wafer level or above relative to those within wafer to accurately predict expected disposition levels thatmay impact finan- cial or operational planning activities for those products. IMPACTS ON WET ASSESSMENT FOR IMPROVEMENT When striving to identify improvements to theprocess that enhance performance or remediate issues, we will be drawing and compar- ing results across samples drawn from alternate pro- (a) (b) (c) (d) Fig. 5 Wafer defectivity and binomial probability across w2w v ariation offsets. Fig. 6 W2W Monte Carlo aligns with integrated binomial density projections.