August_EDFA_Digital

edfas.org 5 ELECTRONIC DEV ICE FA I LURE ANALYSIS | VOLUME 24 NO . 3 vectors together produces an overall standard normal population distribution (μ = 0 and σ 2 = 1) but now with 25% l2l , 25% w2w , and 50% s2s variance components. Looking at their overall distributions, each of these three populations produce virtually identical summary statistics (Table 1). Knowing that the MC populations are standard normal distributions, we can define threshold limits that correspond to selected overall population fractionnoncompliant levels ranging from 0.001 to 0.5 (Fig. 1). IMPACTS ON WET ASSESSMENT FOR CONTAINMENT Once a failuremode has been identi- fied, one may want to take immediate actions for containment by isolating the most-impacted wafers for remedial processing or removal from the produc- tion stream. WET sampling can provide an early read point to flag those wafers and anticipating the volume of expected disposition activities can be critical for financial and operations management. WET sampling consists of selecting some subset (n) of the total number of viable WET test sites imaged on that wafer (N) for compliance to some test criteria. Within a given sample, wemight find k of those n sites to be noncom- pliant and the wafer will be rejected subject to disposition when k ≥ q, some maximum allowed threshold estab- lished for that test. We can simulate that process using our MC populations by first counting the number of sites on each wafer that are noncompliant (i.e., the number of simulated site values on a given simulatedwafer that fall beyond the designated noncompliance thresh- old limits) and then computing the fractional frequencies of wafers having a given number of more noncompliant sites. Multiplying that fraction by 1000 yields the expected disposition rate per 1000 wafers (Disp/K). Under an assumption that each wafer is produced subject to some overall fraction defective that is randomly distributed throughout the entire population, we can Table 1 Comparison of MC population percentiles Percentiles MC Version Ῡ S Y N wafer 0.01% 0.1% 1% 10% 25% 50% 75% 90% 99% 99.9% 99.99% S2S 0.00 1 100,000 -3.76 -3.09 -2.33 -1.28 -0.67 0.00 0.68 1.28 2.33 3.10 3.72 W2W 0.00 1 100,000 -3.72 -3.09 -2.33 -1.28 -0.68 0.00 0.67 1.28 2.33 3.09 3.73 L2L 0.01 1 100,000 -3.70 -3.07 -2.32 -1.27 -0.67 0.01 0.68 1.29 2.33 3.11 3.72 Fig. 1 Simulated Monte Carlo population with fractional noncompliance limits. Fig. 2 Sample S2S Monte Carlo results align with binomial projections.