WHICH PROCESS IS BETTER? Carol Dixon, Freeflowe Technical Services Lean Six Sigma, Quality, Continuous Improvement 2/27/2014

Introduction Carol Dixon, ASQ CQE, CSSGB Principal, Freeflowe Technical Services Phone: Lean Six Sigma, Quality, Continuous Improvement Consulting and Blended Training for small manufacturing businesses Bachelor’s Engineering Physics Graduate studies in Statistics Member of ASQ since 2011 Manufacturing and Quality Engineer in electronics, primarily Hewlett-Packard and medical devices

Overview Process Improvements in manufacturing frequently involve comparisons What is a process? – Process Map Which process is better?  Options  Criteria for evaluation o Comparison criteria for each option and between options o Statistical Hypothesis testing (t-test)  Complete evaluation  Results  Conclusions  Making decision

Raw materials Process step 1 Process step n Final Product Test/ inspect Process step 2 New Supplier, material property, actual material change Tooling, Process or metrology, operator change Pass Process Map Process

EXAMPLE – Which material is better? R_mean ( Ω) R_mean_sigma ( Ω) Yield (%)CpK estimate OldHypothesis testing to determine if there is a difference in means. Ho: μ (A)= μ (B) H a: μ (A) ≠ μ (B) Higher is better Hypothesis test difference in means. Ho: μ (A)= μ (B) H a: μ (A) ≠ μ (B) >1.33 Higher is better New Two different semiconductor materials are being evaluated within a batch for impact to electrical resistance. (Typical batch size 25 wafers). Each wafer for this example will have 100 die. Specification (Target R) = 1250 Ω, LSL= 1200 Ω, USL1300 Ω Wafers 1-6 – Old material Wafers 7-25 – New material After processing all wafers are tested electrically and parameters are compared between the old and new.

Hypothesis testing Courtesy: H O : µ 1 = µ 2 H A :µ 1 ≠ µ 2 H O :µ 1 = µ 2 (α)

Results Go to JMP Test for normality – R_mean  Hypothesis testing for R_mean and Yield  Process capability, CpK R_mean ( Ω) R_mean_sigma ( Ω) Yield (%) (90% acceptable) R_mean CpK estimate (Target >1.00) Old New Compare resultsDifferences in means are not statistically significant Difference in means are not statistically significant but data for new was bimodal. New has higher CpK

Conclusions Is the new material better than the old material? Based on comparison criteria Y or N? What if the new material costs 20% less? What if the new material costs 20% more?