paper 1.4 Application of Non-parametric Statistics of the Parametric Response for Defect Diagnosis

Abstract
I. INTRODUCTION
II. ORDINAL INVARIANT AND RANK STATISTICS
The statistics calculated using rank order are called ordinal invariant statistics.
A. Kendall τ
III. IDDQ DEPENDENCE ON EXTRINSIC VARIABLES
ProdA data is from several lots totaling approximately 2,500 die.
IV. CORRELATION MODEL FOR IDDQ
A. True Value, TV
B. Process Noise, PN
For IDDQ process noise is dominated by sub-threshold current which is controlled by the temperature and threshold voltage.
C. Tester Noise
D. IDDQ Discordant Plots
V. MONTE-CARLO ESTIMATION
A. Shape (Homogeneity) of τ Extreme Values
B. Critical Value Tables
C. Effectiveness of Kendall τ
VI. KENDALL τ CONTRIBUTION TO DIAGNOSIS
A. Elevated versus Non-Elevated Vectors
VII. RESULTS
A. Outlier Screening
B. Elevated and Non-Elevated Vector Selection
C. Inter-Die Correlation
VIII. CONCLUSION