A Practical, Innovative Method for Electro-Static Discharge Data Analysis

A practical method to identify outliers in interval data mixed with some right censored data is proposed. It is based on a simple concept of determining the probability that a deviation of an outlier, by chance, from the probability line which was fitted by the data set with the outlier deleted. The median rank method is employed to determine its sample failing fraction to construct the probability plot. A special treatment is employed to treat the suspect outlier when making probability plots and regression with the suspect outlier deleted. The Fisher Matrix is used to determine the confidence bound, and to assist the estimation of the probability. A critical interval is then defined based on the calculated confidence bound in ESD (Electrostatic Discharge) data, and then the complicated methods are simplified with a simple linear regression using Excel to determine the distance that the possible outlier deviates from the regression line. The remaining work is to simply compare the distance with the critical interval before making a decision about what to do next, such as re-do ESD tests or refine ESD circuits, perform troubleshooting on the manufacturing process, and/or implement process changes for the IC product's ESD performance improvement.