Author: Tom Clifford Company: Industry Consultant (formerly Lockheed-Martin) Date Published: 1/24/2008
Pan Pacific Symposium
Abstract: Accelerated life tests (usually thermal-cycle) are increasingly necessary in microelectronics development and marketing, but are troublesome and expensive. There is never enough time, and samples are typically rare and precious. The number of samples on test is known to be crucial, but important decisions must often be made based on testing only a few samples. This paper offers some insight into the quality of tiny-N statistics, as well as guidance on testing and decisions involving life-tests of few samples. Examples and case histories are offered for test planning as well as for interpretation of sales pitches. Be careful not to lead yourself expensively astray, and be skeptical of reported "data", based on few samples.