Department of Defense MIL-HDBK-816 Notice 1 — MIL-HDBK-816 Notice 1 Radiation Hardness Assurance Guidelines Manual PDF

Downloaded from https://www.everyspec.com MIL-HDBK-816 NOTICE 1 Obtain characterization data and documentation. Select desired lot acceptance probability. Confidence level and sample size for sample series tests. Select parameters to be tested and test conditions. Examine data for adequacy, errors and outliers. Examine data for abrupt failures. Correct or discard bad data. Recast the data to the specified conditions. Compare within-lot and lot-to-lot variations. Identify the type of data analysis that will apply. Calculate RHA end-point limits (RHEPL) to meet objectives. Document all steps in the end-point limit calculations. Incorporate RHEPL values in device specification. FIGURE 4.

Steps for calculating end-point limits for lot acceptance tests. 5.2.1.3 Statistical uncertainties associated with lot acceptance testing. The sample size series sampling table in MIL- PRF-38535, Appendix D and other military standard documents are based on statistical formulas for sampling with replacement i.e., the sample parts used for the tests is replaced into the lot of parts from which it was drawn. Because radiation tests are damaging, the parts tested are not replaced into the lot. The statistical implications of this fact can be complex and are beyond the scope of this document.

In general, statistical uncertainties are reduced when the lot acceptance sample size is large and are further reduced if the lot acceptance history is known. The least desirable situation occurs when only a single lot has been tested and when the sample size is only a few parts, say four or less. In this latter case it will be difficult to estimate what the performance characteristics of the passed lot will be. The recommendation is, therefore, that lot acceptance tests based on four parts or less should not be used. For small sample sizes, information about the uniformity of the lots is also very important.

If a lot were perfectly uniform, for example, then testing a single part would be adequate. If the uniformity of each lot is high then, during the lot acceptance tests, most of the time, either all the parts pass or they all fail. For such uniform lots, a 2/0 or a 4/0 test (presently specified in the group E tests of method 5005 of MIL-STD-883) may be able to give some confidence that the passed lots are of high quality. If lots are mostly being rejected because one part out of the sample is failing, then the passed lots must be considered to be of low quality. SUPERSEDES PAGE 27 OF MIL-HDBK-816 27