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 d. The characterization data was adequate to provide post-irradiation current gain values, at the required collector current, bias voltage, and temperature at 2.5 x 10 13 neutrons/cm 2 . This data is shown in table XIII. Because this is an illustrative example, data is shown for only ten devices. In actual practice, a sample size of 10 should not be considered adequate to support the calculation of an end-point limit. For the values in table XIII, the following quantities may be calculated: MEAN(LN(h FE (PHI(SPEC)))) = 4.591 and STDEV(LN(h FE (PHI(SPEC)))) = 0.107. Assumptions b. and c., make C = 0.90 and P = 0.99.

The size of the characterization sample is 10. For C = 0.90, P = 0.99, n = 10, the one sided tolerance limit factor, K TL , is found in table XII to be 3.5.32. The end-point limit, RHEPL, is thus: RHEPL = 4.591 – 3.532 x 0.107 = 4.213 In more familiar terms, the geometric mean value of the current gain for the ten devices in table XII is 98.6, the plus and minus one standard deviation gains are 109.7 and 88.6 respectively, and the end-point limit for the gain is 67.6 ( = exp(4.213)).

Lot acceptance tests for 2N2222 transistors would then be performed in the future by irradiating 11 devices from each lot to 2.5 x 10 13 neutrons per cm 2 and requiring that each device out of the 11 have a post irradiation gain greater than 67.6; otherwise the lot is rejected. TABLE XIII. Post-irradiation data for ten 2N2222 transistors. PHI(SPEC) = 2.5 x 10 13 neutrons/cm 2 (1 MeV Si damage equivalent) Device number h FE (PHI(SPEC)) LN(h FE (PHI(SPEC))) 1 108.8 4.689 2 105.4 4.657 3 93.7 4.540 4 97.7 4.582 5 102.6 4.630 6 104.5 4.650 7 84.1 4.431 8 101.1 4.616 9 112.5 4.723 10 81.1 4.396 5.2.4.4.5 Case 2. Lot-to-lot variations larger than within lot variations.

Case 2 is applicable when within-lot standard deviations are small compared to lot-to-lot variations. In this case, the effective sample size for the RHEPL calculations is the number of lots and not the number of devices used for the characterization measurements. An approximate analysis method for this case was developed by I. Arimura and A. Namenson and published in IEEE Trans. Nucl. Sci., NS-30, 4322, December 1983. The method assumes that the parameters follow a normal or lognormal probability distribution but it can be modified for a different distribution if necessary.

Once again the assumption described in 5.2.3.7 is made, namely that the object of the lot acceptance tests is, with 90 percent confidence, to have at least 90 percent of lots pass. It is assumed also that an 11/0 sample size series test will be used for lot acceptance. This last assumption means that, again, the individual part survival probability of the lot being tested for acceptance must be P = 0.99. 5.2.4.4.5.1 Sampling for multi-lot analysis. Ideally, there should be at least five lots with at least five parts in each lot. If the within-lot variations are very small compared to the lot-to-lot variations, the requirement on having at least five parts per lot may be waived.

SUPERSEDES PAGE 34 OF MIL-HDBK-816 34